Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
69 C h A P T e R 4 The SmartGAP software was shared with three agencies who were asked to test the software by implementing it in their regions, while a parallel implementation and further testing were performed. The findings of the pilot tests are summa- rized here and recommendations for further enhancements to SmartGAP based on those findings are also presented. Pilot Test Objectives The pilot tests were intended to produce implementations of the SmartGAP software in three varying agency settings in order to provide a range of feedback on the usability and use- fulness of the software. The three agencies that agreed to par- ticipate in the pilot tests were: ⢠Thurston Regional Planning Council (TRPC); ⢠Atlanta Regional Commission (ARC); and ⢠The Maryland DOT. The agencies were selected to represent a small- to medium- sized MPO (TRPC falls into this category), a large MPO (ARC falls into this category), and a department of transportation (the Maryland DOT falls into this category). The three catego- ries were designed to represent a range of institutional capabil- ity and planning needs that covers that of the target audience for SmartGAP. The specific objective of the pilot tests that was communi- cated to the participating agencies was to apply the software so that the following could be better understood: ⢠The usability of the software; ⢠The complexity of and any difficulties or problems with developing input data; ⢠The usefulness and clarity of the output metrics produced by the software; and ⢠The reasonableness of the results. In addition, an objective of the pilot tests was to generate feedback from the software users that would inform the final updates to the software and the userâs guide that took place as part of this project, and to identify suggestions for future updates and features that could be added to software after this project has been completed. Pilot Test Process The pilot tests took began with a webinar to introduce the three agencies to SmartGAP. The webinar described the objec- tives of the pilot tests, provided an overview of the SmartGAP model, discussed the development of the input data, and included a demonstration of how to use the software. Follow- ing the webinar, the agencies were provided with the software, a draft of the userâs guide, and preprocessed Census popula- tion and County Business Patterns data that simplified the creation of some of the base year model inputs. The agencies were asked to accomplish the following tasks and to provide feedback on their experience at each step: ⢠Install the software and successfully run the demonstration model included with the software; ⢠Develop model inputs for their region; and ⢠Run eight standard scenarios and submit the results. The set of eight standard scenarios were devised so that each agency would evaluate a range of policies that tested how the model represented changes in transportation sup- ply, changes in policy assumptions such as travel demand management policies, changes in land use allocation assump- tions, and combinations of those three types of inputs. Ask- ing each agency to test the same set of eight scenarios was intended to allow for comparisons of the results across the three agencies. The design of the eight scenarios is shown in Table 4.1: ⢠Scenario 1 is the baseline scenario, which was intended to be the agencyâs expected future for their region, assuming exist- ing policies such as those embodied in their long-range Pilot Tests
70 plans. The remaining seven scenarios then introduce some change from that baseline. ⢠Scenarios 2 and 3 evaluate the effects of changes in trans- portation supplyâtesting an increase in transit services and highway construction, respectively. ⢠Scenario 4 tests the impact of a transportation system man- agement policy, where additional ITS is added to the regions highway system to improve traffic flow by managing inci- dents and thereby reduce congestion. ⢠Scenarios 5, 6, and 7 alter the allocation of future growth in housing and commercial development in the region, by moving increasingly larger proportions of that growth from the suburban area type to the close-in community and urban core area types to test the impacts of locating development is denser, more accessible locales. ⢠Scenario 8 was designed to evaluate how the model com- bines the effects of several changes, in this case a large shift in the land use allocation, a change in transportation supply, and additional ITS provision. Over the course of the pilot test period, the agencies were provided with varying degrees of assistance. This included telephone calls; e-mail exchanges, reviews, and corrections to input files; and review of outputs. At the end of the pilot tests, the agencies were asked to provide input and output files for the scenarios that they had run, and written feedback on their experiences. A fourth implementation of SmartGAP was developed in parallel to the three agency implementations. This implemen- tation, based on the Portland metropolitan region, was used for model testing and to provide a fourth set of results from the standard scenarios. The intensive testing that was carried out early in the pilot test period resulted in the release of two new versions of SmartGAP to the three agencies. The agencies all used the third version of SmartGAP for the production of the final pilot test results presented in this section. Maryland Department of Transportation Agency Introduction The Maryland DOT is the statewide agency in Maryland responsible for planning, building, operating, and maintain- ing the stateâs transportation network. The Maryland DOT is responsible for the entire state of Maryland, which com- prises 24 counties and a population of 5.8 million people. Rather than using SmartGAP to evaluate the entire state, the Maryland DOT elected to model two separate counties, Montgomery County and Cecil County. Montgomery County is a populous county situated north of Washington, D.C. In 2005 (the base year that the Mary- land DOT used for modeling purposes) the population was 975,000, and the projected population in 2035 (the future year used for modeling purposes) is 1,117,000. This repre- sents a relatively slow rate of population growth of 20%. Cecil County is a more rural county in the northeast corner of Maryland. Its 2005 population was 100,000 and its 2035 projected population is 170,000, which represents growth of 70%, a much higher rate of population growth than for Montgomery County. The relative locations of Montgomery and Cecil counties are shown in Figure 4.1. Development of Model Inputs The Maryland DOT developed local inputs for two counties, Montgomery County and Cecil County. They did not employ a Table 4.1. Scenarios for Pilot Testing Scenario Land Use Transportation Policy 1. Baseline Baseline Baseline Baseline 2. Increase Transit Supply Baseline +0% in Transit Supply Baseline 3. Increase Roadway Supply Baseline +20% in Roadway Supply Baseline 4. Add ITS Baseline Baseline +20% in Lane Miles with ITS 5. Shift 10% Growth to More Dense Areas Shift 10% Pop, Emp to Close-in Community, 10% to Urban Core, from Suburban Area Baseline Baseline 6. Shift 20% Growth to More Dense Areas Shift 20% Pop, Emp to Close-in Community, 20% to Urban Core, from Suburban Area Baseline Baseline 7. Shift 30% Growth to More Dense Areas Shift 30% Pop, Emp to Close-in Community, 30% to Urban Core, from Suburban Area Baseline Baseline 8. Shift 30% Growth to More Dense Areas and Add ITS and Transit Supply Shift 30% Pop, Emp to Close-in Community, 30% to Urban Core, from Suburban Area +20% in Transit Supply +20% in Lane Miles with ITS Note: Pop = population; Emp = employment.
