Failed Winter: Cape Town’s 2026 Warning Before the Next Day Zero

One extraordinary deluge in May masked what followed: a three-month breakdown in the normal rhythm of the Cape winter. The evidence points not to an absence of Southern Hemisphere winter, but to something more spatially peculiar: a persistent atmospheric wave pattern that repeatedly rebuilt high pressure in the wrong part of the South Atlantic. The unanswered question is what kept that pattern returning.

There is a difference between a dry winter and a winter that fails.

A dry winter still behaves like winter, just more grudgingly. Fronts arrive. Cold maritime air follows. Mountain showers persist behind them. Snow appears at elevation when the air is cold enough. Rainfall may be disappointing, but the seasonal machinery is recognisable.

The Cape winter of 2026 increasingly stopped behaving that way.

Winter arrived violently in May. By 22 May, Ceres had received 507 mm, including 333.2 mm on 11 May alone. Rivers flooded, catchments filled and dams rose quickly.

In retrospect, May may have been less the beginning of winter than a dramatic interruption to what followed.

June weakened. July collapsed. August did not recover.

Cape Town International recorded only 14.4 mm in July, its driest July since records began there in 1958. Observatory recorded 15.6 mm, Malmesbury 3.2 mm, Hermanus 12 mm and Paarl 27.6 mm. SAWS attributed the July collapse to a strong blocking Atlantic high-pressure system west of the country, which drove cold fronts farther south and left only three weak fronts crossing the south-western Cape during the month.

A reconstruction from archived SAWS daily rainfall bulletins tells the same story across the core winter. The archive is incomplete, so its monthly totals are lower bounds rather than final gauge totals. Even so, it captures 50 mm, 14 mm and 12 mm at Observatory in June, July and August, and 161 mm, 63 mm and 53 mm at Kirstenbosch. At Kirstenbosch, those three months amount to 277 mm against a long-term June to August reference of roughly 671 mm. Missing bulletin dates mean the true percentage was somewhat higher than the captured 41%, but the July reconstruction closely matches the official final totals at several stations.

The distribution was almost more striking than the amount. Of Kirstenbosch’s 53 mm captured in August, 43 mm fell on 6 August. At Observatory, the archive captures the entire August total on that same date.

This was not simply a rainfall shortage. It was a failure of recurrence.

Winter did not disappear. It missed the Cape.

Winter remained active elsewhere. Southern Chile continued receiving Pacific systems, and NOAA’s July reanalysis shows above-normal precipitation there. Australia and New Zealand experienced repeated frontal and polar outbreaks. Southern Namibia, normally deep in its dry season, received anomalous winter rain. Within southern Africa, cold, rain and snow repeatedly reached the Eastern Cape, Free State, Lesotho and KwaZulu-Natal.

So the problem was not that winter circulation could no longer reach Cape Town’s latitude.

It repeatedly reached farther north elsewhere.

The Cape was not too far north for winter. It was repeatedly beneath the wrong part of the wave.

SAM explains the start, but not the persistence

The Southern Annular Mode, or SAM, is the broad hemispheric measure of whether the westerly winds and associated storm systems circling Antarctica are displaced poleward or equatorward. Positive SAM shifts the frontal track south towards Antarctica; negative SAM allows it farther towards the equator. Neutral SAM is closer to the usual seasonal position.

For the winter-rainfall Cape, the geometry matters. A strongly positive SAM can reduce the odds of Atlantic fronts penetrating far enough north to deliver normal winter rain.

June fits that explanation remarkably well. SAM became exceptionally positive from late May, reaching +5.03 on 27 June, the second-highest daily value for any month in the Bureau of Meteorology record. A poleward-contracted storm belt is entirely consistent with strong Southern Ocean lows continuing east while their rain-bearing fronts struggle to reach south-western Africa.

But then the explanation breaks down.

By 28 July, the Bureau of Meteorology reported SAM as negative. By 11 August it was neutral. The Cape circulation barely responded.

That is one of the most important clues of the season. If positive SAM alone had been sustaining the failed winter, the return towards negative and neutral values should at least have improved the odds of northward frontal penetration.

Instead, the South Atlantic ridge kept returning.

SAM is a hemispheric index. It describes the broad ring of westerlies around Antarctica, but it can smooth over enormous differences by longitude. A single number can therefore say that the hemispheric belt has relaxed while one longitude sector remains stuck in a positive-SAM-like configuration.

