Data Clustering Reveals Climate Impacts on Local Wind Phenomena

Abstract

AbstractThe authors demonstrate the utility of k-means clustering for identifying relationships between winds at turbine heights and climate oscillations, thereby developing a method suited for predicting the impacts of climate change on wind resources. Fourteen years of data from an 80-m tower at the National Wind Technology Center (NWTC) in Colorado have been reduced to four dominant flow phenomena using k-means clustering. At this location, this method identifies two clusters of westerly inflow (strong and weak), another cluster of flow from the north, and one of flow from the south. Similar clusters are found for the data at all heights on the tower, and each follow distinct seasonal cycles. Time series of each cluster, as well as the mean wind speed at the NWTC, are retained for comparison with climate oscillations along with the local 500-hPa pressure gradient. The mean wind speed in the surface layer is strongly correlated with the local north–south pressure gradient. The frequency of strong westerly flow is also negatively correlated with the Niño-3.4 index, whereas weaker westerly winds are negatively correlated with the Pacific–North American pattern (PNA) and Arctic Oscillation (AO). Northerly winds at the NWTC did not strongly correlate with any of the investigated climate indices (AO, PNA, and Niño-3.4). These northerly winds occur more frequently in the summer months, suggesting that these winds are more influenced by local conditions than by mesoscale forcing. This method of identifying clusters in wind data allows objective identification of wind phenomena that may benefit the deployment of wind turbines, for example, in choosing combinations of wind speed and direction to investigate for turbine siting.