Estimating intraseasonal intrinsic water‐use efficiency from high‐resolution tree‐ring δ¹³C data in boreal Scots pine forests

Summary

• Intrinsic water‐use efficiency (iWUE), a key index for carbon and water balance, has been widely estimated from tree‐ring δ¹³C at annual resolution, but rarely at high‐resolution intraseasonal scale.

• We estimated high‐resolution iWUE from laser‐ablation δ¹³C analysis of tree‐rings (iWUE_iso) and compared it with iWUE derived from gas exchange (iWUE_gas) and eddy covariance (iWUE_EC) data for two Pinus sylvestris forests from 2002 to 2019.

• By carefully timing iWUE_iso via modeled tree‐ring growth, iWUE_iso aligned well with iWUE_gas and iWUE_EC at intraseasonal scale. However, year‐to‐year patterns of iWUE_gas, iWUE_iso, and iWUE_EC were different, possibly due to distinct environmental drivers on iWUE across leaf, tree, and ecosystem scales. We quantified the modification of iWUE_iso by postphotosynthetic δ¹³C enrichment from leaf sucrose to tree rings and by nonexplicit inclusion of mesophyll and photorespiration terms in photosynthetic discrimination model, which resulted in overestimation of iWUE_iso by up to 11% and 14%, respectively.

• We thus extended the application of tree‐ring δ¹³C for iWUE estimates to high‐resolution intraseasonal scale. The comparison of iWUE_gas, iWUE_iso, and iWUE_EC provides important insights into physiological acclimation of trees across leaf, tree, and ecosystem scales under climate change and improves the upscaling of ecological models.

Introduction

How trees respond to climate change has profound impact on the carbon and water balances in forest ecosystems (Mathias & Thomas, 2021). This is because trees regulate stomata to control carbon dioxide uptake during photosynthesis and to control water loss that occurs with transpiration. The trade‐off between carbon gain and water loss can be quantified as intrinsic water‐use efficiency (iWUE), expressed as the ratio between net assimilation rate and stomatal conductance (Osmond et al., 1980). A common approach to estimate iWUE is via analysis of stable carbon isotope composition (δ¹³C) of annual tree rings (Farquhar et al., 1982, 1989). However, since tree‐ring δ¹³C reflects a time‐integrated signal (Cernusak, 2020), this method has been rarely applied for detecting high‐resolution intraseasonal variations in iWUE (but see Michelot et al., 2011).

Owing to recent advances in online methods for obtaining carbon isotope data from wood, via laser ablation coupled to an isotope ratio mass spectrometry (hereafter LA‐IRMS), the ease of determining the intraseasonal tree‐ring δ¹³C has been greatly improved (Soudant et al., 2016; Rinne‐Garmston et al., 2022). A pioneer study, which compared intraseasonal δ¹³C data of leaf sucrose and tree rings for Larix gmelinii (Rupr.), found similar low‐frequency trends for the two records but a systematic isotopic offset, caused by postphotosynthetic isotopic fractionation (Rinne et al., 2015). In future, it needs to be verified that such consistent, common patterns between sucrose and tree‐ring δ¹³C are also found for other species and site conditions. But considering that δ¹³C of leaf sucrose can accurately record iWUE at leaf level (Tang et al., in press) and that the sucrose δ¹³C signal is transported to and laid down in tree rings (Gessler et al., 2009), there is high potential to apply LA‐IRMS‐derived tree‐ring δ¹³C data for estimating iWUE at intraseasonal scale. As δ¹³C signal can be potentially retrieved from tree‐ring archives which cover a wide range of areas and periods, such applications, if successful, will be of high value to study the short‐term dynamics of CO₂ and H₂O trade, especially for areas and periods without instrumental records.

For a reliable and accurate estimation of iWUE from tree‐ring δ¹³C (hereafter iWUE_iso), it is important to quantify how a δ¹³C signal may be altered from leaf assimilates to tree rings in postphotosynthetic processes (Fiorella et al., 2022). These processes include the use of reserves in early growing season (McCarroll et al., 2017; Fonti et al., 2018), isotopic fractionation associated with metabolic processes (Gessler et al., 2009; Rinne et al., 2015), and integration of phloem sugars that are assimilated at different canopy heights (Schleser, 1990; Bögelein et al., 2019). These processes are suggested to be species‐specific and site‐specific. For example, the use of reserves has been detected for Quercus petraea at a temperate forest in Fontainebleau‐Barbeau (Vincent‐Barbaroux et al., 2019), but not for Larix gmelinii Rupr. in the permafrost zone of Central Siberia (Rinne et al., 2015). Furthermore, whereas phloem sugars at breast height originated largely from the upper crown for Fagus sylvatica, they originated mainly from the inner and self‐shaded crown parts for Pseudotsuga menziesii (Bögelein et al., 2019). For a certain site or tree species, concomitant high temporal tracking of leaf sucrose δ¹³C along with tree‐ring δ¹³C measurements can help to quantify the extent of postphotosynthetic δ¹³C modification (f _post) and the consequent impact on the iWUE_iso estimates. Such information can also be used as a guideline in studies conducted under similar growth conditions for the same species.

The accuracy of iWUE_iso calculation may also be improved by the use of a complex photosynthetic discrimination model, for instance, via the explicit consideration of mesophyll (Gimeno et al., 2021; Ma et al., 2021) and photorespiratory effects (Keeling et al., 2017; Schubert & Jahren, 2018). Nevertheless, implementing the complex version of photosynthetic discrimination model can be difficult (Lavergne et al., 2019, 2022), partly due to a limited understanding of mesophyll conductance (g _m) and photorespiration dynamics, both of which are dependent on plant species and leaf environment (Sun et al., 2014; Schubert & Jahren, 2018). Hence, the simplified model (Farquhar et al., 1982, 1989) has been applied in the majority of iWUE_iso reconstruction studies (e.g. Frank et al., 2015; Guerrieri et al., 2019). However, it is worthwhile to evaluate how mesophyll and photorespiratory effects may impact iWUE_iso estimates, as this may help to reconcile the trends and absolute values of iWUE derived from different methods.

