The study was conducted in a permanent site located in the Monts-Valin National Park (QC, Canada), called GP, Gaspard (48°34′N; 70°53′W). The site is characterized by a mixed natural stand of black spruce [Picea mariana (Mill.) B.S.P.] and balsam fir [Abies balsamea (L.) Miller] at 227 m above sea level (4–8% slope) located in the balsam fir-yellow birch forest bioclimatic domain (Saucier et al., 2009). The undergrowth is a typical mixed vegetation with herbaceous and ericaceous shrub species (Cornus canadensis L., Gaultheria hispidula L., Kalmia angustifolia L., Rhododendron groenlandicum Oeder Kron and Judd, and Vaccinium myrtilloides Ait.). As the study area is in the transition zone between northern temperate forest and boreal forest, the climate is dry and moderately cold during the winter, with moist and warm summers. Mean annual air temperature and precipitation are 2.8°C and 931 mm, respectively (1981–2010 Climate normals of nearby station with data, Environment Canada). Extreme minimum and maximum temperature reaches −43.3 and 38.4°C, respectively, and the average frost-free period length is 123 days (1981–2010 Climate normals, Environment Canada).

Bud and larval phenology and stem radius variation were monitored over a 4-year period (2016–2019). Three dominant or co-dominant trees per species were randomly selected in the site stands, having low defoliation level of eastern spruce budworm (SBW) [Choristoneura fumiferana (Clemens)] (Ministère des Forêts de la Faune et des Parcs, 2015).

A weather station was installed at the site in a forest gap in order to measure the microclimatic conditions. Air temperature (°C) and rainfall (mm) were collected at 15-min intervals and recorded as hourly averages by means of CR1000 data-loggers (CAMPBELL SCIENTIFIC Corporation, Logan, UT, United States). For technical problems during winter months, absolute maximum and minimum air temperature (°C) were referred to different duration’s observation (Table 1).

From 2016 until 2019, we measured budbreak phases between May and July once a week on apical shoots of the two north-facing and two south-facing permanent branches per tree in the low-crown (three tree species), located generally between a height of 1.5 to 2.5 m above the ground. On each apical shoot, we monitored bud development following Dhont et al. (2010): (1) open bud; (2) elongated bud; (3) swollen bud; (4) translucent bud; (5) split bud; (6) exposed shoot. At the same time, we determined the larval development stages of eastern spruce budworm based on the size of the head capsule of the larvae (from larval stage development L2 to L6), three larvae were randomly observed in each branch (McGugan, 1954). Larva L2 spends autumn and winter in hibernation and emerges in the spring, the insect intensively feeds and grows on shoots during larval stages L4 to L6 (Pureswaran et al., 2015). We also visually assessed the defoliation levels on current-year foliage of apical shoots collected at the mid-crown in each tree during the growing season (except in 2018 when for a technical problem we monitored defoliation only in September) using the shoot-count method based on six defoliation classes (Piene et al., 1981; MacLean and Lidstone, 1982) and calculated the median percentage of defoliation for each tree.

We used point dendrometers Radius type (Ecomatik, Munich, Germany) to monitor stem radius variation from January 2016 until December 2019. We installed the dendrometers at 1.3 m height on the north side of the trunk on three tree species. Measurements were taken every 15 min and stored in CR1000 data-loggers (CAMPBELL SCIENTIFIC Corporation, Logan, UT, United States). The resolution and accuracy of the datalogger-dendrometer equipment was 0.2 and 1.5 μm, respectively. To compare the same temporal interval among years, we selected from each annual record measurements from April to October corresponding to the growing season in our study region, as reported in Belien et al. (2014).

We estimated the probability of co-occurrence of time-windows of needles flushing (phases 3–6) and feeding larvae (fourth to sixth instars) for each tree species during each growing season from 2016 until 2019 using SAS (SAS 9.4, SAS Institute, Inc., Cary, NC). The classification method adopted a parametric approach based on multivariate normal distributions within each group to derive discriminant quadratic functions. We assumed unequal variances between groups using the observed within-group covariance matrices (Ren et al., 2020). For referring to the host phenology in the statistical analyses, we considered the budburst in three periods, based on the days before, during and after the occurrence of the time-window of needles flushing, when buds are susceptible to defoliation (phases 3–6).