71 complex, GIS-based, place type allocation process such as that described in the summary of the ARC pilot test. However, the general differences in existing and expected future land use pat- terns between the two counties were represented in their input files. The graph on the left in Figure 4.2 compares population by area type for Cecil County and Montgomery County. Mont- gomery County is more largely suburban with a significant proportion of people living in areas that the Maryland DOT identified as close-in communities and urban cores, while Cecil Countyâs population lives in predominantly rural and suburban areas. The employment comparison between the two counties (see the graph on the right) shows a similar difference in the distribution, with a much higher proportion of employment in Montgomery County in more urban area types. Figure 4.1. Map of Montgomery and Cecil counties, Maryland. Figure 4.2. Summaries of 2035 population and employment by area type for Cecil and Montgomery counties (percentage of total county population and employment).
72 Scenario Testing Results The Maryland DOT provided inputs for the two counties and completed a full set of eight standard scenarios runs for each county. Figure 4.3 compares the changes in daily VMT by sce- nario for the two counties that were modeled, in the form of an index chart with the base scenario set to zero and the values for other scenarios expressed as percentage changes relative to the base scenario. In the case of Cecil County (to the left), no transit service was modeled and so Scenario 2 was not included (and Scenario 8 only differs from Scenario 7 in its inclusion of addi- tional ITS for incident management of the countyâs highways). Cecil County is predicted to have proportionally higher growth than Montgomery County (shown to the right), and so smart growth policies that are implemented between 2005 and 2035 have larger potential effects: Scenario 7, where approxi- mately 30% of the predicted growth in suburban areas is moved to close-in communities and urban core area types results in an a reduction of 8% VMT compared to the base scenario. The provision of additional transportation supply in the form of more roads (Scenario 3) has relatively little impact on VMT in Cecil County, indicating that its relatively rural and uncongested road system is imposing few constraints on travel. Montgomery County is relatively more developed than Cecil County and less growth is predicted, so the impacts of reallocating future growth have less overall impact. Scenario 7, where approximately 30% of the predicted growth in sub- urban areas is moved to close-in communities and urban core area types results in a reduction of VMT that is between 1% and 1.5%, a much smaller impact than in Cecil County. Increasing transit services was tested in Scenario 2, and resulted in a daily VMT reduction of more than 0.5%. Sce- nario 8, which tests the combined effect of transit service improvements and smart growth land use policies, resulted in a 2% reduction in daily VMT compared to the base scenario. SmartGAP includes various performance metrics that describe aspects of livability, including the number of traffic accidents and the amount of walking. The number of acci- dents is based on rates that are in terms of accidents per mil- lion miles of VMT, so the relative change in each accident severity category tracks the changes in daily VMT shown above. The percentage change in accidents in Montgomery County by accident severity is shown in Figure 4.4. Montgomery County sees a 2% reduction in accidents for Scenario 8, which produced the largest reduction in daily VMT. Because Scenario 3 (increase in transportation supply) leads to an increase in daily VMT, it also leads to an increase in accidents. This is only apparent in the injury and property accident severity categories. The number of accidents in each category is calculated as an integer, and because the number of fatal accidents is relatively small, a relatively large change in daily VMT is required to change the number of fatal accidents. The walking metric is the amount of walking above or below a common zero point (based on the expected amount of walking by residents of the suburban TOD place type) that will take place by residents of new housing and employees of new jobs. Therefore, it is only indicative of the effect of newly developed land uses on the people who live and work in them and not on any (possible) secondary effects on walking by residents and employees in existing areas. Figure 4.5 shows a Cecil Montgomery Figure 4.3. Comparison of percentage change in daily VMT from the base by scenario for Cecil and Montgomery counties. Note that there is no transit Scenario 2 run in Cecil County.
