A Dynamics-Based MJO Index

Indices for monitoring the MJO are often based on real-valued principal component time series obtained by projecting observational fields onto a small number of empirical orthogonal functions. In this work, we examine a method of identifying the MJO by using the eigenmodes of a linear inverse model trained on reanalysis data. These modes are, in general, not orthogonal, and so in contrast to traditional EOF based indices (which are based on orthogonality of EOF modes), an index is obtained by projecting data onto the adjoints, or complex conjugates, of these eigenmodes. This results in a single complex time series whose real and imaginary components form an index for filtering the MJO. By projecting onto the adjoints, we effectively filter the MJO from other dynamical modes of tropical variability, such as those associated with ENSO.