INTRODUCTION
The variability of the Asian summer monsoon ranges from intraseasonal variation to thousands of years, driven by different mechanisms for different time scales (Wang et al., Reference Wang, Wang and Cheng2017). Investigating Asian summer monsoon variations across time scales is helpful for understanding current monsoon changes and predicting possible future changes. The onset of the summer monsoon is crucial to society and ecology, not only because it signifies the end of the dry season, but also due to its link to extreme weather. The monsoon in Asia experiences a rapid warming in the surface air mass, followed by episodic increases in humidity with occasional thunderstorms, and, subsequently, the steady monsoon rains. That last phase, with its combined increases in air temperature and humidity, forms a brief period of elevated convective instability that is prone to lightning, rainstorms, and associated flooding (Joseph et al., Reference Joseph, Eischeid and Pyle1994; Ullah and Shouting, Reference Ullah and Shouting2013; Wang et al., Reference Wang, Davies, Huang and Gillies2011; Siingh et al., Reference Siingh, Buchunde, Gandhi, Singh, Singh, Patil and Singh2015; Li et al., Reference Li, Wang, Gillies, Buckley, Yoon and Cho2019). Depending on its location across monsoonal Asia, the rainy season may be punctuated by one or two “break” phases with a notable decrease of rainfall; this is due to the evolution of tropical intraseasonal oscillations (Krishnan et al., Reference Krishnan, Ramesh, Samala, Meyers, Slingo and Fennessy2006; Kulkarni et al., Reference Kulkarni, Kripalani, Sabade and Rajeevan2011) and/or the passage of the North Pacific subtropical anticyclone (Wang et al., Reference Wang, Lin and Wu2016). Reconstructing the evolution of monsoonal precipitation has been challenging, despite the rich literature of using tree rings to reconstruct the seasonal monsoon rains.
Tree ring–based precipitation reconstructions are usually seasonal or annual, because annual tree growth or isotopes were affected by climate in the current or previous growing seasons (Cleaveland et al., Reference Cleaveland, Stahle, Therrell, Villanueva-Diaz and Burns2003; Shao et al., Reference Shao, Huang, Liu, Liang, Fang and Wang2005; Treydte et al., Reference Treydte, Schleser, Helle, Frank, Winiger, Haug and Esper2006; Yang et al., Reference Yang, Qin, Wang, He, Melvin, Osborn and B2014; Xu et al., Reference Xu, Pumijumnong, Nakatsuka, Sano and Guo2018). Occasionally, tree-ring widths and isotopes can be used for depicting the intraseasonal precipitation, for example, May–June precipitation reconstruction based on tree-ring width and oxygen isotope (Touchan et al., Reference Touchan, Akkemik, Hughes and Erkan2007; Xu et al., Reference Xu, Shi, Zhao, Nakatsuka, Sano, Shi and Guo2019a). Earlywood and latewood represent the different growth periods, and a careful parsing of the signals can extract intraseasonal precipitation (Griffin et al., Reference Griffin, Woodhouse, Meko, Stahle, Faulstich, Carrillo, Touchan, Castro and Leavitt2013; Young et al., Reference Young, Loader, McCarroll, Bale, Demmler, Miles and Whitney2015). However, few studies have obtained a robust reconstruction of monsoon precipitation during its onset or peak phase, and none have done a reconstruction of these in combination for the same location.
Tree ring cellulose oxygen isotope (δ18Oc) is mainly controlled by relative humidity and the precipitation oxygen isotope (Roden et al., Reference Roden, Lin and Ehleringer2000). Higher relative humidity may reduce evapotranspiration, resulting in less enriched leaf-water oxygen isotope and δ18Oc, and the precipitation oxygen isotope signal can be transferred into δ18Oc (Roden et al., Reference Roden, Lin and Ehleringer2000). Monsoon season relative humidity is usually related to precipitation amount in monsoonal Asia, so higher precipitation is associated with reduced δ18Oc. A negative relationship between precipitation amount and precipitation oxygen isotope ratio in the monsoon season has been identified in Asian summer monsoon areas (Dansgaard, Reference Dansgaard1964; Vuille et al., Reference Vuille, Bradley, Werner, Healy and Keimig2003; Kurita et al., Reference Kurita, Ichiyanagi, Matsumoto, Yamanaka and Ohata2009), and higher monsoon precipitation is correlated with depleted δ18Oc. Therefore, δ18Oc in monsoonal Asia is a demonstrated proxy for monsoon season precipitation (Grießinger et al., Reference Grießinger, Bräuning, Helle, Thomas and Schleser2011; Sano et al., Reference Sano, Tshering, Komori, Fujita, Xu and Nakatsuka2013; Xu et al., Reference Xu, Zheng, Nakatsuka and Sano2013, Reference Xu, Pumijumnong, Nakatsuka, Sano and Li2015).
