Land Surface Temperature From the Advanced Along-Track Scanning Radiometer: Validation Over Inland Waters and Vegetated Surfaces

The land surface temperature (LST) product of the Advanced Along-Track Scanning Radiometer (AATSR) was validated with ground measurements at the following two thermally homogeneous sites: Lake Tahoe, CA/NV, USA, and a large rice field close to Valencia, Spain. The AATSR LST product is based on the split-window technique using the 11- and 12-mu m channels. The algorithm coefficients are provided for 13 different land-cover classes plus one lake class (index i). Coefficients are weighted by the vegetation-cover fraction (f). In the operational implementation of the algorithm, i and f are assigned from a global classification and monthly fractional vegetation-cover maps with spatial resolutions of 0.5 degrees x 0.5 degrees. Since the validation sites are smaller than this, they are misclassified in the LST product and treated incorrectly despite the fact that the higher resolution AATSR data easily resolve the sites. Due to this problem, the coefficients for the correct cover types were manually applied to the AATSR standard brightness temperature at sensor product to obtain the LST for the sites assuming they had been correctly classified. The comparison between the ground-measured and the AATSR-derived LSTs showed an excellent agreement for both sites, with nearly zero average biases and standard deviations <= 0.5 degrees C. In order to produce accurate and precise estimates of EST, it is necessary that the land-cover classification is revised and provided at the same resolution as the AATSR data, i.e., 1 km rather than the 0.5 degrees resolution auxiliary data currently used in the LST product.