Introduction

Summer monsoon precipitation is very important to agricultural and social development in monsoon regions, and drought and flood disasters caused by abnormal changes of the monsoon affect the socioeconomic status of the people living in these regions, which accounts for two-thirds of the world’s population. To better adapt to the complex and fluctuating natural environment in monsoon areas, it is necessary to understand the temporal and spatial variations of monsoon precipitation. However, although precipitation in the monsoon area is mainly controlled by monsoons (An et al. 2000), spatial divergence of monsoon precipitation exists in the East Asian summer monsoon (EASM) region (Ding et al. 2008; Zhou et al. 2009a). Therefore, it is necessary to investigate the history of hydrological climate change in North China.

Considering the time span, it is often not enough to understand past climate change using only instrumental data. Using paleoclimatic data to place recent climate changes in the context of centuries or millennia is helpful for grasping the climatic variation pattern and then predicting future climate change (Ljungqvist et al. 2016). Tree rings are one of the most ideal proxy materials for recording climate change over the past century and millennium because of their dating accuracy, temporal continuity, temporal resolution (annual and even seasonal resolution), and wide geographical distribution. Climatic information of the past century or even millennium (Yang et al. 2014) has been obtained from tree rings, covering most parts of the world (Mann et al. 2009). Cai et al. (2010) reconstructed the average temperature from May to July based on tree-ring widths of Chinese pine on Mt. Beiwudang. However, besides traditional tree-ring width, many other research methods have been developed in dendroclimatology (Pearl et al. 2020), such as tree-ring density (Briffa et al. 2001), stable isotope ratio (McCarroll and Loader 2004), and wood anatomy (Fonti et al. 2010). Due to its clear physiological mechanism and the ability to maintain low-frequency climatic signals, tree-ring stable isotope climatic reconstruction has developed rapidly in the past few decades (McCarroll and Loader, 2004). Furthermore, because many tree-ring δ18O reconstruction results around the study area have shown significant correlations with hydroclimatic parameters (Li et al. 2011; Liu et al. 2020), we established a tree-ring δ18O record of Mt. Beiwudang to investigate the hydroclimatic variations in adjacent regions.

Materials and methods

Study area and field sampling

The sampling site is located on Mt. Beiwudang (37° 46′ N, 111° 20′ E, 1740–1900 m above sea level, Fig. 1a), situated in North China and belonging to the East Asian Monsoon Region (Ding and Chan 2005). This mountainous area is far from cities, where the dominant coniferous tree species is Chinese pine (Pinus tabulaeformis Carr.), growing with little human disturbance. The sampling strategy follows dendroclimatological practices (Fritts 1976; Leavitt 2010), targeting climatic-sensitive trees growing on steep rock slopes with a thin soil layer, with discontinuous canopies and no pests or diseases. To facilitate cross-dating, 19 Chinese pines of different ages were sampled using a 10-mm-diameter increment borer at breast height (two cores per tree).

Fig. 1
figure 1

a Map showing the locations of the sampling site in this study (Mt. Beiwudang, green triangle), meteorological stations (black circles), comparison sites from previous studies (yellow block and box), and the regional instrumental data covering the Meiyu/Baiu/Changma rainfall belt (blue box). Spatial correlation analysis result of the RHJA based on the CRU TS 4.04 (land) precipitation dataset during July-August from 1961 to 2013 AD is shown in the map, and only the significantly correlating region (p < 10%) is colored, where red denotes positive and blue denotes negative. b Monthly distribution of precipitation, temperature, and relative humidity for three meteorological stations during 1961–2013. c The reconstructed RHJA, regional observed rainfall of the blue box shown in Fig. 1a and their 21-year moving correlation, where the arrow denotes the recent rainfall trend

Full size image

Climatic data

In consideration of the large spatial representativeness of tree-ring oxygen isotope proxy data (Liu et al. 2020), three national meteorological stations, i.e., Yan’an (36° 36′ N, 109° 30′ E), Taigu (37° 26′ N, 112° 32′ E), and Xingxian (38° 28′ N, 111° 08′ E), were selected to determine the climatic variations across the studied region. They are evenly distributed in different directions near the sampling site (Fig. 1a). The instrumental data were obtained from the China Meteorological Science Data Sharing Service Network (http://data.cma.cn/). From 1961 to 2013 AD, there were no records of relocation or monthly data missing from the three meteorological stations, and data variations conformed to a normal distribution. The similar monthly distribution of climatic data (averaged over 1961–2013 AD) suggests that the three meteorological stations are located in the same climate area with a temperate continental monsoon climate (Fig. 1b). Therefore, the arithmetic average of the instrumental data from the three stations can be used to represent the regional climatic variation.

