Extreme climate events can cause property damage, injury and loss of life and understanding their occurrence is very important to natural and human systems (Katz and Brown, 1992; Easterling et al., 2000; Aguilar et al., 2009). As a consequence, the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) paid more attention to climate extremes change (IPCC, 2007). Recent studies on global, regional and national scales have significantly improved the understanding of temperature and precipitation extremes (Peterson et al., 2002; Aguilar et al., 2005, 2009; Alexander et al., 2006; Klein Tank et al., 2006; New et al., 2006; Brown et al., 2008; Peterson and Manton, 2008; Peterson et al., 2008; You et al., 2008a,b; Choi et al., 2009; Caesar et al., 2011; You et al., 2011a). Most of these studies have been fostered by the World Meteorological Organization (WMO) Joint Expert Team on Climate Change Detection and Indices (ETCCDI) (Peterson and Manton, 2008). They have revealed that cold extremes are generally changing more rapidly than warm extremes, but the exact reasons have not been explored in detail.
The Arctic Oscillation (AO), currently known as Northern Annular Mode, is one of the dominant patterns of Northern Hemisphere climate variability, and it is most prevalent in winter and in the mid and high latitudes. It strongly influences surface air temperatures over the Eurasian continent, especially Europe (Hurrell, 1995; Thompson and Wallace, 1998; 2001; Hurrell et al., 2001; Hurrell and Deser, 2010). AO is a major controlling factor in basic meteorological variables such as surface wind, temperature and precipitation (Bojariu and Gimeno, 2003). AO is defined as a hemispheric mode whose dipole has suffered a displacement to the West during the last decades (Ramos et al., 2010). The AO index has been used to describe the variability of AO in this study.
Recent studies have shown that the winter AO index has a strong positive correlation with temperatures in northern China (Gong and Wang, 2003) and is also correlated with the strength of the East Asian winter monsoon and Siberian Higher pressure system (Gong et al., 2001; Wu and Wang, 2002). Since the 1980s, China has experienced significant temperature increases (Wang and Gong, 2000; Ding et al., 2007), and warming is projected to continue. Although trends in temperature extremes on the annual basis have been studied (Zhai et al., 1999; Zhai and Pan, 2003; Ren et al., 2011; You et al., 2011a), there have been little investigations of how the AO influences winter temperature extremes. Thus we quantify changes in winter temperature extremes during 1961–2003 throughout China, based on indices designed by the Commission for Climatology/Climate Variability and Predictability/Joint WMO Intergovernmental Oceanographic Commission Technical Commission for Oceanography and Marine Meteorology ETCCDI. The relationships between the AO index and winter temperature extremes are also examined.
2. Data and methods
Daily maximum and minimum temperatures for 303 stations in China are provided by the National Meteorological Information Center, China Meteorological Administration. Both the spatial density of stations and the quality of observational data in China meet the World Meteorological Organization's standards. Stations were selected according to procedures described in our recent papers (You et al., 2011a). The selected stations should have the long-term data records and good data quality. The calculation of indices is facilitated using the information provided by ETCCDI (see http://cccma.seos.uvic.ca/ETCCDI for available calculated station-level indices) (Peterson and Manton, 2008). We concentrate on the winter (DJF) variation of five temperature indices (Table I), which have been shown to be most sensitive to climate change in previous studies (You et al., 2008a, 2011a). The winter temperature extremes have the same definition as in previous studies (Aguilar et al., 2005, 2009; Alexander et al., 2006; Klein Tank et al., 2006; New et al., 2006; You et al., 2008a,b; Caesar et al., 2011; You et al., 2011a). RClimDex software was used to perform data quality control and calculate the indices, and RHtest was used to assess homogeneity. Details about data quality control and homogeneity tests are described in our previous papers (You et al., 2008a, 2011a).
Table I. Definitions of five winter temperature indices used in this study. All indices are calculated by RClimDeX software
TX is the daily maximum temperature; TN is the daily minimum temperature.
