Tree-Ring Isotopes Provide Clues for Sink Limitation on Treeline Formation on the Tibetan Plateau

: Identifying what determines the high elevation limits of tree growth is crucial for predicting how treelines may shift in response to climate change. Treeline formation is either explained by low-temperature restriction of meristematic activity (sink limitation) or by the photosynthetic constraints (source limitation) on the trees at the treeline. Our study of tree-ring stable isotopes in two Tibetan elevational transects showed that treeline trees had higher iWUE than trees at lower elevations. The combination of tree-ring δ 13 C and δ 18 O data further showed that photosynthesis was higher for trees at the treeline than at lower elevations. These results suggest that carbon acquisition may not be the main determinant of the upper limit of trees; other processes, such as immature tissue growth, may be the main cause of treeline formation. The tree-ring isotope analysis ( δ 13 C and δ 18 O) suggests that Tibetan treelines have the potential to beneﬁt from ongoing climate warming, due to their ability to cope with co-occurring drought stress through enhanced water use efﬁciency.


Introduction
Understanding what determines the high elevation limits of trees is crucial for predicting how treelines may shift in response to climate change. On a global scale, low temperature is found to impact the growth and regeneration of forests at the treeline [1][2][3]. The low temperature limitation on tree metabolism is related to either carbon gain or photosynthesis, known as the "source limitation hypothesis" [4][5][6], or tissue formation, known as the "sink limitation hypothesis" [7,8].
Carbohydrate concentrations of trees have been used to test these two hypotheses [9][10][11]. If treeline trees are able to acquire photoassimilates more efficiently than they can be used for growth, carbohydrates would accumulate in tissues, and, therefore, trees would have more available carbon (non-structural carbon, "NSC") than sinks are able to consume [7,12]. Evidence for the sink limitation hypothesis would be the following scenario: the carbohydrate reserves of treeline trees are not severely depleted at any time of the year, and the carbohydrate concentrations are higher in trees growing at the treeline than in trees growing at lower altitudes in the Himalayas [13,14]. Otherwise, the source limitation hypothesis would be supported. In addition to direct measurements of the NSC of tree tissues, carbon sink-source relationships have been studied using manipulation experiments [6,[15][16][17]. One way to reduce the strength of sinks in trees is to remove buds [4,15]. Similarly, the sources can be manipulated, either by increasing the CO 2 concentration of air [18,19] or by removing the photosynthetically active tissues, such as defoliation [15,16,20,21].
However, measurements of NSC using both natural methods and manipulations suffer great uncertainties in testing these two competing hypotheses. First, high carbohydrate igure 1. Location map of the two elevational transects on the Tibetan Plateau (a), and (b) the climate diagrams based on e climate records from the nearest CRU TS 4.04 grids for the period 1955-2014 (https://catague.ceda.ac.uk/uuid/89e1e34ec3554dc98594a5732622bce9, accessed on 10th June, 2020). The monthly temperatures inlude maximum temperature (Tmax), mean temperature (Tmean), minimum temperature (Tmin) and monthly precipitaon.
The studied tree species include Picea likiangensis var. rubescens and Abies specta For the GBL transect, P. likiangensis var. rubescens forms a pure forest stand in the s or semi-shady slopes, but in the DJ transect, A. spectabibis is the dominant tree spe mixed with other broad-leaf tree species at their lower elevations, and exists as a forest at the treeline elevation [39].
Each transect consisted of forest stands of two elevations, i.e., a treeline plot (U) low-elevation plot (D). Only predominant trees were cored at breast height from a d tion parallel to the contour of the slope, one core per tree, using an increment borer an inner diameter of 5.15 mm. A total of 62 trees were sampled at the GBL transec cluding 29 trees from the treeline elevation (GBLU) and 33 from the low-elevation f (GBLD). At the Dingjie transect, we cored 19 fir trees at the treeline elevation (DJU) 13 trees from the low-elevation forest (DJD). In total, we sampled 200 tree cores from alpine treelines and low-elevation forests. These sampling sites were located in re forests where no evidence of massive human disturbances exists, partly because these ests are under strict protection from logging. There was no clear evidence of fire or o major disturbances.