There are many climate indices used in this study, where the East Asian Summer Monsoon Index (EASMI) has different definitions. The EASMI (Shi and Zhu 1996) is defined as the sum of the standardized sea-level air pressure (SLP) differences (110° E minus 160° E) at seven parallels between 20° N and 50° N, and then the sum is standardized again. Because there is low pressure over the Asian continent and a subtropical high over the ocean in summer, a more negative EASMI denotes a stronger monsoon flow. The EASMI (Lau et al. 2000) is defined as a shear vorticity index of 200-hPa zonal winds, mainly focusing on the upper-tropospheric westerly jet stream. The EASMI (Huang and Yan 1999) reflects 500-hPa vorticity at three grids in the East Asia-Pacific (EAP) region, focusing on the EAP teleconnection. NINO1+2, NINO3, NINO3.4, and NINO4 are sea surface temperature anomalies in the NINO1+2 region (80° W–90° W, 0°–20° S), NINO3 region (90° W–150° W, 20° S–20° N), NINO3.4 region (120° W–170° W, 20° S–20° N), and NINO4 region (150° E–160° W, 20° S–20° N), respectively, based on the ERSST v5 dataset. The SOI is the Southern Oscillation Index from standardized NCEP Tahiti-Darwin SLP data, historical data, and recent data (http://www.cpc.ncep.noaa.gov/data/indices/soi). The MEI is the Multivariate ENSO Index (http://www.cdc.noaa.gov/people/klaus.wolter/MEI/ESRL).

The developing of tree-ring cellulose δ18O sequences

The tree-ring cellulose δ18O sequences were developed after following experimental steps.

Cross-dating

The tree-ring widths were measured by a Lintab Tree-ring Width Measurement System at a resolution of 0.01 mm. The calendar year of each ring was confirmed by matching all tree-ring width sequences and then dating according to the sampling time. The quality of cross-dating was controlled by the COFECHA program (Holmes 1983).

Splitting samples

Four cores with clear ring boundaries and few missing rings were selected for the isotope study. Each ring was split into small pieces (thickness < 0.5 mm) using a razor blade under a microscope and loaded into clean glass bottles marked with the year and core identifier.

Extracting α-cellulose

The α-cellulose was extracted from the samples using a modified Jayme-Wise method (Loader et al. 1997): (i) removed organic impurities using a mixture of methylbenzene and anhydrous alcohol; (ii) decomposed the lignin using acidified NaClO2; (iii) removed hemicellulose using NaOH solution; (iv) homogenized cellulose using an ultrasonic cell disruptor and dried samples using a vacuum freeze-dryer.

Measuring δ18O values

Each α-cellulose sample was packed in a silver capsule and measured automatically one by one. After every 8 samples, a laboratory-standard sample (Merck) was measured to calibrate the measurement results. The Merck sample was a thin-layer cellulose microcrystalline (EMD Millipore Corporation) standard, and its δ18O value (27.70‰VSMOW) was measured with IAEA-601 (Coplen et al. 2006) as a reference. IAEA-601 (benzoic acid) is a reference material recommended by the International Atomic Energy Agency (IAEA). The δ18O values of the samples were measured using a temperature conversion elemental analyzer (TC/EA) interfaced with a Delta V Advantage Isotope Ratio Mass Spectrometer. The analytical precision of the δ18O values of the laboratory-standard samples in this study was ± 0.19‰ VSMOW at a 95% confidence level. Finally, the measured δ18Otree value was calibrated in reference to Merck samples.

The compositing of δ18Otree chronology

In this study, the early period of the result shows low cohesiveness. This is mainly because of the dating error due to low sample depth in the early years. A threshold of expressed population signal (EPS) above 0.85 was selected as a cut-off to delete these years. It is worth noting that the EPS threshold of 0.85 is prevalent but not compulsory for tree-ring climatic studies (Baker et al. 2008; Fowler et al. 2004; Chen et al. 2015). The interval from 1700 to 1764 was ignored due to low cohesiveness (EPS < 0.85) and no duplications (less than 2 cores), and the data from 1765 to 2013 were used to develop the δ18Otree chronology using a “numerical mix method” (Liu et al. 2012): The δ18Otree chronology is composited from the arithmetic average of four normalized sequences.

Results

δ18Otree sequences and composite chronology

Four δ18Otree sequences and the composite chronology were finally obtained (Fig. 2a). EPS is above 0.85 for most of the period (Fig. 2a), suggesting that these four cores are sufficient for tree-ring isotope studies at this site (Wigley et al. 1984; Leavitt 2010). The statistical characteristics of the individual δ18Otree series and the correlation coefficients among them are shown in Table 1.

