Introduction

The root systems of trees are at the source of many of the beneficial effects on soil expected from agroforestry practices, as well as anticipated negative effects on crop yields, mechanised operations, and plugging of drainage tiles. Yet tree roots in agroecosystems are still poorly understood. For example, these systems are able to sequester carbon, improve soil quality by the return of organic matter and microorganisms (Doblas-Miranda et al. 2014), and improve water quality by acting as a safety net to limit the leaching of excess organic or chemical inputs (Alam et al. 2014; Bergeron et al. 2011). They may also function as a hydraulic pump, drawing to the surface both water and nutrients stored so deeply that they are often inaccessible to crops (Allen et al. 2004). However, tree roots can also compete with the intercrop root system for soil water and nutrients (Casper and Jackson 1997; Zamora et al. 2008). In temperate zones, where agricultural lands usually have high levels of fertility, light may become the limiting factor (Jose et al. 2004). For example in China, Chirko et al. (1996) found that shading significantly reduced winter wheat yield near tree rows in a Paulownia—winter wheat temperate alley cropping system.

However, studies in North America have demonstrated the importance of soil resources competition for the tree-crop interface, primarily in terms of water use, and thus the absorption of nutrients (Rivest et al. 2010; Wanvestraut et al. 2004).

According to the “central hypothesis” supporting agroforestry practices, TBI systems demonstrate improved total productivity (compared with monocultures) if trees have the ability to capture soil resources in horizons other than those used by crops (Cannell et al. 1996). Therefore research and development of agroforestry, especially for TBI systems, should focus on the selection of trees with genetically determined deep roots, as well as phenotypic plasticity, i.e. with the ability to avoid competition by growing deeper fine roots (Mulia and Dupraz 2006; Schroth 1995). The choice of species to be used as an intercrop is also important, because its competitiveness should influence the development of deeper tree roots.

The main objective of this research project was to explore the belowground interactions governing an intercropping system by observing the spatial distribution of fine roots of two species of deciduous trees (red oaks and hybrid poplars) with different ecological characteristics and a perennial crop in a temperate northern climate. We also evaluated agricultural yields and light microclimate with respect to the presence of and distance from trees, to assess the effect of trees on yield, and partition that into above and belowground competition.

Our hypotheses were that (1) the root systems of trees associated with an intercrop develop more deeply than those of trees not associated with an intercrop; so that (2) there is a spatial separation between the root systems of trees and hay crop. Finally, (3) competition for light, not belowground, explains the reduction in hay yield nearest tree rows.

Materials and methods

Site and climate

The study site is located near the town of Saint-Paulin (46°27′6″N; −72°59′26″E), in the administrative region of Mauricie, southern Quebec (Canada). The average temperature is 19.6 °C in July, and −13.2 °C in January; average annual precipitation is 1,087 mm (Meteorological Service of Canada 2006). The soil is a mixture of Pontiac silt loam and Chapleau clay loam (Pageau 1967). Soil analysis revealed three distinct layers. The topsoil (excluding the organic layer and to a depth of 31 cm ± 7 cm) is a sandy loam (80 % sand, 16 % silt and 3 % clay) with a pH of 5.0. The second layer is sand (86, 10 and 4 %, respectively) with a pH of 5.5, and extends to a depth of 63 cm (±21 cm). The deepest layer is a loam (49, 36 and 15 %) with a pH of 5.5.

Nutritive elements in the soil (Table 1) were determined in 2012 in the first 15 cm (including the organic layer) using the Mehlich III method (Mehlich 1984). These analyses indicated a soil from relatively poor to moderately fertile, as mineral elements (P, K, Ca and Mg) are present in low quantities (CRAAQ 2010). There is also a probability of aluminum toxicity, as soil pH is relatively low and aluminum concentration is high (Table 1). Phosphorus saturation rate is quite low (4 %), indicating a possible immobilization of phosphate ions by aluminum, which could result in an absorption deficiency, especially for hay (Delhaize and Ryan 1995).

