As Californians are faced with the challenges of reducing greenhouse gas emissions, the limited capability of most travel models to estimate the emissions reduction of new bicycle facilities presents a challenge in that the models tend to be biased towards the automobile. The Association of Monterey Bay Area Governments (AMBAG) and the Monterey Bay Unified Air Pollution Control District (MBUAPCD) sought to develop a new tool in order to make more informed decisions about how to invest limited financial resources in the bicycle mode. The tool (1) operates independently of other models, (2) runs entirely on an open-source platform, and (3) provides user-friendly access to planners in multiple agencies with varying levels of technical skill.

To satisfy these requirements, we developed an incremental nested logit mode choice model that pivots off of static exports of trip tables from AMBAG’s four-step model. The model is implemented in an Adobe ActionScript graphical user interface (GUI) for independence and ease of use. After the user edits the bicycle network in the GUI, new bicycle levels-of-service are skimmed, and new shares for each mode are calculated from their original shares in the no-build alternative and the change in the bicycle level-of-service. Finally, the emissions reduction is estimated based on the distance and average speed of the vehicle trips substituted by bicycle travel.

The mode choice utility function for the bicycle alternative was derived in a series of steps. First, GPS traces collected from smartphone users in the region were analyzed to estimate a path size logit route choice model. Next, the California Household Travel Survey and the original base-year modal trip tables output by the model before the introduction of the bicycle level-of-service were used to calibrate an alternative-specific constant for each trip purpose, and a scaling coefficient which is applied to the best utility from the route choice model, to account for differences in sensitivity between route and mode choice. The result is an accurate, fast, freely distributable, user-friendly tool that is consistent with the forecasts produced by the four-step model.