Data Sources for Hybrid Machine Learning Framework for Spatial Estimation of Weibull Wind Parameters over Coastal and Offshore Sites in Algeria: A Multi-Model and Multi-Period Approach

1.        Raster Data Sources

The national wind resource rasters spanning the whole Algerian region were applied. We can get these datasets as GeoTIFF files (DZA_wind-speed_10m.tif and DZA_power-density_10m.tif). They show the mean wind speed V (m/s) and wind power density P (W/ m²) at 10 m above the ground.

In the Global Wind Atlas (GWA) version 3.0 (2019), wind resources were calculated with improved accuracy applying novel modeling methodologies and high-quality input data (“Global Wind Atlas,” 2023)data. In this version, Vortex completed roughly 10 years of mesoscale time-series simulations at a spatial resolution of 3 km, pushed by the latest ERA5 reanalysis data, instead of depending on ensemble-based methodologies (Olauson, 2018).

2.  Station Measurement Data

Ground-truth Weibull parameters (C, k, v̅) for 17 time periods p (12 monthly periods, four seasonal periods, and one annual period) were produced using wind speed data spanning a range of 30 years from 1995 to 2025 from a network of 62 weather stations scattered over Algeria (dataset: data23.txt)(“Analyse Vent · Streamlit,” 2025). Each record comprises the geographical coordinates—longitude (x) and latitude (y)—as well as the fitted Weibull parameters for each monthly, seasonal, and annual period. These variables represent the target outputs used in training the supervised machine learning model.