Under the background of global warming, prominent decreases have been observed in the near-surface (10-m) wind speed (NWS) in China over the past five decades (Wang et al., 2004). The weak surface winds observed in recent years are considered to be one of the main factors causing the increase in the number of hazy days in China (Niu et al., 2010; Guo et al., 2015; Pei et al., 2018; Zhong et al., 2018), because low NWSs favor the accumulation of air pollutants (Wu P. et al., 2017). The number of hazy days during autumn has increased in the urban regions of North China from 19.6 days in the 1960s to 35.6 days in the 2000s (Chen and Wang, 2015). China has a large population and there is much concern about environmental protection. The dominant atmospheric circulation in East Asia is different in the four seasons and changes in NWSs in China might be affected by different factors in different seasons. Previous studies have found that the decrease in the winter NWS in China is related to stronger warming in northern China, whereas the decline of the summer monsoon is linked to summer cooling in central China and warming over the South China Sea (Xu et al., 2006).

Climate models are important tools, which can be used to study the mechanisms of past climate and project future climate changes. However, few models are able to reproduce the summertime cooling over central South China and only two models from Phase 3 of the Coupled Model Intercomparison Project (CMIP3) acceptably reproduced the meridional gradients of the wintertime warming trends (Zhou and Yu, 2006). How is the capability of the models to reproduce the changes in the NWS over the last half century? Jiang et al. (2009) found that most of the CMIP3 models are able to simulate the spatial distribution of the NWS in China, but unable to simulate the decreasing trend of the NWS in recent decades. The CMIP5 models are superior to the CMIP3 models in hindcasting the magnitude and spatial pattern of the seasonal mean NWS in China (Jiang et al., 2017), but the decreasing trends in the NWS are still not well represented in the CMIP5 models. The ensemble mean of the CMIP5 models captures the spatial variability of the annual maximum NWS, but not the historical trend of extreme wind speeds (Kumar et al., 2015). Therefore, improvements in simulating the changes in the NWS in China are still demanding in the current climate models.

Two climate models (BCC-CSM1.1 and BCC-CSM1.1m) developed by the Beijing Climate Center were used in CMIP5. Jiang et al. (2017) found that the BCC-CSM1.1 model was unable to reproduce the decreasing trend of the annual mean NWS. The BCC-CSM1.1m model has a higher horizontal resolution (110 km) in both the atmosphere and land models, but its ability to simulate the NWS has not been evaluated. The impact of horizontal resolution on the simulation of the NWS in the two models requires investigation because the only difference between the two models is the horizontal resolution in the land and atmospheric components. We also explored future projections of the NWS in China based on the models’ ability to reproduce historical changes in the NWS. This will provide a reference for the possible influence of greenhouse gas emissions on future changes in the NWS in China.

This paper is organized as follows. Section 2 describes the models, experiments, and data. Section 3 gives the results of the simulation of historical changes in the NWS in China by the two models. Future projections of the NWS in China are presented in Section 4 and the highlights of our results are summarized in Section 5.

The climate models used in this study were BCC-CSM1.1 and BCC-CSM1.1m, developed by the Beijing Climate Center, which include four (atmosphere, land, ocean, and ice) component models. The atmosphere (BCC_AGCM2.1, Wu et al., 2010) and land (BCC-AVIM1.0, Ji et al., 2008) models have different horizontal resolutions, with spectral discretization at T42 resolution (about 280 km) in BCC-CSM1.1 and T106 resolution (about 110 km) in BCC-CSM1.1m. The vertical layers (26 levels) and atmospheric physical processes are the same in both models. BCC-CSM1.1 and BCC-CSM1.1m share the same ocean model (MOM4-L40) and sea ice simulator. The MOM4 model has 40 vertical levels with a tripolar grid of 1° × 1/3° in the horizontal direction (Griffies et al., 2005). More detailed descriptions of the model components of BCC-CSM1.1 have been presented by Wu et al. (2013).

The historical simulations provided by BCC-CSM1.1 and BCC-CSM1.1m for CMIP5 were analyzed in this study. The historical experiments were carried out from 1850 to 2005 driven by the time-varying radiative forcing from greenhouse gases, ozone, aerosols, volcanic activity, and solar variability. The period 1961–2005 was adopted for comparison with the observations. Future projections by BCC-CSM1.1m investigated under the Representative Concentration Pathway (RCP) include RCP8.5, RCP4.5, and RCP2.6. The number following the abbreviation RCP represents the future radiative forcing stabilized at 8.5, 4.5, and 2.6 W m^–2 in 2100, respectively. RCP2.6 is the mitigation scenario required to keep the increase in the global mean temperature below 2°C relative to pre-industrial conditions (van Vuuren et al., 2011). The concentrations of greenhouse gases, ozone, aerosols, and the solar constant vary with time in the RCP experiments. The solar constant has a stable cycle of 11 yr. These experiments have been described in detail by Xin et al. (2013).

