Drought impacts on tree carbon sequestration and water use - evidence from intra-annual tree-ring characteristics.

The impact of climate extremes on forest ecosystems is poorly understood but important for predicting carbon and water cycle feedbacks to climate. Some knowledge gaps still remain regarding how drought-related adjustments in intra-annual tree-ring characteristics directly impact tree carbon and water use. In this study we quantified the impact of an extreme summer drought on the water-use efficiency and carbon sequestration of four mature Norway spruce trees. We used detailed observations of wood formation (xylogenesis) and intra-annual tree-ring properties (quantitative wood anatomy and stable carbon isotopes) combined with physiological water-stress monitoring. During 41 d of tree water deficit, we observed an enrichment in 13 C but a reduction in cell enlargement and wall-thickening processes, which impacted the anatomical characteristics. These adjustments diminished carbon sequestration by 67% despite an 11% increase in water-use efficiency during drought. However, with the resumption of a positive hydric state in the stem, we observed a fast recovery of cell formation rates based on the accumulated assimilates produced during drought. Our findings enhance our understanding of carbon and water fluxes between the atmosphere and forest ecosystems, providing observational evidence on the tree intra-annual carbon sequestration and water-use efficiency dynamics to improve future generations of vegetation models.

Introduction

Understanding the response of forest ecosystems to drought is increasingly important in order to assess the magnitude and impacts of a warming climate more accurately (Werner et al., 2021). By regulating evaporative cooling (Bala et al., 2007) and sequestrating atmospheric carbon (Bonan, 2008), forests play a major role in the global carbon and water cycles (Betts et al., 2007; Friedlingstein et al., 2020). However, the contribution of forests to the mitigation of climate change remains uncertain (Reichstein et al., 2013; Friedlingstein et al., 2014) because forest ecosystems themselves are also vulnerable to climate warming, and current dynamic global vegetation models (DGVMs) still need to be refined to take proper account of certain processes, such as those related to secondary growth (Zuidema et al., 2018).

Tree physiology intimately connects the carbon and water cycles via two distinct processes. At the stomatal level, carbon is absorbed from the atmosphere in exchange for water, while assimilated carbon is fixed in wood at the cambial level during tree growth. During the process of carbon uptake in the stomata, trees unavoidably lose large amounts of water, contributing, on average, c. 60% of the total annual evapotranspiration at the ecosystem level (Schlesinger & Jasechko, 2014). At the same time, through the process of wood formation, forests capture a substantial amount of anthropogenic carbon emissions (Pan et al., 2011; Harris et al., 2021; Xu et al., 2021) and serve as one of the main terrestrial carbon reservoirs. However, one of the main reasons for the uncertainty concerning the role of trees in the mitigation of climate change is the lack of empirical data with a high temporal resolution. These kinds of data are needed to improve the understanding of processes such as growth (Lovenduski & Bonan, 2017; Babst et al., 2020), and they will therefore help us to represent these processes in current DGVMs appropriately (Zuidema et al., 2018; Friend et al., 2019). For instance, most DGVMs still estimate growth from the difference between photosynthesis and plant respiration, without an explicit representation of growth processes themselves (Friend et al., 2019). However, there is clear evidence that the environmental sensitivity differs between photosynthesis (source) and secondary growth (sink) (Fatichi et al., 2013; Körner, 2015). Recent efforts using tree rings to benchmark the impact of extreme drought events on forest productivity (Anderegg et al., 2015; Trotsiuk et al., 2020) not only demonstrated the inability of current models to account for the pervasive impact on the global carbon cycle, but also emphasized the need for a more mechanistic inclusion in climate‐vegetation models of the full cascade of causes–effects and legacies that influence tree growth performance. A tree‐centered approach – the study of individual trees as a main source of information for understanding variability in growth (Sass‐Klaassen et al., 2016) – could contribute to a more in‐depth understanding of causal processes occurring within trees in response to environmental changes.

Under limited water conditions, like those forecast in future climate scenarios (IPCC, 2021), tree physiology is strongly disturbed, which indirectly affects the mitigation capacity of forest ecosystems to sequestrate carbon. For example, based on the measurement of fluxes at the ecosystem level, the 2003 summer drought that occurred in central Europe caused a 30% reduction in net primary productivity, temporarily turning some terrestrial ecosystems into a net carbon source (Ciais et al., 2005). However, as a result of the high sensitivity of cell division and enlargement to both cell turgor and sugar availability (Fatichi et al., 2013; Körner, 2015; Cabon et al., 2020b; Peters et al., 2020), wood formation could be constrained while photosynthesis still occurs. Owing to the complexity of the processes involved in wood formation – which is characterized by cells undergoing several climate‐sensitive phases until they become functional (Rathgeber et al., 2016) – a drought event can also have several repercussions on the anatomical structure of the forming wood (Eilmann et al., 2011; Fonti et al., 2013). This could imply that the carbon fixed in wood does not represent the net amount between carbon absorbed at the leaf level and carbon respired; consequently, the impact of extreme droughts on tree carbon budgets could be systematically underestimated.

