Reconstruction of horizontal tidal current fields in a shallow water with model-oriented coastal acoustic tomography
Reconstruction of horizontal tidal current fields in a shallow water with model-oriented coastal acoustic tomography
复制标题
DOI：
10.3389/fmars.2023.1112592
复制
发表时间：
2023-02
期刊：
影响因子：
--
通讯作者：
N. Taniguchi;Hidemi Mutsuda;M. Arai;Y. Sakuno;Kunihiro Hamada;Toshiyuki Takahashi;Kengo Yoshiki;H. Yamamoto
中科院分区：
文献类型：
--
作者：
N. Taniguchi;Hidemi Mutsuda;M. Arai;Y. Sakuno;Kunihiro Hamada;Toshiyuki Takahashi;Kengo Yoshiki;H. Yamamoto
互易声传输和海岸声波层析成像(CAT)是一种测量浅海潮流的有力工具，特别是在采用数据同化的情况下。在以前的CAT数据同化研究中，已经实现了集合卡尔曼滤波(EnKF)来同化观测的路径平均速度，但称为局部化和膨胀的特别程序并不总是被实施，这些特别程序弥补了使用集合近似所带来的问题。在这项研究中，EnKF被应用于同化路径平均的海流，该海流是在2020年在日本内海的Mihara-Seto与4个声学观测站进行的相互作用的声传输实验中获得的，以重建观测点的潮流的时空变化。我们使用膨胀、本地化和合奏成员数量的不同值的几个组合来执行EnKF。将所得到的数据同化速度重建结果与声学多普勒流速廓线(ADCP)结果进行了比较。结果表明，与模式预报相比，EnKF资料同化改善了速度再现，协方差膨胀的实施有助于进一步的改进。在我们的案例中，协方差局部化并没有改善结果。在协方差膨胀为1.01的980个集合成员的情况下，ADCP速度之间的分数误差方差(FEV)获得了最好的结果，FEV为7.9%。协方差膨胀为1.01的98个合奏成员的情况导致了类似的表现；FEV值为8.2%。因此，在协方差膨胀的情况下，以往CAT研究中使用的集合成员数量是合理的。在研究中，我们还在初步实验中阐明了观测到的路径平均电流高频变化的原因；路径平均电流捕捉到了与岛屿尾迹相关的涡旋生成的时空变化。EnKF的互易声传播可以捕捉到长周期的短周期变化，因此可以用于不同时间尺度的海岸物理过程研究。
Reciprocal acoustic transmission and coastal acoustic tomography (CAT) is a powerful tool for measuring tidal currents in shallow coastal water, especially if data assimilation is employed. In previous CAT data assimilation studies, the ensemble Kalman filter (EnKF) has been implemented to assimilate observed path-averaged velocity, but ad-hoc procedures called localization and inflation, which compensate for issues associated with using ensemble approximation, were not always implemented. In this study, EnKF is applied to assimilate the path-averaged currents obtained from a reciprocal acoustic transmission experiment conducted at Mihara-Seto in the Seto Inland Sea, Japan, with four acoustic stations in 2020 to reconstruct spatiotemporal variations of tidal currents at the observation site. We executed EnKF with several combinations of different values of the inflation, localization, and the number of ensemble members. The resulting data assimilated velocity reconstructions are compared with acoustic Doppler current profiling (ADCP) results. The results show that data assimilation with EnKF improved the velocity reproduction compared with the model prediction and that implementing covariance inflation contributed to additional improvements. The covariance localization did not improve the results in our case. The best result in terms of fractional error variance (FEV) between the ADCP velocity was obtained from the case with 980 ensemble members with a covariance inflation of 1.01; the FEV was 7.9%. The case of 98 ensemble members with a covariance inflation of 1.01 resulted in similar performance; the FEV value was 8.2%. Thus, with the covariance inflation, the number of ensemble member used in previous CAT studies are reasonable. In the study, we also clarified the reason for the high-frequency variation in the observed path-averaged currents in a preliminary experiment; the path-averaged currents had captured the spatiotemporal variation of vortex generation associated with island wakes. The reciprocal acoustic transmission with EnKF can capture short-period variation over a long period; thus, it can be used in studies of coastal physical processes with various time scales.