Wave Guides in the ITF and Their Relation to Remote Wind Forcing

As was shown by Meyers (1996), the interannual variability along IX1 is strongly related to the large-scale wind-stress fields over the Pacific and Indian Oceans. The nature of the relation is qualitatively accounted for by linear wave theory, predicting that equatorial Kelvin and Rossby waves excited by variation in large-scale wind stresses excite coastal waves, which carry the signal to higher latitudes. Here we examine the simulated and observed details of the relationship between these remote wind forcing with sea level and temperature variability in the region at seasonal and longer timescales.

As recognized by Clarke and Liu (1994), the zonal wind stresses along the equator of the Pacific and Indian Oceans are significantly correlated in the ENSO band, such that during ENSO the weakened Pacific Trades are often accompanied by easterly wind anomalies along the Indian Ocean equator. However, while Pacific winds are dominated by the ENSO band, the Indian Ocean winds have much more energy at higher frequencies with spectral peaks near periods of 3 years and between 1 and 2 years. This higher frequency "signature" of equatorial Indian Ocean winds versus the lower frequency energy of the Pacific winds is crucial in distinguishing the impact of these two sources of remote energy within the Indonesian seas.

To identify the large-scale variations of sea level anomaly (SLA) as a response to wind forcing, a lagged multiple regression is performed at each point of a mapped altimetry data set on wind stresses 11 (Wijffels and Meyers, 2004):

SLA(t) - ap[-tp(t + Lp)] + aIti(t + Li)+ aLtL(t + Ll) (1)

where the 4-month low-passed seasonal SLA, at time t, is expressed as the sum of three wind indices, ti, at lag Li with coefficients ai. The wind indices used are the zonal wind averaged along the Pacific equator (note the reverse sign, so that positive wind anomalies give positive temperature anomalies); the along path winds on a path extending along the coast from Sunda Strait to the equator and then along the equator to Africa; and a local wind index. These indices - Pacific, Indian and Local - are denoted by P, I, and L, respectively in equation (1). Each wind index had the seasonal cycle removed, was low-pass filtered at 4 months, and normalized before the multiple correlation was performed. Thus coefficients in equation (1) can be interpreted as the sea-level change associated with a single standard deviation change in the wind index.

All possible lags within 718 months were searched to find the combination that accounted for the maximum variance at each grid point. Since the wind indices used in equation (1) are normalized, the calculated coefficients are in real physical units (cm). Hence, the response during a peak in the wind index (normalized amplitude of 3) is then given by three times the coefficient plotted. Data used by equation (1) account for up to 90% of 4-month low-passed seasonal anomalies of sea-level height collected since late 1992 throughout the throughflow region and over large parts of the Pacific and Indian Oceans (not shown). We find that various estimates of "local" winds are unable to capture much variance (and hence are not shown). The results based on sea level described below have also been confirmed by results from a similar analysis of the temperature fields along the XBT lines from both the observations and the model. For further details about the method, the reader is referred to Wijffels and Meyers (2004).

Observation-based coefficient and lag of sea level for the Pacific wind index (Fig. 10a, b) show that high sea levels occur in the western Pacific in response to easterly wind anomalies, and much of this signal "leaks" along the New Guinea—Irian Jaya-Australian wave guide down to the southern tip of western Australia, with some suggestion of offshore radiation of energy north of 22°S. Coefficients are large and positive near the west Australian coast. The sense of the multiple correlation is that a large warming and high

Figure 10a, b: Coefficient in centimeter (top) and lag in months (bottom) of low-frequency anomalies of SSH for Pacific wind index: (a, b) observations and (c, d) model (For colour version, see Colour Plate Section).

sea levels off the western Australian coast are associated with easterly wind anomalies along the Pacific equator, a result obtained previously by Meyers (1996). Downwelling equatorial Rossby waves excited by the wind anomalies in the central and western Pacific propagate westward and excite coastally trapped waves off Papua New Guinea, which then propagate anticlockwise around Australia. Part of this signal also continues around the southwest tip of Australia and along the south coast (obscured by the spatial smoothing in the plot). Lags off western Australia are between 0 and 2 months. In sea level, the maximum lags occur near 18°S.

In the subtropics, the Pacific wind response dies out near 100°E, in agreement with Masumoto and Meyers (1998) and Potemra (2001), who

found that near this longitude, regional Ekman pumping along Rossby wave characteristics begin to dominate over variability originating at the eastern boundary. The simulated correlation pattern looks similar but is significantly weaker in amplitude (Fig. 10c, d). This is particularly visible in small amplitudes in the far western Pacific and also in the weak wave propagation from the Pacific into the Indian Ocean and along the western Australian coastline. The simulated phase lags off western Australia are between 2 and 4 months. This "weak" ENSO response in the model is also confirmed by analysis of its temperature field, showing that the response of the modeled thermocline along the Australian coast to ENSO winds is too diffusive and too deep (Fig. 4).

Interestingly, in both model and observations there is also a strong response to the Pacific winds northeast of Madagascar with lags that suggest a southward propagating response, due to either local wind changes that correlate with ENSO or expressions of later arrivals of Rossby waves at higher latitudes. However, both the relatively short satellite altimetric record and the short model simulation are likely to be dominated by the very large 1997/1998 ENSO event and strong Indian Ocean dipole that occurred during the same years (Saji et al., 1999) - a situation not likely a representative of a longer record.

The simulated response in sea level to Indian Ocean winds agrees well with observation-based results (Fig. 11): westerly wind anomalies



Figure 11a, b: As for Fig. 10, but for the Indian wind index (For colour version, see Colour Plate Section).


Figure 11a, b: As for Fig. 10, but for the Indian wind index (For colour version, see Colour Plate Section).

60 80 100 120 140 160 180 Longitude

Figure 11c, d Continued (For colour version, see Colour Plate Section).

giving high sea levels in the eastern Indian Ocean equatorial and coastal waveguides. This signal also penetrates into the western portion of the Banda Sea, Savu Sea, and Makassar Strait at near-zero lag. The propagation of the equatorially excited waves around the Bay of Bengal is also evident in the observation-based plot, as is the radiation of energy westward from the eastern boundary off Sumatra and Java. The western and central Indian Ocean response to the equatorial zonal wind anomalies is also evident with lower sea level along 8°S at near zero lag in both model and observations.

Despite the overall quite successful simulation of this response by the model, there are some features that are missing. The radiation of equatorially forced waves around the Bay of Bengal is entirely missing in the model but is clear in the observations. Also, the response east of Lombok Strait is very weak in the model, unlike in the observations, where much energy is seen to "jump" the strait and reach Ombai Strait near 125°E. In the model, Lombok Strait is known to be too wide and "dynamically slippery.'' This is apparent in the fact that the bulk of the throughflow transport passes through Lombok in comparison to reality where only 20-30% actually does. Interestingly, despite this, little of the remotely forced wave energy appears to pass through Lombok Strait into the internal seas of the Indonesian archipelago.

Was this article helpful?

0 0

Post a comment