Dynamic, Integrated Model System: Jacksonville-Area Application (2013) / Chapter Skim
Currently Skimming:
Pages 17-93
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 17... ...
17 C h A P T e r 1 SHRP 2 Project C10A, Partnership to Develop an Integrated, Advanced Travel Demand Model and a Fine-Grained, TimeSensitive Network: Jacksonville-Area Application was undertaken to develop a dynamic, integrated model and to demonstrate its performance through validation tests and policy analyses. This chapter describes the data requirements and steps necessary to implement the integrated model system.
|
From page 18... ...
18 Figure 1.1. PopGen project setup wizard.
|
From page 19... ...
19 feature. If the synthetic population is found to be appropriate, PopGen tools can export it to specific file formats for use in travel demand microsimulation applications, such as DaySim.
|
From page 20... ...
20 Table 1.1. Household Control Data for Permanent and Seasonal Households Household Attribute Control Column Categories Source Data Householder age 1 18–44 CTPP 1-70 2 45–64 3 65+ Household size, number of workers, and income Size categories 1–4: 1, 2, 3, 4+; Workers categories 1–3: 0, 1, 2+; Income categories 1–4: Under $30,000, $30,000– $59,999, $60,000–$99,999, $100,000 & over (Specified as joint distribution using a composite attribute)
|
From page 21... ...
21 not available. However, the seasonal population is clustered in certain areas, such as along the coast.
|
From page 22... ...
22 control information is so simple that an IPF procedure is not necessary. However, if using PopGen to generate the sample, it can be set up to run only the household-level IPF, which will converge quickly, and avoid entirely the person-level iterative proportion fitting (IPF)
|
From page 23... ...
23 population, the three controls in Table 1.3 need to be included in a household marginals file, but no person marginals file is required. All of these steps have been coded in an R-script.
|
From page 24... ...
24 Table 1.6. PopGen Input Sample File Data Items Data Item and Description Values Control Variable ACS 2006–2008 Item Census 2000 5% PUMS Item Household Sample File State ST STATE Pumano PUMA PUMA5 Hhid (same as serialno)
|
From page 25... ...
25 current version requires only one unique ID, but the code still requires that two fields be present in the input file.) Population Synthesis In this step, PopGen is run separately for each of the three subpopulations, using the specified input control and sample files.
|
From page 26... ...
26 settings in the control file. For Jacksonville, three files are used, one for each population segment.
|
From page 27... ...
27 persons in the synthetic population and observed controls. Overall, the synthetic population has about 1.4% fewer persons.
|
From page 28... ...
28 pErmanEnt housEholDs. The household and person controls and their categories shown in Table 1.1 and Table 1.2 for Jacksonville were also used for PopGen synthesis of the permanent household population in Burlington.
|
From page 29... ...
29 Table 1.17. Jacksonville Household Workers Distributions for Permanent Households, by County Household and County Household Workers 0 1 2+ Total Observed Clay 12,231 24,171 24,949 61,352 Duval 74,193 141,661 112,496 328,350 Nassau 6,783 8,053 8,695 23,531 St Johns 17,824 25,939 22,259 66,022 Synthesized Clay 12,233 24,154 24,982 61,369 Duval 74,116 141,841 112,395 328,352 Nassau 6,810 8,053 8,682 23,545 St Johns 17,871 25,963 22,198 66,032 Table 1.18.
|
From page 30... ...
30 Table 1.20. Jacksonville Household Income Distribution for Seasonal Households, by County Household and County Household Income <$30K $30K–$60K $60K–$100K >$100K Total Observed Clay 748 1,033 430 854 3,066 Duval 6,214 8,579 3,575 7,094 25,463 Nassau 641 885 369 732 2,626 St Johns 1,021 1,410 588 1,166 4,185 Synthesized Clay 738 1,061 405 863 3,067 Duval 6,149 8,718 3,412 7,204 25,483 Nassau 643 896 353 734 2,626 St Johns 998 1,450 566 1,177 4,191 Table 1.21.
|
From page 31... ...
31 Distributions for all of the control attributes except those for presence of children were derived from CTPP tables. The distributions for the presence of children attribute was obtained from the Census SF1.
|
From page 32... ...
32 Permanent population controls were estimated in a straightforward manner. An approximate average of 4.5 persons was assumed for the highest household size category (households with four or more persons)
|
From page 33... ...
