ENSOscope

About

Methodology

How the forecasts, the ENSO classification, the skill verification, and the teleconnection composites are produced.

ENSO index & 7-class classification

For each ensemble member we compute area-weighted NINO sea-surface-temperature indices (NINO3.4, NINO3, NINO4) and the NINO3 precipitation anomaly. Each SST index is a relative index: from the box mean we subtract the tropical mean over 20°S to 20°N in the same month. That is the index of L'Heureux et al. (2024), which NOAA CPC publishes as the Relative Niño index (RONI). A warm world lifts the whole tropical ocean together, so removing the tropical mean leaves the east-to-west contrast that actually drives the teleconnections and removes the background warming with it, without fitting anything.

Each index is then deseasonalised against a 30-year rolling climatology and standardised per calendar month over 1980-2014, every dataset and every model by its own spread. Until 14 September 2026 the warming was instead removed by regressing each index on a global mean temperature curve. That estimator drifts: measured against the relative index it runs about 0.2 standard deviations low in the early 1980s and 0.3 high in the 2020s, in four independent SST products alike, so recent events read too strong and old ones too weak. It has been withdrawn.

When an event starts and ends. Detection follows NOAA CPC's own episode rule, written in standardised units. We take the centred three-month running mean of the standardised relative NINO3.4 index, which is the same overlapping "season" CPC tabulates. An episode is at least five consecutive overlapping seasons at or beyond a threshold that depends on the calendar month, t(m) = 0.5 °C divided by the month's ONI spread: from 0.44 standard deviations in December to 0.94 in May. That is NOAA's 0.5 °C test and its variance adjustment, rewritten in standard deviations rather than re-tuned; the twelve numbers are computed once on COBE-SST 2 and shared by every dataset and model, so a product with weaker variance is never handed an easier test. There is no seasonal phase-locking requirement and no bridging of gaps inside the rule.

Multi-year events. Two episodes of the same sign are treated as one event when the gap between them is three months or less and no month in that gap reaches 0.25 standard deviations on the opposite side. The inner boundaries are kept as sub-episodes, so the 1998-2001 and 2020-2023 La Niñas are single multi-year objects that still carry the two pieces NOAA lists, and a genuine swing to the other phase always ends the event however short the gap.

How strong an event is is read on the monthly standardised index inside the event window, with three consecutive months required at each level. The tiers are at or above, so the highest tier reached wins:

  • Extreme: NINO3 (El Niño) or NINO4 (La Niña) at or beyond 1.75, and for El Niño also a NINO3 precipitation anomaly ≥ 5 mm/day of observed (GPCP) rainfall; a forecast run's rainfall is first put on this observed scale, see Bias & amplitude correction below
  • Strong: otherwise, at or beyond 1.25, with no upper bound
  • Moderate: an event that clears detection but not the strong tier
  • Neutral: months inside a catalogued neutral period. The short gaps between events are transitions and count as neither, and so do the latest months the catalogue cannot classify yet (see below)

The tiers were disjoint bands until 14 September 2026, and a single month past 1.75 could break the run of three and drop a genuinely strong event to moderate. Cumulative tiers cannot do that.

The precipitation requirement is deliberate: a truly extreme El Niño is distinguished not by SST alone but by the eastward shift of deep convection into the eastern Pacific, captured by the NINO3 rainfall response.

The strip under the page header. "Now" is the latest month we have sea surface temperature for, usually last month. The strip gives that month's conditions: how warm or cool the central Pacific was that month on its own, sorted into the seven classes above with the same thresholds as the forecast. The number behind it, the relative Niño 3.4 index, shows when you hover over the scale. The second line says since which month the conditions have been on the same side (El Niño, La Niña or neutral), each month classed on its own.

An event is more than a few warm or cool months: the conditions have to last five overlapping three-month seasons in a row before a period counts as an El Niño or La Niña event. So a new El Niño can show strong El Niño conditions for months before it counts as an event. Whether it counts yet shows when you hover over the second line, and screen readers read it out.

