陆地碳数据同化中的模型“异参同效”问题研究

The objectives of this project proposal include three points. Firstly, to review the various factors affecting terrestrial ecosystem carbon cycle model parameter identification rate; Secondly, a carbon data assimilation system based on Bayesian inference framework will be developed to evaluate the uncertainty due to parameterization equifinality, and which associated with representativeness errors in flux observation and remote sensing observations in both spatially and seasonally. To deepen our understanding of the "equifinality"; Thirdly, to explore way through the integration of multi-source observations to reduce uncertainty in model parameters, and quantitative measure and control the uncertainty of terrestrial ecosystem carbon cycle modeling and forecasting.

在Bayes数据同化框架下，系统梳理影响陆地生态系统碳循环模型参数识别率的各种因素，充分考虑通量观测和遥感观测的代表性误差的基础上，评估模型"异参同效"对系统模拟和预报不确定性的影响，深化对陆地生态系统碳循环模型"异参同效"这一极具挑战性问题的认识，促进陆地生态系统碳通量模拟和预报的精度并定量化其不确定性。

基于黑河遥感试验通量观测矩阵试验数据，研究了EC碳通量观测的空间代表性误差，发展了代表性误差的纠正方法，提出了增强地面观测空间代表性的方法建议。基于密集的土壤水分无线传感器网络，系统评估了时间稳定性分析方法的优缺点，发现了该方法性能的尺度依赖性，其在时空结构受系统因素主导的尺度上表现较好，在随机因素主导的尺度上表现较差。给出了高效执行时间稳定性分析的原则建议；基于一维的土壤-植被-大气综合模型CoupModel，研究了其模拟土壤水热碳的参数敏感性、异参同效问题，并测试了利用不同观测控制其不确定性的效果。资助发表期刊论文7篇，其中，国际TOP期刊论文3篇。获发明专利获取1项。