Heat–Drought Resonance in a Tropical Highland City: Local Climate Zone–Scale Explainable Machine Learning and Shared Socioeconomic Pathway Projections for Kunming, China

Abstract Escalating compound heat–drought risks in cities remain poorly understood in terms of seasonal differentiation and spatial heterogeneity, limiting the development of targeted adaptation strategies. Urban heat and surface drought intensified markedly in Kunming during 2000–2024, with clear seasonal asymmetry. Summer land surface temperature (LST) increased from approximately 22.5°C to 26.8°C, accompanied by a 49% rise in the mean temperature vegetation dryness index (TVDI), whereas winter warming and drying were weaker. A significant city-scale summer heat–drought resonance was observed ( R 2 = 0.62, P < 0.01), while the winter LST-TVDI relationship was notably weaker. Strong spatial heterogeneity was identified across local climate zones (LCZs): bare soil/sand showed the strongest summer LST-TVDI coupling ( R 2 = 0.60), whereas densely vegetated zones exhibited weak coupling, indicating a clear ecological buffering effect. A driver analysis further showed that the summer TVDI was mainly controlled by LST, digital elevation model, and the enhanced vegetation index (EVI), while the winter TVDI was more strongly influenced by the EVI, precipitation, and soil moisture; nighttime lights contributed only marginally in both seasons. Future projections revealed substantially greater late-century warming under SSP5-8.5, with summer temperatures increasing by about 5°C and winter temperatures by about 4°C by 2100, whereas warming under SSP1-2.6 was limited to approximately 1°C in both seasons. These findings demonstrate that heat–drought risks in tropical highland cities are seasonally differentiated and strongly LCZ-dependent, highlighting the need for targeted heat mitigation and moisture-retention strategies in exposed urban zones.