That appears to be what happened over the Cape sector.

The ridge moved. The pattern kept coming back.

It would be wrong to say that one South Atlantic High sat stationary for three months.

Individual highs moved. Fronts formed. Lows crossed the Southern Ocean. The atmosphere was not frozen in place.

What repeatedly re-established itself was the preferred ridge longitude over the south-eastern South Atlantic. At Cape longitude, the upper-level westerlies repeatedly bowed poleward around that ridge. A few thousand kilometres east or west, troughs could amplify towards the equator and produce rain, cold and snow.

Rossby waves provide the useful mental model: enormous undulations in the westerly circulation, with stable high pressure favoured beneath ridges and stormier weather beneath troughs.

NOAA reanalysis shows enhanced South Atlantic ridging already present in April, an amplified Southern Hemisphere wavetrain in May, and a broadly annular mid-latitude height pattern in July. The daily Cape sequence then added the local signature: fronts passing far south, weak tails clipping the south-west, little sustained post-frontal weather, rapid rebuilding of high pressure, Berg-wind warmth, and then the next promising system doing much the same thing.

The missing cold was part of the same story. Significant Cape mountain snow was scarce after May. ERA5 contains a zero-degree-level field that could quantify this properly, but this investigation has not yet produced a verified May to August series. No numerical freezing-level anomaly is claimed here.

What might have kept the wave in place?

This is where the evidence becomes a hypothesis rather than attribution.

As the Cape became disconnected from the simple SAM explanation, the tropical oceans were developing an unusual contrast.

The 2026 Pacific El Niño strengthened rapidly. The Bureau of Meteorology’s relative Niño3.4 index was around +1.2°C in late June, +1.94°C in late July, and +2.45°C by the end of August. By 1 September, SAWS formally described a strong El Niño as established and expected to strengthen further into spring, independently confirming that the Pacific forcing discussed here was still intensifying as the Cape winter ended.

At almost the same time, NOAA’s relative ATL3 index, which measures the eastern equatorial Atlantic relative to the surrounding tropics, moved from −0.19°C in May to −0.46°C in June and −0.94°C in July. That is a strong Atlantic-Niña-like relative cooling signal. NOAA had not yet published a settled August relative ATL3 value when this article was prepared, so none is used here.

Cold Atlantic water did not simply create the South Atlantic ridge. That would be far too crude.

But tropical ocean temperatures help determine where deep convection releases heat into the upper atmosphere. Changes in that tropical heating can alter the source and propagation of Rossby waves far into the extratropics.

The working hypothesis is therefore this:

Extreme positive SAM helped establish the Cape’s dry regime in June. As SAM weakened, a zonally asymmetric stationary-wave pattern continued to favour ridging over the south-eastern South Atlantic. The unusually strong contrast between a rapidly warming tropical Pacific and a relatively cool eastern equatorial Atlantic may have helped reinforce that preferred wave phase through changes in tropical convection and Rossby-wave forcing.

That is physically plausible.

It is not formal attribution.

The distinction is the difference between a serious question and a convenient story.

The Day Zero lesson comes before the dams

The extraordinary May rainfall protected Cape Town from the immediate hydrological consequences of what followed. It put a large deposit into the water system before the core winter mechanism deteriorated.

That should not make 2026 reassuring.

It should make it useful.

Nor does the risk end with the winter calendar. SAWS entered September forecasting broadly above-normal temperatures across South Africa through spring and early summer, while the strengthening El Niño increasingly favours drier conditions across much of the country as the summer season develops.

The 2015–17 Day Zero drought showed what happens when poor recharge compounds across seasons. Research from that period linked Cape winter drying to changes in the South Atlantic anticyclone, frontal behaviour and post-frontal high pressure. In 2026, May prevented a meteorological failure from becoming an immediate water-supply crisis.

But reservoir levels are lagging indicators. So are failed crops, stressed urban landscapes and ecological damage. By the time those outcomes prove beyond doubt that a winter has failed, much of the useful planning window has already gone.

The earlier warning is atmospheric.

For SAWS, universities, water planners, the City of Cape Town, the Western Cape Government, agricultural agencies, growers, nurseries and ecological managers, the useful question is not only how much rain has fallen. It is whether the machinery that normally delivers Cape winter is still functioning.

