Tree organ growth and carbon allocation dynamics impact the magnitude and δ13C signal of stem and soil CO2 fluxes

Abstract Incomplete knowledge of carbon (C) allocation dynamics in trees hinders accurate modeling and future predictions of tree growth. We studied C allocation dynamics in a mature Pinus sylvestris L. dominated forest with a novel analytical approach, allowing the first comparison of: (i) magnitude and δ13C of shoot, stem and soil CO2 fluxes (Ashoot, Rstem and Rsoil), (ii) concentration and δ13C of compound-specific and/or bulk non-structural carbohydrates (NSCs) in phloem and roots and (iii) growth of stem and fine roots. Results showed a significant effect of phloem NSC concentrations on tracheid growth, and both variables significantly impacted Rstem. Also, concentrations of root NSCs, especially starch, had a significant effect on fine root growth, although no effect of root NSC concentrations or root growth was detected on Rsoil. Time series analysis between δ13C of Ashoot and δ13C of Rstem or δ13C of Rsoil revealed strengthened C allocation to stem or roots under high C demands. Furthermore, we detected a significant correlation between δ13C of Rstem and δ13C of phloem sucrose and glucose, but not for starch or water-soluble carbohydrates. Our results indicate the need to include C allocation dynamics into tree growth models. We recommend using compound-specific concentration and δ13C analysis to reveal C allocation processes that may not be detected by the conventional approach that utilizes bulk organic matter.


Introduction
Forests assimilate carbon dioxide (CO 2 ) from the atmosphere during photosynthesis, store carbon (C) in tree tissues and release C back to the atmosphere via tree and soil respiration, playing an important role in the global C cycle. However, our knowledge of the C allocation dynamics in trees is incomplete (Hartmann et al. 2020). One of the least understood aspects is not only the interaction between C allocation dynamics and structural growth but also the coupling of the latter with respiration (Darenova et al. 2020). These hinder accurate modeling of tree growth (Guillemot et al. 2017) and thus make it difficult to predict how forest ecosystems will respond to climate change.
A key component of the C balance of trees and forests is stem respiration, which accounts for up to 42% of the aboveground total C budget in mature forests (Waring and Schlesinger 1985). Seasonal variation of stem respiration, which is often derived from stem CO 2 efflux (R stem ) , is driven mainly by temperature (Amthor 2000) and, albeit less understood, by the intensity of stem growth (Lavigne et al. 2004, Chan et al. 2018. The latter process is highly dependent on C supply from leaves and therefore corresponds to the strength of C allocation via phloem unloading, which is the process of sugar transport from phloem to sink tissues to fuel sink development or resource storage (Milne et al. 2018). There have been studies investigating the linkage between stem respiration and stem growth rate, measured as stem increment rate (Zha et al. 2004, Chan et al. 2018, and the coupling between stem respiration and rate of C allocation to stems (Darenova et al. 2020). However, variation in stem radial increment does not reflect how the C demand changes between tracheid phenophases (Darenova et al. 2020). For example, the rapid tracheid enlargement phase causes most of the diameter increment but requires less C than the secondary wall thickening and lignification stage, which results in less diameter increment (Rossi et al. 2006, Darenova et al. 2020). More accurate quantification of wood formation dynamics can be obtained by analyzing micro-cores (Mäkinen et al. 2008), which, however, is labor intensive and thus rarely applied in C allocation studies (Darenova et al. 2020).
Another key component of the C cycle in forest ecosystems is soil respiration, which is often derived from soil CO 2 efflux (R soil ). Both components of soil respiration, autotrophic (root and rhizosphere) respiration and heterotrophic respiration, are driven by a seasonally variable supply of recent photosynthates from leaves to belowground (Brüggemann et al. 2011, Hopkins et al. 2013). Belowground C allocation is impacted by, for example, soil temperature (Domisch et al. 2001), soil moisture (Joseph et al. 2020) and soil nutrient availability (Schönbeck et al. 2020). All these factors are linked with root growth (George et al. 1997, Ding et al. 2020). However, the links between root growth, belowground C allocation and soil respiration still remain poorly understood, partly because of the challenges in monitoring root growth dynamics (Perret et al. 2007). The efficiency and ease of determining root growth have been greatly improved by the root scanner method (Dannoura et al. 2008), which nevertheless is still rarely applied (Ding et al. 2020).
A pivotal tool to examine C allocation dynamics in trees is via tracing changes in concentrations of non-structural carbohydrates (NSCs), which serve as both energy carriers and building blocks for amphibolic processes (Hartmann et al. 2020), including growth and respiration. Tracing the seasonal dynamics of NSC concentrations in different organs has improved our understanding of C dynamics at both the wholetree and ecosystem levels (e.g., Furze et al. 2019, Schiestl-Aalto et al. 2019. Nevertheless, many of the published studies on NSC concentration dynamics have separated NSCs only to its components starch and soluble sugars (e.g., Furze et al. 2019, Davidson et al. 2021, although individual sugars serve distinct functions in trees (Hartmann and Trumbore 2016). For instance, sucrose serves as the predominant transport sugar in trees (Rennie and Turgeon 2009), and thereby changes in sucrose content can provide more accurate information on C allocation compared with changes in total NSC content (Hasibeder et al. 2015).
New insights into C allocation patterns have been obtained from studies tracing stable C isotope composition (δ 13 C) of the assimilates from leaves to respired CO 2 or tree biomass (Brüggemann et al. 2011). Such studies, sometimes combined with 13 CO 2 -pulse labeling technique, have identified, for example, the time delay in the coupling of assimilation and R stem or R soil (Barbour et al. 2005, Wingate et al. 2010 and the dependency of C transfer rate on the phenological stage (Masyagina et al. 2016, Desalme et al. 2017. However, several studies have reported decoupling of δ 13 C signal between respired CO 2 and assimilates or respiratory substrates (e.g., Brandes et al. 2006, Kodama et al. 2008, Maunoury-Danger et al. 2013). This decoupling is probably because δ 13 C of bulk organic matter, such as total organic matter or water-soluble compounds, was analyzed in these studies, where the distinct functions (Hartmann and Trumbore 2016) and isotopic disparity (Bowling et al. 2008, Rinne et al. 2015 of individual compounds cannot be accounted for. Compared with δ 13 C analysis of bulk matter, compound-specific isotope analysis (CSIA) may provide more accurate knowledge of the temporal dynamics of down-stem allocation of assimilates and its linkage to R stem or R soil . CSIA has been applied together with 13 CO 2 -pulse labeling to explore C allocation patterns in trees upon changes in environmental conditions, such as elevated CO 2 and soil warming  or drought (Galiano et al. 2017). However, despite the advantage of quantifying C fluxes within trees, whole-tree labeling experiments are mostly confined to small-sized trees (Epron et al. 2012), and the environmental and physiological information in the natural abundance variability is lost in the pulse labeling experiments, motivating the current study of using CSIA to trace natural δ 13 C signal in field-grown mature trees.
In this study, we applied a novel methodological approach to explore how organ growth of mature Scots pine (Pinus sylvestris L.) interacts with C allocation dynamics during a growing season and how this determines the magnitude and δ 13 C of R stem and R soil in a boreal forest. For the first time, we combined continuous CO 2 flux measurements for shoot, stem and soil with high temporal resolution analysis of stem and root growth, which were determined by micro-coring technique and root scanners, respectively. Our interpretations were further supported by data on dimensional growth of shoots and needles and by compound-specific and/or bulk concentration and δ 13 C analysis of NSCs in phloem and roots. We had the following hypotheses: • HP1: Phloem NSC content is linked to tracheid production rate, and the two factors are coupled with R stem dynamics. • HP2: Root NSC content is linked to root growth rate, and the two factors are coupled with R soil dynamics.
Tree Physiology Online at http://www.treephys.oxfordjournals.org • HP3: CSIA provides a better estimate of δ 13 C value of the respiratory substrate, in comparison to bulk isotope analysis, and can identify a coupling between δ 13 C of phloem NSCs and δ 13 C of R stem .

