1172 JOURNAL OF CLIMATE VOLUME 26

1980–2010 Variability in U.K. Surface Wind Climate

NICK EARL AND STEVE DORLING School of Environmental Sciences, University of East Anglia, Norwich,

RICHARD HEWSTON Department of Meteorology, University of Hawaii at Manoa, Honolulu, Hawaii

ROLAND VON GLASOW School of Environmental Sciences, University of East Anglia, Norwich, United Kingdom

(Manuscript received 6 December 2011, in final form 29 August 2012)

ABSTRACT

The climate of the northeast Atlantic region comprises substantial decadal variability in storminess. It also exhibits strong inter- and intra-annual variability in extreme high and low wind speed episodes. Here the authors quantify and discuss causes of the variability seen in the U.K. wind climate over the recent period 1980–2010. Variations in U.K. hourly mean (HM) wind speeds, in daily maximum gust speeds and in asso- ciated wind direction measurements, made at standard 10-m height and recorded across a network of 40 stations, are considered. The Weibull distribution is shown to generally provide a good fit to the hourly wind data, albeit with the shape parameter k spatially varying from 1.4 to 2.1, highlighting that the commonly assumed k 5 2 Rayleigh distribution is not universal. It is found that the 10th and 50th percentile HM wind speeds have declined significantly over this specific period, while still incorporating a peak in the early 1990s. The authors’ analyses place the particularly ‘‘low wind’’ year of 2010 into longer-term context and their findings are compared with other recent international studies. Wind variability is also quantified and discussed in terms of variations in the exceedance of key wind speed thresholds of relevance to the insurance and wind energy industries. Associated interannual variability in energy density and potential wind power output of the order of 620% around the mean is revealed. While 40% of network average winds are in the southwest quadrant, 51% of energy in the wind is associated with this sector. The findings are discussed in the context of current existing challenges to improve predictability in the -Atlantic sector over all time scales.

1. Introduction coastline). Seasons dominated by blocking or cyclonic weather types, especially winter, can strongly skew the Located in one of the most common regions for at- magnitude of annual insured losses (Munich Re 2002), mospheric blocking, while also situated toward the end as well as have profound effects on the variability of point of a major midlatitude track, the United wind power generated by the expanding U.K. wind en- Kingdom has one of the most variable wind climates and ergy sector (Sinden 2007). northwest Europe is a challenging region for prediction The cold European winter of 2009/10 and the extreme on all time scales (Barriopedro et al. 2006, 2008; Dacre cold of December 2010 have prompted much discussion and Gray 2009; Woollings 2010). Regional wind climate about long-term climate variations and their possible variability in the United Kingdom is large, governed by impacts. However, Cattiaux et al. (2010) show that the latitude (proximity to storm track), altitude, and type of cold European surface temperature anomaly of up to fetch (the United Kingdom has an exceptionally long 68C for winter 2009/10 was in fact not as great as might have been expected given the associated record-breaking North Atlantic Oscillation (NAO) and blocking fre- Corresponding author address: Nick Earl, School of Environ- mental Sciences, University of East Anglia, Norwich, Norwich quency indices. These authors concluded that the event Research Park, Norwich NR4 7TJ, United Kingdom. was a cold extreme that was not in any way inconsistent E-mail: [email protected] with an otherwise generally warming climate. Focusing

DOI: 10.1175/JCLI-D-12-00026.1

Ó 2013 American Meteorological Society Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1173 on predictability at the monthly, seasonal, and decadal , the relative vulnerability of the affected areas, time scale, many forcing agents are thought to modulate whether trees are in leaf or not, and the relative dryness European climate, such as sea surface temperatures, or wetness of the ground at the time of windstorm pas- stratospheric circulation, and solar variability (Rodwell sage (Hewston and Dorling 2011). et al. 1999; Lockwood et al. 2010, 2011; Woollings et al. Wang et al. (2009) demonstrated that storminess in 2010). Regional responses also arise from the dynamical the North Atlantic–European region, based on atmo- reaction of the climate system to this forcing (Woollings spheric sea level pressure gradients, undergoes sub- 2010; Jung et al. 2011) and internal atmospheric dy- stantial decadal and longer time scale fluctuations and namics can be an important source of low-frequency that these changes have a seasonality and regionality to atmospheric interannual variability. Solar activity in them. In particular, these authors showed that winter 2009/10 fell to values unknown since the start of the storminess reached an unprecedented maximum in the twentieth century and Lockwood et al. (2010), linking early 1990s in the North Sea and showed a steady in- this to the occurrence of recent cold European winter crease in the northeastern part of the North Atlantic– months, estimate an 8% chance that the decline, which European region, significantly correlated with variability began around 1985, could continue to Maunder mini- in the NAO index. The link to the NAO is found in all mum levels within 50 years, from the previous grand seasons except autumn. As the NAO swings from one solar maximum. On the other hand, European Centre phase to the other, large changes to windstorm intensity for Medium-Range Weather Forecasts (ECMWF) ex- and track and to mean wind speed and direction are periments (Jung et al. 2011), testing the sensitivity to observed over the Atlantic (Hurrell et al. 2003). Both reduced ultraviolet radiation of the onset of the cold Atkinson et al. (2006), analyzing the period 1990–2005, 2009/10 European winter, show that the unusually low and Boccard (2009), for 1979–2007, showed that the solar activity contributed little, if any, to the observed NAO is a good approximation for synoptic weather type NAO anomaly. Much research is ongoing to improve indices such as Grosswetterlagen (Hess and Brezowsky our predictive capability in Europe. 1952; James 2007) and the Jenkinson–Collison weather In Europe, windstorms remain the most economically type classification (Jenkinson and Collison 1977; Jones significant weather peril when averaging over multiple et al. 1993) and for wind indices in Northern Europe years. The winter storms of the early 1990s had some over the respective periods. A decrease in post-1990 dramatic effects on the United Kingdom, the winter of northern European windiness is clearly revealed in these 1989/90 being one of the most damaging on record, ex- studies. By considering the longer-term Grosswetterla- emplified by windstorm Daria on 25 January (McCallum gen and Jenkinson variability through the twentieth 1990). The storm tracked across a large swath of En- century, these authors concluded that care is needed in gland and Wales, causing widespread damage amount- selecting the most appropriate long-term period on ing to £1.9 billion (equivalent to £3.2 billion in 2010 which to base wind energy investment decisions and that values) of U.K. insured losses (Munich Re 2002). A access to reliable and longer-term wind speed mea- second storm, Vivian, buffeted the United Kingdom surements is highly desirable. Recent industry discus- between 26 and 28 February 1990 and contributed to sion of the low-wind year of 2010 requires further U.K. weather-related property losses that year reaching supporting analysis and discussion of the wider context. their highest mark on record. In the winter of 1991/92 As greater reliance on wind power for the United the New Year’s Day Storm affected northern Scotland Kingdom’s electricity generation needs increases, so and (far more severely) Norway (Gronas 1995), pro- will the magnitude of risk due to exposure of the per- ducing stronger U.K. surface winds than Daria and formance of the turbines to climate change (Harrison Vivian, though causing less U.K. damage because of et al. 2008). reduced vulnerability to insurance losses in the affected Both the wind energy and insurance industries are regions. Meanwhile, winter storm Xynthia in February sensitive to wind speed distributions. The Weibull dis- 2010 caused insured losses totaling almost $3 billion in tribution function has become widely used in meteorol- , , and Spain, representing the world’s ogy to estimate how observed wind speeds tend to vary third most costly catastrophe of that year (Swiss Re around their mean at sites where only a long-term av- 2011), more costly than any 2010 North Atlantic hurri- erage is known. Originally used to describe the size cane. Indeed total damage is distribution of particles, the Weibull distribution has considerable, equivalent to that of worldwide hurricanes numerous applications, including in general insurance to when averaged over longer time scales (Malmquist model reinsurance claim sizes (Kremer 1998). The use 1999). Total annual losses attributed to windstorms de- and importance of the Weibull distribution has grown pend, for example, on the precise track and intensities of immensely in the wind power industry and has been used