Cold day frequency
Percentage of days when TX < 10th percentile of 1961–1990
Cold night frequency
Percentage of days when TN < 10th percentile of 1961–1990
Warm day frequency
Percentage of days when TX > 90th percentile of 1961–1990
Warm night frequency
Percentage of days when TN > 90th percentile of 1961–1990
Diurnal temperature range
Annual mean difference between TX and TN
The AO index is defined as the difference in the normalized monthly zonal-mean sea level pressure (SLP) between 35 and 65°N (Li and Wang, 2003), derived from http://web.lasg.ac.cn/staff/ljp/data-NAM-SAM-NAO/ NAM-AO.htm. Monthly mean geopotential height, air temperature, zonal and meridional wind were obtained from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/ NCAR) reanalysis (available from their website at http://www.cdc.noaa.gov/) (Kalnay et al., 1996). The relationships between solar activity and winter temperature extremes are studied in the study and the solar activity is derived from the studies in Kodera (2002) and Ogi et al(2003).
The Mann–Kendall test for trends and Sen's slope estimates are used to detect and quantify trends in winter temperature extremes (Sen, 1968), with magnitudes of trends and slopes assessed at the 0.05 significance level (p < 0.05).
3.1. Winter temperature extremes (TX10, TN10, TX90, TN90 and DTR)
Figure 1 shows the spatial patterns of trend for five winter temperature indices (for 303 meteorological stations) along with the time series of the entire country. Aggregated regional trends of winter temperature extremes are listed in Table II (third column), calculated as the arithmetic mean of all station. The number of stations with negative, no trend and positive trends, as well as the number of stations passing the significant level for each index is also shown in Table II.
Table II. Trends per decade (with 95% confidence intervals in parentheses), and the number of stations with positive (significant at the 0.05 level), non-trend, and negative (significant at the 0.05 level) trends for winter temperature indices in the entire country
Values for trends significant at the 5% level (t-test) are set in bold.
− 1.33 (−2.70 to − 0.25)
−2.98 (−3.96 to − 1.90)
0.92 (0.05 to 1.85)
2.35 (1.30 to 3.27)
−0.25 (−0.39 to − 0.14)
For cold days (TX10) and cold nights (TN10), about 97 and 98% of stations have decreasing trends, whereas 31 and 84% of stations are statistically decreasing trends. For TX10, stations in the northern China (such as Gansu province) show larger trend magnitudes, and significant decreasing trends are shown over most regions in China for TN10 (Figure 1). TX10 has shown some slight increasing changes since 1990, but the decrease in TN10 has been much more consistent before the 1990s, with only a slight levelling off after that. The countrywide trend (in % of days) for these two indices are − 1.33 and − 2.98 d/decade, respectively (p < 0.05).
For the percentage of days exceeding the 90th percentiles (TX90 and TN90), about 79 and 98% of stations have increasing trends, and about 30 and 70% of stations show statistically significant and increasing trends. Stations in the northern China show larger trend magnitudes for both TX90 and TN90 (Figure 1). Some stations in the southern China have decreasing trends for TX90 and non-significant increasing trends for TN90. Before the mid-1980s, both TX90 and TN90 have fluctuant (decreasing and increasing) changes, and show statistically increasing trends after that. The trends in the entire country for these two indices are 0.92 and 2.35 d/decade, respectively (p < 0.05).
For diurnal temperature range (DTR), about 87% of stations show decreasing trends, while 47% of stations decrease significantly. Similarly, stations in the northern China between 40° and 50°N show larger trend magnitudes, where have more pronounced warming. This illustrates that more warming leads to larger decreases for DTR (You et al., 2008a, 2011a). DTR has shown a significant decreasing trend before the 1990s, with only a slight levelling off after that 1990. The overall trend in the entire country for DTR is − 0.25 °C/decade (p < 0.05), which is larger than the annual DTR trend in the Tibetan Plateau (−0.20 °C/decade) during 1961–2005 (You et al., 2008a) and entire China (−0.18 °C/decade) during 1961–2003 (You et al., 2011a).
