Significance of a CALPUFF Near-Field Analysis

Size: px
Start display at page:

Download "Significance of a CALPUFF Near-Field Analysis"

Transcription

1 Significance of a CALPUFF Near-Field Analysis Paper #70052 Authors: Elizabeth A. H. Carper Trinity Consultants nd Avenue South, Suite 610 Kent, WA Eri Ottersburg Trinity Consultants nd Avenue South, Suite 610 Kent, WA Co-Sponsor: Steve Duvall Longview Fibre Company 300 Fibre Way Longview, WA A&WMA's 96th Annual Conference and Exhibition June 2003 ABSTRACT Historically CALPUFF air dispersion modeling has been a useful tool for determining ambient concentrations, regional haze, and deposition in Class I areas for Prevention of Significant Deterioration (PSD) applications. Until recently, CALPUFF has not been utilized by minor sources for small permitting projects or for near-field regulatory analyses due to the costs encountered from the complexity of the model and the extensive computer resources it requires. However, as shown in this case study, CALPUFF dispersion modeling for near-field analyses can provide significant cost savings and additional permit flexibility, even for minor sources of emissions. As the capabilities of computers grow, CALPUFF will most-likely play an increasingly important role in air permitting and regulatory review of near field analyses, even in minor source reviews. This case study provides a dramatic comparison of a variety of air dispersion models including ISCST3, CALPUFF-Lite, CALPUFF, and CALPUFF with Complex Terrain Sub-Grid (CTSG) receptors. Furthermore, the results of this case conclude that CALPUFF s ability to calculate chemical transformation and deposition as well to incorporate terrain affects on plume dispersion allow for more realistic dispersion representation and reduce overestimation of dispersion impacts. 1

2 INTRODUCTION This case study presents the application of CALPUFF A to near-field modeling of deposition impacts on a Class I area. The modeling was completed as part of a Notice of Construction (NOC) permit application for the Longview Fibre Company s (Longview s) Leavenworth Mill. Although, the Leavenworth Mill is a minor source (i.e. facility-wide emissions less than 100 tons per year (tpy) of all criteria pollutants), modeling was required by the regulatory authority to predict deposition impacts on the nearby Class I area, Alpine Lakes Wilderness Area (Alpine Lakes) because its nearest boundary is only 3.5 km away from the source. This modeling study shows that CALPUFF produces the most accurate and representative prediction of deposition impacts as compared to ISCST3 and CALPUFF-Lite for near-field transport. B Furthermore, CALPUFF s advanced modeling abilities may be beneficial for minor sources with small emissions. BACKGROUND Class I Areas Class I areas, which include certain National wilderness areas, National memorial parks, and National parks, are afforded the highest level of protection under the Clean Air Act. Ambient air requirements that apply within Class I areas are more stringent than those that apply to other areas (i.e., Class II areas). In addition to more stringent ambient air requirements, Class I areas are also protected by the regulation of Air Quality Related Values (AQRVs) within their borders. For example, under the PSD regulations, Federal Land Managers (FLMs), who are those individuals responsible for the management of Class I areas, are charged with an affirmative responsibility to protect the air quality related values of such lands and to consider whether a proposed source or modification will have an adverse impact on such values. [1] FLMs typically use two quantifiable impacts to assess whether or not a facility s emissions will have adverse impacts on Class I areas. These two quantifiable impacts are the degradation of visibility and the deposition of acidic species (e.g., nitrogen and sulfur). Only deposition is considered in this case study. The fundamental methods and criteria for determining and interpreting impacts to Class I areas are set forth in several United States Environmental Protection Agency (EPA) and FLM documents, including the Interagency Workgroup of Air Quality Modeling (IWAQM) Phase I report, the IWAQM Phase II report, the FLM s Air Quality Related Values Workgroup (FLAG) Phase I report, and guidance from the National Park Service and the U.S. Forest Service. [2,3,4] This analysis follows such guidance, as well as guidance from the Washington Department of Ecology. A Class I area analysis is performed using an appropriate dispersion model depending on the distance between the source and the Class I area of interest. For sources more than 50 km away from the Class I areas of interest, the CALPUFF modeling system, a long-range transport model, A CALPUFF is available from Earth Tech, B ISCST3 is available from EPA s Support Center for Regulatory Air Models, 2

