Does Central Asian forest growth benefit from a warming-wetting climate? Insights from tree-ring records

Climate warming has pronounced impacts on high-elevation regions, including arid Central Asia, and has multiple impacts on the environment. Forests in these mountainous areas provide essential services by regulating regional climate, sequestering carbon, and supporting soil and water conservation. However, trends in forest productivity and their response to climate change remain unclear. To address this knowledge gap, we collected tree cores from five sample plots in the western Tianshan region. We used tree-ring widths to reconstruct net primary productivity (NPP) and investigated the sensitivity of forest NPP to climate change by analysing weather station data and employing LASSO regression to identify climatic factors influencing forest growth. Our results demonstrate that the reconstructed forest NPP effectively captured significant carbon shifts and revealed a non-significant increase in forest productivity associated with climate warming and higher precipitation between 1970 and 2020 at low and middle elevations in the Tianshan mountains. Humidity is the primary limiting factor affecting forest growth in this region. Conversely, the relationship between temperature and forest growth is not consistent as precipitation increases. Our findings suggest that continued warming will exacerbate water stress in forests.


Introduction
Predicting the effects of climate change on terrestrial ecosystems, particularly forests, is a significant challenge in global climate change research (McDowell et al 2020, Forzieri et al 2022).Despite potential adaptation measures, the occurrence of drought and heat-induced tree mortality has increased in recent years, indicating that elevated temperatures and water stress are the predominant factors affecting forest growth (Allen et al 2010, Chaste et al 2019).
Tianshan Schrenk spruce (Picea schrenkiana (P.schrenkiana)) is a prominent coniferous species with significant ecological value in the Tianshan Mountains of Central Asia.Its extensive coverage and abundance make it crucial for regulating the regional water cycle, soil and water conservation, and maintaining mountain-desert-oasis ecosystem balance (Bonan 2008, Mori et al 2017).P. schrenkiana growth is influenced by climate change, large-scale disturbances, and anthropogenic activity.Accurately quantifying the interplay between these factors is crucial for predicting future forest growth patterns (Schurman et al 2019, Harvey et al 2020, Trotsiuk et al 2020), particularly in ecologically fragile mountainous regions.
Previous studies on forest productivity and climate change have relied heavily on process-based models, but their accuracy is compromised by the coarse meteorological data (Zhao andRunning 2010, Robinson et al 2018).The forest productivity changes in the Tianshan from 1961 to 2000 have been previously addressed (Yuan and Shao 2005, Su et al 2007, Sang and Su 2009).However, due to the intensified climate fluctuations and increased extremes post-1990, reconstructing forest productivity data over longer time scales has become essential to investigate the mechanisms of impact following climate shifts.Large-scale tree chronology networks have been utilized to analysis the impact of climate change on forest growth (Babst et al 2019).However, in order to obtain more climatic information from the tree ring widths to build the chronologies used to reconstruct historical climate change, some sample cores were primarily taken from the trees at forest edges or under limiting conditions, which may amplify climate effects (Bowman et al 2013, Klesse et al 2018).Forests exhibit resilience to climate change, with higher species diversity generally been associated with greater resilience (De Keersmaecker et al 2015, Morin et al 2018).Trees growing at forest edge and under limiting condition are not accurately reflect the overall growth patterns across the entire forest.Tree chronology networks can improve future impact model, particularly in temperate complex forests, however, these uncertainties must be considered (Charney et al 2016, Babst et al 2018, Schurman et al 2019).Ideally, comprehensive forest and climate data covering the entire forest distribution region should be made available.However, owing to data limitations, inference in the relationship between forest growth and climate change often rely on limited data from specific sites.Therefore, tree-ring sampling schemes must be redesigned to understand the effects of forest feedback on climate change across forest systems.
This study introduces a novel method to accurately capture long-term trends in forest net primary productivity (NPP) and examines the relationship between forest growth and climate in the western Tianshan Mountains.First, the historical diameter at breast height (DBH) of trees was calculated using existing DBH data and tree-ring widths.Subsequently, the forest biomass increment was determined using allometric equations to establish a reliable time interval.Long-term forest NPP was reconstructed by relating tree chronology to biomass increment.Using reconstructed forest NPP data, we investigated the relationship between forest growth and various climatic variables.The study aimed to: (1) evaluate whether tree-ring width is an ideal proxy for forest NPP reconstruction in the western Tianshan Mountains, (2) determine the degree to which forest growth is influenced by climate change, and (3) identify the most critical climatic variables for forest growth.

