Development, implementation and evaluation of a data-assimilative ocean forecasting system off the central California coast

Abstract

Abstract The development and implementation of a real-time ocean forecast system based on the Regional Ocean Modeling System (ROMS) off the coast of central California are described. The ROMS configuration consists of three nested modeling domains with increasing spatial resolutions: the US West coastal ocean at 15-km resolution, the central California coastal ocean at 5 km, and the Monterey Bay region at 1.5 km. All three nested models have 32 vertical sigma (or terrain-following) layers and were integrated in conjunction with a three-dimensional variational data assimilation algorithm (3DVAR) to produce snapshots of the ocean state every 6 h (the reanalysis) and 48-h forecasts once a day. This ROMS forecast system was operated in real time during the field experiment known as the Autonomous Ocean Sampling Network (AOSN-II) in August 2003. After the field experiment, a number of improvements were made to the ROMS forecast system: more data were added in the reanalysis with more careful quality control procedures, improvements were made in the data assimilation scheme, as well as model surface and side boundary conditions. The results from the ROMS reanalysis are presented here. The ROMS reanalysis is first compared with the assimilated data as a consistency check. An evaluation of the ROMS reanalysis against the independent measurements that are not assimilated into the model is then presented. This evaluation shows the mean differences in temperature and salinity between reanalysis and observations to be less than 1 °C and 0.2 psu (practical salinity unit), respectively, with root-mean-square (RMS) differences of less than 1.5 °C and 0.25 psu. Qualitative agreement is found between independent current measurements and the ROMS reanalysis. The agreement is particularly good for the vertically integrated current along the offshore glider tracks: the ROMS reanalysis can realistically reproduce the poleward California Undercurrent. Reasonably good agreement is found in the spatial patterns of the surface current as measured by high-frequency (HF) radars. Preliminary results concerning the ROMS forecast skill and predictability are also presented. Future plans to improve the ROMS forecast system with a particular focus on assimilation of HF radar current measurements are discussed.