Intrinsic water‐use efficiency can be also estimated from gas exchange and eddy covariance (EC) measurements (iWUE_gas and iWUE_EC, respectively; e.g. Keenan et al., 2013; Medlyn et al., 2017). iWUE_gas, iWUE_iso, and iWUE_EC represent signals at different scales: leaf level, whole‐tree level, and ecosystem level, respectively. Comparisons between these ‘scale‐specific’ methods not only reveal the limitations of each method (Medlyn et al., 2017) but also show promise to cross‐validate different sources of iWUE data (Guerrieri et al., 2019). Since the gas exchange and EC data are of high temporal resolution, they can overall help verify the intraseasonal pattern of iWUE derived from tree‐ring δ¹³C data, albeit uncertainties exist in each iWUE estimation method (Medlyn et al., 2017; Knauer et al., 2018; Lavergne et al., 2019). The three methods have been scantly compared for their absolute iWUE values, at a global scale for different plant functional types (Medlyn et al., 2017) or at a local site for different tree species (Yi et al., 2019). Studies that compare temporal changes in these iWUE estimates have not been published at intraseasonal scale, and few exist at interannual resolution (Guerrieri et al., 2019; Lavergne et al., 2019). Previous studies have demonstrated a site‐specific (Martínez‐Sancho et al., 2018; Marchand et al., 2020) and temporally dynamic (Liu et al., 2014; Wieser et al., 2018) response of interannual iWUE to environmental drivers. Further knowledge on the scale‐specific temporal trends of iWUE is not only valuable for in‐depth understanding of tree physiological responses to environmental change but also can improve the upscaling of ecological models.

The main objective of this study was to evaluate the reliability of high‐resolution tree‐ring δ¹³C data to estimate intraseasonal changes in iWUE. For this purpose, we (1) compared the intraseasonal trends and absolute values of iWUE_iso, derived from tree‐ring LA‐IRMS δ¹³C analysis, with that of iWUE_EC and iWUE_gas. The comparison was made using a unique set of EC and gas exchange data covering the period from 2002 to 2019, at two Scots pine‐dominated boreal forests with contrasting growth conditions. Next, in the effort to reconcile differences between absolute values of the three iWUE series, we (2) quantified the impact of f _post, g _m, and photorespiration on iWUE_iso estimates. Furthermore, we (3) discussed the environmental and physiological controls on iWUE_iso, iWUE_gas, and iWUE_EC at both intraseasonal and interannual scales, and strengths and weaknesses of each iWUE estimation method.

Materials and Methods

Site description and environmental data

The study was conducted at two boreal forests dominated by Scots pine (Pinus sylvestris L.) in northern and southern Finland (Fig. S1), both of which belong to the Stations for Measuring Ecosystem‐Atmosphere Relations (SMEAR) network. The northern site, Värriö, is close to the arctic‐alpine timberline for Scots pine. The growth conditions are harsher in Värriö, evident in the lower tree heights, sparser canopy (Fig. S1), lower temperature, and shorter growing season, compared with Hyytiälä (Table S1). More characteristics of the study sites are listed in Table S1.

Air temperature (T) and relative humidity (RH) at the canopy height (16.8 m in Hyytiälä and 9 m in Värriö), precipitation, soil moisture at the topsoil, and air pressure (P _a) at ground level were retrieved from the AVAA Smart SMEAR portal (https://smear.avaa.csc.fi/). Vapor pressure deficit (VPD) was calculated from T and RH. Cumulative precipitation and the means of other environmental variables at half‐hourly scale and during the daytime, which was defined as the period from 2 h after sunrise to 2 h before sunset, were calculated.

Sampling and δ¹³C analysis

Sampling

To determine f _post from leaf sugars to phloem sugars and eventually to tree rings, we collected needle and phloem samples during the season 2018 at both sites. One‐year‐old needles (1 N) and current‐year needles (0 N) were collected every 1 or 2 wk from the sun‐exposed top canopy of five mature trees for sugar δ¹³C analysis. Sampling started before the onset of radial growth (early May in Hyytiälä and late May in Värriö) and ended after the cessation of radial growth (October in both sites), conducted all together 20 times per site. Phloem samples were collected at breast height from five mature trees on 6 d per site and season, twice in May and once per month from June to September. Needle and phloem samples were put in a cool box immediately upon collection, and microwaved at 600 W for 1 min within 2 h to stop metabolic activities (Wanek et al., 2001). Tree‐ring samples were taken at breast height after the cessation of growth in 2019. In Värriö, one 5‐mm‐diameter core sample was collected from five mature trees, while in Hyytiälä, five trees were felled, and cross sections were obtained. Average δ¹³C from five trees was used to represent the average conditions experienced by trees at the study sites (Leavitt & Long, 1984). All sampled trees were within the 80% footprint boundaries of the EC towers, which are c. 400 m in Hyytiälä (Launiainen et al., 2022) and 200 m in Värriö (Kulmala et al., 2019).

δ¹³C of leaf and phloem sugars

Water‐soluble carbohydrates (WSCs) were extracted and purified from homogenized needle and phloem samples, according to Wanek et al. (2001) and Rinne et al. (2012). Briefly, the supernatant was separated from the water extraction at 85°C and then purified by three types of sample preparation cartridges (Dionex OnGuard II H, A and P; Thermo Fisher Scientific, Waltham, MA, USA). The purified WSCs were lyophilized, dissolved in deionized water, and filtered through a 0.45‐μm syringe filter (Acrodisc).

δ¹³C of WSCs was measured using an elemental analyzer (EA; Europa EA‐GSL; Sercon Ltd, Crewe, UK) coupled to an IRMS (20–22 IRMS; Sercon Ltd) at the Stable Isotope Laboratory of Luke (SILL) at Natural Resources Institute Finland (Luke, Helsinki). Before δ¹³C analysis, aliquots of solubilized WSCs were pipetted into individual tin capsules (IVA Analysentechnik, Meerbusch, Germany), freeze‐dried, and wrapped. Three reference materials were used to calibrate the δ¹³C values of WSCs, IAEA‐CH3 (cellulose, −24.724‰), IAEA‐CH7 (polyethylene, −32.151‰), and an in‐house sucrose reference (Sigma‐Aldrich, −12.22‰). Repeat measurement of a quality control material indicates a measurement precision of 0.1‰ (SD). δ¹³C of leaf WSCs was calculated as the average δ¹³C of WSCs in 1 N and 0 N.

δ¹³C values of sucrose were analyzed at WSL (Birmensdorf, Switzerland), using a Delta V Advantage IRMS (Thermo Fisher Scientific) coupled with a high‐performance liquid chromatography (HPLC) system with a Finnigan LC Isolink interface (Thermo Fisher Scientific) (Rinne et al., 2012). External sucrose standards, with known δ¹³C values and comparable concentrations to samples (from 20 to 180 ng C μl⁻¹), were analyzed every 10 samples. Correction for HPLC‐IRMS δ¹³C values was performed according to Rinne et al. (2012). The measurement precision of sucrose standards was 0.26‰ (SD). δ¹³C of leaf sucrose was calculated as the average δ¹³C of sucrose in 1 N and 0 N.

