Predicting urban air temperatures using land cover type and satellite observations of surface temperatures

Extreme urban heat is known to adversely impact humans and the environment, where certain land use/land cover (LULC) types may amplify temperatures. However, sparsely available air temperature (Ta) data limits study of these impacts. Attempts to map air temperature from satellite land surface temperature (LST) data are often highly empirical and lack sufficient data for robust evaluation. In particular, we do not know: 1) how well do predictions perform across diverse land cover characteristics? And 2) what insights can we gain from predictions based on expected biophysical air surface temperature relationships? In this study, we derived an LST-Ta relationship from a surface energy balance to fit LULC-specific LST-Ta predictive relationships (biophysically-based), benchmarked against a simple linear regression fit. We used satellite LST data and spatially-extensive (\u3e 1 million samples) air temperature maps from sampling campaigns during heat wave days in five U.S. cities, including Portland. Results showed LULC had a large impact on LST and Ta values, e.g., more developed areas had higher temperatures than forested (10 ℃ LST and 1 ℃ Ta differences). Both the linear and biophysical models performed well in predicting air temperatures (RMSE 0.50 and 0.49 ℃, respectively); however, biophysical fitted model coefficients corresponded better to LULC characteristics (i.e. vegetation or imperviousness). Using this approach, this suggests some ability to resolve differences in underlying mechanisms of heat transfer among LULCs. Quantifying such relationships in urban landscapes is critical in adapting and managing cities that often face inequitable exposure to heat from historical disinvestment and segregation