Probabilistic high-temporal resolution surface wind nowcasting with deep learning models

High temporal frequency wind speed and direction predictions are critical for short-term operational decisions in the aviation sector, including runway selection, aircraft arrival sequencing, and departure planning. These decisions benefit substantially from having reliable forecasts several hours in advance, allowing operators to evaluate potential scenarios and formulate robust strategies. This results in a demand for forecasts with long forecast horizons at the minute-level time scale, a challenging problem since wind fluctuations are highly random. This study develops probabilistic wind nowcasts at individual sites using well-established deep learning (DL) models. By directly modelling the predictive distributions or quantiles of wind variables, the proposed DL approaches capture both the mean behaviour and the inherent variability of the wind field. Two complementary probabilistic models with distinct architectures and mechanisms are investigated. Both models outperform current Numerical Weather Prediction (NWP) hourly predictions, achieving reductions in forecast error of approximately 10%–30%. Their outputs are further analysed to characterize their respective strengths and limitations. Finally, we demonstrate that a weighted ensemble of the two models can potentially provide additional performance gains, outperforming each individual model, even for the demonstrated case in which one of the individual model has overall significantly worse performance than the other.