Implications of Transient Mode Duration for Spatially Disaggregated High-Resolution Emission Inventory Studies

Transcription

1 TRANSPORTATION RESEARCH RECORD 1587 Paper No Implications of Transient Mode Duration for Spatially Disaggregated High-Resolution Emission Inventory Studies MOHAN VENIGALLA AND DON PICKRELL High resolution emission inventory studies, such as urban air-shed modeling, often use spatially disaggregated emission parameters in the form of link-specific emissions. A key input variable that influences operating mode mix on highway links is the transient mode duration (TMD) in which start-related emissions occur. The TMD variable also establishes the emission rates to be embedded in emission factor models. The basic emission rates built into the MOBILE model are based on a TMD of 505 seconds, which was based on the federal test procedure (FTP). For newer vehicles, it is indicated that the TMD is much less than 505 seconds. This study examines the effects of variations in TMD on the distribution of cold and hot transient VMT on network links, as well as their effect on overall emissions. A network assignment model developed to track operating modes of vehicles on network links was used on a large California city network. The MOBILE model was used to derive bag-specific emission rates. The experiment includes a series of traffic assignment runs with varying TMD and postprocessing of the assignment output. Link-specific distribution of operating mode mix and total emissions were derived. It is indicated that TMD significantly influences the distribution of operating modes, as well as total emissions. The results indicate that as the TMD decreases, the share of cold transient VMT from freeways diminishes, while that of the local roads increases. It was indicated that total emissions on various link groups may fall within a range of 90 to 300 percent of the total emissions derived from the present FTP duration of 505 seconds. It is recommended that the transient mode duration variable be reexamined for deriving emission rates to be embedded in the future versions of emission factor models. Traditionally, emission inventory studies are performed at a regional level or subarea level giving little consideration to details at a finer spatial resolution. Many studies have focused on temporal disaggregation of the inventories based on peak and off-peak periods and other seasonal factors. Since the passage of the Clean Air Act Amendment of 1990 and the Intermodal Surface Transportation Efficiency Act of 1991, spatially disaggregated modeling is evolving as a viable approach to addressing some of the new regulatory issues arising out of these acts. Also, the Environmental Protection Agency (EPA) mandated that serious carbon monoxide and serious, severe, and extreme ozone areas use network models to support conformity determination by January 1, The level of detail in a disaggregation approach using network models is based on spatial and temporal factors. Inventories measured on each urban network link would facilitate spatial grouping by functional class, geographic region, traffic analysis zone, corridor, subarea, and so forth. Time-of-day and seasonal factors determine the temporal disaggregation of emission inventories. Volpe National Transportation Systems Center, 55 Broadway, DTS-49, Cambridge, Mass Databases developed for and from travel demand modeling (TDM) processes are the primary sources of input to developing such disaggregate emission inventories. Travel parameters, such as total vehicle miles traveled (VMT) and average speed, are routinely used as proxies representing the travel patterns that determine emission rates over the entire study area. Recently, Allen and Davies (1) and Chatterjee et al. (2) adopted the traffic assignment technique for deriving vehicle operating modes for each link. It is argued whether the benefits of a disaggregate modeling effort are justified when pitted against the high costs of additional personnel and financial resources associated with this approach. A study by Cambridge Systematics (3) has concluded that the differences in emission estimates at various levels of disaggregation can, at best, be described as insignificant. However, this study did not take into account the variations in operating mode mix, namely, percentage of VMT in cold transient, hot transient, and stabilized modes of operation, by link. Some studies have used traffic assignment procedures to derive operating mode mix for each network link. In addition to inputs related to breakdown of start modes at trip origins, disaggregate modeling using network traffic assignment techniques requires that a transient mode duration (TMD) be specified. While questions remain about the effectiveness of disaggregate analysis for air quality modeling purposes, it is not known the degree to which the TMD affects the operating mode fractions, emission rates, and, hence, the total emissions. The purpose of this paper is to study the effects of various transient mode durations on the spatial distribution of VMT in various operating modes and total emissions. BACKGROUND Historically, EPA has defined a cold start as any start that occurs 4 hours or more after the end of the preceding trip for noncatalystequipped vehicles; and 1 hour or more after the end of the preceding trip for catalyst-equipped vehicles. Hot starts are those starts that occur less than 4 hours after the end of the preceding trip for noncatalyst-equipped vehicles and less than 1 hour after the end of preceding trip for catalyst-equipped vehicles (4). The duration associated with restarting the engine after the end of the preceding trip is called a cold-soak or simply a soak period. Before attaining a stabilized operating mode at which the rate of emissions is uniform and significantly lower, the vehicle will be operating either in a cold transient mode or hot transient mode, depending on the actual starting mode. The time duration associated with a vehicle turning from transient mode operation to stabilized mode operation is called a transient mode duration.