Intrinsic water‐use efficiency (iWUE) is the associated ratio of carbon uptake to stomatal conductance (Farquhar et al., 1982) and it can be measured using stable carbon isotope ratios (i.e. δ13C). The anthropogenic atmospheric CO2 fertilization partially counteracts water dependencies by increasing leaf‐scale photosynthesis and iWUE, implying that trees need less water to absorb the same amount of carbon. Indeed, during the 20th century, broadleaf and coniferous trees across Europe displayed increases in iWUE of 14 ± 10% and 22 ± 6%, respectively (Frank et al., 2015), fostering a positive terrestrial carbon sink (Walker et al., 2020). However, increases in iWUE are also related to the active role of trees in reducing stomatal conductance as a strategy to cope with water limitations (see Shestakova & Martínez‐Sancho, 2021 and references therein). Although several studies have demonstrated that this effect occurs in dry years (Battipaglia et al., 2014; Olano et al., 2014), still little is known about how iWUE varies around a drought event and its recovery dynamics.

Despite potential mismatches based on meteorological quantifications of the intensity of a drought event (Slette et al., 2019; Zang et al., 2020), the earlier‐mentioned physiological processes and their climatic thresholds also depend on the timing, duration and strength of the event. Further, a single drought event might impact individual trees differently according to their site‐specific conditions (e.g. soil depth, topography), species characteristics (e.g. rooting system, tree height, tree age, drought adaptations) and physiological status (Anderegg et al., 2013; Bennett et al., 2015). The severity of a given water stress as it is experienced by an individual tree is indeed a primary factor determining stomatal regulation (Klein, 2014), as well as the processes of cell division, enlargement and wall thickening (Cuny et al., 2014; Rossi et al., 2014). This ultimately affects the amount of water transpired and the carbon sink capacity. The different environmental sensitivities of wood formation and photosynthesis, including their potential interactions, together with potential iWUE adjustments and a discrepancy between meteorological drought and physiological water stress, might partly explain the observed discrepancies between DGVMs and tree‐ring observations (Babst et al., 2013; Klesse et al., 2018).

In this study, we used highly resolved data on xylogenesis, quantitative wood anatomical parameters and stable carbon isotopes to assess the impact of an extreme summer drought on the resilience of the intra‐annual carbon sequestration and water‐use dynamics of mature Norway spruce trees growing in a natural subalpine forest. We hypothesized that: growth processes during drought, including cell formation and anatomical characteristics, would have a direct impact on the ability of trees to sequestrate carbon; tree water losses during drought would be dramatically reduced as a consequence of fast physiological adjustments; and resilience patterns would be distinct for anatomical parameters and stable carbon isotope ratios as a result of differences in source and sink dynamics during and after drought.

Materials and Methods

Study site and selected growing seasons

We conducted our study at the Lötschental tree growth monitoring transect (https://www.wsl.ch/loetschental‐monitoring), where most of the required environmental and growth‐related data were available. We selected a drought‐exposed site located at the valley bottom (46°23′40′′N, 7°45′35′′E), at an elevation of 1360 m above sea level (asl). We analyzed four adult individuals of Norway spruce (Picea abies (L.) Karst.) (Supporting Information Table S1), which is a central European species that has been classified as drought sensitive (Lebourgeois et al., 2010).

We selected two years with distinctly different meteorological conditions: 2014 as a ‘normal’ year not significantly different from the climatic long‐term mean, and 2015 as a year characterized by an extreme summer drought in central Europe (Orth et al., 2016). According to the April–September mean temperature and precipitation sum, the two selected growing seasons were within the 1961–2016 average (2014) and at its hot and dry margin (2015) (Fig. S1). The anomalies observed in 2015 corresponded to precipitation that was lower by almost two‐thirds (24 vs 60 mm) and temperature that was 4°C higher (18.2 vs 14.2°C) than the long‐term July mean.

Monitoring environmental and physiological water stress

Highly resolved environmental and tree physiological water stress data were used to track the intensity of drought over time. Site environmental measurements aggregated during daytime hours (06:00–22:00 h Central European Time) included: maximum air temperature (T air, °C) and average relative humidity (RH, %) at 2 m above ground (U23‐002 Hobo Pro; Onset, Bourne, MA, USA), and average soil volumetric water content at 10 cm depth (SVWC, %; ECH2O EC‐5; Meter Group, Pullman, WA, USA). Furthermore, vapor pressure deficit (VPD, kPa) was also calculated based on these data. Tree physiological water stress was quantified for the local vegetation using the normalized difference vegetation index (NDVI) and the tree water deficit (TWD) measured in the stem (Zweifel et al., 2005). Specifically, the crown stress level was assessed using 4 d composite NDVI values retrieved from the Moderate Resolution Imaging Spectroradiometer (MODIS/Terra) of 250 m resolution centered on the study site. Relative stem shrinkage caused by low water potentials during drought and measured as stem radius variation is commonly referred to as TWD, and it represents the physiological response of trees to changing climate conditions in terms of growth and water status (Zweifel et al., 2005). When TWD lasts more than 24 h it means that the tree could not recover its water status during the night. Stem radius variations were measured at 1.3 m stem height with a sub‐hourly resolution using high‐precision point dendrometers (DR model; Ecomatik, Munich, Germany).

