Dynamical seasonal predictions of tropical cyclone activity: Roles of sea surface temperature errors and atmosphere-land initialization

This study proposes an evaluation framework and analyzes three sets of experiments conducted using a seasonal prediction model developed at the Geophysical Fluid Dynamics Laboratory (GFDL). These experiments apply the nudging technique to the model integration and/or initialization to estimate possible improvements from nearly perfect model conditions.

The results suggest that reducing sea surface temperature (SST) errors remains important for better predicting TC activity at long forecast leads-even in a flux-adjusted model with reduced climatological biases. Other error sources also contribute to biases in simulated TC activity, with notable manifestations on regional scales. A novel finding is that the coupling and initialization of the land and atmosphere components can affect seasonal TC prediction skill.