You are on page 1of 17

REVIEWS

Climate change impacts on wind


power generation
Sara C. Pryor 1 ✉, Rebecca J. Barthelmie , Melissa S. Bukovsky
2 3
, L. Ruby Leung 4

and Koichi Sakaguchi 4


Abstract | Wind energy is a virtually carbon-​free and pollution-​free electricity source, with global
wind resources greatly exceeding electricity demand. Accordingly, the installed capacity of wind
turbines grew at an annualized rate of >20% from 2000 to 2019 and is projected to increase by a
further 50% by the end of 2023. In this Review, we describe the factors that dictate the wind resource
magnitude and variability and illustrate the tools and techniques that are being used to make
projections of wind resources and wind turbine operating conditions. Natural variability due to the
action of internal climate modes appears to dominate over global-​warming-​induced non-​stationarity
over most areas with large wind energy installations or potential. However, there is evidence for
increased wind energy resources by the end of the current century in northern Europe and the
US Southern Great Plains. New technology trends are changing the sensitivity of wind energy to
global climate non-​stationarity and, thus, present new challenges and opportunities for innovative
research. The evolution of climate modelling to increasingly address mesoscale processes is
providing improved projections of both wind resources and wind turbine operating conditions, and
will contribute to continued reductions in the levelized cost of energy from wind power generation.

The potentially exploitable wind energy resource exhib­ where CapEx describes capital expenditures; Fin, financ­
its marked spatial variability. Global estimates vary ing costs; O&M, levelized total annual operations and
from 70 to 3,050 EJ per year (or 19,400 to 840,000 TWh maintenance costs; AEP, revenues that are a function of
per year)1–5, according to assumptions regarding atmos­ the annual energy production; n, project lifetime; and r,
pheric conditions (for example, wind regime), techno­ discount rate10. LCoE values are higher for WTs deployed
logy (for example, type of wind turbines (WTs) deployed, offshore due to higher installation and transmission
WT installed density and electrical power production expenses, more complex and expensive foundations, and
efficiency), economic constraints and exclusions (that is, higher O&M costs10. The average global LCoE from wind
areas where development is prohibited). However, there energy is now ~$50 per MWh onshore and <$120 per
is consensus that the resource greatly exceeds both the MWh offshore8. It is projected to continue to decline
current global electricity demand of 26,700 TWh (ref.6) due to increased capacity factors (the ratio of electrical
1
Department of Earth and and the total primary energy supply of 585 EJ (ref.7), power produced to that if the WT constantly operated
Atmospheric Sciences, Cornell based on 2017 values. at the rated capacity), longer project lifetimes and reduc­
University, Ithaca, NY, USA.
Global installed capacity (IC) of WTs increased at a tions in O&M costs derived from enhancements to WT
2
Sibley School of Mechanical
mean annualized rate of ~26% from 2002 to 2018, and durability plus reduced financing costs10.
and Aerospace Engineering,
Cornell University, Ithaca,
19% over the period 2009–2018, reaching ~600 GW at the In 2017, WTs generated approximately 6% of global
NY, USA. end of 2018 (ref.8). WTs were deployed in 90 countries, electricity demand, including 44% of electricity sup­
3
National Center for 30 of which had an IC above 1 GW and 12 had an IC above ply in Denmark, 16.1% in Germany and 14.2% in the
Atmospheric Research, 4.8 GW (ref.8) (Fig. 1). Projections for 2019–2023 indicate UK11 (Fig. 1a). Thus, the wind power industry strongly
Boulder, CO, USA. additional global installations of over 300 GW (ref.8). influences the decarbonization of the global energy
4
Earth Systems Analysis & Reduction in the levelized cost of energy (LCoE)9 is system. WTs repay their lifecycle carbon burden within
Modeling, Pacific Northwest a primary driver behind the rapid expansion of the WT 3–7 months of operation12 and provide 20–30 years of
National Laboratory,
Richland, WA, USA.
IC, and is calculated as: virtually air-​pollution-​free and carbon-​free electricity
✉e-​mail: sp2279@ CapEx t + Fin t + O & M t supply1,13,14.
∑ nt =1 (1 + r )t Scenarios from the Global Wind Energy Council
cornell.edu
LCoE = Income(f (AEP))t
(1) (GWEC) 15 — including New, 450, Moderate and
https://doi.org/10.1038/
∑ nt =1
s43017-020-0101-7 (1 + r )t Advanced — indicate increases in wind energy IC from

Nature Reviews | Earth & Environment


Reviews

13 12
6 14 11
16
4.8
6.1 10
6.3 4.8 0.6
9

Wind speed (m s–1)


4 4.4 8
7
6
5
7.4
4
6.7
2.2 3
2
1
0

200 100 20 10

Installed capacity (GW)

b 600

500

Energy density (W m–2)


400

300

200

100

c
1.2
∆E

1
0.8
1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010
N. America: SGP N. Europe: Denmark S.E. India N. China
0.25
Energy density interannual variability

0.2

0.15

0.1

1.2
∆E

1
0.8 0.05
1980 1990 2000 2010
N.E. Brazil

www.nature.com/natrevearthenviron
Reviews

◀ Fig. 1 | Global wind resources, variability and wind turbines installed capacity. on the topic1,17,18. We begin by describing factors that
a | Annual mean wind speeds at 100 m above ground level taken from the Global Wind determine wind resources and operating conditions,
Atlas. Purple circles indicate the top 15 onshore installed capacity at the end of 2018 and guide WT selection. Then, we summarize how
(ref.8): China (207 GW), USA (96.6 GW), Germany (53.2 GW), India (35.1 GW), France
greenhouse-​gas-​induced climate change might impact
(15.3 GW), Brazil (14.7 GW), UK (13 GW), Canada (12.8 GW), Turkey (7.4 GW), Sweden
wind power generation and the LCoE of wind-​derived
(7.2 GW), Australia (5.4 GW), Mexico (4.9 GW), Japan (3.7 GW), South Africa (2.1 GW) and
Egypt (1.2 GW). The numbers indicate the fraction of national electricity supply that is
electricity via changes in wind resource magnitude, vari­
derived from wind turbines as of 2017 (ref.127). Data for Egypt are currently not available. ability and/or the conditions in which WTs will operate.
b | Mean annual energy density at 100 m above ground level over land areas from 1979 We end with a forward-​looking synthesis of prospects
to 2018, derived using output from the ERA5 reanalysis28,29. c | Interannual variability in for advancing knowledge in this arena.
energy density described using the standard deviation of annual values to the 40-​year
mean value. The insets depict time series of annual deviations (quantified as the annual Statement of needs
value divided by the long term mean) in five ERA5 grid cells located in regions with high Electricity production from wind. Scales of motion rele­
wind energy penetration. This figure illustrates the high spatiotemporal variability vant to wind resources span many orders of magnitude
in wind resources and the degree to which individual countries are harnessing the wind and vary in terms of their sensitivity to global climate
resource for electricity production. SGP, Southern Great Plains.
change (Table 1). The instantaneous power (P) in the
wind that can potentially be harnessed by a WT and
the baseline value of 433 GW in 2011 to 2,870–5,806 GW converted into electricity is a function of air density (ρ),
by 2050. Indeed, in the Advanced scenario, wind energy swept area of the blades (A) and cube of the wind speed
contributes 36% (15,258 TWh) of projected global elec­ (U) across the rotor plane19:
tricity demand in 2050 (ref.15) (Fig. 2a). Even this most
ambitious scenario requires a lower annual growth rate 1
P= ρAU 3 (2)
in global electricity generation from WT (<8%) than was 2
observed between 2000 and 2018, and remains below Wind speeds close to the surface typically increase
conservative estimates of the global resource. approximately exponentially with height (z) and
Avoided CO2 emissions — an assessment of climate also exhibit a dependence on atmospheric stability.
change mitigation potential — are similarly projected This height variation can be approximated using the
to increase in the future through expanded wind energy stability-​corrected logarithmic profile:
production. Using the approach described in ref.13,
u⁎   z  
GWEC projections can be incorporated within climate
U (z ) = ln   + Ψ  z   (3)
forcing (Representative Concentration Pathway (RCP))  
κ   z 0  
 L  


scenarios from the Intergovernmental Panel on Climate
Change (IPCC)16. The results indicate that, depend­ where u* is the friction velocity, κ is the von Karman con­
ing on the wind scenario used, avoided CO2 emissions stant (0.4), z0 is the surface roughness length (f(land use
increase from ~ 0.5 GtCO2 in 2011 to 4.5–9.2 GtCO2 in
L ()
land cover (LULC))) and Ψ z is the stability correc­
2050 (Fig. 2a). Cumulative avoided emissions from 2011 tion term (L is the stability parameter, Monin–Obukhov
to 2050 range from 88 to 154 GtCO2. Accordingly, WTs length) . Alternatively, the height dependence of wind
20

further help mitigate global temperature rise, the impact speeds can be approximated using the power law:
of which is dependent on the RCP scenario. None avoid α
passing the 1.5 °C threshold (Fig. 2b), but the Advanced z
U (z ) = U (r )   (4)
scenario delays exceeding the 2 °C threshold by up to  z r 
5 years, and, by 2050, reduces cumulative emissions
by ~154 GtCO2 (the equivalent of almost 5 years of where U(r) is the wind speed at the reference height (zr)
anthropogenic CO2 emissions) and, thus, offsets global and α is a coefficient often approximated as 1/7. It is
warming by approximately 0.1 °C. Assuming no fur­ also a function of stability conditions and LULC21. Both
ther expansion of wind energy IC so annual avoided approximations are only valid in the surface layer (lowest
CO2 emissions from wind energy do not increase from 10% of the atmospheric boundary layer, to heights of
their 2050 annual values, by 2100, wind energy offsets 50–200 m), where fluxes are constant with height. WT
approximately 0.3 °C of warming (Fig. 2b). dimensions are increasing so more of the rotor plane is
There are, thus, substantial climate mitigation bene­ extending beyond the surface layer22 to heights where
fits from wind energy expansion. However, wind energy wind speed profiles are no longer primarily dictated
is both a potential mechanism to reduce climate forcing by surface forcing, and the height dependence of wind
as well as a climate-​dependent energy source, so climatic speeds becomes more complex23,24.
changes may influence the conditions in which WTs The wind resource is often described using the mean
operate and the resource they are designed to harness. energy density (E in W m−2, over a given time period;
Further, the wind energy industry is rapidly evolving in j = 1, 2,…, n). It is also dependent on the cube of the
terms of the size and numbers of WTs being deployed incident wind speed (assuming ρ is constant)19:
and the locations where they are being deployed. These
industry changes may amplify the risks or opportunities 11 n 3
E= ρ∑U (5)
posed by climate change. n2 1
In this Review, we outline the reciprocal relation­
ships between anthropogenic forcing of climate and Actual electrical power production as a function of
wind-​derived electricity, updating previous reviews incident wind speed is described by a WT-​specific power

Nature Reviews | Earth & Environment


Reviews

a b
40 5

Cumulative world CO2 emissions (GtCO2)


8,000

Wind energy contribution to


30 Scenarios 2050 4

ΔT cf. 1861–1880 (°C)


world electricity (%)
6,000
20 3

Cumulative avoided 4,000


10 emissions from wind 2
energy 2050 (GtCO2)
3 103 154
88 116 2,000
0 1
0 2,000 4,000 6,000 2020 2040 2060 2080 2100
Wind energy IC (GW) Year
2015 New 450 New 450 Moderate Advanced (filled)
Moderate Advanced RCP2.5 RCP4.5 RCP6.0 RCP8.5

Fig. 2 | Projections of wind energy installations and associated climate change mitigation potential. a | Wind energy
installed capacity (IC) as of the end of 2015 and percent contribution to world electricity supply, along with projections
of wind energy IC and contribution to world electricity in 2050 for four Global Wind Energy Council (GWEC) scenarios15:
New, 450, Moderate and Advanced. Cumulative avoided emissions from 2006 to 2015 or 2050 are depicted in the legend.
b | Representative Concentration Pathway (RCP) scenarios presented in terms of cumulative global CO2 emissions and
associated temperature change16 (solid lines and solid circles). Also shown are the results of reduced cumulative CO2
emissions from 2011 onwards derived from application of the four GWEC wind scenarios to the RCP. To avoid double-​
counting, wind energy from each RCP scenario is removed and replaced by the GWEC scenarios under the assumption
that wind-​generated electricity displaces an average carbon-​based generation source that has an equivalent CO2
emission rate of 600 g per kWh following the approach used in ref.16. Avoided annual emissions 2050–2100 are estimated
by assuming wind energy IC does not increase beyond 2050 and are 8.5, 5.9, 5.2 and 4.3 GtCO2 per year for the Advanced,
Moderate, 450 and New scenarios, respectively. This figure illustrates that following an aggressive expansion of wind
energy could reduce warming by 2100 by up to 0.3 °C.

