提 要:青藏高原积雪分布及其时空变化对地表能量交换、水文过程以及气候环境变化具有重要影响,通过NDSI指数阈值法从卫星数据中获取积雪覆盖信息是当前主流技术手段,但受限于高原积雪偏浅、消融快、下垫面类型多样和斑块化分布显著等特征,采用NDSI固定阈值往往不能精确提取高原积雪分布状况.文中以青藏高原东部青海省为研究区,利用MOD10A1数据分析典型下垫面类型积雪区的NDSI与站点自动观测雪深间对应关系,以确定不同下垫面类型积雪的判识参数NDSI阈值范围,得出研究区典型下垫面草地、裸地、耕地、城镇用地的积雪反演参数NDSI最优阈值分别为0.32、0.19、0.20、0.36,用 Landsat8 OLI 数据判识的积雪空间分布作为"真值"对使用NDSI最优阈值判识出积雪的精度进行验证,结果表明草地、裸地、耕地、城镇用地四种下垫面积雪判识提取的总体精度(OA)分别为92.88％、92.56％、97.19％、99.81％,表明考虑不同下垫面类型下的NDSI阈值率定优化可以有效地提高青藏高原地区积雪反演判别精度.
Calibration of snow inversion parameters for different underlying surfaces
The distribution and spatiotemporal changes of snow cover on Qinghai-Tibet Plateau have significant impacts on surface energy exchange,hydrological processes,and climate and environmental changes.The NDSI index threshold method is one of the mainstream techniques to obtain snow cover information from satellite data.However,the snow cover in this area is characterized by being shallow,rapidly melting,diverse underlying surface types,and significantly patchy distribution,the fixed NDSI thresholds method usually cannot accurately extract the distribution of snow cover on the plateau.So,this paper takes Qinghai Province in the eastern part of Qinghai-Tibet Plateau as the research area,based on MOD10A1 data,to study and analyze the corresponding relationship between NDSI and snow depth on typical underlying surface types in snow covering areas,to determine the NDSI threshold range of snow identification parameters for different underlying surface types.The results indicate that the optimal NDSI threshold values of snow retrieval parameters for typical underlying surfaces,such as grassland,barren,cropland,urban land in the research area are 0.32,0.19,0.20,and 0.36,respectively.Using Landsat8 OLI data as the"true value"of snow spatial distribution to evaluate and quantitatively validate the optimal NDSI thresholds for identifying snow,it is showed that the overall accuracy(OA)of snow recognition extraction by optimal NDSI threshold for different underlying surfaces of grassland,barren,cropland,urban land are 92.88％,92.56％,97.19％,and 99.81％,respectively.The optimization of NDSI threshold calibration considering different types of underlying surfaces can effectively improve the accuracy of snow retrieval and discrimination on the Qinghai-Tibet Plateau region.
NETL
NSTL
万方数据