The monthly mean of the NWS from the simulation is the monthly mean wind speed calculated at a daily frequency. The climatology in this study is adopted from the mean during the time period 1961–2005 and the anomalies are the differences relative to the climatology.

The observed NWS data are from the gridded monthly dataset of CN05.1 (Wu and Gao, 2013), which have been widely used in model evaluations in China (Sun and Wang, 2015; Bucchignani et al., 2016; Wu J. et al., 2017). The spatial resolution of this dataset is 0.25° × 0.25° (longitude × latitude) based on 2416 station observations. The sea level pressure (SLP) and surface temperature data used in this study are derived from the NCEP–NCAR dataset with a resolution of 2.5° × 2.5° (Kalnay et al., 1996).

Figure 1 shows the climatology of the annual mean NWS over China for the observations and the BCC-CSM1.1 and BCC-CSM1.1m simulations. The observational data show a distinctive local minimum (< 1.5 m s^–1) in Sichuan and Tarim basins and a local maximum (> 6 m s^–1) over the Tibetan Plateau. These features are better captured by the BCC-CSM1.1m simulation than by the BCC-CSM1.1 simulation. This could be because the BCC-CSM1.1m simulation provides a better resolution of the detailed topography as a result of its higher horizontal resolution. Both models reproduce relatively stronger surface winds in northern China than in Southeast China, although both models overestimate the wind speed in these regions. In the following analysis, the anomalies relative to the climatology are calculated for both the simulations and observations to remove the model bias in the climatology.

Figure 1 Climatology of annual mean NWS (m s–1) over China in (a) the observation, (b) BCC-CSM1.1m simulation, and (c) BCC-CSM1.1 simulation.

Figure options

The correlation coefficients between the simulations and observational data during the time period 1961–2005 are calculated for the seasonal (March–April–May, June–July–August, September–October–November, December–January–February) mean NWS to show the model’s ability to simulate variations in the NWS. The observational data are interpolated onto the T42 and T106 grids for comparison with the BCC-CSM1.1 and BCC-CSM1.1m simulation, respectively. The autocorrelation effect is removed in the significance test following the method of Trenberth (1984). Figure 2 shows that there is almost no significant correlation between the BCC-CSM1.1 simulation and the observation in each season. The BCC-CSM1.1m simulation shows little improvement in spring, summer, and winter, but shows a prominent improvement for the autumn NWS in most areas of China, especially North China.

Figure 2 Correlation of the NWS between the observation and (a, c, e, g) BCC-CSM1.1 and (b, d, f, h) BCC-CSM1.1m simulations in (a, b) spring, (c, d) summer, (e, f) autumn, and (g, h) winter. The black dots show the correlations above the 90% confidence level.

Figure options

Haze frequently occurs over North China in autumn, accounting for 24.2% of the annual occurrence of haze and shows a clearly increasing trend in recent decades (Chen and Wang, 2015). This corresponds to the weakening trend of the autumn NWS in North China. We explore why the BCC-CSM1.1m simulation reproduces well the variations in the NWS in this region. Figure 3 shows the time series of the observed and simulated NWS anomalies averaged over North China (35°–42°N, 105°–120°E; namely, the black box in Fig. 2f). The observations show a clearly decreasing trend (–1.82 m s^–1 per century) over the last 45 years. The decreasing trend can be partly reproduced by the BCC-CSM1.1m simulation (–0.6 m s^–1 per century), which is about 1/3 of the observed data. The underestimation is similar to those reported for other CMIP5 models by Jiang et al. (2017). There is almost no linear trend for the NWS (0.08 m s^–1 per century) in North China in the BCC-CSM1.1 simulation. The correlation coefficient is 0.3 for the North China NWS between the observations and the BCC-CSM1.1m simulation, which is significant at the 90% confidence level. However, after removing the linear trend, the correlation coefficient is reduced to –0.02. This implies that the BCC-CSM1.1m simulation partly reproduces the linear trend, but not the interannual variation in the NWS in North China observed in recent decades.

Figure 3 Time series of autumn NWS anomalies (solid lines; m s–1) and the linear trends (dashed lines) in North China (35°–42°N, 105°–120°E; namely, the black box in Fig. 2f) during 1961–2005 from the observation and the BCC-CSM1.1 and BCC-CSM1.1m simulations. The correlation coefficient between the simulation and observation is shown in the bracket.