δ¹³C of tree rings

Tree‐ring samples were air‐dried before the preparation for isotope analysis. Each sample was sanded with progressively finer grades of sandpaper until ring boundaries and individual cells were clearly identifiable. To ensure that any sawdust collected in voids and intercellular spaces did not affect the isotope signal, each sample was placed in distilled water in an ultrasonic bath for 30 min to remove the sawdust. Samples were then visually inspected under a binocular microscope. Tree rings were measured using WinDendro™ and statistically crossmatched against local chronologies to ensure the correct year was assigned to each ring. Mobile resin and extractives of the samples were removed using a 2 : 1 mixture of toluene and ethanol in a Soxhlet extractor for a duration of 48 h (Loader et al., 1997). After extraction, any residual toluene and ethanol in the samples were removed by rinsing the samples with distilled water in the Soxhlet extractor. Samples were then air‐dried. Resin‐extracted wood was used for LA‐IRMS analysis, as suggested by Schulze et al. (2004).

Intraseasonal δ¹³C was analyzed for each tree‐ring sample using LA‐IRMS at SILL (Methods S1), following the operation principle by Schulze et al. (2004) and Loader et al. (2017). In brief, ablated dust particles were carried by helium flow through a combustion device, the resulting CO₂ was collected with liquid nitrogen, and subsequently, the CO₂ was released upon heating and purified in a GC‐column before its introduction to IRMS. A series of 40 μm tangential laser tracks were sampled along the same radial direction on tree rings at an interval of 40 μm (years from 2010 to 2019) or 80 μm (Fig. 1a). Depending on the ring width, between 5 and 33 laser tracks per tree ring were ablated for δ¹³C analysis. Each sample was run against an in‐house CO₂ reference gas. The raw δ¹³C values were calibrated against USGS‐55 (Mexican ziricote tree powder, −27.13‰) and an in‐house reference (yucca plant powder, −15.46‰), which were both measured concurrently with the tree‐ring samples. The USGS‐55 and the in‐house reference were in the form of a 10‐mm‐diameter disk, which had been prepared by compressing powder using a manual hydraulic press, providing a smooth, solid surface for ablation. In addition, IAEA‐C3 cellulose paper was measured multiple times during each run for quality control of the produced δ¹³C values. The LA‐IRMS measured δ¹³C value for IAEA‐C3 was −24.69 ± 0.24‰, which is in line with the certified value of −24.91 ± 0.49‰. Spot sizes and track lengths of the reference materials were varied to produce variation in signal size, which enabled monitoring of a size effect on δ¹³C values, and the data were corrected when needed (Werner & Brand, 2001).

Fig. 1.

Example of how the formation period of a tree‐ring δ¹³C measurement of Scots pine was defined. (a) Relative position of a tree‐ring δ¹³C measurement within a tree ring; (b) formation period of the tree‐ring δ¹³C measurement based on the growth curves of tracheid production and tracheid maturation for the specific year and site. The growth curves were modeled via the Carbon Allocation Sink Source Interaction model (Schiestl‐Aalto et al., 2015). First, the relative position of a δ¹³C measurement within a tree ring was defined, for example from 0.37 to 0.42 in (a). Then, the initial development date of the tracheids representing that δ¹³C measurement was determined, according to the tracheid production curve in (b) (day of year, DOY) 177 for relative position 0.37). Next, the date was defined when the tracheids for that δ¹³C measurement were fully mature, according to the tracheid maturation curve (DOY 202 for relative position 0.42). Finally, the obtained period representing a tree‐ring δ¹³C value and intrinsic water‐use efficiency derived therefrom (iWUE_iso ), in this case from DOY 177 to DOY 202, was used to align iWUE_iso with iWUE derived from eddy covariance (iWUE_EC ) and gas exchange (iWUE_gas ) data.

Averaging and timing intraseasonal δ¹³C

Individual tree‐ring δ¹³C series had in general similar intraseasonal patterns (Figs S2, S3), enabling the calculation of site‐representative tree‐ring δ¹³C series. First, considering tree‐to‐tree differences in tree‐ring widths, we aligned the intraseasonal tree‐ring δ¹³C data against their relative position within a tree ring (from 0 to 1) per tree, year, and site. Then, we interpolated tree‐ring δ¹³C per tree, year, and site against relative position from 0.05 to 1 at an interval of 0.05 (Figs S2, S3). Finally, we calculated site‐representative tree‐ring δ¹³C as averages of five tree rings at the interpolated relative position.

To time intraseasonal tree‐ring δ¹³C, tree‐ring growth curves against day of year (DOY) were simulated per year and site via a dynamic growth model Carbon Allocation Sink Source Interaction (CASSIA; Schiestl‐Aalto et al., 2015). The performance of CASSIA model results was evaluated by comparison with xylogenesis observational results for years 2007, 2008, 2009, 2018, and 2019, for both sites (Methods S2; Fig. S4). With simulated growth curves, the start DOY of tracheid production and the end DOY of tracheid maturation could be determined for each site‐representative tree‐ring δ¹³C data point (Fig. 1).

iWUE estimates

iWUE from leaf gas exchange (iWUE_gas )

CO₂ fluxes (A) and H₂O fluxes (E) were measured online with leaf gas exchange systems, as described by Altimir et al. (2002) and Aalto et al. (2014) from 2002 to 2019 at both sites. However, no data were available for years 2005 and 2014 in Hyytiälä. In brief, transparent acrylic plastic chambers were installed at the top canopy of one to four mature trees with a debudded 1‐ or 2‐yr‐old shoot enclosed. Four chamber designs were employed over the years in Hyytiälä, whereas the chamber design in Värriö was the same all the time. The nonairtight chambers were automatically closed intermittently for 50 to 80 times (in Hyytiälä) or 150 to 180 times (in Värriö) per day, with sample air drawn to gas analyzers (URAS‐4; Hartmann & Braun, Siek, Germany; LI‐840; Li‐Cor, Lincoln, NE, USA). CO₂ fluxes and E were calculated from instantaneous CO₂ and H₂O records taken during the first 30–40 s of chamber closure (Kolari et al., 2012). Flux data were omitted when RH exceeded 85% to avoid biased results due to adsorption of water on chamber walls and tubing (Altimir et al., 2006). Small fluxes (A < 0.5 μmol m⁻² s⁻¹ and E < 0.1 mmol m⁻² s⁻¹) were also discarded given the uncertainties they may cause in the calculation of iWUE_gas. Half‐hourly A and E data series were produced and applied to calculate iWUE_gas (Eqn 1; Beer et al., 2009), where g is stomatal conductance. In the well‐stirred chambers, boundary layer conductance is high (Uddling & Wallin, 2012) and thus for simplicity not considered in the calculation of iWUE_gas.

$${\text{iWUE}}_{\mathrm{gas}}=A/g=A/E·\left(\mathrm{VPD}/P_{\mathrm{a}}\right)$$ (Eqn 1)

iWUE from eddy covariance (iWUE_EC )

