Evaluation of the Princeton Ocean Model Using South China Sea Monsoon Experiment (SCSMEX) Data

Abstract

The Princeton Ocean Model (POM) has been implemented in the South China Sea for hindcast of circulation and thermohaline structure. A two-step technique is used to initialize POM with temperature, salinity, and velocity for 1 April 1998 and integrate it from 1 April 1998 with synoptic surface forcing for 3 months with and without data assimilation. Hydrographic and current data acquired from the South China Sea Monsoon Experiment (SCSMEX) from April through June 1998 are used to verify, and to assimilate into, POM. The mean SCSMEX data (Apr?Jun 1998) are about 0.5°C warmer than the mean climatological data above the 50-m depth, and slightly cooler than the mean climatological data below the 50-m depth, and are fresher than the climatological data at all depths and with the maximum bias (0.2?0.25 ppt) at 75-m depth. POM without data assimilation has the capability to predict the circulation pattern and the temperature field reasonably well, but has no capability to predict the salinity field. The model errors have Gaussian-type distribution for temperature hindcast, and non-Gaussian distribution for salinity hindcast with six to eight times more frequencies of occurrence on the negative side than on the positive side. Data assimilation enhances the model capability for ocean hindcast, if even only conductivity?temperature?depth (CTD) data are assimilated. When the model is reinitialized using the assimilated data at the end of a month (30 Apr; 31 May 1998) and the model is run for a month without data assimilation (hindcast capability test), the model errors for both temperature and salinity hindcast are greatly reduced, and they have Gaussian-type distributions for both temperature and salinity hindcast. Hence, POM gains capability in salinity hindcast when CTD data are assimilated.