3 is traditionally recommended. Sources located less than 50 km away from the Class I area customarily use short-range transport models such as ISCST3, AERMOD, or CTDM. Deposition and Chemical Transformation Acid deposition is caused by secondary pollutants that form from the oxidation of nitrogen oxides (NO x ) or sulfur dioxide (SO 2 ) gases. When oxidized forms of nitrogen and sulfur come into contact with water, they become acidic and produce nitric acid (HNO 3 ) and sulfuric acid (H 2 SO 4 ). Deposition of nitrogen and sulfur causes acidification of waters and soils, also known as acid rain. Deposition is measured as mass flux of nitrogen or sulfur atoms (e.g., kilograms per hectare per year (kg/ha/yr) or grams per meter squared per second (g/m 2 /s)). In deposition analyses for PSD applications, the source s contribution to the deposition of acidic chemical species in Class I areas is evaluated against the Deposition Analysis Threshold (DAT) set by the FLMs. The DAT is kg/ha/yr for both sulfur and nitrogen deposition. [5] The proposed project that this case study is based on was not subject to PSD review, therefore, thresholds were set by the local agency at kg/ha/yr for sulfur and 0.01 kg/ha/yr for nitrogen deposition, which are based on 0.2 percent of the 5 kg/hectare/year nitrogen and 3 kg/hectare/year sulfur concern thresholds. IWAQM calculation The ISCST3, AERMOD, and CTDM models are not able to compute deposition or chemical transformation. C,D The IWAQM Phase I report provides some recommendations for calculating and estimating deposition and chemical transformation. [2] These calculations are applied to concentration outputs from the model. This approach is considered a screening analysis and includes the following conservative assumptions: 1. No conversion of SO 2 or NO x into other species. 2. All concentrations of SO 2 and NO x are deposited as SO 2 and HNO 3. These assumptions are conservative because natural atmospheric reactions will convert SO 2 and NO x into other species, whose deposition rates vary. Since SO 2 and HNO 3 are deposited faster than any of the other sulfur or nitrogen species, it is conservative to assume that all pollutants are deposited at the rate of SO 2 and HNO 3. Deposition is calculated as an annual average based on the modeled concentration of SO 2 and NO x. Therefore, the modeled concentration is multiplied by the number of seconds in a year ( x 10 7 ). This result is then multiplied by the respective deposition velocity for each pollutant (0.005 m/s for SO 2 and 0.05 m/s for HNO 3 ). Since the value of concern is the mass flux of nitrogen atoms, the result must be multiplied by the ratio of the molecular weight (m.w.) of the secondary species to the m.w. of the primary species. Finally, the deposition value is converted to kilograms per hectare per year by multiplying by C ISCST3 in regulatory mode has the ability to model dry deposition of particles, but not deposition of gases. D AERMOD and CTDM are available from EPA s Support Center for Regulatory Air Models, 3

4 Equation 1. For nitrogen deposition the equation looks like: 14 ( m. w. of N ) 46 ( m. w. of NO ) [ NO ]( µ g m ) ( s yr ) 0.05( m s) 10 ( kg ha) Nitrogen Deposition( kg ha yr ) x = 2 Equation 2. For sulfur deposition: 32 ( m. w. of S) µ [2] 64 ( m. w. of SO ) [ SO ]( g m ) ( s yr ) 0.005( m s) 10 ( kg ha) Sulfur Deposition( kg ha yr ) 2 = 2 These calculations only represent dry deposition of species. Typically it is assumed that dry deposition equals wet deposition and the results of the above calculations are doubled to predict total deposition. Calculation of deposition using this method leads to very conservative and unrepresentative results. It is not realistic to assume all SO 2 and NO x emitted from the stack is deposited as SO 2 and HNO 3. However, when modeling concentrations with ISCST3, AERMOD, or CTDM, this is the only method recommended by IWAQM and FLAG guidance. CALPUFF Calculation CALPUFF has the ability to model chemical transformation, dry deposition, and wet deposition. For this case study, the MESOPUFF II chemical transformation scheme is used to simulate linear chemical transformation effects by using pseudo-first-order chemical reaction mechanisms. The MESOPUFF II chemical transformation algorithm accounts for the conversion of SO 2 to sulfate (SO 4-2 ) and NO x, which includes nitric oxide (NO) and nitrogen dioxide (NO 2 ), to nitrate (NO 3-1 ) and HNO 3. In further atmospheric reactions, SO 4-2 and NO 3-1 form the compounds ammonium nitrate (NH 4 NO 3 ) and ammonium sulfate ((NH 4 ) 2 SO 4 ) in the presence of ammonia (NH 3 ). In the MESOPUFF II algorithms, ambient ozone (O 3 ) is required for certain oxidation reaction pathways and ambient NH 3 dictates equilibrium between NH 4 NO 3 and HNO 3 and (NH 4 ) 2 SO 4 and H 2 SO 4. Consequently, the MESOPUFF II scheme requires background concentrations of O 3 and NH 3 to simulate these chemical processes. These processes are represented by the following reactions: [2] SO x SO 4-2 (NH4 ) 2 SO 4 NO x NO 3-1 NH4 NO 3 Within CALPUFF, the computation of dry and wet deposition rates of gases and particulate matter are a function of geophysical parameters, meteorological conditions, and pollutant species. A full resistance model is provided in CALPUFF for the computation of dry deposition rates of gases and particulate matter. An empirical scavenging coefficient approach is used in CALPUFF to compute the depletion and wet deposition fluxes due to precipitation scavenging. Gas-phase dry deposition fluxes are modeled for SO 2, NO X, and HNO 3. Particulate-phase dry deposition fluxes are modeled for SO 4, NO 3, and PM 10. Wet deposition is modeled for SO 2, SO 4, HNO 3, and NO 3. 4

