A Method to Derive Long-Term Global Hourly Near-Surface Air Temperature by Combining Remote Sensing and Reanalysis Datasets

Near-surface air temperature (Ta2M), defined as the temperature at 2 m above the ground, is influenced by various dynamical, radiative, and surface-atmosphere exchange processes. Despite numerous methods are developed to estimate daily or monthly mean temperatures, high spatiotemporal resolution products remain scarce. Longwave radiation, land surface temperature (LST), and total column water vapor (TCWV) are considered driving factors for retrieving Ta2M. This study proposes a new method that integrates the strengths of multiple products to generate a balanced global hourly averaged Ta2M dataset. Initially, this study proposes a reconstruction approach to derive longwave downward radiation (LWDR) based on random forest regression (RFR) by combining remote sensing and reanalysis datasets. Furthermore, a multiple linear regression approach is employed to model the interactions among Ta2M, longwave radiation, LST, and TCWV, resulting in a global long-term Ta2M product characterized by a spatial resolution of 0.05 degrees and an hourly temporal resolution. The produced dataset exhibits reasonable accuracy, with root mean square error (RMSE) less than 3.1 K and bias less than 0.4 K. The method proposed in this study offers innovative strategies for estimating Ta2M from remote sensing and reanalysis datasets. It provides valuable insights into the dynamics of radiative and surface-atmosphere exchange processes, contributing to a more comprehensive understanding of Earth's climate system.