Variability in the central equatorial Pacific , 1985 – 1989

We describe variability in the equatoriM Pacific Ocean near 160øW during the 5-year period 1985-1989, encompassing "normM", E1 Nifio, and La Nifia conditions. This description is based on conductivity-temperature-depth and acoustic Doppler current profiler data acquired during five cruises between 21øN and 4øS and on dynamic-height time series from an array based mainly on the Line Islands. At Jarvis Island, near the equator, the time series of dynamic height and near-surface temperature go back to 1981 and show the 1986-1987 E1 Nifio anomalies starting later in the year and having longer duration than those of the 1982-1983 E1 Nifio. Dynamic-height anomaly was less strong for the 1986-1987 event, but the near-surface temperature anomaly was of similar magnitude for the two E1 Nifios. The Jarvis near-surface temperature drop from 1986-1987 E1 Nifio maximum to 1988-1989 La Nifia minimum was 8øC. EmpiriCal orthogonM function analysis of the time series shows that interannual and interseasonal variability in dynamic height was dominated by a mode with meridional form similar to a first-verticM-mode Kelvin wave, while intraseasonal variability had a primary mode with • single pe•k •t 6øN •nd • secondary mode with pe•k •t 6øN •nd trough •t 2øN. While the equatorial thermocline deepened to the east and shoaled to the west during the 1986-1987 E1 Nifio, at 160øW it did not change depth during either this E1 Nifio or the subsequent La Nifia. Nevertheless, just befor e E1 Nifio and just after La Nifia, the thermocline was observed to be about 50 m deeper than at other times. The South Equatorial Current and North Equatorial Countercurrent had markedly reduced (increased) transports during this E1 Nifio (La Nifia). However, the Northern Tsuchiya Jet strengthened during E1 Nifio and weakened during La Nifia.

In discussing interannual variability, especially with regard to the onset and timing of ENSO signals, it is convenient to refer to some baseline of "normal" or average conditions.Details of the baseline calculation are given in the appendix.

Dynamic Height EOFs
A consistent dynamic-height data set from the LIA gauges was produced by converting subsurface pressure, acoustic travel time, and sea level to dynamic height.Five major and several minor data gaps caused by loss or failure of the instruments were filled by various means, depending on the site.Figure 3 ,,i,•,1,,,i,,,i,,,i,,,I,,,i,,,i ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,i,,,,,,,,,, ,,,,,,,,,,:,,,':',,,'   At the time of the final cruise in December 1989, atmospheric conditions in the equatorial Pacific had returned to a nearly normal state.Equatorial SST is within 1øC of that in the shuttle section.However, as in March 1986, the therrnocline has developed a downward tilt everywhere to the south of 8øN.This time, there is no thermocline trough, and the therrnocline slopes almost linearly downward to the south of 8øN all the way to the southern limit of the section at 4øS.In particular, at the equator the therrnocline is 50 rn deeper than during the normal conditions of March 1985.

Salinity (Figure 5b)
The  2 (see Table 3 for ranges of these currents).As with temperature, annual variability in geostrophic current in the shuttle is small compared with that in LIA.
Again as with temperature, the shuttle geostrophic section shows classic equatorial structure.The three main surface currents, westward SEC, eastward NECC, and westward NEC, have well-defined strengths, thick-  These ranges were chosen from Figure 5c.The shuttle section shows well-defined geostrophic currents [Lukas and Firing, 1984] because the influence of internal tides and nonsynoptic sampling has been strongly suppressed through averaging.This is not the case for our sections, and we expect that geostrophic calculations of velocity will become especially unreliable The NECC also has reduced meridional extent, lying between 3øN and 6.5øN (Figure 5c

