Over the last century, the earth’s climate has beenwarming(Wang and Ye, 1995; IPCC, 2007). Globalclimate changes from 1906 to 2005 have led to increasesin the annual mean surface temperature of approximately0.74℃(0.56–0.92℃), with a slightly largerincrease in China compared to that of the global average.Climate model studies indicate that climatewarming trends will continue over the next 50–100years(IPCC, 2007; PRCC, 2007). Extreme climatechange due to global warming is a unique and valuableresearch area(Yan and Yang, 2000). Researchregarding changes in maximum wind speeds(WS_max)at the national level is important to sustainable utilizationof wind energy and mitigation of disasters causedby extreme winds. It is also useful for planning windenergy industry development, maintaining electricitygrid, and predicting the price of wind electricity.

Relevant studies have focused on two areas regardingthe near-surface(10-m level)annual, seasonal, monthly, and daily changes of WS_avg(average wind) and WS_max. The first area involves observational dataanalysis of the cause of wind changes, and the secondarea involves mathematical simulation and projection.Regarding observational data analysis, previous worksuggested that there had been a decreasing trend of annualWS_avg over the past 50 years(Wang et al., 2004;Ren et al., 2005; Zhu and Liang, 2005; Xu et al., 2006;Zhou and Yu, 2006). Li et al.(2007)further indicatedthat the mean wind energy density in China between1960 and 2000 exhibited declining trends, on both annual and seasonal scales. Yan and Yang(2000)showedthat the maximum wind speed decreased by 1–3 m s^-1during the last few decades. Jiang et al.(2010a)indicatedthat WS_avg showed declining trends over broadareas of China, except the regions from the Hetao areato Yunnan Province and Guangxi Region. The annualWS_avg over China showed a decreasing trend with valuesof –0.12, –0.10, and –0.02 m s^-1 per decade in thesounding data, the NCEP/NCAR and the ERA40 reanalysisdata at 925 hPa, respectively, over the last 50years. The primary reason for this decreasing trendis changes in the atmospheric circulation. Li et al.(2011)suggested that the major cause of the reducedwind speed observed during the past half century is thedecreasing sea level pressure(SLP)gradients in NorthChina. Most of the above-mentioned studies have focusedon WS_avg in China, while some of the studiesaddress WS_max but do not provide a clear explanationfor the related changes. None of the previous studiesinclude a future projection of WS_max(such as for the21st century). Therefore, this study aims to examinethe WS_max changes and the underlying reasons inboth present and future China.

Since the 1990s, research on climate models hasdeveloped rapidly. Climate models have been employedto simulate and project variations in temperature and precipitation(Qian and Lin, 2004; Cook and Vizy, 2006; Phillips and Gleckler, 2006; IPCC, 2007;Zhang et al., 2008; Zhou et al., 2008; Yang et al., 2010;Hao et al., 2011). Scientists also used climate modelsto reproduce and project changes of near-surface wind and the corresponding energy. For example, Pryor etal.(2005), Rockel and Woth(2007), and Bloom etal.(2008)independently examined variations in nearsurfacewind and wind power density during the 21stcentury over northern Europe, Europe, southwesterncoast of North America, and Mediterranean region.Jiang et al.(2009a, b, 2010b, c, d)projected changesin WS_avg and wind power density over China. However, there have been few reports on the projectedchanges of WS_max in China. In this paper, basedon features of future WS_avg changes, we analyze and project future changes in near-surfaceWS_max throughoutChina.2. Data and methods2.1 Data processing2.1.1 Observation data

To investigate the effect of urbanization on longtermwind variations, two observational datasets wereused. One dataset is from 745 basic and st and ardweather stations, and the other is from 2425 genericweather stations. Both datasets conform to the guidelinesof the WMO global observation systems(WMO, 2004) and the China Meteorological Administration(CMA)technical observation st and ards(CMA, 2003).Wind data from 535 stations, with no missing recordsfrom 1956 to 2004, were used. These stations areevenly distributed, with the exception of a sparse distributionover the Tibetan Plateau. These datasetstogether with dust storm data and sounding data from140 stations were obtained from instruments operatedby the NMIC(National Meteorological InformationCenter of CMA). The number of typhoons originatedin western Pacific that either made l and fall or influencedwind speeds was also obtained from the NMIC.Cold wave days were defined based on the NationalClimate Center of CMA st and ards(Wang and Ding, 2006).