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1174 JOURNAL OF CLIMATE VOLUME 26 to help site many thousands of wind turbines (Petersen The results presented in this paper include analysis et al. 1998; see section 2c). and discussion of wind speed threshold exceedance Numerous authors have also been considering the frequencies, the proportion of time that the hourly possible impact of climate change over the twenty-first winds or daily gust speeds are above a set of specific century on the wind climate of northwest Europe, in the speeds, at individual sites and on average across the context of the decadal variability seen over the last network of 40 (39) hourly wind speed (gust speed) sites. century (Brown et al. 2009; Ulbrich et al. 2009; Pryor This follows the approach adopted by Vautard et al. et al. 2012). While this is clearly a complex question, one (2010) but provides detail for the United Kingdom point which models do seem to currently agree on for the rather than a more general continental or global scale. future climate of the region is an increasing frequency The further novelty of the study presented here comes of intense cyclones in the region of the British Isles from using hourly data, rather than 6-hourly, by con- (Ulbrich et al. 2009) and increased winter storminess sidering a high spatial density of stations in the United (Scaife et al. 2012). Kingdom, by incorporating gusts and wind directions Hewston (2008) and Hewston and Dorling (2011) in- with mean wind speeds, and by including the anomalous troduced for the first time an hourly wind speed data- conditions of 2010. Furthermore, we present the impli- base for a network of 43 U.K. surface stations, extending cations of a variable wind climate for wind energy den- through the period 1980–2005 and providing good spa- sity and wind power output, building on the work of tial coverage. Based on this they presented a climatology previous U.K. wind resource studies (e.g., Sinden 2007). of the strongest wind gusts in the context of insurance weather perils. These authors presented evidence of an apparent downward trend in the strongest wind gusts 2. Data, methods, and tools over the United Kingdom since the early 1990s. In ad- a. Observed wind data dition, Vautard et al. (2010), also using surface station data, reported that mean wind speeds have also been This study extends the 1980–2005 database described declining over the same period across most areas of the by Hewston and Dorling (2011) of hourly surface wind world, including Europe, a phenomenon they termed speed observations (measured at the standard 10-m ‘‘global stilling’’ and linked to changes in land-based height) from the Met Office (UKMO) stations across biomass. However, while a decline was also found in the United Kingdom, to the end of 2010, incorporating Australian 2-m wind speed observations by Troccoli the anomalous European winter months in 2010. Wind et al. (2012), their equivalent 10-m measurements ac- data for all 31 years were extracted from the Met Office tually showed a positive tendency. Integrated Data Archive System (MIDAS) Land Sur- Here we build on the earlier U.K.-focused work of face Observations Station database (UKMO 2011), ar- Hewston and Dorling (2011) described above by also chived at the British Atmospheric Data Centre (BADC). considering mean wind speeds in the United Kingdom. Unfortunately, three of the 43 sites used in the original The objectives of this paper are to network (Coltishall, Durham, and St Mawgan) have been discontinued since 2005 and have been removed d update analysis of temporal variability to 2010 and from the database. The hourly mean (HM; i.e., the extend the quality control of the Hewston and Dorling 10-min average, recorded from 20 to 10 min prior to the database; hour in question) wind speeds and daily maximum gust d deepen understanding of each of the stations in the speeds (DMGSs; maximum 3-s average), with their as- network by investigating applicability of the Weibull sociated wind directions, are extracted as described in distribution across locations, interpreting the results detail by Hewston and Dorling (2011). The site at from a topographic perspective; Ringway (Manchester Airport) no longer records gusts, d analyze variations of exceedances of a wider range of only mean wind speeds, leaving a 31-yr (1980–2010) U.K. wind speed thresholds of interest to both the insurance network of 40 sites for HM wind speeds and 39 sites for and wind energy sectors, compare these with the larger- DMGSs whose geographical locations are displayed in scale findings of Vautard et al. (2010), and discuss them Fig. 1. Hewston and Dorling’s (2011) primary focus was in the context of key features of the regional-scale the DMGSs, whereas this study makes more use of the atmospheric circulation; and HM wind speeds. The 40 sites used in this study have on d quantify the impact of the observed spatial and tempo- average 98.5% HM data completeness, substantially ral variations in wind power on output from a synthetic higher than previous studies using HM MIDAS data network of 3.6-MW wind turbines, one located at each (e.g., Sinden 2007; 77% HM data completeness). All of of the monitoring stations. the sites used in this study meet the stringent UKMO

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1175

FIG. 1. Location of observation stations in the network. Note that Ringway (23) has no DMGS data, only recording hourly mean wind speed. site exposure requirements (available at http://badc. 0 or 1 kt. This leads to an overrepresentation of HM nerc.ac.uk/data/ukmo-midas/ukmo_guide.html). Since wind values of 2 kt and an underrepresentation of 0 and the sites in this study possess such a wide variety of to- especially 1 kt at many sites. We have, however, made pographies and therefore wind regimes, it is thought that no attempt to redistribute these extra 2-kt values into when averaged together they give a good representation neighboring bins. of the U.K. wind regime as a whole. c. Weibull distribution b. Data quality The Weibull distribution came to prominence in me- The wind speed and direction data has undergone teorology during the 1970s (Takle and Brown 1978). As rigorous quality control, with checks on the equipment a two-parameter density function it can be calculated as and raw data performed at the UKMO and the BADC.     Further information on quality control performed on U k P(U) 5 1 2 exp 2 , (1) the MIDAS database and other possible sources of error A is available at the BADC website (http://badc.nerc.ac. uk/data/ukmo-midas/ukmo_guide.html; UKMO 2011) where P(U) is the probability distribution of wind speed and in Hewston and Dorling (2011). Once downloaded, U, A is the Weibull scale parameter, and k is the shape a series of steps were followed to further test the re- parameter (Pryor and Barthelmie 2010). For a narrow liability of the information, removing duplicate data, distribution, with a marked peak, k will take a relatively detecting missing values, and checking data consistency. high value. Numerous statistical methods have been Analysis of Weibull distributions, discussed below, was proposed to calculate Weibull scale and shape parame- also helpful in highlighting potential anomalies. The ters (Pryor et al. 2004), with Seguro and Lambert (2000) MIDAS data do not normally include an HM value recommending the maximum likelihood method when 2 of1kt(0.515ms 1) and often use a value of 2 kt wind speed data is available in a time series format. 2 (1.03 m s 1) when the wind vane indicates gusty con- When the Weibull shape parameter has a value of 2, it is ditions (BADC website) to represent a mean speed of known as the Rayleigh distribution, and this is often

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1176 JOURNAL OF CLIMATE VOLUME 26

