A Multi‐Scale Approach to Assessing Positioning Accuracy in High‐Density Urban Environments of Hong Kong

ABSTRACT This study examined GPS positioning accuracy in high‐density urban environments using Hong Kong as a representative case study. Through field measurements along 32 testing routes using three GPS devices, we evaluated GPS performance across diverse urban morphologies to assess associations between complex urban characteristics and positioning errors. A three‐phase methodology integrated experimental data collection, device performance evaluation, and multi‐scale environmental association analysis using route‐level Kendall's tau correlation and point‐level machine learning models with SHAP interpretation. Results showed that multi‐device measurements improved accuracy by 7.52%, while traditional GNSS indicators showed limited predictive power ( R 2 = 0.020). The assessment framework identified critical environmental thresholds associated with GPS accuracy. Mean building heights above 20 m and terrain elevations between 20 and 40 m within 100 m buffer zones showed strong associations with increased GPS positioning errors. Based on these findings, we recommend environment‐specific buffer sizes with 5–10 m buffer in open areas and 20–40 m buffer in high‐density areas. This framework supports accurate environmental exposure assessment and privacy protection strategies by enabling differentiated spatial masking approaches adjusted to local positioning accuracy patterns.