Effects of land use and cover change on the near-surface wind speed over China in the last 30 years

Abstract

The changes of near-surface wind speed (SWS) were induced by the combination effects from anthropogenic activities and natural climate changes, thus the research of the long-term changes and cause of SWS is very important to recognize the effects from natural changes and anthropogenic activities to SWS. Some studies have shown that land use and cover change (LUCC) over China was very distinct in the last 30 years. However, the possible effects of LUCC to the slowdown in SWS are still uncertain. The European Centre for Medium-Range Weather Forecasts Reanalysis from January 1989 onward (to be extended back to January 1979) (ERA-Interim hereafter) dataset includes little information about LUCC, so the effects of LUCC on SWS in China during 1979–2010 have been estimated by using the difference between SWS per the ERA-Interim dataset and that from 492 meteorological stations. The results show the following. (1) The effects of LUCC on the SWS are distinct, and could have caused a decrease in the SWS of 0.12 m s^–1 per decade, which could account for decreases of 0.57 and 0.30 m s^–1 wind speed for large and small cities, respectively. In addition, a decrease of 0.1 m s^–1 in the SWS could be induced by a 10% rise in the urbanization rate. (2) The impacts of LUCC on the SWS in the Beijing-Tianjin-Tangshan region, the Yangtze River Delta region and the Pearl River Delta region are more significant than those for the entire region. (3) The bias in the ERA-Interim dataset could cause a 51% error in the estimation for the entire region during the study period, but had a non-significant effect on the decreasing trend in the SWS. The decrease rate of SWS induced by LUCC based on the traditional observation minus reanalysis method had an error of 0.01 m s^–1 decade^–1. (4) The results of different methods to quantify the impacts of LUCC on the SWS are also compared in this study.

Keywords

Land use and cover change near-surface wind speed observation the ERA-Interim dataset China

I Introduction

The near-surface wind speed (SWS) is driven by the pressure gradient, boundary layer mixing and surface friction. Some studies show a reduction in the SWS of –0.11 m s^–1 decade^–1 over continental areas of the Northern Hemisphere during the past few decades (Berrisford et al., 2015; Peterson et al., 2011; Tobin et al., 2014; Vautard et al., 2010, 2012). In Central Asia, Eastern Asia and North America, the annual mean SWS decreased on average at a rate of –0.16, –0.12 and –0.07 m s^–1 decade^–1, respectively, and strong winds showed a greater decrease than weak winds (Vautard et al., 2010). Remarkably, a reduction in wind speed was also reported for many local regions (McVicar and Roderick, 2010; McVicar et al., 2010; Pryor et al., 2009; Rayner, 2007; Roderick et al., 2007); for example, the 10th and 90th percentile hourly mean wind speeds declined significantly over the UK during 1980–2010 (Earl et al., 2013), and the change in annual average wind speed was –0.14 m s^–1 decade^–1 over Turkey during a 32-year period (Dadaser-Celik and Cengiz, 2014). Azorin-Molina et al. (2014, 2016) noted that the SWS had a downward trend of –0.016 m s^–1 decade^–1 from 1961 to 2011 and that the average daily peak wind gust frequency declined by 1.49 days decade^–1. A reduction in SWS was also found in France from 1974 to 2002 (Najac et al., 2009, 2012), in the Czech Republic from 1961 to 2005 (Brazdil et al., 2009), in Canada from 1945 to 1995 (Tuller, 2004) and in North America from 1979 to 1999 (He et al., 2010).

In China, a slowdown in the SWS was also reported. Hu et al. (2011) found SWS showed a distinct downward trend from 1961 to 2007. Jiang et al. (2010) reported that the annual mean SWS declined by 0.12 m s^–1 per decade in the last 50 years. You et al. (2010) noted that both the surface observations and the National Center for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR hereafter) reanalysis dataset over the Tibetan Plateau revealed significant decreasing trends, with rates of –0.24 and –0.13 m s^–1 decade^–1, respectively. McVicar et al. (2010) showed a pronounced decrease in the SWS in central China from 1960 to 2006 and that the SWS declined more rapidly at higher than at lower elevations. Wu et al. (2016a) revealed a distinct decrease in the SWS in station observation data, with a rate of –0.23 m s^–1 decade^–1 over the Eastern China Plain (ECP) region from 1980 to 2011. At the same time, Zha et al. (2016) indicated that the observed SWS decrease was primarily attributed to a decrease in strong wind episodes in the ECP region.

The reduction in the SWS has been thoroughly demonstrated in previous studies, but the reasons for the SWS decrease remain unclear. Some studies showed that the decrease in the SWS was attributed to variation in the large- or regional-scale circulation (Dadaser-Celik and Cengiz, 2014; Jiang et al., 2010; McVicar et al., 2012; Sušelj et al., 2010; Xu et al., 2006). The variability of the circulation fields can induce changes in the pressure–gradient force (PGF), and the PGF is the driving force of wind motion. Therefore, some studies also addressed the reduction in SWS induced by the PGF. Klink (2007) showed that a monthly variation in the SWS was primarily determined by the north–south pressure gradient, which accounted for between 22% and 47% of the SWS variability over America. Guo et al. (2011) noted that the main cause of weakening winds has been shown to be a weakening of the lower-tropospheric PGF in China from 1969 to 2005. You et al. (2010) speculated that the most likely cause of diminishing wind speed was the asymmetrically decreasing latitudinal surface pressure gradient over the Tibetan Plateau. Wu et al. (2016b) quantitatively calculated PGF values of 925 and 850 hPa in the ECP region, and found the phase of wind speed was consistent with that of the PGF; at the same time, the correlation coefficient between the wind speed and PGF can be as high as 0.9; nevertheless, the changes in the wind speed were not in agreement with the changes in the PGF at the near-surface level. Other additional probable causes for a reduction in the SWS include an increase in aerosol particles (Jacobson and Kaufman, 2006), the extraction of wind power to generate electricity (Miller et al., 2011), discontinuities in the ground wind time series (Thomas and Swail, 2011), etc. A number of possible causes of global stilling have also been reported by McVicar et al. (2012).

It is interesting and worth noting that an increasing trend in the SWS has been observed over the ocean, but a decrease has been observed over the land (McVicar and Roderick, 2010; Vautard et al., 2010). Some studies have advocated that the reduction in the SWS over the land has been mainly induced by an increase in the land surface roughness (Schwiesow and Lawrence, 1982; Vautard et al., 2010). Bichet et al. (2012) uncovered that the globally increasing vegetation roughness length in the ECHAM5 model by factors of 1.5, 2 and 4 decreases the global land annual 10 m wind speed on average by –0.15, –0.26 and –0.55 m s^–1, respectively. Wever (2012) showed that land use and cover change (LUCC), including urbanization, forestation and a decrease in pasture land area, are probable causes of increasing surface roughness in Europe, and a doubling of the local roughness length was found for 1962–2009, with the greatest increase after 1981. Furthermore, 70% of the wind speed trend can be attributed to surface roughness changes. Wu et al. (2016a) found that the drag force, induced by LUCC, showed an increasing trend in the ECP region from 1980 to 2011 and that the decrease of the SWS caused by LUCC was also significant.

It is noteworthy that LUCC can cause a decrease in the SWS, especially over regions in which the surface roughness increases rapidly, but the real issue is figuring out a way to quantify the effect of LUCC. To isolate the effects of LUCC on the SWS, some studies have compared the observed wind speed in cities with those in rural areas (Guo et al., 2011; Jiang et al., 2010; Klink, 1999; Li et al., 2011). However, most stations are located near cities, with only a few in mountainous or remote regions or on small islands. A city generally has only one meteorological station, such as most cities in the Chinese meteorological observation network especially before 2005, so it is very difficult to find a corresponding rural station to compare with an urban station. Hence, reliable conclusions are hard to achieve. Wever (2012) quantified the trends in surface roughness and their effects on surface wind speed observations based on a conceptual boundary layer model (BLM). Wu et al. (2016a) used a frictional wind model (FWM) to separate the impact of LUCC from other factors. However, a disadvantage of the FWM and BLM is that they can only be used in the plain regions, and they ignore some complicated dynamic processes, such as the turbulence in the urban canopy layer, the blocking effect of buildings and turbulent vertical mixing.