Table 1 Nutritive elements present at a 15 cm depth at Saint-Paulin, Quebec, in 2012 (Mehlich-III method)
Full size table

Experimental design

The tree-based intercropping system (TBI) was initiated in 2004 and comprised two hardwood species, red oak (Quercus rubra L. (QUR)) and black cherry (Prunus serotina Ehrh), and two fast-growing hybrid poplar clones (HYP), DN3333 (Populus deltoides × nigra, cv. Stormont, Ontario, Canada) and DN3570 (Populus deltoides × nigra, no cv. name, Belgium). In 2011, QUR and HYP were on average 3.2 and 9.3 m tall, respectively, with diameters at breast height (DBH) of 4 cm and 12.7 cm. Tree rows, spaced 12 m apart, were covered with a black plastic mulch (90 cm wide), in an approximately North–South orientation (314 °N), with hybrid poplars and hardwoods on alternate rows. Hybrid poplars were spaced 2 m apart within the row in alternating groups of 12 or 15 individuals of the same clone, whereas hardwoods were spaced 3 m (for rows with QUR and black cherries, in groups of four) or 4 m apart (for rows with QUR only). The experimental design comprises areas with no trees (minimal size of 48 m × 24 m), for crop monoculture (CM), in each of four replication blocks (Fig. 1).

Fig. 1
figure 1

Example of an area used for the present experiment indicating tree and crop fine root sampling positions in different growth conditions. One such area was designated in each of the four experimental blocks

Full size image

The intercrop, initiated in 2009, was an organic hay of Thimoty grass (Phleum pratens L.) and red clover (Trifolium pratense L.). Crops cultivated prior to 2009 included buckwheat (2005, 2006 and 2008), canola (2007) and oat (2004). The soil is not irrigated and has been fertilized for the last time in 2007 with horse manure.

Analysis of tree fine roots was conducted only for QUR and one HYP clone (DN3570). In each block, trees and hay were studied as a function of growth conditions: combined in TBI or separated in CM and TWC (trees without crop). TWC were situated at row ends and therefore partially not associated with the crop (Fig. 1).

Fine root sampling

Fine roots (≤2 mm) were sampled with the Trench Profile Method (Böhm 1976; Van Noordwijk et al. 2000) to a depth of 1 m during the summer of 2011. Trenches, ~1.5 m wide × 1.2 m deep, were dug using a small excavator at each of the 4 positions illustrated on Fig. 1, roughly 3 (CM), 6 (TWC) and 12 m (TBI) in length (Fig. 2). Trench walls were first smoothed with a sharp object and cleaned with a wide paintbrush. Sampling involved root intersection counting (roots intersections per m2 of observation plane) on a vertical profile using grids 50 × 100 cm. Grid cells near the surface were 10 cm wide × 5 cm high, whereas at depths below 20 cm, grid cells were 10 × 10 cm. Fine roots of TBI and TWC were sampled at several distances from the target tree stem: 1.5, 3.5 and 5.75 m, corresponding to the center of the grid (Fig. 2). In TBI, the distance position 5.75 m from QUR is the same position as that of 6.25 m from HYP (Fig. 2). Also, to compare crop FRLD in TBI and in CM, data at 1.5 and 3.5 m for each tree species were pooled into mean values TBI-QUR and TBI-HYP (Fig. 2). Data at 5.75 m from both tree species were pooled to compute mean value TBI-CENTER (Fig. 2). These combinations were possible because values were statistically similar for most depth levels (data not shown).

Fig. 2
figure 2

Sketch of one transect in a tree-based intercropping system (TBI) between a red oak (QUR) and a hybrid poplar (HYP), showing sampling distances and pooling (TBI-QUR, TBI-HYP and TBI-CENTER)

Full size image

For each grid cell, root intersection counting is noted for each tree species and crop (with no distinction for the two hay crop components). QUR, HYP and CROP roots were distinguished according to color and shape criteria. Hay roots are usually white with a translucent aspect. Tree roots have a fine bark, brown or light yellow for HYP, and reddish for QUR.

Soil volumes (650 mL) containing fine roots were taken with a round recipient with sharp edges in three cells from seven grids (one in CM, four in TBI, two in TWC) in each experimental block, to obtain samples with low to high density of fine roots for each species. Samples were stored in a freezer (−35 °C) until laboratory analysis. They were then washed with tap water using two sieves (2 × 2 and 1 × 1 mm mesh size) to avoid fine root losses.

