Introduction

Variations in summer precipitation play an important role in modulating regional moisture conditions in monsoon Asia, and significantly impacts water supplies, agricultural production, and natural hazards through flood and drought events (Chen et al. 2021). Spatiotemporal variations in summer precipitation must be investigated to help understand and better adapt to these variations. However, the limited span of records, often only available for approximately the past century, makes it difficult to study the long-term variability in summer precipitation. In recent decades, multiple climate proxies, including stalagmite, loess deposits, lake sediments, tree rings, and historical documents have been used to study summer monsoon precipitation at multitemporal scales (Ming et al. 2020; Lin et al. 2021; Zaw et al. 2021; Guo et al. 2022). Among these various proxy records, tree-ring chronologies are widely employed to reconstruct the history of monsoon precipitation due to their high quality calibrations and high temporal resolution (Liu et al. 2021; Yang et al. 2023). However, due to seasonal movement of the monsoon rain belt, water availability may restrict tree growth during part of the growing season, affecting radial growth of partial width rather than the total ring width (Zhao et al. 2019; Peng Z et al. 2023; Wu et al. 2023).

Latewood width (LWW), the portion of an annual ring which formed in the mid to late growing season and exhibited a clear transition from the earlywood width (EWW) (Zhu et al. 2021), has been measured separately and widely used as a unique proxy of summer hydroclimatic features (Crawford et al. 2015; Zhao et al. 2019). Nevertheless, there is a strong autocorrelation between earlywood and latewood that are formed in the current year (Torbenson et al. 2016). Therefore, the possible overlap in the seasonal climate responses of EWW and LWW makes the detection of a pure summer climate signal based on LWW and not affected by EWW quite difficult (Pompa-García et al. 2021; Song et al. 2022). Fortunately, studies on summer precipitation reconstructions associated with the North American Monsoon System in the southwestern United States have shown that the linear effect of EWW on LWW can be removed using a mathematical unitary regression model to isolate the climate signals unique to the LWW (Meko and Baisan 2001; Stahle et al. 2009). Studies have demonstrated that this method effectively enhances the sensitivity of LWW to climate variations in the warm season and is useful for detecting a separate summer precipitation signal (Maxwell et al. 2021; Bregy et al. 2022).

In East Asia, research on the partial width of annual rings has focused on climate responses and paleoclimate reconstructions to explore underlying forcing mechanisms (Zhao et al. 2019; Bing et al. 2022). Few attempts have been made to determine the dependence of latewood on the antecedent earlywood and to extract a clean warm season climate signal specifically encoded in latewood. Forests on the Hasi Mountains on the northern fringe of the Asian monsoon are sensitive to variations in the strength of atmospheric circulation (Chen et al. 2008; Yang et al. 2014). Previous research has reconstructed summer (July–August) precipitation based on tree-ring stable oxygen isotopes of Chinese pine (Wang et al. 2020). A recent study on latewood density indicated that it records hydroclimate rather than temperatures in the Hasi Mountains (Yang et al. 2023). However, most of these studies failed to address the impact of latewood on the previous formed earlywood.

The dominant tree species in the Hasi Mountains, Chinese pine, Pinus tabulaeformis Carriere, was selected to develop tree-ring width chronologies. The annual rings exhibit a distinct transition from earlywood to latewood. There is a rapid decrease in cell size and lumen area, usually appearing as a noticeable color change, making the mark of the earlywood-latewood boundary and the measurement of separate EWW and LWW possible. Using tree-ring widths (TRW), EWW, and LWW chronologies, we compared their response to regional climate variables to identify the main factor controlling the radial growth and the EWW effect on subsequent LWW.

Materials and methods

Study area

The site was located in the Hasi Mountains (104°18.40′–104°35.00′ E, 36°58.20′–37°02.40′ N, 2400–2740 m a.s.l.) in Jingyuan County, Gansu Province (Fig. 1). This region is characterized by a semi-arid, continental climate with relatively wet and warm summers and dry, cold winters as precipitation is concentrated in summer and early fall. Chinese pine grows on northern slopes at altitudes from 2200 to 2700 m a.s.l. at relatively low densities on well-drained soils. Chinese pine in this area is sensitive to climate variations and plays a crucial role in regional climate reconstruction and radial growth monitoring studies (Kang et al. 2012; Zeng et al. 2018).

