Characterization, Classification and Prediction of Forecast Dropouts in the NAVGEM Model

Forecast “busts” or “dropouts” are typically defined in terms of the 500mb Anomaly Correlation (AC) in the Northern Hemisphere dropping below a predefined threshold. While all models experience forecast dropouts, in comparison to the NOAA Global Forecast Model (GFS) and European Centre for Medium-Range Weather Forecasts (ECMWF), it has been observed that Navy Global Environmental Model (NAVGEM) typically drops further and, at times, drops sooner and experiences a slower recovery time. Factors that may contribute to dropout cases in NAVGEM include modeling errors, such as poorly represented weakening of the polar vortex associated with Sudden Stratospheric Warming (SSW) and track and/or intensity biases associated with Extratropical Transition (ET) of tropical cyclones. In addition to model errors, initial condition errors can contribute to dropout cases. In some cases, observations which could correct the initial state are inappropriately assimilated due to overly stringent quality control, or inappropriately specified model background and observation error statistics.

In this work we characterize dropouts over a multi-year climatology covering several major transitions, characterizing “dropouts” by their seasonality, recovery time and magnitude. We correlate model performance between centers and assess the impact of initial condition error. Additionally, we investigate the relationship between forecast dropouts and large scale pattern variability, finding dropouts tended to cluster between shifts in the Arctic Oscillation (AO), North Atlantic Oscillation (NAO) and Pacific North American (PNA) indices, which transition from negative phase to positive phase in advance of a dropout event. Finally, we assess the performance of the ensemble in predicting such shifts in forecast skill.