1 Introduction

Southwest China is situated to the southeast of the Tibetan Plateau, and is influenced by the Indian Summer Monsoon (ISM) and the southern branch of the westerlies separated by the plateau (Broccoli and Manabe 1992; Wang et al. 2014). Under the current global warming, most of China has experienced warming since the 1950s (CPNARCCC 2015). However, a cooling trend has been observed in a few regions of southwestern China (CPNARCCC 2015). Previous studies suggested that this cooling trend may have been caused by an increase in aerosols due to intensified industrial activities (Zhou et al. 1996) or sea surface temperature (SST) anomalies in the Pacific Ocean (Tsai et al. 2014). One prominent region with a cooling trend in southwestern China is the Sichuan Basin which has an area of ~ 260 km2 and a population of ~ 90 million. However, most of the instrumental records in the Sichuan Basin started in the 1950s, which limits our ability to comprehend the temperature regimes in historical periods prior to the industrial era.

Long-term temperature records inferred from natural proxies provide long time perspectives to evaluate the current temperature variations (Hartmann et al. 2013). They may help to identify effects of external forcings or internal feedbacks on timescales longer than the instrumental records (Ruddiman 2014). Besides that, long climate records allow for the detection of climate regimes with high statistical confidence. Tree rings are the most widely used proxy for temperature reconstruction over the past millennium due to their long duration, annual resolution, and high temperature sensitivity at many locations (Hughes et al. 2011).

Most tree-ring based climate reconstructions in China have been conducted in arid and/or cold regions (Gou et al. 2014; Liang et al. 2008; Liu et al. 2009, 2011; Shao et al. 2010; Yang et al. 2014; Yuan et al. 2003; Zhang et al. 2003). Tree-ring studies in hot and/or humid regions, such as the Sichuan Basin, are highly needed because they may provide key information on oceanic modes (Buckley et al. 2010; Chen et al. 2012; Duan and Zhang 2014; Fang et al. 2012; Liu et al. 2017). One major challenge for tree-ring studies in hot and humid regions is that it is often more difficult to extract climate signals from tree rings as tree growth tends to be influenced by multiple factors in such regions. This study introduces an alternative method to efficiently extract large-scale climate signals from tree rings.

Mount Emei (29.50N, 103.48E) is an UNESCO World Heritage site located at the southwestern part of the Sichuan Basin, and steeply descends from 3099 m a.s.l. down to around 400 m a.s.l. To date, there is only one tree-ring chronology in the Sichuan Basin at Mount Emei that spans 116 years (Fang et al. 2010). In this study, we sampled temperature sensitive, old-growth Abies fabri trees at the cliff areas. The resulting tree-ring chronology starts in 1590 with its reliable period since 1758. This is by far the longest tree-ring chronology in the Sichuan Basin. Our aims are to: (1) present a new method on tree-ring chronology development; (2) develop the longest tree-ring chronology in the Sichuan Basin; and (3) study the temperature regimes based on the chronology in relation to large-scale oceanic modes.

2 Data and methods

2.1 Study area and tree-ring samples

The tree-ring samples were taken from Mount Emei in the Sichuan Basin, southwest China. Mount Emei located in the transitional area between the low-lying Sichuan Basin and the southeastern Tibetan Plateau, where the elevations decrease from ~ 4000 m on the Tibetan Plateau to ~ 500 m in the Sichuan Basin (Fig. 1). Although far from the ocean, the Sichuan Basin is characterized by a maritime climate due to its small ranges of annual and diurnal temperatures (Gorczynski 1922). The mean annual total precipitation in the Sichuan basin is over 1000 mm and the highest precipitation is found to the west with over 1500 mm, where Mount Emei is located (Fig. 1). Over 70% of the annual precipitation occurs from June to October. Moreover, Mount Emei is the foggiest place and has the highest relative humidity in China (Figure S1).

Fig. 1
figure 1

Location of the sampling sites and meteorological station at Mount Emei and the previously published tree-ring chronologies near the study region. The chronologies that have significant (p < 0.05) correlations with the Mount Emei chronology are indicated by solid triangles. Detailed information on the tree-ring chronologies are shown in Table S1

Full size image

According to the closest meteorological station from our sampling sites at Mount Emei station, the annual mean temperature and the annual total precipitation are 16 °C and 1345 mm at the foot of the Mount Emei, and 3 °C and 1841 mm at the peak. Accordingly, natural vegetation varies from evergreen broadleaf forests from the foot to coniferous forests above ~ 2500 m (Zhu et al. 1997). The studied species of Abies fabri is an endangered, endemic coniferous species in Mount Emei (Guan 1983). Abies fabri trees are mixed with other tree species from 2200 to 2600 m, and form pure forests above 2600 m (Guan 1983). This species is most tolerant to shade and moisture among the 22 fir species, allowing it to adapt to the wettest conditions in the Sichuan Basin (Guan 1983). This species largely grows within a cloud belt on Mount Emei ranging from about ~ 2000 to ~ 2500 m with frequent rains below the belt but sunny above (Zhu et al. 1997).