Study region
The Tianshan Mountains, in Central Asia (figure 1), are the farthest mountain range from the ocean in the world.They serve as the primary water source for oases in arid regions.The western Tianshan Mountains have a distinctive '>' shape and are crucial in influencing the regional climate.The southern side acts as a natural barrier hindering hot, dry airflow from the Taklamakan Desert to the north, whereas the northern side restricts cold airflow from the Arctic Ocean to the north.Only the western side allows the influx of humid air from the Atlantic Ocean.Consequently, this unique geography gives rise to a humid temperate continental climate characterized by ample precipitation and mild weather.The Tianshan Mountains are the origin of numerous rivers, emphasizing their importance in maintaining regional water resources.

Climate data
To examine the impact of climatic factors on forest growth, data were collected from weather stations near the sampling plots.Four climate variables were selected: two original (average temperature and precipitation) and two derived variables (accumulated temperature and heat-moisture index).The accumulated temperature is the sum of the average daily temperature >5 • C, as forests can photosynthesize at 5 • C (Yuan and Shao 2005).The heat-moisture index (HM = (temperature + 10)/(precipitation/1000)) was applied to reflect the heat and water supply, larger HM values indicate drier conditions and vice versa (Wang et al 2012).
To examine the impact of climate change on forest growth during different periods, the analysis was divided into four periods: (1) the growing year from January to December; (2) the growing season from May to September; (3) the dormant season from the previous October to the current April; and (4) the previous growing season from May to September.Division between growing and dormant seasons were based on the forest photosynthesis threshold, an average daily temperature >5 • C. Table 1 summarizes the climate indicators selected for each period.

Experimental design and data collection
Sampling plots were selected considering forest density, climate, and management conditions as criteria to roughly pinpoint the sampling sites while avoiding locations with specialized growth and significant human impacts.The sampling route was determined using high-resolution images and topographic maps.High-quality plots were chosen that represented the characteristics of surrounding forest stands.
During selection, we considered different elevations and set the plot size at 30 × 30 m.Each plot had corresponding geographical coordinates near its centre.Table 2 provides relevant information about the sampling sites.
In each plot, we measured the DBH of all trees with DBH values >5 cm.Trees were categorized into different grades based on their DBH values.Five grading categories were utilized.In each grade, a minimum of two trees were selected as samples, ensuring they accurately represented the growth patterns of other trees in that grade.Specifically, five trees were selected from DK as there were only 20 trees in total within the sample plot and similar growth statuses.To minimize the impact of external factors, such as light, on the radial growth of trees, we collected two cores from each selected tree in both the north and south directions and calculated the average tree-ring width of the corresponding DBH grade.
This approach offers the advantage of gathering samples from trees with different DBHs, thereby reducing sampling costs.As a substitute for measuring the annual biomass of each tree in each plot, we aimed to sample trees that represented the entire range of DBH grades to measure the annual forest NPP.
All sample cores were subjected to natural drying, fixing, and polishing following established dendrochronological standards.Sample cores with noticeable defects were excluded from further analysis to ensure accuracy.
The skeleton method was employed to determine the exact age of the trees, which involved identifying tree ring patterns and assigning calendar years to each ring (Douglass 1941).Tree ring widths were measured using the specialized tool LINTAB.In addition, the tree ring series were verified for dating using the COFECHA (Holmes 1983).Tree ring series that did not pass dating verification were re-observed, measured, and corrected until all tree ring series passed dating verification, and then chronologies were build using ARSTAN.

