A Large Ensemble Seasonal Forecasting System: GloSea6

The Global Seasonal forecasting system (GloSea) is the Met Office's monthly to seasonal ensemble prediction system. The current version, GloSea5, which has been operational since 2014, has undergone a significant scientific upgrade. This includes an update to the coupled model, which contains a treatment of convective entrainment, improvements to sea-ice physics and a scheme for iceberg advection. In addition, we include a more realistic treatment of land-surface initialization, using a forced land model to initialize soil moisture instead of a climatology. We also use CMIP6 forcing data to replace a) a constant Solar forcing with variable fluxes that represent the Solar Cycle and b) zonal mean climatology of Ozone concentrations with three-dimensional, time-varying fields. We describe the scientific performance of GloSea6 using re-forecasts covering the period 1993-2016, with Winter and Summer start dates and 100 members per season. Results are compared with GloSea5 as a baseline. Changes in skill are generally positive or neutral, and we also investigate the impact of the larger ensemble on the skill. GloSea6 better reproduces both the near-surface temperature and Z500 anomalies during the 2003 European heatwave, although there is no significant change for the 2010 Russian event. Additionally, we see reduced sea-surface temperature (SST) biases for both El Nino and La Nina during December-January-February (DJF). Furthermore, the SST ensemble mean standard deviation for El Nino during DJF is closer to the observed value. The NAO skill is similar to that of GloSea5. We note, however, that the September sea-ice extent bias is larger than found in GloSea5. This could be a consequence of the treatment of melt-ponds and sea-ice drag.