Materials and methods
Site description and environmental data  (Kulmala et al. 2008, Schiestl-Aalto et al. 2019. The mean annual temperature (T) is 3.5 • C, and the mean annual precipitation 711 mm, almost evenly distributed throughout the year (period 1981(period -2010(period , Pirinen et al. 2012. Environmental data in Hyytiälä from 15 April to 30 September 2018 ( Figure 1) were obtained from the Smart SMEAR AVAA portal (https://smear.avaa.csc.fi/; see Method S1 available as Supplementary data at Tree Physiology Online). During 2018, this site experienced a dry period lasting from 16 August to 11 September (Figure 1), when soil moisture content dropped close to the wilting point and soil water potential below −0.5 MPa in the A horizon, i.e., 2-6 cm depth in the mineral soil.
Fluxes and δ 13 C of CO 2 from shoot, stem and soil Automated chamber systems were implemented for measuring shoot CO 2 influx (A shoot ), R stem and R soil from May to October in 2018 (see Table S1 available as Supplementary data at Tree Physiology Online). A transparent shoot chamber of 2.1 l was installed in the top canopy of a mature Scots pine tree ('cuvette tree'), with a debudded 1-year-old shoot inserted (see Figure  S1A available as Supplementary data at Tree Physiology Online, Kolari et al. 2009). A custom-made transparent stem chamber was built around the stem of the cuvette tree at a height of 15 m (see Figure S1B available as Supplementary data at Tree Physiology Online, Rissanen et al. 2020). A transparent soil chamber of 80 l was placed nearby the cuvette tree with all ground vegetation removed (see Figure S1C available as Supplementary data at Tree Physiology Online, Aaltonen et al. 2013). The seasonal pattern of R soil was verified by manual chamber measurements according to Ryhti et al. (2021), conducted every 2 or 4 weeks from April to September in 2018 on six control plots located nearby, which were with intact tree roots but with all ground vegetation removed.
The shoot chamber was intermittently closed 50-80 times per day for 65 s, the stem chamber 48 times per day for 90 s and the soil chamber eight times per day for 840 s. Sample air was taken to gas analyzers (G2201-I, Picarro, USA) that measured 12 CO 2 and 13 CO 2 concentrations and replaced by ambient air leaking freely into the chambers. CO 2 fluxes were calculated via non-linear regression fitted to concentrations of 12 CO 2 and 13 CO 2 (Kolari et al. 2012) during the first 5-50 s, 10-40 s and 40-200 s after the closing of the shoot, stem and soil chamber, respectively, to exclude transient (pressure pumping) effects in chamber CO 2 . For the soil chamber system, the mass flow effect caused by the differences between the replacement air flow rate and the sample flow rate is insignificant, as discussed in Pumpanen et al. (2001). The C-isotope composition of CO 2 fluxes was determined from the ratio of 13 CO 2 flux to 12 CO 2 flux. All isotope data in this paper are reported in δ-notation, in per mil, relative to the international Vienna-Pee Dee Belemnite (V-PDB) standard, defined as where R sample and R standard are the 13 C/ 12 C ratios in a sample and standard, respectively. The δ 13 C values of CO 2 were calibrated by reference CO 2 gases (Air Liquide, Houston, USA) with δ 13 C of −19 and −3.1‰. Mean daytime A shoot , nighttime R stem and daily R soil were calculated. Nighttime R stem was used in subsequent analysis to avoid the influence of the upward transport of stem respired CO 2 via sap streams at daytime ) and the depression of daytime R stem due to low water status (Saveyn et al. 2007). The impact of nighttime sap flow on R stem should be negligible, as the magnitude of nighttime sap flow of the cuvette tree is small (data not shown). As the contribution of CO 2 released from xylem storage pools to R stem was suggested to be insignificant Teskey 2004, Saveyn et al. 2008), its impact on R stem and δ 13 C of R stem was Tree Physiology Volume 42, 2022 not considered here. Nighttime was defined as the time when photosynthetically active radiation above the canopy was less than 30 μmol m −2 s −1 (Kolari et al. 2006). Flux weighted δ 13 C of A shoot , R stem and R soil were used in subsequent analysis.

