Quantifying distance overestimation from global positioning system in urban spaces

Objectives. To investigate accuracy of distance measures computed from Global Positioning System (GPS) points in New York City. Methods. We performed structured walks along urban streets carrying Globalsat DG-100 GPS Data Logger devices in highest and lowest quartiles of building height and tree canopy cover. We used ArcGIS version 10.1 to select walks and compute the straight-line distance (Geographic Information System-measured) and sum of distances between consecutive GPS waypoints (GPS-measured) for each walk. Results. GPS distance overestimates were associated with building height (median overestimate = 97% for high vs 14% for low building height) and to a lesser extent tree canopy (43% for high vs 28% for low tree canopy). Conclusions. Algorithms using distances between successive GPS points to infer speed or travel mode may misclassify trips differentially by context. Researchers studying urban spaces may prefer alternative mode identification techniques.