Timing of individual cell formation

To determine the period that each individual wood cell spent in the differentiating phases of cell enlargement and wall thickening, weekly microcores were collected at about 1–1.5 m stem height using a Trephor increment puncher (Rossi et al., 2006a) during two growing seasons (n = 244 samples). All samples were stored in 70% ethanol. Microcores were processed to obtain microsections that allow retrospective observations. The samples were first embedded in paraffin with a HistoCore PEARL tissue processor (Leica Biosystems, Nussloch, Germany) to stabilize the unlignified tissues. The microcores were then cut at a thickness of 7 μm using a rotary microtome (RM2245; Leica Biosystems). The sections obtained were then stained with a solution of Astra blue and Safranin and permanently mounted on glass slides for digital imaging in both bright and polarized light (Axio Scan.Z1; Zeiss). Observations were then performed along three radial files per sample by counting the number of cells in the cambial, enlargement, wall‐thickening and mature phases, according to Rossi et al. (2006b).

The changes in the number of cells in the cambium and in the consecutive phases of xylem differentiation throughout the growing season were analyzed by fitting generalized additive mixed models (GAMMs) assuming a quasi‐Poisson distribution of residuals (Wood, 2006). Generalized additive mixed models were fitted using the package mgcv in R (R Development Core Team, 2015). Diagnostic plots were checked for heteroscedasticity, nonnormal distribution and autocorrelation of the residuals from all GAMMs and were corrected when required. Day of year (DOY) was the main predictor and tree ID was included as a random effect to account for nonindependent data (i.e. repeated measures over time).

Kinetics, including the timing (date of initiation and cessation) and duration (number of days), of the phases (cambial, enlargement, wall thickening and mature phases) for each cell in each ring were calculated in a similar manner to the approach of Cuny et al. (2013) and Pérez‐de‐Lis et al. (2021) (see details in Notes S1; Fig. S2). The outputs of cell kinetics of the two years (2014 and 2015) were compared with parametric t‐tests. Afterwards, phase durations were combined with wood anatomical measurements to calculate rates of enlargement and wall thickening.

Cell anatomy and rates of cell formation

One 1‐cm‐diameter increment core was collected at 1.3 m stem height in 2020 from each of the four trees included in the section described earlier. Cell anatomical measurements of mature tracheids were performed from high‐resolution images (2.26 μm pixel−1) of 12‐μm‐thick transversal microsections processed following the same method as for the microcores. Microsections were obtained from one longitudinal half of the collected cores. Using the software Roxas (von Arx & Carrer, 2014) we measured the transversal cell lumen area (CLA), cell wall area (CWA), total cell area (CTA = CLA + CWA), cell radial wall thickness (CWTrad), cell radial lumen diameter (CLArad), and cell radial diameter (Crad = 2 × CWTrad + CLArad), and positional coordinates of each tracheid included in the annual rings from 2014 and 2015. The R package raptor (Peters et al., 2018) was then used to assign each tracheid to its radial file and to obtain an average value for each position. Afterwards, CTA, CWA, Crad and CLA were fitted with GAMMs to assess their dynamics across the ring width, with DOY as the main predictor and tree ID as a random effect. To calculate the daily rates of cell enlargement and wall thickening, we subsequently divided CTA and CWA of each cell by the corresponding time spent in the enlargement (dE) and cell wall thickening (dW) phases, as described in Cuny et al. (2013) (see Notes S1; Fig. S2).