3.2. Comparison with the annual temperature extremes
In the previous study, the spatial and temporal distributions of temperature extremes on the annual basis have been analysed using the same datasets (You et al., 2011a). Countrywide, the annual trends for TX10, TN10, TX90, TN90 and DTR are − 0.47 d/decade, − 2.06 d/decade, 0.62 d/decade, 1.75 d/decade, − 0.18 °C/decade, respectively. For TX10, TN10 and DTR, about 77, 97, 80% of stations have decreasing trends, and about 83 and 94% of stations have increasing trends for TX90 and TN90, respectively. Compared with the results at the annual scale, the absolute trend magnitudes of winter temperature extremes are higher, and the proportions of stations with positive/negative trends are larger with the exception of TX90. Thus, the spatial and temporal patterns of winter temperature extremes are broadly similar to those on the annual basis, but the trends of temperature extremes in winter are generally higher, indicating pronounced climate warming in winter.
3.3. Correlation with the AO
The correlation between winter temperature indices and the AO in China during 1961–2003 are shown in Figure 2. National linear correlations and coefficients are listed in Figure 3. Strongest correlations occur in the northern China for TX10 and TN10 (some values lower than − 0.5), and the correlations are slight in the southeastern part of the Tibetan Plateau for TX10 and TN10 (Figure 2). Taking China as a whole, the AO index is significantly correlated with the winter cold temperature extremes (TX10 and TN10). It is negatively correlated with TX10 (R = − 0.42, p < 0.01) and TN10 (R = − 0.65, p < 0.01) during the studied period. For the winter warm temperature extremes, the AO index is positively correlated with TX90 and TN90, but only the correlations with TN90 (R = 0.47, p < 0.01) pass the significant level. In most cases, it is clear that the northern and northwestern China have larger correlation coefficients, and the southeastern China have lower values for the winter warm temperature extremes (TX90 and TN90, especially for TN90). Meanwhile, the AO index also has significantly negative correlation with DTR with the value of − 0.51 (Figure 3), and correlation coefficients in most regions are more than − 0.3. Thus winter temperature extremes are strongly connected with the AO index, especially in the northern China.
3.4. Atmospheric circulation composite analysis
In order to examine the influence of AO on climate extremes, the differences of mean winter temperature extremes in positive and negative winter AO years during 1961–2003 are presented (Figure 4). The 23 positive and 19 negative winter AO years are based on whether the AO index is above or below the mean value. The differences (positive minus negative AO years) show that the majority of stations have negative values for TX10, TN10 and DTR, and positive values for TX90 and TN90, while there has spatial variability (Figure 4). Thus winter temperature extremes are significantly different during winters with positive versus negative AO phases, which is consistent with that there are significant relationships between AO and temperature extremes (Figures 2 and 3).
To show the influence of atmosphere circulation on winter temperature extremes, Figure 5 shows the differences (positive minus negative AO years) of mean geopotential height and wind field (m s−1) at 850 hPa during 1961–2003. The selected region covers the domain 10°—70°N and 40°—160°E. The largest negative differences in geopotential height are approximately 30 geopotential meter (gpm), with enhanced cyclonic circulation over the region (focused near 60°N and 60°E) (You et al., 2011b). This generates an anomalous south-westerly flow in the Siberian region and northern China which carries relative warm air into inland, reducing the intensity of Asian winter monsoon. These results suggest that decreasing trends for winter cold temperature extremes and increasing trends for winter warm temperature extremes are highly related to the circulation change (You et al., 2011b).
To examine the atmosphere circulation, the spatial trends of geopotential height (A), air temperature (B), zonal wind (m s−1) (C) and meridional wind (m s−1) (D) at 850 hPa in winter during 1961–2003 are presented in Figure 6. The geopotential height at 850 hPa has decreasing trends at high latitude between 40° and 60°N and increasing trends at low latitude between 10° and 40°N (Figure 6(A)), suggesting that the asymmetrical changes between high latitude and mid latitude will begin to reduce the winter monsoon system, which is consistent with the asymmetrical global warming (IPCC, 2007). Although, the air temperature has increasing trends and more pronounced warming in the northeastern China (Figure 6(B)), the zonal and meridional wind increases significantly between 40° and 60°N (Figure 6(C) and (D)), revealing that the western and southern wind are increasing. The atmosphere conditions support the hypothesis that the increasing contrast between high and mid latitude will reduce the winter monsoon and influence the outbreak of winter temperature extremes.