curve. Typically, WTs cut in (commence electrical power distribution is often used to derive long return-​period
production) at wind speeds ~4 m s−1, power production values27.
increases approximately linearly with increasing wind The non-​linear dependence on wind speed means
speed until ~12–15 m s−1 to the rated (or nameplate) that 90% of electrical power production is typically pro­
capacity and remains constant for wind speeds up to duced by the upper half of the probability distribution
~25 m s−1, after which the WT is shut down to protect of wind speeds, and up to 25% of electricity is generated
it from damage17. A simple approximation of gross from the highest 10% of wind speeds17. Thus, develop­
expected AEP derived using a WT power curve is: ing accurate projections of wind resources requires high
fidelity of climate projections, particularly in the upper
Inf
portions of the wind speed probability distribution that
AEP = T ⋅ f ∫ p(U )P(U )dU (6)
dominate power production.
0

where T is the length of one year, f is frequency, p(U) Energy density spatial variability. Figure 1 shows a
is the probability distribution of hub-​height wind global overview of the mean annual mean energy den­
speeds, and P(U) is the WT power curve that relates sity at 100 m above ground level (a.g.l.) and its inter­
incident wind speed to power production. The viabil­ annual variability (IAV), derived using output from
ity of potential WT deployments is often characterized the ERA5 high-​resolution (approximately 30 × 30 km)
using P50(AEP) (median AEP) and P90(AEP), which reanaly­sis for 1979–2018 (refs28,29). It shows good spatial
is the AEP for which there is a 10% chance of not being agreement with mean wind speeds in the Global Wind
reached or exceeded25. Atlas and indicates that high wind energy potential is
Most wind speed time series conform to a apparent in regions with substantial current penetration,
two-​parameter Weibull distribution26: such as the Central Plains of the USA, Canada, north­
ern Europe, north China and Australia, and in countries
k  U  k−1   U  k that do not yet currently exhibit substantial wind energy
p(U ) =   exp −   (7)
A  A    A  
  deployments, such as Argentina and Chile30,31.
The IAV of the mean annual energy density and
where A is the scale parameter and k is the shape param­ expected AEP are critical to determining the ‘risk’
eter. These wind speeds have an exponential tail, leading associated with a wind project. Lower IAV represents
to the accumulated probability of extreme wind speeds lower risk and is associated with lower cost of capital25
conforming to a double exponential. Hence, a Gumbel and, hence, lower LCoE. The IAV of energy density is

www.nature.com/natrevearthenviron
Reviews

comparatively low in high-​resource regions of the USA sustained wind speed and turbulence intensity) are
(such as the Southern Great Plains) and northern China, used to describe fatigue and extreme mechanical loads,
moderate in high-​resource areas of northern Europe and and guide the selection of an appropriate class of WT
south-​western India, and high in north-​eastern Brazil for a given location. Engineering requirements, and,
(Fig. 1c). Time series of annual energy density for five thus, WT purchase price and O&M costs, increase as
high wind energy penetration subdomains emphasize these external conditions become increasingly harsh.
regional differences in both the relative magnitude of External conditions and WT wind classes are described
IAV and the duration of relatively low or high energy using design standards articulated by the International
density periods due to spatial variations in the dominant Electrotechnical Commission (IEC 61400-1 (ref.36) or
internal climate models and their inherent periodicity 61400-3-1 (ref.37)) and naturally contribute to the pro­
(Fig. 1c; Table 2). jected LCoE. Specific power is the ratio of its nameplate-​
Most regions with high wind resources are located in capacity rating to its rotor-​swept area and is used as an
the mid-​latitudes close to the major storm tracks (Fig. 1). optimization parameter for selecting the WT type and
The dominance of synoptic and mesoscale pheno­mena dimensions for a given wind climate and location38.
(Table 1) in dictating regional-​s cale wind resources Generally, WTs with higher specific power will be more
(and variability therein) has important implications for economic at higher wind speed locations39.
grid-​integration of wind-​derived electricity and for cli­ WTs operate across a wide range of environmental
mate change research. It means there are inherent spa­ conditions, experiencing both fatigue (normal) and
tial scales of coherence on timescales of a few hours to extreme (ultimate) mechanical loading on various
days32 and geographic separation over which wind energy components. Fatigue loading on WTs (and the result­
density variations are decoupled or inversely coupled33. ing mechanical stress) is primarily a function of local
Hence, it may be possible to balance load and supply of wind conditions40, and is, therefore, likely to have a low
electricity from wind across a region (such as Europe) if sensitivity to global climate change (Table 1). However,
the grid is sufficiently strong and the electricity markets extreme loading cases associated with extreme wind
are integrated and harmonized34,35. Further, it is the basis phenomena may differ under climate change as a result
of the a priori assumption that global climate change will of, for example, changes in the tracking and intensity of
likely influence wind resources primarily through changes synoptic-​scale and mesoscale atmospheric phenomena,
in larger scales of atmospheric motion (mesoscale to plan­ such as tropical cyclones41 and polar lows42.
etary scale) that can be — at least partly — resolved in Maintaining WT blade aerodynamic efficiency is key
current-​generation and future-​generation climate models. to maximizing electrical power production. This effi­
ciency can be reduced by ice accretion, which also causes
WT design and operating conditions. During the uneven mechanical loads, resulting in increased O&M
design phase of wind power plants, estimates of opera­ costs43. Measures to reduce ice accumulation (coat­
ting (or external) conditions including a range of wind ing or heating of the WT blades) add to costs and/or
parameters (such as wind gusts, 50-​year return period reduce power production and, thus, increase the LCoE44.

Table 1 | Scales of atmospheric motion relevant to the wind energy industry


inherent scale in Spatial scale temporal scale relevance to wind Climate change
atmosphere energya sensitivity
Micro δ-​γ: fine-​scale 3 mm to 30 m <1 h Loading, materials, Low
turbulence operating conditions
Micro β: surface-​layer 30–300 m Minutes to hours Resource, loading, Low
turbulence (strong diurnal design, wakes, operating
variability) conditions, control
Micro α: boundary-​layer 300 m to 3 km Hours (strong Resource, loading, Moderate where flows
turbulence and flows diurnal variability) design, wakes, operating are generated by thermal
conditions, control gradients
Mesoscale (γ-​α) 3–700 km Hours to days Resource, wakes, Moderate where flows
forecasting, operating are generated by thermal
conditions, maintenance gradients (LULC) or
in regions with strong
mesoscale phenomena
(such as tropical cyclones)
Macro β: synoptic scale 700–4,000 km Days to weeks Resource, forecasting, High
operating conditions
Macro α: planetary 4,000–40,000 km Weeks to decades Resource, operating High
scale (including internal conditions
climate modes)
LULC, land use land cover. aControl = operation of WT or wind farms in specific modes to optimize performance; design = WT
dimensions and construction; forecasting = short-​term (hours to days) estimates of future power production (and variability
therein); loading = mechanical forces on wind turbines (WTs); materials = impact damage and wear; operating conditions = external
conditions such as wind gusts, icing, extreme wind speeds; resource = energy density; wakes = flow within and downwind of WT and
WT arrays (implications for power loss and loading).

Nature Reviews | Earth & Environment


Reviews

Table 2 | Précis of research on iav in WS or wind resources in the contemporary climate


region variablea Data set used magnitude of iav and sign of dependence ref.
on climate modes
North America
Eastern USA AEP RCM (12 km) 9 of 10 years AEP ±6% of long-​term mean. 25

in reanalysis IAV(AEP) = f(PNA/NAO/ENSO)


CONUS WS GCM Northern USA: WS+ when ENSO+ North-​east: WS+ when 67

PNA+Rest of CONUS: WS+ when ENSO−, WS+ when AO+,


WS+ when PNA−
CONUS WS In situ Pacific NW = f(PNA). Midwest = f(PNA, ENSO). 70

North-​east = f(NAO). South-​west = f(ENSO).


Central Plains = f(PNA, ENSO). South-​east = f(ENSO)
CONUS Monthly mean WS (80 m) Reanalysis Western USA: IAV(WS) up to 40% from modulated annual 156

cycle. Great Plains: WS+ when ENSO−


Western USA WS Reanalysis f(ENSO) moderated by sea-​surface temperature anomalies 157

in tropical Pacific
High Plains of central USA E RCM Lower values (~5%) in the 1970s and 1990s than in the 1980s 158

Minnesota Mean monthly WS (70 m) In situ 6–15% of variability explained by ENSO and AO. 159

No dependence on PNA
Colorado WS (80 m) In situ, one site Strong westerly flow+ when ENSO− Weaker westerly 160

winds+ when PNA− and AO−


Canada Mean monthly WS In situ f(ENSO) 161

South America
Brazil Mean monthly, seasonal In situ f(ENSO). El Niño WS+, La Niña WS− 162

and annual WS
Southern S. America WS (50 m) Reanalysis AAO+: WS−. ENSO influence weaker. AAO and ENSO variance 163

over timescales up to decades


Chile WS, CF RCM in reanalysis f(ENSO). El Niño WS−, La Niña WS+ 164

Europe
Europe Normalized AEP GCM ensemble Current IAV(AEP) dominated by NAO, exceeds projected 165

changes
Europe Normalized AEP Reanalysis Winter AEP (Finland, Norway, England and Germany)+ when 166

NAO+. Converse in S. Europe


Europe E (80 m) Reanalysis E factor of three variability f(NAO and AO). NAO+: E+ in north 167

and WPD− in south


South-​western Europe WS RCM in reanalysis NAO−: WS+ by 10–15% 168

Europe WS Reanalysis and GCM NAO+: WS+ north of 45°N, WS− south of 45°N 33

UK WS and AEP f(NAO) In situ (two sites) NAO phase = ±10% variations in mean AEP 169

UK AEP In situ IAV(AEP) = ±20%. ‘Wind drought’ in 2010 170

Norway WS Reanalysis NAO+ enhances wintertime strong winds 171

Scotland AEP, CF Reanalysis and NAO+1σ = CF + 2.6 in winter 172

in situ
Asia and Australasia
Malaysia WS (10 m and 60 m) In situ at three sites La Niña: WS+. El Niño: WS− 173

India CF (50 m) Reanalysis Intensification of monsoon: CF+ 174

AAO, Antarctic Oscillation; AEP, annual energy production; AO, Arctic Oscillation; CF, capacity factor; CONUS, contiguous USA; E, energy density; ENSO,
El Niño–Southern Oscillation; GCM, global climate model; IAV, interannual variability; NAO, North Atlantic Oscillation; PNA, Pacific–North American teleconnection
pattern; RCM, regional climate model; WPD, wind power density; WS, wind speed. aUnless otherwise stated, the estimates of wind speed or energy density are for a
nominal height of 10 m above ground level.