A. fabri can seldom grow over 200 years old (Guan 1983). In order to develop long chronologies, we collected samples from sites near cliffs or with very shallow soil, where trees may be over 200 years. We sampled 30 cores of 14 trees, 33 cores of 12 trees and 19 cores of 11 trees from three cliff sites at elevations of 2380, 2430, and 2480 m, respectively. These sites are near the Leidongping temple. In addition, 25 cores of 12 trees and 30 cores of 13 trees were taken from cliff areas from the other two sites near Xixiangchi Temple with elevations of 2070 and 2170 m, respectively. In total, we collected 137 cores of 62 trees from Mount Emei.

The tree-ring samples were mounted, air dried, and polished for cross-dating and ring-width measurement (Stokes and Smiley 1968). The visually cross-dated samples were measured to 0.001 mm precision. The program COFECHA was used to check the quality of the cross-dating (Holmes 1983). We discarded poor quality samples and retained 101 crossdated cores of 51 trees for further studies. The mean segment length of the crossdated series is 168 years and the longest core is 423 years. The mean correlation among all series (i.e. intercorrelation) is 0.49, which is similar to the intercorrelations (0.50–0.54) for series at five individual sites. As these sites are very close to each other with similar growing environment, we merged the five sites to build one composite chronology to enhance the climate signal encoded in tree rings.

2.2 Development of tree-ring chronologies

Age-related growth trends of the cross-dated series were removed to build a dimensionless tree-ring index series. This is achieved by dividing raw measurements by the growth curves of straight lines or negative exponential curves (Cook 1985). For those tree-ring series with strong disturbances, a smoothed cubic spline with a 50% frequency cutoff of 2/3 of the series length was employed to represent the age-related growth trends and disturbances. The standard (STD) chronology was developed by averaging the dimensionless tree-ring indices (Cook 1985). This is to cancel out the tree-specific disturbances and enhance regional climate signal. The recommend threshold of the expressed population signal (EPS) statistic (Wigley et al. 1984) was used to determine the reliable portion of the chronology with sufficient replication.

The “signal-free” method (Melvin and Briffa 2008) based chronology (SFC) was applied for a comparison. Traditional curve fitting methods have difficulties in distinguishing the medium frequency climate signals and age-related growth trends. Therefore it is possible that the fitted age-related trend contains climate signals (Fang et al. 2012; Melvin and Briffa 2008). The signal-free method was designed to create the fitted age-related growth curve free of climate signals by reiteratively dividing the raw measurements by the chronology developed in the last iteration (Melvin and Briffa 2008). This method was proved to better preserve medium-frequency climate-related variations in the final tree-ring chronology (Melvin and Briffa 2008).

Additionally, a chronology was generated by using the frequency based method (Fang et al. 2017). Disturbances of tree rings at interdecadal (10–100 years) scales are more difficult to be canceled out relative to those at interannual scales. Therefore, this method firstly separates the tree-ring indices into two groups based on interannual (1–10 years) and longer (> 10 years) timescales. All the interannual variations of indices are then averaged, while only the interdecadal variations of the indices with coherent variations are retained. More strict control on the development of the interdecadal component of the chronology allows to better retain the coherent interdecadal variations. The frequency-based chronology (FBC) is the mean of the interannual and interdecadal tree-ring indices. The filtering analysis used in this study is the Gaussian filtering method, which has been widely applied in paleoclimate studies (Mann et al. 2009).

2.3 The large-scale tree-ring chronology (LSC)

The first step is to generate the tree-ring chronologies at the study site, herein Mount Emei, and in regions where climate is potentially linked to the study site (Fig. 2). In this study, we selected tree-ring chronologies on the Tibetan Plateau and the surrounding which are located to the west of our site (Fig. 1). These chronologies are located along the trajectory of the westerlies and the ISM that should have a strong influence on the climate of our study site. Although climate patterns to the east of Mount Emei can also influence climate of the study area, there are few tree-ring chronologies available (Cook et al. 2013).

