MODIS数据反演地表温度的应用研究

Application of MODIS data to retrieve land surface temperature

在近年来提出的两种温度劈窗算法中,研究采用Sobrino等(2003)的辟窗算法和NDVI阈值法获取地表比辐射率,并结合Mao等(2005)劈窗算法中对大气水汽含量的估算,反演了山西省三个时相的地表温度。从反演的LST分布图上发现,该反演算法与LST产品比较符合。同时以其中的一个时相为例,利用LST产品对反演的精度进行了验证,结果发现反演值和产品二者的最大值、最小值和平均值的误差值分别为1.08℃、0.16℃和0.75℃,其均方根误差RMSE达1.30℃,具有较高的精度,从而表明该法反演地表温度的可行性,并化简了地表温度遥感估算的难度,在不依赖于地面实测资料的情况下,可准确快速地大面积遥测农业环境温度,有助于监测农业旱情。

In two recently spilt window algorithms for land surface temperature （LST） inversion, on the basis of Sobrino et al （2003） for LST inversion and the estimation of surface emissivity by NDVI （ Normalized Differential Vegetation Index） threshold method, combined with the algorithm for calculating atmosphere water vapor content in method of Mao et al （2005）, the LST of Shanxi province was retrieved on three dates, 05/28/2006, 10/09/ 2006 and 11/14/2006. The LST distribution maps indicated that the retrieved results were similar to LST products, which were used as a standard to validate this method. According to the validating result of one temporal data on 10/09/2006, the errors of maximum value, minimum value and average value between retrieved LST and LST product were 1.08℃ ,0.16℃ and 0.75℃, respectively, and a very good agreement between retrieved LST and LST product was observed （ the Root Mean Square Error was 1.30℃ ）, which all demonstrated the high accuracy of this retrieval algorithm. Therefore, by using this simplified method, the LST in large scale could be obtained quickly and accurately, without depending on the ground truth data, which provide a convenient path to monitor agricultural onvironment, especially for drought monitoring.