It has been proposed that increasing the temporal resolution in δ18Oc can potentially reveal monthly precipitation; for example, Xu et al. (Reference Xu, Zheng, Nakatsuka, Sano, Li and Ge2016) found that δ18Oc deposited in July is highly correlated with July precipitation. Previous studies using intraseasonal δ18Oc from different tree species also showed how the seasonal evolution of monsoonal precipitation can be reconstructed based on relatively limited years of δ18Oc data (Managave et al., Reference Managave, Sheshshayee, Bhattacharyya and Ramesh2010a, Reference Managave, Sheshshayee, Borgaonkar and Ramesh2010b; Xu et al., Reference Xu, Zheng, Nakatsuka, Sano, Li and Ge2016; Zeng et al., Reference Zeng, Liu, Evans, Wang, An, Xu and Wu2016, Reference Zeng, Liu, Treydte, Evans, Wang, An, Sun, Xu, Wu and Zhang2017). However, the use of intra-annual δ18Oc analysis for the rigorous reconstruction of monsoon onset and peak precipitation has not been attempted.
In this study, we built seasonal δ18Oc time series for Abies georgei on the southeastern Tibetan Plateau by dividing each ring into four segments (S1, S2, S3, S4) and measuring the δ18Oc in each segment. We then used seasonal δ18Oc variations to reconstruct the onset and mature phase of summer monsoon precipitation.
MATERIALS AND METHODS
Sampling site and oxygen isotope analysis
We present the long-term seasonal δ18Oc time series from one A. georgei core (sample number: XC02a) in the southeastern Tibetan Plateau (Fig. 1). Abies georgei and Hippophae tibetana grow on the terminal moraines in Xincuo (30.09°N, 94.27°E, 3930 m above sea level), and we collected samples using 5 mm increment borers. The samples were dried, polished, and cross-dated.
(color online) Map of the sampling site in this study (left) and the sampled Abies georgei (right).
A modified plate method was used to extract α-cellulose (Xu et al., Reference Xu, Sano and Nakatsuka2011, Reference Xu, Zheng, Nakatsuka and Sano2013). We cut the core sample into 1-mm-thick wood plates for chemical treatment. We removed lignin, hemicellulose, lipids using a 17 wt% NaOH solution, an acidified NaClO2 solution, toluene and ethanol (1:1), respectively. After cellulose extraction, the cellulose plate was dried in an oven for 2 hours. The separation of each ring from the cellulose plate was performed under a binocular microscope. Each cellulose ring was divided into four parts with the same width, a method that follows Managave et al. (Reference Managave, Sheshshayee, Bhattacharyya and Ramesh2010a); these four components of the cellulose ring are referred to as S1, S2, S3, and S4, representing cellulose formed during each fraction of the growing season. To compare the climatic response of seasonal δ18Oc with annual δ18Oc, the annual δ18Oc chronology based on four trees (XC2, XC6, XC12, and XC14) that was produced by Xu et al. (Reference Xu, Zhu, Nakatsuka, Sano, Li, Shi, Liang and Guo2019b) was used in this study and is termed “annual(4).”