73 comparison of the walking metric for the full set of standard scenarios for each of Cecil County (to the left) and Mont- gomery County (to the right). The metric is in term of a pro- portional change in walking relative to the zero point of development taking place (on average) in the suburban TOD place type. For Cecil County, the base scenario is a general continuance of development in rural and suburban area types, which are in general less walkable than the suburban TOD place type and so the scenario shows in excess of a 10% reduction in walking among new residents and workers in the county. For scenarios with the same allocation of future resi- dential and employment development, the metric is the same, indicating that (as designed) it is only sensitive to land use changes and does not measure possible changes in walking that may results from changes in transportation supply. As land use growth is shifted to more walkable (more urban) place types in Scenarios 5, 6, and 7, the amount of walking by new residents and employees increases. In Scenario 7, growth is taking (on average) in place types that are more walkable than the suburban TOD place type, and so the walking metric is positive. A comparison between Scenario 7 and the base scenario shows around a 15% increase in the amount of walking by new residents and employees. The range of the change in the amount of walking between the base scenario and Scenario 7 by new residents and employees is similar for Montgomery County, which is expected, given a similar shift in the land use allocation. Of note is that all of the scenarios return a positive walking metric, indicating that even in the base scenario with growth allocated in least walkable manner, on average growth is still predicted to take place in place types that are more walkable than the suburban TOD place type. Agency Comments In addition to providing a complete set of input files for both Montgomery and Cecil counties, the Maryland DOT provided additional feedback on SmartGAP. The Maryland DOT installed the software locally on a desktop computer and was Montgomery Figure 4.4. Comparison of percentage change in accidents by severity for standard scenarios for Montgomery County. Cecil Montgomery Figure 4.5. Percentage change relative to the suburban TOD place type in walking metric for Cecil and Montgomery counties. Note that there is no transit Scenario 2 run in Cecil County.
74 able to successfully run the demonstration scenarios. Follow- ing some assistance, the Maryland DOT created input data for the two counties that they chose to study. The Montgomery County implementation, with a population of approximately 1 million, has run times of around 20 minutes, while the much smaller Cecil County only takes a couple of minutes to run. One aspect of the pilot test that caused some difficulty for Maryland DOT staff was receiving software and transmitting results. The Maryland DOTâs computer network security pre- vents access to external FTP sites and prevents receipt of zipped files attached to e-mail. The Maryland DOT provided other feedback on the pilot tests as well: ⢠Software installation. The Maryland DOT found that instal- lation of software is easy as the steps are clearly outlined in the userâs guide. ⢠Development of input files. The Maryland DOT also stated that the input file preparation was easy to follow using the descriptions in the userâs guide. For the employment data (employment.csv) input, the DOT recommended included more information to create area specific (say for different counties) employment files. The DOT did find that the input file formatting and naming is very precise and can be difficult to debug if errors are made. ⢠Connections with travel demand models. The Maryland DOT recommended that there should be some guidance or methodology described so that regions with travel demand models can use their standard model input/output files for better and easier representation of transportation supply and travel demand. ⢠Adjustment and calibration of the model. The Maryland DOT commented that it would be interesting to investigate how to calibrate each of the individual modules and pro- vide guidance on this issue. ⢠Overall. The Maryland DOT considered that the SmartGAP software offers a great tool to perform high-level scenario planning work with macroscopic formulations. In terms of applicability, the Maryland DOT commented that SmartGAP should act as a good resource for preliminary âwhat-ifâ analysis for agencies, particularly smaller MPOs and local jurisdictions without advanced travel demand models, while bigger MPOs and state agencies can use this tool for prescreening policy scenarios before undertaking extensive travel demand modeling exercises that are resource inten- sive. SmartGAP can help short-list a longer list of scenarios to a reasonable number with relatively less effort. Atlanta Regional Commission Agency Introduction ARC is the regional planning agency for a 10-county area in Georgia, which includes the City of Atlanta. ARC also covers a larger, 20-county area for air quality purposes; the ARC Travel Demand Model covers the 20-county area. It is this larger 20-county region that ARC used as the model region for the SmartGAP pilot test. The 20-county area is shown in Figure 4.6. Figure 4.6. ARC 20-county region used for pilot testing SmartGAP.
