TModel Corporation staff have extensive experience in development of travel forecasting models, transportation plans, and training as well as basic traffic analysis. The principals of TModel Corporation first organized as Professional Solutions, Inc. in 1982 and formed TModel Corporation as a separate entity in 1991. The firm's primary purpose is to provide software tools, training, and application of innovative and progressive techniques for analyzing and solving transportation problems. TModel Corporation is located in Vashon, Washington.
TModel Corporation is the developer of the TMODEL2 computer transportation modeling system, the NCAP iNtersection Capacity Analysis Package, TGAP Traffic Gap Analysis Package, and other transportation analysis tools. TMODEL2 has the ability to model traffic from the regional level down to specific turning movements on intersections. TModel Corporation programs are in widespread use with approximately 600 user sites in over 50 states and provinces, and in nineteen countries. As developers of the TMODEL2, NCAP, TGAP and other transportation analysis software, TModel staff are uniquely qualified to construct and calibrate transportation models and to provide training in all of the features of TMODEL2 and transportation forecasting models.
The staff include the authors of the program as well as engineers and planners with experience on many TMODEL projects. Their knowledge and dedication have been demonstrated in the creation, support, and application of user-friendly, state-of-the-art transportation system planning and evaluation programs.
There are different approaches to creating a transportation forecasting model (hereafter referred to as a model). One is not necessarily better than another. All, however, have common features which should be understood.
TMODEL2 is a set of tools which allows the modeler to develop a mathematical model of traffic. A model consists of a set of "rules" which includes data, functions, function coefficients, assumptions and documentation. Some of its elements are listed and considered below.
The team of people constructing the model are key to its ultimate quality and usefulness. This team must consist of two parts: 1) TModel Corporation staff with in-depth understanding of the traffic modeling process and of the tools being used; and 2) client staff with in-depth understanding of the data, i.e., the area, the transportation network, the traffic, and the growth elements to be modeled.
The goal of a model is simple. Once constructed and calibrated, the model is a predictive tool which can be altered to simulate changes in the roadway network, land uses, and/or traveler behavior characteristics. It should then be able to project traffic volumes and, in many cases, turn counts. It may also be used for other post distribution and assignment calculations such as V/C, VMT, vehicle hours, cost-benefit analyses, emissions modeling, Level of Service, etc.
Some important elements in construction and use of a traffic model are:
For modeling purposes the network consists of all the streets and intersections which carry significant traffic movements. Local streets and driveways which serve to load traffic onto the arterial streets are typically either omitted or are aggregated and represented by zone "connectors" which appear as additional streets in the TMODEL system. It is important to develop a network which supports the analysis of significant traffic flows and yet remains efficient.
The location of TAZ boundaries and centroids (or load points) is critical to the operation of the network. There must be an appropriate balance between zone detail and network detail. Compromises in the zone boundary layout are often necessary due to the limitations of data availability and the inappropriateness of census boundaries for traffic analysis. When possible, however, a traffic-shed (analogous to watershed) notion is used to develop the zones. This assures that the land uses which produce the traffic flows feed into the proper arterials.
The required street descriptions include only length, hourly capacity and free-flow speed. However, TMODEL2 allows inclusion of several other variables such as class, area, type, 1 or 2 way, and number of lanes. Some of these can be used to describe additional capacity-altering conditions such as transit routes, parking, bicycle routes, driveway frequencies or others.
The intersection descriptions and the program's ability to attribute delay appropriately are what sets TMODEL2 apart from other modeling software. Delays can be dynamically (according to traffic flows at the time) determined by and distributed to a) the intersection as a whole and/or to b) any turn movement in the intersection as a function of the movement's traffic and of opposing traffic. The intersections also include class, area and type variables to take advantage of other grouped capacity-altering conditions.
Using full descriptions is important because a) it allows the modeler to take advantage of efficiencies in assigning and altering these values in appropriate groups rather than a street or intersection at a time, and b) these additional data provide the justifications for setting the capacities which can be understood by other users or evaluators of the model.
On the other hand, it is not necessary to constrain every movement or approach at every intersection individually. The general rule is to make the network as simple as possible; complexity is added where needed to better calibrate and justify the model operation.
