Built to a Class, Not a Number

A line of severe thunderstorms crossed central South Dakota in the early hours of 29 June 2026. Storm Prediction Center and National Weather Service preliminary reports list measured gusts of 131 mph in Hyde County around a quarter past six in the morning: one from an Ambient Weather station east-southeast of Holabird, and a second about ten minutes later from a South Dakota Mesonet station west-northwest of Highmore, with further readings of 112 mph near Joe Creek and 114 mph at Highmore. Preliminary reports also note two tornadoes further along the track. In Highmore the wind took roofs, grain bins, power lines and the roof of a church; residents reported lost livestock but, as far as has been reported, no injuries.

The images that travelled fastest were aerial photographs of a wind farm with several turbines folded to the ground and their blades scattered across the fields. One widely shared post put it simply: 130 mph winds destroyed the turbines, and above roughly 110 mph they begin to crumble. The sympathy in that reading is understandable. The engineering in it is not quite right, and the distance between the two is the subject of this note.

What the record shows

The system was a fast-moving complex of severe thunderstorms carrying an embedded swath of extreme straight-line wind, with preliminary reports of two tornadoes separately along the track. The damaging gusts near Highmore are the kind a downburst or bow-echo wind field produces rather than the tornadoes themselves, which the preliminary reports tie to only minor damage.

Hyde County is wind-farm country, and it holds machines of several generations. The older South Dakota Wind Energy Center, the state's first major wind farm, was commissioned in 2003 with twenty-seven 1.5 MW units on towers of around sixty-five metres. The newer Triple H project came online in 2020 with ninety-two 2.72 MW machines, and the North Bend project in 2023 with seventy-one 2.82 MW machines, both far larger. The turbines in the photographs appear to be modern utility-scale units, but photographs alone cannot establish which project was struck, the turbine model, the hub height, or the IEC design class of the machines that failed. At the time of writing the operator has confirmed none of these, nor how many were lost.

A turbine is built to a class, not a wind speed

Large wind turbines are certified to the international standard IEC 61400-1. Rather than a single survival wind speed, the standard sorts machines into classes by a reference wind speed, the ten-minute mean wind at hub height that the machine is built to withstand. Class I is the most robust and Class III the least, with the annual mean wind taken as one fifth of the reference figure. The three standard classes are built to reference winds of 50, 42.5 and 37.5 metres per second, which is to say about 112, 95 and 84 mph as a sustained mean. Turbulence categories sit on top of this, a bespoke Class S covers sites that do not fit the standard envelope, and a 2025 amendment adds a tropical-cyclone class.

One key survival benchmark is not the mean but the fifty-year extreme gust, Ve50, which the standard derives as 1.4 times the reference wind at hub height. That puts it at roughly 157 mph for Class I, 133 mph for Class II and 117 mph for Class III. It is a benchmark, not the whole design: certification also checks turbulence, the parked and idling states, yaw misalignment, loss of grid, and a range of control and fault conditions, any of which can govern instead. Above the cut-out wind speed, typically around 25 metres per second or 56 mph, the machine stops generating, feathers its blades and idles, and it is designed to ride out that extreme gust in the parked state with the yaw system holding the rotor within a bounded misalignment. So the useful question is not whether 131 mph exceeds some universal limit. It is how the gust compares with the design envelope for the particular class, at the height that matters.

The headline number was measured near the ground

The 131 mph figures are near-surface station readings, not hub-height rotor measurements. South Dakota Mesonet documentation notes that current stations measure wind at ten metres while older generations used three-metre sensors, and the private Ambient Weather station that logged one of the two readings may have its own mounting height and exposure. Hub height for large modern machines is in the order of eighty to ninety metres, so comparing any of these surface readings directly with a hub-height figure is an approximation, and the correction can run either way.

