Final Technical Report The impact of climate change on the U.S. wind energy resource

Here we use the new North American Regional Climate Change Assessment Program climate projections to estimate the change of the wind power resource under various carbon dioxide loading s

Final Technical Report: The impact of climate change on the U.S wind energy resource

Applicant/Institution: University of Maryland

Street Address/City/State/Zip: 2133 Lee Building, College Park, MD 20742

Principal Investigator: Dr Daniel Kirk-Davidoff

Address: 3423 Computer and Space Sciences Building,

College Park, MD 20742

Telephone Number: 301-405-5386

Email: dankd@atmos.umd.edu

Co-Investigator: Dr Daniel Barrie, NOAA Climate Program Office,

Daniel.barrie@noaa.gov

DOE/Office of Science Program Office: Office of Biological and Environmental

Research

DOE/Office of Science Program Office Technical Contact: Mr Robert W Vallario

Abstract

The growing need for low-carbon emitting electricity sources has resulted in rapid growth

in the wind power industry The size and steadiness of the offshore wind resource has attracted growing investment in the planning of offshore wind turbine installations Decisions about the location and character of wind farms should be made with an eye not only to present but also future wind resource, which may change as increasing carbon dioxide forces reductions in the poleward temperature gradient, and thus potentially in the mean tropospheric westerly winds

Here we use the new North American Regional Climate Change Assessment Program climate projections to estimate the change of the wind power resource under various carbon dioxide loading scenarios and for a range of climate models We compare our assessment with both our assessment based on the IPCC AR4 model runs, to explore the extent to which improved model resolution changes the prediction for the wind power resource

Recent Results

Figures 1 and 2 below show the IPCC AR4 model ensemble prediction of wind power change over the first 50 and first 100 years of the 21st century, respectively in the

SRESA1B forcing scenario Wind power potential is assumed to be proportional to the third power of the wind, and the density is assumed constant in these figures Power is averaged over diurnal mean wind speed for a decade at the beginning and end of the period in question, and the percent increase or decrease of the later decade over the earlier decade is shown Only those seven models for which daily surface winds are archived are used (BCCR, CNRM, CSIRO, GFDL, MIROC medium resolution,

ECHAM2 and MRI) The modeled power changes are first re-gridded onto a common

grid (globally 128 by 64 grid points, about 2.8° resolution), and then averaged Thatched areas show regions where the standard deviation of the seven model results is smaller in magnitude than the model mean percent change in wind power For the United States, these results indicate model agreement on an increased wind resource over the south-central plains by the end of the century, a region with ample present wind resource and very active construction These results show somewhat better overall agreement than those found by Breslow and Sailor (2001), benefiting from the large set of model runs available in the WCRP CMIP3 archive

Coastal winds are shown to decrease by several percent, but this result is subject to substantial model disagreement, which focuses attention on our need to critically evaluate each model’s ability to represent the mean and variability of the surface winds

Figure 3 shows a first look at the NARRCAP data The figure shows the change in multiyear average daily maximum surface winds in the CRCM regional model forced by the CGCM3 GCM, both produced by the Canadian Center for Climate Modeling and Analysis The impact on increased resolution is immediately apparent, as there are substantial regions across the western United States where the mean wind speed is found

to increase, despite the general trend to reduced winds in that region found by both this high resolution model and the ensemble of IPCC AR4 models

Figure 1: Multimodel ensemble estimate of change in wind power resource (percentage change in monthly mean of cube of surface wind) from the decade 1990-2000 to the decade 2050-2060, across seven of the IPCC AR4 SRESA1B models runs (BCCR, CNRM, CSIRO, GFDL, MIROC (medium resolution), ECHAM2, and MRI) Models chosen are those which provided daily surface wind data on the WCRP CMIP3 data archive Colors in indicate the mean percentage increase, while thatching indicates that this mean change exceeds the standard deviation among the model simulations

Figure 2: As in Figure 1, for the difference between the decades 1990-2000 and 2090-2100.