5 Equation 3. The sum of wet and dry deposition fluxes for SO 2 and SO 4 represents the total sulfur deposition as follows: ( flux SO + flux SO ) + ( flux SO flux SO ) 2 4 wet 2 4 dry Sulfur Deposition (kg/ha/yr) = + The NO 3 and SO 4 will combine with ammonia (NH 3 ) to form (NH 4 ) 2 SO 4 and NH 4 NO 3. The nitrogen molecules from NH 4 NO 3 and both nitrogen molecules from (NH 4 ) 2 SO 4 are added to the total nitrogen deposition. Equation 4. The sum of wet and dry deposition fluxes for NO X, NO 3, HNO 3, SO 4 represent the total nitrogen deposition as follows: ( flux ) ( NOx NO3 HNO3 SO ) NOx + 2fluxNO + fluxhno + 2flux 3 3 SO + flux + 2flux + flux 2flux 4 wet 4 dry Nitrogen Deposition( kg/ha/yr) = + MODEL COMPARISON In regulatory applications for a near-field modeling analysis ISCST3 or AERMOD are typically used to evaluate ambient concentrations of various pollutants. ISCST3 is EPA s approved regulatory model for near-field analyses. [6] AERMOD is increasingly accepted or required by state and federal agencies in regulatory review determinations such as PSD and may soon be approved as the regulatory model in place of ISCST3. The Complex Terrain Dispersion Model (CTDMPLUS or CTSCREEN) is the recommended model for sources located in complex terrain. E These models have relatively simple inputs, meteorology, and dispersion algorithms. Due to their simplicity the models require few computer resources and less run time than more complex models. Although these models are adequate when modeling dispersion of pollutants, they have limitations in assessing deposition impacts. No algorithms are included to account for the chemical transformation of pollutants in the atmosphere. The transformation of SO 2 to SO 4 and NO x to NO 3 and HNO 3 is of particular interest for modeling deposition of acidic species, as discussed above. Furthermore, these models do not include the necessary algorithms to calculate wet and dry deposition. F Deposition is calculated from the model output using conservative equations in accordance with the IWAQM Phase I guidance document. [2] The CALPUFF model is most widely used for Class I area deposition impact analyses. It is the recommended model in the IWAQM Phase II and FLAG guidance for long-range transport. [3,4] CALPUFF is also likely to be approved by the EPA as a regulatory model for long-range transport modeling analyses. CALPUFF is a multi-layer, multi-species, non-steady-state puff dispersion model, which can simulate the effects of time- and space- varying meteorological conditions on pollutant transport, transformation, and removal. CALPUFF is a more accurate and representative model than ISCST3, AERMOD, or CTDM especially in its treatment of meteorology and complex terrain. CALPUFF can model the chemical transformation of pollutants in a plume and has the ability to compute dry and wet deposition of gases and E Complex terrain is defined as terrain exceeding the height of the stack being modeled. F ISCST3 in regulatory mode has the ability to model dry deposition of particles, but not deposition of gases. 5

6 particulate matter. CALPUFF may be run using full meteorology developed using CALMET or in screening mode with meteorology in a format similar to ISC meteorological data, commonly referred to as CALPUFF-Lite. CALPUFF is the recommended tool for assessing AQRV impacts. However, due to its complex modeling inputs and meteorology, it requires expert knowledge and experience. The model also requires extensive computer resources to store and process data, especially when using fully developed meteorology. Conducting a CALPUFF modeling analysis can be an expensive undertaking for any source. Nevertheless, using CALPUFF to model AQRV impacts may create long-term savings as demonstrated in this case study. The following sections describe details on the differences between the models mentioned above. There are many differences between these models, but this discussion will focus on these three: 1. Meteorology driving the model dispersion. 2. Dispersion characteristics within the model algorithms. 3. Treatment of terrain within the model algorithms. Meteorology One major difference between the models mentioned above is the meteorology that drives the pollutant dispersion. Meteorology defines how the plume is transported through the modeling domain and contains parameters used to define the dispersion of the plume. ISCST3, AERMOD, and CTDM all use single point meteorology. Meteorological data information is taken from the nearest surface and upper air stations or from an on-site observation station. Meteorological conditions are assumed to be constant over the entire X, Y, and Z expanse of the modeling domain. For ISCST3, meteorology is developed using two types of data as input: hourly surface data and twice-daily mixing height data. PCRAMMET is used to process this data and develop hourly meteorology composed of these five parameters: wind speed, wind direction, Pasquill stability class, temperature, and mixing height. Wind speed at stack top is approximated from the observed wind speed and anemometer height using the power-law wind profile. A vertical temperature gradient approximates temperature at the stack top. [7] AERMOD differs from ISCST3 in that it includes a vertical wind structure where each vertical slice may have a different wind speed and temperature. AERMOD has a more advanced system for processing meteorology using AERMET. AERMET is able to compile surface observations, vertical atmospheric profiles of wind, temperature, and turbulence, and user defined micrometeorological parameters (based on local land-use types). In AERMET, the micrometeorological parameters can vary spatially, defined in 30 degree sectors, and temporally, by month. AERMET produces two output files for the AERMOD dispersion model. The first of these files contains the surface profile and the second contains the upper air profile. [8] Table 1 summarizes AERMET parameters. 6

7 Table 1. Summary of AERMET Meteorological Parameters. User Defined Micrometeorological Parameters Surface roughness Bowen ratio Albedo Surface Turbulence Parameters Sensible heat flux Surface friction velocity Convective velocity scale Convective boundary layer height Vertical potential temperature gradient Stable boundary layer height Monin-Obukhov length Upper Air Turbulence Parameters Wind speed and direction Temperature Standard deviations of horizontal wind direction Standard deviations of vertical wind speed On-site Data (optional but preferred) Temperature Wind speed and direction Pressure Relative humidity Solar radiation The CTDM model can be run using fully developed meteorology, referred to as CTDMPLUS, or with screening meteorology, referred to as CTSCREEN. CTDMPLUS uses very involved meteorological processing requiring extensive on-site data. Since the meteorological input requirements of the CTDMPLUS model can limit its application, EPA developed a screening version called CTSCREEN. CTSCREEN uses the advanced techniques of the CTDMPLUS in screening mode where actual source and terrain characteristics are modeled with an extensive array of predetermined meteorological conditions. CTSCREEN meteorology is composed of a matrix of conditions that can occur in the atmosphere, which are specified internally in the model. CTSCREEN considers 96 combinations of wind speed and direction, standard deviation of lateral (σ v ) and vertical (σ w ) wind speed, and vertical potential temperature gradient neutral conditions, and 108 combinations of wind speed, friction velocity, Monin-Obukhov length, ambient temperature, potential temperature gradient, and mixing height for unstable conditions. [9] CALPUFF has the ability to use meteorology developed from a single meteorological station in the same format as ISC meteorological data (CALPUFF-Lite) or use multiple observation stations (full CALPUFF). CALPUFF-Lite uses ISC-type data with an extended data set. The extended data set includes meteorological variables to perform the chemical transformation and deposition calculations. These variables are: Surface friction velocity Monin-Obukhov length Surface roughness length Precipitation rate Precipitation type code Short-wave solar radiation 7