Stable carbon isotope measurements and iWUE calculations

Intrinsic water‐use efficiency (iWUE) calculations were derived from high‐resolution measurements of tree‐ring δ13C. Samples for isotope analysis were prepared using tangential cuts from the second longitudinal half of the 1‐cm‐diameter cores. The thickness of the tangential sections (60 μm in the earlywood and 40 μm in the latewood) were selected to account for tracheid size changes across the ring width. Cuts were performed with a rotary microtome (RM2245; Leica) surrounded by two digital cameras to guide the orientation of the cuts and document their thickness. A total of 398 tangential samples were collected (c. 50 consecutive intra‐annual samples per ring and tree). Subsequently, holocellulose was extracted from each wood sample (Boettger et al., 2007), homogenized using an ultrasonic treatment (Laumer et al., 2009), freeze‐dried, and then inserted into tin capsules for combustion to CO2 using a Euro Elemental Analyzer (EuroVector, Milan, Italy). Stable carbon isotope ratios were measured by isotope ratio mass spectrometry (IRMS) using a Delta V Advantage IRMS (Thermo Fisher Scientific, Bremen, Germany) with a precision of 0.02‰ (derived from parallel measurements of reference material). Carbon isotope ratios were reported with the delta (δ) notation indicating the 13C/12C ratio of a sample relative to that of the international standard VPDB (Vienna Pee Dee Belemnite) expressed in per mil.

Using the δ13C values of the cellulose samples of the microsections (δ13CS), we calculated the discrimination against 13C (∆13C) during carbon diffusion and fixation by plants, the estimated leaf intercellular CO2 concentration (C i), and the iWUE following the formulas described in Notes S2. Afterwards, the δ13 CS, ∆13C, C i and iWUE values were then fitted with GAMMs to assess their dynamics across the ring width. The position occupied by the tangential section was included as a predictor and tree ID as a random effect.

Space‐for‐time conversion of cell features

To align the cell anatomical features, their corresponding rate of development and the iWUE measurements along the time axis, we translated the cell positions (analyzed as % of the ring width) into the corresponding time of cell formation (analyzed as DOY) using the calculated time of cell formation (see Notes S1; Fig. S2; space‐for‐time conversion adapted from Pérez‐de‐Lis et al., 2021). Specifically, the outcomes of the GAMMs of the anatomical features (cell area, cell wall area and cell lumen area) and intra‐annual δ13C‐derived variables (δ13C, ∆13C, C i and iWUE) were analyzed accounting for their relative position in the ring width by using percentage of the ring width as a predictor. The space‐for‐time conversion could then be performed because we knew the exact position of the cell within the tree ring and could extract the time it spent in the enlargement and cell wall‐thickening phases from the time models. In the specific case of δ13C‐derived variables, we used the timing of the cell wall‐thickening phase, as we assumed that the cell isotopic signature mainly represented the content of the secondary cell wall and the environmental drivers that contributed to its formation.

Calculations of tree‐level daily carbon sequestration

The intra‐annual cell anatomical assessments provided the basis to upscale tree daily aboveground woody biomass production (kg carbon) to the tree level. Following Cuny et al. (2015), the daily woody biomass production per tree was calculated based on the sum of the rates of wall deposition occurring in all cells growing during a given day. The upscaling to the tree level was performed by considering the number of radial cell files composing the tree ring and the tree allometry, and by accounting for the wall density and its carbon content (see detailed information in Notes S3). All statistical analyses were performed in the R environment v.4.1.2 (R Development Core Team, 2015) and graphics were designed using the R package ggplot2 (Wickham, 2009).

Results

The 2015 summer drought at the study site

The 2015 summer drought was identified as a distinct anomaly from the long‐term annual mean values of all considered climate variables, as well as from the 2014 growing season values (Figs 1a, S1). For the period June–August, the year 2015 displayed an increase of 7.4°C in maximum daytime temperatures, a decrease of 18.6% in average RH, an increase of 0.8 kPa in average VPD, and a decrease of 7.3% in average soil water content in comparison to the same period in 2014 (Fig. 1). The extreme drought in 2015 clearly affected the local vegetation, which showed an NDVI reduction during the period 19 June to 6 August, compared with both the long‐term average (Fig. S3) and the year 2014 (Fig. 1b). Moreover, dendrometer data indicated 41 consecutive days of tree stem shrinkage during 2015, from 19 June to 30 July, with the maximum amount of shrinkage occurring on 19 July (Figs 1c, S4). This period will be referred to as the drought period throughout the ‘Results’ section.

Fig. 1

Environmental and physiological conditions during the 2014 (normal) and 2015 (dry) growing seasons. (a) Maximum daytime air temperature (T air), average daytime relative humidity (RH), average daytime vapor pressure deficit (VPD) and soil volumetric water content (SVWC). (b) Normalized difference vegetation index (NVDI). Dots represent 4 d composite NDVI values. (c) Daily dendrometer‐derived tree water deficit (TWD), measured as a negative change in stem radius.