Figure options

The NWS in North China is related to the pressure gradient in the middle and high latitudes of the Asian continent. Figure 4a shows that the observed NWS in North China is positively correlated with the SLP to the north of 55°N in East Asia, but is negatively correlated with the SLP over Mongolia and North China. Both the BCC-CSM1.1 and BCC-CSM1.1m simulations reproduce the significant negative correlation over Mongolia and North China, but do not capture the positive correlation at latitudes of 55°–70°N in Asia (Figs. 4b, c). The BCC-CSM1.1m simulation gives a better capture of the spatial pattern of the correlation map and the strength of the negative correlation center over Mongolia and North China.

Figure 4 Correlation between the North China NWS and SLP in autumn in (a) observational data and (b) BCC-CSM1.1m and (c) BCC-CSM1.1 simulations. The black dots show the correlations above the 95% confidence level.

Figure options

The linear trend of the autumn SLP, NWS, and surface temperature is explored further in Fig. 5. The SLP over Mongolia and North China in the observational data increases in recent decades, whereas the SLP in Siberia decreases (Fig. 5a). This indicates that the north–south pressure gradient has weakened in Mongolia and North China, which partly explains the weakening of the NWS in North China. The BCC-CSM1.1m simulation gives a better reproduction of the distribution of the linear trend of the SLP, especially the weakening of the SLP to the north of Lake Baikal (51°–55°N) (Figs. 5c–e).

The change in the SLP is related to the heterogeneous change in surface temperatures over the Asian continent. The autumn surface temperatures over Mongolia and North China in the observational data show less warming than the neighboring region to the north of 50°N and in the Pacific Ocean (Fig. 5b). The relatively cooler surface corresponds to the increase in the SLP in Mongolia and North China. The relatively warmer surface to the north of 50°N in Asia is consistent with the decrease in the SLP. The BCC-CSM1.1m model gives a better simulation of the heterogeneous change in surface temperature in Asia than the BCC-CSM1.1 model (Figs. 5d–f). The spatial correlation coefficients between the simulated and observed linear trends of surface temperature over the region (15°–70°N, 70°–160°E) are 0.65 for the BCC-CSM1.1m and 0.52 for the BCC-CSM1.1 simulation, respectively. This partly explains why the BCC-CSM1.1m model gives a better simulation of the changes in SLP in the midlatitudes of the Asian continent.

Figure 5 Linear trend of the autumn SLP (shaded; hPa per century) and surface wind speed (contours; m s–1 per century) in (a) observation, (c) BCC-CSM1.1m simulation, and (e) BCC-CSM1.1 simulation. Linear trend of the autumn surface temperature (°C per century) in (b) observation, (d) BCC-CSM1.1m simulation, and (f) BCC-CSM1.1 simulation. The observed surface wind speed is from the NCEP–NCAR reanalysis data.

Figure options

The autumn SLP over Mongolia and North China (40°–50°N, 105–120°E) is averaged to show the evolution with time from 1961 to 2005. Figure 6a shows that the observed SLP over Mongolia and northern China has had an increasing trend in recent decades. This is the opposite to the observed decreasing trend of the NWS in North China. The correlation coefficient between the two time series is –0.68 (above the 95% confidence level). A similar opposite change in the NWS and SLP was also reported by Wang et al. (2004). The significant negative correlation between the SLP and the NWS can be better reproduced by the BCC-CSM1.1m simulation (–0.66) than by the BCC-CSM1.1 simulation (–0.41) (Figs. 6b, c).

Similar to the observation, the SLP over Mongolia and North China in the BCC-CSM1.1m simulation also shows an increasing trend (1.48 hPa per century) from 1961 to 2005 (Fig. 6b), although it is smaller than the observed trend (8.96 hPa per century). The linear trend (0.98 hPa per century) simulated by BCC-CSM1.1 is much weaker than that simulated by BCC-CSM1.1m. Therefore, it is concluded that the more reasonable simulation of the increased SLP over Mongolia and North China and its relationship with the North China NWS by the BCC-CSM1.1m simulation explains why this model gives a better simulation of the decrease in the NWS in North China.

Figure 6 Anomalies of autumn SLP (hPa) in Mongolia and North China (red, right-hand axis) and anomalies of NWS (m s–1) in North China (black, left-hand axis) relative to the climatology during 1961–2005 from (a) observational data and simulations by (b) BCC-CSM1.1m and (c) BCC-CSM1.1. The correlation coefficent (cor) between the autumn SLP and NWS is shown in the lower right corner of each panel.

