An OLR-based Index of MJO Evolution and Activity

There have been many attempts to define an index of the Madden-Julian Oscillation (MJO) for statistical, climate monitoring, and model verification studies. These various indices have been proven to be very useful for the particular applications that they have been adapted to, but many of them involve the use of circulation information as well as satellite-derived measures of convection, such as OLR. Perhaps the most commonly used metric used to monitor the MJO is the so-called RMM index of Wheeler and Hendon (2004). Among the advantages of RMM, it is relatively simple to calculate, and is robust to changes in the seasonal cycle and interannual variability. However in the course of our research we have found it necessary to use a more convective based MJO index that does not contain large scale circulation information as RMM does. We take a standard approach to calculating an alternative index by calculating the leading EOFs of the filtered OLR field using the entire 20S to 20N global grid using a covariance matrix. The key initial step to our approach involves first filtering the OLR field for the “MJO band” in space-time, retaining only fluctuations with periods from 30-96 days and eastwards zonal wavenumbers from 0-9. This approach results in an EOF pair that represents the propagating pattern of the MJO OLR field, and faithfully reproduces the statistical behavior of the MJO during all seasons. In order to produce EOFs and an index that vary smoothly through the entire annual cycle, the EOFs are calculated for each day of the year using a 120 day sliding window, then the Principal Components for each individual day are assembled into a index pair for each day. The projection of this index onto the MJO OLR field is much stronger than when using RMM, and the correlation between the new index and RMM is around .70. The pros and cons of such an index are discussed, and we compare the results of using alternative indices, such as an RMM-like index using OLR only, which is much easier to calculate. We will demonstrate several examples of when the loading on RMM and other indices is very high yet the accompanying OLR signal and the loading on our index is minimal, and vice-versa. Dynamical signals from reanalysis associated with the evolution of the MJO are shown to vary substantially, depending on the index used. The utility of the using an alternate OLR based index when one is primarily interested in the convective signal of the MJO will be discussed in detail.