Crown architecture of Populus spp. is differentially modified by free-air CO2 enrichment (POPFACE)


  • B. Gielen,

    Corresponding author
    1. University of Antwerp, UIA, Department of Biology, Research Group of Plant and Vegetation Ecology, Universiteitsplein 1, B-2610 Antwerpen (Wilrijk), Belgium;
      Author for correspondence: B. Gielen Tel: +32 3820 22 89 Fax: +32 3820 22 71 Email:
    Search for more papers by this author
  • C. Calfapietra,

    1. Universita degli Studi della Tuscia, Department of Forest Environment and Resources (DISAFRI), Via San Camillo de Lellis, I-01100 Viterbo, Italy;
    Search for more papers by this author
  • A. Claus,

    1. University of Antwerp, UIA, Department of Biology, Research Group of Plant and Vegetation Ecology, Universiteitsplein 1, B-2610 Antwerpen (Wilrijk), Belgium;
    2. Present address: Institute for Vegetables and Ornamental Plants, Theodor-Echtermeyer-Weg 1, D-14989 Grossbeeren (Berlin), Germany
    Search for more papers by this author
  • M. Sabatti,

    1. Universita degli Studi della Tuscia, Department of Forest Environment and Resources (DISAFRI), Via San Camillo de Lellis, I-01100 Viterbo, Italy;
    Search for more papers by this author
  • R. Ceulemans

    1. University of Antwerp, UIA, Department of Biology, Research Group of Plant and Vegetation Ecology, Universiteitsplein 1, B-2610 Antwerpen (Wilrijk), Belgium;
    Search for more papers by this author

Author for correspondence: B. Gielen Tel: +32 3820 22 89 Fax: +32 3820 22 71 Email:


  •  Although canopy architecture is a prime determinant of forest functioning and productivity, it has received little attention when examining forest responses to rising atmospheric CO2 concentrations. In this study different characteristics of crown architecture of two Populus species and one hybrid were investigated within a high-density poplar plantation. A free-air CO2 enrichment (FACE) facility was used to mimic future elevated CO2 concentrations.
  •  Canopy depth and branching patterns were studied, and detailed branch characteristics such as branch dimension, inclination, and internodal length were assessed for the three poplar species in the FACE and control treatments.
  •  Effects of elevated CO2 were restricted to a significantly increased canopy depth and longer internodal lengths after two years of CO2 enrichment. Additionally, branch dimensions and sylleptic branch numbers were increased by FACE, but responses were variable among growing seasons and species. However, FACE did not affect branch angles of origin and termination.
  •  Crown architecture was modified mainly through a growth stimulation in response to FACE. Nevertheless, important differences among species were observed which may influence future CO2 responses when competition will become more important.


Attempts to predict forest ecosystem functioning in response to rising atmospheric CO2 concentration typically emphasize tree growth responses through a photosynthetic enhancement. Mean stimulations of 66% and 73% (median = 55%) for photosynthesis and above-ground woody biomass, respectively, were reported for hardwood trees grown in the field under elevated CO2 concentrations (Norby et al., 1999). However, evidence that growth stimulation may be limited to the initial phase of CO2 enrichment is emerging (Ward & Strain, 1999). To understand this limited effect, and adequately predict future forest functioning and productivity, responses other than tree growth per se need careful attention.

Dry-matter production of forests is related to the amount of intercepted light, which is influenced by leaf area, display and distribution (Halléet al., 1978). These canopy characteristics are in part determined by woody crown architecture. Despite its importance, experimental studies on canopy architecture responses to elevated CO2 are scarce. A better understanding of crown architecture may help scaling from the leaf level to the stand level, and is particularly crucial for predictions of elevated CO2 responses of mixed stands (Diaz, 1995; Norby et al., 1999; Pritchard et al., 1999). Species composition of temperate forests, for example, might change in response to rising atmospheric CO2 concentrations, but shifts wil depend on light patterns, hence on canopy architecture (Hättenschwiler & Körner, 2000).