The net ecosystem CO₂ exchange (NEE) and H₂O flux (ET) were measured using a closed‐path EC system above the stand at 24 m height from 2002 to 2017 and at 27 m height from 2018 to 2019 in Hyytiälä, and at 16.6 m height from 2012 to 2019 in Värriö. Briefly, the EC data were: screened for outliers and erroneous measurements using standard methods (Aubinet et al., 2012); filtered by the turbulence criteria (Markkanen et al., 2001); averaged to half‐hourly scale and gap‐filled (Kulmala et al., 2019); and corrected for the storage of CO₂ below the measuring height (Kolari et al., 2009; Launiainen et al., 2016). Half‐hourly gross primary production (GPP) was calculated by subtracting the modeled total ecosystem respiration from NEE (Kulmala et al., 2019). Furthermore, half‐hourly GPP and ET data were discarded, when precipitation occurred before the measurements, or when RH was higher than 85% to minimize the effect of condensation on canopy surfaces or instruments. More detailed description of the EC systems can be found in Vesala et al. (2005) and Kulmala et al. (2019), and EC data processing in Launiainen et al. (2016) and Mammarella et al. (2016). Assuming infinite aerodynamic conductance and no contribution of nontranspiratory water fluxes, half‐hourly iWUE_EC was calculated from GPP and ET data by Eqn 2 (Beer et al., 2009), where G _s is surface conductance. Twenty‐three percentage of half‐hourly EC data were gap‐filled. Using the gap‐filled data had limited impact on the temporal trends and absolute values of iWUE_EC, considering that the iWUE_EC series with and without days that had high percentage (≥ 50%) of gap‐filled flux records were highly correlated (r = 0.98, P < 0.001).

$${\text{iWUE}}_{\mathrm{EC}}=\mathrm{GPP}/G_{\mathrm{s}}=\mathrm{GPP}/\mathrm{ET}·\left(\mathrm{VPD}/P_{\mathrm{a}}\right)$$ (Eqn 2)

iWUE from tree‐ring δ¹³C (iWUE_iso )

Intrinsic water‐use efficiency can also be estimated from isotope data via the photosynthetic discrimination model of Farquhar et al. (1982, 1989) (Eqns (Eqn 3), (Eqn 4), (Eqn 5), (Eqn 6)).

$${\text{iWUE}}_{\mathrm{iso}}=\left(c_{\mathrm{a}}-c_{\mathrm{i}}\right)/1.6$$ (Eqn 3)

$$\Delta =a+\left(b-a\right)·\frac{c_{\mathrm{i}}}{c_{\mathrm{a}}}-\left(b-a_{\mathrm{m}}\right)·A/\left(g_{\mathrm{m}}·c_{\mathrm{a}}\right)-f·{\mathrm{\Gamma }}_{*}/c_{\mathrm{a}}$$ (Eqn 4)

$${\mathrm{\Gamma }}_{*}=42.75·\exp \left[37830·\left(T_{\mathrm{k}}-298\right)/\left(298·R·T_{\mathrm{k}}\right)\right]$$ (Eqn 5)

$$\Delta =\left({\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{air}}-{\mathrm{\delta }}^{13}{\mathrm{C}}_{\text{tree}}\right)/\left(1+{\mathrm{\delta }}^{13}{\mathrm{C}}_{\text{tree}}/1000\right)$$ (Eqn 6)

c _a and c _i are the atmospheric and intercellular CO₂ concentrations, respectively; Δ is the photosynthetic discrimination; a (4.4‰) is the fractionation due to diffusion of CO₂ through stomata; b (29‰) is the fractionation due to carboxylation; a _m (1.8‰) is the fractionation during the mesophyll CO₂ transfer; f is the fractionation during photorespiration; ${\mathrm{\Gamma }}_{*}$ is the CO₂ compensation point in the absence of dark respiration in μmol mol⁻¹, estimated according to Eqn 5 (Bernacchi et al., 2001); T _k is the leaf temperature in K, taken as the air T measured inside the chamber; $R$ is the universal gas constant (8.3145 J mol⁻¹ K⁻¹); δ¹³C_air is the δ¹³C of atmospheric CO₂; and δ¹³C_tree is the site‐representative δ¹³C of the tree rings (resin‐extracted whole wood). Event‐based c _a and δ¹³C_air values in a northern Finnish site, Pallas (67°58′N, 24°7′E, 565 m asl (above sea level), https://gml.noaa.gov/dv/site/PAL.html), the closest site with continuous records for both c _a and δ¹³C_air, were interpolated to daily scale and used. As there were no δ¹³C_air data observed during 2015 to 2019 in Pallas, we estimated δ¹³C_air from c _a and the linear regression between δ¹³C_air and c _a in Pallas (Fig. S5). The average c _a and δ¹³C_air for the formation period of each tree‐ring δ¹³C measurement were used as the input for Eqns (Eqn 3), (Eqn 4), (Eqn 6). Fractionation associated with day respiration was not considered here, considering that day respiration is intensively inhibited (Keenan et al., 2019) and has insignificant impact on $\Delta$ when net assimilation rate is high (Busch et al., 2020).

To test the impact of mesophyll and photorespiratory terms on iWUE_iso, we took following assumptions: with no explicit consideration of g _m and f, by setting g _m = ∞, f = 0, and b = 27‰, which is the simplified model widely used for estimating iWUE_iso; constant g _m of 0.127 mol m⁻² s⁻¹, corrected for all‐sided leaf area (Stangl et al., 2019) and f = 8‰ (Ghashghaie et al., 2003); constant g _m and f = 16‰ (Evans & von Caemmerer, 2013); dynamic g _m varying with T (Methods S3; Sun et al., 2014) and f = 8‰; dynamic g _m and f = 16‰. We compared how the absolute values of iWUE_iso changed across different assumptions, and evaluated whether the intraseasonal variations or interannual trends of iWUE_iso varied between different assumptions. In later sections, iWUE_iso from the simplified model was reported, unless otherwise specified.

Data analysis

To align iWUE_gas, iWUE_EC, and iWUE_iso at intraseasonal scale, we averaged daytime iWUE_gas and iWUE_EC over the period representing each iWUE_iso estimate (Fig. 1). For simplicity, we did not consider changes in carbon allocation rate over time. Phloem transport time from top canopy to breast height was set to 2 d in Hyytiälä and 1 d in Värriö, according to Mencuccini & Hölttä (2010), that is, each period representing an iWUE_iso estimate (Fig. 1) was corrected for these lags. Due to differences in absolute values of iWUE and a more dampened trend in iWUE_iso in comparison with the other records, iWUE series were z‐scored per site, year, and method for intraseasonal comparisons. Pearson correlations between the three iWUE series were calculated per year and site.

For annual values of iWUE, we compared the means of each iWUE series for the growing periods of earlywood, latewood, and the whole tree ring. To examine the significance of interannual iWUE trends, we applied the Mann–Kendall trend test with R package ‘Kendall’ (McLeod, 2011). All statistical analyses were made in R v.4.0.0 (R Core Team, 2020).