Historical forest NPP reconstruction 2.4.1. Calculation of forest NPP
To estimate forest NPP, we used the total annual aboveground biomass increment of all trees within the plots as a proxy (Davis et al 2009).As historical tree height data are challenging to obtain, we focused on utilizing historical DBH as a substitute, as it be determined using existing DBH measurements and collected trees-ring widths.
We therefore employed an allometric growth equation specific to P. schrenkiana that relates DBH to aboveground biomass (Jie 2020).By applying this equation and the historical DBH values based on tree ring widths, we aimed to calculate the historical forest aboveground biomass (figure 2(1)).This allowed us to assess long-term changes in forest productivity and to better understand the response of P. schrenkiana to climate change.
We estimated the historical DBH values of individual trees by subtracting twice the mean value of the tree-ring widths from the DBH value of the current year.This allowed the establishment of a continuous record of DBH changes over time where a in refers to the ring width for the nth year of the ith core.The DBH in year n-1 (D n−1 ) was calculated based on the DBH in year n (D n ) minus the incremental diameter.
To calculate the annual above-ground biomass values for each tree based on historical DBH values, we employed the following equation: where W is the aboveground biomass (kg) and D is the DBH (cm), n is the nth year.
The NPP of each tree in the current year was calculated by subtracting the above-ground biomass of the previous year from that of the current year where NPP n represents the NPP of an individual tree in year n (t/ha * year), and W n signifies the aboveground biomass of an individual tree in year n (t/ha).
The forest NPP for each year was obtained by summing the NPP of all surviving trees within the sample plot, thereby providing a sample-level measure that minimized the influence of individual tree anomalies where NPP ns is the total NPP value of the surviving trees in year n (t/ha * year) and NPP ni is for the NPP value of the individual trees in year n.
The measured forest NPP was calculated using the tree-ring width and an allometric equation, which allowed accurate estimation of forest NPP for approximately 20 years in each plot.Calculating forest NPP over longer periods is a challenge.Generally, there are different allometric equation for mature and juvenile forests (Peichl and Arain 2007, Yang et al 2019, Mao et al 2022).When we trace back in time to calculate the historical biomass of each tree, some trees enter their juvenile years, rendering our allometric equation unsuitable for calculating their biomass during that period.Moreover, as we extend further into the past, the records of some trees disappear, which brings uncertainty to the calculation of the historical biomass on the plot.Gathering relevant historical records on a large timescale is challenging, and this factor must be addressed in our research to account for potential biases.

Forest NPP reconstruction
The method used to correlate the measured forest NPP values with tree-ring chronology provides a reliable way to reconstruct forest NPP on longer timescales (figure 2(2)).Tree-ring chronology represents the trend in tree-ring width after accounting for age-related growth and it serves as an indicator of tree growth under the influence of climate change.As forest NPP in our plots was calculated using tree-ring width, a strong correlation was expected between forest NPP and tree-ring chronology.This approach can produce a chronology on longer timescales without requiring data from all trees within a plot.
To eliminate the age-related growth component from a specific tree-ring series, we implemented the negative exponential curve method implemented in ARSTAN (Cook 1985).This approach allowed us to separate climate signals from age-related growth patterns in tree-ring data (Blasing andFritts 1976, Cook andKairiukstis 1990).By combining each individual tree-ring series and constructing a regional chronology, we obtained a standardized chronology using ARSTAN.The standardized chronology represents tree-ring variations closely associated with climate change and served as a key component in our analysis (Cook 1985).
To establish a forest NPP reconstruction model, we examined the correlation between tree-ring chronology and measured forest NPP using the Pearson correlation coefficient method.Statistical significance was set at p < 0.05.The reconstructed model can be considered accurate only when the correlation is significant.The measurement results and tree-ring chronologies were then inserted into the calibration period model for linear regression analysis.
To validate the reliability of the reconstructed model, a calibration test was performed using independent samples, following the approach of Michaelsen (Michaelsen 1987).The calibration results indicate that the model satisfied reliability requirements (Wang et al 2023).

Climate-forest growth relationship analysis
In this study, we analysed forest NPP, mean air temperature (TmeanGY), accumulated temperature (TaccuGY), total precipitation (P_GY), and heatmoisture index (HM_GY) in growing year to investigate the relationship between forest growth and climatic variables in western Tianshan Mountains.To examine the trends in forest growth and climate change, we calculated the 5 year moving average for each indicator.Additionally, we calculated the magnitude of change in each indicator during the study period.
To facilitate comparison between different indicators, we standardized the data using z-score processing.This transformation eliminated the unit limitations of the data and allowed direct comparison of the indicators.By transforming the climate variables into dimensionless values, we assessed their relative contributions to forest growth and compared effects on the overall system where Standardized (i, j) represents the standardized value of indicator j in year I, Y (i, j) is the indicator j value in year i, Mean ( Y j ) and Std ( Y j ) denote the average standard deviation of indicator j.
To explore the performance of forest NPP in years with different combinations of heat and humidity, we applied standardized TmeanGY and HM_GY to define the year's characteristics, detailed standards are listed in table 3.
The sensitivity of forest productivity to climate change was expressed using Pearson's correlation coefficients.The absolute value of the correlation coefficient indicates forest NPP sensitivity to climate variables, and the positive or negative of the correlation coefficient indicates that NPP responds positively or negatively to climate variables, respectively.An 'unstable' change in forest productivity in response to climate variables between the two periods is defined by fulfilling one of the following criteria: (a) the sign obvious response occurs change while the response remains significant, (b) the response changes from insignificant to obvious in every direction and (c) the changes from significant to insignificant (Harvey et al 2020).
We employed LASSO regression to select the most influential climatic variables for our regression model.The magnitude of the standardized coefficient for each selected variable indicated its contribution to the predictive model, with positive and negative coefficients representing positive and negative effects, respectively.
All statistical analyses were conducted using R (version 4.2.1)(R Core Team 2015), as well as several packages, including Hmisc, car, and glmnet to implement LASSO regression and ensure robust statistical analysis.