Figure 3: Difference in 2066-2069 and 2038-2040 means of daily maximum wind in the CRCM regional climate model forced by the CGCM3 GCM, both models of the Canadian Center for Climate Modeling and Analysis, run as part of the NARCCAP program

Intermodel variability is the dominant feature of the CMIP3 and NARCCAP datasets Especially in the case of the GCMs run for CMIP3, large-scale models are typically not tuned to faithfully simulate surface wind speeds, but rather to accurately simulate

synoptic-scale features of the atmospheric flow The pathway to generating a 10 m surface wind field in a GCM is lengthy; the large-scale circulation drives the winds, but near the surface, the winds are also influenced by surface friction (which is incorporate in the boundary and surface layer routines of a GCM) By the time the 10 m wind field is generated, the winds have been influenced by a long string of model routines, so

intermodel differences are to be expected However, there are some points of agreement between the models

Most of the CMIP3 models and the NARCCAP 2060 projections agree that wind speeds and wind energy will increase in the south central United States, a region that is currently the focal point of domestic wind energy development (Figure 4 and Figure 5) Most of the increase in wind power was found to be related to the transient component, in contrast

to the projection that the transient component of wind speed is projected to decrease This finding is consistent with the definition of the transient component, which is related to passing synoptic systems that produce short-term high-speed events Because power is related to the cube of wind speed, a substantial portion of the integrated power at a particular wind farm is the result of this transient component, with the qualification that wind turbines are either operated at only their maximum capacity or not at all during high winds, a condition that was not considered here The power estimates presented here are higher than the actual power output that could be generated by turbines installed 10 m above the ground

Figure 4: CMIP multimodel wind speed anomaly average (top row) and standard deviation (bottom row) for the period 1990-2050 (left column) and 1990-2090 (right column) Contours are scaled by a factor of 10-2 m/s.

Figure 5: CMIP multimodel count of models that show an increase in wind speeds over the period

1990-2050 (top) and 1990-2090 (bottom).

In the literature (McCabe et al., 2001; Collins et al., 2005; Yin, 2005; Dommenget, 2009), climate change is generally expected to decrease storminess, shift the storm track

to the north, and increase high-speed wind events Each of these findings is consistent with the analysis carried out here Transient wind speeds are projected to decrease (due to decreased storminess), as seen in both the CMIP3 and NARCCAP results This finding,

in addition to the 500 hPa geostrophic wind analyzed for the NARCCAP CRCM-cgcm3 combination indicate a northward shift in the storm track Finally, extreme wind speed events are expected to increase This is best demonstrated by the large increases in wind power, particularly the transient component, that are simulated by the CMIP3 and

NARCCAP models (Figure 6)

Figure 6: Total (top row), transient (middle row), and stationary (bottom row) average wind power anomaly for 1990-2060 for the following NARCCAP simulations: CRCM-cgcm3 (left column), HRM3-hadcm3 (center column), and RCM3-cgcm3 (right column).

The large suite of CMIP3 models that are available for comparison somewhat offsets the substantial intermodel differences in the magnitude and spatial configuration of projected changes in wind speed In the comparatively small suite of currently available

NARCCAP results, regional patterns are better resolved, enabling a more detailed

discussion of climate change impacts on the wind resource that will be a better aid for wind farm developers and operators However, the three available model data sets

contained a wide range of projections, even when forced by the same GCM data; the extent to which RCMs in the NARCCAP experiment modified regional wind patterns with the same forcing conditions was surprising

An incidental result showed that the NARCCAP (50km) results contain model biases of 0.87 to 1.39 m/s when compared with NARR (32km) The biases are highest in regions of large topographic variability; however, there are also large biases associated with varying storm track locations and strengths (Figure 7)

Figure 7: Biases for each NARCCAP model compared to NARR data Average anomaly over the pictured domain as well as the average error are shown Wind speeds at 10 m are compared here The contours run from -3.2 to 3.2 m/s.

References:

Collins, M.: El Nino or La Nina-like climate change?, Clim Dynam., 24, 89-104, 2005

Dommenget, D.: The ocean’s role in continental climate variability and change, J Clim.,

22, 2939-4952, 2009

McCabe, G., Clark, M., and Serreze, M.: Trends in northern hemisphere surface cyclone frequency and intensity, J Clim., 14, 2763-2768 2001

Yin, J.: A consistent poleward shift of the storm tracks in simulations of 21st century climate, Geophys Res Lett., 32, L18701, 2005