| 题名 | 湖泊水生态模型多源数据同化 |
| 作者 | 李志杰 |
| 学位类别 | 博士 |
| 答辩日期 | 2015-05 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 北京 |
| 导师 | 陈求稳 |
| 关键词 | 蓝藻水华，预测模型，数据同化，集合卡尔曼滤波，遥感,algae bloom, numerical eutrophication model, data assimilation, Ensembel Kalman Filter, remote sensing |
| 其他题名 | Data assimilation for numerical eutrophication model using multi-resouces of data |
| 学位专业 | 环境工程 |
| 中文摘要 | 水是人类社会最重要的基础性自然资源、经济资源和战略资源，是环境的良性运行和经济社会实现可持续发展的重要物质基础。湖泊是世界上重要的淡水资源库之一，但湖泊污染和富营养化问题日益严重，因此，湖泊富营养化和水华预测是湖泊研究的热点。 本研究以太湖为主要实验区，建立了湖泊二维水动力模型，对太湖水动力情况进行模拟；在此基础上建立二维水生态模型，对主要营养盐及产生水华的主要藻类（蓝藻、绿藻、硅藻）进行了模拟；通过Morris 筛选法对模型参数进行了灵敏度分析，选定模型关键参数；针对关键参数对模型进行率定并验证；采用GLUE 方法对模型进行不确定性分析，以确保模型的可靠性；采用数据同化方法（集合卡尔曼滤波），实现观测数据和MERIS 遥感数据与水生态模型的结合，动态更新模型的参数及状态，以提高模型的可靠性和对预测的精度。最 终实现准确、及时地获悉整个湖泊的蓝藻时空分布、变化趋势以及增殖速度等，并对水华爆发的范围和强度进行预测。研究主要取得了以下成果： 1) 采用 Morris 筛选法，对Delft3D-BLOOM 中与蓝藻、绿藻、硅藻生物量关系最为密切的6 类参数（包括0℃时的生长速率，0℃时的死亡速率，0℃时的维持性呼吸速率，生长速率温度系数，死亡速率温度系数，维持性呼吸速率温度系数）进行了灵敏度分析，确定了对蓝藻、绿藻、硅藻生物量影响最大的前10 个参数。有效的降低了模型参数率定的工作量，提高模型精度，提升建模效率。 2) 采用 GLUE 方法对模型进行不确定性分析，经分析，在2009~2011 年36 个观测值中，大多数观测值位于90%的置信区间之内，或非常接近90%的置信区间。说明本研究建立的水生态模型是可靠的，在此基础上进行水华预测是可行的。 3) 在采用站点实测数据进行太湖水生态模型单点同化时，无论是只同化状态变量，还是同时同化状态变量和参数，模拟结果与观测值的一致性系数IOA值都有了一定程度的提高，但采用同时同化状态变量和参数的方案时，模拟结果的IOA 值比只同化状态变量时的IOA 值更高。这说明，同时同化状态变量和参数的同化方案比只同化状态变量的同化方案更优。 4) 在采用 MERIS 遥感反演数据进行太湖水生态模型区域状态变量同化时，同化时段内所有月份的数据同化结果的REMS 均低于BLOOM 模拟结果的RMSE，这说明采用MERIS 遥感数据对BLOOM 模型进行区域状态变量同化可以综合利用水生态模型和遥感数据的优势，改进模型的模拟效果，提高模型的模拟精度。 5) 在采用站点实测数据和 MERIS 遥感反演数据这两种数据源进行太湖水生态模型状态变量同化时，若两种数据源的监测结果较为一致，模型的同化效果较为明显；若两者监测结果差别较大，采用站点实测值的同化效果不尽理想，这说明观测数据的误差估计对同化效果具有明显影响。但是多源数据同化具有观测频率高的优势，更有利于改进模型的预测精度。 |
| 英文摘要 | Water is the most important natural, economic and strategic resource of human society, and is the essential material foundation for the benign operation of environment and sustainable development of economy and society. Lake is one of the most important fresh water resources in the world, while the water pollution and eutrophication of lake is a growing problem. Therefore, lake eutrophication and algae bloom prediction is a research hotspot on lake. In this study, Taihu Lake is selected as the main study area. A two-dimensional hydrodynamic model is established to simulate the hydrodynamic condition of Taihu Lake; Based on which, a two-dimensional water environment model is built up to caculate the main nutrient and algae groups (bluegreen algae, green algae, diatoms). A globle sensitivity analysis method, Morris screening method, is employed to carry out sensitivity analysis, and key parameters are identified. Based on this, the water quality model is caliberated and validated. The generalized likelihood uncertainty estimation (GLUE) method is used to analyze the uncertainty of the two-dimensional water environment model, in order to analyze and quantify the uncertainty in this model. The data assimilation method Ensemble Kalman Filter is taken to assimilate multi-source data, including in site observation and MERIS (MEdium Resolution Imaging Spectrometer Instrument), into the two-dimensional water environment model to update both of the state variable and model parameter, which help to improve the model’s predictive precision and reliability. Main achievements have been gained as follows: 1) A globle sensitivity analysis method, Morris screening method, is employed to carry out sensitivity analysis of model parameters in six categories, including growth rate at 0°C, mortality rate at 0°C, maintenance respiration rate at 0°C, temperature coefficient for growth, temperature coefficient for mortality and temperature coefficient for maintenance respiration. A final rank of ten most influential parameters for bluegreen algae, green algae, and diatoms are obtained. 2) The GLUE method is used to analyze the uncertainty of the two-dimensional water environment model, and results show that the 90% confidence interval of the simulated results can enclose most of the observations, which can certify the effectiveness of this model. 3) Ensemble Kalman Filter is taken to assimilate the in site observation into the single point water environment model to update ①state variable only and ②both of the state variable and model parameter. Results show that the precision of the simulations that aftre data assimilation are significantly higher than that before data assimilation, while the Index of Agreement of scheme ② is higher than scheme ①. As a consequence, the scheme of update both of the state variable and model parameter is better. 4）Ensemble Kalman Filter is taken to assimilate the MERIS data into the twodimentional water environment model to update state variable only. Results show that the RMSE of the simulations that aftre data assimilation are significantly decreased than that before data assimilation. It shows that using the MERIS data to assimilation BLOOM model for regional state variables can improve the simulation precision of the model. 5) When doing data assimilation using two kind of data souces to update the state variable only for the two-dementional water quality model, the model precision can be improved more apparent when the two data source are consistent. When there are obvious difference between these two kinds of data source, the assimilation effect with in site measured values is poorer, which indicate that proper error estimation of observation data has obvious impact on the assimilation effect. The assimilation process with multi-sources of data can help to improve prediction accuracy of the model, with the higher density of observational data and more frequent assimilate process. |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.rcees.ac.cn/handle/311016/34500]  |
| 专题 | 生态环境研究中心_环境水质学国家重点实验室 |
| 推荐引用方式 GB/T 7714 | 李志杰. 湖泊水生态模型多源数据同化[D]. 北京. 中国科学院研究生院. 2015. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论