For analysis of wind speed characteristics in themid and low troposphere, wind speed at the surfacelevel, 925, 850, and 700 hPa from the NCEP and ERA40 reanalysis datasets were also used(seeTable 1).

Table 1. Data time periods under SRES A2 for various simulation and observation datasets

Table options

The near-surface layer is defined as 10 m aboveground, and WS_avg represents the average wind speedfrom four-time daily observed values and was averagedover a particular time period(day, month, oryear). The maximum value from all 10-min averagewind speeds in a particular time period(day, month, oryear)is defined as observed(daily, monthly, or yearly)WS_max. Gale days are defined as days with wind speed≥ 17 m s^-1, as determined with direct measurements, or those at the eighth grade or higher, as determinedwith visual observations when there is instrumentfailure.2.1.2 Model data

Since the three global climate models, i.e., GFDL−CM2−0 from the NOAA Geophysical FluidDynamics Laboratory, CCCMA−CGCM3 from theCanadian Centre for Climate Modeling and Analysis, and MRI−CGCM2 from the Japan Meteorological ResearchInstitute, have strong ability to simulate currentwind speeds in China(Jiang et al., 2009b), thesethree models with an emission projection of SRES A2were employed to project annual and seasonal WS_maxvalues over China for periods 2046–2065 and 2080–2099.

Considering the limitations of the simulated outputdatasets from the GCMs, the yearly(or monthly)WS_max is defined as the maximum value among thedaily average wind speeds. EBGCM denotes the ensembleof the above three global climate models.2.2 Computation method

Global climate models(GCMs)differ in their resolution.For comparative purposes, outputs of GCMs and observations of related stations were interpolatedonto 1.0°×1.0° grids using the inverse distance weightmethod.

To determine the 21st century WS_max, projectionsof each model were compared to corresponding20th century data, and the difference of wind speedwas calculated between the 20th and 21st century. Toexamine the long-term changes of WS_max and galedays, we use linear regression trend analysis(Wei, 1999)followed by a significance test of the WS_maxtrend in terms of space-time variability correlation coefficients.

To address the detailed characteristics of WS_maxchanges in various regions and to compare these valueswith those ofWS_avg(Jiang et al., 2010a), we divide theexperimental area into 12 sub-regions based on topography and physiognomy. The stations in every subregionare evenly distributed. The 12 sub-regions are:(1)Xinjiang, (2)west Inner Mongolia, (3)east InnerMongolia, (4)Northeast China, (5)Tibetan Plateau, (6)west–central China, (7)east–central China, (8)the northeastern coast of China, (9)Yunnan, (10)Guizhou, (11)South China, and (12)southeasterncoast of China.

The WS_max in January, April, July, and Octoberwas used to represent seasonal values of winter, spring, summer, and autumn, respectively.3. Observed maximum wind changes in China during 1956–2004 and associated causes3.1 Variations of maximum wind

We found that the 50-yr annual WS_max exhibiteda noticeable decline(the solid line in Fig. 1a)ofapproximately –1.46 m s^-1 per decade, with a 23.6%decrease per decade and a significant linear trend(atthe 0.001 level). The annual WS_avg over China declinedafter 1956 by approximately –0.124 m s^-1 perdecade, with a 5.17% decrease per decade(Jiang et al., 2010a). The amplitude of the WS_max decrease islarger than that of the WS_avg decrease. This trend isconsistent with that found by Yan and Yang(2000).The mean frequency of annual gale days displayed aremarkable decrease over the last 50 years, with –3.0gale days per decade, corresponding to a –12.7% decreaseper decade.

Fig. 1. Observed annual(seasonal)maximum wind(m s-1) and gale days(day)in China during 1956–2004.(a)Annual maximum wind(solid line) and gale days(dottedline), (b)seasonal maximum wind, and (c)seasonal galedays.