To describe the stem radius variation among different years and between species, we fitted a Gompertz function defined as:

where y is the daily stem radius variation, τ is the time computed in day of the year (DOY), A is the upper asymptote of maximum stem radius variation, β is the x-axis placement parameter, as time at inflection point of the curve, κ is the growth rate of the curve (Camarero et al., 1998; Rossi et al., 2003). The selection of the Gompertz function was based on the lower Akaike information criterion (AIC) and the goodness of fit of the curves was evaluated graphically using the residuals plots of each curve. We then performed a Linear Mixed Models (LMM) with PROC MIXED procedures of SAS 9.4 (SAS Institute, Cary, NC) to assess the effect of year and species, and their interaction the estimates of the coefficients of the Gompertz functions and the stem radius values at the beginning of budburst (SRF). Year and species were considered fixed factors. Then, in the linear mixed models, trees (classified as numeric codes) were included as a random factor and defoliation was used as a covariate.

To assess stem variation radius phases, the stem cycle was extracted using a three-step procedure composed of two SAS routines (SAS 9.4, SAS Institute, Inc., Cary, NC, United States) (Deslauriers et al., 2011). The procedure divides the series into two distinct phases: (1) contraction, the period between the first maximum radius and the following minimum radius; and (2) expansion, considered as recovery period, the period from minimum radius until the return to the previous maximum value or when the stem begins another contraction phase (Turcotte et al., 2009). SAS routines calculated the amount of stem radius variation and the duration of stem cycle phases (Deslauriers et al., 2011). Linear Mixed Model (LMM) was used to analyze differences in the amplitude and duration of the contraction and expansion phases between species and among needles flushing period (before, during, and after), and the species × needles flushing period interaction. Species and periods were defined as fixed factors, whereas trees (classified as numeric codes) were used as random factor. Defoliation (%) was included as covariate. For the analysis, we used a linear MIXED model procedure in SAS (SAS 9.4, SAS Institute, Inc., Cary, NC, United States). The first-order autoregressive [AR(1)] provided the suitable correlation structure (Wolfinger, 1993). The selection of the first-order autoregressive [AR(1)] structure was based on the lower AIC. We evaluated normality and homoscedasticity graphically using the residuals plots of the linear models (Quinn and Keough, 2002).

We also ran linear mixed models to evaluate the effect of years on temperature and rainfall. In all analyses, when fixed-effects were significant, we ran least squares means (LS-means) with Student’s t-tests for paired comparisons and Tukey-Kramer tests for comparing multiple means.

From 2016 until 2019, mean air temperature reached around 10°C, with total rainfall between 620 and 750 mm during the growing season (Table 1). During the growing season (May–October), temperatures were similar overall the years (Table 1). The absolute minimum and maximum temperature reached −37 and 35.3°C in 2018 compared to other years.

Over the years, the range of the optimal stage of feeding larvae (SBW) and optimal needle flushing stages for colonizing buds varied between about 41 and 29 days, respectively (Figure 1). On black spruce, the occurrence of optimum larvae stage was observed a few days after the optimal needle flushing stage (5.75 ± 3.3 days), while there was a longer gap on balsam fir; indeed, the SBW optimal stage arrived 12.5 ± 4.2 days after host flushing (Figure 1). In 2016 and 2017, we observed slightly earlier needles flushing phases than the beginning of larval stages in black spruce. In balsam fir, while the flushing period and the feeding larvae began at the same time, the flushing period ended ∼24 days before the end of feeding larvae. In 2018, probability of occurrence of both opening and ending of black spruce flushing was well synchronized with larval phenology, while in balsam fir flushing was 11 days before larvae started intensive feeding activity. In 2019, we observed a low and, respectively, high (2–3 weeks delay) phenological mismatch in black spruce and balsam fir.

From 2016 to 2019, there was an increase in the defoliation levels for both species (Figure 1). In 2016, we observed the first signs of defoliation in balsam fir (20% of new shoots defoliated), while it remained negligible in spruce (well under 10%). Later it increased to moderate values in 2017, and even more in 2018, when median annual values reached 20 and 30% in black spruce and balsam fir, respectively. Finally, in 2019, both species were severely defoliated, with 80% in black spruce and as much as 100% in balsam fir (Figure 1).

