基于数据同化的杭州湾悬浮泥沙遥感反演方法研究

Suspended sediment concentration (SSC), which has an important influence on the port construction and tidal flats utilization, etc. in estuarine and coastal waters, is one of the main parameters of water quality evaluation. Two large rivers, the Changjiang and the Qiantang, drain into the Hangzhou Bay (HZB). Affected by the Changjiang river and Qiantang river’s discharge, runoff, sediment into the sea, tides and other factors, HZB waters with high SSC and high dynamic characteristic. Existing retrieval models are greatly influenced by time and space constraints, it is difficult to accurately estimate high, wide range of SSC. On the other hand, affected by the temporal resolution of remote sensing data and weather factors is difficult to meet the continuous dynamic monitoring of suspended sediment. This project intends to through indoor experiments and numerical simulation and combination of field in situ observation, analysis the response mechanism between remote sensing data and SSC, establish appropriate retrieval model. Research each retrieval models’ applicability and error response between SSC and water bio-optical property, realize multi-model collaborative retrieval based on data assimilation method. Build an assimilation system about suspended sediment estimating by Kalman filter algorithm based on ensemble. Through assimilating retrieval results and numerical simulate results to realize space-time continuum monitoring of suspended sediment in HZB. The research is expected to improve accuracy of suspended sediment retrieval model and reduce the uncertainty of retrieval model. Get continuous dynamic distribution of SSC, which has very important value in theoretical and application for estuary water environment monitoring, water resources management and development.

悬沙浓度是水环境评价的重要参数，对河口地区的港口建设、滩涂利用等具有重要影响。受长江、钱塘江等径流、入海泥沙、潮汐等因素的影响，杭州湾水体悬沙含量高、水体动态特性强，现有反演模型受时空约束影响较大，很难准确估算高浓度、宽浓度范围的悬沙；另一方面，遥感数据受时间分辨率及天气因素的影响很难满足对悬沙的连续动态监测。本项目尝试通过野外采样、室内分析和数值模拟相结合的方式，分析遥感数据与悬沙浓度之间的响应机制，建立适宜的反演模型；研究各反演模型与水体生物光学特性、悬沙浓度间的适用性及误差响应，借鉴数据同化思想，实现多模型协同反演；采用基于集合的卡尔曼滤波同化方法，构建悬沙估算同化系统，通过遥感反演与数值模拟结果的同化，实现杭州湾悬沙含量的时空连续监测。项目的研究，有望提高悬沙反演精度，降低反演模型的不确定性。获取悬沙浓度的连续动态分布，对河口水环境监测、水资源管理与开发具有重要的理论与应用价值。

本研究以杭州湾为研究区，通过多次野外原位观测，分析了杭州湾悬沙浓度及其水色光学特性；结合多源遥感数据（GF-1 WFV数据，GOCI数据和Sentinel 3 OLCI数据），构建了杭州湾悬沙浓度估算的多种经验和半经验反演模型，分析了不同模型的适用性和稳健性；利用数据同化方法，建立了多模型协同反演算法，实现了杭州湾悬浮物浓度的多模型协同反演；基于EFDC数值模型，对杭州湾悬沙浓度进行了数值模拟，结合GOCI反演结果，实现了遥感反演与数值模拟结果的数据同化。具体结果如下：.（1）利用GOCI和Sentinel 3 OLCI数据，构建了杭州湾悬沙浓度反演的单波段模型、波段比值模型、多元回归模型、三波段模型、SAI指数模型、3S指数模型、SERT模型和Nechad模型，对比了模型精度，分析了模型适用性。结果表明，波段比值模型具有较好的模型精度和稳健性。基于颗粒物的后向散射和吸收系数，构建了杭州湾悬沙浓度估算的半分析算法，该算法具有较高的估算精度，适用于浑浊水体悬浮物浓度估算；.（2）基于GOCI数据，进行了6模型协同反演测试。结果表明，多模型协同反演精度优于任意单一模型，MAPE为14.87%，RMSE为47.27 mg/L；.（3）杭州湾悬沙浓度日变化具有很强的动态特征。杭州湾北部区域悬沙浓度较低，湾口区域悬沙浓度较高。受水文周期的影响，杭州湾悬浮物浓度在枯水季要显著高于丰水季；.（4）基于稀疏表达算法，耦合辐射传输模型，提出了一种针对水体遥感反射率的高光谱重构算法。该算法在杭州湾进行了验证，算法可以有效提升杭州湾悬沙浓度的估算精度；.（5）以EFDC数值模型为基础，建立了基于切应力的长江口二维水沙数值模型，对杭州湾悬沙输移扩散进行了模拟。结合GOCI反演结果，进行了杭州湾悬沙浓度数据同化实验。该方法可以提升悬沙浓度估算精度，为悬沙浓度的动态、连续估算提供了经验借鉴。.整体上看，本项目完成了预期目标。在项目的资助下，已发表论文6篇，其中SCI论文5篇，还有3篇论文已收录。