How the 1997–98 El Niño developed

The 1997–98 El Niño peaked at +2.28 on NOAA's Relative Oceanic Niño Index (RONI) in NDJ 1997 (+2.37 on the traditional ONI) and was documented from its rapid onset to its sudden demise by a newly completed tropical Pacific observing system [15]. We reconstruct its development from those observations, the published event studies and new reanalysis diagnostics, and grade each proposed cause by the strength of event-specific evidence. The equatorial Pacific recharged after the 1995–96 La Niña: warm water volume was positive from December 1995 and peaked in April 1997. Onto that charged state a sequence of westerly wind events from December 1996 to June 1997 — organized by two exceptionally strong Madden–Julian oscillations and amplified by twin cyclones linked to East Asian cold surges — forced downwelling Kelvin waves and pushed the edge of the warm pool east. An ocean-model decomposition attributes about half of the eastern anomaly on 1 June 1997 to that wind stress and about 40% to the initial ocean state. Coupling was established by May–June and the event grew to a December peak. Its end was set by the seasonal southward shift of the westerlies from November 1997 and came abruptly in May 1998. Forecast models predicted warming but none predicted its strength before it was under way. The Indian Ocean Dipole of 1997 followed the Pacific warming, so it played no part in onset; whether it was forced by El Niño or arose internally remains disputed. We find no basis for attributing the event's amplitude to background warming.

1. Precondition (grade A−; high confidence). Recharge after the 1995–96 La Niña: warm water volume positive from December 1995, +3.12 × 10¹⁴ m³ in April 1997; about 40% of the 1 June 1997 Niño 3 anomaly comes from the initial thermal state in one ocean model [8] [14] [25].
2. Trigger (grade A; high confidence). Westerly wind events from December 1996 to June 1997 — calendar 1997 had the most reanalysis bursts of any year in 1979–2001 — confirmed by TAO moorings; wind stress supplies about half the 1 June anomaly and about 90% of the difference from 1996, when no El Niño developed [25] [10] [9] [18].
3. MJO and extratropical amplification of the trigger (grade B; medium–high confidence). Two exceptionally strong MJOs in December 1996 and February 1997 organised the bursts; East Asian cold surges and twin cyclones intensified them [29] [23] [16].
4. Onset mechanism (grade A−; high confidence). Kelvin-wave currents advected warm water east (central Pacific warming in late March – early April 1997), and the March event's own SST response drew in further westerlies [29] [27] [28].
5. Amplifier (grade B; high confidence). Bjerknes feedback from May–June 1997, with zonal advection important in the central Pacific [8] [18] [54].
6. Termination (grade A; high confidence). Seasonal southward shift of the westerlies from November 1997, eastward-propagating thermocline shoaling and basin discharge; abrupt surface cooling in May 1998, with an MJO implicated in its timing [57] [58] [59] [56].
7. Indian Ocean (not a cause of onset; origin disputed; grade B). The 1997 dipole began after Pacific coupling. One line of work traces it to El Niño's effect on the Indian Ocean winds; another argues it was mainly internal Indian Ocean dynamics [51] [50] [47].
8. Predictability. No model forecast even half the peak amplitude at 6–11 months' lead; the rapid succession of bursts in early 1997 was a rare draw from the atmosphere's noise [62] [60] [40].

By 1997 the tropical Pacific was far better observed than in 1982. The Tropical Ocean–Global Atmosphere (TOGA) programme had built the TAO array of moored buoys, an island tide-gauge network and satellite missions for surface winds, SST and sea level [3]; TOPEX/Poseidon altimetry covered 1992 onward [19]; and 1997–98 was the first major event predicted in real time by numerous forecasting groups [62]. The 1997–98 El Niño nevertheless surprised forecasters in both its speed and its strength, and its abrupt end in May 1998 was also unexpected [15] [18] [60].

The event therefore offers a test that 1982–83 could not: whether the processes proposed for El Niño onset — recharge of equatorial heat, wind-forced Kelvin waves, intraseasonal westerly bursts and coupled feedback — can be separated and quantified for a single event when the ocean is observed directly. A lively debate followed over whether the intense westerly wind events or the slower oceanic preconditioning mattered more [25]. This paper reconstructs the sequence month by month and grades each factor, using the same framework, grades and index (RONI) as Paper 1.

