What should I do when two wind-resource models disagree at the same site?

The short answer

When models disagree, first identify whether they actually represent the same location, height, period and wind quantity. Averaging unexplained outputs does not resolve their differences. The useful response is to trace the disagreement to inputs, resolution, assumptions or evidence and then assess how it affects the project decision.

Build a like-for-like comparison

Record each model's coordinates, grid resolution, reference height, source period and treatment of local terrain or obstacles. Check whether one output is a long-term mean while another describes a particular year. Similar map labels can conceal different definitions.

Also confirm whether values are wind speeds, energy indicators or predicted turbine output. A capacity-factor map for one generic turbine should not be treated as a measured wind-speed record for every product.

Investigate the physical reasons

Compare the site plan and available observations with the assumptions of each product. A coarse model may not resolve a local ridge or shelter belt; a detailed model can still depend on uncertain input geometry or calibration. Detail alone is not proof of correctness.

Use independent suitable evidence where available and retain unresolved differences. Do not select the model with the highest forecast simply because it supports the desired payback. Likewise, an arbitrary average can create a precise-looking result without a defensible physical basis.

Make uncertainty visible in the next decision

Show how the credible model range affects useful energy and equipment selection. If all reasonable cases lead to the same screening conclusion, document that robustness. If they lead to different actions, scope measurements or specialist analysis targeted at the disagreement.

The handover should preserve the model versions, comparison table and reason for the selected interpretation. This gives the buyer a transparent resource basis while keeping modelled evidence separate from on-site measurements and from any product-specific generation commitment.