57
Résumé
Les limitations hydrauliques chez les arbres dominants pourraient jouer un rôle dans le déclin en croissance des forêts au cours du temps. A l’aide de mesures directes (flux de sève du tronc) et indirectes (composition isotopique foliaire) de la transpiration des arbres, combinées à des estimations de croissance des arbres et des peuplements (dominance de croissance), nous avons testé l’hypothèse qu’une limitation hydraulique allait affecter seulement les arbres dominants et participerait ainsi aux changements de dominance de croissance des peuplements au cours du temps dans la forêt boréale du nord-est du Canada. Les résultats suggèrent en effet qu’une limitation hydraulique augmente chez les arbres dominants au cours du temps et qu’elle est en partie responsable des changements de dominance de croissance des peuplements. Une limitation hydraulique chez les arbres dominants contribue clairement à la dominance de croissance inverse observée chez les vieilles forêts, étant ainsi probablement impliquée dans le déclin en croissance des forêts.
Mots-clés : Dominance de croissance, Flux de sève du tronc, Composition isotopique foliaire, Développement d’un peuplement, Vieilles forêts
58
Abstract
Hydraulic limitations in dominant trees may play a role in the growth decline that is commonly observed in aging forest stands. We hypothesized that hydraulic limitations affect dominant trees but not non-dominant trees in aging forests, resulting in changes in growth dominance during stand development. To test this hypothesis, we used a 1067-year-long post-fire chronosequence that was established in the eastern Canadian boreal forest. Within each stand, we estimated transpiration of many sized trees using direct (stem sap flow) and indirect (leaf isotope composition) measurements, which were combined with tree and stand growth estimates (growth dominance coefficient, GD). Stem sap flow measurements indicated that transpiration rate per unit leaf area (EL) of dominant trees
indeed decreased with increasing stand age and, based upon temporal changes in leaf C and O isotope ratios, was related to an increase in stomatal closure. These results suggest that hydraulic limitation of dominant trees increased with stand age. Also, the slope of the relationship between EL and tree
diameter that was calculated for each stand decreased with decreasing GD, implying that hydraulic limitation was responsible for the shift in stand growth dominance that was observed over time. Therefore, hydraulic limitation in dominant trees clearly contributes to reverse growth dominance of old-growth boreal forests, and could be involved in age-related declines in forest productivity.
Keywords: Growth dominance, Stem sapflow, Leaf isotopic composition, Stand development, Old- growth stands.
59
Introduction
The growth rate of most forests initially increases in the early stages of their development, peaks usually at the time of canopy closure, and declines thereafter (Assmann, 1970; Gholz and Fisher, 1982; Ryan et al., 1997). Such age-related declines in forest growth have been widely observed in empirical studies (Binkley and Greene, 1983; Taylor and MacLean, 2005; Xu et al., 2012). However, the underlying mechanisms are still a matter of debate (Ryan et al., 2004). Because the hypothesis of increased wood respiration causing the observed decline in growth with stand age (Kira and Shidei, 1967) has been refuted (Ryan and Waring, 1992; Ryan et al., 2004; Drake et al., 2010; Tang et al., 2014), numerous investigations have attempted to identify the mechanisms involved in such growth decline. Several hypotheses have emerged, including increases in nutrient (Murty et al., 1996) or hydraulic (Ryan and Yoder, 1997; Ryan et al., 2006) limitations to photosynthesis, a shift in growth dominance among trees during stand development (Binkley et al., 2002; Binkley, 2004) and an increase in belowground carbon allocation (Gower et al., 1996; Baret et al., 2015). All of these potential mechanisms have been experimentally supported, but no consensus has yet been reached, which suggests that a combination of site-specific processes rather than a universal mechanism could explain the decline in forest productivity with age.