Figure options

The performance of the BCC-CSM1.1m simulation in reproducing the weakening trend of the autumn NWS in North China enhances our confidence in this model for projecting future changes in the 21st century. Three RCP scenarios—RCP2.6 (low), RCP4.5 (middle), and RCP8.5 (high)—were selected to study the changes in the autumn NWS in China projected by the BCC-CSM1.1m simulation. Consistent with the observational data, there were few trends in the autumn wind speed anomalies in North China simulated by BCC-CSM1.1m in different RCPs during the time period 2006–17. This lack of trend is partly because the concentrations of greenhouse gases are very similar in all three RCPs in the first 20 years of the simulation (Van Vuuren et al., 2011). There is still much uncertainty in the interannual variation in the autumn wind speeds simulated by BCC-CSM1.1m in the three RCPs as a result of the internal variability of the atmosphere and the biases in the model. Because this study was focusing on the long-term trends in wind speed, we therefore excluded the time period 2006–17 and selected the time period 2018–2100 to explore the future projection of autumn surface wind speeds.

In the low-emission scenario (RCP2.6), the autumn NWS increases in North China and southeast China (Fig. 7a). In RCP4.5 and RCP8.5 (Figs. 7b, c), however, the autumn NWS decreases in Northwest China, North China, and Northeast China. The decreasing trend is greater in RCP8.5 than in RCP4.5. The surface wind speed over the Tibetan Plateau also clearly decreases in the high emission scenario (RCP8.5).

Figure 7 Linear trend of the autumn NWS (m s–1 per century) projected by BCC-CSM1.1m under (a) RCP2.6, (b) RCP4.5, and (c) RCP8.5 scenarios during the time period 2018–2100.

Figure options

Figure 8 shows the time series for autumn NWS in North China during the time period 2018–2100 in the RCPs projected by the BCC-CSM1.1m model. The NWS shows a slight increase in RCP2.6 (Fig. 8a) and decreases of 0.13 and 0.18 m s^–1 per century, respectively, in RCP4.5 and RCP8.5 (Figs. 8b, c). Because the model underestimates the historical weakening trend of the autumn NWS in North China by about two-thirds, it is inferred that there will be a much larger decreasing trend for the autumn NWS in North China in the 21st century than the trend projected by the BCC-CSM1.1m model in RCP4.5 and RCP8.5. In the 83 yr from 2018 to 2100, the BCC-CSM2-MR simulation in RCP4.5 and RCP8.5 projects 67 and 52 yr, respectively, with a NWS lower than the climatology (1961–2005). The mean NWS in North China is 0.21 m s^–1 lower in RCP4.5 and 0.13 m s^–1 lower in RCP 8.5 than the climatology by the end of the 21st century (2081–2100). The weakening of the autumn NWS in China by the end of the 21st century projected by the BCC-CSM1.1m model is consistent with the projections of climate models under the IPCC Special Report on Emissions Scenarios reported by Jiang et al. (2010). There is almost no change (–0.01 m s ^–1) for the NWS in North China during the late 21st century (2081–2100) under RCP2.6. Thus, it is important to reduce future emissions because weak NWS is one of the most important factors in the development of haze in North China (Niu et al., 2010; Guo et al., 2015; Pei et al., 2018; Zhong et al., 2018).

Figure 8 Autumn NWS anomalies (m s–1) in North China during 2018–2100 projected by BCC-CSM1.1m under (a) RCP2.6, (b) RCP4.5, and (c) RCP8.5.

Figure options

We evaluated simulations of the NWS over China by the BCC-CSM1.1 and BCC-CSM1.1m climate models with different horizontal resolutions. The BCC-CSM1.1m simulation was better in reproducing the spatial distribution of the NWS over China than the BCC-CSM1.1 simulation, including the weak winds over the Sichuan and Tarim basins and the strong winds over the western Tibetan Plateau. The BCC-CSM1.1 simulation was unable to reproduce the changes in the NWS in China during the time period 1961–2005 in all seasons, whereas the BCC-CSM1.1m simulation improved its ability to simulate the autumn NWS in North China, which showed a decreasing trend. This is related to the better performance of the BCC-CSM1.1m model in reproducing the increased SLP over Mongolia and North China. These results suggest that increasing the horizontal resolution in the BCC-CSM model improves its ability to simulate the spatial distribution and long-term changes in the NWS over China.

Future projections of the autumn NWS in North China for the time period 2018–2100 were explored in different RCPs with the BCC-CSM1.1m model. The autumn NWS in North China is projected to decrease at rates of 0.12 and 0.18 m s^–1 per century in RCP4.5 and RCP8.5, respectively. This indicates that increased future emissions of greenhouse gases will cause further weakening of the NWS in North China. Under the mitigation scenario RCP2.6, the autumn NWS in North China will change little by the end of the 21st century. Weak surface winds are conducive to the development of frequent haze events that occur in North China. Our results suggest that adopting a mitigation scenario will reduce the impact of changes in wind speed on the development of haze in North China in autumn.

Our comparison of the model results on different resolutions are based on only two models. The attribution of the improvements to the resolution of the models is currently only tentative. In addition, the initial conditions may introduce uncertainties in the simulation and projection of NWS. Further verification is needed with other models and the ensemble mean of the model outputs.