Tree-Ring Width Methods
The increment cores were air-dried indoors, mounted on wooden slots, and then ished by progressively finer sandpaper until tree-ring boundaries became clearly vis With the help of a microscope, tree rings of each sample were cross-dated by compa The studied tree species include Picea likiangensis var. rubescens and Abies spectabibis. For the GBL transect, P. likiangensis var. rubescens forms a pure forest stand in the shady or semi-shady slopes, but in the DJ transect, A. spectabibis is the dominant tree species, mixed with other broad-leaf tree species at their lower elevations, and exists as a pure forest at the treeline elevation [39].
Each transect consisted of forest stands of two elevations, i.e., a treeline plot (U) and low-elevation plot (D). Only predominant trees were cored at breast height from a direction parallel to the contour of the slope, one core per tree, using an increment borer with an inner diameter of 5.15 mm. A total of 62 trees were sampled at the GBL transect, including 29 trees from the treeline elevation (GBLU) and 33 from the low-elevation forest (GBLD). At the Dingjie transect, we cored 19 fir trees at the treeline elevation (DJU) and 13 trees from the low-elevation forest (DJD). In total, we sampled 200 tree cores from the alpine treelines and low-elevation forests. These sampling sites were located in remote forests where no evidence of massive human disturbances exists, partly because these forests are under strict protection from logging. There was no clear evidence of fire or other major disturbances.

Tree-Ring Width Methods
The increment cores were air-dried indoors, mounted on wooden slots, and then polished by progressively finer sandpaper until tree-ring boundaries became clearly visible. With the help of a microscope, tree rings of each sample were cross-dated by comparing the ring patterns among samples. Tree-ring widths were measured by using a LINTAB 6 measuring system with a resolution of 0.001 mm. COFECHA software was used to check visual cross-dating [40]. All the cross-dated ring widths were derived from the ARSTAN software to standardize the ring-width series, using a negative exponential or linear growth curve to eliminate non-climatic signals. The bi-weight robust mean method was used to merge the detrended index series into a standard (STD) chronology for each forest stand.

Tree-Ring δ 13 C and δ 18 O Methods
We selected five cores from representative canopy trees at each forest stand that had clear and stiff ring boundaries. Since very thin rings occur for several periods at the treeline sites, the annual rings were pooled from the five samples by combining those from the same year. This pooling method has been proven to be a reliable strategy to extract isotopic signals from tree rings on the southeast Tibetan Plateau [41]. Tree rings with a cambial age of less than 50 years were removed to avoid the effect of juvenile wood on the carbon isotope ratios [42]. To ensure homogeneity and efficiency of α-cellulose extraction, the wood materials were then grounded with a centrifugal mill. We extracted the wood cellulose of annual tree rings following the modified version of the methods [43]. To better homogenize the tree-ring cellulose, we used an ultrasound unit in a hot water bath (JY92-2D, Scientz Industry, Ningbo, China) to disrupt the cellulose [44]. The α-cellulose was then freeze-dried for 72 h using a vacuum freeze dryer (Labconco Corporation, Kansas, MO, USA) before the isotope analysis. The δ 13 C values were determined by an element analyzer (Flash EA 1112; Bremen, Germany) coupled with an isotope-ratio mass spectrometer (Delta-plus, Thermo Electron Corporation, Bremen, Germany) at State Key Laboratory of Vegetation and Environmental Change, Institute of Botany Chinese Academy of Sciences. The analytical errors (standard deviations) of the isotope measurements were less than 0.05‰ and 0.3‰ for δ 13 C and δ 18 O, respectively. Calibration was done by measurement of International Atomic Energy Agency (IAEA) USGS-24 (graphite) and by measurement of IAEA-CH 3 (cellulose). All δ 13 C and δ 18 O values are expressed relative to their respective standard (Vienna Pee Dee Belemnite for carbon isotopes and Vienna Standard Mean Ocean Water for oxygen isotopes).
(1) R = 13 C/ 12 C; R sample and R standard are the 13 C/ 12 C ratios of the samples and the standard, respectively.
To accurately acquire tree ring δ 13 C, climate change effect, i.e., the increasing trend of atmospheric CO 2 concentration, was removed. We estimated changes in atmospheric CO 2 concentration and 13 C values using ice core's bubble CO 2 concentration, and its δ 13 C and monitoring data. The ring carbon isotope fractionation sequence can be calculated through the following equations [30]: The 13 C p and 13 C a in the formula were the 13 C value of δ 13 C and CO 2 value of plant photosynthetic products.
where a and b represent CO 2 in isotope fractionation during stomatal (4.4‰) carbon isotope fractionation and the RuBP enzyme carboxylation process (27‰). The concentrations of CO 2 in the leaves and CO 2 in the atmosphere were C i and C a , respectively. Intrinsic water use efficiency (iWUE) can be estimated using C i and C a according to Ehleringer [45]: where 1.6 is the ratio of diffusivities of water and CO 2 in air. Given that δ 18 O values are a reliable indicator of g s, the photosynthesis rate could be derived by a transformation of Equation (4): The A is then used to infer the photosynthesis rate for each forest stand compared within each elevational transect.
In this study, the δ 13 C series was not detrended because we intended to compare the absolute values among different elevations, rather than infer any trends or correlate with other time series, such as climate variables. However, the significant increase in the atmospheric CO 2 concentrations may affect the comparison of δ 13 C and iWUE [33,46]. Therefore, we divided the research period into many slices, with a window-length of approximately 50 years, and then made comparisons over different periods. By doing so, we aimed to see if increased atmospheric CO 2 concentrations have led to different elevational patterns for the compared data. Tree-ring isotopes were measured over the period from 1850 to 2010 for the GBL transect, and from 1900 to 2006 for the DJ transect.