The hydraulic limitation hypothesis (Yoder et al., 1994; Ryan and Yoder, 1997) proposed a mechanism to explain the slowing of height growth with tree size and the maximum limits to tree height. It predicts that stomatal conductance of water vapour (gs) in tall, old trees would be constrained
by a reduction in leaf-specific hydraulic conductance (kl), causing a decrease in photosynthesis and,
consequently, in primary productivity (Yoder et al., 1994). Reduction in kl with tree size could be
explained by a greater path length of the water column, an increase in sapwood density (Bowman et al., 2005), a decrease in sapwood conductance (Pothier et al., 1989), and the gravitational pull in tall trees. Because there is a strong interdependence between kl and gs (Sperry et al., 1993; Hubbard et
al., 2001), low leaf-specific hydraulic conductance would also decrease leaf conductance to CO2,
thereby reducing photosynthesis and the annual gross primary productivity (GPP) over the long-term (Gower et al., 1996; Ryan et al., 1997). Overall tree transpiration would be reduced, as it depends upon tree leaf area (Granier et al., 2000), stomatal characteristics (Kelliher et al., 1993; Hogg and Hurdle, 1997) ) and leaf-to-air vapour pressure deficit (Oren et al., 1999).
Accordingly, a decrease in hydraulic conductance and water flux with tree height and/or age has been observed in many tree species, such as Scots pine (Pinus sylvestris L.) (Mencuccini and Grace, 1996; Martinez-Vilalta et al., 2006), loblolly pine (Pinus taeda L.) (Drake et al., 2010), maritime pine (Pinus
60
pinaster Aiton) (Delzon et al., 2004), European beech (Fagus sylvatica L.) (Schäfer et al., 2000) and
Oregon white oak (Quercus garryana Douglas ex Hook.) (Phillips et al., 2003). Lower net photosynthetic rate and stomatal conductance have also been observed in tall, old trees compared to smaller trees (Bond, 2000; Kolb and Stone, 2000). While a decrease in whole-tree hydraulic conductance affects stomatal aperture (Sperry et al., 1993; Sperry and Pockman, 1993; Sperry, 2000) and leaf gas exchange (Hubbard et al., 2001), age-related effects on total hydraulic conductance per unit leaf area remain unclear (Mencuccini, 2002; Ryan et al., 2006). ).Inconsistent results could be explained by homeostatic mechanisms that would increase the water transport capacity of the tree relative to the total leaf area (Ryan and Yoder, 1997; Becker et al., 2000). Hydraulic limitation may be not universal and as whole canopy processes are often inferred by measuring small components of the whole system because of the complexity of investigating physiological responses of large canopies, testing the hydraulic limitation hypothesis is challenging (Ryan et al., 2006).
The objective of this study was to investigate how a hydraulic limitation to photosynthesis could contribute to the decline in stand growth that was observed after canopy closure in the northeastern Canadian boreal forest (Ward et al., 2014). This was achieved by using a sub-sample of a chronosequence that covered a period of over 1067 years and, thus, included old-growth stands (Ward et al., 2014; Baret et al., 2015). Previous work in the region has shown that declining growth of dominant trees was observed soon after canopy closure and was responsible for the decline in stand growth (Baret et al., 2017). According to the hydraulic limitation hypothesis, we postulated that the transpiration rate of dominant trees, but not that of dominated trees, should decrease in aging forests. This should result in a shift in growth dominance from dominant to non-dominant trees during stand development. To test this hypothesis, we used stem sap flow measurements, leaf-level isotope composition, and tree and stand growth estimates.
61
Material and methods
Study area
Sites were located north of Baie-Comeau (49°07’N, 68°10’W), Quebec, Canada, in the black spruce- feather moss bioclimatic subdomain (Robitaille and Saucier, 1998). The regional climate is cold maritime, with a mean annual temperature of 1.5 °C and mean annual precipitation of 1014 mm. Snow generally represents 35% of yearly total precipitation and the growing season lasts about 155 days. The fire return interval of the study region was estimated to be 270 years (Bouchard et al., 2008).