2 64 Paper No TRANSPORTATION RESEARCH RECORD 1587 Effect of TMD on Basic Emission Rates The rate of emissions of hydrocarbons (HC) and carbon monoxide (CO) is much higher during the initial few minutes of operation. Also, these rates are dependent on several environmental and vehicular factors. To provide a basis for analyzing the emission characteristics of motor vehicles in a comprehensive and consistent manner, a methodology was developed by EPA that is popularly known as the federal test procedure (FTP). This methodology establishes the standard for conducting tests of automotive emission characteristics. Under this test procedure, the tailpipe emissions are collected in three different bags. Bag 1 and Bag 3 collect emissions from cold- and hot-start operations, respectively. The FTP duration for these two bags is 505 sec, while Bag 2 emissions for stabilized conditions are collected for 866 sec. Basic emission rates (BER) embedded in the popular MOBILE model are derived from the contents of the three FTP test bags. The model makes adjustments to BER for weather, trip speed, operating mode fractions, vehicle mix, and so forth. The transient mode duration of 505 sec was originally adopted for FTP in the mid- 1970s as a time period in which most emissions will be stabilized. It was indicated experimentally that for many vehicles CO and HC emissions are stabilized much before 505 sec. In one set of tests, the effects of ambient temperature on CO and HC emission rates for six different vehicle configurations of 1970 production cars were tested (5). The testing involved measuring the emissions produced by each vehicle as it was operated from a cold condition through the Urban Dynamometer Driving Schedule, which was introduced in the FTP. From this experiment, it was observed that emission rate stabilizes between 200 sec and 300 sec of operations, indicating that the transient mode duration may be in this range for mid-1970 model vehicles. Springer and Dickinson (6) have indicated that for a modern air-puffed fuel injector car, the stabilization of the catalyst could occur as low as 70 sec. Presented in Figure 1 is the HC emission rate by time vis-à-vis engine operating conditions for a recent model year car. For the specific case presented in Figure 1, the engine begins to operate in stoichiometric conditions at about 70 sec, and the rate of HC emissions stabilize at around 140 sec. The trend is similar for a vast majority of new model year cars and for CO emissions. Despite this disagreement with actual rate of stabilization, emission factors are still derived based on the FTP drive cycle with a TMD of 505 sec. Thus, emissions in Bag 1 and Bag 3 are averaged over a longer period of time than necessary. This results in lower-thanusual average emission rates for the initial few seconds of operation. To counter this problem, it is necessary to reexamine the usage of the 505-sec duration for transient mode operation. Effect of TMD on Distribution of Operating Modes on Network In the absence of reliable field data or empirical studies on operating modes of vehicles, emission inventory studies have customarily adopted an operating mode fraction mix derived from the FTP drive cycle (4). The FTP mode mix indicates that 20.6 and 27.3 percent of the vehicles in the fleet represent cold and hot transient modes of operation, respectively, while the remaining 52.1 percent of the vehicles are in a hot-stabilized mode of operation. These operating mode fractions are widely being used in several corridor-level and areawide emission studies at different levels of precision. Despite the fact that the use of this mode mix is adequate for compliance purposes, the research community and practitioners alike have serious reservations about adopting the FTP operating mode mix for all cases. Ideally, the operating mode fractions values should be developed for varying situations. For example, these may be stratified by functional class of highways (freeways or expressways, principal arterials, etc.) and also geographic locations (central business district or CBD, fringe, suburban, etc.) in which highway segments are located. The emission rates of CO and HC estimated by the emission factor models are very sensitive to the variations in the inputs related to operating mode fractions. The concentrations of CO and HC are higher when the proportion of transient mode (especially the cold transient mode) VMT is higher. Venigalla et al. (7) established that FIGURE 1 Cold-start catalyst and HC behavior (6).