Figure 1 .
Figure 1.Map. of Pacific Ocean showing the location of the Line Islands array.Inverted echo sounders (IESs)(crosses) were deployed at 10øN, 8øm, and 6øm along 102ø30'w (after 2 years the 6øN IES was moved to 159ø20'W).Pressure gauges (open circles) were deployed at Palmyra, F•[nning, Jarvis and Malden Islands.Tide gauges (solid circles) at Christmas and Penrhyn Islands are part of the Tropical Ocean-Global Atmosphere Pacific sea level network.
near-surface temperature at Jarvis by about 6 months; this is confirmed by a maximum in their cross-correlation function at 5.8 months.Also, Jarvis near-surface temperature and negative SO1 appear to be approximately in phase, and this too is confirmed by their cross-correlation function.After the strong 1982-1983 E1 Nifio [Lukas et al. 1984] the SOI was steady and close to zero until near the middle of 1986, when it began to fall, indicating a weakening of the trade wind system [Kousky .andLeetmaa, 1989].Winds in the western Pacific became anomalously westerly, and there were periods of strong westerly wind bursts in May and November-December 1986.These anomalous westerlies persisted until October 1987, and in response to them there was an increase in Kelvin wave activity across the Pacific Ocean [Delcroix et al., 1991], with one particular pulse in May 1986 well documented [Miller et al.as high as the 1982-1983 E1 Nifio peak, but this was mostly due to one intraseasonal oscillation: when low-pass filtered to remove variability with periods less than 150 days, the 1986-1987 E1 Nifio had a peak amplitude about half that of the

Figure 2 .
Figure 2. (top) The Southern Oscillation Index (SOI), (middle) Jarvis Island (0ø23'S, 160ø02'W) dynamic-height and (bottom) near-surface temperature anomalies from 1981 to 1989.The SOI data are monthly averaged values of Tahiti minus Darwin sea level pressure normalized by the standard deviation.Dynamic height and temperature have means and annual cycles removed.Also plotted is low-pass-filtered dynamic height (cutoff period 150 days).

Figure 3 .
Figure 3.Time series of dynamic height from the Line Islands array.Gaps have been filled as documented by Donohue et al. [1992] and plotted as dots.The dashed line superimposed on each record is its mean annual component.Means have been removed, and the offset between adjacent time series is 0.34 dynamic meters (dyn.m).

FiguresFigure 4b .Figure 5a .
Figures 5a and 5bshow nearly meridional temperature and salinity sections from the five LIA cruises.In both of these figures we have included a corresponding section reconstructed for June 25 (the midpoint of the LIA 3 section), with the constituents determined byFiring and Lukas [1985] from the Hawaii-to-Tahiti Shuttle Experiment.That experiment[Wyrtki et al., 1981] was conducted slightly to the east at longitudes