Robust estimates of icing probability over the entire pro­ specifically, the occurrence of hail and/or extremely
ject lifetime are essential to evaluating the need for and heavy rainfall45. Leading-​edge protection can be applied
benefits of icing mitigation. but may degrade aerodynamic performance46. Thus,
WT blade damage due to material stresses from pre­ there is a need to carefully assess the relative benefits
cipitation impacts also decreases aerodynamic efficiency in reducing O&M costs versus the AEP loss. In many
and power production and increases O&M costs45. The areas with substantial WT IC, such as the US Southern
timescale to damage is determined by the hydroclimate, Great Plains, hail and heavy downpours are generated

www.nature.com/natrevearthenviron
Reviews

by deep convection at the mesoscale47. Therefore, cor­ decades)25,61,62 associated with internal modes of climate
rect characterization of current and possible future variability33,63,64 (Table 2).
leading-​edge erosion potential requires accuracy of For example, the North Atlantic Oscillation (NAO)
climate simulations at that scale. expresses variance at frequencies from sub-​annual
Other mesoscale phenomena of interest and impor­ to multi-​decadal, with the latter (up to timescales of
tance to WT resource intermittency and operating con­ 60 years) being modulated by North Atlantic sea-​surface
ditions include open-​cell structures generated in the temperatures65 and the Pacific Decadal Oscillation66.
presence of large temperature gradients, such as cold-​air The NAO plays a central role in dictating extratropical
outbreaks in coastal regions such as the North Sea48. cyclone tracks and, thus, wind resources over Europe,
They are associated with strong gusty winds and local particularly during winter (Table 2). For instance, wind
increases in wave height49, and have a scale that makes energy density over northern Europe exhibited high
them relevant for power fluctuations from offshore wind values (+20% above the long-​term mean) during the
farm clusters and an intensity that makes them relevant 1980s and early 1990s due, in part, to the prevalence of
for extreme wind loading50. positive-​phase NAO33 (Fig. 1c).
Foundations represent a substantial fraction of the In North America, by contrast, wind resources
installation cost of offshore WTs10. Foundation type is exhibit region-​specific responses to three dominant
largely dictated by water depth, while design and cost are climate modes67: the Pacific–North American tele­
dictated by hydrodynamic and structural analyses based connection pattern (PNA) and the El Niño–Southern
on projected wind-​wave loading51. Significant wave Oscillation (ENSO) in the west; ENSO in the Central
height (Hs, the mean of the highest one-​third of waves Plains and south-​east; and the PNA and NAO in the
measured from trough to crest) and peak period are north-​east (Table 2) . El Niño events, for example,
key aspects of design conditions. These parameters are typically associated with lower wind speeds in the
vary greatly in space, in part due to bathymetry and, western USA and La Niña with higher wind resources.
also, atmospheric forcing. For example, the 100-​year However, it is important to note that responses to differ­
return-​period Hs is substantially higher off the west coast ent phases of ENSO (as well as other climate modes) are
of Scotland (~20–22 m) than in the North Sea (~14 m) not directly opposite, and that mode–mode interactions
or the eastern seaboard of the USA52, resulting in differ­ also cause variability67.
ent extreme loading estimates and design for offshore ENSO also appears to be an important control on
WTs53. Sea ice (amount, thickness, location, drift) and low-​frequency variability in wind regimes over South
sea currents (and resulting scour) also contribute to America and Asia (Table 2), although less research has
mechanical loading on offshore WTs and foundation been conducted to assess these teleconnections. In
design, and may exhibit climate change sensitivity17. locations where the resource is influenced by monsoon
In the following, we summarize the challenges to wind speeds (such as India), there is both pronounced
undertaking assessments of how and whether anthro­ seasonal reversal of wind direction and IAV linked to
pogenic climate change may influence wind power monsoonal intensity and production of associated
generation through changes in wind resources and transitory, synoptic-​scale phenomena68.
WT operating conditions, describe tools that have The presence of low-​frequency variability greatly
been employed and summarize the findings of those confounds the ability to identify and assign causes of
assessments. long-​term tendencies in wind resources. For example,
low-​frequency variability led, in part, to claims of global
Climate impacts on wind power generation stilling. An analysis of 822 in situ stations reported wind
Challenges. A number of inherent challenges confront speeds at 10 m a.g.l. declined by 5–15% over almost
efforts to quantify how global climate non-​stationarity all continental areas in the northern mid-​latitudes
might influence the magnitude of global wind resources. between 1979 and 2008 (ref.69). However, analysis of
They provide critical context for climate projections. short-​duration time series and use of linear fitting might
In contrast to air temperature16 and total precipitable have led to low-​frequency variability being misinter­
water54, it is unknown whether anthropogenic warm­ preted as ‘trends’ and trend magnitudes being inflated
ing will result in stilling (decreases in wind speed) or by neglecting the presence of temporal autocorrelation70.
increased windiness at either the regional or the global Further, in situ time series are subject to discontinui­
scales. Reductions in the equator-​to-​pole large-​scale ties caused by changes in observational systems71 and
temperature gradient will likely modify tropical cir­ localized changes in LULC around surface observing
culation patterns (Hadley cell, monsoon circulations stations. Thus, detailed diagnostic research has found
and/or tropical cyclone frequency)55 and the behaviour that a large fraction (25–60%) of wind speed declines
of mid-​latitude jet streams and storm tracks56 and, hence, was attributable to increased drag69,72, rather than global
cyclone frequency57, intensity58 and tracking59. Enhanced climate change.
atmospheric water vapour and convective instability There has been a recent ‘recovery’ of wind speeds
may further impact the intensity of some mesoscale in both individual regions (for example, in China)73
to synoptic-​scale systems60. However, detection and and across the Northern Hemisphere74. While global
attribution of changes to mesoscale to synoptic-​scale annual mean wind speeds at 10 m height between the
systems, as well as the resultant wind resources due to 1980s and early 2000s exhibit a negative linear trend
global climate non-​stationarity, is extremely challenging, of <−0.1 m s−1 per decade, the trend has since reversed
given the presence of low-​frequency variability (years to and increases of >0.2 m s−1 per decade were reported for

Nature Reviews | Earth & Environment


Reviews

2010–2017 (ref.74). Consistent with the variability of dom­ analysis, in situ data indicate declines in 90th percen­
inant internal climate mode across the continents (Fig. 1c; tile wind speeds at 10 m a.g.l. between 1973 and 2005
Table 2), the timing of the linear trend ‘reversal’ varies over substantial areas of the eastern USA, but output
by continent. The inflexion point occurs in 2003 over from multiple reanalyses and RCMs indicate no trend71.
Europe, 2010 in North America and 2001 over Asia74. Conversely, analyses of output from the North American
The scales on which wind resources vary are below Regional Reanalysis identified positive trends in wind
that manifest in most current-​generation global climate speeds at 80 m at the annual timescale and during winter
models (GCMs), which typically have horizontal grid and spring over most of the contiguous US (CONUS)
cells of ~100 × 100 km and comparatively low verti­ during 1979–2009; these linear trends have magnitudes
cal resolution. Key phenomena such as extratropical of ~0.6 m s−1 per year in the west, 0.12 m s−1 per year in
cyclones have historically been rather poorly repre­ central regions and 0.9 m s−1 per year in the east86.
sented in GCMs75,76, and changes in mid-​latitude storm Research based on data from Australia also found
tracks exhibit a high sensitivity to the precise global and a clear discrepancy between output from reanalysis
nested regional climate models (RCMs) used in their products and in situ observations. In situ wind speed
simulation77. Thus, although it may be possible to infer measurements at 10 m a.g.l. declined from 1979 to 2001
some information about changes in circulation patterns (mean trend of −0.013 m s−1 per year), and while reanal­
at the synoptic scale (and above), direct output from ysis data sets also showed declines, they were a factor of
global or regional models with resolutions lower than 5–7 smaller in magnitude87.
~50 km is of low utility to the assessment of potential Magnitudes of historical trends also exhibit a very
changes to wind resources. Hence, it is necessary to strong dependence on the precise period of data, owing
downscale using either dynamical (application of lim­ to the presence of low-​frequency modes of variability.
ited area, regional, models) or statistical methods61,78 For example, both in situ data and reanalysis output over
(Table 3). Application of downscaling approaches to the Tibetan Plateau exhibit evidence of declining annual
wind climates has demonstrated discrepancies in skill mean wind speed at 10 m from the mid-1980s to the
and strongly varying differential credibility for different early 2000s, followed by a subsequent recovery between
RCMs in describing wind resources in the contempo­ 2005 and 2015 (ref.73). Similar analyses over China indi­
rary climate79,80. In general, greater credence should be cate declining annual mean wind speed at 10 m from
attributed to wind resource results from such approaches 1970 to the early 2000s, with near-​constant values in
and to simulations conducted at higher resolution. 2001–2007 (ref.88). These analyses, thus, re-​emphasize
There is demonstrable value added by application of the importance of internal climate variability to the
higher-​resolution RCMs over the use of GCMs81,82 interannual and interdecadal variability of wind speeds
and/or lower-​resolution regional models83.This improve­ and wind resources (Table 2).
ment in fidelity and credibility is particularly marked in Nevertheless, regarding projected future changes in
higher moments of the probability distribution (such as wind speeds and corresponding impacts on energy den­
extreme wind speeds and gusts), where there is a strong sity, there is evidence of winners and losers at the regional
response to model resolution84. scale (Table 3). For example, the majority of research
Most wind speed observations are taken at heights at focused on Europe indicates that wind resources in
or below 10 m a.g.l. and, thus, are strongly influenced by the north and over the UK may show small-​magnitude
local surface roughness, obstacles and topography. WTs, increases (<10%) over the current century89–92, while
by contrast, have rotor planes that sweep across a layer those in the regions of lower current resource poten­
of the atmosphere extending from ~40 to 160 m a.g.l. tial to the south (such as the Mediterranean) may
with hub heights of ~100 m. The lack of in situ wind slightly decline91,93 (Table 3). In North America, there is
speed data sets at 100 m a.g.l. against which model fidel­ weaker evidence, but an evolving consensus, that wind
ity can be assessed85, as well as discrepancies between resources might decline by up to 5% in the mean annual
reanalyses28,71, constrain the ability to assign reliabil­ energy density over much of the western USA18,82. In the
ity to model-​based projections of wind resources and Southern Great Plains, by contrast, it is anticipated
operating conditions. that energy density may increase by up to 5–10% by
mid-​century (2050)94,95. Studies over Brazil similarly
Climate change and wind resources. Despite the chal­ project large-​magnitude increases in the mean wind
lenges to developing actionable projections of wind resource over regions with current WT deployments96,97.
resources, there is a clear imperative to understand Moreover, there is evidence that resource seasonality
how the evolution of the global climate system has might amplify — increasing in winter and declining
impacted, and may continue to impact, both the sup­ in summer98.
ply and demand components of the energy system. Many studies have considered climate projection peri­
There is marked spatial bias in the volume of research ods at the end of the current century (Table 3). However,
that has been conducted on these topics, with the over­ this timescale is inconsistent with the primary time hori­
whelming majority of analyses examining Europe and zons relevant to the wind energy industry, which focus
North America. on project lifetimes of approximately 30 years (refs10,99).
Research focused on the historical and contempo­ Differences in the mean wind resource or wind cli­
rary climate over North America has illustrated large mates during future periods (even those at the end of
divergence in wind speed trends computed using the 21st century) versus those in the contemporary cli­
in situ data, reanalyses and output from RCMs. In one mate appear to be of small magnitude (generally <5% of

www.nature.com/natrevearthenviron
Reviews

Table 3 | Précis of research projecting wind resources for the coming decades
region variable models/method Projected change time period ref.
(and resolution)
North America
CONUS E 1GCM, 1RCM (50 km) <2% lower in south-​west; 3% higher (2041–2062) – (1979–2000) 18

in Central Plains
CONUS E 2GCM, 3RCM (50 km) No emergence from natural variability (2041–2062) – (1979–2000 80

CONUS WS (90th 4GCM, 5RCM (50 km) No change in Central Plains. Lower over (2041–2062) – (1979–2000 82

percentile) western USA


CONUS WS 5GCM (1–3°) Up to 5–10% increase in winter; declines (2079–2099) – (1979–1999) 98

in summer
CONUS Seasonal mean WS 1GCM, 1RCM (50 km) Remains within ±10% of contemporary (2040–2069) – (1985–2005) 175

at hub height climate


Eastern USA WS (90th 3GCM, 3RCM (50 km) Up to −5% (2041–2060) – (1981–1998) 100

percentile) plus statistical


downscaling
Eastern USA AEP 1GCM, 1RCM (12 km) Up to +8% in Southern Plains; up to −5% (2075–2099) – (1980–2005) 94

in Northern Plains
S. California Annual mean WS 12 RCM–GCM Within ±2% (2051–2071) – (1980–2000) 176

combinations (50 km)


High Plains of E (80 m) 1GCM, 1RCM (50 km) 2–3% increase in Kansas; 1–2% decrease (2040–2070) – (1970–2000) 95

central USA in central Colorado


N. America AEP 10GCM (1–2°) Increased AEP over Mexico. Declines (2020–2040) – (1980–2005) 177

over western USA. Large model-​to-​model and (2020–2040) – (1980–2005)


divergence
Western E, WS 1GCM, 1RCM (45 km) Change < σ (ensemble model mean) (2031–2060) – (1971–2000) 178

Canada
Offshore areas E (50 m) 1GCM (25 km) Declines exceed IAV in current climate (25 years end of the 21st 179

around Mexico century) – (1985–2011)


South America
Brazil WS (100 m) 1GCM, 1RCM Increases in high-​resource areas (2010–2040), (2041–2070) 96

(20 km)a and (2071–2100)


Brazil E (100 m) 3GCM, 1RCM (20 km) Increases of up to +50% in some wind (2021–2050), (1961–1990) 97

farm locations and (2070–2099) – (1961–1990)


Europe
Europe E (80 m) 1GCM, 2RCM Increases in northern and central Europe (2061–2100) – (1961–2000) 89

(up to 15%). Declines over southern Europe.