We wrapped 80–260 μg of each α-cellulose sample in silver foil for isotope measurements. Tree ring cellulose samples were converted into CO gas by a pyrolysis-type elemental analyzer (TC/EA) at 1375°C, and then oxygen isotope ratios (18O/16O) were determined using an isotope ratio mass spectrometer (Delta plus XL) at the Research Institute of Humanity and Nature, Kyoto, Japan. δ18Oc was calculated by comparison with Merck cellulose (laboratory working standard), which was inserted after every eighth tree ring cellulose sample during the measurements. The analytical uncertainty for repeated measurements of cellulose was approximately ± 0.18‰ (n = 76).
Climate of the study site
According to the pilot study by Xu et al. (Reference Xu, Zhu, Nakatsuka, Sano, Li, Shi, Liang and Guo2019b), which analyzed A. georgei δ18Oc at the same study site (Fig. 1), the annual δ18Oc is highly correlated to the local July–August precipitation. To investigate the relationship between seasonal δ18Oc and precipitation, regional daily precipitation data (29°N–32°N, 90°E–96°E) based on four stations from the Global Historical Climate Network data set (Menne et al., Reference Menne, Durre, Vose, Gleason and Houston2012) were employed. Three out of four meteorological stations ended their records in 1997; therefore, a common period of 1957–1997 was used. The rainy season near the study site lasts from June to September (based on the climatology of 1957–1997; Fig. 2), and this is also the growing season for A. georgei (Li et al., Reference Li, Liang, Gričar, Prislan, Rossi and Čufar2013; Xu et al., Reference Xu, Zhu, Nakatsuka, Sano, Li, Shi, Liang and Guo2019b).
Mean (black line) pentad precipitation covering 29°N–32°N, 90°E–96°E from the Global Historical Climate Network during the period of 1957–1997; pentad precipitation at high (red line) and low (green line) precipitation monsoon onset year. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
The 5 day mean (pentad) precipitation shown in Figure 2 indicates that the onset of the local monsoonal rainfall occurs around 6.6–87.2 mm, which corresponds to the 29th to 33rd pentad (May 21–June 10). To depict the monsoon onset processes in the study site, we examined the strong-onset years versus the weak-onset years by using 1 SD of the 29th–33rd pentad precipitation as a threshold. The composite precipitation during strong-onset years includes 1960, 1963, 1971, 1976, 1977, and 1997; while the weak-onset years are 1958, 1964, 1972, 1983, 1986, and 1995. The 5 day evolution of precipitation during these two groups of years is shown in Figure 2 for the strong and weak onsets. It can be seen that the strong-onset years are equivalent to an early onset, whereas weak-onset years reflect late onset; however, the subsequent amounts of rainfall do not seem to differ throughout the monsoon season. The onset (May 21–June 10) and mature phase (June, July, August) of regional summer monsoons are shown in Figure 2.
Based on strong- and weak-onset years, we plotted the composites of differential winds and precipitable water, which is calculated as the vertical integration of specific humidity for the strong-onset minus weak-onset years using the NCEP/NCAR Reanalysis daily variables (data set from Kalnay et al. Reference Kalnay, Kanamitsu and Kistler1996), in terms of the low-level (850 hPa) wind anomalies at 2.5° × 2.5° resolution. These composite maps were done for three periods of 10 days each spanning the onset season, shown in Figure 3. The low-level wind anomalies show a reversal from mid-May to late May, forming a cyclonic circulation over the northern Indian Ocean and the Indian subcontinent, signifying the formation of the monsoon low. By early June, a well-developed monsoonal circulation appears with anomalously strong westerly winds formed over the Arabian Sea into India, before curving northward into the study site on the southeastern Tibetan Plateau. Correspondingly, the precipitable water anomaly increases during late May through early June, supporting the monsoon onset precipitation. These circulation features, as shown in Figure 3, indicate that the variation of monsoon onset in the southeastern Tibetan Plateau is closely connected with the large-scale South Asian monsoon system.
(color online) Composites of anomalous 850 hPa winds (vectors) and precipitable water (shading) between strong and weak monsoon years (based on Fig. 2) during three periods. The study site is indicated by the black dot.
