HB with the peak of 5,396 m a.s.l. and YL with the peak of 5,596 m a.s.l. are two typical snow-capped mountains in the Central Hengduan Mountain (CHM), separated by the Jinsha River (Figure 1). Different vegetation types are formed along the altitudinal gradient in CHM. The dry valleys are usually covered by scrublands. Warm-temperate coniferous forests, dominated by P. yunnanensis and Pinus armandii, are mainly distributed at altitudes between 1,000 m a.s.l. and 3,300 m a.s.l. Pinus densata is a dominant species at altitudes of 2,600 to 3,500 m a.s.l. on the southern slope. Cold-temperate coniferous forests are dominated by spruce (P. brachytyla, P. likiangensis), fir (A. georgei, Abies forrestii) and larch species (L. potaninii), at elevations ranging from 3,000 m a.s.l. to 4,200 m a.s.l. (Fan et al., 2009).

Although vegetation types along the altitudinal gradient were similar in both HB and YL, the distribution range of P. yunnanensis was a bit different. The species is mainly distributed from 2000 m a.s.l. to 3,500 m a.s.l. in HB, while the distribution of P. yunnanensis in YL ranged from 2000 m a.s.l. to 3,000 m a.s.l. In HB, the soil texture is predominantly loamy, distributed from 1800 to 3,450 m a.s.l., transitioning from red soils to yellow and then brown soils. Above 3,900 m a.s.l., podzolic soil and meadow soil prevail. In YL, the soil texture is dominated by loam and clay, with a gradual transition from clay and loamy clay to sandy clay, sandy loam, and loam as the altitude increases from low to high.

The climate in CHM is strongly influenced by the southeast monsoon and southwest monsoon, with the monsoon climate characteristics of simultaneous rain and heat, and distinct dry and wet seasons. According to the meteorological data (1951–2021) of Lijiang meteorological station (26°51′N, 100°13′E, 2380.9 m a.s.l., Figure 2), annual mean temperature was 12.97°C, while annual maximum temperature (23.92°C) and annual minimum temperature (−0.11°C) appeared in June and January, respectively. The annual total precipitation was 988.93 mm, mainly occurring in summer, as the precipitation from June to September accounted for 75% of the total precipitation. Over the last three decades, the climate of the study area has been characterized by noticeable warming and drying, with a significant increasing trend in temperature (particularly during the period from 1990 to 2016) and a decreasing trend in Palmer Drought Severity Index (PDSI) between 1951 and 2021, while precipitation did not show a significant decreasing trend (Figure 3).

We sampled mature and healthy trees without insect damage at undisturbed forest communities at two adjacent sites with similar climate conditions and growth environments (Table 1). The cores were drilled at the trees’ breast height (1.3 m) with a 5.15-mm-diameter increment borer, one or two cores were extracted per tree from the opposite direction parallel to the contour to avoid growth reaction to the slope. Obtained cores were labelled and placed in plastic straws. In the laboratory, the cores were put into wooden tanks by using white latex and tape. The samples were polished by sandpaper with gradually finer particle sizes (200 mesh to 1,000 mesh) until the growth rings were visible. Then, the cores were visual cross-dating under a microscope and scanned by EPSON (Expression11000XL) scanner. The scanner parameters were set to 24-bit full-color professional mode image types, with a resolution of 2000 dpi. The dated cores were measured using the CDendro and CooRecorder ver. 9.8.1 software at a precision of 0.001 mm to obtain the ring widths. Furthermore, the results of cross-dating were examined by the COFECHA program (Holmes, 1983), and those cores with inadequate quality were removed based on the testing results. Ultimately, 62 trees from 112 cores remained for the further analysis.

To remove the growth trend caused by genetic factors or individual interference while preserving more climate signals, a negative-exponential curve function was applied to standardize the ring width series. Residual chronologies (Figure 4) were developed after autoregressive modelling performed on per standardized series to remove potential autocorrelation and retain high-frequency climate change. Further, all residual series were averaged on a site-by-site basis utilizing the bi-weight robust mean (Cook, 1985), to minimize the influence of outliers. All procedures were executed using the ARSTAN Program (Cook, 1985) and two residual chronologies were generated with chronological eigenvalue (Table 2), including mean sensitivity (MS), signal-to-noise ratio (SNR), standard deviation (SD), and expressed population signal (EPS), etc.

