Landsat cloud masks in western South Dakota

This dataset contains masks for six Landsat scenes in the northern U.S. Great Plains. Masks were created manually to enable analysis of the entire available time series (April-October, 1972-2015). Automatic cloud masking was used as a starting point, with data from the Fmask masking algorithm available for reference. Automatic masks were then edited by hand (by the author) to reduce obvious errors. Clouds, cloud shadow, haze, snow, and instrument artifacts were all targeted for manual masking, with a goal of masking data where the land surface was not the primary source of the measured radiance. Areas near masked features were also masked, with automatic masks extending approximately 10 km from detected clouds. The priorities of the masking were to limit cloud/artifact omission error and to limit commission error that occurred consistently in the same spatial location through time, but limiting commission error for any specific Landsat observation was not a high priority. Details of the implementation are described in further detail in the companion publication.