Effect of High-Occupancy Toll Lanes on Mass Vehicle Emissions

Effect of High-Occupancy Toll Lanes on Mass Vehicle Emissions Application to I-85 in Atlanta, Georgia David N. Kall, Randall L. Guensler, Michael O. Rodgers, and Vishal S. Pandey High-occupancy toll (HOT) lanes were recently proposed for I-85 in Atlanta, Georgia, as a way to relieve congestion and to provide a reliable commute time for single-occupant drivers who are willing to pay a toll. It is important to evaluate the air quality impacts of such a proposal in the context of environmental regulations, such as the National Environmental Policy Act (NEPA), and transportation conformity regulations. Several factors affect mass vehicle emissions, such as vehicle activity, speed, age distributions, and class distributions. These factors are incorporated into a base scenario that models the current conditions on I-85 with high-occupancy vehicle lanes by using data collected in the corridor during the summer of 2007, and a future scenario that models the implementation of HOT lanes on this corridor by using information from cities that already have HOT lanes. The MOBILE-Matrix modeling tool, recently developed by the Georgia Institute of Technology, is used for the emissions analysis by means of the input factors described above. It calculates mass emissions for five pollutants: hydrocarbons (HC), oxides of nitrogen (NO x ), carbon monoxide (CO), particulate matter with an aerodynamic diameter of 2.5 µm or smaller (PM 2.5 ), and PM 10 (with an aerodynamic diameter of 10 µm or smaller) as a function of fleet composition and on-road operating conditions. The modeling work predicts extremely small increases in mass emissions for NO x, CO, PM 2.5, and PM 10, and an extremely small decrease in mass emissions for HC. In addition, the postimplementation emissions changes fall well within the motor vehicle emissions budgets for the facility that are used in air quality planning. Therefore, implementation of HOT lanes on I-85 in Atlanta should not violate the emissions budget requirements of the federal Transportation Conformity Rule. For NEPA purposes, this analysis could be used to make the case that air quality impacts are not significant, and therefore further detailed analyses are not required. High-occupancy toll (HOT) lanes were recently proposed for I-85 in Atlanta, Georgia, as a way to relieve congestion and provide a reliable commute time for three-person carpools and for others willing to pay a toll. The current I-85 HOT lane proposal would not involve construction of new lanes but would instead convert the existing HOV-2 (two-person high-occupancy vehicle) lanes to HOV-3 (three-person HOV) lanes and use the new excess capacity for tollpaying single-occupant vehicles (SOVs). The proposed HOT lane School of Civil and Environmental Engineering, Georgia Institute of Technology, 790 Atlantic Drive, Atlanta, GA 30332-0355. Corresponding author: D. N. Kall, kall.david@gmail.com. Transportation Research Record: Journal of the Transportation Research Board, No. 2123, Transportation Research Board of the National Academies, Washington, D.C., 2009, pp. 88 96. DOI: 10.3141/2123-10 configuration is the same as the current HOV lane, which is a concurrent flow lane without barrier separation from the general purpose (GP) lanes. This study focuses on air quality concerns by evaluating how the HOT lane conversion on I-85 will affect vehicle total mass emissions due to changes in traffic volumes, vehicle fleet characteristics (such as age distributions and class distributions), and changes in operating conditions (such as speeds). Emissions are evaluated for five pollutants: carbon monoxide (CO), hydrocarbons (HC), oxides of nitrogen (NO x ), particulate matter with an aerodynamic diameter of 2.5 µm or smaller (PM 2.5 ), and PM 10 (with an aerodynamic diameter of 10 µm or smaller). With respect to the implications of the project on environmental regulations [the National Environmental Policy Act (NEPA) and the federal Transportation Conformity Rule], the authors conclude that there is no significant impact on air quality. METHODOLOGY The goal of this study is to understand how vehicle mass emissions change as a result of implementing HOT lanes on I-85 in Atlanta. This goal is achieved by calculating the mass emissions for a base scenario and for a future scenario on the basis of four input factors. Fleet differences are modeled in each scenario for peak periods in each direction. The base scenario estimates are performed for the vehicle fleet that currently uses the I-85 corridor, where the fleet has been delineated by using license plate and count data collected by the Georgia Institute of Technology (Georgia Tech) during the summer of 2007. The future scenario employs a different vehicle fleet that is based on observed vehicle fleet characteristics from SR-91 lanes in California. Changes in speeds in GP lanes are taken from the MnPass lanes in Minnesota. In addition, changes to speeds and volumes for HOT lanes are based on the assumption that pricing will hold the lane at effective capacity. Mass emissions are calculated by multiplying vehicle activity, measured in vehicle miles traveled (VMT), by a composite emission rate (g/veh mi): ME h = VMT h CER where ME/h = mass emissions (g/h), VMT/h = vehicle miles traveled per hour (veh mi/h) = traffic volume per hour (veh/h) * roadway distance (mi), and CER = composite emission rate (g/veh mi). 88

