Global warming has become an indisputable phenomenon (Marcott et al., 2013). During 1901–2012, the mean global temperature rose by 0.89°C, and the trend of global warming has since strengthened. The average annual surface temperature increase is the strongest in the middle and high latitudes of the Northern Hemisphere; the warmest 30 years of the past 800–1400 years were probably 1983–2012 (IPCC, 2013). This warming trend also has great uncertainty at the regional climate scale, showing a transition of “warm-dry” to “warm-wet” with a double increase in mean temperature and annual total precipitation since the end of the 1980s in Xinjiang, China (Shi et al., 2007). To better predict climate development in the future, it is necessary to understand the changing rules and driving mechanisms of historical climate. However, the limited instrumental data are not enough to study the evolution of the paleoclimate. Tree rings have the characteristics of high resolution, good continuity, wide sample distribution and accurate dating and have been applied in historical climate reconstructions and in studies on the evolution of the paleoenvironment (Cook et al., 2010; Shao et al., 2010; Büntgen et al., 2011; Pederson et al., 2014).
Climate reconstruction based on tree rings is an important method of studying climatology. A large number of studies have been carried out worldwide, mainly in areas of wide distributions of coniferous trees in North America, northern Asia and Western Europe (Briffa et al., 1992; Popa and Kern, 2009; Matskovsky et al., 2017). Much of the climate history has been reconstructed in China, mainly in the northeastern and northwestern regions and on the Tibetan Plateau (Zhang et al., 2011; Zhu et al., 2011; Peng et al., 2013; Yu et al., 2018). The Tianshan Mountains, which are located in the arid and semi-arid inland area of Northwest China, represent an ideal area for the study of dendroclimatology due to its great climatic fluctuations and the highly sensitive coniferous trees (Cheng et al., 2012). A large number of climate reconstructions have been carried out in the Tianshan area since the 1960s and 1970s. Streamflow reconstructions have focused mainly on the central and western regions of the Tianshan Mountains, such as the Manas River, Jinghe River, Urumqi River and Guxiang River (Yuan et al., 2007; Chen and Yuan, 2016; Yang et al., 2016; Zhang et al., 2016). Precipitation reconstructions have been distributed throughout the Tianshan Mountains, especially in the western Tianshan Mountains (Yuan et al., 2001; Shang et al., 2010; Zhang et al., 2013a; Liu et al., 2015; Zhang et al., 2015; Chen et al., 2017). Drought reconstructions have been concentrated in the central and eastern Tianshan Mountains based on the Palmer drought severity index (PDSI), but fewer drought reconstructions have been based on the better Standardized Precipitation Evapotranspiration Index (SPEI) (containing precipitation, temperature, relative humidity, solar radiation, water vapour pressure, and wind speed) (Chen et al., 2013, 2016; Xu et al., 2015a; Xu et al., 2015b; Jiang et al., 2017). Temperature reconstructions have also been studied throughout the Tianshan Mountains, but they have mainly focused on the reconstruction of the mean temperature and the mean maximum temperature; research on the mean minimum temperature is scarce (Chen et al., 2009, 2012a; Chen et al., 2009; Shang et al., 2011; Yu et al., 2013; Zhang et al., 2013b).