Concentration and δ 13 C analysis of NSCs
Sampling, extraction and purification of WSCs and starch Stem phloem at 1.3-m height was collected from five mature trees nearby the cuvette tree with a 2-cm diameter corer six times: twice in May and once per month from June to September (see Table S1 available as Supplementary data at Tree Physiology Online). Fine roots (<2 mm) were excavated from locations about 100 m away from the cuvette tree but within the same stand. This was done in order to avoid any disturbance to the soil or harm to the roots of our study tree and to prevent any disturbance to other on-going observations conducted at this site. In the root sampling spots, the tree and soil characteristics equal those at our main study site. Roots were sampled from three random spots (see Table S1 available as Supplementary data at Tree Physiology Online), approximately 5-15 cm depth from the soil 11 times from May to October every 2 or 4 weeks. All samples were microwaved at 600 W for 1 min within 2 h to stop enzymatic and metabolic activities (Wanek et al. 2001). Samples were then dried and homogenized into a fine powder.
Extraction and purification of water-soluble carbohydrates (WSCs) were performed according to Wanek et al. (2001) and Rinne et al. (2012). In brief, water-soluble compounds were extracted in a water bath at 85 • C for 30 min, and the separated supernatant was purified by three types of sample preparation cartridges (Dionex OnGuard II H, A and P, Thermo Fisher Scientific, Waltham, MA, USA). The purified WSC samples were subsequently freeze-dried, dissolved in 1 mL of Milli-Q water and filtered through a 0.45-μm syringe filter.
Starch was extracted from the pellet of the hot water extraction by enzymatic hydrolysis (Wanek et al. 2001, Lehmann et al. 2019. In brief, lipids were washed out by a sequence of methanol/chloroform/water (12:5:3, v/v/v) solution and Mill-Q water. Starch in pellets was gelatinized at 99 • C for 15 min and hydrolyzed with purified (by Vivaspin 15R, Sartorius, Göttingen, Germany) α-amylase (EC 3.2.1.1, Sigma-Aldrich, Buchs, Switzerland) solution at 85 • C for 2 h in a water bath. The hydrolyzed starch in supernatant was cleaned by centrifugation filters (Vivaspin 500, Sartorius, Göttingen, Germany). WSC and starch samples were stored at −20 • C until isotope analysis.

Concentration and δ 13 C analysis of WSCs and starch
Aliquots of WSC and starch samples were pipetted into individual tin capsules (5 × 9 mm, Säntis, Teufen, Switzerland), freeze-dried and wrapped. Concentrations of WSCs were determined using the weights of samples in tin capsules and the weights of plant materials used for extraction. Similarly, starch content was calculated using a conversion factor accounting for the efficiency of the enzymatic conversion of starch to hydrolyzed glucose (see Method S2 available as Supplementary data at Tree Physiology Online). δ 13 C values of WSCs and starch were measured using an elemental analyzer (Europa EA-GSL, Sercon Limited, Crewe, UK) coupled to an isotope ratio mass spectrometry (20-22 IRMS, Sercon Limited, Crewe, UK) at the Stable Isotope Laboratory of Luke (Helsinki, Finland). The δ 13 C values of bulk samples were calibrated against IAEA-CH3 (cellulose, −24.724‰), IAEA-CH7 (polyethylene, −32.151‰) and in-house (sucrose, −12.22‰) reference materials. Measurement precision was 0.1‰ (SD), determined from repeat measurement of a quality control material.

Concentration and δ 13 C analysis of individual sugars
Compound-specific concentration and isotope analysis of WSCs was done online using a Delta V Advantage IRMS coupled with high-performance liquid chromatography (HPLC) with a Finnigan LC Isolink interface following Rinne et al. (2012) at the Stable Isotope Research Laboratory of WSL (Birmensdorf, Switzerland). Four sugars or sugar alcohols with concentrations above 20 ng C μl −1 were detected for HPLC-IRMS δ 13 C analysis: sucrose, glucose, fructose and pinitol/myo-inositol. A dilution series (20, 40, 60, 90, 120 and 180 ng C μl −1 ) of external compound-matched standard solutions (mixture of sucrose, glucose, fructose and pinitol) were analyzed between every 10 samples to calculate sample concentrations and to correct δ 13 C values . Only δ 13 C values of sucrose and glucose were used in this research, due to the uncertainties in fructose δ 13 C results and the relatively insignificant temporal variation in pinitol δ 13 C signal (Rinne et al. 2015). The measurement precision is SD <0.35‰ for sucrose standards (−25.370‰) and SD <0.46‰ for glucose standards (−10.240‰).

Growth data
Shoot and needle growth To trace the timing of shoot and needle growth, length increment of 15 shoots from three trees and the length of one needle of each shoot were measured two or three times a week during 2018, according to Schiestl-Aalto et al. (2013). The observed trees included the cuvette tree and two other trees nearby with similar height and canopy appearance (see Table S1 available as Supplementary data at Tree Physiology Online). The monitored shoots were at the top or the middle of the crown. The percentage of shoot and needle length to their final length was calculated for each observation, and the averages were presented.
Stem growth To trace the timing of stem growth, microcores (diameter 2 mm, length 15 mm) were collected using Trephor corer (Costruzioni Meccaniche Carabin C., Belluno, Italy). Micro-cores were sampled from five mature trees nearby the cuvette tree (see Table S1 available as Supplementary data at Tree Physiology Online), to avoid damage to the cuvette tree. Micro-cores were extracted at 1.3-m height once or twice a week from May to early August, and once every 2 weeks Tree Physiology Online at http://www.treephys.oxfordjournals.org thereafter till October 11. In laboratory, micro-core sections were prepared and analyzed according to Jyske et al. (2014) to determine the number of current-year tracheids in the enlargement phase (a), wall-thickening and lignification phase (b) and mature phase (c). The number of tracheid cells in each phenophase was normalized using the tree-ring width of the previous year (see Method S3 available as Supplementary data at Tree Physiology Online, see Figure S2 available as Supplementary data at Tree Physiology Online). The Gompertz fitting function (Zeide 1993) was then applied to the number of total tracheids and the number of tracheids at phase c, respectively, using the 'nlsLM' function of R package 'minpack.lm' (Elzhov et al. 2010). By subtracting the Gompertz fitted number of tracheids at phase c from that of total tracheids, we obtained the daily number of tracheids at phases a and b. The increased number of tracheids at phases a and b relative to the previous day was defined as the growth rate of tracheids.
Root growth Daily growth data of Scots pine roots during 2018 in the study site were published in Ding et al. (2020) and were used here for defining root growth periods. In brief, root elongation was monitored by three flatbed computer scanners (Epson Perfection V37/V39, Seiko Epson, Tokyo, Japan) installed vertically in soil. Scanner 1 was installed in May 2017, and scanner 2 and 3 in April 2018. Daily images captured by the scanners were analyzed by WinRHIZO TRON 2015a software (Regent Instruments Inc., Quebec, Canada) to determine the daily growth of fibrous (absorptive) roots, a type of fine roots chiefly used in the absorption of water and nutrients (Polverigiani et al. 2011). The sum of daily elongation of active fibrous roots per scanner was used in subsequent analysis. More details about scanner installation and image analysis are described in Ding et al. (2020).

