Validation of land surface models using satellite‐derived surface temperature

Abstract

This research examines the feasibility of using remotely sensed surface temperature for validation and updating of land surface hydrologic models. Surface temperature simulated by the Variable Infiltration Capacity (VIC) hydrologic model is compared over the Arkansas‐Red River basin with surface temperature retrievals from TOVS and GOES. The results show that modeled and satellite‐derived surface temperatures agree well when aggregated in space or time. In particular, monthly mean temperatures agree on the pixel scale, and basin mean temperatures agree instantaneously. At the pixel scale, however, surface temperatures from both satellites were found to have higher spatial and temporal variabilities than the modeled temperatures, although the model and satellites display similar patterns of variability through space and time. The largest differences between modeled and remotely sensed surface temperature variability occur at times of maximum net radiation both diurnally and seasonally, i.e., afternoon and summer. Comparison of temporal and spatial patterns of VIC‐predicted surface temperature variability with similar predictions by nine other models involved in the PILPS‐2c experiment show that the VIC patterns are similar to those of the other models. Observed surface temperature and air temperature from FIFE are used to identify possible errors in satellite‐retrieved surface temperatures. The FIFE comparisons show that satellite retrieved surface temperatures likely contain errors that increase variability.