3 Venigalla and Pickrell Paper No the FTP mode mix that was derived based on a small sample of trips several years ago does not provide a true representation of current general driving patterns of the population. Venigalla (8) discussed an algorithm that performs this operation and the computer model TAPES (Traffic Assignment Program for Emission Studies) developed using this algorithm. The traffic assignment techniques use transient mode duration to assign parts of each trip into a transient mode or stabilized mode operation. The longer the duration, the higher the proportion of VMT that will be operating in a transient mode. Allen and Davies (1) and Chatterjee et al. (2) demonstrated the application of traffic assignment technique for deriving operating mode mix. Inputs to this assignment procedure include a trip exchange matrix, which is a composite table of trips for all trip purposes, network geometry, and zone-specific cold-and hot-start mode trips. Another key variable input is the transient mode duration, which is the main subject of this research. Chatterjee et al. used a TMD of 505 sec to be consistent with the FTP duration. Also, the FTP transient mode duration is used as the base case in this research. Stratified start modes of trips discussed by Venigalla et al. (7) were used to derive a composite number of cold and hot starts at each origin by adding the total number of starts in each mode across all trip purposes. The trip exchange matrix was developed by synthesizing trip matrices of all purposes, while accounting for appropriate vehicle occupancy ratios by trip purpose. The 1990 planning network for Sacramento, California, is used for the case study. The network has 875 zones and over 12,000 directional links. Trip tables for daily and a.m. peak period were used. A schematic representation of various steps involved in the study is presented in Figure 2. The process of isolating bag-specific emission rates from the MOBILE model, which is explained in the subsequent sections, is done concurrently with the assignment analysis. The results from Steps 1 and 2 above are used to derive total emissions on individual links. Each iterative assignment run of TAPES on the Sacramento network required approximately 20 min on an Intel Pentium based personal computer with a 60-MHz clock speed. Distribution of Operating Mode Mix Study Objective It follows from the foregoing discussion that the TMD for a typical vehicle is much less than 505 sec. Adopting a different TMD at the test level will have a significant effect on the emission rates. Since TMD is a key input parameter for traffic assignment procedures, distribution of operating mode fractions across the network is expected to vary significantly. In this study, we attempt to measure the impact of an alternative set of transient mode durations on the distribution of different operating modes on the network. Also by isolating the bag-specific emissions from the MOBILE output, we attempt to study the impact of variation in TMD on the estimated networkwide emissions. Since it is a general practice for many planning agencies to classify the road facilities by functional class and geographic location, the results are stratified by these two classes. However, the results can be aggregated in any meaningful classes or stratification levels. For example, links may be classified by number of lanes, median treatment, annual average daily traffic levels, and so forth. The results of this analysis are presented in Table 1 and 2 and are illustrated in Figure 3. The resulting trends confirm an intuitive belief that the portion of VMT operating in different modes varies widely between different functional classes of roads, as well as the location of the facility. It can be observed that higher proportions of VMT are operating in the hot-stabilized mode, followed by the cold transient mode and DESIGN OF EXPERIMENT The study is performed in three primary steps: 1. Conduct traffic assignment analyses for deriving distributions of operating mode fractions on a test network with varying TMD. 2. Isolate bag-specific emissions for various trip speeds from the MOBILE5a model. 3. Derive bag-specific and total CO, HC, and nitrogen oxide (NO X ) emissions on each link of the network. The TAPES model discussed by Venigalla (8) was used to design and conduct an experiment for studying the effects of various transient mode durations on the operating mode mix distribution. Eleven transient mode durations, ranging from 60 sec to 505 sec, were used in the experiment. One of the main inputs to the TAPES model is a zone-specific linear array of trips starting in cold (or hot) mode at each origin zone. When trip exchanges are broken down by trip purpose, it is preferable to derive zone-specific cold and hot starts by accounting for variations in the cold- and hot-start proportions by trip purpose. FIGURE 2 Study design.