Largest changes in winter
Europe E 21GCM (1.125–2.8°) Increases <30% over Baltic region. Declines (2081–2100) – (1986–2005) 102

over western Europe. No change in IAV


Europe E (80 m) 22GCM, 1RCM Increases in ensemble mean over northern (2021–2060) and (2061–2100) 91

(25 km)a and central Europe (2061–2100). Decreases – (1961–2000)


over southern Europe in both periods
Europe E (HH) 6GCM, 10RCM 15-​member ensemble exhibits changes (2071–2100) – (1971–2000) 93

(25 km) of between −20% and +20%. Changes


aggregated over EU <2%
Europe Normalized AEP 5GCM (36 Near-​future spatially varying differences. (2020–2049) – (1979–2005) 180

(78 m) realizations) Ensemble mean change: of between −12%


(1.4–2.8°) and +8%
Northern E 1GCM, 2RCM (25 km) Increases over Denmark and Baltic Sea (2070–2099) and (2036–2065) 90

Europe – (1961–1990)
Northern E 1GCM statistical E small declines (<5%) but remains within (2046–2065) and (2081–2100) 101

Europe downscalinga historical IAV – (1961–1990)


UK WS, E, CF (80 m) 1GCM (2 × 2.5°)a Small increase in IAV. +2% increase in CF (2071–2090) – (1981–2000) 92

in Scotland
Germany E (80 m) 1GCM statistical No change at annual scale. Increases in (2061–2100) – (1961–2000) 181

downscalinga winter (<6%). Declines in summer (<4%)


Black Sea E (120 m) 5GCM, 1RCM No change in resource or seasonality (2061–2090) – (1979–2004) 182

(12 km)a

Nature Reviews | Earth & Environment


Reviews

Table 3 (cont.) | Précis of research projecting wind resources for the coming decades
region variable models/method Projected change time period ref.
(and resolution)
Asia and Australia
China WS 9GCM (0.56–2.8°)a No trend. Large model-​to-​model variability. (2006–2100) v. (1850–2005)b 183

No change in IAV
China WS 1GCM two Lower autumn wind speeds (difference of (2018–2100) v. (1961–2005)b 184

resolutions 0.12–0.18 m s−1) at the end of the century


(110–280 km)a
Australia AEP (80 m) 4GCM, 1RCM (50 km) Large spatial variability. Declines in (2060–2079) – (1990–2009) 185

south (<1%)
Japan AEP (80 m) 1GCM, 1RCM Declines (<6%) over Japan. Increases over (2051–2100) – (1951–2010) 186

(ensemble) South Korea (~3%)


Offshore WS 1GCM (20 km) WS lower off west and south coasts. (2075–2099) – (1979–2003) 187

waters in India Increases off south-​east coast (near current


WT deployments)
Most studies used a single climate forcing scenario. Climate forcing applied in the projections is described using the Representative Concentration Pathway (RCP) for
CMIP5 generation, global climate model (GCM) simulations or Special Report on Emissions Scenarios (SRES) for CMIP3. AEP, annual energy production; CF, capacity
factor; CONUS, contiguous USA; E, energy density; HH, hub height; IAV, interannual variability; RCM, regional climate model; WS, wind speed; WT, wind turbine.
a
Denotes studies with multiple RCP/SRES. Unless otherwise stated, the estimates of wind speed or energy density are for a nominal height of 10 m above ground level
(a.g.l.). bStudies where the time periods are specified with ‘v.’ are those that focused on trends rather than differences between two time periods.

the current mean energy density) and generally do not (MPI-​ESM-​LR104) for RCP8.5. Both project areas with
exceed current interannual/interdecadal variability from lower mean future resource in the north and north-​west,
internal climate modes (Tables 2,3). Further, most studies consistent with previous research (Table 3). There are
use a time windowing approach to examine changes in also some differences; notably, WRF output suggests
wind resources and employ time periods that are rela­ an area of statistically significant higher resource at the
tively short compared to the scales of low-​frequency end of the century over the Southern Great Plains (in an
internal climate variability. The continued presence area of high wind energy penetration). This feature is
of this low-​f requency variability may be aliased as present in earlier simulations with regional models
global-​warming-​related changes in wind resources. nested in a range of GCMs80,105 (Table 3) but is absent
Irrespective of the region under study, there is a from the MPAS simulation (Fig. 3a,b).
very large dependence of climate change projections of A further important source of uncertainty in both cli­
wind resources derived using either RCMs or statistical mate and reanalysis models is the representation of drag
methods on the global model providing lateral bound­ by unresolved features such as orography106 (Fig. 3c). This
ary conditions (LBCs) and/or predictors80,100,101. Sources stress is typically formulated in terms of subgrid-​scale
of projection variability, and, thus, uncertainty, listed height variability (in this case, at a horizontal resolution
in order of importance are: the global model used to ~1 km), but is the subject of considerable uncertainty.
provide the LBCs to limited area (regional) models or Changing stress parameters have a profound impact
predictors to statistical downscaling or from which data on the resulting mean energy density and the degree of
are used directly, specific limited area (regional) model agreement with the spatial patterns of other resource
applied, resolution and formulation of the model(s), estimates (Figs1b,3a,c). There is, therefore, a need to
and specific climate forcing scenario101,102. There is evi­ refine the manner in which this stress is applied in order
dence of larger changes in mean energy density in the not to compromise representation of flow in the lower
higher climate forcing scenarios, but only a relatively boundary layer and, hence, wind resources.
weak dependence on this parameter91, consistent with a
continuing dominance of internal climate variability in Climate change and operating conditions. Compre­
determining IAV of the wind resource. hensive assessment of possible impacts from global cli­
Figure 3 illustrates the magnitude of deviations in mate non-​stationarity on wind power generation must
wind resources under the contemporary climate that also incorporate improved characterization of the cli­
arise due to differences in model resolution (12 km mate sensitivity of key external operating conditions.
versus 25 km) or formulation. It also shows that there However, there is a relative paucity of research into how
are broad similarities in mean energy density during global climate change may impact phenomena at the
1980–2005 and some regions where there is agreement mesoscale to the microscale due, in part, to the difficulty
in the sign of difference in energy density in the cli­ and computational cost of representing them.
mate projection period. These estimates derive from Distinguishing climate change signals for extreme
simulations at 12 km using the Weather Research and wind phenomena is rendered challenging by limited
Forecasting (WRF) regional model and at 25 km using model fidelity18,107,108, uncertainty in extreme-​value
the global variable-​resolution Model for Prediction estimation109 and high temporal variability of rare
Across Scales (MPAS)103 over CONUS. Both simula­ events110. They are, thus, dependent on the precise data
tions use bounding information from the Max Planck set, methodology and data period used to derive them
Institute Earth System Model at Low Resolution and the dynamical causes18,90,101,111. Projections from

www.nature.com/natrevearthenviron
Reviews

a multi-​model GCM ensemble indicate reductions in intensity, at least in some regions, may be increased in a
annual maximum wind speed in 2074–2100 relative globally warmed world. For example, simulations of wind
to 1979–2005 over much of the USA112. In Northern gusts by multiple members from a single RCM using
Europe, 50-​year return-​period wind speeds increase an entrainment-​based approach indicate evidence for
beyond the historical envelope of variability only by the increases of up to +10% in extreme wind gusts over parts
end of the 21st century90. of northern Europe by the end of the current century90. An
Wind gusts are an important driving phenomenon analysis using a large ensemble of models to statistically
of extreme mechanical loads on WTs. However, they are downscale sustained and gust wind speeds also projected
probabilistic, not deterministic, features parameterized an increased frequency of wind gusts in excess of 70 km h−1
in regional models on the basis of downwards entrain­ over parts of Canada under climate non-​stationarity114.
ment of higher momentum air from aloft or in terms of Occurrence of deep convection, although challenging
the turbulent kinetic energy budget113. Despite relatively in both weather-​forecasting and climate-​projection con­
high uncertainty, there is some evidence that wind gust texts115, is relevant as a source of both extreme operating

a Mean energy density (W m–2) Δ Mean energy density (W m–2)


600 60

500 40

400 20
WRF

300 0

200 –20

100 –40

0 –60
b (W m–2) (W m–2)
300 60

40
MPAS

200 20

100 –20

–40

0 –60

c Mean energy density — terrain drag (W m–2) Mean energy density — no terrain drag (W m–2)
600 600

500 500
MPAS

400 400

300 300

200 200

100 100

0 0

Fig. 3 | Contemporary and projected mean annual energy density. a | Mean annual energy density at ~100 m above
ground level for 1980–2005 (left panel) and the difference between 2075–2099 and 1980–2005 (right panel). Results are
derived using Weather Research and Forecasting (WRF) simulations at 12-​km resolution within lateral boundary conditions
from the Max Planck Institute Earth System Model at Low Resolution94. b | As in panel a, but mean annual energy density at
~60 m derived using simulations at 25-​km resolution from the global variable-​resolution Model for Prediction Across Scales
(MPAS). Simulations are conducted using sea-​surface temperature and sea-​ice cover from ERA-​Interim for 1990–2010,
and with the change simulated by the Max Planck Institute Earth System Model at Low Resolution added to ERA-​Interim
sea-​surface temperature and sea-​ice data (2079–2099). The MPAS is configured with a climate-​oriented land model188
and physics parameterizations189, as tested by ref.190. On the right of panels a and b, only grid cells where the difference
in mean energy density in the two periods passes a two-​sided t-​test are shown191. Results of individual tests are corrected
for the false discovery rate with α = 0.05 (ref.191). c | Energy density computed with the MPAS for 2008 with (left panel) and
without (right panel) terrain drag. This figure illustrates some of the key challenges in making robust assessments of wind
resources in the contemporary climate and discrepancies in wind resources in the contemporary and possible future
climate that derive from different model formulations.