33 for Jacksonville, which has three population segments, the TAZ controls file is an input of permanent households, seasonal households, and noninstitutionalized GQ residents living in each TAZ. The file format is shown in Table 1.10.
|
From page 34... ...
34 0 10000 20000 30000 40000 50000 60000 70000 0-15 16-20 21-44 45-64 65+ Age Group Observed-Permanent Synthesized-Permanent Figure 1.10. Burlington person age distribution comparison.
|
From page 35... ...
35 daySim Parcel data A distinguishing feature of DaySim is that it uses parcels as one of the fundamental spatial units. The parcel data input file contains input data at the parcel level of detail, the more detailed of the two spatial levels at which DaySim input data are prepared.
|
From page 36... ...
36 Table 1.25. Parcel Data Input File Format Labela Definition PARCELID Parcel ID number X_COORD X coordinate – state plane feet Y_COORD Y coordinate – state plane feet AREA_SQF Area – square feet TAZ TAZ number HOUSESP Housing units – parcel (× 100)
|
From page 37... ...
37 impute housing units for land-use types such as condominium and multifamily housing developments (because housing unit numbers were not necessarily present in the parcel database) , and verify that the resulting housing unit numbers were reasonable.
|
From page 38... ...
38 Parcel-level calculations of proximity to total housing units within quarter-mile and half-mile buffers are also important urban form measures used in DaySim. They are calculated using a script, as described in the section on parcel-level buffers.
|
From page 39... ...
39 classification. A two-digit SIC code was derived from the sixdigit SIC codes in the employment databases (i.e., the first two digits of the six-digit SIC code)
|
From page 40... ...
40 County employment was subsequently scaled up to match aggregate NERPM-based totals. As with housing units, parcel-level calculations of proximity to total employment by sector, within quarter-mile and half-mile buffers, are used in DaySim and calculated using a script, as described in the section on parcel-level buffers.
|
From page 41... ...
41 0% 10% 20% 30% 40% 50% 60% 0-100 101-500 501-1000 1001 or more Enrollment Private Public Figure 1.13. K–12 enrollment distribution.
|
From page 42... ...
42 enrollment and employment information for all of the universities and colleges, with an indication of whether the enrollment was imputed. For brevity, only those institutions with 10 or more employees are shown in this table.
|
From page 43... ...
43 Figure 1.14. Example of quarter-mile and half-mile buffer areas around a parcel centroid.
|
From page 44... ...
44 Figure 1.17. Total employment per parcel and quarter-mile and half-mile total employment buffers, urban area.
|
From page 45... ...
45 addresses the more practical aspects of the current implementation. In several aspects of this work, the project team has made use of the findings from the SHRP 2 C04 project, Improving Our Understanding of How Highway Congestion and Pricing Affect Travel Demand.
|
From page 46... ...
46 used in discrete choice modeling to capture the expected utility across all available choice alternatives. As much as possible, DaySim uses fully disaggregate logsums, which essentially combine two models into a single joint, simultaneous model.
|
From page 47... ...
47 choice for a given tour or trip, for example, DaySim would evaluate all available paths through the network at each time of day for that given traveler on that given tour or trip. In other words, network path choice would be done "on the fly" in a fully disaggregate manner depending on each traveler's tradeoffs between travel time, toll, distance, and any other important route characteristics that are known in the network.
|
From page 48... ...
48 Binary Route-Type-Choice Model The route-type-choice model implemented in DaySim works as shown in Equation 1.2: , .
|
From page 49... ...
49 simulating normal or lognormal taste variation around this coefficient for each individual would also be possible. However, empirical evidence to go by is lacking, and, statistically, estimating taste variation parameters on both the toll constant and the time or cost coefficient at the same time would be difficult.
|
From page 50... ...
50 This extra information can be used by TRANSIMS to (a) know whether to exclude tolled links from possible paths when assigning the trip to the network, and (b)
|
From page 51... ...
51 designed and developed to dynamically adjust to future conditions rather than be fixed or limited to existing traffic controls. Network Conversion Process The TRANSIMS suite includes a number of tools to synthesize a TRANSIMS network from traditional MPO networks.
|
From page 52... ...
52 Jacksonville Network Development In consideration of the flexibility requirement and the project goal to develop a fine-grained network, the team started the model development by creating three network resolutions from the NERPM regional modeling data sets: 1. PLANNING Network.
|
From page 53... ...
53 Figure 1.24. NERPM master network.
|
From page 54... ...
54 Figure 1.26. NERPM year 2005 network (shown in light blue)
|
From page 55... ...