Kall, Guensler, Rodgers, and Pandey 89 The composite emission rate is the average emission rate per vehicle and is calculated by an emission rate model that uses the emissions rates for different vehicle types (classes), vehicle ages, and operating characteristics for the modeled vehicles. For this analysis, the MOBILE-Matrix CALINE-Grid tool provides the composite emissions rates for both the base and future scenarios. To examine the effect of HOT lane implementation on the I-85 corridor, the vehicle activity and composite emissions rates both before and after conversion of HOV lanes is examined to determine whether emissions will decrease or increase and whether the net change in emissions is significant. Overview of MOBILE-Matrix CALINE-Grid Modeling Tool The MOBILE-Matrix CALINE-Grid, which is Georgia Tech s newly developed tool for assessing air quality impacts, provides emissions rates and mass emissions calculations on a link-by-link basis with a set of prerun emissions rate lookup matrices. The development of this tool closely followed the methods outlined by Guensler et al. (1). One advantage of the MOBILE-Matrix tool over the U.S. Environmental Protection Agency s (EPA s) MOBILE6 model alone is that thousands of MOBILE6.2 modeling runs have already been completed by varying calendar years, evaluation month (January or July), facility type, vehicle speeds, and ambient temperatures. The modeling runs use standard assumptions employed by the Atlanta Regional Commission (ARC) that incorporate local conditions in the 13-county Atlanta area, such as humidity, inspection and maintenance programs, and regional fuel characteristics. The results of these modeling runs have been placed in large lookup tables and can be called up on the basis of inputs provided by the user of the MOBILE Matrix tool that correspond to the varying factors in the lookup tables, such as calendar year and evaluation month. Therefore, speeds from hundreds of roadway links can be automatically matched with the appropriate 28 25 emission rates table, which represents 28 MOBILE vehicle types and 25 vehicle ages. Next, the 28 25 emission rate table is multiplied by the vehicle age distributions and vehicle class distributions provided by the user. This procedure results in a single composite emission rate for each roadway link. Vehicle Activity Volumes for the I-85 base scenario come from count data of a summer field study. These data were collected on JAMAR count boards for 60 to 75 min for three days (Monday, Wednesday, and Friday) for each site, time, and direction combination. The observations occurred between 7:00 and 9:00 a.m. for the a.m. peak and between 4:30 and 6:30 p.m. for the p.m. peak. During this 2-h period, the research team collected both northbound and southbound data, in one direction for the first hour and the opposite direction for the second hour. The data were collected by individual lane; therefore, it is possible to separate the volumes from GP and HOV lanes. Table 1 shows the volumes derived from the count data of the summer field study. The future scenario applies changes to the base scenario volumes on the basis of assumptions about effective capacity of the HOT lane. HOT lanes operate on the concept that pricing can be used to adjust the demand for that particular lane, so that traffic volumes are held right below the capacity of the lane. Research by Guin et al. (2) on I-85 in Atlanta has determined that the capacity of an HOV lane is lower than the capacity of a GP lane. That research used data from Georgia Navigator the local intelligent transportation system that uses video detection system cameras to monitor continuous speed and volume data on the major freeway system to determine the effective capacity for an HOV lane and the adjacent GP lane on a stretch of I-85 from I-285 to SR-316. The researchers found that the effective capacity for the I-85 HOV lane is approximately 1,500 veh/h, occurring at 40 mph. The GP lane s effective capacity is higher, at around 2,500 veh/h. These values are shown on the speed flow diagrams in Figure 1, which were developed by using the Georgia Navigator data. That research cited HOV drivers safety TABLE 1 Volumes for Base Scenario GP HOV Section Observation Site Time Direction (vol./ln/h) (vol./ln/h) 1 Fair Dr. a.m. NB 1,507 1,145 2 Fifth St. a.m. NB 1,924 1,336 3 Chamblee Tucker Rd. a.m. NB 951 161 4 Northcrest Rd. a.m. NB 980 161 5 Beaver Ruin Rd. a.m. NB 1,367 248 1 Fair Dr. a.m. SB 870 231 2 Fifth St. a.m. SB 1,625 1,130 3 Chamblee Tucker Rd. a.m. SB 1,766 872 4 Northcrest Rd. a.m. SB 1,398 1,118 5 Beaver Ruin Rd. a.m. SB 1,824 1,373 1 Fair Dr. p.m. NB 1,154 433 2 Fifth St. p.m. NB 1,753 1,297 3 Chamblee Tucker Rd. p.m. NB 1,529 847 4 Northcrest Rd. p.m. NB 1,170 1,152 5 Beaver Ruin Rd. p.m. NB 1,545 1,181 1 Fair Dr. p.m. SB 1,707 864 2 Fifth St. p.m. SB 1,256 1,365 3 Chamblee Tucker Rd. p.m. SB 1,074 387 4 Northcrest Rd. p.m. SB 909 321 5 Beaver Ruin Rd. p.m. SB 1,404 502 NOTE: NB = northbound; SB = southbound.