Index. RONI (the three-month mean of relative Niño 3.4, i.e. Niño 3.4 minus the 20°N–20°S tropical-mean anomaly, rescaled) is the primary index [1] [2]; the traditional ONI is shown alongside. For this event the two agree closely (peak +2.28 vs +2.37), so the classification as a very strong event does not depend on the index choice.

Observations. TAO/TRITON moorings provide daily surface winds, SST and thermocline depth along the equator; TOPEX/Poseidon sea level gives Kelvin and Rossby wave signals; NOAA CPC and PMEL indices give monthly SST, SOI, 850 hPa trades, OLR, upper-300 m heat content (180°–100°W, 1981–2010 base) and warm water volume (WWV) [8] [3] [14]. SST indices are from OISST v2.1 [5] and ERSSTv5 [4].

Reanalysis diagnostics. To compare with 1982–83 on identical definitions, the westerly-burst and MJO diagnostics of Paper 1 were re-run for July 1995 – December 1998: daily NCEP/NCAR Reanalysis 1 winds and interpolated OLR, anomalies against a 1979–2000 mean plus three annual harmonics; a burst is a 5°S–5°N 1000 hPa anomaly of at least 5 m s⁻¹ for at least 3 days over at least 10° of longitude within 120°E–150°W. MJO activity is described with the Wheeler–Hendon RMM index and OLR-based measures, whose amplitudes can differ for the same event [6] [7]. Daily TAO/TRITON winds and 20 °C isotherm depth on the equator give an independent, directly observed check on the reanalysis bursts and trace the Kelvin waves; the definitions (westerly ≥ 5 m s⁻¹ on ≥ 2 days at 147°E–180°; intraseasonal isotherm peaks ≥ 8 m) were fixed before the results were examined [9] [10].

Evidence grades. A: event-specific observation plus a model or budget study isolating the effect. B: event-specific observation with a plausible mechanism and supporting studies. C: correlation, composite or general theory only. D: speculative, disputed or untested for this event. The grade describes the link to 1997–98, not the general validity of a mechanism.

Limits of this review. For several central studies only the abstract could be read (the reference list records what was read for each); conclusions that depend on details beyond an abstract are marked as such.

The event grew out of the 1995–96 La Niña, which reached RONI −1.02 in OND 1995 [8]. As the trades relaxed after that La Niña the equatorial warm-water reservoir refilled: WWV turned positive in December 1995 and stayed positive throughout 1996 (+0.26 to +0.85 × 10¹⁴ m³), while heat content east of the Date Line hovered near zero (−0.47 to +0.17 °C) [8]. This is the recharge state that recharge-oscillator theory and the 1980–99 WWV record identify as the precondition for El Niño [13] [14] [12].

At the surface, 1996 stayed ENSO-neutral and slightly cool (RONI −0.20 to −0.37 from MAM 1996 to DJF 1997), with the far eastern Pacific cold (Niño 1+2 −0.7 to −1.6 °C) [8]. In autumn 1996 the western Pacific was warm and the warm pool extended east, a pattern also present before the aborted event of 1989–90 and before 1982 [29]. Between November 1996 and January 1997 the eastern Pacific had cold equatorial SST and high sea-level pressure, while off-equatorial warm SST and anomalous cyclones in the far west accompanied a switch of equatorial wind anomalies from easterly to westerly [17].

Recharge alone did not determine the outcome. An adjoint analysis with an ocean general circulation model attributes about 40% of the 1 June 1997 Niño 3 anomaly, relative to an average year, to the initial thermal state on 1 December 1996 — concentrated in the deep layer of warm water along the equator in the western Pacific, "in agreement with a recharge mechanism" — and about 50% to the wind. Compared with 1996, a year with a weaker but real build-up and no El Niño, about 30% of the difference comes from the stronger build-up and about 90% from the strong westerly wind events that 1995–96 lacked (the shares exceed 100% because other terms were negative) [25].

