AbstractThis study presents the development of a method to statistically downscale daily wind speed variations in an extended Great Lakes region. A probabilistic approach is used, predicting a daily-varying probability density function (PDF) of local-scale daily wind speed conditioned on large-scale daily wind speed predictors. Advantages of a probabilistic method are that it provides realistic information on the variance and extremes in addition to information on the mean, it allows the autocorrelation of downscaled realizations to be tuned to match the autocorrelation of local-scale observations, and it allows flexibility in the use of the final downscaled product. Much attention is given to fitting the proper functional form of the PDF by investigating the observed local-scale wind speed distribution (predictand) as a function of the decile of the large-scale wind (predictor). It is found that the local-scale standard deviation and the local-scale shape parameter (from a gamma distribution) are nonconstant functions of the large-scale predictor. As such, a vector generalized linear model is developed to relate the large-scale and local-scale wind speeds. Maximum likelihood and cross validation are used to fit local-scale gamma distribution shape and scale parameters to the large-scale wind speed. The result is a daily-varying probability distribution of local-scale wind speed, conditioned on the large-scale wind speed.
New Approximate Bayesian Confidence Intervals for the Shape and the Scale Parameters of the Two Parameter Gamma Distribution
