Cloud filtering of MODIS SCE dataMo

Cloud filtering of MODIS SCE dataMost existing cloud-filter algorithms designed for theMODIS SCE products use empirical relationships betweensnow cover conditions and ancillary data to predict the snowcover occurrence for cloud-covered pixels (e.g., Parajka andBlöschl, 2008; Gafurov and Bárdossy, 2009; Parajka et al.,2010). The empirical relationships are generally appropriatefor the limited areas or conditions in which they were developed and may not be suitable for other regions with differentclimates or topography. To develop a more general cloudfilter algorithm, we exploited spatial interpolation methodsoriginally designed for generating grid-based surface meteorology from in situ weather station observations. We used asimilar methodology that was used to generate Daymet surface precipitation, which uses a truncated Gaussian weighting filter and accounts for the dependence of precipitation onelevation (Thornton et al., 1997). This method was found togenerate reliable precipitation estimates in complex topography in the western USA (Henn et al., 2018). For our application, we treated the pixels without cloud cover as “stationobservations” and then used the spatial filter to predict theoccurrence of snow in cloud-contaminated pixels and gener