摘要：This article discusses the challenges of developing a regional ocean prediction model for the Philippine Archipelago, a complex area in terms of geometry, bathymetry-dominated dynamics and variability, and strong local and remote wind forcing, where there are limited temporal and spatial ocean measurements. We used the Regional Ocean Modeling System (ROMS) for real-time forecasting during the Philippine Straits Dynamics Experiment (2007–2009) observational program. The article focuses on the prediction experiments before and during the exploratory cruise period, June 6–July 3, 2007. The gathered observations were not available in real time, so the 4-Dimensional Variational (4D-Var) data assimilation experiments were carried out in hindcast mode. The best estimate of ocean state (nowcast) is determined by combining satellite-derived products for sea surface temperature and height, and subsurface temperature and salinity measurements from several hydrographic assets over a sequential five-day data assimilation window. The largest source of forecast uncertainty is from the prescribed lateral boundary conditions in the nearby Pacific Ocean, especially excessive salt flux. This result suggests that remote forcing and inflows from the Pacific are crucial for predicting ocean circulation in the Philippine Archipelago region. The lateral boundary conditions are derived from 1/12° global HYbrid Coordinate Ocean Model (HYCOM) daily snapshots. The incremental, strong-constraint 4D-Var data assimilation successfully decreased temperature and salinity errors of the real-time, nonassimilative control forecast by 38% and 49%, respectively.