Fig. 2
figure 2

Flowing diagram showing the procedures to generate the large-scale tree-ring chronology (LSC)

Full size image

For the second step, we screened the available chronologies to only include those have significant correlations with the chronology at the study site (Fig. 2). This is to make sure that the chosen tree-ring chronologies containing large-scale climate signals that are related to the study site. We then calculated the principal components (PCs) of the chosen tree-ring chronologies. The PCs are orthogonal and independent of each other, which can mitigate the issue associated with correlations among tree-ring chronologies. A regressed tree-ring chronology from the PCs is then calculated from these PCs.

The third step is to produce the LSC that contains large-scale climate patterns related to the study site, which is achieved by averaging the local chronology of the study site and the regressed chronology based on remote sites (Fig. 2).

It should be noted that local climate signals in the final chronology may be biased as the LSC method also includes information from remote regions. To minimize this risk, we suggest three prerequisites to be met prior to the use of this method. First, the chronologies outside the sampling site should not only be significantly correlated with the chronology of the sampling site but also should be in regions climatologically related to the study site. Second, the LSC should have more significant correlations with the local climate near the sampling site than the chronologies developed using traditional methods. This is to justify that the LSC enhances the strength of local climate signals encoded in tree rings of the study site. Third, the LSC should not have correlations with climate in regions where no significant correlations were found between the chronology of the study site and climate. This is to make sure that the LSC does not introduce biased climate signals from distant regions.

2.4 Instrumental records and other climate data

The tree-ring data were compared with the instrumental temperature and precipitation records from a meteorological station (29.52oN, 103.33oE, 3049 ma.s.l.) near the top of Mount Emei. We also used a 0.5° × 0.5° gridded climate data from the Climate Research Unit dataset (CRU TS3.10) spanning from 1901 to 2015 (Harris et al. 2014). The climate data in the Sichuan Basin were derived from the CRU TS3.10, which are longer than those recorded at Mount Emei. We also used the SST data from the HadISST1 dataset from 1870 onwards with a 0.5° × 0.5° gridding system (Rayner et al. 2003).

3 Results

3.1 Tree-ring chronologies

The tree-ring chronology at Mount Emei spans from 1590 to 2012 and its reliable period is from 1758 to 2014 when the chronology includes with at least 14 series (Fig. 3). There are 30 out of 166 chronologies which have significant (p < 0.01) correlations with the STD chronology at Mount Emei, mostly (27 out of 30) located along the southern rim of the Tibetan Plateau (Fig. 1). These chronologies and the STD of our study site were used to produce the LSC. One prominent characteristic of the tree-ring chronologies is a sharp decline during the periods of 1808–1823 and 1964–1976, during which the low-passed (> 10 years) values are below one standard deviation from their mean.

Fig. 3
figure 3

Comparisons of the standard chronology (STD) (Cook 1985), the “signal-free” methods based chronology (SFC) (Melvin and Briffa 2008), the frequency based chronology (FBC) (Fang et al. 2017), and the large-scale tree-ring chronology (LSC) during their reliable period of 1750–2012, with the statistics of subsample signal strength (SSS) over 0.85

Full size image

The four versions (STD, SFC, FBC and LSC) of the chronologies have similar variations and are highly correlated (Table 1). Among these, the LSC chronology has the largest differences as indicated by the lowest correlations with the other three (Table 1). The low correlations between the LSC chronology and the other three are mainly found for the high-passed (< 10 years) data (Table 1), but not on the long timescales (> 10 years). The coherence of low-frequency variations can also be viewed from the visual comparisons shown in Fig. 3.

Table 1 Correlations among the four versions of tree-ring chronologies during their common period 1758–2005
Full size table

3.2 Climate–growth relationships

In general, the four versions of tree-ring chronologies have positive correlations with temperature and precipitation in Mount Emei (Fig. 4). The correlations between the LSC and temperature at Mount Emei are higher than the other three versions of the chronologies (Fig. 4a). Significant (p < 0.01) correlations are observed between the LSC and temperatures for the previous November (r = 0.37) and current June (r = 0.37). Significant (p < 0.01), positive correlations are found between the STD, SFC, and FBC chronologies and precipitation in Mount Emei for current February (r = 0.37, 0.38, and 0.40, respectively) (Fig. 4b).

