Remote sensing modeling of land surface temperature

This study addresses the fundamental science question of modeling land surface temperature (LST) by combining the field experiment, spatial interpolation, and satellite thermal infrared (TIR) remote sensing. Specific objectives of this study are to: (1) calibrate the spatial interpolation of LST using satellite-derived surface emissivity; (2) calibrate the conversion from at-satellite brightness temperature to LST using surface variables; and (3) simulate and predict the air temperature profiles in a forest canopy using field observations and satellite TIR data. To this end, Landsat Enhanced Thematic Mapper Plus (ETM+) TIR data were applied to derive surface emissivity and brightness temperature. Hourly temperature observations at national weather stations (NWS) within the study area were extracted from National Climatic Data Center (NCDC) and used to interpolate temperature surfaces. Field-observed temperatures were used to evaluate the interpolation process and to develop algorithms for predicting air temperature profiles in a forest canopy. The results showed that no single interpolation method was capable of obtaining accurate LST from weather station measurements. The accuracy of interpolated LST for the experimental site was significantly improved by calibration using satellite-derived surface emissivity. Brightness temperature derived from Landsat TIR data were well corrected using surface variables such as surface emissivity and solar zenith angle (SZA). The results also showed that split-window method can be adopted to estimate LST using brightness temperatures derived from Landsat-7 low-gain and high-gain TIR data. Twenty-four polynomial models were developed using field observations to simulate and predict the air temperature profiles in a forest canopy at any hour during a summer day. The results showed that the effects of diurnal and seasonal variations of the temperature had to be considered in predicting air temperatures profiles in forest canopy. This study demonstrated that the combination of satellite thermal remote sensing, spatial interpolation, and field empirical models is capable of obtaining accurate LST from Landsat TIR data and predicting air temperature profiles in forest canopy.