A 40-Year 1-km Daily Seamless Near-Surface Air Temperature Product over Yellow River Basin of China

Abstract

Near-surface air temperature (NSAT) is an important environmental parameter; however, there is a lack of long-term, fine-scale NSAT products that offer complete spatiotemporal continuity. Although several available products cover different periods and regions, obtaining gridded NSAT data outside of the provided regions and periods remains challenging. Using multiple source data (ERA5-Land, global land data assimilation system, remotely-sensed data, and several auxiliary factors), a new high-efficient framework for estimating NSAT is presented using the random forest algorithm and implemented on the Google Earth Engine platform in this study. Thereby, the 1-km daily seamless NSAT product (containing daily maximum, mean, and minimum) from January 1st 1981 to December 31st 2020 over the Yellow River Basin in China is generated. To our knowledge, this is the first product that satisfies the conditions of high resolution, seamless coverage, and long-term continuity simultaneously for the Yellow River Basin. Tenfold cross validation shows that the RMSE, MAE, and R2 for the maximum NSAT are 1.746-1.932 K, 1.351-1.486 K, and 0.968-0.974, respectively; for the mean NSAT are 1.219-1.354 K, 0.940-1.035 K, and 0.984-0.987; and for the minimum daily NSAT are 1.663-1.732 K, 1.280-1.322 K, and 0.975-0.997. In addition, a user-friendly NSAT estimation tool was developed for the first time, enabling users to derive NSAT products for specific areas and time periods of interest. Evaluation of the tool indicates that the accuracy of temporally extended NSAT is satisfactory, while the accuracy of spatially extended NSAT is lower compared to the original region. The generated long-term finer-scale NSAT product and the developed NSAT estimation tool hold potential benefits for ecoclimate researchers and environmental policy makers.