Extreme needs rain. In an extreme El Niño the heavy rain of the western Pacific moves east, so as well as very warm water in the NINO3 region we need that region's rain to be at least 5 mm a day above normal (GPCP rain data). Rain data arrive later than sea temperature, so during an El Niño the second line gives the latest month of eastern Pacific rain next to the 5 mm a day an extreme class needs. Until a month's rain is in, that month cannot be classed extreme. The forecast line gives the class the latest ENSO forecast finds most likely, chosen the same way as on the Forecast page.

Where the observed record stops. The relative index needs nothing but sea surface temperature, so the observed NINO indices now run to the latest month of SST on file. The NINO3 rainfall anomaly is only read inside an event window, and every rainfall-based product on the site (the maps, the country series, the extreme El Niño gate) stops at the end of its own input, which is usually earlier. The two ends are therefore different on purpose, and each panel is labelled with the month it actually reaches. Classification itself stops one month before the SST does, because the centred three-month season of the last month needs a month that does not exist yet.

Bias & amplitude correction: no statistical calibration

Each member's index is corrected to the 1993-2016 common hindcast period with a variance-rescaling bias correction (after Merryfield & Lee): the hindcast mean is removed and the variance is rescaled to match observations,

z = (anomaly − μhindcast) / σhindcast  for each start month and lead time.

This removes both the mean bias and the model's ENSO amplitude bias: seasonal models typically over-amplify ENSO by ~40% (the cold-tongue bias). Probabilities are then the raw fraction of ensemble members in each class. No Platt or isotonic calibration is applied: the spread of the bias-corrected ensemble is the probability.

The SST thresholds (0.5, 1.25, 1.75) are in standard deviations of the observations, so z can be compared with them directly. The extreme El Niño rainfall threshold is in mm/day of observed rainfall, so each run's NINO3 rainfall z is first put back on the observed scale of the month it forecasts, rain = μobs(m) + z × σobs(m), with the mean and standard deviation of observed GPCP rainfall for that calendar month over 1980-2014, the baseline of the observed SST indices. Only then is the ≥ 5 mm/day test applied. The scale is fixed in advance and stored with the code, and the hindcast evaluation uses the same scale, as it uses the same observed standard deviations for the SST indices.

On the 1993-2016 hindcast the extreme classes come out close to their observed frequencies: extreme El Niño in 2.8% of SEAS5 and 2.3% of Météo-France run-months against 2.1% of observed months; extreme La Niña in 6.9% and 6.4% against 8.3%. A calibration layer would obscure that.

Skill verification

Skill is shown directly as the temporal correlation between the hindcast ensemble-mean NINO index and observations, as a function of start month and lead time, for each NINO region including NINO3 precipitation, the standard anomaly-correlation metric used in seasonal-forecast verification. Correlation falls at longer leads and across the boreal-spring predictability barrier. See the Skill tab.

Teleconnection composites

The Teleconnections maps composite climate-impact indicators (heat: Tmax, UTCI, WBGT, and daytime, nighttime and compound heatwave days; precipitation extremes: Rx10day, consecutive dry days) by ENSO phase, from observed rainfall (CHIRPS), heat from the ERA5 reanalysis, and the CESM2 large ensemble, a climate model run 50 times. The ENSO indices used to classify each event are relative indices (the box mean minus the concurrent 20°S to 20°N tropical mean), deseasonalised with a 30-year rolling window and standardised by the ENSO standard deviation over 1980-2014, where the threshold is expressed in standard deviations. The composited fields themselves are left alone: taking the tropical mean out belongs to the classification, so a composite shows the conditions actually observed or simulated during those events, warming included. Anomalies are shown relative to the neutral-season average: the average of past neutral seasons, chosen by the same full-season rule as the El Niño and La Niña seasons, so like is compared with like (one season per neutral period, 11 or 12 in the observations). The Event replay and the country packs use normal instead, every neutral month in the record, so their numbers differ slightly.

