Projected Future Wind Speed and Wind Power Density Trends over the Western US High Plains ()
1. Introduction
With the growing need for different forms of energy outside of fossil fuels, it must be recognized that renewable energy resources will be more commonly used in the future. Unlike fossil fuels, however, renewable energy is closely affected by current environmental conditions. Solar energy, for example, is affected by the amount of sunlight or cloud cover in an area. Wind energy, of course, is susceptible to variations in wind speed. Wind statistics such as mean wind speed and gustiness are affected on a wide range of time scales. This can pose a problem for decision makers on the siting of new wind farms, as well as pose challenges to utility wind power users. Many wind power simulations and projections, such as the US Department of Energy’s 20% wind power plan by 2030, assume a constant or near constant wind resource in the future [1]. However, as this paper shows, wind velocities vary, and will continue to vary and change in the future.
This paper examines potential changes in wind velocities that could affect the wind power industry by comparing projected wind speed patterns from 2040-2070 and comparing them to historical data from 1970-2000. It is hypothesized that wind speeds will experience change in the future. This is because near surface wind speeds are linked to the location of temperature gradients, which will change as the planet warms. This paper examines these changes across the western high plains of the United States in an area known to be a prime location for wind farms. The effect of any wind speed changes in the amount of potential power generated by wind turbines is also investigated. Any change in near surface wind speed will cause an even greater change in wind power potential as the energy created by a turbine is proportional to the cube of the wind speed.
2. Background
Several recent studies have examined the impact of estimated future change in wind speeds (below is a brief summary-see [2-4] for detailed reviews). For exampleSailor et al. [5] researched potential climate change impacts on the wind energy industry for the Pacific Northwest. They found little to no consensus with respect to monthly wind speed changes between four General Circulation Models (GCM) models. However, there was agreement from all four of the GCM models that wind speeds will decrease throughout the year. This is especially true for the summer, where they found a 40% decrease on average in the wind power output throughout the region.
Breslow and Sailor [6] also examined potential changes across the US. They suggest that the majority of the U.S. will experience 1.4% to 4.5% decreases in wind speeds over the next 100 years. As with Sailor et al. [5], there was some inconsistency in the models examined. For example, for summer and fall, the Canadian Climate Center model suggested up to a 1 m/s or 15% decrease in wind speeds by the year 2075. This 15% wind speed decrease would have a potential reduction of 30% to 40% in the wind power output. However, the UK Hadley model showed no large or consistent change in the areas studied in the paper.
Other studies have attempted to use nested downscaling methods of determining future wind climate. Segal et al. [7] used a regional climate model (RegCM2) that had a horizontal grid resolution of 52 km nested within a UK Hadley GCM model. The results showed most of the country will experience some wind power seasonal decreases on the average of 10% - 20%. The most notable decrease occurred during summer along the costal areas of California with decreases around 30%. Other areas with significant areas of projected decreased wind power density included the north central United States. Many locations across the southern plains were found to have no significant changes. Only isolated areas of small increases were found across the northwest and far southern United States [7].
Across many locations across the United States, the overall yearly production of wind power occurs in just a few months of the year [2]. Many studies have looked at seasonal wind speed variability by looking at different large scale circulation patterns such as the positive and negative phases of the El Nino/Southern Oscillation Index [8,9], while others have researched more mesoscale patterns such as the low level jet during the spring months across the central plains states [10]. Across a large portion of the US, including the central plains states, wind power is at its peak during winter due to stronger pressure gradients associated with powerful weather systems and the changing location of the jet stream. The areas with the greatest increase in winter wind speeds are across the northern plains and midwest and throughout the northeast. In summer, the wind is mainly driven by smaller or mesoscale process such as topography or a nocturnal low level jet.
Any major shift in typical synoptic weather patterns will affect the wind industry. The majority of the studies discussed above project a slight decrease in wind speeds. However, there remain many uncertainties, such as low model resolution, model disagreement, and changing climate modeling technology. Under a warmer climate it is hypothesized that there will be a shift in wind speed patterns as winds are strongest where a large thermal gradient exists. With continued warming, the thermal gradient may shift further poleward, thus influencing the location and magnitude of wind speeds across the US Great Plains.
3. Data and Methodology
This paper examines the potential change in near surface wind speed and wind power densities across an area of the Central High Plains of the United States. The area of study is located in a region that is known to have some of the highest wind energy potential in the country. The area includes eastern Colorado, western Kansas, and western Nebraska. For this investigation, two sets of data from the North American Regional Climate Change Assessment Program (NARCCAP) will be used [11]. One set of data simulates past climate from 1970 through 2000. The other dataset predicts future climate from 2040 through 2070 based on the International Panel on Climate Change (IPCC) emissions Scenario A2 [12]. NARCCAP data is useful as it simulates climate at high resolutions needed for regional climate studies. NARCCAP looks to solve the uncertainties of regional scale future climate projections and produce higher resolution climate data than is currently available. The higher resolution modes are created by using regional climate models that are nested within the large-scale general circulation models forced with the A2 emission scenario [13]. Output from NARCCAP is climate data at 50 km grid resolution and three-hour temporal resolution across the study area (Figure 1). Each grid point represents the center latitude and longitude of the grid box.
As is the case with most wind speed datasets, the NARCCAP data do not exhibit a normal distribution and are right skewed. For this reason, the median is the better measure of central tendency and will be used to identify and describe any changes instead of the mean. Once all the medians were calculated for each set of data (monthly, seasonal, etc.) the percent change equation was used to determine the direction (negative or positive) of the trend as well as magnitude. The percent change equation is shown in Equation (1):
(1)
Figure 1. Study domain with the NARCCAP grid points.
