Detection of Urban Heat Persistence Zones Using Random Forest-Based LST Downscaling in Cagayan de Oro City, Philippines

The surface urban heat island (SUHI) is one of the emerging environmental and public health issues in rapidly urbanized tropical areas. As nocturnal heat retention has greater health implications and limits micrometeorological cooling more than diurnal heat retention, nocturnal Urban Heat Persistence Zones are determined by quantitative, numerical analysis in Cagayan de Oro City, Philippines. This study downscaled 1 km coarse-resolution Moderate Resolution Imaging Spectroradiometer (MODIS) LST to a commonly available 30 m resolution using a Random Forest Regression. This prediction scheme combined high-resolution secondary predictors such as NDVI, NDBI (Normalized Difference Built-up Index) from high-resolution Sentinel-2 multispectral imagery, and DEM for correction of anomalies induced by topography. With R² values of 0.86–0.99 and RMSE values of 0.333°C (day) and 0.139°C (night) for both prediction schemes, the downscaled diurnal and nocturnal LST can reliably be used in a spatially detailed study. Based on analysis of the Diurnal LST Difference, locations with limited nighttime thermal dissipation were determined; prominent thermal sinks occurred only within the highly NDBI-characterized, high-built-up-density urban core, with only a 3.15°C nighttime cooling compared to the daytime temperature. These findings of localized, high-resolution spatial data could be utilized by Local Government Units (LGUs) to inform their investment in climate adaptation and green infrastructure development.