Background
Urban trails are present in many communities and provide benefits including open space, recreation, non-motorized transportation and opportunities for increased physical activity. The potential to create additional urban trails is great with over 160,000 miles of abandoned track available for rail-trail conversions and federal Intermodal Surface Transportation Efficiency Act funds available to improve and construct trails. A small number of studies have been conducted to examine built environmental determinants of urban trail use. A better understanding of the variables that explain and predict trail use will assist public health professionals, urban planners and policy makers in making critical decisions about resource allocation and the development of recreational spaces.

Objectives
The present study adds to our understanding of trail use predictors and improves on prior studies by using multiple trails in diverse geographic regions, systematic audits of built environmental characteristics of the trails, and a large sample of trail users. Specific objectives of this analysis were to test built environmental variables as correlates of urban trail use.

Methods
Data collection was provided by two mechanisms. First, we adapted the Systematic Pedestrian and Cycling Environment Scan (SPACES) (Pikora, 2002) to assess characteristics of urban trails. SPACES for trails used 38 items to provide audit information on the path itself (e.g., path quality, slope), built structures surrounding the trail (e.g., institutional buildings, recreational facilities), trailside facilities (e.g., rest areas), vegetation and visibility, lighting, signage, crowding, barriers, and aesthetics. Second, trail use was estimated using trail counts of users over a four-day period. The trail count procedure provided estimates of the number of trail users on each segment, the age and gender of those users and the type of activity observed on the trail. Data were collected on multiple-use trails in Chicago and Dallas. The trails transected communities with heterogeneous land uses and socio-demographic characteristics. A GIS was used to divide the trails into a total of 67 half-mile segments, and a GPS enabled boundary verification. For the SPACES assessment, two data collectors walked the length of the trails, coding each trail segment. For trail count data, two data collectors coded trail use. To estimate inter-rater reliability, coders rated the same users on at least four segments for each day of counting. Agreements were 0.67 for age, 0.90 for gender and 0.94 for type of use. In the present abstract, associations between the audit ratings determined by the SPACES instrument were associated with the trail count data in Chicago and Dallas.

Results
A total of 15,646 users were counted on the 67 trail segments in Chicago and Dallas. Trail counts varied widely by segment, from a low of 18 to a high of 825 (mean = 233). Statistical associations between trail characteristics measured for each segment and the number of users counted on each segment were tested in a general linear model with a Poisson link in S-Plus 2000. Initial analyses were conducted using 13 variables characterizing the built environment of the trails. These include the presence of rest areas and water at the side of the trail, condition of the pathway, density of trailside vegetation, visibility of the trail from surrounding neighborhoods, attractiveness and difficulty of the trails for walking and cycling, ease of navigation and the presence road crossings.

Many of the 13 variables examined from the SPACES audit were predictive of trail counts per segment. Results indicate that the presence of water and rest areas next to the trail, and greater attractiveness of the trail for cycling are significantly related to higher trail counts. Greater visibility of properties adjacent to the trail, greater difficulty for cycling and the need to cross roads were each significantly related to lower trail counts.

Conclusions
These analyses have identified a number of variables associated with trail use using strong measurement methodologies on geographically diverse trails and a large sample. The findings may assist public health professionals and urban planners in allocating resources for the design and development of recreational and transportation resources including urban trails. Some variables lacked sufficient variability to assess associations statistically, suggesting a need to refine the SPACES classification system. Further analyses will be conducted with this dataset using additional SPACES audit variables and new data from Los Angeles