SPATIAL COHERENCE AND DECAY OF WIND SPEED AND POWER IN THE NORTH-CENTRAL UNITED STATES

Hourly wind data from a network of climate stations in the north-central United States (drawn from the states of Illinois, lowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota, and Wisconsin) are analyzed to evaluate the efficacy of spatial analyses of near-surface wind speed and power. Spatial autocorrelation functions (acfs) were calculated at a number of timescales: annual, monthly, daily, and hourly. Annual wind speeds have virtually no coherent distance-decay relationship; monthly data produce a more consistent relationship, but still exhibit a large amount of scatter. Both daily and hourly data have classical decay with increasing distance between stations and there appears to be an optimal level of temporal aggregation, near the daily timescale, for spatial analysis of wind. In general, however, spatial acfs overestimate the spatial coherence of both wind speed and power. Temporal nonstationarities in wind data (i.e., diurnal and annual cycles) bias spatial autocorrelation functions and need to be removed before using spatial acfs to estimate characteristics of wind fields. Because mean absolute differences (MAD) of interstation wind speed and power are less affected by temporal nonstationarities, they produce more-robust representations of the spatial variability of wind speed and power. As a result, spatial MADs are recommended over spatial acfs for analyzing spatial coherence and decay of any spatial variable that contains nonstationarities. Methods for improving the spatial analysis of wind are discussed. Key words: wind energy, spatial autocorrelation, spatial analysis, nonstationarity, north-central United States.]