The Autocorrelation of Hourly Wind Speed Observations

Abstract

The autocorrelation of hourly wind speed observations is estimated for seven stations on the west coast of Canada at selected lags ranging from one hour to two months. The estimated autocorrelation function is fitted by a model that includes a modified exponential term and cosine terms. The autocorrelation function declined more slowly than a simple exponential at six of the seven stations. For lags up to 12 hours autocorrelation coefficients were highest at those stations with the most uniform surrounding topography. The autocorrelation functions did not attenuate to zero at lags up to two months. An assumption of independent observations would underestimate standard errors of series statistics. Abstract The autocorrelation of hourly wind speed observations is estimated for seven stations on the west coast of Canada at selected lags ranging from one hour to two months. The estimated autocorrelation function is fitted by a model that includes a modified exponential term and cosine terms. The autocorrelation function declined more slowly than a simple exponential at six of the seven stations. For lags up to 12 hours autocorrelation coefficients were highest at those stations with the most uniform surrounding topography. The autocorrelation functions did not attenuate to zero at lags up to two months. An assumption of independent observations would underestimate standard errors of series statistics.