Results

Comparison of intraseasonal patterns of iWUE

In Hyytiälä, iWUE_iso aligned in the intraseasonal trends with iWUE_EC and/or iWUE_gas, except for year 2019 (Fig. 2), the year that had the smallest variations in T between June, July, and August (15.7°C, 15.9°C, and 15.2°C, respectively) among the studied years and had a small variability in tree‐ring δ¹³C (−27.0 ± 0.2‰). iWUE_iso aligned clearly better with iWUE_gas than with iWUE_EC in years 2006, 2007, 2009, 2010, 2011, and 2016, which were marked by significantly higher VPD in June compared with other years (0.74 kPa vs 0.56 kPa, P = 0.002). In comparison, iWUE_iso aligned better with iWUE_EC than with iWUE_gas in years 2003, 2004, 2013, and 2018. These years had higher soil temperature in May than the other years (6.9°C vs 5.8°C), although the difference was not significant (P = 0.08).

Fig. 2.

Z‐scores of intrinsic water‐use efficiency (iWUE) of Scots pine from 2002 to 2019 in Hyytiälä. Intrinsic water‐use efficiency was derived from gas exchange (iWUE_gas ), tree‐ring δ¹³C (iWUE_iso ), and eddy covariance (iWUE_EC ) data. For each year, the Pearson correlation coefficient and significance level for iWUE_gas and iWUE_iso (upper right), iWUE_EC and iWUE_iso (lower right), and iWUE_gas and iWUE_EC (lower left) are given: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. Middle day of year (DOY) of the formation period representing each iWUE_iso data point is presented.

In Värriö, iWUE_iso, iWUE_EC, and iWUE_gas aligned with each other in the intraseasonal trends except for year 2012 (Fig. 3). In this year, albeit showing a similar low‐frequency trend, iWUE_EC and iWUE_iso were not significantly correlated. In year 2012, the site experienced a dry period, with lowest precipitation amount in August among the studied years.

Fig. 3.

Z‐scores of intrinsic water‐use efficiency (iWUE) of Scots pine from 2002 to 2019 in Värriö. Intrinsic water‐use efficiency was derived from gas exchange (iWUE_gas ), tree‐ring δ¹³C (iWUE_iso ), and eddy covariance (iWUE_EC ) data. For each year, the Pearson correlation coefficient and significance level for iWUE_gas and iWUE_iso (upper right), iWUE_EC and iWUE_iso (lower right), and iWUE_gas and iWUE_EC (lower left) are given: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. Middle day of year (DOY) of the formation period representing each iWUE_iso data point is presented.

For both sites, the intraseasonal trends of iWUE_iso were clearly more dampened than that of iWUE_EC and iWUE_gas, when absolute values instead of z‐scores were compared. The amplitudes of intraseasonal variations in iWUE_iso, iWUE_gas, and iWUE_EC were 13 ± 6, 24 ± 8, and 29 ± 9 ppm, respectively, in Hyytiälä, and 8 ± 3, 19 ± 6, and 30 ± 14 ppm, respectively, in Värriö. This dampened intraseasonal variation in iWUE_iso corresponded to a lower amplitude of intraseasonal variability in tree‐ring δ¹³C compared with leaf sucrose δ¹³C (1.3‰ vs 4.3‰ in Hyytiälä; 1.3‰ vs 4.2‰ in Värriö; Fig. S6).

Comparison of interannual patterns of iWUE

For each three iWUE estimates, the average absolute value of iWUE did not significantly differ between the growing periods of earlywood, latewood, and the whole ring (P > 0.05, Fig. S7). Hence, for examining interannual iWUE variability, we used the average iWUE values for the whole growing season. Annual iWUE_gas presented a statistically significant increasing trend in both Värriö (1.4 ppm yr⁻¹, P = 0.01) and Hyytiälä (1.5 ppm yr⁻¹, P = 0.03; Fig. 4). Annual iWUE_iso and iWUE_EC did not significantly increase during the studied period in Värriö (0.8 ppm yr⁻¹, P = 0.20 and 0.9 ppm yr⁻¹, P = 0.54, respectively) or Hyytiälä (0.7 ppm yr⁻¹, P = 0.06 and −0.2 ppm yr⁻¹, P = 0.40, respectively; Fig. 4). Annual iWUE_gas, iWUE_iso, and iWUE_EC were not significantly correlated with each other for either site (Fig. 4). Among all tested environmental variables, which were c _a, RH, T, VPD and soil moisture, annual iWUE_gas correlated best with c _a at both sites (Table 1), whereas annual iWUE_iso and iWUE_EC correlated best with VPD (Table 1).

Fig. 4.

Interannual intrinsic water‐use efficiency (iWUE) of Scots pine derived from gas exchange (iWUE_gas ), eddy covariance (iWUE_EC ), and tree‐ring δ¹³C (iWUE_iso ) in (a) Hyytiälä and (b) Värriö from 2002 to 2019. Yearly increase rates of iWUE and P values of Mann–Kendall trend test are given. Pearson correlations between iWUE_iso and iWUE_EC , between iWUE_gas and iWUE_EC , and between iWUE_iso and iWUE_gas were 0.46 (P = 0.05), 0.48 (P = 0.06) and 0.05 (P = 0.87), respectively, in Hyytiälä; and 0.55 (P = 0.16), 0.42 (P = 0.08), and −0.31 (P = 0.46), respectively, in Värriö.

Table 1.

Pearson correlations between environmental variables and annual intrinsic water‐use efficiency (iWUE) of Scots pine.

Site	Variables	c a	RH	SM	T	VPD
Hyytiälä	iWUEgas	0.59*	−0.19 ns	0.11 ns	0.19 ns	0.25 ns
Hyytiälä	iWUEiso	0.53*	−0.62**	−0.50*	0.66**	0.75***
Hyytiälä	iWUEEC	−0.10 ns	−0.76***	−0.54*	0.71***	0.84***
Värriö	iWUEgas	0.69**	0 ns	−0.04 ns	−0.25 ns	0 ns
Värriö	iWUEiso	0.55*	−0.50*	−0.52 ns	0.33 ns	0.58*
Värriö	iWUEEC	0.45 ns	−0.88**	−0.76*	0.86**	0.95***

Intrinsic water‐use efficiency was derived from gas exchange (iWUE_gas), tree‐ring δ¹³C (iWUE_iso), and eddy covariance (iWUE_EC) data. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. The best correlation for each category is in bold. c _a, ambient CO₂ concentration; RH, relative humidity; SM, soil moisture; T, temperature; VPD, vapor pressure deficit.

