Time-Space variability of remote sensing and model predictions of skin temperature

Remote sensing estimates of skin temperature offer an attractive means for updating and/or verifying the predictions of land surface schemes. Skin temperature is an important state variable which is closely linked both to the surface energy balance, and to soil moisture. Therefore, the ability for accurate estimation of skin temperature remotely could have important implications for improving the predictability of land surface fluxes of moisture and energy. This paper examines the variability in space and time of surface temperature as derived by a macroscale hydrologic model and as observed by satellites for the Arkansas-Red River basin of the central U.S. Preliminary results show that remotely sensed skin temperature is more variable, both spatially and temporally, than modeled surface temperature. Studies over the Arkansas-Red River basin using the Variable Infiltration Capacity (VIC) hydrologic model and data from TOVS (TIROS Operational Vertical Sounder) have shown that the variability of the model and the satellite track along the same trends but do not exhibit the same magnitude of variability. Previous evaluations of satellite skin temperatures have been based primarily on monthly mean temperatures, which we find to be in relatively good agreement with model predictions, as are basin average temperatures. However, the potential uses for the satellite data are at much shorter time intervals, for which reason this study is concerned with instantaneous observations on the TOVS pixel (1 degree) scale. At this time scale, the satellite values contain errors that artificially increase the range of remotely sensed surface temperature. For example TOVS Ts will sometimes change by +- 35 degrees C in 24 hours while the VIC range is only about +- 20 degrees C. The largest differences in variability appear to occur at times of high incoming solar radiation both diurnally and seasonally, i.e. afternoon and summer. The higher variability of the TOVS data relative to the model predictions are no doubt due in part to TOVS estimation errors. However, there appears to be a depression of space-time variability of the model predictions in summer which is somewhat counterintuitive. This paper reviews diagnostic studies of sources of space-time variability in both the model and TOVS estimate, utilizing related explanatory variables such as precipitation, vegetation, air temperature and solar radiation. The intent is to gain insight into the validity of the modeled energy balance on a spatially continuous and temporally consistent basis. Before the availability of remote sensing data sets this level of precision in model validation was not possible.