Related to biomass production and stem wood quality with respect to knots (Brazier, 1977), studies of crown architecture are of particular interest in managed forests and short-rotation culture systems (Mäkinen, 1996). Ideotype breeding includes crown architecture traits as selection criteria to improve forest productivity (Dickmann, 1985; Ceulemans et al., 1990; Hinckley et al., 1992; St. Clair, 1994). Studies of crown architecture of poplar (Populus spp.) have tried to identify criteria to further increase biomass yield of short-rotation poplar plantations (Nelson et al., 1981; Isebrands & Nelson, 1982; Burk et al., 1983; Ceulemans et al., 1990). Because symmetric branching patterns result from the phyllotactic control of bud arrangement, few studies have investigated crown architecture in a CO2-enriched atmosphere. Nevertheless, phenotypic plasticity induced by the environment has already been documented, and branching patterns can be altered by branch mortality and abscission beyond current-year stem growth (Fisher, 1986). Moreover, within-stand competition modifies the growth of individual trees and the stand development in several ways (Kellomäki, 1986).

The present study was performed in a high-density poplar plantation with a free-air CO2 enrichment (FACE) system. With this technique stand-level responses can be assessed under natural conditions, and therefore largely eliminating the shortcomings of past CO2 enrichment techniques (Hendrey et al., 1993 and 1999). The questions addressed in the present study are: does CO2 enrichment affect woody crown architecture of Populus?; If so, to which extent is this caused by a growth stimulation, or by increased competition in the FACE treatment? To answer these questions, we measured crown depth, number of branches, branch dimension and inclination. Additionally, apical dominance was studied by determining branch insertion height. Emphasis was given to first-order branches due to their size and disposition dominance, and because they determine the arrangement of higher-order branches (Nelson et al., 1981; Burk et al., 1983). Finally, we examined species differences that may help explain variable responses to elevated CO2 concentration.

Materials and Methods

Site description

The FACE study was located in Central Italy near Viterbo (Tuscania; 42°22′ N, 11°48′ E, alt. 150 m) on 9 ha of former agricultural land. Following detailed soil analysis, six (i.e. three pairs) experimental plots (30 m × 30 m, between-plot distance 120 m) were selected, and a FACE facility was established in three of them. Three control plots were left under natural conditions. The carbon enrichment was achieved through octagonal polyethylene rings (diameter, 22 m) mounted on telescopic poles, extendable to 12 m. Pure CO2 was injected through laser-drilled holes in the rings to reach a target CO2 concentration of 550 µmol mol−1 inside the FACE treatment. The CO2 concentration was 544 ± 48 µmol mol−1 during the first year of treatment, and 532 ± 83 µmol mol−1 during the second year. Due to unexpected internal corrosion of the CO2 bulk container, performance problems occurred in autumn of the second year. However, CO2 concentration was within the 20% deviation of the target 86% of the time, while 89% of all days had full enrichment (F. Miglietta, pers. comm.). Daytime CO2 enrichment was provided from bud burst to leaf fall. A meteorological station in the centre of each ring collected data used to control the directional release of gas along the rings. The released quantity of gas was determined by wind speed, and by an algorithm based on a 3-D gas dispersion model developed for the facility. For details on description and first-year performance of the FACE facility, see Miglietta et al. (2001).

Plant material and plantation lay-out

In late spring 1999 a 9-ha poplar plantation was established using hardwood cuttings at a planting density of 5000 trees per ha (spacing 2 m × 1 m), and 10 000 trees per ha (spacing 1 m × 1 m) within the six experimental plots. The experimental plots were planted with two Populus species and one hybrid, namely P. nigra (black poplar, clone Jean Pourtet), a local selection of P. alba (white poplar, clone 2AS11) and P.×euramericana (euramerican poplar, clone I-214), while the nonexperimental part of the plantation was planted with P.×euramericana (clone I-214). Each 314-m2 plot contained 350 plants, and was divided into six triangular sectors, with two sectors per species. Plantation management included continuous drip irrigation, mechanical weed removal, and a limited application of insecticides. The plantation was designed and managed as a short-rotation forest with typical high plant densities (Mitchell et al., 1999). For plantation details see Scarascia-Mugnozza et al. (2000) and Calfapietra et al. (2001).


All measurements and results reported here refer to the first two growing seasons of the plantation. Height growth increment 1 (HGI 1) and 2 (HGI 2) refer to the vertical height growth of the main stem in the first and second growing seasons, respectively (Isebrands & Nelson, 1982). Similarly, first-year branch growth increment (BGI 1) refers to the branch length growth of the first year. We distinguished sylleptic branches that developed from axillary buds not undergoing a resting period (Remphrey & Powell, 1985), and originating during the first (sylleptic HGI 1) and second (sylleptic HGI 2) growing seasons, and proleptic branches from buds undergoing a resting period (Halléet al., 1978). Proleptic branches were only present in the second growing season, and located on HGI 1.


