Some monsters are real

The satellite picture needs no statistics. Along the equator east of the dateline the water is three, four and in places five degrees warmer than the 1991–2020 average for the date, and the warm tongue reaches the coast of South America. NOAA refreshes this map every afternoon; what you see is the latest one it has published.

A Hovmöller diagram trades the map’s second dimension for time. Each column is one week; each row is one degree of longitude along the equator, from Indonesia at the top to the coast of Ecuador at the bottom, with the sea-surface temperature anomaly averaged across the band from 5°S to 5°N. Read it left to right and the year’s story is plain: a weak La Niña through the northern winter, then warm water appearing off South America in March, spreading west and deepening through the summer until it fills the basin east of the dateline. The brackets at the right mark the regions forecasters average over; the thicker one is Niño 3.4, the box every other chart on this page measures.

The same data as four cross-sections. In early April the equator was close to normal almost everywhere; by August the anomaly was above +2°C across the whole eastern half of the basin; in the latest week the warmest water, more than +4°C above normal, sits against South America, and the Niño 3.4 average is +3.3°C. The heat is concentrated in the east, the signature of a classic eastern-Pacific El Niño rather than a central-Pacific one.

Start with the raw thermometer. Every gray line is one year since 1982, the ochre line is the 1991–2020 average, and the red line is this year. The region has a season: it warms to a peak near 28°C in April and May as the trade winds slacken, then cools through the autumn. El Niño years break the pattern by refusing to cool. In 1997 and 2015 the water held near 29°C into November. In 2026 it passed 29°C in June and reached 29.98°C on October 1, warmer than any day in the record, with the months that normally bring the peak anomaly still ahead.

Subtracting the average for each calendar day flattens the seasonal hump and leaves the anomaly, which is what forecasters quote. Here the past El Niños stand up clearly as the handful of gray lines that climb through the autumn to peaks of +2.4 to +3.0°C in November and December. The 2026 line reached +3.3°C on October 1, two months before the season when those events peaked.

But an anomaly of one degree does not mean the same thing in May as in December. The Niño 3.4 region is far more variable in the northern winter, when El Niño and La Niña events mature, than in the late spring, when the ocean sits close to its average almost every year. The ribbon below is that variability: one standard deviation of the daily values across the thirty years of the baseline.

Divide each day’s anomaly by that day’s standard deviation and you get Jacobson’s chart, the one the Australian climate writer Gregory Andrews said should be front-page news. It answers a sharper question than the anomaly does: not how warm, but how unusual for the date. Before this year, no day since 1982 had been more than about 2.7 standard deviations from normal. In 2026 the region crossed three in June and has stayed there, peaking near 3.9σ in mid-September. For a normal distribution, a 3σ day is roughly a 1-in-740 event; 3.9σ is about 1-in-20,000. The real distribution has fatter tails than that, which is why Jacobson’s 1-in-7,000 is a rhetorical figure as much as a statistical one. But the comparison that matters is with the other lines on the chart, and none of them comes close.

Each event is the season from May of its first year through April of the next. Anomaly is the departure from the 1991–2020 average for that calendar day; σ is that anomaly divided by the standard deviation of the thirty baseline years on the same day. The last two columns measure persistence as well as height: the number of days the anomaly stayed at or above the +0.5°C El Niño threshold, and the anomaly above that threshold summed over those days. 2026 is still in progress, with the months that usually hold the peak still to come.

The chart at the top of the page is Fig. 5 and Fig. 8 drawn as one. Its vertical axis is a thermometer, in plain degrees. Its curved grid is the standard-deviation scale from Jacobson’s chart, bent to fit on that thermometer. It takes three steps to build.

Read the curve, not the gap

On ordinary graph paper, the distance between two gridlines means the same thing everywhere. On this paper it doesn’t, and that is the point. One degree above normal in May nearly reaches the +2σ curve, at +1.8σ. One degree above normal in December gets only to +0.9σ. On the thermometer they are the same gap, but the May day is far stranger.

That gives the chart three readings, one for each thing your eye does with a line.

1. Height is temperature. Read the left axis for degrees, as on any thermometer chart.
2. The curve a line sits on is how unusual it is. Count the curves between the line and the ochre average. A line between the +2σ and +3σ curves is two to three typical swings above normal for that date.
3. Crossing curves means the departure is getting rarer. A line that runs parallel to a curve is equally unusual day after day. A line that climbs across the curves is becoming more extreme, even when the thermometer barely moves. From May 1 to June 30, 2026, the water warmed only 0.4°C, but its line climbed from 1.7σ to 3.3σ, because those are the months when the region normally holds still.

Everything else on the chart

Why bend the grid

Each of the usual ways to chart this data answers one question well and drops another.

The thermometer is worth keeping. Over the tropical Pacific, deep thunderstorms build where the sea surface is warmer than about 28°C. When El Niño spreads that warmth east, the rain moves with it, and that is how one box of ocean changes the weather on other continents. A chart in standard deviations can say the water is strange. Only the degrees say it is past the temperature where the atmosphere starts to respond. Degrees are also the unit that forecasts, news reports and readers already use.

Rarity is worth keeping too, because the region’s normal variability more than doubles between late spring and midwinter. Without it, a warm May and a warm December look alike. You could show both charts side by side, as this page does further up, but then the reader has to match dates across two panels. Fil’s version, after an idea from Toph Tucker, leaves the data alone and bends the paper, so one line carries both readings.

Two caveats. The grid is built from a 15-day smoothed average and standard deviation, so the curves don’t wobble with the noise of a 30-year sample. Fig. 8 uses the exact daily values, so a line can sit a hair off from where Fig. 8 puts it. And standard deviations describe typical swings, not odds. The region’s extremes have fatter tails than a bell curve, so read +3σ as three typical swings above normal, not as a precise probability.

El Niño is classified on a slower clock than the daily charts above. NOAA’s Oceanic Niño Index is the three-month running mean of the Niño 3.4 anomaly, and its informal ladder runs from +0.5°C, the threshold for El Niño itself, through +1.5°C for a “strong” event to +2.0°C for a “very strong” one. Only three events since 1950 have cleared the top rung: 1982–83, 1997–98 and 2015–16, which peaked at +2.59°C. The index for July–September 2026 is +2.16°C, already past that line in the season when most El Niños are still gathering themselves. The chart below lays every event since 1950 on the same two-year calendar, from January of the year it formed to December of the year after, so the shape of this one can be compared with all the rest. (A peak-aligned version would be the more common way to rank events, but an event that has not peaked cannot be aligned on its peak; the calendar is the honest axis while it is still growing.)

El Niño anomalies typically peak between November and January. On September 10, NOAA’s Climate Prediction Center kept its El Niño Advisory in place, with subsurface water in the eastern Pacific running more than 10°C above normal at depth, and gave a better than 90 percent chance of a very strong event this winter and a 75 percent chance of a historic one, stronger by its seasonal index than any El Niño since 1950. On October 1, Columbia University’s climate school reported that all 22 models in its monthly plume call for a very strong event through February, most of them above +3°C for the season, against the previous record seasonal peak of +2.59°C. Jacobson’s July post quoted NOAA’s CFSv2 model near +4.2°C and ECMWF near +3.9°C for the peak. Those are forecasts, not observations, so they appear on the chart only as a labeled bracket.

Godzilla is going to return, and return, and return… we won’t be able to walk out of the theater.

Eliot Jacobson, The Climate Casino, July 8, 2026