|
Abstract:
|
Wind energy is considered one of the more promising renewable sources of energy that could curtail the dependency of several countries from oil while also mitigating global warming. To maximize the production of wind energy, it is fundamental to understand where to locate wind turbines. This requires an understanding of the spatial patterns of wind speed across the spatial domain of interest. As observations of wind speed are available only at few, isolated locations, researchers from both government and industries often rely on the output of computer models, such as the Weather and Research Forecasting (WRF) model. The latter, although provides simulated wind speed over the entire domain, suffers from issues of calibration. In this paper, we present various Bayesian hierarchical spatio-temporal models that aim to calibrate the output of max wind speed yielded by the WRF model. All the proposed models address several issues that characterize the distribution of max wind speed and the WRF output: namely, an inflated number of zeroes, extremal spatial dependence, temporal autocorrelation, and a non-linear relationship between the WRF output and the observed point-referenced data.
|
