集合四维变分同化叶面积指数和蒸散发的区域冬小麦产量估测

Remote sensing data assimilation approach that integrate crop growth simulations with remote sensing information have been proposed as a hot spot and developed as a potential approach to improve the accuracy of regional crop yield estimation. In this research field, two critical issues need to be solved, such as assimilation observations with coarser spatio-temporal scales, and single assimilation mode with inefficient assimilation algorithm. An ensemble-based four-dimension variational method (POD4DEnVar), and crop growth simulation with data assimilation as well as field observation on regional scale were employed to deal with the above reasons. Using a spatial-temporal data fusion algorithm, the data set of winter wheat leaf area index and evapotranspiration with high spatio-temporal scale were firstly obtained. Then based upon the crop model, the framework of ensemble-based four-dimension variational assimilation was established, and the reasonable assimilation modes with multivariable-parameter such as leaf area index, evapotranspiration and both of them were developed, the best assimilation mode is also got, and the accuracy of winter wheat yield estimation is eventually improved. It is a new idea for remote sensing data assimilation based on crop models, and also gives the valuable references in theory and methods for accurately and efficiently estimating regional crop yield in the future.

作物模型遥感数据同化是解决区域作物产量估测的有效途径之一。针对作物模型遥感数据同化研究存在的同化观测时空尺度较粗、同化模式单一低效等亟待解决的关键问题，本研究拟引入集合四维变分的同化算法，采用作物同化模拟和区域田间观测试验验证相结合的技术手段，在基于多源遥感数据融合获取冬小麦叶面积指数和蒸散发高时空数据集的基础上，重点开展以作物模型为核心，基于集合四维变分算法的同化框架构建；然后，选取叶面积指数、蒸散发及其两者组合为同化变量，发展多状态变量动态优化的同化模式，提高区域冬小麦产量估测精度和效率。本项目为作物模型数据同化研究提供新方法和新思路，为今后准确高效的大区域作物估产提供有价值的理论和方法参考。

作物模型遥感数据同化是解决区域作物产量估测的有效途径之一。针对作物模型遥感数据同化研究存在的同化观测时空尺度较粗、同化模式单一低效等亟待解决的关键问题，本研究重点开展多状态变量动态优化的作物模型遥感同化研究，拟提高区域冬小麦产量估测精度和效率。主要工作涉及三个方面：（1）基于多源遥感数据冬小麦LAI和ET的高时空反演。课题实现了SEBAL和TSEB耦合的ET遥感反演，构建了PROSAIL模型结合LUT的LAI反演算法，分别得到LAI和ET的高空间分辨率反演结果；结合高时间分辨率、低空间分辨率的MODIS遥感产品，基于HSTAFM遥感数据融合方法，实现了高空间低时间反演结果与高时间低空间MODIS遥感产品的时空融合，得到了高时空分辨率的LAI和ET反演结果。（2）LAI和ET动态优化的作物模型同化模式构建。选取CERES-Wheat模型，结合试验站长期观测数据，进行了作物模型的本地化；利用高时空LAI数据结果，实现了POD4DEnVar算法的作物模型同化模式构建；结合LAI和ET高时空数据集，实现了多观测动态优化的POD4DEnVar作物模型遥感同化模式模拟。（3）作物模型不同同化模式的精度及效率评估。针对不同同化模式，分别从同化算法内部参数、观测误差、时间尺度和空间尺度，分别进行同化估产精度、效率的不确定性分析，最终得到最优的区域同化方案。通过深入的多状态变量动态优化的区域冬小麦产量同化估产研究，形成部分研究成果，主要包括：（1）发表相关研究论文7篇；（2）获得授权相关专利4项；（3）登记软件著作权2项。通过课题的研究，提高了区域冬小麦产量估测的精度、效率和区域应用能力，为今后准确高效的大区域作物估产提供有价值的理论和方法参考。