March 28, Bryce Bird Utah Division of Air Quality Multi Agency State Office Building 195 North 1950 West Salt Lake City, UT 84116

Transcription

1 March 28, 2014 Bryce Bird Utah Division of Air Quality Multi Agency State Office Building 195 North 1950 West Salt Lake City, UT Re: Western Energy Alliance Comments on the General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery Dear Mr. Bird: The intended goal of a General Approval Order (GAO) for similar source types is to provide certainty and consistency for the sources and to streamline the permitting process while maintaining environmental protections. Western Energy Alliance (Alliance) members share these goals and would like to seek coverage under this GAO as it would be more efficient than applying for individual Approval Orders (AOs). As proposed, however, the GAO contains very prescriptive engineering design limitations that will prevent this GAO from achieving the intended goals, and in some instances do not result in environmental benefit. Five proposed requirements in particular will prevent our members from using the GAO, and many other requirements severely limit its usefulness for the majority of our members who operate in Utah. Throughout the stakeholder process, Alliance members hoped this GAO would be developed in a manner in which coverage could be sought for all types of oil and natural gas sites. After detailed analysis of the proposed GAO, however, we have concluded that the only practical application is limited to crude or condensate tank batteries servicing one or two wells, unlike as indicated by the title and abstract. With that in mind, we tailored our comments to reflect just that limited application. We identify the five elements of the GAO that render it unusable by our members (highlighted in green on the attached Comment Table) and discuss the main arguments for them below. Additional arguments for these five can be found in the enclosed Comment Table (see Attachment A). We also highlight other general technical concerns in the Comment Table that could pose problems for use, especially if this GAO is applied to other types of oil and natural gas sites beyond crude or condensate tank batteries servicing one or two wells. Modeling and Stack Height Requirements If stack heights are prescribed in the GAO, operators will not be able to use it and will have to file NOIs for individual AOs. The proposed GAO states that stack heights will be determined based on dispersion modeling for NO 2. Relying on 1-hr NO 2 modeling to determine stack heights will result in

2 Western Energy Alliance Comments on Proposed GAO March 28, 2014 Page 2 of 4 unnecessary costs as well as stack height constraints that may raise safety and operational issues (e.g. back pressure). The inability of air quality models to accurately predict 1-hr NO 2 concentrations is well documented. A letter from the Western States Air Resources Council (WESTAR) to EPA requests that EPA conduct critically needed field studies to resolve 1-hr NO 2 modeling issues (see Attachment B). The background document (see Attachment B), written by the WESTAR 1-hr NO 2 modeling ad hoc committee (of which UDAQ staff were members) points out the model s tendency to overestimate 1-hr NO 2 impacts, and says, it is possible that modeled concentrations exceed the standard when monitoring indicates compliance with the standard. Given these concerns with the accuracy of 1-hr NO 2 models, UDAQ should not rely solely on model results to determine NAAQS compliance. There is sufficient information to demonstrate that a facility authorized under the GAO will not interfere with the attainment of the 1-hr NO 2 standard. Western Energy Alliance suggests contacting other States regarding their 1-hr modeling issues, such as Wyoming. Wyoming conducted an extensive amount of modeling for several facilities to determine the impact of the 1-hr standard on permitting. None showed compliance with the 1-hr standard unless unreasonable stack heights were used. Based on this modeling, it was clear that requiring applicants to demonstrate compliance with the 1-hr standard via modeling was not a viable path for minor sources going forward. Instead, they rely on the extensive ambient monitoring program data to make a demonstration that the proposed facility will not prevent attainment with the 1-hr NO 2 ambient standard. We believe Utah s extensive ambient monitoring program is equivalent and allows UDAQ to adopt this same approach (Table 3-4 in Attachment F). We request that UDAQ consider these and other state regulations that will allow for reasonable GAO requirements while still demonstrating NAAQS compliance. Pneumatic Controllers and Pumps Unless applicability of this requirement is limited to continuous bleed pneumatic controllers, operators will not be able to use the GAO and will have to file NOIs for individual AOs. Sections II.B.5 and II.B.6 address pneumatic controllers and pneumatic pumps, respectively. The GAO does not specify the type of pneumatic controllers that are subject to the requirements. We suggest UDAQ limit the GAO requirements to continuous bleed pneumatic controllers. Recognizing the difficulty of determining emissions from intermittent controllers, EPA did not regulate intermittent bleed emissions related to process control in NSPS OOOO. 1 The GAO requirements for pneumatic controllers go far beyond what is required of operators in NSPS OOOO as pneumatic pumps are not even covered in NSPS OOOO. 1 Oil and Natural Gas Sector: New Source Performance Standards and National Emission Standards for Hazardous Air Pollutants Reviews: Response to Public Comments (76 FR 52738)

3 Western Energy Alliance Comments on Proposed GAO March 28, 2014 Page 3 of 4 Additionally, UDAQ lacks any cost benefit analysis or BACT analysis on requiring controls for pneumatic pumps demonstrating the cost effectiveness of such a requirement. We request UDAQ eliminate specific requirements for pneumatic pumps and reiterate our request that only continuous bleed controllers be subject to requirements under the GAO. We are also concerned about the inconsistent requirement in Sec. II.B.5.a.a that controllers shall have a bleed rate less than or equal to 6 standard cubic feet per hour [scf/hr] and shall comply with 40 CFR (d). If a pneumatic controller bleed rate is less than 6 scf/hr, it is not subject to the referenced Federal regulation. Similarly, the GAO requires compliance with 40 CFR (d) in Section II.B.6.a.a for pneumatic pumps; however, pneumatic pumps are not subject to 40 CFR Moreover, that specific rule was not designed for pneumatic pumps. In light of these conflicts for both controllers and pumps, we suggest UDAQ remove all reference to 40 CFR Malfunction Section I.5 requires that equipment be maintained during periods of startup, shutdown and malfunction. It is not possible to maintain all equipment during periods of malfunction. Additionally, this language should be consistent with the language suggested by the Alliance for the Preliminary Draft for Discussion Purposes, General Provisions, R proposed on July 30, Unless the term malfunction is removed, operators will not be able to use the GAO and will have to file NOIs for individual AOs. Broad Applicability of NSPS JJJJ UDAQ is attempting to apply BACT to all engines, not just new or modified engines. The Alliance questions UDAQ s authority to require controls on engines that are not new or modified. UDAQ references NSPS JJJJ in Section II.B.8, Engine Requirements, but the GAO is far more stringent than NSPS JJJJ. It should be noted that in NSPS JJJJ EPA took careful consideration of manufacture date, engine type (lean or rich) and capacity to ensure that the resulting requirements incorporated appropriate cost effectiveness and technical feasibility. Section II.B.8.b.a references NSPS JJJJ limits and II.B.8.b.b references Table 1 of NSPS JJJJ (see Attachment K) before listing emissions standards for NOx, CO and VOCs. The standards listed in the GAO in II.B.8.b.a are only applicable to engines manufactured after July 1, 2008 and those in II.B.8.b.b are only applicable to engines manufactured after January 1, 2011 in NSPS JJJJ. In some cases, it is not technically feasible to modify an engine to meet these stringent emissions standards. Where it is technically feasible, it is often cost prohibitive to modify an engine. Therefore, EPA does not apply one set of emission standards to all engines. We suggest UDAQ strike the Section II.B.8.b and refer to Section III of the GAO, which includes NSPS JJJJ in the list of federal requirements with which owners/operators must

4 Western Energy Alliance Comments on Proposed GAO March 28, 2014 Page 4 of 4 comply. If UDAQ retains such stringent emissions standards for all engines, our members will not use this GAO and will be required to file NOIs for individual AOs. Dehydrator and Tank Controls Sections II.B.2 and II.B.3. require VOC controls for tanks and dehydrators, respectively, no matter their size. UDAQ lacks any cost benefit analysis or BACT analysis demonstrating cost effectiveness. We are unaware of any Federal or State requirement for dehydrator and tank controls on all units regardless of size. The airshed would likely observe more emissions from the control device than total emissions from smaller dehydrators and tanks. The Alliance proposes that the requirement to control dehydrators and tanks regardless of size be stricken from the GAO. Where a dehydrator or tank control requirement is pursued, an appropriate de minimus threshold should be included based on technical and economic feasibility. Without a threshold, operators will not be able to use the GAO and will have to file NOIs for individual AOs. The comments in this introductory letter are just those that make the GAO unusable for our members. We provide more detailed information and supporting documentation in the attachments, which include the Comment Table (see Attachment A). Our full comments comprise this letter and the ten attachments. Thank you for the opportunity to comment on the GAO and for working with Western Energy Alliance and considering our concerns. Please do not hesitate to contact us if you questions on these comments. Sincerely, Kathleen M. Sgamma Vice President of Government & Public Affairs Cc: Cody Stewart, Energy Advisor, Office of Governor Herbert Amanda Smith, Executive Director, Utah Department of Environmental Quality Brock LeBaron, Deputy Director, Utah Division of Air Quality Alan Humphreys, Minor New Source Section Manager, Utah Division of Air Quality Martin Gray, Major new Source Section Manager, Utah Division of Air Quality Attachments (11)

5 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language Abstract P1 A General Approval Order (GAO) may be issued under the authority of UAC R This GAO is for a Crude Oil and/or Natural Gas Well Site and/or Tank Battery. Produced fluids will be brought to the surface from a single well or multiple wells. Oil, condensate, water, and gas will be separated from the produced fluid. The oil, condensate, and water will be stored in tanks prior to being transported off site by trucks. The gas may pass through a dehydrator on site. The gas shall either be used as fuel for onsite equipment or be routed to a gas gathering system and sent off site. This GAO will cover a facility that processes up to 50,000 barrels of crude oil and condensate combined per year. A dispersion modeling analysis was conducted for NO2. Conditions in this GAO reflect the results of this modeling analysis and will ensure protection of the NAAQS. The HAP emissions are limited by emission controls and equipment specification to ensure the requirements in R (1)(c)(ii) or (iii) will not be triggered. The proposed GAO states that stack heights will be determined based on dispersion modeling for NO 2. Relying on 1-hr NO 2 modeling to determine stack heights will result in unnecessary costs as well as stack height constraints that may raise safety and operational issues (e.g. back pressure). The inability of air quality models to accurately predict 1-hr NO 2 concentrations is well documented. A letter from the Western States Air Resources Council (WESTAR) to EPA requests that EPA conduct critically needed field studies to resolve 1-hr NO 2 modeling issues (Attachment B). The background document, written by the WESTAR 1-hr NO 2 modeling ad hoc committee (of which UDAQ staff were members) points out the model s tendency to overestimate 1-hr NO 2 impacts, and says, it is possible that modeled concentrations exceed the standard when monitoring indicates compliance with the standard. (Attachment B), Given these concerns with the accuracy of 1-hr NO 2 models, UDAQ should not rely solely on model results to determine NAAQS compliance. There is sufficient information to demonstrate that a facility authorized under the GAO will not interfere with the attainment of the 1-hr NO 2 standard. Western Energy Alliance suggests contacting other States regarding their 1-hr modeling issues, such as Colorado and Wyoming. Colorado has acknowledged and adopted EPA s approach to NO 2 modeling (EPA Guidance, Attachment D) in a memo that states, ambient air quality impact analyses are not necessary for either NO 2 or SO 2 emissions below the 40 tpy [significant emissions rate] (CDPHE, Attachment C). This is also consistent with UDAQ R , which states that NO 2 modeling requirements are limited to sources with a total controlled emission rate per pollutant greater than or equal to amounts specified in Table 1, which is 40 tpy. Wyoming conducted an extensive amount of modeling for several facilities to determine the impact of the 1-hr standard on permitting. None showed compliance with the 1-hr standard unless unreasonable stack heights were used. Based on this modeling, it was clear that requiring applicants to demonstrate compliance with the 1-hr standard, via modeling, was not a viable path for minor sources going forward. Instead they rely on the extensive ambient monitoring program data to make a demonstration that the proposed facility will not prevent attainment with the 1-hr NO 2 ambient standard. The drastic discrepancy between measured and modeled data is illustrated in Table A-1 in EPA s March 1, 2011, memorandum (EPA Guidance, Attachment E). It is also important to note that this successful Wyoming program monitors facilities with equipment far in excess of the GAO capacities. Overall, we believe Utah s extensive ambient monitoring program is comparable, shows compliance with the 1-hr NO 2 standard (Table 3-4 of the Redhorse Modeling Study, Attachment F) and allows UDAQ to adopt this same approach. We request that UDAQ consider these and other state regulations that clearly demonstrate NAAQS compliance and allow for reasonable GAO requirements. If stack heights are so high as to be technically infeasible, operators will not be able to use the GAO and will have to file NOIs for individual approval orders (AO). Attachment B: Western States Air Resources Council (WESTAR) letter to EPA and 1-Hour NO 2 Ad Hoc Committee background document Attachment C: Colorado Department of Public Health and Environment (CDPHE) PS Memo Attachment D: EPA Guidance Memo June 29, 2010 Attachment E: EPA Clarification Memo March 1, 2011 Attachment F: Redhorse Utah NO2 Modeling Study Report Western Energy Alliance ATTACHMENT A: Page 1 of 16

6 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language P2 A source must comply with the requirements of R (4) to be subject to this GAO. If a source is not able to construct within the requirements of this GAO, the source must submit a NOI under R and obtain an AO under R P3 NSPS 40 CFR 60 Subpart A, Dc, JJJJ, and OOOO, and MACT 40 CFR 63 Subpart A, HH, and ZZZZ regulations may apply to this source. NESHAP 40 CFR 61 regulations do not apply to this source. Title V of the 1990 Clean Air Act does not apply to this source. P4 The potential emissions, in tons per year, are estimated to be as follows: PM10 = 0.52 (which includes PM2.5), PM2.5 = 0.52, NO x = 8.45, CO = 12.94, VOC = 13.55, HAPs = 2.55 and CO2e = 6,348. P5 The NOI for the above-referenced project has been evaluated and has been found to be consistent with the requirements of UAC R307. Air pollution producing sources and/or their air control facilities may not be constructed, installed, established, or modified prior to the issuance of an AO by the Director of the Utah Division of Air Quality. Form 1 General Information (Application to the GAO) requires the GAO approval letter from UDAQ be issued before construction or installation, but there are several problems with combining that requirement with other data requests in Form 1. Item 12b requires confirmation that the site will have an annual throughput of crude oil and condensate less than or equal to 50,000 bbls/year, but we cannot confirm the throughput of a site before the well has been drilled. Also, Form 1 requires the requested information to be accurate and complete. We cannot verify the accuracy of the information required in Item 12a and 12b until the site has been constructed and we know the well production rate. We suggest that UDAQ remove the specific data requirements from the Form 1 application and request that data within the records section of the GAO (Section I.4). Additionally, many things are unknown during the first days of production that influence emissions including the following. A source must comply with the requirements of R (4) to be subject toeligible for this GAO. If a source is not able to construct within the requirements of this GAO, the source must submit a NOI under R and obtain an AO under R Flowrate uncertainties What type of separator can be used (high/low) Composition of the production Decline curve determination Well pressure Production is not stabilized These issues could result in significant differences in the initially estimated production, which in turn could affect the applicability of the General Approval Order and estimated emissions. In order to limit the risk of underestimating production, applicants would have to significantly overestimate production, and therefore overestimate emissions and valuable\critical emission offsets. This rationale for allowing operators time after start of production to file paperwork on production and emissions was included in the preamble in NSPS OOOO. According to the petitioners, in many cases at well sites and at other locations, emissions cannot be estimated until the storage vessel is in operation, given the uncertainties in flowrate and other characteristics of the liquid flowing to the vessel. When a new well comes online, even at a location where wells are already in production, liquids from the new well can have significantly different characteristics than liquids from the existing wells. Western Energy Alliance ATTACHMENT A: Page 2 of 16

7 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language The IPAA letter on NSPS OOOO provided the following rationale for allowing operators time after start of production: As currently proposed, owners and operators of Group 2 storage vessels must determine their VOC emissions by April 15, 2014 or 30 days after startup, whichever is later. Id (c)(1) and (2). If VOC emissions are projected to be equal or greater than 6 TPY, then controls must be installed by April 15, 2014 or 60 days after startup, whichever is later. Id (d). These time periods are simply too short. At a minimum, 90 days is necessary to conduct the required emissions calculation and install controls. The first 30 days of production normally are not representative of stabilized production from a well, and are subject to variation that could result in the overestimation or underestimation of the emissions from storage vessels associated with that well. Thus, at least 45 days is needed to evaluate and accurately calculate projected annual emissions from a storage vessel. Another 45 days again, at a minimum would be needed to engage a contractor and install the necessary controls. Providing a total of 90 days to make the initial emissions determination and install any necessary controls will ensure a more reliable emissions estimate and afford the regulated community sufficient time to contract for the testing/modeling of emissions and installation of controls. Accordingly, IPAA recommends that EPA extend this compliance period to 90 days. CDPHE s partial adoption of NSPS OOOO had a generic explanation in their preamble: Second, the Division proposes to adopt the requirements for storage vessels at well sites, associated with exploration and production, only after the first 90 days of production has occurred. This is consistent with the Division s approach towards exploration and production activities, allowing owners and operators time to determine if exploration and production activities will result in reportable emissions. P6 Section I: GENERAL PROVISIONS A 30-day public comment period will be held in accordance with UAC R A notification of the intent to approve will be published in the Salt Lake Tribune and Deseret News on February 25, 2014; the Sun Advocate on February 27, 2014; Times Independent on February 27, 2014; Uintah Basin Standard on February 25, 2014; and the Vernal Express on February 26, During the public comment period the proposal and the evaluation of its impact on air quality will be available for the public to review and provide comment. If anyone so requests a public hearing within 15 days of publication, it will be held in accordance with UAC R The hearing will be held as close as practicable to the location of the source. Any comments received during the public comment period and the hearing will be evaluated. The proposed conditions of the AO may be changed as a result of the comments received. I.1 All definitions, terms, abbreviations, and references used in this Western Energy Alliance ATTACHMENT A: Page 3 of 16

8 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language GAO conform to those used in the Utah Administrative Code (UAC) Rule 307 (R307) and Title 40 of the Code of Federal Regulations (40 CFR). Unless noted otherwise, references cited in these GAO conditions refer to those rules. [R ] I.2 The limits set forth in this GAO shall not be exceeded without prior approval. [R ] I.3 Modifications to the equipment or processes approved by this GAO that could affect the emissions covered by this GAO must be reviewed and approved in accordance with UAC R [R ] I.4 All records referenced in this GAO or in other applicable rules, which are required to be kept by the owner/operator, shall be made available to the Director or Director's representative upon request, and the records shall include the two-year period prior to the date of the request. Unless otherwise specified in this GAO or in other applicable state and federal rules, records shall be kept for a minimum of two (2) years. [R ] I.5 At all times, including periods of startup, shutdown, and malfunction, owners and operators shall, to the extent practicable, maintain and operate any equipment approved under this GAO including associated air pollution control equipment in a manner consistent with good air pollution control practice for minimizing emissions. Determination of whether acceptable operating and maintenance procedures are being used will be based on information available to the Executive Secretary which may include, but is not limited to, monitoring results, opacity observations, review of operating and maintenance procedures, and inspection of the source. All maintenance performed on equipment authorized by this GAO shall be recorded. [R ] I.6 The owner/operator shall comply with UAC R General Requirements: Breakdowns. [R ] I.7 The owner/operator shall comply with UAC R Series. Inventories, Testing and Monitoring. [R ] I.8 The owner/operator shall comply with UAC R (4), General Approval Order: Application, and receive approval according to R (5), General Approval Order: Approval, to become subject to this GAO. [R ] Section II: II.A The approved installations shall consist of the following SPECIAL equipment: We propose UDAQ include language that specifies that once a facility recognizes that it will not meet the GAO requirements, it will apply for a NOI within a certain time. It is not possible to maintain all equipment during periods of malfunction. Additionally, this language should be consistent with the language suggested by the Alliance for the Preliminary Draft for Discussion Purposes, General Provisions, R proposed on July 30, Unless the term malfunction is removed, operators will not be able to use the GAO and will have to file NOIs for individual approval orders (AO). Operators installing new equipment in the Uinta Basin attempt to consolidate sites and equipment as much as possible to increase operational efficiency, decrease surface disturbance and reduce their Western Energy Alliance ATTACHMENT A: Page 4 of 16 If an owner or operator finds that they are exceeding the The limits set forth in this GAO shall not be exceeded without prior approval, the owner or operator shall be covered under the GAO, but shall apply for an NOI. [R ] At all times, includingduring periods of startup, shutdown, and malfunction, owners and operators shall, to the extent practicable, shutdown the facility and any required maintain and operate any equipment approved under this GAO including associated air pollution control equipment under this GAO must be maintained and operated in a manner consistent with good air pollution control practices for minimizing VOC emissions. Determination of whether acceptable operating operation and maintenance procedures are being used will be based on information available to the Executive Secretary which may include, but is not limited to, monitoring results, opacity observations, review of operating and maintenance procedures, and inspection of the source.

9 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language PROVISIONS environmental impact. Both EPA and BLM are also encouraging this trend through their regulations and requirements. By focusing on prescriptive capacity requirements, UDAQ is actually discouraging consolidation in the GAO. We suggest site-wide emission limits rather than prescriptive equipment specifications. II.A.1 II.A.2 Crude Oil and Natural Gas Well Site and/or Tank Produced Fluids Storage Tanks Contents: Crude Oil, Condensate, and/or Produced Water Maximum Site-Wide Capacity: 2,200 barrels Maximum Individual Capacity: 550 barrels Throughout the stakeholder process, Alliance members hoped that this GAO would be developed in a manner in which coverage could be sought for all types of oil and gas sites. We have concluded, however, that the scope of the coverage is limited to tank batteries with 1 or 2 wells, unlike as indicated by the title and abstract, and our many conversations with UDAQ permitting staff. With that in mind, we scoped our comments to reflect that limited application. We first identify four elements of the GAO that render it unusable (highlighted in green in this comment table). Additionally, in this table you will find other concerns we have with this GAO. The proposed GAO has a total site-wide produced fluids capacity of 2,200 bbls and max individual or emergency/overflow tank capacity of 550 bbls. Operators are moving towards larger tank batteries in an effort to consolidate their operations and reduce their surface impacts. By limiting the site-wide capacity to 2200 bbls, UDAQ is discouraging consolidation of tank batteries, which is counter to EPA s recent NSPS OOOO rule for storage vessels. Even if we apply the constraints we view as the scope of this GAO of tank batteries with 1 or 2 wells, we recommend a site-wide storage capacity of at least 3000 bbls. Currently, a typical production location is comprised of 3-4 tanks. A site-wide capacity limit of 3,000 bbls allows for these typical locations with a small contingency for up to 2 additional 500 bbl tanks for smaller centralized batteries or batteries located at multi-well pads. Crude Oil, Condensate, and/or Produced Water Maximum Site-Wide Capacity: 2,200 3,000 barrels Maximum Individual Capacity: barrels Western Energy Alliance also suggests removing limits on individual tank sizes. Some operators are moving to 600 bbl tanks, and operational flexibility can be retained without increasing site-wide emissions. II.A.3 Dehydrators: Maximum Site-Wide Capacity: 2.0MMscf/day The draft proposed GAO has capacity limit of 2.0 MMscf/day and we support that approach. The most common sizes of field natural gas dehydration units range from 1 to 2 MMscf/day. Establishing a maximum capacity of 2 MMscf/day would encompass most field installations and will also coordinate with the MACT HH applicability threshold of 2 MMscf/day. Although many of the dehydrators in the basin are currently at or below 1.0 MMscf/day, operators are often installing 2.0 MMscf/day dehydrators, which allows for greater site consolidation and reduced surface impact. Additionally, data demonstrates that the emissions profile differences between a 1.0 MMscf/day unit and a 2.0 MMscf/day unit are negligible (Dehydrator Calculations, Attachment G). Attachment G: Dehydrator Calculations II.A.4 VOC Control Device: Minimum Control Efficiency: 98% We understand that UDAQ does not intend to apply 98% efficiency to VRUs, however we feel the language in the current draft of the GAO is ambiguous in this regard and we request clarifying language be inserted. VOC Combustion Control Device: Minimum Control Efficiency: 98% For background, during the development of NSPS OOOO, EPA clearly disagreed with comments asserting that 98% control is technically achievable on a continuous basis (EPA Response to NSPS OOOO Comments) further states that data clearly supports that other technologies can only achieve 95% reduction (EPA, Attachment H). While 98% is achievable for some combustion devices such as flares and vapor combustors, other existing and innovative technologies may not be able to achieve 98%. The 98% control requirement used in generic terms could reduce operational flexibility and Western Energy Alliance ATTACHMENT A: Page 5 of 16

10 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language require operators to flare, which causes further emissions. It also discourages innovation of new control technology that could eliminate the emissions associated with flaring. Western Energy Alliance further asserts that in some applications the use of a combustion device is not cost effective. II.A.5 II.A.6 II.A.7 II.A.8 Natural Gas-Driven Pneumatic Controllers Natural Gas-Driven Pneumatic Pumps Truck Loading Operations Pumpjack, Gas Lift, and Generator Engines: Maximum Site-Wide Rating: 130hp, Fuel: Natural Gas or LPG Attachment H: EPA GASTAR VRU Lessons Learned For this engine capacity, we understand that Utah has state-only regulations addressing toxic air pollutant (TAP) impacts. The Utah regulations include requirements for impact screening assessments that are dependent upon TAP emission rates. The Utah GAO is written to keep TAP emissions below the screening levels, assuring that more complex analyses are not necessary. The Utah TAP screening requirements limit site-wide horse-power capacity to 130 horse-power, with higher horse-power capacities triggering complex site-specific analyses. As such, the site-wide horse-power limits are driven by state-only screening criteria. It is possible to have higher site-wide horse-power capacity in a Utah permit, but in Utah the higher permit limits would require site-specific analyses that go beyond the scope of doing a generic screening analysis for a general permit. Pumpjack, Gas Lift, and Generator Engines: Maximum Site-Wide Rating: 130hp, Fuel: Natural Gas or LPG However, the draft proposed GAO limits of 130hp forces operators away from consolidation of sites and equipment. It should be noted that consolidated sites would have more engine controls and thereby, fewer emissions per horsepower. In addition, the GAO requires all engines to meet EPA NSPS JJJJ requirements, which are the same for an engine equal to 100hp and an engine less than or equal to 500hp. II.A.9 Various Boilers/Heaters: Maximum Site-Wide Capacity: 10.0 MMBtu/hr combined, Fuel: Natural Gas or LPG II.A.10 Methanol &Ethylene Glycol Storage Vessels: Maximum Site-Wide Capacity: 1,000 gallons combined Also, many tank batteries are in remote areas with no power infrastructure available, and the addition of a vapor recovery unit (VRU), as may be required by other sections of the GAO, could increase onsite horsepower needs. As a result of these cited issues, the site-wide horsepower limit of 130hp will severely limit the number of GAO eligible sites. Additionally, in some areas due to the BTU content of the gas, natural gas or LPG would not be feasible and other fuels would need to be considered. Unless we apply the constraints we view as the scope of this GAO for tank batteries with 1 or 2 wells, this capacity would need to be increased to allow for other oil and gas equipment. The proposed GAO has a total site-wide methanol and glycol storage capacity of 1,000 gallons and for tank batteries with 1 or 2 wells, we agree with this approach. Where present, methanol tanks and glycol storage tanks are typically 500 gallons. This is a standard size in the industry and a standard size provided by the methanol and ethylene glycol suppliers who also frequently provide us with the tanks to store their product. A larger site-wide capacity may be needed for sites outside that narrowed scope. II.A.11 Heater Treaters: Oil/Water Separator, - listed for informational purposes only - Furthermore, emissions from methanol and glycol tanks are negligible so limiting the site-wide capacity is unnecessary to for emissions reduction. For example, one operator calculated the annual emissions from glycol and methanol tanks to be 0.02 pounds/year and 8 pounds/year, respectively, under typical operations (Storage Tank Emissions Calculations, Attachment I). We suggest these tanks be treated as ancillary equipment listed for informational purposes and that a site-wide storage capacity limit not be included for them. Attachment I: Storage Tank Emissions Calculations Western Energy Alliance ATTACHMENT A: Page 6 of 16

11 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language II.A.12 Compressors & Pumps: centrifugal and/or reciprocating, - listed for informational purposes only - II.A.13 One (1) Emergency/Overflow Storage Tank: Maximum Capacity: 550 barrels, - listed for informational purposes only - II.B Requirements and Limitations II.B.1 Site-Wide Requirements II.B.1.a The owner/operator shall not exceed 50,000 barrels (1 barrel = 42 gallons) of crude oil and condensate throughput combined per rolling 12-month period. [R ] (See comments on capacities, decreased surface disturbance and reduced environmental impact above.) The proposed GAO 50,000 bbl/year limit may not be suitable for some horizontal wells that produce more or for multi-well tank batteries. 50,000 bbl/year equates to ~137 bbl/day, and this limit will not be practical in the future given the push for centralized tank batteries and tank emission controls. As we have mentioned before, industry prefers a site-wide emissions limit, but if a throughput limit is required, we request that it be increased to accommodate the trend of increased consolidation. If we apply the constraints we view as the scope of this GAO for tank batteries with 1 or 2 wells, although it will limit the number of GAO eligible sites we support the use of 50,000 bbls. II.B.1.a.1 II.B.1.b II.B.1.c II.B.1.d II.B.1.d.1 To determine compliance with a rolling 12-month total, the owner/operator shall calculate a new 12-month total by the twentieth day of each month using data from the previous 12 months. Records of crude oil and condensate throughput shall be kept for all periods when the plant is in operation. Crude oil and condensate throughput shall be determined by process flow meters, load tickets, sales meters, and/or sales records. The records of crude oil and condensate throughput shall be kept on a monthly basis. [R ] All gas produced from the Heater Treater shall either be used as fuel on site or be routed to a gas gathering system and sent off site. [R ] A sign shall be located at the site entrance that indicates the presence of oil and gas operations and the potential for exposure to emissions from oil and gas operations. [R ] Unless otherwise specified in this GAO, visible emissions from any stationary or fugitive emission source on site shall not exceed 10 percent opacity. [R ] Unless otherwise specified in this GAO, opacity observations of fugitive and non-fugitive emissions from stationary sources shall be conducted in accordance with 40 CFR 60, Appendix A, Method 9. For intermittent sources and mobile sources, opacity observations shall be conducted using Method 9; however, the requirement for observations to be made at 15 second intervals over a six-minute period shall not apply. [R ] UDAQ lacks any cost benefit analysis or BACT analysis demonstrating the cost effectiveness of applying this requirement to all new and existing oil and gas sites in all areas and modes of operation. If we assume gathering infrastructure is available and apply this requirement only to new installations during normal operations, this requirement could potentially be technically practicable and cost effective in some instances, however, this is not a cost effective measure for existing facilities in all oil and gas applications during all modes of operation.. Western Energy Alliance asserts that even UDAQ s own rules allow for 20% opacity. Unless UDAQ has an analysis justifying the lowering of the opacity for oil and gas, we recommend that the GAO remain consistent with R We suggest striking this requirement and instead reference R For new installations, Aall gas produced during normal operations from the Heater Treater shall either be used as fuel on site, flared or be routed to a gas gathering system and sent off site. [R ] Unless otherwise specified in this GAO, visible emissions from any stationary or fugitive emission source on site shall not exceed 10 percent opacity. [R ] Unless otherwise specified in this GAO, opacity observations of fugitive and non-fugitive emissions from stationary sources shall be conducted in accordance with 40 CFR 60, Appendix A, Method 9. For intermittent sources and mobile sources, opacity observations shall be conducted using Method 9; however, the requirement for observations to be made at 15 Western Energy Alliance ATTACHMENT A: Page 7 of 16

12 II.B.1.e II.B.1.f II.B.1.g II.B.2.a ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language second intervals over a six-minute period shall not apply. [R ] The owner/operator shall notify the Director in writing when the equipment listed in this GAO has been installed and is operational within 30 days after startup. To ensure proper credit when notifying the Director, send your correspondence to the Director, attn: Compliance Section. If the owner/operator has not notified the Director in writing of the installation and operation of the equipment listed in this GAO within 18 months of a source being granted approval under this GAO, the owner/operator shall submit documentation of the continuous construction and/or installation of the operation to the Director. If a continuous program of construction and/or installation is not proceeding, the Director may require the source to submit a NOI according to R [R ] The owner/operator shall submit a list of the actual equipment installed on site and the potential emissions from this equipment to the Director within 180 days after startup. [R ] The owner/operator shall submit an annual inventory of the actual equipment on site and the actual emissions from the site to the Director on or before April 15 of each year following the first full calendar year of operation. [R ] II.B.2 Tank Requirements VOC emissions from the produced fluids storage tanks shall either be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered or be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] The term startup is not clearly defined, and we suggest replacing it with commencement of normal operation. The term startup is not clearly defined, and we suggest replacing it with commencement of normal operation. Western Energy Alliance asserts that UDAQ lacks any cost benefit analysis or BACT analysis for this requirement demonstrating the cost effectiveness. We are unaware of any Federal or State requirement that requires controls on all tanks regardless of size or content. It should also be noted that the air shed would likely observe more emissions from the control device than total emissions for smaller tanks. Western Energy Allaince proposes that this requirement be stricken from the GAO. If a tank control requirement is pursued, a selection of an appropriate de minimis threshold based on technical and economic feasibility is necessary. The owner/operator shall notify the Director in writing when the equipment listed in this GAO has been installed and is operational within 30 days after commencement of normal operationstartup. The owner/operator shall submit a list of the actual equipment installed on site and the potential emissions from this equipment to the Director within 180 days after commencement of normal operationstartup. [R ] VOC emissions from the produced fluids storage tanks shall either be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered or be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] II.B.2.b II.B.2.b.1 At least once every month, the thief hatches on the produced fluids storage tanks shall be inspected to ensure the thief hatches are closed and latched and the associated gaskets, if any, are in good working condition. If the gaskets are not in good working condition, they shall be replaced within 15 days of identification of the deficient condition. [R ] Records of thief hatch inspections shall include the following: The date of the thief hatch inspection, The status of the thief hatches, Any corrective action taken, and The date of corrective action. [R ] II.B.3 Dehydrator Requirements Unless a de minimis is applied, operators will not be able to use the GAO and will have to file NOIs for individual approval orders (AO). Western Energy Alliance ATTACHMENT A: Page 8 of 16

13 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language II.B.3.a VOC emissions from dehydrators shall either be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered or be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] Western Energy Alliance asserts that UDAQ lacks any cost benefit analysis or BACT analysis for this requirement demonstrating the cost effectiveness. We are unaware of any Federal or State requirement that requires dehy controls on all units regardless of size. It should also be noted that the air shed would likely observe more emissions from the control device than total emissions for smaller dehys. Western Energy Alliance proposes that this requirement be stricken from the GAO. If a dehy control requirement is pursued, a selection of an appropriate de minimis threshold based on technical and economic feasibility is necessary. VOC emissions from dehydrators shall either be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered or be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] II.B.4.a II.B.4.a.1 II.B.4.a.2 II.B.4.b II.B.4.b.1 II.B.5.a II.B.4 VOC Control Device Requirements Any VOC control device shall have a control/destruction efficiency of no less than 98%. [R ] To show compliance with the control/destruction efficiency, the VOC control device shall be operated according to the manufacturer's written instructions when gases/vapors are vented to it. [R ] The owner/operator shall keep and maintain records of the following: The VOC control device's control/destruction efficiency guaranteed by the manufacturer, The manufacturer's written operating and maintenance instructions, and The date and type of any maintenance conducted by the owner/operator. [R ] The VOC control device shall operate with no visible emissions. [R ] Visual determination of emissions from the VOC control device shall be conducted according to 40 CFR 60, Appendix A, Method 22. [R ] II.B.5 Natural Gas-Driven Pneumatic Controller Requirements Each natural gas-driven pneumatic controller shall comply with either a or b: a. A natural gas-driven pneumatic controller shall have a bleed Unless a de minimis is applied, operators will not be able to use the GAO and will have to file NOIs for individual approval orders (AO). [SEE COMMENTS IN II.A.4.] [SEE COMMENTS IN II.A.4.] [SEE COMMENTS IN II.A.4.] [SEE COMMENTS IN II.A.4.] Western Energy Alliance questions the notion that a combustion device can technically be operated at all times with no visible emissions. We suggest looking to other states for solutions that are workable. For example, most governing bodies, including UDAQ allow for a minimal time period for visible emissions to be present given the nature of these devices. Examples include, but are not limited to, periods in excess of 1 minute in any 15 minute period. Additionally, it is not practical to apply these visible emissions requirements outside of normal operation. The GAO does not specify the type of pneumatic controllers that are subject to the requirements. We suggest UDAQ limit the GAO requirements to continuous bleed pneumatic controllers. Recognizing the difficulty of determining emissions from intermittent controllers, EPA did not regulate intermittent II.B.4 VOC Combustion Control Device Requirements Any VOC combustion control device shall have a control/destruction efficiency of no less than 98%. [R ] To show compliance with the control/destruction efficiency, the VOC combustion control device shall be operated according to the manufacturer's written instructions when gases/vapors are vented to it. [R ] The owner/operator shall keep and maintain records of the following: The VOC combustion control device's control/destruction efficiency guaranteed by the manufacturer, The manufacturer's written operating and maintenance instructions, and The date and type of any maintenance conducted by the owner/operator. [R ] The VOC control device shall operate with no visible emissions at a duration greater than or equal to 1 minute in any 15 minute period during normal operation. Visible emissions do not include radiant energy or water vapor. [R ] Visual determination of emissions from the VOC control device shall be conducted according to 40 CFR 60, Appendix A, Method 22. [R ] II.B.5 Continuous Bleed Natural Gas- Driven Pneumatic Controller Requirements Each continuous bleed natural gas-driven pneumatic controller shall comply with either a or b: Western Energy Alliance ATTACHMENT A: Page 9 of 16

14 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language rate less than or equal to 6 standard cubic feet per hour and shall comply with 40 CFR (d). b. The VOC emissions from a natural gas-driven pneumatic controller shall either: i. be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered; or ii. be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] bleed emissions related to process control in NSPS OOOO. (EPA Response to NSPS OOOO Comments). The GAO requirements for pneumatic controllers go far beyond what is required of operators in NSPS OOOO as pneumatic pumps are not even covered in NSPS OOOO. Pneumatic pumps are not covered in NSPS OOOO. Additionally, UDAQ lacks any cost benefit analysis or BACT analysis on requiring controls for pneumatic pumps demonstrating the cost effectiveness of such a requirement. We request UDAQ eliminate specific requirements for pneumatic pumps and reiterate our request that only continuous bleed controllers be subject to requirements under the GAO. We are also concerned about the inconsistent requirement in Sec. II.B.5.a.a that controllers shall have a bleed rate less than or equal to 6 standard cubic feet per hour [scf/hr] and shall comply with 40 CFR (d). If a pneumatic controller bleed rate is less than 6 scf/hr, it is not subject to the referenced Federal regulation. Similarly, the GAO requires compliance with 40 CFR (d) in Section II.B.6.a.a for pneumatic pumps; however, pneumatic pumps are not subject to 40 CFR Moreover, that specific rule was not designed for pneumatic pumps. In light of these conflicts for both controllers and pumps, we suggest UDAQ remove all reference to 40 CFR a. A natural gas-driven pneumatic controller shall have a continuous bleed rate less than or equal to 6 standard cubic feet per hour and shall comply with 40 CFR (d). b. The VOC emissions from a continuous bleed natural gas-driven pneumatic controller shall either: i. be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered; or ii. be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] II.B.6.a II.B.6 Natural Gas-Driven Pneumatic Pump Requirements Each natural gas-driven pneumatic pump shall comply with either a or b: a. A natural gas-driven pneumatic pump shall have a bleed rate less than or equal to 6 standard cubic feet per hour and shall comply with 40 CFR (d). b. The VOC emissions from a natural gas-driven pneumatic pump shall either: i. be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered; or ii. be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] Unless applicability of this requirement is limited to continuous bleed pneumatic controllers, operators will not be able to use the GAO and will have to file NOIs for individual AOs. The GAO requires compliance with 40 CFR (d) in Section II.B.6.a.a for pneumatic pumps; however, pneumatic pumps are not subject to 40 CFR Moreover, that specific rule was not designed for pneumatic pumps. In light of this conflict for controllers, we suggest UDAQ remove all reference to 40 CFR Pneumatic pumps are not covered in NSPS OOOO. Additionally, UDAQ lacks any cost benefit analysis or BACT analysis on requiring controls for pneumatic pumps demonstrating the cost effectiveness of such a requirement. We request UDAQ eliminate specific requirements for pneumatic pumps and reiterate our request that only continuous bleed controllers be subject to requirements under the GAO. Regardless of application to our narrowed scope of tank batteries with 1 or 2 wells, operators will not be able to use the GAO and will have to file NOIs for individual approval orders (AO). Each natural gas-driven pneumatic pump shall comply with either a or b: a. A natural gas-driven pneumatic pump shall have a bleed rate less than or equal to 6 standard cubic feet per hour and shall comply with 40 CFR (d). b. The VOC emissions from a natural gasdriven pneumatic pump shall either: i. be routed to a process unit where the emissions are recycled, incorporated into a product, and/or recovered; or ii. be routed to a VOC control device where the emissions are consumed and/or destroyed. [R ] II.B.7.a II.B.8.a II.B.8.b II.B.7 Truck Loading Requirements The owner/operator shall load the tanker trucks on site by the use of submerged loading or bottom fill loading. [R ] II.B.8 Engine Requirements Any stationary engine on site shall only use natural gas or LPG as fuel. [R ] Any stationary engine on site shall comply with the following emission standards: a. For engines rated less than 100 hp: [40 CFR (c)], HC+NO x = 3.8 g/kw-hr (2.84 g/hp-hr), CO = 6.5 g/kw-hr (4.85 g/hp-hr), UDAQ is attempting to apply BACT to all engines, not just new or modified engines. The Alliance questions UDAQ s authority to require controls on engines that are not new or modified. UDAQ references NSPS JJJJ in Section II.B.8 Engine Requirements but the GAO is far more stringent than NSPS JJJJ. It should be noted that in NSPS JJJJ EPA took careful consideration of manufacture date, engine type (lean or rich) and capacity to ensure that the resulting requirements incorporated Any stationary engine on site shall comply with the following emission standards: a. For engines rated less than 100 hp: [40 CFR (c)], HC+NO x = 3.8 g/kw-hr (2.84 g/hp-hr), CO = 6.5 g/kw-hr (4.85 Western Energy Alliance ATTACHMENT A: Page 10 of 16

