基于神经网络的近地面气温遥感反演研究  

Study of remote sensing inversion of temperature near ground based on neural network

作　　者：赵登忠[1] 谭勇[1] 汪朝辉[1] 

机构地区：[1]长江科学院空间信息技术应用研究所,湖北武汉430010

出　　处：《人民长江》2012年第8期32-37,共6页Yangtze River

基　　金：国家自然科学基金(51009011)

摘　　要：为了获取近地面气温的空间分布格局,鉴于遥感技术在获取地表时空分布信息方面所具有的优势,提出了利用BP神经网络进行遥感反演近地面日平均气温、日最高和最低气温的算法。采用汉江上游流域的气象站点观测数据和Landsat ETM+遥感图像进行了试验研究。研究结果表明,遥感反演近地面日平均气温和日最高气温时,单纯利用遥感信息或者地形信息,都不能得到精度最高的结果,只有把两者结合起来,才能准确地获取近地面气温的空间分布信息。In order to acquire the spatial distribution pattern of temperature near the ground,we take the advantages of remote sensing technology in information acquiring of temporal and spatial distribution pattern near the ground,and propose an algorithm for remote inversion calculation of daily average,highest and lowest temperatures near the ground using BP neural network.The monitoring data of meteorological observation stations in upper Hanjiang River Basin and Landsat ETM＋ remote sensing imagines are adopted for experimental study.It shows that the precision of the simulation results of daily average and highest temperature near the ground would be the highest by neither remote sensing information nor terrain information,and only by the combination of 2 methods,the higher accurate result can be obtained.

关 键 词：近地面气温 BP神经网络 遥感反演 汉江上游流域 

分 类 号：P407[天文地球—大气科学及气象学]