Development of an operational coastal model of the Seto Inland Sea, Japan

We have developed a coastal model of the Seto Inland Sea, Japan, for a monitoring and forecasting system operated by the Japan Meteorological Agency (JMA). We executed a hindcast experiment using reanalysis datasets for the atmospheric and lateral boundaries without ocean initialization by data assimilation. The seasonal variability is verified to be realistic by comparing sea surface temperature and salinity of the hindcast experiment with observations. With a horizontal resolution of approximately 2 km, the model represents explicitly various coastal phenomena with a scale of 10-100 km, such as the Kuroshio water intrusion into Japanese coasts. This leads to good representation of intramonthly variations. For example, intensity of the sea level undulations with a period shorter than 23 days shows 1.6-fold improvement, as compared to the present model of JMA with the horizontal resolution of approximately 10 km. In addition to the increased resolution, the model is optimized for coastal modeling as follows. Incorporation of a tidal mixing parameterization reduces a high temperature bias in the Bungo Channel (a western channel of the Seto Inland Sea) and contributes to formation of a frontal structure. An accurate dataset of the river discharges is used for runoff, which has a strong impact on salinity. Enhancement of coastal friction improves surface currents. Owing to the increased resolution and these optimizations, the model shows realistic variability in a wide temporal range from several days to seasons. Root-mean-square errors of sea surface temperature and heights are evaluated as 1-2 K and 7-10 cm, respectively, without data assimilation. In the eastern part, however, the predictability is relatively low, which might be related to representation of an eastward mean flow in the Seto Inland Sea.