8 Potential temperature lapse rate Wind speed power law exponent Relative humidity As with ISCST3, CALPUFF-Lite meteorological fields do not vary spatially. [10] The key feature of full CALPUFF that distinguishes its meteorological processing is its use of multiple meteorological observation stations and its ability to vary both spatially and temporally. For full CALPUFF, meteorological data is processed using the CALMET meteorological model. CALMET creates three-dimensional wind fields using multiple surface stations, upper air stations, and precipitation stations. CALMET can also incorporate Mesoscale Modeling System (MM4 or MM5) data to generate the wind fields. Adjustments are made to the wind fields based on geophysical (i.e. land-use and elevation) characteristics of the modeling domain. The CALMET model requires expert knowledge and experience to ensure proper results. [10] Table 2 summarizes CALMET parameters. Table 2. Summary of CALMET Meteorological Parameters. Geophysical Elevation Land use type Surface roughness length Albedo Bowen ratio Soil heat flux Anthropogenic heat flux Leaf area index Surface Observations Wind speed Wind direction Ceiling height Opaque sky cover Air temperature Relative humidity Station pressure Precipitation code Upper Air Observations Pressure Height above sea level Temperature Wind direction Wind speed Precipitation Observations Precipitation rate MM5 Data a Wind direction Wind speed a MM5 data contains more parameters than those listed, but wind speed and wind direction are the only two parameters read by CALMET. Dispersion Characteristics ISCST3, AERMOD, and CTDM are Gaussian steady-state models. They assume that the plume disperses in the horizontal and vertical direction, resulting in Gaussian (bell-shaped) concentration distributions. For the steady-state assumption, emission rates are assumed to be constant and continuous. Furthermore, all of the pollutants released in the atmosphere remain in the atmosphere. No allowances are made for pollutant loss due to chemical conversion, surface deposition, or removal from precipitation. Portions of the plume dispersing toward the ground are assumed to be dispersed back away from the ground by turbulent eddies (eddy reflection). No variations occur in wind speed, wind direction, or Pasquill stability class during transport from the source to the receptor. Also, there is no memory of the previous hours emissions. Consequently, for each hour the plume is dispersed in the direction of the given hourly meteorology in a straight-line trajectory as seen in Figure 1. 8

9 CALPUFF is a multi-layer, multi-species, non-steady-state Lagrangian puff dispersion model. Dispersion is simulated using discrete puffs of material emitted from the modeled source. These puffs are tracked until they have left the modeling domain while calculating dispersion, transformation, and removal along the way. Furthermore, time- and space-varying meteorological conditions are explicitly incorporated in the dispersion of the puff. An important effect of the non-steady-state dispersion is that the puff can change direction with changing winds, allowing a curved trajectory as shown in Figure 1. These mechanisms are the same when conducting full CALPUFF or CALPUFF-Lite analyses. For full CALPUFF, the puff responds to space-varying surface characteristics, such as surface roughness and soil moisture, as it moves through the domain. For example, there is a continuous transition between rural and urban dispersion coefficients as the puff moves into and out of an urban area. The gridded meteorological and terrain data used in full CALPUFF meteorology are not incorporated when using CALPUFF-Lite. Instead, a single value for the land use category, surface roughness, and leaf area index is specified for the entire modeling domain. Therefore, the ability to vary dispersion spatially according to local surface characteristics is lost when modeling with CALPUFF-Lite. [10] Figure 1: Comparison of Guassian and Puff Models Dispersion Characteristics Hour 1 Hour 2 Hour 1 Hour 2 Hour 1 Hour 2 Hour 3 Hour 4 Hour 3 Source Hour 3 Source + Chemistry Hour 4 Hour 4 LOCAL WINDS GAUSSIAN PLUME DISPERSION PUFF PLUME DISPERSION Figure 2 illustrates the importance of the difference between the steady-state and non-steadystate models. In this simplified example, there is a valley located between the source and the Class I area of interest. Terrain induced channeling causes the flow vector to change direction by nearly 90 degrees. In the Gaussian model, this change in spatial flow is not recognized and the plume is directed in a straight-line trajectory toward the Class I area. With the puff model, the puff is allowed to follow the spatially varying flow and is taken down the valley and away from the Class I area. Therefore, the Guassian model predicts impacts while the puff model does not. Although simplified, this situation is realistic and exemplified by the case study to follow. 9

