Sources of Tropical Cyclone Track Bifurcation in the Coamps-TC and GFS Ensembles

As tropical cyclone (TC) official and model track forecasts improve, it can be increasingly difficult to get further improvement. One technique to circumvent this problem is to retrospectively find cases where models performed unusually poorly and/or had unusually high forecast uncertainty, and focus subsequent model development on these challenging cases. In this study, forecast track bifurcations are identified in the 21-member experimental Coupled Ocean-Atmosphere Mesoscale Prediction System-Tropical Cyclone (COAMPS-TC) ensemble. In all, thirty-four bifurcating cases are identified over a time period spanning the 2020-2022 tropical cyclone seasons in the Northern Hemisphere. The identified forecasts are then compared to the Global Ensemble Forecast System (GEFS) with the goal of identifying sources for the increased model spread and/or degraded model accuracy.

The bifurcating cases are primarily ensemble forecasts initialized with either tropical depression (TD) or tropical storm (TS) intensity, with both COAMPS and GEFS either keeping majority of the member TCs at or below TS status during the forecast, or dissipating them entirely before the end of the forecast. While forecast spread at late lead times for bifurcating cases is statistically significant from the larger set of cases in both ensembles, late lead time ensemble mean track error is only statistically significantly higher in the COAMPS-TC ensemble. Positioning of an ensemble member TC relative to the ensemble mean at early lead times showed little correlation to the member’s positioning relative to the ensemble mean later in the forecast. Rather, the TCs forecast location relative to the ensemble mean in the bifurcating cases shows high sensitivity to the tropospheric deep-layer steering flow, particularly in the COAMPS-TC ensemble. On average, a region of weak flow, or a ‘col’ is to the northwest of the TC’s at initialization. Minor differences in the steering flow at early lead times induced by variations in the synoptic-scale environment cause the TC’s relative position to the ‘col’ to vary, leading to increasingly different tracks, and eventually, a bifurcation.