Changes in kinetic cell formation

The analyses of the weekly collected microcores revealed that the extreme summer drought of 2015 strongly altered the kinetics of cell formation and, hence, the tree‐ring formation. The drought was mirrored by a significant reduction in the number of cells in each phase of cell formation compared with the previous growing season (Fig. 2). Indeed, the decreases in the number of cells in the cambial phase (−31% on average) and wall‐thickening phase (−30%) were concomitant with the occurrence of the TWD, whereas TWD effects were visible earlier in the case of cells in the enlargement phase (−36%). Its phenology was less affected, however, as the timing of the beginning of the cambial, enlargement, wall‐thickening and mature phases was rather similar in the two years (Fig. 2; Table 1). A similar final number of cells was produced in the two years. However, the kinetics of cell formation were different between years (Fig. S5). In 2014, the duration of the cell enlargement phase increased slightly from the first to the last tracheid (Fig. S5a,c). This trend was even more evident for the duration of the cell wall‐thickening phase. In comparison to 2014, the phases of cell enlargement and wall thickening lasted longer during the 2015 drought period (Fig. S5b,d). However, the cell enlargement phase was shorter compared with 2014 values for the cells formed after the 2015 drought (Fig. S5b,d). Thus, the extreme summer drought influenced 60% of the 2015 ring by affecting at least one of the phases of cell development.

Fig. 2

Comparison of the number of cells in the cambial, enlargement, wall‐thickening and mature phases in the two study years for the four studied Norway spruce trees. Colored lines indicate the fitted values obtained from generalized additive mixed models and the corresponding shaded colored areas indicate the 95% confidence intervals. The gray band indicates the period when there was a tree water deficit (TWD; negative change in stem radius in μm) for more than 24 h during the extreme summer drought in 2015.

Table 1

Critical dates for each phase of wood formation of Norway spruce, expressed as mean day of year.

	2014	2015
bE	138 ± 1.7	138.3 ± 6.4
bW	155 ± 2.9	158 ± 0.0
bM	165 ± 10.5	162 ± 3.8
cE	260 ± 2.9	251 ± 4.3
cW	285 ± 3.5	290 ± 5.0
dE	121 ± 2.2	113 ± 9.0
dW	130 ± 4.6	131 ± 5.0
dX	147 ± 1.9	151 ± 7.4

bE, beginning of the cell enlargement phase; bW, beginning of the cell wall‐thickening phase; bM, beginning of the mature cell phase; cE, cessation of the cell enlargement phase; cW, cessation of the cell wall‐thickening phase; dE, duration of the cell enlargement phase; dW, duration of the cell wall‐thickening phase; dX, total duration of wood formation.

A t‐test was performed to assess differences between the two years; bold values indicate significant differences (P < 0.05).

Changes in the rate of cell formation and anatomical characteristics

The total cell areas were significantly smaller in 2015 than in 2014 for the 45–75% ring portion (Fig. 3a). This portion of the ring was associated with cells that underwent the enlargement phase during the summer drought of 2015 (Fig. 3b). The pattern of the rate of cell enlargement also differed between years (Fig. 3c). Both rings showed an increase in the rate of enlargement in the first cells, and this rate was higher in 2015. Afterwards, during the 2015 summer drought, the rate of cell enlargement decreased by > 50%. The rate increased again after the cessation of the dry period, whereas it decreased continuously in 2014.

Fig. 3

Anatomical features (cell area, cell wall area and cell lumen area) and rates of cell differentiation in the years 2014 and 2015 for the four studied Norway spruce trees. The anatomical features are aligned by position within the ring width (a, d, g) and by time of cell formation (b, e, h). Differences in rates of cell enlargement and wall thickening are also shown (c, f), where each rectangle represents a single modeled cell. Colored lines (a, c, d, f, g) indicate the fitted values obtained from generalized additive mixed models and the corresponding shaded colored areas indicate the 95% confidence intervals. The gray band (b, c, e, f, h) indicates the period when there was a tree water deficit (TWD; negative change in stem radius in μm) for more than 24 h during the extreme summer drought in 2015.

Further, cells located in the middle of the 2015 ring (25–85%) also had significantly smaller wall areas than those from the same portion of the 2014 ring, despite the fact that cells located in the first 25% of the 2015 ring had larger wall areas than those from 2014 (Fig. 3d). The cells with a reduced wall area in the 2015 ring completed the wall‐thickening phase during the summer drought (Fig. 3e). The rate of cell wall thickening was higher in 2015 than in 2014 at the beginning of the growing season (Fig. 3f). However, this situation was reversed as soon as the summer drought started (i.e. rates of wall thickening dropped in mid‐July 2015, but they reached their maximum values at about the same time in 2014). The rates of wall thickening almost reached pre‐drought levels after the cessation of the summer drought, and showed similar values to those of 2014 at the end of the growing season (Fig. 3f). Similarly, significant differences between years in cell lumen area were detected in the first 35–80% of the ring width (Fig. 3g), but these were not clearly associated with the period of summer drought (Fig. 3h). The observed changes in the anatomical characteristics of the cells formed during drought (Fig. 3) mean that the reduced rates of formation could not be compensated by increased phase durations (Fig. S6).

