Comparison of Spatial Interpolation Methods for Extreme Wind Speeds over Canada

Abstract

Specified wind speeds or their corresponding wind velocity pressures tabulated in structural design codes are spatially interpolated from the extreme value analysis results of wind records at meteorological stations. Commonly used spatial interpolation methods are compared in this study to determine their suitability to interpolate the surfaces or contour maps of the statistics of the annual maximum wind speed. For the analysis, multiyear wind records from 235 meteorological stations in Canada are considered. The interpolation methods include several deterministic methods and geostatistical methods [kriging (KO) and cokriging (Co-KO)]. To compare the performance of the methods, cross-validation statistical analysis was carried out for two approaches to estimate the T-year return period value of the wind speed: (1) directly interpolate it from the T-year return period value of the wind speeds at the meteorological stations, and (2) calculate it using the mean and coefficient of variation of the extreme wind speed interpolated from those at meteorological stations. Analysis results indicate that among the considered interpolation methods, the completely regularized spline, spline with tension, KO and Co-KO are the preferred methods. The first two lead to interpolated surfaces with more details; the last two result in relatively smooth interpolated surfaces.