Indonesia’s complex archipelagic waters require high-resolution and dynamically consistent marine forecasts. The Indonesian Coupled Atmosphere–Wave–Ocean Model (InaCAWO), developed by BMKG in collaboration with North Carolina State University, integrates WRF, ROMS, and SWAN through the Model Coupling Toolkit. The system provides forecasts up to 14 days at 3 km resolution over Indonesia, with 1 km nested domains for selected narrow straits.

To improve its initial ocean state and forecast performance, BMKG and the National Center for Atmospheric Research developed a ROMS-based ensemble data assimilation system using the Data Assimilation Research Testbed and the Ensemble Kalman Filter. The assimilated observations include surface currents from high-frequency radar, Marine Automatic Weather Station data, temperature and current measurements from Surface Velocity Program drifters, temperature and salinity profiles from NKE–ARVOR floats, blended sea-surface temperature from the Copernicus Marine Service, and satellite sea-surface height. An 80-member ROMS ensemble was used to update temperature, salinity, currents, and sea-surface height through flow-dependent background-error covariances.

The assimilation substantially improved the analyzed ocean state. The root mean square error of HF-radar currents decreased from 0.293 m s⁻¹ in the prior to 0.170 m s⁻¹ in the posterior. HF-radar assimilation also corrected unrealistic current patterns occurring over island and coastal grid areas. For sea-surface temperature, the average bias was reduced from 0.37°C to 0.01°C after assimilation. The observational influence extended horizontally up to approximately 300 km and vertically to about 500 m, indicating that the assimilation effectively corrected both surface and subsurface ocean conditions. These results demonstrate that multi-platform ensemble data assimilation can substantially improve InaCAWO initial conditions and strengthen its capability to provide more accurate marine forecasts over Indonesian waters