高分辨地表反照率遥感模型构建与反演方法研究

Global climate models require the absolute accuracy of surface albedo should be 0.02-0.05 units, while surface albedos with spatial resolution finer than 30m are in urgent need by regional climate models. Due to failure to take the directional distribution of skylight into account, current surface albedo products are still unable to meet the accuracy and resolution requirements for climate studies. Inasmuch as the amount of effective information provided by high-resolution satellite data is very limited, suface albedo retrievals always face common difficulties of instability, and high-resolution surface albedo is deficient. We propose a new method for satellite remote sensing of surface albedo with high-resolution data in this project.We will explore spatial expressions of mutiresolution surface spectral reflectance data, study the characterizations and applications of a priori knowledge at the scale of remote sensing pixels, and empoly regularization theory to overcome the problem of surface BRDF models inversion. Based on high accuracy radiative transfer simulations of remote sensing processes, an approach will be established for describing the directional distribution of diffuse skylight. Then, a novel remote sensing model for surface albedos will be constructed. Validation of the constructed model will be performed by means of experimental intercomparision between multiscale remote sensing data. The focus is also on solving the problem of model applications at multiple regions and various climatic condtions. This study will provide more accurate properties of atmospheric boundary-layer of underlying surfaces for regional climate changing studies and weather forcasts, and facilitate applications of new generation high-spatial-resolution satellite surface albedo products with model supports and algorithm references.

全球气候模型要求地表反照率的绝对精度必须达到0.02-0.05。而区域模型则强烈需求空间分辨率高于30m的地表反照率。目前的地表反照率产品，并未考虑天空光方向性分布特征，精度和分辨率均无法满足气候研究需求。高分辨卫星数据提供的有效信息量非常有限，地表反照率反演面临不稳定的共性难题，高分辨率的地表反照率产品匮乏。本项目开展高分辨地表反照率卫星遥感方法研究，探索多分辨率地物光谱数据的空间表达，研究先验知识在遥感像元尺度上的表征与应用，运用正则化理论解决地物BRDF模型反演难题，通过对遥感过程的高精度辐射传输模拟，形成天空光方向性分布的描述，构建新的地表反照率遥感模型，采用多尺度遥感数据相互比对的实验方式，检验所构建模型的真实性，解决模型多区域多气候条件应用问题，为区域气候变化研究与天气预报提供更为准确的大气边界层下垫面地表参数，为建立和应用新一代高分辨卫星地表反照率产品提供模型支持和算法参考。

地表反照率等反射特性参数对太阳入射辐射重返大气的方向和能量分布起着调节作用，是气候和数值天气预报模型的重要输入参量。利用高分辨卫星数据反演地表反照率存在两个方面的主要问题：（1）地表反射模型反演时，因卫星数据信息量不足导致的模型参数不唯一、不稳定甚至不存在，限制了高分辨卫星数据的遥感应用水平；（2）在太阳低角度入射和大气较浑浊条件下，天空光各向同性分布的假设，导致地表反照率计算精度不高，无法满足区域应用需求。. 本项目针对高分辨卫星数据信息量不足的问题，建立了地表反射特性参数的正则化反演算法，掌握了卫星数据在地表反射模型反演中的误差传播机制，解决了正则化参数优化选取的技术难题。通过三维大气辐射传输模拟，获得了天空光场的三维分布结果，验证了项目提出的关于天空光各向异性分布特征观点的正确性，在此基础上，通过考虑地表-大气之间的多次相互作用以及天空光的实际分布，实现了地表反照率遥感新模型的构建。项目研究结果表明，卫星数据噪声水平未知情况下地表反射模型的定量反演，先验知识和正则化方法的联合利用是可行的。当太阳天顶角大于50°时，忽略天空光分布会给地表反照率的计算带来平均-1.55%的不确定度。项目分别对植被地表和冰雪地表开展了模型验证，表明所构建的模型精度达到了优于3.2%的水平，满足目前气候模型对地表反照率5%的精度要求，同时通过对模型的参数化将反照率的计算速度提高了约2000倍。. 本项目研究将推动高纬度地区尤其是冬春季节地表反射特性遥感模型精度不足问题的解决，促进高分辨对地观测数据的定量应用水平，为卫星对地观测目标进行方向性校正和大气辐射校正提供指导依据，也可为空间对地观测光学成像系统的标校提供模型支持。