TABLE 1. Power produced by a present-day state-of-the-art being well represented by the Weibull distribution. The 3.6-MW wind turbine and energy density [from Eq. (2)] for wind 21 DMGS set, however, by definition, should be more speeds in the range 0–26 m s converted to 100 m using the Weibull compatible. This study examines the capability power-law approximation [Eq. (3)]. of the Weibull distribution to represent the variance of Energy land-based wind monitoring sites by calculating the Surface wind Wind speed density 2 2 2 31-yr shape parameter at each site for both HM wind speed (m s 1) (m s 1) at 100 m Power (kW) (W m 2) speed and DMGSs. This also reveals how well the 00 00commonly used Rayleigh distribution approximates the 1 1.38 0 1.61 sites’ wind speed variance. There have been numerous 2 2.76 0 12.88 3 4.14 102 43.46 methods and modifications to the Weibull distribution 4 5.52 361 103.02 to deal with zero and low wind speed values; however, it 5 6.90 770 201.21 is not the intention here to assess which of these best 6 8.28 1386 347.69 represents the DMGS and HM datasets, so this study 7 9.67 2175 553.84 simply uses the commonly adopted basic maximum 8 11.04 2965 824.16 9 12.42 3411 1173.47 likelihood method (Seguro and Lambert 2000). It must be 10 13.80 3565 1609.69 noted that the basic method used is unable to accom- 11 15.18 3595 2142.50 modate calm conditions, although the approach can be 12 16.56 3600 2781.55 modified to account for these (Wilks 1990). Tests were 2 13 17.95 3600 3542.42 carried out assigning a negligible value (0.000 01 m s 1) 14 19.33 3600 4423.86 21 15 20.71 3600 5440.59 to reports of 0 m s ; however, the results for HM wind 16 22.09 3600 6602.27 speeds (not shown) displayed strong positively skewed, 17 23.47 3600 7918.54 poorly fitting Weibull distributions and k values as low 18 24.85 3600 9399.08 as 0.3. 19 26.23 0 11 050.50 20 27.61 0 12 888.75 d. Wind turbine power 21 28.99 0 14 920.34 22 30.37 0 17 154.93 The 31-yr U.K. HM wind speed database enables an 23 31.75 0 19 602.18 assessment of the potential impact of spatial and tem- 24 33.13 0 22 271.77 25 34.51 0 25 173.35 poral variations in the U.K. wind regime on the wind 26 35.89 0 28 316.59 energy sector. Power generated is proportional to the cube of the wind speed and the variability of the wind around the mean is therefore critical to the amount of power produced. Wind power generation potential can used as the standard for wind turbine manufacturers’ be quantified using the concept of energy density (or performance figures (Weisser 2003). The Weibull dis- power density): tribution, however, has been found to produce a better fit to observed wind speeds than the simpler Rayleigh 1 E 5 rU3 (2) distribution (Celik 2004). 2 Nevertheless it is problematic fitting a Weibull dis- 2 tribution at low wind speeds, as highlighted by Justus where E is energy density (W m 2), r is air density 2 2 et al. (1976), who assessed potential output from wind- (kg m 3), and U is the hub-height wind speed (m s 1) powered generators. On the other hand, it is generally (Pryor et al. 2012). For this study, the energy density for accepted that sites with regular moderate or high wind each of the 40 HM observation sites is calculated to first 2 speeds can almost always be approximated by the order with Eq. (2), using an air density of 1.225 kg m 3 Weibull distribution (Petersen et al.1998); Jamil et al. (158C at sea level) and assuming negligible density vari- (1995) estimated this moderate wind speed threshold to ations (Pryor et al. 2004; Jamil et al. 1995), ignoring alti- 2 be 12 m s 1 or higher. It would therefore be expected tude and temperature variability between sites (which that a Weibull distribution would more realistically could theoretically lead up to an associated 68% air simulate a DMGS distribution than an HM distribution. density variation compared to the average value adopted). Both the 31-yr U.K. HM wind speed and DMGS data A limitation of the applicability of the energy density can be used to assess whether the Weibull distribution quantity is that even the most modern wind turbines function is a good fit to these observations. The HM data cannot harvest power below and above specific wind contains periods of low wind speeds (including many speed thresholds (Table 1). Outside this range, the wind calm hours–periods) that have been highlighted as not speed is either too low to turn the blades or too high,

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1177

21 FIG. 2. The 10th, 50th, and 90th percentiles of annual average HM wind speeds (m s ), 1980–2010, from the 40-station network. forcing the turbine to be shut down in order to prevent and averaged across the network, weighting for any damage (Forster et al. 2011). Based purely on the cubic missing data, and the observed temporal variability is relationship between wind speed and power generation, discussed. energy density returns an overestimation of wind tur- e. North Atlantic Oscillation bine performance, especially during stormy periods such as the early 1990s. For comparison, another method is The HM and DMGS 1980–2010 wind speed database also used to quantify wind turbine performance to second presents an excellent opportunity to investigate the re- order, including cut-in and cut-out wind speed thresh- lationship between the NAO index and U.K. wind speeds olds and sensitivity to wind speed variations within that and assess the impacts of the phase changes of the NAO range (Oswald et al. 2008). For each of the 40 HM ob- on land-based wind measurements and wind energy servation sites, a synthetic state-of-the-art 3.6-MW wind output estimates. This furthers the work of Cheng et al. turbine is considered for the duration of the recorded (2011), who used satellite observations to investigate observations and the 10-m winds are adjusted to the interannual variability of high wind occurrence in the typical hub height of 100 m using the power-law ap- North Atlantic over the period 1988–2009. The par- proximation, ignoring the important effect of variable ticular NAO index used for this study is based on atmospheric stability and surface roughness (z0) for this normalized sea level pressure observations made at simple estimate (Petersen et al. 1998; Motta et al. 2005): Gibraltar and Reykjavik in Iceland, with homogeneous   records that date back to the 1820s, allowing for a long- U(z ) z p term monthly NAO index (Jones et al. 1997) [available 1 5 1 , (3) on the University of East Anglia’s Climatic Research U(z2) z2 Unit (CRU) website, http://www.cru.uea.ac.uk/;timo/ datapages/naoi.htm; hereafter called the CRU website]. where U(z ) and U(z ) are the wind speeds at heights z 1 2 1 There are numerous methods to calculate the NAO in- and z , respectively, and p is the power-law exponent 2 dex; however, this monthly index has the advantage of taken to be equal to 0.14 (Petersen et al. 1998) (giving the longest record, helping place the 1980–2010 U.K. U 5 U 3 1.38) . The value of p typically ranges from 100 10 HM wind variability into context (see http://www.cru. 0.05 (very unstable atmosphere with z 5 0.01 m) to 0.69 0 uea.ac.uk/cru/info/nao/ for more details). (stable atmosphere with z0 5 3 m), with the adopted value 0.14 representing a neutral atmosphere for a small z0 (0.01–0.1 m) and a typical value for areas with vari- 3. Results and discussion able stability (Irwin 1979). Once the height conversion a. Interannual variability has been performed, the power output is then estimated for each hour at each site based on the power output Figure 2 shows a time series of annual average 10-m HM curve of the 3.6-MW wind turbine (Table 1). Energy wind speeds in the form of the 10th, 50th, and 90th density and power output are calculated for each site percentiles, quantifying the intersite variability. The

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1178 JOURNAL OF CLIMATE VOLUME 26

