How ENSO reads India's rain
One question, counted the same way for every district: in the years the Pacific ran a strong El Niño, did the season's rain fall short of what a neutral year would have brought? Maplore makes no forecast of its own; the Pacific forecast on the map is the forecast models' of NOAA and of Europe's Copernicus service, shown as they run.
Strong El Niño years
8since 1950
1957, 1965, 1972, 1982, 1991, 1997, 2009, 2015. November to January relative index at or above 1.5 °C.
Neutral years
26the benchmark
Years whose November to January index stayed within half a degree of zero.
Rain
0.25°IMD gridded daily rain, 1950 to 2025
A district's rain is the mean of the grid cells whose centres fall inside it; 33 small districts take the nearest cell.
The Pacific now
1.7 °CONI, JAS 2026
The index for the latest three-month season NOAA has published.
All India, event by event
percent of a typical neutral year; under 100 is drier
The monsoon of the El Niño year
8 of 8drier
244 of 779 districts were drier in more than three quarters of the events; 188 in half or fewer.
October to December of the El Niño year
6 of 8drier
85 of 697 districts were drier in more than three quarters of the events; 303 in half or fewer.
The monsoon of the year after
2 of 8drier
22 of 779 districts were drier in more than three quarters of the events; 583 in half or fewer.
La Niña, and the test against chance
the mirror reading, and how far either count can be trusted
Strong La Niña years
6since 1950
1973, 1988, 1998, 1999, 2007, 2010. November to January index at −1.5 °C or lower. Same rain, same neutral-year line; the count is of seasons wetter than it.
Against its own other years
246 · 87districts that stand apart: El Niño · La Niña monsoon
For each district: how often it is on that side of the line in every other year, then the odds of its event count arising from those odds. Chance alone gives up to 78 of 779.
What that licenses
0districts that hold alone, either phase
With 8 and 6 events no single district can be told from luck. The map shows where the count leans; the pattern is in the group.
Which index
On the classic index the strong El Niño years would be 1957, 1965, 1972, 1982, 1997, 2009, 2015, 2023 and the strong La Niña years 1955, 1973, 1975, 1988, 1999, 2007. Both sets are counted each week; the Guide puts the two side by side.
The Pacific picture
the map's opening view: the ocean this week
What is drawn
7days, averaged
Sea-surface temperature minus the normal for the date, 17 Oct to 23 Oct 2026. Satellites, ships, buoys and floats, blended by NOAA on a quarter-degree grid.
Normal
1971–2000the product's own base years
NOAA's three-month index uses a later, moving base, so the two numbers differ by a few tenths of a degree.
The box
+3.5 °Cthis week, inside it
5° north to 5° south, 170° to 120° west: the stretch of ocean the index is read in. The figure is the area mean of the picture inside the box.
Kept honest
7 of 7days still preliminary
NOAA revises each day within two weeks; the picture is redrawn daily. It is not published if fewer than five of the last seven days arrived, or the newest is over six days old.
The forecast
where the weekly line is headed, by NOAA's and Copernicus's models
What is drawn
624runs of 13 models
6 from NOAA's multi-model ensemble and 7 from Copernicus, all started October 2026. Each model counts once. Dashed: their mean. Band: the range of the models' means, drawn to Mar 2027.
The peak
+3.9 °CNovember to January
Model means run from +3.1 to +4.7 °C, against the 1991 to 2020 normal. NOAA's models alone: +3.7 °C. Their mean matches NOAA's own published mean to 0.01 °C.
The record
+2.6 °CNovember 2015 to January 2016
The highest three-month value in NOAA's table since 1950. 100 % of this month's runs reach it.
Kept honest
2models not moved to the normal
Each model is moved from its own base years to 1991 to 2020 by the observed difference; two publish no base years and are drawn as they come. Models that run in both families are taken once, from NOAA's.
The Harvest reading
the map's second reading: yields, for regions on 1966 boundaries
Regions
276counted for rice
Districts as they stood in 1966, so one series runs unbroken. 12 crops, 72 to 276 regions each.
El Niño years
12index +1.0 °C and over
1968, 1972, 1976, 1977, 1982, 1986, 1991, 1994, 1997, 2002, 2009, 2015. A wider list than the rain reading's, because the table ends in 2017.
The test
398region-crops stand apart
Yield below the region's own three years before and after, more often in El Niño years than in its ordinary years. Chance alone: up to 215.
Two classes
40hold; the rest lean
Holds: still stands apart when the number of regions tested is allowed for. Leans: stands apart alone; some of these are chance.
How exactly
- A typical neutral year. For each district and season, a Theil-Sen line is drawn through the neutral years' rain totals and read at the event year. That is the benchmark, so a long drying or wetting trend is not mistaken for El Niño.
- Drier, counted. An event season is drier when its total is below the benchmark. Each district gets a count out of the events and the median percent. The map colours the count, not a prediction.
- Seasons with almost no rain are left out. Where a typical neutral season brings under 20 mm, a percent means nothing; those districts are blank for that season.
- Where this departs from the source. The method is Zeke Hausfather's (The Climate Brink, 2026), built on GPCC monthly rain at 1° for world regions. Here it runs on IMD's 0.25° daily rain at district grain. The 8 events are the same as his.
Where sources disagree
checked on 3 October 2026
- Two copies of the Pacific index. NOAA's own table and the copy at its Physical Sciences Laboratory differ by up to 0.35 °C in some winters. On the second, 1991 would count as a ninth strong event. This atlas uses NOAA Climate Prediction Center's own table.
- The grid against IMD's official season figure. All-India rain from the grid runs up to 7 points above IMD's published percent of normal (in 2009: 85 against 78). IMD's official figure comes from a station series, not this grid. The direction never differs.
- A second rain record agrees. GPCC, an independent gauge analysis, gives all-India monsoon figures within 3 points of the ones here in every event year it covers.
What it can't tell you
- What 2026 or 2027 will bring. Eight events is a small record. It says what has happened, and how often it did not.
- The Indian Ocean's part. A positive dipole can cancel a Pacific signal, as in 1997. It is not separated out here yet.
- Anything finer than a district, or than the 25 km grid the rain comes on. Hill districts with few gauges are the least certain.
Sources
Computed 8 October 2026.