Abstract
Moderate resolution imaging spectroradiometry (MODIS) snow cover accuracy has been assessed in the past at different scales, with various approaches and in relation to the many factors influencing the remote observation of snow-covered areas (SCA). However, the challenge of fully characterizing MODIS accuracy over forest sites is still open. In this study, we exploit 5 years of data from the upper river Adige basin at Ponte Adige (Eastern Italian Alps) to condition an enhanced temperature index snowpack model accounting for model parameter uncertainty by using the Generalized Likelihood Uncertainty Estimation (GLUE) methodology. The simulated SCA is then compared with MODIS retrievals through a range of different statistical metrics to investigate how land use and solar illumination conditions affect such comparison. In particular, the Overall Accuracy index (OA) is used to quantify the agreement between satellite-derived and simulated SCA on a pixel-by-pixel basis. Analyzing the spatial variability either of the median OA and its range shows that illumination conditions over forested canopies represent a major source of uncertainty in MODIS SCA. Exploiting this finding, we identify the minimum level of incoming short-wave radiation for accurate use of MODIS SCA in forest areas.