基于地面联网观测的粗分辨率叶面积指数遥感产品验证方法

Coarse leaf area index (LAI) is a critical biophysical parameter in the research area of global change. In the current validation scheme, the high resolution remotely sensed data are used as a bridge to link the ground observation and remote sensing data, then the coarse resolution reference LAI is produced by spatial aggregation. Though such a validation scheme has been carried out in the last two decades when the LAI ground data mainly relied on the point-wise measurement, under the circumstance of new emerging ground networked observation, it is the time to reconsider rational of this scheme. Three issues should be addressed in the context of ground networked observation while validating remote sensing LAI products: the nonlinear impact of the inversion function and the scale transfer function, the impact of the spatial heterogeneity on the quality of the spatial aggregated products, and the method to evaluate products of uncertainty features. Supported by an innovated LAI ground networked observation experiment, which will be carried out in the spatial heterogeneity landscape environment, this proposal aims to validate coarse resolution LAI products (i.e. MODIS LAI) by combining the information coming from spectrum of remotely sensed data and vegetation variation in spatial and temporal scale to generate the reference LAI products, and thus to make a scaling-linkage between the ground observation and remotely sensed products. The closeness degree is proposed, which is calculated from the “probability distance” between the posterior probability of the LAI products and the reference products. And then, this indication is selected to evaluate the performance of remote sensing LAI products. The contribution of this research lies in the following aspects. The scale linkage is established by associating the field observation data with the coarse resolution satellite products through ground observation network by considering the spatial pattern of surface landscape in a Bayesian non-parameters regression model. This new proposed validation work will improve the scheme in the aspects of the method to generate reference ‘true’ values and the method of processing uncertainty information.

粗分辨率叶面积指数（leaf area index，LAI）是研究全球变化重要指标，当前LAI验证是以高分辨率数据为桥梁建立地面与卫星尺度关联，通过空间聚合生成LAI参考产品。这种策略与传统LAI单点测量方法相适应。本项目研究在LAI地面联网观测条件下粗分辨率LAI产品新的验证策略，重点解决三个方面问题：反演函数与转换函数非线性影响、地表空间异质性对聚合产品影响、带有不确定性产品评价方法。基于自主研制LAI地面观测网络，在空间异质性环境下设计面向粗分辨率LAI验证的联网观测实验，融合光谱与时空变异信息构建点-面尺度扩展模型，计算遥感产品与参考产品概率分布贴近度来评估遥感产品不确定性。创新性体现在，基于地面联网观测数据捕捉LAI空间变异，考虑地表空间格局，通过贝叶斯非参数回归模型将地面观测与粗分辨率卫星产品尺度关联。项目将会改进传统方案中对参考真值与不确定性信息的处理方法。

叶面积指数（leaf area index，LAI）是表征植被冠层结构特征的一个重要参数，基于地面设备与卫星遥感平台是实现对陆表植被LAI测量的两个重要手段。利用地面观测数据对卫星遥感产品进行精度与稳定性评价构成了对遥感产品验证的主要工作内容。本项目利用一种LAI联网观测系统（LAINet）获取地面观测数据，通过时间尺度与空间尺度转换，形成与卫星观测尺度一致的参考产品，对中低分辨率的MODIS第6版500米分辨率的LAI产品进行了验证。以黑河流域农作物种植区为主要研究区域，获取了2018-2019年的连续的联网观测数据，基于过渡尺度卫星（ASTER，Landsat等）数据，通过非线性回归模型实现了从点到面的尺度转换，然后进行空间聚合生成了与待验证产品时空尺度一致的参考真值，完成了对多年MODIS的LAI产品质量验证。结果表明，时间序列上，MODIS的LAI产品能够刻画植被生长和凋落的季节特征，但MODIS的LAI在植被生长季与衰落季具有不同质量特征。与参考真值比较，MODIS的LAI总体低估明显，平均低估约为41%。本项目实现了基于地面联网观测的中低分辨率LAI产品验证，论证了引起MODIS产品不确定性的主要因素，定量评价了从观测到尺度转换的操作对产品验证的影响。研究发现了MODIS产品在不同生长阶段具有明显不同的质量特征，为开展更为精细的地面试验提供了一个新的研究方向。项目共取得13项研究成果，包括11项学术论文，1项发明专利，1项中国地理信息产业协会科技进步奖。