On the role of the MJO in exciting

The role of intraseasonal variability in the tropics in exciting El Niño is investigated. The most important component is the zonal wind in the equatorial wave guide, 5◦S–5◦N, in the west and central Pacific, 130◦E–160◦W. This is correlated to the MJO, but there are components that are independent. We subtract the influence of ENSO on the indices of intraseasonal variability as well as from the evolution of the Niño3 index itself. Statistically, there is a strong lag correlation of zonal wind with deviations of the standard ENSO cycle four months later. The connection with the MJO is similar but weaker. The MJO only influences El Niño through its mean, the intraseasonal variability does not seem to play a role. In contrast, the correlation with the monthly mean zonal wind is roughly equal to the correlation with its monthly variability. The first effect can be understood as linear integration of high-frequency variability into lower-frequency variations by the ocean, the effect of the variability must be due to non-linear interactions. Finally, the mean zonal wind stress is correlated to central Pacific SST not in phase with eatsern Pacific SST, so that models that represent the central Pacific well can be expected to predict the onset of El Niño better than models that do not include warm pool physics correctly. 1 Ideas wind stress thermocline depth SST A A A AUA A A AK weather noise Figure 1: A schematic picure of the ENSO cycle A schematic picture of the ENSO cycle is given in Fig. 1. Wind stress anomalies on the equator in the western and central Pacific cause thermocline depth anomalies; wave dynamics transport these to the east where they are advected up as SST anomalies. These in turn affect the wind. Wind anomalies also influence SST directly, especially in the western and central Pacific. 2 Case study: 1997/98 The onset of the very strong 1997/98 El Niño was preceded by very strong westerly wind events in the western Pacific (e.g. McPhaden, 1999). These in turn coincided with strong Madden-Julian oscillations. Although a causal relationship seems obvious, modelling studies diverge on its influence. Among the many studies, van Oldenborgh (2000) concluded that the steep onset was mostly due to the westerly wind events in February, 1 VAN OLDENBORGH: ON THE ROLE OF THE MJO IN EXCITING EL NIÑO M A M F J 1997 D 1996 (∂N3/∂τx) ·δτx [K/week] 0.3 0.2 0.1 0 −0.1 δτx [Nm−2] 0.06 0.04 0.02 0 −0.02 Figure 2: The influence of zonal wind stress on the Niño3 index on Jun 1, 1997 (solid line) and the zonal wind stress in the area 5◦S–5◦N, 130◦E–160◦E (dashed line). March and April 1997, and hence unpredictable. However, the actual forecast from the December 1, 1996 analysis of the ECMWF seasonal forecast System-1, based on essentially the same ocean model, were in fact very good, although they did not include these wind events. This discrepancy will be investigated here. The spatial and temporal patterns influencing the strength of El Niño can be studied using an adjoint model (van Oldenborgh et al., 1999; Galanti and Tziperman, 2002). Only the adjoint ocean model is considered, which means that coupled interactions on time scales shorter than the run are neglected. During the onset phase this is a reasonable assumption for the eastern Pacific, but it is not so good in the central Pacific (see also Boulanger, these proceedings). Using this adjoint, deviations of the Niño3 index on a given date from climatology can be computed as ∆N3 = ∫ T 0 dt ∂N3 ∂τx(t) ·∆τx(t)+ ∂N3 ∂T0 ·∆T0 + smaller terms (1) where the partial derivatives are computed by the adjoint model, ∆τx(t) is the anomalous zonal wind stress and ∆T0 the anomalous 3D ocean temperature field at the start of the experiment. The dots indicate integration over the surface and ocean volume respectively. Specifically, for the 6-month forecast starting Dec 1, 1996 the influence of zonal wind stress on the Niño3 index at Jun 1, 1997 is shown in Figure 2. The zonal wind stress term explained about half of the rise of the Niño3 index, the other half was due to the initial state temperature: a deeper thermocline in the west Pacific. As can be seen in Fig. 2, the time evolution of the contribution of the zonal wind stress parallels the zonal wind stress in the equatorial wave guide, so the sensitivity is quite constant in this area. The spatial structure is illustrated in Fig. 3, which shows the week of the large westerly wind event around March 11, 1997. The spatial sensitivity is also quite constant in the area of interest (the large sensitivities in the eastern Pacific are due to model error caused by a target function with sharp corners, see Galanti and Tziperman (2002) for a better function). 3 Statistics The connections between the MJO, westerly wind in the equatorial wave guide and ENSO are investigated further using statistical methods over the last 15–25 years. For the MJO we use the 5-daily MJO01 index of the CPC minus MJO06 (1978–2003, Xue, 2003) , the westerly wind has been measured by the TAO array (∼1990–2002, McPhaden et al., 1998) and the Niño3 index is derived from the weekly Reynolds OIv2 analyses (1981–2003, Reynolds et al., 2002). These are plotted in Fig. 4 for the years 1996–1998. One immediately sees 2 VAN OLDENBORGH: ON THE ROLE OF THE MJO IN EXCITING EL NIÑO