55 Figure 1.28. TransimsNet controls, by area type.
|
From page 56... ...
56 Figure 1.31. Example location showing short link resulting from facility-type change.
|
From page 57... ...
57 lanes are the primary source of capacity within the simulation and lane changing is one of the primary reasons for congestion and lost vehicles, these types of errors are extremely problematic. The last type of discontinuity was a large change in speeds as seen in Figure 1.33.
|
From page 58... ...
58 Figure 1.34. Nodes collapsed as result of removing discontinuity in link attributes.
|
From page 59... ...
59 For example, when a divided arterial is represented as a pair of two one-way links, the intersections are represented by two nodes. The tagging process was useful in two respects.
|
From page 60... ...
60 48867 75017 49031 49086 49134 49417 49252 49067 3586935887 35873 35877 36046 36462 3650836455 36050 51371 51853 5185151388 b. If a single-link roadway intersected a double-link roadway, then the nodes at the two intersection points were attached to the traffic signal.
|
From page 61... ...
61 Figure 1.36. Network conversion process.
|
From page 62... ...
62 ALLSTREETS Network The PLANNING network and the ALLSTREETS network were envisioned as the two ends of network resolution for this project. While the PLANNING network was limited to the NERPM modeling links for the 2005 scenario, the ALLSTREETS network used links that were not defined in any of the NERPM scenarios.
|
From page 63... ...
63 2. Additionally, not all streets bring value to the TRANSIMS simulation modeling.
|
From page 64... ...
64 given the PLANNING and ALLSTREETS networks using PathSkim. The accessibility scores for the resulting FINEGRAINED network were computed by measuring the Euclidean distance of each parcel in the region from its spatially nearest roadway.
|
From page 65... ...
65 Table 1.41. Accessibility Computation S
|
From page 66... ...
66 Figure 1.43. Employment accessibility.
|
From page 67... ...
67 locations. To synthesize TRANSIMS simulation network data, the number of links into and out of a given node was used along with intersection logic to construct turn pockets, lane connectivity, and traffic controls -- both signs and signals.
|
From page 68... ...
68 locations. This provided an automated means of importing new and future-year four-step planning networks and automatically correcting and updating the activity locations synthesized by TransimsNet.
|
From page 69... ...
69 Figure 1.46. Resulting FINEGRAINED network.
|
From page 70... ...
70 NERPM model system currently used in Jacksonville, with spatial and temporal detail added to support integration with the detailed demand and supply simulation models. Because this demand is exogenous to the DaySim-TRANSIMS model system, the total demand -- and the spatial distribution, mode, and timing of those trips -- is fixed within a given forecast or horizon year, though of course it will vary across model run years.
|
From page 71... ...
71 Note that, regarding special generators, the original NERPM model contains more special generators than were ultimately included in the DaySim-TRANSIMS model system (e.g., state parks, military bases, and malls)
|
From page 72... ...
72 Figure 1.49. Jacksonville auxiliary demand time-of-day distribution.
|
From page 73... ...
73 Studio Components The integrated model implemented in TRANSIMS Studio comprises four software components: 1. TRANSIMS Studio user interface and application management software; 2.
|
From page 75... ...
75 • var.NUM_PATHSKIM_THREADS = 8 This variable applies only to the TRANSIMS Version 5 software PathSkim, which is used to create skims. It specifies the number of threads to use in its application.
|
From page 76... ...
76 impedance path between the origin and destination based on time-dependent link travel times and turning movement delays. The paths or travel plans for all of the trips over a 24-hour period are loaded onto the network and simulated by the Microsimulator.
|
From page 77... ...
77 number of partitions can be specified independent of the number of machines/threads or nodes at the user's disposal. When the number of partitions is specified higher than the number of machines/threads, TRANSIMS Studio processes the partitions as sequential sets of applications according to the number of machines/threads available.
|
From page 78... ...
78 VDFs for computing link delays rather than the second-bysecond simulation of individual vehicles. Thus, the Routeronly process does not consider traffic signal operations, lane changing, or vehicle interactions.
|
From page 79... ...
79 the composite plan file. This plan file is then aggregated into link volumes by time of day; delays are calculated using a facility-type-based VDF to the 5-min or 15-min link volumes to estimate the travel time for each time increment; and a weighted average or MSA procedure is used to combine the travel time estimates with previous travel time estimates for feedback to the next Router iteration.
|
From page 80... ...
