基于去云处理的祁连山积雪覆盖遥感监测研究

Estimating snow cover accuracy from MODIS and AMSR-E with cloud removal methodology in Qilian Mountains

雪盖信息在生态研究、水资源评价管理以及灾害防治中有重要的作用,MODIS利用冰雪指数(NDSI)和阈值提供全球每日积雪产品,微波遥感传感器AMSR-E提供南北半球不受云影响的雪水当量数据。通过融合同一天不同时间过境的MODIS积雪产品MOD10A1和MYD10A1为MOYD,融合MOYD和AMSR-ESWE积雪当量产品产生MODAM,以祁连山区气象站观测雪深数据为"真值",检验了2010-2011年积雪季MODIS积雪产品和AMSR-E识别积雪的精度,结果表明:MOYD产品和MODAM使云量减少了15%和100%,积雪精度和总体精度分别达到了24%、59%和88%、80%,通过融合多时相和多传感器数据大大提高了积雪监测精度,此外对祁连山积雪时间分布和不确定进行了分析。

Snowcover information is very crucial to ecological research, water resource evaluation and disaster pre- vention for 40%-60% surface of the northern hemisphere covered by snowcover. Monitoring of the snowcover extent has been a hotspot in snowcover studies. MODIS sensor provides daily snowcover products with high temporal reso- lution, using snow and ice index （ NDSI ） and classification threshold, but most of these products are disturbed by the clouds that are very common in the alpine areas. The passive microwave remote sensing, AMSR-E, provides dai- ly snowcover products without clouds disturbance, however, the spatial resolution of AMSR-E products are very coarse with 25 km. So the aim of this study is to explore the method to combine the product of MODIS and the AM- SR-E and to get the accurate and high special resolution snowcover products. To achieve this purpose, firstly, MOYD data were obtained by combing MODIS snowcover products （i.e., MOD10Aland MYD10A1） at different transit time in a day, which came from MODIS sensors in TERRA and AQUA satellites. Secondly, MODAM data were derived by combining MOYD and AMSR-E. Lastly, the snowcover depth data observed in 26 weather stations in the Qilian Mountains area were used to test the accuracy of the snowcover products. The results show that the combination of MOD and MYD can reduce the clouds by 15%, while for the MOYD and AMSR-E; this value can be 100%. The results also show that this combination method can improve the classification accuracy of snowcover; the accuracy of snowcover classification of MOYD data is 24% and the overall classification accuracy is 59%. For the MODAM data, the accuracy of snow classification can get to 88% and the overall classification accuracy is 80%. The analysis of the temporal distribution of snowcover in the Qilian Mountains shows that the peak season of snowcover, almost 75% area of the Qilian Mountains is covered by snowcover, from the beginning of January to the late February because of low temperature and the large amounts of snowfall in this period. Through assessing of the uncertainty of snowcover product, the paper finds that the combination of remote sensing products from different time and different remote sensor is very effective in monitoring of the snowcover distribution.