The MS quantifies the degree of inter-annual variation in ring width, serving as an indicator of the sensitivity of tree growth to climate change. A higher value indicates a greater inter-annual variation in ring width, suggesting that the trees are highly responsive to environmental changes. The SNR serves as an indicator of the proportion of climatic signals to non-climatic noise within the ring width data. A high SNR signifies that the ring-width data contains a richer climatic information. The SD is a statistical measure that assesses the dispersion of tree ring width data, reflecting the variation in ring width across different growth years. The EPS is a statistical metric that assesses the representativeness of a sample. A high value of EPS means that the sample data are well represented and can be used to infer growth and climate change in broader regions.

Monthly mean maximum temperature (T_max), monthly mean temperature (T_mean), and monthly mean minimum temperature (T_min), sourced from the neareast Lijiang Meteorological Station (26°50′N, 100°13′E, 2,382 m a.s.l), were acquired through the National Oceanic and Atmospheric Administration (NOAA).¹ Precipitation data were obtained from the Climatic Research Unit (CRU TS v. 4.07; https://crudata.uea.ac.uk/cru/data/hrg) at a resolution of 0.5° × 0.5° (Harris et al., 2020). The PDSI is a comprehensive meteorological drought index that considers precipitation, temperature, evapotranspiration, and drought duration (Alley, 1985). The PDSI data was extracted from the monthly data set compiled by the global CRU grids ² at 0.5° × 0.5° spatial resolution (Dunn et al., 2022). Its index ranges from −4 (indicating extremely dry conditions) to +4 (indicating extremely wet conditions), where a smaller value represents a more severe relative drought.

To capture the lag effect of climate on tree radial growth (Fritts et al., 1965), each tree ring was paired with 14 months of climate data spanning from the previous year’s September through the following October. The climate timeseries interval was from 1987 to 2017 for the data analysis. Five climate variables were selected for analyzing relation to tree radial growth, including T_mean, T_max, T_min, the monthly total precipitation, and PDSI.

The response function was applied to analyze correlations between climate variables and chronologies by using DendroClim2002 (Biondi and Waikul, 2004). The coefficients of the response function are multivariate estimates from a principal component regression model (Morzuch and Ruark, 1991) and the method solves the inter-correlation problems among climatic variables (Fritts et al., 1971). To further identify common climate variables affecting the radial growth of P. yunnanensis, the redundancy analysis (RDA) was also conducted using CANOCO 5.0 software. RDA is the canonical form of principal component analysis (PCA) and constitutes the direct extension of multiple regressions applied to multivariate data (Legendre and Legendre, 1998). In RDA, the ordination axes are constrained to be linear combinations of supplied environmental variables (ter Braak, 1994), and it is an effective method in quantifying the relationship between tree-ring width and climate variables (Girardin et al., 2004; Drobyshev et al., 2013; Zhang et al., 2020). Like the response function, T_mean, T_max, T_min, the monthly total precipitation, and PDSI were selected for RDA. Following a Monte Carlo permutation test involving 999 random permutations, only those climate variables showing significant correlations (p < 0.05) with the two residual chronologies were presented in the RDA.

For tracing the dynamic relationship over time between the annual ring width index and climate, the Moving Forward method (with a window of 36 years) in the Evolutionary and Moving Response and Correlation module of DendroClim 2002 software was also applied to calculate temporal stability of relationships between chronologies and climate variables. The Special Report on Emissions Scenarios (SRES) B2 scenario was considered as future changes in temperature and precipitation, it was published by the Intergovernmental Panel on Climate Change to provide an assessment of the climate change. The graphs were plotted using the ggplot2 program package in R.