15 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language b. For engines rated greater than or equal to 100 hp: [40 CFR 60 Subpart JJJJ - Table 1] NO x = 1.0 g/hp-hr, CO = 2.0 g/hp-hr, VOC = 0.7 g/hp-hr. [40 CFR 60 Subpart JJJJ, R ] appropriate cost effectiveness and technical feasibility. Section II.B.8.b.b references Table 1 of NSPS JJJJ (EPA, Attachment K) before listing emissions standards for NO x, CO and VOCs. The standards listed in the GAO are only applicable to engine manufactured after January 1, 2011 in NSPS JJJJ. In some cases, it is not technically feasible to modify an engine to meet these stringent emissions standards. Where it is technically feasible, it is often cost prohibitive to modify an engine. Therefore, EPA does not apply one set of emission standards to all engines. A specific example of cost prohibitiveness is that Ajax has developed the E-565 JJJJ compliant model, which is a new version of their old non-jjjj model E-42 engine. The new equivalent of the E-42 is the E565, which costs approximately $35,000. Replacing the E-42 with a new E-565 in order to meet NSPS JJJJ standards would result in a reduction in emissions of 0.65 tons per year of NOx, which equates to $54,000 per ton of NOx, which is not cost effective. This would place an unacceptable financial burden on the operator. At a foreseeable pace of adding 100 new wells per year, that is a potential cost of $3.5 million. Another option is to install a control system on an existing non-jjjj compliant engine, but that can also be cost prohibitive. One specific example would be the AFR controller system for 5.9L Cummins engines. Based on an estimate for work for one of our members, the cost to install this system on one engine, including equipment and labor, is $10,400. This does not include the costs of bringing a power supply to the engine site, which would be a common requirement in some remote areas, or other ancillary costs associated with downtime or travel to the site. We suggest UDAQ strike the Section II.B.8.b and refer to Section III of the GAO, which includes NSPS JJJJ in the list of federal requirements with which owners/operators must comply. If UDAQ retains such stringent emissions standards for all engines, our members will not use this GAO and will be required to file NOIs for individual AOs. g/hp-hr), b. For engines rated greater than or equal to 100 hp: [40 CFR 60 Subpart JJJJ - Table 1] NO x = 1.0 g/hp-hr, CO = 2.0 g/hp-hr, VOC = 0.7 g/hp-hr. [40 CFR 60 Subpart JJJJ, R ] See Section III: Applicable Federal Requirements II.B.8.b.1 II.B.8.c The owner/operator shall keep and maintain the following records: a. The emission rate guaranteed by the manufacturer for: HC+NO x and CO for engines rated less than 100 hp, or NO x, CO, and VOC for engines rated greater than or equal to 100 hp, b. The manufacturer's written operating and maintenance instructions, c. Any maintenance conducted by the owner/operator, and d. The date of the maintenance activities. [R ] Each stationary engine stack on site shall vent no less than 4 feet above ground level. [R ] Attachment K: NSPS JJJJ Table 1 Older engines may not have manufacturer guarantees. Western Energy Alliance would suggest alternatives such as manufacturer publications or estimates are included. An alternative would be to take into consideration the engine type; lean or rich burn. Engine manufacturer data confirms that extending exhaust stacks from the pumpjack engines is not technically feasible. Modeling analyses that indicate tall stack heights are required for the engines to comply with the NO2 NAAQS are based on overly conservative model assumptions (Redhorse Modeling Study, Attachment F). For more explanation and detail, see comments on Abstract Paragraph 1 above). For engines subject to 40 CFR 60 Subpart JJJJ, Tthe owner/operator shall keep and maintain the following records: a. The emission rate estimated, published or guaranteed by the manufacturer for: HC+NO x and CO for engines rated less than 100 hp, or NO x, CO, and VOC for engines rated greater than or equal to 100 hp, b. The manufacturer's written operating and maintenance instructions, c. Any maintenance conducted by the owner/operator, and d. The date of the maintenance activities. [R ] Western Energy Alliance ATTACHMENT A: Page 11 of 16

16 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language II.B.9 Boilers/Heater Requirements II.B.9.a All boilers/heaters on site shall only use natural gas or LPG as fuel. [R ] II.B.9.b Each boiler stack and each heater stack on site shall vent at least 1 foot above the height of the Produced Fluids Storage Tanks. [R ] II.B.10 Leak Detection and Repair Requirements II.B.10.a The owner/operator shall conduct an inspection of each valve, flange or other connection, pump, compressor, pressure relief device or other vent, process drain, open-ended valve, pump seal, compressor seal, and access door seal or other seal that contains or contacts a process stream with hydrocarbons according to the following schedule: a. No later than 90 days after startup. b. For sources with at least one crude oil or condensate storage tank on site: 1. At least once every 12 months, for sources that have a projected annual throughput of crude oil and condensate combined that is greater than or equal to 10,000 barrels, 2. At least once every 3 months after the initial inspection, for sources that have a projected annual throughput of crude oil and condensate combined that is greater than or equal to 25,000 barrels. Inspection frequency, for sources that have a projected annual throughput of crude oil and condensate combined that is greater than or equal to 25,000 barrels, shall change according to the following: i. If no leaks are detected during inspections for one year, inspection frequency shall be reduced to at least once every 6 months, ii. If no leaks are detected during inspections for two years, inspection frequency shall be reduced to at least once every 12 months, iii. If two or more leaks are detected during any inspection, inspection frequency shall be conducted at least once every 3 months, c. At least once every 12 months, for sources that do not have a crude oil or condensate storage tank on site. [R ] If stack heights are so high as to be technically infeasible, operators will not be able to use the GAO and will have to file NOIs for individual approval orders (AO). Modeling analyses that indicate tall stack heights to comply with the NO2 NAAQS are based on overly conservative model assumptions (Redhorse Modeling Study, Attachment F). For more explanation and detail, see comments on Abstract Paragraph 1 above). Section II.B.10.a.a requires all inspections of virtually every connection at a well site or tank battery be completed within 90 days of start-up. The term start-up is not clearly defined, and we suggest replacing it with commencement of normal operation. All Federal LDAR programs, including KKK and refinery LDAR programs, allow for a reduced monitoring frequency if it is demonstrated that a site has very little potential for leaks. Western Energy Alliance proposes that an alternative inspection frequency may be applied if a company is able to demonstrate a specified low number of leaks in their monitoring history (two consecutive monitoring events). The requested changes are performance based and the inspection frequencies revert back to the original frequencies if the number of leaks identified rise above the identified thresholds; ensuring production facilities maintain a low number of new component leaks. Western Energy Alliance also asserts that a standard of no leaks would render this reduced frequency alternative unusable. As KKK allows, we propose that a 2% or less or 10 or less components is an appropriate standard. Western Energy Alliance also proposes that the no tank monitoring requirement, paragraph c., be stricken as the potential for leaks at these sites would fall well below any cost effective criteria and be conducted for little benefit. a. No later than 90 days after commencement of normal operationstartup. b. For sources with at least one crude oil or condensate storage tank on site: 1. At least once every 12 months, for sources that have a projected annual throughput of crude oil and condensate combined that is greater than or equal to 10,000 barrels, 2. At least once every 3 6 months after the initial inspection, for sources that have a projected annual throughput of crude oil and condensate combined that is greater than or equal to 25,000 barrels. Inspection frequency, for sources that have a projected annual throughput of crude oil and condensate combined that is greater than or equal to 25,000 barrels, shall change according to the following: i. If no 10 or less component leaks or 2% or less of total components are detected during inspections for one yearin each of 2 monitoring events, inspection frequency shall be reduced to at least once every 6 months, ii. If no leaks are detected during inspections for two years, inspection frequency shall be reduced to at least once every 12 months, iii. If two or more leaks are detected during any inspection at a number greater than i. above, inspection frequency shall be conducted at leastresume at once every 36 months, At least once every 12 months, for sources that do not have a crude oil or condensate storage tank on site. [R ] II.B.10.a.1 Inspections shall be conducted with an analyzer meeting U.S. EPA We request flexibility in the method used to detect leaks and suggest adding or any approved method Inspections shall be conducted with an analyzer Western Energy Alliance ATTACHMENT A: Page 12 of 16

17 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language Method 21, 40 CFR Part 60, Appendix A, a tunable diode laser absorption spectroscopy (TDLAS), or an infrared camera that can detect hydrocarbons. to Section II.B.10.a.1 of the GAO. As technology changes and given the diversity of our companies operations we ask to UDAQ to leave open the possibility of using other instruments to detect leaks. meeting U.S. EPA Method 21, 40 CFR Part 60, Appendix A, a tunable diode laser absorption spectroscopy (TDLAS), any Executive Secretary approved instrument based monitoring device or method, or an infrared camera that can detect hydrocarbonsleaks as defined in this section, except as provided in A reading of 500 ppm or greater with an analyzer or a TDLAS shall be considered a leak. Any emissions detected with an infrared camera shall be considered a leak unless the owner/ operator evaluates the leak with an analyzer meeting U.S. EPA Method 21, 40 CFR Part 60, Appendix A no later than 5 calendar days after detection and the analyzer's reading is less than 500 ppm. Emissions detected from tank gauging, load-out operations, or other maintenance activities shall not be considered leaks. [R ] The proposed GAO s definition of a leak is a reading of 500pppm or greater with an analyzer or [tunable diode laser absorption spectroscopy]. UDAQ informed us this definition came from EPA s New Source Performance Standard for the Oil and Gas Sector, Section OOOO (NSPS OOOO); however, that leak definition applies only to gas processing plants in Quad O, not well sites or tank batteries. EPA s definition of a leak for pumps, valves, and connectors, which apples to well sites and tank batteries is 10,000ppm or greater, found in NSPS, Sec. VVa (40 CFR a(2)(6)). Under most New Source Performance Standards, including subpart KKK and VVa, leaks are defined as greater than 10,000 ppm and under those programs, fugitive emissions below 10,000 ppm are not leaks, and do not require repair. 40 C.F.R. Part 60, Subparts KKK and VV. Additionally, Alberta has a well-established LDAR program where operators use a screening value of 10,000 ppm to determine if a component is leaking and thus may warrant repair (CAPP, Attachment J). Also, based on data provided by API member companies for several natural gas processing plants that are currently subject to 40 CFR 60 Subpart KKK, the percent of valves that leak above 10,000 ppm ranges between %; however, these leaking valves contribute 82-99% of the total mass emissions from facility valves when using the Leak/No-Leak method from EPA s 1995 protocol for estimating emissions. This is consistent with an earlier API study (API # 310, November 1997) of petroleum refineries equipment leaks that showed that 92% of reducible emissions are due to only ~ 0.13% of components. For all these reasons, Western Energy Alliance believes that 10,000ppm leak definition is protective and appropriate for this GAO. II.B.10.a.2. A reading of ,000 ppm or greater with an analyzer or a TDLAS shall be considered a leak. Any emissions detected with an infrared camera shall be considered a leak unless the owner/ operator evaluates the leak with an analyzer meeting U.S. EPA Method 21, 40 CFR Part 60, Appendix A no later than 5 calendar days after detection and the analyzer's reading is less than ,000 ppm. Emissions detected from tank gauging, load-out operations, or other maintenance activities shall not be considered leaks. [R ] II.B.10.a.2 The owner/operator is exempt from inspecting a valve, flange or other connection, pump or compressor, pressure relief device, process drain, open-ended valve, pump or compressor seal system degassing vent, accumulator vessel vent, agitator seal, or access door seal under any of the following circumstances: a. the contacting process stream only contains glycol, amine, methanol, or produced water, b. monitoring could not occur without elevating the monitoring personnel more than six feet above a supported surface or without the assistance of a wheeled scissor-lift or hydraulic type scaffold, c. monitoring could not occur without exposing monitoring personnel Attachment J:, Canadian Association of Petroleum Producers, Best Management Practice, Management of Fugitive Emissions at Upstream Oil and Gas Facilities, January 2007 Western Energy Alliance ATTACHMENT A: Page 13 of 16

18 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language to an immediate danger as a consequence of completing monitoring, or d. the item to be inspected is buried, insulated in a manner that prevents access to the components by a monitor probe, or obstructed by equipment or piping that prevents access to the components by a monitor probe. [R ] II.B.10.b If a leak is detected at any time, the owner/operator shall attempt to repair the leak no later than 5 calendar days after detection. Repair of the leak shall be completed no later than 15 calendar days after detection, unless parts are unavailable or unless repair is technically infeasible without a shutdown. The owner/operator shall inspect the repaired leak no later than 15 calendar days after the leak was repaired to verify that it is no longer leaking. If replacement parts are unavailable, the replacement parts must be ordered no later than 5 calendar days after detection, and the leak must be repaired no later than 15 calendar days after receipt of the replacement parts. If repair is technically infeasible without a shutdown, the leak must be repaired by the end of the next shutdown. If a shutdown is required to repair a leak, the shutdown must occur no later than 6 months after the detection of the leak unless the owner/operator demonstrates that emissions generated from the shutdown are greater than the fugitive emissions likely to result from delay of repair. [R ] II.B.10.c Records of inspections and leak detection and repair shall include the following: a. The date of the inspection, b. The name of the person conducting the inspection, c. Any component not inspected and the reason it was not inspected d. The identification of any component that was determined to be leaking, e. The analyzer's or TDLAS reading (if used), f. The date of first attempt to repair the leaking component, g. Any component with a delayed repair, h. The reason for a delayed repair, 1. For Unavailable Parts: i. The date of ordering a replacement component, ii. The date the replacement component was received, 2. For a Shutdown: i. The reason the repair is technically infeasible, ii. The date of the shutdown, iii. Emission estimates of the shutdown and the repair if the delay is longer than 6 months, i. Corrective action taken, j. The date corrective action was completed, and k. The date the component was verified to no longer be leaking. Western Energy Alliance ATTACHMENT A: Page 14 of 16

19 ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Section Requirement Comment Alternate Language [R ] Section III: APPLICABLE FEDERAL REQUIREMENTS In addition to the requirements of this AO, all applicable provisions of the following federal programs have been found to apply to this installation. This AO in no way releases the owner or operator from any liability for compliance with all other applicable federal, state, and local regulations including UAC R307. NSPS (Part 60), A: General Provisions NSPS (Part 60), Dc: Standards of Performance for Small Industrial-Commercial-Institutional Steam Generating Units NSPS (Part 60), JJJJ: Standards of Performance for Stationary Spark Ignition Internal Combustion Engines NSPS (Part 60), OOOO: Standards of Performance for Crude Oil and Natural Gas Production, Transmission and Distribution MACT (Part 63), A: General Provisions MACT (Part 63), HH: National Emission Standards for Hazardous Air Pollutants From Oil and Natural Gas Production Facilities MACT (Part 63), ZZZZ: National Emissions Standards for Hazardous Air Pollutants for Stationary Reciprocating Internal Combustion Engines ADMINISTRATIVE CODING The following information is for UDAQ internal classification use only: State Wide County CDS B MACT (Part 63), NSPS (Part 60) ACRONYMS The following lists commonly used acronyms as they apply to this document: 40 CFR Title 40 of the Code of Federal Regulations AO Approval Order BACT Best Available Control Technology CAA Clean Air Act CAAA Clean Air Act Amendments CDS Classification Data System (used by EPA to classify sources by size/type) CEM Continuous emissions monitor CEMS Continuous emissions monitoring system CFR Code of Federal Regulations CMS Continuous monitoring system CO Carbon monoxide CO2 Carbon Dioxide CO2e Carbon Dioxide Equivalent - 40 CFR Part 98, Subpart A, Table A-1 COM Continuous opacity monitor DAQ Division of Air Quality (typically interchangeable with UDAQ) DAQE This is a document tracking code for internal UDAQ use EPA Environmental Protection Agency FDCP Fugitive Dust Control Plan GHG Greenhouse Gas(es) - 40 CFR (b)(49)(i) GWP Global Warming Potential - 40 CFR Part (a) HAP or HAPs Hazardous air pollutant(s) ITA Intent to Approve LB/HR Pounds per hour MACT Maximum Achievable Control Technology MMBTU Million British Thermal Units NAA Nonattainment Area NAAQS National Ambient Air Quality Standards NESHAP National Emission Standards for Hazardous Air Pollutants NOI Notice of Intent Western Energy Alliance ATTACHMENT A: Page 15 of 16

20 NO x Oxides of nitrogen NSPS New Source Performance Standard NSR New Source Review PM10 Particulate matter less than 10 microns in size PM2.5 Particulate matter less than 2.5 microns in size PSD Prevention of Significant Deterioration PTE Potential to Emit R307 Rules Series 307 R Rules Series Section 401 SO2 Sulfur dioxide Title IV Title IV of the Clean Air Act Title V Title V of the Clean Air Act TPY Tons per year UAC Utah Administrative Code UDAQ Utah Division of Air Quality (typically interchangeable with DAQ) VOC Volatile organic compounds ATTACHMENT A: Comment Explanations and Alternate Language Table INTENT TO APPROVE: General Approval Order for a Crude Oil and Natural Gas Well Site and/or Tank Battery, February 20, 2014 Western Energy Alliance ATTACHMENT A: Page 16 of 16

21 W E S T E R N S T A T E S A I R R E S O U R C E S C O U N C I L December 13, 2012 Mr. Richard Wayland, Director Air Quality Assessment Division Office of Air Quality Planning & Standards 109 T.W. Alexander Drive (Mail Code: C304-02) Research Triangle Park, NC Dear Mr. Wayland, The Western States Air Resources Council (WESTAR) highly commends the U.S. Environmental Protection Agency (EPA) for establishing the NO 2 /NOx in-stack ratio database on August 30, 2012 to assist in modeling compliance with the 1-hour Nitrogen Dioxide (NO 2 ) National Ambient Air Quality Standard (NAAQS). WESTAR will strongly encourage WESTAR members to participate by contributing to the database. WESTAR is also grateful for the opportunity to submit suggestions to the EPA on the 1-hour NO 2 NAAQS. Specifically, WESTAR requests that EPA:1) adopta higher significant impact level (SIL) in pending rulemaking; 2) support the use of higher interim SILs for current permitting activities by WESTAR s member agencies; and 3) assume responsibility for further field studies of isolated sources to improve AERMOD accuracy, or provide technical and financial assistance for such studies. Additionally, WESTAR is reviewing the treatment of intermittent emissions and may offer comments in the future. BACKGROUND: On April 12, 2010, the EPA enacted a new 1-hour NO 2 NAAQS. This new standard has presented state and local air agencies with a host of challenges when implementing the new standard under their New Source Review (NSR) permitting programs. EPA attempted to address the problems by issuing guidance memorandums (dated June 28 and June 29, 2010, and March 1, 2011) to provide further clarification and guidance on the application of Appendix W for this standard. Last year, WESTAR s Air Directors directed WESTAR staff to form an ad hoc committee to review the new modeling requirements, identify key issues related to its implementation, Alaska Arizona California Colorado Hawaii Idaho Montana Nevada NewMexico NorthDakota Oregon SouthDakota Utah Washington Wyoming WESTAR, rd Ave, Seattle, WA (206)

22 and to determine possible solutions to those issues, as they relate to the use of EPA dispersion model and Appendix W. In response, a 1-hour NO 2 modeling ad hoc committee was convened. The committee consists of Phil Allen (Oregon DEQ), Clint Bowman (WA DOE), Cyra Cain (MT DEQ), Tom Orth (UT DEQ), David Prey (UT DEQ), Alan Schuler (AK DEC), and Jeff Gabler (WESTAR). Many conference calls were conducted to discuss the issues. Air modeling experts from EPA and other state agency staff were, at times, included in the discussions as the group attempted to better understand the complexity of the issues associated with implementing the new standard. SIGNIFICANT IMPACT LEVEL (SIL) A problematic issue identified by the ad hoc committee relates to EPA-proposed SIL for 1- hour NO 2 of 4ppb or 7.5µg/m 3. The EPA has suggested that states can use the proposed SIL as an interim value until a new 1-hour NO 2 SIL is adopted. In its comments, EPA acknowledges that air agencies are not bound by the value, and that states may choose a different interim value until the 1-hour NO 2 SIL is finalized. Prior to EPA s recommendation, the Northeast States for Coordinated Air Use Management (NESCAUM) recommended a value of 10 µg/m 3. While EPA s choice of a SIL value is consistent with other criteria pollutant SILs (usually 1-4% of the NAAQS), the current SILs are based on 24-hour and annual-averaging periods, where multiple single-hour concentration predictions are averaged over time. Compliance with the new 1-hour NO 2 NAAQS is based on a single hour s predicted concentration within the AERMOD modeling system, the one-hour daily maximum. Single-hour predictions generally consume a greater percentage of the standard they are compared to, whereas, multi-hour predictions will be less sensitive to the standard, since the hourly contribution varies widely from hour to hour. In short, a one-hour modeling prediction is considerably more sensitive to the SILs than those based on multiple hours. The primary concern identified by the ad hoc committee is that nearly all of their permit applicants will trigger the requirement for a cumulative analysis at the proposed 1-hour NO 2 SIL level. Impacts from sources modeling under the new 1-hour NO 2 standard are generally very local to the subject source, and are less sensitive to surrounding sources, especially when sources are separated by several kilometers or more. Requiring sources to perform unnecessary cumulative analyses under these conditions consumes valuable state resources and places an undue burden on the permitting source. WESTAR recommends that EPA propose a higher 1-hour NO 2 SIL that is less sensitive to hourly predictions, or if possible, propose another methodology to determine the necessity of a cumulative analysis when modeling for the new 1-hour NO 2 NAAQS. Alaska Arizona California Colorado Hawaii Idaho Montana Nevada NewMexico NorthDakota Oregon SouthDakota Utah Washington Wyoming WESTAR, rd Ave, Seattle, WA (206)

23 In the meantime, until the 1-hour NO 2 SIL is finalized, WESTAR s member agencies will, at the states discretion, usehigher interim SIL values, or other methodologies for determining the need for a cumulative NAAQS analysis. The SIL value or methodology chosen to determine the necessity for a cumulative analysis will be left to the individual memberagency. MODEL ACCURACY EPA has limited field data to test the accuracy of AERMOD s NO 2 algorithms. A field study developed around a relatively isolated source could partially provide the data needed to advance refined NO 2 modeling techniques. The objective would be to compare the estimates from a variety of NO 2 modeling techniques to actual ambient NO 2 concentrations. WESTAR believes that field studies are critically needed to help resolve some of the 1-hour NO 2 modeling concerns, and that these field studies are EPA sresponsibility. If EPA is unwilling to accept the responsibilities associated with the field studies, WESTAR requests that EPA provide technical and financial assistance to support the studies needed to improve NO 2 modeling techniques. If you have any questions or require further clarification on our comments, please contact WESTAR Executive Director Dan Johnson at or djohnson@westar.org. Sincerely, Greg Remer, President Western States Air Resources Council Alaska Arizona California Colorado Hawaii Idaho Montana Nevada NewMexico NorthDakota Oregon SouthDakota Utah Washington Wyoming WESTAR, rd Ave, Seattle, WA (206)

24 WESTAR 1-HOUR NO 2 MODELING AD HOC COMMITTEE BACKGROUND INFORMATION: Computer models are imperfect attempts to estimate an existing or future air quality impact from a given emissions activity. The refined models promulgated by EPA, such as AERMOD, have been shown to not be biased towards underestimating air quality impacts, but according to EPA s Guideline on Air Quality Models (Guideline) errors in the highest estimated concentrations of ± 10 to 40 percent are typical. The Guideline further states, models are also more reliable for estimating longer time-averaged concentrations than for estimating short-term concentrations at specific locations. This is likely due to the averaging of both overand under-estimates that occurs with a longer period. Accurately modeling 1-hour impacts is therefore more challenging than modeling impacts over longer averaging periods. The current techniques for estimating the amount of atmospheric conversion of oxide of nitrogen (NOx) into nitrogen dioxide (NO 2 ) may also be conservative, especially in the nearfield. This apparent tendency to overestimate 1-hour NO 2 impacts, coupled with an extremely stringent 1-hour NO 2 standard, leads to hurdles that are proving difficult to overcome. Some of these difficulties are further described in this report. To demonstrate compliance with EPA s new 1-Hour NO 2 NAAQS air quality dispersion modeling analysis must be performed which shows that emissions from a source will not cause or contribute to a violation of the standard. Initial performance of air quality dispersion modeling for the 1-hour standard has found that demonstrating compliance with the new standard is challenging, and can result in significant delays and hurdles in the permitting process and in granting approvals. The 1-hour NO 2 standard is more stringent than the previous NAAQS, and as such the margin for error is smaller than it has been in the past, and when combined with the conservatism to modeling guidelines it is possible that modeled concentrations exceed the standard when monitoring indicates compliance with the standard. Such results can lead to uncertainty and unnecessary commitment of scare state resources to solve nonexistent issues. EPA is aware of the difficulties surrounding these complex issues and has attempted to address the problems by issuing guidance memorandums (dated June 29, 2010 and March 1, 2011) to provide further clarification and guidance on the application of Appendix W guidance for the 1-hour NO 2 standard. Nevertheless, WESTAR s Air Directors asked WESTAR staff to form an ad hoc committee to review the modeling requirements in order to identify the issues causing the difficulties and to determine possible solutions because without a better understanding of these issues the challenges of demonstrating compliance will continue. In response, a 1-hour NO 2 Modeling ad hoc committee was convened. The committee consisting of Phil Allen (Oregon DEQ), Clint Bowman (Washington Department of Ecology), Cyra Cain (Montana DEQ), Tom Orth (Utah DEQ), Alan Schuler (Alaska DEC), and Jeff Gabler (WESTAR). Conference calls were conducted to discuss issues. Guest speakers included EPA staff and state staff.

25 The committee has attempted to provide the best recommendations possible. However, there is no quick and optimal solution for improving model accuracy or substantively streamlining the process. Accuracy and streamlining also tend to have opposite effects. Streamlining typically leads to short-cuts at the cost of accuracy. Improving accuracy typically requires more detailed information (which takes time to collect) and typically requires longer processing times. Therefore, the general modeling difficulties associated with the 1-hour NO 2 standard will likely be around for a while. Following are the salient issues as identified by the committee are: Temporary/Portable/Intermittent/Seasonal In-Stack Ratios Significant Impact Level (SIL) Background Ambient Ozone Concentrations TEMPORARY/PORTABLE/INTERMITTENT/SEASONAL: The promulgation of the 1-Hour NO 2 standard has focused attention on the issue of intermittent versus continuously operating emission units. When demonstrating compliance with the annual standard, intermittent emissions were generally not considered significant contributions to total emissions, and an annual operating limit, such as 500 hours/year, could be included as a permit condition for many intermittent sources. As a result, modeled impacts from intermittent sources were not in general significant when compared to the annual standard. However, when intermittent emissions, for example from emergency generators or startup-shutdown operations, are modeled for compliance with the 1-Hour standard, the modeled concentrations can be high, and in many case can be significantly higher than what might be "realistically expected," in EPA's language. This is because the intermittent hourly emission rate is treated as continuous over multiple years in order to calculate the design value, which is the 98th percentile, averaged over three years. In order to clarify earlier guidance to show compliance with the 1-hr NO 2 standard, EPA issued additional guidance in a March 1, 2011 memorandum that specifically addressed the question of intermittent emissions. In part, the guidance stated that certain types of intermittent sources could be excluded from the 1-hr compliance demonstration. As a result, unrealistically high modeled impacts from some intermittent sources could be avoided. This raises the issue of the criteria to distinguish an intermittent source from one that is considered to operate continuously. The guidance states that a source is considered to operate continuously when its emissions "contribute significantly to the annual distribution of daily maximum 1-hour concentrations." The guidance gave as an example a large, baseload power generator that operates continuously with relatively infrequent start up and shut downs. This is compared to a peaking unit that may go through frequent startup/shutdown cycles over the course of a week, or even of a day. In this case the guidance would exclude the startup-shutdown emissions from the base load plant, but not from the peaker unit. In the guidance, EPA states "that case-specific issues and factors may arise that affect the application of this guidance," and that not all facilities will fit within a "clearly defined continuous/normal operations vs. intermittent/infrequent emissions" scenario, such as that between baseload and peaker operations

26 described above. However, it is not clear all the factors that a state regulatory agency might consider in determining intermittent from continuous facilities and their emissions. It is recommended that this workgroup work request that EPA develop more detailed criteria, with examples, of operations that could reasonably fall within the scope of this guidance and be excluded as intermittent sources from demonstrating compliance with the 1-hr NO 2 standard. ONE-HOUR N0 2 MODELING AND THE NOX/NO 2 IN-STACK RATIO: The AERMOD Model includes two non-regulatory options for refining modeled NO 2 impacts. These options are known as the Ozone Limiting Method (OLM) and the Plume Volume Molar Ratio Method (PVMRM). Both options use a two part process to estimate the NO 2 component of a NO x impact. The first part requires the user to provide the assumed NO 2 /NOx in-stack ratio for each source, which is defined as the fraction of NOx gas that is thermally converted to NO 2 prior to its release from a stack or point source. The second part uses available O 3 information and methodologies to estimate the portion of the remaining NO x that will mix with available O 3 and be converted to NO 2 during transport. Prior to the new 1-hour NO 2 NAAQS, a commonly used in-stack ratio for purposes of modeling the annual average NO 2 impact was The EPA s most current guidance for 1-hour modeling proposes to use an instack ratio of 50% conversion of NOx to NO 2 in the stack. The limited amount of measured NO 2 /NO x ratio data currently available suggests that most industrial processes have a NO 2 /NOx ratio of between 0 and 30%. Permitted sources are periodically required to conduct stack test, either for initial testing or compliance tests. Larger sources are also typically required to install and operate Continuous Emissions Monitors (CEMs) to instantaneously monitor the portion of NO x in the gas stream. Current NOx stack testing equipment is capable to differentiating NO 2 from NOx in the gas stream, however minor adjustments must be made to the equipment in able to extract the NO2/NOx in-stack ratio. CEMs are capable of measuring a NO 2 /NOx in-stack ratio if the equipment were programmed to report this value. The NO 2 /NOx in-stack ratio is critical since it defines the portion of the model predicted NOx concentration that will be automatically converted to NO 2. The remaining portion released into the air may or may not undergo conversion to NO 2 prior to it reaching a receptor point. In the case of lower-level releases, the transport distance may be a few hundred meters or less. In this case, the predicted concentration would be in-stack ratio dependent with minimal NO 2 formation due to reactions with O 3. Hence, the user s choice of an in-stack ratio could be the determining factor in model predictions. The group did not attempt to identify suitable in-stack ratios for modeling 1-hour NO2. Rather, it was the group s conclusion that these ratios should be derived from measurements taken during stack testing of permitted in-place equipment operating under normal conditions Therefore, it is the recommendation of this workgroup that the State agencies explore enhancements to their compliance testing and CEM programs that would allow the instruments to report the in-stack ratio. This information could then be compiled in a national database to assist EPA in proposing more representative NO 2 /NOx in-stack ratios. This information could be used to identify default ratios that are process based, or it may provide support for a lower 1-hour NO 2 /NOx in-stack default ratio.

27 SIGNIFICANT IMPACT LEVEL (SIL): The significant impact level (SIL) is the threshold used to determine when a modeled impact may be considered de minimis in a regulatory modeling analysis. Modeled impacts above this threshold are considered large enough to cause or contribute to a modeled violation of an air quality standard or increment. Modeled impacts below this threshold are considered insignificant. The SIL allows new source review (NSR) applicants to conduct the relatively simpler project impact assessment if they believe their project impacts to be de minimis. If not, they then have to conduct the more complex cumulative impact analysis, where nearby sources must also be included in their modeling analysis. EPA has previously codified the SILs for each criteria pollutant (except ozone), so that the SILs may be used in regulatory assessments. EPA intends to promulgate a SIL for the 1-hour NO 2 standard, but in the meantime has issued a recommended interim value that states may consider when carrying out their new source review modeling assessments. EPA stated that the interim value does not bind state and local governments and the public as a matter of law. EPA also acknowledged that several states have adopted interim 1-hour NO 2 SILs that differ (both higher and lower) from their recommended value, and that the EPA-recommended value is not intended to supersede any interim SIL that is now or may be relied upon to implement a state PSD program that is part of an approved SIP, or to impose the use of the SIL concept on any state that chooses to implement the PSD program. EPA s interim value is 4 parts per billion (8 micrograms per cubic meter). Prior to EPA s recommendation, the Northeast States for Coordinated Air Use Management (NESCAUM) recommended a value of 10 micrograms per cubic meter (µg/m 3 ). The 1-hour NO 2 standard is 188 µg/m 3. Modeling staff in some states consider EPA s interim value as the expected value that they should use, in spite of EPA s comments. They also question the reasonableness of the interim value. They feel that the value is so small that it has become effectively moot i.e., most new PSD project impacts would likely be considered significant. PSD sources with taller stacks may model under the SIL if the emissions are just above the significant emissions rate trigger, but the modeled impacts for sources with larger or lower-level releases would likely exceed the SIL. Minor sources, which typically release their pollutants at lower levels and typically have short transport distances to ambient air, would likely always exceed the interim SIL. If true, then all NSR applicants, including minor source applicants, will be required to conduct cumulative impact analysis. In summary, the 1-hour NO 2 standard is extremely stringent and the 1-hour modeling techniques tend to be conservative (for the reasons described in the Additional Background section of this report). Therefore, a larger SIL may be needed to provide relief from the false positives that will likely come from the current NO 2 modeling techniques. The committee request that WESTAR seek EPA confirmation that they are willing to accept higher interim values, or even alternative approaches (such as a population-dependent range of values), in state NSR programs. The committee wants to uphold the new standard, but fears the current modeling techniques can be overly conservative. Allowing states to use higher SILs would therefore allow them to cull smaller projects out of the cumulative impact requirement, without jeopardizing air quality. The committee further request that EPA consider higher SILs in their pending rulemaking.

28 BACKGROUND AMBIENT OZONE CONCENTRATIONS: Background ambient ozone (O 3 ) concentrations are required for the applications of the OLM and PVMRM options in AERMOD. Ozone concentrations can be entered into the model as a single (most conservative) or hourly values covering an entire year (modeling requires five years of data). Current Sources of Ambient Background Ozone Data: Clean Air Status and Trends Network (CASTNET): Hourly; 80 sites in U.S. (including Alaska) and Canada in remote areas USEPA AirData: 1-hour values (first, second, third, and fourth highest); in most cases, monitoring occurs in high population areas The committee recommends that WESTAR establish a database of all rural hourly ozone concentrations, and corresponding NO 2 and NOx values from participating states. Additional monitoring may be required through WESTAR financial assistance. Background concentrations could be obtained through a fusion (as available in BenMAP) of observations and modeling. Modeling domain could cover the entire western state region with incorporation and validation of monitored values. Once established, modeling could be expanded to include other pollutants and interstate transport. Committee is willing to provide recommendations and additional details upon request. OTHER ISSUE: EPA has inadequate field data to promulgate PVMRM/OLM modeling options as approved techniques for modeling 1-hour NO 2 impacts. Therefore, the use of these techniques must be approved by EPA on a caseby-case basis, which takes time and delays permit actions. There is also concern that the techniques are overly conservative; however, they can t be refined without additional data. A field study developed around a relatively isolated source could provide some of the data needed to advance refined NO 2 modeling techniques. The objective would be to compare the estimates from a variety of NO 2 modeling techniques to actual ambient NO 2 concentrations. The committee did not develop a cost estimate. It would not be cheap since a number of source and ambient air parameters would need to be measured and processed. The study would also take time to develop and conduct. EPA would need to be included in the study design. Since the concerns are national, there may be merit in sharing the cost with other groups, if not in expanding the study to include a variety of sources. However, the committee cannot explore this option without Director support. The committee stands ready though to provide additional details as to what this type of study would entail.

29 COLORADO DEPARTMENT OF PUBLIC HEALTH AND ENVIRONMENT Stationary Sources Program / Air Pollution Control Division PS Memo INTER-OFFICE COMMUNICATION TO: FROM: Stationary Sources Staff, Local Agencies, Regulated Community Kirsten King and Roland C. Hea DATE: September 20, 2010 RE: Permit Modeling Requirements for the 1-Hour NO 2 and SO 2 NAAQS The Division is establishing this guidance for use by minor stationary sources of nitrogen dioxide (NO 2 ) and sulfur dioxide (SO 2 ) in evaluating whether modeling is necessary for permitting purposes to determine whether a permit applicant s emissions will comply with the new 1-hour NO 2 and/or the new 1-hour SO 2 National Ambient Air Quality Standard (NAAQS). The United States Environmental Protection Agency (EPA) published implementation guidance on June 28, 2010 and August 23, 2010 regarding demonstrating compliance with the new standards for Prevention of Significant Deterioration (PSD) sources. 1 The Division finds it useful to publish this supplemental state guidance to ensure that minor sources are addressed in a manner consistent with the EPA guidance for PSD sources. Under federal rules, an ambient air quality impact analysis is required for each pollutant that a PSD source has the potential to emit in significant amounts. Such analysis includes modeling. The metric used by EPA to measure significant amounts is the significant emissions rate (SER). Federal rules currently define the SER for NO X and SO 2 as 40 tons per year (tpy). (40 CFR 52.21(b)(23)(i); 40 CFR (b)(23)(i)). EPA recently evaluated and decided to apply on an interim basis the 40 tpy SER to major source permitting compliance demonstrations for the hourly NO 2 and SO 2 standards. EPA concludes and states that an ambient air quality impact analysis is not necessary for PSD sources with projected NO 2 or SO 2 emissions rates below the SER. (Wood Memoranda at p.11 and p.4) 1 See June 28, 2010, Anna Marie Wood, Acting Director, Air Quality Policy Division, Office of Air Quality Planning and Standards Memorandum General Guidance for Implementing the 1-hour NO 2 National Ambient Air Quality Standard in Prevention of Significant Deterioration Permits, Including an Interim 1-hour NO2 Significant Impact Level and August 23, 2010 Memorandum General Guidance for Implementing the 1-hour SO2 National Ambient Air Quality Standard in Prevention of Significant Deterioration Permits, Including an Interim 1-hour SO2 Significant Impact Level ( Wood Memoranda ).

30 The Division has evaluated EPA s rationale for establishing NO 2 and SO 2 SERs for modeling the 1-hour NO 2 and SO 2 standards. The Wood Memoranda guidance set forth EPA s reasoning that its SER for SO 2 (a pollutant with shorter-term 3-hour and 24- hour averaging times) is 40 tpy, and, for this pollutant, ambient air quality impact analyses have not been necessary at levels below the SER. EPA has concluded that this reasoning applies to the one-hour NO 2 and SO 2 standards on an interim basis. EPA states it intends to conduct an evaluation of screening tools available to permitting agencies. In the interim, it recommends the continued use of the existing SER for NO x and SO 2 emissions with respect to the 1-hour NO 2 and SO 2 standards, and thus ambient air quality impact analyses are not necessary for either NO 2 or SO 2 emissions below the 40 tpy SER. EPA s Wood Memoranda guidance address PSD sources. The Division believes that the same principles apply to minor sources, in part, to ensure consistency of treatment in permitting and to ensure that it is not imposing different requirements on minor sources than those to which PSD sources are subject. The Division is aware of no factual basis to impose more stringent requirements on minor sources than EPA would impose on the largest air pollution sources. Therefore, the Division will apply EPA s SERs for NO X and SO 2 to the 1-hour NO 2 and 1-hour SO 2 standards for all stationary source permitting activities, including determining when ambient air quality impact analyses are necessary for permitting, pending the consideration of any further guidance issued by EPA on this subject.

31 Further clarification of this guidance and application of Appendix W for the 1 hour NO2 standard was published March 1, 2011 and is available in the Region 7 NSR Policy & Guidance database.