Another possible method is the establishment of a reference SWS dataset without LUCC effects, and then, the effects of LUCC on the SWS can be taken as the difference between the observation and the reference data. Obviously, numerical models with good performance can be used to establish this reference wind speed dataset for a large region (Li et al., 2008). Reanalysis data include good long-term natural climate change signals as well as some important anthropogenic signals, such as an increase in greenhouse gases and aerosols. Otherwise, reanalysis data are insensitive to local and regional surface processes associated with different land types (National Research Council (NRC), 2005), and little or no surface data or information about land–surface changes were used in the reanalysis data assimilation process (Lim et al., 2005) because land–surface change is a gradual process and reliable data are difficult to collect in a timely manner for use in the production process of the reanalysis data. Therefore, the difference in the SWS between the observation and the reanalysis data should reflect the effects of LUCC on the SWS. Kalnay and Cai (2003) and Zhou et al. (2004) investigated the effects of LUCC on near-surface climate change by comparing observations with NCEP/NCAR reanalysis data. Frauenfeld et al. (2005) and Lim et al. (2005) also discussed the same subject by comparing observations with a 45-year European Centre for Medium-Rang Weather Forecasts Reanalysis from September 1957 to August 2002 (ERA-40 hereafter) reanalysis data. Hence, the difference between observational and reanalysis data (observation minus reanalysis (OMR)) could be used to investigate the impacts of LUCC on the global and regional climate (Kalnay and Cai, 2003).

In the last 30 years, the gross domestic product of China has increased at an average annual rate of 9.5%, compared with 2.5% for developed countries and 5% for developing countries (Zhou et al., 2004). In the meantime, China had lost more than 10% of its cropland and showed an expansion in urban area of greater than 20% (Liu and Tian, 2010). Liu et al (2014) showed that approximately 3.18 × 10⁶ hm² of cropland were used for construction from 1990 to 2010; in particular, the built-up land increased by 1.76 × 10⁶ hm² before 2000, and the built-up land expansion accelerated after 2000 and increased by 3.76 × 10⁶ hm². Asselen and Verburg (2013) used a land change model on a global scale to simulate the intensification of LUCC, and found that LUCC over China was very distinct in the last 30 years. However, the possible effects of LUCC on the slowdown in the SWS are still uncertain.

Wind speed changes may affect the variability of wind energy operations (Li et al., 2008; Pryor and Barthelmie, 2010; Pryor et al., 2006), so an investigation of the wind speed change is essential for surface flux estimation, wind power estimation and wind risk assessments (He et al., 2010). Meanwhile, changes in wind speed can also affect soil moisture, evaporation and water resources (McVicar et al., 2008, 2010, 2012; Rayner, 2007) and may also influence the evolution of arid and semi-arid environments (Okin et al., 2006). Therefore, it is worth investigating changes in the wind speed and quantifying the effects of LUCC on the SWS. To explore the impacts of LUCC on the SWS in China, the ERA-Interim dataset reanalysis is used as the reference data in this work, and the effects of LUCC on the SWS were estimated using the difference in the wind speed between meteorological observations and the reanalysis data. The datasets and methods used for this study are discussed in Section II. Section III describes the results and analysis. A discussion is provided in Section IV, and the conclusions are summarized in Section V.

II Data and methods

Wind speed data were used at a height of 10 m above the surface, measured four times per day (at UTC 00, 06, 12 and 18) from the ERA-Interim reanalysis dataset at a 0.75°×0.75° resolution from 1979 to 2010. The ERA-Interim dataset was produced with a sequential four-dimensional variational assimilation scheme and can be advantageous, especially for those stations in topographically complex, data-sparse areas (Dee et al., 2011; Frauenfeld et al., 2005; Rabier et al., 2000). Comparison with the ERA-40 dataset (Uppala et al., 2005), the representation of the hydrological cycle, the quality of the stratospheric circulation data and the consistency in time of the reanalysis fields have been improved in the ERA-Interim dataset (Dee et al., 2011). Berrisford et al. (2011) demonstrated that the ERA-Interim reanalysis dataset had superior quality to the ERA-40 dataset and that the wind data were improved in the ERA-Interim dataset. In addition, the land surface observations assimilated in the ERA-Interim dataset include only the 2-m temperature, the 2-m humidity and snow (Poli, 2010). That is to say, little land surface observed data were used in the data assimilation process for the ERA-Interim dataset. In addition, Dee et al. (2011) also showed that the near-surface wind observations over land were not selected and that surface pressure observations over high terrain with elevations higher than 1500 m were also not selected for assimilation into the ERA-Interim reanalysis data. The effects of the regional geographical distribution were controlled by the standard deviation of a small-scale orographic dataset, and the wind speed was better verified with respect to surface synoptic observations in mountainous regions (Dee et al., 2011). In brief, the data in the ERA-Interim dataset has self-contained large-, meso- and small-scale characteristics, but the effects of the urban underlying surface variability and LUCC on the SWS are not involved. The ERA-Interim reanalysis data has often been used in studies of climate change and can capture inter-annual and inter-decadal changes very well (Simmons et al., 2007, 2010, 2014).

The daily mean wind speed data from 514 climatic stations in China from 1979 to 2010 were also compared with the SWS from the ERA-Interim dataset. These are standard national ground meteorological stations, and the wind speed is measured by a cup anemometer at a standard 10 m exposure height above the ground according to the China Meteorological Administration (CMA) on the Observing System and Technical Regulations on Weather Observations (CMA, 2003). Missing data were found for 181 stations during the study period, which accounts for 35.2% of the total stations used in our research, and these 181 stations were mainly located in the middle reaches of the Yangtze River and the Yellow River. In addition, the total days of missing data account for only less than 1% of the length of total data series, and no continuous missing data days are present. An investigation of the homogeneity of the annual wind speed at 347 stations selected from all of the stations in China has been conducted (Liu, 2000), which included the stations used in this study. All of the points of discontinuity in the annual SWS dataset from the 347 stations were found. The results showed points of discontinuity in the SWS dataset included replacement of the wind measurement instrument, relocation of the station and changes in the observation height (Liu, 2000). At the same time, all of the points of discontinuity existing in the SWS dataset found by the homogeneous test, the extreme test and the temporal consistent test carried out by the Chinese National Meteorological Information Center (NMIC) have been calibrated. Therefore, the SWS dataset is considered to be a credible dataset (CMA, 2003). To maintain the continuity of the dataset, the SWS for missing data days was interpolated by using the day before and the day after the day with the missing data in our study. Only the 492 stations that passed a significance t-test of the correlation coefficient between observation and the ERA-Interim dataset SWS at a confidence level of 0.01 were used in the following analysis. The terrain height and the spatial distribution of the 492 stations throughout China are shown in Figure 1(a).

Figure 1.

The spatial distribution of the terrain height in China (a) (unit: m), as well as the correlation coefficient between the observation and the ERA-Interim wind speed in 492 stations (b). The threshold of the significance t-test at the 0.01 significance level is 0.13; white circles in (a) represent the sounding stations; Regions 1, 2 and 3 in (a) represent the Beijing-Tianjin-Tangshan region (BTTR), Yangtze River Delta region (YRDR) and Pearl River Delta region (PRDR), respectively; squares and circles in (b) denote large and small cities, respectively.

In addition, we also compared the monthly mean wind speed from 140 sounding stations of 850 and 500 hPa in China from 1980 to 2010 with the ERA-Interim dataset wind speeds. These 140 sounding stations are shown in Figure 1(a). The continuity of the sounding wind speeds of 850 and 500 hPa are both good; the total days with missing data only account for less than 5%, and 2% of the length of the total data series for 850 and 500 hPa, respectively. The sounding data were evaluated and processed by the NMIC, and passed the climate extreme value test, the temporal consistency test, the wind shear test and statistical tests. Therefore, the sounding observations are also considered to be a credible dataset.

