Spatial Heterogeneity in Urban Heat and PM2.5 Co-variation: A Multi-site Sensor Study in Pittsburgh

TZUCHI LIN, Albert Presto, Carnegie Mellon University

     Abstract Number: 103
     Working Group: Urban Aerosols

Abstract
Although urban heat islands (UHIs) and urban pollution islands (UPIs) often spatially coincide, their underlying drivers can vary substantially across urban subregions and may also influence each other. This study explores spatial heterogeneity in the co-variation of PM2.5 and meteorological factors using multi-year, multi-site data from a network of low-cost sensors (RAMPs) deployed across Pittsburgh. The sensor data are supplemented by PM2.5 composition and size distribution data from a central monitoring site. Focusing on temperature and relative humidity as key atmospheric drivers, we examine how PM2.5 and gas-phase pollutants such as CO and NO2, which serve as proxies for emission activity and background accumulation, vary under different diurnal and seasonal conditions. Rather than assuming uniform responses to meteorology, we apply machine learning models to characterize spatial variability and use explainable techniques to assess the relative contribution and interaction of drivers under varying conditions. This helps clarify whether PM2.5 increases are more strongly influenced by thermodynamic partitioning, precursor accumulation, or local emission patterns. Despite similar meteorological settings, subregional differences in PM2.5 behavior are evident, highlighting the interplay between emissions, atmospheric dynamics, and urban form. These findings emphasize the need for spatially resolved approaches to understand UHI and UPI interactions and to inform targeted mitigation strategies.