32 UNITED STATES ENVIRONMENTAL PROTECTION AGENCY RESEARCH TRIANGLE PARK, NC JUN MEMORANDUM OFFICE OF AlA QUALITY PLANNING AND STANDARDS SUBJECT: Guidance Concerning the Implementation of the I-hour N02 NAAQS for the Prevention of Significan~tete i, ration pro~r FROM: Stephen D. Page, Directo ~L.lA / Office of Air Quality Plil' ni g ;;tcisfa~da rd TO: Regional Air Division Directors On January 22, 2010, the Environmental Protection Agency (EPA) announced a new 1- hour nitrogen dioxide (N02) National Ambient Air Quality Standard (hereinafter, either the 1- hour N02 NAAQS or I-hour N02 standard) of 100 parts per billion (Ppb), which is attained when the 3-year average of the 98th-percentile of the annual distribution of daily maximum 1- hour concentrations does not exceed 100 ppb at each monitor within an area. EPA revised the primary N02 NAAQS to provide the requisite protection of public health. The final rule for the new I-hour N02 NAAQS was published in the Federal Register on February 9, 2010 (75 FR 6474), and the standard became effective on April 12, EPA policy provides that any federal Prevention of Significant Deterioration (PSD) permit issued under 40 CFR on or after that effective date must contain a demonstration of source compliance with the new I-hour N02 standard. EPA is aware of reports from stakeholders indicating that some sources- both existing and proposed- are modeling potential violations of the I-hour N02 standard. In many cases, the affected units are emergency electric generators and pump stations, where short stacks and limited property rights exist. However, larger sources, including coal-fired and natural gas-fired power plants, refineries, and paper mills, could also model potential violations of the new N02 NAAQS. To respond to these reports and faci litate the PSD permitting of new and modified major stationary sources, we are issuing the attached guidance, in the form of two memoranda, for implementing the new I-hour N02 NAAQS under the PSD permit program. The guidance contained in the attached memoranda addresses two areas. The first memorandum, titled, "General Guidance for Implementing the I-hour N02 National Ambient Air Quality Standard in Prevention of Significant Deterioration Permits, Including an Interim I-hour N02 Significant Impact Level," includes guidance for the preparation and review of PSD permits with respect to the new I-hour N02 standard. This guidance memorandum sets forth a recommended interim 1- hour N02 significant impact level (SIL) that states may consider when carrying out the required Internet Address (URL). hhp:/iwww.epa.gov RecycledIRecyclable Printed whh Vegetable Oit Based Inks on Recycled Paper (Minimum 25% Poslconsumer)

33 PSD air quality analysis for N02, until EPA promulgates a I-hour N02 SIL via rulemaking. The second memorandum, titled "Applicability of Appendix W Modeling Guidance for the I-hour N02 National Ambient Air Quality Standard," includes specific modeling guidance for estimating ambient N02 concentrations and determining compliance with the new I-hour N02 standard. This guidance does not bind state and local governments and the public as a matter of law. Nevertheless, we believe that state and local air agencies and industry will find this guidance useful when carrying out the PSD permit process. We believe it will provide a consistent approach for estimating N02 air quality impacts from proposed construction or modification of NO x emissions sources. For the most part, the attached guidance reiterates existing policy and guidance, but focuses on how this information is relevant to implementation of the new I-hour N02 NAAQS. Please review the guidance included in the two attached memoranda. If you have questions regarding the general implementation guidance contained in the first memorandum, please contact Raj Rao (rao.raj@epa.gov). If you have questions regarding the modeling guidance in the second memorandum, please contact Tyler Fox (fox.tylercfll,epa.gov). We are continuing our efforts to address permitting issues related to N0 2 and other NAAQS including the recently-signed I-hour sulfur dioxide NAAQS. We plan to issue additional guidance to address these new I-hour standards in the near future. Attachments: I. Memorandum from Anna Marie Wood, Air Quality Policy Division, to EPA Regional Air Division Directors, "General Guidance for Implementing the I-hour N02 National Ambient Air Quality Standard in Prevention of Significant Deterioration Permits, Including an Interim I-hour N02 Significant Impact Level" (June 28, 2010). 2. Memorandum from Tyler Fox, Air Quality Modeling Group, to EPA Regional Air Division Directors, "Applicability of Appendix W Modeling Guidance for the I-hour N02 National Ambient Air Quality Standard" (June 28, 2010). cc: Anna Marie Wood Richard Wayland Raj Rao Tyler Fox Dan deroeck Roger Brode Rich Ossias Elliott Zenick Brian Doster 2

34 June 28, 2010 MEMORANDUM UNITED STATES ENVIRONMENTAL PROTECTION AGENCY Office of Air Quality Planning and Standards Research Triangle Park, North Carolina SUBJECT: FROM: TO: General Guidance for Implementing the I-hour N02 National Ambient Air Quality Standard in Prevention of Significant Deterioration Permits, Including an Interim I-hour N0 2 Significant Impact Level Anna Marie Wood, Acting Director /s/ Air Quality Policy Division Regional Air Division Directors INTRO])UCTION We are issuing the following guidance to explain and clarify the procedures that may be followed by applicants for Prevention of Significant Deterioration (PSD) permits and permitting authorities reviewing such applications to properly demonstrate that proposed construction will not cause or contribute to a violation of the new I-hour nitrogen dioxide (N0 2 ) National Ambient Air Quality Standard (hereinafter, either the I-hour N0 2 NAAQS or I-hour N0 2 standard) that became effective on April 12,2010. EPA revised the primary N02 NAAQS by promulgating a I-hour N0 2 NAAQS to provide the requisite protection of public health. Under section I 65(a)(3) of the Clean Air Act (the Act) and sections 52.21(k) and (k) of EPA's PSD regulations, to obtain a permit, a source must demonstrate that its proposed emissions increase will not cause or contribute to a violation of any NAAQS. This guidance is intended to: (1) explain the recommended procedures for stakeholders to follow to properly address concerns over high preliminary modeled estimates of ambient N0 2 concentrations that suggest potential violations of the new I-hour N0 2 standard under some modeling and permitting scenarios; (2) help reduce the burden of modeling for the hourly N02 standard where it can be properly demonstrated that a source will not have a significant impact on ambient I-hour N02 concentrations; and (3) identify approaches that allow sources and permitting authorities to mitigate, in a manner consistent with existing regulatory requirements, potential modeled violations of the I-hour N02 NAAQS, where appropriate. Accordingly, the techniques described in this memorandum may be used by permit applicants and permitting authorities to configure projects and permit conditions in order to reasonably conclude that a proposed source's emissions do not cause or contribute to modeled I-hour N02 NAAQS violations so that permits can be issued in accordance with the applicable PSD program requirements. This guidance discusses existing provisions in EPA regulations and previous guidance for applying those provisions but focuses on the relevancy of this information for implementing the 3

35 new NAAQS for N02. Importantly, however, this guidance also sets forth a recommended interim I-hour N0 2 significant impact level (SIL) that EPA will use for implementing the federal PSD program, and that states may choose to rely upon to implement their PSD programs for NOx if they agree that these values represent de minimis impact levels and incorporate into each permit record a rationale supporting this conclusion. This interim SIL is a useful screening tool that can be used to determine whether or not the emissions from a proposed source will significantly impact hourly N0 2 concentrations, and, if significant impacts are predicted to occur, whether the source's emissions "cause or contribute to" any modeled violations of the new I-hour N0 2 NAAQS. BACKGROUND On April 12, 2010, the new I-hour N02 NAAQS became effective. EPA interprets its regulations at 40 CFR (the federal PSD program) to require permit applicants to demonstrate compliance with "any" NAAQS that is in effect on the date a PSD permit is issued. (See, e.g., EPA memo dated April I, 2010, titled "Applicability of the Federal Prevention of Significant Deterioration Permit Requirements to New and Revised National Ambient Air Quality Standards.") Due to the introduction of a short-term averaging period for the I-hour N02 NAAQS, we anticipate that some stationary sources with relatively short stacks may experience increased difficulty demonstrating that emissions from new construction or modifications will not cause or contribute to a violation of the I-hour N02 NAAQS. We are responding to reports from stakeholders which indicate that some sources, existing and proposed, are modeling high hourly N02 concentrations showing violations of the 1- hour N0 2 NAAQS-based only on the source's projected emissions of NO x under some modeling and permitting scenarios. We find that, in many cases, the modeled violations are resulting from emissions at emergency electric generators and pump stations, where short stacks and limited property rights exist. In other cases, the problem may occur during periods of unit stmlup, particularly where controls may initially not be in operation. Finally, certain larger sources, including coal-fired and natural gas-fired power plants, refineries, and paper mills could also experience problems in meeting the new I-hour N0 2 NAAQS using particular modeling assumptions and permit conditions. We believe that, in some instances, the projected violations result from the use of maximum modeled concentrations that do not adequately take into account the form of the 1- hour standard, and are based on the conservative assumption of 100% NOx-to-N02 conversion in the ambient air. To the extent that this is the case, it may be possible to provide more accurate projections of ambient N0 2 concentrations by applying current procedures which account for the statistical form of the I-hour N0 2 standard, as well as more realistic estimates of the rate of conversion of NO x emissions to ambient N0 2 concentrations. See EPA Memorandum from Tyler Fox, Air Quality Modeling Group, to EPA Regional Air Division Directors, "Applicability of Appendix W Modeling Guidance for the I-hour N02 National Ambient Air Qnality Standard" (June 28, 2010) for specific modeling guidance for estimating ambient N02 concentrations consistent with the new I-hour N02 NAAQS. In addition, where short stacks are currently being used, or are under design, it may be possible to lessen the source's air quality impacts without improper dispersion by implementing "good engineering practice" (GEP) stack heights to 4

36 increase the height of existing or designed stacks to avoid excessive concentrations due to downwash, as described in the guidance below. It is EPA's expectation that the guidance in this memorandum and available modeling guidance for N02 assist in resolving some of the issues arising from preliminary analyses that are reportedly showing potential exceedances of the new I-hour N0 2 NAAQS that would not be present under more refined modeling applications. In addition, the techniques described in this memorandum may also help avoid violations of the standard through design of the proposed source or permit conditions, consistent with existing regulatory requirements, which enable the source to demonstrate that its proposed emissions increase will not cause or contribute to a modeled violation of the I-hour N0 2 standard. Moreover, the interim I-hour N0 2 SIL that is included in this guidance will provide a reasonable screening tool for efficiently implementing the PSD requirements for an air quality impact analysis. The following discussion provides guidance concerning demonstrating compliance with the new NAAQS and mitigating modeled violations using air quality-based permit limits more stringent than what the Best Available Control Technology provisions may otherwise require, air quality offsets, the use of GEl' stack heights, possible permit conditions for emergency generators, and an interim I-hour N0 2 SIL. AIR-QUALITY BASED EMISSIONS LIMITATIONS Once a level of control required by the Best Available Control Technology provisions is proposed by the psd applicant, the proposed source's emissions must be modeled at the BACT emissions rate(s) to demonstrate that those emissions will not cause or contribute to a violation of any NAAQS or psd increment. EPA's 1990 Workshop Manual (page B.54) describes circumstances where a source's emissions based on levels proposed through the top-down process may not be sufficiently controlled to prevent modeled violations of an increment or NAAQS. In such cases, it may be appropriate for psd applicants to propose a more stringent control option (that is, beyond the level identified via the top-down process) as a result of an adverse impact on the NAAQS or psd increments. DEMONSTRATING COMPLIANCE WITH THE NEW NAAQS & MITIGATING MODELED VIOLATIONS WITH AIR QUALITY OFFSETS A 1988 EPA memorandum provides procedures to follow when a modeled violation is identified during the PSD permitting process. See Memorandum from Gerald A. Emison, EPA OAQpS, to Thomas J. Maslany, EPA Air Management Division, "Air Quality Analysis for Prevention of Significant Deterioration (1'SD)." (July 5,1988). In brief: a reviewing authority may issue a proposed new source or modification a 1'SD permit only if it can be shown that the proposed project's emissions will not "cause or contribute to" any modeled violations. To clarify the above statement, in cases where modeled violations of the I-hour N0 2 NAAQS are predicted, but the permit applicant can show that the NOx emissions increase from the proposed source will not have a significant impact at the point and time of any modeled violation, the permitting authority has discretion to conclude that the source's emissions will not 5

37 contribute to the modeled violation. As provided in the July 5, 1988, guidance memo, in such instances, because of the proposed source's de minimis contribution to any modeled violation, the source's impact will not be considered to cause or contribute to such modeled violations, and the permit could be issued. This concept continues to apply, and the significant impact level (described further below) may be used as part of this analysis. A 2006 decision by the EPA Environmental Appeals Board (EAB) provides detailed reasoning that demonstrates the permissibility of finding that a PSD source would not be considered to cause or contribute to a modeled NAAQS violation because its estimated air quality impact was insignificant at the time and place of the modeled violations 1 See In re Prairie State Gen. Co., 13 E.A.D.,_, PSD Appeal No , Slip. Op. at (EAB 2006) However, where it is determined that a source's impact does cause or contribute to a modeled violation, a permit cannot be issued without some action taken to mitigate the source's impact. In accordance with 40 efr (b )2, a major stationary source or major modification (as defined at (a)(1)(iv) and (v» that locates in an N0 2 attainment area, but would cause or contribute to a violation of the I-hour N02 NAAQS anywhere may "reduce the impact of its emissions upon air quality by obtaining sufficient emission reductions to, at a minimum, compensate for its adverse ambient [N0 2 ] impact where the major source or major modification would otherwise cause or contribute to a violation... " An applicant can meet this requirement for obtaining additional emissions reductions by either reducing its emissions at the source, e.g., promoting more efficient production methodologies and energy efficiency, or by obtaining air quality offsets (see below). See, e.g., In re Intel]Jower a/new York, Inc., 5 E.A.D. 130, 141 (EAB 1994)3 A State may also provide the necessary emissions reductions by imposing emissions limitations on other sources through an approved State Implementation Plan (SIP) revision. These approaches may also be combined as necessary to demonstrate that a source will not cause or contribute to a violation of the NAAQS. Unlike emissions offset requirements in nonattainment areas, in addressing the air quality offset concept, it may not be necessary for a permit applicant to fully offset the proposed emissions increase if an emissions reduction of lesser quantity will mitigate the adverse air quality impact on a modeled violation. ("Although full emission offsets are not required, such a source must obtain emission offsets sufficient to compensate for its air quality impact where the violation occurs." 44 FR 3274, January 16,1979, at 3278.) To clarify this, the 1988 guidance memo referred to above states that: offsets sufficient to compensate for the source's significant impact must be obtained pursuant to an approved State offset program consistent with State Implementation Plan (SIP) requirements under 40 efr (b). Where the source is contributing to an I While there is no I-hour NO, significant impact level (SIL) currently defined in the PSD regulations, we believe that states may adopt interim values, with the appropriate justification for such values, to use for permitting purposes. In addition, we are recommending an interim SIL as part of this guidance for implementing the NO, requirements in the federal PSD program, and in state programs where states choose to use it. 2 The same provision is contained in EPA's Interpretative Ruling at 40 efr part 51 Appendix S, section 1Il. J In contrast to Nonattainment New Source Review permits, offsets are not mandatory requirements in PSD permits if it can otherwise be demonstrated that a source will not cause or contribute to a violation of the NAAQS. See, In re Knauf Fiber Giass, GMBH, 8 E.A.D. 121, 168 (EAB 1999). 6

38 existing violation, the required offset may not correct the violation. Such existing violations must be addressed [through the SIP]. In addition, in order to determine the appropriate emissions reductions, the applicant and permitting authority should take into account modeling procedures for the form of the I-hour standard and for the appropriate NOx-N0 2 conversion rate that applies in the area of concern. As pati of this process, existing ambient ozone concentrations and other meteorological conditions in the area of concern may need to be considered. Note that additional guidance for this and other aspects of the modeling analysis for the impacts of NOx emissions on ambient concentrations ofn0 2 are addressed in EPA modeling guidance, including the June 28, 2010, Memorandum titled, "Applicability of Appendix W Modeling Guidance for the I-hour N02 National Ambient Air Quality Standard." "GOOD ENGINEERING PRACTICE" STACK HEIGHT & DISPERSION TECHNIQUES If a permit applicant is unable to show that the source's proposed emissions increase will not cause or contribute to a modeled violation of the new I-hour N0 2 NAAQS, the problem could be the result of plume downwash effects which may cause high ambient concentrations near the source. In such cases, a source may be able to raise the height of its existing stacks (or designed stacks if not yet constructed) to a GEP stack height of at least 65 meters, measured li'om the ground-level elevation at the base of the stacie While not necessarily totally eliminating the effects of down wash in all cases, raising stacks to GEP height may provide substantial air quality benefits in a manner consistent with statutory provisions (section 123 of the Act) governing acceptable stack heights to minimize extensive concentrations due to atmospheric downwash, eddies or wakes. Permit applicants should also be aware of the regulatory restrictions on stack heights for the purpose of modeling for compliance with NAAQS and increments. Section 52.21(h) of the PSI) regulations currently prohibits the use of dispersion techniques, such as stack heights above GEP, merged gas streams, or intermittent controls for setting NOx emissions limits or to meet the annual and I-hour NAAQS and annual N02 increments. However, stack heights in existence before December 31, 1970, and dispersion techniques implemented before then, are not affected by these limitations. EPA's general stack height regulations are promulgated at 40 CFR 51.1 OO(ff), (gg), (hh), (kk) and (nn), and 40 CFR a. Stack heights: A source cannot take credit for that portion of a stack height in excess of the GEP height when modeling to develop the NOx emissions limitations or to determine source compliance with the annual and I-hour N0 2 NAAQS. It should be noted, however, that this limitation does not limit the actual height of any stack constructed by a new source or modification. The following limitations apply in accordance with 52.21(h): For a stack height less than GEP, the actual stack height must be used in the source impact analysis for NOx emissions; 7

39 For a stack height equal to or greater than 65 meters, the impact on NOx emission limits may be modeled using the greater of: o A de minimis stack height equal to 65 meters, as measured from the groundlevel elevation at the base of the stack, without demonstration or calculation (40 efr 51.1 OO(ii)(1»; o The refined formula height calculated using the dimensions of nearby structures in accordance with the following equation: GEl) = H + 1.5L, where H is the height of the nearby structure and L is the lesser dimension of the height or projected width of the nearby structure (40 efr (ii)(2)(ii».4 A GEP stack height exceeding the refined formula height may be approved when it can be demonstrated to be necessary to avoid "excessive concentrations" of N02 caused by atmospheric downwash, wakes, or eddy effects by the source, nearby structures, or nearby terrain features. (40 efr 51.l00(ii)(3), (jj), (kk»; For purposes ofpsd (and NOx/N02), "excessive concentrations" means a maximum ground-level concentration ofn02 due to NOx emissions from a stack due in whole or in part to downwash, wakes, and eddy effects produced by nearby structures or nearby terrain features which individually is at least 40 percent in excess of the maximum N02 concentration experienced in the absence of such effects and (a) which contributes to a total N0 2 concentration due to emissions from all sources that is greater than the annual or I-hour N02 NAAQS or (b) greater than the PSD (annual) increment for N0 2. (40 efr (kk)(I». Reportedly, for economic and other reasons, many existing source stacks have been constructed at heights less than 65 meters, and source impact analyses may show that the source's emissions will cause or contribute to a modeled violation of the annual or I -hour N02 NAAQS. Where this is the case, sources should be aware that they can increase their stack heights up to 65 meters without a GEP demonstration. b. Other dispersion techniques: The term "dispersion technique" includes any practice carried out to increase final plume rise, subject to certain exceptions (40 efr 51. I OO(hh)(I )(iii), (2)(i) - (v». Beyond the noted exceptions, such techniques are not allowed for getting credit for modeling source compliance with the annual and I-hour N0 2 NAAQS and annual N02 increment. 4 For stacks in existence on January 12, 1979, the GEl' equation is GEl' H (provided the owner or operator produces evidence that this equation was actually relied on in establishing an emission limitation for NOx (40 CFR (ii)(2)(i) 8

40 OPERATION OF EMERGENCY EQUIPMENT & GENERAL STARTUP CONDITIONS In determining an emergency generator's potential to emit, existing guidance (EPA memo titled "Calculating Potential to Emit (PTE) for Emergency Generators," September 6, 1995) allows a default value of 500 hours "for estimating the number of hours that an emergency generator could be expected to operate under worst-case conditions." The guidance also allows for alternative estimates to be made on a case-by-case basis for individual emergency generators. This time period must also consider operating time for both testing/maintenance as well as for emergency utilization. Likewisc, existing EPA policy does not allow NOx emissions to be excluded from the source impact analysis (NAAQS and increments) when the emergency equipment is operating during an emergency. EPA provides no exemption from compliance with the NAAQS during periods of emergency operation. Thus, it is not sufficient to consider only emissions generated during periods of testing/maintenance in the source impact analysis. If during an emergency, emergency equipment is never operated simultaneously with other emissions units at the source that the emergency equipment will back up, a worst-case hourly impact analysis may very well occur during periods of normal source operation when other emissions units at the facility are likely to be operating simultaneously with the scheduled testing of emergency equipment. To avoid such worst-case modeling situations, a permit applicant may commit to scheduling the testing of emergency equipment during times when the source is not otherwise operating, or during known off-peak operating periods. This could provide a basis to justify not modeling the I-hour impacts of the emergency equipment under conditions that would include simultaneous operation with other onsite emissions units. Accordingly, permits for emergency equipment may include enforceable conditions that specifically limit the testing/maintenance of emergency equipment to certain periods of time (seasons, days of the week, hours of the day, etc.) as long as these limitations do not constitute dispersion techniques under 40 CFR 51.1 (hh)(l )(ii). We also note that similar problems associated with the modeling of high I-hour N0 2 concentrations have been reported to occur during startup periods for certain kinds of emissions units--often because control equipment cannot function during all or a portion of the stm1up process. EPA currently has no provisions for exempting emissions occurring during equipment startups fl'om the air quality analysis to demonstrate compliance with the NAAQS. Startup emissions may occur during only a relatively small portion of the unit's total annual operating schedule; however, they must be included in the required PS]) air quality analysis for the NAAQS. Sources may be willing to accept enforceable permit conditions limiting equipment startups to certain hours of the day when impacts are expected to be lower than normal. Such permit limitations can be accounted for in the modeling of such emissions. Applicants should direct other questions arising concerning procedures for modeling startup emissions to the applicable permitting authority to determine the most current modeling guidance. 9

41 SCREENING VALUES In the final rule establishing the hourly N02 standard, EPA discussed various implementation considerations for the PSD permitting program. 75 FR.6474, 6524 (Feb. 9, 20 I 0). This discussion included the following statements regarding particular screening values that have historically been used on a widespread basis to facilitate implementation of the PSD permitting program: We also believe that there may be a need to revise the screening tools currently used under the NSRlPSD program for completing N02 analyses. These screening tools include the significant impact levels (SILs), as mentioned by one commenter, but also include the significant emissions rate for emissions of NO x and the significant monitoring concentration (SMC) for N02. EPA intends to evaluate the need for possible changes or additions to each of these important screening tools for NOx/N02 due to the addition of a I-hour N02 NAAQS. If changes or additions are deemed necessary, EPA will propose any such changes for public notice and comment in a separate action. 75 FR EPA intends to conduct an evaluation of these issues and submit our findings in the form of revised significance levels under notice and comment rulemaking if any revisions are deemed appropriate. In the interim, for the reasons provided below, we recommend the continued use of the existing significant emissions rates (SER) for NOx emissions as well as an interim I-hour NOz SIL that we are setting forth today for conducting air quality impact analyses for the I-hour N02 NAAQS. As described in the section titled Introduction, EPA intends to implement the interim I-hour NOz SIL contained herein under the federal PSD program and offers states the opportunity to use it in their PSD programs if they choose to do so. EPA is not addressing the significant monitoring concentrations in this memorandum. SIGNIFICANT EMISSIONS RATE Under the terms of existing EPA regulations, the applicable significant emissions rate for nitrogen oxides is 40 tons per year. 40 CFR 52.21(b)(23); 40 CFR (b)(23). The significant emissions rates defined in those regulations are specific to individual pollutants but are not differentiated by the averaging times of the air quality standards applicable to some of the listed pollutants. Although EPA has not previously promulgated a N02 standard using an averaging time of less than one year, the NAAQS for S02 have included standards with 3-hour and 24-hour averaging times for many years. EPA has applied the 40 tons per year significant emissions rate for S02 across all of these averaging times. Until the evaluation described above and any associated rulemaking is completed, EPA does not believe it has cause to apply the NOz significant emissions rate any differently than EPA has historically applied the S02 significant emissions rate and others that apply to standards with averaging times less than I year. Under existing regulations, an ambient air quality impact analysis is required for "each pollutant that [a source] would have the potential to emit in significant amounts." 40 CFR (m)(1 )(i)(a); 40 CFR (m)(1 )(i)(a). For modifications, these regulations require this analysis for "each pollutant for which [the modification] would result in a significant net 10

42 emissions increase." 40 CFR.S2.21(m)(l)(i)(b); 40 CFR.SJ.J66(m)(l)(i)(b). EPA construes this regulation to mean that an ambient impact analysis is not necessary for pollutants with emissions rates below the significant emissions rates in paragraph (b)(23) of the regulations. No additional action by EPA or permitting authorities is necessary at this time to apply the 40 tpy significant emissions rate in existing regulations to the hourly N0 2 standard. INTERIM I-HOUR N0 2 SIGNIFICANT IMPACT LEVEL A significant impact level (SIL) serves as a useful screening tool for implementing the PSD requirements for an air quality analysis. The primary purpose of the SIL is to serve as a screening tool to identify a level of ambient impact that is sufficiently low relative to the NAAQS or PSD increments such that the impact can be considered trivial or de minimis. Hence, the EPA considers a source whose individual impact falls below a SIL to have a de minimis impact on air quality concentrations that already exist. Accordingly, a source that demonstrates that the projected ambient impact of its proposed emissions increase does not exceed the SIL for that pollutant at a location where a NAAQS or increment violation occurs is not considered to cause or contribute to that violation. In the same way, a source with a proposed emissions increase of a particular pollutant that will have a significant impact at some locations is not required to model at distances beyond the point where the impact of its proposed emissions is below the SILs for that pollutant. When a proposed source's impact by itself is not considered to be "significant," EPA has long maintained that any further effort on the part of the applicant to complete a cumulative source impact analysis involving other source impacts would only yield information of trivial or no value with respect to the required evaluation of the proposed source or modification. The concept of a SIL is grounded on the de minimis principles described by the court in Alabama Power Co. v. CosrIe, 636 F.2d 323, 360 (D.C. Cir. 1980); See also Sur Contra La Contaminacion v. EPA, 202 FJd 443, (I st Cir. 2000) (upholding EPA's use of SIL to allow permit applicant to avoid full impact analysis); In re: Prairie State Gen. Co., PSD Appeal No. OS-OS, Slip. Op. at 139 (EAB 2006) EPA has codified several SILs into regulations at 40 CFR 5 I. I 65(b). EPA plans to undertake rulemaking to develop a I -hour N0 2 SIL for the new NAAQS for N0 2. However, EPA has recognized that the absence of an EPA-promulgated SIL does not preclude permitting authorities from developing interim SILs for use in demonstrating that a cumulative air quality analysis would yield trivial gain. Response to Comments, Implementation of New Source Review (NSR) Program for Particulate Matter Less Than 2.5 Micrometers in Diameter (PM 25 ), pg. 82 (March 2008) [EPA-HQ-OAR J. Until such time as a I -hour N0 2 SIL is defined in the PSD regulations, we are herein providing a recommended interim SIL that we intend to use as a screening tool for completing the required air quality analyses for the new I -hour N02 under the federal PSD program at 40 CFR To support the application of this interim SIL in each instance, a permitting authority that utilizes this SIL as part of an ambient air quality analysis should include in the permit record the analysis reflected in this memorandum and the referenced documents to demonstrate that an air quality impact at or below the SIL is de minimis in nature and would not cause a violation of the NAAQS. I I

43 Using the interim I-hour N02 SIL, the permit applicant and permitting authority can determine: (I) whether, based on the proposed increase in NOx emissions, a cumulative air quality analysis is required; (2) the area of impact within which a cumulative air quality analysis should focus; and (3) whether, as part of a cumulative air quality analysis, the proposed source's NOx emissions will cause or contribute to a modeled violation of the I-hour N02 NAAQS. In this guidance, EPA recommends an interim I-hour N02 SIL value of 4 ppb. To determine initially whether a proposed project's emissions increase will have a significant impact (resulting in the need for a cumulative air quality analysis), this interim SIL should be compared to either of the following: The highest of the 5-year averages of the maximum modeled I-hour N02 concentrations predicted each year at each receptor, based on 5 years of National Weather Service data; or The highest modeled I -hour N02 concentration predicted across all receptors based on I year of site-specific meteorological data, or the highest of the multi-year averages of the maximum modeled I-hour N0 2 concentrations predicted each year at each receptor, based on 2 or more, up to 5 complete years of available site-specific meteorological data. Additional guidance will be forthcoming for the purpose of comparing a proposed source's modeled impacts to the interim I-hour N02 SIL in order to make a determination about whether that source's contribution is significant when a cumulative air quality analysis identifies violations of the I-hour N02 NAAQS (i.e., "causes or contributes to" a modeled violation). We derived this interim I-hour N02 SIL by using an impact equal to 4% of the I-hour N02 NAAQS (which is 100 ppb). We have chosen this approach because we believe it is reasonable to base the interim I-hour N02 SIL directly on consideration of impacts relative to the I-hour N02 NAAQS. In 1980, we defined SER for each pollutant subject to PSD. 45 FR 52676, August 7,1980 at For PM and S02, we defined the SER as the emissions rate that resulted in an ambient impact equal to 4% of the applicable short-term NAAQS. The 1980 analysis focused on levels no higher than 5% of the primary standard because of concerns that higher levels were found to result in unreasonably large amounts of increment being consumed by a single source. Within the range of impacts analyzed, we considered two factors that had an important influence on the choice of de minimis emissions levels: (I) cumulative effect on increment consumption of multiple sources in an area, each making the maximum de minimis emissions increase; and (2) the projected consequence of a given de minimis level on administrative burden. As explained in the preamble to the 1980 rulemaking and the supporting documentation,5 EPA decided to use 4% of the 24-hour primary NAAQS for PM and S02 to define the significant emissions rates (SERs) for those pollutants. It was noted that, at the time, only an annual N02 NAAQS existed. Thus, for reasons explained in the 1980 preamble, to define the SER for NOx emissions we used a design value of2% of the annual N02 NAAQS. See 45 FR Looking now at a short-term NAAQS for N02, we believe that it is reasonable as an interim approach to use a SIL value that represents 4% of the I-hour N02 5 EPA evaluated de minimis levels for pollutants for which NAAQS had been established in a document titled "Impact of Proposed and Alternative De Minimis Levels for Criteria Pollutants"; EPA-4S n, June

44 NAAQS. EPA will consider other possible alternatives for developing a I-hour N02 SIL in a future rulemaking that will provide an opportunity for public participation in the development of a SIL as part of the PSD regulations. Several state programs have already adopted interim I-hour N02 SILs that differ (both higher and lower) from the interim value being recommended herein. The EPA-recommended interim I-hour N02 SIL is not intended to supersede any interim SIL that is now or may be relied upon to implement a state PSD program that is part of an approved SIP, or to impose the use of the SIL concept on any state that chooses to implement the PSD program-in particular the ambient air quality analysis-without using a SIL as a screening tool. Accordingly, states that implement the PSD program under an EPA-approved SIP may choose to use this interim SIL, another value that may be deemed more appropriate for PSD permitting purposes in the state of concern, or no SIL at all. The application of any SIL that is not reflected in a promulgated regulation should be supported by a record in each instance that shows the value represents a de minimis impact on the I-hour N0 2 standard, as described above. In the event of questions regarding the general implementation guidance contained iu this memorandum, please contact Raj Rao (rao.raj@epa.gov). cc: Raj Rao, C504-0 I Dan deroeck, C Tyler Fox, C Roger Brode, C Richard Wayland, C Elliot Zenick, OGC Brian Doster, OGC EPA Regional NSR Contacts 13

45 June 28, 2010 MEMORANDUM UNITED STATES ENVIRONMENTAL PROTECTION AGENCY Office of Air Quality Planning and Standards Research Triangle Park, North Carolina SUBJECT: FROM: TO: Applicability of Appendix W Modeling Guidance for the I-hour N02 National Ambient Air Quality Standard Tyler Fox, Leader Air Quality Modeling Group, C Regional Air Division Directors INTRODUCTION On January 22, 2010, EPA announced a new I-hour nitrogen dioxide (N02) National Ambient Air Quality Standard (I-hour N02 NAAQS or I-hour N02 standard) which is attained when the 3-year average of the 98th-percentile of the annual distribution of daily maximum I-hour concentrations does not exceed 100 ppb at each monitor within an area. The final rule for the new I-hour N02 NAAQS was published in the Federal Register on February 9, 2010 (75 FR ), and the standard became effective on April 12, 2010 (EPA, 2010a). This memorandum clarifies the applicability of current guidance in the Guideline on Air Quality Models (40 CFR Part 51, Appendix W) for modeling N0 2 impacts in accordance with the Prevention of Significant Deterioration (PSD) permit requirements to demonstrate compliance with the new I-hour N02 standard. SUMMARY OF CURRENT GUIDANCE While the new I-hour NAAQS is defined relative to ambient concentrations ofn0 2, the majority of nitrogen oxides (NOx) emissions for stationary and mobile sources are in the form of nitric oxide (NO) rather than N0 2. Appendix W notes that the impact of an individual source on ambient N02 depends, in part, "on the chemical environment into which the source's plume is to be emitted" (see Section 5.1.j). Given the role of NO x chemistry in determining ambient impact levels ofn02 based on modeled NOx emissions, Section of Appendix W recommends the following three-tiered screening approach for N02 modeling for annual averages: Tier I - assume full conversion of NO to N02 based on application of an appropriate refined modeling technique under Section of Appendix W to estimate ambient NOx concentrations; Tier 2 - multiply Tier I result by empirically-derived N0 2 /NOx ratio, with 0.75 as the annual national default ratio (Chu and Meyer, 1991); and 14

46 Tier 3 - detailed screening methods may be considered on a case-by-case basis, with the Ozone Limiting Method (OLM) identified as a detailed screening technique for point sources (Cole and Summerhays, 1979). Tier 2 is often referred to as the Ambient Ratio Method, or ARM. Site-specific ambient N0 2 INOx ratios derived from appropriate ambient monitoring data may also be considered as detailed screening methods on a case-by-case basis, with proper justification. Consistent with Section 4.2.2, AERMOD is the current preferred model for "a wide range of regulatory applications in all types of terrain" for purposes of estimating ambient concentrations of NOz, based on NOx emissions, under Tiers I and 2 above. We discuss the role of AERMOD for Tier 3 applications in more detail below. APPLICABILITY OF CURRENT GUIDANCE TO I-HOUR N0 2 NAAQS In general, the Appendix W recommendations regarding the annual N02 standard are also applicable to the new I-hour N02 standard, but additional issues may need to be considered in the context of a I-hour standard, depending on the characteristics of the emission sources, and depending on which tier is used, as summarized below: Tier I applies to the I-hour N02 standard without any additional justification; Tier 2 may also apply to the I-hour N02 standard in many cases, but some additional consideration will be needed in relation to an appropriate ambient ratio for peak hourly impacts since the current default ambient ratio is considered to be representative of "area wide quasi-equilibrium conditions"; and Tier 3 "detailed screening methods" will continue to be considered on a case-by-case basis for the I-hour N02 standard. However, certain input data requirements and assumptions for Tier 3 applications may be of greater importance for the I-hour standard than for the annual standard given the more localized nature of peak hourly vs. annual impacts. In addition, use of site-specific ambient N02INOx ratios based on ambient monitoring data will generally be more difficult to justify for the I-hour N02 standard than for the annual standard. While Appendix W specifically mentions OLM as a detailed screening method under Tier 3, we also consider the Plume Volume Molar Ratio Method (PVMRM) (Hanrahan, 1999a) discussed under Section 5.1.j of Appendix W to be in this category at this time. Both of these options account for ambient conversion of NO to N02 in the presence of ozone, based on the following basic chemical mechanism, known as titration, although there are important differences between these methods: (Eq. I) As noted in Section 5.1.j, EPA is currently testing the PVMRM option to determine its suitability as a refined method. Limited evaluations of PVMRM have been completed, which show encouraging results, but the amount of data currently available is too limited to justify a designation of PVMRM as a refined method for N02 (Hanrahan, 1999b; MACTEC, 2005). EPA is currently updating and extending these evaluations to examine model performance for IS

47 predicting hourly N02 concentrations, including both the OLM and PVMRM options, and results of these additional evaluations will be provided at a later date. A sensitivity analysis of the OLM and PVMRM options in AERMOD has been conducted that compares modeled concentrations based on OLM and PVMRM with Tiers 1 and 2 for a range of source characteristics (MACTEC, 2004). This analysis serves as a useful reference to understand how ambient N02 concentrations may be impacted by application of this three-tiered screening approach, and includes comparisons for both annual average and maximum I-hour N02 concentrations. Key model inputs for both the OLM and PVMRM options are the in-stack ratios of N0 2 /NOx emissions and background ozone concentrations. While the representativeness of these key inputs is important in the context of the annual N0 2 standard, they will generally take on even greater importance for the new I-hour N0 2 standard, as explained in more detail below. Recognizing the potential importance of the in-stack N02/NOx ratio for hourly N02 compliance demonstrations, we recommend that in-stack ratios used with either the OLM or PVMRM options be justified based on the specific application, i.e., there is no "default" in-stack N02INOx ratio for either OLM or PVMRM. The OLM and PVMRM methods are both available as non-regulatory-default options within the EPA-preferred AERMOD dispersion model (Cimorelli, el ai., 2004; EPA, 2004; EPA, 2009). As a result of their non-regulatory-default status, pursuant to Sections c, a, and A.l.a(2) of Appendix W, application of AERMOD with the OLM or PVMRM option is no longer considered a "preferred model" and, therefore, requires justification and approval by the Regional Office on a case-by-case basis. While EPA is continuing to evaluate the PVMRM and OLM options within AERMOD for use in compliance demonstrations for the I-hour N0 2 standard, as long as they are considered to be non-regulatory-default options, their use as alternative modeling techniques under Appendix W should be justified in accordance with Section 3.2.2, paragraph (e), as follows: "e. Finally, for condition (3) in paragraph (b) of this subsection [preferred model is less appropriate for the specific application, or there is no preferred model], an alternative refined model may be used provided that: I. The model has received a scientific peer review; 11. The model can be demonstrated to be applicable to the problem on a theoretical basis; 111. The data bases which are necessary to perform the analysis are available and adequate; IV. Appropriate performance evaluations of the model have shown that the model is not biased toward underestimates; and v. A protocol on methods and procedures to be followed has been established. " Since AERMOD is the preferred model for dispersion for a wide range of application, the focus of the alternative model demonstration for use of the OLM and PVMRM options within AERMOD is on the treatment of NO x chemistry within the model, and does not need to address basic dispersion algorithms within AERMOD. Furthermore, items i and iv of the alternative 16

48 model demonstration for these options can be fulfilled in part based on existing documentation (Cole and Summerhays, 1979; Hanrahan, 1999a; Hanrahan, 1999b; MACTEC, 200S), and the remaining items should be routinely addressed as part of the modeling protocol, irrespective of the regulatory status of these options. The issue of applicability to the problem on a theoretical basis (item ii) is a case-by-case determination based on an assessment of the adequacy of the ozone titration mechanism utilized by these options to account for NOx chemistry within the AERMOD model based on "the chemical environment into which the source's plume is to be emitted" (Appendix W, Section S.I.j). The adequacy of available data bases needed for application of OLM and PVMRM (item iii), including in-stack N0 2 /NOx ratios and background ozone concentrations, is a critical aspect of the dcmonstration which we discuss in more detail below. It should also be noted that application of the OLM or PVMRM methods with other Appendix W models or alternative models, whether as a separate post-processor or integrated within the model, would require additional documentation and demonstration that the methods have been implemented and applied appropriately within that context, including model-specific performance evaluations which satisfy item iv under Section e. Given the form of the new I-hour N02 standard, some clarification is needed regarding the appropriate data periods for modeling demonstrations of compliance with thc NAAQS vs. dcmonstrations of attainment of the NAAQS through ambient monitoring. While monitored design values for the I-hour N0 2 standard are based on a 3-year average (in accordance with Section 1 (c )(2) of Appendix S to 40 CFR Part SO), Section of Appendix W addresses the length of the meteorological data record for dispersion modeling, stating that "[T]he use of S years ofnws [National Weather Service] meteorological data or at least 1 year of site specific data is required." Section b further states that "one year or more (including partial years), up to five years, of site specific data... are preferred for use in air quality analyses." Although the monitored design value for the I-hour N02 standard is defined in terms of the 3-year average, this definition does not preempt or alter the Appendix W requirement for use of S years of NWS meteorological data or at least 1 year of site specific data. The S-year average based on use of NWS data, or an average across one or more years of available site specific data, serves as an unbiased estimate of the 3-year average for purposes of modeling demonstrations of compliance with the NAAQS. Modeling of "rolling 3-year averages," using years 1 through 3, years 2 through 4, and years 3 through 5, is not required. Furthermore, since modeled results for N02 are averaged across the number of years modeled for comparison to the new I-hour N02 standard, the meteorological data period should include complete years of data to avoid introducing a seasonal bias to the averaged impacts. In order to comply with Appendix W recommendations in cases where partial years of site specific meteorological data are available, while avoiding any seasonal bias in the averaged impacts, an approach that utilizes the most conservative modeling result based on the first complete-year period of the available data record vs. results based on the last complete-year period of available data may be appropriate, subject to approval by the appropriate reviewing authority. Such an approach would ensure that all available site specific data are accounted for in the modeling analysis without imposing an undue burden on the applicant and avoiding arbitrary choices in the selection of a single complete-year data period. The form of the new I-hour N02 standard also has implications regarding appropriate methods for combining modeled ambient concentrations with monitored background 17

49 concentrations for comparison to the NAAQS in a cumulative modeling analysis. As noted in the March 23, 2010 memorandum regarding "Modeling Procedures for Demonstrating Compliance with PM 25 NAAQS" (EPA, 201 Ob), combining the 98 th percentile monitored value with the 98 th percentile modeled concentrations for a cumulative impact assessment could result in a value that is below the 98 th percentile of the combined cumulative distribution and would, therefore, not be protective of the NAAQS. However, unlike the recommendations presented for PM2S, the modeled contribution to the cumulative ambient impact assessment for the I-hour N02 standard should follow the form of the standard based on the 98 th percentile of the annual distribution of daily maximum I-hour concentrations averaged across the number of years modeled. A "first tier" assumption that may be applied without further justification is to add the overall highest hourly background N02 concentration from a representative monitor to the modeled design value, based on the form of the standard, for comparison to the NAAQS. Additional refinements to this "first tier" approach based on some level of temporal pairing of modeled and monitored values may be considered on a case-by-case basis, with adequate justification and documentation. DISCUSSION OF TECHNICAL ISSUES While many of the same technical issues related to application of Appendix W guidance for an annual N0 2 standard would also apply in the context of the new I-hour N0 2 standard, there are some important differences that may also need to be considered depending on the specific application. This section discusses several aspects of these technical issues related to the new I-hour N02 NAAQS, including a discussion of source emission inventories required for modeling demonstrations of compliance with the NAAQS and other issues specific to each of the three tiers identified in Section of Appendix W for N02 modeling. Emission Inventories The source emissions data are a key input for allmodcling analyses and one that may require additional considerations under the new I-hour N02 standard is the source emissions data. Section 8.1 of Appendix W provides guidance regarding source emission input data for dispersion modeling and Table 8-2 summarizes the recommendations for emission input data that should be followed for NAAQS compliance demonstrations. Although existing NOx emission inventories used to support modeling for compliance with the annual N02 standard should serve as a useful starting point, such inventories may not always be adequate for use in assessing compliance with the new I-hour N0 2 standard since some aspects of the guidance in Section 8.1 differs for long-term (annual and qumierly) standards vs. short-term (:s: 24 hours) standards. In particular, since maximum ground-level concentrations may be more sensitive to operating levels and startup/shutdown conditions for an hourly standard than for an annual standard, emission rates and stack parameters associated with the maximum ground-level concentrations for the annual standard may underestimate maximum concentrations for the new I-hour N02 standard. Due to the importance of in-stack N0 2 /NOx ratios required for application of the OLM and PVMRM options within AERMOD discussed above, consideration should also be given to the potential variability of in-stack N02/NOx ratios under different operating conditions when those non-regulatory-default options are applied. We also note that source emission input data recommendations in Table 8-2 of Appendix W for "nearby sources" and "other sources" that 18

