The Rivers Were Already Low. El Niño Starts From There.

The World Meteorological Organization published State of Global Water Resources 2025 on 17 September. Its headline finding: 2025 was one of the driest years for river discharge worldwide in 35 years, with below-normal conditions across about 36% of global catchment area.

The year that report describes was not an El Niño year. WMO's own climate summary records neutral ENSO conditions for most of 2025 with a weak La Niña re-emerging toward year end, alongside a predominantly negative Indian Ocean Dipole. The dryness in it is not El Niño damage. It is the hydrological state the 2026-27 event inherits, which the report's foreword says plainly: "a strengthening El Niño increases risk of drought in some areas, and further flooding in others."

So the report is the baseline, and the forecast is the forecast. What follows is what the baseline actually says, because in at least one place the number that reached the press is not the number in the document.

Two 36 percents, and the one that changed units

The modelled portion of global catchment area falls into three categories for 2025. Around 7% of global catchment area is not represented because of model-resolution limitations.

Two different quantities, the same figure. That is a coincidence of one year's arithmetic, and it is the root of the trouble. Overall, discharge deviated from normal across roughly 57% of global catchment area.

The report's seven-year framing is explicit about its unit: "The share of global basin area experiencing normal discharge conditions has remained below the historical mean in each of the past seven years (34%-38%, against a 1991-2020 mean of 46%)."

WMO's press release restated that as "the number of river basins with 'normal' conditions was in a clear minority: ranging between 34-38%." Area became basin count. Wire coverage then carried a basin-count framing onto the other figure, reporting below-normal flows in "over a third of the world's river basins."

Nobody invented a number here, and the values are unchanged at every step. But area and basin count are different denominators, and they diverge because a handful of very large basins carry a great deal of area while a great many small basins carry very little. The report's own wording is the one to quote. Where a figure in the release and a figure in the document disagree on units, the document wins.

How the 36% was made, and where it is weakest

The discharge assessment is not a gauge reading. It comes from an ensemble of 13 global hydrological models driven by observed weather data, validated against station records. Both halves of that sentence matter.

WMO validated against 3,611 stations: 3,115 with in situ observations and 496 infilled using space-based observations. Modelled anomalies agreed with observations in 63% of validated basins in 2025. WMO names where they did not agree: parts of eastern North America, the Amazon basin, and parts of Australia.

And the station network is not evenly spread. Europe holds 36.9% of it; North America, Central America and the Caribbean 27.2%; the South-West Pacific 19.2%; South America 9.6%. Africa and Asia together hold 7.1%, and those are the regions that accounted for more than 80% of reported deaths from water-related extremes over the past five years. WMO states this gap directly rather than burying it, which is to its credit, and it is the single most important caveat on any global figure in the report.

So the 36% is a defensible number with a stated method, a published validation rate and a known coverage problem. Treating it as a direct measurement overreads it. Dismissing it because models are involved underreads it.

Which basins, and over what period

The report mixes two timeframes. Some findings describe calendar 2025. Others describe the five reporting years, 2021 to 2025. Reading a five-year statement as a 2025 statement is the easiest error in the document.

Note what that does to the Amazon, because it runs against the assumption. In 2025 the Amazon basin and the coastal basins of Peru recorded normal to above- and much-above-normal discharge. The Amazon appears in the five-year recurring below-normal list, not the 2025 one, and it is also on WMO's list of basins where the models and the gauges disagreed. The long-running deficit in parts of that system is real and we covered it in the drought debt piece, but anyone citing this report for "the Amazon was low in 2025" has the wrong document.

The Zambezi is the clean case: much-below-normal discharge in 2025, and Kariba much below normal while almost every other major African lake and reservoir sat much above normal.

Reading the list against the ENSO map

The 2025 map should not be laid over one generic El Niño impact map. WMO's regional categories combine places with different seasonal rainfall regimes, and 2025 was shaped by neutral ENSO conditions, a weak La Niña late in the year, and a predominantly negative Indian Ocean Dipole. The useful comparison is therefore regional and seasonal: the Zambezi and Kariba entered the current event from observed 2025 deficits, while much of South, East and South-East Asia entered from above-normal discharge. Neither condition alone determines the next season's outcome. We laid out the regional picture in nine regions, one El Niño.