To analyze the effects of cities of different sizes on the winds, the stations were further classified into large, medium and small cities based on a population size of greater than 1,000,000, between 500,000 and 1,000,000 and under 500,000, respectively (Wu et al., 2012, 2016a, 2016b). The population data for 2005 are the only data currently available, so temporal changes in city scales cannot be considered in this study. In addition, the population data and the urbanization rate are derived from the National Bureau of Statistics of China from 1979 to 2010. The urbanization rate is defined as the ratio between the urban population and the total population and is regarded as an index of urbanization (Wu et al., 2016a, 2016b; Zhou et al., 2004). To focus on the difference in the rough underlying surfaces between different cities and to better highlight the impact of LUCC on the SWS, we primarily considered the differences in the SWS among 136 large cities and 240 small cities (Figure 1(b)), because these differences should reflect the effects of LUCC on the SWS to the greatest extent.

All the ERA-Interim datasets were interpolated to the surface climatic stations over China using the bilinear interpolation method, which has been judged as the most suitable for the transfer of a grid forecast field to another or discrete observation field (Accadia et al., 2003; Mastylo, 2013). Ensemble Empirical Mode Decomposition (EEMD) was also used to obtain the long-term trend in the evolution of wind speed, in which a basis function is unnecessary, so the method can exclude some anthropogenic assumptions that are embodied in the moving average and the linear fitting methods (Huang and Wu, 2008; Wu and Huang, 2009). EEMD has also been demonstrated to be a powerful tool for the analysis of non-linear and non-stationary data in many geophysical fields (Breaker and Ruzmaikin, 2011; Franzke and Woollings, 2011; Qian et al., 2011). To quantify the consistency of the phase between two data series, the probability of extrema appearing at the same time point (represented by the acronym PEST) in the two data series was also calculated (Wu et al., 2016a, 2016b). In addition, the Student’s t-test was used to determine the significance of the data. The linear trend coefficient was calculated using the least-squares method.

III Results and analysis

1 Spatio-temporal variation of SWS in observation and the ERA-Interim dataset

The correlation coefficient for the observations and the ERA-Interim dataset 10 m wind speed is shown in Figure 1(b). The correlation coefficient exceeds 0.4 for most regions of China and is significant at the 99% confidence level according to a t-test. The lowest correlation was found in the eastern part of the Tibetan Plateau and the middle reaches of the Yangtze River. The correlation coefficients of 492 stations were significant at the 99% confidence level, so only these stations were used in the subsequent analyses. There were only 22 stations removed in the total of 514 meteorological stations; the removed stations account for only 4% of total stations.

The spatial distributions of the SWS values from the observations and the ERA-Interim dataset are shown in Figure 2. Higher SWS values can be observed north of the Yellow River, where the SWS was higher than 2 m s^–1, but the SWS was generally lower than 2 m s^–1 south of the Yellow River (Figure 2(a)). The regions with a lower SWS were located in the eastern part of the Tibetan Plateau, the middle regions of the Yangtze River and some regions of Southeast China, in which the SWS is generally lower than 1.6 m s^–1. The regions with a SWS exceeding 2.8 m s^–1 include some parts of Northern China and the Tibetan Plateau. The distinct difference in the SWS between Southern and Northern China per the observations (Figure 2(a)) is similar to those reported by Jiang et al. (2010). The distribution characteristics of the ERA-Interim dataset wind speeds were similar to the observations with consistent maximum and minimum regions (Figure 2(b)), but the ERA-Interim dataset wind speeds were distinctly higher than the observations for almost all of China. In addition, the observed SWS showed a noticeable decline in most parts of China. The rate of decrease in the SWS reached 0.2 m s^–1 decade^–1 in Eastern China, which is more rapid than in other regions (Figure 2(c)). However, the ERA-Interim dataset 10 m wind speed showed an indistinctive decrease, which was non-significant at the 0.01 level (Figure 2(d)). Similar results were also reported by Vautard et al. (2010). The ERA-Interim dataset reanalysis data captured the primary signals of the large-scale circulation well (Simmons et al., 2014; Vautard et al., 2010), and the ERA-Interim dataset 10 m wind speed showed no distinct decreasing trend from 1979 to 2010, indicating that the pronounced decrease in the observed SWS should not be mainly attributed to a weakening of the large-scale circulation field.

Figure 2.

Spatial distribution of surface wind speed in observation (a) and ERA-Interim (b) averaged from 1979 to 2010 (unit: m s^–1), and their linear trend (c), (d) (unit: m s^–1 decade^–1) (light blue, yellow, red in (c) and (d) represents the linear trend passes the 90%, 95%, 99% significance levels, respectively).

The temporal variation of the SWS in the ERA-Interim dataset and the observations is shown in Figure 3. Figure 3(a) indicates that the SWS in the ERA-Interim dataset is higher than the observational data (Figure 3(a)) and an increasing trend in the difference between the surface observations and the ERA-Interim dataset. In addition, the two datasets show good agreement between the inter-annual and seasonal variability with a PEST of 73.3%, indicating that the large-scale climate background cycle field is uniform for the wind speeds in both datasets. A decrease in the observed SWS can be found from 1979 to 2007, but the ERA-Interim dataset wind speed is smooth and did not show a distinct trend during the same period. The difference of the EEMD trends in Figure 3(a) reached a minimum of 0.5 m s^–1 in 1979, and then showed an increasing trend until 2010, and the mean wind speed difference between the ERA-Interim 10 m wind speed and the observed SWS (ERA-Interim 10 m wind speed minus observed SWS) was 0.71 m s^–1 from 1979 to 2010.

Figure 3.

Temporal variability of the surface wind speed (SWS) with Ensemble Empirical Mode Decomposition (EEMD) trend (comparison of observation and ERA-Interim (a); observation SWS in large and small cities (b); (c) is same as (b) but for ERA-Interim).

The SWS for small cities was higher than that for large cities in the observations (Figure 3(b)), with an average difference of 0.33 m s^–1. However, a distinct decreasing trend was also apparent in the observation data for both large and small cities, accompanied by a decrease in the difference between them. A rapid decrease in the SWS in small cities from 1979 to 1993 was also present and was followed by a smaller and more steady decreasing trend after 1993, but simultaneously, the decreasing trend in large cities showed less variation than in small cities. The annual and seasonal cycle of the SWS in large and small cities has the same phase, with a PEST of 80%, indicating that the large-scale wind cycle field has the same effects in large and small cities. Figure 3(c) shows that the phase of the ERA-Interim dataset wind speed in large cities is similar to that in small cities, with a PEST of 81%. At the same time, compared to the observations, a smaller difference between the SWS of the ERA-Interim dataset in large and small cities with the mean difference of 0.01 m s^–1 was presented, and a decrease of SWS in large cities was also presented, but was less than that of the observations. The SWS in small cities was nearly constant during that period, which means that the ERA-Interim dataset might include less information pertaining to LUCC, especially for small cities. Frauenfeld et al. (2005) also showed that significant trends in the station data could reflect extensive land use changes and the industrialization that occurred, but the European Centre for Medium-range Weather Forecast (ECMWF) reanalysis data were not affected at all by land cover changes because a static map was used.

2 Impacts of land use and cover change on the SWS

Figure 3 shows that the long-term decrease of the SWS is not mainly attributable to the variability of the large-scale climate cycle field. In addition, Wu et al. (2016a, 2016b) also found that changes in the PGF, the East Asian monsoon and the regional- or local-scale climate are not the primary factors responsible for the distinct reduction in the SWS. Li et al. (2008) demonstrated that the OMR method could be used to estimate the effects of anthropogenic changes in LUCC on the near-surface potential wind energy in China. Therefore, we thought that the SWS difference between the ERA-Interim dataset and the observations (represented by the SWSD) at these stations could be considered as the representative of the impact of the LUCC, and analyzed the long-term changes in the SWSD.

