干旱区生态环境知识资源中心

In semi-arid Mediterranean areas, the snow in the mountains represents an important source of water supply for many people living downstream. This study assessed the daily MODIS fractional snow-covered area (FSC) products over seven catchments with a mixed snow-rain hydrological regime, covering the Atlas chain in Morocco. For this purpose, more than 4760 daily MODIS tiles (MOD10A1 version 5) from September 2000 to June 2013 were processed, based on a spatio-temporal filtering algorithm aiming at reducing cloud coverage and the problem of discrimination between snow and cloud. The number of pixels identified as cloudy was reduced by 96% from 22.6% to 0.8%. In situ data from five snow stations were used to investigate the relative accuracy of MODE snow products. The overall accuracy is equal to 89% (with a 0.1 m. threshold for snow depth). The timing of the seasonal snow was also correctly detected with 11.4 days and 9.4 days of average errors with almost no bias for onset and ablation dates, respectively. The comparison of the FSC products to a series of 15 clear sky FORMOSAT-2 images at 8 m resolution in the Rheraya sub-basin near to Marrakech showed a good correlation of the two datasets (r = 0.97) and a reasonable negative bias of -27 km(2). Finally, the FSC products were analyzed through seasonal indicators including onset and melt-out dates, the Snow Cover Duration (SCD) and the maximum snow cover extent (SCAmax) at the catchment level: (1) the dynamic of the snow cover area is characterized by a very strong inter-annual signal with a variation coefficient of the SCAmax reaching 77%; (2) there is no evidence of a statistically significant long-term trend although results have pointed out that the SCD increased in February-March and, to a lesser extent, decreased in April-May for the 2000-2013 period. The study concludes that the daily MODIS product can be used with reasonable confidence to map snow cover in the South Mediterranean area despite difficult detection conditions. (C) 2015 Elsevier Inc. All rights reserved.