Geospatial modeling of urban lake cooling based on morphometric and urban contexts using tree ensembles and SHAP interpretability

In light of increasing heatwave frequency and intensity in Da Nang City, urban lakes play a critical role in mitigating urban heat stress. Focusing on urban lakes in a rapidly urbanizing core of Da Nang, Vietnam, this study establishes a geospatial modeling framework to quantify and explain their cooling intensity. The cooling intensity is defined as the difference between the mean land surface temperature (LST) of all built‑up and bare‑land areas and the LST at a given location. In the study area, the cooling performance of urban lakes has not yet been systematically quantified. To address this gap, we develop an interpretable machine learning framework for lake‑cooling assessment that accounts for both the specific properties of lakes and their surrounding urban context. A set of influencing factors, including distance to shoreline, morphometric metrics of lakes, land use/land cover, urban morphology, proximity features, and topographical factors, is used to characterize the spatial distribution of cooling intensity within 500 m buffers around the lakes. The Categorical Boosting regressor is used to generalize a functional mapping between the influencing factors and the target variable. Experimental results show that the proposed model is capable of effectively estimating the cooling intensity within the sampling zones. This framework also relies on SHapley Additive exPlanations for interpreting the impact of the employed features. By jointly modeling lake morphometry and the surrounding urban context with an interpretable ensemble approach, this study moves beyond purely descriptive assessments of urban lake cooling and identifies model‑inferred patterns in vegetation, bare land, and built‑up densities within the riparian zone. Overall, the findings provide a data‑driven basis for blue–green space planning and for alleviating urban heat stress in Da Nang.