Divergent transpiration regulation among 20 urban tree species under extreme heat revealed by UAV multisource remote sensing

Urban areas face increasing extreme heat risk, yet satellite remote sensing struggles to quantify crown-scale transpiration in heterogeneous urban landscapes. Consequently, species-specific heat responses, key regulatory factors, and tipping points of urban tree transpiration remain poorly understood. Here, we integrated unmanned aerial vehicle (UAV) multispectral, thermal infrared, and LiDAR observations with field surveys of 665 trees across 20 species to estimate crown-scale transpiration for nearly 50,000 trees. The UAV-based single-source energy balance model captured diurnal evapotranspiration (ET) dynamics and agreed well with eddy covariance measurements (R 2 = 0.86, RMSE = 0.04 mm h −1 ). At the individual-tree scale, UAV-derived crown-scale transpiration agreed closely with sap-flow-derived transpiration (R 2 = 0.88, RMSE = 0.05 mm h −1 ) and leaf area index (LAI)-upscaled portable-photosynthesis-system transpiration estimates (R 2 = 0.88, RMSE = 0.06 mm h −1 ). Compared with the normal day, landscape ET was higher on the extreme-heat day, mainly reflecting higher tree transpiration (0.47 vs. 0.40 mm h −1 ), whereas grassland and shrub ET remained relatively stable. At the species level, three event-specific response groups were identified: enhanced with significantly higher transpiration under extreme heat (12 species), neutral with no significant change (3 species), and suppressed with significantly reduced transpiration (5 species). Random forest analysis identified the temperature vegetation dryness index (TVDI) and leaf-to-air vapor pressure deficit (VPD L ) as dominant controls of crown-scale transpiration, with tree height (TH) also exerting a strong influence. SHAP-derived tipping points for TH were identified at 7.01, 4.42, and 3.65 m in the enhanced, neutral, and suppressed response groups, respectively. SHAP-derived tipping points for TVDI (0.23–0.34) and VPD L (5.34–5.84 kPa) suggested the transition toward soil-water limitation and atmospheric-demand regulation. Overall, this framework enables spatially explicit assessment of urban tree water use and provides a basis for species selection and irrigation management during extreme heat.