遥测地表温度与比辐射率的迭代反演方法──理论推导与数值模拟

首先在地表比辐射率为已知的条件下，提出一个非线性迭代温度反演模型，我们对不同的地表和大气条件进行了模拟计算，结果表明当大气温度廓线误差－２Ｋ－２Ｋ，水汽廓线误差±２０％时的温度均方根误差为０．４８Ｋ。当大气模式误差一个模式时反演的温度均方根误差为０．８５Ｋ。在此基础上，引人相邻像元的概念，相邻像元的大气状况可以认为是相同的，应用两个时相的遥感影像数据，假定在两个相近时相之间地表比辐射率值不变，建立地表比辐射率与温度的反演模型。我们对不同的地表和大气条件进行了模拟计算，结果表明当大气温度廓线误差－２Ｋ－２Ｋ，水汽廓线误差±２０％时地表温度均方根误差小于１．５Ｋ，地表比辐射率均方根误差小于０．０２，地表辐射均方根误差为１％；当大气温度廓线误差－２Ｋ－２Ｋ，水汽廓线误差±１０％时，地表温度均方根误差小于１．０Ｋ，地表辐射均方根误差小于０．６％。

The Retrieval of Land Surface Temperature and Emissivity byRemote Sensing Data: Theory and Digital Simulation

The land surface temperature (LST) is one of the most important parameters required to monitor many environmental phenomena such as the Land-Atmosphere Interaction. It is difficult to retrieve LST, because: 1) the emissivity of the land surface is unknown and is not equal to 1 which make atmospheric correction more challenging than before; 2) the land surfaces are usually inhomogenous and changeable LSTs may exit within a pixel, which make the definition of effective surface temperatures and emissivities difficult; 3) the air temperature near the land surface are usually quite different from the actual LST, which may weaken the linear assumpdons in the atmosphedc correction; 4) the skin effect of the LST is more obvious than SST. At first, we assume that the emissivity of the land surface is known. Based on the atmospheric radiative transfer equation, a non-linear atmospheric correction model had been worked out in order to retrieve LST from NOAA/AVHRR aided by the lowtran-7 Program. We have simulated different atmospheric conditions and different land surface condition. Accurate radiative transfer simulations show that: 1)the accuracy of the LST is about 0.48K if the error of the atmospheric profile is less than 2K and the error of the water vapor profile is less than 20%; 2) the accuracy of the LST is about 0.85K for one atmospheric model error; 3) non-linear atmospheric correction model has better accuracy of LST. Then. we assume that the neighbored pixel has the same atmospheric condition and the difference of the emissivity of two NOAA/AVHRR channels is constant for every type of surface. A model to retrieve not only the LST but also the emissivity from two channels is and two neighbored pixels of two temporal images. We simulate different atmospheric conditions and different land surface condition by Lowtran-7. Results show that: 1) the accuracy of the LST is about 1.5K the accuracy of the emissivity is about 0.02 and the accuracy of the ground radiance about 1% can be obtained. if the error of the atmospheric profile is less than 2K and the error of the vapor profile is less than 20%; 2)we may reach the accuracy of LST about 1.0K and the accuracy of the ground radiance about 0.6%, if the error of the atmospheric profile is less than 2K and the error of the water vapor profile is less than 10%.