A Two-Step Spatio-Temporal Framework for Turbine-Height Wind Estimation at Unmonitored Sites using Sparse Meteorological Data

A Two-Step Spatio-Temporal Framework for Turbine-Height Wind Estimation at Unmonitored Sites using Sparse Meteorological Data.

James Sweeney, University of Limerick.

9th of April 2026

Abstract: 

Accurate wind speed estimates at turbine hub heights are essential for wind resource assessment and energy grid management. Because direct wind measurements at turbine heights are expensive to gather at greenfield sites, wind speeds are typically extrapolated from lower heights using parametric models, though recent extensions allowing spatially or temporally varying vertical gradients still require co-located reference and hub-height data. To address this limitation, we propose a two-step spatio-temporal modelling framework that relies solely on open-access meteorological station data. First, a generalized additive model trained on reanalysis data performs non-parametric vertical extrapolation of wind speeds across multiple heights. Second, a spatial Gaussian process interpolates the resulting hub-height estimates to wind farm locations while propagating uncertainty from the first stage. This enables real-time, high-resolution, sub-hourly turbine-height wind speed estimates and associated uncertainty maps at national scale, capabilities not available from existing reanalysis products. Validation against hub-height measurements from seven Irish wind farms shows improved accuracy relative to realanalysis datasets while requiring only sparse, real-time observations.