Requirements for satellite land surface temperature validation using a silt playa

We investigate the validation of satellite land surface temperature (LST) products. In particular, the planned EOS/MODIS global LST accuracy of 1°C must be confirmed by independent measurements for a range of cover types. Simple fat surfaces are the most easily instrumented and characterized and so provide a logical starting point. Such surfaces include inland water, sand, snow, ice, and playa. Validation with water is straight-forward, but other land surfaces are more difficult to measure accurately. The analysis we present is based on a sequence of MODIS Airborne Simulator (MAS) images collected over a playa (silt dry lake bed). Atmospheric effects in the MAS images are corrected by an in-scene method and are then converted to a time sequence of temperature maps. These form the basis for estimating the spatial and temporal sampling requirements to achieve a surface measurement accuracy of better than 0.5°C for 1km2 MODIS pixels. We conclude that a grid of thermistors appears to be the must practical instrumentation method for ,such a large area. Depending on the environmental conditions, between 4 and 16 points are required per km2 and we recommend that these be sampled at intervals ranging from 1 min to 2.5 min. A MAS underfight validation is possible if it is synchronized to within several minutes of the MODIS overpass. Finally, radiometers should be used at several of the thermistor locations to verify the skin temperature corrections and to characterize the temperature transients and angular effects.