These data comprise residential and employment land uses. Single family dwelling units (SFDU) and multi family dwelling units (MFDU) are the typical residential land uses and are easy to extract from census data. The employment data can be any mix of numbers of employees by generalized business sector or in square feet of building area or acres of land use type. For whatever categories are used, there must exist or be derivable, hourly trip generation rates for the model hour.
The land uses and the trip generation rates are obviously the underpinning of the traffic volumes on the network. The employment data are typically the most difficult to acquire. Any extraordinary efforts in this area are worthwhile.
Also important are the assumptions being used to characterize a) the travelers' views of the roadway network and b) their behavior. These assumptions must then be appropriately quantified through the gravity model coefficients and the link and node delay function coefficients. It should be recognized that applying regional parameters to a specific area is often not appropriate.
We judge a model on a) its ability to replicate base data conditions and b) its documentation; the justifications for and existence of the set of data, rules, assumptions and conditions. Either of these without the other is of little value.
Calibrating the model consists of refining the data, parameters and rules until the model volume assignments sufficiently replicate base date traffic counts. Two major issues arise here.
First, count data must be directional and specific to the hour to be modeled, generally AM or PM peak. It must exist for all streets which enter or leave the modeled area and connect to the outside world, the external zones. And it must exist in locations which capture major traffic movements aggregated into screenlines. The count data is one of the first places to look when reconciling error in a model. Better data produces better models.
Second, calibrating or refining the data means replacing the original data and assumptions with better data and assumptions as more and more is learned about the area through assessment of the inadequacies of early model runs. We require all data, parameters and rules to be justifiable, that is, to be backed by reason and reasonableness. While it is always possible to "adjust" certain data or parameters to help "calibrate" the model, the model's ability to predict will be compromised when justifications are compromised.
Accuracy, that is, how closely the ground counts are replicated by the model, is related to the volumes and traffic densities. Accuracy and requirements for accuracy should increase as traffic volumes and V/C ratios increase. We calibrate to exceed the goals shown for "allowable calibration errors" for screenlines and for links or streets where counts exist as set forth on graphs in NCHRP 255 (pp. 41, 49). We are willing in appropriate situations to accept a less accurate calibration for a more reasoned model which will be a better predictor.
Replication of those conditions can be assessed through comparison of assignment volumes to base data conditions using the following tools:
This is the ultimate use of the model and purpose for its creation. This is a task which will be easily accomplished by agency staff who are familiar with TMODEL2 and the documentation for this particular model. The study team will use the model to forecast future traffic using the projected land use growth scenarios. TModel Corporation will also perform preliminary forecast runs during calibration so that anomalies in the calibration can be found and to assure that the calibration is right for the right reasons.
The calibrated model can be used to test the effects of land use growth or of network change. Network changes may simulate either permanent street construction such as widening, intersection improvements, new roadways, or closures; or temporary changes such as reduced capacity and/or speed during maintenance.
The calibrated model can project absolute volumes or changes in volumes on streets or turn counts. It can also analyze traffic flows to and from specific developments or cumulative developments or help to schedule improvements in an area of cumulative development. The model can allocate responsibility for traffic on a specific street to its zones of origin and/or destination.
Documentation, the creation of a notebook of data and methods, is critical. The test of documentation is whether another TModeler understands the procedures, assumptions and data from the existing documentation sufficiently to use it to test scenarios, recalibrate or explain the model to interested persons.
In addition to a knowledge of TMODEL2 which no other firm can match, TModel Corporation has in-house programs and techniques for assessing the traffic simulation models developed using TMODEL2. We provide the service of evaluating and recommending improvements to existing models for agencies as well as other consultants. For any modeling project that we participate in, we bring this expertise as well as experience gained from reviewing the techniques of others.
There are many different approaches in the creation of a traffic simulation model. One is not necessarily more right or wrong than another. Most important are the assumptions being used to characterize a) the travelers' view of the roadway network and b) their behavior. It follows that these assumptions must then be appropriately quantified. TModel Corporation staff can evaluate your model and its accompanying documentation and advise you as to its capability of assessing impacts of development. The philosophy and method of our evaluation of models constructed by ourselves and others is discussed below.
We judge models on a) their ability to replicate base data conditions and b) the documented justifications for and existence of the set of data, rules, assumptions and conditions. Either of these without the other is of little value.
