Satellite-view biases in retrieved surface temperatures in mountain areas

Ground surface temperatures retrieved from satellite data taken in mountain areas may have biases that depend strongly on the satellite viewing angle and the time of day. Biases occur whenever there is a correlation between the viewing angle and the temperatures of subpixel mountainsides, such as when the satellite's view directly faces the most brightly sunlit slopes. This paper reports simulations of this satellite-view bias effect for mountain terrain in central and southwestern Colorado and the southern Sierra Nevada in California. Surface temperatures for the simulations were computed by using a mesoscale numerical weather prediction model with a parameterization for terrain features that are at a scale finer than the model grid resolution. The representation of terrain in the simulations was based on elevation data with high resolution (30 m) from U.S. Geological Survey Digital Elevation Map data sets. Biases as large as 9 C degrees were found while assuming a satellite sensor resolution of 14 km. Even larger biases were found with finer sensor resolutions. These biases are substantially larger than those that had previously been found from modeling with terrain data at about 90 m resolution (3'' latitude/longitude) from the Defense Mapping Agency. The simulations suggest that satellite-view biases could cause substantial local errors when surface temperature retrievals are used for climate diagnosis, weather analysis, soil moisture estimation, or geologic mapping. (C) Elsevier Science Inc., 1997.