Potential biases in remotely sensed urban climates as revealed by urban thermal anisotropy mapping across global cities

Urban thermal anisotropy (UTA) significantly distorts satellite-derived land skin-surface temperatures (LST) and surface flux estimates, posing a major challenge to understanding global urban climate dynamics. Despite decades of research, a comprehensive map of urban thermal anisotropy across global cities with varying climatic contexts remains a knowledge gap in Earth observation. This gap hinders accurate estimation of critical urban climate variables and undermines the reliability of a wide range of urban climate studies that increasingly rely on satellite-based thermal data. Leveraging the extensive archive of multi-angle thermal remote sensing data, here we present a novel, statistically robust, data-driven approach that departs from traditional complex model-based methods to directly quantify urban thermal anisotropy across global cities. Our findings reveal that the global mean UTA intensity exhibits seasonal variation, peaking at 5.1 K during summer daytime. Compared to nadir LST measurements, UTA-induced biases in satellite-derived urban sensible heat flux and surface urban heat island intensity can lead to substantial underestimations (> 40.0%) when using LST data from sensor viewing zenith angles (VZAs) of ±60°. However, using LST data from sensor VZAs within ±30° can limit these errors to within ±10%. Finally, we formulate a data-driven, globally applicable approach to correcting angle-dependent biases in satellite thermal observations across global cities.