Fig. 2
figure 2

a δ18Otree chronology (thick line) and four δ18Otree sequences (thin line) from Mt. Beiwudang, EPS, Rbar (mean of the correlation coefficients among the samples) and the number of cores, where Rbar and EPS were calculated for every 50 years with a lag of 25 years. b Correlation coefficients between instrumental data and δ18Otree chronology during the overlapping period (1961–2013), where pApr represents the previous-year April for example. c The δ18Otree and calibration data (observed Relative Humidity of July to August for 1961–2013) scatter plot. d RHJA during 1765–2013, where the blue dotted line is the average value of each relative dry/wet period, the blue number denotes the shift year of each period, and the arrow indicates the drought trend after the 1970s. e The observed data and RHJA. f Multi-taper method (MTM) spectral analysis

Full size image
Table 1 Statistical parameter of δ18Otree series and correlation coefficients among them
Full size table

Climatic response

A correlation analysis (Fig. 2b) was conducted between the instrumental data and δ18Otree chronology during the overlapping period (1961–2013 AD). The δ18Otree chronology negatively correlated with precipitation in most months, positively correlated with temperature in all months and significantly negatively correlated with relative humidity in April and July-September (p < 0.05). Among all monthly combinations, the correlation between δ18Otree chronology and relative humidity in July-August was the highest (r = − 0.70, n = 53, p < 0.05), suggesting that the δ18Otree chronology for this study area mainly reflects the relative humidity signals from July to August.

July-August relative humidity reconstruction for Mt. Beiwudang

Based on the climatic response analysis, the relative humidity from July to August was selected for reconstruction. The δ18Otree chronology and calibration data (relative humidity of July to August for 1961–2013) scatter plot presents a linear distribution (Fig. 2c); hence, we selected a simple linear regression model to design the transfer function as follows:

$$ {\displaystyle \begin{array}{c}\kern1.25em {RH}_{\mathrm{JA}}=70.81-3.97\times {\updelta}^{18}{\mathrm{O}}_{\mathrm{tree}}\kern2.25em \\ {}\Big(n=53;r=0.702;{R}^2=0.49;{R^2}_{\mathrm{adj}}=0.48;\\ {}F=49.58,p<0.0001;D/W=2.09\Big)\end{array}} $$

where δ18Otree is the composite chronology of tree-ring cellulose stable oxygen isotope and RHJA is the reconstructed relative humidity from July to August (Fig. 2d). The RHJA explains 49% (R2) of the instrumental variance during the period of 1961–2013, and it reduces to 48% considering the loss of freedom (R2adj). An F-test (F = 49.58) with a confidence level of p < 0.0001 indicates that the linear relationship between the δ18Otree chronology and observed data is significant. A Durbin-Watson test (Durbin and Watson 1950) suggests that the residual of the reconstruction does not have significant autocorrelation. As shown in Fig. 2e, the RHJA have similar variability with the instrumental data during 1961–2013, and the correlation coefficient (r) is − 0.70 (n = 53). In addition, the correlation coefficient of the first-order difference series between the RHJA and instrumental data is − 0.67 (n = 52), which indicates that the correlation between RHJA and instrumental data does not simply trend similarity.

The stability and reliability of the transfer function were tested using the split calibration-verification model (Cook et al. 1999; Fritts 1991). The results of sign test and t test of the calibration period, verification period, and the entire period pass the significant level (Table 2), indicating that the RHJA exhibits high consistency with the instrumental data (Fritts 1976). Both the reduction of error (RE) and coefficient of efficiency (CE) are above zero, indicating robust and reliable reconstruction skills (Cook et al. 1999). In summary, the reconstructed model passed the split calibration-verification test, and the RHJA can be used to represent the local relative humidity from July to August. The RHJA is shown in Fig. 2d.

Table 2 Statistics of the split calibration-verification
Full size table

Discussions

Climatic response of δ18Otree

The negative correlation between δ18Otree and the hydroclimatic index is explicable by the fractionation of tree-ring isotopes (McCarroll and Loader 2004). Constituent elements of tree-ring cellulose mainly come from CO2, H2O, and O2, while δ18O of tree-ring cellulose only depends on H2O (Deniro and Epstein 1979). δ18O in the source water of trees varies with the isotopic signal of precipitation. The critical fractionation process of water is transpiration in leaves, where the sucrose forms by fixing the leaf water (McCarroll and Loader 2004). The intensity of transpiration is mainly influenced by relative humidity: a low relative humidity leads to strong transpiration, resulting in a high enrichment of 18O in leaf water and it is vice versa for high relative humidity. During the formation of cellulose, the oxygen of sugars may be re-exchanged with source water in xylem, and the proportion of the exchanged oxygen atoms determines the proportion of relative humidity and δ18Oprecipitation signals in the cellulose (Anderson et al. 2002). The proportion is difficult to predict (Barbour and Farquhar 2000), but it can be seen from Fig. 2b that the correlation of δ18Otree and relative humidity is significant in this study.