Over the four study years, both species showed similar seasonal patterns in radial growth (Figure 1) with a fast increase from the end of April and beginning of May to mid-July (from DOY 120–130 to DOY 201), then a progressive transition to a plateau reached in fall (DOY 304). No significant effects of species, year, of their interaction and of defoliation were observed on two coefficients of the Gompertz function, the upper asymptote (A) and non-linear rate of growth (κ) (Table 2). Despite the low number of trees, the high accuracy of the dendrometer measurements allowed to highlight significant differences. In particular, significant effects of years (P = 0.0476, Table 2) and defoliation (P = 0.0208, Table 2) were found on the coefficient β (inflection point of the growth response curve). However, no significant differences among years were observed anymore when multiple comparison correction by Tukey-Kramer’smethod was applied. Defoliation had a significant and negative effect on β, meaning that an increase in defoliation corresponded to an earlier β (P = 0.0208, Table 2). Furthermore, the stem radius expansion at the beginning of needle flushing (SRF) was significantly affected by all factors except the interaction (Table 2), meaning that although species responded similarly at the end of the growing season, stem radius responded strongly to species and defoliation effects in the first phases of needle flushing. The values of stem radius expansion at the beginning of needle flushing was significantly lower in balsam fir compared to black spruce, with 0.06 and 0.42 mm, respectively (P = 0.0038, Table 2). On the contrary, the defoliation had a significant and positive effect on stem radius values (P = 0.004, Table 2): an increase in defoliation corresponded to an increase in stem radius at the beginning of needles flushing (Figure 2). Even if a significant effect of the year was observed on SRF (P = 0.0435, Table 2), no difference in stem radius expansion at the onset of needles flushing was detected from 2016 to 2019 (after multiple comparison correction by Tukey-Kramer’s method).

The species significantly modified the amplitude of contraction and expansion according to the different periods, before, during and after needles flushing (Table 2). The amplitude of stem contraction and expansion was significantly higher in black spruce compared to balsam fir (LSM, 0.09 and 0.06 mm in spruce vs. 0.04 and 0.02 mm in fir, respectively) (Figure 3). The needles flushing period influenced only the amplitude of stem expansion, but not the contraction. The amplitude of stem expansion was significantly lower before flushing compared to during flushing (LSM, 0.03 vs. 0.05 mm) (Figure 3). However, we observed no significant difference in the amplitude of contraction and expansion in the combination of flushing × species, and according to the defoliation level (Table 2).

The duration of stem contraction and expansion was influenced by the flushing period and defoliation effects (Table 3). The duration of stem contraction after flushing was significantly lower than before and during bud development (LSM, 7.5 h vs. 10.6 and 9.5 h, respectively) (Figure 3). A significant difference in stem expansion duration was also found between, before and during the flushing as well as between, before and after the flushing (Table 3). The duration of expansion before flushing was significantly higher than during bud development (LSM, 13 h vs. 9) (Figure 3). However, we found no differences in the duration of contraction and expansion between species and for the flushing × species interaction (Table 3 and Figure 3).

Furthermore, significant differences in the duration of stem contraction and expansion were found according to the defoliation level. In the pooled raw data, mean values of the duration of contraction were 8.5 h in the low-moderate defoliation levels, whereas they slightly increased to 9 h in higher levels of defoliation (Figure 3). In the pooled raw data, mean values of duration of expansion were around 10–11 h in the low defoliation levels, whereas they decreased to 9 h in the severe defoliation levels (Figure 3).

