Antarctic Wind Atlas: Gradient-Boosting based Quantile Mapping on 15 years of Climatological Reanalysis

The Antarctic Wind Atlas provides a high-resolution assessment of wind resource statistics across Antarctica. As climate concerns and the logistic costs of fuel in Antarctica rise, many research stations are looking to expand their renewable energy portfolios. Wind energy is a particularly promising option, with tens of wind turbines already operating on the continent. However, identifying suitable locations for these projects remains difficult due to the lack of reliable wind statistics. Utilizing 15 years (2010–2024) of ECMWF ERA5-Land reanalysis data, we fit Weibull distributions to derive wind speed characteristics, wind energy density, and turbine-specific power potential at 0.1°x0.1° spatial resolution. Our study leverages 178 existing automatic weather stations for parametrized quantile mapping, correcting reanalysis biases using a novel statistical transformation based on the inversibility of the Weibull function, which allows for analytical mapping of wind distributions. We apply an ensemble of gradient-boosting machine learning algorithms to determine the optimal mapping for each grid-cell across the continent based on ERA5 time-invariant parameters, resulting in a predicted average wind speed RMSE of 1.41 m/s and correlation of 0.86. The Antarctic Wind Atlas is a novel refinement of reanalysis-based Antarctic wind resources, providing an indication of wind energy potential before any on-site measurements are conducted, an important factor given the high costs of deploying measurement campaigns in Antarctica. This atlas lowers the barrier for stations to consider integrating wind power into their operations by offering our validated statistics. The resulting wind resource maps provide insights for scientific research, logistical planning, and sustainable energy development in the most pristine environment on Earth.