Figure options

Seasonally, WS_max declined within the study period, and it decreased more significantly during spring(Fig. 1b). The frequency of gale days decreased as well(Fig. 1c), with spring exhibiting the strongest winds, the highest gale frequency, and a marked decrease inthe value of strongest wind speed and the number ofwindy days. In this case, WS_max does not follow thetrends of WS_avg.

Both annual WS_max and WS_avg decreased overmost of China during 1956–2004(see Fig. 2). The decliningtrend of WS_max is more significant than thatof WS_avg. However, there is a difference in the trenddistribution between WS_max and WS_avg. The yearlyWS_avg exhibited a large decline across China, exceptin the Hetao area southward to Yunnan and Guangxi and in eastern Tibet, where there were no significanttrends in wind change(Jiang et al., 2010a). Examinationof the distribution of WS_avg showed that the zoneswith a decline of WS_avg(< –0.12 m s^-1 per decade)also exhibited greater long-term average wind speeds(> 2.2 m s^-1), and the zones with no noticeable trendscorresponded to those of multiple-year average weakwinds, implying that the yearly WS_avg decreases occurredprimarily in regions with greater wind velocities(Jiang et al., 2010a). However, WS_max has differenttrend distribution characteristics. WS_max exhibited anobvious reduced trend across China. WS_max showedrelatively small decreases(approximately –0.8 m s^-1per decade)from western Inner Mongolia to SichuanProvince. WS_max exhibited relatively large decreases(> –1.6 m s^-1 per decade)in the regions from easternInner Mongolia to Fujian and Hainan provinces and across most regions of Northeast China. The regionswith the smallest and largest WS_max decreasesmismatched the regions with the smallest and largestdecreases in WS_avg.

Fig. 2. Distributions of observed maximum and average wind speeds and trends during 1956–2004.(a)Maximum wind(m s-1) and (b)trend of the maximum wind speed(m s-1(10 yr)-1).

Figure options

To further investigate regional wind characteristicsduring 1956–2004, the experimental area(i.e., China)was separated into 12 sub-regions(Jiang et al., 2010a). Both annual WS_max and the frequencyof annual gale days exhibit declining trends(Table 2).WS_max decreased over northern(sub-regions 1–4) and southeast coastal China(sub-region 12), where themulti-year average wind speed is significantly greaterthan that in sub-regions 6 or 10. This is similar tothe results of the annual WS_avg, but sub-regions 6 and 10 show decreases in WS_max and no significanttrend or a slight increasing trend for WS_avg. However, the frequency of gale days differs between cases(Table 2), e.g., changes in WS_avg are closely related tochanges in WS_max but not clearly related to the frequencyof gale days. The latter denotes the frequencyof extreme wind rather than the wind speed value.WS_max over sub-regions 1–4 is mainly impacted bycold waves during winter and spring. The decliningtrend over sub-regions 1–2 is more obvious thanthat over sub-regions 3–4. This finding represents theslightly different changes of cold wave intensity betweenthe western(1–2) and eastern(3–4)sub-regions.For the regions along the southeast coast(8 and 12), WS_max is primarily impacted by typhoons from theNorthwest Pacific Ocean that either make l and fall orinfluence wind speeds over l and . The decrease inWS_max over the southern sub-region(12)is more significantthan that over the northern sub-region(8), which is related to the decrease of typhoon intensity and frequency as well as the trajectories of typhoonstraveling northward(Lei et al., 2009; Wang et al., 2006).

Table 2. Annual maximum wind speed, average wind speed, number of gale days and their trends for the 12sub-regions based on observation data in 1956–2004

Table options

Wind instrument renewal occurred in China duringapproximately 1969–1970, consistent with a timeof significant changes in average wind measurementscompared with wind measurements before and afterthe renewal events(Jiang et al., 2010a). However, thechange in maximum wind speed during the shift periodis not clear. Ignoring the impact of nationalwind instrument renewal from 1969 to 1970, the maximumwind speed and gale day frequency significantly and gradually decreased from 1971 to 2004(Fig. 1a).Additionally, the small number of stations undergoinginstrument renewal and shifting did not affect thetrends for the other 535 stations in the country. Thus, the effects of instrument renewing and station shiftingare unlikely to significantly influence WS_max changingtrends.3.2.2 Impact of urbanization