Fig. 1
figure 1

Locations of sampling sites and the Jingyuan meteorological station

Full size image

Sampling and chronology development

Increment cores were taken from healthy Chinese pine at two sites in October 2012. A total of 146 cores from 72 trees were extracted at breast height using an increment borer (Table 1). All specimens were processed in the laboratory using standard dendrochronological techniques (Stokes and Smiley 1968). Samples were air-dried, mounted, and sanded until the cells were clearly visible under a stereomicroscope. The EWW and LWW of all samples were measured using a Velmex Measuring System (Velmex Inc, Bloomfield, NY), with an accuracy of 0.001 mm. The earlywood and latewood boundary was identified under the microscope based on the cross-sectional area of the tracheids, the thickness of the cell walls and an obvious color change (Stahle et al. 2009) (Fig. 2). The measured tree rings were statistically cross-dated using the software COFECHA (Holmes 1983) to test the quality of the dating technique. Eight cores that showed low correlations (< 0.3) with the master series were excluded from the subsequent site-level chronologies.

Table 1 Information on increment core collection sites
Full size table
Fig. 2
figure 2

Photomicrographs of earlywood and latewood growth in Chinese pine

Full size image

The ARSTAN software (Cook 1985) was used to develop the chronologies. There is a biological trend inherent in raw data series, which is removed by a two-step process. In most cases, negative exponential functions or negative linear functions were adopted to detrend the tree-ring series. A few of the series that did not fit the negative exponential or linear models were detrended with a cubic spline function subjected to a 50% frequency–response cutoff equal to 67% of the series length. The detrended series were processed to produce the mean chronology using a biweight mean method. Standard (STD), residual (RES) and arstan (ARS) chronologies of EWW and LWW were created, and a TRW chronology used as a reference. In this study, the STD chronology was used to preserve low-frequency signals. The mean inter-tree correlation (Rbar) and the expressed population signal (EPS) (Wigley et al. 1984) were used, calculated over 50-year windows with a 25-year lag to validate the common signal strength over time and representativeness of the final chronology. Based on an EPS threshold value that exceeded 0.85, the reliable portion of the chronology extended from 1700 to 2012 for the TRW and EWW chronologies and from 1712 to 2012 for the LWW chronology. Since the two chronologies for the two sites were consistent with each other in terms of both annual and decadal variations during the common period with correlations of more than 0.70 (Fig. 3), all the increment cores collected at the two sites were integrated to develop a composite chronology to obtain better sample depth and reliability (Fig. 4).

Fig. 3
figure 3

Comparison of the TRW, EWW and LWW chronologies between two sites; a the TRW chronologies between HSS01 and HSS02; b and c are the same as a but for the EWW chronologies and LWW chronologies, respectively. Red lines represent the 11-year smooth curves to show decadal variations, and blue lines the sample depth of each chronology

Full size image
Fig. 4
figure 4

Plots of the three composite standard chronologies and their coefficients; a composite chronology of the TRW, the vertical dashed line represents the reliable portion of the chronology. b and c are the same as in a but for the EWW and LWW chronologies, respectively. d Sample depth. e Rbar and EPS for the TRW chronology. f and g are the same as in e but for the EWW and LWW chronologies, respectively

Full size image

Climate data

The Jingyuan Meteorological Station (104.67° E, 36.57° N, 1398 m a.s.l.), which is the closest station to the study sites, provided the climate records. Total monthly precipitation and mean temperature recorded by the Jingyuan Meteorological Station during 1952–2012 were used in this study, and the monthly self-calibrating Palmer Drought Severity Index (scPDSI) from the Royal Netherlands Meteorological Institute (KNMI) Climate Explorer (http://climexp.knmi.nl/) from 1951 to 2012 were also used (Wells et al. 2004). The long-term climate records from the Jingyuan station showed that mean annual temperature was 9.1 ℃ and annual precipitation 231.1 mm during 1952–2012 (Fig. 5). January and July were the coldest and warmest months, with temperatures of − 7.2 and 22.7 ℃, respectively. Monthly precipitation has an uneven temporal distribution over the year, with more than 80% falling during May–September.

Fig. 5
figure 5

Mean monthly temperature, precipitation, and relative humidity recorded at the Jingyuan Meteorological Station during 1951–2012

Full size image

Statistical analysis

An LWW adjustment was made using simple linear regression to remove its systematic dependence on EWW as in Meko and Baisan (2001). This established a linear regression model with the LWW chronology on the EWW chronology first. The residual of the regression equation with a constant of 1.0 added to restore the same mean value as that of the original chronology was reported as the ‘adjusted latewood index’ (LWWadj) (Meko and Baisan 2001; Stahle et al. 2009). The LWWadj indices were computed over the common reliable period (1712–2012) assessed by the EPS that exceeded 0.85 of the EWW and LWW chronologies.