Fig. 4
figure 4

Correlations of the four versions of tree-ring chronologies with monthly mean temperature and precipitation from Mount Emei and the Sichuan Basin (29–32°N; 103–107°E) during their common periods (1951–2005 and 1901–2005) from the previous March to the current November, as well as the mean temperature and precipitation from current February to July. Note that the threshold values for the significant (p < 0.01) correlations with climate in Mount Emei and the Sichuan Basin are different due to the different length of the common periods. The four versions of chronologies are: STD standard tree-ring chronology, SFC “signal-free” methods based chronology, FBC frequency based chronology, LSC large-scale tree-ring chronology

Full size image

The correlations with climate in the Sichuan Basin (29-32oN; 103-107oE) for different versions of the chronologies are similar (Fig. 4c, d). Therefore, we only report the correlations with the LSC chronology. Significant (p < 0.01), positive correlations with temperature in the Sichuan Basin are found for the previous March (r = 0.26), April (r = 0.36), June (r = 0.28), and July (r = 0.29), as well as current April (r = 0.27), June (r = 0.37), and July (r = 0.36) (Fig. 4c). The correlations with precipitation in the Sichuan Basin are non-significant (Fig. 4d).

Correlation with the mean temperature of their growing season from current February to July in the Sichuan Basin for the period of 1901–2005 is 0.46, which is even higher than the correlations with temperature in Mount Emei for the period of 1951–2005. Strong correlation with the Sichuan Basin temperature indicates that tree rings on Mount Emei can largely represent the Sichuan Basin temperature. This may be because the sampling site is located near the trajectory of the air current from southeastern Tibetan Plateau to the Sichuan Basin, and that the Sichuan Basin is flat with relatively even climate conditions. We found the correlations with temperature is unstable through time with peak correlation in the period from 1901 to 1950 (r = 0.70) (Fig. 5a). This indicates that the lower correlations with instrumental data on Mount Emei relative to the Sichuan Basin are probably because the former correlations were calculated after the 1950s when the linkages of tree growth to temperature have weakened. Spatial correlation patterns indicate that our tree rings have high correlations with mean temperature from current February to July in the Sichuan Basin and nearby regions prior to year 1950 (Fig. 5). Correlations between tree-ring chronologies and precipitation near our study region are non-significant and are therefore not shown.

Fig. 5
figure 5

The a running correlations between the LSC chronology and the mean temperature of the Scihuan Basin from current February to July derived from the CRU TS3.10 dataset starting in 1901 based on a 51-year window, and b map of the correlations between the LSC and the mean temperature from current February to July. Non-significant (p > 0.1) correlations are masked out and the square indicates the location of the Sichuan Basin (29–32°N; 103–107°E)

Full size image

Apart from the high correlations with regional climates, the tree-ring chronologies are highly correlated with the SSTs of the northern Atlantic Ocean and the northeastern boundaries of the Pacific Ocean (Fig. 6). The correlations with northern Atlantic SST are particularly high in its northern and tropical areas. Moderately high correlations are observed with SSTs in the eastern rim of the northern Pacific Ocean. This correlation pattern is very similar to the correlation between the Atlantic Multi-decadal Oscillation (AMO) and the global SSTs (Gray et al. 2004; Schlesinger and Ramankutty 1994; Wang et al. 2017), suggestive of the influence of the AMO on tree growth in our study area. We therefore compared the LSC chronology with the instrumental and reconstructed AMO indices. The tree-ring chronology matches more closely with the instrumental AMO than the temperature of northern hemisphere (NH), particularly on interdecadal scales (Fig. 7a). To further investigate their linkages over the historical periods, we compared our tree-ring data with two AMO reconstructions that were derived from completely independent proxy records and have close matches (Gray et al. 2004; Wang et al. 2017), particularly on interdecadal scales (Fig. 7b). Close matches are observed between the AMO reconstructions and the LSC, as well as the NH temperature reconstruction on interdecadal scales (Fig. 7). Taken together, tree rings in Mount Emei are mainly controlled by temperature of the growing season from current February to July, but their relationships weakened in modern periods after ~ 1950. Our tree rings match well the AMO reconstructions over the whole reconstruction period on interdecadal scales, except for the late nineteenth century and after ~ 1980.