Nature Reviews | Earth & Environment


Reviews

wind gusts116 and hail, the latter of which is a major driver zero (Table 3). In Europe, there is some evidence for
of WT blade leading-​edge erosion45. According to some emerging consensus that the mean annual energy density
estimates, deep convection is likely to become more fre­ will increase in the north (for example, over Denmark
quent in regions with high WT IC densities, such as the and the UK) and slightly decrease over the south (includ­
Southern Great Plains in the USA117. ing the Mediterranean). In North America, there is some
Mechanisms of ice accretion and the atmospheric evidence of regional declines in the wind resource over
processes responsible for ice accumulation on WT blades much of the western USA, but increases in the region of
are complex and difficult to predict, even with short lead highest penetration (the Southern Great Plains).
times43. Further, icing frequency exhibits high IAV in
the contemporary climate118, rendering trend detection Industry trends. Projections indicate enhanced longevity
and attribution to global climate change very difficult. of wind farm projects. Wind power plants being con­
Nevertheless, one analysis indicated that surface icing structed today are anticipated to have lifetimes beyond
over Northern China (including the region of highest historical expectations of 20 years to nearly 30 years99.
WT IC) decreased over the period 1961–2015, with an There is also an increased prevalence of repowering126
accelerating trend to 4.8 days per decade over the last two (where WTs are replaced by newer-​generation and
decades119. This change is largely attributable to increas­ large-​capacity WTs). Use of repowering implies a degree
ing near-​surface night-​time temperatures119. Empirical of entrenchment in areas of traditionally high resource
approximations based on climate model output have suggesting value added from assessment of medium-​
also indicated declining probability of icing over most term (up to 40 years) resource stability. One of the key
of Scandinavia by the end of the current century120, and disjuncts that has hampered efforts to supply wind energy
over much of CONUS, except near the Great Lakes79. stakeholder-​relevant science is that climate science typi­
These projections may imply shifts in the cost–benefit cally focuses on long-​term changes (Table 3). Even with
analysis for de-​icing technologies, but more mechanisti­ the evolution towards enhanced project lifetimes, for
cally advanced methods are needed to provide actionable future changes in resource and/or operating conditions
information to the wind energy industry. to be of actionable importance to the wind energy indus­
At high latitudes, most climate projections indicate try, they must either occur on timescales that mean the
amplification of global warming and resultant degra­ resource at currently operating or planned wind farms
dation of permafrost121. These changes in permafrost will change or be of sufficient magnitude that the industry
conditions may have profound impacts on both WT should consider migrating from regions with historically
foundation design122 and road construction and repair high wind density to those that are projected to have the
for wind farm installation and maintenance123. highest resource in the future. Both require that climate
Changes in extreme wind-​wave loading on offshore projections have higher fidelity than is currently observed.
WTs associated with tropical cyclones may occur in Projected expansion of WT IC and evolution of new
some regions where major WT deployments are pro­ wind energy technologies is amplifying the imperative
posed. There has been only very limited work on this for research into how climate change may impact wind
topic and the results are likely to be highly regionally power generation and affords new opportunities for
specific. For example, one study suggested that the stakeholder-​relevant geoscience research.
20-​year return-​period Hs in the North Atlantic under Projections for global IC in 2040 are 2,400–3,320 GW
RCP8.5 will increase from 15.7 m (1970–1999) to 17.3 m and 342–562 GW for onshore and offshore facilities,
in 2071–2100, and from 18.5 to 20.3 m for the 100-​year respectively, a large increase from the 2017 values of 591
return period, although the mean Hs is higher in the and 23 GW (ref.8). Much of the expansion may occur in
historical period124. In a separate study, extreme Hs in non-​traditional areas such as forested, complex terrain,
the western tropical Pacific was projected to decrease which poses particular challenges to accurate model­
(by ~10%) under RCP8.5 by the end of the century125. ling of the wind climate and, thus, wind resources and
operating conditions111. For example, 69% of Sweden,
Summary and future perspectives a country with large resource potential, is forested127.
Summary. Wind energy is an increasingly important source To harness the power of higher wind speeds aloft
of clean, renewable electricity. Installed capacity is rapidly and increase swept area, higher hub heights and longer
expanding. Following an aggressive projected growth blades are being employed, and the rated capacity
scenario of IC, passing the 2 °C warming threshold can of WTs continues to increase22,128. The average rated
be delayed by up to 5 years, and cumulative emissions capacity of WTs installed in the USA during 2017 was
can be reduced by 154 GtCO2 (the equivalent of almost 2 MW (ref.129), while in Europe, it was 2.7 MW onshore
5 years of anthropogenic CO2 emissions) by 2050 (Fig. 2). and 6.8 MW offshore130,131. US projections for 2030 are
Global climate non-​stationarity is likely to change 4.5 MW onshore and 10 MW offshore132,133. Mean WT
the resources WTs harness and/or the conditions hub heights are projected to increase from 2017 values
in which they operate, at least in some regions of the of 86 m in the USA to 102 m in 2030 and rotor diame­
world. Research to date has largely focused on potential ters are expected to increase from a mean of 113 m in
changes in regions with high WT penetration. Generally, 2017 to 167 m by 2030 (refs132,133). The implied increase
any change in regional wind resources over the course in rotor-​tip speeds may increase rates of leading-​edge
of the current century is smaller than contemporary IAV erosion, adding to the imperative to better quantify
(Tables 2,3). The uncertainty envelope encompassing drivers of precipitation-​induced blade damage in the
projections made using different model tools includes contemporary and future climate.

www.nature.com/natrevearthenviron
Reviews

Growth of the physical dimensions of WTs coupled improve simulation of operating conditions to compre­
with expansion into complex terrain onshore and deploy­ hensively assess the climate change sensitivity of wind
ments offshore mean that WTs are no longer operating power generation.
in the atmospheric surface layer. Many of the studies Some of these needs are consistent with evolving foci
summarized in Table 3 vertically extrapolate modelled and initiatives within climate science. There is an increas­
wind speeds from a nominal height of 10 m to WT hub ing focus on long-​term very-​high-​to-​high-​resolution,
heights (80 m or above), most frequently using the sim­ convection permitting, simulations within regional cli­
ple power law with an exponent of 1/7. The continued mate science initiatives115,143. Use of a kilometre-​scale
physical growth of WTs is rendering use of these types of resolution will likely improve representation of the mean
approximations even less appropriate22. Advances in the flow84 and should improve representation of phenomena
representation of atmospheric boundary-​layer dynamics such as derechos and the resulting wind gusts and extreme
are needed to capture the first-​order and second-​order precipitation (including hail) and, thus, WT blade ero­
moments of the flow and better characterize wind sion conditions47,144. Improved simulations of such
profiles23, including the occurrence of low-​level jets24, pheno­mena that are responsible for extreme and fatigue
and to project how such phenomena may change as a mechanical stresses and how those may evolve under
result of both climate evolution and changing LULC. global climate non-​stationarity will reduce uncertainty
Major climate science research initiatives need to for the wind energy industry. It will be instrumental in
archive output at WT-​relevant heights (~100 m a.g.l.) at enabling production of carbon-​free electricity at the low­
high spatiotemporal resolution to negate the need for est achievable LCoE, by avoiding suboptimal siting, costly
post-​processing vertical extrapolation of wind speeds over-​engineering of WT arrays, unacceptably high rates of
and, thus, decrease associated uncertainty. structural failures or underperformance of wind projects.
Expansion of WT deployments offshore and expansion There is increasing emphasis on assessing the fidel­
into deeper water will require improved estimates of joint ity of climate projections, and a recognition that models
wind-​wave loading134 and innovations in foundations135 need to exhibit both statistical skill in the contemporary
or floating technologies136, particularly in areas prone to climate with respect to some target data sets and that
intense mesoscale to synoptic-​scale phenomena such as they are ‘getting the right answer for the right reason’94,145.
tropical cyclones that have potential for very high struc­ Enhanced model credibility will help develop and convey
tural loading137. There is, thus, a the need for improved actionable projections to the wind energy industry.
coupling of atmospheric and oceanic models for charac­ GCMs with 25–50-km horizontal grid spacings are
terization of current and projected wind resources, WT being included in CMIP6 under HighResMIP146 and
design properties and operating conditions138,139. those with <5-​km grid spacing are being actively devel­
WTs are being deployed in larger clusters and across oped and tested using month-​long simulations147. Global
larger areas. This clustering will increase the probabil­ models with variable-​resolution capability (such as MPAS,
ity of array–array interactions and illustrates the need Fig. 3) afford important new opportunities, since they
for improved characterization of whole-​wind-​farm can employ mesoscale and even convection-​permitting
wakes both onshore128,140 and offshore141. Given the resolutions in the refined domain while using coarser
propagation of these wakes is a function of atmos­ (30–100-​km) grid spacings over the rest of the globe148.
pheric stability and wind direction140, analyses designed High-​resolution coupled wind-​w ave RCMs are
to optimize system-​wide power production should be increasingly able to simulate offshore winds and waves
conducted under the contemporary and possible future and to accurately depict the influence of complex orog­
climate states. raphy. For example, in the Pacific coast and the Salish
Increased penetration of this intermittent resource Sea bordered by Washington State, USA and British
also requires improvements in short-​term forecasts of Columbia, Canada, surface winds are much better sim­
power production 142 and improved understanding ulated by a regional model at 6-​km grid spacing com­
of the scales of coherence of wind fluctuations in the pared to a reanalysis product at ~25-​km grid spacing149.
contemporary and possible future climate. Simulated wave height driven by the 6-​km regional simu­
lation also exhibits enhanced fidelity (Fig. 4). This research
Recommendations for future research. A number of demonstrates feasibility for improved wind-​wave loading
outstanding research questions remain before projec­ on offshore WTs in the contemporary and possible future
tions of future changes in wind climates can be taken climate. Such coupling may also enhance the fidelity of
into account when planning future WT deployments. modelled extreme wind speeds138.
Further, there needs to be greater emphasis placed on Increasingly large ensembles of high-​resolution
quantifying the fidelity of projections of wind resources climate simulations are likely to better sample uncer­
and operating conditions. Access to data from operating tainties related to internal variability and model struc­
wind farms would greatly benefit efforts to evaluate and tural uncertainty150. There remains a need for carefully
improve our modelling of wind resources and address curated subsets of GCMs and reanalyses to create an
these challenges and other research questions85. ensemble that encompasses known uncertainties and
There is a need to better understand drivers and mag­ minimizes the use of LBCs that contain biases that
nitudes of resource variability at interannual to inter­ decrease the credibility of regional wind simulations.
decadal timescales in order to contextualize projected Improved, higher-​resolution simulations from GCMs
changes and enable attribution and detection studies. are expected to reduce this problem151, and the availabil­
There is also a need to undertake research designed to ity of GCM large ensembles will facilitate examination of

Nature Reviews | Earth & Environment


Reviews

9
46131
5

Modelled Hs (m)
4 8
50°N
3
2 7
1
0 6
0 1 2 3 4 5
Observed Hs (m)
49°N 5

Hs
46146
4 4
Modelled Hs (m)

3
2 3
48°N
1
2
0
0 1 2 3 4
Observed Hs (m) 1

46087 47°N 0
10 126°W 125°W 124°W 123°W 122°W
Modelled Hs (m)

8 46041 46088
6 12 4
10
Modelled Hs (m)

Modelled Hs (m)
4 3
8
2
6 2
0 4
0 2 4 6 8 10 1
Observed Hs (m) 2
0 0
0 2 4 6 8 10 12 0 1 2 3 4
Observed Hs (m) Observed Hs (m)

Fig. 4 | Simulations of Hs from a coupled regional climate model. Spatial distribution of significant wave height (Hs)
simulated by a wave model during a storm event at 10:00 on 13 December 2015. This storm event brought 9-​m waves to
the entrance of the Strait of Juan de Fuca near buoy 46087. Wave height is drastically reduced east of this location and is
strongly influenced by the local winds modulated by the complex orography. Buoy locations are shown with black and
white dots, with corresponding scatter plots of simulated and observed Hs at those locations for a 5-​year period between
2011 and 2015. The wave model was driven by winds simulated by a regional model at 6-​km grid spacing. This figure
illustrates the importance of topographic forcing on Hs close to coastlines and model skill for Hs. Adapted with permission
from ref.149, Elsevier.

uncertainties associated with natural variability152. New Improvements in wind climate and mesoscale pro­
research using regional refinement of global models jections will not only benefit the wind energy industry
(as such the MPAS) may offer solutions to discontinu­ but also a wide array of stakeholders from different
ities arising during application of RCMs within LBCs socio-​economic sectors (such as the insurance indus­
from offline GCMs153,154. try and the transportation sector). An increased
Continued collaboration with the integrated assess­ focus on characterizing and improving model fidelity
ment modelling community will also allow additional with respect to atmospheric flow will also fundamen­
uncertainties related to future changes in LULC under tally advance climate science by improving represen­
different socio-​economic scenarios to be incorporated tation of the surface energy balance and moisture
in RCMs155 and these may help inform constraints on, advection.
for example, land areas available for expansion of wind
energy IC128. Published online xx xx xxxx