The SWSD was negative over most of China with an average value of –0.71 m s^–1 (Figure 4(a)), and showed an evident decline over the last 30 years (Figure 4(b)). More distinct changes in the SWSD were observed in Northeast China, the middle and low regions of the Yellow River and the Yangtze River, in which the linear trend of SWSD was –0.2 m s^–1 decade^–1 and was significant at the 0.01 level (Figure 4(b)). The decline of the SWSD in Figure 5(a) is consistent with the results in Figure 4(b). The SWSD was –0.47 m s^–1 in 1979, and reached –0.89 m s^–1 in 2010 with a standard deviation of –0.12 m s^–1 during the study period. The highest decrease of –0.32 m s^–1 decade^–1 was found from 1979 to 1991, followed by a smooth phase of –0.074 m s^–1 decade^–1 from 1992 to 2004, and a rapid decline of –0.20 m s^–1 decade^–1 after 2004.

Figure 4.

The spatial distribution of the SWS difference between the ERA-Interim dataset and the observations (represented by SWSD) (a) (unit: m s^–1) and (b) SWSD linear trend (unit: m s^–1 decade^–1) (light blue, yellow, red in (b) denotes linear trend can pass the significant t-test at the 90%, 95%, 99% confidence levels, respectively).

Figure 5.

Temporal variability of the SWS difference between the ERA-Interim dataset and the observations (represented by SWSD) (a) and mean urbanization rate (b) and the regression between the urbanization rate and the SWSD (c) (R is the correlation coefficient, Rc is threshold of the correlation coefficient, P is the significance level and the red rectangle in (a) represents annual mean SWSD). EEMD: Ensemble Empirical Mode Decomposition.

LUCC in China includes urbanization, forest cutting, farmland irrigation, desertification, returning farmland to forest and other types of changes, so investigating the effects of different types of LUCC on the SWS is difficult. However, most stations in China are located near cities, with only a few in mountainous or remote regions or on small islands (Zhou et al., 2004), so the impact of urbanization on the SWS should be evident (Zhang et al., 2015). To quantify the relationship between urbanization and a decrease of the SWS, the correlation between the urbanization rate and the SWSD was computed based on a regression analysis. The urbanization rate has often been used to indicate trends in urbanization, which is defined as the ratio of the population in cities to the total population of a country (Wu et al., 2016a, 2016b; Zhou et al., 2004). Figure 5(b) shows an increasing trend in the urbanization rate in China, with a mean increase rate of 9.36% decade^–1 from 1979 to 2010. Figure 5(c) shows a distinct negative correlation between the urbanization rate and the SWSD, with a correlation coefficient 0.86, which is significant at the 0.01 level, and a linear relationship with a slope of –0.011. According to the linear relationship, a 10% increase in the urbanization rate could cause a 0.11 m s^–1 decrease in the SWS. In addition, the SWSD of cities of different sizes was also evident: the average SWSD was –0.84 and –0.52 m s^–1 for large and small cities, respectively. The mean SWSD of large cities is lower than that of the entire region mean and small cities at –0.13 and –0.32 m s^–1, respectively.

The potential impacts of urbanization on the SWS, which means the decrease of the SWS during the urbanization process, was shown by the relationship between the SWSD and the urbanization rate, as indicated by a regression analysis. According to the relationship, a larger SWSD should be observed in a developed city cluster, so the Beijing-Tianjin-Tangshan region (BTTR), the Yangtze River Delta region (YRDR) and the Pearl River Delta region (PRDR) (shown in Figure 1(a)) were selected for further analysis. Beijing, the capital of China, is included in the BTTR; Shanghai, the largest city in China, is included in the YRDR; and Guangzhou, the largest city in Southern China near to Hong Kong and Macao, is included in the PRDR. The three regions are developed economic areas in China and located in different plains, where the urban area and population are much larger than in other cities in China (Liu et al., 2013). Studies have shown that farmland, woodlands and wetlands decreased, while urban, industrial and residential land notably increased in the BTTR, YRDR and PRDR (Liu and Tian 2010; Liu et al., 2003, 2014); simultaneously, the total population in each of the three regions reached 71.85, 77.99 and 28.68 million, respectively. The non-agricultural industries output value ratios were 93.91%, 90.97% and 97.56%, the urbanization rates were 52.12%, 48.11% and 53.56% and the economic densities were 13.74, 44.84 and 46.39 million RMB km⁻² in the BTTR, YRDR and PRDR, respectively. Therefore, economic and social development, along with LUCC in these three regions should be higher than in other parts of China (Liu et al., 2013). There are 21, 21 and 19 meteorological stations in the BTTR, YRDR and PRDR, respectively, so it is easy to assess the impact of anthropogenic LUCC on the SWS.

A decreasing trend of the SWSD was found in the BTTR and YRDR, but the SWSD had a slight increase in the PRDR (Figure 6). The minimum of SWSD of –1.26 m s^–1 occurred in the PRDR and, simultaneously, SWSDs of –0.95 and –0.73 m s^–1 were evident in the BTTR and the YRDR, respectively. Zhang et al. (2010) used the weather research and forecasting (WRF) model to investigate the influence of land cover change on the regional climate in the YRDR; the results pointed out that the slowdown in the SWS may reach 1.5 m s^–1 in the high-density urban area, and the urban land cover can cause 50% wind speed loss over the urbanized area. Wang et al. (2014) used the same model to study the urbanization impacts on the SWS in the PRDR, and found that the SWS over the urban area decrease by roughly 1.2–1.5 m s^–1. Hence, we can find that the estimation results in the YRDR and PRDR are close to the results of Zhang et al. (2010) and Wang et al. (2014), and the decreases in SWS induced by LUCC in the BTTR, YRDR and PRDR should be reliable based on the OMR method compared to the previous numerical model studies in these regions.

Figure 6.

Temporal variability of the SWS difference between the ERA-Interim dataset and the observations (represented by SWSD) in the Beijing-Tianjin-Tangshan region (BTTR) (black line), Yangtze River Delta region (YRDR) (blue line) and Pearl River Delta region (PRDR) (gray line) (red, pink and green lines denote the Ensemble Empirical Mode Decomposition (EEMD) non-linear trend of the BTTR, YRDR and PRDR, respectively).

The SWSD in the BTTR, YRDR and PRDR was lower than the mean of all 492 stations, with values of –0.24, –0.02 and –0.55 m s^–1, respectively. Therefore, the impacts of LUCC on the SWS were more significant in the three regions than in all of China. The decrease of the SWS in these three regions was induced by the total effects of LUCC, but the enhancement of vertical mixing caused by the taller canopy and the increase of anthropogenic heat release in the urban regions can increase the wind speed at the low level (Oceal and Belcher, 2005; Zhang et al., 2010, 2015). Therefore, the decreasing trends of the SWS in the three city clusters are relatively small. Guo et al. (2011) also showed that the SWS is smaller in urban stations than that in rural stations, but the decreasing trend of the SWS is also lower in urban stations than in the rural stations. In addition, we think that the differences of SWSD in the three regions could be due to the difference in LUCC, because the intensification of LUCC was different in each of the three regions, which could have caused different degrees of change in the land surface roughness and thermodynamics. The influences of urbanization on the SWS are complex, so it is hard to use the statistics diagnosis methods to analysis the physical mechanism of urbanization and quantify its impacts on the SWS. These specific influences will be further examined and identified using the detailed numerical model and such work is beyond the scope of this study.

IV Discussion

1 Estimation of errors in the results

The influences of LUCC on the SWS have been quantified by the OMR method, but the reliability of the wind speeds in the ERA-Interim reanalysis dataset needs to be further investigated. In this section, we will discuss how reliably the ERA-Interim reanalysis dataset captures large-scale signals and try to estimate the error in the results.

