利用实测资料评估被动微波遥感雪深算法

利用SSM/I微波亮温数据,结合地面站点实测资料,比较Chang算法和Che算法在前苏联、中国及蒙古境内6种不同积雪类型的反演精度,结果表明:被广泛应用于全球雪深反演的Chang算法低估了前苏联境内雪深7.6cm,相对误差为-24.3%,而分别高估中国及蒙古境内雪深9.2cm与11.4cm,相对误差分别为108.8%和180.9%,区域反演效果很差;针对中国境内积雪的Che算法严重低估前苏联境内雪深,整体低估21.3cm,相对误差为-68.6%,RMSE为31.4cm;在中国及蒙古境内反演效果有所改善。6个积雪类型中,植被较单一,地形较平坦的苔原型积雪和草原型积雪雪深的反演效果较好。随着纬度和积雪深度的增加被动微波雪深反演有由高估变为低估的趋势。Che算法反演的雪深大体以40°N为界,以北表现为低估,以南表现为高估,另一方面,整体上该算法在雪深低于6.7cm时表现为低高估,高于6.7cm表现为低估;因此,全球算法应用到局部地区需要进行修正,不同下垫面性质以和气候条件下形成的积雪的被动微波反演应区别对待。

Evaluation of Snow Depth Products Derived from Passive Microwave Satellite Remote Sensing Data Using Ground\|based Snow Measurements

Snow depth products derived from SSM/I passive microwave remote sensing data were evaluated against the ground\|based measurements across the Eurasian continent.The preliminary results show that the Chang algorithm underestimates and overestimates snow depth over the former USSR,China and Mongolia.The bias is 7.6,9.2 and 11.4 cm,respectively.In other words,the average error can be up to -24.3% for the former USSR,108.8% for China and 180.9% for Mongolia.Although the Che algorithm has relatively better results in China and Mongolia than the Chang algorithm,it still underestimates snow depth of the USSR by on average 21.3 cm or -68.6%,For the six snow types,the tundra snow and prairie snow,which always have less vegetation and relatively flat surface.Both the Chang and the Che algorithms have a trend to underestimate snow depth for thick snow cover and overestimate snow depth for thin snow.For the Che algorithm ,it underestimate snow depth for regions north of 40°N and overestimate the snow depth for regions south of 40°N.Similarly,the Che algorithm overestimates the snow depth for snow cover with <6.73 cm,in thickness and underestimate for snow cover with >6.73 cm in thickness.Accordingly,there is no single algorithm which can be used for global application in snow depth estimation,especially,in mountain and forested regions.