10 Figure 2: Comparison of Guassian and Puff Models Predicted Impacts Gaussian Plume Class I Area Wind direction near the source Source Puff Plume Dispersion Wind direction in Terrain Adjustments When the source is surrounded by complex terrain, the treatment of terrain by the different models is very important in obtaining the most accurate and representative results. EPA considers that receptor locations can fall into two categories with regard to terrain: simple or complex terrain. Simple terrain applies where the receptors have elevations at heights between the elevation of the stack base and the elevation of the stack top (release height). Complex terrain is the situation where the receptors are at elevations above the stack top. ISCST3 and CALPUFF-Lite make no adjustments to account for terrain. In ISCST3, concentrations are calculated at receptor elevations up to the elevation of the stack top. However, if receptors are entered with elevations above stack top, calculations of concentration will be made with the elevation of that receptor set to the elevation of stack top. This is commonly referred to as chopped terrain. The plume is modeled as if terrain above the stack height were removed. [7] AERMOD and CTDM account for the three-dimensional nature of plume and terrain interaction. They both use the concept of the critical dividing streamline height. The dividing streamline is dependant on the hourly meteorology and the nature of the terrain feature. Releases that take place below the dividing streamline are assumed to seek a path around the obstruction and remain at the same elevation. Releases that take place above the dividing streamline are assumed have sufficient kinetic energy to pass over the top of the obstruction. These models differ, however, in the preprocessor used to characterize the surrounding terrain. AERMOD uses a preprocessor called AERMAP, which calculates a hill height for each receptor within the modeling domain. AERMAP searches for the terrain height and location that has the greatest influence on dispersion for each individual receptor. The hill height is assumed 10

11 to be directly proportional to the difference between the elevation of the receptor and the height of the local terrain feature and inversely proportional to the distance between the receptor and the terrain feature. AERMOD employs the hill height to select the correct critical dividing streamline and concentration algorithm for each receptor. [11] CTDM requires detailed terrain and meteorological data that are representative of the modeling domain. The terrain preprocessor produces a hill file that contains the parameterization of individual hill shapes. A separate receptor file contains the association of each model receptor with a particular hill. These files are input for the calculation of the dividing streamline. The computed concentration at each receptor is then derived from the receptor position on the hill and the resultant plume position and shape. Within CALPUFF, adjustments are made within CALMET to the three-dimensional wind fields. The adjustments are made to account for kinematic effects of terrain, slope flows, and terrain blocking effects. The kinematic effects of terrain are displayed when a flow vector encounters a hill. When this occurs, the hill causes a horizontal flow to turn to a vertical flow. The height of the puff above ground is reduced as it goes over the hill and as the terrain height increases, vertical displacement decreases. Buoyancy of the air drives slope flow. During the day hot air rises creating upslope flow, whereas at night cold air sinks creating downslope flow. The slope flow is parameterized in terms of the terrain slope, distance to the crest, and local sensible heat flux. The local Froude number defines terrain blocking effects. If the Froude number at a particular grid point is less than a critical value, the wind has an uphill component, and the wind direction is adjusted to be tangent to the terrain. If the Froude number is less than the critical value, the flow goes around the terrain. [10] LEAVENWORTH MILL CASE STUDY Longview proposed to install a 10 megawatt (MW) boiler at its Leavenworth Mill, located 3.5 km from a designated Class I area, Alpine Lakes. The potential emissions from the boiler are less than 0.63 g/s (22 tons per year (tpy)) NOx, 0.24 g/s (8.4 tpy) SO 2, and 0.13 g/s (4.4 tpy) PM 10. The Leavenworth Mill is a minor source and not subject to PSD New Source Review (NSR). However, due to the short distance between the mill and Alpine Lakes, the local regulatory authority required Longview to demonstrate that the proposed boiler would not cause significant nitrogen or sulfur deposition impacts to Alpine Lakes as part of the NOC permit application. In addition to modeling the nitrogen and sulfur deposition impacts within Alpine Lakes, the regulatory authority also requested that an all-inclusive dispersion plot to show the deposition contours around the Leavenworth Mill. The Longview facility is 3.5 kilometers from the Alpine Lakes boundary. Although the boiler s potential to emit is small, the boiler s distance to Alpine Lakes is so close that most screening models will over predict any impacts from the boiler on the Class I area. Furthermore, because of its close proximity to Alpine Lakes, terrain features that lie near the source could greatly influence the air dispersion, which is not adequately evaluated in screening models as discussed in the model comparison section of this report. Specifically, the Alpine Lakes boundary follows the ridgeline of a hill rising 570 meters above the facility. This hill plays an important role in 11

12 assessing impacts to Alpine Lakes. Figure 3 shows the Leavenworth Mill relative to the Alpine Lakes and the complex terrain of the modeling domain. Figure 3: Modeling Domain 2400 m 2300 m 2200 m 2100 m 2000 m 1900 m 1800 m 1700 m 1600 m 1500 m 1400 m 1300 m 1200 m 1100 m 1000 m 900 m 800 m 700 m 600 m 500 m 400 m 300 m 200 m Modeling Methodology Leavenworth Mill Alpine Lakes Wilderness Area In this analysis, three different models were used to determine the deposition impacts on Alpine Lakes. First, a conservative screening model was used to give an initial estimate of the deposition impacts from the proposed boiler. In this case ISCST3 was chosen. Although deposition results based on the ISCST3 model may vary slightly from those calculated based on the AERMOD model, these differences will be relatively small in comparison to differences that would result between ISCST3 and CALPUFF Light modeled deposition, due to the conservative IWAQM assumptions used in calculating deposition. After the initial ISCST3 modeling results were reviewed, it was determined that ISCST3 drastically over predicted deposition impacts because of the conservative assumptions used in calculating chemical transformation and deposition. A more refined screening model, CALPUFF-Lite, was used to better represent the chemical transformation of the boiler s emissions and calculate deposition. However, CALPUFF-Lite was not able to incorporate the effects of the complex terrain on the local meteorological parameters and plume dispersion. Thus, it was determined that a more refined model would be needed for a more realistic understanding of the plume dispersion within the area immediately surrounding the Leavenworth Mill. Therefore, full CALPUFF was used to depict a more realistic representation of chemical transformation, deposition, and the affects of the complex terrain. Finally, CTSG was enabled in CALPUFF to resolve important terrain features even further as a check to determine CALPUFF s ability to capture the effects of the nearby complex terrain. 12