The terrain height in eastern China (eastern part of 106°E) is lower than that in western China (western part of 106°E), so we think that the wind speeds of 850 and 500 hPa were representative of the large-scale wind field in eastern and western China, respectively. Figure 7 shows that the spatial distribution of the ERA-Interim dataset is consistent with that in the sounding wind speed for 850 hPa. Both the sounding and the ERA-Interim dataset wind speeds for 850 hPa were greater in the northeast part of China than in other regions. At the same time, the regions that had smaller sounding and ERA-Interim dataset wind speeds were located in the middle reaches of the Yellow and Yangtze Rivers (Figures 7(a) and (b)). The spatial distribution of the sounding wind speed was also in accord with those in the ERA-Interim dataset for 500 hPa, which had a lower wind speed of 10.0 m s^–1 for the Tibetan Plateau region (Figures 7(e) and (f)). In addition, the sounding wind speed was higher than the ERA-Interim dataset wind speed for 850 and 500 hPa in most regions of China, but the wind speed differences were not distinct for 850 and 500 hPa and were not significant at the 99% confidence level for most regions of China (Figures 7(c) and (g)). The relative error of the ERA-Interim dataset wind speed for 850 hPa is lower than 10% for most regions of China, except for the middle and lower reaches of the Yellow River and southern China coastal area, and the lowest relative error was found in northeast China, which was lower than 6% in this region. The relative error of the ERA-Interim dataset wind speed for 500 hPa is less than 5%, except for the southern regions of the Tibetan Plateau. The mean relative error for 850 and 500 hPa is only 8.2% and 5.7%, respectively. Figure 8 shows that the inter-annual and seasonal changes of the ERA-Interim dataset wind speeds are consistent with the sounding wind speed for 850 and 500 hPa, respectively. The mean wind speed difference between the ERA-Interim dataset and sounding observation on 850 and 500 hPa is 0.43 and 0.55 m s^–1, respectively, and the correlation coefficients for 850 and 500 hPa were higher than 0.9, with PEST values greater than 90%. These results demonstrate that the ERA-Interim dataset reanalysis data could capture the primary signals of the large-scale circulation fields very well. Previous studies have also shown that the applicability of the wind speed data in the ERA-Interim reanalysis dataset was appropriate for climate change studies for most regions of China. Bao and Zhang (2013) reported that the vertical profiles of the mean and standard deviation for U and V in the ERA-Interim dataset products closely followed those averaged over the verifying sounding observations, indicating that the ERA-Interim dataset could adequately capture the mean and variation of the horizontal wind fields. On the other hand, the relatively small overall bias of both U and V suggests that the horizontal winds are very reliable for plateau-scale averages over seasonal or longer time scales and that the ERA-Interim dataset has the smallest sub-regional variability of the mean biases. Chen et al. (2014) demonstrated that the ERA-Interim dataset wind speed is reliable for describing the regional mean climate on a seasonal scale and representing the temporal and spatial variation of the wind speed throughout the troposphere.

Figure 7.

The spatial distribution of sounding and ERA-Interim wind speed (a), (b), (e), (f), as well as their wind speed differences (c), (g) and relative errors (d), (h) on 850 and 500 hPa ((a)–(d) for 850 hPa; (e)–(h) for 500 hPa; black dots in (c) and (g) denote that the wind speed difference can pass the significance t-test at the 99% confidence level).

Figure 8.

The temporal change of sounding and ERA-Interim wind speed on 850 hPa (a) and 500 hPa (b) (the black and red lines in (a) and (b) denote the sounding and ERA-Interim wind speed, respectively).

The ERA-Interim reanalysis data adequately capture the inter-annual, seasonal and inter-decadal characteristics of the large-scale wind field, but we must admit that the reanalysis data are also somewhat biased, despite the recent improvement of the assimilation systems and the numerical models. Hence, quantifying the effects of LUCC on the SWS with the traditional OMR method could include errors from the reanalysis data (Wu et al., 2016b). The terrain height in eastern China is lower than that in western China, so we think that the relative error of the ERA-Interim dataset for 850 hPa is a systematic error for eastern China, and the relative error for 500 hPa is the systematic error for western China. To deduct the error in the SWSD and obtain a more accurate SWSD value (represented by the AWSD hereafter), the AWSD was calculated using equation (1)

AWSD = V_{obs} - \frac{V_{ERA - Interim}}{1 + {(RE)}_{i}}

where $V_{obs}$ denotes the observed SWS, $V_{ERA - Interim}$ denotes the ERA-Interim dataset 10 m wind speed, ${(RE)}_{i}$ denotes the relative error between the ERA-Interim dataset and sounding observation wind speeds and i denotes 850 and 500 hPa, respectively. There are fewer sounding stations than surface weather stations, so some surface stations have no sounding observations. The surface stations located in a circle with a radius of 2° in latitude and longitude centered in the center of the sounding station were defined as the adjacent stations to the sounding stations. If the adjacent stations had no the sounding data, the error of the central station was regarded as the systematic error. In addition, if the adjacent station had more than one central station, the maximum relative error of all of the central stations was considered as a systematic error.

Figure 9(a) shows that the spatial distribution of the AWSD is consistent with that of the SWSD (Figure 4(a)). The AWSD is negative for most regions of China, with an average value of –0.47 m s^–1; however, the AWSD has shown a pronounced decrease in the last 30 years, in accordance with the SWSD results in Figure 4(b). The most distinct decrease of AWSD was observed in Northeast China, and the middle and low regions of the Yangtze River, in which the decrease of AWSD reached –0.2 m s^–1 decade^–1, which was significant at the 99% confidence level. In addition, the inter-annual and seasonal changes of the AWSD were also consistent with those of the SWSD. The PEST between the SWSD and the AWSD could be as high as 99%, with a correlation coefficient of 0.99 (Figure 10(b)). The SWSD was lower than the AWSD by –0.24 m s^–1, but the decreasing trend of the SWSD is just lower than that of the AWSD by 0.01 m s^–1 decade^–1. The AWSD was –0.57 and –0.30 m s^–1 for large cities and small cities, respectively, so the AWSD for large cities was lower than the entire region and small cities with values of –0.10 and –0.27 m s^–1, respectively. In addition, the AWSD was –0.50, –0.50 and –0.90 m s^–1 for the BTTR, YRDR and PRDR, respectively, which was higher than the SWSD in the same regions with values of 0.45, 0.23 and 0.36 m s^–1, respectively.

Figure 9.

The spatial distribution of the more accurate near-surface wind speed difference between the ERA-Interim dataset and the observations (represented by AWSD) (a) (unit: m s^–1) and its linear trend coefficient (b) (unit: m s^–1 decade^–1) (light blue, yellow, red in (b) denotes that the linear trend can pass the significant t-test at the 90%, 95%, 99% confidence levels, respectively).

Figure 10.

Temporal variability of the more accurate near-surface wind speed difference between the ERA-Interim dataset and the observations (represented by AWSD) (a) and the relationship between the AWSD and SWSD (b) (R is the correlation coefficient, Rc is threshold of the correlation coefficient, P is the significance level and the red rectangle in (a) represents annual mean AWSD). EEMD: Ensemble Empirical Mode Decomposition.

According to these results, the bias in the ERA-Interim dataset could cause a 51% error in the estimation for the entire region during the study period, but had an indistinctive effect on the decreasing trend in the SWS. The decrease rate of the SWS induced by LUCC based traditional OMR method had an error of 0.01 m s^–1 decade^–1, which is significantly lower than the linear trend of the SWSD itself. The bias in the ERA-Interim reanalysis data will not cause significant changes in the estimation results, so the long-term decrease rate of the SWS induced by LUCC based on the OMR method should be reliable. Wu et al. (2016b) used the statistical downscaling method (SDM) to investigate the impacts of LUCC on the SWS in the ECP region, and found that the regional mean linear trends of the SWSD are –0.19 and –0.16 m s^–1 decade^–1 for the SDM and OMR methods, respectively. The variances of the SWSD based on the two methods have the similar spatial patterns, with regional average values of 0.076 and 0.067 m^–2 s^–2 for the SDM and OMR methods, respectively. The applicability and reliability of the OMR method have also been verified in previous studies (Hua et al., 2014; Li et al., 2008). In addition, the AWSD could be considered to be a more accurate estimation of LUCC impacts on the SWS.

