Separating snow and forest temperatures with thermal infrared remote sensing

Thermal infrared sensing from space is a well-developed field, but mixed pixels pose a problem for many applications. We present a field study in Dana Meadows, Yosemite National Park, California to scale from point (similar to 2-m resolution) to aerial (similar to 5-m resolution gridded, 1 km x 6 km extent) to satellite (MODIS, similar to 1000-m resolution, global extent) observations. We demonstrate how multiple thermal bands on MODIS can be used to separate snow and forest temperatures and determine the fractional snow-covered area (f(scA)) over a 3 km x 3 km array of 9 MODIS grid cells. During the day, visible, near-infrared, and shortwave-infrared bands provide a first guess of f(scA) and help to constrain the solution. This technique, which has estimated errors < 2 degrees C and 10% f(scA) for many expected conditions, enables better understanding of the snowpack energy balance, atmospheric inversions and cold air pools, and forest health.