RESULTS AND DISCUSSION
Seasonal tree-ring oxygen isotope variations
When all phases of δ18Oc throughout each year during 1950–1980 (Fig. 4a) and 1980–2010 (Fig. 4b) are plotted, it becomes apparent that the S1–S4 segments of δ18Oc show a distinct seasonal cycle. The maximum δ18Oc occurs in the beginning of the growing season and then decreases throughout the growing season (Fig. 4c). The mean values for S1, S2, S3, and S4 are 27.60‰, 24.21‰, 21.59‰, and 20.04‰, respectively. A similar seasonal δ18Oc pattern that followed seasonal precipitation oxygen isotope and relative humidity changes was also found in Abies forrestii in the southeastern Tibetan Plateau (Gao et al., Reference Gao, Masson-Delmotte, Risi, He and Yao2013; Zeng et al., Reference Zeng, Liu, Evans, Wang, An, Xu and Wu2016, Reference Zeng, Liu, Treydte, Evans, Wang, An, Sun, Xu, Wu and Zhang2017). The correlation coefficients between δ18Oc values for S1, S2, S3, S4 and the annual δ18Oc are 0.77, 0.87, 0.90, and 0.70, respectively, during 1950–2010. The correlation between S4 (δ18Oc) in the previous year and S1 in the current year is −0.03 (n = 58), suggesting that δ18Oc in the previous year does not influence δ18Oc in the current year (Xu et al., Reference Xu, Pumijumnong, Nakatsuka, Sano and Guo2018). In addition, the correlation between June-July-August (JJA) precipitation in the previous year and δ18Oc of S1 in the current year is not significant (r = −0.03, n = 41) during the period of 1957−1997, again indicating that monsoon precipitation in the previous year does not affect δ18Oc in the current year.
Intra-annual δ18Oc time series from Abies georgei during the period of 1951–2009.
Climatic response of different parts of tree-ring oxygen isotope
It is known that δ18Oc formed in different segments of a tree ring records precipitation oxygen isotope and relative humidity at different times of the year (Treydte et al., Reference Treydte, Boda, Graf Pannatier, Fonti, Frank, Ullrich, Saurer, Siegwolf, Battipaglia and Werner2014; Xu et al., Reference Xu, Zheng, Nakatsuka, Sano, Li and Ge2016; Zeng et al., Reference Zeng, Liu, Evans, Wang, An, Xu and Wu2016, Reference Zeng, Liu, Treydte, Evans, Wang, An, Sun, Xu, Wu and Zhang2017). In our study area, the precipitation oxygen isotope is negatively correlated with rainfall and relative humidity during the monsoon season (Gao et al., Reference Gao, Masson-Delmotte, Risi, He and Yao2013), with δ18Oc depleted when rainfall is high and increased when rainfall is low.
Next, we computed the correlation between S1–S4 δ18Oc and pentad precipitation from May to October (Fig. 5). We found that δ18Oc of S1 significantly correlates with the late May to mid-June precipitation (29th–33rd pentad) that corresponds to the regional monsoon onset (Fig. 2). The correlation coefficient between S1 (δ18Oc) and the onset precipitation is −0.69 (n = 41, P < 0.001). Early monsoon onset is usually associated with decreased precipitation oxygen isotope values and high relative humidity (Tian et al., Reference Tian, Masson-Delmotte, Stievenard, Yao and Jouzel2001; Yang et al., Reference Yang, Yao, Yang, Xu, He and Qu2012), which would further deplete δ18Oc of S1 (i.e., when cellulose is formed). Therefore, δ18Oc of S1 is a promising proxy for the onset of the regional monsoon rains both statistically and physically.
Correlations between S1–S4 δ18Oc and pentad precipitation from May to September during the period of 1957–1997; the dashed line indicates the 95% confidence level.
The Indian summer monsoon (ISM) brings moisture into the southeastern Tibetan Plateau. Precipitation oxygen isotope as one controlling factor of δ18Oc on the southeastern Tibetan Plateau becomes depleted when the ISM onset is early (Tian et al., Reference Tian, Masson-Delmotte, Stievenard, Yao and Jouzel2001; Gao et al., Reference Gao, Masson-Delmotte, Risi, He and Yao2013). The relationship between ISM onset and δ18Oc was found in several individual years (Zeng et al., Reference Zeng, Liu, Evans, Wang, An, Xu and Wu2016). To examine the precipitation linkage with the ISM, we compared the relationship between the ISM onset at Kelera (Pai and Nair, Reference Pai and Nair2009) and δ18Oc from different segments during the period of 1971–2007 and found that the ISM onset at Kelera is negatively correlated with δ18Oc of S1 (r = −0.36, P < 0.01). These results reveal that δ18Oc of S1 records the summer monsoon onset.