75 The ARC 20-county area is a very large region, with a 2010 (base year) population of 5.3 million people and a 2040 (future year) projected population of 8.3 million people. This projection represents population growth of 57%. In 2010, there were 2.1 million jobs in the region, with growth of 68% projected in 2040, giving a total of 3.5 mil- lion jobs. Development of Model Inputs The ARC provided a detailed description of its approach to developing the model input data. In general, it followed a somewhat detailed process to derive input data from land use data as presented in its Unified Growth Policy Map (UGPM), and from its regional travel demand model. It developed heuristics to align its land use with the 13 place types that SmartGAP uses. Population and Jobs by Place Type (place_type_ existing.csv and place_type_growth.csv) The conversion of land use data to the place type scheme used in SmartGAP involved taking ARCâs UGPM areas and converting them to the 13 SmartGAP place types: 1. The first step was to allocate the UGPM areas to the four area types used in SmartGAP. The urban core area type includes region core, region employment centers and Aero tropolis UGPM areas; close-in community includes maturing neighborhoods; suburban includes developing suburbs and established suburbs; and rural includes rural areas and developing rural. 2. The ARC traffic analysis zone (TAZ) system was overlaid with the area types and the centroid of the TAZ was used to determine its area type. 3. The SmartGAP development type, the other dimension of the place type matrix, which included residential, mixed-use, employment, and TOD development types was determined for each TAZ by using the base year per- centage of the TAZâs employment in relation to the total of the population and employment in the TAZ. The mix between the employment and employment was used to determine the TAZâs development type using the follow- ing cut points: ⢠Residential: <33.33% ⢠Mixed Use: 33.33% to 66.67% ⢠Employment: >66.67% 4. Only one TAZ was determined to be TOD as a develop- ment type, Lindbergh Center, in the urban core area type. 5. The combination of the area type and the development type was then used to allocate all TAZs to one of the 13 place types. 6. The 2010 TAZ employment and population totals were summed by the 13 place types and then scaled to total 1 for both employment and population as called for by the file format for place_type_existing.csv. 7. The population and employment growth amounts between 2010 and 2040 were determined for the 13 place types and were scaled to total 1 for both employment and population as called for by the file format for place_type_ growth.csv. Figure 4.7 shows summaries of 2040 population (on the left) and employment (on the right) by area type for the ARC Figure 4.7. ARC summaries of 2040 population and employment by area type.
76 region base scenario (i.e., the expected future described in its UGPM), as produced by SmartGAP based on the two place type input files. About half of the population is expected to live in suburban areas in 2040, with 40% split between the two denser, more urban area types, and the remainder in rural areas. Employment is more heavily concentrated in the urban core. Figure 4.8 shows similar summaries of 2040 pop- ulation and employment, this time by development type. The charts indicate the level of mixing of residential and employment locations, with approximately 40% of each land use located in the residential and employment devel- opment types, respectively: approximately 20% in the mixed-use areas and 20% in the opposite development type (i.e., residential development in employment areas and vice versa). There is relatively little existing or planned TOD development in the region. Base Daily Vehicle Miles Traveled (base_vmt.csv) This input file includes the total light vehicle daily VMT in the region and the proportion that takes place on freeway and arterial roads. To develop the light vehicle VMT, ARC obtained the single occupant vehicle, high occupancy toll, and drive-to-transit VMTâs from the ARC 2010 Plan 2040 Model Summary. These VMTs were summed together and displayed in thousands of miles, as required by the file format of base_vmt.csv. To develop the freeway and arterial percent- age of light vehicle VMT, the ARC summarized VMT by facil- ity type for from the loaded network TOTAL10 in its travel demand model, and then aggregated it to freeway, arterials, and other roads. The freeway and arterial VMTs were then added and convert to a percentage of the total VMT. Truck and Bus Vehicle Miles Traveled (truck_bus_vmt.csv) This input file includes the split of VMT by bus and truck that takes place on freeways, arterials, and other roads, and includes the proportion of total VMT in the region that is driven by trucks. The data were developed by ARC using its 2010 Plan 2040 model. To summarize the bus data, ARC used data on transit buses by line joined with the loaded highway network and followed these steps: 1. Used the networkâs facility type attribute to create total distance of freeways, arterials and other roads by bus line. 2. Computed bus VMT by freeway, arterial, and other: ⢠Number of Local Buses by Peak = 8 hours â 60/peak headway. ⢠Number of Express Buses by Peak = 6 hours â 60/peak headway. ⢠Number of Local Buses by Off Peak = 10 hours â 60/peak headway. ⢠Number of Express Buses by Off Peak = 2 hours â 60/peak headway. ⢠If a Local Bus, Total Number of Buses by Line = Num- ber of Local Buses by Peak + Number of Express Buses by Peak. ⢠If an Express Bus, Total Number of Buses by Line = Number of Local Buses by Peak + Number of Express Buses by Peak. ⢠Total Bus VMT by Line = Total Line Distance â Total Number of Buses by Line. ⢠Total Bus VMT is the sum of all Total Bus VMT by Line. Figure 4.8. ARC summaries of 2040 population and employment by development type.
