Preliminary Validation of the Global Hourly, 5-km, All-sky Land Surface Temperature (GHA-LST) dataset

Land Surface Temperature (LST), mainly retrieved by Thermal Infrared (TIR) remote sensing, is a key component of the Surface Energy Budget (SEB) and plays a vital role in atmospheric cycling and hydrological balance, and the Diurnal Temperature Cycle (DTC) is an essential driving factor for estimating evapotranspiration, soil moisture, and heatwaves; however, due to the cloud cover, numerous data gaps exist in current TIR LST datasets, severely restricting their applications. Studies have been focusing on reconstructing cloud gaps and producing all-sky LST products, either by interpolation or fusing TIR LST with modeled skin temperature or passive microwave-based LSTs. However, no studies are working on recovering diurnal hourly LST at the global scale. Considering its high temporal variability and importance, we revised the current SEB-based cloudy-sky LST estimation scheme and released a global hourly, 5-km, all-sky land surface temperature (GHA-LST) dataset from 2011 to 2021. In the first step, a temperature time-evolving model was built by the temperature series from ERA5 reanalysis data; in the second step, clear-sky LSTs products from multiple geostationary and polar-orbiting satellites were filtered by the time-evolving model using a regression-Kalman filter spatiotemporally; in the last step, the cloud effect was estimated and superposed to the fusion results from global surface radiation products.

The preliminary validation was implemented using ground measurements from the SURFRAD network from 2011-2020. The overall RMSE of the clear-sky samples was 2.76 K with a bias of -0.70 K and R2 of 0.97, and the RMSE of the cloudy-sky samples was 2.73 K with a bias of -0.17 K and R2 of 0.95, producing higher and more stable accuracy than the MYD21C1. DRA and TBL had larger RMSEs mainly due to the high elevation that reduced the site representativeness at the 5-km scale. Temporal continuity analysis suggests that GHA-LST had a consistent temporal variation at hourly and daily-mean scales, which captured both regular diurnal patterns and the variation in continuous cloudy days. GHA-LST will be comprehensively assessed and potentially used in improving hydrological analysis, weather modeling, and meteorological forecasting. GHA-LST is available at the University of Maryland (glass.umd.edu/allsky_LST/GHA-LST).