Uncertainty associated with the Remote-sensing of stream temperatures at multiple spatial scales

Stream temperature is an important water quality indicator, particularly in the Pacific Northwest where endangered fish populations are sensitive to elevated water temperature. Regional assessment of stream temperatures from the ground is limited by sparse sampling of temperatures in both space and time. Remotely sensed thermal infrared (TIR) images can be used to derive spatially distributed estimates of the radiant skin temperature of streams. Even for fully resolved streams, however, our limited ability to compensate for atmospheric and emissivity effects constrains the accuracy of stream temperature measurements. When the stream is not fully resolved, it is the effect of mixed image pixels and thermal scattering from the near-bank environment that dominates both the accuracy and uncertainty of stream temperature measurements. Whereas airborne thermal sensors can resolve smaller streams than spaceborne sensors, their limited spatial coverage and cost makes satellite-derived TIR data an attractive option. We examine the accuracy and uncertainty associated with recovered stream temperatures for a range of stream sizes and pixel resolutions, using remotely sensed airborne (MASTER) and satellite (ASTER, LANDSAT-TM) TIR measurements. We find that when the stream width is resolved by fewer than 3 pixels, the accuracy of TIR measurements significantly decreases, and the uncertainty is greater, largely due to the reduction of mixed pixels (bank and stream). This limits the use of satellite TIR to temperature recovery for large rivers (currently, ~180 m stream widths with Landsat- TM)