植被叶面积指数时序贝叶斯网络反演方法及应用

项目名称： 植被叶面积指数时序贝叶斯网络反演方法及应用

项目编号： No.41271348

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 天文学、地球科学

项目作者： 屈永华

作者单位： 北京师范大学

项目金额： 75万元

中文摘要： 中国定量遥感的发展成果应该在国产卫星数据上得到体现与应用。然而，现状并非如人所预期。制约中国生产出高分辨率定量产品的因素不仅仅是数据质量问题，更深层次的是没有针对我们数据特点的创新性反演算法。以陆表植被叶面积指数（leaf area index，LAI）为例，一般来说，LAI的动态变化信息可以为定量遥感反演提供基础背景知识。然而，这些信息在当前反演算法里还没得到充分重视与合理利用。面对以上问题，本项目拟对中低分辨率叶面积指数产品再分析来获取叶面积指数变化规律，发展一种基于时序贝叶斯网络（或称之为动态贝叶斯网络）的反演方法，通过融合遥感观测与植被动态信息来生成时空连续叶面积指数产品。选择2010-2015年的经过处理后的环境星数据作为应用方向，以生成典型实验区时空连续的高分辨（30m）叶面积指数产品为应用目标。本项目的研究将能够从理论与应用两个层面推动国产遥感数据的定量化发展。

中文关键词： 叶面积指数；时序贝叶斯网络；时间序列；不确定性分析；

英文摘要： The achievement of China quantitative remote sensing should be fully demonstrated by its successful application on China satellite data. However, the things are not going on the expected way. Behind the conflict of rich data and lacking higher level products is not only the quality of remote sensing data, but the lack of innovative inversion algorithm. Taking the vegetation leaf area index(LAI) as an example, in general, ground LAI dynamic information can be employed as background knowledge in retrieving LAI from remotely sensed data, however, such information has yet not been paid enough attention. In this proposal, such background knowledge is extracted from re-analysis on long term middle-coarse resolution LAI which are available through MODIS or SPOT VEGETATION LAI products. On this context, we proposed a dynamic Bayesian network(or time series Bayesian network) inversion method to produce spatial and temporal continue LAI products with the fusion of remote sensing observation and vegetation dynamic information. Application of our proposed methodology will be carried on the China HJ-1 CCD data, which has a 30m pixel resolution and 4 days revisit period. Thus, besides the inversion algorithm, another output of this proposed project is the 6 yeas (2010-2015) time-series higher resolution(30m) LAI products. So

英文关键词： leaf area index (LAI)；Dynamic Bayes Network；time series；uncertainty analysis;；