Westerly wind events, December 1996 – June 1997. A series of westerly wind bursts from December 1996 to June 1997, notably in March 1997, was instrumental in setting up the event [18]. In the adjoint analysis three wind-stress peaks — early March, late March and early April 1997 — contribute about 0.6, 0.3 and 0.5 °C to the modelled 1 June Niño 3 anomaly, and a strong December 1996 event about 0.4 °C over a negative baseline; each generated downwelling Kelvin waves that deepened the eastern warm layer two to three months later [25]. The March event was the strongest of the winter in stress (up to 0.2 N m⁻²), fetch (more than 30° of longitude) and duration (about a month, mid-February to mid-March) [27]. Satellite (SSM/I) winds show westerly anomalies almost every month from September 1996 to March 1997, with the late-December and late-February episodes particularly intense and long-lived; they were embedded in the active phase of the MJO but greatly enhanced over the western Pacific by equatorial twin cyclones, themselves induced by northerly cold surges from East Asia and the western North Pacific [23]. A best-track study of the same period finds twin tropical cyclogenesis tied to MJO westerly bursts in mid-October 1996 over the Indian Ocean and in late May 1997 over the western Pacific [24]. Two exceptionally strong MJOs, in December 1996 and February 1997, first cooled the western Pacific through cloud shading and evaporation, with off-equatorial upwelling adding to the cooling in February [29]. On Paper 1's definitions the reanalysis contains 10 bursts between December 1996 and June 1997 and 16 in calendar 1997 — the most of any year in 1979–2001 (mean 4.2 a year; 1982 had 13, while 1995 and 1996 had 0 and 3). The strongest were on 20–31 December 1996 (+11.8 m s⁻¹ at 137.5°E), 3–19 March 1997 (+10.3 m s⁻¹ at 150°E) and 10–21 April 1997 (+10.8 m s⁻¹ at 155°E). TAO moorings observed the same episodes directly, with daily westerlies of up to 8.7 m s⁻¹ at 156°E in late December, 9.9 m s⁻¹ at 147°E in late February – early March and 11.0 m s⁻¹ at 165°E in mid-April [10] [9].

Kelvin waves and the eastward jump of the warm pool. In TAO records from October 1996 to December 1997, intraseasonal variability east of the Date Line was dominated by first-baroclinic Kelvin waves travelling east at 2.4–2.7 m s⁻¹ [20], consistent with the 2.4 m s⁻¹ found for the decade before [31]. The onset was linked to such waves forced by intraseasonal oscillations originating over the Indian Ocean [16]. The TAO thermocline record traces two of the resulting downwelling Kelvin waves: an intraseasonal deepening at 165°E on 02 January reached 110°W on 23 February (about 2.1 m s⁻¹), and one at 165°E on 16 March reached 110°W on 23 April (about 2.88 m s⁻¹) [9]. The first wave's sea-level signal reached the South American coast in February, and a second, larger group arrived in April–May, greatly depressing the eastern thermocline [23]. In late March and early April 1997 the Kelvin wave forced by the February MJO warmed the central Pacific, mainly by zonal temperature advection; afterwards winds, currents and SST gradients did not return to their earlier state [29]. In ocean-model experiments the March event displaced the eastern edge of the warm pool about 2000 km in a month, cooled the far west by about 0.8 °C and warmed the Kelvin-wave path by about 0.5 °C [26] [27]; a nonlinear interaction with the salinity and temperature fronts at the warm-pool edge tripled the surface currents relative to a linear response [28]. In an atmospheric model forced with those SST changes, two responses favoured rapid growth: weaker trades in the central-eastern Pacific, and an eastward shift of convection that promoted further frequent and intense westerly events in the following months [27].

Warming in two places. SST warmed first near the Date Line in January–February 1997 and in the far east from March (Niño 1+2 from −0.45 °C in February to +0.80 °C in April) [17] [8]. The eastward-moving and westward-spreading warm anomalies merged; once the central and eastern warm regions joined, coupling between SST and winds became fully effective [17] [18]. WWV peaked at +3.12 × 10¹⁴ m³ in April 1997 [8].