50 may be needed to conduct a cumulative impact assessment include further differences between emission data for long-term vs. short-term standards which could also affect the adequacy of existing annual NOx emission inventories for the new I-hour N02 standard. The terms "nearby sources" and "other sources" used in this context are defined in Section of Appendix W. Attachment A provides a more detailed discussion on determining NOx emissions for permit modeling. While Section of Appendix W emphasizes the importance of professional judgment by the reviewing authority in the identification of nearby and other sources to be included in the modeled emission inventory, Appendix W establishes "a significant concentration gradient in the vicinity of the source" under consideration as the main criterion for this selection. Appendix W also indicates that "the number of such [nearby] sources is expected to be small except in unusual situations." See Section b. Since concentration gradients will vary somewhat depending on the averaging period being modeled, especially for an annual vs. I-hour standard, the criteria for selection of "nearby" and "other" sources for inclusion in the modeled inventory may need to be reassessed for the I-hour N0 2 standard. The representativeness of available ambient air quality data also plays an important role in determining which nearby sources should be included in the modeled emission inventory. Key issues to consider in this regard are the extent to which ambient air impacts of emissions from nearby sources are reflected in the available ambient measurements, and the degree to which emissions from those background sources during the monitoring period are representative of allowable emission levels under the existing permits. The professional judgments that are required in developing an appropriate inventory of background sources should strive toward the proper balance between adequately characterizing the potential for cumulative impacts of emission sources within the study area to cause or contribute to violations of the NAAQS, while minimizing the potential to overestimate impacts by double-counting of modeled source impacts that are also reflected in the ambient monitoring data. We would also caution against the literal and uncritical application of very prescriptive procedures for identifying which background sources should be included in the modeled emission inventory for NAAQS compliance demonstrations, such as those described in Chapter C, Section IV.C.l of the draft New Source Review WorhhojJ Manual (EPA, 1990), noting again that Appendix W emphasizes the importance of professional judgment in this process. While the draft workshop manual serves as a useful general reference regarding New Source Review (NSR) and PSD programs, and such procedures may playa useful role in defining the spatial extent of sources whose emissions may need to be considered, it should be recognized that "[ijt is not intended to be an official statement of policy and standards and does not establish binding regulatory requirements." See, Preface. Given the range of issues involved in the determination of an appropriate inventory of emissions to include in a cumulative impact assessment, the appropriate reviewing authority should be consulted early in the process regarding the selection and proper application of appropriate monitored background concentrations and the selection and appropriate characterization of modeled background source emission inventories for use in demonstrating compliance with the new I-hour N02 standard. Tier-specific Technical Issues 19

51 This section discusses technical issues related to application of each tier in the threetiered screening approach for N0 2 modeling recommended in Section Appendix W. A basic understanding of NO x chemistry and "of the chemical environment into which the source's plume is to be emitted" (Appendix W, Section 5.1.j) will be helpful for addressing these issues based on the specific application. Tier I: Since the assumption of full conversion of NO to N0 2 will provide the most conservative treatment of NO x chemistry in assessing ambient impacts, there are no technical issues associated with treatment of NO x chemistry for this tier. However, the general issues related to emission inventories for the I-hour N02 standard discussed above and in Attachment A apply to Tier I. Tier 2: As noted above, the 0.75 national default ratio for ARM is considered to be representative of "area wide quasi-equilibrium conditions" and, therefore, may not be as appropriate for use with the I-hour N0 2 standard. The appropriateness of this default ambient ratio will depend somewhat on the characteristics of the sources, and as such application of Tier 2 for I-hour N0 2 compliance demonstrations may need to be considered on a source-by-source basis in some cases. The key technical issue to address in relation to this tier requires an understanding of the meteorological conditions that are likely to be associated with peak hourly impacts hom the source(s) being modeled. In general, for low-level releases with limited plume rise, peak hourly NOx impacts are likely to be associated with nighttime stable/light wind conditions. Since ambient ozone concentrations are likely to be relatively low for these conditions, and since low wind speeds and stable atmospheric conditions will further limit the conversion of NO to N0 2 by limiting the rate of entrainment of ozone into the plume, the 0.75 national default ratio will likely be conservative for these cases. A similar rationale may apply for elevated sources where plume impaction on nearby complex terrain under stable atmospheric conditions is expected to determine the peak hourly NOx concentrations. By contrast, for elevated sources in relatively Hat terrain, the peak hourly NOx concentrations are likely to occur during daytime convective conditions, when ambient ozone concentrations are likely to be relatively high and entrainment of ozone within the plume is more rapid due to the vigorous vertical mixing during such conditions. For these sources, the 0.75 default ratio may not be conservative, and some caution may be needed in applying Tier 2 for such sources. We also note that the default equilibrium ratio employed within the PVMRM algorithm as an upper bound on an hourly basis is 0.9. Tier 3: This tier represents a general category of "detailed screening methods" which may be considered on a case-by-case basis. Section 5.2.4(b) of Appendix W cites two specific examples of Tier 3 methods, namely OLM and the use of site-specific ambient N0 2 /NOx ratios supported by ambient measurements. As noted above, we also believe it is appropriate to consider the 20

52 PVMRM option as a Tier 3 detailed screening method at this time. The discussion here focuses primarily on the OLM and PVMRM methods, but we also note that the use of site-specific ambient N02INOx ratios will be subject to the same issues discussed above in relation to the Tier 2 default ARM, and as a result it will generally be much more difficult to determine an appropriate ambient N02INOx ratio based on monitoring data for the new I-hour N02 standard than for the annual standard. While OLM and PVMRM are both based on the same simple chemical mechanism of titration to account for the conversion of NO emissions to N02 (see Eg. I) and therefore entail similar technical issues and considerations, there are some important differences that also need to be considered when assessing the appropriateness of these methods for specific applications. While the titration mechanism may capture the most important aspects ofno-to-n02 conversion in many applications, both methods will suffer from the same limitations for applications in which other mechanisms, such as photosynthesis, contribute significantly to the overall process of chemical transformation.' Sources located in areas with high levels ofvoc emissions may be subject to these limitations of OLM and PVMRM. Titration is generally a much faster mechanism for converting NO to N02 than photosynthesis, and as such is likely to be appropriate for characterizing peak I-hour N02 impacts in many cases. Both OLM and PVMRM rely on the same key inputs of in-stack N02/NOx ratios and hourly ambient ozone concentrations. Although both methods can be applied within the AERMOD model using a single "representative" background ozone concentration, it is likely that use of a single value would result in very conservative estimates of peak hourly ambient concentrations since its use for the I-hour N02 standard would be contingent on a demonstration of conservatism for all hours modeled. Furthermore, hourly monitored ozone concentrations used with the OLM and PVMRM options must be concurrent with the meteorological data period used in the modeling analysis, and thus the temporal representativeness of the ozone data for estimating ambient N02 concentrations could be a factor in determining the appropriateness of the meteorological data period for a particular application. As noted above, the representativeness of these key inputs takes on somewhat greater importance in the context of a I-hour N0 2 standard than for an annual standard, for obvious reasons. In the case of hourly background ozone concentrations, methods used to substitute for periods of missing data may playa more significant role in determining the I-hour N0 2 modeled design value, and should therefore be given greater scrutiny, especially for data periods that are likely to be associated with peak hourly concentrations based on meteorological conditions and source characteristics. In other words, ozone data substitution methods that may have been deemed appropriate in prior applications for the annual standard may not be appropriate to use for the new I-hour standard. While these technical issues and considerations generally apply to both OLM and PVMRM, the importance of the in-stack N02INOx ratios may be more important for PVMRM than for OLM in some cases, due to differences between the two methods. The key difference between the two methods is that the amount of ozone available for conversion of NO to N02 is based simply on the ambient ozone concentration and is independent of source characteristics for OLM, whereas the amount of ozone available for conversion in PVMRM is based on the amount of ozone within the volume of the plume for an individual source or group of sources. The plume volume used in PVMRM is calculated on an hourly basis for each source/receptor 21

53 combination, taking into account the dispersive properties of the atmosphere for that hour. For a low-level release where peak hourly NOx impacts occur close to the source under stable/light wind conditions, the plume volume will be relatively small and the ambient N02 impact for such cases will be largely determined by the in-stack N02/NOx ratio, especially for sources with relatively close fenceline or ambient air boundaries. This example also highlights the fact that the relative importance of the in-stack N0 2 /NOx ratios may be greater for some applications than others, depending on the source characteristics and other factors. Assumptions regarding instack N0 2 INOx ratios that may have been deemed appropriate in the context of the annual standard may not be appropriate to usc for the new I-hour standard. In particular, it is worth reiterating that the 0.1 in-stack ratio often cited as the "default" ratio for OLM should not be treated as a default value for hourly N02 compliance demonstrations. Another difference between 0 LM and PVMRM that is worth noting here is the treatment of the titration mechanism for multiple sources of NOx. There are two possible modes that can be used for applying OLM to multiple source scenarios within AERMOD: (l) apply OLM to each source separately and assume that each source has all of the ambient ozone available for conversion of NO to N02; and (2) assume that sources whose plumes overlap compete for the available ozone and apply OLM on a combined plume basis. The latter option can be applied selectively to subsets of sources within the modeled inventory or to all modeled sources using the OLMGROUP keyword within AERMOD, and is likely to result in lower ambient N02 concentrations in most cases since the ambient N02 levels will be more ozone-limited. One of the potential refinements in application of the titration method incorporated in PVMRM is a technique for dynamically determining which sources should compete for the available ozone based on the relative locations of the plumes from individual sources, both laterally and vertically, on an hourly basis, taking into account wind direction and plume rise. While this approach addresses one of the implementation issues associated with OLM by making the decision of which sources should compete for ozone, there is only very limited field study data available to evaluate the methodology. Given the importance of the issue of whether to combine plumes for the OLM option, EPA has addressed the issue in the past through the Model Clearinghouse process. The general guidance that has emerged in those cases is that the OLM option should be applied on a sourceby-source basis in most cases and that combining plumes for application of OLM would require a clear demonstration that the plumes will overlap to such a degree that they can be considered as "merged" plumes. However, much of that guidance was provided in the context of applying the OLM method outside the dispersion model in a post-processing mode on an annual basis. The past guidance on this issue is still appropriate in that context since there is no realistic method to account for the degree of plume merging on an hourly basis throughout the modeling analysis when applied as a post-processor. However, the implementation of the OLM option within the AERMOD model applies the method on a source-by-source, receptor-by-receptor, and hour-byhour basis. As a result, the application of the OLMGROUP option within AERMOD is such that the sources only compete for the available ozone to the extent that each source contributes to the cumulative NOx concentration at each receptor for that hour. Sources which contribute significantly to the ambient NOx concentration at the receptor will compete for available ozone in proportion to their contribution, while sources that do not contribute significantly to the ambient NOx concentration will not compete for the ozone. Thus, the OLMGROUP option 22

54 implemented in AERMOD will tend to be "self-correcting" with respect to concerns that combining plumes for OLM will overestimate the degree of ozone limiting potential (and therefore underestimate ambient N02 concentrations). As a result of these considerations, we recommend that use of the "OLMGROUP ALL" option, which specifies that all sources will potentially compete for the available ozone, be routinely applied and accepted for all approved applications of the OLM option in AERMOD. This recommendation is supported by model-tomonitor comparisons of hourly N0 2 concentrations ii-om the application of AERMOD for the Atlanta N02 risk and exposure assessment (EPA, 2008), and recent re-evaluations of hourly N02 impacts fl'om the two field studies (New Mexico and Palaau) that were used in the evaluation of PVMRM (MACTEC, 2005). These model-to-monitor comparisons of hourly N02 concentrations show reasonably good performance using the "OLMGROUP ALL" option within AERMOD, with no indication of any bias to underestimate hourly N0 2 concentrations with OLMGROUP ALL. Furthermore, model-to-monitor comparisons based on OLM without the OLMGROUP option do exhibit a bias to overestimate hourly N0 2 concentrations. We will provide further details regarding these recent hourly N0 2 model-to-monitor comparisons at a later date. SUMMARY To summarize, we emphasize the following points: I. The 3-tiered screening approach recommended in Section of Appendix W for annual N0 2 assessments generally applies to the new I-hour N0 2 standard. 2. While generally applicable, application of the 3-tiered screening approach for assessments of the new I-hour N02 standard may entail additional considerations, such as the importance of key input data, including appropriate emission rates for the I-hour standard vs. the annual standard for all tiers, and the representativeness of in-stack N0 2 /NOx ratios and hourly background ozone concentrations for Tier 3 detailed screening methods. 3. Since the OLM and PVMRM methods in AERMOD are currently considered nonregulatory-default options, application of these options requires justification and approval by the Regional Office on a case-by-case basis as alternative modeling techniques, in accordance with Section 3.2.2, paragraph (e), of Appendix W. 4. Applications of the OLM option in AERMOD, subject to approval under Section e of Appendix W, should routinely utilize the "OLMGROUP ALL" option for combining plumes. 5. While the I-hour NAAQS for N02 is defined in terms of the 3-year average for monitored design values to determine attainment of the NAAQS, this definition does not preempt or alter the Appendix W requirement for use of 5 years of NWS meteorological data or at least I year of site specific data. REFERENCES Cimorelli, A../., S. G. Perry, A. Venkatram, J. C. Weil, R. J. Paine, R. B. Wilson, R. F. Lee, W. 23

55 D. Peters, R. W. Brode, and J. O. Paumier, AERMOD: Description of Model Formulation with Addendum, EPA-4S4/R U.S. Environmental Protection Agency, Research Triangle Park, NC. Cole, B.S. and J.E. Summerhays, A Review of Techniques Available for Estimation of Short-Term N02 Concentrations. Journal of the Air Pollution Control Association, 29(8): Chu, S.B. and E.L. Meyer, Use of Ambient Ratios to Estimate Impact of NO x Sources on Annual N0 2 Concentrations. Proceedings, 84th Annual Meeting & Exhibition of the Air & Waste Management Association, Vancouver, B.C.; June (l6pp.) (Docket No. A , II-A-9) EPA, New Source Review Workshop Manual: Prevention of Significant Deterioration and Nonattainment Area Permitting - DRAFT. U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, User's Guide for the AMS/EPA Regulatory Model- AERMOD. EPA-4S4/B U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, Risk and Exposure Assessment to Support the Review of the N0 2 Primary National Ambient Air Quality Standard. Office of Air Quality Planning and Standards, Research Triangle Park, NC. EPA, Addendum - User's Guide for the AMS/EPA Regulatory Model- AERMOD. EPA-4S4/B U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA,2010a. Applicability of the Federal Prevention of Significant Deterioration Permit Requirements to New and Revised National Ambient Air Quality Standards. Stephen D. Page Memorandum, dated April I, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA,2010b. Modeling Procedures for Demonstrating Compliance with PM 2. 5 NAAQS. Stephen D. Page Memorandum, dated March 23,2010. U.S. Environmental Protection Agency, Research Triangle Park, NC. Hanrahan, P.L., I 999a. The Plume Volume Molar Ratio Method for Determining N0 2 /NOx Ratios in Modeling - Part I: Methodology..! Air & Waste Manage. Assoc., 49, Hanrahan, P.L., 1999b. The Plume Volume Molar Ratio Method for Determining N0 2 /NOx Ratios in Modeling - Part II: Evaluation Studies..! Air & Waste Manage. Assoc., 49, MACTEC,2004. Sensitivity Analysis ofpvmrm and OLM in AERMOD. Final Report, Alaska DEC Contract No MACTEC Federal Programs, Inc., Research Triangle Park, NC. 24

56 MACTEC,2005. Evaluation of Bias in AERMOD-PVMRM. Final Report, Alaska DEC Contract No MACTEC Federal Programs, Inc., Research Triangle Park, NC. cc: Richard Wayland, C Anna Wood, C Raj Rao, C Roger Brode, C Dan deroeck, C Elliot Zenick, OGC Brian Doster, OGe EPA Regional Modeling Contacts 25

57 Introduction ATTACHMENT A Background on Hourly NOx Emissions for Permit Modeling for the I-hour N0 2 NAAQS The purpose of this attachment is to address questions about availability of hourly NOx emissions for permit modeling under the new N0 2 NAAQS. It summarizes existing guidance regarding emission input data requirements for NAAQS compliance modeling, and provides background on the historical approach to development of inventories for N02 permit modeling and computation of hourly emissions appropriate for assessing the new I -hour N0 2 standard. Although the NAAQS is defined in terms of ambient N0 2 concentrations, source emission estimates for modeling are based on NOx. Under the PSD program, the owner or operator of the source is required to demonstrate that the source does not cause or contribute to a violation of a NAAQS (40 CFR (k)(1) and 40 CFR (k)(1» and/or PSD increments (40 CFR (k)(2) and (k)(2». However, estimation of the necessary emission input data for NAAQS compliance modeling entails consideration of numerous factors, and the appropriate reviewing authority should be consulted early in the process to determine the appropriate emissions data for use in specific modeling applications (see 40 CFR 51, Appendix W, 8.l.l.b and b) Summary of Current Guidance Section 8. I of the Guideline on Air Quality Models, Appendix W to 40 CFR Part 5 I, provides recommendations regarding source emission input data needed to support dispersion modeling for NAAQS compliance demonstrations. Table 8-2 of Appendix W provides detailed guidance regarding the specific components of the emission input data, including the appropriate emission limits (pounds/mmbtu), operating level (MMBtu/hr), and operating factor (e.g., hr/yr or hrlday), depending on the averaging time of the standard. Table 8-2 also distinguishes between the emission input data needed for the new or modified sources being assessed, and "nearby" and "other" background sources included in the modeled emission inventory. Based on Table 8-2, emission input data for new or modified sources for annual and quarterly standards are essentially the same as for short-term standards (:'0 24 hours), based on maximum allowable or federally enforceable emission limits, design capacity or federally enforceable permit conditions, and the assumption of continuous operation. However, there are a few additional considerations cited in Appendix W that could result in diffcrent emission input data for the I -hour vs. annual N02 NAAQS. For example, while design capacity is listed as the recommended operating level for the emission calculation, peak hourly ground-level concentrations may be more sensitive than annual average concentrations to changes in stack parameters (effluent exit temperature and exit velocity) under different operating capacities. Table 8-2 specifically rccommends modeling other operating levels, such as 50 percent or 75 percent of capacity, for short-term standards (see footnote 3). Another factor that may affect maximum ground-level concentrations differently between the I -hour vs. annual standard is

58 restrictions on operating factors based on federally enforceable permit conditions. While federally enforceable operating factors other than continuous operation may be accounted for in the emission input data (e.g., if operation is limited to 8 am to 4 pm each day), Appendix Walso states that modeled emissions should not be averaged across non-operating time periods (see footnote 2 of Table 8-2). While emission input data recommendations for "nearby" and "other" background sources included in the modeled emission inventory are similar to the new or modified source emission inputs in many respects, there is an important difference in the operating factor between annual and short-term standards. Emission input data for nearby and other sources may reflect actual operating factors (averaged over the most recent 2 years) for the annual standard, while continuous operation should be assumed for short-term standards. This could result in important differences in emission input data for modeled background sources for the I-hour N02 NAAQS relative to emissions used for the annual standard. Model Emission Inventory for NOz Modeling For the existing annual N02 NAAQS, the permit modeling inventory has generally been compiled from the annual state emission inventory questionnaire (EIQ) or Title V permit applications on file with the relevant permitting authority (state or local air program). Since a state uses the annual EIQ for Title V fee assessment, the state EIQ typically requires reporting of unit capacity, total fuel combusted, and/or hours of operation to help verify annual emissions calculations for fee accuracy purposes. Likewise, Title V operating permit applications contain all of the same relevant information for calculating emissions. While these emission inventories are important resources for gathering emission input data on background sources for NAAQS compliance modeling, inventories which are based on actual operations may not be sufficient for short-term standards, such as the new I-hour N02 NAAQS. However, appropriate estimates of emissions from background sources for the I-hour N02 standard may be derived in many cases from information in these inventories regarding permitted emission limits and operating capacity. Historically, it has not been a typical practice for an applicant to use the EPA's national emission inventory (NEI) as the primary source for compiling the permit modeling inventory. Since the emission data submitted to the NEI represents annual emission totals, it may not be suitable for use in NAAQS compliance modeling for short-term standards since modeling should be based on continuous operation, even for modeled background sources. Although the NEI may provide emission data for background sources that are more appropriate for the annual N02 standard, the utility of the NEI for purposes of NAAQS compliance modeling is further limited due to the fact that additional information regarding stack parameters and operating rates required for modeling may not be available from the NEI. While records exist in the NEI for reporting stack data nccessary for point source modeling (i.e., stack coordinates, stack heights, exit temperatures, exit velocities), some states do not report such information to the NEl, or there are may be errors in the location data submitted to the NEI. Under such conditions, default stack information based upon SIC is substituted and use of such data could invalidate modeling results. Building locations and dimensions, which may be required to account for building down wash influences in the modeling analysis, may also be missing or incomplete in many cases. 27

59 A common and relatively straightforward approach for compiling the necessary information to develop an inventory of emissions from background sources for a permit modeling demonstration is as follows, patterned after the draft New Source Review Workshop Manual (EPA, 1990). The applicant completes initial modeling of allowable emission increases associated with the proposed project and determines the radii of impact (ROI) for each pollutant and averaging period, based on the maximum distance at which the modeled ambient concentration exceeds the Significant Impact Level (SIL) for each pollutant and averaging period. Typically, the largest ROI is selected and then a list of potential background sources within the ROI plus a screening distance beyond the ROI is compiled by the permitting authority and supplied to the applicant. The applicant typically requests permit applications or ElQ submittals from the records department of the permitting authority to gather stack data and source operating data necessary to compute emissions for the modeled inventory. Once the applicant has gathered the relevant data from the permitting authorities, model emission rates are calculated. While this approach is fairly common, it should be noted that the draft workshop manual "is not intended to be an official statement of policy and standards and does not establish binding regulatory requirements" (see, Preface), and the appropriate reviewing authority should be consulted early in the process regarding the selection of appropriate background source emission inventories for the I-hour N02 standard. We also note that Appendix W establishes "a significant concentration gradient in the vicinity of the source" under consideration as the main criterion for selection of nearby sources for inclusion in the modeled inventory, and further indicates that "the number of such [nearby] sources is expected to be small except in unusual situations." See Section b. As mentioned previously, modeled emission rates for short-term NAAQS are computed consistent with the recommendations of Section 8.1 of Appendix W, summarized in Table 8-2. The maximum allowable (SIP-approved process weight rate limits) or federally enforceable permit limit emission rates assuming design capacity or federally enforceable capacity limitation are used to compute hourly emissions for dispersion modeling against short-term NAAQS such as the new I-hour N02 NAAQS. If a source assumes an enforceable limit on the hourly firing capacity of a boiler, this is reflected in the calculations. Otherwise, the design capacity of the source is used to compute the model emission rate. A load analysis is typically necessary to determine the load or operating condition that causes the maximum ground-level concentrations. In addition to 100 percent load, loads such as 50 percent and 75 percent are commonly assessed. As noted above, the load analysis is generally more important for short-term standards than for annual standards. For an hourly standard, other operating scenarios of relatively short duration such as "startup" and "shutdown" should be assessed since these conditions may result in maximum hourly ground-level concentrations, and the control efficiency of emission control devices during these operating conditions may also need to be considered in the emission estimation. Emission Calculation Example The hourly emissions are most commonly computed from AP-42 emission factors based on unit design capacity. For a combustion unit, the source typically reports both the unit capacity and the actual total amount of fuel combusted annually (gallons, millions of cubic feet 28

60 of gas, etc.) to the permitting authority for the EIQ. Likewise, Title V operating permit applications will contain similar information that can be used to compute hourly emissions. For example, assume you are modeling an uncontrolled natural gas package boiler with a design firing rate of 30 MMBtu/hr. The AP-42 emission factor for an uncontrolled natural gas external combustion source (AP-42, Section 1.4) for firing rates less than 100 MMBtu/hr is 100 lbs. NOx/1 0 6 SCI' natural gas combusted. The hourly emission rate is derived by converting the emission factor expressed in terms of lbs. NOx/l 0 6 SCF to lbs. NOx/MMBtu. The conversion is done by dividing the 100 lbs. NOx/l 0 6 SCF by 1,020 to convert the AP-42 factor to lbs. NOx/MMBtu. The new emission factor is now lbs. NOx/MMBtu. For this example, the source has no limit on the hourly firing rate of the boiler; therefore, the maximum hourly emissions are computed by multiplying the design firing rate of the boiler by the new emission factor. Ehoudy = Ibs/MMBtu x 30MMBtu/hr = 2.94 lbs/hr Thus 2.94 lbs/hr represents the emission rate that would be input into the dispersion model for modeling against the I-hour N0 2 NAAQS to comport with emission rate recommendations of Section 8.1 of Appendix W. It is important to note that data derived for the annual state emission inventory (EI) is based on actual levels of fuel combusted for the year, and is therefore different than how allowable emissions are computed for near-field dispersion modeling. For the annual EI report, a source computes their annual emissions based upon the AP-42 emission factor multiplied by the actual total annual throughput or total fuel combusted. In the 30 MMBtu/hr boiler example, the annual NOx emissions reported to the NEI is computed by: Eam",al = (AP-42 emission factor) x (total annual fuel combusted) E wl ""al = (100 lbs/l 0 6 SCF) x ( SCF/yr) = 10,000 Ibs. NOx/yr or 5 tons NOx/yr 29

61 UNITED STATES ENVIRONMENTAL PROTECTION AGENCY RESEARCH TRIANGLE PARK, NC MAR OFFICE OF AIR QUALITY PLANNING AND STANDARDS MEMORANDUM SUBJECT: FROM: TO: Additional Clarification Regarding Application of Appendix W Modeling Guidance for the 1-hour N02,National Ambient Air Quality Standard ~(_ 'f--d:.c Tyler Fox, Leader / d J 1 - Air Quality Modeling Group, C Regional Air Division Directors INTRODUCTION On January 22, 2010, EPA announced a new 1-hour nitrogen dioxide (N~) National Ambient Air Quality Standard (1-hour N02 NAAQS or 1-hour N02 standard) that is attained when the 3-year average of the 98th-percentile of the annual distribution of daily maximum 1-hour concentrations does not exceed 100 ppb at each monitor within an area. The final rule for the new 1-hour N02 NAAQS was published in the Federal Register on February 9, 2010 (75 FR ), and the standard became effective on April12, 2010 (EPA, 2010a). A memorandum was issued on June 29, 2010, clarifying the applicability of current guidance in the Guideline on Air Quality Models (40 CFR Part 51, Appendix W) for modeling N02 impacts in accordance with the Prevention of Significant Deterioration (PSD) permit requirements to demonstrate compliance with the new 1-hour N02 standard. This memorandum supplements the June 29, 2010 guidance memo by providing further clarification and guidance on the application of Appendix W guidance for the!-hour N02 standard. Note that while the discussion ofnox chemistry options in this memo is exclusive to the 1-hour N02 standard, the discussion of other topics in this memo should apply equally to the 1-hour S~ standard, accounting for the slight differences in the form of the 1-hour N02 and S02 standards 1 In summary, the memo: 1. Clarifies procedures for demonstrating compliance with the 1-hour N02 NAAQS based on the form of the standard, including significant contribution analyses using the interim Significant Impact Level (SIL) established in the June 29, 2010 memo, 1 The!-hour N02 standard is based on the 98~ 1 -percentile (8'h highest) of the annual distribution of maximum daily 1-hour values, whereas the 1-hour so2 standard is based on the 991h-percentile (4 1 h-highest) of the annual distribution of maximum daily!-hour values. Internet Address (URL) Recycled/Recyclable Printed whh Vegetable Oil Based Inks on Recycled Paper (Minimum 25% Postconsumer)

62 and details updates to the AERMOD model with an internal post-processor option that supports such analyses. 2. Provides clarification on the use and acceptance of Tier 2 and Tier 3 options for NO 2, including updated model evaluation results for the OLM and PVMRM options incorporated in the AERMOD model. 3. Recommends that compliance demonstrations for the 1-hour NO 2 NAAQS address emission scenarios that can logically be assumed to be relatively continuous or which occur frequently enough to contribute significantly to the annual distribution of daily maximum 1-hour concentrations based on existing modeling guidelines, which provide sufficient discretion for reviewing authorities to not include intermittent emissions from emergency generators or startup/shutdown operations from compliance demonstrations for the 1-hour NO 2 standard under appropriate circumstances. 4. Provides additional clarification and a more detailed discussion of the factors to consider in determination of background concentrations as part of a cumulative impact assessment including identification of nearby sources to be explicitly modeled. 5. Recommends an appropriate methodology for incorporating background concentrations in the cumulative impact assessment for the 1-hour NO 2 standard and details updates to the AERMOD model with an option to include temporally-varying background concentrations within the modeling analysis. PROCEDURES FOR DEMONSTRATING COMPLIANCE WITH 1-HOUR NO 2 NAAQS Compliance with the 1-hour NO 2 NAAQS is based on the multiyear average of the 98 th - percentile of the annual distribution of daily maximum 1-hour values not exceeding 100 ppb. The 8 th -highest of the daily maximum 1-hour values across a year is an unbiased surrogate for the 98 th -percentile 1. The AERMOD dispersion model, EPA s preferred model for near-field applications under Appendix W, was recently modified (version dated 11059) to fully support the form of the 1-hour NO 2 NAAQS, as well as other analyses that may be needed in order to demonstrate that a source does not cause or contribute to a violation of the NAAQS based on the interim SIL established in the June 29, 2010, memorandum. Application of Interim SIL to Project Impacts Using the interim 1-hour NO 2 SIL, a permit applicant can determine: (1) whether, based on the proposed increase in NO x emissions, a cumulative air quality analysis is required; (2) the area of impact within which a cumulative air quality analysis should focus; and (3) whether the proposed source s NO x emissions will contribute to any modeled violation of the 1-hour NO 2 NAAQS identified in the cumulative analysis. To determine initially whether a proposed project s emissions increase will have a significant impact (resulting in the need for a cumulative impact assessment), the June 29, 2010, memorandum recommended that the interim SIL should be compared to either of the following: 2

63 The highest of the 5-year averages of the maximum modeled 1-hour NO 2 concentrations predicted each year at each receptor, based on 5 years of National Weather Service data; or The highest modeled 1-hour NO 2 concentration predicted across all receptors based on 1 year of site-specific meteorological data, or the highest of the multi-year averages of the maximum modeled 1-hour NO 2 concentrations predicted each year at each receptor, based on 2 or more years, up to 5 complete years of available sitespecific meteorological data. Since the form of the standard is based on the annual distribution of daily maximum 1-hour values, the maximum contribution that a project could make to the air quality impact at a receptor is the multiyear average of the highest 1-hour values at that receptor. If the multiyear average of the highest 1-hour values is below the SIL at all receptors, then the project could not contribute significantly to any modeled violations of the 1-hour NO 2 NAAQS, thus exempting that project from the cumulative impact assessment. Application of Interim SIL to Cumulative Impact Assessment If a project s impacts exceed the SIL at any receptors based on this initial impact analysis, then a cumulative impact assessment should be completed to determine whether the project will cause or contribute to any modeled violations of the NAAQS. While not common practice in the past, given the more complex analysis procedures associated with the form of the 1-hour NO 2 NAAQS, we deem it appropriate and acceptable in most cases to limit the cumulative impact analysis to only those receptors that have been shown to have significant impacts from a proposed new source based on the initial SIL analysis, assuming that the design of the original receptor grid was adequate to determine all areas of ambient air where the source could contribute significantly to modeled violations. This may especially be appropriate for the 1-hour NO 2 standard since the initial modeling of the project emissions without other background emission sources may have a tendency to overestimate ambient NO 2 concentrations, even under Tier 3 applications, by understating the potential ozone limiting influence of the background NO x emissions. If modeled violations of the NAAQS are found based on the cumulative impact assessment, then the project s contribution to all modeled violations should be compared to the interim SIL to determine whether the project causes or contributes to any of the modeled violations. In past guidance (EPA, 1988), EPA has indicated that the significant contribution analysis should be based on a source s contribution to the modeled violation paired in time and space. The form of the 1-hour NO 2 NAAQS complicates this analysis since the modeled violation is based on a multiyear average of the annual distribution of daily maximum 1-hour values, i.e., a particular modeled violation at a particular receptor represents an average based on specific hours on specific days from each of the five years of meteorological data (for National Weather Service (NWS) data). It is important to point out here that the significant contribution analysis is not limited to analyzing the source s contribution associated only with the modeled design value based on the 98 th -percentile cumulative air quality impact at the receptor, but rather must examine all cases where the cumulative impact exceeds the NAAQS at or below the 98 th - 3

64 percentile. In some cases a source s contribution to the 98 th -percentile of the daily maximum 1- hour values from the cumulative impact (i.e., the cumulative impact value or modeled design value that is compared to the NAAQS) may be below the SIL, while the source s contribution to cumulative impacts below the 98 th -percentile but above the NAAQS could exceed the SIL. Therefore, the significant contribution analysis should examine every multiyear average of daily maximum 1-hour values, beginning with the 8 th -highest (98 th -percentile) 2, continuing down the ranked distribution until the cumulative impact is below the NAAQS. Since the form of the standard is based on the annual distribution of daily maximum 1-hour values, the significant contribution analysis should be limited to the distribution of daily maximum 1-hour values, i.e., the 2 nd, 3 rd, 4 th -highest 1-hour values during the day, and so on, are not considered in this analysis. In addition, for applications with more than one year of meteorological data, the significant contribution analysis should only examine ranks paired across the years, i.e., the multiyear average of the N th -highest values across each of the years processed. The recent update to the AERMOD model (dated 11059) includes an option (the MAXDCONT keyword) to automatically perform this contribution analysis (EPA, 2010b), examining the contribution from project emissions to the cumulative impacts at each receptor across a user-specified range of ranked values, paired in time and space, as an internal post-processor within the model. Other options are available in the recent AERMOD update that identify the specific data periods contributing to the cumulative modeled impacts at each receptor. Applicability of Ambient Monitoring Requirements to Modeling Demonstrations The June 29, 2010 memo addressed one aspect of the applicability of ambient monitoring requirements, set forth in Appendix S to 40 CFR Part 50 in relation to the 1-hour NO 2 standard 3, to modeling applications to demonstrate compliance with the NAAQS, namely the use of 3 years of ambient monitoring data as the basis for attainment of the NAAQS using monitoring vs. the use of 5 years of meteorological data for modeling demonstrations of compliance with the NAAQS. Specifically, the June 29, 2010 memo indicated that Although the monitored design value for the 1-hour NO 2 standard is defined in terms of the 3-year average, this definition does not preempt or alter the Appendix W requirement for use of 5 years of NWS meteorological data or at least 1 year of site specific data. The 5-year average based on use of NWS data, or an average across one or more years of available site specific data, serves as an unbiased estimate of the 3-year average for purposes of modeling demonstrations of compliance with the NAAQS. Modeling of rolling 3-year averages, using years 1 through 3, years 2 through 4, and years 3 through 5, is not required. We would also like to emphasize that other aspects of the ambient monitoring requirements for the 1-hour NO 2 standard should not be applied for modeling analyses to demonstrate compliance with the NAAQS. For example, Appendix S addresses the data completeness requirements for monitored NO 2 concentrations, procedures for handling missing data periods, and conventions for rounding of monitored values. Appendix S specifies that a sampling day is complete if at least 75 percent of the hourly values are valid and a quarter is complete if at least 75 percent of the sampling days have complete data, and establishes calculation procedures for identifying the monitored design value that should be compared to the 2 For the 1-hour SO 2 standard the analysis should begin with the 4 th -highest, or 99 th -percentile value. 3 Appendix T to 40 CFR Part 50 addresses ambient monitoring requirements for the 1-hour SO 2 standard. 4

65 NAAQS. While the requirements of Appendix S are appropriate in the context of ambient monitoring, application of these requirements and procedures to a dispersion modeling analysis is not appropriate and may conflict with modeling guidance in many cases. Appendix W provides guidance on data completeness for meteorological data which specifically addresses the needs of dispersion modeling, including procedures that are explicitly implemented within the meteorological processor and dispersion model to account for missing data due to calm winds or other factors. Adjustments to the calculation procedures for determining the modeled design value for comparison to the NAAQS based on Appendix S data completeness criteria is not appropriate. The EPA Model Clearinghouse has also issued guidance in the past that modeled concentrations should not be rounded before comparing the modeled design value to the NAAQS. The fundamental point to recognize here is that ambient monitoring requirements/procedures and dispersion modeling guidance/procedures address different issues and needs relative to each aspect of air quality assessment, and are often motivated by different concerns and exigencies. APPROVAL AND APPLICATION OF TIERING APPROACH FOR NO 2 Given the stringency of the 1-hour NO 2 standard relative to the annual standard, many more permit applicants may find it necessary to use the less conservative Tier 2 or Tier 3 approaches in order to demonstrate compliance with the new NAAQS rather than relying on the Tier 1 assumption of full conversion. The June 29, 2010 memo highlighted some of the potential issues that may need to be addressed in the application of these less conservative assumptions for estimating ambient NO 2 impacts, relative to the Tier 1 option of full conversion, and clarified the status of the Tier 3 PVMRM and OLM approaches available as non-regulatory-default options within the AERMOD model. In order to ease the burden on permit applicants in addressing the need to demonstrate compliance with the 1-hour NO 2 NAAQS, as well as the burden on the permitting authority in reviewing such applications, we offer additional discussion and recommendations in relation to the use of Tier 2 and Tier 3 options. Specifically, we recommend the following: Use of 0.80 as a default ambient ratio for the 1-hour NO 2 standard under Tier 2 without additional justification by applicants; and General acceptance of 0.50 as a default in-stack ratio of NO 2 /NO x for input to the PVMRM and OLM options within AERMOD, in the absence of more appropriate source-specific information on in-stack ratios. The following sections explain these recommendations in more detail and also discuss the relative merits of the PVMRM and OLM options, clarifying that we have not indicated any preference of one option over the other. We also provide updated model evaluation results for the PVMRM and OLM options in AERMOD that lend further credence to the use of these Tier 3 options for 1-hour NO 2 compliance demonstrations. We anticipate that these recommendations and updated model evaluations will simplify and facilitate the process of gaining approval for use of these non-regulatory default options in AERMOD. 5

66 Tier 2 Ambient Ratio Method (ARM) for NO-to-NO 2 Conversion Regarding the Tier 2 option of applying an ambient ratio to the Tier 1 result, the June 29, 2010 memo cautioned against use of the 0.75 national default ratio recommended in Appendix W for the annual standard for estimating hourly NO 2 impacts, without some justification of the appropriateness of that assumption. We still do not consider 0.75 as an appropriate default ambient ratio for the 1-hour standard, but several references cite ambient ratios of about 0.80 for hourly NO 2 /NO x (e.g., Wang, et al., 2011; Janssen, et al., 1991), and we believe it would be appropriate to accept that as a default ambient ratio for the 1-hour NO 2 standard. Consideration was given to adopting the default equilibrium ratio of 0.90 incorporated in the PVMRM option as an hourly ARM, but we do not consider that to be an appropriate choice since it is the maximum ratio applied on a source-by-source and hourly basis, irrespective of the predicted hourly NO x concentration, whereas the Tier 2 ARM of 0.80 would be applied to the maximum cumulative hourly NO x concentration. Tier 3 Options for NO-to-NO 2 Conversion The June 29, 2010 memo clarified that the OLM and PVMRM options in the AERMOD model should be considered as Tier 3 applications under Section of Appendix W. Also, since the OLM and PVMRM methods are currently implemented as non-regulatory-default options within the AERMOD dispersion model (Cimorelli, et al., 2004; EPA, 2004; EPA, 2010b), their use requires justification and approval by the Regional Office on a case-by-case basis, pursuant to Sections c, a, and A.1.a(2) of Appendix W. The June 29 memo also highlighted the importance of two key model inputs for both the OLM and PVMRM options in the context of the 1-hour NO 2 standard, namely the in-stack ratios of NO 2 /NO x emissions and background ozone concentrations. This section provides additional discussion of these key inputs for OLM and PVMRM and also clarifies the similarities and differences between these methods and discusses their relative merits for purposes of demonstrating compliance with the 1- hour NO 2 standard. As noted in the June 29, 2010 memo, limited evaluations of PVMRM have been completed which show encouraging results, but the amount of data currently available is too limited to justify a designation of PVMRM as a refined method for NO 2 (Hanrahan, 1999; MACTEC, 2005). Furthermore, the original evaluations focused on model performance for annual averages since the only NO 2 standard in effect at the time was annual. We have recently updated the evaluations to reflect the current AERMOD modeling system components and extended them to examine model performance for hourly NO 2 concentrations. Preliminary results from these recent evaluations are presented in Attachment A. While the limited scope of the available field study data imposes limits on the ability to generalize conclusions regarding model performance, these preliminary results of hourly NO 2 predictions for Palaau and New Mexico show generally good performance for the PVMRM and OLM/OLMGROUP ALL options in AERMOD. We believe that these additional model evaluation results lend further credence to the use of these Tier 3 options in AERMOD for estimating hourly NO 2 concentrations, and we recommend that their use should be generally 6

