正则化滤波在反演地表光谱反照率中的应用
Application of Regularized Filtering Technique to Retrieval of Land Surface Spectral Albedo
-
摘要: 利用有限的卫星观测数据,通过反演地物二向性反射分布函数(BRDF)模型提取复杂地表的真实信息,是一个不适定的地球科学反演问题。为了克服离散不适定性带来的困难,在分析引起模型解不稳定的原因和数据误差传递机制的基础上,使用构建滤波函数的方法实现BRDF模型的正则化约束反演。实验结果表明,正则化滤波算法与MODIS AMBRALS算法精度相当,完全适用于具有充足观测和稀疏观测情况下的地物参数反演。Abstract: Surface bi-directional reflectance distribution function(BRDF) model can be applied to retrieve land surface climatological or biological parameters.However,in geosciences,it is often an ill-posed inverse problem to extract true information of complex surface through BRDF model inversion with limited satellite observations.Based on analysis of the cause of instability in model inversion and the mechanism of data error propagation,regularization filtering technique was adopted to overcome the difficulty due to the discrete ill-posedness in inverting the BRDF model.In order to test the retrieval ability and validity of the presented algorithm,multi-angular POLDER-3/PARASOL BRDF data and MODIS satellite data with deficient looks over the Tibetan Plateau were used.Retrieval results reveal that accuracy of the proposed regularized filtering algorithm is equivalent to that of MODIS AMBRALS code and thus it can be applied to retrieve land surface parameters using sufficient or even sparse observations.
下载:
