PLATFORM VARIABILITY IN PEDESTRIAN ACCESSIBILITY RESEARCH: DOCUMENTATION, PREDICTION, AND POLICY IMPLICATIONS

Abstract

Researchers increasingly rely on web-based routing platforms like Google Maps, ArcGIS, and OpenRouteService (ORS) to conduct pedestrian accessibility studies, but the interchangeability of these tools for consistent research findings remains unexamined. This dissertation addresses a fundamental methodological question: do different routing platforms produce consistent accessibility estimates, and if not, what are the consequences for research and policy? Using residential-to-bus-stop routes in Athens-Clarke County, Georgia, this dissertation investigates platform variability through three interconnected studies. The first study reveals that while platforms agree on relative rankings of accessible areas, they produce different absolute travel estimates. Google Maps generates travel times higher than ArcGIS and ORS due to variable walking speed assumptions. Distance estimates show stronger agreement but still exhibit meaningful variation. The second study develops a machine learning framework to predict whether platforms will agree or disagree with 81% accuracy using only open-source predictors. The models identify network complexity metrics as the strongest predictors for distance outcomes, while topographic features best predict time outcomes. This predictive capability enables researchers to anticipate where platform choice matters most and target validation efforts accordingly. The third study demonstrates that platform selection can affect equity assessments for vulnerable populations. Platform selection produced statistically significant differences in equity assessments in two-thirds of comparisons, with equity gap estimates varying by up to 6.4 percentage points between platforms. Time-based accessibility thresholds proved particularly unstable, showing disagreement in 83% of comparisons, while distance-based measures exhibited less variability. These findings establish that routing platform selection represents an essential methodological choice with direct consequences for both pedestrian accessibility research and policy decisions. For researchers conducting critical equity analyses, this dissertation recommends validating findings across multiple platforms and prioritizing distance-based accessibility measures when possible.