Why a 3 km wind map hides your best sites

Early in a wind project, before anyone has put up a measurement mast, a wind atlas is often the first look at the resource. The New European Wind Atlas offers its maps at more than one resolution, and the coarse one is easier to work with. I wanted to know how much that convenience costs.

Same average, different answer

I compared the 3 km and 50 m maps over one project area at the same hub height.

Averaged over the whole area, the two maps agree closely. If the only question were “how windy is this region?”, either would do.

But nobody builds turbines on an area average. They go on specific spots, usually the high ground. At the candidate turbine locations, the maps disagreed sharply:

• the 50 m map gave a noticeably higher mean wind speed at those sites;
• more than twice as many sites cleared 6 m/s on the 50 m map as on the 3 km map.

Same data provider, same height, same area. Two different conclusions about whether the project works.

Why it happens

A 3 km grid cell is an average over nine square kilometres. In hilly terrain that cell holds a ridge, the valleys either side and the slopes between. Wind speeds up over ridges and slows in valleys, and the cell average sits somewhere in the middle.

The area average survives that smoothing almost untouched, because averaging an average changes little. What gets lost is the spread: the coarse map cuts the peaks and fills the troughs.

And the peaks are exactly where turbines go. Developers pick ridgelines and exposed high ground because they’re windier. So the error isn’t random noise that cancels out across a project. It points the same way at nearly every site that matters: the coarse map understates the resource where you’d build.

What changed in practice

The reason people reached for the coarse map wasn’t ignorance. Getting the 50 m maps for a project area and preparing them for GIS work took about half a day by hand, while the coarse map was ready to use.

So I built a small tool for the team: give it a site outline or a bounding box and a hub height, and it extracts the 50 m maps for that area, writes them in the formats the GIS workflow expects, and checks its own output. Half a day became about a minute, and the high-resolution map stopped being the expensive option.

The general lesson

When a coarse and a fine dataset agree on the mean, it’s tempting to conclude the coarse one is good enough. Check what question you’re actually asking. If the decision depends on the extremes, the ridgelines in wind or the heavy days in rainfall, resolution isn’t a detail. The average can be right while every decision built on the map is wrong.

A useful habit: before trusting a coarse product, compare it with a finer one at the points you care about, not just over the area.