The partial correlation coefficients of the RHJA with temperature and precipitation from July to August are − 0.15 (1961–2013, p = 0.29) and 0.50 (1961–2013, p < 0.001), respectively. Hence, the RHJA in this region is mainly related to the amount of precipitation. In addition, the precipitation is highest in July-August (Fig. 1b), and July-August can be regarded as the summer rainy season over North China. The seasonal advance of the EASM and the monsoon rainfall belt is stepwise rather than continuous (Ding and Chan 2005; Wu and Wang 2001). In late July, the monsoon rainfall belt jumps northward to North China from Meiyu/Baiu/Changma regions, and after a stationary period, it retreats back in mid-August (Wang and Lin 2002). Furthermore, the RHJA is significantly correlated with various monsoon indices (Table 3). There are various definitions of the monsoon index, which focus on various aspects of the monsoon circulation (Wang et al. 2008) due to the complexity of the East Asian monsoon system. As a result, the RHJA is mainly influenced by precipitation in the summer rainy season and further by the EASM.

Table 3 Correlation coefficients between RHJA and climate indices
Full size table

Spatial representation of RH JA

A spatial correlation analysis of the instrumental calibration period (Fig. 1a) and reconstructed period (Fig. 3b) shows a significant positive correlation with the dry/wet variations in most parts of North China around the sampling point. The RHJA was significantly correlated with other proxy series near the sampling site in previous studies (Fig. 3a). However, the dryness/wetness index (DWI-37) has a significant correlation with the RHJA at high frequency but not at low frequency, which may be because historical documents are difficult to capture the interdecadal trend in climate change. Therefore, to represent the climate situation more accurately, the characteristics of the selected proxy data (Schneider et al. 2019) should be taken into account when investigating past climate change. All in all, the RHJA can represent the dry/wet variations in a large area covering most parts of North China.

Fig. 3
figure 3

a RHJA and other proxy series near the sampling point in previous studies: DWI-37 (38.75° N, 111.25° E) is the nearest point to the sampling site in the Yang et al. (2013) dataset; Mt.LY is the tree-ring δ18O data from Mt. Luya (38° 44′ N,111° 50′ E), where a smaller δ18O value represents wetter conditions and vice versa (Li et al. 2011); WLP is the regional precipitation series from previous-year July to current-year June reconstructed from 10 tree-ring width chronologies in the Loess Plateau Region (33.8° N–40.5° N,100° E–107° E) of the central-western edge of the Asian summer monsoon (Liu et al. 2019). The related coefficients of RHJA with other proxy series are marked in the figure. b Spatial correlation analysis of the RHJA using the gridded DWI dataset from 1765 to 2013 AD, only coloring the significant correlation region (p < 10%). c Drought (DWI > 3) distribution in four extensive drought years (1877, 1900, 1928, and 1929 AD). All grid values are based on the DWI, denoting drought, and flood as 1-very wet, 2-wet, 3-normal, 4-dry, and 5-very dry (Yang et al. 2013)

Full size image

Extensive drought events and the ENSO impact on RH JA

The mean value of the relative humidity reconstruction sequence is 70.71%, and the standard deviation is 3.45%. From an interannual perspective, we defined extremely wet years as having a relative humidity higher than 74.16% (mean + α) and extremely dry years as having a relative humidity below 67.26% (mean − α). In the reconstructed 249 years, extremely wet and dry years accounted for 14.46% (36 years) and 16.47% (41 years), respectively (Table 4). According to Table 4, the RHJA recorded many widely reported serious drought events, such as the Dingwu Extraordinary Drought from 1876 to 1878 (Cook et al. 2010; Zhang and Liang 2010; Hao et al. 2010), drought in the 1900 (Dong et al. 2015) and 1920s drought from 1922 to 1932(Liang et al. 2006). The droughts in 1877, 1900, and 1928–1929 are widely distributed, demonstrating by the DWI gridded dataset (Fig. 3c) and other proxy records (Fig. 3a), suggesting that they are extensive drought events.