δ18Oc of S2 shows negative correlations with June and mid-July precipitation, and the June signal is stronger (Fig. 5b). δ18Oc of S3 is correlated with July and August precipitation, and July precipitation has a stronger influence on δ18Oc of S3 (Fig. 5c). δ18Oc of S4 has the strongest correlation with August precipitation. Such negative correlations between δ18Oc and precipitation are mainly derived from negative correlations between precipitation amount and precipitation oxygen isotope, as is evident for monsoon season precipitation on the southeastern Tibetan Plateau (Tian et al., Reference Tian, Masson-Delmotte, Stievenard, Yao and Jouzel2001; Yang et al., Reference Yang, Yao, Yang, Xu, He and Qu2012; Gao et al., Reference Gao, Masson-Delmotte, Risi, He and Yao2013).
In summary, regional monsoon onset and mature-phase precipitation (as outlined in Fig. 2) have the highest correlations with δ18Oc of S1 and averaged δ18Oc of S2, S3, and S4, respectively (Table 1). In addition, precipitation in June, July, and August showed the strongest correlation with δ18Oc of S2, S3, and S4, respectively (Table 1). Although the seasonal δ18Oc data are based on one tree, the monsoon onset/mature-phase precipitation and monthly precipitation signal from seasonal δ18Oc using one tree is stronger than from annual δ18Oc using four trees (Table 1). Compared with annual δ18Oc, seasonal δ18Oc variations provide not only more detailed climatic information but also a stronger climatic signal.
Correlations between precipitation and tree-ring δ18O from parts S1 to S4 and annual(4).
a S2–4 indicates average of S2, S3, and S4.
b Annual(4): the annual δ18Oc chronology based on four trees (XC2, XC6, XC12, and XC14) that was produced by Xu et al. (Reference Xu, Zhu, Nakatsuka, Sano, Li, Shi, Liang and Guo2019b).
* P < 0.01.
bold type indicates the highest correlations between precipitation and tree-ring δ18O from parts S1 to S4 and annual(4).
Reconstruction of precipitation for monsoon onset and mature phase
Given the aforementioned analysis, regional monsoon onset can be reconstructed based on a linear relationship with δ18Oc of S1. We examined the mature phase of the monsoon precipitation (JJA) by reconstructing it using the combination of S2, S3, and S4. We constructed a set of linear regression models between monsoon onset and mature-phase precipitation and δ18Oc from different segments:
Here, PREonset and PREmature represent regional monsoon onset and mature-phase precipitation, respectively. S1 to S4 represent δ18Oc values from segments S1 to S4. Split calibration–verification tests for monsoon onset and mature-phase precipitation reconstruction are shown in Tables 2 and 3, respectively. For both subperiods of monsoon onset and mature-phase precipitation, most rigorous statistics, such as reduction of error (RE) and coefficient of efficiency (CE) are positive, suggesting the reconstruction is robust (Cook et al., Reference Cook, Meko, Stahle and Cleaveland1999). The observed and reconstructed monsoon onset and mature-phase precipitation showed consistent variations at interannual and decadal scales (Fig. 6). The long-term seasonal δ18Oc data set covering several decades showed the fidelity of monsoon onset and mature-phase precipitation reconstruction based on seasonal δ18Oc.
Comparison of the reconstructed JJA precipitation (red line) and precipitation data (black line) from Global Historical Climate Network during the period of 1957–1997. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Calibration and verification statistics for monsoon onset precipitation reconstruction.
a RE, reduction of error.
b CE, coefficient of efficiency.
Calibration and verification statistics for monsoon mature-phase (JJA) precipitation reconstruction.
a RE, reduction of error.
b CE, coefficient of efficiency.
