PowerVeritas
Insights / Field note · Wind

Is climate change cutting wind yields? 50 years of data says no

Climate change is sometimes blamed when a wind farm consistently underperforms its budget P50. Fifty years of reanalysis at one Scottish site tells a different story, a near-flat trend and very large year-to-year swings.

← All insights

Rising global temperatures are well documented. The effect on wind speeds is far less discussed, yet it increasingly appears in budget conversations, usually as a suggested reason why a site keeps missing its P50.

To test that, we pulled 50 years of hourly ERA5 reanalysis wind for the Hill of Towie wind farm in Moray, Scotland, and ran every hour through the turbine power curve, density adjusted, to simulate energy production.

Two panel chart. Top, annual mean wind speed 1976 to 2025 with a flat trailing 10 year mean around 7.3 m/s. Bottom, simulated annual energy versus the 50 year mean, swinging between plus and minus 16 percent with a flat trailing mean.
Fifty years at Hill of Towie. Annual mean wind speed (top) and simulated energy versus the 50-year mean (bottom), with trailing 10-year means and the P90–P10 band of annual outcomes.

What the data shows

  • The linear trend over 1976–2025 is −0.1% per decade for wind speed and −0.4% per decade for energy. Statistically indistinguishable from zero.
  • The 10-year trailing mean is essentially flat for both wind and energy across the whole record.
  • Individual years are anything but flat. Energy swings up to ±16%, roughly twice the wind speed swing, because the power curve amplifies wind variations.
  • The worst years, 1985, 2010 and the 2021 “wind drought”, sit 12–16% below the long-term mean. Nothing was wrong at the site in those years. There was simply less wind.

That variability, not a climate trend, is what breaks single-year budgets. A perfectly healthy site can sit 14% under its long-term mean for a year, and two or three soft years in a row are well within normal.

What about storms

The same record answers the follow-up question. The highest hourly wind of each year trends at just +0.06 m/s per decade, and the number of storm hours above 20 m/s shows decade-to-decade variation but no long-term trend. Notably, at this site Storm Éowyn (January 2025) ranks below both the February 1989 storm and the 1993 Braer storm.

Two panel chart. Top, highest hourly wind speed of each year 1976 to 2025 with a flat trailing mean around 21 m/s and labels for the 1989 storm, the 1993 Braer storm and Storm Eowyn in 2025. Bottom, hours per year above 20 m/s with no long term trend.
Storm view of the same 50 years. Highest hourly wind of each year (top) and hours per year at or above 20 m/s (bottom).
Method notes. Simulated energy is hourly ERA5 wind passed through the OEM power curve, density adjusted. It is illustrative, not metered production. ERA5 is an hourly grid-cell mean, so maximum values sit below peak gusts recorded on site, but it gives a reliable view of long-term trends. This is one grid point at one site; the balance of trend and variability will differ elsewhere.

What this means for budgets

If a site is consistently underperforming its budget, the cause is usually closer to home. Curtailment, downtime, degradation, icing or control issues. And an annual budget quoted as a single P50 number invites a miss, the P90 context belongs alongside it because normal wind variability alone can produce a double-digit shortfall.

PowerVeritas quantifies the real causes of underperformance from operational SCADA data and provides P50/P90 energy yield forecasts built on what a site has actually produced, combined with 20+ years of historical meteorological data.

Data, ERA5 reanalysis via the Open-Meteo archive. Analysis and charts by PowerVeritas.