【摘要】 地表温度与地表比辐射率是陆地表层系统的两个重要特征物理量,它们反演的精度将在很大程度上影响间接导出的地表参量的准确性和相关遥感应用的有效性。多光谱热红外反演温度和比辐射率受到陆地表面类型复杂而观测信息不足的限制,很难同时反演出精确的地表温度数值和地表比辐射率数值。高光谱热红外传感器的出现,为更好地解决这一难题带来了机遇。利用高光谱热红外数据的优势和特性,提出了一种基于大气下行辐射残余指标(DRRI)的方法,实现了地表温度与地表比辐射率的准确分离。通过高光谱热红外模拟数据的反演实验表明DRRI方法具有运算速度快、结果精度高、抗噪声干扰能力强等优点。该方法能够应用于野外测量的高光谱热红外数据以及经准确大气校正后的星载高光谱热红外数据。更多还原

【Abstract】 Land surface temperature(LST) and land surface emissivity(LSE) are two important and basic characteristics of the land surface physical system.Precision of the retrieval for LST and LSE will have an obvious impact on the accuracy of those derivate parameters such as land surface long-wave radiation,water deficit index,et al,and then influence effects of a variety of remote sensing applications.However,land surface temperature/emissivity separation(TES) and retrieval is mathematically a non-determinate problem.So an extra constraint must be introduced into TES procedures in order to obtain a deterministic resolution.Quite a few TES methods for multispectral thermal data have been developed based on a variety of extra constraints.But those constraints are all not stable enough or universal enough,due to complicated landcover types and insufficient observation channels.The occurrence of hyperspectral and ultraspectral thermal infrared sensors brings possibility for the better way to deal with TES problem.A concept named as atmospheric "downwelling radiance residual index"(DRRI) was defined in this paper,and a new method based on DRRI concept was designed to give an additional constraint to the inverse problem and simultaneously realize accurate retrievals of LST and LSE.The DRRI is defined on an estimated emissivity spectrum.Since the natural earth surface could not be a blackbody,the at-surface leaving radiance will contain both the surface thermal emission and the reflected atmospheric downwelling radiance,then if the land surface temperature is not accurately estimated,the corresponding derived emissivity spectrum will exhibit "downwelling radiance residual feature"(that is,the estimated emissivity spectrum is smooth on major parts but at certain areas there lie sharp convexities or concavities,which are caused by atmospheric downwelling radiance).An index value has been designed,which is constructed by the estimated emissivity values of N×3 well-chosen channels,to depict the direction and magnitude of the downwelling radiance residual feature.These N×3 channels are divided into N groups(three in each group),and DRRI=∑Ni=1(ε～i,2-w(ε～i,1,ε～i,3)),each group can produce a DRRI component,then all the N DRRIcomponents will sum to the total DRRI.When the land surface temperature estimation increases step by step(initial temperature value can be obtained through supposing some proper channel’s emissivity as unity),the corresponding DRRI value can be calculated and tend to change from a positive value to a negative valueor vice versa,where the DRRI=0 exactly implies the optimal temperature estimation(it can be determined in practice by interpolation on two nearest temperature samples and their DRRI values).Tests on simulated hyperspectral thermal infrared data proved this method to be successful and effective.Furthermore,through retrieval tests on simulated data,the DRRI method was compared with the Spectrally Smooth Emissivity Retrieval Algorithm,which is the most popular hyperspectral TES method in the world.The comparison results indicate that the DRRI method has three fine features: fast speed,high precision,and good stability.更多还原