Introduction

Competition is often viewed as the main driver of interactions between trees, but this idea is being challenged with examples of cooperation or facilitation through natural root grafting (Bertness and Callaway 1994; McIntire and Fajardo 2014). Root grafts are morphological and physiological unions formed between roots of distinct trees (Tarroux and DesRochers 2010). Solid contact and pressure between thick roots developing parallel or across each other can initiate root grafting and lead to vascular continuity and cambial fusion (Bormann and Graham 1959; Eis 1972). Research on root grafting has provided important insights into how ubiquitous it is in both conifers and deciduous species, but the struggles of surveying belowground tree attributes often hindered its assessment in comparison to aboveground characteristics (Lev-Yadun 2011). As a result, scientists are still speculating on the processes leading to root graft formation, the factors involved and how they operate. Several factors have been suggested to increase root grafting occurrence including rooting depth, root morphology and density, soil moisture and texture (Tarroux and DesRochers 2010), genetic relatedness (Graham and Bormann 1966; Loehle and Jones 1990), exposure to wind (Jelínková et al. 2009), stand density and age (Bormann and Graham 1959). However, there is growing evidence that distance between trees constitutes the most important factor as it leads to intimate contact between roots of different trees (Fraser et al. 2005; Külla and Lõhmus 1999; Reynolds and Bloomberg 1982; Tarroux and DesRochers 2010).

The importance of root grafting resides in its ability to affect physiology and ecology of the paired trees (Lev-Yadun 2011). It can allow survival of stumps connected to residual trees after thinning (Bader and Leuzinger 2019; Tarroux et al. 2010), provide mechanical support and windthrow resistance to connected trees (Basnet et al. 1993) or enable trees to better cope with the negative impacts of insect outbreaks (Salomón et al. 2016). Connected trees can directly share water, nutrients or photosynthates (Fraser et al. 2006; Schultz and Woods 1967; Stone and Stone 1975), in turn affecting radial growth of connected neighbors (Adonsou et al. 2016b; Tarroux and DesRochers 2011).

Natural root grafts have also been found in tree plantations (Tarroux and DesRochers 2010). Given that tree spacing affects tree growth (Benomar et al. 2012), one can wonder on the role that root grafting could play in tree relationships in plantations under different spacing conditions. Such relationships may play an important role in fast-growing hybrid poplar (Populus spp.) plantations, considering their shorter rotation length and the extensive financial investments that are made to maximize growth rates and productivity of trees. To our knowledge, root grafting in hybrid poplar has only been mentioned once (McIvor et al. 2008) in a Veronese clone (Populus euramericana: [Populus deltoides L. × Populus nigra L.]), but its frequency and timing of initiation has never been investigated. Since root grafting is frequent in natural stands of balsam poplar (Populus balsamifera) and trembling aspen (Populus tremuloides) (Adonsou et al. 2016a; Jelínková et al. 2009), we hypothesized that root grafting would also occur in hybrid poplar and that reduced spacing between trees would increase its occurrence. This study thus aimed at determining the occurrence of natural root grafts in hybrid poplar plantations and if present, how they would affect tree growth. Fifteen-year-old plantations were hydraulically excavated to reveal root systems and grafts, and their presence was characterized for two clones.

