Comparing Surface and Canopy Urban Heat Islands in Hong Kong with Population Heat Exposure: Integrating Satellite LST and Ground Weather Observations

Abstract

Satellite land surface temperature (LST) is widely used to map urban heat, yet it characterizes the radiative state of surfaces rather than the near-surface air that people experience. This study quantitatively compares the surface urban heat island (SUHI) and the canopy-layer (air-temperature) urban heat island (CUHI) over Hong Kong, a high-density coastal subtropical city, and evaluates their implications for population heat exposure. SUHI is derived from Moderate Resolution Imaging Spectroradiometer (MODIS) day/night LST (1 km) and Landsat Collection-2 surface temperature (100 m) for the 2023–2024 warm season, after water masking and an elevation-corrected non-built reference; CUHI is derived from 25 Hong Kong Observatory (HKO) automatic weather stations. Population heat exposure is then modeled with WorldPop 100 m population under two metrics: a surface-temperature proxy and a station-interpolated air-temperature field (leave-one-out mean absolute error 0.3–0.4 °C). Three findings emerge. (1) The surface heat island is much larger in magnitude than the canopy heat island (SUHI spatial range 6–14 °C versus CUHI ~2 °C). (2) Their spatial coupling is weak and scale-dependent: at 1 km no statistically robust station-level association is found (day and night bootstrap confidence intervals include zero), and only a moderate daytime rank association emerges at 100 m (Spearman ρ=0.53, p=0.01). (3) The two exposure metrics diverge: a relative surface proxy ranks 79% of residents in upper-quartile surface heat, whereas the air-temperature metric—tied to local heat-health thresholds—identifies 10% of residents (7% of the elderly) exposed to very-hot-day air temperatures (≥33 °C). Overall, LST usefully maps the surface thermal environment but cannot, without calibration, substitute for canopy-level heat exposure in a high-density coastal city.