Validation of the land-surface temperature products retrieved from Terra Moderate Resolution Imaging Spectroradiometer data

Abstract

This paper presents the status of land-surface temperature (LST) standard products retrieved from Earth Observing System (EOS) Moderate Resolution Imaging Spectroradiometer (MODIS) data. Based on estimates of the channel-dependence error and noise equivalent temperature difference (NEDT) and the calibration accuracy of MODIS thermal–infrared data, the impact of instrument performance on the accuracy of LST is discussed. A double-screen scheme based on the difference between the 5-km LST retrieved by the day/night LST algorithm and the aggregated 1-km LST retrieved by the generalized split-window algorithm, and the difference between daytime and nighttime LSTs, is proposed to remove the LSTs contaminated with cloud effects. The accuracy of daily MODIS LST product at 1-km resolution, which was produced by the generalized split-window algorithm, was validated in 11 clear-sky cases with in situ measurement data collected in field campaigns in 2000 and 2001. The MODIS LST accuracy is better than 1 K in the range from 263 to 300 K over Lake Titicaca in Bolivia, Mono Lake, Bridgeport grassland, and a rice field in Chico, CA, and Walker Lake, Nevada (NV), in the atmospheric column water vapor range from 0.4 to 3.0 cm. It is difficult to validate the daytime LST product over land sites rather than lakes with ground-based measurements alone because of the high spatial variations in the in situ LST measurement data, which was verified by the daytime data of the MODIS Airborne Simulator (MAS) over a grassland in Bridgeport, CA, on October 6, 2000. In six cases over a silt playa in Railroad Valley, NV, the 1-km MODIS LSTs are a few Kelvin degrees lower than the in situ measured LSTs because the surface emissivities inferred from land cover types in the split-window LST method are often overestimated in semi-arid and arid regions. After a correction with the difference between the 5-km LST retrieved by the day/night LST method and the LST aggregated from 1-km LSTs retrieved by the split-window method, the MODIS LSTs agree with in situ measured LSTs within ±1 K in the range 263–322 K for the six cases in Railroad Valley and one case of snowcover in Bridgeport, CA, leading a recommendation for use of the 5-km LST product retrieved by the day/night LST method in bare and sparse vegetated areas.