Materials and methods

Study site and experimental design

The study area was located in the boreal region of Abitibi-Témiscamingue, Western Quebec, Canada near the town of Amos (48° 34′ N, 78° 08′ W, 310 m). This site was chosen because it was located on a gentle slope near the Harricana River that allowed hydraulic excavation of root systems of a hybrid poplar plantation. The latter was established in 2003 on an abandoned agricultural field, dominated by herbaceous vegetation and shrubs [speckled alder (Alnus incana subsp. Rugosa (Du Roi) Clausen), willow (Salix spp.) and sparse trembling aspen (Populus tremuloides Michx.)] and was growing on a heavy clay soil. The mean annual temperature and precipitation are respectively 1.5 °C and 929 mm and the mean number of growing degree days are 1  423 [above 5 °C, 30 years mean (1981–2010)] (Environment and Natural Resources Canada 2018). Mechanical site preparation consisted of ploughing at 30 cm depth and then disking to level the soil surface and incorporate organic matter. Weedy vegetation was controlled mechanically twice a year during the first five growing seasons. Complete site preparation and planting techniques were reported by Guillemette and DesRochers (2008). Planting material consisted of dormant bareroot hybrid poplar trees of two clones: 747215 (Populus balsamifera L.  × Populus trichocarpa Torrey & A. Gray) and 915319 (Populus maximowiczii A. Henry × Populus balsamifera L.), planted at 30-cm depth. Thirty-six trees of each clone were planted in monoclonal blocks across two spacings (1 × 1 m and 3 × 3 m), equivalent to approximatively 10,000 and 1000 stems ha−1. After 15 years, trees had reached a mean diameter at breast height (diameter at 1.3 m, DBH) under the 1 × 1 m spacing of 9 cm and 11.1 cm respectively for clones 747215 and 915319; and 13.2 and 14.8 cm under 3 × 3 m spacing. Mean height was respectively 10.1 m and 13 m for clones 747215 and 915319 under the 1 × 1 m spacing and 9.2 m and 13.2 m under the 3 × 3 m spacing. Mean stem volume was respectively 0.03 m3 and 0.05 m3 for clones 747215 and 915319, corresponding to a yield of approximatively 308.6 and 570 m3 ha−1, under the 1 × 1 m spacing. Under the 3 × 3 m spacing, mean stem volumes were 0.06 and 0.10 m3 for clones 747215 and 915319, corresponding to yields of 64.4 and 112.2 m3 ha−1. Root characteristics were detailed for each clone and spacing (Supplementary Tables 1 and 2).

Fieldwork

Sampling was done between May and July 2018. DBH, basal diameter and height of each tree were measured to determine the influence of tree size on root grafting (or vice-versa). We calculated each tree stem volume outside the bark as the sum of the volume of two sections—from the base to breast height and from breast height to the top the tree—using the following equations:

$$V_{1} \left( {{\text{m}}^{3} } \right) = \left( {\frac{\pi }{12}} \right)\left( {d_{1}^{2} + d_{2}^{2} + d_{1} d_{2} } \right)L$$
(1)
$$V_{2} \left( {{\text{m}}^{3} } \right) = \left( {\frac{\pi }{12}} \right)d_{2}^{2} L$$
(2)

where V is the volume of the section, d1 is the basal diameter, d2 is the DBH, and L is the length of the section. In Eq. (1), L corresponds to 1.30 m (Perron 1996), while it is the tree height minus 1.30 m in Eq. (2) (West 2009).

Slenderness index, an indicator of tree stability and resistance to windthrow (Wang et al. 1998), was also computed as the ratio of tree height (m) to DBH (m). Trees were cut down with a chainsaw and cross-sectional disks of the stems were collected at ground-level (0 m). Hydraulic excavation with a high-pressure forestry water pump (Mark III; Wajax, Lachine, Quebec) was done to expose root systems and root grafts of each tree (Tarroux and DesRochers 2010) (Fig. 1a). Small hybrid poplar roots (< 2 cm) and herbaceous roots were cut and pulled by hand to ease the hydraulic excavation. The 1 × 1 m plots—containing 35 and 32 surviving trees in clones 747215 and 915319, respectively, were excavated (25 m2). Since it would have been too cumbersome to excavate the whole planted areas in the 3 × 3 m plots (324 m2), we excavated the roots of ten randomly chosen trees by following each root from the stump to where it became < 2 cm in diameter for each tree. In total, the root systems of 87 trees were excavated and we decided to exclude from the analyses nine trees that had died some years before the excavation. Root systems were mapped by hand and cross-sections of each proximal root > 2 cm nearest to the stump were collected and its diameter recorded. Proximal roots (Domenicano et al. 2011) are referred here as the roots that start directly from the stump and which further branch into subsequent roots or secondary segments. These roots are larger close to the stem base allowing easier growth ring visibility and avoiding missing rings (Krause and Morin 2005).

