Answers are grounded in this paper and cite their sources.
The Austrian Windatlas provides an hourly, observation-constrained wind climatology for Austria on a 1 km grid, covering 2000–2020 at 10 m and at hub heights of 80–250 m AGL. In total, 23 interpolation configurations within a shared EOF/rPCA framework using 20 annual cross-validation folds and 54 permanently withheld stations are evaluated for the 10 m fields. Seven physical formulations extend the fields to different hub heights, where they are evaluated against SCADA from 251 turbines and compared with the New European Wind Atlas (NEWA) for a subset of five years. The lowest station RMSE (1.42 m s−1) is achieved by a deep-learning configuration that produces substantially smoother spatial fields. A Bayesian additive model with ERA5 covariates (GAM + ERA5) has higher station RMSE (1.66 m s−1), but retains ridge–valley contrasts and achieves the lowest hub-height RMSE (2.39 m s−1), 15–17% below NEWA in flat and hilly terrain. GAM + ERA5 is selected for the delivered atlas using station skill, hub-height skill, spatial structure and capacity-factor comparisons. The atlas supports regional planning and downscaling activities. Its Alpine hub-height estimates remain experimental, and its spread layers describe differences between methods rather than calibrated hourly error intervals.
No takes yet. Share an insight, caveat, or question.
Schicker et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: