中国典型积雪区MODIS积雪产品精度评价及影响因素分析  

Accuracy evaluation and analysis of MODIS snow cover products in typical snow cover areas of China

作　　者：沈言龙 沈姣姣 郭慧 郑照军[2,3] 王晓艳 SHEN Yanlong;SHEN Jiaojiao;GUO Hui;ZHENG Zhaojun;WANG Xiaoyan(College of Earth and Environmental Sciences,Lanzhou University,Lanzhou 730000,China;Key Laboratory of Radiometric Calibration and Validation for Environmental Satellites,National Satellite Meteorological Center(National Center for Space Weather),China Meteorological Administration,Beijing 100081,China;Innovation Center for FengYun Meteorological Satellite(FYSIC),Beijing 100081,China)

机构地区：[1]兰州大学资源环境学院,甘肃兰州730000 [2]中国气象局、中国遥感卫星辐射测量和定标重点开放实验室/国家卫星气象中心(国家空间天气监测预警中心),北京100081 [3]许健民气象卫星创新中心,北京100081

出　　处：《冰川冻土》2024年第5期1551-1564,共14页Journal of Glaciology and Geocryology

基　　金：国家自然科学基金项目(42271373);中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室开放基金项目“2001—2018年中国东北地区积雪物候变化及其对气候的响应”资助。

摘　　要：MODIS V6不再提供二值积雪分布及积雪面积比例产品,而是仅仅给出像元的归一化差值积雪指数NDSI。因此,基于MODIS V6进行积雪制图时,NDSI阈值的选取及相应的积雪制图精度有待研究。本文基于2013—2021年间250个地面气象站点逐日实测积雪深度数据,对中国三大典型积雪区内1927景MOD10A1和1936景MYD10A1影像中的NDSI_Snow_Cover波段进行评价,分别计算了逐站点像元在积雪产品制图中的最优精度及对应的最优NDSI阈值,并对影响精度的因素进行了分析。基于站点雪深的评价结果表明:(1)1 cm雪深阈值下,MOD10A1和MYD10A1的最优NDSI阈值均值±标准差分别为0.16±0.09和0.17±0.10,对应的总体精度OA、FS指数和CK指数的均值±标准差分别为0.96±0.05和0.94±0.05、0.84±0.19和0.75±0.24、0.81±0.20和0.71±0.24,MOD10A1的精度要优于MYD10A1。(2)MODIS积雪产品精度具有明显的空间异质性,青藏高原地区要远小于东北-内蒙古地区和北疆地区。(3)基于站点的积雪制图精度评价中,站点雪深阈值将会影响评价结果。采用2 cm和4 cm雪深阈值评价MOD10A1和MYD10A1时对应的积雪制图精度最高。(4)积雪存在率SCO、积雪持续时间SDI与积雪产品MOD10A1和MYD10A1的精度CK存在显著的正相关关系,相关系数分别为0.72和0.77、0.67和0.71。(5)青藏高原地形复杂,积雪以浅雪为主,站点积雪信息不能很好地代表像元。因此,对青藏高原积雪产品的精度评估,最好采用同步的无人机观测或者更高分辨率的遥感影像。MODIS V6 no longer provides binary and fractional snow cover products,but only gives the Normal⁃ized Difference Snow Index(NDSI)of pixels.Therefore,when snow mapping based on MODIS V6,the selec⁃tion of NDSI threshold and the corresponding snow mapping accuracy need to be studied.In this paper,based on the daily measured snow depth data of 250 ground meteorological stations from 2013 to 2021,we evaluate the NDSI_Snow_Cover band in 1927 scenes of MOD10A1 and 1936 scenes of MYD10A1 images in three typi⁃cal snow cover areas in China.The optimal accuracy and corresponding optimal NDSI threshold of station-bystation pixels in snow cover product mapping were calculated respectively,and the factors affecting the accuracy were analyzed.The evaluation results based on station snow depth show that:(1)The mean±standard devia⁃tion of the optimal NDSI thresholds of MOD10A1 and MYD10A1 were 0.16±0.09 and 0.17±0.10,respective⁃ly.The mean±standard deviation of the corresponding OA,FS and CK were 0.96±0.05 and 0.94±0.05,0.84±0.19 and 0.75±0.24,0.81±0.20 and 0.71±0.24,respectively.The accuracy of MOD10A1 was better than that of MYD10A1.(2)The optimal accuracy of MODIS snow cover products has obvious spatial heterogeneity.The accuracy in the Qinghai-Xizang Plateau is much smaller than that in the Northeast-Inner Mongolia region and the northern Xinjiang region.(3)In the station-based snow mapping accuracy evaluation,the threshold of snow depth will affect the evaluation results.The accuracy of snow mapping was the highest when the snow depth threshold of 2 cm and 4 cm was used to evaluate MOD10A1 and MYD10A1.(4)SCO and SDI have a sig⁃nificant positive correlation with CK of MOD10A1 and MYD10A1,and the correlation coefficients were 0.72 and 0.77,0.67 and 0.71,respectively.(5)The terrain of the Qinghai-Xizang Plateau is complex,and the snow is mainly light snow.The snow information of the stations cannot represent the pixels well.Therefore,it is better to use synchronous UAV observation or higher resolution remot

关 键 词：MODIS 积雪 气象站点 精度评价 最优阈值 

分 类 号：P468.025[天文地球—大气科学及气象学]