1. Wiser, R. et al. in IPCC Special Report on Renewable 4. Possner, A. & Caldeira, K. Geophysical potential for 9. Veers, P. et al. Grand challenges in the science of wind
Energy Sources and Climate Change Mitigation wind energy over the open oceans. Proc. Natl Acad. energy. Science 366, eaau2027 (2019).
(eds Edenhofer, O. et al.) 535 (Cambridge Univ. Sci. USA 114, 11338–11343 (2017). Emphasizes important trends within the wind
Press, 2012). 5. Jung, C., Schindler, D. & Laible, J. National and global energy industry and highlights resulting key
A comprehensive analysis of the technical potential wind resource assessment under six wind turbine research avenues.
wind resource, the role of wind energy in climate installation scenarios. Energy Convers. Manag. 156, 10. Wiser, R. et al. Expert elicitation survey on
change mitigation and wind energy technologies. 403–415 (2018). future wind energy costs. Nat. Energy 1, 16135
2. Eurek, K. et al. An improved global wind resource 6. International Energy Agency. Global energy & CO2 (2016).
estimate for integrated assessment models. status (IEA, 2019). 11. Global Wind Energy Council. Global wind report:
Energy Econ. 64, 552–567 (2017). 7. International Energy Agency. Key world energy annual market update 2017 (GWEC, 2018).
3. Marvel, K., Kravitz, B. & Caldeira, K. Geophysical statistics 2019 (IEA, 2019). 12. Smoucha, E. A., Fitzpatrick, K., Buckingham, S. &
limits to global wind power. Nat. Clim. Change 3, 8. Global Wind Energy Council. Global wind report 2018 Knox, O. G. Life cycle analysis of the embodied carbon
118–121 (2013). (GWEC, 2019). emissions from 14 wind turbines with rated powers

www.nature.com/natrevearthenviron
Reviews

between 50KW and 3.4Mw. J. Fundam. Renew. Energy 37. International Electrotechnical Commission. IEC and wind energy potentials over Central Europe.
Appl. 6, 1000211 (2016). 61400-3-1:2019. Wind energy generation systems. Tellus A 68, 29199 (2016).
13. Barthelmie, R. J. & Pryor, S. C. Potential contribution Part 3-1: design requirements for fixed offshore wind 65. Higuchi, K., Huang, J. & Shabbar, A. A wavelet
of wind energy to climate change mitigation. turbines (IEC, 2019). characterization of the North Atlantic oscillation
Nat. Clim. Change 4, 684–688 (2014). 38. Johansson, V. et al. Value of wind power–implications variation and its relationship to the North Atlantic sea
Quantifies the potential role of wind energy from specific power. Energy 126, 352–360 (2017). surface temperature. Int. J. Climatol. 19, 1119–1129
to climate change mitigation. 39. Jackson, K., Van Dam, C. & Yen-​Nakafuji, D. (1999).
14. Wiser, R. et al. Long-​term implications of sustained Wind turbine generator trends for site-​specific 66. Schwing, F. B., Jiang, J. & Mendelssohn, R. Coherency
wind power growth in the United States: Potential tailoring. Wind Energy 8, 443–455 (2005). of multi-​scale abrupt changes between the NAO, NPI,
benefits and secondary impacts. Appl. Energy 179, 40. Hansen, M. O. L. Aerodynamics of Wind Turbines and PDO. Geophys. Res. Lett. 30, 1406 (2003).
146–158 (2016). (Routledge, 2015). 67. Schoof, J. T. & Pryor, S. C. Assessing the fidelity
15. Global Wind Energy Council. Global wind energy 41. Knutson, T. et al. Tropical cyclones and climate of AOGCM-​simulated relationships between large-​
outlook 2016 (GWEC, 2016). change assessment: Part I: Detection and attribution. scale modes of climate variability and wind speeds.
16. Intergovernmental Panel on Climate Change. Bull. Am. Meteorol. Soc. 100, 1987–2007 (2019). J. Geophys. Res. 119, 9719–9734 (2014).
Climate Change 2013. The physical science basis 42. Romero, R. & Emanuel, K. Climate change and 68. Sandeep, S., Ajayamohan, R., Boos, W. R., Sabin, T.
(Cambridge Univ. Press, 2013). hurricane-​like extratropical cyclones: projections for & Praveen, V. Decline and poleward shift in Indian
17. Pryor, S. C. & Barthelmie, R. J. Climate change North Atlantic polar lows and medicanes based on summer monsoon synoptic activity in a warming
impacts on wind energy: A review. Renew. Sustain. CMIP5 models. J. Clim. 30, 279–299 (2017). climate. Proc. Natl Acad. Sci. USA 115, 2681–2686
Energy Rev. 14, 430–437 (2010). 43. Davis, N., Hahmann, A. N., Clausen, N. E. & Zagar, M. (2018).
A comprehensive review on the topic of climate Forecast of icing events at a wind farm in Sweden. 69. Vautard, R., Cattiaux, J., Yiou, P., Thepaut, J. N. &
change impacts on the wind energy industry. J. Appl. Meteorol. Climatol. 53, 262–281 (2014). Ciais, P. Northern hemisphere atmospheric stilling
18. Pryor, S. C. & Barthelmie, R. J. Assessing the 44. Sabatier, J., Lanusse, P., Feytout, B. & Gracia, S. partly attributed to an increase in surface roughness.
vulnerability of wind energy to climate change and in Informatics in Control, Automation and Robotics Nat. Geosci. 3, 756–761 (2010).
extreme events. Clim. Change 121, 79–91 (2013). Vol. 495 (eds Gusikhin, O. & Madani, K.) 641–663 70. Pryor, S. C. & Ledolter, J. Addendum to “Wind speed
Addresses climate change impacts on the conditions (Springer, 2020). trends over the contiguous United States”. J. Geophys.
in which WTs operate. 45. Letson, F. W., Barthelmie, R. J. & Pryor, S. C. Res. Atmos. 115, D10103 (2010).
19. Manwell, J. F., McGowan, J. G. & Rogers, A. L. RADAR-​derived precipitation climatology for wind 71. Pryor, S. C. et al. Wind speed trends over the
Wind Energy Explained: Theory, Design and turbine blade leading edge erosion. Wind Energy Sci. contiguous United States. J. Geophys. Res. Atmos.
Application (Wiley, 2010). 5, 331–347 (2020). 114, D14105 (2009).
20. Stull, R. B. An Introduction to Boundary Layer 46. Herring, R., Dyer, K., Martin, F. & Ward, C. 72. Wu, J., Zha, J. L., Zhao, D. M. & Yang, Q. D.
Meteorology (Springer, 2012). The increasing importance of leading edge erosion Changes in terrestrial near-​surface wind speed and
21. Irwin, J. S. A theoretical variation of the wind profile and a review of existing protection solutions. their possible causes: an overview. Clim. Dyn. 51,
power-​law exponent as a function of surface roughness Renew. Sustain. Energy Rev. 115, 109382 (2019). 2039–2078 (2018).
and stability. Atmos. Environ. 13, 191–194 (1979). 47. Letson, F., Shepherd, T. J., Barthelmie, R. J. & 73. Chen, L., Pryor, S. C., Wang, H. & Zhang, R.
22. Barthelmie, R. J., Shepherd, T. J., Aird, J. A. & Pryor, S. C. WRF modelling of deep convection and Distribution and variation of the surface sensible
Pryor, S. C. Power and wind shear implications of hail for wind power applications. J. App. Meteorol. heat flux over the central and eastern Tibetan
large wind turbine scenarios in the US Central Plains. Climatol. https://doi.org/10.1175/JAMC-​D-20-0033.1 Plateau: comparison of station observations and
Energies 13, 4269 (2020). (2020). multireanalysis products. J. Geophys. Res. Atmos.
23. Gryning, S.-E., Batchvarova, E., Brümmer, B., 48. Feingold, G., Koren, I., Wang, H., Xue, H. & Brewer, W. A. 124, 6191–6206 (2019).
Jørgensen, H. & Larsen, S. On the extension of the Precipitation-​generated oscillations in open cellular 74. Zeng, Z. et al. A reversal in global terrestrial stilling
wind profile over homogeneous terrain beyond cloud fields. Nature 466, 849–852 (2010). and its implications for wind energy production.
the surface boundary layer. Bound. Layer Meteorol. 49. Pleskachevsky, A. L., Lehner, S. & Rosenthal, W. Nat. Clim. Change 9, 979–985 (2019).
124, 251–268 (2007). Storm observations by remote sensing and influences Reports evidence for the presence of low-​frequency
24. Nunalee, C. G. & Basu, S. Mesoscale modeling of of gustiness on ocean waves and on generation of variability in near-​surface wind speeds.
coastal low-​level jets: implications for offshore wind rogue waves. Ocean Dyn. 62, 1335–1351 (2012). 75. Poan, E., Gachon, P., Laprise, R., Aider, R. &
resource estimation. Wind Energy 17, 1199–1216 50. Larsén, X. G., Vincent, C. & Larsen, S. Spectral Dueymes, G. Investigating added value of regional
(2014). structure of mesoscale winds over the water. climate modeling in North American winter storm
25. Pryor, S. C., Shepherd, T. J. & Barthelmie, R. J. Q. J. R. Meteorol. Soc. 139, 685–700 (2013). track simulations. Clim. Dyn. 50, 1799–1818 (2018).
Interannual variability of wind climates and wind 51. Passon, P. & Branner, K. Condensation of long-​term 76. Trzeciak, T. M., Knippertz, P., Pirret, J. S. &
turbine annual energy production. Wind Energy Sci. wave climates for the fatigue design of hydrodynamically Williams, K. D. Can we trust climate models to
3, 651–665 (2018). sensitive offshore wind turbine support structures. realistically represent severe European windstorms?
26. Pryor, S. C., Nielsen, M., Barthelmie, R. J. & Mann, J. Ships Offshore Struct. 11, 142–166 (2016). Clim. Dyn. 46, 3431–3451 (2016).
Can satellite sampling of offshore wind speeds 52. Young, I., Vinoth, J., Zieger, S. & Babanin, A. V. 77. Hodges, K. I., Lee, R. W. & Bengtsson, L. A comparison
realistically represent wind speed distributions? Part II: Investigation of trends in extreme value wave height of extratropical cyclones in recent reanalyses ERA-​
Quantifying uncertainties associated with sampling and wind speed. J. Geophys. Res. Oceans 117, Interim, NASA MERRA, NCEP CFSR, and JRA-25.
strategy and distribution fitting methods. J. Appl. C00J06 (2012). J. Clim. 24, 4888–4906 (2011).
Meteorol. 43, 739–750 (2004). 53. Sun, C. & Jahangiri, V. Fatigue damage mitigation 78. Pryor, S. C., Schoof, J. T. & Barthelmie, R. J.
27. Mann, J., Kristensen, L. & Jensen, N. O. in Bridge of offshore wind turbines under real wind and wave Winds of change?: Projections of near-​surface winds
Aerodynamics (eds Larsen, A. & Esdahl, S.) 49–56 conditions. Eng. Struct. 178, 472–483 (2019). under climate change scenarios. Geophys. Res. Lett.
(Balkema, 1998). 54. Trenberth, K. E., Dai, A., Rasmussen, R. M. & 33, L11702 (2006).
28. Ramon, J., Lledó, L., Torralba, V., Soret, A. & Parsons, D. B. The changing character of precipitation. 79. Pryor, S. C. & Barthelmie, R. J. in Climate Change
Doblas-​Reyes, F. J. What global reanalysis best Bull. Am. Meteorol. Soc. 84, 1205–1217 (2003). in the Midwest: Impacts, Risks, Vulnerability, and
represents near-​surface winds? Q. J. R. Meteorol. Soc. 55. Ma, J., Xie, S.-P. & Kosaka, Y. Mechanisms for tropical Adaptation Ch. 16 (ed. Pryor, S.C.) 213–229
145, 3236–3251 (2019). tropospheric circulation change in response to global (Indiana Univ. Press, 2013).
29. Olauson, J. ERA5: The new champion of wind power warming. J. Clim. 25, 2979–2994 (2012). 80. Pryor, S. C. & Barthelmie, R. J. Assessing climate
modelling? Renew. Energy 126, 322–331 (2018). 56. Shaw, T. et al. Storm track processes and the change impacts on the near-​term stability of the
30. Garcia-​Heller, V., Espinasa, R. & Paredes, S. opposing influences of climate change. Nat. Geosci. 9, wind energy resource over the USA. Proc. Natl Acad.
Forecast study of the supply curve of solar and wind 656–664 (2016). Sci. USA 108, 8167–8171 (2011).
technologies in Argentina, Brazil, Chile and Mexico. 57. Bengtsson, L., Hodges, K. & Roeckner, E. Storm tracks 81. Winterfeldt, J. & Weisse, R. Assessment of value
Renew. Energy 93, 168–179 (2016). and climate change. J. Clim. 19, 3518–3543 (2006). added for surface marine wind speed obtained from
31. Bianchi, E., Solarte, A. & Guozden, T. Spatiotemporal 58. Catto, J. L. et al. The future of midlatitude cyclones. two regional climate models. Mon. Weather Rev. 137,
variability of the wind power resource in Argentina Curr. Clim. Change Rep. 5, 407–420 (2019). 2955–2965 (2009).
and Uruguay. Wind Energy 22, 1086–1100 (2019). 59. O’Gorman, P. A. Understanding the varied response 82. Pryor, S. C., Barthelmie, R. J. & Schoof, J. T. Past and
32. Pryor, S. C., Conrick, R., Miller, C., Tytell, J. & of the extratropical storm tracks to climate change. future wind climates over the contiguous USA based
Barthelmie, R. J. Intense and extreme wind speeds Proc. Natl Acad. Sci. USA 107, 19176–19180 on the North American Regional Climate Change
observed by anemometer and seismic networks: (2011). Assessment Program model suite. J. Geophys. Res.
An eastern US case study. J. Appl. Meteorol. Climatol. 60. McCabe, G., Clark, M. & Serreze, M. Trends in Northern 117, D19119 (2012).
53, 2417–2429 (2014). Hemisphere surface cyclone frequency and intensity. 83. Larsen, X. G., Mann, J., Berg, J., Gottel, H. & Jacob, D.
33. Pryor, S. C., Barthelmie, R. J. & Schoof, J. T. J. Clim. 14, 2763–2768 (2001). Wind climate from the regional climate model REMO.
Inter-​annual variability of wind indices across Europe. 61. Pryor, S. C. & Hahmann, A. N. in Oxford Research Wind Energy 13, 279–296 (2010).
Wind Energy 9, 27–38 (2006). Encyclopedias: Climate Science (ed. von Storch, H.) 84. Pryor, S. C., Nikulin, G. & Jones, C. Influence of spatial
34. Grams, C. M., Beerli, R., Pfenninger, S., Staffell, I. (Oxford Univ. Press, 2019). resolution on regional climate model derived wind
& Wernli, H. Balancing Europe’s wind-​power output 62. Jung, C., Taubert, D. & Schindler, D. The temporal climates. J. Geophys. Res. Atmos. 117, D03117 (2012).
through spatial deployment informed by weather variability of global wind energy–Long-​term trends 85. Kusiak, A. Share data on wind energy: giving
regimes. Nat. Clim. Change 7, 557–562 (2017). and inter-​annual variability. Energy Convers. Manag. researchers access to information on turbine
35. Reichenberg, L., Wojciechowski, A., Hedenus, F. 188, 462–472 (2019). performance would allow wind farms to be optimized
& Johnsson, F. Geographic aggregation of wind 63. Bett, P. E., Thornton, H. E. & Clark, R. T. Using the through data mining. Nature 529, 19–22 (2016).
power-​an optimization methodology for avoiding low Twentieth Century Reanalysis to assess climate 86. Holt, E. & Wang, J. Trends in wind speed at wind
outputs. Wind Energy 20, 19–32 (2017). variability for the European wind industry. Theor. Appl. turbine height of 80 m over the contiguous United
36. International Electrotechnical Commission. IEC Climatol. 127, 61–80 (2017). States using the North American Regional Reanalysis
61400-1:2019. Wind energy generation systems – 64. Moemken, J., Reyers, M., Buldmann, B. & Pinto, J. G. (NARR). J. Appl. Meteorol. Climatol. 51, 2188–2202
Part 1: design requirements (IEC, 2019). Decadal predictability of regional scale wind speed (2012).