Black spruce (Picea mariana (Mill.) BSP) and balsam fir (Abies balsamea (L.) Mill.) are the dominant canopy species in these forests, with relatively minor components of white spruce (Picea glauca (Moench) Voss.), paper or white birch (Betula papyrifera Marsh.), jack pine (Pinus banksiana Lamb.), tamarack or eastern larch (Larix laricina (Du Roi) K. Koch), and trembling aspen (Populus
tremuloides Michx.). Low fire frequency in the area has led to the creation of a forest landscape that
is composed of 65-70 % old-growth, uneven-aged stands (Côté et al., 2010).
Site and tree characteristics
To test the hydraulic limitation hypothesis in boreal forest stands, we used the same post-fire chronosequence as that of Ward et al. (2014) and Baret et al. (2015). This chronosequence was composed of 30 stands, which were aged from 17- to 1277-years-since-fire. The sites were selected to be as similar as possible in terms of surface deposits, topographic position, exposure and drainage. Particular attention was given to select sites that were characterized by deep glacial tills with good drainage, which are the dominant biophysical features of the study area (Bouchard et al., 2008). In each stand, we established one 0.04-ha circular plot. Within each plot, we measured the diameter at breast height (DBH, 1.3 m) of all trees with a DBH > 9.0 cm. Tree height of all trees was also measured with an electronic dendrometer (Vertex III, Haglöf, Sweden). Foliage biomass, stem biomass and stem increment were estimated for each tree using their DBH, together with the biomass equations of Lambert et al. (2005) for Canadian tree species. Five-year wood biomass production was estimated from increment cores that had been taken at 1.3 m and oriented toward the plot centre for all trees with a DBH > 9.0 cm. Increment cores were used to estimate tree sapwood area (SA), while tree projected leaf areas (LA) were estimated using relationships between LA and SA, as described by Ward et al. (2014).
62 Time since last fire (TSF) for stands that were < 200-years-old were determined according to the historical fire map of the region, which had been prepared by Bouchard et al. (2008). They extracted basal discs from fire-scarred trees or cored several dominant trees, which were generally extracted from trembling aspen and jack pine because of their early establishment after fire. TSF was calculated by subtracting the year of inventory from the year of the last fire event. For older stands, the lifespan of individual black spruce and balsam fir trees was exceeded (Burns and Honkala, 1990); therefore, extraction of tree increment cores would not provide a precise TSF measurements, as individuals from the first cohort had likely disappeared. In such cases, 14C dating of charcoal samples from the last fire
was performed (for more details, see Ward et al. 2014).
The relationship between tree size and growth rate within the sampled forest stands was characterized according to their growth dominance patterns (Binkley, 2004; Binkley et al., 2006). Growth dominance in individual stands was evaluated based upon the method that was described by West (2014). For each stand, trees were arranged in ascending order of DBH; the cumulative stem increment was plotted as a function of the cumulative stem mass to form a growth dominance curve. A growth dominance coefficient (GD), similar to a Gini coefficient, was calculated for each stand as:
GD = 1 − ∑ni=1(si− si−1)(di+ di−1) [1]
where 𝑠𝑖is the cumulative proportional size of tree 𝑖 , and 𝑑𝑖 is the cumulative proportional growth of tree 𝑖. Consequently, a stand with strong growth dominance (GD > 0) means that the contribution of dominant trees to stand growth is greater than their contribution to stand mass. Conversely, a stand showing reverse growth dominance (GD < 0) indicates that dominant trees account for a larger share of stand biomass than does current growth. When growth of each tree is linearly proportional to tree size, GD is equal to zero.