As a further validation, we used the reconstructed onset precipitation (as in Fig. 6a) to produce a regression map of low-level winds and precipitable water during the May 21 through June 10 period, corresponding to Figure 3. The regressed low-level atmospheric circulations with the onset precipitation reconstruction (Fig. 7) reveals a pattern similar to the strong-onset minus weak-onset composite in early June (Fig. 3), indicating increased monsoonal westerly winds and the deepened monsoonal trough over the India subcontinent. This is another line of evidence that the tree ring δ18Oc-recorded onset precipitation signal in the southeastern Tibetan Plateau is closely associated with the large-scale South Asian monsoon variation.
(color online) Anomalous winds at 850 hPa and precipitable water during May 21–June 10 regressed on the reconstructed onset precipitation (normalized) from 1957 to 1997 based on the NCEP/NCAR Reanalysis data set (Kalnay et al., Reference Kalnay, Kanamitsu and Kistler1996). Anomalous wind speed greater than 3 m/s and precipitable water greater than 4.5 kg/m2 are significant (P < 0.05). Study site is indicated as black dot.
Seasonal δ18Oc from one tree showed great potential for robust monsoon onset reconstruction, which cannot be done using annual δ18Oc from four trees. Moreover, monthly precipitation reconstruction has been considered challenging when it comes to the analysis of annual tree-ring proxies; however, we found applicability based on the negative correlation between August precipitation and δ18Oc of S4 (Table 1). It should be noted that building a δ18Oc record based on a single tree should be treated cautiously. Two or more trees should be included for cross-checking purposes. Increased sampling resolution per year (e.g., six or eight segments per year) may allow us to extract a stronger and more accurate monthly precipitation signal. However, the exponential increase in the cost of analysis as the sampling resolution increases can be prohibitive.
Considering that tree-ring oxygen isotopes record monsoon season moisture variations in monsoonal Asia, seasonal δ18Oc analysis in monsoonal Asia will shed more light on the Asian summer monsoon variability for the long term. For example, seasonal δ18Oc-based monsoon onset and mature-phase precipitation reconstruction at different locations in monsoonal Asia can provide a spatial variation of the different phases of monsoon precipitation during the past several hundred years. Millennial seasonal δ18Oc records in monsoonal Asia can potentially reveal external forcing (such as solar activity, greenhouse gases) and internal forcing (such as ENSO) influences on monsoon precipitation. Seasonal δ18Oc chronologies in monsoonal Asia in a past warm period (such as the mid-Holocene) can be useful to understand possible precipitation changes in the context of global warming.
CONCLUSIONS
In this study, we took a novel step, dividing the seasonal tree-ring oxygen isotope in one tree ring from the southeastern Tibetan Plateau into four segments (S1 to S4) and analyzing the segments for the period of 1951–2009. We found that δ18Oc of S1 has the strongest correlations with the regional monsoon onset precipitation, while δ18Oc values for S2, S3, and S4 correlate strongly with the June, July, and August precipitation, respectively. The linear regression models between δ18Oc of S1 and monsoon onset precipitation and δ18Oc S2–S4 and monsoon mature-phase precipitation pass all statistical tests commonly used for calibration and verification. The results suggest that using seasonal δ18Oc from one tree alone can achieve robust monsoon onset and monthly precipitation reconstruction, a task that is difficult using only annual δ18Oc and tree-ring width variations. Future studies should consider using seasonal δ18Oc collected at different sites throughout monsoonal Asia to reconstruct the distinct monsoon life cycle as revealed in daily precipitation for a better understanding of variation of the South Asia summer monsoon in the pre-instrumental period.
ACKNOWLEDGMENTS
This study was supported by the National Natural Science Foundation of China, grants 41888101, 41630529, 41672179, 41671193, 41430531, and 41690114; the National Key R&D Program of China, grant 2017YFE0112800; the Chinese Academy of Sciences (CAS) Pioneer Hundred Talents Program; and the Strategic Priority Research Program of the Chinese Academy of Sciences, grants XDB26020000 and XDA13010106. SYW is supported by a U.S. DOE HyperFACET grant. We deeply appreciate helpful comments from the editors and two anonymous reviewers that improved the article.