90 Transportation Research Record 2123 80 70 60 Vehicle Speed (mph) 50 40 30 20 GP1 HOV 10 0 0 500 1000 1500 2000 2500 3000 Traffic Flow (vehicles/hour) FIGURE 1 Georgia navigator speed flow diagram for I-85 north of Atlanta (2). concerns that were associated with vehicles jumping into or out of the HOV lane as the main reason for the lower effective capacity in the HOV lane (2). It is possible that electronic enforcement of entrance and exit points for the HOT lanes could reduce these weaving patterns and increase the effective capacity slightly beyond 1,500 veh/h; however, this analysis assumes that the effective capacity of the HOT lane will be the same as the effective capacity of the HOV lane. To estimate the volume reductions on the GP lanes, a comparison is made between the volume increases on the HOT lanes and the volume reductions on the GP lanes. In theory, much of the volume increase on the HOT lane is due to vehicles in the GP lanes paying to switch to the faster HOT lanes. The remaining volume increase on the HOT lane likely comes from the latent demand for extra capacity in the corridor. For example, travelers who would not have taken the trip or would have taken a different route are now choosing to use the extra capacity in the HOT lane. Data from the MnPass HOT lane project in Minneapolis, Minnesota, which has a similar configuration for much of its length (concurrent flow lanes not barrier separated) to that envisioned for the Atlanta HOT system, revealed that the GP lane volume decrease is 77.2% of the HOT lane volume increase (3). Therefore, the change in GP lane volume is calculated as 77.2% of the change in HOV lane volume, as shown in Table 2. TABLE 2 Volume Changes for Future Scenario Volume per Hour Section Observation Site Time Direction HOV Base HOT Future HOV Change GP Change 1 Fair Dr. a.m. NB 1,145 1,500 355 274 2 Fifth St. a.m. NB 1,336 1,500 164 127 3 Chamblee Tucker Rd. a.m. NB 161 161 0 0 4 Northcrest Rd. a.m. NB 161 161 0 0 5 Beaver Ruin Rd. a.m. NB 248 248 0 0 1 Fair Dr. a.m. SB 231 231 0 0 2 Fifth St. a.m. SB 1,130 1,500 370 286 3 Chamblee Tucker Rd. a.m. SB 872 1,500 628 485 4 Northcrest Rd. a.m. SB 1,118 1,500 382 295 5 Beaver Ruin Rd. a.m. SB 1,373 1,500 127 98 1 Fair Dr. p.m. NB 433 433 0 0 2 Fifth St. p.m. NB 1,297 1,500 203 157 3 Chamblee Tucker Rd. p.m. NB 847 1,500 653 504 4 Northcrest Rd. p.m. NB 1,152 1,500 348 269 5 Beaver Ruin Rd. p.m. NB 1,181 1,500 319 246 1 Fair Dr. p.m. SB 864 1,500 636 491 2 Fifth St. p.m. SB 1,365 1,500 135 104 3 Chamblee Tucker Rd. p.m. SB 387 387 0 0 4 Northcrest Rd. p.m. SB 321 321 0 0 5 Beaver Ruin Rd. p.m. SB 502 502 0 0

Kall, Guensler, Rodgers, and Pandey 91 Vehicle Operating Speeds Speeds for the base scenario were obtained from a 2006 run of the travel demand model (TDM) from the ARC, which is the local metropolitan planning organization. This particular run was conducted for the 2006 state implementation plan (SIP) for 8-h ozone nonattainment (4). Speeds for the a.m. and p.m. periods were selected from the TDM to correspond to the periods of volume data collection during the summer field study. Changes to speed are made for the future scenario only for the links in the peak time and direction, such as a.m. southbound for northern observation sites. Fifth Street is considered to have peak conditions during all times and directions. All HOT links in the peak time and direction are assumed to operate at 40 mph, which is the speed that corresponds to the effective capacity of 1,500 veh/h applied to the HOT links on the basis of previous research on I-85 (2). For the GP links in the peak time and direction, data on changes in speed on I-394 associated with the HOT lane implementation in Minneapolis are applied. These data show a 4.2% increase in speed for GP lanes during the morning and a 4.1% increase in speed for GP lanes during the afternoon (3). These values are taken from the diamond-lane section of I-394, which resembles the proposed configuration on I-85 in Atlanta. On-Road Vehicle Age Distribution MOBILE modeling typically uses registration distribution and mileage accumulation as inputs, but these two inputs can be replaced by an on-road vehicle age distribution if the distribution is known. This study uses the on-road vehicle age distribution because it is available for the I-85 corridor from the summer field study license plate data for light-duty vehicles (LDVs) and light-duty trucks (LDTs). For the remaining vehicle classes (heavy-duty trucks and buses), age distributions from the ARC and the Georgia Environmental Protection Division are used. These data are based upon registration data from R. L. Polk and Co. for the 13-county area (except for heavy-duty vehicle Classification 8B, which uses national defaults), and national mileage accumulation defaults from the EPA. The on-road vehicle age distributions for the base scenario are developed by using the license plate data for passenger cars and LDTs, which were recorded by site, lane, time, and direction. These license plates are matched to the Georgia