Non-Linear Responses of Urban Land Surface Temperature to Vapor Pressure Deficit and Vegetation Density in an Arid Oasis City

Urban warming threatens arid oasis cities, yet vegetation cooling efficiency is non-linearly moderated by atmospheric dryness and structural density. Using a multi-year summer satellite composite from 2020 to 2025 in Urumqi, China, this study calibrated a locally continuous fractional vegetation cover (FVC) model, controlled for desert thermal advection, impervious surface fraction (ISF), and elevation (DEM) using a spatial error model, and detected statistical vapor pressure deficit (VPD) breakpoints consistent with transpirational cooling attenuation via segmented regression. Mean summer land surface temperature reached 43.16 °C under a satellite-overpass hourly VPD of 2.13 kPa. Spatial error correction significantly improved model fit (R2 = 0.5585 for OLS; SEM λ = 0.8003, p < 0.001), reducing the Akaike Information Criterion by 1752.86. We identified a critical satellite-overpass VPD breakpoint at 2.83 kPa [95% CI: 2.73, 2.92 kPa], beyond which the vegetative cooling efficiency experienced severe attenuation, with absolute temperatures rising due to overarching atmospheric heating. High-density canopies (FVC > 0.6) showed an estimated tendency to reach this attenuation threshold earlier at at 2.83 kPa [95% CI: 2.59, 3.07 kPa] compared to sparse vegetation at 2.96 kPa (95% CI: [2.83, 3.51] kPa). Although the confidence intervals overlapped considerably, this tendency reflects plausible earlier water stress in dense stands under extreme moisture demand. A non-vegetated control group (FVC < 0.05) exhibited monotonic warming (+5.34 °C/kPa, p < 0.001) without any cooling branch, confirming that the observed threshold reflects statistical associations plausibly linked to vegetative biophysical regulation rather than thermodynamic co-dependency. These findings provide vital, threshold-guided design principles for climate-adaptive urban greening and sustainable water management in dryland cities.