A novel wind speed modeling approach using atmospheric pressure observations and hidden Markov models

Modeling the wind speed data has important implications in wind studies, providing valuable insight and parametric quantities for further engineering analysis. The classical modeling approach is to fit the probability distribution to a known model and estimate statistical parameters like mean and variance. Such models lack the time variation properties and ignore cross-dependencies between other meteorological data. In this paper a procedure is developed to model the wind speed data using a dependent process of atmospheric pressure in the form of hidden Markov models (HMMs). Consequently, the inherent dependencies between the wind speed and pressure are exploited. HMMs relate the two quantities in a framework which eliminates the necessity of direct sample-wise correlations, and avoid direct time-series analysis complications of the stochastic wind speed data at a marginal expense of easy pressure measurements. The experimental data were obtained from recordings of hourly atmospheric pressure and wind speed values for two cities in Turkey, namely Izmir and Kayseri. Model details and numerical results are presented.