Land-surface temperature measurement from space: physical principles and inverse modeling

To apply the multiple-wavelength (split-window) method used for satellite measurement of sea-surface temperature from thermal-infrared data to land-surface temperatures, the authors statistically analyze simulations using an atmospheric radiative transfer model. The range of atmospheric conditions and surface temperatures simulated is wide enough to cover variations in clear atmospheric properties and surface temperatures, both of which are larger over land than over sea. Surface elevation is also included as the most important topographic effect. Land covers characterized by measured or modeled spectral emissivities include snow, clay, sands, and tree leaf samples. The empirical inverse model can estimate the surface temperature with a standard deviation less than 0.3 K and a maximum error less than 1 K. A band in the region from 10.2 to 11.0 mu m will usually give the most reliable single-band estimate of surface temperature. A band in either the 3.5-4.0- mu m region or in the 11.5-12.6- mu m region must be included for accurate atmospheric correction.<
>