4 66 Paper No TRANSPORTATION RESEARCH RECORD 1587 TABLE 1 Operatating Mode Distribution by Urban Location for A.M. Peak Period (Percentage) hot transient mode. A lower proportion of VMT is associated with the hot transient mode because the percentage of hot starts are fewer than the percentage of cold starts for trip purposes that make up a majority of the trip interchange matrix. As indicated by Figure 2, a higher proportion of VMT operates in cold transient mode for lower classes of roads. For example, collectors have, in general, a higher proportion of VMT operating in a cold transient mode than freeways or expressways. This can be expected as more trips originate near these facilities during the morning peak period. Freeways and principal arterials are, in general, major thoroughfares and are used largely by through traffic. Therefore, lower percentages of VMT operating in a cold transient mode on freeways (including expressways) and principal arterials may be justified. With regard to the location of roadway facilities, the CBD area, in general, has fewer VMT in the cold transient mode. Again, this trend conforms to a general trend that, in the morning peak hours, more trips are destined toward the CBD than originate in the CBD. By plotting Bag 3 distribution (not shown), it was observed that hot transient trips have a trend similar to that of cold transient trips. Freeways and expressways exhibit the least concentrations of hot transient trips, which are closely followed by principal arterials. Local streets and frontage roads are among those that have higher concentrations of hot transient trips. As would be expected, hot-stabilized VMT has a trend opposite that of cold transient VMT. The percentage of VMT operating in the hot-stabilized mode is the highest for freeways and expressways, which are closely followed by principal arterials. The proportion of VMT in the hot-stabilized mode of operation has a trend opposite that of the cold transient mode VMT. Higher classes of roads, such as freeways/expressways and principal arterials, exhibit higher concentrations of stabilized mode VMT. The trends are identical for

5 Venigalla and Pickrell Paper No TABLE 2 Operatating Mode Distribution by Facility Type for A.M. Peak Period (Percentage) both the a.m. peak period and the 24-hr period (not shown). However, the share of cold-start VMT is relatively less for the 24-hr analysis period. Effect of Transient Mode Duration The results indicate that the transient mode duration has a noticeable effect on the operating mode mix by functional class or urban location. The trends confirm an intuitive belief that the lower the transient mode duration, the lower the VMT operating in cold transient mode. Conversely, the stabilized portion of the VMT increases with a decrease in transient mode duration. In Figure 4, the contribution of cold-start VMT by facility type as a percentage of overall cold-start VMT is depicted. This figure reveals that the contribution of major arterials to the overall coldstart VMT can be observed as a steady decrease with an increase in transient mode duration. Up to about a 150-sec duration, it can be observed that minor arterials and collector roads increase their contribution to the overall cold-start VMT with an increase in transient mode duration. This trend is followed by a decrease in their contribution to cold-start VMT with a decrease in transient mode duration. On the other hand, freeways have a trend exactly opposite that of the arterials and collectors. The rate of increase in freeway contribution to cold-start VMT as a percentage of overall cold starts is rather steep with other facilities. Also, at a 300-sec or higher transient mode duration, the freeway contribution of cold-start VMT is much higher than the contribution from any other functional class of roads. In a modeling network, not every road segment in the urban network is represented. It is conceivable that the amount of travel represented by the local streets and other lower-order roads is

