Assessment of High-Resolution Wind Fields in the Basque Country Using Diagnostic Models

Accurate representation of near-surface wind fields in complex terrain is essential for understanding and forecasting atmospheric hazards such as strong wind events, gap flows, and local wind extremes. In this study, we assess the capability of two diagnostic models to generate high-resolution wind fields over the Basque Country, a region characterized by complex orography and heterogeneous land use.

Euskalmet’s operational numerical weather prediction (NWP) systems currently operate at horizontal resolutions of up to 1 km. However, resolving microscale wind features relevant for hazard assessment requires finer spatial detail, often at a high computational cost. To address this limitation, CALMET and GRAMM are evaluated as computationally efficient downscaling tools capable of producing wind fields at 100 m resolution, with potential for further refinement. Both models, widely used within the CALPUFF and GRAL modelling frameworks, adjust meteorological input fields using terrain elevation, land use, and observational data, enabling an enhanced representation of local wind patterns. Simulations were performed on a 100 × 100 m grid using wind observations from automatic weather stations under different dominant wind regimes. The temporal evolution of hourly mean wind fields was analysed to assess the models’ ability to reproduce local flow structures.

Special attention is given to extreme wind situations, including episodes of strong synoptic forcing and locally enhanced flows associated with complex terrain features such as valleys and coastal gaps. Comparative analyses were carried out for selected high-impact events to evaluate the performance of both models in capturing wind intensity, spatial gradients, and the location of local maxima. These results provide insight into the strengths and limitations of each approach under adverse conditions, where accurate wind representation is particularly critical for risk assessment and early warning systems. A preliminary validation was conducted using standard statistical metrics, including bias, root mean square error, and correlation coefficient, complemented by graphical diagnostics such as scatterplots and Taylor diagrams. Results indicate that both models capture the main spatial patterns of wind flow, with differences in their sensitivity to terrain-induced effects and local variability, particularly under extreme conditions.

The proposed methodology is transferable to other regions with complex terrain and can be driven by either observational data or NWP outputs. These results highlight the potential of diagnostic downscaling approaches to improve the representation of wind-related hazards at microscale, supporting both forecasting and risk assessment applications.