67 accepted provided some reasonable demonstration can be made of the appropriateness of the key inputs for these options, the in-stack NO 2 /NO x ratio and the background ozone concentrations. Although well-documented data on in-stack NO 2 /NO x ratios is still limited for many source categories, we also feel that it would be appropriate in the absence of such source-specific instack data to adopt a default in-stack ratio of 0.5 as being adequately conservative in most cases and a better alternative to use of the Tier 1 full conversion or Tier 2 ambient ratio options. This value appears to represent a reasonable upper bound based on the available in-stack data. We hope that over time the range of source categories for which in-stack ratio information is available increases and the quality of such information will improve. These preliminary model evaluation results also serve to highlight a point worth emphasizing, which is that the PVMRM option in AERMOD is not inherently superior to the OLM option for purposes of estimating cumulative ambient NO 2 concentrations. The June 29, 2010 memo indicated that both PVMRM and OLM should be considered as Tier 3 options, but did not indicate any preference between these two options. Both PVMRM and OLM simulate the same basic chemical mechanism of ozone titration, the interaction of NO with ambient ozone (O 3 ) to form NO 2 and O 2. The main distinction between PVMRM and OLM is the approach taken to estimate the ambient concentrations of NO and O 3 for which the ozone titration mechanism should be applied. For isolated elevated point sources, the PVMRM option does represent a more refined treatment of ozone titration since it estimates the NO and O 3 available for conversion based on simulating the actual volume of the instantaneous plume as it is transported downwind. As a result, this method will generally provide a more realistic simulation of the NO-to-NO 2 conversion rate along the path of the plume for a particular source, accounting for the influence of meteorological conditions on the entrainment of O 3 associated with growth of the plume. However, the algorithm incorporated in PVMRM for determining which plumes compete for available ozone for multi-source applications has not been thoroughly validated, and as shown in the model evaluation results for New Mexico, PVMRM may not always provide a better answer than the OLM option. The PVMRM algorithm as currently implemented may also have a tendency to overestimate the conversion of NO to NO 2 for low-level plumes by overstating the amount of ozone available for the conversion due to the manner in which the plume volume is calculated. The plume volume calculation in PVMRM does not account for the fact that the vertical extent of the plume based on the vertical dispersion coefficient may extend below ground for low-level plumes. This overestimation of the volume of the plume could contribute to overestimating conversion to NO 2. The PVMRM option has further limitations for area source applications, especially for elongated area sources that may be used to simulate road segments. In these cases, the lateral extent of the plume used in calculating the plume volume depends on the projected width of the area source, even if only a portion of the area source actually impacts a nearby receptor. This again would tend to overestimate the volume of the plume for purposes of determining the amount of ozone available for conversion of NO to NO 2, and would likely overestimate ambient NO 2 concentrations. In light of these issues, a series of volume sources rather than elongated area sources is recommended for simulating NO 2 impacts from roadway emissions with PVMRM, especially for receptors located relatively close to the roadway. Furthermore, the OLM option with OLMGROUP ALL was used to estimate NO 2 concentrations from mobile source emissions modeled as area sources for the Atlanta area as part of the EPA s 7

68 Risk and Exposure Assessment (REA) for the most recent NO 2 NAAQS review (EPA, 2008). Results of model-to-monitor comparisons from the REA show generally good performance, suggesting that use of OLM with OLMGROUP ALL is appropriate for modeling such emissions. TREATMENT OF INTERMITTENT EMISSIONS Modeling of intermittent emission units, such as emergency generators, and/or intermittent emission scenarios, such as startup/shutdown operations, has proven to be one of the main challenges for permit applicants undertaking a demonstration of compliance with the 1- hour NO 2 NAAQS. Prior to promulgation of the new 1-hour NO 2 standard, the only NAAQS applicable for NO 2 was the annual standard and these intermittent emissions typically did not factor significantly into the modeled design value for the annual standard. Sources often take a 500 hour/year permit limit on operation of emergency generators for purposes of determining the potential to emit (PTE), but may actually operate far fewer hours than the permitted limit in many cases and generally have not been required to assume continuous operation of these intermittent emissions for purposes of demonstrating compliance with the annual NAAQS. Due in part to the relatively low release heights typically associated with emergency generators, an assumption of continuous operation for these intermittent emissions would in many cases result in them becoming the controlling emission scenario for determining compliance with the 1-hour standard. EPA s guidance in Table 8-2 of Appendix W involves a degree of conservatism in the modeling assumptions for demonstrating compliance with the NAAQS by recommending the use of maximum allowable emissions, which represents emission levels that the facility could, and might reasonably be expected to, achieve if a PSD permit is granted. However, the intermittent nature of the actual emissions associated with emergency generators and startup/shutdown in many cases, when coupled with the probabilistic form of the standard, could result in modeled impacts being significantly higher than actual impacts would realistically be expected to be for these emission scenarios. The potential overestimation in these cases results from the implicit assumption that worst-case emissions will coincide with worst-case meteorological conditions based on the specific hours on specific days of each of the years associated with the modeled design value based on the form of the hourly standard. In fact, the probabilistic form of the standard is explicitly intended to provide a more stable metric for characterizing ambient air quality levels by mitigating the impact that outliers in the distribution might have on the design value. The February 9, 2010, preamble to the rule promulgating the new 1-hour NO 2 standard stated that it is desirable from a public health perspective to have a form that is reasonably stable and insulated from the impacts of extreme meteorological events. 75 FR Also, the Clean Air Science Advisory Committee (CASAC) recommended a 98 th -percentile form averaged over 3 years for such a standard, given the potential for instability in the higher percentile concentrations around major roadways. 75 FR To illustrate the importance of this point, consider the following example. Under a deterministic 1-hour standard, where the modeled design value would be based on the highest of the second-highest hourly impacts (allowing one exceedance per year), a single emission episode lasting 2 hours for an emergency generator or other intermittent emission scenario could 8

69 determine the modeled design value if that episode coincided with worst-case meteorological conditions. While the probability of a particular 2-hour emission episode actually coinciding with the worst-case meteorological conditions is relatively low, there is nonetheless a clear linkage between a specific emission episode and the modeled design value. By contrast, under the form of the 1-hour NO 2 NAAQS only one hour from that emission episode could contribute to the modeled design value, i.e., the daily maximum 1-hour value. However, by assuming continuous operation of intermittent emissions the modeled design value for the 1-hour NO 2 NAAQS effectively assumes that the intermittent emission scenario occurs on the specific hours of the specific days for each of the specific years of meteorological data included in the analysis which factor into the multiyear average of the 98 th -percentile of the annual distribution of daily maximum 1-hour values. The probability of the controlling emission episode occurring on this particular temporal schedule to determine the design value under the probabilistic standard is significantly smaller than the probability of occurrence under the deterministic standard; thereby increasing the likelihood that impact estimates based on assuming continuous emissions would significantly overestimate actual impacts for these sources. Given the implications of the probabilistic form of the 1-hour NO 2 NAAQS discussed above, we are concerned that assuming continuous operations for intermittent emissions would effectively impose an additional level of stringency beyond that intended by the level of the standard itself. As a result, we feel that it would be inappropriate to implement the 1-hour NO 2 standard in such a manner and recommend that compliance demonstrations for the 1-hour NO 2 NAAQS be based on emission scenarios that can logically be assumed to be relatively continuous or which occur frequently enough to contribute significantly to the annual distribution of daily maximum 1-hour concentrations. EPA believes that existing modeling guidelines provide sufficient discretion for reviewing authorities to exclude certain types of intermittent emissions from compliance demonstrations for the 1-hour NO 2 standard under these circumstances. EPA s Guideline on Air Quality Models provides recommendations regarding air quality modeling techniques that should be applied in preparation or review of PSD permit applications and serves as a common measure of acceptable technical analysis when supported by sound scientific judgment. 40 C.F.R. Part 51, Appendix W, section 1.0.a. While the guidance establishes principles that may be controlling in certain circumstances, the guideline is not a strict modeling cookbook so that, as the guideline notes, case-by-case analysis and judgment are frequently required. Section 1.0.c. In particular, with respect to emissions input data, section 8.0.a. of Appendix W establishes the general principle that the most appropriate data available should always be selected for use in modeling analyses, and emphasizes the importance of the exercise of professional judgement by the appropriate reviewing authority in determining which nearby sources should be included in the model emission inventory. Section b. For the reasons discussed above, EPA believes the most appropriate data to use for compliance demonstrations for the 1-hour NO 2 NAAQS are those based on emissions scenarios that are continuous enough or frequent enough to contribute significantly to the annual distribution of daily maximum 1-hour concentrations. Section b of the guideline also provides that [t]he appropriate reviewing authority should be consulted to determine appropriate 9

70 source definitions and for guidance concerning the determination of emissions from and techniques for modeling various source types. When EPA is the reviewing authority for a permit, for the reasons described above, we will consider it acceptable to limit the emission scenarios included in the modeling compliance demonstration for the 1-hour NO 2 NAAQS to those emissions that are continuous enough or frequent enough to contribute significantly to the annual distribution of daily maximum 1-hour concentrations. Consistent with this rationale, the language in Section d of Appendix W states that [i]t is appropriate to model nearby sources only during those times when they, by their nature, operate at the same time as the primary source(s) being modeled. While we recognize that these intermittent emission sources could operate at the same time as the primary source(s), the discussion above highlights the additional level of conservatism in the modeled impacts inherent in an assumption that they do in fact operate simultaneously and continuously with the primary source(s). The rationale regarding treatment of intermittent emissions applies for both project emissions and any nearby or other background sources included in the modeling analysis. However, this rationale does not apply to the load analysis recommended in Table 8-2 of Appendix W, since various operating loads are not by design intended to be intermittent. Appendix W, Section a. With respect to the operating level, for the proposed new or modified source, Table 8-2 calls for using [d]esign capacity or federally enforceable permit condition. With respect to nearby sources, the guidelines call for estimating emissions based on [a]ctual or design capacity (whichever is greater), or federally enforceable permit condition. Footnote 3 to the table notes that [o]perating levels such as 50 percent and 75 percent of capacity should also be modeled to determine the load causing the highest concentration. The justification for not including certain intermittent operations described in this memo does not apply to these guidelines that address analyzing the load causing the highest concentration. We recognize that case-specific issues and factors may arise that affect the application of this guidance, and that not all facilities required to demonstrate compliance with the 1-hour NO 2 NAAQS will fit within the scenario described above with clearly defined continuous/normal operations vs. intermittent/infrequent emissions. Additional discretion may need to be exercised in such cases to ensure that public health is protected. For example, an intermittent source that is permitted to operate up to 500 hours per year, but typically operates much less than 500 hours per year and on a random schedule that cannot be controlled would be appropriate to consider under this guidance. On the other hand, an intermittent source that is permitted to operate only 365 hours per year, but is operated as part of a process that typically occurs every day, would be less suitable for application of this guidance since the single hour of emissions from each day could contribute significantly to the modeled design value based on the annual distribution of daily maximum 1-hour concentrations. Similarly, the frequency of startup/shutdown emission scenarios may vary significantly depending on the type of facility. For example, a large baseload power plant may experience startup/shutdown events on a relatively infrequent basis whereas as a peaking unit may go through much more frequent startup/shutdown cycles. It may be appropriate to apply this guidance in the former case, but not the latter. Another aspect of intermittent emissions worth noting is the distinction between intermittent emissions that can be scheduled with some degree of flexibility vs. intermittent emissions that cannot be scheduled. For example, a portion of emissions from an emergency 10

71 generator are likely to be associated with regular testing of the equipment that may be required to ensure its reliable operation, while that portion of emergency generator emissions associated with actual emergency use typically cannot be scheduled. In this case it may be appropriate to include a permit condition that restricts operation of the emergency generator during testing to certain hours of the day, which may mitigate that source s contribution to ambient NO 2 levels based on dispersion conditions. Limiting operation to specific time periods is an appropriate permit condition under Appendix W guidance and would not constitute a dispersion technique subject to Section 123 of the CAA. In this case the portion of the emissions associated with scheduled testing can be accounted for more realistically by limiting the hours modeled to account for meteorological conditions that are more representative of actual operations. Another approach that may be considered in cases where there is more uncertainty regarding the applicability of this guidance would be to model impacts from intermittent emissions based on an average hourly rate, rather than the maximum hourly emission. For example, if a proposed permit includes a limit of 500 hours/year or less for an emergency generator, a modeling analysis could be based on assuming continuous operation at the average hourly rate, i.e., the maximum hourly rate times 500/8760. This approach would account for potential worst-case meteorological conditions associated with emergency generator emissions by assuming continuous operation, while use of the average hourly emission represents a simple approach to account for the probability of the emergency generator actually operating for a given hour. Also note that the contribution of intermittent emissions to annual impacts should continue to be addressed as in the past to demonstrate compliance with the annual NO 2 standard. A final point of clarification regarding intermittent emissions that deserves some emphasis is that the guidance provided here in relation to determining compliance with the 1- hour NO 2 NAAQS through dispersion modeling has no effect on or relevance to the existing policies and guidance regarding excess emissions that may occur during startup and shutdown, where such excess emissions violate applicable emission limitations 4. In other words, all emissions from a new or modified source are subject to the applicable permitted emission limits and may be subject to enforcement action regarding such excess emissions, regardless of whether a portion of those emissions are not included in the modeling demonstration based on the guidance provided here. Given the added complexity of the technical issues that arise in the context of demonstrating compliance with the 1-hour NO 2 NAAQS through dispersion modeling, we strongly encourage adherence to the recommendations in Section of Appendix W that [e]very effort should be made by the Regional Office to meet with all parties involved in either a SIP revision or a PSD permit application prior to the start of any work on such a project. During this meeting, a protocol should be established between the preparing and reviewing parties to define the procedures to be followed, the data to be collected, the model to be used, and the analysis of the source and concentration data. 4 While excess emissions during malfunctions are also addressed in the policy related to excess emissions, Appendix W explicitly excludes emissions due to malfunction from the modeling analysis to demonstrate compliance with the NAAQS, unless the excess emissions are the result of poor maintenance, careless operation, or other preventable conditions. See Section a, footnote a. 11

72 DETERMINING BACKGROUND CONCENTRATIONS Unless a facility can demonstrate that ambient impacts associated from its emissions will not exceed the appropriate SIL, a cumulative analysis of ambient impacts will be necessary, and the determination of background concentrations to include in that cumulative impact assessment will be a critical component of the analysis. The June 29, 2010 memorandum addressed some aspects of this issue, but given the stringency of the new 1-hour NO 2 standard, the margin for error in this aspect of the analysis is much smaller than it has been in the past. As a result, we believe it is necessary to provide additional clarification and a more detailed discussion of the factors associated with this aspect of the permitting process. We hope that this additional discussion will serve to more clearly define some of the key steps and considerations in the process that could form the basis of a generic modeling protocol. We also provide suggestions regarding some of the documentation related to this component of the modeling analysis that may facilitate and expedite the review process. The goal of the cumulative impact assessment should be to demonstrate with an adequate degree of confidence in the result that the proposed new or modified emissions will not cause or significantly contribute to violations of the NAAQS. In general, the more conservative the assumptions on which the cumulative analysis is based, the more confidence there will be that the goal has been achieved and the less controversial the review process will be from the perspective of the reviewing authority. As less conservative assumptions are implemented in the analysis, the more scrutiny those assumptions may require and the review process may tend to be lengthier and more controversial as a result. We expect that by providing a more detailed discussion of the factors to be considered in the cumulative impact assessment, permit applicants and permitting authorities will be able to find the proper balance of the competing factors that contribute to this analysis. Identifying Nearby Sources to Include in Modeled Inventory As noted in the June 29, 2010 memo, Section of Appendix W emphasizes the importance of professional judgment by the reviewing authority in the identification of nearby and other sources to be included in the modeled emission inventory, and establishes a significant concentration gradient in the vicinity of the source under consideration as the main criterion for this selection. Appendix W also suggests that the number of such [nearby] sources is expected to be small except in unusual situations. See Section b. In light of this guidance, the June 29, 2010 memo cautioned against the literal and uncritical application of very prescriptive procedures for identifying which background sources should be included in the modeled emission inventory for NAAQS compliance demonstrations, such as those described in Chapter C, Section IV.C.1 of the draft New Source Review Workshop Manual (EPA, 1990). This caution should not be taken to imply that the procedures outlined in the NSR Workshop Manual are flawed or inappropriate in themselves. Cumulative impact assessments based on following such procedures will generally be acceptable as the basis for permitting decisions, contingent on an appropriate accounting for the monitored contribution. Our main concern is that following such procedures in a literal and uncritical manner may in many cases result in cumulative impact assessments that are overly conservative and could unnecessarily complicate the permitting 12

73 process in some cases. Such procedures might be characterized as being sufficient in most cases, but not always necessary to fulfill the requirements of a cumulative impact assessment. A fundamental challenge in developing more detailed general guidance on the issue of determining background concentrations as part of a cumulative impact assessment is that the factors that need to be considered are very case-specific in nature. These factors include foremost the nature of the source being permitted, including the source characteristics and local meteorological and topographical factors that determine the spatial and temporal patterns of the source s ambient impacts. The initial significant impact assessment should serve to characterize these factors, and we would suggest the following: 1. As a standard practice contour plots of modeled concentrations should be prepared which clearly depict the impact area of the source, preferably overlaid on a map of the area that identifies key geographical features that may influence the dispersion patterns. The concentration contour plot also serves to visually depict the concentration gradients associated with the source s impact. 2. We also recommend that the controlling meteorological conditions for the project impacts be identified as clearly as possible. The probabilistic form of the 1-hour NO 2 standard complicates this assessment somewhat, but the recent update to the AERMOD model includes new model output options (MAXDAILY and MXDYBYYR keywords) that identify the specific time periods on which the modeled design value is based. 3. As an aid to interpreting this information, we also suggest including the location of the meteorological monitoring station used in the modeling analysis on the plot of source impacts, as well as a wind rose depicting general flow patterns. If a cumulative impact assessment is required due to the source s impacts exceeding the interim SIL, the applicant will need to identify and acquire data on the two main components of the cumulative impact assessment, namely the location and emissions from nearby background sources that may need to be included in the modeled component of the cumulative ambient impact assessment, and the location and magnitude of air quality data from ambient NO 2 monitors located within the area. Section b of Appendix W states that [t]ypically, air quality data should be used to establish background concentrations in the vicinity of the source(s) under consideration. Section c further states that [i]f the source is not isolated, it may be necessary to use a multi-source model to establish the impact of nearby sources. While many applications will be required to include both monitored and modeled contributions to adequately account for background concentrations in the cumulative analysis, we believe that these statements imply a preference for use of ambient air quality data to account for background concentrations where possible. Many of the challenges and more controversial issues related to cumulative impact assessments arise in the context of how best to combine a monitored and modeled contribution to account for background concentrations. Addressing these issues requires an assessment of the spatial and temporal representativeness of the background monitored concentrations for purposes of the cumulative impact assessment and the potential for double counting of impacts from modeled sources that may be contributing to the monitored concentrations. This assessment may 13

74 involve significant technical details which could complicate the review process. Therefore, the more thoroughly and clearly these issues are documented the more efficient and effective the review process is likely to be. A key point to remember when assessing these issues is their interconnectedness the question of which nearby background sources should be included in the cumulative modeling analysis is inextricably linked with the question of what ambient monitoring data is available and what that data represents in relation to the application. Furthermore, the question of how to appropriately combine monitored and modeled concentrations (temporally and spatially) to determine the cumulative impact depends on a clear understanding of what the ambient monitored data represents in relation to the modeled emission inventory. A more detailed temporal pairing of monitored and modeled concentrations may be acceptable in one case given the extent of the modeled emission inventory, while a more conservative assumption for combining monitored and modeled concentrations using high ranked monitored concentrations may be sufficient to justify a more limited modeling inventory. As noted above, the stringency of the new standard may require a more detailed and refined analysis of these issues in order to demonstrate compliance with the standards than was necessary in the past, and these refinements will generally increase the burden on the applicant to adequately demonstrate that the net result of the analysis is protective of the standard. A detailed analysis and explanation of any potential bias to the net result introduced by proposed refinements will be important to facilitate the review process. The issues associated with determining an appropriate method for combining modeled and monitored contributions to a cumulative impact assessment are discussed in more detail in the next section. Building on the geographical information recommended above for the initial SIL analysis, we suggest including the following documentation: 1. A geographical depiction of the location and magnitude of nearby emission sources, along with the location and magnitude of any ambient monitored data as part of the documentation submitted with a cumulative impact assessment. 2. Depicting the impact area and pattern of the project impacts on such a figure along with a wind rose should be useful in assessing many of the issues touched on above, such as what nearby sources are likely to cause significant concentration gradients in the vicinity of the project source, or more specifically in the areas of high impacts associated with the project source. This figure should also help to identify what nearby source s impacts are likely to be adequately represented in the available monitored data and the potential for double counting of impacts from modeled background sources if certain ambient background data are used. 3. In addition to a standard wind rose, pollution roses (i.e., a depiction of monitored pollutant concentrations as a function of wind direction and/or other meteorological factors) should also be useful for purposes of assessing the representativeness of the monitoring background concentrations in relation to the cumulative impact assessment. 14

75 Finally, we reiterate the importance of close coordination with the appropriate reviewing authority in the determination of nearby or other sources to include in the modeled emission inventory. Significant Concentration Gradient Criterion While Appendix W (Section b) identifies a significant concentration gradient in the vicinity of the source as the sole criterion in relation to determining which nearby sources should be explicitly modeled as part of the cumulative impact assessment, little else has been written to explain what significant means in this context or even what the relevance of a significant concentration gradient is for this purpose. In fact, Appendix W states that no attempt was made to comprehensively define the term, owing to both the uniqueness of each modeling situation and the large number of variables involved in identifying nearby sources. Section b. Nothing has fundamentally changed to alter this characterization, but given the issues and challenges arising from the implementation of the new 1-hour NO 2 standard, we feel compelled to offer some additional explanation regarding what this guidance means and how it should be applied. One definition of the term gradient that applies in this context is the rate of change of a physical quantity... with distance 5. In this case the physical quantity is the ground-level concentration of the pollutant being assessed. The first point worth noting is that the gradient of the ground-level concentration has two dimensions, a longitudinal (along-wind) gradient and a lateral (cross-wind) gradient. Appendix W makes no distinction as to which gradient is more important or whether both gradients should be considered. Before offering any suggestions on that question, it might be helpful to offer some thoughts on the question of why a significant concentration gradient is mentioned as the sole criterion. Since an ambient monitor is limited to characterizing air quality at a fixed location, the impact from a nearby source that causes a significant concentration gradient in the vicinity of the project source is not likely to be characterized very well by the monitored concentration in terms of its potential for contributing to the cumulative modeled design value due to the high degree of variability of the source s impact. In this sense both the longitudinal and lateral gradients could be of importance. However, since the location of impacts from a particular source relative to other sources being modeled or relative to the ambient monitor location is strongly influenced by the transport wind direction, relatively minor changes in wind direction can result in significant changes in modeled concentrations at a particular time and point in space, such as the monitor location. The longitudinal gradient will also vary as a result of changes in wind speed and atmospheric stability, but in general the impact of this longitudinal variability on concentrations at a particular time and point in space will be less significant than the variability associated with the lateral gradient. From this perspective it would appear that the lateral gradient may be more important to consider for purposes of assessing which background sources should be explicitly modeled. Concentration gradients associated with a particular source will generally be largest between the source location and the distance to the maximum ground-level concentrations from the source. Beyond the maximum impact distance, concentration gradients will generally be much smaller and more spatially uniform. A general rule of thumb for estimating the distance 5 Webster's New World College Dictionary, Copyright 2010 by Wiley Publishing, Inc., Cleveland, Ohio. 15

76 to maximum 1-hour impact and the region of significant concentration gradients that may apply in relatively flat terrain is approximately 10 times the source release height. For example, the maximum impact area and region of significant concentration gradients associated with a 100 meter stack in flat terrain would be approximately 1,000 meters downwind of the source, with some variation depending on the source characteristics affecting plume rise. However, the potential influence of terrain on maximum 1-hour pollutant impacts may also significantly affect the location and magnitude of concentration gradients associated with a particular source. Even accounting for some terrain influences on the location and gradients of maximum 1-hour concentrations, these considerations suggest that the emphasis on determining which nearby sources to include in the modeling analysis should focus on the area within about 10 kilometers of the project location in most cases. The routine inclusion of all sources within 50 kilometers of the project location, the nominal distance for which AERMOD is applicable, is likely to produce an overly conservative result in most cases. The relative importance of the lateral vs. the longitudinal gradient will also depend on terrain effects and other factors, such as the atmospheric stability associated with worst-case impacts. The importance of the lateral gradient relative to the longitudinal gradient will generally increase for sources where maximum hourly impacts occur under stable conditions due to the narrowness of the plume under such conditions. The contour plots of modeled design values suggested above provide a method for examining concentration gradients more explicitly. The AERSCREEN model should also serve as a useful tool for identifying the worst-case meteorological conditions for individual sources, as well as determining locations of maximum impact and areas of significant concentration gradients. A final point to mention in relation to this topic is that the pattern of concentration gradients can vary significantly based on the averaging period being assessed. In general, concentration gradients will be smaller and more spatially uniform for annual averages than for short-term averages, especially hourly averages. The spatial distribution of annual impacts around a source will typically have a single peak downwind of the source based on the prevailing wind direction, except in cases where terrain or other geographical effects are important. By contrast, the spatial distribution of peak hourly impacts will typically show several localized concentration peaks with more significant gradients. The number of peaks and the magnitude of the gradients will be somewhat smaller for modeled design values based on the form of the 1-hour NO 2 standard than for overall peak hourly values, due to the smoothing effect of using a multiyear average of the 98 th -percentile from the annual distribution of daily maximum values. One implication of these differences between long-term and short-term concentration patterns is that the factors affecting which sources should be included in the modeled inventory and the method for combining modeled with monitored concentrations are more complex for the 1-hour NO 2 standard than for the annual standard. While we hope this discussion provides some useful insight into this issue, we also caution against interpreting this guidance too literally or too narrowly, and emphasize that a large number of variables (Appendix W, Section b) are involved in this assessment. 16

77 COMBINING MODELED RESULTS AND MONITORED BACKGROUND TO DETERMINE COMPLIANCE One important aspect of the cumulative impact assessment that also deserves further discussion and entails new challenges with the 1-hour NO 2 NAAQS is the method for combining modeled concentrations with monitored background concentrations to determine the cumulative ambient impact. The June 29, 2010 memo indicated that a first tier assumption for a uniform monitored background contribution that may be applied without further justification is to add the overall highest hourly background NO 2 concentration (across the most recent three years) from a representative monitor to the modeled design value 6 for comparison to the NAAQS. Use of a single uniform monitored background contribution is the simplest approach to implement since it can be applied outside of the modeling system. We recognize that use of the overall highest hourly background concentration may be overly conservative in many cases, but that conservatism also provided the basis for indicating that this approach could be used without further justification. As explained above, the more conservative the assumptions on which the cumulative analysis is based, the more confidence there will be that the goal of demonstrating that the source will not cause or contribute to violations of the NAAQS has been achieved and the less controversial the review process will be from the perspective of the reviewing authority. The June 29, 2010 memo also indicated that additional refinements to this first tier approach based on some level of temporal pairing of modeled and monitored values may be considered on a case-by-case basis, with adequate justification and documentation. Given the importance of this aspect of the analysis and the challenges that have arisen in application of the guidance to date, we feel compelled to offer additional guidance on this issue. While the first tier assumption from the June 29, 2010 memo of using a uniform monitored background contributions based on the overall highest hourly background NO 2 concentration should be acceptable without further justification in most cases, we recognize that this approach could be overly conservative in many cases and may also be prone to reflecting source-oriented impacts from nearby sources, increasing the potential for double-counting of modeled and monitored contributions. Based on these considerations, we believe that a less conservative first tier for a uniform monitored background contribution based on the monitored design value from a representative monitor should be acceptable in most cases. The monitored NO 2 design value, i.e., the 98 th -percentile of the annual distribution of daily maximum 1-hour values averaged across the most recent three years of monitored data 7, should be used irrespective of the meteorological data period used in the dispersion modeling. This somewhat less conservative first tier for a uniform monitored background contribution retains the advantage of being relatively easy to implement. 6 The 1-hour NO 2 modeled design value refers to the highest (across all modeled receptors) of the 5-year average of the 98 th -percentile (8 th -highest) of the annual distribution of daily maximum 1-hour values based on NWS meteorological data, or the multiyear average of the 98 th -percentile of the annual distribution of daily maximum 1- hour values based on one or more complete years (up to 5 years) of site-specific meteorological data. The1-hour SO 2 modeled design value follows the same form except that the multiyear averages of the 99 th -percentile (4 th - highest) values are used. 7 The monitored design value for the 1-hour SO 2 standard is based on the 99 th -percentile of the annual distribution of daily maximum 1-hour values averaged across the most recent three years of monitored data. 17

78 Depending on the circumstances of a particular application, use of a first tier assumption for a uniform monitored background contribution may represent a level of conservatism that would obviate the need to include any background sources in the modeled inventory if, for example, the number of nearby sources which could contribute to the cumulative impact is relatively few and the available ambient monitor would be expected to reflect their cumulative impacts reasonably well or conservatively in relation to the modeled design value based on the project emissions. At the other extreme, if the background source inventory included in the modeling is complete enough and background levels due to mobile sources and/or minor sources that are not explicitly modeled is expected to be small, an analysis based solely on modeled emissions and no monitored background might be considered adequate for purposes of the cumulative impact assessment. One of the important factors to consider in relation to this issue is that the standard is based on the annual distribution of daily maximum 1-hour values, which implies that diurnal patterns of ambient impacts could play a significant role in determining the most appropriate method for combining modeled and monitored concentrations. For example, if the daily maximum 1-hour impacts associated with the project emissions generally occur under nighttime stable conditions whereas maximum monitored concentrations occur during daytime convective conditions, pairing modeled and monitored concentrations based on hour of day should provide a more appropriate and less conservative estimate of cumulative impacts than a method that ignores this diurnal pattern. This situation could occur for applications dominated by low-level sources and for elevated releases subject to plume impaction on nearby complex terrain. It is also important to consider the role of NO x chemistry for applications using the Tier 3 options in AERMOD since diurnal patterns of background ozone concentrations may also factor into the diurnal patterns of modeled impacts. Given the potential contribution of background ozone levels to the temporal variability of modeled impacts, the seasonal variability of background monitored values could also be important. Incorporating a seasonal component to the variability of background monitored concentrations will also account for some of the variability in meteorological conditions that may contribute to high hourly impacts. Another situation where understanding the temporal variability of modeled vs. monitored concentrations could be important in determining the most appropriate method for combining modeled and monitored concentrations is where contributions from mobile source emissions contribute significantly to either the monitored background concentrations and/or the modeled concentrations. In these cases, diurnal variability of emissions associated with morning and afternoon rush hours could contribute to the temporal variability of ambient impacts in addition to meteorological factors associated with the dispersion and conversion of NO x emissions. Since rush hours tend to be relatively fixed in terms of time of day and also occur near the transitions from nighttime stable to daytime convective conditions, and vice versa, incorporating a seasonal or monthly element to the temporal variability should account for the variable effect that dispersion conditions may have depending on whether rush hour occurs during stable or convective hours. With these general considerations in mind, we now examine the following guidance in relation to the use of background monitored concentrations in a cumulative impact assessment, from Section of Appendix W, which applies to applications for isolated sources and for the 18

79 contribution of other sources consisting of [t]hat portion of the background attributable to all other sources (e.g., natural sources, minor sources and distant major sources) in a multi-source area: b. Use air quality data collected in the vicinity of the source to determine the background concentration for the averaging times of concern. Determine the mean background concentration at each monitor by excluding values when the source in question is impacting the monitor. The mean annual background is the average of the annual concentrations so determined at each monitor. For shorter averaging periods, the meteorological conditions accompanying the concentrations of concern should be identified. Concentrations for meteorological conditions of concern, at monitors not impacted by the source in question, should be averaged for each separate averaging time to determine the average background value. Monitoring sites inside a 90 sector downwind of the source may be used to determine the area of impact. One hour concentrations may be added and averaged to determine longer averaging periods. c. If there are no monitors located in the vicinity of the source, a regional site may be used to determine background. A regional site is one that is located away from the area of interest but is impacted by similar natural and distant man-made sources. The key principle in this guidance in relation to short-term averaging periods is to determine background concentrations associated with meteorological conditions accompanying the concentrations of concern. The concentrations thus determined should be averaged for each separate averaging time to determine the average background value. Based on this guidance, we believe that an appropriate methodology for incorporating background concentrations in the cumulative impact assessment for the 1-hour NO 2 standard would be to use multiyear averages of the 98 th -percentile 8 of the available background concentrations by season and hour-of-day, excluding periods when the source in question is expected to impact the monitored concentration (which is only relevant for modified sources). For situations involving a significant mobile source component to the background monitored concentrations, inclusion of a day-of-week component to the temporal variability may also be appropriate. The rank associated with the 98 th -percentile of daily maximum 1-hour values should be generally consistent with the number of samples within that distribution for each combination based on the temporal resolution but also account for the number of samples ignored in specifying the 98 th -percentile based on the annual distribution. For example, Table 1 in Section 5 of Appendix S specifies the rank associated with the 98 th -percentile value based on the annual number of days with valid data. Since the number of days per season will range from 90 to 92, Table 1 would indicate that the 2 nd -highest value from the seasonal distribution should be used to represent the 98 th -percentile. On the other hand use of the 2 nd -highest value for each season would effectively ignore only 4 values for the year rather than the 7 values ignored from the annual distribution. Balancing these considerations we recommend that background values by season and hour-of-day used in this context should be based on the 3 rd -highest value for each season and hour-of-day combination, whereas the 8 th -highest value should be used if values vary by hour-of-day only. For more detailed temporal pairing, such as season by hour-of- 8 The 99 th -percentile should be used for the 1-hour SO 2 standard. 19

80 day and day-of-week or month by hour-of-day, the 1 st -highest values from the distribution for each temporal combination should be used. 9 Figure 1 shows the background monitored concentrations by season and hour-of-day for the Salt Lake City, UT monitor for the period based on these recommendations. The values labeled Average Winter, Average Spring, etc. are the 3-year averages of the 3 rd - highest values by hour-of-day for each season; the values labeled Average 98 th % (the dashed line) are the 3-year average of the 8 th -highest values by hour-of-day only; and the values labeled Overall Average are the averages across all values by hour-of-day. These results illustrate the significant temporal variability captured by the multiyear averages of the 98 th -percentile values by season and hour-of-day. Also note that values for the 98 th -percentile by hour-of-day only show little variation by hour-of-day, while values by season and hour-of-day show significant diurnal variability for some seasons. 100 Figure 1. Monitored Background Concentrations for Salt Lake City, UT Monitor One-Hour NO 2 Concentrations NO 2 Concentration (ppb) NAAQS 98th % Winter 98th % Spring 98th % Summer 98th % Fall 98th % Annual Overall Average 1-hr DV Hour It should also be noted here that the conventions regarding observation reporting time differ between ambient air quality monitoring, where the observation time is based on the hourbeginning convention (EPA, 2009; see Section 3.20), and meteorological monitoring where the observation is based on the hour-ending convention (EPA, 2000; see Section 7.1). Thus, ambient monitoring data reported for hour 00 should be paired with modeled/meteorological data for hour 01, etc. The recent update to the AERMOD model (dated 11059) provides an option (the BACKGRND keyword on the SO pathway) to include temporally-varying background concentrations within the cumulative impact assessment based on these temporal factors, similar 9 For 1-hour SO 2 analyses, use the 2 nd -highest value for each season and hour-of-day combination, or the 4 th -highest value for hour-of-day only. Use the 1 st -highest value for more detailed pairing, such as month by hour-of-day or season by hour-of-day and day-of-week. 20

81 to the options that have been available in previous versions of the model to vary source emissions using the EMISFACT keyword. We believe that this technique provides a reasonable and efficient method for ensuring that the monitored contribution to the cumulative impact assessment will be representative of the meteorological conditions accompanying the concentrations of concern since the monitored values will be temporally paired with modeled concentrations based on temporal factors that are associated with meteorological variability, but will also reflect worst-case meteorological conditions in a manner that is consistent with the probabilistic form of the 1-hour NO 2 standard. The use of multiyear-averaged monitored values for the meteorological conditions of concern is consistent with the language in Appendix W related to this issue, and also consistent with the intent of using monitored background concentrations, which is to reflect the contribution from natural or regional levels of pollution and the net contribution of minor emission sources which are not explicitly accounted for in the modeled inventory. Since several applications have come to our attention proposing to combine monitored background and modeled concentrations on an hour-by-hour basis, using hourly monitored background data collected concurrently with the meteorological data period being processed by the model, we feel compelled to include a discussion of the potential merits and concerns regarding such an approach. On the surface this approach could be perceived as being a more refined method than what is recommended above, and therefore more appropriate. However, the implicit assumption underlying this approach is that the background monitored levels for each hour are spatially uniform and that the monitored values are fully representative of background levels at each receptor for each hour. Such an assumption clearly ignores the many factors that contribute to the temporal and spatial variability of ambient concentrations across a typical modeling domain on an hourly basis. Therefore we do not recommend such an approach except in rare cases of relatively isolated sources where the available monitor can be shown to be representative of the ambient concentration levels in the areas of maximum impact from the proposed new source. Another situation where such an approach may be justified is where the modeled emission inventory clearly represents the majority of emissions that could potentially contribute to the cumulative impact assessment and where inclusion of the monitored background concentration is intended to conservatively represent the potential contribution from minor sources and natural or regional background levels not reflected in the modeled inventory. In this case, the key aspect which may justify the hour-by-hour pairing of modeled and monitored values is a demonstration of the overall conservatism of the cumulative assessment based on the combination of modeled and monitored impacts. Except in rare cases of relatively isolated sources, a single ambient monitor, or even a few monitors, will not be adequately representative of hourly concentrations across the modeled domain to preclude the need to include emissions from nearby background sources in the modeled inventory. REFERENCES Cimorelli, A. J., S. G. Perry, A. Venkatram, J. C. Weil, R. J. Paine, R. B. Wilson, R. F. Lee, W. D. Peters, R. W. Brode, and J. O. Paumier, AERMOD: Description of Model Formulation with Addendum, EPA-454/R U.S. Environmental Protection Agency, Research Triangle Park, NC. 21

82 EPA, Air Quality Analysis for Prevention of Significant Deterioration (PSD). Gerald A. Emison memorandum, dated July 5, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, New Source Review Workshop Manual: Prevention of Significant Deterioration and Nonattainment Area Permitting DRAFT. U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, Meteorological Monitoring Guidance for Regulatory Modeling Applications. EPA-454/R U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, User's Guide for the AMS/EPA Regulatory Model AERMOD. EPA-454/B U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, Risk and Exposure Assessment to Support the Review of the NO 2 Primary National Ambient Air Quality Standard. EPA-452/R a. U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, AQS Data Dictionary. Version U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, 2010a. Applicability of the Federal Prevention of Significant Deterioration Permit Requirements to New and Revised National Ambient Air Quality Standards. Stephen D. Page Memorandum, dated April 1, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA, 2010b. Addendum User's Guide for the AMS/EPA Regulatory Model AERMOD. EPA-454/B U.S. Environmental Protection Agency, Research Triangle Park, NC. Hanrahan, P.L., The Plume Volume Molar Ratio Method for Determining NO 2 /NO x Ratios in Modeling Part II: Evaluation Studies. J. Air & Waste Manage. Assoc., 49, Janssen, L.M.J.M., F. Van Haren, P. Bange, and H. Van Duuren, Measurements and modelling of reactions of nitrogen oxides in power-plant plumes at night. Atmos. Env., 25A, No. 5/6, MACTEC, Evaluation of Bias in AERMOD-PVMRM. Final Report, Alaska DEC Contract No MACTEC Federal Programs, Inc., Research Triangle Park, NC. Wang, Y.J., A. DenBleyker, E. McDonald-Buller, D. Allen and K. Zhang, Modeling the chemical evolution of nitrogen oxides near roadways. Atmos. Env., 45, cc: Richard Wayland, C Scott Mathias, C Lydia Wegman, C

83 Raj Rao, C Roger Brode, C Dan deroeck, C Elliot Zenick, OGC Brian Doster, OGC Melina Williams, OGC EPA Regional Modeling Contacts 23