Coupled growth, May – December 1997. RONI rose from +0.83 (AMJ) to +1.20 (MJJ), with SOI −1.3 and −1.4 and central-Pacific OLR near −26 W m⁻² [8]. Growth continued to SON (+2.22) and the peak in NDJ (+2.28); in October the central-Pacific trade index reached −4.1 and OLR −49.8 W m⁻² [8]. Much of the western warm water was carried east by anomalous surface currents [18]. Interannual heat budgets from TAO moorings show zonal advection important everywhere and most significant in the central Pacific, vertical heat flux especially important in the east, and surface heat fluxes generally damping the anomalies [54]. Eastern equatorial SST set a new monthly record every month from June to December 1997, the cold tongue failed to form, and the strongest westerlies and deep convection occurred only over water warmer than about 29 °C [15]; by December cold-tongue anomalies averaged nearly 4 °C, with maxima above 5 °C, and the eastern thermocline was more than 90 m deeper than normal [15] [16].

Discharge began before the surface peak. WWV fell to +0.66 × 10¹⁴ m³ in November and turned negative in December 1997 [8]. From November, near-Date-Line westerly anomalies shifted south of the equator, a shift reproduced as the atmosphere's deterministic response to the observed SST field; the oceanic response was thermocline shoaling that began in the central Pacific and propagated east [58] [57]. The same seasonal migration of the warmest water appears as a general mechanism of El Niño termination in observations and coupled models [55] [59].

Eastern SST stayed warm from December 1997 to May 1998 despite strong subsurface cooling [57]. Upwelling signals arrived from several directions: Kelvin waves generated by easterlies and boundary reflection from the west, Rossby waves from westerly forcing in the east, and the general uplift of the thermocline as the basin discharged [18] [16] [19]. In the west, off-equatorial pressure turned high and easterly anomalies developed over the far western Pacific [17]. When the trades returned to near-normal strength in the eastern Pacific in mid-May 1998, SST at 0°, 125°W fell 8 °C in 30 days, more than ten times the normal rate of cooling for the season [15], and eastern equatorial SST fell by more than 4 °C within two weeks [57]; the cooling began between 125°W and 170°W [16]. The cause of the abrupt return of the trades was unknown when McPhaden wrote, and no forecast anticipated it [15]; Takayabu et al. relate the termination to an MJO [56], and an MJO was indeed active from late April to the end of May 1998 [10]. La Niña followed (RONI JJA 1998 −1.05), with the largest equatorial phytoplankton bloom then observed [8] [22].

Intraseasonal variability and bursts. The MJO was unusually active through the onset: the PSL OLR MJO index stayed at or above 1 almost continuously from late October 1996 to mid-June 1997, intraseasonal OLR variance over the Indian Ocean and western Pacific in October 1996 – March 1997 was the largest of 22 seasons in 1979–2001, and the bursts of 25 December, 13 March and 28 March occurred at index amplitudes of 2.7–3.3 [10]. Indices disagree for some bursts: on 13 April the OLR-based index was 2.0 but a circulation-weighted index only 0.9, as expected when indices weight convection and winds differently [7]. In a wave model forced with observed winds, MJO-forced downwelling Kelvin waves reached 15–20 cm in sea level, an order of magnitude larger than the waves reflected from the western boundary [16]. Yet an active MJO in January–April 1996 produced no El Niño, and the late-1996 bursts intensified to over 9 m s⁻¹ in the Pacific while the MJO's own westerlies had fallen below 2 m s⁻¹ over the Maritime Continent [23]. Across many events, MJO activity in the western Pacific in late boreal spring leads El Niño the following autumn [33], and stronger MJO Kelvin-wave forcing preceded larger eastern anomalies by 6–12 months in 1980–99 [32]. Yet waveguide warming follows MJO events only when they contain a westerly wind event, and an MJO does not make a burst more likely [35]. For 1997 the two meet: the bursts that mattered were embedded in strong MJOs [23] [29]. An intermediate coupled model initialised on 1 January 1997 produced a weak warm event on its own, and one about 50% stronger when the rectified effect of MJO-like winds was imposed [30].

