基于GF-1的青藏高原东南部积雪分布及MODIS积雪产品适用性研究

青藏高原东南部海拔高,地形复杂,云量大,准确掌握该地区的积雪分布特征对于积雪灾害防治非常重要。论文以2013—2019年冬季积雪积累期云量符合要求的35景高分一号(GF-1)影像为基础,将全色影像和多光谱影像融合为2 m分辨率影像,通过目视解译获取了研究区积雪的空间分布特征,结合改进后的30 m分辨率SRTM DEM,探讨了地形对积雪分布的影响。结果表明：积雪像元在研究区范围内占比为33.1%。积雪的垂直分布特征明显：积雪在高程带4000~5000 m(高海拔)处分布较集中,积雪面积占比为18.1%;在高程带0~2000 m、2000~3000 m和6000~7000 m处积雪面积占比均不到0.1%。积雪在北坡、东北坡的分布比例较高,均为15%以上;在南坡、西坡、西南坡、东南坡分布比例较低,均为10%左右。将基于GF-1影像获取的积雪分布分别与同日获取的根据MODIS V6积雪产品计算的积雪比例(MODIS FSC)和积雪分布的对比表明,64.4%的MODIS FSC像元绝对误差不超过10%,MODIS积雪分布产品对含雪像元的漏分率和误分率平均为33.8%和32.7%,说明MODIS积雪产品在研究区的精度还具有较高的不确定性,其对低覆盖积雪反演的精度较差。这表明利用MODIS积雪产品研究青藏高原东南部积雪的时空变化特征时还需要对其积雪反演算法进行改进,同时亟需加强地面观测和基于多源遥感数据的积雪研究。研究结果可为青藏高原东南部雪冰灾害防治提供支撑。

Characteristics of snow cover in the southeastern Tibetan Plateau based on GF-1 and applicability of MODIS snow cover products

The southeastern Tibetan Plateau is high in elevation, with complex terrain and high cloud cover. It is very important to understand the characteristics of the snow cover distribution in this area for railway snow disaster prevention. Based on 35 GF-1 images during the snow accumulation period from 2013 to 2019, 2 m-resolution images were obtained by fusing panchromatic band and multispectral bands. Characteristics of the spatial distribution of snow cover within study area were obtained by visual interpretation. Combined with the modified 30 m-resolution SRTM DEM, the influence of topography on the distribution of snow cover was analyzed. The proportion of snow pixel in the study area is 33.1%. The vertical characteristics of snow cover are obvious. Snow cover is concentrated in the elevation zone of 4000-5000 m (high altitude), accounting for 18.1% of the total snow cover area. The distribution ratio of snow cover on the north slope and northeast slope is higher, which is more than 15%, and the distribution ratio of snow cover on the south slope, west slope, southwest slope and southeast slope is lower, which is all about 10%. The comparison between the snow cover distribution obtained based on GF-1 images and the snow distribution proportion (MODIS fractional snow cover, FSC), and the snow cover distribution calculated based on the MODISV6 snow cover product (MODIS snow cover) obtained on the same day suggests that for the absolute error, 64.4% of MODIS FSC snow pixels are less than 10%, the producer accuracy and user accuracy of MODIS snow distribution products snow pixels are 33.8% and 32.7%, respectively, which indicates that the accuracy of MODIS snow products is relative low in the areas on the study area, and it is more pronounced with low snow coverage. The results indicate that when using MODIS snow cover products to study the temporal and spatial variation characteristics of snow cover in the study area, it is necessary to improve the retrieval algorithm of snow cover and to strengthen ground observations and snow cover research based on multi-source remote sensing data. This study provides support for snow and ice disaster prevention in the southeast Tibetan Plateau.