Climate reconstruction is based on the ‘uniformitarian principle’, with a stable relationship between tree radial growth and climatic limiting factors (Fritts, 1976). However, some studies have discovered a ‘divergence problem’, with decreasing sensitivity of radial growth to the temperature response in both the middle-high latitudes and the high-altitude regions of the Northern Hemisphere as a result of the strengthening trend of global warming, which has put forward a great challenge to the reconstruction of the historical climate based on tree rings (Briffa et al., 1998; D’Arrigo et al., 2008; Franceschini et al., 2012). Moreover, the results of divergent responses are relatively complex, such as those involving
Samples were collected from the western region of the Tianshan Nature Reserve of Kurdistin in the Tianshan Mountains, Xinjiang, China (
In the laboratory, the tree cores were mounted on woody supports, air dried and sanded to produce clearly visible boundaries for identification, and three cores with ambiguous rings and large growth-curve variations were removed. The rings of the remaining 47 cores were measured with 0.001-mm resolution using the LINTAB measurement system (TM5, Rinntech, Heidelberg, Germany) (Schweingruber, 1996; Stokes and Smiley, 1968). The quality of all measurement series was verified using the COFFECHA program (Holmes, 1983). The measured ring widths were detrended with negative exponential curves or straight lines (Melvin and Briffa, 2008; Cook et al., 2013). A cubic spline with a 50% frequency-response cut-off equal to 67% of the series length was also used in several cases when both types of previous curves failed. All the detrended series were then combined into a composite chronology using a robust (biweight) mean function (Cook and Kairiukstis, 1990). After all these processes, three kinds of chronologies of the KRD sampling site were obtained: standard chronology (STD), residual chronology (RES), and arstan chronology (ARS). The ultimate reliable standard chronology (STD) began in 1680 AD, which was determined based on a subsample signal strength (SSS) value that exceeded 0.85 (Wigley et al., 1984) (
The temperature and precipitation data were obtained from Zhaosu (ZS) meteorological stations (43.15° N, 83.13° E, 1854 m a.s.l.) of the National Meteorological Information Center (NMIC) of China near our sampling site (149.3 km) (
In this study, Pearson correlation coefficients were calculated to identify the relationships between tree-ring chronology and the major controlling climate factors. The temporal stability of the growth-climate relationship was tested by a moving correlation function with a 30-year time window using the Dendroclim2002 program (Biondi and Waikul, 2004). A linear regression model was used as a transfer function to derive the reconstruction, and the statistical fidelity of the regression model was validated with split-sample calibration-verification tests (Cook and Kairiukstis, 1990). The statistics calculated for testing the accuracy of the reconstruction include the explained variance (
To determine the radial growth and dominant climate factors of trees, we analyzed the relationship between tree-ring width chronology and climate factors via Pearson correlation analysis (
The response stability of the radial growth to the mean minimum temperature was tested via a moving correlation (1959–2012) (
Based on the results of the correlation analysis and moving correlation, we reconstructed the mean minimum temperature in the growing season of the previous year using the standard chronology. A linear regression model was used to describe the relationship between the tree-ring chronology (STD) and mean minimum temperature. The model was designed as follows:
Where Tmin5–9 represents the mean minimum temperature value from May to September at a time scale of 5 months and STD represents the tree-ring index for the next year.
The model explained that the variance of the reconstruction was 62.7% (62% after adjusting for the degrees of freedom) for the calibration period (1959–2012), and the reconstruction series of the mean minimum temperature and observations were significantly correlated (
Calibration and verification statistics for the Tmin5–9 reconstruction in the western Tianshan Mountains.
Calibration | Verification | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
Period | Period | |||||||||
1960–1985 | 0.655** | 0.429 | 0.405 | 17.9** | 23+/3-** | 1986–2012 | 0.710** | 0.534 | 0.485 | 19+/8- |
1986–2012 | 0.710** | 0.503 | 0.484 | 25.3** | 19+/8- | 1960–1985 | 0.655** | 0.744 | 0.458 | 23+/3-** |
1960–2012 | 0.792** | 0.627 | 0.62 | 85.8** | 46+/7-** |
Based on the response analysis and regression model, the Tmin5–9 series of the KRD region during 1680–2012 was rebuilt, which can reflect the interannual variation in the mean minimum temperature (
Some significant low-and high-frequency cycles of the reconstruction series of the mean minimum temperature were revealed by the MTM analysis in the growing season of the previous year (
Climate reconstruction based on tree-ring enables us to understand not only the changes in the past climate but also the distinct climatic change rules and mechanisms and allows speculation of the trend of future climate development. In this paper, the mean minimum temperature of the growing season of 333 years in the western Tianshan Mountains was rebuilt by the samples of Schrenk spruce tree rings collected in the Western Tianshan National Nature Reserve of Kurdistan, and its variation and driving factors were analyzed.