Are bursts a cause or a response? Bursts become more likely as the warm pool extends east, and treating them as partly SST-dependent roughly doubles modelled event amplitude [37] [38]; statistics also show bursts migrating east ahead of the El Niño peak [36]. In 1997 the warm pool had extended east by autumn 1996 [29], so some burst activity may already reflect the developing state. This does not remove the trigger role: the decomposition finds the December, March and April events to be the main difference from 1996 [25], and the impact of a burst depends on its timing relative to the oscillation and on the background state [39].

Extratropical influence. The best-documented pathway is synoptic: East Asian cold surges induced the twin circulations that intensified the December and February bursts [23]; best-track twin cyclones, in contrast, appear in October 1996 and late May 1997 [24], so the two studies differ in what they count as a twin-cyclone event. The Pacific meridional mode and seasonal footprinting precede many El Niños [42] [43] [44], but no study read here isolates them for 1997 (grade C).

Decadal background. The PDO was strongly positive through 1997, and the shallow meridional overturning circulation was at a low during 1992–98 before rebounding [8] [45] [46]. These describe the background; their contribution to this event is not quantified (grade C).

Indian Ocean. Cool SST appeared in the eastern Indian Ocean along and south of the equator from strong upwelling along the equator and Sumatra, and warm anomalies developed in the west from July 1997, peaking in February 1998 [49]; the CPC dipole index rose from +0.55 °C in July to +1.39 °C in November [8]. One analysis attributes the basin's SST changes largely to ENSO's effect on the atmosphere — the reversal of the Walker circulation and displaced south-east trades [50]; Webster et al., by contrast, concluded that the 1997–98 anomalies "may primarily be an expression of internal dynamics" of the Indian Ocean: the Indian Ocean rim climate differed from the usual El Niño pattern, the ENSO correlation with the along-equator SST gradient is not significant (+0.19), and of 16 years with a reversed gradient in 1950–98 only three were El Niño years [47]. Saji et al. defined a dipole mode of the tropical Indian Ocean the same year (title only read) [48]. A reanalysis and altimetry study traces the triggering to El Niño itself: divergent easterly anomalies over the equatorial Indian Ocean in summer 1997 weakened the western westerlies and strengthened the eastern easterlies, producing the east–west SST contrast by autumn through evaporation and upwelling, which ocean Rossby and Kelvin waves then reinforced [51]. Whatever its origin, the dipole developed after Pacific coupling (western warming from June, eastern cooling from July 1997) [47], so it cannot have contributed to onset; coupled forecast experiments suggest an extreme IOD can enhance El Niño [52], but this was not tested here for 1997.

Tropical Atlantic. The tropical North Atlantic was near or slightly below normal in early 1997 (−0.37 °C in March) and warm in 1998 (+0.51 °C in May) [8]. Spring cooling in the north tropical Atlantic has been proposed as an El Niño trigger in general [53]; for 1997 this is untested (grade C/D), and the 1998 warming is the usual lagged response.

Volcanic and external forcing. The Smithsonian Global Volcanism Program catalogue lists no eruption of VEI ≥ 4 anywhere in 1995–97; the most recent tropical VEI 4 eruption was Rabaul (Papua New Guinea, 4°S) in September 1994, more than two years before the December 1996 burst [11]. Volcanic forcing is therefore not a candidate for this event.

Background warming. The tropical-mean adjustment in RONI changes the peak by only 0.09 °C (RONI +2.28, ONI +2.37) [2] [8]. Contemporary debate concerned whether the post-1976 run of warm events was unusual [63]; projections of more frequent extreme El Niños under greenhouse warming [64] do not attribute this event. No event-specific attribution is made here.

The most plausible sequence is: (1) recharge after the 1995–96 La Niña created a deep warm-water reservoir by late 1996; (2) a run of strong, MJO-organised and extratropically amplified westerly wind events from December 1996 to June 1997 forced downwelling Kelvin waves and moved the warm pool's edge east; (3) the resulting central-Pacific warming in March–April and coastal warming from March merged by May–June, when the Bjerknes feedback took over; (4) the seasonal southward shift of the westerlies from November and basin discharge ended the event abruptly in May 1998.