2 Comparison of the results of different methods to quantify the effects of LUCC on the SWS

Some studies have investigated the effects of LUCC on the SWS by comparing the wind speeds observed in rural areas with those in urban stations (represented by the CRU hereafter) method, and their results have revealed the effects of LUCC to the SWS in the last 40 years (Guo et al., 2011; Jiang et al., 2010). The CRU method was also used in this study as a comparison with the OMR method. The SWSD in Figure 11(a) represents the surface wind speed difference between wind speeds in a large city minus those in a small city. Figure 11(a) shows that the SWS in large city is lower than that in small city, with a mean SWSD of –0.47 m s^–1, but the SWSD shows an increase of 0.04 m s^–1 decade^–1 with a standard deviation of 0.05 m s^–1. However, the CRU method may produce uncertainties in the results. Firstly, long-term meteorological observational data include very few strictly rural stations in China, so some small cities and towns were designated as rural stations in some previous studies. Otherwise, the criteria for rural stations may be different in different studies, which can introduce error in the results. Secondly, the effects of LUCC on the SWS in a large city can only be achieved if LUCC in a small city is negligible in the CRU method; in fact, LUCC and urbanization are becoming more distinct than previously for small cities. Therefore, the changes in the SWS due to LUCC in a large city assessed by the CRU method may include effects from the neglected LUCC in a small city. Furthermore, the CRU method cannot be used to assess the impact of LUCC on the SWS in a small city because the reference is missing in this case. Thirdly, economic and social development is unbalanced in different areas of China, so the large cities are mainly located in the middle and eastern parts of China, and the small cities are in Western and Northern China. This lack of homogeneity in the spatial distribution of stations of different scales may introduce complicated uncertainties in the assessment results.

Figure 11.

The temporal change of the annual SWSD based on the CRU method (a), and annual AWSD in large cities and small cities based on the observation minus reanalysis (OMR) method (b) (R is the correlation coefficient, Rc is the threshold and P represents the significance level). SWSD: near-surface wind speed difference between wind speeds in a large city minus those in a small city. CRU: compare the wind speeds observed in rural areas with those in urban stations. AWSD: the more accurate near-surface wind speed difference between the ERA-Interim dataset and the observations.

That the uncertainty of LUCC affects the SWS based on the CRU method is evident (Hua et al., 2014), so a revised CRU method was proposed by the comparison of adjacent large and small cities. Small cities, located within a circle with a radius of 1° in latitude and longitude centered in the center of a large city, were selected and compared with the large city. According to that criterion, 63 large cities and 76 small cities could be selected and designated as adjacent large and small cities, homogenizing the spatial distribution of stations of different scales. The results showed that the mean SWSD between adjacent small and large cities was –0.55 m s^–1, and an increasing rate of 0.08 m s^–1 decade^–1 with the standard deviation of 0.09 m s^–1 was noted (Figure 11(a)). Therefore, compared to the original CRU method, the assessment results for the effects of LUCC on the SWS in large cities based on the revised CRU method are more distinct than that of the original CRU method, but this method can only be used to determine the impact of LUCC on the SWS for 63 large cities. Both the original and revised CRU methods indicated that the effects of LUCC on the SWS are more significant in large cities than that in small cities, but the SWSD showed a significant increasing trend, which means that the effects of LUCC on the SWS in small cities increase faster than that in large cities, and LUCC in small cities might be more distinct in recent years. These results suggest a new aspect of urban effects on wind speed changes (Guo et al., 2011).

The assessment results for the OMR method are shown in Figure 11(b). SWS decreases of 0.57 and 0.30 m s^–1 were induced by LUCC in a large and small city in the last 30 years, respectively. A slowdown in the SWS induced by LUCC in large cities, as estimated by the OMR method, is higher than the results from the original and revised CRU methods by 0.10 and 0.02 m s^–1, respectively. Figure 11(b) also shows that the AWSD in large and small cities has distinctly decreased by –0.1 and –0.13 m s^–1 decade^–1, respectively; however, the AWSD for large and small cities is becoming smaller, which means that the recent effects of LUCC on the SWS are greater in a small city than that in a large city. Similar results can also be found for both the original and revised CRU methods. Therefore, the above analysis indicates that the OMR estimate for a large city is close to that of the revised CRU method. However, unlike the CRU method, the OMR method can estimate the effects of LUCC on the SWS in a small city because the OMR method can exclude the error brought by the lack of homogeneity of stations of different scales. In addition, the OMR method can also be used easily for complex terrain, and is more convenient than the FWM method (Wu et al., 2016a). However, the reanalysis data is assimilated using different models and data sources, and also ignores some regional- or local-scale climate information (Farajzadeh et al., 2015), which could also cause uncertainty in the estimates.

V. Conclusion

The ERA-Interim reanalysis data and the station observation data in China were compared, and their differences, regarded as an effective representation of the impacts of LUCC on the SWS, were analyzed for 492 stations and the mean for large and small cities and three designated city clusters were determined from 1979 to 2010. The following conclusions can be reached.

(1) The observed SWS declined by –0.11 m s^–1 decade^–1 in China in the last 30 years, and this result was significant at the 99% confidence level. The ERA-Interim dataset 10 m wind speed decreased by –0.006 m s^–1 decade^–1 and was not significant at the 99% confidence level. In addition, the ERA-Interim reanalysis data captured the large-scale characteristics adequately, and the spatial pattern and the inter-annual changes of the ERA-Interim dataset wind speeds are in agreement with the observed wind speeds. However, a larger difference in the SWS between large and small cities was found from the observational data, and this difference in the ERA-Interim dataset was very small, so the SWSD between the ERA-Interim dataset and observation could be regarded as a representative of the impact of LUCC on the SWS.

(2) The effects of LUCC on the SWS have been distinct in China in the last 30 years. LUCC could have caused a decline in the SWS of 0.12 m s^–1 per decade during the last 30 years. Urbanization is an important type of anthropogenic LUCC, and has a significant impact on the SWS. LUCC could have caused a decline of the SWS of 0.57 and 0.30 m s^–1 in large and small city stations, respectively. LUCC may also contribute to the observed decreases of 0.50, 0.50 and 0.90 m s^–1 for the BTTR, YRDR and PRDR, respectively, during the study period. In addition, the relationship between the urbanization rate and the decrease of the SWS is significant, and the decrease of 0.11 m s^–1 in the SWS could be induced by a 10% increase in the urbanization rate.

(3) A decline in the SWS due to LUCC for a large city as estimated by the OMR method is higher than the results obtained with the original and revised CRU methods by 0.10 and 0.02 m s^–1, respectively. However the estimates via the OMR method for a large city are close to those for the revised CRU method. In addition, a bias in the ERA-Interim dataset could cause a 51% error in the estimates for the entire region during the study period, but this would have a non-significant effect on the decreasing trend in the SWS. Therefore, the OMR method could indicate the correct decreasing trend of the SWS due to LUCC, and could be regarded as a credible method to evaluate the effects of LUCC on the SWS.

Some uncertainties are associated with our research. The weather stations employed were distributed in different-sized cities and the actual rural stations were lacking, so we did not use actual rural station data in our analyses. If corresponding rural stations become available and are matched with urban stations, a more accurate comparison of urban and rural observational data could be used to assess the impact of LUCC on the SWS. LUCC in China includes urbanization, forest cutting, irrigation of farmlands, desertification, returning farmland to forest as well as other activities, but we cannot distinguish the contributions of the individual factors to the observed SWS decrease, and we hope to discuss this issue with reliable numerical models in the near future. In addition, while general circulation models (GCMs) are sufficiently fine to predict the principal features of global climate change, GCMs are too coarse to provide the regional-scale information required for regional impact assessments (Giorgi and Mearns, 1999). Therefore, downscaling methods to extract regional-scale information from GCM output and reanalysis data should be considered in future studies (Curry et al., 2011; Huang et al., 2015; Kirchmeier et al., 2014). We also argue that the other methods, such as the numerical simulation and dynamical downscaling method, can be used to quantify the impacts of LUCC on the SWS and explain the reliability of the OMR method further. These will need to be further investigated in the near future.

Footnotes

Acknowledgements

The authors cordially thank the anonymous reviewers for their thorough comments and constructive suggestions, which improved the paper quality distinctly. At the same time, we also appreciate Divya’s help so much. Daily meteorological data is available from the China Meteorological Data Sharing Service System, and the ERA-Interim dataset come from the ECMWF. We thank all the data providers.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Chinese Natural Science Foundation (grant number 41675149, 41275162, 41305103) and the Chinese Academy of Sciences Strategic Priority Program (grant number XDA05090206). This work is also supported by Chinese Jiangsu Collaborative Innovation Center for Climate Change.

