Landsat-based snow persistence map for northwest Alaska

Landsat imagery for northwest Alaska from 1 February to 31 August, 1985-2011 was used to map snow persistence at high spatial resolution. We analyzed 11,645 scenes covering 505,800 km2, including five Arctic National Park units and the range of the Western Arctic caribou herd (85 Landsat path/rows). A cloud mask was created using the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS). Terrain shadows were calculated from ASTER G-DEM2 and solar incidence angle. The presence of snow cover was determined using separate Snowmap algorithms for non-shadowed and shadowed pixels. Resulting snow cover data were reformatted into 562 30 × 30 km tiles, with an average sample size per pixel of 216 cloud-free observations. A binary classification tree was used to successfully determine the day of the year that best marked the change from snow to snow-free conditions for 99.8% of the study area. An internal consistency check evaluating the occurrence of snow-free data earlier than that day or snow data later than that day, showed that 98.7% of the land pixels were consistently classified ≥ 90% of the time. Comparison with MODIS end of snow season data showed an average difference of 4.2 days. The snow persistence map was strongly correlated with the few SNOTEL stations in the study area (r2 = 0.856). Broadly, most snowmelt over the study area occurs from late April through early June, with timing delayed farther north and at higher elevations. Many local-scale snow patterns are evident in the detailed, 30-m product. The snow persistence map was co-registered to Landsat land cover mapping, creating a powerful, publicly available resource for ecosystem and land use analyses (https://irma.nps.gov/App/Reference/Profile/2203863).