Variations in δ and iWUE

The intra‐annual tree‐ring δ13C variations were significantly different between the two years (Figs 4, S7). The δ13C values started at a slightly higher level in 2014 (−25.2‰) compared with in 2015 (−25.7‰). Despite an increasing trend during the first 20% of the ring in both years, δ13C values then decreased steadily in 2014 but continued to increase towards a maximum (−24.0‰) at 55% of the ring in 2015. High δ13C values in 2015 coincided with the timing of the drought (Fig. 4b), when many cells were in the wall‐thickening phase (40% of the ring). The δ13C values in 2015 decreased again at the end of the summer drought but remained significantly higher than in 2014 for the rest of the ring.

Fig. 4

Intra‐annual tree‐ring cellulose δ13C patterns for 2014 and 2015 and derived parameters aligned by ring position (a) and time of cell wall formation (b–e) for the four studied Norway spruce trees. In (b–e), each rectangle represents a single modeled cell. ∆13C, discrimination against 13C; C i, leaf intercellular CO2 concentration; iWUE, intrinsic water‐use efficiency. In (a), the colored lines indicate the fitted values obtained from generalized additive mixed models and the corresponding shaded colored areas indicate the 95% confidence intervals. The gray band indicates the period when there was a tree water deficit (TWD; negative change in stem radius in μm) for more than 24 h during the extreme summer drought in 2015.

Similarly, the δ13C‐derived variations in ∆13C, C i and iWUE differed between years. During 2014, we observed a decrease in ∆13C and C i from the beginning of the growing season to the end of June (Fig. 4c,d), which resulted in an increase in iWUE (from 93.03 to 97.13 μmol CO2 mol−1 H2O; Fig. 4e). Subsequently, the ∆13C and C i increased and iWUE decreased steadily until the end of the growing season. Although ∆13C and C i were higher at the beginning of 2015 than in 2014, we observed a decrease in both parameters during the summer drought. This resulted in an enhanced iWUE over a longer period in 2015 than in 2014 (maximum iWUE values of 107.8 and 97.4 μmol CO2 mol−1 H2O in 2014 and 2015, respectively, corresponding to an increase of 10.6%).

Impacts on tree‐level daily carbon sequestration

The intra‐annual dynamics of daily tree carbon sequestration through woody biomass production showed a strong fingerprint of the 2015 extreme drought. We observed an earlier growth start and more favorable woody biomass production at the beginning of 2015 compared with 2014. While daily woody biomass production showed a bell shape in the reference year 2014, reaching its maximum during the first weeks of July, the 2015 dry spell clearly had a negative effect on woody biomass production (Fig. 5). During the 41‐d drought period, the average tree gained 67% less carbon (2.692 vs 4.033 kg) than in the 2014 reference period. Carbon sequestration resumed strongly after the drought, however, and reached even higher values than in 2014 (+23%), indicating an important compensating activity (Fig. 5). Notably, this effect resulted in an equivalent total amount of carbon being sequestered in woody biomass in the two years (7.261 kg in 2014 and 7.349 kg in 2015).

Fig. 5

Seasonal dynamics of daily carbon sequestration of Norway spruce in 2014 and 2015. The tree carbon gain is represented by carbon in woody biomass production and was calculated based on daily wall thickness increments. The gray band indicates the period when there was a tree water deficit (TWD; negative change in stem radius in μm) for more than 24 h during the extreme summer drought in 2015.

Discussion

In this study, we quantified the impact of an extreme summer drought on Norway spruce water‐use efficiency and carbon sequestration using a unique set of detailed retrospective observations of wood formation and intra‐annual tree‐ring properties (xylem cell anatomy and stable carbon isotopes), combined with measurements of physiological water stress. Our measurements of physiological stress clearly indicated that both the local ecosystem (via NDVI values) and the selected trees (via TWD) responded strongly to the meteorological extreme drought event in 2015. We observed imprints of drought on the cell kinetics, anatomical features and isotopic signatures of the formed cells, as a consequence of physiological adjustments to cope with water limitations. During the drought period, fewer cells were produced, were smaller with thinner cell walls, cells remained longer in the formation phases, and trees lost less water during gas exchange than during average growing seasons. Such physiological adjustments indirectly impacted the capacity of trees to sequester carbon and use water in the drought year 2015 compared with the reference year 2014. However, the different resilience patterns observed in the anatomical characteristics and stable carbon isotopes suggest that secondary growth might not be constrained by carbon availability.

The physiological responses triggered by water deficit

In our study, we made use of a physiological definition of drought based on a proxy for stem dehydration, that is, TWD (Zweifel et al., 2005). This is a measure of stem radial shrinkage caused by water depletion from living tissues when transpiration exceeds water uptake (Dietrich et al., 2018). This measure made it possible to identify precisely when and for how long the 2015 extreme summer drought impacted the water status of the studied trees, and this 41‐d period matched well with the period with altered growth processes. This period can still be considered moderate in length compared with, for instance, the 91 d of TWD observed at a low‐elevation site (490 m asl) in southern Germany in the same year (Schäfer et al., 2019).