Fig. 1
figure 1

Photographs of a hydraulic excavation of a 1 × 1 m plot of clone 747215 (Populus balsamifera × Populus trichocarpa), and b root graft example between hybrid poplar trees under the 1 × 1 m spacing for clone 915319 (Populus maximowiczii × Populus balsamifera)

Full size image

Laboratory work

Stem and root discs as well as grafts were collected and brought to the laboratory for dendrochronological analyses. All discs were air-dried and sanded progressively from 80 to 400 grit to enhance the visibility of growth rings. Ring counts were done to confirm tree and root ages using a binocular microscope. Root diameters were measured using a Velmex unislide table (Bloomfield, New York, USA). As hybrid poplar roots were mostly elliptical with a T-shape, the largest and shortest perpendicular diameter were measured and averaged to obtain root diameter (Ruel et al. 2003). Root cross-sectional area (CSA) was calculated using the surface area of a circle:

$${\text{CSA}} = 2\pi \left( \frac{d}{2} \right)^{2}$$
(3)

where d is the averaged root diameter (Ruel et al. 2003).

We also computed the mean diameter of roots per tree and per plot. Total root surface of one tree was calculated as the product of the number of roots and the average root diameter (Bilodeau-Gauthier et al. 2013). Height-root ratio of each tree was calculated as the ratio between height and total root surface indicating the importance of tree height compared to the roots below-ground (Bilodeau-Gauthier et al. 2013).

Root grafting was assessed at three levels: the presence, frequency and age of each root graft. First, root grafts were validated with visual confirmation of vascular fusion of the cambium of the two roots by cutting through each graft with a wood saw. Root grafting was recorded when absent (0) or present (1) for all trees during the survey. The number of root grafts per tree was assessed by noting 0 (when absent) or the counted number. Since root grafting can take several years to complete (Tarroux and DesRochers 2010), the first year of graft formation was identified as the ring where tissues began to merge, i.e., the year following the last complete and independent growth ring in both roots. The last year of graft formation was identified as the first complete common growth ring surrounding the two roots (Tarroux and DesRochers 2010). As root grafting might also be related to tree exposure to wind (Keeley 1988), the location of the grafted trees within each plot was noted as interior or exterior: exterior trees were part of the bordering row of plots, while interior trees grew inside the plots. The age of root grafts allowed us to determine that all root grafts found in this study were only at the beginning of their formation; the effect of root grafting on tree growth could thus not be assessed at this time.

Statistical analysis

Statistical analyses were conducted with R software 3.6.0 version (R Core Team 2019), using a significance level of p < 0.05. The influence of spacing, clone, tree size and root size on root grafting was investigated. To reduce the number of variables, an exploratory visualization of correlations between variables was done using principal component analysis (PCA). The PCA results confirmed that some variables could be excluded as they were strongly correlated with one another (Fig. 2). Slenderness Index was strongly negatively correlated with mean root diameter and CSA. Basal diameter and DBH were also positively correlated with stem volume and root surface and negatively with height-root ratio. Finally, the number of roots was positively correlated with tree height. We thus only retained tree volume, number roots > 2 cm and CSA as they included tree and root sizes that might influence root grafting (Reynolds and Bloomberg 1982). Differences in tree volume, number of roots and CSA between grafted and non-grafted trees were statistically tested using a two-way analysis of variance with clone and grafting status as categorical predictors, followed by a Tukey–Kramer multiple comparisons test when independent variables were significant. Tree volume, number of roots and CSA were further added as predictors to test the models explaining the presence of root grafts (PRESENCE), the number of grafts per tree (NUMBER) and the timing of root grafting initiation (TIMING).

