Combined effects of blue-green space on land surface temperature and PM2.5 and its spatio-temporal heterogeneity in Urumqi City, China

Blue-Green Space (BGS) cools cities and reduces air pollution through local climate regulation. However, the interactions between BGS, thermal conditions, and air quality in arid regions remain insufficiently understood. In this study, after descriptive analysis, we applied Geographically and Temporally Weighted Regression (GTWR) to quantify how Blue-Green Space Indices (BGSIs) influence LST and PM2.5 and to assess their spatiotemporal heterogeneity. Bootstrapping mediation analysis was further applied to examine the indirect effects of BGSIs on LST and PM2.5. The results indicated an overall reduction in LST and PM2.5 levels within the study area during the investigation period. Compared with Ordinary Least Squares Regression (OLS) and Geographically Weighted Regression (GWR), GTWR demonstrated superior fitting performance. All three BGSIs including patch density (PD), percentage of landscape (PLAND), and mean shape index (SHAPE_MN) reduced LST. PD exhibited a positive overall effect on PM2.5, while PLAND and SHAPE_MN showed negative effects. PM2.5 mediated part of the effects of PLAND and SHAPE_MN on LST, whereas LST also mediated their influence on PM2.5. PLAND exerted a significant negative impact on both LST and PM2.5 at threshold values of 62 % and 72.46 %, respectively, whereas no significant threshold effect was detected for PD and SHAPE_MN. • Blue–green spaces simultaneously mitigate land surface temperature and reduce PM2.5 concentrations in arid regions. • GTWR model delivers optimal performance in simulating air pollution, due to its consideration of spatiotemporal heterogeneity. • Landscape metrics exert distinct impacts on land surface temperature and air pollution, and the two exhibit mutual mediation. • Critical thresholds exist for blue-green space coverage, with shape complexity being a key factor.