77 ⢠Total Bus VMT by Freeway = Total Bus VMT â (Freeway Mileage/Total Mileage). ⢠Total Bus VMT by Arterial = Total Bus VMT â (Arterial Mileage/Total Mileage). ⢠Total Bus VMT by Other = Total Bus VMT â (Other Mileage/Total Mileage). Peak headway is the number of minutes in the peak period divided by the average number of buses in the peak period. ARC computed truck VMT by freeway, arterial, and other roads by using the following steps: 1. From the 2010 loaded highway network, Truck VMT by Segment = length of the segment â volume of trucks. 2. Summarized all Truck VMT by facility type: ⢠Truck VMT Freeway % = Truck VMT Freeway/Truck VMT Total. ⢠Truck VMT Arterial % = Truck VMT Arterial/Truck VMT Total. ⢠Truck VMT Other % = Truck VMT Other/Truck VMT Total. 3. The overall Truck VMT percentage of total VMT was obtained from the ARC 2010 Plan 2040 Model Summary, Truck VMT Percentage = (Commercial Vehicle VMT + Medium Truck VMT + Heavy Truck VMT)/Total Daily VMT. Auto and Transit Trips per Capita (trips_per_cap.csv) This input file contains average number of auto and transit trips per day per person in the region. ARC obtained population, total vehicle trips, and total transit trips from the ARC 2010 Plan 2040 Model Summary, and calculated the two data items as follows: 1. Auto Transit Trips per Capita = Total Vehicle Trips/ Population. 2. Transit Trips per Capita = Total Transit Trips/Population. Scenario Testing Results ARC successfully installed the software in a network location, developed input data for their region as described above, ran the eight standard scenarios, and provided a complete set of results for the scenarios. The three scenarios that involved alternative land use assumptions were Scenarios 5, 6, and 7. The proportions of population and employment by area type are shown in Figure 4.9. ARC chose to define relatively similar changes between Scenarios 5, 6, and 7 in terms of the reallo- cation of population, with larger differences in the location of employment growth. All three scenarios embody the objec- tive of these test scenarios: to locate increasingly higher pro- portions of growth to denser and more urban place types. The direct travel performance metrics presented by SmartGAP include daily VMT, vehicle hours of travel and delay vehicle hours. Figure 4.10 shows daily VMT by scenario, in the form of an index chart with the base scenario set to zero and the values for other scenarios expressed as percentage changes relative to the base scenario. The chart shows that in Scenario 2, an increase in transit services leads to a reduction in daily VMT, in this case by a little more than 1%. Scenario 3, Figure 4.9. ARC percentages of 2040 population and employment by area type for base scenario and Scenarios 5, 6, and 7. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.
78 where road supply is increased, induces an increase in daily VMT. Scenario 4, where additional highway lane miles are provided with the ITS for incident management, does not affect daily VMT as the ITS policy affects the calculation of policy-adjusted congestion, which is after the final calcula- tion of travel demand. Scenarios 5, 6, and 7, show increasingly larger reductions in VMT as more and more growth is located in denser, more urban area types, culminating in an almost 5% reduction in VMT in Scenario 7. Combining the land use allocation in Scenario 7 with an increase in transit services, gives a VMT reduction in Scenario 8 that approaches 6%. The changes appear to be directionally consistent and reasonable in magnitude. Figure 4.11 shows both a comparison of changes in total vehicle hours for the eight standard scenarios (to the left) and a comparison of changes in delayed vehicle hours (to the right). Scenario 2 (increase in transit service) and Scenarios 5, 6, and 7 (land use growth shifts to more urban areas) shows reductions in vehicle hours that follow the patterns of reduc- tions in VMT. More striking, however, are the changes in Sce- nario 3 (increase in road supply) and Scenario 4 (more ITS for incident management). Both scenarios model changes that decrease the effects of congestion, with the first increas- ing capacity (and while some of that capacity is used up by induced demand, not all of it is) and the second improving traffic flow given the same capacity. In both scenarios, there is a significant reduction in congestion, with an almost 25% reduction in hours of delay in Scenario 3 and more than 15% reduction in Scenario 4. These translate to overall reduction in vehicle hours of 4% and more than 3%, respectively. Agency Comments In addition to providing detailed descriptions of their input data development process and a complete set of inputs files and results for the eight standard scenarios, ARC provided some additional feedback on SmartGAP: ⢠Input data development. ARC found some of the input development to be easy and some to be more difficult to Figure 4.10. ARC percentage change from the base of daily VMT for eight scenarios. Figure 4.11. ARC percentage reduction of vehicle hours and delayed vehicle hours by scenario.
79 obtain or calculate. The processes ARC followed to allocate land use to place types and to calculate the VMT by facility type inputs based on travel model inputs were somewhat time-consuming. One of the policy tests, which fell outside the eight standard scenarios, was travel demand manage- ment policies. ARC expressed difficulty in translating their detailed household travel survey results, that categorized work schedules into many categories, into the simpler categories used to represent compressed work schedules in SmartGAP. ⢠Software installation. ARC faced some initial problems when trying to install R and SmartGAP on a desktop with- out admin rights, but was able to install R and SmartGAP on a flash drive and copy everything to a folder on a desk- top or a server with user rights. ARC was able to install R and SmartGap easily on a server with admin rights. ⢠Running the software. ARC found that the model would not run to completion on a desktop with 2GB of RAM due to insufficient memory, but it completed with no problem when installed a server with more RAM. ⢠Software performance. ARC found that each scenario took approximately 1 hour and 45 minutes to run, and gener- ated approximately 850 MB of data. ⢠Userâs guide content. ARC commented that the content of the userâs guide was helpful for installing and using the software. ⢠Other comments. ARC found that there are many policies that SmartGAP could test that cannot be evaluated with the current version of their travel demand model. Thurston Regional Planning Council Agency Introduction Thurston Regional Planning Council (TRPC) is the regional council of governments and MPO for Thurston County, Wash- ington, which includes Washingtonâs capital, Olympia. The region the TRPC chose for their implementation of SmartGAP covers the whole of their jurisdiction, which is the single county of Thurston. Thurston Countyâs population in 2010 (the base year used by TRPC) was 250,000 and the projected popula- tion in 2040 (the future year used by TRPC) is 425,000, which represents population growth of 69%. The 2010 employment in Thurston County was 130,000, with projected growth by 2040 of 100%. Figure 4.12 shows the location