基于Landsat8多光谱数据与气象因子的磴口县地表温度反演

Land surface temperature retrieval in Dengkou county based on meteorologil factors and multispectral data of Landsat 8

传统手段获取的地表温度数据已经难以满足现代大尺度的研究需要,遥感数据以其多时相、全天候、大范围等特点已被广泛用于地表温度的反演。Landsat8卫星自2013年升空以来,已有不少学者尝试使用TIRS数据进行地表温度反演研究,但大多集中于对以往卫星研制的方法改进上,针对Landsat8遥感数据的地表温度反演研究仍比较少见。将Landsat8多光谱数据与实测地表温度进行相关性分析,综合光谱特征选取最优波段作为输入参量,利用三种常用的数据同化方法对研究区地表温度进行演变模拟。结果表明:EnSRF模型的反演精度最高,其次是PF模型,EnKF模型的反演精度一般。EnSRF模型在水体地类上的模拟R2和RMSE分别为0.91和0.687,在裸土地类上的模拟R2和RMSE分别为0.92和0.711,在自然地表上的模拟R2和RMSE分别为0.82和1.552,该模型对热环境比较敏感,适用于水体和裸土较多区域。PF模型在自然地表上的模拟R2和RMSE分别为0.93和0.886,更适用于自然地表较多区域。

The land surface temperature data obtained by traditional methods has difficulty in meeting the needs of modern large-scale research. Remote sensing data has been widely used in land surface temperature retrieval since it was born due to its multi-temporal,all-weather and large-scale features. Since the launch of the Landsat 8 satellite in 2013,many scholars have tried to study the land surface temperature retrieval based on TIRS data,but most of them have focused on the improvement of previous methods which were made for the other satellites. The algorithm of land surface temperature retrieval for Landsat 8 is still infrequent. This research carried out a correlation analysis between Landsat 8 multispectral data and the measured land surface temperature,selected optimal bands which took spectral features into consideration as input parameters,and Used three commonly used data assimilation methods to simulate the land surface temperature in the study area. Results show that the retrieval accuracy of EnSRF model is the highest,followed by the PF model,and the retrieval accuracy of EnKF model is the lowest among the three models. The R2 and RMSE of the EnSRF model on water are 0. 91 and 0. 687,with 0. 92 and 0. 711 on bare soil and 0. 82 and 1. 552 on natural surface,which is sensitive to thermal environment and is suitable for the areas mostly covered by water and bare soil. The R2 and RMSE of the PF model on the natural surface are 0. 93 and 0. 886,which is more suitable for the areas mostly covered by natural surface.