84 ATTACHMENT A Summary of AERMOD Model Performance for 1-hour NO2 Concentrations As noted in the June 29, 2010 memo, limited evaluations of the Plume Volume Molar Ratio Method (PVMRM) for estimating conversion of NO to NO 2 have been completed which show encouraging results, but the amount of data currently available is too limited to justify a designation of PVMRM as a refined method for NO 2 (Hanrahan, 1999; MACTEC, 2005). The original evaluations of PVMRM also focused on model performance for annual averages since the only NO 2 standard in effect at the time was annual. These evaluations have recently been updated to reflect the current AERMOD modeling system components and extended to examine model performance for hourly NO 2 concentrations and to include the Ozone Limiting Method (OLM). Preliminary results from these recent evaluations are presented below in the form of Q- Q plots of ranked hourly NO 2 concentrations for the two monitors included in the New Mexico Empire Abo field study and for the single monitor included in the Palaau, HI field study. Evaluation results are also summarized in the form of predicted vs. observed 1-hour Robust Highest Concentrations (RHC), a model evaluation metric that represents an exponential tail fit to the top 26 ranked values in the distribution of hourly concentrations. Note that the OLM results presented here incorporate an equilibrium NO 2 /NO x ratio of 0.90, consistent with the PVMRM option. Figures A-1 and A-2 show results in the form of hourly Q-Q plots for the North monitor and the South monitor, respectively, from the New Mexico field study based on the Tier 1 option of full conversion of NO to NO 2, the OLM option applied on a source-by-source basis, the OLM option applied using OLMGROUP ALL (OLMGRP), as recommended in the June 29, 2010, NO 2 clarification memorandum, and the PVMRM option. The New Mexico results clearly show the conservatism associated with the Tier 1 assumption of full conversion and the OLM option on a source-by-source basis, with both options showing a significant bias to overpredict hourly NO 2 concentrations. The OLMGRP option exhibits the best performance for both New Mexico monitors, with nearly unbiased results for the North monitor and a slight bias to overpredict for the South monitor. The PVMRM option shows significantly better performance than full conversion or source-by-source OLM for both monitors, but not as good performance as the OLMGRP option. Figure A-3 shows the hourly Q-Q plot for Palaau based on the same range of options shown in Figures A-1 and A-2. Similar to the New Mexico results, the Tier 1 option of full conversion and the OLM option applied on a source-by-source basis show a significant bias to overpredict hourly NO 2 concentrations at Palaau. The PVMRM option shows the best performance for this field study with very good agreement between predicted and observed concentrations. The use of the OLMGRP option clearly improves model performance as compared to application of the OLM option on a source-by-source basis, with the peak predicted concentrations within a factor of 2 higher than observed. These Q-Q plot comparisons are consistent with the comparisons of RHCs summarized in Table A-1, where the average (geometric mean) ratios of Predicted/Observed RHCs for PVMRM and OLMGRP are about 1.5 and 1.2, respectively, and the average RHC ratios for OLMGRP and FULL conversion are much higher at 4.5 and 5.0. A-1

85 Since these Tier 3 options in AERMOD are intended to estimate the conversion of ambient NO to NO 2, it is also useful to compare the modeled vs. observed NO 2 /NO x ratios since offsetting errors in dispersion vs. conversion could mask poor model performance. Table A-2 summarizes the observed vs. predicted NO 2 /NO x ratios for the three monitors included in these Palaau and New Mexico field studies. These results are generally consistent with the hourly Q-Q plots of NO 2 concentrations, and clearly indicate that the OLM option on a source-by-source basis significantly overestimates the conversion of NO to NO 2. However, results for the New Mexico South monitor are interesting in that the PVMRM option shows much better agreement with observed NO 2 /NO x ratios than the OLMGRP option, whereas the OLMGRP option indicates better performance than PVMRM in terms of hourly NO 2 concentrations. These preliminary model evaluation results of hourly NO 2 predictions for Palaau and New Mexico show generally good performance for the PVMRM and OLMGROUP ALL options in AERMOD; however, it should be emphasized that these results are very limited in terms of the number of monitors. Although the scope of the field study data is limited, this level of model performance on a paired-in-space basis is impressive, especially for the PVMRM option at Palaau and for the OLMGROUP ALL option for the North monitor at New Mexico. We believe that these additional model evaluation results lend further credence to the use of these Tier 3 options in AERMOD for estimating hourly NO 2 concentrations and to the recommendation to use the OLMGROUP ALL option whenever OLM is applied. A-2

86 Figure A-1. AERMOD Model Evaluation - New Mexico North Monitor - Hourly NO2 Q-Q Plot 1000 Predicted Conc (µg/m 3 ) 100 FULL OLM OLMGRP PVMRM Observed Conc (µg/m 3 ) Figure A-2. AERMOD Model Evaluation - New Mexico South Monitor - Hourly NO2 Q-Q Plot 1000 Predicted Conc (µg/m 3 ) 100 FULL OLM OLMGRP PVMRM Observed Conc (µg/m 3 ) A-3

87 Figure A-3. AERMOD Model Evaluation - Palaau, HI - Hourly NO2 Q-Q Plot 1000 Predicted Conc (µg/m 3 ) FULL OLM OLMGRP PVMRM Observed Conc (µg/m 3 ) New Mexico Abo North Monitor RHC New Mexico Abo South Monitor RHC Hawaii Palaau Monitor RHC Geometric Mean Pred/Obs RHC Table A-1. 1-hour NO 2 Robust Highest Concentrations (µg/m 3 ) Observed PVMRM OLMGRP OLM FULL Table A-2. Average Unpaired NO 2 /NO x Ratios for Monitored Values of NO x > 20 ppb Monitored NO 2 /NO x PVMRM NO 2 /NO x OLMGRP NO 2 /NO x OLM NO 2 /NO x New Mexico Abo North Monitor (n=772) New Mexico Abo South Monitor (n=262) Hawaii Palaau Monitor (n=672) Geometric Mean Pred/Obs Ratios A-4

88 November 18, 2013 Redhorse Corporation Notes on Draft UDAQ Modeling for GAO: Redhorse Corporation (Redhorse) reviewed the modeling information related to the State of Utah s modeling analyses of nitrogen dioxide (NO 2 ) from various engines at well locations in Utah. Redhorse has the following comments on these analyses: 1. It appears that an earlier version of AERMOD (ver 12060) was used in the modeling analyses. We recommend using the latest version of AERMOD (ver 12345). It should also be noted that AERMOD Version is currently being revised by EPA to incorporate options that are expected to reduce model over-predictions during low wind conditions under stable atmospheric conditions. The over-predictions are particularly evident for low-level emission units such as small engines with low stacks. The updated AERMOD model is expected to be released by the end of the 2013 calendar year. 2. Two AERMOD input files were provided - GWCN10065.adi and GWCN10065vr.adi. It appears that both model input files are identical except for the IC engines stack and diameter and velocity parameters. The exhaust flow rate used to determine the stack velocity for both input files are about 180 cubic feet per minute (cfm) (3 cubic feet per second). Generally, these engine types exhaust flowrate is about 575 cfm. The modeled flowrates (i.e. velocities) appear very low. We suggest verifying the flow rates at the sites, as the higher flow rate and resulting higher exit velocity is likely to produce more accurate model results. 3. In GWCN10065vr.adi the stack diameters for the engines are entered as 1 meter. The typical diameter of engines this size is about 4 inches (consistent with the diameter used in GWCN10065.adi). 4. All stack temps are modeled at 600K (620 F). These temperatures are lower than the typical IC engine exhaust temps of 800 to 1000 F). A review of the engine specification sheets shows that the Arrow L795 engine exhaust temperature is 900 F, while several of the Ajax model engines have exhaust temperatures closer to 600 F. 5. The receptor grid in the model input is a polar grid, with the first ring starting at 25 meters. Because the BLM will not allow extensive fencing, Redhorse recommends the model consider a pseudo fenceline beginning at 100 meters. The public should be well aware that they are potentially impacted by a source within this distance East Plumb Lane, Suite 100 Reno Nevada Phone:

89 6. The modeling was performed using 5 years of meteorological data ( , assumed National Weather Service data) from Canyonlands (with Grand Junction UA data). It is unclear how representative this met data is or how far away it is. There are a significant number of calm hours in the data set (about 6% of the data), and more than 3300 missing hours (more than 7% of the data). Without performing additional modeling, it is difficult to determine if one or more of these calm periods are contributing to the elevated concentrations predicted. In addition, the header of the.sfc met file indicates it has been processed with AERMET (which is inconsistent with the version of AERMOD used 12060). It is unclear if the data have been processed using AERMINUTE (and if so, was it the most recent version), taking advantage of any of the newer AERMET processing options. These can have a profound effect on the modeled concentrations. 7. It appears Utah utilized Tier 3 assumptions (PVMRM) for NO 2 chemistry. Utah utilized a 0.25 NO 2 /NOx ratio and a default value of 0.9 for the equilibrium ratio. Seasonal ozone values were also input. However, the air quality assessment performed by ENVIRON for the Southern Ute Indian Tribe (3/2013) used an in-stack NO 2 /NOx ratio of 0.1 and hourly ozone values from monitoring data (pp ) to implement PVMRM. Redhorse recommends that these parameters also be used. In addition to the specific recommendations above, Redhorse notes that available monitoring information in the surrounding region does not indicate any significant NO 2 ambient compliance issues. Rather, a recent study conducted by Utah State University concluded that elevated NO 2 concentrations were not significantly correlated with oil and gas sites (but are significantly correlated with populated areas). Based on some limited model simulations Redhorse performed (using AERMOD ver and starting with Utah s modeling inputs), Utah s pseudo-stack approach for simulating low momentum stack releases appears to yield concentrations on the order of 25% higher when compared to AERMOD s builtin horizontal stack release option (POINTHOR). Since building parameters were not included in Utah s modeling, the use of the POINTHOR beta option is appropriate and will likely result in a significantly lower impact (well below the standard). Redhorse recommends that this option be utilized if any modeling is performed by Utah. Utah s Emission Impact Guidelines state that the EIA will include a review of previous modeling, an evaluation of site specific conditions, application of a conservative impact assessment, or an in-house modeling exercise. Because of the current modeling uncertainty for low-level NO 2 releases, other meteorological factors as discussed above and the very low site NO 2 emissions from these well locations, Redhorse believes that Utah should exercise its discretion to rely on other information (such as the low ambient concentrations and studies) to support the issuance of the permit East Plumb Lane, Suite 100 Reno Nevada Phone:

90 Redhorse Corporation Comments on UDAQ Proposed GAO Approval and Supporting Analysis. March 20, 2014 Technical and Modeling Analysis 1. The Utah Division of Air Quality's (UDAQ) emission analysis determines NO x emissions from pumpjack engines using the maximum New Source Performance Standards (NSPS) NOx+HC emission limit (2.84 g/hp-hr). Since this is a combined limit for both NOx and hydrocarbons, it clearly overestimates the NO x emissions. UDAQ's NO 2 modeling also utilized this combined emission limit. UDAQ also calculates VOC emissions from pumpjack engines using emission factors from U.S. EPA's AP-42 emission factor resource. Based on information from UDAQ's web site ( the hourly VOC emissions are approximately 13% of the combined hourly NO x +HC emissions. In order to avoid excessively conservative modeling results for NO 2, it is recommended that UDAQ model a decreased NO x emission of 2.47 g/hp-hr (87% of 2.84 g/hp-hr). 2. Because the modeling input files were not included on the UDAQ's web site, it is impossible to verify UDAQ's modeling results. 3. The modeling memo dated January 27, 2014 indicates that building downwash affects were considered due to the storage tanks. However no further technical information was discussed. Since the modeling input files were not posted on UDAQ's web site along with the rest of the supporting information, relative location and configuration of these tanks cannot be confirmed. 4. The modeling memo discusses UDAQ's treatment of NO 2 background values. The memo simply states that: [P]redicted impacts for the wellhead sources themselves at each site should be low enough to accommodate the 40 to 65 μg/m3 one-hour ambient NO2 values that have been recorded at various monitoring sites in the Uintah Basin region. It appears that UDAQ is over-simplifying the background value by creating a maximum modeled target concentration for the GAO emission sources, which accomodates this maximum background value. As a result, in UDAQ's analysis, the worst case NO 2 model result can be no more than 123 µg/m 3 (188 µg/m 3 65 µg/m 3 ). This over-simplification almost certainly results in a needlessly conservative impact assessment. EPA guidance (T. Fox Memo, March 1, 2011) states that: [W]e believe that a less conservative first tier for a uniform monitored background contribution 1

91 based on the monitored design value from a representative monitor should be acceptable in most cases. Therefore, a less conservative implementation of the background value would be to identify a representative background monitor and determine the design value for that monitor. Based on the rural nature of the oil and gas well sites (as determined by UDAQ), the most appropriate rural monitor in the Uinta Basin is the Ouray site (AQS site I.D ). This site is unaffected by urban influences that are common in the other monitoring sites in the Uinta Basin. The most recent three years of data from this site were obtained from EPA's AQS data mart and are summarized in Table 1. Table 1 Ouray Design Value Average ppb µg/m 3 ppb µg/m 3 ppb µg/m 3 ppb µg/m 3 Annual 98 th %: As can be seen, the design value is 54.5 µg/m 3. This simple change, if implemented, would reduce the background by over 10 µg/m 3 and may result in more flexibility for the minimum stack height requirements proposed for the pumpjack engines in UDAQ's technical support document. 5. UDAQ's modeling included the Plume Volume Molar Ratio Method (PVMRM) model option and assumes an in-stack NO x to NO 2 conversion of 0.25 and an ambient equilibrium of 0.9. However, very recent work by researchers (and incorporated by U.S. EPA into AERMOD 13350) focused on the ambient ratio method (ARM/ARM2). In-stack NO x /NO 2 ratios are difficult to determine and ambient equilibrium values require lengthy justification. Conversely, EPA considers ARM as a Tier 2 method for NO 2 modeling (more simple to implement than PVMRM) and does not suffer from the additional required parameters, as does PVMRM. U.S. EPA included updated algorithms in AERMOD specifically to address concerns that previous AERMOD versions did not predict NO 2 concentrations well. Testing indicates that ARM (and specifically ARM2) performs better than PVMRM. Although, ARM2 is currently officially classified as a BETA option, personnel at U.S. EPA Headquarters have generally indicated it is acceptable for use with no further justification. UDAQ should revise its modeling to utilize the ARM2 option. 6. UDAQ's modeling analysis utilized the maximum allowed NO x emissions (see Comment 1 above) as input to the modeling. However, the modeling memo also indicates that exhaust parameters (flow and temperature) were taken from stack test data submitted to the Division from various sites in the Uintah basin for the most commonly used Arrow L HP engine. In essence, 2

92 UDAQ is modeling maximum worst-case emissions in combination with actual tested stack parameters. Because the modeling memo does not indicate whether these tests were conducted at maximum capacity, the stack information was likely obtained during partial load operating conditions (i.e. significantly less than maximum operating capacity) and should not be paired with maximum emission conditions. If stack parameters based on less than full load tests are to be modeled, the emission rate should be decreased, as well, to properly simulate a partial load operating mode. Typical manufacturer full load operating temperature and flow values for the L-795 engine are 625 acfm and 900 F. As a result, UDAQ's modeling is likely overly conservative and contributes to erroneous conclusions related to minimum stack height requirements contained in the proposed GAO. 7. The proposed GAO includes a minimum 4 foot stack height requirement for the pumpjack engines. While this requirement on the surface does not appear extreme, in reality the physical design limitations of most of the engines in operation at these well pad sites would cause operators to exceed manufacturer design criteria. Two of the most commonly used NSPS JJJJ pumpjack engines in the Uinta Basin are the Arrow L-795 and the Cameron Ajax E-565. The exhaust port on these engines is oriented downward and located near the base of the engine, not at the top of the engine in the vertical. An elbow is placed on the exhaust pipe immediately after leaving the engine to make a 90º turn and bring the exhaust pipe out from under the engine to a horizontal orientation. As such the exhaust pipe is initially horizontal near the base of the engine, not pointing horizontal at the top of the engine. Portions of the specifications included in manufacturer s operation\maintenance\service manuals are summarized below. Arrow L-795 This engine is of two cycle design, and LENGTH AND SIZE of exhaust pipe are very important. Good performance, satisfactory service, and power cannot be expected unless the exhaust pipe has the proper length and size for the speed at which the engine is to be operated. Not more than one 90º elbow can be used and obtain good results. The exhaust should be as straight as possible. (Arrow Engine Company 2013) Ajax E-565 The exhaust system must be properly designed for the operating conditions of the engine, both for proper scavenging of the engine cylinder, and for correct dissipation of exhaust heat. Recommended exhaust pipe size and length are established for each engine at various operating speeds. Muffler type and size are also critical to good operation, and recommendations are also established for this equipment. (Cameron 2013) The L-795 document underscores the importance that the manufacturer places on these parameters. At full load the Arrow L-795 specifications include an exhaust length of

93 The manufacturer specifications include a total length as small as 6-3 beyond the engine silencer, depending on load. In addition, the manufacturer specifies that not more than one 90º elbow should be used in the exhaust pipe. The exhaust pipe already requires one 90º turn to reach a horizontal orientation after leaving the bottom of the engine. Ajax E-565 manufacturer specifications are for a total exhaust length of 9-6 to properly operate. The engine specifications reviewed depict 7-4 of horizontal exhaust, including elbows, immediately after leaving the bottom of the engine. Complying with manufacturer specifications reviewed would leave a maximum of 2-2 of exhaust that could be added in the vertical, above the horizontal exhaust pipe section. Requiring a minimum 4 foot stack height for the pumpjack engines is almost certainly outside manufacturer design specifications and could result excessive back-pressure, over-heating, excess emissions and premature failure of the engine. This could cause industry to avoid obtaining the GAO, resulting in little to no value to the State or the industry. 8. As discussed above, the manufacturers' design specifications for engine exhaust preclude the installation of elevated releases as required by the proposed GAO. Attempting to comply with the minimum stack height requirement may result in improper operation of the pumpjack engines, cause excess emissions, increased maintenance costs and potentially void the warranty. However, by modifying the emissions rate, determining a more reasonable and appropriate background value, utilizing the ARM2 model option, and combining maximum emissions with maximum stack operating parameters, the cumulative effect will allow UDAQ to determine decreased concentration impacts. The result may obviate the need for stack height requirements or at least demonstrate the need for stack height requirements for the pumpjack engines that are within the manufacturers' design. 4

94 OIL AND GAS TANK BATTERY NITROGEN DIOXIDE MODELING STUDY REPORT Prepared by: Redhorse Corporation King St. Westminster, Colorado January 2014

95 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT Section CONTENTS Page 1.0 Introduction Monitored Ambient NO 2 Concentrations Modeled Air Quality Impacts Model Options/Source Emissions and Stack Characteristics Meteorological Data Model Receptors Background NO 2 Data Selection Summary of Model Results Conclusions References TABLES Table Description Page 3-1 Comparison of Modeled Emission Source Parameters Comparison of AERMOD Model Inputs Manufacturers Engine Specifications for the Arrow L-795 Engine Summary of Daily 1-Hour 98 th Percentile NO Summary of Modeled Impacts with Engines Operating at Maximum Conditions Summary of Modeled Impacts with Engines Operating at Normal Conditions FIGURES Figure Description Page 2-1 Uinta Basin NO 2 monitoring sites Vernal Windrose Model Receptors PAGE i

96 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT 1.0 INTRODUCTION The Utah Division of Air Quality (UDAQ) is developing a General Approval Order (GAO) for oil and gas tank battery sites. A key aspect of conditions that will be in the GAO involves nitrogen dioxide (NO 2 ) emission sources and the resulting NO 2 air concentration impact of the allowable emissions under the GAO. The simulated tank battery site being evaluated by UDAQ includes the following NO 2 emissions sources which would be allowed under the Utah GAO: Pumpjack engines consisting of a combined 100 horsepower (hp). Gas-fired open combustion sources, such as hot oil heaters and tank heaters, consisting of a combined 5 million British thermal units (mmbtus) per hour (mmbtu/hr) of fuel combustion. Gas-fired flares consisting of a combined 2 mmbtu/hr of fuel combustion. There are numerous important considerations in determining conditions that would appear in the GAO as a result of NO 2 emission sources, including the following: Existing NO 2 ambient conditions in the oil and gas fields; Expected NO 2 ambient conditions in the oil and gas fields under the GAO; and Equipment design specifications. This report provides a discussion of important issues as they relate to development of the GAO, including the presentation of modeling analyses that were conducted using reasonable assumptions designed to protect air quality by over-predicting impacts. This document demonstrates the following: Existing NO 2 ambient concentrations are well below the National Ambient Air Quality Standard (NAAQS) in the oil and gas development area ambient monitoring data show that NO 2 concentrations in the oil and gas fields are in compliance with and well below the NAAQS (even with current well pad sources in operation and without extended exhaust stacks); Continued compliance with NO 2 NAAQS can be demonstrated the revised modeling demonstrates that when maximum emissions are paired with maximum operating parameters or normal emissions are paired with normal operating parameters, the maximum modeled impact (including background) will be less than the NO 2 NAAQS (UDAQ s analysis paired maximum emissions with normal operating parameters); and Manufacturer specified exhaust pipe restrictions preclude the installation of tall engine stacks. The revised modeling using more realistic model inputs demonstrate that oil and gas tank battery sites will comply with the NO 2 NAAQS without the need to extend engine stack heights beyond manufacturer-designed parameters. PAGE 1

97 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT 2.0 MONITORED AMBIENT NO 2 CONCENTRATIONS Extensive NO 2 monitoring has been conducted in the Uinta Basin. Figure 2-1 shows Uinta Basin NO 2 monitoring sites. Ambient monitoring data show that NO 2 concentrations in the oil and gas fields are in compliance with the NAAQS with pumpjack engines in operation without extended exhaust stacks. Moreover, in most cases, extending exhaust stacks from these engines is not technically feasible. One-hour NO 2 impact issues are primarily related to urban areas. The U.S. Environmental Protection Agency s (EPA s) initial intent for the 1-hour NO 2 standard was to address air quality in urban areas. As a result of the recognized status of 1-hour NO 2 as an urban air quality issue, EPA only requires monitoring in large urban areas. Several years of 98 th percentile 1-hour NO 2 values from the Ouray and Red Wash monitoring sites are shown below. The 1-hour NO 2 NAAQS is 100 parts per billion (ppb) or approximately 188 micrograms per cubic meter of air (µg/m 3 ). There is not a rural NO 2 impact issue associated with oil and gas equipment. Utah State University s 2013 winter ozone and air quality report (Lyman, Mansfield, and Shorthill 2013) concludes that the highest NO 2 impacts are likely due to traffic. The 2013 Utah State University report shows that highest NO 2 impacts occur during the Vernal and Roosevelt rush hour. The report identifies less pronounced peaks around noon for the Red Wash, Ouray, and Horsepool sites. Traffic counts at Horsepool show highest traffic occurring mid-day. The Red Wash monitor is about 250 meters from a well-traveled road. These data support the finding that the NO 2 impacts in the basin are primarily associated with traffic, and urban traffic in particular, unlikely from oil and gas stationary sources. Ouray: 2010: 56 µg/m : 58 µg/m : 49 µg/m 3 Red Wash: 2010: 56 µg/m : 64 µg/m : 47 µg/m 3 Monitoring data in the Uinta Basin demonstrate that the highest NO 2 measurements are associated with urban areas such as Vernal and Roosevelt and are not an issue in the oil and gas development areas. NO 2 concentrations in areas away from Vernal and Roosevelt are less than 35 percent (%) of the ambient standard. NO 2 measurements at Ouray, in the middle of the oil and gas development, are approximately 30% of the ambient standard, demonstrating that NO 2 is not an issue in the oil and gas development area. Higher NO 2 concentrations in Vernal and Roosevelt can be traced to urban rush-hour traffic, not oil and gas operations. PAGE 2

98 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT Figure 2-1: Uinta Basin NO 2 monitoring sites PAGE 3

99 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT 3.0 MODELED AIR QUALITY IMPACTS A key consideration in assuring compliance with the NAAQS using dispersion modeling is how far to go in using over-prediction assumptions to assure that air quality is protected. Going too far with overprediction assumptions in a dispersion modeling analysis may result in modeled violations and can lead to excessive requirements being dictated in the GAO. Having excessive requirements in the GAO may cause it to be unusable by the regulated industry. UDAQ has conducted a preliminary air quality impact analysis using EPA s AERMOD dispersion model to evaluate the need for stack height requirements in the GAO. The preliminary NO 2 modeling performed by UDAQ has resulted in, among other things, unreasonably tall stack height requirements. The stack heights dictated by the UDAQ modeling are actually beyond the specifications of the engine manufacturers, as explained later in this document. The initial GAO conditions dictated by UDAQ from their modeling are largely considered excessive by oil and gas operators. Redhorse Corporation (Redhorse) has performed air quality dispersion modeling using the EPA AERMOD dispersion model to predict NO 2 air quality impacts from emission sources representing a generic tank battery and associated onsite oil production equipment. Details of the dispersion modeling analysis conducted by Redhorse are presented in the remaining sections of this document. It is important to note that the options and tools used in the Redhorse analysis are options that EPA has made available in their dispersion model. The EPA and recognized experts have generally acknowledged that the AERMOD model system tends to over-estimate concentrations for the NO 2 NAAQS. Because of this, several recent updates to the AERMOD modeling system have been made in an attempt to reduce the amount of over-prediction in the model output (EPA 2013a). A review of the initial UDAQ modeling was performed and revealed that adjustment of several key inputs within the modeling system would likely result in more reasonable impacts from the sources. The key model inputs identified are: Model Options/Source emission and stack characteristics; Meteorological data; Model receptor grid ; and Background NO 2 data selection. 3.1 Model Options/Source Emissions and Stack Characteristics Redhorse adjusted several model and source parameter inputs to more closely characterize the well pad operations. Table 3-1 compares the modeled stack parameter input assumptions used in the UDAQ model with the values used in the revised modeling. Table 3-2 compares the AERMOD model input assumptions used in the UDAQ model with those used in the revised modeling. The AERMOD model option revisions were made largely to take advantage of model corrections and upgrades made by EPA in the AERMOD system. PAGE 4

100 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT Source Table 3-1 Comparison of Modeled Source Parameters (Engines with Horizontal Releases and at Maximum Operational Conditions) NOx Emission Height Diameter Temperature Flow Rate Rate (lb/hr) (feet) (feet) ( F) (cfm) UDAQ Redhorse UDAQ Redhorse UDAQ Redhorse UDAQ Redhorse UDAQ Redhorse Flare ,180 6, Heater (5 MMBtu/hr) 100 HP Engine 65 HP Engine 40 HP Engine 0.63 N/A 12* N/A 3.31 N/A 620 N/A 172 N/A ** N/A 0.21 N/A 3 N/A 4.92 N/A 600 N/A 380 * In addition to 12, UDAQ also modeled 15 and 20 stack heights. ** In addition to 8, UDAQ also modeled 12 and 15 stack heights. Table 3-2 Comparison of AERMOD Model Inputs Model Input Parameter UDAQ Model Redhorse Model Rationale for Change Meteorological Surface Data Source Meteorological Data Processing Statistical Concentration Averaging Method Model Receptor Grid 4 or 5 years data from Canyonlands, Price, and Vernal (uncertain which years were used but include 2006) Typical EPA approach 4 or 5-year average of the 98 th percentile of the 1-hour daily maximum, for the 3 stations Polar grid with receptor rings at 50, 62.5, 75, 100, 150, 200, and 1000 meters 5 years Vernal data ( ) New EPA approach using the stablebl_adj_u* adjustment 5-year average of the 98th percentile of the 1-hour daily maximum, using the 5-year meteorological dataset from Vernal Polar grid with receptor rings at 91.4, 103.9, 116.4, 141.4, 166.4, 191.4, 291.4, 5 years of the most representative, recent meteorological data was selected in accordance with 40 CFR Part 51, Appendix W, Section The new EPA approach was released December 24, 2013 and addresses inaccuracies in model results during stable, low wind atmospheric conditions. Redhorse approach follows EPA guidance for modeling analyses to evaluate the form of the 1- hour NO 2 NAAQS. Takes into account a required 300 foot setback distance from well pads to buildings representing public receptors and meters AERMOD Version Version Version The latest version of AERMOD is the most appropriate for regulatory modeling. NO x to NO 2 Conversion in the atmosphere Plume Volume Molar Ratio Method (PVMRM) Ambient Ratio Method 2 (ARM2) The ARM2 method is classified by EPA as a Tier 2 approved method and does not rely on the uncertain in-stack NO 2 /NO x ratio required when using PVMRM. PAGE 5

101 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT The assumptions made in the dispersion modeling conducted by Redhorse were purposely developed to protect air quality by over-predicting NO 2 impacts. However, where possible, more reasonable assumptions were made for a generic simplified oil tank battery site than those assumed in the UDAQ model. Examples of more reasonable modeling assumptions include: The GAO engines are limited to the New Source Performance Standard (NSPS) JJJJ combined nitrogen oxides (NO X )-plus-hydrocarbon (NO X +HC) emission rate. The NO X only fraction of the emissions were modeled rather than the combined NO X +HC emission; Existing background NO 2 concentrations added to modeled concentrations are represented by the three year average of the annual 98 th percentile of the daily maximum 1-hour observed concentrations, measured at a monitoring site located in the oil and gas development area; UDAQ modeled a single 100 HP engine in an attempt to allow for more than one engine at a site. Redhorse modeled two separate engine configurations (two 65 hp models at a site and a 65 hp and 40 hp engine at a site). This was done to account for differences in the stack parameters and release heights between the engines; and Redhorse maintained the same equipment location assumptions contained in the UDAQ modeling. The receptor grid was centered on the engines, while the tank heater was located 20 meters north of the engines and the flare was located 20 meters south of the engines. However, two engines were assumed to be emitted from the same location but with individual stack parameters. The receptor grid was modified to start at the required minimum setback for well pads (300 feet or approximately 91 meters). A polar grid was centered on the engine location with radials every 10 and distances along each radial of m, m, m, m, m, m, m and m. The assumptions in the Redhorse analysis are more reasonable than those used in the initial UDAQ modeling, while being adequately conservative to assure compliance with the NO 2 NAAQS. The Redhorse assumptions are more representative of a generic tank battery site and are sufficient to assure NO 2 NAAQS are protected. Pumpjack Engine Emissions Under the draft GAO, pumpjack engines would be allowed to operate at a maximum 100 hp of capacity, either from one engine or multiple engines, and the NSPS JJJJ specified combined NO x +hydro-carbon (NO X +HC) emission limit of 2.8 grams per horsepower-hour (g/hp-hr). As seen in Table 3-3, the uncontrolled NO X emission rate of the Arrow L hp 2-stroke lean burn (2SLB) engine is 1.89 g/hp-hr, according to the manufacturer of this frequently used engine. As a result, the dispersion modeling performed in this analysis uses 1.89 g/hp-hr for the 65 hp engine type. PAGE 6

102 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT AP-42 Section 3.2 produces a VOC emission rate of 0.29 g/hp-hr (all organics less methane and ethane) for a 65 hp engine. AP-42 also produces a non-methane HC (NMHC) emission rate of 0.46 g/hp-hr for a 65 hp engine. Therefore, it is arguable that the NO X only emissions for a NSPS JJJJ engine is 2.4 g/hphr. Therefore, 2.4 g/hp-hr for NO x emissions was used for the 40 hp engine type. Table 3-3 Manufacturers Engine Specifications For the Arrow L-795 Engine ENGINE MODEL: L 795 Rich/Lean Burn Lean 2 or 4 Cycle 2 Bore 7.50 Stroke 9.00 Displacement (CI.) 795 No. Cylinders 2 RPM Max/Min. 600/300 Max HP (cont.) 65 BMEP 54 BSFC (BTU/HP HR) Exhaust Stack NPT Dia. (in.) 4" Height (in.) ** 7" Temp. (Deg. F) 900 Flow (acfm) 625 Emissions (g/hp hr) Pre Cat Nox 1.89 Pre Cat CO 2.58 Pre Cat VOC N/A Pre Cat HCHO N/A Post Cat Nox *2.4 Post Cat CO *4.8 Post Cat VOC N/A Post Cat HCHO N/A Max. Exhaust Back Pressure ("W.C.) TE Weight (lb.)dry 4510 * = EPA emission regulation limits as of March 1, Check with your local DEQ, as they may be lower than the EPA requirements. BSFC (BTU/HP max rated RPM ** = Stack height is from the base of the mounting feet to the exhaust manifold outlet. += Catalyst equiped engines. ⁿ = MUF 1 standard muffler outlet height. TE = Tuned Exhaust. IP = In Process Emissions vary depending on AFR set point and emission equipment from engine to engine. This information is for reference only. PAGE 7

103 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT Pumpjack engines typically operate at less than the full capacity for which they are designed. UDAQ modelers contend that the pumpjack engines exhaust parameters (e.g. temperature and velocity) in the model should be simulated at a load level representing less than full capacity. However, the UDAQ also contends that the emissions from the engine should be modeled at full capacity, even though the engine is assumed to operate at less than full load. UDAQ s modeling assumed a combined 100 hp engine, representing the possible operation of multiple small engines up to this cap. Redhorse s analysis considered a 65 hp engine and a 40 hp, modeled separately (combined 105 hp) and two 65 hp engines (combined 130 hp). These combinations of engines would exceed the 100 hp maximum proposed limit and result in slightly more emission potential. The UDAQ modelers have incorrectly set up their modeling analysis of the engines as follows: Stack parameters: representing the normal reduced load parameters (i.e. lower temperature and velocity - based on a 65 hp engine); and Emissions: 2.8 g/hp-hr X 100 hp = 280 grams per hour (g/hr) (i.e. full load). While it is not uncommon to perform dispersion modeling at less than maximum operating conditions (e.g. periods of startup) for large sources, it is not common to evaluate this for very small sources, such as these well pad sites. However, if less than maximum conditions are modeled, any such model is typically performed with stack parameters and emissions, both, at less than full load. Assuming an engine operates at less than full capacity, while emissions are at a full capacity level, is not a reasonable assumption. Instead, if modeling is to be done at less than full capacity, the engines above should be modeled as follows: Stack parameters: representing normal load parameters (e.g. 50% load). Emissions: normal reduced capacity (e.g. 65 hp * 50% *2.4 g/hp-hr = 78 g/hr) Modeling the emissions at full load should be done with stack parameters at full load as well, for example: Stack parameters: representing the 100% load parameters. Emissions: 2.4 g/hp-hr X 65 hp = 156 g/hr The dispersion modeling in this analysis evaluates a scenario with the pumpjack operating at less than full capacity, including emissions representing emissions less than full capacity. Additionally this analysis assumes the engines operate with full load stack parameters and full load emissions. Pumpjack Engine Stack Parameters Two commonly used NSPS JJJJ pumpjack engines used in the Uinta Basin are the Arrow L-795 and the Cameron Ajax E-565. The exhaust port on these engines is oriented downward and located near the base of the engine, not at the top of the engine in the vertical. An elbow is placed on the exhaust pipe immediately after leaving the engine to make a 90⁰ turn and bring the exhaust pipe out from under the engine to a horizontal orientation. As such the exhaust pipe is initially horizontal near the base of the engine, not pointing horizontal at the top of the engine. PAGE 8

104 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT The manufacturers of these engines spell out exhaust piping requirements (specifications) including the number of turns in the exhaust pipe and required maximum exhaust pipe length, in combined horizontal and vertical orientations. Specification included in manufacturer s operation\maintenance\service manuals are summarized below. Arrow L-795 This engine is of two cycle design, and LENGTH AND SIZE of exhaust pipe are very important. Good performance, satisfactory service, and power cannot be expected unless the exhaust pipe has the proper length and size for the speed at which the engine is to be operated. Not more than one 90º elbow can be used and obtain good results. The exhaust should be as straight as possible. (Arrow Engine Company 2013) The use of capital letters for the words LENGTH AND SIZE in the above text is as written in the L-795 document and underscores the importance that the manufacturer places on these parameters. At full load the Arrow L-795 specifications include an exhaust length of 9-4. The manufacturer specifications include a total length as small as 6-3 beyond the engine silencer, depending on load. In addition, the manufacturer specifies that not more than one 90⁰ elbow should be used in the exhaust pipe. The exhaust pipe requires one 90⁰ turn to reach a horizontal orientation after leaving the bottom of the engine (Arrow Engine Company 2013). To illustrate this, the manufacturer lists the stack height of the exhaust pipe as 7- inches in Table 3-3. Ajax E-565 The exhaust system must be properly designed for the operating conditions of the engine, both for proper scavenging of the engine cylinder, and for correct dissipation of exhaust heat. Recommended exhaust pipe size and length are established for each engine at various operating speeds. Muffler type and size are also critical to good operation, and recommendations are also established for this equipment. (Cameron 2013) Ajax E-565 manufacturer specifications are a total exhaust length of 9-6 to properly operate. The engine specifications reviewed depict 7-4 of horizontal exhaust, including elbows, immediately after leaving the bottom of the engine. Complying with manufacturer specifications reviewed would leave approximately 2-2 of exhaust that could be added in the vertical, above the horizontal exhaust pipe section (Source: Cameron (manufacturer) design drawings). Correspondence with manufacturer s representatives has confirmed the L-795 and E-565 specifications. This review only examined the L-795 and E-565 engines and it is possible that other engines could have more restrictive exhaust stack requirements. Manufacturer specifications for the common L-795 and E- 565 NSPS JJJJ compliant engines reveal a vertical exhaust pipe restriction as small as 2-2 above the horizontal section of the exhaust pipe. If the height of the horizontal section of the exhaust pipe were assumed to be as low as 1 above ground level, manufacturer specifications for the E-565 would restrict stack heights to 3-2 above ground level. It would be inappropriate for UDAQ to include permit conditions requiring companies to operate outside of manufacturer specifications. PAGE 9

105 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT 3.2 Meteorological Data Dispersion modeling was conducted using hourly surface meteorological data from the National Weather Service (NWS) station located in Vernal, Utah. The Vernal dataset consists of five complete years of data for the period January 1, 2008 through December 31, The hourly surface data were supplemented with one-minute averaged wind data in the pre-processing step, following EPA guidance. Use of the oneminute averaged wind data to supplement hourly wind data is a recent processing method developed and recommended by EPA to reduce the number of hours in a dataset with calm wind conditions. While this method is not yet required for regulatory purposes, EPA believes that use of the minute data produces more accurate modeled concentrations. EPA documentation on AERMINUTE is located online at The AERMET meteorological data pre-processor software Version was used to process the dataset into a format usable by AERMOD. Surface data were combined with upper air meteorological data from Grand Junction, Colorado for the same period using the AERMET software. AERMET was run with a new meteorological data processing option that is designed to address problems in AERMOD with model performance during stable, low wind atmospheric conditions. Recent studies have shown large model over-predictions during stable, low wind conditions, especially for low-level emission sources (Paine and others 2012). The new processing option is referred to as the AERMET Surface Friction Velocity Adjustment (STABLEBL ADJ_U*). When activated, the option adjusts the surface friction velocity to increase turbulence during stable, low wind atmospheric conditions (EPA 2013b). This adjustment option is listed as a beta option by EPA and therefore requires approval by UDAQ to use in a regulatory context. EPA has conducted significant model evaluations with the new option and therefore additional model equivalency evaluations typically required for alternative models should not be necessary. EPA model evaluations are available on the EPA SCRAM website ( Surface characteristics required as input to AERMET were calculated using the AERSURFACE software. This program calculates surface roughness, Bowen ratio, and albedo parameters using USGS land cover datasets. A wind rose plot of the surface meteorological data is presented in Figure 3-1. This figure shows that the predominant winds in the area blow from the west through west-northwest. 3.3 Model Receptors In the draft GAO modeling analysis conducted by UDAQ, a polar receptor grid was constructed with radials every 10 degrees for a total of 36 radials. Receptors were placed along those radials at distances of 50, 62.5, 75, 100, 150, 200, and 1000 meters. It is believed that receptors elevations and hill heights were determined by UDAQ using the AERMAP software and U.S. Geological Survey (USGS) digital terrain data. The modeling study described herein is based on a similar model receptor grid, but starts at meters and places additional receptors at radial distances of , , , , , and meters. A plot of the model receptors relative to the emission sources is presented in Figure 3-2. PAGE 10

106 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT PAGE 11

107 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT PAGE 12

108 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT 3.4 Background NO 2 Data Selection Redhorse reviewed available monitoring information and found that Utah had reported NO 2 monitoring for the Ouray station (AQS Site ID , see Figure 2-1). This site is more representative and preferable because it is rural and not affected by vehicle traffic patterns. Data from the site were available for part of 2009 and all of 2010, 2011 and 2012 (almost 3.5 years). Redhorse tabulated daily maximum 1-hour NO 2 values at the Ouray station. Consistent with the form of the NO 2 NAAQS, the 98 th percentile value of each year ( ) was determined. The average of the 3 years was then calculated to determine a background value (see Table 3-4). Model results were combined with the background value to determine compliance with NAAQS. Table 3-4: Summary of Daily 98 th Percentile 1-Hour NO Average ppb µg/m 3 ppb µg/m 3 ppb µg/m 3 ppb µg/m 3 Annual 98th%: Source: AQS Data Mart for Ouray Monitor (Site ID ) 3.5 Summary of Model Results As a result of reasonable adjustments to the model in the key areas described above, the NO 2 impact (including added background NO 2 concentrations) is demonstrated to be less than the NAAQS without requiring significant changes to the current operations of the oil and gas sources. Table 3-5 shows a summary of the various inputs and the resultant predicted impacts when pumpjack engines are at maximum operating conditions. Table 3-6 shows a summary of model results when pumpjack engines are at normal operating conditions. All resulting modeled impacts comply with the 1-hour NO 2 NAAQS of 188 µg/m 3 (100 ppb). Table 3-5 Summary of Modeled Impacts with Engines at Maximum Operating Conditions Model Run Description Modeled Impact (µg/m 3 ) Background (µg/m 3 ) Flare+Heater+2-65HP Engines Max. Emissions, Horizontal Release, 1.89 g/hp-hr, 300' Setback Receptor Grid, u*met Flare+Heater+2-65HP Engines Max. Emissions, Vertical Release, 1.89 g/hp-hr, 300' Setback Receptor Grid, u*met Flare+Heater+65HP+40HP Max. Emissions, Horizontal Release, (65HP=1.89 g/hp-hr), 300' Setback Receptor Grid, u*met Flare+Heater+65HP+40HP Max. Emissions, Vertical Release, (65HP=1.89 g/hp-hr), 300' Setback Receptor Grid, u*met Note: µg/m 3 = micrograms per cubic meter Total Impact (µg/m 3 ) PAGE 13