Fine root length density (FRLD, km m−3) for each soil sample and for each species was determined with WinRhizo (Regent Instrument, Quebec, Canada, 2012a version) after scanning roots at 400 dpi (Epson Perfection V700, Nagano, Japan). FRLD of each species was then estimated for all cells of all grids following the core break method (Van Noordwijk et al. 2000) by calibration (simple linear regression with n = 33 for QUR and HYP, n = 34 for CROP) according to root intersection counting. The relationship between RIC and FRLD is FRLD = 0.002 × X × root intersection counting, where X is a factor representing preferential orientation of roots in soil (Aigner and Ziegler 2006; Mulia and Dupraz 2006; Van Noordwijk et al. 2000). Root intersection counting data were not distributed normally and had a positive asymmetry; they were thus transformed by 4√ to fit normality requirements.

Fine roots were dried for 48 h at 38 °C then weighed to determine specific root length (SRL, m g−1).

Crop aboveground dry biomass

Aboveground crop biomass was collected in the 4 experimental blocks from 0.25 m2 quadrats, on June 27–28, 2011. The center of each quadrat was disposed at different distances (1.5, 3.5 and 5.75 m) between a QUR and a HYP. In CM areas, 2 samples of the same size per block were collected. The crop was cut with scissors, deposited in paper bags, dried for 48 h at 38 °C and then weighed (g m−2, later converted to t ha−1).

Light conditions

In order to estimate light availability for the intercrop, hemispherical photographs were taken on July 26–27, 2011, along transects in each experimental block. Pictures were taken at 1 m from the ground with a digital camera (Nikon Coolpix 990, Tokyo, Japan) equipped with a fisheye lens (Nikon fisheye converter FC-E8, Tokyo, Japan) at different distances from tree rows (0, 1, 2, 3 and 6 m, east and west of tree row). Hemispherical photographs were also taken in the CM areas. All pictures (n = 72) were analysed with gap light analyser (GLA) Version 2.0 (Frazer et al. 1999) to determine the gap light index (% of total light transmittance, PTLT). Since the pictures were not taken at the same distances than those for root and biomass sampling, the light indices obtained in TBI were interpolated by a quadratic linear regression of light according to distances (data not shown).

Data analysis

Statistical analysis was conducted with JMP 8.0 (SAS Institute Inc., Cary, NC, USA). Analyses of variance were performed for each depth using a mixed model with a randomized complete block design. Possible factors were agronomic system (TBI, TWC, or CM), species (QUR, HYP or CROP) and distances from the tree rows (1.5, 3.5 and 5.75 m). Box Cox transformation was applied to tree FRLD data (FRLD+1) in order to satisfy normality and homogeneity of variance requirements. Results were considered significant at p < 0.10. Means comparison tests were performed by contrasts or Tukey HSD tests. Figures illustrating root profiles present means without transformation.

Variance partitioning was used to determine the fractions of crop yield variance that could be attributed to the single and combined effects of light and tree FRLD using adjusted R2 ratios (Peres-Neto et al. 2006). The probabilities associated with the different fractions were computed using partial regressions and 999 permutations of the residuals under the reduced model, except for the common fraction which cannot be tested (Anderson and Legendre 1999). These analyses were carried out within ‘R’ using ‘varpart’, ‘rda’ and ‘anova.cca’ functions from the ‘vegan’ library (Oksanen et al. 2013).

An index of tree FRLD by unit of DBH (diameter at breast height, km m−3 cm−1) and an index of crop FRLD by unit of aboveground crop biomass (km g−1) were calculated and compared according to multiple factors (growth conditions, distances, species).

Results

Calibration

Results of the simple linear regressions of FRLD and root intersection counting are summarized in Table 2. R2 was 0.23, 0.62 and 0.97 for QUR, HYP and CROP respectively. Constants for each equation were not significantly different from 0 (two-tailed t test with α = 0.05).