registration database, which provides vehicle identification numbers that are decoded to obtain make and model year information for over 110,000 vehicles. The make of these vehicles, such as four-door sedan, are mapped to MOBILE vehicle types. Because the license plate data were collected only for small passenger vehicles, almost all the vehicles fell into the MOBILE vehicle type LDV or LDT2. On-road vehicle age distributions are created for the base scenario individually for each of the 80 combinations of vehicle type, lane type, time, direction, and site, by using gamma distribution values to fill in data gaps for vehicles from older model years. However, due to low volumes for HOV sites during off-peak times, these 20 distributions are condensed into eight distributions by combining from surrounding sites, to prevent the lack of data from producing unreliably shaped distributions. For the future scenario, changes are applied to the on-road vehicle age distributions in the peak time and direction to reflect the change in vehicle fleet characteristics during pricing periods on the HOT lanes. A case study of SR-91 in southern California contains information on the differences in vehicle fleet characteristics between the GP and HOT lanes (5). SR-91 is of interest to this study because it is a variably priced tollway. The SR-91 toll fluctuates on the basis of demand but always allows carpools of three or more persons to travel in the lanes for free; these are the same payment conditions envisioned for the HOT lanes in Atlanta. In their study, Barth et al. (5) used license plate data to distribute the vehicle fleet on the GP and HOT lanes into technology categories used in the Comprehensive Modal Emissions Model, as shown in Table 3. To apply the Barth et al. (5) distributions to I-85 in Atlanta, a mapping procedure is used to convert the vehicle technology categories into model years, which are required by the MOBILE- Matrix emissions model used in this analysis. This mapping procedure, which is outlined by Barth et al. in a separate paper (5), uses data that describe the vehicle technologies that were used during the production of certain model year vehicles. For example, vehicles from the 1974 model year and earlier do not have catalytic converters; therefore, the category for no catalyst can be applied to this group of model year vehicles. While the mapping procedure produced a distribution for 1973 to 1997 model year vehicles, it was assumed that these distributions could be generalized to vehicle ages instead of model years. While this assumption ignores several real-world occurrences, such as changing vehicle lifetimes over time and vehicle sales patterns that are dependent on particular model years, it is believed that these items would create only small differences over the 13 years between the Barth et al. (5) study and the proposed I-85 HOT lane implementation. Therefore, these differences were ignored for the purposes of this analysis. Instead of using the actual distributions found for HOT lanes on SR-91, the relationships between the GP and HOT lanes for various vehicle ages on SR-91 are applied to the age distributions for the GP lane observed on I-85 in Atlanta. Use of this method reflects the local vehicle age distributions in Atlanta and at the same time considers differences in the age of vehicles that drive in the HOT lane. Therefore, for each vehicle age, a conversion factor for GP lane to HOT lane is created by dividing the distribution percentage for the HOT lane by the distribution percentage for the GP lane. These conversion factors are simply multiplied by an age distribution for the GP lane to obtain a projected age distribution for the HOT lane. The magnitude of the conversion factors for GP lane to HOT lane can be interpreted to describe the extent of the difference in number of vehicles of a certain age that can be expected in the HOT lane compared with the GP lane. For example, a conversion factor of 1.0 means that there are an equal number of vehicles from that model year in the GP and HOT lanes, while a conversion factor of 2.0 means that there are twice as many vehicles from that model year in the HOT lane than in the GP lane. Figure 2 shows that 9-year-old and newer vehicles have factors above the 1.0 solid dividing line, which correspond to more vehicles in the HOT lane than in the GP lane. The 10-year-old and older vehicles have a factor below the 1.0 solid dividing line, which corresponds to fewer vehicles in the HOT lane. This result is expected due to the known strong relationship between higher personal income and newer vehicle ages (7) and the expectation that HOT lane drivers will be higher income individuals who are more willing to pay the associated toll. These conversion factors are applied to age distributions for the GP lane to estimate the HOT lane distributions for the peak time and direction. Figure 3 shows an example of an on-road vehicle age distribution for a HOT lane calculated by applying the conversion factors to the GP lane distribution.

