Improving forecast skill of a global coupled forecasting system by accounting for model error

Abstract:

The Navy Earth System Prediction Capability (Navy ESPC; Barton et al. 2020) model is a global coupled (atmosphere-ocean-ice) forecasting system used operationally to produce 45-day ensemble forecasts. The Navy ESPC model couples the Navy Global Environmental Model (NAVGEM) atmospheric model to the Global Ocean Forecasting System (GOFS), which is comprised of both the HYbrid Coordinate Ocean Model (HYCOM) and the Los Alamos Community Ice CodE (CICE) model. We test the impact of including methods to account for systematic model biases and stochastic model uncertainty in the atmospheric component of the Navy ESPC model. These methods include stochastic kinetic energy backscatter (SKEB) and analysis correction-based additive inflation (ACAI; Crawford et al. 2020). Both methods are able to address issues of model bias and stochastic model uncertainty, but in very different ways. Our formulation of SKEB uses a moisture convergence mask to determine regions in which to perturb vorticity with stochastic noise. ACAI uses atmospheric analysis corrections from the Navy ESPC weakly-coupled data assimilation system to address model error by combining a seasonal mean analysis correction (as a representation of model bias) with a randomly sampled analysis correction drawn from the same season (as an added source of model uncertainty). Both methods (SKEB and ACAI) indicate significant improvement of many atmospheric forecast metrics out to weeks 4 and 5. We will present results using each method in isolation and the impact of using the two in concert, as well as, a cross-interface analysis describing the impact of applying these methods in the atmosphere to forecast skill in the ocean and ice components of Navy ESPC.