Table 2 Simple regression of fine root length density and root intersection counts results and calibration factors (X) of the core break method (Van Noordwijk et al. 2000) for red oak (QUR), hybrid poplar (HYP) and the hay crop (CROP)
Full size table

Fine root profiles of tree species in TBI and TWC

Results of variance analysis are presented in Table 3. Values of FRLD obtained for HYP were higher than those for QUR (Fig. 3b–d). This was particularly true in topsoil, where QUR FRLD ranged from 1.4 to 12.9 km m−3 (Fig. 3b) and HYP FRLD ranged from 15.4 to 55.7 km m−3 (Fig. 3b, c). Size differences between the two tree species explained the difference in FRLD as HYP were three times higher than QUR. However, HYP still produced more fine roots than QUR per unit size; FRLD per unit DBH was significantly higher (0.57 for QUR and 0.83 for HYP, p = 0.0131, no treatment or interaction effects). SRL was also higher for HYP than QUR, with respectively 80 and 37 m g−1 (p < 0.001).

Table 3 Per depth level p values* of variance analysis of red oak (QUR) and hybrid poplar (HYP) fine root length density (FRLD) with treatment** (T, agroforestry or not) and distance (D, 1.5, 3.5 and 5.75 m (only for HYP)) as factors and cumulative percentage of fine roots with tree species (S) and treatment as factors
Full size table
Fig. 3
figure 3

Fine root distribution according to depth of a cumulative percentage of red oak (QUR) and hybrid poplar (HYP) in tree-based intercropping system (TBI) and for trees without crop (TWC); b QUR fine root length density (FRLD) in TBI and TWC at 1.5 and 3.5 m; c HYP FRLD in TBI and d in TWC at all distances. For each depth, different letters indicate significant differences at p < 0.05 (Tukey HSD test). Ovals regroup similar values. Shades represent the three soil horizons

Full size image

HYP fine roots extended further laterally (up to 6.25 m in TBI system) than those of QUR, with no fine roots present at 5.75 m in the TBI system and very few in TWC (not included in the analysis).

QUR and HYP showed a high proportion of fine roots in the first 10 cm of soil in treatments (TBI and TWC), representing respectively 43 and 40 % of the total root system (Fig. 3a). Cumulative percentage of fine roots at each depth level tended to differ between tree species from a depth of 25 cm, and this difference was significant (p < 0.10) at 35, 45 and 65 (Table 3 and Fig. 3a). Within the first 25 cm, QUR displayed 95 % of its fine roots, whereas HYP fine roots went deeper, with 95 % within the first 45 cm. Distance had no influence on root cumulative percentages in TBI or TWC (p > 0.10, results not shown), but FRLD decreased with distance from the tree stems for both species (Table 3 and Fig. 3). For HYP, FRLD values were similar at 1.5 m and 3.5 m, but decreased at 5.75 m (Fig. 3c, d).

HYP fine root cumulative percentage was very similar at each depth level in TBI and TWC (Fig. 3a), but there was a tendency for interactions between treatments (TBI or TWC) and species (QUR and HYP) at 2.5 cm depth (Table 3). Actually, in topsoil, QUR fine root cumulative percentage tended to be lower in TBI than in TWC. The same result appeared when comparing QUR FRLD in TBI and TWC: at 1.5 m from the tree stems, FRLD was lower in TBI than in TWC at 2.5 cm depth (p = 0.08), and this difference persisted to a depth of 20 cm (though not significant—Fig. 3b and Table 3). There was no difference at a distance of 3.5 m from the tree stems.

Spatial separation of red oak and crop roots in TBI

In addition to the reduction of QUR FRLD in topsoil at 1.5 m in the TBI system, we found a spatial separation between the fine root systems of QUR and the crop (Table 4). Results of ANOVA at different depth levels did not indicate interaction between distances and root species (QUR or CROP), but there were significant differences between root species. QUR had a lower relative root density than CROP at depths from 0 to 5 cm and from 0 to 10 cm. On the contrary, QUR relative root density was higher than crop at depths from 10 to 30 cm. This spatial separation was not observed for HYP and crop, with relative root densities statistically similar at each depth level, except from 5 to 10 cm and from 10 to 15 cm. HYP relative root density was significantly higher than that of crop, but only at 6.25 m from the tree trunk at 5 to 10 cm deep and at 10 to 15 cm deep (Table 4).