109 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT Table 3-6 Summary of Modeled Impacts with Engines at Normal Operating Conditions Model Run Description Modeled Impact (µg/m 3 ) Background (µg/m 3 ) Flare+Heater+2-65HP Engines Normal Emissions, Horizontal Release, Engines = 65% of maximum emissions, 300' Setback Receptor Grid, u*met Flare+Heater+2-65HP Engines Normal Emissions, Vertical Release, Engines = 65% of maximum emissions, 300' Setback Receptor Grid, u*met Flare+Heater+65HP+40HP Normal Emissions, Horizontal Release, Engines = 65% of maximum emissions, 300' Setback Receptor Grid, u*met Flare+Heater+65HP+40HP Normal Emissions, Vertical Release, Engines = 65% of maximum emissions, 300' Setback Receptor Grid, u*met Note: µg/m 3 = micrograms per cubic meter Total Impact (µg/m 3 ) CONCLUSIONS The UDAQ s preliminary engine stack requirements resulting from air quality modeling analyses represents an unrealistic conclusion for the GAO for the following reasons: 1. Monitored actual NO 2 concentrations in the oil and gas fields shows that these areas are in compliance with the NO 2 NAAQS. 2. Engine manufacturer data confirms that extending exhaust stacks from the pumpjack engines is not technically feasible. 3. Modeling analyses that indicate tall stack heights are required for the engines to comply with the NO 2 NAAQS are based on overly conservative model assumptions. PAGE 14

110 OIL AND GAS WELL NITROGEN DIOXIDE MODELING STUDY REPORT 5.0 REFERENCES Arrow Engine Company L-795 Operation Maintenance Parts. Document no. L795-POM-C- APR13. April. Cameron Ajax E-565 Gas Engine Service Manual. Lyman,S., M. Mansfeld, and H. Shorthill Final Report, Uintah Basin Winter Ozone & Air Quality Study. Utah State University. October 24. Paine, B., J. Connors, C. Szembek, and S. Hanna AERMOD Low Wind Speed Evaluation Study. Presented at the 10 th EPA Modeling Conference. March. U.S. Environmental Protection Agency (EPA). 2013a. Addendum to the User s Guide for the AMS/EPA Regulatory Model AERMOD. EPA-454/B Office of Air Quality Planning and Standards, Air Quality Assessment Division. Research Triangle Park, North Carolina. December. EPA. 2013b. Addendum to the User s Guide for the AERMOD Meteorological Preprocessor (AERMET). EPA-454/B Office of Air Quality Planning and Standards, Air Quality Assessment Division. Research Triangle Park, North Carolina. December. EPA Revision to the Guideline on Air Quality Models: Adoption of a Preferred General Purpose (Flat and Complex Terrain) Dispersion Model and Other Revisions; Final Rule. Fed. Reg./Vol. 70, No. 216/Wednesday November 9, 2005, Rules and Regulations, pp CFR Part 51, FRL RIN AK60. PAGE 15

111

112

113

114

115

116

117 Lessons Learned from Natural Gas STAR Partners Installing Vapor Recovery Units on Storage Tanks Executive Summary There are about 500,000 crude oil storage tanks in the United States. These tanks are used to hold oil for brief periods of time in order to stabilize flow between production wells and pipeline or trucking transportation sites. In addition, the condensate liquids contained in produced gas that are captured by a mist eliminator filter/ coalescer ahead of the first compressor station in transmission pipelines are often directed to a storage tank as well. During storage, light hydrocarbons dissolved in the crude oil or condensate including methane and other volatile organic compounds (VOC), natural gas liquids (NGLs), hazardous air pollutants (HAP), and some inert gases vaporize or "flash out" and collect in the space between the liquid and the fixed roof of the tank. As the liquid level in the tank fluctuates, these vapors are often vented to the atmosphere. One way to prevent emissions of these light hydrocarbon vapors and yield significant economic savings is to install vapor recovery units (VRUs) on storage tanks. VRUs are relatively simple systems that can capture about 95 percent of the Btu-rich vapors for sale or for use onsite as fuel. Currently, between 7,000 and 9,000 VRUs are installed in the oil production sector, with an average of four tanks connected to each VRU. Natural Gas STAR partners have generated significant savings from recovering and marketing these vapors while at the same time substantially reducing methane and HAP emissions. Partners have found that when the volume of vapors is sufficient, installing a VRU on one or multiple storage tanks can save up to $606,800 per year and payback in as little as two months. This Lessons Learned study describes how partners can identify when and where VRUs should be installed to realize these economic and environmental benefits. Technology Background Underground crude oil contains many lighter hydrocarbons in solution. When the oil is brought to the surface and processed, many of the dissolved lighter hydrocarbons (as well as water) are removed through a series of high-pressure and low-pressure separators. The crude oil is then injected into a storage tank to await sale and transportation off site; the remaining hydrocarbons in the oil are emitted as vapors into the tank. The same principles apply for condensate, which accumulates as a result of the conditions within the pipelines and is removed ahead of the first compressor station. The recovered condensate, which contains dissolved light hydrocarbons, is routed to a storage tank where the dissolved light hydrocarbons are emitted as vapors. These vapors are either vented, flared, or recovered by vapor recovery units (VRUs). Losses of the remaining lighter hydrocarbons are categorized in three ways: Flash losses occur when the separator or heater treater, operating at approximately 35 pounds per square inch (psi), dumps oil into the storage tanks, which are at atmospheric pressure. Working losses refer to the vapor released from the Economic and Environmental Benefits Method for Reducing Natural Gas Losses Volume of Natural Gas Savings (Mcf/yr) Value of Natural Gas Savings ($/yr) 1 $3 per Mcf $5 per Mcf $7 per Mcf Implementation Cost ($) Other Costs ($) $3 per Mcf Payback (Months) $5 per Mcf $7 per Mcf Installing Vapor Recovery Units (VRUs) on Oil Production Storage Tanks 4,900 96,000 $13,965 $273,600 $23,275 $456,000 $32,585 $638,400 $35,738 $103,959 $7,367 $16, Assumes 95% of the annual volume of gas lost can be recovered using a VRU. 1

118 Installing Vapor Recovery Units on Storage Tanks (Cont d) changing fluid levels and agitation of tank contents associated with the circulation of fresh oil through the storage tanks. Standing losses occur with daily and seasonal temperature changes. The volume of gas vapor coming off a storage tank depends on many factors. Lighter crude oils (API gravity>36 ) flash more hydrocarbon vapors than heavier crudes (API gravity<36 ). In storage tanks where the oil is frequently cycled and the overall throughput is high, more working vapors will be released than in tanks with low throughput and where the oil is held for longer periods and allowed to weather. Finally, the operating temperature and pressure of oil in the vessel dumping into the tank will affect the volume of flashed gases coming out of the oil. The makeup of these vapors varies, but the largest component is methane (between 40 and 60 percent). Other components include more complex hydrocarbon compounds such as propane, butane, and ethane; natural inert gases such as nitrogen and carbon dioxide; and HAP like benzene, toluene, ethyl-benzene, and xylene (collectively these four HAP are referred to as BTEX). VRUs can recover over 95 percent of the hydrocarbon emissions that accumulate in storage tanks. Because recovered vapors contain natural gas liquids (even after condensates have been captured by the suction scrubber), they have a Btu content that is higher than that of pipeline quality natural gas (between 950 and 1,100 Btu per standard cubic foot [scf]). Depending on the volume of NGLs in the vapors, the Btu content can reach as high as 2,000 Btu per scf. Therefore, on a volumetric basis, the recovered vapors can be more valuable than methane alone. Exhibit 1 illustrates a VRU installed on a single crude oil storage tank (multiple tank installations are also common). Hydrocarbon vapors are drawn out of the storage (stock) tank under low-pressure, typically between four ounces and two psi, and are first piped to a separator (suction scrubber) to collect any liquids that condense out. The liquids are usually recycled back to the storage tank. From the separator, the vapors flow through a compressor that provides the low-pressure suction for the VRU system. (To prevent the creation of a vacuum in the top of a tank when oil is withdrawn and the oil level drops, VRUs are equipped with a control pilot to shut down the compressor and permit the back flow of vapors into the tank.) The vapors are then metered and removed from the VRU system for pipeline sale or onsite fuel supply. Exhibit 1: Standard Stock Tank Vapor Recovery System 2

119 Installing Vapor Recovery Units on Storage Tanks (Cont d) Economic and Environmental Benefits VRUs can provide significant environmental and economic benefits for oil and gas producers. The gases flashed from crude oil or condensate and captured by VRUs can be sold at a profit or used in facility operations. These recovered vapors can be: Piped to natural gas gathering pipelines for sale at a premium as high Btu natural gas. Used as a fuel for onsite operations. Piped to a stripper unit to separate NGLs and methane when the volume and price for NGLs are attractive. VRUs also capture HAPs and can reduce operator emissions below actionable levels specified in Title V of the Clean Air Act. By capturing methane, VRUs also reduce the emissions of a potent greenhouse gas. Decision Process Companies using fixed roof storage tanks can assess the economics of VRUs by following five easy steps. Step 1: Identify possible locations for VRU installation. Virtually any tank battery is a potential site for a VRU. The keys to successful VRU projects are a steady source and adequate quantity of crude oil or condensate vapors along with an economic outlet for the collected product. The potential volume of vapors will depend on the makeup of the oil or condensate and the rate of flow through the tanks. Pipeline connection costs for routing vapors off site must be considered in selecting sites for VRU installation. meters, however, might not be suitable for measuring total volumes over time due to the low pressures at tanks. Calculating total vapor emissions from oil tanks can be complicated because many factors affect the amount of gas that will be released from a crude oil tank, including: 1. Operating pressure and temperature of the separator dumping the oil to the tank and the pressure in the tank; 2. Oil composition and API gravity; 3. Tank operating characteristics (e.g., sales flow rates, size of tank); and 4. Ambient temperatures. There are two approaches to estimating the quantity of vapor emissions from crude oil tanks. Both use the gas-oil ratio (GOR) at a given pressure and temperature and are expressed in standard cubic feet per barrel of oil (scf per bbl). This process is applicable to all compressor designs. The less common overhung compressors have a single seal, and switching from wet to dry seals would yield half the savings of doing the same for a beam type compressor. The first approach analyzes API gravity and separator pressure to determine GOR (Exhibit 2). These curves were constructed using empirical flash data from laboratory studies and field measurements. As illustrated, this graph can be used to approximate total potential vapor emissions from a barrel of oil. For example, given a certain oil API Exhibit 2: Estimated Volume of Storage Tank Vapors Step 2: Quantify the volume of vapor emissions. Emissions can either be measured or estimated. An orifice well tester and recording manometer (pressure gauge) can be used to measure maximum emissions rates since it is the maximum rate that is used to size a VRU. Orifice Five Steps for Assessing VRU Economics: 1. Identify possible locations for VRU installation; 2. Quantify the volume of vapor emissions; 3. Determine the value of the recovered emissions; 4. Determine the cost of a VRU project; and 5. Evaluate VRU project economics. 3

120 Installing Vapor Recovery Units on Storage Tanks (Cont d) gravity (e.g., 38 ) and vessel dumping pressure (e.g., 40 psi), the total volume of vapors can be estimated per barrel of oil (e.g., 43 scf per bbl). Once the emissions rate per barrel is estimated, the total quantity of emissions from the tank can be determined by multiplying the per barrel estimate by the total amount of oil cycled through the tank. To continue the example above, assuming an average throughput of 1,000 barrels per day (bbl per day), total emissions would be estimated at 43 Mcfd (Exhibit 3). Exhibit 3: Quantity (Q) of Hydrocarbon Vapor Emissions Given: API Gravity = 38 Separator Pressure = 40 psi Oil Cycled = 1,000 bbl/day Vapor Emissions rate = 43 scf/bbl (from Exhibit 2) Q = 43 scf/bbl x 1,000 bbls/day = 43 Mcfd The shortcoming of this approach is that it does not generate information about the composition of the vapors emitted. In particular, it cannot distinguish between VOC and HAP, which can be significant for air quality monitoring, as well as determining the value of the emitted vapors. The second approach is to use the software package E&P Tank version 2.0. This is the modified version of the previous software; the American Petroleum Institute (API) introduced several changes in this model which made it more user-friendly. Partners in the Natural Gas STAR Program have recommended E&P Tank as the best available tool for estimating tank battery emissions. Developed by API and the Gas Research Institute (now the Gas Technology Institute), this software estimates emissions from all three sources flashing, working, and standing using thermodynamic flash calculations for flash losses and a fixed roof tank simulation model for working and standing losses. An operator must have several pieces of information before using E&P Tank, including: 1. Separator pressure and temperature. 2. Separator oil composition. 3. Reference pressure. 4. Reid vapor pressure of sales oil. 5. Sales oil production rate. 6. API gravity of sales oil. E&P Tank also allows operators to input more detailed information about operating conditions, which helps refine emissions estimates. With additional data about tank size, shape, internal temperatures, and ambient temperatures, the software can produce more precise estimates. This flexibility in model design allows users to employ the model to match available information. Since separator oil composition is a key input in the model, E&P Tank includes a detailed sampling and analysis protocol for separator oil. Future versions of the software are being developed to estimate emissions losses from production water tanks as well. Step 3: Determine the value of the recovered emissions. The value of the vapors recovered from VRUs and realized by producers depends on how they are used: 1. Using the recovered vapors onsite as fuel yields a value equivalent to the purchased fuel that is displaced typically natural gas. 2. Piping the vapors (NGL enriched methane) to a natural gas gathering pipeline yield a price that reflects the higher Btu content per Mcf of vapor. 3. Piping the vapors to a processing plant that will strip the NGLs from the gas stream and resell the NGLs and methane separately should also capture the full Btu content value of the vapors. Exhibit 4 illustrates a method of calculating the value of the recovered vapors using an average price of $7.00 per Mcf (for pipeline quality natural gas at 1,000 Btu per scf). Where the Exhibit 4: Value of Recovered Vapors R = Q x P R = The gross revenue Q = The rate of vapor recovery (Mcf/day) P = The price of natural gas Calculate: Q = 41 Mcfd (95% of 43 from Exhibit 3) P = $7.00/Mcf R = 41 Mcfd x $7/Mcf = $287/day $8,800/month $105,600/year 4

121 Installing Vapor Recovery Units on Storage Tanks (Cont d) Methane Content of Natural Gas The average methane content of natural gas varies by natural gas industry sector. The Natural Gas STAR Program assumes the following methane content of natural gas when estimating methane savings for Partner Reported Opportunities. Production 79 % Processing 87 % Transmission and Distribution 94 % Btu content of the vapors is higher, the price per Mcf would be higher. Step 4: Determining the cost of a VRU project. The major cost elements of VRUs are the initial capital equipment and installation costs and operating costs. VRU systems are made by several manufacturers. Equipment costs are determined largely by the volume handling capacity of the unit; the sales line pressure; the number of tanks in the battery; the size and type of compressor; and the degree of automation. The main components of VRUs are the suction scrubber, the compressor, and the automated control unit. Gas measurement is an add-on expense for most units. Prices for typical VRUs and related costs are shown in Exhibit 5. When sizing a VRU, the industry rule-of-thumb is to double the average daily volume to estimate the maximum emissions rate. Thus, in order to handle 43 Mcfd of vapor (Exhibit 3), a unit capable of handling at least 86 Mcfd should be selected. Nelson Price Indexes In order to account for inflation in equipment and operating & maintenance costs, Nelson-Farrar Quarterly Cost Indexes (available in the first issue of each quarter in the Oil and Gas Journal) are used to update costs in the Lessons Learned documents. The Refinery Operation Index is used to revise operating costs while the Machinery: Oilfield Itemized Refining Cost Index is used to update equipment costs. To use these indexes in the future, simply look up the most current Nelson-Farrar index number, divide by the February 2006 Nelson-Farrar index number, and, finally multiply by the appropriate costs in the Lessons Learned. Partners who have installed VRUs and VRU manufacturers report that installation costs can add as much as 50 to 100 percent to the initial unit cost. Installation costs can vary greatly depending on location (remote sites will likely result in higher installation costs) and the number of tanks (larger VRU systems will be required for multiple tanks). Expenses for shipping, site preparation, VRU housing construction (for cold weather protection), and supplemental equipment (for remote, unmanned operations) must also be factored in when estimating installation costs. Operations and maintenance (O&M) expenses vary with the location of the VRU (sites in extreme climates experience more wear), electricity costs, and the type of oil Exhibit 5: Vapor Recovery Unit Sizes and Costs Design Capacity 1 (Mcfd) Compressor Horsepower 2 Capital Costs 3 ($) Installation Costs 3 O&M Costs ($/year) ,421 10,207 20,421 7, ,327 13,164 26,327 8, ,728 15,864 31,728 10, ,529 21, ,529 11, ,405 29,703 59,405 16,839 1 Assumes design capacity is double average vapor recovery rate. 2 Assumes compressor discharge to a 100 psi or less sales line or fuel gas system. 3 Cost information provided by Natural Gas STAR partners and VRU manufacturers. 5

122 Installing Vapor Recovery Units on Storage Tanks (Cont d) produced. For instance, paraffin based oils can clog the VRUs and require more maintenance. Step 5: Evaluate VRU Project Economics. Installing a VRU can be very profitable, depending on the value of the recovered vapors in the local market. Exhibit 6 calculates the simple payback and Internal Rate of Return (IRR) for VRU sizes and costs listed in Exhibit 5. Using an estimate of the value of recovered vapors of $7.00 per Mcf, the potential returns are attractive, particularly for the larger units. When assessing VRU economics, gas price may influence the decision making process; therefore, it is important to re-examine the economics of installing vapor recovery units as natural gas prices change. Exhibit 7 shows an economic analysis of installing a 100 Mcfd vapor recovery unit at different gas prices. Lessons Learned The use of VRUs can profitably reduce methane emissions from crude oil storage tanks. Partners offer the following lessons learned: E&P software can be an effective tool for estimating the amount and composition of vapors from crude oil tanks. Vapor recovery can provide generous returns due to the relatively low cost of the technology and in the cases where there are market outlets for the high BTU vapors. VRUs should be installed whenever they are economic, taking into consideration all of the benefits environmental and economic. Because of the very low pressure differential between Exhibit 6: Financial Analysis for VRU Project Design Capacity (Mcfd) Installation & Capital Costs 1 ($) O&M ($/Year) Value of Gas 2 ($/Yr) Payback 3 (months) Internal Rate of Return 4 (%) 25 35,738 7,367 30, ,073 8,419 60, ,524 10, , ,425 11, , ,959 16, , Unit cost plus estimated installation cost of 75% of unit cost. Actual costs might be greater depending on expenses for shipping, site preparation, supplemental equipment, etc. 2 95% of total gas recovered at $7 per Mcf x 1/2 design capacity x 365 days 3 Based on 10 percent discount rate. 4 Calculated for 5 years. Exhibit 7: Gas Price Impact on Economic Analysis Value of Gas Saved $3/Mcf $5/Mcf $7/Mcf $8/Mcf $52,011 $86,686 $121,360 $138,697 $10/Mcf $173,371 Payback Period (Months) Internal Rate of Return (IRR) NPV (i=10%) 70% 136% 200% 231% $93,947 $213,440 $332,934 $392, % $512,174 6

123 Installing Vapor Recovery Units on Storage Tanks (Cont d) the storage tank and the compressor, large diameter pipe is recommended to provide less resistance to the gas flow. A VRU should be sized to handle the maximum volume of vapors expected from the storage tanks (a rule-of-thumb is double the average daily volume). Rotary vane compressors are recommended for VRUs to move the low volume of gas to low pressures. It is very important to choose reliable, sensitive control systems, because the automated gas flow valves must be opened and closed on very low pressure differences. Include methane emissions reductions from installing VRUs in annual reports submitted as part of the Natural Gas STAR program. One Partner s Experience Chevron USA Production Company installed eight vapor recovery units in 1996 at crude oil stock tanks. As a result, Chevron has realized an estimated reduction in methane emissions of 21,900 Mcf per year from each unit. At today s gas price of $7 per Mcf, this corresponds to approximately $153,300 in savings per unit, or $1,226,400 for all eight units. The capital and installation costs were estimated to be $240,000 ($30,000 per unit) in 1996 or the equivalent of $324,000 ($40,500 per unit) in 2006 dollars. This particular project would have realized a payback in just over 3 months in References Bigelow, Tom and Renee Wash "VRUs Turn Vented Gas Into Dollars." Northeast Oil Reporter. October pp Choi, M.S API Tank Vapors Project. Presented at the 1993 SPE Technical Conference, Houston, TX, October 3-6, SPE Technical Paper No Dailey, Dirk, Universal Compression, personal contact. Evans, G.B. and Ralph Nelson Applications of Vapor Recovery to Crude Oil Production. Hy-Bon Engineering Company. Midland, TX. SPE Technical Paper No Griswold, John A., Power Services, Inc. and Ted C. Ambler, A & N Sales, Inc A Practical Approach to Crude Oil Stock Tank Vapor Recovery. Presented at the 1978 SPE Rocky Mountain Regional Meeting, Cody, WY, May 7-9, SPE Technical Paper No Henderson, Carolyn, U.S. EPA Natural Gas STAR Program, personal contact. Hy-Bon Engineering Company, Inc Product Bulletin: Vapor Recovery Systems. Liu, Dianbin and J.V. Meachen Jr., The Use of Vapor Recovery Units in the Austin Chalk Field. Presented at the 1993 SPE Technical Conference, Houston, TX, October 3-6, SPE Technical Paper No Lucas, Donald, David Littlejohn, Ernest Orlando, Lawrence Berkeley National Laboratory; and Rhonda P. Lindsey, U.S. Department of Energy The Heavy Oil Storage Tank Project. Presented at the 1997 SPE/EPA Exploration and Production Environmental Conference, Dallas, TX, March SPE Technical Paper No Martin, Mark, UMC Automation, personal contact. Moreau, Roland, Exxon-Mobil USA, personal contact. Motley, Jack, V.R. Systems, Inc., personal contact. Newsom, Vick L Determination of Methane Emissions From Crude Oil Stock Tanks. Presented at the SPE/EPA Exploration & Production Environmental Conference, Dallas, TX, March 3-5, SPE Technical Paper No Presley, Charles, A.G. Equipment, personal contact. Primus, Frank A., Chevron USA, personal contact. Tims, Arnold, Hy-Bon Engineering Company, Inc., personal contact. Tingley, Kevin, U.S. EPA Natural Gas STAR Program, personal contact. U.S. Department of Commerce Control of Volatile Organic Compound Emissions from Volatile Organic Liquid Storage in Floating and Fixed Roof Tanks. Available through NTIS. Springfield, VA PB U.S. Environmental Protection Agency Methane 7

124 Installing Vapor Recovery Units on Storage Tanks (Cont d) Emissions from the U.S. Petroleum Industry (Draft Document). DCN: Visher, Stuart, A.C. Compressors, personal contact. Watson, Mark C "VRU Engineered For Small Volumes." The American Oil & Gas Reporter (Special Report: Enhanced Recovery). March pp Webb, W.G Vapor Jet System: An Alternate Vapor Recovery Method. Presented at the 1993 SPE/EPA Exploration & Production Environmental Conference, San Antonio, TX, March 7-10, SPE Technical Paper No Weldon, R.E. Jr., "Could You Recover Stock Tank Vapors at a Profit?" The Petroleum Engineer. May pp. B29-B33. Weust, John, Marathon Oil, personal contact. 8

125 Installing Vapor Recovery Units on Storage Tanks (Cont d) United States Environmental Protection Agency Air and Radiation (6202J) 1200 Pennsylvania Ave., NW Washington, DC October 2006 EPA provides the suggested methane emissions estimating methods contained in this document as a tool to develop basic methane emissions estimates only. As regulatory reporting demands a higher-level of accuracy, the methane emission estimating methods and terminology contained in this document may not conform to the Greenhouse Gas Reporting Rule, 40 CFR Part 98, Subpart W methods or those in other EPA regulations. 9

126 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 1 of 6 2/28/2012 TANKS 4.0.9d Emissions Report - Detail Format Tank Indentification and Physical Characteristics Identification User Identification: City: State: Company: Type of Tank: Description: 500 gal glycol Horizontal Tank Tank Dimensions Shell Length (ft): 6.00 Diameter (ft): 4.00 Volume (gallons): Turnovers: 0.00 Net Throughput(gal/yr): 2, Is Tank Heated (y/n): N Is Tank Underground (y/n): N Paint Characteristics Shell Color/Shade: Shell Condition Gray/Light Good Breather Vent Settings Vacuum Settings (psig): Pressure Settings (psig) 0.03 Meterological Data used in Emissions Calculations: Salt Lake City, Utah (Avg Atmospheric Pressure = psia)

127 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 2 of 6 2/28/2012 TANKS 4.0.9d Emissions Report - Detail Format Liquid Contents of Storage Tank 500 gal glycol - Horizontal Tank Daily Liquid Surf. Temperature (deg F) Liquid Bulk Temp Vapor Pressure (psia) Vapor Mol. Liquid Mass Vapor Mass Mol. Basis for Vapor Pressure Mixture/Component Month Avg. Min. Max. (deg F) Avg. Min. Max. Weight. Fract. Fract. Weight Calculations Propylene glycol All Option 2: A=8.2082, B=2085.9, C=203.54

128 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 3 of 6 2/28/2012 TANKS 4.0.9d Emissions Report - Detail Format Detail Calculations (AP-42) 500 gal glycol - Horizontal Tank Annual Emission Calcaulations Standing Losses (lb): Vapor Space Volume (cu ft): Vapor Density (lb/cu ft): Vapor Space Expansion Factor: Vented Vapor Saturation Factor: Tank Vapor Space Volume: Vapor Space Volume (cu ft): Tank Diameter (ft): Effective Diameter (ft): Vapor Space Outage (ft): Tank Shell Length (ft): Vapor Density Vapor Density (lb/cu ft): Vapor Molecular Weight (lb/lb-mole): Vapor Pressure at Daily Average Liquid Surface Temperature (psia): Daily Avg. Liquid Surface Temp. (deg. R): Daily Average Ambient Temp. (deg. F): Ideal Gas Constant R (psia cuft / (lb-mol-deg R)): Liquid Bulk Temperature (deg. R): Tank Paint Solar Absorptance (Shell): Daily Total Solar Insulation Factor (Btu/sqft day): 1, Vapor Space Expansion Factor Vapor Space Expansion Factor: Daily Vapor Temperature Range (deg. R): Daily Vapor Pressure Range (psia): Breather Vent Press. Setting Range(psia): Vapor Pressure at Daily Average Liquid Surface Temperature (psia): Vapor Pressure at Daily Minimum Liquid Surface Temperature (psia): Vapor Pressure at Daily Maximum Liquid Surface Temperature (psia): Daily Avg. Liquid Surface Temp. (deg R): Daily Min. Liquid Surface Temp. (deg R): Daily Max. Liquid Surface Temp. (deg R): Daily Ambient Temp. Range (deg. R): Vented Vapor Saturation Factor Vented Vapor Saturation Factor: Vapor Pressure at Daily Average Liquid: Surface Temperature (psia): Vapor Space Outage (ft): Working Losses (lb): Vapor Molecular Weight (lb/lb-mole): Vapor Pressure at Daily Average Liquid Surface Temperature (psia): Annual Net Throughput (gal/yr.): 2, Annual Turnovers: Turnover Factor:

129 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 4 of 6 2/28/2012 Tank Diameter (ft): Working Loss Product Factor: Total Losses (lb):

130 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 5 of 6 2/28/2012 TANKS 4.0.9d Emissions Report - Detail Format Individual Tank Emission Totals Emissions Report for: Annual 500 gal glycol - Horizontal Tank Losses(lbs) Components Working Loss Breathing Loss Total Emissions Propylene glycol

131 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 6 of 6 2/28/2012

132 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 1 of 6 2/29/2012 TANKS 4.0.9d Emissions Report - Detail Format Tank Indentification and Physical Characteristics Identification User Identification: City: State: Company: Type of Tank: Description: 525 gal Methanol Tank Horizontal Tank Tank Dimensions Shell Length (ft): 6.00 Diameter (ft): 4.00 Volume (gallons): Turnovers: 3.81 Net Throughput(gal/yr): 2, Is Tank Heated (y/n): N Is Tank Underground (y/n): N Paint Characteristics Shell Color/Shade: Shell Condition Red/Primer Good Breather Vent Settings Vacuum Settings (psig): Pressure Settings (psig) 0.03 Meterological Data used in Emissions Calculations: Salt Lake City, Utah (Avg Atmospheric Pressure = psia)

133 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 2 of 6 2/29/2012 TANKS 4.0.9d Emissions Report - Detail Format Liquid Contents of Storage Tank 525 gal Methanol Tank - Horizontal Tank Daily Liquid Surf. Temperature (deg F) Liquid Bulk Temp Vapor Pressure (psia) Vapor Mol. Liquid Mass Vapor Mass Mol. Basis for Vapor Pressure Mixture/Component Month Avg. Min. Max. (deg F) Avg. Min. Max. Weight. Fract. Fract. Weight Calculations Methyl alcohol Jan Option 2: A=7.897, B= , C= Methyl alcohol Feb Option 2: A=7.897, B= , C= Methyl alcohol Mar Option 2: A=7.897, B= , C= Methyl alcohol Oct Option 2: A=7.897, B= , C= Methyl alcohol Nov Option 2: A=7.897, B= , C= Methyl alcohol Dec Option 2: A=7.897, B= , C=229.13

134 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 3 of 6 2/29/2012 TANKS 4.0.9d Emissions Report - Detail Format Detail Calculations (AP-42) 525 gal Methanol Tank - Horizontal Tank Month: January February March April May June July August September October November December Standing Losses (lb): Vapor Space Volume (cu ft): Vapor Density (lb/cu ft): Vapor Space Expansion Factor: Vented Vapor Saturation Factor: Tank Vapor Space Volume: Vapor Space Volume (cu ft): Tank Diameter (ft): Effective Diameter (ft): Vapor Space Outage (ft): Tank Shell Length (ft): Vapor Density Vapor Density (lb/cu ft): Vapor Molecular Weight (lb/lb-mole): Vapor Pressure at Daily Average Liquid Surface Temperature (psia): Daily Avg. Liquid Surface Temp. (deg. R): Daily Average Ambient Temp. (deg. F): Ideal Gas Constant R (psia cuft / (lb-mol-deg R)): Liquid Bulk Temperature (deg. R): Tank Paint Solar Absorptance (Shell): Daily Total Solar Insulation Factor (Btu/sqft day): , , Vapor Space Expansion Factor Vapor Space Expansion Factor: Daily Vapor Temperature Range (deg. R): Daily Vapor Pressure Range (psia): Breather Vent Press. Setting Range(psia): Vapor Pressure at Daily Average Liquid Surface Temperature (psia): Vapor Pressure at Daily Minimum Liquid Surface Temperature (psia): Vapor Pressure at Daily Maximum Liquid Surface Temperature (psia): Daily Avg. Liquid Surface Temp. (deg R): Daily Min. Liquid Surface Temp. (deg R): Daily Max. Liquid Surface Temp. (deg R): Daily Ambient Temp. Range (deg. R): Vented Vapor Saturation Factor Vented Vapor Saturation Factor: Vapor Pressure at Daily Average Liquid: Surface Temperature (psia): Vapor Space Outage (ft): Working Losses (lb): Vapor Molecular Weight (lb/lb-mole): Vapor Pressure at Daily Average Liquid Surface Temperature (psia): Net Throughput (gal/mo.): Annual Turnovers: Turnover Factor:

135 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 4 of 6 2/29/2012 Tank Diameter (ft): Working Loss Product Factor: Total Losses (lb):

136 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 5 of 6 2/29/2012 Emissions Report for: January, February, March, October, November, December 525 gal Methanol Tank - Horizontal Tank TANKS 4.0.9d Emissions Report - Detail Format Individual Tank Emission Totals Losses(lbs) Components Working Loss Breathing Loss Total Emissions Methyl alcohol

137 TANKS 4.0 Report file://c:\program Files (x86)\tanks409d\summarydisplay.htm Page 6 of 6 2/29/2012

138 BEST MANAGEMENT PRACTICE Management of Fugitive Emissions at Upstream Oil and Gas Facilities January

139 The Canadian Association of Petroleum Producers (CAPP) represents 150 companies that explore for, develop and produce natural gas, natural gas liquids, crude oil, oil sands, and elemental sulphur throughout Canada. CAPP member companies produce more than 95 per cent of Canada s natural gas and crude oil. CAPP also has 130 associate members that provide a wide range of services that support the upstream crude oil and natural gas industry. Together, these members and associate members are an important part of a $100-billion-a-year national industry that affects the livelihoods of more than half a million Canadians. Review by January 2009 Disclaimer This publication is issued by the Canadian Association of Petroleum Producers (CAPP). While it is believed that the information contained herein is reliable under the conditions and subject to the limitations set out, CAPP, its consultants Clearstone Engineering Ltd and the Project Steering Committee do not guarantee its accuracy. The use of this report or any information contained will be at the user s sole risk, regardless of any fault or negligence of Clearstone Engineering Ltd., the Project Steering Committee, CAPP or its co-funders. 2100, th Ave. S.W. Calgary, Alberta Canada T2P 3N9 Tel (403) Fax (403) , 235 Water Street St. John s, Newfoundland Canada A1C 1B6 Tel (709) Fax (709) communication@capp.ca Website:

140 Section Table of Contents Page List of Acronyms...i Acknowledgment... ii Forward... iii 1 Applicability Implementation And Schedule For Review Schedule for review Basic Control Strategy Technology and Standards Management Systems Directed Inspection & Maintenance (DI&M) Program Leak Definition Leak Detection Leak Quantification Target Components Monitoring Frequency Inaccessible Components Tagging Components Leak Repairs Personnel Training Primary Calibrations and Field checks Inspection, Monitoring & Record-Keeping Corporate Commitment Appendices References Cited...15 Appendix 1 Table 4. Examples Of Leak Monitoring Frequencies For Leak-Prone Equipment Components...17 Appendix 2 Appendix 3 Sample Leaker Tag...18 Appendix III Leak Survey Forms...19 Appendix 4 Economic Analysis...23 Appendix 5 Emissions Inventory From Method 21 Data Leak/No-leak Emission Factors Three-stratum Emission Factors Published Leak-Rate Correlations Unit-Specific Leak-Rate Correlations...32 January 2007 Management of Fugitive Emissions At Upstream Oil and Gas Facilities

141 Appendix 6 Component-Specific Control Options Reciprocating Compressors Vent Monitoring Systems Emission-Controlling Vent Systems High Performance Packing Systems Barrier Fluid Systems Purge Gas Systems Unit Shutdown Practices Static Packing Systems Valve Cap Leakage Centrifugal Compressors Emission-Controlling Vent Systems Used with Conventional Seals Dry Gas Seals Valves Sewers and Drains Pumps Threaded and Flanged Connections Pressure Relief Devices Open-ended Valves and Lines Sampling Points...44 Appendix 7 US EPA Method January 2007 Management of Fugitive Emissions At Upstream Oil and Gas Facilities

142 List of Tables Table Page 1 Leak detection and measurement methods. 6 2 Summary of screening and measurement techniques 7 3 Default mean life of repairs for economic analysis of repair costs Examples of leak monitoring frequencies for leak prone equipment components Sample calculations of the simple payback period for individual leak repairs Leak/No-leak emission factors for estimating fugitive leaks at UOG facilities Three-stratum emission factors for estimating fugitive leaks at UOG facilities Correlation parameters for estimating leak rates from equipment components Default zero emission rates Common causes of leakage from flanged and threaded connections 43 Figures 1. DI&M Decision Tree 3 2. Schematic diagram of a piston-rod packing-case system on a reciprocating compressor 35 January 2007 Management of Fugitive Emissions At Upstream Oil and Gas Facilities

143 List of Acronyms AENV - Alberta Environment API - American Petroleum Institute ASME - American Society of Mechanical Engineers BMP - Best Management Practice CAC - Criteria Air Contaminant CAPP - Canadian Association of Petroleum Producers CASA - Clean Air Strategic Alliance CCME - Canadian Council of Ministers of the Environment DI&M - Direct Inspection and Maintenance (DI&M) EUB - Alberta Energy and Utilities Board EC - Environment Canada GHG - Greenhouse Gases MTBF - Mean Time Between Failure NACE - National Association of Corrosion Engineers NFPA - National Fire Protection Association NMHC - Non-Methane Hydrocarbon NPT - National Pipe Thread PTAC - Petroleum Technology Alliance of Canada PTFE - Polytetrafluoroethylene SEPAC - Small Explorers and Producers Association of Canada THC - Total Hydrocarbons TOC - Total Organic Compounds TNMOC - Total Non-methane Organic Compounds UOG - Upstream Oil and Gas U.S. EPA - U.S. Environmental Protection Agency VOC - Volatile Organic Compound January 2007 Management of Fugitive Emissions i At Upstream Oil and Gas Facilities

144 Acknowledgment Special thanks are given to the following individuals and companies who participated on the project steering committee and/or provided review and comments at different times throughout the duration of the project: M. Brown - Alberta Energy and Utilities Board B. Forsyth - BP Canada C. Chamberland - Shell Canada Limited W. Hillier - Husky Energy M. Layer - Environment Canada S. McIntyre - Canadian Natural Resources Limited L. Miller - BP Canada R. Pettipas - Conoco-Phillips Canada S. Reilly - Talisman Energy B. Ross - Nexen Inc. S. Sian - CAPP R. Sikora - KEYERA Energy J. Squarek - CAPP R. Sundermann - EnCana J. Tatham - BP Canada A. Varley - Petro-Canada January 2007 Management of Fugitive Emissions At Upstream Oil and Gas Facilities

145 Forward The issue of fugitive emissions management has and continues to be important to the upstream oil and gas (UOG) industry. The genesis of this fugitive emissions Best Management Practice (BMP) can be traced back to earlier undertakings of the Air Research Planning Committee (ARPC) of the Petroleum Technology Alliance of Canada (PTAC). The Canadian Association of Petroleum Producers (CAPP), the Small Explorers and Producers Association of Canada (SEPAC), Environment Canada (EC) and the Alberta Energy and Utilities Board (EUB) have coordinated efforts to develop this BMP which also satisfies recommendations No. 43 and 44 of the Clean Air Strategic Alliance (CASA) Flaring and Venting Project Team (FVPT): 43) CAPP and SEPAC develop a best management practices document by December 31, 2005 to assist the upstream oil and gas industry in managing fugitive emissions and targeting sources that are most likely to have larger volume emissions and which would be more cost effective to address. CAPP and SEPAC will incorporate improvement to emission factors into the best management practices document as they become available. 44) Once a best management practices document has been developed by CAPP and SEPAC, the EUB should require licensees to develop and implement leak detection and repair programs to minimize fugitive emissions from upstream petroleum industry facilities. The aim of this BMP is to assist the UOG industry in meeting the requirements under section 8.7 of the EUB Directive 060 (Upstream Petroleum Industry Flaring, Incinerating, and Venting) and in cost effectively managing the most likely sources of significant fugitive emissions. The emissions of primary concern are methane (CH 4 ) and non-methane volatile organic hydrocarbons (NMVOC). This BMP: identifies the typical key sources of fugitive emissions at UOG facilities, presents strategies for achieving cost-effective reductions in these emissions (e.g., through improved designs, Directed Inspection and Maintenance (DI&M) practices, improved operating practices, and the application of new and retrofit technologies), and summarizes key considerations and constraints. While this BMP is specific to fugitive equipment leaks, it considers leakage directly to the atmosphere and unintentional gas carry-through to storage tanks. The BMP also considers emerging technologies that have the potential to improve the efficiency and effectiveness of leak detection. The overall aim is to provide practical guidance to operators for developing focused approaches to manage and reduce fugitive emissions at individual oil and gas facilities, while giving consideration to each facility s specific circumstances. January 2007 Management of Fugitive Emissions iii At Upstream Oil and Gas Facilities

146 1 Applicability This BMP provides guidance for the management of fugitive emissions at UOG facilities from leaks (i.e., the loss of process fluid to the environment past a seal, threaded or mechanical connection, cover, valve seat, flaw or minor damage point) on equipment components in hydrocarbon service. A component is considered to be in hydrocarbon service when the process fluid being handled contains greater than 10 percent hydrocarbons on a mass basis. Fugitive emissions from equipment leaks are unintentional losses and may arise due to normal wear and tear, improper or incomplete assembly of components, inadequate material specification, manufacturing defects, damage during installation or use, corrosion, fouling and environmental effects. Components also tend to have greater average emissions when subjected to frequent thermal cycling, vibrations or cryogenic service. Only a small percentage of the equipment components have any measurable leakage, and of those only a small percentage contributes to most of the emissions. Thus, the control of fugitive emissions is a matter of minimizing the potential for big leaks and providing early detection and repair. The UOG industry is characterized by many small widely dispersed facilities rather than a few large facilities so it is appropriate to apply a directed approach that targets the sources most practicable to control, components most likely to result in big leaks. At each target facility, efforts should be focused on the areas most likely to offer significant, cost-effective control opportunities (e.g., on specific component types and service applications). This BMP is designed to apply to components in sweet gas service which are expected to represent the greatest opportunity for emission reductions.1 Existing mechanisms to address odour, health and safety concerns in sour facilities are deemed to meet or exceed the purpose of this BMP.2 Furthermore, this BMP is designed to apply primarily to fugitive equipment leaks from components in natural gas or hydrocarbon vapour service. This reflects the greater difficulty in containing a gas than a liquid (i.e., due to the greater mobility or fluidity of gases), and the general reduced visual indications of gas leaks. 1 For the purpose of this BMP, sweet gas is defined as natural gas with an H2S concentration of less than 10 mol/kmol in accordance with the approach adopted in the EUB Directive Odour emissions are addressed under section 14 of the EUB Directive 064, which states that in accordance with the OGC Act, part 1(4)(f), the EUB has jurisdiction to control pollution above, at, or below the surface in the drilling wells and in operations for the production of oil and gas. In this regard, migration of H2S emissions off lease is considered pollution and may be defined as a major unsatisfactory inspection. Appendix 1, 3 and 11 of Directive 064 also provide additional information regarding off-site regulatory requirements for H2S odour management. January 2007 Management of Fugitive Emissions 1 At Upstream Oil and Gas Facilities

