Stream-temperature estimation from thermal infrared images

Stream temperature is an important water quality indicator. Spatial gradients of stream temperature can also be used to identify groundwater and surface water input locations in channel systems. Endangered fish populations are sensitive to elevated stream temperature, especially in the summer when low precipitation and high solar insolation increase temperatures beyond established thresholds. The removal of riparian vegetation and increases in surface run-off, that results from land-use change, also contribute to elevated stream temperatures. Thus, if critical watersheds are to be managed properly, accurate and spatially extensive temperature measurements are needed. For these purposes, it is necessary to know water surface temperature within 1/spl deg/C. Thermal infrared images (TIR) have long been used to estimate water surface temperatures, especially of the ocean where split-window techniques have been used to compensate for atmospheric effects. Streams are a more complex environment because 1) most are unresolved in typical thermal infrared images, and 2) stream corridors may consist of tall trees that irradiate the stream surface. Therefore, key additional problems to solve in measuring stream temperatures include both subpixel unmixing and multiple scattering. Over a watershed In the Cascade mountains of southern Washington, USA, we use fine-resolution airborne MODIS/ASTER Simulator (MASTER) data, and coarse-resolution ASTER data, to develop an approach for successful extraction of stream temperatures.