To explore the long-term effect of urbanizationnear climate stations on wind changes by using themethod presented by Jiang et al.(2010a), we selected174 of the 535 stations located near highly urbanizedareas(urban stations) and 180 stations surrounded byareas with little or no urbanization(rural stations).These two sets of stations are evenly distributed, respectively, with the exception of a sparse distributionover the Tibetan Plateau. A comparison of the longtermchanges in annual WS_max(Fig. 3a)between theurban and rural stations reveals that WS_max of urbanstations was commensurate with that of rural stationsbefore the 1970s, where both showed slightly decliningtrends over time. However, after the 1970s, WS_maxshowed a prominent decreasing trend at both urban and rural stations. Additionally, the reducing trendwas larger at urban stations than at rural stations(–1.75 versus –1.30 m s^-1 per decade). This is becausethe WS_max trends did not differ initially between urban and rural stations due to the low constructionfrequency around both types of stations before the1970s, but after the 1970s, the urban wind decline wasslightly greater than the rural wind decline due to enhancedurbanization that occurred rapidly post 1970in China. However, even the stations that underwentlittle or no urbanization displayed a declining trend ofannual WS_max. This suggests that urbanization is unlikelyto be a major factor affecting the yearly WS_maxdecline in China, which is consistent with the findingsof Jiang et al.(2010a)regarding WS_avg and those of Li et al.(2011)regarding extreme wind speeds in Beijing.

Fig. 3. Observed annual maximum wind speed(a; m s-1) and frequency of gale days(b; day)over the urban stations(solid line) and rural stations(line with squares)during 1956–2004.

Figure options

The average wind trends over China were calculatedfor winds measured at 10 m above ground and atdifferent levels of the troposphere using NCEP/NCAR(1980–2004), ERA40(1980–2001), and atmosphericsounding(1980–2004)data. Despite the differencesin these data sources, consistent decreasing trends inwind speed are shown from 150 hPa to the surface.The decreasing wind speed detected from atmosphericsoundings was the most significant at 250 hPa, witha decrease of –0.23 m s^-1 per decade. The decliningwind trend gradually receded until the 850-hPalevel, where there was a decrease of –0.04 m s^-1 perdecade(Zhang et al., 2009). The three datasets obtainedfrom NCEP/NCAR, ERA40, and atmosphericsoundings yielded similar conclusions and exhibitedsimilar trend values at the same level. Therefore, thedecreasing winds in the near-surface layer are relatedto the decreasing winds in the troposphere and lowerstratosphere over China(Fig. 4).

Fig. 4. Wind speed changes over China based on theobserved station data(dark blue), sounding data(red), NCEP reanalysis data(yellow), and ERA-40 reanalysisdata(light blue)in the mid-low troposphere from 1956to 2004.

Figure options

Most of China is affected by the East Asian monsoons.The Asian zonal(meridional)circulation hasweakened over the past 50 years and resulted in weakenedmonsoons, which is consistent with the resultsthat show reduced intensity of the winter and summermonsoons(Guo, 1983; Shi, 1996; Wang et al., 2001;Zhao and Zhou, 2005; IPCC, 2007). This suggeststhat, under the global warming, the air-sea thermalcontrast decreases, the strength of the winter Siberianhigh is reduced(Guo et al., 2001; Guo et al., 2004) and thus, the atmospheric circulation varies accordingly.These factors are most likely the forces thatcause the decline of yearly and seasonal maximal windspeed trends across China.3.2.5 Association of cold waves, dust storms, and extratropical cyclones

