Using HADGEM3-AO initialised case studies to assess the impact of an interactive ocean on medium-range hindcast skill and MJO prediction

The UK Met Office (UKMO) currently uses atmosphere only models for operational short-range (1-5day) and medium-range (5-15day) forecasts. Recent research has been directed at setting up and testing a new medium-range coupled system, which couples the UKMO atmospheric model to the fully interactive ocean (NEMO) and sea-ice (CICE) models. The atmosphere component is initialised from UKMO operational analyses, while the ocean has been initialised from analyses generated from the latest UKMO ocean and ice data assimilation system (FOAM-NEMOVAR). An extensive set of case studies have been carried out to assess potential benefit to be gained from the coupled model over atmosphere control and ocean control configurations. The atmosphere control persists SST anomaly and the ocean control obtains forcing fluxes from the atmosphere control. These configurations have been rigorously compared across a number of standard numerical weather prediction (NWP) cases and in a range of weather situations.

The coupled model is broadly competitive with control simulations at most forecast lead times, but achieves notably higher skill for some variables in the Tropics. For example, precipitation and relative humidity biases around the Maritime continent are improved with an interactive ocean. Comparisons with longer seasonal and climate coupled simulations using the same model configuration indicate that the 15 day hindcasts display similar biases within a few days. Analysis of growth of these errors in computationally cheap and directly verifiable 15-day hindcasts provides a powerful tool to diagnose and potentially fix persistent climate model biases. A subset of cases (>70) were selected to examine the impact of coupling on representation of the MJO. Preliminary results indicate somewhat stronger convection and slightly improved teleconnections in the coupled forecasts over that of the atmosphere control. Work is underway to develop metrics which will help to determine the processes which govern the observed improvement.