The tropospheric temperature contrast between high and mid latitude is also of great importance to form the winter monsoon, supporting the atmospheric circulation for the change of temperature extremes. The tropospheric temperature (unit is °C) is defined as the average of air temperature vertically integrated between 200 and 1000 hPa based on the NCEP/NCAR reanalysis. The differences of composite tropospheric temperature during the period of 1961–1982 and 1983–2003 (latter minus former) are show in Figure 7 (top plot). The tropospheric temperature has larger values at the higher latitude, especially near the 55°N and 100°E (1.2 °C). These atmospheric patterns will reduce the transport of energy through baroclinic waves, diminish the strength of troughs and ridges, and increase the occurrence of calm atmospheric conditions (Niu et al., 2010). Thus, the transport of cold air originating from high latitude around 70°N will become less powerful and influence the frequency of warm temperature extremes (Gong et al., 2001; Niu et al., 2010).
The AO may influence the warm temperature extremes in China through the contrast of atmosphere conditions. Differences of composite tropospheric temperature in strongly positive and strongly negative winter AO years during 1961–2003 are shown in Figure 7 (bottom plot). Strongly positive (1982, 1988, 1989, 1991 and 1994) and strongly negative (1962, 1964, 1968, 1976 and 1978) winter AO years are those with index anomalies exceeding ± 1σ. During the positive AO years, the tropospheric temperature is positive in most southern China (near 1 °C), and there has negative anomaly in the north of China (almost 0.8 °C). Thus, the enlarging contrast of tropospheric temperature between high (around 60°N) and mid latitude (around 30°N) are helpful to bring more warm air flow from the ocean and prevent the cold air flow from the north, which will weaken the Asian winter monsoon and reduce the cold outbreaks. The patterns are similar to the correlation map of vertical-latitude from the 1000 to 10 hPa (Figure 8), which the AO has significant negative/positive correlations with atmospheric variables (geopotential height, temperature and wind) at high/mid latitude at both troposphere and stratosphere.
3.5. Influenced by the solar activity
Previous studies have shown that the extension of the AO differs significantly for different phases of solar activity (Kodera, 2002; Gimeno et al., 2003; Ogi et al., 2003). In order to investigate the effect of solar activity on winter temperature, the studied period has been separated into two phases for maximum and minimum solar activity, depending on whether the solar fluxes are above or below the mean value (Kodera, 2002; Gimeno et al., 2003; Ogi et al., 2003). The 18 years (1967–1971, 1979–1983, 1989–1992 and 1999–2002) are classified as the solar maximum years, and the 24 years (1961–1966, 1972–1978, 1984–1988 and 1993–1998) as the solar minimum years. The classification is the same as the studies in Kodera (2002) and Ogi et al(2003).
During solar maximum years, about 70, 49, 17 and 47% of stations for TX10, TN10, TX90 and TN90, respectively, have larger values than that during solar minimum years. It is clear that the larger negative differences between solar maximum and minimum years for both TX10 and TN10 are shown in the western and northwestern China, while stations in the western and northwestern China show larger positive values for both TX90 and TN90 (Figure 9). In most regions, DTR is sensitive to the change of solar activity, and about 93% of stations have larger values during solar minimum years than that during solar maximum years, resulting the negative differences between them (Figure 9). This is probably because more solar activity will heat the surface and increase the winter surface temperature, thus influencing the winter temperature extremes. Moreover, solar activity can also influence the atmospheric circulations which are memorized in the snow-cover, ice and permafrost regions (Ogi et al., 2003). The western China especially in the Tibetan Plateau is more sensitive to climate change due to the larger cryospheric area, and shows stronger signal. This probably suggests that solar activity can influence the winter temperature extremes to some extents and vary with the surface conditions, while the detailed mechanism needs to be investigated in future studies.