Detailed architecture measurements were made on four trees per plot and per species at the end of the first growing season. Insertion heights of the lowest and highest sylleptic branch were measured, and the number of branches counted. Branch characteristics assessed were length, diameter at 1 cm above point of attachment, and the number of buds. Additionally, the number of buds on the stem was counted. Height, branch diameter and branch length measurements were performed with an extendable pole, a digital calliper (Mitutoyo, type CD-15DC, Telford, UK), and a ruler, respectively.

Similar data were collected at the end of the second growing season on three trees per plot and per species. Insertion heights of the lowest and the highest branches were determined for all branch categories. The diameter and length of all branches in each category, or of a subsample (n = 8) when the number of branches exceeded 10, were measured. Second-order branches were counted on sylleptic branches of HGI 1. Internodal lengths on proleptic branches, on sylleptic branches of HGI 2, and on the stem (HGI 2) were calculated from leaf counts during the growing season (n = 96, 30, and 12 per treatment and species for proleptic branches, sylleptic branches and the stem, respectively). Internodal length on the stem was determined as:

inline image

(H, tree height at the end of the second growing season; Ns1 and Ns2, number of sylleptic branches on HGI 1 and HGI 2, respectively; Np, number of proleptic branches; and Nb, number of buds on the stem. Internodal length for HGI 1 at the end of the growing season was calculated in a similar way using Ns1 and Np. Canopy depth was determined as H – Ilow, with Ilow being the insertion height of the lowest branch. Insertion heights of branches were calculated in proportion to total tree height; the relative canopy depth was equivalent to the relative insertion height of the lowest branch.

Angles of origin and termination (Polk, 1974) of a sample of branches were measured (0°= stem axis) with a protractor and a custom inclinometer in the first and second growing seasons, respectively (Kockelbergh & Assissi, 2000). Sample size was identical to that described in the previous paragraph, except for the second growing season when additional measurements were made in mid summer.

Angles of origin and termination were grouped in arcs of a circle (0°–90°) with length 5° (λ = 90°/18); this frequency distribution was used in all further calculations and analyses. For each sample, a mean vector with mean angle and mean length was calculated (Batschelet, 1981). The correction factor for grouping had a minimal effect because of the large number of groups, and was therefore omitted (Batschelet, 1981). The length of each branch was also measured as the straight line between branch insertion and branch tip, and was used to correlate branch angles to branch dimensions.

Clonal phenograms

Simplified, schematic two-dimensional representations of the woody architecture were constructed for all species in the FACE and control treatments. Phenograms were based on mean values of height, insertion height of specific branch categories, number of branches, branch length, and angles of origin and termination. Leaves and higher-order branches are not shown, but information on internodal lengths and second-order branching are represented in some figures and tables.

Statistical analysis

To determine the main effects of FACE and species on canopy architecture, an ANOVA was used. A randomized-complete-block design with treatment, species, and the treatment–species interaction as fixed factors, and block as a random factor was applied. All statistical analyses were performed in SAS (SAS system 6.12, SAS Institute Inc., Cary, NC, USA) using the mixed procedure (Littell et al., 1996) and plot as the replicate. Satterthwaite’s procedure was applied to obtain the denominator degrees of freedom. Where the ANOVA F-test highlighted a significant interaction between CO2 treatment and species, a posteriori comparison of means was performed using parameter estimates as given by SAS. The Bonferroni method was used to correct for the multiple comparisons. Differences between parameter means were considered significant when the P-value of the ANOVA F-test was < 0.05. Data were tested for normality using the Shapiro-Wilk statistic (proc univariate in SAS).

Branch angles were analysed according to the rules of circular statistics (Batschelet, 1981). The Raleigh test was applied to test for randomness. This test was significant in all cases, proving unimodality and a concentration of the angles around the mean direction (Batschelet, 1981). Data met the assumptions of the Watson-Williams test for significant difference between mean angles of two samples. If the statistic F (for calculation see Batschelet (1981), p. 96) was greater than F1,n–2 at a significance level of 0.05, the null hypothesis of equal mean directions was rejected. The Pearson correlation coefficients were calculated between branch length and branch angles for different branch categories of all species pooled among CO2 treatments. This was performed in SAS (proc corr) after testing for normality.


