Lessons Learned from Backcasting and Forecasting Exercises (December 14, 2016)

Recording: https://connectdot.connectsolutions.com/p609mzn3nyf/

Date:  December 14, 2016

Description: basic part of travel demand model validation is running the model for a “base year” and comparing the outputs to observed data such as traffic counts, travel times and speeds, transit ridership, and other measures of travel demand. The Federal Highway Administration (FHWA) recently completed a project to provide information for agencies performing this part of the validation process, sometimes known as dynamic validation. The models for two U.S. metropolitan areas, Baltimore and Cincinnati, were chosen as case studies for this work, and the agencies responsible for these models, the Baltimore Metropolitan Council and the Ohio-Kentucky-Indiana Regional Council of Governments, provided the model files and data. For each region, each of the current and previous model versions were run for the base year for that model, and for the base year of the other model. This means that the previous model was run for the base year and a forecast year, and the current model was run for the base year and a backcast year.

The study demonstrated that backcasting and forecasting can be challenging, but several lessons and recommendations for modeling practitioners became evident:

  1. If possible, temporal model validation should include a backcast and/or a forecast year application.
  2. Recognize that in a model update, changes in model procedures, assumptions, and input data can result in changes in model results that can go well beyond changes in travel behavior over time.
  3. Model inputs, including networks and socioeconomic data, need to be thoroughly checked during model development and validation.
  4. Whenever model parameters are changed or recalibrated during validation, the effects of these changes should be estimated if possible, and should be recognized in any case. Sensitivity tests can be structured to examine such effects.
  5. Recognize the effects of error propagation from model components to subsequent components, and test for error propagation whenever possible. Understand that downstream components may have more error associated with them than upstream components.
  6. If it is necessary to use fixed factors or constants in models, recognize and test for the effects of model sensitivity of such factors. When using results of models that use fixed factors, recognize the limitations associated with insensitivity to factors that are not included in the models.

Presenters: Thomas F. Rossi is a Principal of Cambridge Systematics with over 30 years of management experience in travel demand modeling and transportation planning. He has developed and applied travel demand models throughout the U.S., including conventional trip based models, advanced activity based models, and integrated transportation demand-supply models. He has been an expert advisor to federal agencies and metropolitan planning organizations (MPO) in the development of travel models and survey data collection efforts. Mr. Rossi has worked with the U.S. DOT for over 20 years conducting research, authoring reference documents, and developing and teaching training courses on travel demand modeling. He has been the Principal Investigator on several national research projects.

We are committed to providing equal access to TMIP Webinars for everyone. If you need alternative formats or other reasonable accommodations, please send an e-mail to moderator@tmip.org with your request. We will respond to your request within 24 hours.