Data analysis
We examined the dependence of: (i) tracheid growth on phloem NSC concentration. Since both variables were performed repeatedly on the same sampling trees, a linear mixed-effects model was applied for analysis (i), using R package 'NLME' (Pinheiro et al. 2021), with sampling tree identifier used as a random term. (ii) R stem on tracheid growth and phloem NSC concentration. Because R stem was measured on a different tree from which tracheid growth and phloem NSC concentration were measured, linear models were used with average values of the independent variables. To remove the effect of temperature on R stem , we used the residual of R stem from the exponential fitting between nighttime R stem and temperature (Saveyn et al. 2007) as the response variable. Chan et al. (2018) have found that stem respiration consistently lags behind stem growth by 1 day for mature Scots pine at our study site, as within-stem diffusion resistances cause R stem to lag actual stem respiration. Thereby, the response variable, R stem residuals, was shifted by 1 day to account for this lagged response. Interaction between tracheid growth and phloem NSC concentration was not significant, thus removed from the final model. (iii) δ 13 C of R stem on δ 13 C of phloem NSCs. δ 13 C of R stem was tested against the average δ 13 C values of individual phloem NSCs from five trees using linear models. (iv) root growth on root NSC concentration. As no repeated measurements were done for root NSC concentration, root growth and R soil , we applied linear models with average values of the response variables and the independent variables, assuming that the average values represent the conditions of the stand. To remove the temperature and soil moisture effects on growth of fibrous root growth, we tested the effect of root NSC concentration on the residual of the regression between root growth and soil temperature and soil moisture (Ding et al. 2020). (v) R soil on root growth and root NSC concentration. As in (iv), we tested the residual of the regression between R soil and soil temperature and soil moisture against the independent variables: the residuals of fibrous root growth (see iv) and concentrations of root NSCs. For all analyses from (i) to (v), time (day of year, DOY) was originally included as a covariate and removed from the final model when not having a significant effect on the response variable.
To test the temporal variability in the correlation between δ 13 C of A shoot and δ 13 C of R stem or δ 13 C of R soil , we applied the wavelet coherence analysis, using the R package 'WaveletComp' (Rösch and Schmidbauer 2016). Wavelet coherence analysis can not only determine the multi-temporal correlations between two time series but also quantify the phase differences (or time lags) between the two time series (Vargas et al. 2011). This method has been applied in R soil (Vargas et al. 2011) or R stem (Chan et al. 2018) studies, because of its advantage in analyzing nonstationary and heteroscedastic time series signals. Thereby, this approach is robust for time series signals with unexpected noises, rising from, e.g., rain pulses or heat waves. Here, we determined correlation and phase difference (or time lag) between daily δ 13 C of A shoot and daily δ 13 C of R stem or δ 13 C of R soil at different time periods or frequencies. Any gaps (2%) in the input data were linearly interpolated. The output results include cross-correlation between x (daily δ 13 C of A shoot ) and y (daily δ 13 C of R stem or δ 13 C of R soil ) time series at any given time period (or frequency) and phase difference (or time lag) of x over y. As δ 13 C signal is transported from photosynthesis to respiration, we considered only the phase differences when δ 13 C of A shoot led δ 13 C of R stem or δ 13 C of R soil as meaningful.
All statistical analysis was done in R version 4.0.0 (R Core Team 2020).

HP1-Relationships between tracheid growth, phloem NSC content and R stem
Tracheid growth periods were defined as follows (Figure 2A): tracheid production period (May 7 to July 31), tracheid maturation period (June 6 to October 11), period with maximum Tree Physiology Volume 42, 2022 tracheid production rate (June 1 to June 15) and period with maximum tracheid maturation rate (July 14 to August 13), according to the Gompertz fitted tracheid production and maturation curves (see Method S3 available as Supplementary data at Tree Physiology Online). Earlywood (EW) growth period of year 2018 was from May 7 to July 14, latewood (LW) growth period from August 2 to October 11 and EW-LW transition period from July 15 to August 1 ( Figure 2B, see Method S4 available as Supplementary data at Tree Physiology Online).
During the EW-LW transition period, between July 4 and August 1, a dramatic depletion of phloem starch content from 72 to 26 mg g −1 and an increase in phloem sucrose content were observed, indicating significant re-mobilization of phloem starch during this period ( Figure 2B). As starch accumulated over a period and had rather stable δ 13 C values which differed from that of glucose and sucrose ( Figure 2B and C), the degradation of starch would very likely blur the environmental signal imprinted in δ 13 C of phloem sugars. Thereby, August 1 was excluded from the linear models for testing the correlation between δ 13 C of R stem and δ 13 C of individual phloem NSCs ( Table 1).
The nighttime R stem increased when tracheid production initiated and decreased during the EW-LW transition period and LW maturation period (Figure 2A and B), showing a synchrony with air T. R stem varied between 0.1 and 1.2 μmol CO 2 m −2 s −1 during the entire period of study, in agreement with the earlier reports (<1 μmol m −2 s −1 in general) by Kolari et al. (2009) for the study site.
Tracheid growth rate was positively correlated with the concentration of glucose, negatively to that of sucrose in phloem (Table 2). Tracheid growth rate had a positive linear regression with R stem residuals (Table 3), whereas phloem NSC concentration showed diverse effects on R stem residuals (Table 3). For instance, concentration of phloem sucrose had a negative correlation with R stem residuals (P = 0.02). In contrast, phloem starch concentration was positively correlated with R stem residuals (P = 0.01).
The temporal correlation between δ 13 C of A shoot and δ 13 C of R stem via wavelet coherence analysis is shown in Figure 3A. At any time period, the regions with significant wavelet coherence (red areas within the white contours) were not continuous throughout the observation period, with gaps existing in between. It denotes that temporal correlations between δ 13 C of A shoot and δ 13 C of R stem were not constant throughout the growing season. The phase differences when δ 13 C of A shoot led changes in δ 13 C of R stem with significant coherence levels (marked by arrows) mainly appeared in the maximum tracheid growth period, maximum tracheid maturation period and dry period. It indicates intensified correlations between δ 13 C of A shoot and δ 13 C of R stem during these periods, compared with other periods.

