Human mobility is essential for accurately estimating individual heat exposure, yet most assessments still rely on residence-based measures that neglect daily movements. How mobility patterns influence heat exposure across different urban contexts remains underexplored. This study used mobile phone signaling data to capture individual mobility and employed a numerical model to simulate mean radiant temperature. Residence-based exposure (RBE) and mobility-based exposure (MBE) to heat were estimated in Chongqing and Chengdu, and the differences between the two were examined. Interpretable machine learning was then applied to explore the nonlinear effects of mobility indicators on estimated differences (|MBE–RBE|). Estimated results show that RBE can both overestimate and underestimate heat exposure, leading to biases relative to MBE. These differences were larger in Chongqing, reflecting its more complex mobility patterns. In Chongqing, human flows were more dispersed toward scattered subcenters, increasing exposure, whereas in Chengdu, flows concentrated toward a few employment centers, reducing exposure. Machine learning analysis revealed that mobility indicators substantially influenced the estimated differences. Out-of-home duration had a positive effect once exceeding 6 h, stabilizing beyond 12 h. Travel frequency exerted a positive effect within an effective range from 2 to 4 trips, particularly in Chongqing. Radius of gyration showed a positive effect beyond 3 km, with diminishing marginal effects after 10 km, especially in Chengdu. These findings highlight the methodological importance of incorporating mobility into heat-related assessments and provide evidence for designing targeted planning and adaptation strategies to reduce urban heat risks.
扫码关注我们
求助内容:
应助结果提醒方式:
