热带地理 ›› 2017, Vol. 37 ›› Issue (3): 434-442.doi: 10.13284/j.cnki.rddl.002954

• 论文 • 上一篇    

一种针对旱季与雨季差异的AMSR-E被动微波 遥感地表温度反演经验模型

刘礼杨1,3,苏泳娴2,陈修治3,邵怀勇1   

  1. (1.成都理工大学 a.地球科学学院;b.国土资源部地学空间信息技术重点实验室,成都 610059; 2.广州地理研究所,广州 510070;3.中国科学院 华南植物园,广东省应用植物学重点实验室,广州 510650)
  • 出版日期:2017-05-05 发布日期:2017-05-05
  • 作者简介:刘礼杨(1992―),男,四川泸州人,硕士,研究方向为地表干旱模型的构建与改进,(E-mail)lly_cdut@163.com。

AMSR-E Passive Microwave Remote Sensing Surface Temperature Inversion Experimental Model Focused on Differences between Dry Season and Rainy Season

LIU Liyang1,3,SU Yongxian2,CHEN Xiuzhi3,SHAO Huaiyong1   

  1. (1.a.College of Earth Science;b.Key Laboratory of Geoscience Spatial Information Technology of Ministry of Land and Resources, Chengdu University of Technology,Chengdu 610059,China;2.Guangzhou Institute of Geography,Guangzhou 510070,China; 3.Guangdong Provincial Key Laboratory of Applied Botany,South China Botanical Garden,Guangzhou 510650,China)
  • Online:2017-05-05 Published:2017-05-05

摘要: 基于2009年的AMSR-E/Aqua L2A亮度温度数据,以中国中部和南部地区为研究对象,根据地区干湿差异,以秦岭―淮河线为界,结合旱季(10月―次年3月)和雨季(4―9月),将研究区划分为北方雨季、北方旱季、南方雨季和南方旱季4种情况。利用逐步回归分析方法,分别进行地表温度反演独立建模,构建了基于AMSR-E四波段亮度温度的地表温度多元回归反演模型,回归方程的RMSE除北方雨季外均优于3.0 K。以2010年MODIS LST产品对模型进行验证,结果表明:该模型的地表温度反演平均误差约80%被控制在2.5 K以内,南方旱季平均误差<2.5 K的区域面积甚至达到88.64%。文章主要是针对水汽对微波的影响,根据地区干湿差异和旱季、雨季差异而构建的地表温度经验模型,可为其他经验性方法构建提供新的思路。

关键词: 被动微波遥感, AMSR-E, 地表温度, MODIS, 回归分析

Abstract: Based on the AMSR-E/Aqua L2A brightness temperature data and MODIS MYD11A2 Land Surface Temperature (LST) product in 2009, we retrieved LST in Central and Southern China. Firstly, the study area was divided into the southern (humid) and northern (semi-humid) regions along Qinling Mountain and Huaihe River based on local humidity. Next, according to the temperature difference between the dry season (October to March) and the rainy season (April to September), the study area was further divided into four types: northern rainy season, northern dry season, southern rainy season, and southern dry season. Finally, using the stepwise regression method, LST inversion models were built for each type based on AMSR-E 4-channel brightness temperatures. The decision coefficients (R2) of the regression models are 0.437 (southern rainy season), 0.663 (southern dry season), 0.701 (northern rainy season), and 0.682 (northern dry season), and the corresponding root mean square errors (RMSE) are 3.052 K, 2.637 K, 3.510 K, and 2.931 K, respectively. We selected the MODIS MYD11A2 LST product and the AMSR-E brightness temperature data in 2010 to validate the model. Firstly, the LST of different regions and different periods were simulated, and then the simulation errors of each period were calculated. Finally, the average errors and their spatial distributions in different regions and different periods were obtained. The validation results showed that, for the areas with an error ranging from 2.5 K to 5 K, their distribution in the northern region is relatively of disorder, with no prominent features, while their distribution in the southern region is mainly concentrated in the southwest of the Sichuan-Chongqing-Yunnan area. And the areas with an error greater than 4 K, such as the western Sichuan Plateau and the Hengduan Mountains, are mainly distributed at the edge of the Qinghai-Tibet Plateau. From the statistical results, the average error of the model is less than 2.5 K in about 80% of our study area, and the average error is less than 2.5 K in 88.64% of Southern China in the dry season. This study mainly investigated the impact of water vapor on microwave. An empirical model of surface temperature was constructed by taking into count the difference in regional humidity and the difference between dry season and rainy season, which provided an improvement on previous empirical methods.

Key words: passive microwave remote sensing, AMSR-E, land surface temperature, MODIS, regression analysis