This study investigated seasonal stem radius variation of black spruce and balsam fir during 4 years of natural defoliation by spruce budworm. We expected that a higher level of defoliation would reduce seasonal stem radius variations such as in young trees in controlled conditions (Deslauriers et al., 2019). However, in the mature trees in natural conditions of this study, we detected a decreasing trend in seasonal stem radius variation only in balsam fir during the fourth defoliation year, thus partially contradicting the first hypothesis. Since branches perform as semi-independent compartments (Sprugel et al., 1991), water reserve stored in roots and stems sustains woody and leaf tissues with a given pressure during high evaporative periods (Scholz et al., 2007; Meinzer et al., 2008). Under longer term droughts, stored water acts as a safety buffer more efficiently in adult trees than in saplings (Phillips et al., 2003), limiting the damage to stem growth. Specific soluble sugars (i.e. sucrose raffinose and pinitol) were then required for osmoregulation to prevent hydraulic failure (Hölttä et al., 2006; Deslauriers et al., 2014; Sevanto et al., 2014). However, under different stress conditions such as defoliation, needles damage or loss after larval feeding had a minimal or no effect on water loss (Quentin et al., 2011; Bouzidi et al., 2019). An increased leaf water content and reduced evapotranspiration (Balducci et al., 2020) are required to counteract the reduction of turgor pressure due to needle removal. Therefore, in the case of defoliation, increased water uptake and stem growth could take place until carbon remobilization from plants parts could act also as a safety buffer for stem growth in the first years of consecutive defoliation. Indeed, defoliation did not affect two of the three estimated Gompertz parameters (A and κ), whereas the third one, β, was significantly affected by year and defoliation. Furthermore, a significant increase in stem radius expansion (SRF) at the beginning of the needles flushing was observed in the increased defoliation level. In fact, during and after flushing, heavy defoliation removes all or most of current-year needles. Previous studies observed a temporary slowdown of stem growth and hypothesized an effect of budburst phenology (Belien et al., 2014; Balducci et al., 2019) due to C-sink competition between bud development and stem growth (Deslauriers et al., 2016; Martínez-Vilalta et al., 2016).

There was no significant variation over time for parameter A, showing that in the first 4 years of defoliation both species were able to maintain their stem radius without major effects of defoliation for the remainder of the growing season. Despite this, the sharp decrease of stem radius in balsam fir in the fourth year could be due to intensity and duration of defoliation and age response. Previous work in young trees showed that stem radius decreased when starch levels were significantly depleted in the needles, branchlet, and roots after 2 years of heavy defoliation (Deslauriers et al., 2019). We could expect a more delayed response after several successive years of defoliation in mature than in young trees, due to age differences in different hydraulic safety behaviors and C-dynamic (Domec and Gartner, 2002; Hartmann and Trumbore, 2016). Indeed, under favorable conditions, a bigger fraction of water and C-reserve is generally stored in mature trees than in young trees (Domec and Gartner, 2002; Martínez-Vilalta et al., 2016). But mature trees operate nearest to xylem dysfunction compared to young trees (Domec and Gartner, 2002). Moreover, during defoliation mature trees showed a narrower safety margin to cavitation than young trees, compromising hydraulic integrity and efficiency (Hillabrand et al., 2019), although C-stocks can be remobilized for regrowth after stress conditions (Carbone et al., 2013).

Contrary to the other Gompertz parameters, β was responsive to year and defoliation, indicating an earlier occurrence of the inflection points at higher defoliation levels. Our finding suggests an earlier end of the exponential increase of stem radial growth, particularly when defoliation is high. Interannual changes in stem radius variation are regulated by a mixture of current or delayed abiotic effects (temperature, precipitation, cloud cover) (King et al., 2013; Vieira et al., 2013). A previous study suggests that the convergence observed in host and insect phenology under warmer temperatures increases food quantity and quality for the foraging insects (Ren et al., 2020). In this study, we observed co-occurrence of host-insect stages over the years, although their peaks did not perfectly overlap in 2018 and 2019 in balsam fir. Reduced synchrony in balsam fir could protect the species against the severe -defoliation (Nealis and Régnière, 2004). The year effect on β could be potentially masked and influenced by the occurrence of insect feeding. However, it is known that defoliation reduces water safety and efficiency (Hillabrand et al., 2019) and could reduce the mechanical strength of the thick-walled tracheids via impairing the C source-sink relationships. Defoliation advanced needle flushing and reduced radial growth in young trees (Deslauriers et al., 2019). Needle flushing is a high priority event to ensure new carbon assimilation for the coming growing season. In case of defoliation, newly assimilated C is primarily allocated to primary growth and crown recovery rather than to secondary growth. However, growth reduction in tree rings, with a typical “U” or “V” shape damage of tree rings due to lower ring width, was observed only after 5 years of heavy defoliation (Simard et al., 2008). This suggests that xylem conduits could be deformed and this deformation could affect water transport, although they could have enough C-resources to grow for several years, as suggested by our findings on parameter A. However, similarly to young trees, mature trees experienced an earlier needle flushing after 1 year of defoliation (Fuentealba et al., 2017a). The feeding activity may induce high pressure on bud and needles availability in the next season, predetermining growth conditions. In our study, the earlier time of inflection of the stem radial growth curve could signal a pattern of growth compensation, which could be changed by an adverse effect of defoliation.