The main competing view weights the slow ocean more heavily: with recharge and western-boundary reflection already favourable [18] [19], an El Niño was likely regardless of the bursts. The available quantitative answer splits the difference — roughly 40% initial state and 50% wind stress at 1 June [25] — but comes from one ocean model forced with one wind product. A noise-driven ensemble started at the end of 1996 favours a moderate event over the observed extreme [40], which supports the view that the bursts set the amplitude rather than the occurrence. Bergman et al. reach a similar conclusion for the MJO: relevant to timing and initial growth rather than responsible for the event itself [29].

Unresolved: how much of the burst activity was itself a response to the eastward-extended warm pool; whether the Indian Ocean dipole was forced or internal, and whether it fed back on the peak; and the role of extratropical modes beyond the documented cold surges.

This was the first very strong El Niño predicted in real time by many groups. Most models forecast some warming one to two seasons before the boreal-spring onset, but none predicted its strength until it was already becoming very strong in late spring; the best forecasts of Niño 3.4 before positive anomalies appeared were about +1 °C against an observed +2.5 to +3 °C [60]. Against a climatology-and-persistence baseline, no model anticipated even half the peak amplitude at 6–11 months' lead, and none was skilful for both onset and decay at 6–8 months [62]. Once the event had developed, more models captured the late-1997 peak and the 1998 reversal [60]. The two assessments differ in emphasis: NCEP's own review finds that its coupled model's spring-1997 forecasts, "while somewhat too weak", clearly suggested a developing strong El Niño, and that comprehensive coupled models (NCEP, COLA, ECMWF) outperformed simple dynamical and most statistical models [61]; the comparison with a persistence baseline finds no model skilful at 0–8 months [62]. Both agree that the rapid spring onset was not foreseen in its full strength. The rapid succession of bursts in early 1997 was a relatively rare event, which limits predictability [40] [25].

• The 40/50% decomposition comes from one ocean model and wind product [25].
• Saji et al. (1999) and Takayabu et al. (1999) were read only as bibliographic records; several other studies only as abstracts (see the reference list).
• Reanalysis burst and MJO diagnostics depend on index and definition; RMM and OLR-based indices can disagree for the same event [7].
• Off-equatorial and sub-thermocline conditions before Argo are less certain than the TAO line itself.

The 1997–98 El Niño developed from a recharged Pacific, but its speed and size came from a rare run of westerly wind events in December 1996 – June 1997, organised by strong MJOs and intensified by twin cyclones tied to East Asian cold surges. Those events forced Kelvin waves and an eastward jump of the warm pool that joined central and eastern warming by May–June, after which coupled feedback carried the event to a December peak. Its demise, set in motion by the seasonal southward shift of the westerlies, was as abrupt as its onset. Compared with 1982–83, the trigger is far better documented: the ocean was observed directly, and a quantitative separation of precondition and wind forcing exists for this event.

Primary index: NOAA's Relative Oceanic Niño Index (RONI), official since 1 February 2026 [1] [2]. On RONI the 1997–98 peak is +2.28 (NDJ 1997); on the traditional ONI it is +2.37. The two agree within 0.1 °C for this event, unlike 1982–83 (RONI +2.40, ONI +2.14).

Observing system in 1997. The first strong El Niño observed from start to finish by the TOGA observing system — the TAO array of moored buoys, drifting buoys, island tide gauges, ship XBT lines — and by satellite winds, SST and TOPEX/Poseidon sea level (from 1992) [3] [15] [18] [19]. Onset, growth and termination are therefore documented daily along the equator, which allows heat-budget and wave analyses that were impossible for 1982. Argo did not yet exist; the subsurface off the equator and outside the array remains less certain.

Values from NOAA CPC, PMEL and NCEI files as held in the site's data archive; read from the source files, not retyped.

Reanalysis bursts on Paper 1's definitions (1000 hPa anomaly ≥ 5 m s⁻¹, ≥ 3 days, ≥ 10° of longitude, 5°S–5°N, 120°E–150°W), with the PSL OLR MJO index (OMI) and the circulation-weighted index at the burst peak [10].

