Importance of ocean initial conditions of late autumn on winter seasonal prediction skill in atmosphere-land-ocean-sea ice coupled forecast system

Abstract

In coupled general circulation models, which are widely used to study weather and climate in subseasonal-to-seasonal timescales, forecast errors pertaining to the systematic model biases could occur. These errors can be overcome using initialization processes, but imperfections in these processes often lead to initialization shock. To investigate this issue, this study examines the impact of imbalanced ocean initial conditions (ICs) on seasonal prediction skills via an atmosphere-land-ocean-sea ice-coupled forecasting system. Three types of ensemble experiments were conducted from 1980 to 2016: (1) the noDA experiment with no data assimilation but with dynamic balance among oceanic variables through the ocean component model (MOM3); (2) the DA_IB experiment with data assimilation but with the dynamic imbalance between assimilated and non-assimilated variables in the IC; (3) the DA_B experiment with data assimilation and balanced states among the ICs. The results show that the prediction skill of DA_B is noticeably higher over most of the globe than the others in winter forecasts, particularly in the middle-high latitudes for ocean, sea ice, and atmospheric variables. The inferior performance of DA_IB compared to noDA at high latitudes, including the Arctic, indicates that the initialization shock caused by spatial discontinuity and dynamic imbalance among variables can cause low prediction skill. We present quantitative evidence via numerical experiments to indicate the negative impact of imbalanced oceanic ICs on other component systems, consequently limiting the seasonal prediction skills. The results of this study will contribute to the understanding of the shock effects on sub-seasonal- to seasonal-scale forecasts.