A gridded wind speed observation product using artificial intelligence for Eastern Iberian Peninsula

Abstract: The demand for high spatially distributed wind speed data is substantially increasing during the last decades. This increase stems from its crucial role for climate change studies and many socioeconomic and environmental issues, e.g., wind power generation. However, observed wind speed records from weather stations do not cover this demand due to their coarse spatial resolution and inhomogeneous time scales, limiting the possibility of developing accurate gridded wind speed products using traditional geostatistical gridding methods. Moreover, wind speed from reanalyses and climate simulations does not accurately reproduce observed wind speed and gusts at regional scales. For instance, it lacks capturing the multidecadal variability of wind speed, e.g., the stilling (decline in winds) vs. the reversal (increase in winds) phenomena.

Artificial Intelligence is a powerful tool that can overcome these data availability and quality limitations of wind observations. In this study, we apply artificial intelligence to reconstruct wind speed data from in situ weather observations in the Eastern Iberian Peninsula, focusing on the Valencia region. The generated time series are then implemented to develop a high spatial resolution gridded wind speed observations at a regional scale. This new gridded wind speed dataset will allow computing, e.g., wind indices as a climate service for multiple socioeconomic and environmental sectors.

Keywords: wind speed; time series reconstruction; machine learning; gridded wind speed product