The growth of trees is concentrated mainly in the growing season, so climate change during spring and summer has a great influence on the radial growth of trees (Chen et al., 2009; Peng et al., 2005; Yu et al., 2013; Linderholm et al., 2015; Zhang et al., 2018). Our study also confirmed these results, showing that the radial growth of Schrenk spruce has a significant correlation with the temperature from May to September of the previous year; the strongest correlation was with the mean minimum temperature (
Tree rings are important substitute materials for paleoclimate reconstruction, which has been confirmed in a large number of studies on tree-ring climatology (White et al., 2014; Jiao et al., 2016; Ponocna et al., 2016; Jiang et al., 2017). However, divergent phenomena have taken place with global warming, and dynamic responses have a great impact on the accuracy of the results of climate reconstruction based on tree rings (D’Arrigo et al., 2008; White et al., 2014; Jiao et al., 2015; Shen et al., 2016). Therefore, it is necessary to test the stability between radial growth and the main limiting climatic factors before climate reconstruction. There was no divergent response, and the response of Schrenk spruce to the climate factors was stable based on examination of the dynamic growth-climate relationship. Reconstruction results by divergent tests were relatively more reliable during the past 500 years of precipitation reconstruction in the southern part of Germany based on
To understand the spatial representation of the reconstruction results, the correlation coefficients were calculated between the measured values and reconstruction values of the mean minimum temperature and the regional CRU lattice from 1959–2012 (
To further verify the reliability of the results, the reconstruction series were compared with other reconstruction series near the study regions. Our reconstruction series significantly correlated with the mean minimum temperature series of Gongnaisi (
1938, 1957–1974 and 1708–1722, 1758–1790, 1908–1934 (Zhang et al., 2017). These results all indicated the accuracy of our reconstruction of the mean minimum temperature in the growing season from another aspect.
In addition, the reconstruction results were compared with the drought events recorded in the historical documents of Xinjiang regions and the volcanic eruption years by high/low mean percentage, it is the reconstructed value minus the reconstructed average divided by the reconstructed average. (Wen et al., 2006; Chen et al., 2012a). The mean minimum temperature reconstruction had a good correspondence with the drought events in the Xinjiang region, and the values of the drought events were significantly higher than the average reconstructed temperatures, especially in some extreme drought years, such as 1775, 1919, and 1945 (Chen et al., 2013; Liu et al., 2013; Zhang et al., 2017). Radial growth of trees is also responsive to volcanic eruptions and temperature changes had been importantly affected by many upper westerly circulation and low latitude tropics massive volcanic eruptions in central Asia (Chen et al., 2012b). After the eruption, the width of the growth-rings increased due to reduced temperature and intensified precipitation. Therefore, it was found that the temperature in the year of volcanic eruptions was obviously lower than the average value of the reconstructed temperature (
The dry and volcanic years recorded in historical documents (high/low mean percentage were calculated with the reconstructed value minus the reconstructed average divided by the reconstructed average).