The relationships between the chronologies and the seasonal climate variables were determined. The climate response of each chronology was analyzed with monthly total precipitation, mean temperature and scPDSI. Considering the lagged effects of climate in the previous year on tree growth (Yin et al. 2023), climate data for the correlation analysis ranged from the previous July to the current September.

Results

Characteristics of the chronologies

The composite chronologies for the TRW, EWW, and LWW showed strong significant relationships with each other, with correlations of 0.993 for TRW versus EWW, 0.915 for TRW versus LWW, and 0.865 for EWW versus LWW. In addition, the EWW proportion ranged from 63.7 to 96.7% of the TRW with a mean of 80.9% based on calculations with the raw data series.

The statistical characteristics of the three composite chronologies are shown in Table 2. The values of mean sensitivity (MS) and standard deviation (SD) are higher for the LWW chronology than for the TRW and EWW chronologies, indicating a higher relative change between consecutive years for LWW than for TRW and EWW indices. At the same time, the LWW chronology exhibited a greater year-to-year persistence (AC1), while the TRW and EWW chronologies showed a higher common signal among the series (R1, R2, R3). For the three chronologies, the EPS values were all above 0.97, implying that they were dominated by a coherent stand-level signal rather than an individual tree-level one. Therefore, it was adequate to analyze the linkages between the TRW, EWW and LWW chronologies and climate variables.

Table 2 Statistical characteristics of the composite TRW, EWW and LWW chronologies
Full size table

In addition, the linear regression fit between LWW and EWW was good as measured by the regression R2 exceeding 0.733; the explained variance did not increase significantly when the relationship was modeled with a polynomial function (for a two-order polynomial, R2 = 0.742) (Fig. 6a). After adjusting for the time series, a lower departure from the average occurred for LWWadj than for LWW (Fig. 6b). In addition, LWWadj and LWW had a correlation coefficient of r = 0.535, while the correlation coefficient was 0.04 between LWWadj and EWW, suggesting that there was little dependence of LWWadj on EWW.

Fig. 6
figure 6

Comparison of LWW, EWW and LWWadj. a Scatter plot of LWW versus EWW during the common reliable period 1712–2012; b Comparison of the LWW and LWWadj chronologies

Full size image

Climate signals in tree-ring widths

All the TRW, EWW, and LWW chronologies exhibited significant (p < 0.01) positive correlations with precipitation in August and September of the previous year, and significant (p < 0.05) negative correlations with temperature in the previous September and the current January (Fig. 7). However, the LWWadj was significantly correlated with precipitation in July (Fig. 7). The positive correlations with precipitation combined with the negative ones with temperature indicate a moisture stressed growth pattern. Therefore, the relationships between the chronologies and drought index were tested to verify whether there was a summer drought signal in the chronologies. The results showed that the TRW, EWW, and LWW chronologies were all positively correlated with the scPDSI, significantly from the previous August to the current September (Fig. 7). However, no significant correlation for the LWWadj chronology with the monthly scPDSI occurred.

Fig. 7
figure 7

Correlation coefficients between the chronologies and climatic factors during the 1951–2012. *, p < 0.05

Full size image

In addition, the relationships between the chronologies and the seasonal climate variables were examined. The highest positive correlation was found between total precipitation from the previous July to the current May and the TRW (r = 0.52, p < 0.01), EWW (r = 0.52, p < 0.01) and LWW (r = 0.48, p < 0.01) chronologies. In contrast, the highest negative correlation occurred between mean temperatures from January to May of the current year and the TRW (r = –0.33, p < 0.01), EWW (r = –0.34, p < 0.01) and LWW (r = –0.30, p < 0.01) chronologies. For the LWWadj chronology, the highest correlation coefficient was found between it and total precipitation from June to July (r = 0.29, p < 0.05). Additionally, the highest positive correlation between seasonal mean scPDSI and tree-ring indices were from April to May of the current year, with the correlations of 0.65, 0.63, and 0.64 in TRW, EWW, and LWW chronologies, respectively. As expected, the seasonally averaged climatic variables may be more representative than climatic variables averaged over a single growing month for radial growth. Notably, however, LWW was insensitive to climatic variables in the summer than EWW, regardless of the single month or multi-month average.