Leaf carbon and oxygen isotope composition
The isotopic ratio of 13C to 12C (δ13C) was used as a time-integrated measure of g
s (Farquhar et al.,
1989). To separate the effects of stomatal dynamics and photosynthetic capacity (Amax) on δ13C
temporal variation, we simultaneously determined the isotopic ratio of 18O to 16O (δ18O) (Scheidegger
et al., 2000). We used a sub-sample of the chronosequence to determine δ13C and δ18O of sample tree
foliage. This sub-sample was composed of stands that were randomly selected within each of five age classes (0-50 y, 51-100 y, 101-150 y, 151-200 y, > 200 y) of the chronosequence, for a total of 12 stands. From the end of September to the beginning of October 2012, we collected foliage samples in each stand. Sampled trees were separated in two social classes based on their diameter at breast height (Table 3-1). In each stand, dominant trees corresponded to the eight largest trees in the plot of
63 400 m2 (i.e., the 200 largest trees per hectare), while the remaining trees corresponded to the non-
dominant social status. We sampled three branches from the upper third of the crown that faced south for up to three trees per social status. About 100 current year needles per tree were collected, and their area was determined with WinSeedle Pro v.2001a (Regent Instruments Inc., Québec, QC, Canada). The needles were oven-dried at 65 °C for 48 h. The samples were ground prior to being analyzed by the UC Davis Stable Isotope Facility (University of California, Davies, CA, USA) to determine δ13C
64 Table 3-1. Characteristics of the stands sampled for measurements of leaf carbon and oxygen isotope composition.
Note: ANPPwood is the aboveground net primary productivity of wood and DBH is the diameter of
tree at breast height. Sampled trees were composed of black spruce (bS), balsam fir (bF) and white spruce (wS), each of which being composed of two social classes: dominant (D) and non-dominant (ND).
Stand
Time since last
fire ANPPwood Growth Efficiency Species
Social
Class DBH Height
year kg.yr-1 kg.m-2.LA-1.ha-1.yr-1 cm m
S-07 37 48.704 0.0433 bS D 9.5-11.5 9.8-10.3 S-21 67 78.582 0.0353 bS ND 6.6-16.5 7.1-14.4 bS D 17.3 16.04 S-25 90 78.838 0.0526 bS ND 9.7-13.6 10.8-12.5 wS ND 10-13.3 10.81-12.5 wS D 20.3 17.31 bS D 20.3-22.6 16.8-17.3 S-38 112 49.692 0.0289 bS ND 9.3-22.9 13.2-21.6 bS D 28.4 21.6 S-41 112 59.023 0.0439 bS ND 9.6-18.1 10.3-19.0 bS D 26.5-28.1 22.8-24.3 S-42 112 45.008 0.0232 bS ND 9.5-16.5 10.8-17.2 bS D 23.4 19.8 S-14 153 36.631 0.0234 bS ND 9.1-18.7 8.8-18.8 bS D 23.9-26.9 20.7-20.9 S-28 198 47.566 0.0594 bF ND 11.1-19.9 7.81-16.9 bS D 30.1-33.1 19.7 S-39 198 47.628 0.0434 bF ND 9.8-17.5 8.4-11.9 bF D 24.5-31.3 15.1-22.1 bS D 34.5 21.4 S-19 553 68.178 0.0451 bF ND 9.2-15.9 8.01-13.7 bS ND 19-22.1 17.9-19.0 bS D 24.1-29.5 19.0-20.7 S-18 676 47.655 0.0417 bS ND 9.8-15.9 8.01-14.7 bF ND 11.7-19.0 7.5-15.4 bF D 20.4-21.3 15.9-16.9 bS D 27.2 19.61 S-02 1070 53.258 0.0547 bF ND 11.3-23.35 13.6-18.3 bF D 25.5-28.65 19.2-19.29
65
Water flux estimations
Sap flow measurements
We selected five stands that were distributed along the sub-sample of the chronosequence to measure transpiration at the tree scale with sap flow sensors. In each stand, five black spruce trees were selected to cover the range of each stand tree diameter (Table 3-2). Trees with large apparent defects, such as rot or cankers in the lower part of the bole, were never selected. Sap flows were measured with a constant heating radial flowmeter according to the original design of Granier (Granier, 1985, 1987).