Forest NPP trends from 1950 to 2020
Forest NPP trends in the five plots from 1950 to 2020 exhibited similar patterns.These findings indicate synchronized variations in NPP due to similar climatic conditions experienced by the trees within the plots.Notable, low NPP values were observed in 1957, 1963, 1974, 1985, 1992, 1997, 2001, 2008, and 2014, whereas notably high NPP values were observed in 1970, 1989, 1994, 1999, 2002, 2010, and 2017 (figure 3(a)) in each plot.
The overall forest NPP trend was determined by calculating the average NPP of the four plots and analysing the 5 year moving average results (figure 3(b)), ZS sample plots could not be included in the mean analysis due to temporal inconsistencies with the other sample plots.The average NPP of the plots, fluctuated around 2.2 (t•hm −2 •y −1 ), and exhibited distinct patterns over time.Noticeable upward trends were observed in 1950-1970, 1975-1985, and 2010-2020. Conversely, clear downward trends were observed in 1970-1975, 1985-2000, and 2005-2010. .Since 1990, forest NPP has fluctuated significantly.

Relationship between forest NPP and climate variables
We analysed temperature, precipitation, and derived variables (accumulated temperature and heatmoisture index) from nearby weather stations, focusing on 1970-2020 because of limited data availability  before 1970.Analysis of weather station data near the four plots revealed significant increases in TmeanGY, TaccuGY, and P_GY over the past 50 years (figure 4).In contrast, HM_GY has declined, while forest NPP has tended to increase but not significantly.Figure 5 shows the gradual increase in the number of warm years and years with abundant precipitation in the western Tianshan Mountains.Overall, the regional climate has become warmer and more humid since 1998.In years with higher forest NPP, TmeanGY, TaccuGY, and P_GY were higher, while HM_GY was lower.Although TmeanGY and TaccuGY were not necessarily high, higher P_GY promoted tree growth.Conversely, years with lower forest NPP were generally associated with lower P_GY.Additionally, high TaccuGY led to lower forest NPP.
Figure 6 shows the shift in climatic patterns in the western Tianshan Mountains.Prior to 1998, the region was characterized by a warm-dry and cold-wet climate; after 1998, this transitioned to a warm-wet climate.Warm-wet years were associated with higher forest NPP, whereas cold-dry and warm-dry years were associated with lower forest NPP.Prolonged dryness and cold conditions have significant detrimental effects on tree growth.In the past two decades, the region experienced two significant droughts in 2008 and 2014.These years were characterized by warmdry climate, which had a severe impact on forest growth.Tree growth exhibited a lagged response to climate events, with the climatic conditions of the previous year continuing to influence tree growth in the present year.This indicates a temporal delay in the relationship between climate and tree growth, with the effects of climate on tree growth extending beyond the immediate year.

Sensitivity of forest growth to climate change
The sensitivity of forest NPP to climatic variables was examined using Pearson's correlation analysis of the five plots.The results reveal that the limiting factors influencing the growth of P. schrenkiana varied across different periods (figure 7).
The relationship between forest NPP and temperature was unstable across different periods.Before Forest NPP had an unstable relationship with accumulated temperature across different time periods.Before 1998, TaccuGS, TaccuPGS and forest NPP were negatively associated, and higher accumulated temperatures during the growing season were not beneficial to tree growth.However, after 1998, TaccuGS and TaccuPGS did not substantially affect forest NPP.TaccuDS was positively correlated with the latter, and the lower accumulated temperature during the dormant season was unfavourable for tree growth.
Furthermore, the relationship between forest NPP, precipitation, and the heat-moisture index remained stable.Regardless of period, precipitation was positively correlated with forest NPP, indicating that higher precipitation levels were conducive to tree growth.Conversely, there was a negative correlation between heat-moisture index and forest NPP, suggesting that drought conditions had a detrimental effect on tree growth.
Figure 8 presents a comparison between the reconstructed and predicted forest NPP based on the LASSO regression model using the selected climate variables.The plot shows a strong relationship between the reconstructed and predicted values, indicating a high level of agreement.The correlation coefficient (R 2 ) between the reconstructed and predicted forest NPP was 0.87, indicating that >87% of the variability in forest NPP from 1970 to 2020 could be explained by the selected climate variables.

