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.
153 Proposed Validation and Sensitivity Testing of Denver Region Activity- Based Models David L. Kurth, Parsons Corporation Suzanne Childress, Parsons Corporation Erik E. Sabina, Denver Regional Council of Governments Thomas Rossi, Cambridge Systematics, Inc. Traditional four- step travel modeling procedures haveevolved over the last half of the 20th century. Manyimprovements were made incrementally and sub- jected to either formal or informal validation and sensitiv- ity tests. Formal validation tests were normally applied at the end of the model calibration process and, quite fre- quently, focused on the âsuper testââthe concept that reproduction of observed traffic volumes and transit board- ings at some reasonable level of aggregation somehow showed that the models were, in fact, valid. Informal vali- dation and sensitivity tests, unfortunately, too often con- sisted of discovering modeling problems after illogical travel forecasts were produced. In the late 1990s, FHWA commissioned the develop- ment of a Model Validation and Reasonableness Check- ing Manual through the Travel Model Improvement Program (1). This manual summarized validation stan- dards and recommended a process that focused on the validation of the individual four- step model components as well as the traditional overall model system validation focused on traffic volumes and transit boardings. Sensitivity testing has been somewhat less formal. It has frequently focused on the sensitivity of individual model components using measures such as elasticity. Sen- sitivity testing of modeling systems by validating model results over time has not been as common because it requires observed travel data from more than one point for the same region. Even if such data did exist, many regions do not have measurable changes in their trans- portation system such as the addition of new roadway capacity or the opening of a new transit line. Without major changes to the underlying transportation system, it is difficult to test the veracity of the underlying models (unless the test proved that the models were, in fact, poorly calibrated). In 1997, the Denver Regional Council of Govern- ments (DRCOG) initiated the collection of travel survey data to update its traditional four- step travel model and for the longer- term development of state- of- the- art mod- eling techniques. The ârefreshâ of the traditional model using these data took place from 2002 to 2004 and included model component validation, validation to the 1997 base year, and validation to travel conditions in 2001. Between 1997 and 2001, Denverâs light rail sys- tem more than doubled in length with the opening of the 8- mi- long Southwest LRT line, enhancing the effective- ness of transit component calibration and validation. As with its current trip- based model, DRCOG is committed to rigorous validation and sensitivity testing of its activity- based modeling (ABM) system that will be developed over the next 18 months. The system is designed to make use of the most recent developments in ABM theory to better represent the travel decision- making process and provide reasonable sensitivity to a wider range of future travel options and constraints. The downfall is that there are a number of places where the models can fail. DRCOG has addressed this concern by committing approximately the same budget to the validation and sensitivity testing of the ABM that it committed to the entire refresh of the existing tradi- tional four- step travel model. Perhaps a statement made by Chandra Bhat and Frank Koppelman in a recent