There is a sign trap here, and the report contains its own worked example. WMO notes that drought returned to the Greater Horn of Africa in 2025 after one of the driest October to December seasons on record, "linked to La Niña and a negative Indian Ocean Dipole," prompting a state of emergency in Somalia in November. Two paragraphs earlier in the same chapter, describing the same region: "Late in 2023 the region experienced significant flooding due to heavy rains linked to El Niño and the positive Indian Ocean Dipole."

Same basin, opposite drivers, opposite outcome, both documented by WMO. Nothing about the Greater Horn in 2025 should be carried forward into 2026 without stating which phase is doing the work.

So the map does not uniformly darken under an El Niño. It rearranges. What has changed is the level it rearranges from.

The slow reserves underneath

The annual discharge figure is the headline, but the slow series carry the argument. WMO Secretary-General Celeste Saulo framed it as a balance sheet: "Traditionally these slow reserves have acted as the planet's savings account: a buffer that absorbs bad years so that a single drought or a single dry season does not translate directly into a water crisis. We are depleting that savings account."

• Terrestrial water storage shows a persistent decline since 2014-2016, with increasing area under below-normal conditions. Between 2021 and 2025 the repeat offenders were the south-western United States, Patagonia and the subtropical Andes, the Ganges-Indus headwaters, northern Africa, the Middle East, Central Asia and eastern China.
• Groundwater: in 2025, 65% of groundwater observation stations recorded annual mean levels outside the normal range. Below to much-below-normal at 34% of stations, up from 25% in 2024; above to much-above-normal at 31%, down from 37%. The balance moved toward deficit, not recovery.
• Glaciers lost 408 ± 132 Gt in hydrological year 2025, equal to 1.1 ± 0.4 mm of sea-level rise, the fourth consecutive year in which every major glaciated region recorded a net loss. Central Asia, South Asia West, the Russian Arctic, Iceland, and Western Canada and USA all posted annual mass balances among the three most negative on record for those regions. Cumulative loss across 2023-2025 is about 1,400 Gt.
• Peak water: the Caucasus, Western Canada and USA, and Central Europe saw little or no increase, or even a decrease, in 2025 summer balance despite greater annual mass loss than the 1991-2020 norm. WMO reads that as those regions having already reached peak water, the point past which glacier runoff declines rather than grows.

The savings-account metaphor is useful and it has a limit worth naming. A savings account has a balance you can read to the cent. Terrestrial water storage is a satellite gravimetry anomaly series with real observational uncertainty, and WMO publishes glacier loss as 408 ± 132 for a reason: that error bar is a third of the central value. The direction is well supported across independent measurement systems. A precise remaining balance is not available, and nobody should quote one.

What this does, and does not, tell you about the next six months

What it does tell you: 36% of modelled global catchment area had below-normal discharge in 2025, while terrestrial water storage has shown a persistent global decline since the mid-2010s. In regions already carrying low reservoir levels, soil moisture or groundwater, a further rainfall deficit can produce a larger water-supply impact. That is a vulnerability, not a forecast of drought magnitude.

What it does not tell you: any drought magnitude, anywhere. There is no route from the size of an ENSO index to a regional drought probability, and this site does not publish one.

The Pacific numbers, for the record: CPC's El Niño Advisory, unchanged in the 10 September Diagnostic Discussion; weekly relative Niño 3.4 at +2.0°C for the week centred 9 September, with Niño 1+2 at +3.7°C and Niño 4 at -0.1°C; a 98% probability of a very strong event in OND on CPC's strength table, defined on 3-month RONI at or above +2.0°C; and the 75% historic-event probability for OND, defined on 3-month RONI at or above +2.5°C. Those two probabilities come from different CPC products and should not be merged. We took them apart in the odds piece.

What to Watch

• CPC Diagnostic Discussion, 8 October 2026. Whether the OND historic-event probability moves off 75%, and whether the alert status changes.
• The weekly relative Niño table, every Monday. Whether Niño 3.4 holds at or above +2.0°C while Niño 4 stays negative. That east-weighted structure is the shape of the event, not a detail.
• Kariba lake levels into the October to March wet season. WMO puts the world's largest reservoir by volume much below normal while nearly every other major African lake sits much above, which makes it a useful gauge to watch through the southern African wet season.
• Colorado basin snowpack from November. WMO recorded much-below-normal seasonal peak snow mass there in 2025. Compare observed snowpack with CPC's formal winter outlook rather than treating an El Niño composite as a local forecast.
• WMO's next El Niño/La Niña Update, and whether it revises the persistence figure published on 3 September.