The country panel under the map. Picking a country opens a table with one row per month of the ENSO forecast (L1, the month the forecast is issued, to L6). For each forecast model it names the most likely class by the rule of the Forecast page: the family with the highest chance (El Niño, neutral or La Niña), then its most likely strength. Beside it is what past events of that class did to the heaviest 10-day rainfall (or, on the dry tab, the longest dry spell) in the season that month falls in: the composite anomaly averaged over the map cells whose centres lie inside the country, each weighted by its area (the cosine of its latitude). The value is one number per season: the heaviest 10-day total (or the longest dry spell) anywhere in the three months, crossing month ends, one season per event (the one the event covers most), minus the neutral-season average. The Event replay shows the same index inside each calendar month against normal, so a season there reads smaller and can differ in sign; for an event that starts late in the year it also draws the first year's months, where the map takes the next year's season (2014-16: SON 2015, DJF 2015-16), and the replay page names those seasons under the charts. A change under 0.5 mm or 0.5 days either way, as printed, reads "no clear signal". The number of past events is shown with the season each one gave and how many of them went the way of the average ("n = 4: 1986, 1991, 2015, 2023; heavier in 2 of 4"), because one or two events make a weak guide and an average can rest on one event. The map tiles store the composite as computed; the colour scale saturates at its ends, the numbers do not. The panel does not average a whole region, which can hold opposite changes, and it is not a forecast of rain.

How the heatwave days are defined. Following the daytime, nighttime and compound framing of Zhang et al. (2025), a daytime heatwave day is one whose daily maximum temperature exceeds the local calendar-day 90th percentile (TX90p); a nighttime heatwave day is one whose daily minimum exceeds the 90th percentile (TN90p); and a compound heatwave day is hot by both measures at once. The percentiles use a recent multi-decade baseline (1981-2020 for the observations, 1980-2014 for the CESM2 historical ensemble) with a 5-day calendar-window smoother, so a heatwave day is defined relative to what is locally and seasonally normal rather than a fixed temperature. Counts are summed over the season each event contributes and then composited by ENSO phase.

Which season each event contributes. One rule, applied to the observations and to CESM2, to heat and to rainfall alike. An event, or a neutral period, enters a seasonal composite only if it spans every month of that season, from the first day of its first month to the last day of its last. It contributes one season even when it lasted several years. And that season must have at least 90 per cent of its days in the data record, which is why the 2025-26 La Niña is not yet in the December to February maps: the observed record used here ends in December 2025.

Because the teleconnection composites use a long observational/ensemble baseline, they use a different normalization from the forecasts (which are constrained to the 1993-2016 hindcast period).

Absolute heat-stress thresholds

Alongside the percentile-based heatwave days, the heat indicators are also counted against fixed temperature thresholds that carry a direct health meaning, so a number can be read as "days in the 32 to 38 °C band" rather than "days above the local 90th percentile". These were added at the request of the humanitarian teams.

  • UTCI: days in three bands: 32 to 38 °C (strong heat stress), 38 to 46 °C (very strong) and 46 °C and above (extreme), following the standard UTCI thermal-stress scale (Bröde et al. 2012, Table 3).
  • WBGT: days in 29 to 31 °C, 31 to 32 °C and 32 °C and above, approximately the US National Weather Service category edges (29.4, 31.1 and 32.2 °C).
  • Nighttime heat: nights whose minimum is 20 to 25 °C (tropical night), 25 to 30 °C (equatorial night) or 30 °C and above (torrid night).

The bands do not overlap: each day lands in exactly one, and the three add up to the days at or above the lowest cut-off. "Heat-stress days" in the country packs are that sum, days with UTCI 32 °C or more.

As absolute counts these come from the ERA5 reanalysis only, and that is deliberate. A percentile threshold is defined from the field's own distribution, so a uniform warm or cold bias cancels out: a model running 2 °C warm still has the same number of days above its own 90th percentile. An absolute threshold has no such protection, because it reads the actual value. Computed on ERA5, a 1 °C warm bias adds roughly 6 days per season over the Horn of Africa, 7 over South Asia and 10 over Southeast Asia, which is as large as the El Niño signal itself. From the CESM2 climate model the fixed-threshold layers are therefore offered only as the change against neutral seasons (how many days El Niño adds), where the model's bias is largely shared by both composites, though it does not cancel exactly as it does for a mean. The percentile-based layers remain the safer choice for the model.

One further caution on WBGT. It is computed here with a simplified formula that uses temperature and humidity only, assuming light winds and fairly sunny conditions, and so cannot respond to the actual radiation or wind. A seasonal mean tolerates that approximation well; a count of threshold crossings tolerates it less well, because a count is a step function at the cut-off. UTCI, computed with the full Bröde polynomial using actual radiation and wind, is the more reliable of the two for this purpose.