13 ISCST3 Setup Modeling Domain and Receptor Setup The ISCST3 modeling domain includes both the Leavenworth mill and the entire Alpine Lakes area. Ring receptors are used at two different radii. These two radii represent the shortest and longest distance between the Leavenworth Mill and the Alpine Lakes boundary. Receptors are placed at every two degrees around the ring. The use of ring receptors is widely accepted by agencies for conservative screening analyses. Figure 4 shows the receptor layout in relation to the Leavenworth Mill and Alpine Lakes. Meteorological Data The meteorological data used is in ISC format. The meteorological data is composed of the nearest, most representative surface observation station, Wenatchee Pangborn Airport, and the nearest, most representative upper air observation station, Spokane International Airport. Terrain Data In this analysis terrain data is not imported, however elevations are assigned to all structures within the Leavenworth Mill and to the ring receptors. Each ring of receptors is modeled at three specific elevations. These three elevations included the lowest elevation, the highest elevation, and the average of the lowest and highest elevation within the Alpine Lakes area. Chemical Transformation and Deposition ISCST3 does not have algorithms to calculate chemical transformation or to calculate deposition for gases. Deposition is calculated using the modeled concentrations and calculations from the IWAQM Phase I report. IWAQM guidance assumes worst-case deposition as is discussed in the Introduction section of this report. [2] Figure 4: Ring Receptors UTM Cooridinates (km) Leavenworth Mill Alpine Lakes Wilderness Area Receptors 13

14 ISCST3 Deposition Results The maximum modeled nitrogen and sulfur depositions from the screening analysis are 1.95 kg/ha/yr and 0.12 kg/ha/yr, respectively. Figure 5 shows the dispersion capabilities of ISCST3 and the predicted deposition contours immediately surrounding the Leavenworth Mill. Deposition is calculated using conservative equations recommended by the IWAQM Phase I report that predict nitrogen and sulfur deposition based on the modeled concentration of NO X and SO 2, respectively. 14

15 Figure 5: ISCST3 Deposition Contours Nitrogen Deposition kg/ha/yr Sulfur Deposition kg/ha/yr a The domain of the contour analysis extends 50 km from the Leavenworth mill. The receptors are placed every 500 m. Receptor elevations are determined by importing terrain data. b Contour scale is not linear. c The lowest deposition contour represents the threshold that the local regulatory authority used to determine significance. Note that state agencies and EPA are currently requiring significance thresholds of kg/ha/yr for both nitrogen and sulfur deposition for PSD projects based on Federal Land Managers recommendations. 15

16 The ISCST3 sulfur deposition contours show that sulfur deposition from the proposed boiler would significantly impact Alpine Lakes. However, there are several factors that must be taken into consideration. First, ISCST3 does not have chemical transformation algorithms. Therefore, based on IWAQM guidance, it is assumed that none of the pollutant emitted from the boiler is removed or transformed to other species. Second, ISCST3 does not calculate deposition for gases. Therefore conservative assumptions are made regarding the speed and amount of deposition that occurs. Finally, as demonstrated in Figure 5, ISCST3 is unable to incorporate the effects of the terrain on dispersion modeling. Although it appears that the contours follow the terrain in some areas, these effects are due to the elevation of the receptors and not to terrain adjustments. The lack of terrain influence is demonstrated by the contours to the left of the mountains in Alpine Lakes. ISCST3 algorithms allow concentrations to travel through terrain rather than above or around it. Therefore it is concluded that the ISCST3 model is not an appropriate representation of plume dispersion or deposition for this particular case study. CALPUFF-Lite Setup Modeling Domain and Receptor Setup In the CALPUFF-Lite screening analysis, the modeling domain and receptor setup is identical to the set up described for the ISCST3 screening analysis. Figure 4 shows the receptor layout in relation to the Leavenworth Mill and Alpine Lakes. Meteorological Data The meteorological data used is in extended ISC format, which includes parameters such as relative humidity used in chemical transformation. The meteorological data is composed of the nearest, most representative surface observations station, Wenatchee Pangborn Airport, the nearest, most representative upper air observation station, Spokane International Airport, and the most representative precipitation station, Lake Wenatchee. Terrain Data The terrain assumptions are the same as those described for the ISCST3 screening analysis. Chemical Transformation and Deposition Both chemical transformation using MESOPUFF II and deposition are included in the model algorithms. CALPUFF-Lite Deposition Results The maximum modeled nitrogen and sulfur depositions from the screening analysis are 0.09 kg/ha/yr and 0.12 kg/ha/yr, respectively. Figure 6 shows the dispersion capabilities of CALPUFF-Lite and the predicted deposition contours immediately surrounding the Leavenworth Mill. The chemical transformation and deposition algorithms used in CALPUFF-Lite provide more representative deposition results. However, like ISCST3, CALPUFF-Lite does not consider terrain effects, which is demonstrated by the contours going through the hills located near the Leavenworth Mill. Therefore, it is concluded that a more refined modeling analysis is needed to adequately define the effect of the 16