Table 4 p values of ANOVA (S root species, D distance and SD interaction) and average percentages of tree fine roots (red oak, QUR or hybrid poplar, HYP) and hay (CROP) fine root at different depths in a TBI system
Full size table

Crop fine root length density

Absolute FRLD values for crop were higher than those of trees. In the first 5 cm, they varied from 105 km m−3 near HYP (TBI-HYP) to 190 km m−3 in CM (Fig. 4). Hay, like trees, had a superficial rooting pattern, with more than 50 % of its fine roots located in the first 10 cm of soil and 95 % within 35 cm.

Fig. 4
figure 4

Hay fine root length density (FRLD) in monocrop (CM) and in tree-based intercropping system (TBI), near red oaks (QUR), in the center of the alley (CENTER) and near hybrid poplars (HYP). For each depth, different letters indicate significant differences with p < 0.05 (Tukey HSD test). Ovals regroup similar values. Shades represent the three soil horizons

Full size image

The presence of trees in TBI caused a decrease in crop FRLD with depth, but distinctions occurred mainly from 0 to 5 cm depth depending on the position in the alley (Fig. 4). Near HYP (distances 1.5 and 3.5 m combined), crop FRLD decreased significantly all along the vertical profile (p < 0.05) when compared with CM. In the first 5 cm, crop FRLD decreased by 45 % compared to CM. On the contrary, near QUR (TBI-QUR) and in the middle of the alley, the adaptation of crop roots to the presence of trees as FRLD in topsoil (5 cm) was maintained and statistically similar to CM (p > 0.10 with respectively 139, 178 and 191 km m−3; Fig. 4). However, at depths >5 cm, crop FRLD was in both cases lower than in CM (p < 0.05).

The index of crop FRLD by unit of aboveground biomass (km g−1) indicated that crop invested more resources in the production of fine roots at the expense of aboveground biomass when associated with HYP. Actually, the index was statistically similar in the middle of the alley, near QUR and in CM (respectively 0.12, 0.17 and 0.12), but it was three times higher near HYP (0.32, p < 0.05 with Tukey HSD).

Aboveground crop biomass and importance of light

In the center of the intercropped alley (5.75 m from QUR and 5.75 m from HYP) and at 3.5 m from QUR, hay yield was not statistically different to the yield obtained in CM (4 t ha−1, p > 0.10, Fig. 5). At 1.5 m from QUR, mean productivity was 2.3 t ha−1 (−42 %). Near HYP (1.5 and 3.5 m from trunk), hay productivity was strongly affected, as yield decreased by 75 % compared to CM (4 t ha−1, Fig. 5).

Fig. 5
figure 5

Above ground hay dry biomass (t ha−1) and percentage of total light transmittance at 1 m height from soil in monocrop (CM) and in tree-based intercropping system (TBI) according to distance (1.5, 3.5 and 5.75 m) from red oak (QUR) and hybrid poplar (HYP). Different letters indicate significant differences at p < 0.05 (Tukey HSD test)

Full size image

Tree presence also contributed to a decrease in the amount of light available (percentage of total light transmittance, PTLT) for the intercrop in comparison to CM (97 %) and according to distance from the tree line (Fig. 5). As HYP were taller and tighter on the rows, light reduction was stronger (PTLT from 65 % at 1.5 m to 75 % at 3.5 m). QUR also reduced light availability, but less so than HYP (e.g. 76 % at 1.5 m). Finally, TLTP was highest at 3.5 and 5.75 m from QUR, at 95 and 90 %, respectively (Fig. 5).

To explore the overlap and singular contributions of light availability and length density of tree fine roots accounting for variation in crop productivity, we used variance partitioning in ‘R’ (Oksanen et al. 2013). Both components together explained roughly 42 % of the total variance (Fig. 6). Nearly half of that (20 %) was shared between the two, as expected since they are indeed correlated. However, the fraction that could be attributed to the singular effect of light accounted for the better part of the remaining variance explained (22 %), leaving nothing to the singular contribution of tree roots.