HP2-Relationships between fibrous root growth, root NSC content and R soil
Three fibrous root growth periods were identified ( Figure 4A): period I (June 6 to July 12), period II (July 13 to August 16) and period III (August 17 to September 30), according to changes in the number of active fibrous roots captured by root growth scanners. The fibrous root flush in period I was captured by Tree Physiology Online at http://www.treephys.oxfordjournals.org Table 1. Results of linear models for testing the dependence of δ 13 C of stem CO 2 efflux (R stem ) on δ 13 C of different phloem NSC pools, i.e., sucrose, glucose, starch and WSCs. Average δ 13 C values of phloem NSCs from five trees were used in the models. Data from four observation days were used in the models.

Response variable (Y)
δ 13 C of R stem Independent variable (X) δ 13 C of sucrose δ 13 C of glucose δ 13 C of starch δ 13 C of WSCs  Table 3. Results of linear models for testing the dependence of stem CO 2 efflux (R stem ) on tracheid growth rate (G tracheid ) and concentration of different phloem NSC pools, i.e., sucrose (C Suc ), glucose (C Glu ), starch (C Sta ), water-soluble carbohydrates (C WSC ) and total NSCs (C NSC ). Average concentrations of phloem NSC from five trees were used in the models. Data from six observation days were used in the models.  Figure 3. Wavelet coherence analysis and phase difference between (A) δ 13 C of shoot CO 2 influx (A shoot ) and δ 13 C of stem CO 2 efflux (R stem ), (B) δ 13 C of A shoot and δ 13 C of soil CO 2 efflux (R soil ) in Hyytiälä during the growing season of 2018 at daily intervals for any given time periods (y axis). Coherence levels are indicated by the colors from blue (low values) to red (high values). White contours represent the 5% significance level, and shaded white regions indicate the region that is potentially impacted by edge effects. Phase differences (i.e., time lags) when δ 13 C of A shoot leads δ 13 C of R stem or R soil with significant wavelet coherence are indicated by arrows. Arrows pointing right indicate no lag; down, ∼0.25 period lag; and left, ∼0.5 period lag.
Tree Physiology Volume 42, 2022 scanner 1 but not by scanner 2 or 3. This scanner-specific fibrous root growth pattern in period I may be associated with the fact that scanners 2 and 3 were not installed until April of the study year, whereas the installation of scanner 1 in May 2017 enabled undisturbed root growth in the observation spot. Both scanners 1 and 2 captured the fibrous growth in period II, whereas root growth around scanner 3 might still have been disturbed by the installation of scanner. The fibrous growth in period III was significantly recorded only by scanner 3, reflecting a spatial heterogenous growth pattern initiated by water stress (Ding et al. 2020). For all three root growth periods, the root growth patterns were to some extent impacted by soil heterogeneity.
Root NSC concentration shared a similar seasonal pattern with fibrous root growth ( Figure 4A). In period I of fibrous root growth, which initiated directly after shoots had developed to their full length (Figure 2A), the maximum number of active fibrous roots corresponded to the first peak in root NSC concentration. Simultaneous increase in fibrous root growth and root NSC content was observed also for period II and for period III, which included the dry period ( Figure 4A). During the entire studied period, root starch content varied in a similar manner to root WSC content (Pearson r = 0.64, P = 0.05, Figure 4A).
Both the absolute values and seasonal variation of R soil captured by the automated chamber were in agreement with that observed by the manual chambers ( Figure 4B). The daily R soil captured by the automated soil chamber varied from 0.5 to 2.9 μmol CO 2 m −2 s −1 during the observation period ( Figure 4B). R soil peaked at the end of July, coinciding with the peaks of fibrous root growth and root NSC content in period II.
Fibrous root growth and root NSC concentrations had no significant effect on R soil residuals. However, positive linear relationships existed between root NSC concentrations and residuals of fibrous root growth (Table 4). The linear model with root starch concentration explained more of the variance than the other models in Table 4.
The temporal dynamics of C allocation to soil was tested by wavelet coherence analysis between δ 13 C of A shoot and δ 13 C of R soil ( Figure 3B). The temporal correlations between the two variables were not constant over the growing season at any time period, as indicated by the gaps between the regions with significant coherence in Figure 3B. The phase differences when δ 13 C of A shoot led δ 13 C of R stem (marked by arrows) mainly appeared in the fibrous root growth period I and period III, which included the dry period. It indicates a tighter relationship between δ 13 C of A shoot and δ 13 C of R stem for these periods than other periods.

