应用FY-3 MERSI数据反演地表温度

Retrieval of Land Surface Temperature from FY-3 MERSI Data

地表温度是反映土壤—植被—大气系统能量流动与物质交换的重要参数,是地球观测类卫星数据反演地面参数的主要内容之一。气象卫星为快速获取大范围辐射面瞬时温度提供了重要手段,其中FY-3 MERSI在保留了极轨气象卫星高时间分辨率观测的基础上,其远红外波段空间分辨率提高到了250 m,提高了温度反演的空间精度。应用FY-3 MERSI第5通道长波辐射观测数据,并结合1 km分辨率高光谱数据,获取反演地温所需的2个重要参数——大气水汽含量及地表比辐射率,实现了地表温度反演。反演算法利用两通道比值法得到大气透过率,并计算出大气水汽含量;应用NDVI阈值方法获得地表比辐射率;根据FY3 MERSI通道特点改进和构造了各参数反演算法;并利用热红外通道数据,使用单通道算法反演地表温度。通过对辽宁地区3个时次的地表温度反演,并与MODIS分裂窗地表温度算法进行了比较分析。结果表明温度精度达到了预期水平,空间精度有显著提高。

Land surface temperature（LST）, one of the most important ground parameter that can be retrievedfrom earth observation satellite data, is of great importance in studying energy flow and material exchange insoil- vegetation- atmosphere system. To acquire instantaneous LST of large area, radiation surface sensorscarried by meteorological satellite platforms are often used. On board FY-3 satellite which was launched byChina in 2008, MERSI is one of the sensors that can be used to get large area of LST with significantenhancement of spatial resolution to 250 m in far infrared band and leads to a higher LST spatial precision. Inthis paper, far infrared data of MERSI band 5 together with other MERSI hyper-spectral 1 km band data wereused to calculate atmospheric transmissivity and land surface emissivity（LSE）—2 important parameters in LSTretrieval. The LST retrieval algorithms were as follows. Atmospheric transmissivity which could be used toobtain water vapor content via radiation transmission model was calculated using two-channel ratio weighted method; surface emissivity was obtained via NDVI threshold method; parameters used in retrieving algorithmswere improved to better suite FY- 3 MERS; mono- window algorithm was used to retrieve LST. Two LSTretrieving experiments were carried out in Liaoning provincial area. The retrieved products were compared withthat retrieved from MODIS data using split- windows algorithm. The results demonstrate that temperatureprecision can reach the desired accuracy, and the spatial precision can be improved significantly.