Fig. 6
figure 6

Venn diagram of variation partitioning between components of competition with intercrop: percentage of total light transmittance (PTLT) and tree root FRLD. Adjusted R2 fractions are printed for each partition (and associated probabilities of testable fractions; n = 12,333) (Anderson and Legendre 1999). Residuals are shown at the bottom right. Note that some fraction may be negative due to two strongly correlated predictors with strong effects on Y of opposite signs (one positive and the other negative), as is the case here for the singular Tree FRLD fraction (true value was 0.03), making the overall sum of all fractions, including residuals, equal to 100 (Peres-Neto et al. 2006)

Full size image

Discussion

Agronomic conditions of the study site are not optimal for many agricultural crops, mainly because of a relatively reduced pH. However, this low level of fertility is characteristic of marginal farmlands. Proponents of agroforestry often claim to increase the overall productivity of such sites, by means of plantations of marginal and extensive crops cultivated in a sustainable manner (Ashton and Montagnini 2000). The study of root distribution of trees and crops under conditions of low resource availability is particularly relevant for evaluating the competition between trees and intercrop.

The study did not reveal deeper rooting of trees in the presence of crop (first hypothesis), but indeed a different rooting pattern in one of the tree species, resulting in better vertical separation of the two root systems (second hypothesis). As expected, crop yield decreased as a function of tree proximity, but competition for light seemed to be the main driver of that, not soil resources (third hypothesis).

Fine root length density calibration factor

The calibration factor of the core break method (X) estimated for hay was close to 1, which corresponds to the range of values published for durum wheat (1.1; Mulia and Dupraz 2006) and for maize (0.8–1.1; Van Noordwijk et al. 1995). It also indicates that the fine roots of hay have no preferential orientation. Nonetheless, Vepraskas and Hoyt (1988) reported a value of 1.9 for maize, and Logsdon and Allmaras (1991) observed a much higher value (8). Values of the calibration factor for HYP and QUR were significantly higher, respectively at 14 and 4.56 (Table 2). Mulia and Dupraz (2006) reported values of 0.46 for hybrid poplars (R2 = 0.32) and 0.61 for hybrid walnuts (R2 = 0.56) in TBI systems in the south of France. Moreno et al. (2005) reported a calibration factor of 0.75 (R2 = 0.42) for Holm oaks in Spain. For both of these studies the core break method was used on horizontal observation planes (in this study, observation planes were vertical) and the authors concluded in a preferential vertical orientation of trees fine roots, which our results also suggested. Actually, vertical roots have fewer chances to intersect with a vertical wall. Several other factors can explain high calibration factor values, including intensive root ramifications, as it is possible that not all the branched roots would intersect the observation plane (Bengough et al. 1992). High calibration coefficients can be a consequence of having underestimated the number of root tips during field surveys, since roots with very small diameters are not readily observable and can dry out and retract in open trenches (Chopart and Siband 1999).

The coefficients of determination R2 resulting from the regressions conducted in this experiment were relatively high, except for red oak. A low R2 has previously been observed for Mediterranean oaks (Moreno et al. 2005). The low values of R2 usually obtained in this type of study can be explained by the tendency of fine roots of trees to occur in aggregates (Graves et al. 2007).

Tree fine root distribution in a TBI system in a temperate climate

This study of tree fine root profiles in an agroforestry context showed no deeper or inverted root distribution as it has been shown for trees in a TBI system in the south of France, a region with a summer drought period (Mulia and Dupraz 2006). Also, in Spain, a study of widely-spaced oaks showed a uniform distribution of fine roots across all depths evaluated, up to 2 m (Moreno et al. 2005). In these areas, rainfall was insufficient to maintain surface moisture levels during the main growing season. Under such conditions, tree roots tended to extend into deeper, more humid horizons (Casper and Jackson 1997).