21 FIG. 3. Network average threshold exceedance percentages for 11 and 13 m s HM wind speeds.

10th and 50th percentile 5-yr moving averages exhibit hub height, 3.6-MW wind turbines begin to work at full peaks in the early 1980s and early 1990s, with a general capacity (Table 1). Furthermore, all of the 40 sites in the statistically significant decrease visible over the full network experience such wind speeds, unlike for higher 1980–2010 period (confidence levels of 99.9% and 95% thresholds that are only exceeded at a minority of sites. for the 10th and 50th percentiles, respectively, using Throughout the paper, we have chosen to focus on wind ordinary least squared linear regression analysis). The speed thresholds that are consistent with those high- 90th percentile shows a much more pronounced early lighted by Vautard et al. (2010) and, especially, on those 1990s peak, without the general decline seen in the 10th for which it is known that building damage of varying and 50th percentiles, but with a statistically significant degrees would be expected. It is acknowledged, how- decrease since 1990 (at the 99% level). The 10th and ever, that the latter actually varies with geography ac- 50th percentiles show that in the mid to late 2000s wind cording to build quality as shown by Klawa and Ulbrich speeds began to recover; however, the anomalously low (2003) and so the implications of our threshold results winds of 2010, discussed in detail below, are at odds with should be seen as indicative of potential damage only. this recovery. Figure 2 highlights large year-to-year The proportion of time when the network average 2 variability in wind speeds for all percentiles—for ex- HM wind speed exceeds the 11 m s 1 threshold ranges 2 ample, the median varying from 4.3 to 5.3 m s 1. Our from just over 2% of the time in 2010, due to the cold results and those of other authors highlight the presence and relatively calm months of January and December of strong decadal variability and we include linear trend that year (see 2010 wind speed and direction in Fig. 4d), analyses here only for completeness. Behind these re- to 6.7% in 1990, associated with the storminess of January sults from the network as a whole, it should be noted that and February. The interannual variation is striking 32 of the 40 sites display a decrease in annual mean wind with, for an extreme example, 1986 experiencing winds 2 speed over the full period, 15 of which are statistically in excess of 11 m s 1 for twice as many hours as in the significant (95% confidence level), while 8 show an in- previous and following years, a feature also reported by crease, 2 of which are statistically significant. There is no Vautard et al. (2010) for Europe as a whole, though less 2 clear geographical pattern to the distribution of stations pronounced. The 13 m s 1 threshold exceedances ex- 2 exhibiting statistically significant changes. hibit a similar pattern to that of 11 m s 1, ranging be- To learn more about the nature of winds experienced tween just below 1% and just below 3% also in 2010 and in the United Kingdom over the 1980–2010 period, sev- 1990, respectively. The early 1980s and early 1990s, eral HM wind speed exceedance thresholds were selected particularly the latter, have the highest proportion of and the frequency of exceedance at each site calculated. HM wind speeds over each threshold, with a statistically Figure 3 displays results, expressed as a network average, significant decrease from 1980 (95% and 99% confi- 2 2 for two particular thresholds, 11 and 13 m s 1, a ‘‘strong dence for 13 and 11 m s 1 exceedances, respectively). breeze’’ on the Beaufort scale. These thresholds have The more intense threshold exceedance peak in the early been chosen here because when adjusted to wind turbine 1990s compared with that of the early 1980s is in keeping

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1179

FIG. 4. Network average HM wind roses for (a) 1980–2010, (b) 1986, (c) 1987, and (d) 2010. with the 90th percentile of the HM annual average wind Fig. 5, although the early 1980s peak is not as pro- speed shown in Fig. 2. This reinforces the findings of nounced as in the 10th and 50th percentiles of site HM Wang et al. (2009), suggesting a more volatile wind re- wind speeds shown in Fig. 2. In Fig. 5, in addition to the 2 gime in the early 1990s with more 10-m winds reaching 11 and 13 m s 1 exceedance thresholds shown in Fig. 3, 2 2 in excess of 11 and 13 m s 1 but with a lower average further thresholds of 3, 5, 7, 9, and 15 m s 1 are also wind speed compared to the early 1980s. included. Although the logarithmic scale somewhat re- Figures 2 and 3 reveal a large change between the duces the visual impact of the variability, nevertheless adjacent years 1986 and 1987, with 1986 recording far a statistically significant decrease ($99% confidence) higher wind speeds. To further investigate this difference, over the last 20 years remains visible for exceedance 2 network average wind roses were produced for both years thresholds in the range 7–15 m s 1. As expected, the [Figs. 4b,c; also shown are the 1980–2010 climatology contribution of individual sites to the total exceedance (Fig. 4a) and the extreme year of 2010 (Fig. 4d)], with percentage varies throughout the network, especially as 1986 revealing a much more pronounced tendency for the exceedance thresholds rise and become of interest southwesterly winds. This is to be expected with stron- for the insurance sector. [This is discussed in detail be- ger southwesterly winds associated with the extra- low (section 3e), with Fig. 10a highlighting the site 2 tropical cyclone storm track. Increased southwesterly contribution variations for the 15 m s 1 threshold.] winds are positively correlated with the NAO (Cheng One of the findings of Vautard et al. (2010) was a et al. 2011) and the monthly NAO index is significantly general decline in European wind speeds over the last 30 more positive in January, October, November, and years, especially for extreme winds, whereas U.K. results December in 1986 than in the equivalent 1987 months. presented here more strongly emphasize an early 1990s The peaks of the early 1980s and early 1990s are fur- peak and a marked decline over the last 20 years, high- ther highlighted by the 5-yr running mean of network lighting the importance of not assuming a simple overall average HM wind speed threshold exceedance shown in linear trend. We might not be surprised by this difference

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1180 JOURNAL OF CLIMATE VOLUME 26

FIG. 5. Network average 5-yr running mean HM threshold exceedance percentages for 3, 5, 7, 9, 2 11, 13, and 15 m s 1 HM wind speeds. given the United Kingdom’s location on the edge of threshold 0.5% of days (at all sites), compared to 2001 Europe, more exposed to the Atlantic, compared to the and 2010 when this threshold was not breached at all (not continental scale of the Vautard et al. (2010) study. Re- shown). Lerwick (station 40) and Kirkwall (39), in the sults presented here extend and are consistent with the Northern Isles (Fig. 1), contributed to 16 and 15 days, U.K., NAO, and Grosswetterlagen indices presented by respectively, of the total 69 DMGS values in excess of this Atkinson et al. (2006) and with the broader spatial-scale extreme wind threshold in 1993 (not shown). Note that 2 findings of Wang et al. (2009) and Boccard (2009). 20 m s 1 is generally accepted as a starting DMGS The DMGS exhibits a similar long-term variability to threshold for minor structural damage in connection with that of the HM as depicted by the 5-yr moving average of insurance claims. network average DMGS threshold exceedance shown in Sensitivity tests of the interannual variability of Fig. 6. Higher thresholds are included here compared threshold exceedances to the network configuration have 2 with the HM analysis, ranging from 9 to 35 m s 1, re- been carried out (not shown), based on the removal of the vealing peaks in the early 1980s and early 1990s with the most significant contributor stations to the 15 (HM) and 2 2 exception of the highest 35 m s 1 exceedance threshold, 25 m s 1 (DMGS) exceedance thresholds in Figs. 5 and 6, which does not have such a marked peak in the early respectively. While the removal of these stations leads to 1980s but a more extreme maximum in the running mean inevitable quantitative changes of exceedance percent- 2 around 1991/92. The 35 m s 1 1980–2010 decline is sta- age, the interpretation of the periods of enhanced and tistically significant (with 99% confidence), accommo- reduced exceedance remains unchanged, indicating low dating a peak in 1993, with the wind speed exceeding the sensitivity to specific station choice.

FIG. 6. Network average 5-yr running mean threshold exceedance percentages for 9, 11, 13, 15, 2 20, 25, 30, and 35 m s 1 DMGS.