Both residual chronologies with high values of SNR and variance in first eigenvector contained large climate information (Table 2). Tree growing in YL with greater MS and SD showed higher interannual fluctuations than those in HB (Table 2). The high EPS (>0.85) confirmed that the sampled population for each site was well-represented, and two residual chronologies could reflect the overall growth characteristics of P. yunnanensis in the study area and then the chronologies could be applied to conduct correlation analysis with climate variables.

We showed correlations between radial growth and climate variables in Figure 5. In HB, the radial growth of P. yunnanensis was significantly associated with T_max in May and precipitation in February, by showing negative and positive correlations, respectively. In addition, PDSI in the previous September negatively affected P. yunnanensis growth. In YL, the radial growth was significantly and negatively correlated with May and June temperatures (both T_mean and T_max), and significantly and positively correlated with T_mean and T_min of the current October. While P. yunnanensis radial growth was found to be positively correlated with precipitation in May and June, and previous October. For PDSI, significant and positive correlations were found in June and July.

The results of RDA (Figure 6) revealed that three climate variables significantly affected the radial growth of P. yunnanensis. Specifically, May T_mean, current October T_mean, and June precipitation explained 17.9, 6.7, and 4.7% of the radial growth, respectively. The first axis and the second axis accounted for 24.49 and 4.85% of the response variables, respectively. May T_mean exerted the most considerable influence on radial growth with negative impacts at both sites, while current October T_mean positively affected the radial growth. June precipitation presented inverse impacts on tree growth at two sites, by showing negative and positive influences in HB and YL, respectively.

The moving response analysis results (Figures 7A,C,E) showed that the growth response of P. yunnanensis to precipitation in May remained relatively stable in HB, by showing positive correlations over the decade with significant periods between 1987 and 1999. The negative response to May temperature reached a significant level before 2000 (p < 0.05) and then weakened. The drought condition in previous September was likely to affect P. yunnanensis growth, as the negative influences of PDSI strengthened in recent 20 years.

In YL, the response of P. yunnanensis to precipitation (positive) and temperature (negative) in current May, temperature in current October (positive), and PDSI (negative) in current October were stable, by showing significant correlations at many years. For current June, the negative impacts of temperature and positive effects of precipitation on tree growth were enhanced after 1992. In Addition, the positive response to PDSI in current June and July fluctuated due to climate change and reached a significant level from 2012 to 2016 (Figures 7B,D,F).

Both response function analysis and RDA revealed that high temperature (T_max) in May restricted the radial growth of P. yunnanensis at two sites (Figures 5, 6). In general, temperatures rose rapidly during the early growing season (April to May), but the region does not experience a rapid increase in precipitation during these months (Figure 2). This may cause drought stress for tree growth induced by the acceleration of plant transpiration and soil moisture evaporation, which lead to insufficient water supply (Denmead and Shaw, 1962). Similar results have been reported in other conifers in CHM, such as the A. georgei in HB (Sun et al., 2021), A. forrestii and P. likiangensis in YL (Yang et al., 2022), and T. dumosa in YL (Guo G. et al., 2009), indicating the importance of sufficient moisture condition of this period on tree growth in the area. This water-demand effect was also supported by the results of response function analysis, by showing positive relation to May precipitation (Figure 5). High precipitation could compensate for the loss of soil water caused by rapid increase in temperature with high evaporation, consequently alleviate the plant transpiration and avoid stomatal closure. Moreover, the increase in precipitation could supply soil moisture for xylem cell production and the onset of xylogenesis (Shen et al., 2020) during the cambium active period of tree growth in the early growing season (Rossi et al., 2016).

Current October T_mean positively affected the radial growth of P. yunnanensis (Figure 6), suggesting the importance of the impact of post-growing season temperature on tree growth. During the post-growing season, cambial activity tended to be weakened (Gričar et al., 2007). However, elevated temperatures could stimulate photosynthesis and generate nutrients for tree growth, and prolong the growing season, consequently leading to the formation of wider annual ring widths (Dawadi et al., 2013). Similar results have also been found in the growth of A. georgei (Sun et al., 2021) in HB and Abies fargesii (Kharal et al., 2017) in adjacent Manan River Valley of central Himalayas.