92 Transportation Research Record 2123 TABLE 3 SR-91 Vehicle Fleet Distribution by Technology Category (5) No. Comprehensive Modal Emissions Model Category GP HOT 1 No catalyst 12.4 0.53 2 2-way catalyst 4.81 0.64 3 3-way catalyst, carbureted 3.74 2.53 4 3-way catalyst, FI, >50K mi, low power/weight 9.17 9.66 5 3-way catalyst, FI, >50K mi, high power/weight 12.22 16.63 6 3-way catalyst, FI, <50K mi, low power/weight 0.98 1.18 7 3-way catalyst, FI, <50K mi, high power/weight 1.53 2.6 8 Tier 1, >50K mi, low power/weight 1.54 2.29 9 Tier 1, >50K mi, high power/weight 2.99 6.41 10 Tier 1, <50K mi, low power/weight 3.25 4.36 11 Tier 1, <50K mi, high power/weight 7.01 14.51 12 Pre-1979 ( 8,500 GVW) 7.01 0.66 13 1979 1983 ( 8,500 GVW) 1.59 0.61 14 1984 1987 ( 8,500 GVW) 1.84 1.69 15 1988 1993, 3,750 LVW 3.07 4.46 16 1988 1993, >3,750 LVW 2.9 6.39 17 Tier 1 LDT2/3 (3,751 5,750 LVW or alt. LVW) 0.79 1.5 18 Tier 1 LDT4 (6,001 8,500 GVW, >5,750 alt. LVW) 0.8 2.5 19 Runs lean 3.8 3.58 20 Runs rich 1.45 1.6 21 Misfire 7.9 6 22 Bad catalyst 5.52 4.39 23 Runs very rich 1.38 0.73 24 Tier 1, >100K mi 0.1 0.17 25 Gasoline-powered, LDT (>8,500 GVW) 2.14 4.24 40 Diesel-powered, LDT (>8,500 GVW) 0.04 0.17 Total 100 100 On-Road Vehicle Class Distribution To create on-road vehicle class distributions (known as VMT fractions in MOBILE) local data are used from the summer field study volume counts, which provided four vehicle types: passenger vehicles, small trucks, large trucks, and buses. Because the summer field study data do not provide sufficient detail to break the vehicles into 28 MOBILE vehicle types, a combination of the summer field study data and the national default VMT fractions are used to create local VMT fractions. To do this, summer field study vehicle types, which are based on FHWA vehicle class definitions, are matched to MOBILE6 vehicle types through a mapping procedure 3.50 3.00 2.50 Multiplier 2.00 1.50 LDGV LDGT 1.00 0.50 0.00 24+ 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 Vehicle Age FIGURE 2 GP to HOT lane conversion fractions. (LDGV = light-duty gasoline vehicles; LDGT = light-duty gasoline trucks.)

Kall, Guensler, Rodgers, and Pandey 93 0.175 0.15 0.125 Percent 0.1 0.075 Base HOV Base GP Future HOT 0.05 0.025 0 1984 1986 1988 1990 1992 1994 1996 1998 Model Year (a) 2000 2002 2004 2006 2008 0.4 0.35 National Defaults GP AM NB Fifth St. 0.3 0.25 Percent 0.2 0.15 0.1 0.05 0 LDGV LDGT1 LDGT2 LDGT3 LDGT4 HDGV2B HDGV3 HDGV4 HDGV5 HDGV6 HDGV7 HDGV8A HDGV8B (b) LDDV LDDT12 HDDV2B HDDV3 HDDV4 HDDV5 MOBILE6 Vehicle Type HDDV6 HDDV7 HDDV8A HDDV8B MC HDGB HDDBT HDDBS LDDT34 FIGURE 3 Example HOT lane age distribution and vehicle class distribution. (HDGV = heavy-duty gas vehicles; LDDV = light-duty diesel vehicles; LDDT = light-duty diesel trucks; HDDV = heavy-duty diesel vehicles; MC = motorcycles; HDGB = heavy-duty gasoline bus; HDDBT = heavy-duty diesel and urban transit buses; HDDBS = heavy-duty diesel school buses, and LDDT = light-duty diesel trucks.) from Yoon (8). With knowledge of which MOBILE6 vehicle types are included in each of the summer field study categories, the distributions collected for that study can be further distributed to all 28 MOBILE6 vehicle types. This distribution is done by using the relative share of each of the MOBILE6 vehicle types for each summer field study category on the basis of national defaults. An example of a full set of local VMT fractions compared with national defaults can be found in Figure 3. Because the observation site for this example lies within I-285 perimeter (on GP lanes), which is the boundary for restricting heavy through-trucks in Atlanta, the percentage of heavy trucks is substantially less than the national defaults. Heavy trucks are not currently allowed on any HOV lanes (inside or outside of I-285) and will continue to be prohibited on future HOT lanes.