One focus of this paper is how a changing wind climate may affect the wind power industry. A certain change in the wind speed does not necessarily illustrate how much change there may be in wind power. Much of the previous research thusfar has examined changes or trends in wind speed. Turning these trends into values that will have meaning to utility companies and policy makers is important. Thus, wind power density analysis, in addition to wind speed, will be shown. The wind power density equation is a useful way to evaluate a wind resource at a given site. Wind power density is measured in watts per square meter and is calculated using the following equation [14]:
(2)
where Pw is in watts per square meter, is the power of the wind, v is the wind speed, while ρ is the density of air, and n is the number of observations in the dataset. Density of air is a function of elevation. Using the common sea surface air density of 1.225 kg/m3 is not appropriate for this study as most of the area is well above sea level. An average elevation of 900m was chosen for the entire study area. Using the equation [14]:
(3)
the density of the air for the region will be an estimated 1.118 kg/m3, slightly lower than at sea level, which will decrease the overall the wind power density.
For a better understanding in trends in the amount of power output, the NARCCAP data must be vertically extrapolated to hub height. The NARCCAP wind speed data is at 10 meters. This paper assumes hub height to be at 80 meters, typical of large utility wind farms. The power law will be used for this extrapolation [15]:
(4)
where α is the power law index, and its value is based on the roughness of the surrounding terrain [16]. As the only available data is at one height (10 m) this project will use the “1/7 power law” where α will be estimated to be 1/7 [17], Uref is the reference wind speed at 10 meters (zref), and z is the desired height of 80 meters. It should be noted that this is an estimate, and there have been observed values of α measuring 0.15 to 0.25, which are higher than the 1/7 power law, at thirteen tall towers across the plains states [18]. With this in mind, this height extrapolation from 10 meters to 80 meters may be a slight underestimate for the study area, but will still give a good idea of the trends across the area. After extrapolating the NARCCAP data to 80 meter hub heights, wind power densities (Equation (2)) are calculated for the grid points and changes are analyzed. Three different model grid points were chosen for further analysis as they were relatively close to large utility wind farms currently in use. These points are shown in Figure 1. The three points are the following: 1) located 10 miles north of the Colorado Green wind farm, which is 10 miles south of Lamar, CO which has a 162 MW capacity, 2) a data point located 4 miles north of Logan/Peetz Table wind farm that has a 174 MW capacity and approximately 10 miles northeast of Peetz Table and Spring Canyon, which has a combined 700 MW of capacity, and 3) a data point in southwest Kansas near the Gray County wind farm with a 112 MW capacity.
4. Results and Discussion
This section investigates the projected difference in 10 meter wind speed and wind power densities for the NARCCAP time slice from the 2040-2070 compared to the data from 1970-2000. Any changes during this timeframe will be very important as there will likely be a much greater reliance on wind energy for electricity production in the future [1].
4.1. 1970-2000 to 2040-2070 Comparison
Analysis was first performed to compare the historic and future NARCCAP datasets. Figure 2 shows the percent change in median wind speed between the two datasets. There was little change in 10 meter wind speeds across the study area; however, there was a southeast to northwest gradient in the direction of the change in median wind speed. The majority of the region saw increases or decreases of less than 1% in median wind speed between the two time periods. Areas across south central Kansas are predicted to have 2% to 3% increases, while areas across central Colorado are predicted to have 1% to 2% decreases.
Figure 2. 1970-2000 to 2040-2070 percent change in 10 meter wind speeds.
The seasonal difference between the two datasets are seen in Figure 3, and quantified for the three wind farms described above in Table 1. During winter, the entire region will see a decrease of at least 3% and these decreases became greater further west with decreases of nearly 5% closer to the front range of the Rocky Mountains. During summer across the southeastern portion of the study area 4% to 5% increases are projected. During summer wind speeds are typically much weaker than other seasons so this percent increase may seem significant; however, when looking at wind power production it is not. Smaller changes were noted during spring and fall.
The overall changes become more apparent when looking at wind power density between the two timeslices (Figure 4). Wind power density at 80 meter was computed using the equations described above. Wind power density changes vary from 6 to 7 percent decreases in the far northwest portion of the study area to a 5 to 6 percent increase across the southeast portions of the region. Explanations for these decreasing values with further westward extent could be delayed development of storm systems due to baroclinic zones setting up further east. Another possibility is less winter precipitation or snowpack over the Rocky Mountains and southwest United States. With jet stream patterns developing due to temperature gradients, and with mountain snowpack playing a big role in these temperature gradients, more ridging could be possible across the western portion of the study area in the future [19]. The increases in the southeast portion
Figure 3. Seasonal change in median wind speed between 1970-2000 and 2040-2070.
Table 1. Percent change in overall and seasonal median wind speeds between past and future datasets at three studied wind farms.
Figure 4. 1970-2000 to 2040-2070 percent change in wind power density.
of the study area were mainly found during summer. This could possibly be due to an earlier transition to a fall weather pattern with more storm systems developing across the central plains later in the summer in the future.
Table 2 shows a comparison of the two datasets in terms of the monthly percent change for each month from the first decade from each dataset (e.g., the 1970s and 2040s) to the last (e.g., the 1990s and 2060s) for the Gray County, Kansas wind farm. It can be seen that there are slightly less positive values in the future (2040-2070) dataset than the past as in the future dataset only 8.3% of the months show a 10% to 20% positive changes and only 2.1% are in the 20% to 30% range compared to 17.1% and 4.5% in the past dataset. Both datasets have similar amounts for negative outliers. There are slightly more negative months in the future dataset as well, as the past had 46.4% of the months having a negative trend compared to 60.3% of the months seeing negative trends in the future dataset.