Nature Reviews | Earth & Environment


Reviews

87. McVicar, T. R. et al. Wind speed climatology and trends 110. Weisse, R. & von Storch, H. in Marine Climate and offshore wind turbines. Mar. Struct. 64, 174–185
for Australia, 1975–2006: Capturing the stilling Climate Change: Storms, Wind Waves and Storm (2019).
phenomenon and comparison with near-​surface Surges Ch. 5 165–203 (Springer, 2010). 135. Luengo, J., Negro, V., García-​Barba, J., López-​
reanalysis output. Geophys. Res. Lett. 35, L20403 111. Lombardo, F. T. & Ayyub, B. Approach to estimating Gutiérrez, J.-S. & Esteban, M. D. New detected
(2008). near-​surface extreme wind speeds with climate change uncertainties in the design of foundations for offshore
88. Chen, L., Li, D. & Pryor, S. C. Wind speed trends over considerations. J. Risk Uncertain. Eng. Syst. A 3, wind turbines. Renew. Energy 131, 667–677 (2019).
China: quantifying the magnitude and assessing A4017001 (2017). 136. Igwemezie, V., Mehmanparast, A. & Kolios, A. Current
causality. Int. J. Climatol. 33, 2579–2590 (2013). 112. Kumar, D., Mishra, V. & Ganguly, A. R. Evaluating trend in offshore wind energy sector and material
89. Hueging, H., Haas, R., Born, K., Jacob, D. & Pinto, J. G. wind extremes in CMIP5 climate models. Clim. Dyn. requirements for fatigue resistance improvement in
Regional changes in wind energy potential over Europe 45, 441–453 (2015). large wind turbine support structures – A review.
using regional climate model ensemble projections. 113. Born, K., Ludwig, P. & Pinto, J. G. Wind gust estimation Renew. Sustain. Energy Rev. 101, 181–196 (2019).
J. Appl. Meteorol. Climatol. 52, 903–917 (2013). for Mid-​European winter storms: towards a probabilistic 137. Dai, K. S. et al. Nonlinear response history analysis
90. Pryor, S. C. et al. Analyses of possible changes in view. Tellus A 64, 17471 (2012). and collapse mode study of a wind turbine tower
intense and extreme wind speeds over northern 114. Cheng, C. S., Lopes, E., Fu, C. & Huang, Z. Possible subjected to tropical cyclonic winds. Wind Struct. 25,
Europe under climate change scenarios. Clim. Dyn. impacts of climate change on wind gusts under 79–100 (2017).
38, 189–208 (2012). downscaled future climate conditions: updated for 138. Larsen, X. G. et al. Estimation of offshore extreme
91. Reyers, M., Moemken, J. & Pinto, J. G. Future changes Canada. J. Clim. 27, 1255–1270 (2014). wind from wind-​wave coupled modeling. Wind Energy
of wind energy potentials over Europe in a large 115. Prein, A. F. et al. A review on regional convection-​ 22, 1043–1057 (2019).
CMIP5 multi-​model ensemble. Int. J. Climatol. 36, permitting climate modeling: Demonstrations, 139. Du, J. T., Bolanos, R. & Larsen, X. G. The use of a
783–796 (2016). prospects, and challenges. Rev. Geophys. 53, wave boundary layer model in SWAN. J. Geophys. Res.
92. Hdidouan, D. & Staffell, I. The impact of climate 323–361 (2015). Oceans 122, 42–62 (2017).
change on the levelised cost of wind energy. Emphasizes key trends towards higher-​fidelity, 140. Pryor, S. C., Shepherd, T., Volker, P., Hahmann, A.
Renew. Energy 101, 575–592 (2017). high-​resolution simulations. & Barthelmie, R. J. “Wind theft” from onshore
93. Tobin, I. et al. Assessing climate change impacts 116. Orwig, K. D. & Schroeder, J. L. Near-​surface wind wind turbine arrays: Sensitivity to wind farm
on European wind energy from ENSEMBLES high-​ characteristics of extreme thunderstorm outflows. parameterization and resolution. J. Appl. Meteorol.
resolution climate projections. Clim. Change 128, J. Wind Eng. Ind. Aerodyn. 95, 565–584 (2007). Climatol. 59, 153–174 (2020).
99–112 (2015). 117. Hoogewind, K. A., Baldwin, M. E. & Trapp, R. J. 141. Jiménez, P. A., Navarro, J., Palomares, A. M. &
Includes one of the most comprehensive RCM The impact of climate change on hazardous convective Dudhia, J. Mesoscale modeling of offshore wind
ensembles and a focus on medium-​term projections weather in the United States: insight from high-​ turbine wakes at the wind farm resolving scale:
of wind resources. resolution dynamical downscaling. J. Clim. 30, a composite-​based analysis with the Weather
94. Pryor, S. C., Shepherd, T. J., Bukovsky, M. & 10081–10100 (2017). Research and Forecasting model over Horns Rev.
Barthelmie, R. J. Assessing the stability of wind 118. Di, Y., Lu, J., Xu, X., Feng, T. & Li, L. A response Wind Energy 18, 559–566 (2015).
resource and operating conditions. J. Phys. Conf. Ser. characteristics study of widespread power grid icing 142. Wilczak, J. et al. The Wind Forecast Improvement
1452, 012084 (2020). to El Nino. Math. Probl. Eng. 2019, 6589410 (2019). Project (WFIP): A public–private partnership
95. Greene, J. S., Chatelain, M., Morrissey, M. & 119. Yu, Y., Ren, Z., Gao, F. & Meng, X. Changes in surface addressing wind energy forecast needs. Bull. Am.
Stadler, S. Projected future wind speed and wind icing duration over north china during 1961–2015. Meteorol. Soc. 96, 1699–1718 (2015).
power density trends over the Western US High Plains. Atmos. Sci. Lett. 19, e827 (2018). 143. Gutowski, W. J. Jr. et al. The ongoing need for
Atmos. Clim. Sci. 2, 32–40 (2012). 120. Clausen, N. E. et al. in Impacts of Climate Change on high-​resolution regional climate models: Process
96. Ruffato-​Ferreira, V. et al. A foundation for the strategic Renewable Energy Sources (ed Fenger, J.) 105–128 understanding and stakeholder information. Bull. Am.
long-​term planning of the renewable energy sector in (Nordic Council of Ministers, 2007). Meteorol. Soc. 101, E664–E683 (2020).
Brazil: hydroelectricity and wind energy in the face of 121. Zhang, Y., Chen, W. & Riseborough, D. W. Transient 144. Kendon, E. J. et al. Do convection-​permitting regional
climate change scenarios. Renew. Sustain. Energy Rev. projections of permafrost distribution in Canada climate models improve projections of future
72, 1124–1137 (2017). during the 21st century under scenarios of climate precipitation change? Bull. Am. Meteorol. Soc. 98,
97. de Jong, P. et al. Estimating the impact of climate change. Glob. Planet. Change 60, 443–456 (2008). 79–93 (2017).
change on wind and solar energy in Brazil using a 122. Zheng, M., Yang, Z. J., Yang, S. & Still, B. Modeling 145. Bukovsky, M. S., Gochis, D. J. & Mearns, L. O.
South American regional climate model. Renew. Energy and mitigation of excessive dynamic responses of wind Towards assessing NARCCAP regional climate model
141, 390–401 (2019). turbines founded in warm permafrost. Eng. Struct. credibility for the North American monsoon: Current
98. Kulkarni, S. & Huang, H. P. Changes in surface 148, 36–46 (2017). climate simulations. J. Clim. 26, 8802–8826 (2013).
wind speed over North America from CMIP5 123. Cheng, G. Permafrost studies in the Qinghai–Tibet 146. Haarsma, R. J. et al. High resolution model
model projections and implications for wind energy. Plateau for road construction. J. Cold Reg. Eng. 19, intercomparison project (HighResMIP v1. 0) for
Adv. Meteorol. 2014, 292768 (2014). 19–29 (2005). CMIP6. Geosci. Model Dev. 9, 4185–4208 (2016).
99. Wiser, R. H. & Bolinger, M. Benchmarking anticipated 124. Vanem, E. Non-​stationary extreme value models to 147. Stevens, B. et al. DYAMOND: The DYnamics of the
wind project lifetimes: results from a survey of U.S. account for trends and shifts in the extreme wave Atmospheric general circulation Modeled On Non-​
Wind industry professionals. Electricity Markets & climate due to climate change. Appl. Ocean Res. 52, hydrostatic Domains. Prog. Earth Planet. Sci. 6, 61
Policy https://emp.lbl.gov/publications/benchmarking-​ 201–211 (2015). (2019).
anticipated-wind-​project (2019). 125. Shope, J. B., Storlazzi, C. D., Erikson, L. H. & 148. Hagos, S., Leung, L. R., Zhao, C., Feng, Z. &
100. Pryor, S. C. & Barthelmie, R. J. Hybrid downscaling Hegermiller, C. A. Changes to extreme wave Sakaguchi, K. How do microphysical processes
of wind climates over the eastern USA. Environ. Res. climates of islands within the Western Tropical Pacific influence large-​scale precipitation variability and
Lett. 9, 024013 (2014). throughout the 21st century under RCP 4.5 and RCP extremes? Geophys. Res. Lett. 45, 1661–1667
101. Pryor, S. C. & Schoof, J. T. Importance of the SRES in 8.5, with implications for island vulnerability and (2018).
projections of climate change impacts on near-​surface sustainability. Glob. Planet. Change 141, 25–38 (2016). 149. Yang, Z. et al. Modeling analysis of the swell and
wind regimes. Meteorol. Z. 19, 267–274 (2010). 126. Dvorak, P. A look at repowering older wind farms wind-​sea climate in the Salish Sea. Estuarine Coast.
102. Carvalho, D., Rocha, A., Gomez-​Gesteira, M. & to 2020. Windpower Engineering & Development Shelf Sci. 224, 289–300 (2019).
Santos, C. S. Potential impacts of climate change https://www.windpowerengineering.com/look-​ 150. Coppola, E. et al. A first-​of-its-​kind multi-​model
on European wind energy resource under the CMIP5 repowering-older-​wind-farms-2020/ (2018). convection permitting ensemble for investigating
future climate projections. Renew. Energy 101, 29–40 127. International Energy Agency. Wind technology convective phenomena over Europe and the
(2017). collaboration programme: annual report 2017 Mediterranean. Clim. Dyn. 55, 3–34 (2020).
103. Sakaguchi, K. et al. Exploring a multiresolution (IEA, 2018). 151. Roberts, M. et al. The benefits of global high resolution
approach using AMIP simulations. J. Clim. 28, 128. Pryor, S. C., Barthelme, R. J. & Shepherd, T. 20% of for climate simulation: process understanding and the
5549–5574 (2015). US electricity from wind will have limited impacts on enabling of stakeholder decisions at the regional scale.
104. Giorgetta, M. A. et al. Climate and carbon cycle system efficiency and regional climate. Sci. Rep. 10, Bull. Am. Meteorol. Soc. 99, 2341–2359 (2018).
changes from 1850 to 2100 in MPI-​ESM simulations 541 (2020). 152. von Trentini, F., Leduc, M. & Ludwig, R. Assessing
for the Coupled Model Intercomparison Project phase 129. American Wind Energy Association. Wind industry natural variability in RCM signals: comparison
5. J. Adv. Model. Earth Syst. 5, 572–597 (2013). annual market report, year ending 2017. AWEA of a multi model EURO-​CORDEX ensemble with a
105. Johnson, D. L. & Erhardt, R. J. Projected impacts of https://www.awea.org/resources/publications-​and- 50-member single model large ensemble. Clim. Dyn.
climate change on wind energy density in the United reports/market-​reports/2017-u-​s-wind-​industry- 53, 1963–1979 (2019).
States. Renew. Energy 85, 66–73 (2016). market-​reports (2018). 153. Marbaix, P., Gallée, H., Brasseur, O. & van Ypersele, J.-P.
106. Lindvall, J., Svensson, G. & Caballero, R. The impact 130. Dalla Riva, A. et al. IEA wind TCP Task 26 – Lateral boundary conditions in regional climate
of changes in parameterizations of surface drag and Wind technology, cost, and performance trends in models: a detailed study of the relaxation procedure.
vertical diffusion on the large-​scale circulation in the Denmark, Germany, Ireland, Norway, Sweden, the Mon. Weather Rev. 131, 461–479 (2003).
Community Atmosphere Model (CAM5). Clim. Dyn. European Union, and the United States: 2008–2016 154. Giorgi, F. Thirty years of regional climate modeling:
48, 3741–3758 (2017). (National Renewable Energy Laboratory, 2019). Where are we and where are we going next? J. Geophys.
107. Larsén, X. G., Ott, S., Badger, J., Hahmann, A. N. & 131. Wind Europe. Wind energy in Europe in 2018 Res. Atmos. 124, 5696–5723 (2019).
Mann, J. Recipes for correcting the impact of effective (Wind Europe, 2019). Summarizes the history of, challenges to and new
mesoscale resolution on the estimation of extreme 132. National Renewable Energy Laboratory. Annual prospects for regional climate modelling.
winds. J. Appl. Meteorol. Climatol. 51, 521–533 (2012). technology baseline: electricity. NREL https://www. 155. Gao, J. & O’Neill, B. C. Data-​driven spatial modeling of
108. Lombardo, F. T. Improved extreme wind speed nrel.gov/analysis/data-​tech-baseline.html (2019). global long-​term urban land development: The SELECT
estimation for wind engineering applications. J. Wind 133. Musial, W., Beiter, J., Spitsen, P., Nunemaker, J. & model. Environ. Model. Softw. 119, 458–471 (2019).
Eng. Ind. Aerodyn. 104–106, 278–284 (2012). Gevorgian, V. 2018 offshore wind technologies market 156. Hamlington, B. D., Hamlington, P. E., Collins, S. G.,
109. Cook, N. J. Confidence limits for extreme wind speeds report (Department of Energy, 2019). Alexander, S. R. & Kim, K. Y. Effects of climate
in mixed climates. J. Wind Eng. Ind. Aerodyn. 92, 134. Horn, J. T., Krokstad, J. R. & Leira, B. J. Impact oscillations on wind resource variability in the United
41–51 (2004). of model uncertainties on the fatigue reliability of States. Geophys. Res. Lett. 42, 145–152 (2015).