Fig. 2
figure 2

Principal component analysis between grafted and non-grafted trees under the 1 × 1 m spacing. Dotted horizontal and vertical lines represent the origin of the axes. CSA root cross-sectional area, DBH diameter at breast height

Full size image

Stepwise regression selection was used to find the most appropriate model explaining root grafts PRESENCE, NUMBER and TIMING by testing the model containing all the predictors (clone, spacing, tree volume, CSA and number of roots) and removing the least significant variable (Pekár and Brabec 2016). Additionally, generalized linear models were used to test the influence of tree size and spacing on root graft formation. PRESENCE was best represented with a binomial distribution, while NUMBER and TIMING were characterized by Poisson distributions (Table 3). Generalized linear models create estimated values that correspond to response constraints and allow nonlinear relationships between predictors and fitted values (Pekár and Brabec 2016). To reduce bias, we used the “brglm” package when there was uniform root grafting in some spacings (too many zeros in the matrix) (Kosmidis 2007). Otherwise, only the trees in the 1 × 1 m spacing plots were introduced in the models. Models assumptions such as deviation of residuals from normality, homoscedasticity, and independence of variables were checked prior to doing the tests (Pekár and Brabec 2016). Model selection was performed among all plausible models for each response variable, based on the Akaike information criterion corrected for small sample sizes (AICc). Best fit statistical models were supported by the lowest AICc and the highest Akaike weights which is a model having the greatest statistical support (Burnham and Anderson 2004). If more than one model had a delta AICc < 2, the multimodel inference was done to merge both plausible models.

Results

Intraclonal root grafts were found in the two excavated 1 × 1 m plots, while there were no grafts under the 3 × 3 m spacing. Overall, 38% of trees in the 1 × 1 m spacing were grafted to at least another tree. Grafted trees were connected to one, two or three other trees and the mean number of root grafts per tree was 1.14 (Table 1). Grafts had formed between trees distant of 1 m, while no graft was found between trees that were planted more than 1 m away within plots. There were significantly more grafted trees in clone 915319 than in clone 747215 (70% vs. 17%; Table 1).

Table 1 Root grafting characteristics and tree and root sizes in a 15-year-old hybrid poplar plantation for the 1 × 1 m spacing and the two clones
Full size table

The majority of root grafts (64%) had only started the grafting process, i.e., the last growth ring was shared in part between two roots, but the graft was still incomplete (not a single growth ring around both roots). The other 36% of root grafts were complete but had only recently completed (between 1 and 6 years ago at the time of sampling). For this reason, the influence of root grafting on tree growth could not be assessed at this time. Most of the grafts were formed when lateral roots of trees came in contact with those of other trees (Fig. 1b) or bumped into the stump of a neighboring tree. Roots also grafted when proximal roots of one tree were intercepted between the ramifications of other trees. Root grafts were established when trees were between 10 and 15 years old, for both clones (Table 1). Older (p < 0.05) and wider (p < 0.05) roots grafted later and more frequently than younger and smaller roots (Tables 2, 3).

Table 2 Chosen models to explain root grafting in a 15-year-old hybrid poplar plantation using stepwise regression with significance value for each logistic regression
Full size table
Table 3 Models explaining root graft presence, number of grafts and timing of root grafting according to results of small sample adjusted Akaike Information Criterion
Full size table