94 Transportation Research Record 2123 The same VMT fractions are used for both the base and the future scenarios on the basis of the assumption that the implementation of a HOT lane will not affect the vehicle class distribution. This is believed to be a fairly reasonable assumption as long as the HOT lane implementation also allows transit and school buses in the same manner as the current HOV lane does. It is possible that the demographics of HOT lane drivers and possible commercial uses could cause a different mix of LDV and LDT2 vehicles. However, due to the lack of available data on this effect and the belief that this effect would be fairly small, a possible change in the mix of LDV and LDT2 vehicles is ignored for this analysis. EMISSIONS ANALYSIS AND RESULTS Major Findings After the MOBILE-Matrix tool with the input factors described above for a 2010 analysis year are run, the final results show very small changes in the mass emissions between the base and the future scenarios. For all pollutants except HC, very small increases are predicted for the future scenarios. For HC, a small decrease is predicted for the future scenarios. These results appear in Table 4. When the peak time and direction are isolated, as shown in the columns for those variables in Table 4, the same trends between base and future scenarios can be observed. However, the magnitude of the increase or decrease is slightly higher because all changes in the future scenario are applied during the peak time and direction. To investigate further the reasons for the trends in predicted results, the percentage changes in mass emissions are examined separately for each of the three inputs changed in the future scenario (VMT, vehicle age, and vehicle speeds). Figure 4 shows the relative contributions of each of these three inputs along with the overall percentage difference in mass emissions due to all future changes. Adjustments applied to volumes in the future scenario yield a small predicted increase (less than 1%) in VMT during both the morning and the afternoon. Therefore, all pollutants show an increase in mass emissions due to changes in VMT; however, the magnitudes of these increases are different because of the varying emission rates for each pollutant. Changes in vehicle age and speeds have a less-consistent effect on mass emissions due to different relationships with emission rates for each pollutant built into the MOBILE model. The newer vehicle ages applied in the future scenario reduce the mass emissions for HC, CO, and NO x but have no effect on PM 2.5 and PM 10 emissions. Changes to vehicle speeds have varying effects on mass emissions for HC, CO, and NO x but have no effect on PM 2.5 and PM 10 emissions. Putting Predicted Emissions Changes into Context To put the predicted emissions changes into a broader context, they are presented on the right side of Table 4 as the equivalent change in total pounds per day, VMT per day, and total number of vehicles taken off or put onto the road. PM 2.5 a.m. data, which exhibited the largest increase, are examined here as an example of the magnitude and insignificance of these increases in emissions. These data show that emissions will increase by only 37 lb/day in relation to the total 16,182 lb of PM 2.5 emitted per day on the facility before implementing HOT lanes. This increase is equivalent to about 67,000 additional VMT per day of the total of more than 8 million daily VMT on the facility. It is also equivalent to 2,229 additional vehicles on the region s roadway, assuming vehicles travel 30 mi/day. This quantity constitutes a 0.83% increase in predicted daily emissions for the facility. While quantitative estimates of the MOBILE model s overall accuracy and uncertainty are not available, some studies estimate uncertainty for specific parts of the model. For example, one study estimated uncertainties in LDV exhaust emissions to be 20% to 40% for HC and 25% to 55% for NO x (9). The predicted emissions increases of less than 1% for the I-85 corridor in question should certainly be considered insignificant (a) because they are so much smaller than the uncertainties in the emissions model and (b) given that the project is extremely small relative to the total motor vehicle emissions budget (MVEB). Nevertheless, emissions from the I-85 HOT lane project must be directly assessed to demonstrate that they will not cause the MVEBs TABLE 4 2010 Mass Emissions on I-85 corridor in Atlanta and Equivalent Changes Time Peak Time and Direction All Times and Directions Scenario Base Future % Diff. Base Future % Diff. Equivalent Changes Units g/h g/h % g/h g/h % lb/day veh mi/day veh/day HC a.m. 44,675 43,915 1.70 64,059 63,299 1.19 26 95,398 3,180 p.m. 41,572 41,370 0.49 64,243 64,040 0.32 6 20,281 676 CO a.m. 945,099 948,444 0.35 1,412,701 1,416,046 0.24 113 19,015 634 p.m. 1,517,035 1,530,677 0.90 2,513,693 2,527,334 0.54 385 34,936 1,165 NO x a.m. 277,984 279,071 0.39 479,312 480,398 0.23 37 18,196 607 p.m. 256,196 257,471 0.50 424,874 426,149 0.30 36 19,318 644 PM 2.5 a.m. 6,277 6,359 1.32 9,848 9,930 0.84 3 66,866 2,229 p.m. 5,850 5,908 1.01 9,099 9,158 0.65 2 41,753 1,392 PM 10 a.m. 11,615 11,742 1.10 17,598 17,726 0.73 4 58,426 1,948 p.m. 10,627 10,732 0.99 16,470 16,576 0.64 3 41,427 1,381