Answer. Strong evidence that the MJO set the timing and much of the amplitude of the onset bursts, through Kelvin waves an order of magnitude larger than boundary reflections [16] [29]; but the MJO was not sufficient on its own — an active MJO in early 1996 produced no El Niño, and the bursts that mattered were intensified locally by cold surges and twin circulations [23]. Bergman et al. conclude that the MJO is relevant to the timing and initial growth of El Niño rather than responsible for the event itself [29]. One intermediate coupled model initialised on 1 January 1997 becomes about 50% stronger when rectified MJO winds are included [30]. Statistically, MJO events warm the waveguide only when they contain a westerly wind event [35].

Compared with 1982. The 1996–97 season had 137 days with the circulation-weighted index at or above 1 against 77 in 1981–82, and the bursts in calendar 1997 outnumber those of 1982 (16 vs 13) on identical definitions [10]. In the mature phase (Oct 1997 – Mar 1998) intraseasonal OLR variance over 60°E–150°E fell to 77.8 W² m⁻⁴ (22-season mean 233), consistent with the weak eastward MJO propagation reported for the mature stage [41].

Rainfall in the usual El Niño impact regions, from GPCP v2.3 against 1991–2020. In 1997–98, 7 of the 8 regions had the usual El Niño sign. The most extreme seasons (top three since 1979) were: Central equatorial Pacific (SON, 394% of normal, 1st wettest); Coastal Peru and Ecuador (Dec–Apr, 269% of normal, 1st wettest); Indonesia (Jul–Oct, 50% of normal, 2nd driest); Equatorial East Africa (Oct–Dec, 349% of normal, 1st wettest); California (Nov–Mar, 180% of normal, 2nd wettest). The same comparison across all four events is in Paper 4.

The tropical lower troposphere (NOAA STAR TLT, 20°S–20°N) warmed to +1.12 °C above 1991–2020 by Feb 1998, the usual lag of a few months behind the sea surface.

Data. Every number in the text that comes from the monthly table, the TAO analysis or the reanalysis and MJO diagnostics was recomputed from the source files by tools/papers/verify_1997.py: 75 checks, all passing after the correction below. Diagnostic values in the text are inserted by the build script from data/papers/1997/diag_1997.json, not typed by hand.

Literature. Claims were re-read against the full texts held in the private library: McPhaden (1999), McPhaden and Yu (1999), Yu and Rienecker (1998), van Oldenborgh (preprint of 2000), Lengaigne et al. (2003), Bergman et al. (2001), Webster et al. (1999), Ueda and Matsumoto (2000), Zhu et al. (2003), Barnston et al. (1999, NCEP), Landsea and Knaff (2000), Kessler and Kleeman (2000), Vecchi and Harrison (2006), Vecchi (2006), Kutsuwada and McPhaden (2002) and Picaut et al. (2002). All other claims rest on the abstracts recorded in the reference list. Not read beyond the bibliographic record: Saji et al. (1999), Takayabu et al. (1999), Boulanger and Menkes (1999), Ham et al. (2013), McPhaden and Zhang (2002); each is cited only for what its title or a held paper states.

Corrections made.

• Start of the La Niña phase: “−0.41 (JAS 1995)” → −0.41 (JJA 1995); JAS 1995 was −0.57.
• Eastern Indian Ocean cooling: “below −2 °C in Nov 1997” → an anomaly of about 2 °C at its November maximum (the sign is garbled in the extracted text of Webster et al.; magnitude kept, wording neutral).

Checked and unchanged. The adjoint numbers (0.6, 0.3, 0.5 and 0.4 °C; 40%/50%; 30%/90%) match the preprint and the journal abstract; the 8 °C in 30 days at 0°, 125°W, the June–December monthly records and the >90 m thermocline depression match McPhaden (1999); the 15–20 cm against about 2 cm Kelvin-wave amplitudes match McPhaden and Yu (1999); the forecast statements match Barnston et al. (1999, both papers) and Landsea and Knaff (2000).

Grades. No evidence grade was changed in this pass. Open items: island sea-level records (UHSLC server unreachable from the automated fetch); an ERA5 re-test of the burst list, as for 1982.