www.nature.com/natrevearthenviron
Reviews

157. Lledo, L., Bellprat, O., Doblas-​Reyes, F. J. & Soret, A. 172. Commin, A. N. et al. The influence of the North 187. Kamranzad, B. & Mori, N. Future wind and wave climate
Investigating the effects of Pacific sea surface Atlantic Oscillation on diverse renewable generation projections in the Indian Ocean based on a super-​high-
temperatures on the wind drought of 2015 over in Scotland. Appl. Energy 205, 855–867 (2017). resolution MRI-​AGCM3.2S model projection. Clim. Dyn.
the United States. J. Geophys. Res. Atmos. 123, 173. Albani, A., Ibrahim, M. Z. & Yong, K. H. Influence of the 53, 2391–2410 (2019).
4837–4849 (2018). ENSO and monsoonal season on long-​term wind energy 188. Lawrence, D. M. et al. Parameterization improvements
158. Greene, J. S., Chatelain, M., Morrissey, M. & potential in Malaysia. Energies 11, 2965 (2018). and functional and structural advances in version 4 of
Stadler, S. Estimated changes in wind speed and 174. Dunning, C., Turner, A. & Brayshaw, D. The impact of the Community Land Model. J. Adv. Model. Earth Syst.
wind power density over the western High Plains, monsoon intraseasonal variability on renewable power 3, M03001 (2011).
1971–2000. Theor. Appl. Climatol. 109, 507–518 generation in India. Environ. Res. Lett. 10, 064002 189. Bogenschutz, P. A. et al. The path to CAM6: coupled
(2012). (2015). simulations with CAM5.4 and CAM5.5. Geosci. Model.
159. Klink, K. Atmospheric circulation effects on wind 175. Haupt, S. E. et al. A method to assess the wind and Dev. 11, 235–255 (2018).
speed variability at turbine height. J. Appl. Meteorol. solar resource and to quantify interannual variability 190. Zhao, C. et al. Exploring the impacts of physics
Climatol. 46, 445–456 (2007). over the United States under current and projected and resolution on aqua-​planet simulations from a
160. Clifton, A. & Lundquist, J. K. Data clustering reveals future climate. J. Appl. Meteorol. Climatol. 55, nonhydrostatic global variable-​resolution modeling
climate impacts on local wind phenomena. J. Appl. 345–363 (2016). framework. J. Adv. Model. Earth Syst. 8, 1751–1768
Meteorol. Climatol. 51, 1547–1557 (2012). 176. Rasmussen, D., Holloway, T. & Nemet, G. Opportunities (2016).
161. George, S. S. & Wolfe, S. A. El Niño stills winter winds and challenges in assessing climate change impacts 191. Wilks, D. S. Statistical Methods in the Atmospheric
across the southern Canadian Prairies. Geophys. Res. on wind energy — a critical comparison of wind speed Sciences Vol. 100 (Academic, 2011).
Lett. 36, L23806 (2009). projections in California. Environ. Res. Lett. 6, 024008
162. Torres Silva dos Santos, A., E Silva, S. & Moisés, C. (2011). Acknowledgements
Seasonality, interannual variability, and linear 177. Karnauskas, K. B., Lundquist, J. K. & Zhang, L. This work was supported by the U.S. Department of Energy
tendency of wind speeds in the Northeast Brazil Southward shift of the global wind energy resource (DoE) (DE-​SC0016438 and DE-​SC0016605). The research
from 1986 to 2011. Sci. World J. 2013, 490857 under high carbon dioxide emissions. Nat. Geosci. 11, used computing resources from the NCAR-​CISL programme
(2013). 38–43 (2018). (UCOR0020) and the National Science Foundation’s Extreme
163. Bianchi, E., Solarte, A. & Guozden, T. M. Large 178. Daines, J. T., Monahan, A. H. & Curry, C. L. Model-​ Science and Engineering Discovery Environment (XSEDE)
scale climate drivers for wind resource in Southern based projections and uncertainties of near-​surface (allocation award to S.C.P. is TG-​ATM170024). The authors
South America. Renew. Energy 114, 708–715 wind climate in western Canada. J. Appl. Meteorol. express their appreciation to N. Davis of DTU Wind Energy
(2017). Climatol. 55, 2229–2245 (2016). for providing access to a digital form of the Global Wind Atlas
164. Watts, D., Duran, P. & Flores, Y. How does El Nino 179. Gross, M. & Magar, V. Offshore wind energy climate mean wind speeds shown in Fig. 1a.
Southern Oscillation impact the wind resource in projection using UPSCALE climate data under the
Chile? A techno-​economical assessment of the RCP8.5 emission scenario. PLoS ONE 11, e0165423 Author contributions
influence of El Nino and La Nina on the wind power. (2016). S.C.P. conducted the majority of the analyses presented
Renew. Energy 103, 128–142 (2017). 180. Devis, A., Van Lipzig, N. P. M. & Demuzere, M. herein, made Figs 1,3 and Tables 1,2, undertook the litera-
165. Ravestein, P., van der Schrier, G., Haarsma, R., Should future wind speed changes be taken into ture review and wrote the initial draft of the manuscript.
Scheele, R. & van den Broek, M. Vulnerability of account in wind farm development? Environ. Res. Lett. R.J.B. conducted the analyses presented in Fig. 2 and wrote
European intermittent renewable energy supply 13, 064012 (2018). related text and wrote the initial draft of the wind energy
to climate change and climate variability. Renew. 181. Reyers, M., Pinto, J. G. & Moemken, J. Statistical-​ industry trends. M.S.B. performed the WRF simulations that
Sustain. Energy Rev. 97, 497–508 (2018). dynamical downscaling for wind energy potentials: are analysed and presented herein and wrote part of the
166. Francois, B. Influence of winter North-​Atlantic evaluation and applications to decadal hindcasts section on reconciling climate and wind energy science. L.R.L.
oscillation on climate-​related-energy penetration in and climate change projections. Int. J. Climatol. 35, helped perform the simulations presented in Fig. 4 and con-
Europe. Renew. Energy 99, 602–613 (2016). 229–244 (2015). tributed related text. K.S. performed the MPAS simula-
167. Kriesche, P. & Schlosser, C. A. The association of the 182. Davy, R., Gnatiuk, N., Pettersson, L. & Bobylev, L. tions that are analysed and presented herein and wrote parts
North Atlantic and the Arctic Oscillation on wind Climate change impacts on wind energy potential in of the section on reconciling climate and wind energy science.
energy resource over Europe and its intermittency. the European domain with a focus on the Black Sea. All co-​authors contributed to editing the draft manuscript
Energy Procedia 52, 270–277 (2014). Renew. Sustain. Energy Rev. 81, 1652–1659 (2018). and read and approved the revised manuscript.
168. Jerez, S. et al. The impact of the North Atlantic 183. Chen, L., Pryor, S. C. & Li, D. Assessing the performance
Oscillation on renewable energy resources in of Intergovernmental Panel on Climate Change AR5
Competing interests
southwestern Europe. J. Appl. Meteorol. Climatol. 52, climate models in simulating and projecting wind
The authors declare no competing interests.
2204–2225 (2013). speeds over China. J. Geophys. Res. Atmos. 117,
169. Brayshaw, D. J., Troccoli, A., Fordham, R. & Methven, J. D24102 (2012).
The impact of large scale atmospheric circulation 184. Xiong, Y. J., Xin, X. G. & Kou, X. X. Simulation and Peer review information
patterns on wind power generation and its potential projection of near-​surface wind speeds in China by Nature Reviews Earth & Environment thanks Christopher Jung
predictability: A case study over the UK. Renew. Energy BCC-​CSM models. J. Meteorol. Res. 33, 149–158 and the other, anonymous, reviewer(s) for their contribution to
36, 2087–2096 (2011). (2019). the peer review of this work.
170. Earl, N., Dorling, S., Hewston, R. & von Glasow, R. 185. Evans, J. P., Kay, M., Prasad, A. & Pitman, A.
1980–2010 variability in UK surface wind climate. The resilience of Australian wind energy to climate Publisher’s note
J. Clim. 26, 1172–1191 (2013). change. Environ. Res. Lett. 13, 024014 (2018). Springer Nature remains neutral with regard to jurisdictional
171. Iversen, E. C. & Burningham, H. Relationship between 186. Ohba, M. The impact of global warming on wind energy claims in published maps and institutional affiliations.
NAO and wind climate over Norway. Clim. Res. 63, resources and ramp events in Japan. Atmosphere 10,
115–134 (2015). 265 (2019). © Springer Nature Limited 2020

Nature Reviews | Earth & Environment

You might also like