An atmospheric radiosounding database for generating land surface temperature algorithms

A database of global, cloud-free, and atmospheric radiosounding profiles was compiled with the aim of simulating radiometric measurements from satellite-borne sensors in the thermal infrared. The objective of the simulated data is to generate split-window (SW) and dual-angle (DA) algorithms for the retrieval of land surface temperature (LST) from Terra/Moderate Resolution Imaging Spectroradiometer (MODIS) and Envisat/Advanced Along Track Scanning Radiometer (AATSR) data. The database contains 382 radiosounding profiles acquired over land, with nearly uniform distribution of precipitable water between 0.02 and 5.5 cm. Radiative transfer calculations were performed with the MODTRAN 4 code for six viewing angles between 0 degrees and 60 degrees. The resulting radiance spectra were convoluted with the response filter functions of MODIS bands 31 and 32 and AATSR channels at 11 and 12 pm. By using the simulation database, the SW algorithms adapted for MODIS and AATSR data and the DA algorithms for AATSR data were developed. Both types of algorithms are quadratic in the brightness temperature difference and depend explicitly on the land surface emissivity. The SW and DA algorithms were validated with actual ground measurements of LST collected concurrently to MODIS and AATSR observations in a site located close to the city of Valencia, Spain, in a large, flat, and thermally homogeneous area of rice crops. The results obtained have no bias and a standard deviation around +/- 0.5 K for the SW algorithms at nadir for both sensors. The SW algorithm used in the forward view results in a bias of 0.6 K and a standard deviation of +/- 0.8 K. The worst results are obtained in the other algorithms with a bias close to -1.0 K and a standard deviation close to +/- 1.1 K in the case of the DA algorithms.