东北地区MODIS和AMSR-E积雪产品验证及对比

Verification and Comparison of the MODIS and AMSR-E Snow Cover Products in Northeast China

通过2002-2008年6个积雪季节的Terra-Aqua/MODIS积雪产品(MOD10A2、MOD10C2)和Aqua/AMSR-E雪水当量产品,分析了东北地区积雪覆盖面积的变化特征,以研究区气象站点观测的积雪数据为真实值来验证两种产品积雪信息的精度,探讨了云覆盖、土地利用类型和雪深对积雪覆盖精度的影响.结果表明:云的存在对微波数据积雪识别的影响较小,在积雪量较多的12月至次年的2月随云量百分比的变化,MOD10A2积雪覆盖面积比例大体出现负变化.因此,在有云情况下AMSR-E数据反演积雪精度最好.对比草地、耕地、林地和居民地4种土地覆盖类型对监测积雪覆盖精度的影响,发现林地对其影响最大,在林区3种积雪产品的积雪识别精度分别为55.8%、81.2%、85.4%;雪深对AMSR-E积雪产品识别精度影响较小,总体精度为97.8%;积雪深度对MOD10A2积雪产品识别精度影响较大,总体精度为57.3%.MOD10A2、MOD10C2和AMSR-E 3种积雪产品的总体反演精度分别为69.3%、76.6%、76.3%.有必要开发适用于东北地区的积雪覆盖算法,提高估算精度,为能量平衡估算、气候模型、农业生产、土壤墒情监测服务.

In this study,the snow cover variation characteristics from 2002 to 2008 in Northeast China were analyzed by using the MODIS snow cover products（MOD10A2,MOD10C2） and snow water-equivalent product from passive microwave remote sensing data AMSR-E.The accuracies of the snow-covered-area（SCA） from the two kinds of snow products were verified by observed data.Furthermore,the influence of cloud cover,land-use type and snow depth on the SCA accuracy was also studied in this paper.The results show that owing to the minor impact of cloud on microwave data,the SCA derived from AMSR-E SWE was of relatively high accuracy,the SCA percentage derived from MOD10A2 had close relation to the change of cloud cover,and the SCA,in general,showed an inverse relation with cloud covered area in the higher snowfall and accumulation season starting from December to February in the next year.Therefore,snow cover calculated from AMSR-E has better accuracy when it is cloudy.Through the comparative analysis of land surface impacting the SCA accuracy,it is found that the grassland,farmland,woodland and residents have different effects on the accuracy of SCA.It is revealed that forest has significant impact on the three other snow cover produces,with the identification accuracy of 55.8%,81.2% and 85.4%,respectively.It is also found that snow depth has less impact on the snow cover identification accuracy for the AMSR-E SWE product,with an overall accuracy of 97.8%.However,the snow depth has a great impact on the MOD10A2 product,with an overall accuracy of 57.3% only.The accuracy for the three snow products（MOD10A2,MOD10C2 and AMSR-E） were 69.3%,76.6% and 76.3%,respectively.Thus,it is necessary to develop a regional algorithm to accurately estimate the snow covered area in the future,which can be served as input factor for land surface energy balance calculation,climate modeling,and soil moisture monitoring and agricultural management.