HP3 and seasonal variations in δ 13 C of R stem and R soil
Over the observation period, daily δ 13 C of R stem varied 18.7‰, whereas δ 13 C of phloem WSCs and δ 13 C of phloem sucrose varied only 0.9 and 1.3‰, respectively ( Figure 2C). δ 13 C of R stem exhibited a gradual decline from early June to early August, with extremely low values from mid-July onward, coinciding with the EW-LW transition period of stem growth ( Figure 2B and C). R stem was overall more 13 C-depleted relative to A shoot , phloem sucrose and phloem starch by 4.5, 6.4 and 6.2‰, respectively ( Figure 2C). δ 13 C of R stem was significantly correlated with δ 13 C of sucrose (P = 0.02) and glucose (P = 0.01), but not with δ 13 C of WSCs or starch in phloem (Table 1).
Similar to δ 13 C of R stem , the range of δ 13 C variation of R soil (11.7‰) was much higher than that of root WSCs (1.8‰) during the studied period ( Figure 4C). δ 13 C of R soil exhibited a progressive decrease of 5.6‰ during the dry period, coinciding with a decreasing trend in δ 13 C of A shoot which started before the dry period ( Figure 4C). Overall, δ 13 C of R soil was on average Tree Physiology Online at http://www.treephys.oxfordjournals.org Table 4. Results of linear models for testing the dependence of fibrous root growth rate (G fibrous ) residuals on concentration of different root NSC pools, i.e., starch (C Sta ), water-soluble carbohydrates (C WSC ) and total NSCs (C NSC ). Average NSC concentration of roots collected from three random spots, and average G fibrous residuals measured by three root scanners were used in the models. Data from nine observation days were used in the models 1.0 and 1.8‰ lower than δ 13 C of A shoot and δ 13 C of root WSCs, respectively. δ 13 C of R soil was not significantly correlated with δ 13 C of root WSCs, even though a time lag from 0 to 20 days was considered (not shown).

Discussion
To the best of our knowledge, we present the first quantification of the interaction between NSC pools, organ growth and respired CO 2 of stem and roots based on field observation of mature field-grown trees. We identified (i) negative correlations between phloem sucrose concentration and tracheid growth and R stem residuals, (ii) a positive correlation between concentration of root starch and fibrous root growth rate, (iii) strengthened couplings between δ 13 C of A shoot and δ 13 C of R stem or R soil under high stem or root growth demand periods and (iv) a significant correlation between δ 13 C of R stem and δ 13 C of phloem sucrose and glucose, but not for starch or WSCs. Our results provide evidence that C allocation dynamics in trees are highly coupled with growth demands of sink organs via regulations on NSC pools, and suggest that a better knowledge about these dynamics would improve the current presentation of C allocation scheme in land surface models.

HP1-Phloem sucrose dynamics as an indicator of phloem unloading strength
Concentrations of different phloem NSC pools have diverse relationships with tracheid growth and R stem (Tables 2 and  3), among which the most intriguing finding is the negative correlations between sucrose concentration and tracheid growth and R stem residuals. Sucrose, as the predominant transport sugar in trees (Rennie and Turgeon 2009), is often converted to hexoses in sink cells via sucrose synthase or invertase to increase phloem unloading rate (Roitsch et al. 2003, Stein andGranot 2019). Conversely, as a result of reduced sucrose synthase or invertase activity, high sucrose concentration in sink cells may indicate a curtailed phloem unloading rate. The significant role of sucrose cleavage in long-distance assimilate transport, associated with both enzyme mechanisms, has been shown in controlled experiments on herbaceous plants or saplings using transgenic approach (Roitsch et al. 2003, Dominguez et al. 2021). The observed negative correlations between phloem sucrose concentration and tracheid growth in our study thereby support that phloem unloading rate increases under high tracheid growth rate, providing probably the first piece of evidence from field-grown mature trees. The decreased phloem sucrose content, which can be interpreted as increased phloem unloading rate (or C allocation rate) to stem, is apparently due to high C demands of tracheids at enlargement, thickening and lignification phases for completing maturation (Darenova et al. 2020). This result is also in line with the previous finding that C availability influenced wood formation based on phloem girdling experiments (Winkler and Oberhuber 2017). Wavelet coherence analysis between δ 13 C of A shoot and δ 13 C of R stem ( Figure 3A) further supports that C allocation to stem is intensified under high demands. First, the inconsistency of wavelet coherence between the two signals demonstrates the dynamic nature of C allocation over the growing season. Indeed, the down-stem transport of carbohydrates operates at very different velocities depending on environmental conditions and sink demand (Kuzyakov and Gavrichkova 2010). Second, the phase differences with significant coherence levels appeared predominantly in the maximum tracheid growth period, the maximum tracheid mature period and the dry period. All these periods were under high C demands either for tracheid growth (Darenova et al. 2020) or for coping with water stress (Chaves et al. 2002). This result suggests that C allocation dynamics as a function of sink demands, for example, via a dynamic growth model (Schiestl-Aalto et al. 2015), can be used in land surface models. Currently, the C allocation schemes in the land surface models are set as functions of net primary production, which generates large uncertainties in long-term estimates of aboveground biomass accumulation (Montané et al. 2017).
Our results further denote that the increase in C allocation rate to stem and associated enhanced tracheid growth rate promote R stem , since both phloem NSC concentrations and tracheid growth have significant effects on R stem (Table 3). Likewise, Maier et al. (2010) observed a linear correlation between stem carbohydrate content and temperature corrected R stem for girdled loblolly pine trees. Chan et al. (2018) revealed a significant correlation between tracheid growth rate and stem growth respiration, based on modeled growth signal from Tree Physiology Volume 42, 2022 Tree organ growth, carbon allocation and CO 2 flux 2413 dendrometer measurements. However, we note that the correlations are negative, albeit not significant, if concentrations of phloem WSCs are used in the models (Tables 2 and 3). On one hand, this negative correlation may rise from the abovementioned adverse effect of sucrose, which constitutes a significant portion (34%) of WSCs in phloem. On the other hand, it points out potential confusions that concentration analysis on WSCs may cause in in-depth understanding of C allocation dynamics, tracheid growth and stem respiration. Taken together, our results demonstrate that compound-specific concentration analysis of NSCs can provide new insights on the control of sink metabolic activity on tree C allocation (Gessler 2021).