In the present study, trees had a higher root concentration in surface horizons, with more than 40 % of fine roots in the top 10 cm. Root density decreased rapidly with soil depth, as well as with distance from the stem. This distribution pattern is quite common for trees growing in northern temperate regions, either in natural forest (Canadell et al. 1996) or in plantations (Achat et al. 2008). In Indiana (USA), black walnut (Juglans nigra L.) and red oak in a TBI system also occupied the same soil horizons as corn (Zea mays L.), i.e. from 0 to 30 cm below the surface (Jose et al. 2000). Higher fine root densities at the surface are often associated with higher levels of nutrient-rich organic matter, generally with greater soil porosity than in other soil horizons (Schmid and Kazdas 2002). The organic layer’s capacity to retain moisture provides access to surface water and limits the need to seek water at greater depths, especially in a humid, temperate climate with sufficient rainfall during the growing season.

On the present study site, the soil texture is characterized by a very sandy layer superimposed on another, more clayey layer, which could physically limit trees vertical rooting ability. Possible toxicity due to soil acidity and high level of aluminum (Table 1) may also explain the low penetration of roots of trees (and those of hay as well). Indeed, aluminum is known to impede root elongation (Delhaize and Ryan 1995). Surface organic matter can create complexes with aluminum that can precipitate at greater depths through leaching. The acidic sand layer must certainly build such complexes, explaining the drastic decrease in root density evident in this horizon.

Spatial separation and competition

While rooting depths were similar for hay and red oak (95 % reached 35 and 25 cm respectively), high resolution analysis (vertical sampling steps of 5 cm at the surface) of the relative distribution of fine roots revealed a spatial separation between the two agroforestry components. In the presence of red oak, hay roots were concentrated in the first 10 cm below the surface, while the largest proportion of fine roots of QUR was at lower depths between 10 and 30 cm (Table 4). This spatial separation was not observed for hybrid poplar.

The two tree species have very different ecological characteristics, poplar being an early successional species and red oak a species of intermediate shade tolerance and mid to late successional status (Finér et al. 1997). Morphological and allometric differences between the two succession types have been identified for aboveground parameters (Claveau et al. 2002) but remain largely unknown at the underground level. However, studies have shown that root structure differs along successional gradients (Finér et al. 1997). Bauhus and Messier (1999) compared the ability of shade-tolerant tree species (Abies balsamea (L.) Mill., Picea glauca (Moench) Voss) and pioneer species (Populus tremuloides Michx., Betula papyrifera Marsh.) to exploit soil in a southern boreal forest. They concluded that early successional trees tended to produce thin (larger SRL) fine roots (≤2 mm) that are highly branched, which maximizes occupation of the soil volume. Finér et al. (1997) arrived at the same conclusion in a similar study (roots ≤ 10 mm), in which pioneer species had more branched roots and greater total length, while root biomass was comparable to that of late-successional species. According to Coll et al. (2008), an increase in root length unaccompanied by increasing biomass in early successional species suggests morphological, rather than allometric, differentiation. This strategy encourages greater soil exploitation, while allocating more biomass to aboveground growth. Our results appear to reflect this type of strategy, as poplar root length densities were much higher than those obtained by red oak, even when relativized by tree size. The average SRL for hybrid poplar is also higher than for red oak.

Soil preparation

Several studies have shown a reduction in fine roots due to tillage (Burgess et al. 2005; Livesley et al. 2000). According to Coker (1959), trees exhibit less root extension (lateral and vertical) when associated with agriculture, compared to trees established on uncultivated areas. Gray (2000) showed that root growth of hybrid poplar intercropping was significantly reduced at a depth corresponding to the annual ground work by farm machinery. Mulia and Dupraz (2006) observed that the impact extended beyond the area of tillage influence. The system in our study, a perennial crop requiring no annual soil preparation, is operated according to organic farming guidelines. Intercropping has undergone very little mechanical tillage since its establishment, the most recent one being in 2009, with hay sowing. The integration of activities like root pruning or tillage as practiced for annual crops could lead to an increased spatial separation of fine roots of trees and crop (Jose et al. 2004).