17 local terrain features and to demonstrate that the proposed boiler will not have a significant impact on the Class I area. Figure 6: CALPUFF-Lite Deposition Contours Nitrogen Deposition kg/ha/yr Sulfur Deposition kg/ha/yr a The domain of the contour analysis extends 50 km from the Leavenworth mill. The receptors are placed every 500 m. Receptor elevations are determined by importing terrain data. b Contour scale is not linear. c The lowest deposition contour represents the threshold that the local regulatory authority used to determine significance. Note that state agencies and EPA are currently requiring significance thresholds of kg/ha/yr for both nitrogen and sulfur deposition for PSD projects based on Federal Land Managers recommendations. 17

18 CALPUFF Setup Modeling Domain and Receptor Setup In the refined CALPUFF analysis, the domain is determined based on the grid spacing and the amount of computer resources required to run the model. For long-range CALPUFF analyses meteorological and terrain data is typically calculated every four kilometers. However, the regulatory authority requested a smaller grid spacing, 500 m, for this near-field analysis to minimize the chance of CALPUFF overlooking important terrain features. The smaller grid spacing made it unreasonable to model the entire Class I area with the available computer resources. Therefore, it was agreed by the regulatory authority that a domain extending 50 km in all directions around the Leavenworth Mill was appropriate for this model. The smaller domain is justifiable because the boiler emissions are small and unlikely to cause any long-range impacts. Receptors are placed within the Alpine Lakes boundary with a 500 m spacing. Figure 7 shows the model domain and receptor setup used in this analysis. Meteorological Data CALMET meteorological data was used in the refined CALPUFF analysis. This data incorporated all the meteorological data used in the CALPUFF-Lite analysis with the addition of the Peshastin Telem precipitation station, MM5 data generated with 12 km grid spacing, 1:250,000 scale Digital Elevation Model (DEM) data, and 1:250,000 scale land use data. Terrain Data Terrain data is incorporated into CALMET as described in the Terrain Adjustments Section. Chemical Transformation and Deposition Both chemical transformation using MESOPUFF II and deposition are include in model algorithms. 18

19 Figure 7: CALPUFF Receptors Receptors Inside Alpine Lakes Boundary Leavenworth Mill CALPUFF Deposition Results The maximum modeled nitrogen and sulfur depositions are kg/ha/yr and kg/ha/yr, respectively. Figure 8 shows the dispersion capabilities of CALPUFF and the predicted deposition contours immediately surrounding the Leavenworth Mill. As mentioned above, the Alpine Lakes boundary follows the ridgeline of a hill located southwest of the Leavenworth Mill. The refined CALPUFF analysis demonstrates the significance of the rising terrain between and the Leavenworth Mill and Alpine Lakes and other complex terrain features. Figure 8 shows how the hill improves the dispersion of the boiler s plume as well as acts as a protective barrier to the Class I area. During the course of the modeled time period, the majority of the puffs do not have sufficient energy to reach the top of the hill and impact the Alpine Lakes boundary. Furthermore, the puffs follow the valley induced winds predicted by CALMET, which have a primarily North-South component. The refined modeling analysis provides a significantly better representation of how the emissions from the boiler will be dispersed and deposited. Furthermore, it demonstrates that the proposed boiler would not have a significant impact on Alpine Lakes. 19

20 Figure 8: CALPUFF Deposition Contours Nitrogen Deposition kg/ha/yr Sulfur Deposition kg/ha/yr a The domain of the contour analysis extends 50 km from the Leavenworth mill. The receptors are placed every 500 m. Receptor elevations are determined by importing terrain data. b Contour scale is not linear. c The lowest deposition contour represents the threshold that the local regulatory authority used to determine significance. Note that state agencies and EPA are currently requiring significance thresholds of kg/ha/yr for both nitrogen and sulfur deposition for PSD projects based on Federal Land Managers recommendations. 20

21 Summary of Modeled Deposition Table 3 summarizes the deposition results from the three modeling analyses and compares them to both the threshold requested by the local regulatory authority and the FLM s recommend DATs. G These results show that the use of a more refined and more representative model such as CALPUFF can be valuable to minor sources, especially when dealing with complex terrain. In this analysis CALPUFF is able to present a more realistic representation of dispersion from the boiler at the Leavenworth Mill and demonstrate that the impacts on Alpine Lakes are not significant. Due to the fact that ISCST3 and CALPUFF-Lite do not incorporate terrain effects on dispersion, both of these models over-predict impacts on Alpine Lakes. Table 3. Modeled Deposition Results Sulfur Deposition (kg/ha/yr) Nitrogen Deposition (kg/ha/yr) ISCST3 (kg/ha/yr) CALPUFF- Lite (kg/ha/yr) CALPUFF (kg/ha/yr) Target Thresholds (kg/ha/yr) CTSG SUPPORTING ANALYSIS FLM Thresholds (kg/ha/yr) The grid spacing used in CALMET is 500 meters. Although this grid resolution is adequate for the domain selected, it is expected to not sufficiently resolve the terrain features between the facility and the Class I boundary. Since impacts on the Alpine Lakes boundary are of the most concern, it is important to attempt to account for this terrain feature. The most accurate way to model dispersion around the terrain feature within CALPUFF is to include the complex terrain algorithm for subgrid scale features (CTSG). The CTSG option in CALPUFF allows the model to only calculate concentration impacts, not deposition fluxes, at specific receptors that are placed on significant terrain features. The characteristics of terrain features modeled by CTSG are specified from a data file produced by the CTSCREEN processor. Similarly, CTSG receptors are read from a corresponding CTSCREEN receptor file. 27 CTSG receptors are placed along the boundary of Alpine Lakes nearest to the Leavenworth Mill as shown in Figure 9. G Note that state agencies and EPA are currently requiring significance thresholds of kg/ha/yr for both nitrogen and sulfur deposition for PSD projects based on Federal Land Managers recommendations. 21

