A more predictable MJO?

Claire Valva, now at Caltech, just submitted our paper on the development of a more predictable Madden Julian Oscillation using Koopman spectral analysis. The Madden-Julian oscillation (MJO) is an eastward moving pattern of enhanced and suppressed tropical rainfall that lasts, approximately, from 30 to 60 days. The MJO is connected with many aspects of global weather including tropical cyclones and the monsoon. We use a data-driven method to identify an index of the MJO and show that this is a more predictable index than those traditionally used.

Woah, did we just outperform ECMWF with Claire’s laptop? Not exactly. The point of the paper was to identify a target that isolates the most periodic, and hence predictable, elements of the MJO, as opposed to capturing the most variance, as done with the Realtime Multivariate MJO (RMM) index. In the paper, however, we show that the new Koopman based index captures similar variability.