Event replay

The teleconnection maps answer "what does an El Niño typically do", by averaging every past event of a class into one seasonal composite. The Event replay page answers the opposite question, for one named event, and nothing on it is averaged into seasons.

The window. Each event is replayed from January of the year it began through December of the year it ended. That places the peak near the middle and shows the build-up and the decay either side: the 1997-98 El Niño runs January 1997 to December 1998. It also handles multi-year events without special-casing, so the 2020-2023 triple-dip La Niña is one 48-month window rather than three clipped pieces. Events shorter than 24 months are padded by extending the end, never the start, which would push the onset off the edge.

The comparison. The faint lines behind each panel are the other observed events of the same type, aligned so that month 0 is January of each one's own first year. Without a common origin, events beginning in different calendar months could not be read on one axis.

The country series. Monthly, area-weighted over the country at native 0.25°, from CHIRPS rainfall and ERA5 heat. Only fixed thresholds are used here: a percentile index would need a 0.25° percentile climatology, and dropping to 1° would leave small countries with two or three grid cells. Note that the seasonal maps show the longest dry run (CDD), while this page counts days below 1 mm, because a longest run inside one calendar month is ambiguous when spells cross month boundaries. They are different quantities and are labelled differently.

Kiribati covers the Gilbert Islands only. Kiribati is a chain of atolls spread across the date line, each far smaller than one 0.25° grid cell. ENSOscope covers the Gilbert Islands (168° E to 178° E), Tarawa included, where most people live; the Phoenix and Line Islands are left out, and CHIRPS has no rainfall cells at Kiritimati in any case. No cell centre falls on an atoll, so the country is the average of the cells that overlap the islands, each weighted by how much of it the islands cover: three CHIRPS cells for rainfall, sixteen ERA5 cells for heat. Those ERA5 cells are mostly sea, so the heat figures follow the air over the surrounding ocean and read lower than a thermometer on land would. The country zoom shows these cells at 0.25° rather than 1°.

Normal. The dashed grey line on each panel is normal, what an ordinary year looks like in that country: the average of the catalogued neutral years for the same calendar month, between 15 and 22 years per calendar month since 1981. Comparing an event only against other events answers "was this El Niño worse than the last one"; it cannot answer "was it worse than a normal year", which is the question that decides whether to pre-position anything. The same reference defines every anomaly on the page. A conventional thirty-year normal is deliberately not used: such a window contains El Niños and La Niñas, so an event measured against one is partly measured against itself. Months belonging to no catalogued period at all are ENSO transitions and count as neither.

The most recent months are not counted as neutral. Classification needs five consecutive seasons, so a run that has not yet lasted that long belongs to no event and would otherwise be recorded as neutral by default. In the middle of the record that is safe, because we can see what followed. At the end of the record it is not: the record simply stops, and the run may still be building. Treating those months as neutral would state that conditions are normal when what we actually have is no classification yet, and it would fold them into the very definition of an ordinary year. So the trailing unclassified period is excluded from the reference and is not labelled anywhere on the site. As of August 2026 that period is March to August 2026, which contains an August at +2.4 standard deviations: unambiguously not an ordinary month, and exactly the kind of value that must not be allowed to define what ordinary means.

Past El Niño seasons in the downloads. The Full analysis and the One-page summary say how rain went in past El Niño seasons over the coming three and six months, for the country and each region, as % of normal. Three levels, the same in every language. Drier or wetter: at least 7 in 10 seasons on that side of normal and a median below 95% or above 105% of normal. Tends drier or tends wetter: 6 in 10 seasons (60 to 69%) on that side and a median beyond the same limits; Honduras, November to January, was drier in 9 of 14 seasons with a median of 89% of normal. Otherwise no clear signal, with the count of seasons on the side the median falls on. When 7 in 10 seasons or more went one way but the median stayed between 95% and 105%, the sheet says the change was slight. Counts are of seasons, not events. Both downloads count the same seasons: each past event's window starts as many months after January of its onset year as the forecast's starts after January of the current event's onset year, and it counts when more than half of its months fall inside the event. The same window a year later counts too while more than half of it stays inside the event, so an event that lasts into a second or third year can give two or three seasons. Until 18 November 2026 the One-page summary counted every calendar season most of whose months were El Niño months; for December to May that took in December 1992 to May 1993, which the Full analysis did not count.