Importance of light availability

The TBI system in place on this site is less productive for the intercrop as currently structured. Hay biomass in the vicinity of trees, especially near hybrid poplars, is significantly lower than in the control (monocrop), mainly due to a decrease in the availability of light. Light does indeed seem to be the main driver of intercrop yield, in that it not only accounts for the effect of competition by roots (being correlated), but also has a singular effect that cannot be explained by root density. However there was still an important variance not explained by our simple partitioning model (Fig. 6). Indeed other factors were not evaluated dynamically (along gradients of time and space) in this study, such as air temperature and humidity, and soil moisture, temperature and nutrients, all of which could also have played a role in determining crop yield relative to the presence of trees. Our results are however still very useful in terms of management, as will be discussed below.

Several studies have concluded that competition for light is greater than belowground competition in temperate and humid climatic regions where crops are usually fertilized and rainfall is relatively abundant, ensuring an adequate water supply (Ding and Su 2010; Friday and Fownes 2002). In this study, water is not believed to be limiting as mean precipitations were 108 mm for June and July 2011 according to Environment Canada archives (http://climat.meteo.gc.ca/).

Light availability for intercropping nearby hybrid poplar was the lowest because of tree size and planting density on the row. However, the strong decrease in hay biomass, high ratio of root density per unit of hay biomass (FRLDs/BIOMASS), high values of root density of HYP at the surface and low nitrogen concentration in soil also suggest limitation of soil resources. Actually, as the system has received no input since 2007, limiting nutrients like nitrogen might also be impeding hay growth.

Implications for management practices

Tree species selection is a crucial step in the establishment of an agroforestry system, since shade intolerant species have the potential to develop strong competitive interactions with intercrops, as shown in this study. However, the temporal dynamic aspect of a TBI system involving several species of trees should also be considered. Fast-growing hybrid poplars will mature and be harvested much earlier (20 years) than red oaks (60 years), and their density on the row can also be adjusted (probably too high and not thinned early enough in the present case). Hybrid poplars, with their strong apical dominance, are easy to grow into straight boles and thus can easily be pruned to allow more light in (and increased value).

In much the same way the spatial separation between trees and intercropping can be improved by tillage or root pruning, competition for light can be mastered by thinning and pruning trees. Spacing between trees inside rows should be increased to nearly 6 m (felling 2 out of 3 trees) to support the rapid growth of poplars and sustained crop yields. Wider spacing could also be implemented at planting. In European country like France, it is now recommended that trees be planted at densities of 50–100 trees per hectare, with large spacing within and between rows (Baldy et al. 1993). Also, an uncultivated buffer zone of 2 m on both sides of tree rows is recommended (Rivest and Olivier 2007).

The impact of fast-growing poplars on crop yield at this stage of development of the system (eight growing seasons), both aboveground and belowground, was greater than that of red oak. However, the denser and larger root system can be considered an environmental asset, since it sequesters more carbon and produce more ecosystem services (Alam et al. 2014), and may act as a more effective safety net and hydraulic pump. Also, hybrid poplars would be harvested sooner than red oak and the newly freed space should benefit crop productivity, despite a probable increasing impact of oaks. However, even at their present young age, increased soil microarthropod richness (which supports a series of ecological important processes and services) was observed only in the presence of oaks (Doblas-Miranda et al. 2014). These temporal dynamic aspects and trade-offs of agroforestry systems are similar to those at work in natural ecosystems, and should therefore be subjected to further investigation to ensure the success of TBI systems.

Conclusion

This study did not demonstrate that vertical root profiles of trees were deeper in the presence of intercropping, but rather superficial like those of many planted trees or in natural areas with acidic, glacial soils. However, red oak, unlike hybrid poplar, adapted its root distribution by reducing density at the surface, resulting in a spatial separation from the root system of intercropped hay.

Crop yield decreased near trees, especially near hybrid poplars. However, our results suggest that competition for light predominated over root competition, even though soil resources and fertilization levels were low.

Our results imply that management practices can be optimized to ameliorate the overall system production as branches and roots pruning could lessen competition between crop and trees, as well as fertilization, organic or not. The main objective of an agroforestry system is also to be taken into account before plantation. For example, fast growing tree species may have more environmental assets than slow growing species. On the contrary, slow growing species may allow normal crop yield for longer time before light and root competitions occur.