147 2 Implementation And Schedule For Review Efficient management of fugitive emissions is best achieved through the application of DI&M techniques. DI&M focuses inspection and correction efforts on the areas most likely to offer significant cost-effective control opportunities, with coarse or less frequent screening of other areas for additional opportunities. The Decision Tree reproduced under Figure 1 has been developed to provide a process to effectively manage fugitive emissions. When phasing-in their DI&M program, companies may consider factors such as size and age of facility or type of facility, percentage of components per year, percentage of facilities, business units, geographic area, shutdown schedule, and economic classification of repairs. The implementation of a DI&M program at existing UOG facilities should be completed by December 31, While this BMP provides for a phasing-in period, CAPP members will be requested to provide interim reports on the status of implementation of a DI&M program within their companies by March 31, 2008 and March 31, These status reports will indicate any problem areas or improvements and allow for updating and additions to the BMP as necessary. When implementing this BMP, companies should keep in mind that the EUB Directive 060 imposes a mandatory requirement to implement a program to detect and repair leaks and that such a program must meet or exceed CAPP BMP. In view of the above, the use of the world should in this BMP does not imply that action is not necessary within the context of the EUB Directive 060; i.e., alternative methodologies to those described in the BMP can be used as long as the expected results are achieved or exceeded. No action is not an option. 2.1 Schedule for review Concurrent with the introduction of this BMP is a field investigation program to verify the effectiveness of the methodologies proposed in this document. The field work should provide further knowledge on key detection, monitoring and control technologies and practices, as well as additional data to review and modify as needed the recommended monitoring frequency and default repair lives for different types of equipment. This field investigation work will be conducted in cooperation with EC and EUB and will extend over a two-year period until the fourth quarter of Interim and final reports will be presented as a basis for the continued improvement of this BMP. A review of the document is expected to take place within 24 months from the date of implementation to incorporate any adjustments that CAPP and its members deem necessary as result of the expertise gained during the first year of implementation and the ongoing field investigation work. Thereafter, this BMP will be regularly reviewed and revised as necessary. January 2007 Management of Fugitive Emissions 2 At Upstream Oil and Gas Facilities

148 Figure 1 - DI&M Decision Tree: Implement a DI&M program to focus on big leaks. Then for each leak detected proceed as follows. s.3.2.1, 3.2.2, Are there health, safety, or environmental concerns? s yes Repair no leaking components s no no no Is the leak easy to fix? yes Does the leak require a shutdown to fix? no no or yes Measure or estimate the size of the leak and evaluate the cost effectiveness of fixing it. s Add to repair list and fix during next planned/unplanned shutdown without quantification. s Is the leak economical to fix? s yes no Re-evaluate at the next scheduled leak survey January 2007 Management of Fugitive Emissions 3 At Upstream Oil and Gas Facilities

149 3 Basic Control Strategy The key elements for effective long-term control of fugitive emissions are the application of best available technology and standards, implementation of management systems, and corporate commitment. The application of control technologies and design standards, alone, do not preclude the potential for fugitive emissions. Reliable fugitive emissions control requires: the development of monitoring programs, operating procedures and performance objectives for controlling fugitive emissions, and Management s commitment to the implementation and maintenance of a DI&M program. 3.1 Technology and Standards The first step in controlling fugitive equipment leaks should always be to minimize potential for leaks by applying proper design and material-selection standards, to follow the manufacturer s specifications for the installation, use and maintenance of components and to implement practicable control technologies (e.g., reduction, recovery and treatment systems). 3.2 Management Systems A management system is needed to establish objective performance targets and to implement ongoing monitoring and predictive maintenance programs to ensure that leaks are detected and remain well controlled. The following sections describe the basic elements of a DI&M program Directed Inspection & Maintenance (DI&M) Program The first step is to determine which types of components will be targeted (i.e., subjected to regular screening for leaks). The objective is to minimize the potential for leaks in the most practicable manner possible. This is done by focusing efforts on the types of components and service applications most likely to offer significant cost-effective control opportunities (see Section 3.2.5). Nontarget components are subjected to coarse or less frequent screening. Typically, a facility will phase the DI&M program over a certain number of years by progressively adding to the list of target components until all key potential contributors are being targeted. Once a leak is detected, regardless of whether it is a target or non-target component, the Decision Tree reproduced under Figure 1 should be followed to determine if a leak need to be repaired. Once a leak is determined to need fixing, this should be done within a reasonable period of time (see Section 3.2.9), or at the next facility turnaround if a major shutdown is required. A facility may choose to simply repair or fix the leak. If it is not a simple repair or fix, an operator may choose to program the repair at the next shut down without quantification or, alternatively, the leak should be measured or estimated to determine if it is economical to repair. Where an operator believes that it may not be economical to repair, this should be documented based on reliable quantification of the amount of leakage and the repair costs (see Sections January 2007 Management of Fugitive Emissions 4 At Upstream Oil and Gas Facilities

150 and 3.2.8). If a leak poses a health, safety, or environmental concern, then it needs to be repaired regardless of whether it is economical to fix Leak Definition Fugitive emissions control is becoming more common as a condition of a facility s operating approval. Accordingly, it is useful to consider a definition that corresponds to those typically applied in other industries. Firstly, a leak could be defined as a screening concentration of 10,000 ppm or more 3 for the purposes of deciding whether to measure the emission rate and evaluate the practicability of making repairs. Below this threshold the emissions generally become too small to quantify. Moreover, usually only the top 5 to 10 percent of leaking components account for 80 to 90 percent of the emissions at a facility. Consequently, there is limited value in dedicating resources to measure or estimate emissions from components that do not achieve the screening value identified. However facilities may still choose to repair these below ppm emissions without measurement. There are several new emerging technologies that have the potential to improve efficiency and effectiveness of leak detection programs and replace US EPA Method 21. These technologies include: differential lasers to measure atmospheric concentrations of component gases, computer analysis of ambient air sample trends to estimate leak source location and volumes and infrared optical technology to visually inspect the components. Many of these and other technologies are being developed and can be used to identify leaking components Leak Detection Leak screening should be done on accessible components using a portable organic vapour analyzer in accordance with US EPA Method 21 or using such alternative methods that provide an equivalent result (see Section for the leak definition). In some cases, US EPA Method 21 has been considered too slow and labour intensive and more suited for large facilities. For these reasons, this BMP provides the opportunity to use other methods. There are several new emerging technologies that have the potential to improve efficiency and effectiveness of leak detection programs and replace US EPA Method 21. These technologies include: differential lasers to measure atmospheric concentrations of component gases, computer analysis of ambient air sample trends to estimate leak source location and volumes and infrared optical technology to visually inspect the components. Many of these and other technologies are being developed and can be used to identify leaking components. Using alternatives to US EPA Method 21 may also allow operators to evaluate components that may have not been accessible otherwise. 3 This is the current leak definition applied by the CCME (1993) guidelines for the measurement and control of fugitive volatile organic compound (VOC) emissions from equipment leaks at petroleum refineries and organic chemical plants based on US EPA Method 21. January 2007 Management of Fugitive Emissions 5 At Upstream Oil and Gas Facilities

151 3.2.4 Leak Quantification The quantification by measurement or estimation of leak rates to evaluate the feasibility of repairing or replacing a component should be sufficiently accurate for this purpose (e.g., within ±25 percent or enough to clearly establish a positive net financial benefit). Depending on the type of component and information available, potentially valid quantification methods may include, but are not limited to, process modelling, material balances, flow capture and metering systems, duct sampling techniques, tracer tests and some types of remote sensing methods. Table 1 provides a list of potential methods to detect, and measure or estimate leaks. Table 1. Leak Detection and Measurement Methods 1. Qualitative Methods 2 Quantitative Methods 3 Bubble Tests Portable Organic Vapour Analyzers 4 Optical emissions detection Quantitative remote sensing techniques (Leak imaging) Ultrasonic Leak Detectors Engineered estimates 1 This is not necessarily a complete list of valid methods. 2 A leak detection method is deemed to be qualitative where it provides a leak detection capability consistent with, or better than, the leak definition given in Section but is not able to provide a quantitative output that can be related to the leak definition. If a qualitative method can be enhanced to consistently provide quantitative output that can be related to the leak definition then it may be reclassified as a quantitative method. 3 A leak detection method is deemed to be quantitative where it provides a minimum leak detection capability consistent with, or better than, the leak definition given in Section and provides quantitative output that can be related to the leak definition. 4 Operators note that the sensors may be damaged by vapours at high concentrations or give a false reading depending on the calibration gas. January 2007 Management of Fugitive Emissions 6 At Upstream Oil and Gas Facilities

152 Table 2 provides an indication of cost effectiveness for some of the screening and measurement techniques based on an EPA s Lessons Learned Study. Table 2 Summary of Screening and Measurement Techniques Instrument/Technique Effectiveness Approximate Capital Cost Soap Solution ** $ Electronic Gas Detectors * $$ Acoustic Detection / ** $$$ Ultrasound Detection Toxic vapour analyzer / * $$$ Flame ionization Detector Bagging * $$$ High Volume Sampler *** $$$ Rotameter ** $$ Leak Imaging *** $$$ * Least effective at screening/measurement *** Most effective at screening/measurement $ Smallest capital cost $$$ Largest capital cost Source: EPA s Lessons Learned Study & Presentation to Energy Management, Workshop, Methane to Markets, Directed Inspection and Maintenance, Roger Fernandez, EPA, January 16, Target Components All equipment components on process-, fuel- and waste-gas systems are potential sources of fugitive emissions. The types of components may include flanged and threaded connections (i.e., connectors), valves, pressure-relief devices, openended lines, blowdown vents (i.e., during passive periods), instrument fittings, regulator and actuator diaphragms, compressor seals, engine and compressor crankcase vents, sump and drain tank vents and covers. The amount of emissions from a leaking component is generally independent of the size of the component. Furthermore, as previously mentioned, usually only the top 5 to 10 percent of leaking components account for 80 to 90 percent of the emissions. For equipments in gas service, the most cost-effective types of components to target tend to be, in the order of decreasing cost-effectiveness: compressor seals, open-ended lines, pressure relief valves, regulators, and control valves. The least cost-effective components to target tend to be connectors and block valves. The priority and feasibility of repairing a given component will depend on the leak rate, value of the process fluid being lost, cost of repairs, life expectancy of the repair, and the value of various potential indirect factors such as avoiding safety, January 2007 Management of Fugitive Emissions 7 At Upstream Oil and Gas Facilities

153 health, and environment impacts, avoiding damage to the component, improved process reliability and better performance. Storage tanks at production and processing facilities are potentially a significant source of emissions due to working or evaporation losses; particularly where intentional boiling or flashing losses occur. Other less recognized, and often unaccounted for, contributions to atmospheric emissions from storage tanks may include the following: Leakage of process gas or volatile product past the seats of drain or blowdown valves into the product header leading to the tanks. Inefficient separation of gas and liquid phases upstream of the tanks allowing some gas carry-through (by entrainment) to the tanks. This usually occurs where liquid volume (e.g., produced water) has increased significantly over time resulting in a facility s inlet separators being undersized for current conditions. Piping changes which result in the unintentional placement of high vapour pressure product in tanks not equipped with appropriate vapour controls. Displacement of large volumes of gas to storage tanks during pigging operations. Malfunctioning or improperly set blanket gas regulators and vapour control valves can result in excessive blanket gas consumption and, in turn, increased flows to the end control device (e.g., vent, flare or vapour recovery compressor). The blanket gas is both a carrier of product vapours and a potential pollutant itself (i.e., natural gas is usually used as the blanket medium for blanketed tanks at gas processing plants). Leakage from components in oil or light hydrocarbon liquid service is usually easy to visually detect and often is well controlled by normal maintenance programs. However due to their low average leak rates, less substantive improvements in fugitive emissions reductions are expected for these components Monitoring Frequency The equipment components most likely to leak should be screened most frequently. Studies indicate that components subject to vibration, high use, or temperature cycles are the most leak-prone. Operators should develop a DI&M survey schedule that achieves maximum cost-effective fugitive emissions reductions yet also suits the unique characteristics and operations of their facility. Operators may choose to determine the frequency of follow-up surveys based on different factors such as anticipated life of repairs made during their previous survey, company maintenance cycle, and availability of resources. If subsequent surveys show numerous large or recurring leaks, the operator may chose to increase the frequency of follow-up surveys. These follow-up surveys may focus on components repaired during previous surveys, or on the classes of components identified as most likely to leak. Over time, operators can continue to fine-tune January 2007 Management of Fugitive Emissions 8 At Upstream Oil and Gas Facilities

154 the scope and frequency of surveys as leak patterns emerge. 4 Where the repair frequency is high, permanent leak detection system may be a more practical solution and should be considered. Examples of leak monitoring frequencies for leak-prone components are provided under Appendix 1. Operators should design a frequency monitoring program best suited for its operations while ensuring maximum cost-effective fugitive emissions reductions Inaccessible Components Inaccessible sources or components can be defined as equipment that is more than 2 metres above a permanently available support surface or is cover protected or insulated (CCME 1993). This equipment is excluded from a DI&M program under this BMP. However, if these components are leaking and become accessible during facility shut-downs or turn-around, they should then be repaired Tagging Components All leaking components should be flagged using a tag or an alternative method for identification purposes as well as to ensure that the component is repaired and that it will be given appropriate follow-up attention under the company s DI&M program. This should assist in identifying the proper monitoring frequency for that specific component. An example of a leaker tag is provided under Appendix II while Appendix III provides an example of leak survey forms. Companies should determine the best format for flagging/identifying leaks as well as for leak surveys in accordance with the unique characteristics and operations of their facilities Leak Repairs Decisions to repair or replace leaking components should be made on a case-bycase basis in consideration of health, safety, environmental, and economical concerns. Where feasible, repairs or replacements should be done within 45 days from the time a leak is detected. Where a major shutdown is required to facilitate this work, or there are marginal economics for repairing the component, the repair or replacement may be delayed until the next planned shutdown, provided this does not pose any safety, health, or environmental concerns. A leaking component need not be repaired if the component is shown to be uneconomical to repair and does not pose a safety, health, or environmental concern. In such cases, the components should remain tagged/identified and be rescreened at the next scheduled leak survey. The economics of repairing a leak or replacing the components should be based on the market value of the process fluid being lost, the repair, the replacement cost, and the life expectancy of the applied solution. All leak repairs that have a simple payback period of less than 1 year based on the following equation should 4 Lessons Learned from Natural Gas Star Partners, Directed Inspection and Maintenance at gas processing plants and booster stations, EPA, October January 2007 Management of Fugitive Emissions 9 At Upstream Oil and Gas Facilities

155 be deemed economical to repair and should be repaired as soon as possible but no later than 45 days: PBP = Cost of Control Annual Leak Rate Gas Price Where, PBP = payback period (years). Cost of Control = direct repair or replacement costs + gas vented during repair + cost of lost production due to shutdown Annual Leak Rate = amount of gas/vapour emitted directly to the atmosphere or that leaked into a vent or flare system which does not have vent or flare gas recovery. Gas Price = current market price of the gas based on criteria specified by the EUB Directive or, for the midstream industry, the processing fee or margin received. Components that have a payback period greater than 1 year should be reevaluated with the Cost of Control equal to direct repair or replacement cost only. If this analysis shows the repair or replacement is economic to do during a major shut down, it should then be scheduled for repair at the next shut down. Where the payback period is greater than the anticipated life expectancy of the repair or replacement, the component may be deemed uneconomic to repair or replace and supporting details of this cost evaluation shall be kept on file. Table 3 is provided as an indication of possible mean default life expectancies of component repairs that an operator may choose to use in the absence of official data on the life expectancy of the affected component. Table 3. Default mean life of repairs for economic analysis of repair costs. Source Category Mean Repair Life (years) Compressors - Reciprocating Seals 1 Valve Covers 1 Variable Volume 1 Pocket Governor 1 Cylinder Head 1 Compressors Centrifugal Seals 1 Connectors All 5 5 The commodity price forecasts used in evaluations of conventional gas conservation projects (gas gathered, processed, and sold to market) will be the most recently published by Dobson Resource Management. January 2007 Management of Fugitive Emissions 10 At Upstream Oil and Gas Facilities

156 Table 3. Default mean life of repairs for economic analysis of repair costs. Source Category Mean Repair Life (years) Open-Ended Lines All 2 Pressure Relief Valves All 2 Pumps Seals 1 Regulators All 5 Tank Fittings Hatches 1 Pressure Vacuum 2 Valves Valves Quarter-Turn 4 Rising Stem 2 Vents All 1 Sample calculations for payback periods are presented in Appendix IV Personnel Training Proper personnel training should be part of the DI&M program. This training is needed to ensure that the program achieves the best results Primary Calibrations and Field checks All instruments used to detect and measure leaks should be factory serviced or serviced by a factory authorized technician and should be calibrated regularly as per the specification of the manufacturer or whenever problems arise. January 2007 Management of Fugitive Emissions 11 At Upstream Oil and Gas Facilities

157 4 Inspection, Monitoring & Record-Keeping Operators should have a record program to support the company s DI&M system. Proper record keeping should assist in ensuring that leaking components are identified and repaired and that appropriate follow-up actions are implemented. This information will also assist in identifying the proper monitoring frequency for that component to achieve maximum cost-effective fugitive emissions reductions while suiting the unique characteristics and operations of the facility. Although it remains for each company to define its record keeping system, consideration should be given to the recording of the following information: Records of repairs made on leaking components, including leak repair frequency. The economic analysis performed on all leaking equipment components that have not been fixed on the basis that this is uneconomic to do and do not pose any safety, health, or environmental concerns. Record keeping in support of a company s DI&M program may be audited by the EUB to assess compliance with section 8.7 of the EUB Directive 060. January 2007 Management of Fugitive Emissions 12 At Upstream Oil and Gas Facilities

158 5 Corporate Commitment Corporate commitment should entail full management support including adequate funding and resource allocation. Components that are initially tight may leak, and leaks, once fixed, may reoccur. Consequently, the reduction of fugitive emissions requires a dedicated ongoing commitment. January 2007 Management of Fugitive Emissions 13 At Upstream Oil and Gas Facilities

159 6 Appendices The information provided in the following appendices is included as guidance only. Operators may elect to implement other approaches to develop their DI&M program. January 2007 Management of Fugitive Emissions 14 At Upstream Oil and Gas Facilities

160 7 References Cited Aikin, J.A Valve Packing and Live-Loading Improvements. Valve Magazine. Valve Manufacturers Association of America. Washington, D.C. v4, n3. pp , 52. American Industrial Hygiene Association Odor Thresholds for Chemicals With Established Occupational Health Standards. Stock No. 108-EA-89. Fairfax, Virginia. American Petroleum Institute Standard 610: Centrifugal Pumps for General Refinery Service. 7 th Edition. Washington, D.C. Order No pp American National Standards Institute Standard B-16.5: Piping Flanges and Flange Fittings. New York, NY. pp American National Standards Institute Standard B-31.3: Chemical Plant and Petroleum Refinery Piping. New York, NY. pp American National Standards Institute Standard B-73.1M: Specification for Horizontal End Suction Centrifugal Pumps for Chemical Process. New York, NY. Order No. J American National Standards Institute Standard B-73.2M: Specification for Vertical In- Line Centrifugal Pumps for Chemical Process. New York, NY. Order No. J American Society of Mechanical Engineers ASME Boiler and Pressure Vessel Code: An American National Standard. Section VIII - Rules for Construction of Pressure Vessels, Division 1. New York, NY. Battilana, R.E Better Seals Will Boost Pump Performance. Chemical Engineering Progress. v85, n7. pp (July, 1989) Brestel, R., W. Hutchens, and C. Wood Technical Monograph 38: Control Valve Packing Systems. Fisher Controls International, Inc. Marshalltown, Iowa. pp. 22. Lipton, S Reduce Fugitive Emissions Through Improved Process Equipment. Chemical Engineering Progress. v88, n10. pp (Oct., 1992) The Canadian Council of Ministers of the Environment Environmental Code of Practice for the Measurement and Control of Fugitive VOC Emissions from Equipment Leaks. Prepared by the National Task Force on The Measurement and Control of Fugitive VOC Emissions from Equipment Leaks. Ottawa, ON. pp. 35. U.S. Environmental Protection Agency Protocols for Generating Unit-specific Emission Estimates for Equipment Leaks of VOC and VHAP. Research Triangle Park, NC. Report No. EPA-450/ U.S. Environmental Protection Agency VOC Fugitive Emissions in Synthetic Organic Chemicals Manufacturing Industry - Background Information for Proposed Standards. Report No. EPA-450/ a. Table 4-7. Page January 2007 Management of Fugitive Emissions 15 At Upstream Oil and Gas Facilities

161 U.S. Environmental Protection Agency Lessons Learned from natural Gas Star Partners, Directed Inspection and Maintenance at gas processing plants and booster stations. Wright, J.B Avoid Valve Leaks. Chemical Engineering Progress. v89, n6. pp (June, 1993). January 2007 Management of Fugitive Emissions 16 At Upstream Oil and Gas Facilities

162 Appendix 1 Table 4. Examples Of Leak Monitoring Frequencies For Leak-Prone Equipment Components Table 4. Source Category Process Equipment Examples of leak monitoring frequencies for leak-prone equipment components, presented by component category and type Type of Component Service (sweet gas, light liquid) Frequency Control Valves Gas/Vapour/LPG Annually Block Valves Gas/Vapour/LPG Annually Rising Stem Block Valves Gas/Vapour/LPG Once every 5 years Quarter Turn Compressor Seals 1 All Quarterly Pump Seals 1 All Quarterly Pressure Relief All Annually Valves Open-ended Lines All Annually Emergency Vent 1,2 All Annually Vapour Collection Systems Blowdown Systems 1,2 All Quarterly Tank Hatches 1 All Quarterly Pressure-Vacuum All Quarterly Safety Valves 1 1 Alternatively, institute a predictive maintenance program to monitor seals performance 2 Emergency vents and blowdown systems should be screened during periods when relief or blowdown events are not occurring to determine the amount of leakage into these systems. January 2007 Management of Fugitive Emissions 17 At Upstream Oil and Gas Facilities

163 Appendix 2 Sample Leaker Tag Leak repaired c Date January 2007 Management of Fugitive Emissions 18 At Upstream Oil and Gas Facilities

164 Appendix 3 Appendix III Leak Survey Forms LEAK RATE MEASUREMENTS Site Name: Operating Company: Location: Survey Contractor: Page of Date Technicians: Process Unit Name/Identification Code Type Leaker Tag No. Process Tag No. Component Information Type Process Stream Size Measurement Type Odourized Method Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N Y/N THC Leak Rate (m 3 /h) January 2007 Management of Fugitive Emissions 19 At Upstream Oil and Gas Facilities

165 Direct Measure Datasheet Location: Date: Tag No. Source Data Type of Compone nt Size Process Unit Meter Type Measurement Data Initial Volume Final Volume Time Temp (C) Pressure (kpa) Comments CV Control Valve NV Needle Valve BV Ball Valve GBV Globe Valve GTV Gate Valve PV Plug Valve BFV Butterfly Valve MW Manway PRV Pressure Relief Valve O Open-Ended Line PR Pressure Regulator R - Regulator GOV Govenor PIG Pig Trap Cover FC Filter Cover VC Valve Cap C - Coupling F Flange T Threaded Fitting TB Tube Fitting PS Pump Seal CS Compressor Seal SW/PG Sweet Process Gas SR/PG Sour Process Gas FG Fuel Gas S Sales Gas P Propane C2 - Ethane C Condensate MP Multipahse O Oil CO Crude Oil AG Acid Gas January 2007 Management of Fugitive Emissions 20 At Upstream Oil and Gas Facilities

166 Site Name: Operating Company: Location: Industry Sector: Facility Type: GAS SENSOR CALIBRATION RECORD Page of Date Technicians: Device Manufacturer Serial No. Date of Last Field Calibration Checks Calibration Day Time Zero Check Span Check Units of Measure Factory Office Reading Actual Reading %Error January 2007 Management of Fugitive Emissions 21 At Upstream Oil and Gas Facilities

167 Site Name: Operating Company: Location: Industry Sector: Facility Type: FLOW METER CALIBRATION RECORD Page of Date Technicians: Device Manufacturer Serial No. Date of Last Field Calibration Checks Calibration Day Time Zero Check Span Check Factory Laboratory Reading Refere Reading nce %Error Units of Measure January 2007 Management of Fugitive Emissions 22 At Upstream Oil and Gas Facilities

168 Appendix 4 Economic Analysis Table 5 presents example calculations for determining the simple payback period of individual leak repairs using the equation presented in Section As indicated in the table, the simple payback period is calculated as the estimated repair costs (column C) divided by the product of the leak rate (column A) and the net gas value (column B). For the examples given, the net gas price is taken to be $4.0/GJ or $0.1496/m 3, which, in this case, is the market value of the gas (i.e., determined from the Chenery-Dobson Resource Management Ltd. Survey of Hydrocarbon Price Forecasts Utilized by Canadian Petroleum Consultants and Canadian Banks). The effect of discount rates and inflation rates are neglected for simplification purposes. January 2007 Management of Fugitive Emissions 23 At Upstream Oil and Gas Facilities

169 Table 5. Sample calculations of the simple payback period for individual leak repairs. Tag ID Process Unit / Location Component Type Nominal Size (Inches) Stream Type Hydrocarbon Leak Rate (m 3 /hr) Hydrocarbon Leak Rate (10 3 m 3 /y) Net Value of Lost Gas ($/10 3 m 3 ) Estimated Repair Cost ($) Default Repair Life (y) Payback Period (y) (A) (B) (C) (C/A/B) 1982 Desiccant Dehydrator Open Ended 0.5 Regen Line Gas 1985 Desiccant Dehydrator Regulator 0.5 Dry Gas Desiccant Dehydrator Plug Valve 12 Dry Gas Desiccant Dehydrator Gate Valve 0.75 Dry Gas Desiccant Dehydrator Open Ended 0.5 Dry Gas Line 1990 Desiccant Dehydrator Flange 8 Wet Gas Desiccant Dehydrator Gate Valve 8 Wet Gas Desiccant Dehydrator Flange 8 Wet Gas Desiccant Dehydrator Ball Valve 0.5 Wet Gas Desiccant Dehydrator Gate Valve 8 Wet Gas (Control Valve) 1996 Desiccant Dehydrator Gate Valve 8 Wet Gas (Control Valve) 1997 Desiccant Dehydrator Flange 8 Wet Gas Desiccant Dehydrator 3-Way Control 8 Wet Gas Valve 4087 Desiccant Dehydrator Threaded Connection 1 Wet Gas January 2007 Management of Fugitive Emissions 24 At Upstream Oil and Gas Facilities

170 Appendix 5 Emissions Inventory From Method 21 Data The leak screening data compiled using US EPA Method 21 (i.e., measurements of the vapour concentration at the leakage point on each component) may be used to estimate total fugitive emissions from the site for the purposes of developing an emissions inventory. These data should not to be used for quantifying the emissions from a single component for the purposes of conducting an evaluation of the economics of repairing or replacing the component since they do not provide sufficient accuracy for this purpose. The use of Method 21 data to calculate the emission rate from a single component may easily be in error by 2 to 3 orders of magnitude, while economic justification for not repairing a component should be based on a leak rate measurement accurate to ±25 percent. For component quantification purposes see section Emissions may be estimated from US EPA Method 21 data by applying one of the following methods, presented in the order of increasing sophistication and accuracy: leak/no-leak emission factors, three-stratum emission factors, published leak-rate correlations, and unit-specific leak-rate correlations. There is relatively little difference in the effort required to apply each approach once the screening data have been compiled. Consequently, it may be preferable to use the correlation approach since it gives the most reliable results. Simple spreadsheet or database application may be developed for this purpose. 5.1 Leak/No-leak Emission Factors. To apply this approach, the screening values must be classified as either leaking (i.e., has a maximum screening value of ppm or more) or non-leaking (i.e., has a maximum screening value of less than ppm), and categorized by type of component and type of service. The amount of emissions is then estimated for each source category using the equation: where, ER= i j N i, j EF L nl+ EF nl + nn nn X j ER = total methane leak rate (kg/h) of pollutant k for the target source population, EF L = appropriate leaking emission factor for the source/service category of interest (see Table 6), EF N = appropriate non-leaking emission factor for the components of type i in service j (see Table 6), N i, j January 2007 Management of Fugitive Emissions 25 At Upstream Oil and Gas Facilities

171 nl = number of components screened and determined to be leaking (i.e., give a screening value of ppm or more) for the source category of interest, nn = number of components screened and determined not to be leaking (i.e., give a screening value of less than ppm, including those sources with a screening value of zero) for the source category of interest, N i, j = total number of components of type i in service j (i.e., all components screened plus those not screened), X j = mass fraction of methane in the process stream It is assumed, in using the leak/no-leak method, that components within each screening range leak, on average, at the same as rate as for the rest of the UOG. However, the experience in other industries is this not is not necessarily true (Schaich and Stine, 1989). For example, significant differences have been noted in the relative number of sources with zero screening values. A zero screening value indicates that the true screening value of the source is below the lower detectable limit of the vapour analyzer used to screen for leaks. Table 6. Source Leak/No-leak emission factors for estimating fugitive equipment leaks at UOG facilities. Number of Sources Percent of Sources Emissions (kg/h/src) 95% Confidence Limits Lower Upper Connector 1 No Leak Leak Block Valve 2 No Leak Leak Control Valve 3 No Leak Leak PRV No Leak Leak Regulator No Leak Leak Orifice Meter 4 No Leak Leak Other Flow Meter 5 No Leak Leak Station or No Leak Pressurized Compressor Leak Blowdown System 6 January 2007 Management of Fugitive Emissions 26 At Upstream Oil and Gas Facilities

172 Table 6. Source Leak/No-leak emission factors for estimating fugitive equipment leaks at UOG facilities. Number of Sources Percent of Sources Emissions (kg/h/src) 95% Confidence Limits Lower Upper Compressor No Leak Blowdown System - Depressurized Leak Reciprocating Compressor No Leak Blowdown System - Depressurized Centrifugal Leak Open-Ended Line No Leak Leak Instrument No Leak Controller 7 Leak Compressor Seal No Leak Reciprocating 8 Leak Compressor Seal - No Leak Centrifugal 8 Leak Source: Ross and Picard (1996), Table 6, page No data available. 1 Includes flanges, threaded connections and mechanical couplings. 2 Accounts for emissions from the stem packing and the valve body, and it applies to all types of block valves (e.g., butterfly, ball, globe, gate, needle, orbit and plug valves). Leakage past the valve seat is accounted for by the Open-Ended Line emission category. Leakage from the end connections is accounted for by the connector category (i.e., one connector for each end). 3 Accounts for leakage from the stem packing and the valve body. Emissions from the controller and actuator are accounted for by the Instrument Controller and Open-Ended Line categories respectively. This factor applies to all valves with automatic actuators (including fuel gas injection valves on the drivers of reciprocating compressors). 4 Accounts for emissions from the orifice changer. Emissions from sources on pressure tap lines etc. are not included in the factor (i.e., these emissions must be calculated separately). 5 Accounts for emissions from other types of gas flow meters (e.g., diaphragm, ultrasonic, roots, turbine and vortex meters). 6 Accounts for leakage past a valve seat through an open vent line to the atmosphere. These vents are typically six inches or greater in diameter and are used to blowdown major process units or sections of pipeline. Small diameter open-ended lines such as those used to blowdown chart recorders, meter runs etc. are accounted for by the Open-Ended Line category. 7 The Instrument Controller category accounts for emission from pneumatic control devices that use natural gas as the supply medium. 8 The Compressor Seal categories account for emissions from individual compressor seals (i.e., for a four cylinder reciprocating compressor unit there are four seals so the compressor seal emissions for the unit would be four times the factor in the table). 9 Non-leaking components with screening values of less than ppm. 10 Leaking components with screening values of ppm or greater. January 2007 Management of Fugitive Emissions 27 At Upstream Oil and Gas Facilities

173 5.2 Three-stratum Emission Factors Use of the three-stratum factors offers a further increase in rigour and reliability from use of the leak/no-leak factors. The sources are categorized based on three ranges of screening values: 0 to ppm, to ppm and over ppm. The amount of emissions is estimated for each source category using the equation: ER= i j N i, j n1 EF 1+ n2 EF 2 + n3 EF 3 n1 + n2 + n3 where, EF 1, EF 2, EF 3 = THC emission factors for sources with screening values in the range of 0 to ppm, to ppm, and over ppm, respectively (see Table 7). n 1, n 2, n 3 = Total number of sources surveyed that had screening values in the range of 0 to ppm, to ppm, and over ppm, respectively. The basic assumptions inherent in use of the three-stratum emission factor method are the same as those presented for the leak/no-leak method. 5.3 Published Leak-Rate Correlations Leak-rate correlations provide a method for estimating the leak rates corresponding to individual screening values. The use of this approach is a considerable refinement over the available emission-factor methods in which constants are applied over discrete ranges of screening values. The correlations are given by a two-constant relation of the form given below: where: Log (ER) = B 0 + B 1 Log(SV) (5) B 0, B 1 = Model parameters as given in Table 8. ER = Leak rate in (kg/h/source). SV = Maximum screening value above background measured using a detector calibrated to methane (ppm). January 2007 Management of Fugitive Emissions 28 At Upstream Oil and Gas Facilities

174 Table 7. Source Three-stratum emission factors for estimating fugitive equipment leaks at UOG facilities. Number of Sources Percent of Sources Emissions (kg/h/src) 95% Confidence Limits Lower Upper Connector 1 < > Block Valve 2 < > Control Valve 3 < > PRV < > Regulator < > Orifice Meter 4 < > Other Flow Meter 5 < > Station or Pressurized < Compressor Blowdown System > Compressor < Blowdown System - Depressurized Reciprocating > Compressor < Blowdown System - Depressurized Centrifugal > Open-Ended Line < > Instrument Controller 7 < > January 2007 Management of Fugitive Emissions 29 At Upstream Oil and Gas Facilities

175 Table 7. Source Three-stratum emission factors for estimating fugitive equipment leaks at UOG facilities. Number of Sources Percent of Sources Emissions (kg/h/src) 95% Confidence Limits Lower Upper Compressor Seal - < Reciprocating > Compressor Seal - < Centrifugal > Source: Ross and Picard (1996), Table 7, page No data available. 1 Includes flanges, threaded connections and mechanical couplings. 2 Accounts for emissions from the stem packing and the valve body, and it applies to all types of block valves (e.g., butterfly, ball, globe, gate, needle, orbit and plug valves). Leakage past the valve seat is accounted for by the Open-Ended Line emission category. Leakage from the end connections is accounted for by the connector category (i.e., one connector for each end). 3 Accounts for leakage from the stem packing and valve body. Emissions from the controller and actuator are accounted for by the Instrument Controller and Open-Ended Line categories respectively. This factor applies to all valves with automatic actuators (including fuel gas injection valves on the drivers of reciprocating compressors). 4 Accounts for emissions from the orifice changer. Emissions from sources on pressure tap lines etc. are not included in the factor (i.e., these emissions must be calculated separately). 5 Accounts for emissions from other types of gas flow meters (e.g., diaphragm, ultrasonic, roots, turbine and vortex meters). 6 Accounts for leakage past a valve seat through an open vent line to the atmosphere. These vents are typically six inches or greater in diameter and are used to blowdown major process units or sections of pipeline. Small diameter open-ended lines such as those used to blowdown chart recorders, meter runs etc. are accounted for by the Open-Ended Line category. 7 The Instrument Controller Category accounts for emission from pneumatic control devices that use natural gas as the supply medium. 8 The Compressor Seal categories account for emissions from individual compressor seals (i.e., for a four cylinder reciprocating compressor unit there are four seals so the compressor seal emissions for the unit would be four times the factor in the table). Table 8. Correlation parameters for estimating leak rates from equipment components at UOG facilities. Source B 0 B 1 Number of Sources Correlation (R 2 ) Connectors Valves Open-Ended Lines Pressure Relief Devices Pressure Regulators The correlation for this source is based on screening and bagging data collected by Ross and Picard (1996), by Environment Canada (Williams, 1996), and data collected for U.S. EPA (1995). 2 The correlation for this source is based on screening and bagging data collected by Ross and Picard (1996) and Environment Canada (Williams, 1996). January 2007 Management of Fugitive Emissions 30 At Upstream Oil and Gas Facilities

176 The values of the correlation constants for application to the Canadian UOG industry are summarized in Table 8. The basic approach involves processing each individual screening value as follows and then aggregating the results to determine total emissions: On-scale Screening Values - Components that have screening values within the detection range of the screening instrument and more than 1 ppm above background are assessed using equation (5). Do not average screening values and then enter the result into the correlation to estimate emissions. Zero Screening Values - Components with screening values of 1 ppm or less above background are assigned a default emission rate in accordance with Table 9. If the instrument has a minimum detection limit greater than 1 ppm, the default zero values in Table 8 do not apply. In this situation, all zero screening values should set to one-half the instrument's minimum detection limit and then processed using equation (5). Off-scale or Pegged Values - Components with off-scale or pegged values are either assigned the appropriate average emission rate for a leaking equipment component (i.e., see Table 6 or 7), or are bagged to determine the actual mass emission rate. The level of uncertainty in the total emissions estimated by this approach is a function of the number of components considered and the portion that are pegged sources. For a single component, the uncertainty may be as high as a factor of 100. The uncertainty tends to decrease almost exponentially with the number of components. A facility would need to have at least 3000 components before the uncertainty would be below 50 percent for a 90 percent confidence level. Even more components would be required if there are pegged sources. However, very few facilities will have even 3000 components. The best overall accuracy for a facility that, generally, can be expected from use of the correlation method, including use of pegged emission factors, is about ±300 percent (telephone communication with Mr. R.A. Lott at Gas Research Institute, June 19, 1999). January 2007 Management of Fugitive Emissions 31 At Upstream Oil and Gas Facilities

177 Table 9. Default zero emission rates 1. Source Type Default-Zero Emission Rate 2 (kg/h/source) Gas valve 6.6 (10-7 ) Light liquid valve 4.9 (10-7 ) Light liquid pump (10-6 ) Connector 6.1 (10-7 ) 1 Source: U.S. EPA Protocol for Equipment Leak Emission Estimates. Research Triangle Park, NC. Report No. EPA-453/R Table p Total hydrocarbon emissions. 3 The light liquid pump default zero value can be applied to compressors seals, pressure relief valves, pressure regulators, agitators and heavy liquid pumps. 5.4 Unit-Specific Leak-Rate Correlations Some companies may wish to develop unit-specific leak-rate correlations to achieve better accuracy for their particular operations. The benefit in doing this is often questionable since, in theory, the relationship between a properly corrected screening value and the emission rate for a given component type and service is the same across all industries. Nonetheless, where there is some doubt, various nonparametric statistical methods may be applied to determine the validity of the available correlations. This involves comparing a set of predicted emission rates to actual measured values for a given source type and service category. The sign or a suitable rank-sum test may be applied to determine if the two data sets are statistically different (see almost any statistical analysis text). To apply the sign test, try using a sample size of four. If all four measured values are consistently greater than or consistently less than the predicted values, then the selected correlation probably does not provide reasonable emission estimates for the given application. To develop a leak-rate correlation it is necessary to compile a reasonable number of data points to cover the desired screening range for each target source/service category. Each data point must comprise an actual measured mass emission rate and corresponding screening value. U.S. EPA (1995) suggests using a minimum of six random data points for each of the following ranges that the correlation will span: 1 to 100 ppm, 101 to ppm, to ppm, to ppm and to ppm. The collected data are fit to equation (5) using a least-squares regression analysis. For further details on developing a unitspecific leak-rate correlations, refer U.S. EPA's (1995) protocol for equipment leak emission estimates. January 2007 Management of Fugitive Emissions 32 At Upstream Oil and Gas Facilities

178 Furthermore, use of a portable organic vapour analyzer is much slower than other leak detection techniques; especially if screening data are being recorded for each component when an analyzer is used, and other methods can eliminate this need. For example, use of bubble tests with no data recording, just tagging of leaking components, is at least 3 to 4 times faster than application of US EPA Method 21 using an organic vapour analyzer. January 2007 Management of Fugitive Emissions 33 At Upstream Oil and Gas Facilities

179 Appendix 6 Component-Specific Control Options This appendix presents potential options for eliminating or controlling chronic leaks for each of the following common types of equipment components, respectively: Reciprocating compressors Centrifugal compressors Valve stem packing systems Sewers and drains Pump seals Flanged and threaded connections Pressure relief devices Open-ended valves and lines Sampling points 6.1 Reciprocating Compressors Packings are used on reciprocating compressors to control leakage around the piston rod on each cylinder. A schematic diagram of a conventional packing system is presented in Figure 2. Typically, the distance piece is either left open with the vent piping connected directly to the packing case, or the distance piece is closed and the vents may be connected to both the packing case and the distance piece. The packing and distance piece vents are commonly routed outside the building to the atmosphere if the process gas is sweet, but should be connected to an emission controlling vent system if the gas is sour. The latter approach provides continuous treatment of any emissions and allows for more convenient scheduling of any required maintenance to the packing system Vent Monitoring Systems It is good practice to install instrumentation on the vent lines to indicate excessive vent rates and the need for maintenance. A sensitive rotameter, an orifice and pressure differential indicator providing flow indication, or a temperature element may be used depending on the application Emission-Controlling Vent Systems Where emission-controlling vent systems are employed they should be designed to minimize the potential for either the flow of process gas through the distance piece into the compressor crank case, or air ingress to the vent system through the nose of the packing case or through the air breather on the crank case and past the wiper packing leading to the distance piece (depending on the location of the vent connections). Both conditions pose a potential explosion hazard. Additionally, the leakage of process gas into the crank case could possibly result in contamination of the lubricating oil or corrosion problems (especially if the process gas contains hydrogen sulphide). January 2007 Management of Fugitive Emissions 34 At Upstream Oil and Gas Facilities

180 Figure 2. Schematic diagram of a piston-rod packing-case system on a reciprocating compressor. January 2007 Management of Fugitive Emissions 35 At Upstream Oil and Gas Facilities