The first two principal components accounted for 86.1% of the total variance in the data. The first principal component explained 65.6% of the variance and was associated with tree size (DBH, basal diameter, stem volume). The second principal component explained 20.5% of the variance and was mainly associated with root size (CSA, root diameter and number of roots). Grafted and non-grafted trees were mostly differentiated by size, showing that grafted trees were larger and had larger root systems than non-grafted trees, that had greater slenderness indices and height-root ratios (Fig. 2; Table 1). Grafted trees of clone 915319 were larger (DBH, height, volume) (F1,58 = 17.1, p < 0.001, Fig. 3a; Table 1) than non-grafted trees, while there was no significant difference in size between grafted and non-grafted trees of clone 747215 (Fig. 3a). Though grafted trees had slightly wider and more roots in clone 915319, the difference was not significant from non-grafted trees (F1,58 = 0.091, p = 0.75; Fig. 3b, c; Table 1). Likewise, grafted trees of clone 915319 also had greater stem volumes and number of roots than grafted trees of clone 747215 (Fig. 3a, b). Mean CSA of grafted trees was also greater than for non-grafted trees (F1,58 = 16, p < 0.001, Fig. 3c), but the difference was not significant for both clones (Fig. 3c). The following model (4) predicting the probability of finding root grafts according to tree volume and spacing had the best fit (p < 0.001, pseudo-R2 = 0.35, Fig. 4):

$${\text{Predicted}}\;{\text{logit}}\;{\text{of}}\left({{\text{root}}\;{\text{grafting}}\;{\text{presence}}} \right) = - 3.53 + \left( {70.27*{\text{volume}}} \right) + ( - 7.57*{\text{spacing}})$$
(4)
Fig. 3
figure 3

Mean a tree volume, b number of roots and c root cross-sectional area for grafted and non-grafted trees of two hybrid poplar clones (915319: Populus maximowiczii × Populus balsamifera, 747215: Populus balsamifera × Populus trichocarpa) planted under the 1 × 1 m spacing. Error bars indicate standard errors of the mean (n = 58). Different letters between bars within a graph indicate significant differences. CSA root cross-sectional area

Full size image
Fig. 4
figure 4

Prediction of the probability of root graft presence in relation to stem volume at 1 × 1 m spacing for clones 747215 and 915319 (747215: Populus balsamifera × Populus trichocarpa, 915319: Populus maximowiczii × Populus balsamifera). Points are the observed values of stem volume for trees with or without root grafts. Dashed and full lines show the predicted values of the probability of root grafts presence, respectively for clone 747215 and 915319. Grey colors indicate the 95% confidence interval for each prediction. Horizontal dotted line represents a probability of 50% of finding a root graft

Full size image

According to the model, log of the odds of a tree forming root grafts was positively related to tree volume (p < 0.001) and negatively related to spacing (p < 0.001; Table 2). In other words, larger trees planted close together were associated with a greater frequency of root grafting for both clones (Table 2). An increase in tree volume of 0.01 m3 increased the probability of root grafting by a maximum of 20%. Among grafted trees, clone and volume had significant effects on the number of root grafts per tree (Table 2). Trees of clone 915319 had five times more root grafts per tree than clone 747215 (Table 2). Furthermore, the number of root grafts per tree doubled with a tree volume increase of 0.01 m3 (Table 2). Root grafts were mostly found at the outer part of the plots. For both clones, 86% of the grafted trees were found in the border rows (Table 1). For both clones, tree and root size such as stem volume (F1,58 = 96.07, p < 0.001), slenderness index (F1,58 = 87.1, p < 0.001), number of roots (F1,58 = 13.1, p < 0.001) and root CSA (F1,58 = 18.0, p < 0.001) of trees that were positioned in the exterior of the plots were greater than the interior trees. Self-grafts—i.e., grafts between roots of the same trees—were frequently observed but were not recorded.