Our observations of reduced xylem cell division and smaller cells formed under drought support the current understanding of the biophysical mechanisms underlying the rates of cell enlargement. Cell division and enlargement are limited under water deficit (Génard et al., 2001), as turgor is needed to expand the primary cell wall of the forming cells (Hsiao, 1973). Studies using mechanistic modeling have confirmed the major role of turgor in tracheid enlargement and size (Cabon et al., 2020a; Peters et al., 2020). These results are confirmed by observations in water‐limited environments where long time‐series of earlywood characteristics were found to correlate better with precipitation than with temperature (Martin‐Benito et al., 2013; Castagneri et al., 2018). Our study also showed that cell wall‐thickening rates were reduced during the period of water deficit, consistent with findings from other studies (Olano et al., 2014; Stangler et al., 2021). Limited water availability during xylogenesis can influence the amount of nonstructural carbohydrates (NSCs) available for use in cell wall thickening, as NSCs are instead required to counteract turgor loss by increasing osmotic potential (Deslauriers et al., 2014). The previously described mechanism of compensating for reduced cell development rates by increasing phase duration also occurred in our trees (Cuny & Rathgeber, 2016; Cuny et al., 2019). This alteration mitigates the impact of climatic conditions on the cell anatomical structure (Balducci et al., 2016) and ensures xylem functionality and adequate tree growth (Lachenbruch & Mcculloh, 2014). However, the extended periods spent in both formation phases did not suffice to avoid altered anatomical characteristics. Overall, these anatomical responses to the extreme event of 2015 reduced the amount of carbon sequestrated in the stem during the summer drought: 67% less than in the same period of an average year.

It is also valuable to note that despite the relatively long water deficit of 41 d, we did not observe a complete cessation of the processes of cell division and enlargement, as has been previously documented in drier environments (Camarero et al., 2010; Eilmann et al., 2011). Our observations are in contrast to the zero‐growth concept associated with dendrometer measurements affirming that growth (defined as a permanent radial increment) cannot occur under stem shrinkage (Zweifel et al., 2016). This minor conceptual incongruency might be a result of different definitions of growth (expansive vs structural; Hilty et al., 2021), which can have great relevance when assessing the effect of meteorological variables on cambial activity. Furthermore, the earlier start of the growing season in 2015 compared with 2014 is probably a result of the warmer spring conditions that promote radial growth (Rossi et al., 2008; Delpierre et al., 2019; Huang et al., 2020) and increased annual growth in mesic climates (Babst et al., 2019; Anderson‐Teixeira et al., 2021).

Our results show that the period of TWD also affected the carbon–water exchange of the studied trees, as indicated by an increase in δ13C values during the extreme drought, which translated into lower Δ13C, lower C i, and therefore higher iWUE. δ13C ratios are highly influenced by changes in photosynthetic activity and/or stomatal conductance, which are indeed strongly coupled (Farquhar et al., 1989). Soil and atmospheric water deficits during the extreme drought period are more likely to lead to a closure of stomata than a reduction in photosynthetic rates, which would better explain the δ13C increase (Battipaglia et al., 2014). Indeed, stronger regulation of the stomata is necessary to reduce water loss by transpiration, causing a more parsimonious use of water for a given amount of assimilated carbon. An increase in iWUE is thus often interpreted as a reaction to drought to facilitate the maintenance of a positive carbon balance under dry conditions (Raven, 2002). We found that water deficit increased the seasonal maximum water‐use efficiency by 10.6% in the dry compared with the normal growing season, in line with other studies of intra‐annual δ13C patterns under dry conditions (Sarris et al., 2013; Battipaglia et al., 2014). We cannot exclude post‐photosynthetic fractionation processes as a result of the remobilization of reserves, which would partially decouple the isotopic signal of stem cellulose from the leaves (Gessler et al., 2014; Bögelein et al., 2019). Nonetheless, the tight match between TWD and increased δ13C indicates an immediate and direct incorporation of the assimilates produced under dry conditions into the forming wood structure. Indeed, labeling experiments on mature coniferous trees have shown that the isotopic signal in the leaves can be transferred down the stem within few days (Gao et al., 2021).