Kall, Guensler, Rodgers, and Pandey 95 2.00 1.50 % Difference from Base Scenario 1.00 0.50 0.00-0.50-1.00 AM PM AM PM AM PM AM PM AM PM HC CO NOX PM2.5 PM10-1.50-2.00 Contributions VMT Only Age Only Speed Only All Future Changes FIGURE 4 Relative contributions to changes in mass emissions from individual inputs. for the Atlanta 8-h ozone nonattainment area to be exceeded. Conformity requires that the mass emissions from any transportation plan or project must be demonstrated to conform to the emissions budgets used in air quality planning. Under current regulatory interpretations, any predicted increase in emissions associated with plans or projects that results in the emissions budget being exceeded is considered to cause or contribute to a planning violation (irrespective of the uncertainty issues discussed earlier). Table 5 presents a comparison of the mass emissions for NO x (as one pollutant example) calculated for the base and future scenario of this study and the NO x MVEB from the 2006 SIP (4). The I-85 portion of the regional MVEB is isolated by means of a geographical information system analysis. Table 5 shows that both the base and the future scenarios fall well below the MVEB during both the morning and afternoon. While the amount below the budgets seems quite large at first, it is fairly consistent with the 75% below budget determined for Envision6, the transportation plan for the entire region. This comparison shows that the implementation of HOT lanes on I-85 will not cause the region to exceed its MVEB because the mass emissions results for I-85 fall well below the portion of the MVEBs allocated to the same portion of I-85 for both the base and the future scenarios. Therefore, the implementation of HOT lanes on I-85 should not violate the emissions budget requirements of the Transportation Conformity Rule. SUMMARY AND CONCLUSIONS This study completes a mass emissions analysis on I-85 in Atlanta for a base scenario that represents the current condition and a future scenario that represents the implementation of HOT lanes. Input factors considered for the analysis include vehicle activity, speeds, age distributions, and class distributions. The base scenario mainly uses information from a data collection effort by Georgia Tech during the summer of 2007 on the I-85 corridor, while the future scenario makes alterations to these data by using information from other cities that already have HOT lanes. The MOBILE-Matrix tool, recently developed by Georgia Tech, is used to run the emissions analysis with the input factors from these data sources. The results show very small predicted increases in mass emissions for NO x, CO, PM 2.5, TABLE 5 Comparison of I-85 Mass Emissions Results to MVEB Weighted Avg. NO X g/h Hourly VMT Emission Rate Scenario a.m. p.m. a.m. p.m. a.m. p.m. Base 479,312 424,874 524,475 503,415 0.914 0.844 Future 480,398 426,149 527,673 507,098 0.910 0.840 MVEB (I-85 portion) 987,431 1,136,822 554,162 641,473 1.782 1.772 Base as % of MVEB 48.54 37.37 94.64 78.48 51.29 47.62 Future as % of MVEB 48.65 37.49 95.22 79.05 51.09 47.42

96 Transportation Research Record 2123 and PM 10, and very small predicted decreases in mass emissions for HC along this corridor. The process used to conduct this analysis yields a number of lessons and conclusions: Locally collected corridor data, such as those gathered during the summer field study on I-85, can be used for better representation of corridor-specific conditions such as vehicle age distribution and class distribution. This option is preferable to the use of regional or national default values that can often significantly differ from local corridor values due to specific local conditions. For example, the vehicle class distributions (VMT fractions) for locations inside the I-285 perimeter in Atlanta have fewer heavy trucks and more LDVs than the national defaults due to the ban on through-truck traffic inside the perimeter. When an attempt is made to model a proposed HOT lane that does not yet exist, it is helpful to observe how conditions change in other cities after the implementation of a HOT lane. For example, this study made use of noted changes in vehicle speeds after the implementation of HOT lanes on I-394 in Minneapolis and changes in vehicle fleet characteristics associated with