Dynamical Filtering of Tropical Subseasonal Variability in Observations

The MJO is often identified in observations with spectral or variance-based filters. While effective, these approaches lead to definitions of the MJO that involve other tropical phenomena, such as ENSO. Here, we construct a dynamical filter of tropical variability based on the eigenmodes of a linear inverse model (LIM). We apply our filter to ERA-5 reanalysis data and find that the composite evolution of the MJO based on the RMM index (and similar indices) is composed of the sum of three robust components: a “fast” MJO eigenmode, a “slow” MJO eigenmode, and a set of eigenmodes representing ENSO. In addition, our LIM-based filter is capable of cleanly separating ENSO from the MJO, and provides a view of the MJO as being the sum of fast and slow propagating eigenmodes. This further distinguishes our work from previous principal-oscillation-pattern (POP) or LIM analyses, which have tended to describe the MJO as a single eigenmode.