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1181

FIG. 7. The 1980–2010 network average HM wind roses when NAO index is (a) $2 and (b) #22. b. North Atlantic Oscillation: Driver of temporal c. Intra-annual variability wind climate variations The considerable intra-annual wind variation in the Positive peaks in the NAO index are seen in the early United Kingdom is highlighted in Fig. 8 by the seasonal 2 1980s and particularly in the early 1990s when the 10-yr network averages of HM wind speed for 15 m s 1 thres- Gaussian-weighted filter was at its highest during the hold exceedances. The winter peak of HM wind speeds 2 whole 189-yr time period (CRU website). The decrease exceeding 15 m s 1 during the early 1990s is apparent, since the early 1990s is apparent, and partly explains the displaying the impact of the associated intense winter declining tendency in HM and DMGS U.K. wind ob- storminess (Wang et al. 2009). The statistically significant servations and DMGSs over the last 20 years as shown in winter decline since 1990 (99% confidence) is particularly Figs. 2, 3, 5, and 6. The winter of 2009/10 had sub- marked, generally following a similar progression to that stantially more negative NAO index than any other of the NAO wintertime series. The winter of 1989/90 2 winter measured during the record (Osborn 2011), ex- witnessed the highest 15 m s 1 threshold exceedance plaining the anomalously low wind speeds observed. percentage of ;3.5%, with the lowest (complete) winter 2 The consecutive winters 1994/95 and 1995/96 produced being in 2009/10, exceeding 15 m s 1 just 0.3% of the the greatest year-to-year contrast since the NAO series time, lower than in most autumn and spring seasons. 2 began in 1823; however, this was not seen in the station The spring 15 m s 1 exceedance percentage (Fig. 8) observations (Figs. 2, 3, 5, and 6), showing that winter generally hovers around 0.5%, peaking at over 1% in NAO index is not the only important factor contributing 1994. Autumn, meanwhile, does not reveal a peak dur- to the U.K. wind regime and hence the importance of ing the early 1990s, but was more extreme instead at the studying intra-annual variability as discussed below. start of the observation period during the early 1980s To investigate the effects that the NAO index varia- and also peaked in the late 1990s before declining once tions have on the observed U.K. wind climate, two network more, partially consistent with the findings of Vautard average wind roses are presented in Fig. 7, highlighting et al. (2010), during 1979–2008, that the most substantial the difference in wind speed and direction observed linear decrease in Europe occurred in the autumn season 2 during months when the NAO index is in a strong neg- in this particular period. The relatively high 15 m s 1 ative (#22) and strong positive phase ($2). When the exceedances of the early 1980s in autumn are consistent NAO is in a strong positive phase, observed winds are with the early 1980s peak in U.K. observations (Figs. 2, 3, stronger and very much dominated by the southwest 5, and 6) but are not as apparent in the NAO wintertime sector, whereas during periods of strong negative phase series. Meanwhile, summer season threshold exceed- the speeds are more often lower and the direction is ances remain low and relatively consistent throughout much more evenly spread, with a greater tendency the observation period. From this we can deduce that for northeasterlies. During a negative NAO phase, the the threshold exceedance peak of the early 1980s is as- anomalous increase in pressure over Iceland suppresses sociated with higher winds in both winter and autumn westerly winds, diverting the storm track southward over seasons, whereas the early 1990s peak is caused mainly the Mediterranean and encouraging a more northerly by the winter storminess alone. and easterly flow over the United Kingdom (Hurrell Because the seasonal variation of the HM wind 2 et al. 2003). exceedance threshold of 15 m s 1 is so strong, especially

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1182 JOURNAL OF CLIMATE VOLUME 26

21 FIG. 8. Network average threshold exceedance percentages for 15 m s HM wind speeds during each season, winter [December–February (DJF)], spring [March–May (MAM)], summer [June–August (JJA)], and autumn [September–November (SON)] (note that the winter of 1980 only includes January and February 1980 and the winter of 2011 only includes December 2010). between winter and summer, we show in Fig. 9 the net- to fetch over the Atlantic Ocean and Irish Sea is im- work average wind direction distribution for each season portant, along with the latitude and altitude; the higher over the 1980–2010 period. All of the seasons are dom- and farther north a site is, the stronger the wind due to inated, on average, by winds from the southwest quad- reduced friction and the greater the proximity to the rant, winter unsurprisingly having the strongest such higher storm track density region to the south and east of winds, associated with the storm track moving south Iceland (Dacre and Gray 2009). Surface roughness and during the Northern Hemisphere winter (Dacre and vegetation also play key roles as highlighted by Vautard Gray 2009). Autumn has a similar-looking wind rose to et al. (2010). These points in mind, the relative contri- that of winter, whereas summer and spring have different butions of each site to threshold exceedance, especially appearances, summer having a more influential north- for higher thresholds, are expected to vary significantly. west quadrant (and lower wind speeds overall) and spring Figure 10 shows the relative contributions of each site to 2 a more significant northeasterly component. During the exceedances of HM 15 m s 1 (the speed at which 2 summer the Atlantic westerlies are less dominant with insured property damage begins) and 25 m s 1 wind the storm track pushed north by the Azores high, leading speed thresholds over the period 1980–2010, the circle to climatologically more high pressure systems centered size representing the contribution percentage. The 2 to the west of the United Kingdom producing compara- 15 m s 1 site contributions are dominated by the west tively more northwesterly winds. This means that sum- coast sites exposed to the Atlantic and Irish Sea, such as mer winds are generally less extreme in speed despite the Aberporth (station 13; Fig. 1) and Ronaldsway (27), increase in thunderstorm activity seen in the summer while the two sites farthest north, Kirkwall (39) and and the associated potential for damaging downdrafts Lerwick (40), also make up more than 25% of the ex- (Wheeler and Mayes 1997). Conditions during spring ceedances. This is unsurprising considering that the and early summer are more favorable for blocking sit- latter areas, closer to the Icelandic low, are susceptible uations over northern Europe (Barriopedro et al. 2006), to more intense storms, especially during positive NAO leading to comparatively more wind with a northeasterly (Serreze et al. 1997). Meanwhile the west coast stations component as confirmed in Fig. 9b. experience reduced friction when flow is onshore. This 2 is further highlighted in the 25 m s 1 site contribution d. Spatial variability map (Fig. 10b) with even more weight toward exposed When dealing with the network average of exceed- sites and the most northerly Kirkwall (39) and Lerwick ance thresholds, spatial variability is hidden. Spread (40) stations. across the United Kingdom, the network sites possess Inland sites rarely contribute to either exceedance characteristics that vary considerably, both in topogra- threshold compared with their more coastal neighbors. phy and exposure to the storm track (Fig. 1). Exposure The inland northern sites of Eskdalemuir (31) and

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1183

FIG. 9. Network average HM seasonal wind roses, 1980–2010, for (a) winter, (b) spring, (c) summer, and (d) autumn.

Salsburgh (33) are situated only 50 miles from each flat and open site of Heathrow possesses a similar wind other and have similar altitudes of 242 and 277 m, re- direction distribution to that of the network average spectively; however, Salsburgh contributes far more to with a prevailing southwesterly (Fig. 4a). Table 2 shows 2 the 15 and 25 m s 1 exceedance thresholds (just under the network average proportion of wind direction for 2 10% for each), with Eskdalemuir not exceeding 25 m s 1 each quadrant of the compass, revealing that despite the at all during the 1980–2010 period. Eskdalemuir is situ- southwesterly predominance, there is an easterly com- ated in a north–south-oriented valley, with tree-covered ponent to the U.K. HM wind 38.1% of the time. ridges on either side, whereas the Salsburgh monitoring Wind roses are shown for the directions of HM winds 2 site is located on an exposed grass covered hill with exceeding the thresholds of 15 and 25 m s 1, to confirm a large flat top to the north and east. Centrally located in where the strongest winds originate (Fig. 11). The 2 2 Scotland’s heavily populated central belt, Salsburgh is 15 m s 1 and the 25 m s 1 thresholds are dominated by broadly representative of the insurance risks associated southwesterly winds with the southwest quadrant (1908– with windstorms transitioning across this important 2708) accounting for 59.9 and 78.9%, respectively, as area. The Salsburgh–Eskdalemuir contrast is high- Hewston and Dorling (2011) found for extreme (top lighted in the 1980–2010 HM wind roses in Fig. 10, with 2%) DMGSs. wind direction distribution affected by the site charac- The DMGS 1980–2010 39-site network average wind teristics, meaning that Eskdalemuir is somewhat shel- rose (not shown) is similar to that of the HM (Fig. 4a), tered from the strong westerly winds. Many of the site with the proportion of wind direction for each quadrant characteristics are highlighted by their respective wind (Table 2) also extremely similar. This is the same when roses, with Bala (17) located in a southwest to northeast comparing individual site HM wind roses (Fig. 10) with oriented valley in Snowdonia, dominated by south- equivalent DMGS wind roses (not shown). This suggests westerly and northeasterly winds, whereas the relatively that the factors, whether site aspect, local-scale flow, or

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1184 JOURNAL OF CLIMATE VOLUME 26

21 FIG. 10. Contribution (percentage) of each site to (a) 15 m s (total counts 74 154) and 2 (b) 25 m s 1 (total counts 323) HM wind speed threshold exceedance plus selected all-wind speed 1980–2010 individual site wind roses.