Impact of f _post, g _m, and f on iWUE_iso

To reconcile differences between iWUE_iso, iWUE_gas, and iWUE_gas, we evaluated the impact of f _post, g _m, and f on iWUE_iso.

f _post differed between the two sites, thus posing site‐specific impacts on iWUE_iso. Among all analyzed carbon pools, leaf WSCs had the lowest δ¹³C values, whereas phloem sucrose had the highest δ¹³C values (Fig. 5). Water‐soluble carbohydrates were significantly ¹³C‐depleted in comparison with sucrose in both leaves (P < 0.001 for both sites) and phloem (P < 0.001 for both sites) (Fig. 5), due to the contribution of pinitol/myo‐inositol (33 ± 6% in leaves, and 18 ± 4% in phloem) with low δ¹³C values (−31.4 ± 0.4‰ in leaves, −30.6 ± 0.6‰ in phloem). In Hyytiälä, tree‐ring δ¹³C was 0.9‰ (P < 0.001), 2.8‰ (P < 0.001), and 0.7‰ (P = 0.03) higher than δ¹³C of leaf sucrose, leaf WSCs, and phloem WSCs, respectively, but 0.4‰ (P = 0.24) lower than δ¹³C of phloem sucrose (Fig. 5a). In Värriö, the δ¹³C differences from tree rings to leaf sucrose, leaf WSCs, phloem WSCs, and phloem sucrose were 0‰ (P = 0.99), 1.6‰ (P < 0.001), 0.5‰ (P = 0.10), and − 0.7‰ (P = 0.08), respectively (Fig. 5b). Common seasonal courses existed in δ¹³C variability of leaf sucrose and tree rings for both sites (Fig. S6). In Hyytiälä, both leaf sucrose δ¹³C and tree‐ring δ¹³C presented an inverse ‘V’ shape variation from June to September (Fig. S6a). In Värriö, δ¹³C of tree ring followed the general increasing trend in leaf sucrose for the whole ring formation period (Fig. S6b). Taken together, f _post from leaf sucrose to tree rings was 0.9‰ in Hyytiälä but 0.0‰ in Värriö (Fig. 5). By subtracting the 0.9‰ offset from tree‐ring δ¹³C for Hyytiälä, iWUE_iso decreased by 11% (Fig. 6a).

Fig. 5.

δ¹³C difference between resin‐extracted tree rings and sugar pools of Scots pine in (a) Hyytiälä and (b) Värriö, together with possible underlying mechanisms. PEPc, phosphoenolpyruvate carboxylase; WSCs, water‐soluble carbohydrates.

Fig. 6.

Boxplot showing the intrinsic water‐use efficiency (iWUE) of Scots pine in (a) Hyytiälä and (b) Värriö. iWUE_iso , iWUE_gas, and iWUE_EC were derived from tree‐ring δ¹³C, gas exchange, and eddy covariance data, respectively. iWUE_iso was estimated from different assumptions of mesophyll conductance (g _m) and photorespiration fractionation factor (f) and corrected by postphotosynthetic δ¹³C alteration (f _post). Horizontal line represents the median, box represents the interquartile range, the tails extend to 1.5 times of the interquartile range, and dots represent outliers that are outside 1.5 times of the interquartile range. Mean ± SD value (ppm) for each category is given at the bottom.

The explicit consideration of g _m and f in the calculation of iWUE_iso did not change the intraseasonal or interannual patterns of iWUE_iso (Fig. S8), but it lowered the absolute values of iWUE_iso for both sites (P < 0.001, Fig. 6). Constant g _m (r = 0.96, P < 0.001) and dynamic g _m (r = 0.97, P < 0.001) assumptions both produced iWUE_iso that linearly correlated with iWUE_iso from the simplified model. Similarly, iWUE_iso from the assumptions f = 8‰ (r = 0.93, P < 0.001) and f = 16‰ (r = 0.92, P < 0.001) both linearly correlated with iWUE_iso from the simplified model.

iWUE_gas, iWUE_iso, and iWUE_EC differed significantly in their absolute values for both sites (P < 0.05), in the following order: iWUE_iso > iWUE_gas > iWUE_EC (Fig. 6). In Hyytiälä, iWUE_gas and iWUE_EC were on average 19% and 39%, respectively, lower than iWUE_iso, whereas in Värriö, the differences were 7% and 41%. If applying constant g _m and f = 8‰ and correcting iWUE_iso by f _post, the differences from iWUE_iso to iWUE_gas and iWUE_EC decreased to 1% and 25%, respectively, in Hyytiälä; and to 3% and 39%, respectively, in Värriö (Fig. 6).

Discussion

Validity of tree‐ring δ¹³C for intraseasonal iWUE estimates

We observed a general agreement in the intraseasonal trends of iWUE_iso, iWUE_gas, and iWUE_EC (Figs 2, 3), which clearly supports the validity of using tree‐ring δ¹³C for estimating iWUE at high‐resolution intraseasonal scale. This is partly in line with Michelot et al. (2011), where the latewood section of deciduous Quercus petraea recorded well‐seasonal variations in iWUE. Nevertheless, our results indicate that earlywood of boreal conifers can also be a good recorder for iWUE_iso. This is in line with Kress et al. (2010), although earlywood has been removed by default in some earlier annual iWUE_iso studies (Waterhouse et al., 2004) in case of possible use of reserves for earlywood growth (McCarroll & Loader, 2004). Our conclusion on the suitability of both earlywood and latewood sections for iWUE_iso studies is based on the following observations: variations of iWUE_iso during the earlywood growing period aligned with that of iWUE_EC and/or iWUE_gas (Figs 2, 3); iWUE_iso values averaged for the growing periods of earlywood, latewood, and the whole ring were not significantly different (Fig. S7); we did not observe a previous‐year reserve δ¹³C signal in earlywood (Fig. S9).

Factors affecting iWUE_iso estimates

The impact of f _post, that is, overall apparent isotope fractionation between tree rings and new assimilates, on iWUE_iso has been addressed in earlier studies with a correction factor that has been obtained by measuring δ¹³C difference between tree rings and leaf water‐soluble organic matter (Frank et al., 2015) or total organic matter (Belmecheri & Lavergne, 2020). Considering that leaf bulk materials have varying and significant δ¹³C offsets from new assimilates and that sucrose accurately records assimilate δ¹³C (Tang et al., in press), our comparison between tree rings and leaf sucrose presents a more precise quantification of f _post. From this comparison, we defined f _post of 0.9‰ for Hyytiälä but 0.0‰ for Värriö (Fig. 5), which can reduce iWUE_iso by 11% in Hyytiälä (Fig. 6).