Discussion

Tree growth response to climate variables

There were positive correlations between ring width and precipitation in August and September of the previous year, and negative correlations with temperature in the previous September, and positive correlations with scPDSI in all months, indicating that growth of Chinese pine was limited by drought. Previous climate reconstruction studies in this area have also reported that tree growth was regulated by drought (Kang et al. 2012; Peng et al. 2021). High temperatures increase evapotranspiration and consequently exacerbate water deficits for growth (Zhu et al. 2018). Stomatal closure must occur to avoid water loss, resulting in a constraint in photosynthesis and radial growth (Martin-StPaul et al. 2017). The negative correlation between tree growth and temperature in May also indicates that radial growth is limited by drought. In addition, the high temperature in January has negative impacts on tree growth which might be related to the increased respiration costs caused by higher temperatures (McDowell et al. 2011). Moreover, the climate response patterns in TRW, EWW, and LWW are almost the same, which is related to the high correlation between their width variations. Similar results have also been found in North America (Torbenson et al. 2016).

For Chinese pine in the Hasi Mountains, the strongest climate signal recorded in tree ring width appeared chiefly during the late growing season of the previous year, especially for precipitation (Fig. 7). This temporally lagged response to climate could be attributed to tree growth strategies in dry areas, which use the stored carbohydrates from the previous growing season for growth and the available soil water before the growing season for photosynthesis (Fritts 1966; Koretsune et al. 2009). Previous studies have shown that the primordia of needles in Pinus species are formed during the growing season of the previous year (Sacher 1954; Sucoff 1971) and high precipitation would promote bud formation and consequently increase the total leaf area during the following year (Rico-Gray and Palacios-Rios 1996; Hoff and Rambal 2003; Koretsune et al. 2009). The annual increase in stem mass is strongly correlated with leaf area index in Pinus species (Albaugh et al. 2004). Therefore, the amount of precipitation in the previous year would indirectly affect growth in the current year. Thus, temporally lagged climate responses of tree-ring widths to climatic conditions of the previous year widely exist in nearby regions such as the Changling Mountains (Chen et al. 2012), the Helan Mountains (Liu et al. 2005) and the Qilian Mountains (Gou et al. 2015).

Dependence of LWW on antecedent EWW

EWW and LWW were highly correlated in ring width variation and consistent climate response patterns, while the LWWadj had low correlation and showed differences in climate response patterns with EWW, indicating that EWW has legacy effects on LWW. Silkin and Kirdyanov (2003) showed that the growth potential of both EWW and LWW is determined by environmental conditions in the first part of the growing season. The development of needles plays a leading role. Therefore, the LWW of Chinses pine is preconditioned by earlywood growth (Meko and Baisan 2001), as is reported for many other tree species. For example, in northern Europe, Miina (2000) found that earlywood and latewood indices were highly positive correlated for Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies (L.) Karst) in eastern Finland. In addition, across North America, the relationships between EWW and LWW have been investigated using 197 tree-ring data from seven tree species and showed that correlations between EWW and LWW at the same site are often very high (Torbenson et al. 2016).

Latewood is formed in the mid- to late-growing season (July–August) of Chinese pine in the Hasi Mountains (Zeng et al. 2018). However, correlations between LWW and monthly temperature and precipitation in this period were insignificant. However, mid to late-growing season in monsoon Asia is characterized by a warm and wet climate due to the monsoon rain belt, with a weaker drought limit for the radial growth of Chinese pine so that climate signals in this period were rarely recorded in LWW. Conversely, the late growing season is the crucial phase for cell wall thickening but the effect of EWW on LWW decreased the sensitivity of latewood to moisture variations. Research has indicated that the greater quality of LWW chronologies has higher variance and more discrete growth signal independent of EWW (Torbenson et al. 2016). For instance, in certain regions of the world, LWW is less affected by EWW as climatically sensitive, which documents the summer climate signals and could be used to provide estimates of monsoon precipitation in the Asian summer monsoon regions (Zhao et al. 2017), and in the North America monsoon system regions (Griffin et al. 2013).