Changes in Tree Growth and Cellulose Stable Isotopes
All chronologies had high SNR (signal-to-noise ratio), SD (standard deviation) and EPS (expressed population signal), indicating that the radial growth of different forest stands was responding to common factors. Moreover, the large percentage of variance explained by the first eigenvectors over the common period indicates that common signals were strong among trees in each forest stand. Additionally, autocorrelations were higher at the treelines than at lower elevations in both transects. Except in DJD, the EPS (expressed population signal) was greater than 0.85 among the studied trees (Table 1). Table 1. Statistic characteristics of the tree-ring width chronologies. Each of the two transects (GBL and DJ) consisted of forest stands of two elevations, i.e., a treeline plot (U) and low-elevation plot (D). In general, during the study period, the base area increment (BAI) of the treelines was lower than that at low elevations, but BAI in recent decades was lower in GBLD ( Figure 2). Tree-ring δ 13 C was higher at low elevations than treelines during the study period ( Figure 3). There was no significant difference in δ 18 O between treelines and low elevations (Figure 4).

Conceptual Model of δ 13 C and δ 18 O Relationships
Changes in photosynthetic assimilation rates and/or stomatal conductance in treelines and low-elevation forests were estimated by assessing the shifts in the δ 13 C-δ 18 O space ( Figure 5). Comparing the average isotope values of the treelines and low elevations, δ 13 C was greater in treelines and low-elevation forests at the GBL and DJ site, while no significant difference was found for δ 18 O in treelines and low elevations (Figures 3 and 4). Since δ 18 O remained unchanged between the two altitudes, the observed change in δ 13 C can only be explained by a higher A, but not by a decrease in gs ( Figure 6), according to the dual-isotope approach of Scheidegger et al. (2000).

Conceptual Model of δ 13 C and δ 18 O Relationships
Changes in photosynthetic assimilation rates and/or stomatal conductance in treelines and low-elevation forests were estimated by assessing the shifts in the δ 13 C-δ 18 O space ( Figure 5). Comparing the average isotope values of the treelines and low elevations, δ 13 C was greater in treelines and low-elevation forests at the GBL and DJ site, while no significant difference was found for δ 18 O in treelines and low elevations (Figures 3 and 4). Since δ 18 O remained unchanged between the two altitudes, the observed change in δ 13 C can only be explained by a higher A, but not by a decrease in g s (Figure 6), according to the dual-isotope approach of Scheidegger et al. (2000).

Discussion
Tree-ring δ 13 C is expected to be strongly affected by photosynthetic rates, which ar governed both by temperature and by other factors at the upper treeline [35,47,48]. Fo lowing the theoretical principles of the dual-isotope model [28], we found that photosyn thesis is greater at treeline than at low-elevation forests. The synergistic effect of elevate CO2 and temperature was reported to stimulate forest productivity in high mountainou forest ecosystems of temperature-limited environments [2,49]. Elevated Ca could improv photosynthesis (A) through the enhanced reaction rate of the RuBisco enzyme [50][51][52] Increased A may improve the non-structural carbon (NSC) pool, which could be used fo sink activities such as growth [53].
Our results imply that treeline trees are able to acquire photoassimilates so efficientl that growth consumption will not be able to deplete the carbohydrate pool (Figures 2 an  6). Consequently, carbohydrates accumulate in tissues, and trees have more available car bon than sinks are able to consume. Although the carbon status of trees was improve given the increased iWUE and A, such improvement may not translate into tissue growt because such a process was not carbon-limited. Low temperature limits meristematic ac tivity [6]. Therefore, growth ceases at a relatively higher temperature than photosynthesi [7]. It is likely that trees uphold their carbon uptake under warmer conditions becaus increased CO2 atmospheric concentrations compensate for reduced stomatal conductance which occurs more frequently, possibly due to drier conditions [54][55][56][57][58]. Nonetheless, thi uptaken carbon may not be translated into enhanced tree growth [34,59,60], but might g to other potential sinks, such as root systems [11,61,62].
The winter respiratory costs in treeline trees usually do not exceed c. 10-15% of th carbon acquired during the growing season [63], and the requirement for carbohydrat storage can be assumed to remain fairly stable in the long term. Cold-adapted tree specie reduce stem and shoot growth when the temperature drops below 5-7 °C [64], but the light-saturated photosynthesis still reaches more than 50% of full capacity at such temper atures, and their tissues are fully charged with storage carbohydrates such as starch an soluble sugars [65]. Therefore, when low temperatures constrain tissue production, th rate of photosynthetic assimilate production may exceed demand. An in situ study of P nus cembra at an alpine timberline suggested that total measured carbon loss during th

