Ground-based testing of MODIS fractional snow cover in subalpine meadows and forests of the Sierra Nevada

Abstract

The Moderate Resolution Imaging Spectroradiometer (MODIS) is used widely for mapping snow cover in climate and hydrologic systems, but its accuracy is reduced in forests due to canopy obstruction. Prior validation datasets cannot quantify MODIS errors in forests, because finer-resolution passive sensors (e.g., Landsat) encounter the same canopy errors, and operational ground-based networks sample snow in clearings where snow dynamics differ from those in the forest. To assess MODIS accuracy relative to forest cover, we applied a common canopy adjustment to daily 500m fractional snow-covered area (fSCA) from the physically-based MODIS Snow-Covered Area and Grain size (MODSCAG) algorithm, and tested it at subalpine meadow and forest sites (0.25km2–1km2) in the Sierra Nevada, California during two snow seasons. 37 to 89 sensors monitored hourly ground temperature at these sites. Damped diurnal variations provided a signal for snow presence due to the insulating properties of snow, yielding daily ground-based fSCA at each site. Ground-based fSCA values were validated in a canopy-free area of a meadow site using time-lapse imagery and 15m snow maps from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). Ground-based fSCA had high correlation (R2=0.98) with time-lapse data and was within 0.05 of ASTER fSCA. Comparisons between MODSCAG and ground-based fSCA revealed that an underestimation bias remained in the canopy-adjusted MODSCAG fSCA, ranging from −0.09 to −0.22 at the meadow sites and from −0.09 to −0.37 at the forest sites. Improved canopy adjustment methods are needed for MODIS fSCA.