An alteration of the daily cycle of stem contraction and expansion (especially of their duration) with defoliation is consistent with our first hypothesis. This result suggests that contraction lasted longer, and that water repletion was faster at the higher defoliation levels. The consequence is a continued rapid radial growth at the time of needles flushing. Similar modulation (increase/decrease) of duration of cyclic contraction/expansion was observed under water deficit (Zweifel et al., 2005) in concomitance with a negative/positive shift in predawn water potential (Balducci et al., 2019). An imbalance between transpiration demand and water uptake could explain this (Herzog et al., 1998; Meinzer et al., 2008). However, during and after severe defoliation of apical shoots, a higher transpiration demand is generally combined with a decrease in midday water potential (Eyles et al., 2013; Salmon et al., 2015). Yet the reduction of leaf area due to defoliation, even if it constrains transpiration, was not enough to significantly reduce the relative water content (RWC) (Balducci et al., 2020). Previous works support the idea that RWC reflects the balance between leaf tissue water supply and transpiration rate (Meinzer et al., 2003, 2014). Hence, the respective increase/decrease of contraction/expansion duration may reveal that mechanisms acting within diurnal fluctuation mediate response of internal water status to defoliation.

Our second hypothesis (in defoliated trees, during the needles flushing time-window, the amplitude and duration of daily contraction and expansion are reduced) was not supported by the results. Needles flushing period influenced all phases of stem radius variation, except the amplitude of contraction. In this study, the increase in the amplitude of expansion during flushing explained the higher SRF at the time of flushing (see previous section). Earlier spring stem rehydration associated to earlier snowmelt could explain this (Rossi et al., 2011). A recent study demonstrated that soil temperature and mean daily surface temperature significantly increased the number of enlarging tracheids in our species (Chen et al., 2019). Thus, favorable environmental conditions could increase reversible stem expansion before irreversible stem radial growth.

Differences in the species response to defoliation was partly expected (third hypothesis). The amplitudes of stem radius variation were higher in black spruce than in balsam fir over the 4-year defoliation period. Previous studies showed that mature black spruce and balsam fir showed similar amplitudes of stem variations, ranging between 0.9 and 2 mm (Deslauriers et al., 2003; Belien et al., 2014). In our study, balsam fir reduces the amplitude of stem variation more than black spruce; this suggests that balsam fir could prioritize the production of new epicormic and lateral shoots to overcompensate for needle losses by insect feeding (Piene, 1989a,b; Piene and Little, 1990). The species effect was also significant on SRF at the beginning of needle flushing. Balsam fir had a lower SRF than black spruce at the beginning of needle flushing, indicating that secondary growth slowed down during the flushing period in balsam fir.

However, we did not observe any difference in the durations of the diurnal fluctuation of stem radius variation between the species. Before 2016, when SBW was not observed in our site, the duration of xylogenesis (irreversible radial growth) and number of enlarging cells were also similar between the species (Chen et al., 2019), suggesting that water availability in wood tissues and turgor pressure process (cell expansion) were adequate. However, when defoliation started, balsam fir showed a smaller hydraulic safety margin compared to black spruce (Balducci et al., 2020). This suggests that these species have different behavior of daily radial variation over time to sustain growth water-driven process. An earlier flushing compared to the timing of insect emergence was recently observed in both species after severe defoliation (Deslauriers et al., 2019; Ren et al., 2020). This implies that in the case of higher allocation of sugar compounds to buds, trees might live at the limit of hydraulic failure to favor chemical defense, primary and root growth and defense strategies over replenishment of stem growth reserves (Körner, 2003; Deslauriers et al., 2015, 2019). We observed a lower defoliation level in black spruce than in balsam fir over 4 years. This result suggests that black spruce had not yet attained the critical defoliation values that suppress stem radial variation, as observed in previous studies (Bouzidi et al., 2019; Deslauriers et al., 2019).