Do offshore wind farms take the rain from the coast?
A study published in August 2026 projects that a full build-out of North Sea wind farms after 2050 would cut rain on nearby coasts by 10–15 %, and by more than 15 % in parts of Jutland. We are checking whether the farms that already stand show any effect in 35 years of rain-gauge data. Every rule was locked before a single measurement was read.
What the study actually says
Akhtar, Elizalde, Geyer and Schrum (Helmholtz-Zentrum Hereon), Communications Earth & Environment 7, 651 (2026). A regional climate model, ten years of weather, run with and without wind farms.
10–15 % less rain on the coast
“Precipitation decreases by 10–12% in coastal regions of Germany, the Netherlands, and the United Kingdom, and by more than 15% in parts of Jutland, Denmark, particularly under south-westerly winds.” More rain falls over the farms themselves.
The Jutland effect sits behind farms that don’t exist
The largest decrease lies around Herning, downwind of a cluster the paper places about 80 km off the west coast. That cluster has not been built, so the headline figure can’t be checked against measurements yet.
“Small effects … primarily confined to marine areas”
That is what the paper finds for the farms of 2023 and 2030. It gives no number for land. What can be tested now is whether today’s farms already change rain on the coast, or whether an effect of a stated size can be ruled out.
35 years of offshore wind, turbine by turbine
Every turbine lights up on its own commissioning date, from the registers. Press play, drag through the years, and pick a gauge to see what the test calls upwind.
What the test calls upwind
Gauge near Blåvand. Click any gauge on the map to choose another.
of turbines within 100 km, in a 45° sector towards the wind
The test works this out for every gauge and every day, with the wind direction at 850 hPa from the ERA5 weather model.
Turbines: Marktstammdatenregister (Germany), Energistyrelsen’s turbine register (Denmark), operators’ dates for Gemini (Netherlands) and Lillgrund (Sweden). Coastline: Natural Earth (public domain). The bars below the map are operating capacity at the end of each year: gold for the German and Dutch North Sea, white for the Danish North Sea, grey for Kattegat and the Baltic. No rain values are shown, because none has been read.
Different coasts got their farms in different years
The German islands got farms close by in 2014–2015. Esbjerg and Blåvand have had Horns Rev since 2002, 2009 and 2019. The coast from Ringkøbing to Thyborøn had almost nothing until Vesterhav in 2023–24 and Thor from 2026. Because each coast changed at a different time, a change that follows the farms can be told apart from one that hits the whole region at once.
How the test works
The same gauge, on days when the wind blows from the farms towards it, compared with days when it doesn’t, before and after the farms were built. And all of it measured against a weather model that has no wind farms in it.
Each gauge gets a number for each day: how many gigawatts of operating turbines lie upwind of it. The test asks whether rain falls when that number rises, compared with the same gauge on similar days, and with other gauges at the same distance from the coast in the same year and wind direction.
Four possible readings
The effect is in per cent per gigawatt upwind. “Ruled out” means less than a 10 % change at the most exposed gauge, the size of the paper’s post-2050 decrease. A chance pattern has to beat 199 reshuffles of the farms’ start years before it counts.
Placebos
- Farms behind the gauge. A farm downwind of a gauge is not expected to change its rain. If it seems to, the method is picking up something else.
- Farms switched on early. The same farms moved five years earlier in time should show nothing.
- North-westerly winds. The paper itself says these have negligible impact. They are reported on their own.
Two traps already avoided
- DMI replaced its gauges while the farms were built. In West Jutland the network went from 65 gauges in 1991 to 28 in 2011, mostly manual gauges replaced by automatic ones. A plain before-and-after comparison would measure the new instruments. Every instrument change and move now starts a new series.
- The European farm map had wrong years. One farm is listed as built in 2027, others have no year, one is a quarter of its real size. Dates come from the turbine registers instead.
Where we are
Each step is committed to the public repository before the next one reads anything new. No rain value is read until the method, the farms, the gauges, the weather model and the power check are all in place.
Reading the live status from the repository…
The method
Every rule, the readings, the placebos and the checks, committed on 2 October 2026 and reviewed independently before the first commit.
Read it →Every change, with its reason
Including the day windows of each network, fixed from the metadata before any value was read, and what went wrong on the way.
See the log →Data and code
The turbine and gauge tables, the weather-model values and the scripts that make them, with a SHA-256 for every file.
The repository →A test of the future, locked now
The part nobody else can do afterwards.
Thor and Vesterhav
Vesterhav Syd and Nord (344 MW, 4–10 km off the coast) came online in 2023–24. Thor (1,058 MW, about 22 km off Ringkøbing) sent its first power in 2026 and should be complete in 2027. That is a sharp step right in front of DMI gauges that have been unchanged since 2011.
Once a year, to 2029
The same test, on the new farms only, run each spring when DMI has checked the year’s data. The runs for 2026, 2027 and 2028 show the estimate but no reading. The reading comes with the data to the end of 2029. New farms enter only through a dated change in the log.
What this can’t show
Nothing here confirms or refutes the post-2050 projections: the farms behind them don’t exist. Effects smaller than what the data can detect will look like zero. Rain gauges are on land, while the paper expects more rain over the farms. Gauges catch less rain in strong wind, and a change at a coastal gauge that happens at the same time as a nearby farm can’t always be separated from it.
It goes here, whatever it says
The reading for the farms built so far comes when every step above is done. It will be published on this page and in the repository, together with every placebo and robustness check, whether it finds drier coasts, wetter coasts, nothing, or can’t tell.