Dynamical prediction of western North Pacific seasonal accumulated cyclone energy with experimental SFS reforecasts
Abstract
The National Oceanic and Atmospheric Administration experimental seasonal forecast system (SFS) reforecasts, based on two versions of the unified forecast system (UFS), prototypes Prototype 8 and High-Resolution Prototype 3, are evaluated for skill in predicting western North Pacific (WNP) seasonal tropical cyclone (TC) activity over 40 years of April initialized seasonal reforecasts (1982–2021). TC events are identified using the TempestExtremes event tracking algorithm based on sea level pressure and mid-troposphere temperature fields. Dynamical forecasts using the identified TC events are skillful when estimating seasonal variations of accumulated cyclone energy (ACE), but are not skillful when forecasting the number of TC events. We assess the influence of the large-scale environment on observed and simulated TC activity. The relationship between the large-scale environment and ACE is consistent between observations and the SFS reforecasts, emphasizing the influence of El Niño–Southern Oscillation (ENSO) variations. In contrast, the large scale relationship differs for the number of TC events, with TC events in SFS reforecasts having a ‘too positive’ relationship with ENSO resulting in low forecast skill. This is because the SFS reforecasts underestimate the observed reduction (increase) of TC events in the southwest WNP during positive (negative) ENSO years. These results suggest that these UFS prototypes skillfully represent interannual variations in TC intensity, as represented by ACE, but may require improvement in their representation of weaker events and the TC-genesis-ENSO teleconnection to skillfully forecast TC frequency.