通用劈窗算法的NOAA-18(N)AVHRR/3数据地表温度遥感反演与验证

本文以NOAA-18(N)AVHRR/3数据,运用通用劈窗技术获得地表温度。首先,利用MODTRAN 4模拟不同地表和大气状况下热红外通道(Ch4,10.3~11.3μm和Ch5,11.5~12.5μm)的星上亮温,并建立模拟数据库。其次,按照地表温度、大气可降水汽含量、地表比辐射率和观测天顶角,对模拟数据库分组,确定出各分组的通用劈窗算法系数。然后,将构建的地表温度反演模型应用到NOAA-18(N)AVHRR/3数据,模型所需的地表比辐射率由NDVI阈值法确定,大气可降水汽含量是利用Li等(2003)提出的一种劈窗的协方差与方差比的方法来估算。反演结果表明:在观测天顶角小于30°或者大气可降水汽含量小于3.5 g/cm2时,地表温度反演的均方根误差小于1.0K;在观测天顶角小于45°并且大气可降水汽含量小于5.5g/cm2情况下,均方根误差小于1.5K。最后,利用美国通量站的实测数据对地表温度反演结果进行了验证,结果表明均方根误差小于1.8K。

Retrieval and Validation of Land Surface Temperature with General Split-window Algorithm from NOAA-18(N) AVHRR/3 Data

On the basis of the atmospheric radiative transfer theory, land surface temperature (LST) was retrieved from NOAA-18 (N) AVHRR/3 cloud-free data using a general split-window (GSW) algorithm. A simulated database including the thermal infra-red channels (Channel 4 with spectral range 10.3-11.3μm and Channel 5with spectral range11.5-12.5μm) radiance observed at the satellite level was established with the atmospheric radiative transfer model MODTRAN 4. It covers various land surface types and atmospheric conditions. Then the coefficients of the GSW were determined by grouping the LST, the atmospheric water vapor content (WVC), the land surface emissivity (LSE), and the viewing zenith angle (VZA) by several sub-ranges. Results showed that the root mean square error (RMSE) between the estimated and actual LST is less than 1.0K when the VZA is less than 30° or the WVC is less than 3.5 g/cm². The RMSE is less than 1.5K when the VZA is less than 45° and the WVC is less than 5.5 g/cm². In addition, the proposed algorithm was applied to the NOAA-18 (N) AVHRR/3 data to retrieve LSTs. The LSE was determined by using the NDVI thresholds method, and the WVC was estimated with the transmittance ratio method proposed by Li et al. (2003). Finally, the field measured data of the US-Flux stations were used to validate the retrieved LST. The results showed that the RMSE is less than 1.8K.