Most of China is located in the midlatitudes ofthe Northern Hemisphere. In the last 50 years, the frequencyof cyclones in the midlatitudes has decreased, and cyclone routes show a northward migration trend(Wang et al., 2007), which is closely related to declinesin WS_max and gale day frequency in northernChina. Extremely cold weather is often accompaniedby a sharp temperature drop. During the study period, due to conspicuous warming in many regions ofthe extratropical latitudes, the number of cold wavedays greatly decreased, which was linked to the reducedfrequency of gale days(Wang and Ding, 2006;Qian and Zhang, 2007; Wei and Lin, 2009). Duststorms occur when large volumes of s and and dustare raised by strong winds and lead to sharply diminishedvisibility. The frequency of dust storm days isassociated with the frequency of gale days in much ofnorthern China. The frequency of dust storm dayshas greatly reduced during the last 50 years(Qian et al., 2002; Ding et al., 2005; Wang et al., 2009). Dueto global warming, the frequencies of cold waves, duststorms, and extratropical cyclone days have declined, which is the direct and convincing evidence of decreasingmaximum winds.3.2.6 Influences of northwestern Pacific typhoons along the southeast coast of China

Typhoons in the Northwest Pacific Ocean havean important climatological influence, especially thosethat make l and fall on the southeast coastal regions ofChina. Typhoons often bring disastrous winds. Basedon the observation data, the frequency of typhoons hassignificantly declined(Fig. 5). This finding is consistentwith previous research(Ren et al., 2006; Wang et al., 2006; Wang and Zhang, 2006; Hu and Song, 2009). Previous studies also indicate that the strengthof strong typhoons has weakened during the last 50years(Emanuel, 2005; Wu et al., 2007; James et al., 2008). Figure 5 shows that the gale day frequency insub-regions 8 and 12 correlates with the frequency oftyphoon activity in the Northwest Pacific Ocean withcoefficients of 0.46 and 0.56 at the 99.9% significancelevel, respectively. Thus, the declining trend of boththe frequency and strength of typhoons in the NorthwestPacific Ocean is closely related to the decreasesin WS_max and the frequency of gale days in the southeastcoastal areas of China.

Fig. 5. Changes in the observed annual maximum wind speed(a; m s-1) and frequency of gale days(b; day)forsub-regions 8 and 12 located along the east coast of China, and the number of typhoons in the western Pacific(red line)from 1956 to 2004.

Figure options

Global warming can cause significant extremeevent feature changes(Ding et al., 2002; IPCC, 2007;PRCC, 2007). For example, during the last 50years, studies have shown slight increases in maximumtemperature, significant increases in minimumtemperature, increases in frequency and strength ofextreme precipitation, decreases in dust storm frequency(closely related to increased rainfall), and decreasesin frequency of strong wind events in China(Wang and Zhai, 2004). Li et al.(2011)also showedthat the changes in strong winds or wind speed extremesare due to large-scale climate changes in theregion. Therefore, global warming, especially in themid and high latitudes of the Northern Hemisphere, results in decreasing temperature differences betweenthe Northern and Southern Hemispheres, declinesin atmospheric baroclinicity, decreasing SLP gradient, possible decreases in the frequency of gale and cold wave days, and subsequently, reduced WS_maxin China. This is only the statistical relationshipbetween global(or regional)warming and maximumwind speed changes over China. The mechanism ofmaximum wind speed changes caused by global(orregional)warming is not clear. Further research isneeded to provide a more dynamic explanation.4. Climate model-projected changes in maximum wind speed across China during the late 21st century

Several studies utilizing GCMs over the last severaldecades have indicated that the GFDL−CM2−0, MRI−CGCM2, and CCCMA−CGCM3 models areable to simulate wind(at the 10-m level) and temperaturechanges in China(Wang et al., 2001; Zhou and Yu, 2006; IPCC, 2007; PRCC, 2007; Zhao et al., 2007; Jiang et al., 2009a, 2010b). These three globalclimate models are now used to project wind features, such as WS_max, throughout China during the late 21stcentury.4.1 2046–2065 projection

The annual WS_max throughout China between2046 and 2065, as projected by both GFDL−CM2−0 and MRI−CGCM2, is likely to decrease by –4.80 and –4.55 m s^-1, respectively, relative to the 1981–2000value(7.0 and 6.5 m s^-1, respectively). However, theCCCMA−CGCM3 model projects a slight increase inWS_max of +0.35 m s^-1. The combined results fromthese three models indicate that WS_max will decreaseby approximately –3.0 m s^-1 per year from 2046 to2065 relative to the 1981–2000 value(8.5 m s^-1)(seeTable 3).

