The Argo gridded data have been shown to accurately
resolve the individual equatorial Kelvin waves forced by
the MJO over the four-year study period (Fig. 2). The
average MJO cycle in thermocline depth is shown in
a composite Hovm¨oller diagram (Fig. 9a) where time
is lagged relative to phase 1 of the MJO, when the
active convection is beginning over the Indian Ocean.
Clear eastward propagation can be seen right across
the Pacific. Negative D20 anomalies, corresponding to
an upward displacement of the thermocline, start in
the western Pacific around lag 12 days, which corresponds
approximately to phase 3. These are consistent
with forcing by westward surface wind stress anomalies
(Fig. 4c) and surface Ekman flux divergence along
the equator. This negative D20 anomaly then propagates
eastward into the central and eastern Pacific
with a phase speed of approximately 2.5 m s−1 (thick
dashed line in Fig. 9a) and an amplitude above 3 m.
Similarly, positive D20 anomalies at lag −12 (approximately
phase 6) in the western Pacific are consistent
with eastward wind stress anomalies in phases 5 and