Thresholds. Heat stress and warm nights each carry all three health bands, selectable above the panel: UTCI 32-38, 38-46 and 46 °C and above, and nights of 20-25, 25-30 and 30 °C and above. The bands are disjoint: each day falls in exactly one, so the three add up rather than nesting. They were cumulative until 2026-08-25, which meant a day at UTCI 47 was counted in all three and the series could not be compared or summed. The extreme end is a different question from the mild end rather than a scaled version of it, since a country can gain a fortnight of UTCI 32-38 days in an El Niño year and not a single day above 46. Where a country never reaches a cut-off anywhere in the 1981-2025 record the button says so, because "never happens here" is information and a blank panel is not. The map underneath follows the selected cut-off when it is already showing that family.

Comparing two events in space. Below the world map, the selected country is drawn twice side by side: the chosen event on the left, any other event of the same family on the right, on one shared colour scale. Independent scales would show the same anomaly in different colours and quietly invert the comparison the panels exist to make. A month slider steps both panels together through the same offset from each event's own January, so "November of year 0" is November of 1997 on one side and November of 2023 on the other. This matters because the difference between two seasons is often not how much fell but when: a window average cannot separate a season that started early from one that caught up late. Values are clipped to the country and its border drawn heavy, with neighbours outlined for orientation only. The colour range is set from the two months on screen rather than from the whole record, because on an extreme month a fixed range saturates both panels into identical blocks and destroys the comparison; the bar under each panel carries the numbers, and rescales with the slider.

The axis. Months are named, on two lines, with the relative year under each January. The year is relative rather than absolute because the comparison lines are different events from different decades sharing one axis; the selected event's real years are in the axis title and in every tooltip. Labels are never rotated: a rotated tick label is anchored at its own centre, so it drifts away from the tick it belongs to, and on a 48-month axis that drift is larger than the spacing between months.

Resolution. The country time series are computed at the source resolution, 0.25°, from CHIRPS and ERA5. The maps are 1°. That is a hosting limit, not a scientific one, and it costs nothing where it applies: the world map is about two pixels per degree, so a 0.25° cell there would be half a pixel. It does cost something on the country zoom, where a 1° cell is a visible block, and a 0.25° version of that layer is built and waiting on somewhere large enough to serve it. Note the distinction: the gridded 0.25° map is not yet served, but the sub-national numbers below are computed from that same 0.25° store, so the regional figures already carry the full source resolution even while the picture above them is drawn at 1°.

Sub-national regions. Clicking a province on either country panel swaps every chart below it to that province. The regional figures are the mean of the 0.25° cells whose centre falls inside the admin-1 boundary, taken from the same store the maps are drawn from, so the table and the map above it can never disagree. Charts are shown in absolute units against the same neutral-year reference the country view uses: the store holds anomalies, and the neutral mean per calendar month is carried alongside them so the absolute value can be recovered exactly rather than approximated. Every admin-1 unit is listed. A unit holding fewer than three cell centres (a capital district, a small province) is not a plain average of one or two pixels, which would be noise: it is averaged over every cell whose 0.25° box overlaps it, weighted by the overlap, and its row says "part" in the cells column. A small island that no cell box overlaps takes the nearest cell within half a degree ("nearest"). A unit no data cell reaches, such as northern Germany for rainfall (CHIRPS stops at 50° N) or a remote island, is shown as "none" rather than left out. A unit that reaches past 50° and has rain on under 90% of its cells shows its rain with the count, such as Santa Cruz in Argentina, rain on 353 of its 477 cells, and the downloads list it as partly covered. A unit with rain on under a quarter of its cells shows no rain value anywhere, only "not covered by CHIRPS (rain on 30 of 235 cells)" for Magallanes in Chile; Hessen in Germany and Nord in France are the other two. Their heat values stay. A country's rain is the average of every cell with rain; where that is not every cell, the downloads say how many it is.

The ± beside a regional figure is a spread, not an error bar. It is the standard deviation of the cells inside that region for that month, so it answers whether the region moved as one, and it is not an uncertainty on the mean. The two imply opposite things about confidence and are easily confused. Where the spread is larger than the mean, part of the region went the other way and the average is hiding a split rather than summarising a signal.