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1185

TABLE 2. Network average HM wind direction, energy density, and highlighting the dangers of simply using the Rayleigh daily maximum gust direction divided into compass quadrants. distribution to describe wind distributions for wind farm Quadrant Percentage Percentage Percentage siting. of wind of wind of energy of DMGS wind The Weibull distribution describes the observed HM direction direction density direction winds well as shown by the histograms in Fig. 12. The Northeast (108–908) 17.9 11.1 17.5 Weibull distribution provides a better fit to the sites with Southeast (1008–1808) 20.2 17.8 19.6 comparatively few low wind speeds, as shown when Southwest (1908–2708) 39.8 51.8 40.2 comparing the sites of Lerwick (40) and Kirkwall (39) to 8 8 Northwest (280 –360 ) 22.2 19.3 22.7 Eskdalemuir (31) and East Malling (8). This is partly due to the method of low value recording in the MIDAS data- 2 synoptic-scale flow, that contribute to the direction of base producing an overrepresentation of 2 kt (1.03 m s 1) HM winds are the same for DMGSs. at certain sites [e.g., Eskdalemuir (31) and Heathrow (10)]. This slightly negatively skews the Weibull distribution e. Application of the Weibull function to describe and affects the k values. It is also due to the nature of the wind speed distributions Weibull distribution best approximating well-measured The spatial variation of wind speeds in the United sites with moderate or high wind speeds (Petersen et al. Kingdom is considerable, as shown above, and this 1998). contrast is also seen when the Weibull distribution is Weibull shape parameter (k) values seem to be fitted to the HM and DMGS data. Figure 12 shows the a function of both the strength of the mean wind and the relationship between the Weibull shape parameter (k) impact of site characteristics. Sites with very low wind and mean wind speed at each of the 40 HM locations, speeds such as East Malling (8) produce low values of k, along with histograms for some prominent sites. Gen- due to the high counts of low wind values; however, erally there is a slight positive correlation (not statisti- other sites with higher means but with anomalous wind cally significant) between mean wind speed and k. The roses (varying greatly from that of the network average, spread of k ranges from ;(1.45 to 2.1), values similar to affected by local site characteristics; Fig. 10) such as Bala those reported in the literature by Celik (2004) based on (17) and West Freugh (30) also have low k (not shown), hourly observations in Turkey (1.1–1.89) and by Pryor associated with topographic effects such as local valley et al. (2004) for buoy measurements around the coast of flows. Sites with low means but evenly distributed (similar North America (1.4–2.5). Different Weibull parameter to network average) wind roses such as Heathrow (10) calculation methods and ways of dealing with zero (Fig. 10) and Nottingham (18) (not shown) have rela- values have an effect (see section 2c), along with the tively high k with regard to mean wind (Fig. 12). Valley fact that the locations used in this study are geo- (22) has high mean wind speed but is located in a valley, graphically heterogeneous, leading to highly varied so local topography affects the wind direction and wind wind regimes. Just 6 out of the 40 sites have k values of speed distributions. more than the commonly used Rayleigh distribution The Weibull distribution does not approximate the value of 2 and the majority of sites range from 1.7 to 1.9, DMGS distribution as accurately as for the HM winds as

21 FIG. 11. The 1980–2010 HM wind roses for exceedances of (a) 15 m s (total counts 74 154) and 2 (b) 25 m s 1 (total counts 323) thresholds (all sites).

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1186 JOURNAL OF CLIMATE VOLUME 26

FIG. 12. The HM wind speeds compared with Weibull shape parameter k for each site plus selected site wind distributions. shown by Fig. 13. The k values are much higher than for f. Wind energy implications the HMs, ranging between ;2.4 and ;2.9, which is unsurprising given that the use of the DMGS metric The HM wind speeds have been converted into net- eliminates many low values. The wind speed threshold work average energy density and potential power output 2 of 12 m s 1 required for a good Weibull fit according (PPO) of a synthetic wind turbine network. Table 2 to Jamil et al. (1995) seems not to be reliable for highlights just how important the southwest quadrant is DMGSs, with sites possessing averages above and be- for wind power production. Both methods show signif- 2 low 12 m s 1, being underestimated for the most fre- icant year-to-year variability of power output over the quent values and overestimated for the lower wind 1980–2010 period (Fig. 14), as originally seen in the speeds (Fig. 13). Generally the tails of the distributions annual average percentile HM wind speeds (Fig. 2), in are well approximated for the higher average DMGS the HM threshold exceedances (Figs. 3 and 5), in the sites and slightly overestimated for the sites with lower DMGS threshold exceedances (Fig. 6), and in the NAO average DMGS. index (CRU website). Peaks in energy density and PPO

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1187

FIG. 13. DMGSs compared with Weibull shape parameter for each site, along with selected site DMGS distributions. are seen in the early 1980s and early 1990s and are clearly levels of energy production in the energy density output displayed by the 5-yr moving averages. The anomalous (Table 1) because of the cubic relationship with wind year of 2010 stands out in both energy metrics, repre- speed. The PPO results are in accordance with those of senting the lowest values of the whole period; the extreme Sinden (2007) during corresponding years of study. In variability of consecutive years 1986 and 1987 is also addition the load factor of 30% is in keeping with the clear. The main difference between the two methods is predetermined value used in the Sinden (2007) study. the more marked peak in the early 1990s in energy This load factor was found by Sinden to approximate the density. The unprecedented storminess described by U.K. wind power output figures well, especially since Wang et al. (2009) of the early 1990s produced the most 1997. extreme winds of the period in the United Kingdom, The range of annual mean PPO is large, 867–1265 kW often above the cut-out speed of even the most modern (2010 and 1986 respectively) with an average of and largest turbines. The 10-m wind speeds of above 1087 kW. During the highest production year, the syn- 2 18 m s 1 are too high to be captured by the 3.6-MW thetic 3.6-MW wind turbine network was working on turbines in the PPO, but account for extremely high average at 35% efficiency (or load factor; with the

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1188 JOURNAL OF CLIMATE VOLUME 26

output (Forster et al. 2011). Winter is the season when electrical power output is most important: with colder temperatures and shorter days, domestic and commer- cial users require energy for heating and lighting. So how does our synthetic wind turbine network simulate sea- sonal PPO variation over the 1980–2010 period? Figure 15 shows the evolution of seasonal mean PPO, highlighting the prominence of the winter season, although it is not as dominant in power production as might be expected given the dominance of winter windiness (Fig. 8). The efficiency of synthetic power harnessed is at its greatest in winter 1995 (47% efficiency), and at its lowest (18%) in summer 1983. PPO is very low in the winter of 2009/10 22 FIG. 14. (bottom) Network average energy density (W m ). and is comparable to the summer averages. This shows (top) Network average potential power output (kW) of a synthetic that storage and backup generation schemes will be- network of 100-m hub height 3.6-MW wind turbines. come crucial to energy suppliers in the future, with ever- increasing reliance on wind power and other renewable sources. assumption of steady winds) and at 24% efficiency for the lowest production year. The year 1986 saw 16% more energy generated than the 1980–2010 average 4. Conclusions and outlook whereas 2010 was 20% below. The energy produced in The characteristics of the U.K. HM and DMGS wind 1987 was just 73% of that of 1986, a much larger dif- regimes, with applications to the insurance and wind ference than the interannual variability in wind energy energy industries, are presented here, based on data from density that Petersen et al. (1998) found across many a 40-station wind monitoring network over the continu- regions in Europe [6(10%–15%)]. This shows that ous 1980–2010 period. The main findings are summarized basing wind farm decisions on a single year of monitored as follows: data can be a dangerous practice (Brayshaw et al. 2011). The demand for electricity in the United Kingdom d The 10th and 50th (but not the 90th) percentile HM fluctuates strongly, varying from hourly to annual time wind speeds have declined significantly over this scales (Po¨ yry 2011). Users need electricity at different specific period, while still incorporating a peak in the times of the year for different reasons (e.g., summer early 1990s. 2010 recorded the lowest annual 10th and cooling demand and warming in winter) (Sinden 2007), 90th percentile and second lowest (behind 1987) 50th which may not match the periods of low and high wind percentile wind speed over the whole 1980–2010

FIG. 15. Network average seasonal mean potential power output (kW) of a synthetic network of 100-m hub height 3.6-MW wind turbines (note that the winter of 1980 only includes January and February 1980 and the winter of 2011 only includes December 2010).

