You don't actually need onsite data offshore - correlation with met data from not-too-distant sources is good enough - offshore wind is predictable over long distances and you can use mesoscale studies (modelisation done using NOAA data for instance).

The one tricky issue offshore is wake effect (i.e. the impact of one row of turbines on the production of the row "behind" them) as it can be quite significant (10-30% for individual rows) in some wind directions - both from the wind fair itself as from neighboring ones.

Wind power