Reconstruction and validation of SCA from spectral mixture analysis

As climate continues to change, the empirical methods of managing water, which are based on historical relationships between point measurements and runoff, are likely to become less accurate. Hence the utility of distributed snowmelt models based on a judicious integration of remotely sensed and surface measurements will consequently increase. However, as the analysis in this paper shows, translation of reflectance measurements from MODIS into a product that is useful for hydrologic analyses involves complicated, somewhat arcane knowledge. The daily products have data gaps and errors because of cloud cover and sensor viewing geometry. Rather than make users interpolate and filter these patchy daily maps without completely understanding the retrieval algorithm and instrument properties, we use the daily time series to improve the estimate of the measured snow properties for a particular day. We use a combination of noise filtering, snow/cloud discrimination, and interpolation and smoothing to produce our best estimate of the daily snow cover. We compare the result of smoothed SCA with TM SCA, the precise is 0.98 and RMSE is 0.06, but the RMSE is up to 0.22 and the precise is 0.9 when we compare the result between MYD10A1 and TM.