Our modeling philosophy drives our approach to model development and assessment. We feel that understanding and justification of each element of the model is the critical factor. In order for a model to best project future scenarios, it must be based on solid reasoning. Making unreasoned adjustments to data or parameters in order to make the base data calibration appear to be more accurate is against our philosophy. We are willing to accept a slightly "less accurate" calibration for better future projections.
Replication of base conditions, also known as "calibration" of the model, consists of two issues: a) what are base conditions and b) how do we know when we've replicated them.
Base conditions consist of:
Replication of those conditions can be assessed through comparison of assignment volumes to base data conditions using the following tools:
Documentation of data and methods is critical. The real test is whether or not another Tmodeler can understand the procedures, assumptions, and data from the existing documentation at a level where the new modeler could use it to test scenarios, recalibrate, or explain the model to interested persons. Unless the documentation can pass this test, your model is a precarious investment.
Other modeling services that TModel Corporation provides include the following:
Model Updating - TModel Corporation staff can update your model to reflect current conditions or projected conditions. Revisions can be made to the land use, zone, or networks to enable the model to meet your evaluation needs.
Model Application - We can perform model runs for testing large numbers of different land use and network alternatives. In addition, we can import the turn movement data into NCAP files for analysis with the 1985 Highway Capacity Manual or TRC 212 methodologies.
Model Results Evaluation - We can provide assistance in evaluating the model can be provided. Street and intersection volume/capacity ratios, operating speeds, and delays can be presented in graphical form for rapid evaluation. Turn movements can be evaluated for more precise measures of Level-Of-Service. We can assist you by providing any level of assistance that you desire.
TModel Corporation offers consulting and subconsulting services for any community or consultant wishing to undertake an origin destination survey. Through participation in four surveys for cities and counties in the past four years, we have developed methods which eliminate the geocoding problems and significantly reduce the time and money costs of acquiring travel data to determine local conditions and parameters for computer traffic modeling.
In the course of these activities we have developed a set of programs for data entry, data verification, statistical analysis, and traffic analysis.
Services we offer are:
Some of these programs can be licensed to the ultimate user of the survey. Others can be modified for their use. Still others must be used by us, because they require constant modification for individual needs.
In past projects, TModel Corporation staff have personally developed or contractually advised on transportation planning models for the municipalities of:
|
Arlington, Washington Bay Area Rapid Transit, California Black Diamond, Washington Boulder, Colorado Carpinteria, California Douglas County, Washington Dutchess County, New York Fishkill, New York Herkimer-Oneida Counties, New York Hollister, California Kansas City, Missouri Ketchikan, Alaska Maitland, New South Wales Naperville, Illinois Olympia, Washington Orcutt, California Ottawa, Ontario Paradise, California Paso Robles, California Regina, Saskatchewan Richland, Washington Roswell, New Mexico Santa Cruz, California Santa Maria, California Sedro-Woolley, Washington Skagit County, Washington Shafter, California Stafford County, Virginia Victorville, California Wenatchee, Washington Wilsonville, Oregon Yakima, Washington York County, Pennsylvania.
|
Auburn, Washington Beaverton, Oregon Blaine, Washington Cambria, California Chico, California DuPage County, Illinois Everett, Washington Gallup, New Mexico Hoffman Estates, Illinois Indianapolis, Indiana Kent, Washington Lancaster County, Pennsylvania Marysville, Washington Novi, Michigan Orange County, New York Oroville, California Overland Park, Kansas Park Ridge, Illinois Portland, Oregon Renton, Washington Rockford, Illinois Santa Barbara, California Santa Fe, New Mexico Saskatoon, Saskatchewan Shelton, Washington Sioux Falls, South Dakota Spokane, Washington Taos, New Mexico Washington County, Oregon Whatcom County, Washington Winthrop, Washington Yakima County, Washington |
TModel Corporation staff have developed and conducted over 60 training workshops on transportation modeling and planning. The workshops discuss the elements of transportation modeling and data collection for all levels of experience. Advanced courses and specialized courses have also been conducted. These have included sessions on general transportation modeling practice as well as dealing specifically with TMODEL2. Workshops have been held in British Columbia, California, Illinois, Maryland, New Hampshire, New Mexico, New South Wales, New York, Oregon, Saskatchewan, South Dakota and Washington.