and boundaries of Thurston County. Development of Model Inputs TRPC developed a complete set of inputs for SmartGAP using local data. They followed a GIS-based process very similar to that used by ARC to develop the existing and future baseline allocation of land uses to place types. The results of the process are shown in Figure 4.13. The distribution of population by area type (to the left) in the base scenario is focused on the suburban area type, which accounts for 65% of the population in 2040, with 20% in rural areas, 10% in close-in communities, and only around 2% in the urban core. The distribution of employment (shown to the right) is slightly more even across the area types, with around 50% in suburban, 25% in close-in communities, and 15% in the urban core. Figure 4.14 shows the distribution of population (to the left) and employment (to the right) by development type. The majority of the population is in primarily resi- dential development types, with the largest proportion of employment (approximately a third) in mixed-use areas and slightly smaller proportions in both employment and residential development types. TRPC elected to augment the preprocessed employment data that they were provided with (based on County Business Patterns data) with additional records to reflect the employ- ment types that are not covered by those data. Specifically, TRPC added employment in government, which is a very important element of employment in Olympia, the capital of Washington. Scenario Testing Results TRPC successfully installed the software in a network loca- tion to allow sharing of access among several staff, developed input data for their region, ran the eight standard scenarios, and provided a complete set of results for the scenarios. The three scenarios that involved alternative land use assump- tions were Scenarios 5, 6, and 7. The proportions of popula- tion by area type (to the left) and development type (to the right) are shown in Figure 4.15. TRPC chose to reallocate Figure 4.12. TRPC region used for pilot testing of SmartGAP.
80 population from the suburban area type to the close-in com- munity area type, and from the residential development type to the mixed-use development type. They followed a similar approach to the allocation of employment (except that the reduction was made in the employment development type). TRPC did not allocate any population or employment to the TOD development type. Several of the direct travel impacts and the financial and economic impacts that are related to them are only sensitive to land use allocation changes and not to the transportation supply or other policy changes that were tested in the eight standard scenarios. Figure 4.16 shows a comparison of transit trips (to the left) and vehicle trips (to the right) for the base scenario and the three scenarios that include land use changes (Scenarios 5, 6, and 7). The transit trip metric increases transit use when more growth is allocated to transit accessible locations (i.e., the close-in community area type and mixed-use development type to which TRPC allocated more population and employment). The results show an increase in transit trips of around 3% among new residents Figure 4.13. TRPC percentages of 2040 population and employment by area type. Figure 4.14. TRPC percentages of 2040 population and employment by development type.
81 and employees in Scenario 7 relative to the base land use allo- cation. The vehicle trip metric shows a decrease in the num- ber of vehicle trips made by new residents and employees when more growth is allocated to area types and develop- ment types that are more transit accessible and more walk- able, as the opportunity to make trips by modes other than car increases. The results show this trend, with Scenario 7 showing a reduction in vehicle trips of close to 1% relative to the base scenario. The transit operating costs and capital costs performance metrics are calculated using rates that are proportional, and (as with the transit trip metrics) only measure changes that relate to changes in land use allocations. Therefore, the pat- tern of changes in costs is intended to follow the same pattern Figure 4.15. TRPC 2040 population and employment by area type for base scenario and Scenarios 5, 6, and 7. Figure 4.16. TRPC percentage changes in transit and vehicle trips for base and Scenarios 5, 6, and 7.
82 of changes in the number of trips. Figure 4.17 demonstrates that the performance metrics behave as intended. Agency Comments In addition to providing a complete set of input files and results for the eight standard scenarios, TRPC provided addi- tional information on its experiences during the pilot tests and feedback on SmartGAP: ⢠Software installation. TRPC installed the software locally and then installed the software in a network location. TRPC was able to successfully run the demonstration scenarios from both locations. ⢠Employment data. TRPC found that the preprocessed County Business Pattern employment data supplied with software does not cover enough of the total employment in its region to be accurate. It omits government employ- ment, which is important in Olympia, the state capital, and so requires augmentation with additional records to cover omitted employment types. ⢠ITS strategy. TRPC felt that the ITS strategy/policy is diffi- cult to understand and interpret on the basis of its descrip- tion in the userâs guide and its effects on the performance metrics. ⢠Software performance. TRPC found that software is very easy to prepare input tables for and to run, and runs very quickly. For the TRPC implementation of SmartGAP, sce- narios take approximately 4 minutes on a relatively new desktop. ⢠Software usability. TRPC reported that it experimented with editing the inputs files in the file system rather through the GUI, but found that this caused some problems due to mistakes or typos in the file causing errors when the model was run. The GUI layout and the legibility of output charts can be affected by long scenario names. ⢠Interpretation of results. TRPC found the distinction between the two types of performance metricsâthose that are sensitive to all input changes and those that are only sensitive to land use allocation changesâto be confusing. TRPC found that, when only the transportation supply was changed, the comparative output graphs showed no distinction between the scenarios for several of the metrics (which is as designed), but that differences when land use growth was redistributed were much more interesting across all of the metrics. Test Implementation in Portland Region Introduction A fourth implementation of SmartGAP was developed in par- allel to the three agency implementations. This implementa- tion, based on the Portland metropolitan region, was used for model testing and to provide a fourth set of results from the standard scenarios. The specific region used for this test imple- mentation is the three-county Portland, Oregon, metropolitan area, comprising all of Clackamas, Multnomah, and Washing- ton counties (shown in Figure 4.18). The three-county area had a 2005 (model base year) population of 1.5 million and Figure 4.17. TRPC percentage changes in transit operating costs and transit capital costs for base and Scenarios 5, 6, and 7.