80 AON link delay file. Similarly, PlanSum is used to create a link delay file using the travel plans and the Microsimulator performance file.
|
From page 81... ...
81 The data processing steps required to calculate the three trip-gap convergence measures are outlined in Figure 1.59. The inputs to this process are the same as the relative-gap process.
|
From page 82... ...
82 simulate vehicle-type choice or allocate specific vehicles to person trips. Therefore, a simple procedure is used to treat every vehicle trip as an independent vehicle.
|
From page 83... ...
83 Table 1.47. TRANSIMS Activity File Example HHOLD PERSON ACTIVITY PURPOSE PRIORITY START END DURATION MODE VEHICLE LOCATION PASSENGER 1 1 111110 0 9 1 44520 44519 1 0 5937 0 1 1 11111 4 9 45480 57360 11880 2 1 13688 0 1 1 11121 0 9 58500 97140 38640 2 1 5937 0 the conversion of the time units from hours and minutes (for example, 1222 represents 12:22 in the original DaySim trip list output)
|
From page 84... ...
84 at each occupancy level to assign as driver tours; the method depends on the other trip modes used on the tour. For example, if a tour includes one or more walk or bike trips as well as shared-rider trips, it is designated as a car passenger tour.
|
From page 85... ...
85 Ultimately, time and cost measures may be based on more spatially detailed TRANSIMS activity locations and for specific times that a trip or activity may be routed. As described earlier, the fundamental spatial unit used in DaySim is the individual parcel, which is significantly more fine-grained than the TAZs used for network skimming.
|
From page 86... ...
86 equivalent to vehicle types. Figure 1.61 provides an overview of the core model processing.
|
From page 87... ...
87 Operating Mode Distribution Data Generator The operating mode distribution generator (OMDG) provides a mechanism for defining the distribution of operating modes used to calculate emissions.
|
From page 88... ...
88 very similar to the way most MPOs applied the MOBILE6 software in the past. The primary disadvantage of the lookup method is that the resulting rate tables may be too big and bulky for practical use.
|
From page 89... ...
89 Figure 1.63. MOVES emissions inventory method.
|
From page 90... ...
90 Table 1.51. TRANSIMS Speed Bin MetaData MetaData Description TIME_STAMP The time and date when the file was created BOX_LENGTH The segment length in meters CELL_LENGTH The cell length used in the simulation SAMPLE_TIME The frequency in which data are collected (seconds)
|
From page 91... ...
91 Table 1.53. Collapsed Emissions Rate Table yearID monthID sourceTypeID roadTypeID pollutantID processID avgSpeedBinID emissionRate 2008 1 21 2 1 1 1 1.91824 2008 1 21 2 1 1 2 1.02998 2008 1 21 2 1 1 3 0.608886 2008 1 21 2 1 1 4 0.430296 2008 1 21 2 1 1 5 0.37313 2008 1 21 2 1 1 6 0.318093 2008 1 21 2 1 1 7 0.2814 2008 1 21 2 1 1 8 0.258368 2008 1 21 2 1 1 9 0.241404 2008 1 21 2 1 1 10 0.228209 2008 1 21 2 1 1 11 0.217652 2008 1 21 2 1 1 12 0.206746 2008 1 21 2 1 1 13 0.195869 2008 1 21 2 1 1 14 0.208396 2008 1 21 2 1 1 15 0.237551 2008 1 21 2 1 1 16 0.276734 Table 1.54.
|
From page 92... ...
92 Table 1.56. Ramp Fractions roadTypeID roadDesc rampFraction 1 Off-Network 0 2 Rural Restricted Access 0.056354 3 Rural Unrestricted Access 0 4 Urban Restricted Access 0.084319 5 Urban Unrestricted Access 0 Table 1.57.
|
From page 93... ...
93 Table 1.58. Average Speed Bin Distribution sourceTypeID roadTypeID hourDayID avgSpeedBinID avgSpeedFraction 21 2 12 1 0.004948 21 2 12 2 0.004122 21 2 12 3 0.003 21 2 12 4 0.002265 21 2 12 5 0.002105 21 2 12 6 0.003277 21 2 12 7 0.00927 21 2 12 8 0.019876 21 2 12 9 0.04253 21 2 12 10 0.093737 21 2 12 11 0.152748 21 2 12 12 0.169864 21 2 12 13 0.125502 21 2 12 14 0.072482 21 2 12 15 0.065015 21 2 12 16 0.229258
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.