Fig. 6
figure 6

Map of correlations between the four versions of tree-ring chronologies and mean sea surface temperature (SST) of current February and July from the HadISST1 dataset during the common period from 1870 to 2005. Non-significant (p > 0.1) correlations are masked out. The square indicates the location of the Sichuan Basin (29–32°N; 103–107°E). The four versions of chronologies are the same as those in Fig. 3

Full size image
Fig. 7
figure 7

Comparisons among the interdecadal variations (> 20 years) of the: a instrumental temperatures in the Sichuan Basin, the Northern Hemisphere (NH), and the Atlantic Multi-decadal Oscillation (AMO); b reconstructed AMO indices by Gray et al. (2004) and Wang et al. (2017); c large-scale tree-ring chronology (LSC); and d ensemble mean of the 16 temperature reconstructions for the Northern Hemisphere (NH) derived from Hartmann et al. (2013). All the data are normalized for direct comparisons

Full size image

4 Discussion

4.1 Applications of the LSC

Tree growth can be viewed as a function of age-related growth trends, climate signals, and ecological disturbances (Hughes et al. 2011). The disturbances at tree level can be reduced by averaging a large number of tree-ring cores from a site, but there may site-level non-climatic variations (Fang et al. 2017). Growth disturbances in hot and/or humid regions may be more intense, which impedes the extraction of climate signals from tree rings. In addition, local climate patterns may vary from site to site. The site-specific non-climate variations are expected to be dampened when tree rings from different sites controlled by similar climate patterns are synthesized.

In our study, the LSC matches more closely with the local temperature at Mount Emei than the other three versions of the chronologies. This suggests that the LSC indeed enhanced local climate signals. The LSC chronology has high correlations with the Sichuan Basin temperature and the SSTs of the northern Atlantic Ocean. These correlations are not biased due to the inclusion of distant chronologies because the other three versions of the chronologies show equally high correlations in these regions.

The method for the generation of a LSC chronology is similar to the principal components (PC) based regional climate reconstruction, which extracts the dominant PCs of tree-ring chronologies in a region. The difference is that the LSC aims to extract climate signals that influence the target sampling regions, while the PC based reconstruction aims to reconstruct the regional climate history. The LSC and the other three versions of chronologies are highly coherent on interdecadal timescales and their differences are mainly found on interannual timescales. Large-scale climate patterns are more likely to occur on longer timescales and the differences on interannual scales are more likely due to disturbances (Fang et al. 2018). The LSC chronology has similarities with the so-called atmospheric circulation tree-ring index (ACTI) (Schultz and Neuwirth 2012; Seim et al. 2017), as both aim to identify the relationship between large-scale climate and tree rings. One difference is that the ACTI is calculated from atmospheric modes, while the LSC is based solely on tree-ring data.

Our study indicates that the LSC may help to improve the extraction of climate signals from tree rings in “satisfying” regions. In addition, we suggest examining topographic features to identify stressed sites, such as cliff areas and eroded sides, for fieldwork conducted in regions with “satisfying” macroclimate conditions. This is because trees under such micro-environmental conditions may be older and more sensitive to climate.

4.2 AMO and temperature in southwestern China

Close matches on interdecadal scales have been observed between the AMO and temperature reconstructions in nearby regions of southwestern China (Shi et al. 2017; Wang et al. 2014, 2015). Our study shows that the linkage between the AMO and the Sichuan Basin temperature is evident on interdecadal scales, and that is part of the large-scale teleconnection with the temperature in southwestern China. Indeed, the highest correlations between the AMO and temperature in central and eastern Asia are mostly found in southwestern China, including the Sichuan Basin (Figure S2). It is readily understood that their linkages mainly occur on interdecadal scales as the variability of the AMO is more pronounced on interdecadal scales (Schlesinger and Ramankutty 1994; Wang et al. 2017). We also found linkages between our tree rings and SSTs of the northeastern rim of Pacific Ocean, which was also suggested to be linked to the AMO (d’Orgeville and Peltier 2007). Our results differ from a previous study that suggested the control of Pacific SSTs on Sichuan Basin temperatures (Tsai et al. 2014). We consider that the linkages between the Sichuan Basin temperature and the AMO are more robust because the close match existed over most of the past 250 years.