These physiological adjustments (i.e. reduced stomatal conductance and reduced cell and lumen area) increase the hydraulic resistance to water flow throughout the tree (Sviderskaya et al., 2021), but at the same time they enhance the hydraulic safety against interruptions to sap flow as a result of cavitation and embolism (Hacke & Jansen, 2009; Bouche et al., 2014). Both of these adjustments are therefore important for coping with decreasing water potentials under drought. Our findings are in agreement with results from physiological studies indicating an early physiological response to water limitations in Norway spruce (Lu et al., 1995; Ditmarová et al., 2009). Our observations demonstrate that water deficit does not always completely curtail wood formation processes, but limits growth before a reduction in photosynthesis occurs. These responses, however, might have different impacts depending on the timing and severity of the drought with respect to the tree physiological and phenological status. For instance, a severe drought might induce dramatic hydraulic failures, causing leaf shedding with pervasive legacies (Anderegg et al., 2015) and exposing declining trees to a higher risk of mortality (Timofeeva et al., 2017; Camarero et al., 2018). Additionally, whether the drought event occurs early or late in the season might have different impacts on tree‐ring radial growth (Etzold et al., 2022; Salomón et al., 2022).

Post‐drought recovery

Our findings indicated a rapid and strong recovery of cell enlargement and wall thickening rates after the drought event – a doubling of the minimum rates observed during the drought to reach almost pre‐drought levels – as soon as the drought period was over (TWD turned positive). Moreover, the observed rates were far above the levels observed for the reference year 2014. Although we did not observe an increase in cell production during the hydric resumption, this boost in activity nevertheless affected carbon sequestration at the cellular level in the post‐drought period. Notably, the δ13C values of these cells remained about 1‰ higher than in 2014. This difference was even higher for ∆13C, C i and iWUE because we also accounted for seasonal atmospheric δ13C variations in these calculations.

We interpret these results (i.e. the drought‐induced reduction and the subsequent post‐drought boost in sink activity) as a direct consequence of changes in the interaction between the wood formation processes and carbon assimilate production. Despite a reduction in leaf physiological processes during drought conditions, carbon assimilation still continues with a higher isotopic signal but incorporation into sinks is minimal (Tomasella et al., 2017), owing to the different climatic sensitivities of photosynthesis and growth (Fatichi et al., 2013; Körner, 2015; Hagedorn et al., 2016). Previous findings of an accumulation of NSCs under low water availability (Dietze et al., 2014; Piper et al., 2017) and reduced transfer to storage tissues (Rademacher et al., 2021) support our results. After the resumption of an improved tree hydric status, the assimilates that accumulated during drought are readily available and rapidly incorporated into wood formation processes, which are no longer constrained by water limitation (Gessler et al., 2020). Our results also suggest that carbon assimilates that have not yet been stored are the main driver of sink activity (Rademacher et al., 2021), but only under favorable conditions for cambial activity.

Conclusions

Our study improves the understanding of the causal cascade of detailed physiological processes and structural responses related to tree water use and carbon sequestration using a tree‐centered approach. Precise identification of the timing and severity of the physiological drought allowed us to demonstrate that the reduced capacity for wood formation (reflected in wood structure) induced a net loss of sequestrated carbon of 67% during a drought period in comparison to a regular year. This occurred despite attempts to mitigate drought stress by increasing the efficiency of water use and extending the duration of tracheid development. However, a resumption of the pre‐drought growth rate upon release from the water deficit, through the use of previously unused assimilates, compensated for the growth impacts of the earlier drought stress.

Increments in size and wall thickness are dissociated at the cellular level, and inconsistent bias could therefore occur when inferring tree carbon allocation without accounting for secondary wall deposition (Cuny et al., 2015). This mismatch between xylem processes (enlargement and wall thickening) might partly explain discrepancies observed in studies that try to link ecosystem productivity with stem girth (Zweifel et al., 2010) or tree‐ring width series (Babst et al., 2014). Our study provides novel insights into the mechanisms of responses that could support further development of sink model components of next‐generation DGVMs that include carbon sequestration based on wood processes (Zuidema et al., 2018; Friend et al., 2019; Babst et al., 2020; Eckes‐Shephard et al., 2021). Similar studies designed along ecological gradients and including more variability in impacting events would help to further test and calibrate such models.

Author contributions

EM‐S and PF designed the study. EM‐S performed the analysis and prepared the results. PF, KT, MML and AR contributed to the interpretation of the results. EM‐S and PF led the writing, with many contributions from the other co‐authors.

Fig. S1 Climate conditions of the selected years.

Fig. S2 Methodology and abbreviations applied in this study.

Fig. S3 Seasonal dynamics of the normalized difference vegetation index.

Fig. S4 Average stem radius changes throughout the two growing seasons.

Fig. S5 Mean duration of the enlargement and wall‐thickening phases for individual cells.

Fig. S6 Ratio between duration and rate of cell kinetics.

Fig. S7 Intra‐annual δ13C values of tree‐ring cellulose from Norway spruce from the years 2014 and 2015.

Notes S1 Calculation of cell kinetics (timing and duration) and rates.

Notes S2 Calculation of stable isotope‐derived physiological parameters.

Notes S3 Calculation of daily woody biomass (carbon budget).

Table S1 Allometric characteristics of the four selected trees.

Please note: Wiley Blackwell are 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.

Click here for additional data file.