渤海海表面盐度神经网络反演研究

Retrieval of Sea Surface Salinity in the Bohai Sea by Neural Network

摘要:

海表面盐度是研究海洋生态环境和全球气候变化的重要指标。基于多层神经网络，利用实测海表面盐度数据和 MODIS-Aqua 遥感反射率产品，针对渤海建立了海表面盐度的反演模型，随后通过该模型分析了 2022 年 3 月至 2023 年 3 月 4 个季度的渤海盐度时空变化。研究发现：模型的决定系数（R² ）和均方根误差（RMSE）分别为 0.66 和 0.39，优于先前文献提出的多波段线性模型（0.39 和 0.60）。渤海盐度的时空分析表明：黄河冲淡水是影响渤海海表面盐度重要因素，它致使莱州湾和辽东湾的盐度长期降低；9 月份以后随着渤海冷流沿山东半岛南下，随后向渤海海峡和北黄海方向扩散。

Abstract:

Sea surface salinity is an important indicator for studying marine ecological environments and global climate change. Based on a multi-layer neural network，a retrieval model for sea surface salinity in the Bohai Sea is established by using measured sea surface salinity data along with MODIS-Aqua remote sensing reflectance product. Subsequently，the spatiotemporal variations of salinity in the Bohai Sea over four quarters from March 2022 to March 2023 are investigated by mapping the salinity distribution of the Bohai using this model. The results show that the coefficient of determination（R² ）and root mean square error（RMSE）of the model are 0.66 and 0.39 respectively，which are better than the previously proposed multi-band linear model（0.39 and 0.60）. The mapping of Bohai Sea salinity reveals that the dilution effect of the Yellow River freshwater is a significant factor affecting the sea surface salinity，resulting in a long-term decrease in salinity in Laizhou Bay and Liaodong Bay. After September，followed as the Bohai Cold Water Mass，the Yellow River freshwater moves southward along the Shandong Peninsula，subsequently disperses towards the Bohai Strait and the northern Yellow Sea.