Abstract
Vegetation cooling in cities is often represented as a local, isotropic effect, although atmospheric processes can redistribute cooling beyond individual vegetation patches. This simplification may overlook the nonlocal and directionally structured nature of cooling propagation. We developed a physics-guided source–medium–receiver (SMR) framework for the Greater Bay Area of China to compare multi-source isotropic connectivity with prevailing-wind-conditioned connectivity. Vegetation-source strength was represented by the VCSI, and parameters estimated from 2021 to 2023 were frozen for an independent 2024 holdout and contrasting monsoon assessments. Cooling connectivity was modeled using a spatially varying distance-decay function, controlled by the wind speed, moisture content, surface albedo, terrain, and directional constraints from prevailing winds. In 2024, isotropic connectivity showed a stronger relationship with LST (r = −0.8069) than wind-conditioned connectivity (r = −0.7780) and local VCSI (r = −0.7757), with the same general ordering across monsoon and built-up receiver assessments. The source-level decay lengths ranged from 0.504 to 0.518 km, whereas the wind-conditioned emergent multi-source persistence increased from 4.13 to 4.88 km, reflecting cumulative field behaviors. Wind-relative decomposition further showed that annual connectivity was distributed across downwind, crosswind, and upwind pathways. These findings indicate that regional vegetation cooling is primarily characterized by distance-mediated, multi-source connectivity, while prevailing-wind direction acts as a scale- and condition-dependent modifier. This shift in perspective provides a physically grounded framework for effective urban climate analysis and precision greening strategies.