MODIS雪深反演数学模型验证及分析

Verification and analysis of mathematical model for retrieved snow depth based on MODIS data

在MODIS卫星遥感积雪监测的基础上,利用雪深反演数学模型、积雪指数NDSI和多光谱阈值等相结合的方法,对2004年以来新疆北疆地区的积雪分布状况进行了反演和计算,并利用2004年11月～2005年3月冬季北疆地区气象台站雪深数据和2004年12月～2006年1月加密野外实测雪深数据,对反演雪深数据进行了验证及分析,北疆各地除塔城地区反演精度为83.2%以外,其它地区反演精度达85.2%以上,平均反演精度达86.2%;野外实测数据验证反演精度达92%以上。

This paper introduces a new mathematical model for retrieved snow depth using MODIS data. Based on the model ,the snow depths in the North of Xinjiang Uygur Autonomous Region of China since 2004 have been calculated and also tested using both the snow depth data from the meteorological stations of the same region from Nov. 2004 to Mar. 2005 and the field-measured snow depth data from Dec. 2004 to Jan. 2006. The results show the retrieved snow depth data is well correlated to the measured snow depth data, especially to the field-measured data. Although the snow depth data from the meteorological stations is not representative enough for the discrete distribution and city hot-island effect of the stations themselves, the retrieved snow depth data also has the same coherent variation trend as the field-measured snow depth data. Using snow depth data from the meteorological stations, the precision of the model in the north of Xinjiang reaches 85.2 percent except Tacheng Station＇s, whose precision amounts to 83.2 percent and the average precision is beyond 86.2 percent. Using the field-measured snow depth data, the precision of the model in the same area can reach 92 percent. The retieved snow depth model based on MODIS data, considering snow char- acteristics in different physical states（wet snow, dry snow, frozen snow, old snow, new snow and so on） and various complex underneath conditions including DEM（slope, plain, mountainous area）, land cover such as grassland, forest, cropland, desert, shrubland, wetland, tundra and so on, gets a high retrieval precision of the snow depths in the north of Xinjiang of China, which shows that it is feasible to perform snow depth calculation using the snow depth in-verse model based on MODIS data.