区域作物生长模拟遥感数据同化的不确定性研究

The uncertainty in remote sensing data assimilation into crop growth model is a critical research hotspot in agricultural remote sensing.In this proposal, we devised a sysmatical rsearch experiment based on mathematical modeling analysis and field observation in regional scale to investigate this topic. The main research in this proposal includes, (1) The critical parameters of crop models significantly affected on uncertainty of simulation will be checked using global sensitivity and uncertainty analysis methods. Then the uncertainties from different regionalized methods of driving parameters will be studied so that crop models can be localized and regionalized. (2) Three mainstream crop models will be employed as core dynamical models which will be coupled with four data assimilation algorithms. The uncertainty of data assimilation will be analyzed for different coupled modes between algorithms and crop models so that new data assimilation mode will be developed which calibrates the model parameters and updates model state variables synchronously. (3) With revealing the sensitivity and characters of error propagation for uncertainty of multi-source observed variables in assimilating process, we will propose the possible methods to decrease the uncertainty under spatio-temporal transformation. (4) A trial will be performed on regional scale considering the precision and efficiency so that the operational feasibility will be evaluated. The research in this proposal will provide some new insights into the uncertanty of the remote sensing data assimilation based on crop models. It will contribute to the theory development in this field. It will also develop new methods and trials in order to support operational application of crop model data assimilation in regional and national scales. It will provide some experiences for other relevant studies on remote sensing data assimilation related to simulation of land surface process models.

针对作物模型遥感数据同化亟待解决的关键不确定性问题，本项目采用数学建模分析和区域田间观测试验数据支持和验证相结合的技术手段，重点研究：（1）应用全局敏感性与不确定性分析方法，确定影响作物模型模拟不确定性的关键参数，分析驱动参数不同区域扩展方式对模拟结果的不确定性影响，实现模型"本地化"和"区域化"；（2）以3个典型作物模型为核心，系统分析4种同化算法及耦合模式对同化模拟的不确定性影响，发展新型模型参数和状态同步动态校正和更新的同化模式；（3）探讨作物模型同化模拟过程对多源观测变量不确定性的敏感性和误差传递特征，提出时空尺度变换下同化模拟不确定性的有效消减方法；（4）开展区域同化应用试验，从估测精度和效率角度评估新型同化模式的业务应用可行性。研究成果将为作物模型数据同化研究和应用提供新思路，为今后业务化应用积累理论、方法和试验基础，为其他相关陆面过程模拟的遥感数据同化的不确定性研究提供参考。

区域作物生长模拟是准确预报预测作物长势和产量的重要技术方法，对粮食安全评估、预警与相关决策具有重要价值。全球气候变化加剧，特别是水资源短缺、农业灾害频发等，给农业生产带来巨大冲击，严重威胁国家粮食安全保障。作物模型遥感数据同化是农业遥感应用的研究热点和前沿之一。当前研究主要集中在选择合适的同化和优化算法提高模拟效率和精度上，遥感数据同化对区域作物生长模拟产生的不确定性方面的研究不足。遥感数据本身的不确定性、不同同化算法以及作物生长模型的选择都会对同化结果的不确定性带来影响。针对该领域亟待解决的关键不确定性问题，本研究采用数学建模分析和区域田间观测试验数据支持和验证相结合的技术手段，重点研究：（1）应用全局敏感性与不确定性分析方法，确定影响作物模型模拟不确定性的关键参数，在此基础上，分析驱动参数不同区域扩展方式对模拟结果的不确定性影响，实现作物模型“本地化”和“区域化”扩展；（2）探讨作物模型模拟过程对同化观测变量不确定性的敏感性和误差传递特征，研究时空尺度变换下同化观测变量的精度控制方法；（3）不同同化算法与作物模型耦合模式的同化模拟不确定研究；（4）开展区域同化应用试验，从同化估测精度和效率角度初步评估新同化模式的业务应用可行性。通过项目研究完成了作物生长模型敏感性分析，实现了模型的本地化与区域化，分析了遥感反演不确定性、模型参数、同化算法及其参数对同化不确定性的影响；分析了不同时空尺度对同化精度的影响；并从业务应用出发，分析了同化精度与同化效率的影响因素；大量遥感地面同步实验，获取了大量科学数据，支持了业务应用。