22 5291 Figure 9: CTSG Receptor Locations Leavenworth Mill Alpine Lakes Wilderness Area CTSG Receptors In this analysis, the CTSG option in CALPUFF is selected to provide further information on how local terrain will affect the proposed project s impacts on the Alpine Lakes Class I Area. Although CALPUFF only calculates concentration impacts at CTSG receptors, a comparison between the modeled concentrations at CALPUFF discrete receptors with the modeled concentrations at CTSG receptors can provide additional information regarding CALPUFF s ability to incorporate the effects of local terrain. This additional analysis is used to determine whether CALPUFF s predicted deposition impacts are under-estimated or over-estimated. Table 4 compares the maximum and average modeled concentrations for both 27 discrete CALPUFF receptors and 27 CTSG receptors with the same UTM coordinates. All pollutants are calculated on an annual average. The comparison of concentrations at these receptor locations shows that all of the CTSG receptors predict lower concentrations than the discrete CALPUFF receptors. Therefore, it is believed that the CALPUFF predicted deposition impacts for nitrogen and sulfur in Table 3 are conservative and over-predict the actual deposition impact of the proposed project on Alpine Lakes. These results are logical since terrain usually increases the dispersion of a plume, thereby decreasing concentration. 22

23 Table 4. Comparison of Concentrations for Discrete CALPUFF Receptors and CTSG Receptors Discrete CALPUFF Receptors CTSG Receptors Pollutant Maximum Concentration (mg/m 3 ) Average Concentration (mg/m 3 ) Maximum Concentration (mg/m 3 ) Average Concentration (mg/m 3 ) Average % Change HNO E E E E % NO E E E E % NOx 3.69E E E E % SO E E E E % SO E E E E % Figure 10 shows example concentration comparisons between the 27 discrete CALPUFF receptors and 27 CTSG receptors for NOx and SO 2. Similar trends are seen with HNO 3, NO 3, and SO 4. Figure 10: Comparison of CTSG and Discrete CALPUFF Receptors NO X Concentration Comparison Concentration (m g/m 3 ) 4.00E E E E E E E E E Receptor Number CTSG Receptors Discrete Receptors SO 2 Concentration Comparison 1.20E-04 Concentration (mg/m 3 ) 1.00E E E E E-05 CTSG Receptors Discrete Receptors 0.00E Receptor Number 23

24 REGULATORY SIGNIFICANCE The Leavenworth Mill case study is significant in EPA Region X for the following reasons. 1. It is the first time CALPUFF has been used to demonstrate compliance for a minor source as part of a NOC application. Minor sources, prior to this case study, have not used CALPUFF for dispersion modeling. Often it is not economical for small sources to use refined modeling. Screening models are usually sufficient to demonstrate compliance with regulatory standards. However, as this case study shows, some near-field analyses may require more sophisticated and refined modeling in order to demonstrate compliance, especially for sources located in complex terrain. 2. It is the first time CALPUFF has been used to perform a near-field analysis for regulatory purposes. Traditionally CALPUFF has been used only in long-range Class I impact analysis. Near-field Class I impacts have traditionally been analyzed using screening techniques with less refined models. This case study is important because CALPUFF was the only model capable of incorporating the nearby terrain features and demonstrating that the proposed source would not significantly impact the nearby Class I area. 3. Regulatory authority domain and grid spacing requirements for Class I area analyses were modified to enhance the near-field modeling analysis. In this case study the modeling domain includes only the nearest segment of the Class I area instead of the entire Class I area, and a CALPUFF grid spacing of 0.5 km is used rather than the traditional 4 km grid spacing. 4. It is the first time CTSG receptors have been included in a regulatory CALPUFF analysis. The use of CTSG receptors reaffirms CALPUFF s ability to accurately incorporate small terrain features such as the 570 m hill rising from the Leavenworth mill toward Alpine Lakes. ECONOMIC SIGNIFICANCE CALPUFF dispersion modeling is expensive when compared to screening models used to demonstrate near-field Class I impacts. However, in this particular case study, the Leavenworth Mill CALPUFF analysis proved to be a cost savings for Longview. CALPUFF is the only model able to incorporate the complex terrain surrounding the Leavenworth mill as well as provide representative chemical transformation and deposition calculations. This proved to be essential in demonstrating that the proposed boiler would not have significant impacts on Alpine Lakes. Because, Longview was able to show compliance with the regulatory standards, the NOC for the boiler was approved within a relatively short time frame, allowing for earlier operation and additional cost savings. Furthermore, the CALPUFF analyses demonstrated compliance with regulatory standards without implementing unnecessary add-on controls or operation limits. With the other less refined models, Longview would have had to implement several controls or operation limits in order to demonstrate compliance with the regulatory standards, as well as limit combustion to only ultra-low sulfur fuel rather than low-sulfur fuel. By demonstrating that the impacts of the boiler will be insignificant without the add-on controls, operation limits, or ultra-low sulfur fuel, Longview was relieved of significant and unnecessary costs. For example, the annual cost 24