HP2-Root starch as a dynamic C pool supporting root growth
Starch in sink tissues has been long regarded as the primary 'old' reserve, the counterpart of newly assimilated C, used to buffer growth demands and photosynthetic supply (Kozlowski 1992, Vincent-Barbaroux et al. 2019). However, our observations denote that the root starch mainly originated from the newly assimilated C, as starch content was very low before May and showed a synchrony with WSC content ( Figure 4A). It is not likely that starch and WSCs in roots originated from root lipids, considering a similar seasonal pattern in lipids, starch and soluble sugars in roots (Sanz-Perez et al. 2009). Our results further address that root starch serves as a rapid and easily accessible intra-seasonal reserve fueling growth, according to the following facts. First, root starch content presented a larger amplitude of variation than root WSC content ( Figure 4A), demonstrating its relatively fast turnover rate. Second, root starch concentration had a significant positive correlation with fibrous root growth residuals (P = 0.004, Table 4), in line with the observation for mature Norway spruce (Wang et al. 2018). This finding also aligns with the previous report that fresh assimilates are intricately involved in new root growth of conifers (Villar-Salvador et al. 2015 and reference therein). Overall, the results clearly support the hypothesized coupling between root growth and C allocation to roots.
However, we did not find a significant effect of root growth or root NSC content on R soil , though previous studies have suggested that substrate availability in roots can impact R soil (Högberg et al. 2001, Hartley et al. 2007) by controlling autotrophic (root and rhizosphere) respiration (Hopkins et al. 2013) and stimulating decomposition of old organic C pool in soil via root exudation (Adamczyk et al. 2019). The observed decoupling is probably due to heterotrophic respiration, which may not respond to increases in root NSC content in a straightforward manner. Notable part of heterotrophic organisms is independent of tree roots, and an increase in root NSC content does not directly mean increased allocation to belowground symbionts nor root exudates. In addition, this decoupling may partly rise from the fact that our observed R soil at soil surface does not equal to the CO 2 production rate in soil, as the former depends also on soil CO 2 diffusion rate (Fang and Moncrieff 1999).
Nevertheless, the wavelet coherence analysis between δ 13 C of A shoot and δ 13 C of R soil indicates an intensified C allocation to belowground when C demands of roots increase, notably for the fibrous root growth period I and the dry period. The apparent allocation of new assimilates for the initial fibrous root flush is important to increase the chances of getting access to water before the onset of any potential summer drought (Poot and Lambers 2003). The increased C allocation during the dry period is likely to support water and nutrient acquisition via stronger WSC solution in roots ( Figure 4A) or via development of new roots (Zadworny andEissenstat 2011, Joseph et al. 2020).

HP3-The choice of the substrate for δ 13 C analysis is crucial for understanding δ 13 C of R stem and R soil
The observed 13 C-depletion of R soil compared with root WSCs ( Figure 4C) can be partly ascribed to the 13 C-depletion of root respired CO 2 (Werth and Kuzyakov 2010 and references therein), considering that root respiration accounts for about half of soil respiration at our site (Ryhti et al. 2021). It can also be attributed to heterotrophic respiration, which has been suggested to be even more 13 C-depleted than autotrophic respiration for a Scots pine forest in northern Sweden (Bhupinderpal-Singh et al. 2003). Previous field investigations have reported higher or similar δ 13 C values of R soil relative to soil organic matter or soil litter (Bowling et al. 2008 and references therein;Wingate et al. 2010). Together with our finding, δ 13 C values of R soil reflected the contributions of both newly assimilated C and old soil C stores to soil respiration (Søe et al. 2004). The result further demonstrates that a profound understanding of 13 C-discrimination during soil respiration processes in future studies will depend on thorough definition of δ 13 C values of substrates for soil respiration, which should include not only soil C but also root NSCs.
Previous studies on the δ 13 C difference between R stem and respiratory substrates are contradictory (Damesin and Lelarge 2003, Kodama et al. 2008, Wingate et al. 2010. As was the case for δ 13 C of R soil , also these studies were based on analysis of a mixture of organic compounds to estimate δ 13 C of respiratory substrates. Wingate et al. (2010) reported a 13 C-depletion of R stem relative to phloem water-soluble organic matter for maritime pine, Damesin and Lelarge (2003) observed a 13 C-enrichment of R stem in comparison to stem total organic matter for European beech and Kodama et al. (2008) obtained higher or comparable δ 13 C of R stem relative to phloem exudate organic matter for Scots pine. These contrasting results probably stem from varying δ 13 C signals of different substrates (Bowling et al. 2008). Water-soluble organic matter includes also amino acids, organic acids and phenolic compounds aside from WSCs (Antonova and Stasova 2006), and total organic matter consists of an even bigger mixture of compounds. In contrast, the WSCs analyzed in the present study contain only sugars and sugar alcohols , and the sucrose, obtained via CSIA, is the sugar used for down-stem transport (Rennie and Turgeon 2009). Although other storage compounds, such as lipids, can also provide respiratory substrate, these compounds are used when carbohydrate supply from photosynthesis is significantly reduced, e.g., during drought or shading . Hence our study provides a more accurate estimate for the δ 13 C value of the respiratory substrate under normal conditions. The 13 C-depletion of R stem relative to phloem WSCs and sucrose observed in our data ( Figure 2C) may be related to intensive activities of oxidative pentose phosphate (Bathellier et al. 2008) and phosphoenolpyruvate carboxylase (Gessler et al. 2009). The former decarboxylates the 13 Cdepleted C-1 position of glucose (Rossmann et al. 1991), and the latter refixes CO 2 against 12 C (Farquhar 1983), thus enriching the product and depleting the remaining CO 2 .
Also the decoupling between δ 13 C of stem or phloem substrates and δ 13 C of R stem , which has been reported in previous studies (e.g., Brandes et al. 2006, Kodama et al. 2008, Maunoury-Danger et al. 2013, may be explained by the fact that these studies are based on δ 13 C analysis of a mixture of various types of organic compounds. Using phloem sucrose and glucose as substrates, we observed a tight coupling between δ 13 C of R stem and δ 13 C of substrates (Table 1). This coupling further supports HP1, suggesting that phloem transport sugar (i.e., sucrose) provides substrate for stem respiration, in line with findings from 13 CO 2 -pulse-labeling studies (Epron et al. 2012 and references therein). When using phloem WSCs or starch as substrates, we did not detect such coupling with δ 13 C of R stem (Table 1). This is most likely because the δ 13 C signal of WSCs is dampened by the mixing of different components and that of starch by the mixing of assimilated C formed at different times. It also weakens the assumption that phloem water extracts are close to pure sucrose, which grounded the common use of these extracts as a substitute for phloem sucrose in δ 13 C analysis. Therefore, we recommend going beyond the conventional isotope analysis of bulk matter to CSIA, which can provide us more detailed information about the use of individual NSCs for growth and respiration. For example, the use of CSIA can shed light on changes of respiratory substrates, e.g., under stressed conditions or over a diurnal course, and the associated C isotope fractionation during respiration.

