How Teleconnections Influence Wind Speed and Power Variability in the Upper Midwest through Changing Synoptic Regimes

Over the last few decades, wind power has received increasing interest as a renewable energy source and an economically competitive alternative to traditional fossil fuels. The increasing interest in and reliance on wind energy has led to the need for information on wind variability at multiple scales. Much attention has been given to sub-daily wind variations that affect turbine productivity, yet the 20 – 30 year lifetime of modern wind turbines necessitates information on long-term wind resources and their variations. Of particular interest are the teleconnections within the climate system which are known to affect climate variability on monthly to decadal timescales via changes in atmospheric circulation that impact wind productivity for extended periods of time. The goal of this study is to focus on how teleconnection patterns important to North America impact hub-height wind speed and power in the Upper Midwest (40 – 52°N, 87 – 105°W), a region with abundant wind resources and a fast-growing wind energy sector.

Data for this project were taken from reanalyses (ERA-Interim, CFSR and MERRA) for the period 1979/80 – 2018. The bulk of the analysis was done using the MERRA version 2 dataset, though the results are broadly consistent between models. Daily averaged 80-meter wind speed data were previously derived by extrapolating from 10-meter winds in each reanalysis and assessing modeled wind speed fidelity to several tall towers throughout the region. Wind power was calculated from the wind speeds using a standard wind turbine power curve. Surface pressure data were used to ascribe daily patterns into synoptic types according to the Lamb classification scheme, which categorizes the patterns into one of 26 flow regimes (10 purely directional, cyclonic or anticyclonic, or hybrids of those types). Teleconnections were represented through their monthly average indices as obtained from the NOAA Climate Prediction Center. While there are several teleconnections with potential to impact the Upper Midwest, preliminary analysis indicated that only the Arctic Oscillation (AO), the El Niño-Southern Oscillation (ENSO, represented here by the multivariate ENSO index), the Pacific North American pattern (PNA) and the Pacific Decadal Oscillation (PDO) had statistically significant impacts on wind speeds. Correlations between the regionally average monthly wind speed anomalies and the indices were 0.29, -0.21, -0.13, and -0.14 for the AO, ENSO, PNA and PDO respectively, with each exhibiting statistically significant regressions (p < 0.05) at the regionally-averaged and grid-cell scales. Monthly aggregated wind power was found to vary by 5 – 30% when regressed against the teleconnection indices.

To facilitate finding the impacts of the teleconnection patterns on wind speed through changes in synoptic regimes, the daily wind speeds were subset to produce Weibull distribution components (shape and scale) for each synoptic type in each meteorological season, yielding 104 Weibull distributions. To quantify the intra-type variability, the percentile of each daily wind speed within the Weibull distribution assigned to that day according to its synoptic type are calculated. Thus, each daily wind speed was represented by a shape, scale and percentile based on the synoptically-derived distributions. Daily scale parameters were used to calculate the monthly mean wind speeds and distributions based on the synoptic regimes, yielding a time series of the monthly anomalies caused by changes in synoptic type frequency. Daily percentile variations demonstrated significant spreads in in-type wind speeds, which was found to correlate significantly (0.58) with the strength of the pressure gradient, as approximated in this analysis through the daily difference between the maximum and minimum pressure within the region. The connection between the monthly averaged synoptically-driven wind speed anomalies and percentiles to the teleconnections were assessed using Pearson correlation and multiple linear regression of the monthly averages against the AO, ENSO, PNA and PDO. This was done spatially for each grid cell and for the regional average, yielding information on the collective impact of the teleconnections on wind speeds through changes in the synoptic regimes (synoptic-anomalies) and the relative strengths (percentiles). The impact of the teleconnections on the specific synoptic types was assessed through odds ratios as calculated by a logistic regression model against the teleconnection indices.

Synoptic regimes were shown to have distinct effects on wind speeds. Anticyclonic regimes tend to reduce wind speed (and by extension, wind power production), with anticyclonic centers within the region strongly associated with reduced flow. Directional and cyclonic regimes tend to produce higher wind speeds, though there is significant variation based on the direction of the flow, with southwesterly, westerly and northwesterly flow regimes producing the fastest wind speeds overall. Pure anticyclonic regimes are the most frequent synoptic type over the period of record (~18%), followed by pure cyclonic (~11%), while the collective bulk of days tend to fall into one of the directional regimes (46%).

The linear and logistic regression models yield indications as to the effects of the teleconnections on synoptically-driven aspects of the regional wind speeds. For example, the AO is shown to be positively related to both the synoptically driven anomalies and the percentiles, indicating that higher AO index values tend to produce fast wind speeds by changing the frequencies of certain synoptic types and their relative strengths. Positive phases of the AO reduce anticyclonic flow regimes while increasing the frequencies of directional types, particularly those with a westerly flow. On the other hand, ENSO is shown to have little impact on the frequency of synoptic-types but a significant negative effect on the percentiles, indicating that the effects of ENSO tend to occur primarily through a reduction in pressure gradients (in-type winds). Further, the compounded effects of the teleconnections have important impacts, with the most significant lulls in wind speed across the region (February 2010) being the result of a combination of a positive ENSO phase (El Niño) and a strong, negative AO event. This combination would tend to result in higher counts of anticyclonic types and weak pressure gradients, creating adverse conditions for wind energy suppliers across much of the region. Further research may help utilize this information, as well as increasing skill in predicting future teleconnection phases, to bolster wind resource projections.