Discussion
Tree-ring δ 13 C is expected to be strongly affected by photosynthetic rates, which are governed both by temperature and by other factors at the upper treeline [35,47,48]. Following the theoretical principles of the dual-isotope model [28], we found that photosynthesis is greater at treeline than at low-elevation forests. The synergistic effect of elevated CO 2 and temperature was reported to stimulate forest productivity in high mountainous forest ecosystems of temperature-limited environments [2,49]. Elevated C a could improve photosynthesis (A) through the enhanced reaction rate of the RuBisco enzyme [50][51][52]. Increased A may improve the non-structural carbon (NSC) pool, which could be used for sink activities such as growth [53].
Our results imply that treeline trees are able to acquire photoassimilates so efficiently that growth consumption will not be able to deplete the carbohydrate pool (Figures 2 and 6). Consequently, carbohydrates accumulate in tissues, and trees have more available carbon than sinks are able to consume. Although the carbon status of trees was improved given the increased iWUE and A, such improvement may not translate into tissue growth because such a process was not carbon-limited. Low temperature limits meristematic activity [6]. Therefore, growth ceases at a relatively higher temperature than photosynthesis [7]. It is likely that trees uphold their carbon uptake under warmer conditions because increased CO 2 atmospheric concentrations compensate for reduced stomatal conductance, which occurs more frequently, possibly due to drier conditions [54][55][56][57][58]. Nonetheless, this uptaken carbon may not be translated into enhanced tree growth [34,59,60], but might go to other potential sinks, such as root systems [11,61,62].
The winter respiratory costs in treeline trees usually do not exceed c. 10-15% of the carbon acquired during the growing season [63], and the requirement for carbohydrate storage can be assumed to remain fairly stable in the long term. Cold-adapted tree species reduce stem and shoot growth when the temperature drops below 5-7 • C [64], but their light-saturated photosynthesis still reaches more than 50% of full capacity at such temperatures, and their tissues are fully charged with storage carbohydrates such as starch and soluble sugars [65]. Therefore, when low temperatures constrain tissue production, the rate of photosynthetic assimilate production may exceed demand. An in situ study of Pinus cembra at an alpine timberline suggested that total measured carbon loss during the winter months was small, equaling the photosynthetic production of one to two warm days in spring or summer, when the average air temperature was above 6 • C [63].
Furthermore, we found that tree growth at the two treelines is generally lower than in low-elevation forests, despite the recent increase in growth rate ( Figure 2). Rising atmospheric CO 2 concentration stimulates leaf-level photosynthesis, but not growth [34,55,66,67]. Our results show that the increasing CO 2 concentration during the past century did not affect the tree growth. Not only dependent on photosynthesis rate, tree growth can be affected by other factors, such as nutrient availability, which may confound the effects of atmospheric CO 2 enrichment at treelines [68]. Moreover, water is particularly vital for cell elongation [7]. In arid mountains, trees face simultaneously low temperature and drought, strongly restricting their growth [69,70]. A recent study suggests a sink limitation as the main mechanism behind treeline formation in high, arid Himalayas [13].
Overall, our findings suggest a larger photosynthesis rate and smaller growth rate of trees at treelines as compared with low-elevation forests (Figure 6), providing supportive evidence for the growth limitation on treeline formation. Moreover, Tibetan treelines may have great potential to benefit from the ongoing climate warming due to their ability to cope with the co-occurring drought stress through enhanced water use efficiency.

Conclusions
This study leveraged tree-ring isotope techniques along two treeline transects on the eastern Tibetan Plateau to provide physiological evidence of the sink limitation on treeline formation. Ring widths are overall narrower at treelines than in low-elevation forests. The cellulose-stable isotopes suggest that higher iWUE was observed at treelines, possibly due to their better ability to cope with warming-related drought stress and rising atmospheric CO 2 , as compared with low-elevation forests. Combined with comparable tree-ring δ 18 O at treeline and in low-elevation forests, our results further show that photosynthesis is higher at the treeline than at lower elevations, which is supportive evidence for the sink limitation hypothesis on the formation of the upper treeline in the study region. Therefore, we demonstrate that tree-ring isotopes may hold great potential to further elucidate the underlying mechanisms that form the alpine treelines in a new and physiologically meaningful perspective at leaf level.
Author Contributions: X.P. and L.L. conceived the research, X.P. performed the statistical analyses, X.P., L.L. and X.W. wrote the paper. All authors contributed to the interpretation of the results. The authors declare they have no conflicts of interest regarding this article. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.