No. | Year (high/low mean percentage) | Historical Document Records |
---|---|---|
1 | 1775 (23.98%) | Drought in Xinjiang in 1775 was of such devastating proportions that 80% of the grain crops failed. |
2 | 1874 (3.33%) | Tacheng, Qitai and Jimusal counties drought, harvest apology. Changji drought and locust frequency. |
3 | 1893 (11.36%) | Drought occurred in Dihua (Urumqi now) and Changji regions in 1893. |
4 | 1913 (8.39%) | Droughts in the town of Zhenxi county. |
5 | 1917 (26.55%) | Yili, Yingjisha, Pishan and other counties drought. |
6 | 1918 (27.42%) | Severe drought occurred in more than three regions in 1918, such as Bachu, Jiashi, Yecheng, |
7 | 1919 (27.27%) | Drought occurred in Yanqi region in 1919. |
8 | 1921 (16.32%) | In Yining, Turpan, Shanshan and other counties, summer crops summer failed and to autumn harvest, drought, autumn grain difficult to replant. |
9 | 1938 (4.10%) | In Yili area, flood and drought alternate, and lack of rain in summer and autumn, field seedlings dry account for 9 out of 10. |
10 | 1945 (14.53%) | Drought occurred in Miquan region in 1945. |
11 | 1680 (–27.57%) | Tongkoko in Sulawesi (1680). |
12 | 1696 (–26.04%) | Komaga-Take in Japan (1695). |
13 | 1708 (–19.07%) | Fuji, Japan (1707). |
14 | 1784 (–9.69%) | Laki in Iceland (1783). |
15 | 1800 (–14.83%) | St Helens in the US (1800). |
16 | 1812 (–19.45%) | Awu in Indonesia (1812). |
17 | 1815 (–18.36%) | Tambora in Indonesia (1815). |
18 | 1854 (–2.15%) | Sheveluch, Kamchatka (1854). |
19 | 1883 (–22.86%) | Krakatau, west of Java (1883). |
20 | 1886 (–4.70%) | Okataina in New Zealand (1886). |
21 | 1903 (–5.32%) | Santa Maria, Guatemala (1902). |
22 | 1983 (–8.59%) | El Chichon, Mexico (1982). |
Regional climate change is affected by large-scale climate-forcing factors, such as solar activity and atmospheric–oceanic variability (Fang et al., 2010; Linderholm
On the one hand, based on the MTM analysis results, the temperature changes in the study area are mainly influenced by the solar activity via the mean minimum temperature within approximately 11-year periods (Li et al., 2006; Wang et al., 2015). The cycle similarity of the reconstruction values with the North Atlantic Oscillation Index (NAO) indicated that the climate might be affected by the NAO in the study regions: the values included 2.06 (2.07), 3.90 (3.91), 3.96 (3.95), 5.44 (5.43), 5.56 (5.57), 64.10 (63.69), 72.99 (72.99), 102.04 (102.04) and 113.64 (113.64), respectively. The climate is influenced mainly by the westerly circulations in the Tianshan Mountains, showing that the warm wet air in the Atlantic Ocean enters the Xinjiang regions driven by the westerly wind circulations and affects the trees in the Yili Valley and the western Tianshan Mountains. Moreover, the temperature changes in the cycle in the range of 5.4–5.9 and 3.9 years might be affected by the El Niño-Southern Oscillation (ENSO), and in the range of approximately 2.1 years, these changes might be related to tropical biennial oscillation (TBO) (Allan et al., 1996).
On the other hand, regarding the spatial correlations between the reconstruction values and the global sea surface temperature (SSTs) in May to September, the reconstruction was strongly correlated with the northern Atlantic, western Pacific and Indian Oceans (
1950–2012), which indicates that the change in the sea temperature in the Western Pacific Ocean has an important influence on the temperature of the study area in the summer (
The tree-ring chronology was developed by samples of Schrenk spruce collected from the National Nature Reserve of the Western Tianshan Mountains. The mean minimum temperature in the growing season is the main and stable limiting climate factor. Therefore, the mean minimum temperature series in the growing season during 1680–2012 was reconstructed based on the STD chronology. In the past 333 years, the mean minimum temperature has roughly experienced three relatively cold periods and relatively warm stages (relatively cold periods: 1680–1707, 1802–1911 and 1935–1997; relatively warm periods: 1708–1801, 1912–1934 and 1998–2012). By analyzing similar trends in regional temperature changes in our reconstruction series with drought events, large volcanic eruptions and other reconstruction series around the study regions in Xinjiang and even large-scale regions, we found that the mean minimum temperature of the reconstruction was accurate and reliant. Moreover, the mean minimum temperature was influenced by solar activity (sunspots) and large-scale atmospheric–oceanic fluctuations (NAO, WPO, ENSO, TBO) based on the MTM and spatial correlation analysis.