As an oceanic mode, the AMO can influence the Asian temperature via atmospheric bridges propagating from the Atlantic Ocean. The atmospheric bridges include the westerlies and the wave train that comes with it which can modulate surface air pressures and temperatures (Li and Bates 2007; Si and Ding 2016). The westerlies are separated by the Tibetan Plateau, and their southern branch reaches southwest China, which plays an important role in modulating regional winter climate. This cyclone-like pattern is particularly obvious in the Sichuan Basin (Yanai and Wu 2006). We thus infer that our tree-ring chronology in the Sichuan Basin may be more closely related to the variations of the southern branch of the westerlies. In summer, climate in the North Atlantic Ocean and the ISM region are two centers of action in the atmospheric teleconnection across the NH (Ding and Wang 2005; Si and Ding 2016). Climate in southwestern China and the southern rim of the Tibetan Plateau is strongly influenced by the ISM. This explains the close linkages between the AMO and summer climate in southwest China. Both the trajectory of the southern branch of the westerlies and the wave train connect the AMO and climate in southwestern China covering the southern rim of the Tibetan Plateau. This explains why most of the sites with significant correlations with our tree-ring chronology are located along the southern rim of the Tibetan Plateau (Fig. 1), and supports the selection of tree-ring chronologies in developing the LSC chronology. High correlations between the AMO and climate are also found along the southern rims of the Tibetan Plateau (Figure S2).

Although the AMO plays a dominant role in regulating temperature variations in the Sichuan Basin, its linkages vary over time. Poor matches between the AMO reconstructions and our tree rings in the late nineteenth century may be due to either weakened linkages between them or decreased reconstruction robustness of the AMO in that period. Weakened linkages with the AMO in modern period may be due to enhanced human activities. For the instrumental period, the running correlations between the AMO, temperatures in the NH and the Sichuan Basin, and the tree-ring chronology are reduced for ~ 1940 and ~ 1980 (Fig. 8). These running correlations are calculated for the detrended series. The running correlations based on the non-detrended series are similar except for the less consistent decreases in correlations for ~ 1980 (Figure S3). As the AMO is one key regulator of the temperature in the Sichuan Basin, the linkages between temperature in the Sichuan Basin and the NH also decreased when the influences of the AMO weakened. The shift from warm to cool phases of the AMO might have contributed to the cooling trend in the Sichuan Basin from the 1950s onwards. However, the recent cooling trend in around 1980 may have been caused by other factors such as enhanced aerosol emissions (Zhou et al. 1996) or Pacific SSTs (Tsai et al. 2014), because the linkages between the AMO and temperature in low-lying areas of the Sichuan Basin decreased sharply in that period (Fig. 8a).

Fig. 8
figure 8

Running correlations based on a 31-year window for the common period from 1901 to 2005 among the Atlantic Multi-decadal Oscillation (AMO), temperature for the Sichuan Basin and the Northern Hemisphere (NH) calculated from the CRU TS3. 10 dataset, as well as the large-scale tree-ring chronology (LSC). Note that the linear trends of the temperature series are removed in the correlation analyses

Full size image

Long temperature records in southwestern China are the key to understanding the long-term dynamics of the ISM, because the temperature gradient between southwestern China near the southeastern Tibetan Plateau and the tropical Indian Ocean is an important driver of the ISM (Feng and Hu 2008). It is generally accepted that the ISM tends to intensify in warm periods due to the northward placement of the intertropical convergence zone (ITCZ) and enhanced land–ocean temperature gradients (Ruddiman 2014). However, the ISM has been decreasing since the early 1980s, although the temperature has been increasing (Li and Zeng 2002). Some studies indicated that the weakening of the ISM may be mainly controlled by enhanced anthropogenic activities such as the increased aerosol emissions over Indian subcontinent (Bollasina et al. 2011). We infer that the weakening of the ISM from the early 1980s onwards may also be linked to the cooling or less pronounced warming in southwestern China, which may be related to the reduced linkages between the AMO and temperature in the region.

5 Conclusions

We developed by far the longest tree-ring chronology (1758–2012) in the Sichuan Basin so as to place the recent cooling in southwestern China in a historical context. We introduced the LSC method to enhance regional climate signals encoded in the tree-ring data. The LSC method is shown be able to better extract climate signals in sites with strong local variations in hot and/or humid regions.

The tree-ring data well represent the mean temperature from current February to July of the Sichuan Basin and much of southwestern China. The highest correlations between our tree rings and the global SSTs are found in the northern Atlantic Ocean, suggestive of the modulation of the AMO. Linkages between the AMO and our tree rings are pronounced on interdecadal scales over most of the past 250 years. The AMO may influence temperature in southwestern China via modulating the southern branch of the westerlies and the circumglobal teleconnections. Temperature in southwestern China can influence the land–ocean temperature gradients and thus the strength of the ISM. The linkages between the AMO and our tree rings were weak in the late nineteenth century and after ~ 1980, the latter of which may be caused by enhanced human activities such as enhanced aerosol emissions. The decoupling of the ISM from the global warming trend since the early 1980s may be related to the weakened linkages between the AMO and temperature in our study region.