That means watching the Cape-sector 500-hPa ridge, South Atlantic sea-level pressure, jet latitude by longitude, frontal frequency, post-frontal cold-air residence, freezing level, SAM, tropical forcing and the failure of rainfall to recover when broad hemispheric indices change.

The Cape needs a way to recognise a circulation failure while it is happening, not only once the dams and fields have translated it into consequences.

2026 has given us an unusually clean warning because May bought time.

The next occurrence may not.

And that is why the question of what kept rebuilding the ridge deserves a proper meteorological post-mortem, rather than being filed away as another dry winter.

Sources and methodology

• South African Weather Service rainfall reporting and daily bulletins: SAWS media release, July 2026 was the driest on record in parts of the south-western Western Cape, 31 July 2026, provides the official July station totals, the three-front count and SAWS’s attribution of the month’s frontal suppression to a strong blocking Atlantic high-pressure system west of the country. Archived SAWS daily rainfall bulletins were obtained through Cape Town Data. The compiled workbook contains 2,132 station-day records downloaded on 6 September 2026. Missing bulletin dates mean reconstructed monthly totals are treated as lower bounds. https://capetowndata.com/en/data/daily-rainfall/ and https://www.sanews.gov.za/south-africa/call-use-water-sparingly-parts-western-cape-record-lowest-rainfall
• Cape Town International climatology: JMA ClimatView, WMO station 68816, for 1991–2020 rainfall and temperature normals. https://www.data.jma.go.jp/tcc/tcc/products/climate/climatview/
• Kirstenbosch climatology: SANBI long-term monthly rainfall references used to contextualise the bulletin reconstruction.
• ERA5 / ERA5T: ECMWF Copernicus Climate Data Store single-level and pressure-level reanalysis. ERA5 is a numerical reconstruction that assimilates observations into a consistent atmospheric model, not a satellite photograph. https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels and https://cds.climate.copernicus.eu/datasets/reanalysis-era5-pressure-levels
• Southern Hemisphere circulation: NOAA Climate Prediction Center Climate Diagnostics Bulletin / CDAS Reanalysis, including April and July 2026 circulation diagnostics. https://www.cpc.ncep.noaa.gov/products/CDB/
• Southern Annular Mode and El Niño: Australian Bureau of Meteorology climate-driver monitoring and archive. https://www.bom.gov.au/climate/sam/ and https://www.bom.gov.au/climate/enso/wrap-up/
• SAWS El Niño and seasonal outlook: South African Weather Service, South African Weather Service statement on the 2026/27 El Niño, 1 September 2026, and Seasonal Climate Watch: September 2026 to January 2027, issued 31 August 2026. The latter forecasts broadly above-normal minimum and maximum temperatures through spring and early summer and an increasing dry signal across significant parts of South Africa as El Niño strengthens. https://www.weathersa.co.za/home/seasonal
• Tropical Atlantic: NOAA CPC relative ATL3 monitoring. The term “Atlantic-Niña-like” here describes the strong negative relative ATL3 signal and does not establish Atlantic Niña as the cause of the Cape anomaly. https://www.cpc.ncep.noaa.gov/products/international/ocean_monitoring/IODMI/rATL3.html
• Cape drought dynamics: UCT research and the peer-reviewed analysis of the 2015–17 Cape Town drought, including frontal rainfall, South Atlantic anticyclone behaviour and post-frontal suppression. https://www.news.uct.ac.za/article/-2018-06-29-lingering-high-pressure-cells-limit-winter-rains and https://www.nature.com/articles/s41612-019-0084-6
• May 2026 rainfall context: SAWS reporting and TerraClim / SA Wine climate reporting used to contextualise the exceptional May event. https://www.terraclim.com/

AI-assisted research disclosure: This investigation was directed and written from the author’s research questions and observations. AI-assisted research tools were used to reconcile station observations, sum the downloaded SAWS daily-rainfall archive, compare climatological baselines, cross-check conflicting public datasets and synthesise reanalysis and climate-index material. AI was not treated as a meteorological source. Observations and climate indices used in the article are traceable to the sources listed; calculations derived from incomplete rainfall bulletins are identified as lower-bound reconstructions, and the proposed tropical forcing mechanism is explicitly presented as a hypothesis rather than attribution. No AI-generated meteorological value is presented as an observed measurement.