6 68 Paper No TRANSPORTATION RESEARCH RECORD 1587 FIGURE 3 Distribution of cold transient (Bag 1) VMT by facility type (a.m. peak period). significantly higher than what is represented in a typical travel demand modeling exercise. MODELING EMISSIONS From the discussion on the distribution of the operating mode mix, it is clear that the TMD significantly influences the spatial spread of various operating modes. However, it is important to know how it affects overall emissions. To estimate total emissions of CO, HC, and NO X, the TMD used in both the assignment and emission factor models should be compatible. Emission rates for a TMD other than 505 sec cannot be derived using the MOBILE model. Alternatively, a data source on second-by-second emissions for a large sample of automobiles can be used to derive appropriate emission rates. Such data were not available to the authors at the time this study was being performed. Also, modeling emissions with the MOBILE model requires that the trip speed be known. Tracking trip speeds and operating mode fractions at the same time becomes a computationally intense exercise for the traffic assignment procedures. To address these shortcomings in emission modeling, the following assumptions are made: 1. The composite emission rates output by the MOBILE model can be adjusted to different TMDs. 2. The Bag 2 emission rate will not change with variation TMD. 3. Link speeds may be used as close approximations to the trip speeds. FIGURE 4 Contributions to total Bag 1 emissions by facility type.

7 Venigalla and Pickrell Paper No Due to resource constraints, Assumption 1 is deemed acceptable. Since TMD pertains only to the start-related emissions, the second is a valid assumption to make. Assumption 3 is valid for Bag 2 emissions, while for Bag 1 and Bag 3, it falls short of being completely accurate. Given the complexity of simultaneously tracking trip speeds and operating mode fractions, the third assumption is also deemed acceptable. Isolating Bag-Specific Emissions From MOBILE5a For normal modeling exercises, the results from the traffic assignment model can be used as inputs to the MOBILE model for deriving emission rates. This procedure is valid as long as the TMD used for deriving the operating mode mix (traffic assignment) and the TMD used for deriving emission rates (MOBILE) are the same. Since the basic emission rates embedded in the MOBILE model are based on a TMD of 505 sec, the resulting emission rates need to be adjusted for any other TMD. It should be noted that the variation in TMD is assumed to result in a change in emission rates for Bag 1 and Bag 3 only. The output from the MOBILE model is a composite emission rate of all three bags. Therefore, adjusting the composite rates from the MOBILE model will adjust the rates for Bag 2, which is not appropriate. The alternative to this approach is to isolate bag-specific emissions from the model. The MOBILE model outputs a composite emission rate for CO, HC, and NO X based on the input parameters. The basic emission rate equation embedded in the model is as follows (9): BER D 1 ( = f 1 Bag 1+f2 Bag 3) + D 2 Bag where BER = composite FTP base emission rate (g/mi), Bag i = bag i emission rate (g/mi), (i = 1, 2, and 3), D1 = distance driven for Bag 1 and Bag 3 samples (FTP: D1 = 3.59 mi), D2 = distance driven for Bag 2 sample (FTP: D2 = 3.92 mi), f1 = factor representing the fraction of cold transient VMT in start-related emissions, and f2 = factor representing the fraction of hot transient VMT in start-related emissions. The MOBILE model computes the values of f1 and f2 based on the operating mode fractions supplied. The operating mode mix derived from the FTP drive cycle comprises 20.6 percent cold transient VMT and 27.3 percent hot transient VMT. Thus, for the FTP operating mode mix f1 = 0.43 and f2 = To isolate bag-specific emissions from the equation, one can supply appropriate operating mode mixes. If the model is supplied with a mode mix of 100 percent cold transient VMT and 0 percent hot transient VMT, the resulting emission rate is for Bag 1. Similarly, Bag 3 emissions can be isolated by equating hot transient mode VMT to 100 percent. By equating both Bag 1 and Bag 3 VMT to zero, Bag 2 emissions can be isolated. Networkwide Emissions Using the aforementioned procedure, bag-specific emission rates were isolated for a TMD of 505 sec and other MOBILE5a default conditions. The