83 2035 projected population of 2.3 million (growth of 50%). Table 4.2 shows the breakdown by county. Development of Model Inputs The majority of the input data were derived from existing sources, such as the inputs to the Oregon statewide imple- mentation of the GreenSTEP model. The data for the three- county metropolitan area were extracted from the complete set of GreenSTEP inputs that cover either each county in Oregon individually or each metropolitan area individually. A simple method was used to develop the place type alloca- tion, with density thresholds used to divide households and employment into the four area types and asserted alloca- tions made to the various development types for testing pur- poses. This approach for actual implementations is not recommended; the more detailed approach developed by ARC is preferable. Figure 4.19 shows the distribution of employment (to the left) and population (to the right) by area type for the eight standard scenarios. For both employment and population, the distribution is held static for the first four scenarios and then growth is gradually shifted to toward close-in communities and urban core. Figure 4.20 shows zero-based index charts for the same distributions to show more clearly the positive and negative changes compared to the base scenario. Scenario Testing Results This section of the report presents the results of the eight standard scenarios for the Portland implementation of SmartGAP and also the results of two additional pricing sce- narios that were defined and run. Figure 4.21 shows a com- parison of daily VMT across the eight standard scenarios, with a comparison in terms of miles to the left and a zero- based index chart showing percentage changes to the right. The chart in miles shows that there are relatively small varia- tions in total daily VMT across scenarios. The lowest daily VMT is for Scenario 8 with the most land use growth focused in urban core and additional transit supply. The highest VMT is from Scenario 3, with increased road supply. Given the rela- tively small variation in total daily VMT across scenarios, the percentage change was plotted to show the changes more clearly than the chart to the left that show daily VMT totals. This chart shows that, in comparison to the base: ⢠Scenario 2, with more transit provided, leads to a decrease in VMT; ⢠Scenario 3, with more highway supply, leads to a small increase in VMT; ⢠Scenario 4, with the addition of ITS for incident manage- ment, does not affect VMT (the ITS policy is applied during Table 4.2. Portland Region Population in 2005 and 2035 by County County 2005 2035 Growth (%) Clackamas 361,300 552,800 1.53 Multnomah 692,826 968,700 1.40 Washington 489,786 793,100 1.62 Total 1,543,912 2,314,600 1.50 Figure 4.18. Portland region used for testing SmartGAP.
84 the final estimation of policy-adjusted congestion, after the policy-adjusted VMT is calculated); ⢠Scenarios 5, 6, and 7, which gradually move growth in population and employment to close-in communities and the urban core, result in increasingly larger reductions in VMT; and ⢠Scenario 8 shows the highest reduction, of 3%, as transit supply is increased and a high proportion of the growth is located in close-in communities and the urban core. Figure 4.22 shows the effects on congestion (in terms of vehicle hours to the left and delayed vehicle hours to the right) by scenario. The total vehicle hours chart to the left (showing percentage changes relative to the base scenario) shows that Scenario 4, where ITS is added to sections of highway, has a large impact on total vehicle hours by reducing nonrecurring congestion (ITS is also applied as part of Scenario 8). A similar pattern is seen in the chart to the right, as expected, which plots the absolute number of hours of delay due to congestion. The reductions are due to increased transit and denser, more mixed land uses reducing travel demand, and to increased road supply increasing capacity, with the strongest effects due to ITS being implement to manage incidents and thus reduce nonrecurring congestion. Figure 4.19. Portland 2040 population and employment by area type for eight standard scenarios. Figure 4.20. Portland percentage changes in 2040 population and employment by area type from the base for eight standard scenarios.
85 The transit trips metric reports trips by new residents solely based on land use changes and does not relate to the transit revenue miles supplied as an input. Figure 4.23 shows that transit ridership (to the left) is highest in the urban core, particularly in the scenarios clustering most growth in urban core. The transit operating cost metric develops costs based on forecast usage and, as with the transit trips metric, is not based on the revenue miles supplied. The transit operating cost chart, to the right, shows that the highest operating costs are for the scenarios with growth in the urban core that lead to the highest transit use. The pattern of reductions in fuel use is affected by both changes in daily VMT and also changes in congestion, because that affects travel speeds and hence fuel economy. GHG emis- sions are estimated on the basis of fuel use and so the changes in emissions track the changes in fuel consumption. Fig- ure 4.24 shows a comparison of changes in fuel consumption by scenario (to the left) and changes in GHG emissions by area type for the base and Scenarios 2, 3, and 4 (to the right). The comparison of fuel consumption shows that congestion reduction through ITS provision has a large impact. The total quantities of emissions by area type only change marginally Figure 4.21. Portland daily VMT by scenario (total and percentage change from base). Figure 4.22. Portland congestion effects by scenario (percentage change from base and total).
