Verification and Intercomparison of Global Ocean Eulerian Currents

All ocean prediction systems contain errors. Verification and post-processing of the ocean forecasts is essential and contributes significantly to forecast accuracy. This paper describes the verification of ocean model currents against Eulerian currents derived from the drifting buoys, and intercomparison of currents from various global models. The OceanPredict task team for Intercomparison and Validation (IV-TT) has established the CLASS4 data standard for routine forecast verification against reference observing platforms. The set of CLASS4 reference data has been recently extended to include near-surface currents derived from the trajectories of drifting buoys drogued at 15 m. We have applied these data to the Ocean Model, Analysis and Prediction System (OceanMAPS) at the Australian Bureau of Meteorology for verification and inter-comparison with multiple global ocean models namely: Mercator Océan International ocean forecast system (MOi); the operational models of the Met Office, UK: Forecast Ocean Assimilation (FOAM) and Coupled Atmosphere-Land-Ocean-Ice Data Assimilation (CPLDA) systems; and the Global Ice Ocean Prediction System (GDPS-GIOPS) at the Canadian Centre for Meteorological and Environmental Prediction (CCMEP). The aims for this verification analysis are to extend the routine monitoring of the operational system; to assess the OceanMAPS skill against other models; and to inform our stakeholders of the OceanMAPS performance. 

We have assessed the impacts of adding Stokes drift and tidal currents from separate global wave and global tidal models to the model currents on the verification of currents. Inclusion of surface stokes drift improves the model representation with the observations while inclusion of tides has no significant impact. Overall, the MOi and the new version of OceanMAPS show the best performance against the observations. Although there are significant differences in the model configurations of the eight models under evaluation, all models are shown to be remarkably statistically equivalent with consistent spatial and temporal patterns. Thus, indicating that the main differences are attributable to unrepresented processes. We therefore conclude that there remains scope to further improve the representation of the modelled currents with the observations.