Discussion

Our study showed that intraclonal root grafting occurs in hybrid poplar plantations, with an average of 38% of trees grafted to a neighbor in the 1 × 1 m plots. Although clone and stem size were significant factors in determining the presence of root grafts (Table 2), tree proximity appears to be determinant since no root grafts were found under the 3 × 3 m spacing. Trees further apart have less contact between roots and take longer to reach one another (Tarroux and DesRochers 2010). Moreover, crossing roots from further apart trees probably undergo less friction between them under the action of wind, required to break the bark and allow the fusion of cambium tissues (Cook and Welch 1957). It is thus expected to find more root grafts as trees (and roots) grow older and larger (Armson and van den Driessche 1959; Reynolds and Bloomberg 1982), increasing the chances for roots to cross and fuse. In natural stands of balsam poplar aged 43–103 years old, Adonsou et al. (2016a) found that approximately 50% of trees were connected by root grafts and that the mean distance between grafted trees was 1.14 m, while Tarroux and DesRochers (2010) found that 90% of root grafts occurred between jack pine trees less than 2 m apart. Natural stands also tend to have more clumped spatial arrangements where stems are closer to each other, enhancing the occurrence of root grafts compared to plantations where trees are distributed more evenly (Tarroux and DesRochers 2010).

Tarroux and DesRochers (2011) showed that grafted trees generally had better growth than non-grafted trees. However, since most grafts had only started to form or were just recently completed, we could not assess the impact of root grafting on the growth of our trees. Nonetheless, root grafting initiated relatively early, when trees were 10 years old and roots were around 4 years old (Table 1). Dendrochronological analyses allowed us to determine that grafted trees of clone 915319 were larger than non-grafted trees at the time of root grafting (Table 1), suggesting that trees that would later form root grafts had better growth rates than trees that would not (Tarroux and DesRochers 2011). Trees of clone 747215 were generally smaller and formed less root grafts (Fig. 4). Larger roots also tended to form more root grafts but showed high variability; and differences in root sizes between grafted and non-grafted trees were not statistically significant. The two clones also had different root morphologies with the root system of clone 747215 composed of one large root (great CSA) and other small roots, whereas roots of clone 915319 all having similar CSA. Finally, the root systems of grafted trees also tended to be composed of more roots than the non-grafted trees (Table 1), again increasing the chances for roots of different trees to cross (Reynolds and Bloomberg 1982). Populus maximowiczii hybrids, especially clone 915319, were reported to have higher growth plasticity compared to clone 747215, taking advantage of the available space (Benomar et al. 2012). Moreover, clone 915319 had older roots on average (Supplementary Table 1), meaning that this clone produced more roots earlier in stand development allowing roots to be in contact earlier than for clone 747215.

We found that exterior trees were larger than the interior ones and they formed more root grafts. Ecological conditions at the edges of plantations may substantially vary from the interior and phenomena such as inter-tree competition and mortality may, therefore, be modified (York et al. 2003). Because the border trees receive more light and nutrients, it is not unusual for them to be larger than interior trees in plantations (Langton 1990). Interestingly, 86% of root grafts were formed by exterior trees, even though they have fewer neighbors than trees inside plots, hence less chance for roots to meet. As shown in our predictive models (Tables 2, 3) tree size was linked to root grafting occurrence and as tree DBH and height (stem volume) increased, the probability of finding root grafts increased. Grafted and border trees also had lower slenderness indices, i.e., lower ratios between height and DBH, showing that border trees were more stable from mechanical disturbances than interior trees. It could be argued that if root grafts conferred increased stability to trees (Basnet et al. 1993; Keeley 1988), grafted trees would rather have greater slenderness indices. Perhaps here the root grafts were too recent to have impacted tree shape.

In conclusion, this study showed that root grafting readily occurs and that root grafts are initiated early in hybrid poplars. Roots grafts were more prevalent in clone 915319, in trees that had greater above and belowground growth. Since no root grafts were found in the larger spacing, our study strengthens the postulate that tree proximity enhances the frequency of root grafting. In the event that root grafting is beneficial to tree growth or detrimental in the case of diseases spreading through root grafts (Gleason and Mueller 2005) in hybrid poplars, planting density should be carefully considered by managers.

Author contribution statement

AD and DTG conceived and designed the experiment. DTG collected the data and carried out the laboratory work, analyzed and interpreted the data and prepared the manuscript. AD supervised, reviewed and edited the manuscript. AD approved the final version of the paper to be submitted.