Abstract: We have been developing an integrated system to predict environmental variables in the Japan/East Sea. The nowcast and forecast of the water properties are expected to contribute to the ocean science and societal applications such as fish resource and pollutant control or and marine weather forecast in the region.
Recent developments in the Japan/East Sea (JES) modeling made an operation of the forecasting system possible starting from October 2004. The system produces short-term forecasts up to 5-6 days and provides nowcast and long-term forecast of basic oceanographic variables for the JES. The forecasted variables are the water temperature, sea level anomalies, currents, salinity, and so on. The results are opened for wide range of users through a web site at http://jes.riam. kyushu-u.ac.jp/ .
The system flowchart is shown in Fig. 1. The central part of the system is an original 3D OGCM named RIAM Ocean Model (Lee et al., OPR, 2003), which is accompanied with an oil spill drift and fate model (Varlamov, JMST, 1999) and with an ecosystem model of the JES (Yanagi et al., 2001). The oil spill drift prediction model provides the tools for emergency forecasts of drift and fate of floating and dispersed hazardous materials in the sea. The ecosystem model will supply lower trophic level parameters such as dissolved inorganic nitrogen or phosphorus, phytoplankton, zooplankton, or detritus.
Fig. 1 Flowchart of the Japan/East Sea forecasting system.
The RIAMOM solves the primitive equations assuming the hydrostatic relation at regular latitude-longitude grid in z-coordinates. The free surface is explicitly calculated by a barotropic-baroclinic splitting technique and enables the tidal simulation. The model is further implemented by generalized Arakawa scheme (Ishizaki and Motoi, JAOT, 1999), a turbulent mixing model (Noh and Kim, JGR, 1999), an isopycnal diffusion scheme (Gent and McWilliams, JPO, 1990). The operational version has 1/12° grid spacing and thin top layer of 5 m thickness. The model well reproduces many oceanographic processes in the JES such as the Tsushima and the East Korean Warm Currents (TWC, EKWC) and their variabilities, deep water formations, or tidal variations.
The long-term assimilation model is driven by daily-mean surface fluxes estimated by bulk formulas from the JMA high resolution (~0.1 degree) regional meteorological data. The sequential corrections have been preformed by assimilating the satellite altimeter data on a coarser 1/3° grid for the barotropic and first baroclinic modes, which is known as an approximate Kalman filter (Fukumori et al., JGR, 1999). The surface relaxation is simultaneously made by the high-resolution merged satellite SST data of Tohoku University with one-day time scale. Manda et al. (2005) demonstrated that the restoring the surface temperature to the satellite observation matches the nonlinear ensemble Kalman filtering. The inflow open boundary conditions (Tsushima/Korea Straits) for the volume transport are now given by long-term monthly mean data measured by ship-mounted ADCP, and temperature and salinity are taken from the results of 1/6° Pacific Ocean model.
Figure 2 shows the sea surface temperature and currents at October 1, 2004 estimated by the assimilation system for the southwestern part of the JES. The TWC is relatively stable at the southwestern entrance and separated into two branches by the Tsushima Islands. The two flows generate leeward eddies by interacting each other. The western boundary current of the EKWC and travels north up to 37°N along the coast and contributes meandering of the Polar Front at this time. The warm water transported into the JES tends to be unstable and to form mesoscale eddies, which are easily found by the streamlines on the map. The positions and strength of the eddies are crucial information for fisheries. The present estimates indicate small-scale eddy structures and the cold water along the North Korean coast.
Fig. 2 Estimated surface temperature and currents in the southwestern part of the JES at 10/1, 2004. The temperature variation is shown by colors ranging from 18 to 25°C.
The assimilation of the inflow condition at the Tsushima/Korea Straits is underway. The long-term forecast will be performed by switching the forcing with climatology.
The short-term simulation and forecast version of model includes the tidal prediction module driven by the tidal body forcings and sea level anomalies at the open boundaries. The surface fluxes are estimated with the TOGA-COARE bulk parameterization accompanied with estimation of radiative fluxes with daily variations. It also adopts the relaxation to the Tohoku University SST data at the analysis stage as is done in the assimilation model. The meteorological data are currently supplied by the NOGAPS analysis and forecasts up to 144 h (6 day). Sine one day is consumed by downloading remote data and model computation, the prediction is made for up to 5 days.
Oil spill simulation
The oil spill model starts its forecasting in any emergency cases. It presents the spilled oil as an ensemble of particles. Each particle is characterized by its position on the plane and in the depth, diameter, and physical properties – initial density, viscosity, fraction of water in oil, fraction evaporated, degradated etc. Particles tracking method is adopted, that includes effects of oil transport by wind and sea currents, buoyancy, three dimensional random diffusion and oil - coastal line interaction. Parameterizations for the oil evaporation and emulsification at the sea surface and for the biochemical degradation in the water column are adapted. As result the oil emulsion viscosity and density are changed. The last value as well as particle size influences the oil buoyancy and so the penetration of oil into the water depth (dispersion formation). Random walk technique is used for simulation of vertical and horizontal turbulent mixing. Mixing length scales depended from the surface wind and current, counting for larger rate in the stormy weather. In calm weather oil particles could resurface by the buoyancy effect.
The system is designed to be automatic for operation in real time. The local data server stores the static data of topography and climatological data, and calls remote sources for the meteorological data or satellite data by regular ftp. In present configuration, the short-term and long-term JES simulation and nowcast are produced by the RIAMOM every 12h and one month, respectively. The results are available to the public through the WWW server (http://jes.riam.kyushu-u.ac.jp/). It provides an interactive interface for flexible visualization. It requires approximately 2.5 hours for model run predicting for 6 days on 2 Itanium CPUs workstation with OMP parallelization.