Table 4 Extreme dry/wet years and relevant ENSO events*
Full size table

It is noteworthy that the extreme dry/wet years always co-occurred with ENSO events (Table 4). A correlation analysis shows that RHJA is significantly correlated with different ENSO indices (Table 3). In addition, a multi-taper method (MTM) spectral analysis (Fig. 2f) shows that the RHJA has 2.1–4.2 high-frequency quasi-cycles, which agrees with the quasi-cycle (2–7 years) of the ENSO. To conclude, ENSO plays an important role in the interannual variations of RHJA in North China. In fact, El Niño events always lead to negative precipitation anomalies in North China. It is mainly because the western Pacific subtropical high (WPSH) strengthening and shifting westward, and the weakening of Indian summer monsoon induces moisture inflow to North China (Zhang et al. 1999). The situation is vice versa for La Niña events.

The unprecedented weakening of the Asian summer monsoon since the late 1970s

Regarding the 11-year moving average of the RHJA, there is a significant drying trend after the 1970s (Fig. 1c). Four main transition points were detected using a sliding t test and Mann-Kendall test (Mann 1945), dividing the RHJA into 5 periods (Fig. 2d). The period of 1981–2013 represents an unprecedented drought period throughout the past two and a half centuries. Considering the significant correlation between RHJA and EASM, it can be inferred that the weakening of the EASM since the late 1970s was unprecedented in the past two and a half centuries. In fact, during the middle and late 1970s, abrupt changes in precipitation occurred in the East Asian monsoon region (Ding et al. 2008), which is related to the weakening of the EASM circulation after the end of the 1970s (Wang 2001). The leading factor for the weakening of the EASM since the late 1970s is the recent warming of the tropical ocean centered over the central and eastern Pacific Ocean (Zeng et al. 2007; Li et al. 2010).

In addition, it has been widely reported that the weakening of the EASM since the late 1970s has resulted in a “southern China flood and northern China drought” bipolar rainfall pattern in the monsoon region of China (Ding et al. 2008; Zhou et al. 2009a). A significant negative correlating pattern is present in the lower reaches of the Yangtze River Basin, southern Japan and southern Korea, according to the spatial correlation analysis results using the gridded rainfall dataset (Fig. 1a). The regional land rainfall sequence in this region (30° –37° N,115°–131° E) shows an increasing trend and a significantly negative correlation with RHJA after 1965 (Fig. 1c). This region corresponds to the Meiyu/Baiu/Changma rainfall belt, and the rainfall in this region has increased since the late 1970s (Ho et al. 2003). Following the warming of the tropical Pacific Ocean, the WPSH (one important component of the ESAM) has extended westward since the late 1970s (Zhou et al. 2009b), increasing summer rainfall over the Yangtze River valley (Chang et al. 2000; Gong and Ho 2002).

Moreover, the RHJA shows a significant correlation with the dry/wet variations across India in the spatial correlation analysis (Fig. 1a). This finding can be explained by the teleconnection between rainfall across India and North China, which has been widely reported for the instrumental period (Hu et al. 2005; Wu 2017). Similar to the ESAM, the Indian summer monsoon circulation has also weakened since the late 1970s along with recent warming of the tropical ocean (Wu 2005; Lin et al. 2017).

Conclusions

The tree-ring δ18O of Chinese pine from Mt. Beiwudang mainly reflects the relative humidity signal of July to August. The relative humidity of July and August from 1765 to 2013 was reconstructed using a linear regression model, and the reconstructed RHJA explains 49% of the instrumental variance during the period of 1961–2013. The RHJA can represent the dry/wet variations of the summer rainy season in a large area covering most parts of North China to some extent. There were 41 (36) extremely dry (wet) years over the past two and a half centuries and the extreme dry/wet years always co-occurred with ENSO events. In addition to the significant correlation coefficient between RHJA and ENSO index and the similar quasi-cycle with ENSO, it can be inferred that ENSO played an important role in the interannual variations of RHJA in North China. Based on the good relationship between the RHJA and EASM, the RHJA captures the recent weakening of the EASM since the late 1970s and suggests the recent weakening is unprecedented in the past two and a half centuries. At the same time, the precipitation over the Meiyu/Baiu/Changma rainfall belt region has increased and the precipitation over India has decreased due to the recent warming of the tropical ocean according to previous papers.

In general, the tree-ring cellulose δ18O data show a strong advantage in paleoclimatic reconstruction. However, single-site research is insufficient to testify to the regional interrelation and reflect climate change over the entire Asian summer monsoon (ASM) system, but it can be a clue to further regional research. It is meaningful to build a proxy network reflecting more detail of the interannual to interdecadal ASM climatic changes and even global changes.