Impact of extremely warm and dry weather on CO 2 fluxes and growth
Extreme weather events, for example, heatwaves and droughts, can severely affect tree growth (Rosner et al. 2018, Shi et al. 2021 and CO 2 exchange between trees and the atmosphere (Rodríguez-Calcerrada et al. , Blessing et al. 2016). In the studied year 2018, southern Scandinavia experienced an exceptionally hot and dry summer (Lindroth et al. 2020). Unexpectedly, we recorded limited responses in R stem and R soil to the extreme weather, which, however, is in line with Lindroth et al. (2020) who studied eddy covariance fluxes at our study site. Also, we did not observe a clear impact of the dry period on the tracheid growth of Scots pine, in accordance with the report by Salomón et al. (2022) that the 2018 drought did not lead to consistent growth reduction even in central Europe. The low impact of the hot and dry weather in year 2018 on tree growth and C exchange fluxes can be attributed to the following reasons. First, the impact of the dry weather, which tends to decrease R stem and R soil (Rodríguez-Calcerrada et al. 2014, Blessing et al. 2016, may counterbalance the impact of increased temperatures, which promote respiration rate. Second, Scots pine can tolerate moderate drought well (Cregg and Zhang 2001, Baumgarten et al. 2019), and Matkala et al. (2021 reported that the soil moisture stress in 2018 was not severe enough for this species to suffer from it at a site in northern Finland (Matkala et al. 2021). Third, the dry period appeared at the late part of the growing season, when tracheid production had completed (Figure 2A), thus posing no impact on tracheid growth. Taken together, our data imply that the timing, severity and concurrent occurrence of extreme weather events should be carefully evaluated when forecasting tree growth and forest C exchange, especially in regions with a potential shift toward warmer and drier summers, such as Scandinavia (Christidis and Stott 2021).

Methodological concerns
Only one stem chamber was used to capture R stem and δ 13 C of R stem , which would be insufficient for analyzing absolute values of those parameters. However, when studying processes and their responses to driving factors, such as seasonal variability in R stem , high time resolution data can provide useful information even with a low number of trees. This is also supported by Tarvainen et al. (2018), who demonstrated the similarity in temporal variations of R stem measured from different Scots pine trees growing at the same stand. Furthermore, performance assessment of the chamber system used in our study indicates that it is able to capture small fluxes, such as volatile organic compounds (Rissanen et al. 2020). Moreover, the quality of δ 13 C of R stem data was evaluated by changing the calculating methods (see Figure S3 available as Supplementary data at Tree Physiology Online). The agreement of the results across calculating methods supported the reliability of our data, although the day-to-day variation in δ 13 C of R stem was rather high. We also had only one automated soil chamber to determine R soil . However, we used manual soil chamber data to validate the seasonal course of R soil and found good agreement between the two ( Figure 4B).
Root samples were taken from spots away from the cuvette tree, which may lead to deviation in absolute values of concen-Tree Physiology Volume 42, 2022 tration and δ 13 C of root NSCs between the measured roots and the roots of the cuvette tree. Nevertheless, also in this case, our statistical analysis relies more on the seasonal variations in concentration and δ 13 C of root NSCs rather than the absolute values. Leavitt and Long (1984) suggested that sampling from four trees would provide accurate site-representative δ 13 C trend and absolute values for tree rings. We assume that pooled root samples from three random spots represent the trends in concentration and δ 13 C of root NSCs for our stand.
The stem chamber for δ 13 C of R stem measurement was installed at a higher position than the height of phloem sampling. Previous observation on mature pine trees (Brandes et al. 2006) has shown that δ 13 C of R stem does not change with height. Moreover, it would have been optimal to measure NSC contents both for cambium and phloem, when connecting phloem substrates to R stem and cambium growth. However, since the cambium sampling is detrimental to trees, it is impractical in the study site intended for long-term observation.

Conclusions
In this study, we aimed to examine the interaction between tree organ growth and C allocation dynamics on field-grown mature trees. With this aim, for the first time, we traced the magnitude and δ 13 C signals of shoot, stem and soil CO 2 fluxes with growth measurements of tree organs (i.e., shoots, needles, stem and fine roots), together with compound-specific and/or bulk concentration and δ 13 C analysis of NSCs in phloem and roots of Scots pine (Pinus sylvestris L.) trees over a growing season. The results demonstrate that the allocation of newly assimilated C to stem and roots increases under high growth rate of stem and roots, respectively. The pieces of evidence include: (i) phloem sucrose content, which was conversely linked to phloem unloading rate, had a significant negative correlation with tracheid growth; (ii) δ 13 C of phloem sucrose had a significant correlation with δ 13 C of R stem ; (iii) concentration of root starch, which mainly originated from new assimilates, had a significant positive effect on fine root growth and (iv) time-series analysis revealed significant temporal correlations between δ 13 C of A shoot and δ 13 C of R stem or δ 13 C of R soil for the periods when stem or roots were under high C demands. Increased C allocation and associated growth contributed to higher R stem , which however was not observed for R soil likely due to the disturbance of heterotrophic soil respiration. We suggest that the dynamics of C allocation in response to tree organ growth should be considered in models based on whole tree C allocation for projecting forest growth under climate change. Our results also emphasize the need to carefully evaluate the composition of the substrates, which are used for concentration and δ 13 C analysis, to study tree growth and C allocation dynamics. Where possible, compound-specific concentration and δ 13 C analysis should be used, as these signals can provide more detailed and accurate information on C allocation and utilization processes in trees.

Supplementary Data
Supplementary data for this article are available at Tree Physiology Online.