For its extreme wind model the standard extrapolates with a power-law exponent of 0.11 (the normal wind profile uses a steeper 0.2). Either way wind rises with height, so the gust at hub height would be somewhat higher than the value measured at ten metres, which is worse for the structure. Thunderstorm downbursts, though, do not follow either profile. Their strongest winds often sit at low to mid height in a nose-shaped profile, so the hub-height value could be similar to, or in places lower than, the surface reading. The honest position is that we do not yet know the hub-height gust at the failed machines, and the surface number alone cannot settle it.

Why moderate-wind sites can sit closer to the margin

The point is not size in itself but the class a moderate-wind site invites. The annual mean wind across much of the Great Plains is moderate, in the order of 7.5 to 8.5 metres per second at hub height, which sits in the lower IEC classes. To capture more energy from that steady but unexceptional wind, developers favour low-wind, large-rotor models, which are commonly certified to Class III or to a site-specific Class S. The economics are sound, but the same choice lowers the extreme-gust benchmark, to around 117 mph at hub height for Class III against 157 mph for Class I. If the affected Hyde County machines were of this low-wind, large-rotor type they would belong to that lower-benchmark family, but their class is not confirmed.

Load scales with the square of wind speed, through the dynamic pressure, one half the air density times velocity squared. To see the sensitivity, take air density at roughly 1.2 kilograms per cubic metre, a little lower at this elevation: a 131 mph gust then carries about 2.06 kilopascals, the Class III Ve50 of 117 mph about 1.65 kilopascals, and the 56 mph cut-out about 0.37 kilopascals. So the surface gust corresponds to roughly a quarter more pressure than the Class III benchmark, and to something like five and a half times the pressure at cut-out. A band of a few tens of mph separates the machine idling safely from the machine overloaded, which is why the headline wind speed, on its own, decides so little. These figures are illustrative: they treat the surface gust as the loading gust, while the true rotor-height gust is unknown and the height correction moves them either way.

How the damage reads

Two failure modes are visible in the photographs. Several towers have buckled and folded over, the fold a third to half of the way up the tube. That is consistent with local buckling of the tubular steel shell under combined bending and other loads exceeding the section capacity, but photographs cannot show where the controlling weakness lay: a tower can, a flange or weld, the door opening, a wall-thickness transition, the anchor bolts, or the foundation, and the fold height need not mark the origin. Separately, some machines left standing have lost blades, with composite shredded at the root or tip, which points to blade overload or detachment as a distinct mode, one that can occur while the tower survives.

Both readings are provisional. From aerial images one cannot tell whether a given machine was parked and idling or still rotating, what yaw angle and blade pitch it held when the gust front arrived, the direction the outflow came from, or whether the tower shell, a flange connection, or the foundation bolts governed. The spacing of neighbouring machines may also bear on turbulence, wake exposure or debris impact, but that is a question for the damage survey, not something a photograph can settle, and proximity and wind direction are at most possible contributors until the evidence is in. That evidence is the machines' own SCADA record: yaw angle, pitch angle, rotor speed, parked status, alarms, logged wind direction, and the timing of each failure. This note should be read as informed observation pending the operator's formal post-event assessment.

Where the standard and the storm part company

The extreme-wind model in IEC 61400-1 is built on synoptic extreme-wind statistics and a defined gust and turbulence model. Thunderstorm downbursts behave differently: rapid onset, a strong and fast change of wind direction, a vertical component, and the non-standard vertical profile already noted. A parked turbine is not a passive pole in that wind: its blades pitch, the rotor idles or is held stopped, and the nacelle yaws to keep loads within the design envelope. The parked load cases assume the yaw system keeps the rotor within a bounded misalignment, but an outflow can veer faster than the yaw drive can follow. A rotor caught misaligned as the gust front arrived would not see the clean head-on parked case the envelope is checked against, but a combined one: side load, bending and torsion together, with the rotor and nacelle broadside to the wind. On a parked machine, the direction a gust arrives from can shape the load as much as its speed. The standard is being extended to close gaps of this kind: the 2019 edition and its 2025 amendment add provisions including a tropical-cyclone class and draw in ISO 4354 for wind actions. Downburst loading on tall structures remains an active research question rather than a settled one.

