Zhang Xiaoyu, Li Lingling
Opt Express. 2013 Dec 30;21(26):31907-18. doi: 10.1364/OE.21.031907.
Land surface temperature (LST) is key parameters in the interaction of land-atmosphere system. This paper proposed a method to inverse LST from multi-temporal thermal infrared remote sensing data based on the theory of split-window algorithm and diurnal temperature cycle model. The new method was validated by a diurnal brightness temperatures data sets corresponding to MSG2-SEVIRI, which was simulated by the atmospheric radiative transfer model MODTRAN 4 with several input parameters under clear sky, including near surface air temperature, atmospheric water, surface temperature and emissivity, and viewing angles, and result showed the root mean square error (RMSE) of LST reaches 1.2K for simulated data and most errors are within ± 2K with accurate parameters input. At the same time, comparison of LST estimated using the proposed method from MSG2-SEVIRI data with that from MOD11B1 V5 product displayed that the RMSE equals to 3.0K and most errors are distributed within ± 6K. But, the method is proposed under no cloudy condition and is tested only in mid-latitude and daytime; more validation should be made in different areas and atmospheric conditions.
地表温度（LST）是陆气系统相互作用中的关键参数。本文基于分裂窗算法和昼夜温度周期模型理论，提出了一种从多时相热红外遥感数据反演地表温度的方法。利用大气辐射传输模型MODTRAN 4在晴空条件下，通过几个输入参数（包括近地面气温、大气水汽、地表温度和发射率以及视角）模拟得到的与MSG2-SEVIRI对应的昼夜亮温数据集对新方法进行了验证，结果表明，对于模拟数据，地表温度的均方根误差（RMSE）达到1.2K，在输入准确参数的情况下，大多数误差在±2K以内。同时，将利用所提方法从MSG2-SEVIRI数据中估算得到的地表温度与MOD11B1 V5产品估算得到的地表温度进行比较，结果显示均方根误差为3.0K，大多数误差分布在±6K以内。但是，该方法是在无云条件下提出的，且仅在中纬度和白天进行了测试；应在不同区域和大气条件下进行更多验证。
