摘要
被动微波遥感数据是进行积雪深度反演的重要资料,不同频率微波信号对积雪响应不同。利用人工神经网络(Artificial Neural Network,ANN)方法结合新疆北部地区积雪观测资料建立AMSR2(Advanced Microwave Scanning Radiometer 2)微波亮温(Brightness Temperature,TB)、地理位置、地形因子与雪深的隐含关系,以实现通过亮温、地理位置、地形因子估算北疆地区积雪深度,并分析微波极化方式、位置以及地形的不同组合方式对雪深反演效果的影响。实验结果表明:水平极化对雪深反演的影响大于垂直极化,纬度对雪深的影响大于经度,地表粗糙度和坡向对雪深的影响大于高程和坡度,并且位置和地形因子对雪深影响作用相当。最终通过4种优选模型的误差空间分布对比发现,综合亮温、经纬度、坡度、坡向的ANN输入模型能够较好的反映北疆地区积雪分布状况,训练集的站点平均误差在-7~6cm之间,该组合模型作为神经网络的输入能够较为合理地获取北疆地区雪深模拟值。
Based on the characteristics of the microwave signal responding to the snow depth,we use AMSR2 brightness temperature,geo-location and terrain factor as the inputs of ANN,and snow depth as the desired output to develop an efficiency snow depth retrieve model.We compared the influence of combinations of TB,geo-location and terrain factors on the retrieve of snow depth.It is reviewed in this article that,TB of horizontal polarization,latitude perform better than vertical polarization and longitude respectively.Combination of slope and aspect is superior to other combinations of terrain factors.Besides,there are equivalent influence on snow depth of geo-location and terrain factors.Finally,we compare the performance of four optimal ANN models under different input combinations.At last,we found that the ANN consists TB,latitude,longitude,slope and aspect as inputs is the best model which might fairly simulating the snow depth of Beijiang.
作者
侯海艳
侯金亮
黄春林
王昀琛
Hou Haiyan1,2 , Hou Jinliang1 , Huang Chunlin1 , Wang Yunchen1,2(1.Laboratory of Remote Sensing and Geospatial Science, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences ,Lanzhou 730000, China ; 2.University of Chinese Academy of Sciences ,Beijing 100049,Chin)
出处
《遥感技术与应用》
CSCD
北大核心
2018年第2期241-251,共11页
Remote Sensing Technology and Application
基金
国家自然科学基金项目(41671375
41501412)资助