Table 3. Wind speed changes(WS_avg and WS_max; m s^-1)as projected by three GCMs and their ensembleunder SRES A2 for 2046–2065 and 2080–2099 relative to 1981–2000

Table options

In each season, both GFDL−CM2−0 and MRI−CGCM2 project WS_max to be lower during2046–2065 by a range from –1.90 to –3.71m s^-1 relative to the 1981–2000 value, with thelargest differences occurring in summer and autumn.The combined results from GFDL−CM2−0, MRI−CGCM2, and CCCMA−CGCM3 projectWS_max to be lower in 2046–2065 by a range from–1.35 to –2.46 m s^-1 relative to the 1981–2000 value.The range of WS_max decreases is more significant thanthat of the WS_avg decreases(Table 3).4.2 2080–2099 projectionThe characteristics of the annual WS_max changeduring 2080–2099 are similar to those during 2046–2065. The GFDL−CM2−0 and MRI−CGCM2 models, as well as the combined model results, show a cleardeclining trend compared to that from 1981 to 2000, while the CCCMA−CGCM3 model shows a slight increasingtrend. The range of WS_max decreases duringthe 2080–2099 period is more significant than that ofWS_avg(Table 3).

Notably, the ranges of the four seasonal WS_maxdecreases during the 2080–2099 period, as projectedby the three GCMs and the ensemble of the models, are similar to the projected ranges during the period2046–2065(Table 3).

Several studies showed that, with regard to globalwarming, the frequency of cold waves accompanyingthe East Asian winter monsoon will decrease duringthe 21st century(Hu et al., 2000; Buech et al., 2003;IPCC, 2007). The decreased frequency of cold waveslikely drives the declines in WS_avg and , in particular, WS_max. The changes in WS_max may also be relatedto specific extreme events, such as cold waves, duststorms, and typhoons. The frequency of cold waveswill likely decrease during the 21st century(IPCC, 2007). Approximately 10 climate models that includeanthropogenic emission data have predicted that, afterCO2 doubling, the frequency of tropical cyclonesover the Northwest Pacific Ocean is likely to decreaseby the end of the 21st century(Zhao et al., 2007) and the frequency of typhoons over the western PacificOcean making l and fall in China might decreaseas well(Knutson and Tuleya, 1999; Knutson et al., 2001; Gao et al., 2002, 2003; IPCC, 2007). Therefore, decreases in WS_max over China during the 21st centuryare related to future decreases of cold waves, and decreases in WS_max over the southeast coastal areasof China during the 21st century are corresponding tofuture decreases of typhoons.5. Concluding remarks

Based on the observation data, it is found thatannual and seasonal WS_max, frequency of gale days, and wind speed in the troposphere and lower stratospheredeclined from 1956 to 2004(especially in winter)in China. The amplitudes of annual and seasonalWS_max decreases were more significant than those ofWS_avg, especially in spring.

Wind instrument renewal, station shifting, and urbanization are not major factors affecting the yearlyWS_max decline in China. The decreasing intensity ofthe East Asian winter and summer monsoons resultsin the decrease in the annual and seasonal WS_max and WS_avg. The distinct decrease of cold waves in Chinais a key factor for the distinct decrease in annual and seasonal WS_max in whole China, and the declining frequencyof typhoons making l and fall were key factorsfor the distinct decrease in annual and seasonalWS_maxin the southeast coastal areas of China between 1956 and 2004.

The annual and seasonal WS_max in China during2046–2065 and 2080–2099 are likely to decreaseconsistently relative to those of the 1981–2000 period, as projected by the GFDL−CM2−0, MRI−CGCM2, and EBGCM models. The three models also projectthat the annual and seasonal reductions in WS_max inthe 21st century will be much larger than those ofWS_avg. In the 21st century, the decreasing maximumwind speed over China may be related to decreases inthe intensity of the East Asian winter monsoon and the decline of cold wave frequency. In addition, thedecreasing maximum wind speed over the southeastcoastal areas of China may be related to the decreasing number of tropical cyclones over the NorthwestPacific Ocean.