rates were derived for various trip speeds ranging from 5 mph to 75 mph. For sensitivity analysis, emission rates for TMD other than 505 sec were assumed to be a percentage of the MOBILE model s basic emission rates. For example, if the Bag 1 emission rate for CO is 46 g/mi, a 10 percent increase would indicate a rate of 50.6 g/mi, and a 200 percent emission rate increase would indicate a rate of 138 g/mi. Such categorization of emission variation between 10 percent and 200 percent was necessary because the exact emission rates under varying TMD are unknown. Also, a 200 percent increase over base emission rates (TMD = 505 sec) was assumed to represent an unlikely extreme case with the highest possible basic emission rates. The resulting emission rates were organized into look-up tables. From the traffic assignment analysis, VMT in each bag for each link was computed. Bag-specific emission rates for appropriate speed and TMD were then extracted from the look-up table for computing total emissions on each link. After completing a series of computations in this fashion, link-specific CO, HC, and NO X emissions were derived. It should be noted that areawide emissions are usually grossly overestimated with the current procedures. No attempt was made to model intersection delays, excess idle, acceleration, and deceleration. Although this practice leads to huge errors, it is consistent with other areawide emission inventory studies for comparing emission variations due to alternative TMD. As in the case of operating mode fractions, it would be preferable to study the results by stratified link groups. However, due to the existence of multiple levels of TMD and emission rates categories, interpretation of the results by further stratifying emissions into link groups becomes difficult. Therefore, only total networkwide emissions are discussed. Table 3 presents total networkwide emissions (in metric tons) of the three pollutant gases on the entire network for the a.m. peak period. The numbers from Table 3 were normalized and illustrated in Figures 5, 6, and 7 for CO, HC, and NO X emissions, respectively. It can be observed (Table 3) that for a TMD of 180 sec, the estimated CO emissions could be 8 percent lower than the current MOBILE emission estimates. Should the emission rates for this TMD be 200 percent higher than that of the MOBILE emission rates, the total emissions could be 23 percent higher. Total emissions for TMD less than 505 sec are always lower than the total MOBILE emissions when the same emission rates are used (100 percent case). The overall variation in total emissions, within the experimental setup, is found to range from 8.4 percent to +82 percent. For example, if there were no change in emission rates with a change in TMD, total CO emissions could be lower by as much as 10.6 percent (HC and NO X will be lower by as much as 8.9 and 8.4 percent, respectively). On the other extreme, if the TMD were to be 480 sec instead of the 505 sec used in current studies, total CO emissions could be higher by as much as 80 percent (HC and NO X emissions will be higher by at least 82 and 72 percent, respectively). From Figures 5 though 7, it appears that the potential for reduction in emissions is not as pronounced as the potential for higher emissions if an alternative TMD were to be used. This is mainly because emission rates for Bag 2, which represent a vast majority of the network travel, do not change with TMD. Even if Bag 1 and Bag 3 emissions were to be completely eliminated, emissions due to Bag 2 would still exist on the network. The trends also are similar for HC and NO X emissions. Feasible zones for emission estimates of CO, HC, and NO X are defined in Figures 5, 6, and 7. Each of these zones assumes that the TMD lies