References

Accadia

Mariani

Casaioli

. (2003) Sensitivity of precipitation forecast skill scores to bilinear interpolation and a simple nearest-neighbor average on high-resolution verification grids. Weather Forecasting 18: 918–932. doi: https://dx-doi-org.web.bisu.edu.cn/10.1175/1520-0434(2003)018<0918:SOPFSS>2.0.CO;2.

Asselen

Verburg

(2013) Land cover change or land-use intensification: Simulating land system change with a global-scale land change model. Global Change Biology 19(12): 3648–3667.

Azorin-Molina

Guijarro

McVicar

. (2016) Trends of daily peak wind gusts in Spain and Portugal, 1961-2014. Journal of Geophysical Research-Atmospheres. Prepublished January 3, 2016. doi:10.1002/2015JD024485.

Azorin-Molina

Vicente-Serrano

McVicar

. (2014) Homogenization and assessment of observed near-surface wind speed trends over Spain and Portugal, 1961–2011. Journal of Climate 27(10): 3692–3712.

Bao

Zhang

(2013) Evaluation of NCEP-CFSR, NCEP-NCAR, ERA-Interim, and ERA-40 reanalysis datasets against independent sounding observations over the Tibetan Plateau. Journal of Climate 26(1): 206–214.

Berrisford

Kallberg

Kobayashi

. (2011) Atmospheric conservation properties in ERA-Interim. Quarterly Journal of the Royal Meteorological Society 137(659): 1381–1399.

Berrisford

Tobin

Dunn

RJH

. (2015) [Global climate; Atmospheric circulation; Surface winds] Land surface wind speed [in “State of the Climate in 2014”]. Bulletin of American Meteorological Society 95(7): S33–S34.

Bichet

Wild

Folini

. (2012) Causes for decadal variations of wind speed over land: Sensitivity studies with a global climate model. Geophysical Research Letters 39(L11701): 1–6. doi:10.1029/2012GL051685.

Brazdil

Chroma

Dobrovolny

. (2009) Climate fluctuations in the Czech Republic during the period 1961-2005. International Journal of Climatology 29(2): 223–242.

10.

Breaker

Ruzmaikin

(2011) The 154-year record of sea level at San Francisco: Extracting the long-term trend, recent changes, and other tidbits. Climate Dynamics 36(3–4): 545–559.

11.

Chen

Iwasaki

Qin

. (2014) Evaluation of warm-season diurnal variability over East Asia in recent reanalyses JRA-55, ERA-Interim, NCEP CFSR, and NASA MERRA. Journal of Climate 27(14): 5517–5537.

12.

China Meteorological Administration (CMA) (2003) Ground Surface Meteorological Observation. Beijing: China Meteorological Press, 157.

13.

Curry

Kamp

Monahan

(2011) Statistical downscaling of historical monthly mean winds over a coastal region of complex terrain. I. Predicting wind speed. Climate Dynamics 38(7): 1281–1299.

14.

Dadaser-Celik

Cengiz

(2014) Wind speed trends over Turkey from 1975-2006. International Journal of Climatology 34(6): 1913–1927.

15.

Dee

Uppala

Simmons

. (2011) The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quarterly Journal of the Royal Meteorological Society 137(656): 553–597.

16.

Earl

Dorling

Hewston

. (2013) 1980–2010 variability in U.K. surface wind climate. Journal of Climate 26(4): 1172–1191.

17.

Farajzadeh

Oji

Cannon

. (2015) An evaluation of single-site statistical downscaling techniques in terms of indices of climate extremes for the Midwest of Iran. Theoretical and Applied Climatology 120(1–2): 377–390.

18.

Franzke

Woollings

(2011) On the persistence and predictability properties of North Atlantic climate variability. Journal of Climate 24(2): 466–472.

19.

Frauenfeld

Zhang

Serreze

(2005) Climate change and variability using European Centre for Medium-Range Weather Forecasts reanalysis (ERA-40) temperatures on the Tibetan Plateau. Journal of Geophysical Research Atmospheric 110(D02101): 1–9. doi:10.1029/2004JD005230.

20.

Giorgi

Mearns

(1999) Introduction to special section: Regional climate modeling revisited. Journal of Geophysical Research Atmospheric 104(D6): 6335–6352.

21.

Guo

(2011) Changes in near-surface wind speed in China; 1969-2005. International Journal of Climatology 31(3): 349–358.

22.

Monahan

Jones

. (2010) Probability distributions of land surface wind speeds over North America. Journal of Geophysical Research Atmospheric 115(D4): 288–303.

23.

Zhang

. (2011) Temporal variation of wind speed in China for 1961–2007. Theoretical and Applied Climatology 104(3–4): 313–324.

24.

Hua

Chen

(2014) Review of land use and lover change in China and associated climate effects. Advances in Earth Science 29(9): 1025–1036 (in Chinese).

25.

Huang

Capps

Huang

., (2015) Downscaling near-surface wind over complex terrain using a physically-based statistical modeling approach. Climate Dynamics 44(1–2): 529–542.

26.

Huang

(2008) A review on Hilbert-Huang transform: Method and its applications to geophysical studies. Reviews of Geophysics 46(RG2006): 1–23 doi:10.1029/2007RG000228.

27.

Jacobson

Kaufman

(2006) Wind reduction by aerosol particles. Geophysical Research Letters 33(L24814): 1–6. doi:10.1029/2006GL027838.

28.

Jiang

Luo

Zhao

. (2010) Changes in wind speed over China during 1957–2004. Theoretical and Applied Climatology 99(3–4): 421–430.

29.

Kalnay

Cai

(2003) Impact of urbanization and large-use change on climate. Nature 423(6939): 528–531.

30.

Kirchmeier

Lorenz

Vimont

(2014) Statistical downscaling of daily wind speed variations. Journal of Applied Meteorology and Climatology 53(3): 660–675.

31.

Klink

(1999) Trends in mean monthly maximum and minimum surface wind speeds in the coterminous United States, 1961 to 1990. Climate Research 13(3): 193–205.

32.

Klink

(2007) Atmospheric circulation effects on wind speed variability at Turbine Height. Journal of Applied Meteorology and Climatology 46(4): 445–456.

33.

Wang

Chu

. (2008) The impact of anthropogenic land-use changes on the near-surface wind energy potential in China. Chinese Science Bulletin 53(21): 2646–2653.

34.

Yan

. (2011) Chang in wind speed and extremes in Beijing during 1960-2008 based on homogenized observations. Advances in Atmospheric Sciences 28(2): 408–420.

35.

Lim

Cai

Kalnay

. (2005) Observational evidence of sensitivity of surface climate changes to land types and urbanization. Geophysical Research Letters 32(22): 117–137.

36.

Liu

Shen

Liu

(2013) The financial services industry development level and spatial pattern of urban agglomeration in China. Acta Geographica Sinica 68(2): 186–198 (in Chinese).

37.

Liu

Kuang

Zhang

. (2014) Spatiotemporal characteristics, patterns, and causes of land-use changes in China since the late 1980s. Journal of Geographical Sciences 24(2): 195–210.

38.

Liu

Zhuang

. (2003) Study on spatial pattern of land-use change in China during 1995-2000. Science in China 46(4): 373–384.

39.

Liu

Tian

(2010) China’s land cover and land use change from 1700 to 2005: Estimations from high-resolution satellite data and historical archives. Global Biogeochemical Cycle 24(3): 285–286.

40.

Liu

(2000) The homogenity test on mean annual wind speed over China. Journal of Applied Meteorology Science 11(1): 28–34 (in Chinese).

41.

McVicar

Niel

TGV

Roderick

. (2010) Observational evidence from two mountainous regions that near-surface wind speeds are declining more rapidly at higher elevations than lower elevations: 1960–2006. Geophysical Research Letters 37(L06402): 1–6. doi:10.1029/2009GL042255.

42.

McVicar

Roderick

(2010) Atmospheric science: Winds of change. Nature Geoscience 3(11): 747–748.