Discussion
Temperature and precipitation in the Tianshan Mountains have increased over the past few decades, with greater uncertainty in recent years.These changes have been particularly pronounced since 1990, indicating an accelerating trend in climatic conditions (Chen et al 2015, Jiao et al 2019).Our results also indicate that temperature and precipitation continued to rise in mountainous areas, thus promoting forest growth (figure 4).The warm and humid climate of the Tarim Basin has improved vegetation growth, vitality, and coverage over the past few decades   (Jiapaer et al 2015).In situ observations and satellite-based normalized difference vegetation index data indicate that a favourable climate can lead to an extended vegetation growing season (Hu et al 2016).Increased temperature also accelerates the melting of snow and glaciers, which could alleviate water shortages, promote cell differentiation in the cambium, prolong the growth period, and increase photosynthetic efficiency (Boisvenue and Running 2006).
Continuous warming in the Tianshan Mountains has exacerbated drought and negatively affected forest productivity.Over the past 20 years, the two most serious abnormal forest productivity events occurred in the western Tianshan Mountains in 2008 and 2014.Continuous drought also occurred during the 1970s, causing decreased forest productivity (figure 6).
However, comparison of these two events showed that the droughts in 2008 and 2014 had a more severe impact, which was closely related to higher temperatures.Temperatures above the adaptive threshold of a tree cause reduced water availability and increased respiratory energy consumption, leading to reduced tree growth.Melting snow and glaciers trigger shortterm increases in water flow, which may affect vegetation growth by altering soil moisture.However, in the long term, as glaciers recede and disappear, ice will become scarcer, which will decrease water availability, especially during summer (Barnett et al 2005, Shen et al 2020).Droughts can cause tree death or reduce tree growth, for example, by reducing photosynthesis rates, increasing respiration costs for net nutrient accumulation, causing failure of hydraulic conductivity, and plugging of xylem, carbon sequestration, damage to new shoots and roots, loss of needles, and canopy dieback.These conditions can cause competition among trees for water resources, thereby exacerbating disease and fires (Anderegg et al 2015).In the future, as climate warming and extreme climatic events increase, a long-term negative impact on forest ecosystems is expected (Sánchez-Salguero et al 2017).

Conclusions
Climate fluctuations are becoming increasingly evident, highlighting the need to examine changes in forest growth in ecologically fragile regions.Our study demonstrates the advantages of using treering widths to reconstruct historical forest NPP.By incorporating climate data, we observed a nonsignificant increasing trend accompanied by greater fluctuations in forest NPP, which was attributed to climatic warming and humidification and heightened climate extremes in the western Tianshan Mountains.Humidity is the primary factor limiting forest growth in arid regions.In addition, with a sustained rise in precipitation, the influence of temperature on forest growth shifted from negative to positive, introducing uncertainty in future forest NPP model.The establishment of representative, widespread tree-ring networks is crucial for capturing spatial and temporal heterogeneity and improving predictions.These networks serve as key resources for initial model development and validation, thereby enhancing our understanding of the relationship between climatic conditions and forest productivity.

Figure 1 .
Figure 1.Spatial distribution of sampling sites and weather stations.Acquisition of tree sample cores.

Figure 3 .
Figure 3. Forest net primary productivity (NPP) trends in different sample plots (a) and average forest NPP trend of four sample plots (b).

Figure 4 .
Figure 4. Forest net primary productivity (NPP) and climate variable trends including average annual temperature, annual accumulated temperature, annual precipitation, and annual heat-moisture index in four plots.The black curve indicates the five-year moving average of each indicator.The black dotted line shows the trend line.

Figure 5 .
Figure 5. Temporal variations in standardized forest net primary productivity (NPP) and climatic variables.Blue bars represent periods with high forest NPP; Orange bars represent periods with low forest NPP.

Figure 6 .
Figure 6.Forest growth under different heat and humidity conditions.

Figure 7 .
Figure 7. Responses of forest net primary productivity and climate variables during different time periods.* represents p-values between 0.01 and 0.05.* * * represents p-values less than 0.001.

Table 1 .
Description of climate indicators.

Table 2 .
Basic characteristics of five forest plots.

Table 3 .
Definition of heat and humidity criteria for each year.

Table 4 .
Standardized coefficients of selected variables in LASSO regression.TmeanGS and TmeanPGS had negative effects on forest NPP.However, after 1998, the influence of TmeanGS and TmeanPGS on forest NPP became non-significant, whereas TmeanGY had a significant positive effect.From 1970 to 2020, there was a positive correlation between TmeanGY, TmeanGS, TmeanDS and forest NPP.Conversely, TmeanPGS was significantly negative correlated with forest NPP.