Boundaries are Natural Earth admin-1. For some countries these are the older administrative divisions: DR Congo appears as its eleven pre-2015 provinces rather than the current 26, and Kenya as its former provinces rather than its 47 counties. This is adequate for regional comparison and inadequate for matching a specific current administrative unit; where that matters the OCHA Common Operational Datasets are the right source, and are what humanitarian reporting aligns to. A few Natural Earth features carry the wrong name and are corrected by location: Malawi's Karonga, which the file calls a second Chitipa, Afghanistan's Daykundi and Panjshir, Uganda's Kiruhura, Wirral in England, and Maputo, Cebu, Iloilo and Cotabato cities, named like the provinces around them.

The map. Each event's map is its own, not a phase composite: the field is aggregated over the replay window and expressed against the neutral-year normal for the same calendar months. Matching the months matters, because a 24-month window contains two of every month and a 36-month one does not, so comparing against an annual mean would alias the seasonal cycle into the anomaly and make long events look wet or dry for no physical reason.

Forecasts on this page. An event still running also carries the live ECMWF SEAS5 and Météo-France System 9 sea-surface-temperature forecast, shown dashed. Where the selected event is still running, the rainfall and temperature charts also carry the regional forecast: ECMWF SEAS5 and Météo-France System 9, corrected day by day against CHIRPS rainfall and the ERA5 reanalysis, dashed: the median of the runs and the middle 80% of them, drawn only in months where that model's past forecasts show skill for that region (see Regional forecast). Nothing is derived from the SST forecast itself, which would be a considerably larger claim than that forecast supports.

Regional forecast

For every admin-1 region, every country and the zones partners use, the regional forecast gives rainfall, the heaviest 10-day rainfall, the longest dry spell, dry days, maximum temperature, heat-stress days and warm nights for each lead month and for the next 3 and next 6 months (the months after the forecast month: November to January and November to April for the October forecast). It is built from the daily fields of ECMWF SEAS5 and Météo-France System 9 from the Copernicus Climate Change Service, because every index is a daily statistic (dry days under 1 mm, dry spells, heat-stress bands).

Correction. On a 1 degree grid, each model's daily values are mapped onto the daily values on the same cells, lead month by lead month, by quantile mapping fitted on its 33 past seasons (1993-2016 past forecasts and 2017-2025 real-time forecasts, 25 ensemble members a year): CHIRPS for rain, ERA5 reanalysis for temperature and UTCI. The indices are then computed from the corrected days with the same reductions as the observed series, and averaged over each region with overlap weights.

Normal. A model is compared with its own climate first: the mean of its past forecasts for the seasons that were ENSO neutral (most months inside a confirmed neutral period of the catalogue). The forecast is then put on the site's normal (neutral years since 1982) with the observations alone: rain times the ratio of the observed neutral mean over the model's years to the observed normal, other indices plus the difference of the two. Rain is shown as % of normal, the rest as the difference from normal, from the average of the runs.

Thirds and leans. The chances are the share of the forecast's runs (51) in the driest, middle and wettest thirds of the model's 33 past seasons, El Niño and La Niña years included (25 ensemble members a year, pooled). A share does not depend on the number of runs, though its resolution and sampling noise do. A model leans one way only when at least half its runs fall in one third and its past forecasts picked out that third better than chance: the ROC area of that third, on the leave-one-season-out past forecasts described below, must pass a one-sided 5% Mann-Whitney test (normal approximation, about 0.68 for 11 events in 33 seasons).