86 for the scenarios without redistribution of land uses, reflect- ing the relatively small percentage changes shown in the fuel consumption results. In addition to the eight standard scenarios, two pricing scenarios were tested, as defined in Table 4.3. The first of these, Scenario 9, increased auto operating cost growth by 25% to test the sensitivity of the model to higher fuel costs. The second test, Scenario 10, added a per mile VMT charge at a rate of 10 cents/mile, to test the sensitivity of the model to this form of road pricing. Figure 4.25 shows results for daily VMT by area type (to the left) and delay vehicle hours by vehicle type (to the right) for the base scenario and the two pricing scenarios. The results show that VMT pricing at this rate (10 cents/mile), Figure 4.23. Portland transit trips and costs by scenario. Figure 4.24. Portland percentage changes in fuel consumption and total greenhouse gas emissions by scenario.
87 which is Scenario 10 in the charts, has a stronger effect than the more modest increase in operating costs (i.e., higher fuel price), which is Scenario 9 in the charts. Although truck VMT is not affected by these pricing policies (as the truck VMT model is only sensitive to regional income changes over time and not to transportation supply or other policy inputs), trucks experience less delay as they benefit from lower traffic levels on the roads. This effect is captured in the chart to the right that shows a reduction in delayed vehicle hours for trucks as well as for light vehicles. The model was implemented in Portland and efficiently run for the standard scenarios and other scenarios. For the Portland implementation, scenarios took approximately 25 minutes to run on a relatively new desktop. The testing pro- cess was useful and led to two rounds of revisions to the model code being released to the pilot test agencies during the course of the pilot test. In general, the results of the Portland scenar- ios appear reasonable and in line with expectations based on the intended sensitivity provided by the modelâs algorithms. Summary of Pilot Test Findings The five implementations of the SmartGAP model by three pilot agencies provided some valuable feedback on the per- formance and usability of SmartGAP and the supporting userâs guide. Each agency provided a set of results and also additional comments. Some common findings are: ⢠The agencies were all able to install and run the software with relatively little difficulty, although some comments were provided that will assist with the packaging and dis- tribution of the model. ⢠The performance of the model was good for the smaller agencies, but runtime and hardware (memory) require- ments were more onerous for the large implementation of the model by ARC. ⢠Some of the input data, particularly employment data, was found to need a better introduction and discussion in the userâs guide. The preprocessed employment data, based on County Business Patterns data, which was provided to the agencies, requires improvement as it omits certain employ- ment categories. ⢠Each agency developed an approach, which varied greatly in terms of level of complexity, to allocate their population and housing to place types. The userâs guide should include some information on different practical approaches than an agency might follow to develop the place type inputs. ⢠The results from the five implementations appear to be reasonable and consistent, with varying degrees of sensi- tivity to the policy changes depending on the levels of Table 4.3. Pricing Scenarios Scenario Land Use Transportation Policy 9. Increase Operating Costs Baseline Baseline +25% auto operating cost growth 10. Add VMT Charge Baseline Baseline 10 cents/mile VMT charge Figure 4.25. Portland daily VMT and delay vehicle hours for pricing scenariosâ research findings.
88 growth predicted in a region, the existing distribution of land uses, and the severity of the changes made in the test scenarios. Table 4.4 provides an overall comparison of the percentage change in daily vehicle miles traveled across the five pilot tests completed for all eight scenarios. The greatest reductions in vehicle miles traveled were in Cecil County, Maryland, because it is a rural county with high growth predicted, so smart growth strategies can have a larger impact than in other Table 4.4. Comparison of Percentage Change from Base in Daily VMT by Scenario for each Pilot Test Scenario Cecil County, Maryland (%) Montgomery County, Maryland (%) Atlanta Region (%) Olympia Region (%) Portland Region (%) 2 NA -0.7 -1.1 -0.6 -0.8 3 +0.1 +0.1 +0.6 +0.7 +0.1 4 0 0 0 0 0 5 -3.2 -0.3 -2.9 -0.4 -0.8 6 -5.0 -0.8 -4.0 -0.8 -1.5 7 -9.0 -1.3 -4.5 -1.2 -2.1 8 -9.0 -1.9 -5.7 -1.8 -2.8 9 NA NA NA NA -1.4 10 NA NA NA NA -6.5 Note: NA = not applicable. areas that are already mature. Atlanta also had a higher rate of reduction in VMT, which may be a result of the large size of this region (20 counties) which includes less mature areas of high growth. It should be noted that each agency interpreted the design of the standard scenarios themselves and each incorporated some amount of deviation from the precise scenario definitions, so the comparison presented in the table is illustrative and not a rigorous comparison. The findings of the pilot tests supported the recommended enhancements to SmartGAP discussed in this report.