Figure 5c .
Figure 5c.Geostrophic velocity sections from (top left) the June 25 shuttle section and (right) from the last three LIA Cruises (LiA 3 to LIA 5).(middle, bottom left) Zonal velocity measured with the ADCP during LIA 4 and LIA 5.
general pattern is one of two high-salinity regions in the upper thermocline and two low-salinity regions, one in the lower thermocline and the other at the surface.One high-salinity region is a tongue from the South Pacific extending just north of the equator; the other is a tongue from the North Pacific extending equatorward to about 15øN.One low-salinity region is a tongue, below the northern high-salinity tongue, extending upward and equatorward to about 10øN; the other is a lens of water at the surface centered near 10øN.These features are all seen in a referenced to 1000 dbar was computed from the LIA and shuttle hydrographic data after averaging the half-degree-spaced dynamicheight field with a five-point Hamming filter to eliminate some aliasing caused by internal tides [Moumet al., 1987].Currents on the equator were calculated using L'H6pital'srule [e.g., Lukas and Firing, 1984; Picaut et al., 1989].Geostrophic sections made during LIA 1 and LIA 2 appear unrealistic (perhaps because of the poorer quality CTD measurements made during those cruises, see appendix) and are not shown here.ADCP measurements of current velocity were made only during LIA 4 and LIA 5. Equatorial current transports estimated from these sections are given in Table located at 4øN, has maximum speed a little over 10 cm s -• at 230 m depth.Since the STJ core is typically south of 4øS[Tsuchiya, 1975], only the edge of it appears in this section.Unlike all the LIA sections, the shuttle section shows the SEC as a current whose westward flow is unbroken at the surface, albeit with a pronounced velocity minimum due to EUC influence at 0.5øS.
the equator.It appears individual hydrographic sections cannot be used reliably to infer currents at latitudes below 30 .In June 1987, during E1 Nifio, the ADCP was not yet available, and we have only the geostrophic velocity section.It shows an EUC extending upward to the surface.Although within the equatorial band, where geostrophic calculations become unreliable, this feature may be correct, since it is also seen in the 1988 and 1989 ADCP sections.If we take it as correct, the EUC separates the SEC into two branches: a branch to the south of 2øS and a narrow northern branch centered on 2øN.With this reduced meridional extent the SEC transport is less than half that in the shuttle section.
June 1987 is only two thirds that in the shuttle section.In contrast, we obtain a transport of 54 Sv for the EUC, compared with 30 Sv for the June shuttle section, but because of the uncertainties in the geostrophic method near the equator, we cannot say whether the EUC transport is anomalous in June 1987.(Comparison of the ADCP and geostrophic estimates of the EUC for LIA 4 and LIA 5 suggests the geostrophic calculation grossly overestimates the transport).The EUC had normal strength at this time both at 110øW (Mr2) and at 165øE (Mr1), though at the latter location it weakened and reversed during the next 4 months in response to local westerly wind anomalies in June and July.The June 1987 section shows the NTJ at its usual position, but with double the speed and transport shown in the shuttle section.The November 1988 velocity sections, taken during La Nifia conditions, show strong, deep SEC and NECC flows, with transports about 50% greater than in the shuttle section.The ADCP and geostrophic sections both show that the SEC again appears split into two branches by an EUC which reaches the surface.Both the SEC and NECC have (~100 m) velocity maxima in excess of 60 cm s -•.The boundary between these two currents is at 5øN.The ADCP section (which, near the equator, is the more reliable section) shows an EUC having structure similar to that in the shuttle experiment, except that it breaks the surface between 2.5øS and iøN and carries about 75% more transport.The NTJ is weak.Because the surface flow near the equator is eastward and equatorial SST is warmer to the west [Delcroix et al., 1992], the cold SST anomaly along the equator is not the result of zonal advection.Easterly winds at Christmas Island were stronger than normal during 1988 (MP2), and cool equatorial SST is presumably the result of increased upwelling.For December 1989 the geostrophic and ADCP sections are similar only for latitudes farther than about 20 from the equator; near the equator they differ substantially.For this reason, we base the following discussion on the ADCP section.The EUC again appears to reach the surface, but the region of eastward flow is now especially large; it extends from about 2øS to 5øN (except for a small piece near the surface between 0 ø and IøN), and its core is deep (130 m).Moreover, the EUC carries a large transport (71 Sv).The northern branch of the SEC is very weak near the surface.The NECC is even weaker than in June 1987 but is located about 50 farther north.The NTJ is strong and is located at 3.5øN, about a half-degree closer to the equator than previously.Computed current transports at this