43.

McVicar

Roderick

Donohue

. (2012) Global review and synthesis of trends in observed terrestrial near-surface wind speeds: Implications for evaporation. Journal of Hydrology 416–417(3): 182–205.

44.

McVicar

Thomas

Van Niel

. (2008) Wind speed climatology and trends for Australia, 1975–2006: Capturing the stilling phenomenon and comparison with near-surface reanalysis output. Geophysical Research Letters 35(L20403): 1–6. doi:10.1029/2008GL035627.

45.

Mastylo

(2013) Bilinear interpolation theorems and applications. Journal of Functional Analysis 265(2): 185–207.

46.

Miller

Gans

Kleidon

(2011) Estimating maximum global land surface wind power extractability and associated climatic consequences. Earth System Dynamics Discussions 2(1): 1–12.

47.

Najac

Boé

Terray

(2009) A multi-model ensemble approach for assessment of climate change impact on surface winds in France. Climate Dynamics 32(5): 615–634.

48.

Najac

Lac

Terray

(2012) Impact of climate change on surface winds in France using a statistical-dynamical downscaling method with meso-scale modelling. International Journal of Climatology 31(3): 415–430.

49.

National Research Council (NRC) (2005) Radiative Forcing of Climate Change: Expanding the Concept and Addressing Uncertainties. Washington, DC: National Academies Press.

50.

Oceal

Belcher

(2005) Mean winds through an inhomogeneous urban canopy. Boundary-Layer Meteorology 115(1): 47–68.

51.

Okin

Gillette

Herrick

(2006) Multi-scale controls on and consequences of Aeolian processes in landscape change in arid and semi-arid environments. Journal of Arid Environments 65(2): 253–275.

52.

Peterson

Vautard

McVicar

. (2011) Surface winds over land. Bulletin of American Meteorological Society 92(6): S57.

53.

Poli

(2010) List of observations assimilated in ERA-40 and ERA-Interim. ERA Report Series 4, ECMWF, Reading, UK. Available at: http:/www.ecmwf.int/publications/.

54.

Pryor

Barthelmie

(2010) Climate change impacts on wind energy: A review. Renewable and Sustainable Energy Reviews 14(1): 430–437.

55.

Pryor

Barthelmie

Young

. (2009) Wind speed trends over the contiguous United States. Journal of Geophysical Research 114(D14105): 1–18. doi:10.1029/2008JD011416.

56.

Pryor

Schoof

Barthelmie

(2006) Winds of change? Projections of near-surface winds under climate change scenarios. Geophysical Research Letters 33(L11702): 1–5. doi:10.1029/2006GL026000.

57.

Qian

(2011) Changes in the amplitude of the temperature annual cycle in China and their implication for climate change research. Journal of Climate 24(20): 5292–5302.

58.

Rabier

Järvinen

Klinker

. (2000) The ECMWF operational implementation of four-dimensional variational assimilation. I: Experimental results with simplified physic. Quarterly Journal of the Royal Meteorological Society 126(564): 1148–1170.

59.

Rayner

(2007) Wind run changes: The dominant factor affecting pan evaporation trends in Australia. Journal of Climate 20(14): 3379–3394.

60.

Roderick

Rotstayn

Farquhar

. (2007) On the attribution of changing pan evaporation. Geophysical Research Letters 34(L17403): 1–6. doi:10.1029/2007/GL031166.

61.

Schwiesow

Lawrence

(1982) Effect of a change of terrain height and roughness on a wind profile. Boundary-Layer Meteorology 22(1): 109–122.

62.

Simmons

Poli

Dee

. (2014) Estimating low-frequency variability and trends in atmospheric temperature using ERA-Interim. Quarterly Journal of the Royal Meteorological Society 140(679): 329–353.

63.

Simmons

Uppala

Dee

. (2007) ERA-Interim: New ECWMF reanalysis products from 1989 onwards. ECMWF Newsletter 110(110): 25–35.

64.

Simmons

Willett

Jones

. (2010) Low-frequency variations in surface atmospheric humility, temperature and precipitation: Inferences from reanalysis and monthly gridded observational datasets. Journal of Geophysical Research Atmospheric 115(D01110): 1–21. doi:10.1029/2009JD012442.

65.

Sušelj

Sood

Heinemann

(2010) North Sea near-surface wind climate and its relation to the large-scale circulation patterns. Theoretical and Applied Climatology 99(3–4): 403–419.

66.

Thomas

Swail

(2011) Buoy wind inhomogeneities related to averaging method and anemometer type: application to long time series. International Journal of Climatology 31(7): 1040–1055.

67.

Tobin

Berrisford

Dunn

RJH

. (2014) [Global climate; Atmospheric circulation; Surface winds] Land surface wind speed [in “State of the Climate in 2013”]. Bulletin of American Meteorological Society 95(7): S28–S29.

68.

Tuller

(2004) Measured wind speed trends on the west coast of Canada. International Journal of Climatology 24(11): 1359–1374.

69.

Uppala

Kallberg

Simmons

. (2005) The ERA-40 re-analysis. Quarterly Journal of the Royal Meteorological Society 131(612): 2961–3012.

70.

Vautard

Cattiaux

Yiou

. (2010) Northern Hemisphere atmospheric stilling partly attributed to an increase in surface roughness. Nature Geoscience. Prepublished October 17, 2010. doi:10.1038/NGEO979.

71.

Vautard

McVicar

Thepaut

. (2012) [Global climate; Atmospheric circulation; surface winds] Land surface winds and atmospheric evaporative demand [in “State of the Climate in 2011”]. Bulletin of American Meteorological Society 93(7): S6–S38.

72.

Wang

Liao

Zhang

. (2014) A numerical of regional climate change induced by urban expansion in the Pearl River Delta, China. Journal of Applied Meteorology and Climatology 53(2): 346–362.

73.

Wever

(2012) Quantifying trends in surface roughness and the effect on surface wind speed observations. Journal of Geophysical Research Atmospheric 117(D11104): 1–4 doi:10.1029/2011JD017118.

74.

Zhang

. (2012) Trend of visibility on sunny days in China in the resent 50 years. Atmospheric Environment 55(3): 339–346.

75.

Zha

Zhao

(2016a) Estimating the impact of the changes in land use and cover on the surface wind speed over the East China Plain during the period 1980-2011. Climate Dynamics 46(3–4): 847–863.

76.

Zha

Zhao

(2016b) Evaluating the effects of land use and cover change on the decrease of surface wind speed over China in recent 30 years using a statistical downscaling method. Climate Dynamics. Prepublished March 18, 2016. doi:10.1007/s00382-016-3065-z.

77.

Huang

(2009) Ensemble empirical mode decomposition: A noise-assisted data analysis method. Advances in Adaptive Data Analysis 1(1): 1–41.

78.

Chang

CFY

. (2006) Steady decline of East Asian monsoon winds, 1969-2000: Evidence from direct ground measurements of wind speed. Journal of Geophysical Research Atmospheric 111(D24111): 1–8. doi:10.1029/2006JD007337.

79.

You

Kang

Flugel

. (2010) Decreasing wind speed and weakening latitudinal surface pressure gradients in the Tibetan Plateau. Climate Research 42(1): 57–64.

80.

Zha

Zhao

(2016) Changes of probabilities in different wind grades induced by land use and cover change in Eastern China Plain during 1980–2011. Atmospheric Science Letters 17(4): 264–269.

81.

Zhang

Gao

Wang

. (2010) Modeling the impact of urbanization on the local and regional climate in Yangtze River Delta, China. Theoretical and Applied Climatology 102(3–4): 331–342.

82.

Zhang

Wang

Chen

. (2015) Numerical simulations on influence of urban land cover expansion and anthropogenic heat release on urban meteorological environment in Pearl River Delta. Theoretical and Applied Climatology. Prepublished August 09, 2015. doi:10.1007/s00704-015-1601-0.

83.

Zhou

Dickinson

Tian

. (2004) Evidence for a significant urbanization effect on climate in China. Proceedings of the National Academy of Science 101(26): 9540–9544.