HOT lanes on SR91 in southern California. Predicted changes in mass vehicle emissions are influenced by changes in both VMT, which is due to vehicle activity, and the emission rate, which is due to changes in speed and vehicle fleet characteristics. For this study, emission rates are predicted to decrease or remain unchanged for the future scenario, and VMT are predicted to increase slightly for the future scenario. For four of the five pollutants, the predicted increase in VMT is modeled as having a greater impact on predicted emissions (causing a slight predicted increase in mass emissions). However, for HC, the predicted decrease in emission rates is modeled as having a larger net effect, causing a slight decrease in predicted mass emissions. It is important to determine whether the proposed project is likely to cause a conformity violation for the Atlanta 8-h ozone nonattainment area and the Atlanta PM 2.5 nonattainment area. The very small percentage change in predicted emissions along the corridor is well within the error bounds of current emission rate models. Hence, the small impact cannot be statistically differentiated from a zero predicted increase in emissions. This provides initial evidence that the conversion of HOV lanes to HOT lanes on I-85 will not significantly change the emissions that are modeled by the ARC for transportation conformity. The conformity regulation is currently interpreted such that any predicted increase in emissions (irrespective of modeling uncertainty) associated with plans or projects that result in an emissions budget being exceeded is considered to be a conformity violation. As such, the emissions changes associated with the project were compared with the regional MVEBs and to the allocated subregion (corridor) budgets to demonstrate that the implementation of HOT lanes on I-85 will not cause the region to exceed any of the MVEBs and will therefore have no significant impact. NEPA requires a detailed environmental analysis for proposed federal projects with a potentially significant environmental impact. In relation to air quality, the implementation of HOT lanes on I-85 has no potential impact due to the extremely small changes in mass vehicle emissions shown in this study. Therefore, this study makes a compelling case that further detailed analyses are not needed. REFERENCES 1. Guensler, R. L., K. K. Dixon, V. V. Elango, and S. Yoon. MOBILE Matrix: Application of Georgia Statewide Multimodal Transportation Planning Tool for Rural Areas. In Transportation Research Record: Journal of the Transportation Research Board, No. 1880, Transportation Research Board of the National Academies, Washington, D.C., 2004, pp. 83 89. 2. Guin, A., M. P. Hunter, and R. Guensler. Analysis of Reduction in Effective Capacities on High-Occupancy Vehicle Lanes Related to Traffic Behavior. In Transportation Research Record: Journal of the Transportation Research Board, No. 2065, Transportation Research Board of the National Academies, Washington, D.C., 2008, pp. 47 53. 3. I-394 MnPass Technical Evaluation. Cambridge Systematics. Nov. 2006. www.mnpass.org/pdfs/394mnpass_tech_eval.pdf. 4. Georgia s Draft Early Progress State Implementation Plan Revision for the Atlanta 8-Hour Ozone Nonattainment Area. Georgia Environmental Protection Division. Oct. 26, 2006. www.gaepd.org/files_pdf/plans/sip/ draft_early_progress_sip_links.html. 5. Barth, M., C. Malcolm, and G. Scora. Integrating a Comprehensive Modal Emissions Model into ATMIS Transportation Modeling Frameworks. California PATH Research Report. California Partners for Advanced Transit and Highways (PATH). Research Reports. Paper UCB-ITS-PRR- 2001-19. University of California, Berkeley, Aug. 2001. repositories.cdlib. org/its/path/reports/ucb-its-prr-2001-19. 6. Barth, M., F. An, T. Younglove, G. Scora, C. Levine, M. Ross, and T. Wenzel. NCHRP Web-Only Document 122: Development of a Comprehensive Modal Emissions Model. Transportation Research Board of the National Academies, Washington, D.C., April 2000. onlinepubs.trb.org/ onlinepubs/nchrp/nchrp_w122.pdf. 7. Miller, T. L., W. T. Davis, G. D. Reed, P. Doraiswamy, and A. Tang. Effect of County Level Income on Vehicle Age Distribution and Emissions. In Transportation Research Record: Journal of the Transportation Research Board, No. 1815, Transportation Research Board of the National Academies, Washington, D.C., 2002, pp. 47 53. 8. Yoon, S. A New Heavy-Duty Vehicle Visual Classification and Activity Estimation Method for Regional Mobile Source Emissions Modeling. PhD thesis. Aug. 2005. 9. Committee to Review EPA s Mobile Source Emissions Factor (MOBILE) Model. Modeling Mobile-Source Emissions. National Academy Press, Washington, D.C., 2000. The Transportation and Air Quality Committee sponsored publication of this paper.