Assessment of gridded datasets of various near surface temperature variables over Heihe River Basin: Uncertainties, spatial heterogeneity and clear-sky bias

Abstract

Near-surface temperatures, such as air, land surface, and soil temperatures, play significant roles in surface radiation and energy balance. This study assessed nine gridded near-surface temperature products and analyzed the spatial heterogeneity and clear-sky bias of these temperature variables, using extensive measurements collected at Heihe River Basin. The MXD21 (MOD21 and MYD21) product had the lowest root mean square error (RMSE) (3.35 K) among all skin temperature products but a high percentage of missing values (48.4 %). All-weather skin temperature products had comparable accuracy for the interpolated cloudy-sky cases (RMSE 4.92 K) and observed clear-sky pixels (RMSE 3.42 K). For air temperature, AMSR2 had the lowest RMSE (2.48 K), but a high percentage of invalid data (32.5 %); and ERA5 had a worse accuracy (RMSE 3.87 K) but a high spatial resolution and gap-free data coverage. Comparing products from the same data source, air and soil temperatures had higher accuracies than skin temperature. Among the different variables of temperature, the 0 cm soil temperature and skin temperature had higher spatiotemporal heterogeneity than the air temperature and the soil temperatures at greater depths. The skin temperature, 0 cm soil temperature, and air temperature had higher clear-sky biases compared to soil temperatures.

Keywords

• Temperature
• Near surface temperature
• Land surface temperature
• Air temperature
• Soil temperature
• Validation