Skill. Skill is checked against the region series this site shows (CHIRPS, ERA5) for the same months, on past forecasts corrected with each season left out of its own fit (leave-one-season-out), so that no season helps correct itself; the ROC test and the El Niño counts below use the same past forecasts, while the forecast and its thirds use the fit on all 33 seasons. Skill is the same measure as on the Skill page: the correlation between the average of the past forecasts and the observations. It is shown when it reaches the one-sided 5% limit (n - 2 degrees of freedom: 0.29 for 33 seasons). In the tables a model's cell says "weak skill" above that limit but below 0.6, and "no skill shown" under the limit, with no forecast described; a cell with neither mark has useful skill (0.6 or more, the Skill page's line). The maps show the forecast in every cell and leave skill to the tables; dots mark where the two models disagree on the sign. About one region in twenty passes by chance alone. For temperature, part of the skill can come from the warming trend shared by the forecasts and the observations: where the correlation no longer passes once a linear trend is removed from both (n - 3 degrees of freedom), the text says that much of this skill comes from the warming trend. For each region and model the detailed workbook also counts, in the El Niño seasons of 1993-2025, how often the model leaned the right way, the wrong way or not at all.

What is not said. Rain is not described in windows that normally bring under 8% of a region's yearly rain, or, for a single month, a month bringing under 2.7% of the year's rain: the 8% three-month rule spread over one month. Where at least 9 in 10 runs of each model are past the wettest or driest (or warmest) season since 1982, every cell that shows a number says "beyond the record since 1982" with the record and its season. Where more than half of a season's forecast rain comes from days beyond the range the correction was fitted on, the map's cell detail says "beyond the model's past range; treat as indicative". Where the runs reach past the warmest (or wettest, driest) season since 1982, the text says how many do (more than three in four, more than nine in ten) and names that season and its value. Temperatures are compared with past neutral years since 1982, so part of any warm reading, observed or forecast, is the warming since then and not the current El Niño; the forecast values include this warming, and so does the skill measure. On a 1 degree grid a region smaller than a cell borrows from its neighbours, and day counts run a little different from the 0.25 degree counts of the record, which is why only ratios and differences are compared across columns. An index is compared with this season's column only where it passes a like-for-like check, region by region: over 1982-2025 the record's seasons on the forecast's 1 degree cells and the series this site shows must correlate at 0.9 or more, and the site's year-to-year spread must be 0.8 to 1.25 times the 1 degree spread. Where the check fails (for example warm nights in the Philippines), the tables say "not comparable with this season's column" and describe no forecast. Where no observation cell covers a region, or the record holds no values there (rainfall north of 50 degrees), past forecasts cannot be checked and no forecast is shown.

Data sources and references

ENSOscope builds on the datasets below. Each keeps its own licence and terms. Please cite the original sources if you reuse this material.

Table 1. The datasets and methods ENSOscope builds on, what each is used for, and how to cite it.