The probes were installed on the five selected trees (one probe per tree, inserted on the north side of the stem) in each of the five selected stands. Once the probes were installed, stems were wrapped with reflecting sheets to avoid direct heating of the probes by solar radiation. The probes were then protected from rain and stem flow by wrapping the stem with plastic sheets, sealed at the top with pruning wax and vinyl tape. Sap flow data were collected over three months from June to August 2012. Data were recorded every 10 s and averaged every 10 min with data-loggers (models CR10 and CR10X, Campbell Scientific Ltd., Logan, UT, USA). Days when the sap flux sensors failed or had been shut off during battery replacement were eliminated from the data set. This was also the case during rainy days because heat dissipation sensors have poor signal-to-noise ratios during rainfall (Phillips and Oren, 1998).
66 Table 3-2. Diameter at breast height (DBH) and height of the five black spruce trees selected per stand for sap flow measurements.
Station
Time since last
fire Tree DBH Height
years cm m T1 9.9 12 T2 10.5 12.13 S-21 67 T3 13.3 12.13 T4 15.9 17.27 T5 17 17.27 T1 9.9 10.16 T2 13.3 16.83 S-25 90 T3 13.9 16.83 T4 18.4 23.19 T5 19.2 23.19 T1 14.1 10.3 T2 16.9 13.2 S-38 112 T3 17.4 13.2 T4 19.15 17.35 T5 21.1 21.7 T1 11.6 15.3 T2 14.1 15.3 S-14 153 T3 15.9 15.3 T4 17.95 18.8 T5 19.5 18.8 T1 11.2 11.51 T2 14.1 11.52 S-19 553 T3 17.23 13.77 T4 20.2 18.97 T5 27.6 18.97
Sap flow calculations
Sap flow density was calculated using the original calibration coefficients of Granier (1987). Sap flow density (u, cm3 cm-2 s-1) was computed as:
u = 0.0119 × K1.231 [2]
where K was computed as:
67 where dT is the difference of temperature between the heated upper sensor and non-heated lower sensor; dTM is the maximum temperature difference between the sensors (no flow). Since the probes integrate along their whole length and cover all sapwood depth, total sap flow (g h-1) was obtained
as:
F = u × SA × 3600 [4]
where SA is the total sapwood area (cm2) of trees.
As the probes generally extended deeper into the tree than the active sapwood, the correction that was developed by Clearwater et al. (1999) was used to scale intra-tree sap flux by considering the proportion of the probes that was in contact with non-conductive xylem. The total sap flow of each tree was converted to obtain a transpiration value for each sampled tree per day (litre day-1). To
compare the transpiration of trees of different sizes, we then calculated the transpiration per unit leaf area (EL, g cm-2 day-1) by using the aforementioned estimates of projected leaf area at the tree scale.
For each stand, we fitted a linear regression between monthly averaged EL and tree DBH. The slope
of these relationships (SEl−DBH) was then used to characterize the transpiration dynamics between
dominant and non-dominant trees at the stand scale.
Statistical analysis
Stand level relationships between the whole-tree transpiration per unit leaf area (El) and tree DBH
were fitted using a mixed linear model to take account of the repeated El measures on each tree.
Linear models were used to test the effects of time-since-last fire (TSF) on the SEl−DBH relationship and δ13C values. Logarithmic transformation of TSF was used to improve the model fit. The model
form that was used to estimate the relationship between δ13C and TSF was:
Yi = b0+ b1log10TSF + b2TSF [5]
where Yi is δ13C (‰). Sampled tree species was added as a categorical variable in the model.
