The paper and presentation report work done for an NCHRP project to improve methods for predicting demand for cycling and walking, as a function of infrastructure and surrounding land use and population characteristics. The methods developed for this study were tested using data from the Seattle region. Some specific features of the modeling approach designed for this study include:

• Use of impedance skim matrices for bike developed using a bicycle network, including all bicycle lanes and paths, and created using a cyclist route choice model (transferred from San Francisco).

• Use of detailed sidewalk data along an all-streets network to help characterize path data for walking.

• Use of shortest path distances along an all streets network between every parcel-to-parcel O/D pair, to provide more accurate distances and times than can be obtained using only zone-to-zone skim matrices.

• Use of distance-decay buffering to measure the land uses and attractions surrounding any parcel in the region. The buffering uses the shortest path street distance to any other parcel, rather than simply using straight-line distances.

• Use of various types of measures to represent the mixed-use balance of neighborhoods across housing, various types of business attractions, and parks/open space.

• Use of detailed, along-street distances to measure walking impedances to and from transit stops.

In addition to describing methods for preparing these more detailed input data and presenting the modeling results, as compared to modeling results from more traditional approaches, we will also describe how these methods can be readily applied in a regional forecasting framework.