In our study, the site‐specific f _post values were not associated with site‐specificity in the use of previous‐year reserves for tree‐ring growth (Fig. S9) but may be related to site‐to‐site differences in postphotosynthetic metabolisms. First, f _post can be partly ascribed to the remobilization of ¹³C‐enriched transitory starch during the night (Gessler & Ferrio, 2022), which causes an overall ¹³C‐enrichment in phloem sucrose relative to leaf sucrose (Fig. 5). The proportion of starch‐derived ¹³C‐enriched sucrose in breast‐height phloem depends on the phloem transport velocity and tree height (Gessler & Ferrio, 2022), which differed between Hyytiälä and Värriö (Table S1). Second, f _post can be affected by the mixing of assimilates formed at different canopy heights, which may have a δ¹³C gradient of up to 8‰ (Bögelein et al., 2019). Apparently, the vertical mixing of assimilates should have less impact on f _post for Värriö compared with Hyytiälä, on account of a sparser canopy density and therewith lower intracanopy light gradients and assimilate δ¹³C gradients in Värriö (Fig. S1). Third, f _post is partly determined by the activity of phosphoenolpyruvatecarboxylase (PEPc), which favors ¹³C when catalyzing CO₂ refixation in stems (Farquhar, 1983) and thus enriches the organic matter. Since PEPc activity varies with temperature (Chinthapalli, 2003), which differed remarkably between our study sites (Table S1), f _post may have differed between the two sites. Fourth, f _post may be affected by the ratio between ¹³C‐enriched cellulose and ¹³C‐depleted lignin (Loader et al., 2003), which may vary between the two sites with different environmental conditions (Kilpelainen et al., 2005).

While the explicit consideration of g _m and f generally reduced iWUE_iso values by up to 14% (Fig. 6), it did not change the intraseasonal (Fig. S8) or interannual trends of iWUE_iso. This is in contrast to earlier studies that have reported the potential of g _m and f for modulating isotopic discrimination in long term (Keeling et al., 2017; Lavergne et al., 2022). Our results thereby indicate that the simplified model for iWUE_iso is able to capture the intraseasonal trends of iWUE_iso for our study species, albeit the accuracy of absolute values of iWUE_iso may benefit from a better understanding of g _m dynamics in future studies. However, our results are in contrast to the report of Gimeno et al. (2021) for Eucalyptus, where a better fit between iWUE_gas and iWUE_iso was obtained after incorporating g _m into the calculations of iWUE_iso. This contradiction probably comes from the use of a constant g _m value or temperature‐dependent g _m in this study, considering that g _m can vary with photosynthetic rate and water stress (Schiestl‐Aalto et al., 2021).

Environmental and physiological controls on iWUE

Differences in intraseasonal patterns of iWUE_iso, iWUE_EC, and iWUE_gas existed in some cases, likely associated with environmental control. For instance, the intraseasonal pattern of iWUE_iso did not align with that of iWUE_gas and iWUE_EC for year 2019 in Hyytiälä (Fig. 2), possibly due to a dampened low‐frequency trend in tree‐ring δ¹³C data governed by low variability in T this year. It demonstrates that a certain degree of variability in environment conditions and tree‐ring δ¹³C data within a growing season is crucial for reliable estimation of intraseasonal iWUE_iso. Furthermore, the intraseasonal pattern of iWUE_EC is impacted by water stress. For example, for years with higher VPD in Hyytiälä (Fig. 2) and for the dry year 2012 in Värriö (Fig. 3), iWUE_EC aligned less well with iWUE_iso than iWUE_gas did. This is probably because iWUE_EC integrates the species‐specific physiological response to water stress (Yi et al., 2019) across various plant species over the stand (Table S1). Moreover, high spring temperatures tend to cause high uncertainties in gas exchange measurements, as the measuring shoot may recover faster than the whole canopy, resulting in divergent variability in iWUE_gas for years 2003, 2004, 2013, and 2018 in Hyytiälä (Fig. 2).

Environmental drivers for annual iWUE were diverse across leaf (iWUE_gas), tree (iWUE_iso), and ecosystem (iWUE_EC) scales (Table 1). At leaf level, rising c _a tightly regulates iWUE_gas (Table 1) by enhancing assimilation rate (Streit et al., 2014) and reducing g _s (Brodribb et al., 2009). However, the c _a effect on whole‐tree level iWUE_iso was reduced (Table 1), because the prevalent photosynthesis at lower canopy is limited by RuBP‐regeneration and less sensitive to rising c _a compared with Rubisco‐limited photosynthesis at top canopy (Yang et al., 2020). At ecosystem level, the c _a effect was further dampened (iWUE_EC in Table 1) due to species‐specific responses of iWUE to c _a (Marchand et al., 2020) and changes in leaf area and soil water savings (Lavergne et al., 2019). Instead of c _a, VPD dominated changes in annual iWUE_iso and iWUE_EC (Table 1), as also identified in Kannenberg et al. (2021) and Zhang et al. (2019), respectively. This is probably because carbon uptake decreases less than g _s with increasing VPD (Zhang et al., 2019). We also note that environmental control on iWUE varies between interannual and intraseasonal scales and from year to year, possibly related to changes in leaf area (Launiainen et al., 2016) and soil moisture (Beer et al., 2009).

The strengths and weaknesses of the three iWUE estimation methods

Each of the three iWUE estimation approaches has its own strengths and weaknesses (Table 2). Eddy covariance measurements have the merits in the manner that global networks, such as FLUXNET (Baldocchi et al., 2001), provide EC data at varying temporal coverage up to decades (Medlyn et al., 2017). However, as an ecosystem‐level integrated signal, EC data are not able to discern species‐specific iWUE responses (Yi et al., 2019). Meanwhile, iWUE_EC estimates have multiple sources of uncertainties (Knauer et al., 2018), including within‐canopy gradients, nontranspiratory water fluxes, energy balance nonclosure, issues in NEE partitioning, aerodynamic conductance, and meteorological differences between measurement height and canopy surface. Some of these uncertainties are site‐specific and may vary with time. For example, within‐canopy gradient, which may result in lower iWUE_EC with a higher contribution of fluxes from the understory (Domingues et al., 2007; Sellin et al., 2010), has a site‐specific impact. iWUE_EC calculated from subcanopy flux data was 30.5% lower than the gradient‐integrated iWUE_EC in Hyytiälä, whereas the difference was negligible in Värriö (6%). This site‐specific trait rises from an open stand structure in Värriö (Fig. S1), which results in lower intracanopy light gradients and a higher coupling of air exchange to the atmosphere relative to a closed canopy structure in Hyytiälä (Wieser et al., 2018). Moreover, iWUE_EC is underestimated due to the contribution of nontranspiratory water fluxes (Eqn 2), mainly soil evaporation in this study as canopy evaporation should be minimal after excluding the time periods following precipitation. This underestimation is in the order of 15%, assuming soil evaporation accounted for half of forest floor ET, which contributed to c. 30% of total ecosystem ET at our sites (estimated from subcanopy EC data and Launiainen et al., 2005). However, this proportion would change with the increase in leaf area index in Hyytiälä (Table S1) but stay almost constant over the years with roughly unchanged leaf area index in Värriö.

Table 2.