8 70 Paper No TRANSPORTATION RESEARCH RECORD 1587 TABLE 3 Total Emissions on the Network for Various Emission Rates (in Metric Tons) between 150 and 240 seconds, and the emission rates are no more than 300 percent of the current MOBILE emission rates. These zones are of no practical importance for emission modeling purposes. Rather, they are expected to give an indication of the error in current emission estimates that are based on a TMD of 505 sec. CONCLUSIONS Variation in TMD is indicated to significantly affect the distribution of operating mode mix. Freeways are indicated to contribute lesser to the cold transient VMT if the TMD were lower than 505 sec. The contribution to total cold transient VMT is greater for lower road classes when the TMD is lower. Total emission estimates are also sensitive to the TMD variable. The overall variation in total emissions, within the experimental setup, is found to range from 8.4 to +82 percent. While the exact deviations from the current MOBILE emission estimates depend on actual TMD and corresponding emission rates, approximate ranges for error in emission estimates are defined. Based on these observations, the present emission estimates using the MOBILE model can, at best, be described as inaccurate. Therefore, there is a need to reexamine the transient mode duration adopted in the federal test procedure. It is recommended that the

9 Venigalla and Pickrell Paper No FIGURE 5 Total CO emissions for different TMD as a percentage of emissions for 505-second TMD. future versions of the emission factor models, such as MOBILE6, should be developed to derive emission rates reflecting a more realistic transient mode duration. Accurate determination of true TMD would require an experimental study with a scientific engineering study involving a number of test vehicles. The possibility of a wide variation in the duration raises the question, What is the appropriate value for the transient mode duration? The answer perhaps could be as simple as, The time it takes to stabilize for a typical car that is defined by the 85th percentile model year in the fleet. It should be noted that this study focused only on the networkwide emissions. Project-level emissions could vary much more significantly than the networkwide emissions because the VMT in each bag for a given set of links are very much dependent on TMD. The extent of variation in total emissions for a project-level analysis is currently being studied by the authors. FIGURE 6 Total HC emissions for different TMD as a percentage of emissions for 505-second TMD.

10 72 Paper No TRANSPORTATION RESEARCH RECORD 1587 FIGURE 7 Total NO X emissions for different TMD as a percentage of emissions for 505-second TMD. ACKNOWLEDGMENTS This research was sponsored by the Federal Highway Administration s Office of Environmental Planning. The authors express their sincere thanks to Richard Shoeneberg of FHWA for his guidance and support during the course of the study. REFERENCES 1. Allen, W. A. G. and G. W. Davies. A New Method for Estimating Cold Start VMT. Compendium of technical papers from ITE 63rd Annual Meeting, The Hague, Netherlands, Sept. 22, Chatterjee, A., P. M. Reddy, M. M. Venigalla, and T. Miller. Operating Mode Fractions on Urban Roads Derived by Traffic Assignment. In Transportation Research Record 1520, TRB, National Research Council, Washington, D.C., 1996, pp Data Aggregation Issues in the Application of the MOBILE Emissions Factor Model. Report prepared for U.S. Department of Transportation. Cambridge Systematics, MOBILE 5A User Guide. Environmental Protection Agency, May Midurski, T. P. Determination of Percentages of Vehicles Operating in the Cold Start Mode. Prepared for Office of Air Quality Planning and Standards, Environmental Protection Agency, Springer, K. J., and A. G. Dickinson. Car Cold-Start Hydrocarbon Control Considerations. Presented at Energy-Sources Technology Conference and Exhibition, American Society of Mechanical Engineers, Houston, Tex., Venigalla, M. M., T. Miller, and A. Chatterjee. Start Modes of Trips for Mobile Source Emissions Modeling. In Transportation Research Record 1472, TRB, National Research Council, Washington, D.C., 1995, pp Venigalla, M. M. A Network Assignment Based Approach to Modeling Mobile Source Emissions. Ph.D. dissertation. University of Tennessee, May Evaluation of MOBILE Vehicle Emission Model. Prepared for U.S. Department of Transportation. Sierra Research, Inc., June The opinions, findings, and conclusions presented in this paper are those of the authors and are not necessarily endorsed by FHWA. Publication of this paper sponsored by Committee on Transportation and Air Quality.