Reading this from Nairobi

This matters here for a reason beyond curiosity about turbines. On 11 June 2026 the United States National Oceanic and Atmospheric Administration upgraded its outlook to an El Niño Advisory: El Niño conditions are present and expected to strengthen into the 2026-27 northern winter, with the agency putting the chance of a very strong event by November to January at about 63 per cent, potentially among the strongest in the record since 1950.

For East Africa the consequence is usually felt as rain, not wind. Regional outlooks from the WMO and the IGAD Climate Prediction and Applications Centre tie El Niño to a raised probability of wetter than normal short rains from October to December across much of the Greater Horn, though the signal varies by sub-region and is never uniform across Kenya; the strong 2023 event coincided with severe flooding across the country and its neighbours. The load that tests infrastructure here is water: saturated and weakened ground, scour around foundations and piers, higher river flows, and, on the coast where I have worked on jetty and marine structures at Lamu and Mombasa, wave action and surge layered on top of the tidal range.

The structures differ and the dominant load differs, but the design lesson is one and the same. A structure stands or falls on whether it was designed for the real extremes of its own site, not for a generic or convenient assumption. The South Dakota machines were in all likelihood built to a defensible standard for their wind class; the storm may have approached or exceeded the applicable design load envelope, the more so for Class III units, or under unfavourable yaw, parked, or fault conditions. The same logic governs a coastal jetty sized for a design wave and surge, a drainage system sized for a design storm, or a foundation checked against a design scour depth. Choosing that design value honestly, against the site's real climate and its tail, is the whole task. None of this is an argument against wind energy: turbines can be repaired and replaced, and the case for wind is not altered by a single severe storm. It is an argument for treating extreme-event margins and honest load assumptions with the seriousness we give the routine cases.

For Kenya that turns a forecast into a work programme. The outlook is not the whole story; it is the clock. The question for the months before the rain arrives is a practical one: are we clearing drains and checking culverts, reviewing bridge scour and mapping low river crossings, auditing the slopes above roads and settlements, coordinating dam releases, and staging emergency access routes, while the ground is still dry? The design value is settled at the desk; that preparation is the field half of the same task, and it is what the months before the short rains are for.

What to watch

Three things will sharpen the picture. First, the operator's formal damage assessment for the affected farm, much of it from the machines' SCADA logs: how many were lost, their design class, whether they were parked, the yaw and pitch they held, the wind direction recorded, and where in the structure the failure began. Second, the meteorological reanalysis of the gust front, which may estimate the hub-height wind the surface stations could not. Third, the updated regional outlook for the October to December short rains as any El Niño develops, which sets the timeline for the preparation above. I will return to these as the assessments are published.

Sources: wind measurements for the 29 June 2026 event are per Storm Prediction Center and National Weather Service preliminary reports and the South Dakota Mesonet, with reporting from the Washington Post, Agweek and regional outlets; two gusts of 131 mph were logged in Hyde County, near Holabird and near Highmore. Wind farm details are drawn from South Dakota Public Utilities Commission project records, and identification of the affected project, model and class awaits operator confirmation. The wind turbine classes, the reference wind speed as a ten-minute mean at hub height, and the fifty-year extreme gust Ve50 taken as 1.4 times that speed extrapolated with a 0.11 exponent, are per IEC 61400-1 (2019 edition with 2025 amendment); Ve50 is one of several load benchmarks the standard applies. Dynamic pressure figures use one half the air density times velocity squared, with assumed values stated in the text, and are illustrative of sensitivity rather than a hub-height load. The El Niño status is per the NOAA Climate Prediction Center advisory of 11 June 2026, with East Africa short-rains outlooks from the WMO and ICPAC, phrased as probabilities. Descriptions of how the turbines failed are preliminary, drawn from early aerial footage; formal post-event structural evaluation will take longer. Wind-load design for buildings in Kenya is governed by the National Building Code 2024 alongside applicable wind-action standards, and any turbine works by IEC 61400.