Relationships between GD and SEl−DBH, and between GD and δ13C were also tested with linear
models. Model assumptions (normality of residuals and homogeneity of variance) were validated using a Shapiro-Wilk test, Levene’s test and graphical analysis of residuals. Pearson’s method was used to test the correlation between δ13C and δ18O values. Statistical analyses were performed in R
68
Results
Transpiration of different-sized trees and growth dominance
Mixed linear regressions between monthly values of EL and tree DBH revealed that the slope of these
relationships (SEl−DBH) varies with time that had elapsed since the last fire (Table 3-3). For example,
a positive slope was calculated for a 67-year-old stand, indicating that larger trees had higher EL
values than smaller trees (Figure 3-1a), whereas a 112-year-old stand was associated with a negative slope, meaning that the EL of small trees was higher than that of large trees (Figure 3-1b). Overall,
the two stands that were younger than 100 years were associated with positive slopes, while negative slopes characterized older stands (Figure 3-2).
69 Figure 3-1. Whole-tree transpiration per unit leaf area (EL) in July 2012 as a function of tree diameter
in a stand that was established 67 years after the last fire (a) and 112 years after the last fire (b). Each data point represents a daily EL value and the dashed lines correspond to the linear regressions
70 Figure 3-2. Slope of the relationship between whole-tree transpiration per unit leaf area and tree diameter (SEL-DBH) in June, July and August 2012, as a function of the time that had elapsed since the
71 Table 3-3 Values of SEL-DBH and the related statistics for the five sampled stands over each sampling
period.
Month Time since fire SEl-DBH SE R2 P-value
year 67 2.47×10-03 6.71×10-04 0.32 0.002 90 1.51×10-03 1.75×10-04 0.54 <0.0001 June 112 -3.45×10-04 2.05×10-04 0.14 0.137 153 -6.05×10-05 6.05×10-05 0.05 0.344 553 -3.10×10-04 1.10×10-04 0.10 0.008 67 1.53×10-03 2.56×10-04 0.41 <0.0001 90 1.00×10-03 1.23×10-04 0.42 <0.0001 July 112 -9.01×10-04 9.29×10-05 0.29 <0.0001 153 -9.39×10-05 4.26×10-05 0.20 0.092 553 -1.20×10-04 4.54×10-05 0.03 0.010 67 9.93×10-04 2.87×10-04 0.27 0.003 90 3.75×10-04 5.13×10-05 0.28 <0.0001 August 112 -8.25×10-04 1.36×10-04 0.18 <0.0001 153 -3.68×10-05 6.24×10-05 0.04 0.661 553 -1.10×10-04 4.54×10-05 0.03 0.018
Note: SE is the standard error of the SEl−DBH estimates
Stand-level coefficients of growth dominance were closely related to the values of SEl-DBH that were
calculated for July 2012 (Figure 3-3). A simple linear regression fitted between these two variables indicated that the slope of the relationship was strongly positive (P = 0.0012), with a R2 of 0.98.
72 Figure 3-3. Growth dominance coefficient (GD) as a function of the slope between whole-tree transpiration per unit leaf area and stem diameter (SEL-DBH) that was calculated for July 2012. The
dashed line represents a simple linear regression.
Leaf isotope composition
δ13C of dominant trees increased gradually up to 500 years, then decreased in the two oldest sampled
stands (Figure 3-4). Parameter estimates of the model (Eq. 5) explaining the temporal changes in δ13C
were significantly different from 0 (P < 0.005) for dominant trees, but not for non-dominant trees. No statistical differences in temporal changes in δ13C were found among tree species.
73 Figure 3-4. Leaf C isotope abundance (δ13C) of dominant and non-dominant trees as the time-since-
last-fire. Each data point corresponds to a foliar δ13C value of an individual tree.
Along the whole chronosequence, δ13C increased with increasing values of δ18O in dominant trees
(Figure 3-5); a positive linear correlation coefficient was found between these two variables (r = 0.46,
P = 0.0194). Furthermore, the strength of the relationship increased with increasing time that had
elapsed since the last fire. Prior to 120 years, a weak relationship was observed (r = 0.051, P = 0.8885), whereas between 120 and 200 years, the correlation between the two variables increased (r