The Forecast Performance Evaluation of numerical prediction model of ocean temperature and salt flow (MaCOM)

MaCOM model takes the international advanced numerical model NEMO as the power core, coupled with the sea ice model, with the horizontal grid resolution better than 10 km and a total of 75 layers in vertical direction. On this basis, a comprehensive integrated numerical forecasting system with data collection system as the root, ensemble assimilation system as the backbone, forecasting system as the branch and product production system as the terminal has been developed, forming a distributed and loosely coupled tree operation and maintenance architecture with four subsystems: data collection, data assimilation, numerical forecasting and product distribution.

In order to test the MaCOM model forecasting effect, the MaCOM model is used to make day-by-day forecasts of temperature, salt, current, sea surface height and other variables of the global ocean for the whole year of 2020. This experiment focuses on evaluating the model performance, and to avoid differences in assimilation systems, the global 1/12 resolution day-by-day analysis field of the PSY4 model v3r1 version of the Mercator Center in France is selected as the initial field of the model; the GFS meteorological forecast field data is used as the model upper surface forcing field to drive the model; the model is run from the forecast moment with a forecast time limit of 7 days, and after each forecast process the The forecast results are interpolated to the standard latitude and longitude grid and depth after each forecast process; other settings of the model remain unchanged.The model forecasts are compared using the GOV IV-TT (The GODAE Oceanview Intercomparison and Validation Task Team) Class 4 standard method, which is commonly used to evaluate the performance of forecast systems and forecast skill. The statistics used in the evaluation are based on the comparison of model forecasts with observations, including root mean square error (RMES), bias (Bias), and anomaly correlation, as well as comparing forecasts with climatology and persistence.The following conclusions were obtained from the 2020  evaluation:

• The MaCOM model sea surface temperature forecasts are less biased and closer to the live observations, with RMSE around 0.6℃ and better forecast stability, and PSS and CSS show that the model has obvious positive skill.
• The vertical structure test of the MaCOM model shows that the RMSE is around 0.6℃, and the forecastability of temperature profiles in the Southern Ocean, Indian Ocean, South Pacific, North Pacific and other Southern Hemisphere regions is better than that of the PSY4 model.

• The RMSE of sea surface height anomaly of MaCOM model is around 0.05m, which is smaller than that of PSY4. The PSS test indicates that the forecasting skill of MaCOM model for sea surface height anomaly needs further improvement.
• MaCOM has better forecasts than PSY4 for sea surface temperature, vertical structure of temperature and salt, and sea surface height anomalies; among them, it has effective forecasting techniques for vertical structure of temperature and salt and sea surface temperature, and can better simulate the weather-scale variability, which has good operational application value.