Comparison of land surface temperature (LST) modeled with a spatially-distributed solar radiation model (SRAD) and remote sensing data

Abstract

The solar radiation model SRAD was applied to a Mediterranean mountainous environment in southern California for estimating land surface temperature (LST). The simulated SRAD LST results were compared with high- and meso-resolution satellite-based LST data at daily, monthly and annual temporal scales to identify potential ways of improving the LST accuracy in either the SRAD or satellite-based approaches. It was found that mean monthly and annual LST from SRAD closely matched the MODIS LST observations (but for the fact they were from 0.7 to 1.5 °C lower) while the daily LST from SRAD agreed less well with ASTER observations and were 2.1 °C and 4.8 °C higher for 01/23/05 and 07/21/06, respectively. High mountainous, steep and south/north facing slopes resulted in large discrepancies in LST estimates and the winter LST estimates are more sensitive to terrain factors and their associated land use/cover characteristics than the summer estimates. It is therefore suggested that vertically variable elevation lapse rates, spatially distributed surface albedo and leaf area index for the time-specific simulations that represent the heterogeneity of land surface characteristics be used in SRAD inputs. The results also show how instantaneous model outputs rather than lumped average daily outputs offer better comparisons with satellite based data as well.