time appear more like those during E1 Nifio than those at other times.This is perhaps a response to the western Pacific westerly wind bursts which began in late November 1989 dominated by a mode whose meridional form peaked at the equator and resembled the theoretical first-vertical-mode Kelvin wave.This mode had strong interannual and interseasonal oscillations.Its interseasonal oscillations were strongest during E1 Nifio.At periods less than 80 days there were two significant modes: a principal mode with a single peak at 6øN and a secondary mode associated with meandering of the SEC/NECC dynamic-height ridge.These two modes were strongest around the end of the year and during La Nifia.Those are the times when currents are strongest, and it is probable that both modes represent oscillations of the ridge by tropical instability waves.The temperature section obtained on the cruise been associated with heat buildup on the ,western side of the Pacific prior to E1 Nifio.Dynamic height and near-surface temperature rose rapidly with the onset of E1 Nifio and reached highest values by the end of 1986 before dropping at different rates.In June 1987, 6 months after the peak of the event, dynamic height had fallen to about half its maximum value, although near-surface temperature still remained close to its highest value.By this time the thermocline had relongitude of lO0øW was, as it were, the pivot point for the seesawing equatorial thermocline.In June 1•87, increased rainfall during E1 Nifio resulted in lower surface salinity; also, SST was 4øC warmer than in March 1•85 and lacked an equatorial minimum.The transition from the 1•86-1•87 E1Nifio to the 1•88-1.In all velocity sections the EUC reached the surface, separating the SEC into two branches, but uncertainties introduced by geostrophic calculation near the equator prevented reliable determination of EUC transport in June 1987.Current transports during E1 Nifio and La Nifia were considerably different from those computed from the shuttle data.In June 1987, 6 months after peak E1 Nifio conditions, the SEC transport was reduced by nearly 60% and the NECC by 35% from shuttle values.In November 1988, near the peak in La Nifia, SEC and NECC transports were both almost 50% larger than shuttle values.NTJ transport was affected oppositely, being larger by a factor of 2 during E1 Nifio and smaller by a factor of 2 during La Nifia.In December 1989, shortly after the end of La Nifia, there was a large region of eastward flow encompassing the EUC; also the thermocline trough, which is usually found near 4.5øN was absent and, consequently, the NECC was weak.However, in the next paragraph we present evidence that this absence of the thermocline trough was a brief occurrence which would have lasted only a week or two.To what extent are the differences between our hydrographic sections representative of interannual signals?Our low-pass EOF, which predominantly describes the interannual ENSO signal, is limited to the region within about 50 of the equator.Yet, in the hydrographic sections we see significant changes in thermocline depth extending to 8øN (section 4bottom left) passed through zero, going from high to low dynamic height.Putting these facts together, it appears dynamic height due to intraseasonal waves at the CTD section would have been close to a minimum at the time the CTD casts were taken.Moreover, the meridional structure of the waves (Figure 4b, top left) resembles that of the mean dynamicthat resulted in gradual drifts of up 0.4 dbar in their output signals.During the last year's record (November 1988 to December 1989) the 4øS gauge again showed an apparent drift.Fortunate]y, there were enough other measurements (in situ and satellite) to allow us to show with confidence that the drifts there was only i year in which we had data from all sites, these gaps were filled appropriately to obtain a data set which could be analyzed in a uniform manner (Figure 3).Details of the gap filling are given by Donohue et al. [1992].In particular, three gaps of ,,,1 year each at 4øS (1985-1986), 8øN (1986-and ended in Honolulu.On every cruise, CTD casts were made at each IES site and near each island instrumented with a pressure gauge.Sections were also made with casts at every degree of latitude (every half degree between 3øN and 3øS) SBE-09 on LIA 3. Water samples for conductivity calibrations were collected with Niskin bottles.On LIA 4, and 5 the Sea-Bird SBE-09 CTD was used with shipboard recording of the data and a the central equatorial Pacific are limited, but the Hawaii-to-Tahiti shuttle Experiment [Wyrtki and Kilonsky, 1984] from January 1979 until June 1980 was conducted in years without either pronounced La Nifia or E1 Nifio conditions and therefore provides one description of normal conditions.Firing and Lukas [1985] computed mean, annual, and semiannual constituents of the temperature, salinity and

Table 1 . Locations and Depths of Line Islands Array Instruments.
IES is inverted echo sounder, SSPG is subsurface pressure gauge, and SLG is sea level gauge.•Data obtained from February 22, 1985 to June 8, 1987.b Data obtained from June 14, 1987 to December 13, 1989.½Latitude is south.