This similarity exists for other species in surrounding regions of our study area. For example, EWW and LWW of Picea crassifolia Kom. in the middle and western Qilian Mountains showed a similar response to summer temperature (Chen et al. 2012; Xu et al. 2013). These regions are characterized by a semiarid climate with annual precipitation below 400 mm and summer precipitation the majority of annual precipitation, similar to our region. In addition, it is noteworthy that for Chinese pine in the Funiu Mountains, central eastern China, EWW and LWW did not show similar responses to climate variables (Zhao et al. 2019). This is possibly partially attributed to the proportion of EWW on the TRW (65.8%), which is markedly less than that of the same tree species in this paper (80.9%), and this site is located in a humid region with 757 mm annual precipitation. In addition, there are differences in the climate response of EWW and LWW by Pinus armandii Franch. (Hughes et al. 1994) and Picea schrenkiana Fisch. et Mey. (Chen et al. 2010). All these sites have a semi-humid or humid climate with annual precipitation above 400 mm. Therefore, whether EWW and LWW chronologies respond to similar climatic variables is primarily determined by the genes of specific species and modified by different moisture conditions.

Summer climatic implications of LWW adj

Although the summer climate signal in the LWW of Chinses pine was weak in the Hasi Mountains, the LWWadj chronology was significantly correlated with July precipitation in terms of interannual (r = 0.319, p < 0.05) and first order difference (r = 0.376, p < 0.01) variations after adjusting for the effect of EWW. This indicated that the LWWadj chronology, which recorded the climate signals unique to LWW and hardly affected by EWW, should be sufficient to detect a separate summer signal. The enhanced summer precipitation signal in the LWWadj chronology is consistent with the enhanced seasonal precipitation signal in regions of the North America Monsoon System extracted using the same method as in this study (Griffin et al. 2011; Crawford et al. 2015). LWWadj showed a significant link to July precipitation, especially in the consistency of their extreme values (Fig. 8). The extremely narrow rings of Chinese pine latewood in 1954, 1961, 1965, 1971, 1983, 1989, 2002, and 2010 coincided well with extremely low levels of July precipitation in these years, suggesting that latewood growth in our region is limited by precipitation during dry years. July is the main period for cell wall thickening and latewood formation in Chinese pine in the Hasi Mountains (Zeng et al. 2018). When the precipitation deficit surpasses a functional threshold in July during dry years, photosynthesis and carbohydrate storage is diminished as a result of insufficient soil water, thus affecting cell wall-thickening and leading to an indirect decrease in latewood growth (Zhang et al. 2021).

Fig. 8
figure 8

Contrast of the LWWadj series and precipitation in July of the current year (σ indicates the standard deviation of the value)

Full size image

The LWWadj of Chinese pine in the Hasi Mountains is sensitive to summer precipitation as are other species, such Nothotsuga longibracteata (W. C. Cheng) Hu ex C. N. Page in south China (Zhao et al. 2017) and Pseudotsuga menziesii (Mirbel) Franco and Pinus ponderosa Douglas ex C. Lawson in the southwestern US (Stahle et al. 2009; Crawford et al. 2015). In south China, the latewood growth of N. longibracteata (W. C. Cheng) Hu ex C. N. Page shows the effect of severe summer droughts due to subtropical highs (Ma et al. 2022), and thus the drought-limited latewood of this species responded well to precipitation in the warm season (Zhao et al. 2017). Similarly, in the southwestern US, there is a severe pre-monsoon and early-warm-season drought (Stahle et al. 2009; Griffin et al. 2011). Thus, rainfall amount in the summer monsoon determines whether latewood would remain well-hydrated or still limited by the previous drought, resulting in a clear summer precipitation signal recorded in LWWadj. Similarly, although our study area had a relatively high July precipitation, it is still considered a drought because precipitation and high temperatures (resulting in high evapotranspiration), the LWWadj still adequately record the precipitation variations in summer. These results suggest that the adjusted LWW would work better when applied in summer drought-limited areas.

Conclusions

The EWW and LWW chronologies spanning 1685–2012 based on tree-ring data of Cheng in the Hasi Mountains were developed to investigate their response to climate variables during the growing season. The LWW chronology was significantly correlated with the EWW chronology and showed a similar pattern, with an obvious temporally lagged response, indicating a profound effect of EWW on LWW. After removing this effect using a linear regression model, the LWWadj chronology correlated significantly with July precipitation. In addition, the coincidence of extremely narrow latewood and low levels of July precipitation suggests that latewood formation is mainly influenced by July precipitation in dry years. Therefore, the method of enhancing summer climate signals by removing the EWW effect would work better in summer drought-limited areas, a method that has great potential to isolate the summer climate information recorded in latewood with benefits for monsoon precipitation reconstructions in a network in monsoon Asia.