4. Discussion and conclusions
Winter temperature extremes have been shown to be warming faster than annual mean warm extremes (Aguilar et al., 2009) and are therefore particularly sensitive to future change. We have examined the spatial and temporal distributions of trends for four winter temperature extreme indices, using 303 stations in China over the period 1961–2003. For the majority of stations, significant decreases in cold days/nights (TX10/TN10) are observed with mean rates of − 1.33 and − 2.98 d/decade, respectively, while significant increases in warm d/nights (TX90/TN90) are also observed with mean rates of 0.92 and 2.35 d/decade, respectively. Such changes are consistent with previous studies in other parts of the world (Peterson et al., 2002, 2008; Aguilar et al., 2005, 2009; Alexander et al., 2006), and show that changes in winter temperature extremes reflect the consistent winter warming in China (You et al., 2008a,b, 2011a). The asymmetric changes in minimum and maximum temperature result in the declining DTR with rate of − 0.25 °C/decade. In most cases, stations in the north of China have the largest trend magnitudes, again consistent with rapid warming in the region (Wang and Gong, 2000; Ding et al., 2007).
Changes in winter temperature extremes are consistent with the report of You et al. (2011a) at the annual scale, but the trend magnitudes are higher than those, suggesting the winter is more sensitive to the extremes. The causes about the temperature extremes change have been studied, but require further study. Besides gas greenhouse gas emissions, You et al. (2011a) considered that temperature extremes change is probably associated with rapid urbanization, increased industrial aerosols and non-climate factors such as population, economic activity and local energy usage. The influences become particularly significant in China because of its rapid urbanization and economic activity (Qian and Lin, 2004).
The AO influences surface air temperature not only over the bulk of the Eurasian continent (Hurrell, 1995; Thompson and Wallace, 1998) but also in northern China (Gong and Wang, 2003). The winter AO index is significantly negatively correlated with TX10/TN10 and DTR, and positively correlated with TX10/TN10, indicating that the AO influences winter cold/warm extreme temperature. During the strongly positive AO index years, enhanced cyclonic circulation over the Urals (focused near 50°N and 60°E) brings more warm airflow into northern China, decreasing the strength of the East Asian winter monsoon and limiting its southward extension (Figure 10). This is consistent with previous research that shows that atmospheric circulation changes have contributed to the changes in climate extremes in China (You et al., 2011a). Other work has also suggested that an increase in strong positive AO phases could lead to a decreasing East Asian winter monsoon (Gong et al., 2001; Wu and Wang, 2002). Composites of atmospheric circulation shown in this study also support the relationship between AO variability and the strength of winter cold outbreaks in the northern China.
It is notable that the winter AO index has shifted several phases during 1873–2010 (Figure 11), derived from Li and Wang (2003). Winter AO index increases since the 1960s and has the downward trend since 1990 (Hurrell and Deser, 2010), confirmed by the weakening East Asian winter monsoon (Niu et al., 2010). The asymmetrical change in geopotential height, zonal and meridional wind may reflect a weakening of the East Asian winter monsoon. At the same time, more warming at high latitudes also reduce the thermal contrast, contributing to the weakening the East Asian winter monsoon (Figure 10). This will reduce the invasion of dry and cold air from the northern regions, creating a favourable background for temperature extremes. The limitation of this study is that the study period stops at the end of 2003, and the recent winters have strongly negative AO values in 2009 and 2010, especially in 2010 (Figure 11). In 2010, China has experienced the coldest winter since 1987, with the annual mean temperature of − 4.7 °C (Ren et al., 2011). This supports the hypothesis that the AO can modulate the winter temperature extremes by the atmosphere conditions. Meanwhile, the AO can be modulated by the 11 year solar cycles (Kodera, 2002). In this study, winter temperature extremes have stronger values during solar maximum years in most cases. But the mechanical linkage between solar activities, the AO and temperature extremes need to be investigated in future studies. Our results indicate also that further investigation of the linkage between the AO and climate extremes in China is worthwhile.
This study is supported by the Global Change Research Program of China (2010CB951401), the Chinese Academy of Sciences (KZCX2-YW-145), and the National Natural Science Foundation of China (40870743). The China postdoctoral science foundation (the 49th) is also appreciated. This study became possible through a Sino-Swiss Science and Technology Cooperation (SSSTC) research grant (EG76-032010 and EG23-092011). The authors thank the National Meteorological Information Center, China Meteorological Administration (NMIC/CMA), for providing the data for this study. Qinglong You is supported by the Alexander von Humboldt Foundation.