Dependence of Intraseasonal Tropical Cyclone Prediction on the MJO in the S2S Dataset

The Madden-Julian Oscillation (MJO) modulates tropical cyclone (TC) activity on the intraseasonal timescales. Given the ongoing effort to develop dynamical seasonal-to-subseasonal (S2S) TC predictions, it is important to examine if the MJO status affects the forecast skill. For genesis prediction, our previous results have shown that, without applying any bias correction, even the best model (ECMWF) in the most reliable basins (western north Pacific and Atlantic) has very little skill (Brier skill score close to 0.1) after week 1. For some models and basins, a bias correction can bring up the prediction skill from no skill to slightly skillful.

To identify the dependence of the basin-wide genesis prediction skill on the MJO , we again use Brier skill score with a calibration applied. We will compare the skill scores based on all the forecasts to those using subsets of the forecasts grouped conditional to the MJO phase and magnitude. A positive MJO phase(s) or magnitude(s) is defined when a subset of forecasts associated with that MJO phase or magnitude positively contribute to the forecast skill across the majority of the S2S models. We will also conduct genesis forecasts using a Tropical Cyclone Genesis Index (TCGI) to examine whether a derived parameter can have a prediction skill with a stronger dependence to MJO. Both forecasts (using the direct output and using TCGI) will be classified based on two MJO indices, the Real Time Multivariate (RMM) index and the Real Time OLR (Outgoing Long Wave Radiation) index (ROMI). The ROMI was recently shown to be better at capturing the northward propagation of MJO in the boreal summer. We will examine the dependence of prediction skill on MJO at leads from week one to five. The lag dependence of prediction skill to MJO will also be discussed. In addition to the genesis prediction, we will also show the performance of the S2S models in simulating the TC tracks using cluster analysis.