Dataset or methodUse and citationPeriodResolutionLicence
ECMWF SEAS5(System 51): seasonal forecasts and hindcasts. Johnson et al. (2019), Geosci. Model Dev. 12, 1087-1117. doi:10.5194/gmd-12-1087-2019. Accessed via the Copernicus Climate Data Store (C3S).see sourcesee sourceCC BY 4.0 (C3S, since 2 July 2025)
Météo-France System 9seasonal forecasts and hindcasts, via the Copernicus C3S multi-system seasonal service.see sourcesee sourceCC BY 4.0 (C3S, since 2 July 2025)
ERA5 reanalysisbasis for the heat-stress indices (WBGT, UTCI), maximum temperature and warm nights. Daily maximum and minimum temperature are the highest and lowest of ERA5's hourly maximum and minimum 2 m temperature over the whole day, 00 to 24 UTC. Hersbach et al. (2020), Q. J. R. Meteorol. Soc. 146, 1999-2049. doi:10.1002/qj.3803.see sourcesee sourceCC BY 4.0
CHIRPS v2.0daily precipitation (1981-present) for the rainfall teleconnection composites. Funk et al. (2015), Sci. Data 2, 150066. doi:10.1038/sdata.2015.66.1981-presentsee sourcePublic domain
CESM2 Large Ensemble (LENS2)~50 members for the model teleconnection composites. Rodgers et al. (2021), Earth Syst. Dynam. 12, 1393-1411. doi:10.5194/esd-12-1393-2021.see sourcesee sourceCC BY 4.0
COBE-SST 2observed SST for the ENSO indices. Hirahara et al. (2014), J. Climate 27, 57-75. doi:10.1175/JCLI-D-12-00837.1. Data provided by the NOAA PSL, Boulder, Colorado, USA, from psl.noaa.gov; months from November 2025 come directly from its producer, the Japan Meteorological Agency (JMA).see sourcesee sourceNo usage restrictions (NOAA PSL); JMA terms compatible with CC BY 4.0
GPCP(Global Precipitation Climatology Project, Monthly Analysis Product): observed precipitation for the ENSO precipitation index. Adler et al. (2003), J. Hydrometeorol. 4, 1147-1167. doi:10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2. Data provided by the NOAA PSL, Boulder, Colorado, USA, from psl.noaa.gov; months from January 2026 come from NOAA NCEI (GPCP version 2.3).see sourcesee sourceNo usage restrictions (NOAA PSL, NCEI)
Relative Niño indexthe definition used to classify every event here. L'Heureux et al. (2024), A relative sea surface temperature index for classifying ENSO events in a changing climate, J. Climate 37, 1197-1211. See also NOAA CPC's RONI page, from which the episode rule and its 0.5 °C variance-adjusted threshold are taken.see sourcesee sourcePublic domain (NOAA NWS)
Berkeley Earthglobal mean temperature. No longer used to classify events; kept for the archived runs that predate 14 September 2026. Rohde & Hausfather (2020), Earth Syst. Sci. Data 12, 3469-3479. doi:10.5194/essd-12-3469-2020.see sourcesee sourceCC BY-NC 4.0
UTCIheat-stress index: Bröde et al. (2012), Int. J. Biometeorol. 56, 481-494. doi:10.1007/s00484-011-0454-1.see sourcesee sourcesee source
Heatwave-day framework(daytime / nighttime / compound): Zhang et al. (2025), J. Geophys. Res. Atmos. doi:10.1029/2024JD042446.see sourcesee sourcesee source
UTCI heat-stress thresholds(32 / 38 / 46 °C) applied to population exposure: Emerton et al. (2026), Global heat stress intensification and its expanding footprint on the human population, Nature Climate Change. doi:10.1038/s41558-026-02670-5.see sourcesee sourcesee source
WBGT heat-stress categoriesUS National Weather Service, and the ArcGIS Living Atlas WBGT layer.see sourcesee sourcesee source
Simplified WBGT formulaAustralian Bureau of Meteorology approximation, bom.gov.au/info/thermal_stress.see sourcesee sourcesee source
Tropical / equatorial / torrid nights(20 / 25 / 30 °C): Correa et al. (2024), Int. J. Climatology. doi:10.1002/joc.8510.see sourcesee sourcesee source
Country and coastline boundariesNatural Earth (public domain). Made with Natural Earth. Free vector and raster map data @ naturalearthdata.com.see sourcesee sourcePublic domain (Natural Earth)

Licences as each provider states them, from the project's data licence check of 14 October 2026. "See source": no value checked by the project yet; the dataset's own page or paper, cited in the table, has it.

Seasonal calendar (event replay)

Under the rainfall chart, a thin strip shows a typical year for the selected region: its rainy season and, where a source gives them, planting, harvest and the lean season. Rainy months come from CHIRPS rainfall (observed): a month counts when it holds 10% or more of the year's rain, averaged over the whole years of the served series, and rainy months that follow each other make one season. Planting, harvest and the lean season come from FEWS NET's seasonal calendars. A region takes the calendar of the FEWS NET zone it lies in on FEWS NET's locator map, or the country's own calendar where FEWS NET publishes one for the whole country or for its rural households; in Tanzania, regions are assigned to the zone with the same rain pattern. In four countries without a FEWS NET calendar, FAO's Crop Calendar gives planting and harvest for the region's staple crops. It is used under FAO's terms (non-commercial use, with acknowledgement) and is not covered by ENSOscope's CC BY 4.0. Every cell of the strip traces to a row of the source ledger, data/calendars/sources_ledger.csv.

Disasters and outbreaks (event replay timeline)

Required credits

Contains modified Copernicus Climate Change Service information [2026]. Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains. The seasonal forecasts come from the C3S multi-system seasonal forecast (doi:10.24381/cds.181d637e). ERA5 is used through the NSF NCAR copy (doi:10.5065/BH6N-5N20). COBE-SST2: Source, Japan Meteorological Agency, edited for ENSOscope. CESM2: we acknowledge the CESM2 Large Ensemble Community Project and supercomputing resources provided by the IBS Center for Climate Physics in South Korea.