The common climate variables with inverse effects between two sites were also observed. Excessive precipitation in June had negative impacts on P. yunnanensis growth in HB. Tree began to fast growth in June on HB and would benefit from high temperature for photosynthesis (Zhang et al., 2020). However, greater precipitation is generally combined with enhanced cloudiness, reduced solar radiation input, and lower temperatures, which might decrease the photosynthetic process (Fan et al., 2019). The constraints of early summer precipitation on tree growth have also been reported in previous studies in four conifers (Zhang et al., 2020), P. yunnanensis (Shen et al., 2020), and A. georgei (Yin et al., 2018) in CHM. In comparison with HB, the sampling site is about 300 m lower in YL, water availability was more important than temperature at lower elevation sites (Fan et al., 2009). Xylem cell production benefits from energy and soil moisture during the cambium active phase (Liang et al., 2014) and enough precipitation can satisfy the water demand of the tree growth in the growing season. This was also demonstrated by growth reaction to PDSI in YL, presenting significant and positive correlations with PDSI in current June (Figure 5).

Furthermore, the observed inverse response can be attributed to variations in soil conditions. The soil is relatively infertile with low organic matter due to steep slopes and shallow soil layers in HB (Su and Wang, 2020), elevated precipitation may trigger soil erosion and therefore endanger its growth. While in YL, the soil is comprised of loam and clay with a higher organic matter content, abundant precipitation may create moist soil conditions for tree growth (Guo L. N. et al., 2009).

The radial growth of P. yunnanensis was affected by winter precipitation in HB (Figure 5). The rise in February precipitation facilitated the formation of a thicker snow cover, effectively shielding roots from potential damage induced by low temperatures and harsh winds at higher altitudes (Zhang et al., 2020). As the snow melted, the heightened soil moisture further supported roots in optimizing water and nutrient uptake, ultimately promoting tree growth (Su and Wang, 2020). The positive effects of winter temperature on tree growth were also highlighted by previous studies at high elevations in the area (Fan et al., 2009; Zhang et al., 2018).

Favorable soil moisture conditions in July promoted the radial growth of P. yunnanensis in YL, since adequate soil moisture could accelerate nutrient dissolution and transport, as well as facilitate root absorption, enriching trees with essential nutrients and alleviating drought stress. A similar response pattern to soil moisture was observed by Sun et al. (2020) in Hengduan Mountains, where P. yunnanensis tree rings were positively correlated with summer moisture availability.

In general, the results of moving interval analysis approved the findings obtained through response function and RDA, by showing significant correlations in many years for those months (e.g., May, June, July, and October), which presented significant relations to tree growth. Furthermore, the stability of climate-growth relationships exhibited variability on an interannual scale. For example, the strongest correlation between tree growth and May temperature in HB was detected from 1990 to 2000, then the strength of the relationship became weakened after the 21^st century. This may be attributed to the noticeable temperature surge in Lijiang during that 10-year period (Figure 3), correspondingly, there was a notable strengthening of the positive correlation between growth and May precipitation during this period. The increased precipitation seemed to alleviate the drought stress caused by the sharp temperature rise. Similar results were also presented by the relationship between tree growth and June precipitation, temperature, and PDSI in YL, and other studies that showed the stability of climate-growth relationships varied with warming periods and abrupt climatic changes in the region (Shen et al., 2020).

According to the future climate simulation under the Special Report on Emissions Scenarios (SRES) B2 scenario, the mean temperature in spring, summer, autumn, and winter in Southwest China is projected to increase by 2.6°C, 3.1°C, 2.7°C, and 3.1°C, and precipitation will relatively increase by 8, 7, 6, and 8% in 2071–2,100 (Xu et al., 2005). Incorporating the climate response of tree growth and the prediction of future climate conditions, the productivity of P. yunnanensis forests would likely be enhanced. The radial growth of P. yunnanensis will benefit from increases in precipitation since precipitation is all positively correlated with tree growth (except June precipitation in HB). Although higher temperature in spring and early summer will inhibit tree growth, increases in precipitation of the corresponding seasons is likely to offset the negative impact. Therefore, the growth of P. yunnanensis forests was expected to enhance under future climate change in the area.