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1189

period (Fig. 2). This is all, however, in the context of Future climate projections have a large spread be- longer-term decadal variability. tween models and low signal-to-noise ratio over Europe d The Weibull distribution is more suited to represent- compared with other midlatitude areas (Hawkins and ing HM winds rather than DMGS distributions at Sutton 2009), Europe being one of the hardest regions typical land-based sites, the former revealing site- for which to predict weather and climate on all time specific shape parameter values ranging from 1.4 to scales (Woollings 2010; Ulbrich et al. 2009). Recent 2.1 (Fig. 12), somewhat in contrast with the often extreme events such as the European winter of 2009/10 assumed k 5 2 Rayleigh distribution, with associated have led to alternative causal interpretations, including implications for turbine site selection. an emphasis on the important role of recent declining d As the HM exceedance thresholds rise, the early 1980s solar output (Lockwood et al. 2010, 2011) and on in- peak in exceedance frequency diminishes, while the ternal dynamical responses to varied forcing (Jung et al. early 1990s peak becomes more apparent (Fig. 5), with 2011). While further research seeks to improve models a declining tendency since, confirming the early 1990s and reduce key uncertainties, both in the prediction of unprecedented peak in northeast Atlantic winter extreme event onset and of persistence, it seems wise to storminess reported by Wang et al. (2009). This is anticipate further significant variability in the U.K. wind not fully consistent with Vautard et al. (2010), who climate and concentrate upon building resilience to this. highlighted a temporally broader decline for the whole of Europe over the period 1979–2008. Acknowledgments. This research was kindly funded d The DMGS exceedance thresholds exhibit similar var- by the Worshipful Company of Insurers, and was carried iations to those of the HM, with the highest thresholds out at the University of East Anglia. Thanks must go to 2 (30 and 35 m s 1) displaying the most marked early the BADC and the Met Office for providing the wind 1990s peak and a decline since (Fig. 6), indicating that speed data. Interested parties wishing to access the ob- the decrease of extreme DMGSs highlighted by served wind speed data may, for research purposes, Hewston and Dorling (2011) has continued through to apply for access through the BADC. Thanks must also 2010, contributing to the reduction in U.K. storm-related go to Ben Webber and Jennifer Graham at the Uni- insurance claims. versity of East Anglia for help with data processing and the development of figures. d The network average 1980–2010 HM prevailing wind direction is in the southwest quadrant (40% of the time). However, significant seasonal and interannual variation is apparent in the relative frequency of all REFERENCES wind directions and this needs to be accounted for in Atkinson, N., K. Harman, M. Lynn, A. Schwarz, and A. Tindal, 2006: wind energy assessments. Long-term windspeed trends in northwestern Europe. Garrad Hassan Tech. Rep., 4 pp. [Available online at http://www.gl- d The 40% frequency in southwest quadrant winds garradhassan.com/assets/downloads/Long_term_wind_speed_ translates into a 51% proportion of energy in the wind trends_in_northwestern_Europe.pdf.] (Table 2). Barriopedro, D., R. Garcı´a-Herrera, A. R. Lupo, and E. Herna´ndez, d The range of network average annual mean potential 2006: A climatology of Northern Hemisphere blocking. 19, power output is significant, from 220% to 116% J. Climate, 1042–1063. ——, ——, and R. Huth, 2008: Solar modulation of Northern around the average, with the synthetic energy pro- Hemisphere winter blocking. J. Geophys. Res., 113, D14118, duced in 1987 just 73% of the previous year, 1986, and doi:10.1029/2008JD009789. 2010 the lowest producing year of all (Fig. 14). Boccard, N., 2009: Capacity factor of wind power realized values vs. estimates. Energy Policy, 37, 2679–2688. The recent variability in U.K. mean wind and gust cli- Brayshaw, D. J., A. Troccoli, R. Fordham, and J. Methven, 2011: mate, including the particularly anomalous atmospheric The impact of large scale atmospheric circulation patterns on circulation patterns of 2010, quantified and discussed wind power generation and its potential predictability: A case 36, here, naturally leads to related questions about the fu- study over the U.K.. Renewable Energy, 2087–2096. Brown, S., P. Boorman, R. McDonald, and J. Murphy, 2009: In- ture, both within the scientific community and from other terpretation for use of surface windspeed projections from the stakeholders. 2010 was an anomalously low wind year, 11-member Met Office Regional Climate Model ensemble. a relatively bad year for wind energy production but UKCP09 Tech. Note, 22 pp. [Available online at http:// a good year for the insurance industry in terms of reduced ukclimateprojections.defra.gov.uk/media.jsp?mediaid=87947& claims volumes. The two sectors are, however, also pos- filetype=pdf.] Cattiaux, J., R. Vautard, C. Cassou, P. Yiou, V. Masson-Delmotte, itively related if one considers the growing underwriting and F. Codron, 2010: Winter 2010 in Europe: A cold extreme role that insurance is now playing, reducing the risk of in a warming climate. Geophys. Res. Lett., 37, L20704, weather-sensitive wind energy revenue streams. doi:10.1029/2010GL044613.

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 1190 JOURNAL OF CLIMATE VOLUME 26