Method comparison between intrinsic water‐use efficiency (iWUE) derived from gas exchange (iWUE_gas), tree‐ring δ¹³C (iWUE_iso) and eddy covariance data (iWUE_EC).

	iWUEgas	iWUEiso	iWUEEC
Factors affecting iWUE estimates	(1) Possible chamber artifacts, for example, occasional mechanical flaws and inconsistence in measurement systems over years (2) Uncertainties arising from limited sampling coverage	(1) Uncertainties arising from the photosynthetic discrimination model due to, for example, limited knowledge of mesophyll conductance (2) δ13C alteration from leaf assimilates to tree rings, due to vertical mixing of assimilates, postphotosynthetic isotopic fractionation, and possible use of reserves	(1) Energy balance nonclosure (2) Uncertainties in net ecosystem CO2 exchange (NEE) partitioning (3) Within‐canopy gradient impacted by stand structure (4) Contribution of nontranspiratory water fluxes (5) Other sources of uncertainties, for example, aerodynamic conductance
Strength of the method	(1) High temporal resolution data, for example, at daily or diurnal scale (2) Information on species‐specific dynamics	(1) Signal is archived in tree materials and can be retrieved for sites (periods) where (when) no instrumental data are available years after tree‐ring formation (2) Suitable for long‐term iWUE reconstructions (3) No need for on‐site δ13C measurements (4) Information on species‐specific dynamics	(1) High temporal resolution data (2) Continuous and long‐term EC records are available for many sites globally
Weakness of the method	(1) Labor‐intensive and requires accessing leaves of tall trees (2) Restricted in their spatial and temporal coverage (3) Low consistency in chamber systems in long term (4) Chamber systems are prone to mechanical flaws	(1) May be influenced by other sources of isotopic discrimination	(1) Can not resolve species‐specific leaf‐ or tree‐scale dynamics (2) Subject to noise and errors in, for instance, unclosed energy balance problem and NEE partition

Leaf gas exchange measurements have the advantages of tracing instantaneous changes in iWUE, but this method is labor‐intensive and requires accessing leaves of tall trees (Yi et al., 2019). Meanwhile, even though a global compilation of gas exchange measurements is available (Lin et al., 2015), there is currently a lack of long‐term continuous datasets. Also, iWUE_gas estimates are subject to uncertainties regarding chamber artifacts, limited sampling coverage, and low consistency in measurement systems in long term. Occasional mechanical flaws, for example leaks, and possible damages to the measuring shoots may affect the observed seasonal cycle. Limited sampling coverage on one or several measuring shoots on the top canopy may induce uncertainties, for example, during warm springs. Changes in measurement systems over years may bias the interannual trend of iWUE_gas. Nevertheless, not including boundary layer conductance in calculating iWUE_gas overall has a limited impact on absolute values (Seibt et al., 2008) and intraseasonal patterns of iWUE_gas (Figs 2, 3).

Tree‐ring δ¹³C can be retrieved even decades or centuries after tree‐ring formation without laborious work on site (Cernusak, 2020). Hence, a major advantage of tree‐ring δ¹³C records is their potential for reconstructing long‐term iWUE_iso. Accuracy of iWUE_iso estimates can be further improved by a better understanding of f _post, g _m, and f. More importantly, our study shows that iWUE can be obtained from tree‐ring δ¹³C at intraseasonal scale with reasonable effort using LA‐IRMS, extending the application of this iWUE estimation method from annual resolution to intraseasonal resolution. This finding is inspiring in the way that it provides a valuable method for intraseasonal iWUE estimates, especially for sites and periods where and when no gas exchange or EC data are available.

Conclusions

This work presented the first comparison between intraseasonal and interannual iWUE signal derived from leaf gas exchange, tree‐ring δ¹³C, and EC data, resting on a unique set of 18‐yr‐long records in two boreal forest sites. The alignment in intraseasonal iWUE trends across different methods demonstrated the reliability of tree‐ring δ¹³C derived intraseasonal iWUE estimates. This result is of special significance to studies, which seek to detect intraseasonal tree physiological dynamics in terms of iWUE but with no access to instrumental data. The absolute values of iWUE across different datasets can be reconciled by taking into account an overestimation in iWUE_iso of up to 11% due to f _post, and of up to 14% due to nonexplicit consideration of mesophyll and photorespiratory effects. A significant increasing interannual trend existed in iWUE_gas, but not in iWUE_iso or iWUE_EC, for both sites, possibly resulting from a predominant control of c _a on iWUE_gas but VPD control on iWUE_iso and iWUE_EC. We encourage more across‐method comparisons of iWUE at various temporal and spatial scales in the future. Such studies will not only deepen our understanding of how trees physiologically adapt to climate change but also provide insights into ecological models in respect of linking ecological information across scales.

Competing interests

None declared.

Author contributions

KTR‐G and YT planned and designed the study. YT, PS‐A, KTR‐G, ES and GY conducted fieldwork. YT prepared sugar samples for δ¹³C analysis. GY prepared tree‐ring samples for δ¹³C analysis. ES conducted bulk and LA‐IRMS δ¹³C analysis. MS conducted HPLC‐IRMS δ¹³C analysis. PK calculated eddy covariance and leaf gas exchange data. PS‐A modeled tree‐ring growth via CASSIA. YT and TJ conducted xylogenesis observations. YT conducted data analysis. YT was responsible for writing the manuscript. JB and all other authors contributed to the interpretation of data and the writing of the manuscript at various stages.

Supporting information

Fig. S1 Locations and photographs of the study sites in Finland.

Fig. S2 Intraseasonal tree‐ring δ¹³C of Scots pine from 2002 to 2019 in Hyytiälä.

Fig. S3 Intraseasonal tree‐ring δ¹³C of Scots pine from 2002 to 2019 in Värriö.

Fig. S4 Comparison of growth curves of Scots pine from CASSIA model and from xylogenesis observations.

Fig. S5 Relationship between event‐based δ¹³C of atmospheric CO₂ (δ¹³C_air) and concentration of ambient CO₂ (c _a) in Pallas.

Fig. S6 Comparison of δ¹³C signal in leaf sugars, phloem sugars, and resin‐extracted wood of Scots pine in Hyytiälä and Värriö in 2018.

Fig. S7 Boxplot showing the intrinsic water‐use efficiency of Scots pine averaged for the growing periods of earlywood, latewood, and whole ring.

Fig. S8 Boxplot showing correlations between intraseasonal intrinsic water‐use efficiency of Scots pine derived from different methods under different mesophyll and photorespiratory assumptions.

Fig. S9 Across‐border correlations in tree‐ring δ¹³C of Scots pine, which denotes the degree of use of previous‐year reserves.

Methods S1 LA‐IRMS systems.

Methods S2 Tracheid growth curves from xylogenesis observations and CASSIA model.

Methods S3 Dynamic g _m assumption.

Table S1 General description, site characteristics, and data availability for our study sites.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

The data that support the findings of this study are openly available in Figshare at doi: 10.6084/m9.figshare.21267963.v1.