Quantifying urban tree canopy cooling capacity using Bayesian hierarchical models and satellite imagery

Abstract

The urban heat island phenomenon significantly affects thermal comfort in cities. Urban trees offer ecosystem services that can help cool built-up areas, making cities more livable environments. Regarding the cooling capacity associated with increasing urban trees canopy cover, it is often analyzed at the city level. This study aimed to quantify the local-scale cooling benefits of increasing tree canopy cover to inform urban planning. Using Landsat 8 satellite imagery to measure land surface temperature, we applied a Bayesian hierarchical models (BHMs) with Integrated Nested Laplace Approximation (R-INLA) to estimate cooling effects in 900 m2 urban units, accounting for other factors and spatial autocorrelation through a Gaussian field. The model achieved a root mean squared error of 0.989°C ± 0.255°C and 0.833 ± 0.06 of observations were covered by the Highest Posterior Probability Interval of the Posterior Predictive Distribution (p-PPD-HPDI), both metrics calculated in the test sets of 10-folds spatial cross-validation. Considering other factors responsible of land surface temperature distribution and spatial autocorrelation using a Gaussian field, the model shows that increasing canopy cover in an observational unit by 450 m2 is associated with a reduction of about 0.268°C (Quantile0.025 0.241, Quantile0.975 0.295), ceteris paribus. These findings highlight the significant local cooling potential of urban tree planting at scales relevant to planning decisions. By quantifying these effects, the study underscores the value of integrating tree canopy increase into urban design and policy to enhance comfort, resilience, and overall urban sustainability.