HWGEN: An hourly wind stochastic GENerator

Extreme wind speeds are vital for applications such as wind erosion assessment and wind power generation. However, their high spatial and temporal variability limits the accuracy of observations and models in capturing extremes. This study develops the Hourly Wind stochastic GENerator (HWGEN), a single-site stochastic model designed to improve the simulation of daily and hourly wind, particularly extreme events. HWGEN includes two versions: HWGEN_D, which simulates both wind speed and direction, and HWGEN_ND, which simulates only wind speed for data-sparse regions. Using hourly data from 388 stations in northwest China (2000–2020), HWGEN_D adaptively selects between Weibull and Gumbel distributions for daily wind speed and applies a lognormal model to simulate daily maximum hourly wind (U max ). Compared with WINDGEN, HWGEN_D reduces mean absolute percentage errors in wind erosion force and wind power density from 72.8% to 36.5% and from 35.1% to 8.1%, respectively. HWGEN_ND performs slightly worse but requires only 15% of the parameters. Moreover, HWGEN reproduces the observed relationship between daily maximum 10-minute wind speed (D max ) and U max , missing in existing models. HWGEN can stochastically generate infinitely long hourly wind sequences, improving the representation of extreme wind events and enhancing wind erosion and wind power risk assessments.