Celik, A. N., 2004: A statistical analysis of wind power density Kremer, E., 1998: Largest claims reinsurance premiums for the based on the Weibull and Rayleigh models at the southern Weibull model. Bla¨tter Dt. Ges. Versicherungsmath., 23, 279–284. region of Turkey. Renewable Energy, 29, 593–604. Lockwood, M., R. G. Harrison, T. Woolings, and S. K. Solanki, Cheng, X., S. Xie, H. Tokinaga, and Y. Du, 2011: Interannual 2010: Are cold winters in Europe associated with low solar variability of high-wind occurrence over the North Atlantic. activity? Environ. Res. Lett., 5, 024001, doi:10.1088/1748-9326/ J. Climate, 24, 6515–6527. 5/2/024001. Dacre, H. F., and S. L. Gray, 2009: The spatial distribution and ——, ——, M. J. Owens, L. Barnard, T. Woollings, and F. Steinhilber, evolution characteristics of North Atlantic cyclones. Mon. 2011: The solar influence on the probability of relatively cold Wea. Rev., 137, 99–115. U.K. winters in the future. Environ. Res. Lett., 6, 034004, Forster, D., M. Benzie, S. Winne, and R. Milnes, 2011: Evaluation doi:10.1088/1748-9326/6/3/034004. of the climate risks for meeting the UK’s carbon budgets. Malmquist, D. L., Ed., 1999: European windstorms and the North AEA Technology plc, Rep. for Committee on Climate Atlantic Oscillation: Impacts, characteristics, and pre- Change ED56732- 3, 151 pp. [Available online at http://hmccc. dictability. RPI Series 2, Risk Prediction Initiative/Bermuda s3.amazonaws.com/Progress%202011/ED56732_FinalReport_ Biological Station for Research, 21 pp. FINALv2.pdf.] McCallum, E., 1990: The Burns’ day storm, 25 January 1990. Gronas, S., 1995: The seclusion intensification of the New Year’s Weather, 45, 166–173. Day storm 1992. Tellus, 47A, 733–746. Motta, M., R. J. Barthelmie, and P. Vølund, 2005: The influence of Harrison, G., L. C. Cradden, and J. P. Chick, 2008: Preliminary non-logarithmic wind speed profiles on potential power output assessment of climate change impacts on the U.K. onshore at Danish offshore sites. Wind Energy, 8, 219–236. wind energy resource. Energy Sources, 30, 1286–1299. Munich Re, 2002: Winter storms in Europe (II): Analysis of 1999 Hawkins, E., and R. Sutton, 2009: The potential to narrow un- losses and loss potentials. Munich Reinsurance Company, 72 pp. certainty in regional climate predictions. Bull. Amer. Meteor. Osborn, T. J., 2011: Winter 2009/2010 temperatures and a record- Soc., 90, 1095–1107. breaking North Atlantic Oscillation index. Weather, 66, 19–21. Hess, P., and H. Brezowsky, 1952: Katalog der Grosswetterlagen Oswald, J., M. Raine, and A. Ashraf-Ball, 2008: Will British weather Europas. Berichte das Deutschen Wetterdienstes in der US- provide reliable electricity? Energy Policy, 36, 3212–3225. Zone 33, 39 pp. Petersen, E. L., N. J. Mortensen, L. Landberg, J. Højstrup, and P. F. Hewston, R., 2008: Weather, climate and the insurance sector. Ph.D. Helmut, 1998: Wind power meteorology. Part I: Climate and dissertation, University of East Anglia, 312 pp. turbulence. Wind Energy, 1, 25–45. ——, and S. R. Dorling, 2011: An analysis of observed maximum Po¨ yry, 2011: The challenges of intermittency in North West Eu- wind gusts in the U.K.. J. Wind Eng. Ind. Aerodyn., 99, 845– ropean power markets. Summary Rep., 16 pp. 856, doi:10.1016/j.jweia.2011.06.004. Pryor, S. C., and R. J. Barthelmie, 2010: Climate change impacts on Hurrell, J. W., Y. Kushnir, G. Ottersen, and M. Visbeck, 2003: The wind energy: A review. Renewable Sustainable Energy Rev., North Atlantic Oscillation: Climate Significance and Environ- 14, 430–437. mental Impact. Geophys. Monogr., Vol. 134, Amer. Geophys. ——, M. Nielsen, R. J. Barthelmie, and J. Mann, 2004: Can satellite Union, 279 pp. sampling of offshore wind speeds realistically represent wind Irwin, J. S., 1979: A theoretical variation of the wind profile power- speed distributions? Part II: Quantifying uncertainties asso- law exponent as a function of surface roughness and stability. ciated with sampling strategy and distribution fitting methods. Atmos. Environ., 13, 191–194. J. Appl. Meteor., 43, 739–750. James, P. M., 2007: An objective classification method for Hess and ——, R. J. Barthelmie, N. E. Clausen, M. Drew, I. MacKellar, and Brezowsky Grosswetterlagen over Europe. Theor. Appl. Cli- E. Kjellstro¨ m, 2012: Analyses of possible changes in intense matol., 88, 17–42. and extreme windspeeds over northern Europe under climate Jamil, M., S. Parsa, and M. Majidi, 1995: Wind power statistics and an change scenarios. Climate Dyn., 38, 189–208, doi:10.1007/ evaluation of wind energy density. Renewable Energy, 6, 623–628. s00382-010-0955-3. Jenkinson, A. F., and F. P. Collison, 1977: An initial climatology of Rodwell, M. J., D. P. Rowell, and C. K. Folland, 1999: Oceanic gales over the North Sea. Met Office, Synoptic Climatology forcing of the wintertime North Atlantic Oscillation and Eu- Branch Memo. 62, 18 pp. ropean climate. Nature, 398, 320–323. Jones, P. D., M. Hulme, and K. R. Briffa, 1993: A comparison of Scaife, A. A., and Coauthors, 2012: Climate change projections and Lamb circulation types with an objective classification scheme. stratosphere–troposphere interaction. Climate Dyn., 38, 2089– Int. J. Climatol., 13, 655–663. 2097, doi:10.1007/s00382-011-1080-7. ——, T. Jonsson, and D. Wheeler, 1997: Extension to the North Seguro, J. V., and T. W. Lambert, 2000: Modern estimation of the Atlantic oscillation using early instrumental pressure obser- parameters of the Weibull wind speed distribution for wind vations from Gibraltar and south-west Iceland. Int. J. Clima- energy analysis. J. Wind Eng. Ind. Aerodyn., 85, 75–84. tol., 17, 1433–1450. Serreze, M. C., F. Carse, R. G. Barry, and J. C. Rogers, 1997: Jung, T., F. Vitart, L. Ferranti, and J. Morcrette, 2011: Origin Icelandic low cyclone activity: Climatological features, link- and predictability of the extreme negative and NAO winter ages with the NAO, and relationships with recent changes in of 2009/10. Geophys. Res. Lett., 38, L07701, doi:10.1029/ the Northern Hemisphere circulation. J. Climate, 10, 453–464. 2011GL046786. Sinden, G., 2007: Characteristics of the U.K. wind resource: Long- Justus, C. G., W. R. Hargraves, and A. Yalcin, 1976: Nationwide term patterns and relationship to electricity demand. Energy assessment of potential output from wind powered generators. Policy, 35, 112–127. J. Appl. Meteor., 15, 673–678. Swiss Re, 2011: Natural catastrophes and man-made disasters in Klawa, M., and U. Ulbrich, 2003: A model for the estimation of 2010: A year of devastating and costly events. Sigma Rep. 1/ storm losses and the identification of severe winter storms in 2011, 36 pp. [Available online at http://media.swissre.com/ Germany. Nat. Hazards Earth Syst. Sci., 3, 725–732. documents/sigma1_2011_en.pdf.]

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC 15 FEBRUARY 2013 E A R L E T A L . 1191

Takle, E. S., and J. M. Brown, 1978: Note on the use of Weibull Wang, X. L., F. W. Zwiers, V. R. Swail, and Y. Feng, 2009: Trends statistics to characterize wind-speed data. J. Appl. Meteor., 17, and variability of storminess in the northeast Atlantic region, 556–559. 1874–2007. Climate Dyn., 33, 1179–1195. Troccoli, A., K. Muller, P. Coppin, R. Davy, C. Russell, and A. L. Weisser, D., 2003: A wind energy analysis of Grenada: An esti- Hirsch, 2012: Long-term wind speed trends over Australia. mation using the ‘Weibull’ density function. Renewable En- J. Climate, 25, 170–183. ergy, 28, 1803–1812. UKMO, cited 2011: Met Office surface data users guide. [Available Wheeler, D., and J. Mayes, 1997: Regional Climates of the British online at http://badc.nerc.ac.uk/data/ukmo-midas/ukmo_guide. Isles. Routledge, 437 pp. html.] Wilks, D. S., 1990: Maximum likelihood estimation for the gamma Ulbrich, U., G. C. Leckebusch, and J. G. Pinto, 2009: Extra-tropical distribution using data containing zeros. J. Climate, 3, 1495–1501. cyclones in the present and future climate: A review. Theor. Woollings, T., 2010: Dynamical influences on European climate: Appl. Climatol., 96, 117–131. An uncertain future. Philos. Trans. Roy. Soc., 368A, 3733– Vautard, R., J. Cattiaux, P. Yiou, J. The´paut, and P. Ciais, 2010: 3756, doi:10.1098/rsta.2010.0040. Northern Hemisphere atmospheric stilling partly attributed to ——, M. Lockwood, G. Masato, C. Bell, and L. Gray, 2010: En- an increase in surface roughness. Nat. Geosci., 3, 756–761, hanced signature of solar variability in Eurasian winter climate. doi:10.1038/ngeo979. Geophys. Res. Lett., 37, L20805, doi:10.1029/2010GL044601.

Unauthenticated | Downloaded 10/04/21 09:19 PM UTC