Crown architecture

Mean canopy depth slightly increased in response to FACE. Differences in canopy depth between the control and FACE treatments ranged from 9% (P. alba and P.×euramericana) to 14% (P. nigra) in the first growing season (Fig. 1a), and from 5% (P. nigra) to 16% (P. alba) in the second growing season (Fig. 1c). The overall effect of FACE on canopy depth was only significant in the second growing season (Table 1). This effect persisted after correcting for the larger tree height in the FACE treatment, that is, when relative canopy depth was considered (Fig. 1b,d). This was evident from the significantly lower relative insertion height of the lowest branch at the end of the second growing season (Table 1). For example, branching of P.×euramericana started, on average, at 2% of tree height in the FACE treatment vs at 4% in the control treatment. The insertion height of the top sylleptic branch was increased by the FACE treatment in the first growing season for both P. nigra and P.×euramericana (Table 1). However, this observation was not statistically different, and disappeared in the second growing season. The first proleptic branch was located relatively higher in the FACE treatment for P. nigra and P.×euramericana, while the insertion of the highest proleptic branch was only slightly affected. The overall treatment effect was not significant because P. alba showed the opposite trend (Table 1). Mean stem internodal length was significantly affected by the FACE treatment in the second growing season. For P.×euramericana internodal length was 4.87 cm and 5.55 cm for the control and FACE treatments, respectively (Fig. 2a, Table 1).

Figure 1.

Canopy depth and relative canopy depth of three Populus species at the end of the first (a,b) and the second (c,d) growing season in the control (open bars) and FACE (hatched bars) treatments. Average (± SE) values are shown with n= 12 and n= 9 in the first and the second growing season, respectively. Canopy depth and relative canopy depth were significantly affected by FACE in the second growing season. Both variables were always affected by species, with the exception of canopy depth in the second growing season.

Table 1.  Summary of ANOVA results for different crown architecture characteristics (measured variables) of three Populus species in the control and FACE treatments
 Measured variableSource of variation
 CO2SpeciesCO2 × species
  1. Levels of significance (P-values) are indicated as: *P < 0.05; **P < 0.01;***P < 0.001; ****P < 0.0001. BGI, branch growth increment; H, highest branch; HGI, height growth increment; L, lowest branch.

First growing season
 TreeCanopy depth ** 
 Insertion height L ** 
 Insertion height H< 0.1**** 
 Internodal length< 0.1**** 
 Sylleptic branchesRel. no. branches ****< 0.1
 Length< 0.1* 
 Diameter< 0.1** 
 Internodal length******
Second growing season
 TreeCanopy depth*  
 Internodal length HGI 1 **** 
 Internodal length HGI 2***** 
 Internodal length (HGI 1 + HGI 2)***** 
 Sylleptic branches HGI 1Rel. no. branches< 0.1*****
 Length < 0.1 
 Length BGI 1*  
 Length BGI 2 < 0.1< 0.1
 Rel. no. second-order branches ** 
 Insertion height L*** 
 Insertion height H< 0.1**** 
 Proleptic branchesRel. no. branches< 0.1**< 0.1
 Length ****< 0.1
 Diameter * 
 Internodal length***** 
 Insertion height L **< 0.1
 Insertion height H < 0.1 
 Sylleptic branches HGI 2Rel. no. branches **** 
 Length * 
 Diameter ** 
 Internodal length **** 
 Insertion height L   
 Insertion height H **** 
Figure 2.

Internodal length of stem (a), proleptic branches (b), and sylleptic branches of height growth increment (HGI) 2 (c) for three Populus species in the second growing season in the control (open bars) and FACE (hatched bars) treatments. Average (± SE) values are shown, n= 9 (stem), n= 96 (proleptic branches), n= 30 (sylleptic branches of HGI 2). Treatment effects were significant for both stem and proleptic branches; species effects were significant in all cases.

Trees in the FACE treatment produced more sylleptic branches on HGI 1 compared to those in the control treatment (Fig. 3). During the first growing season the number of branches per unit of tree height increased nonsignificantly (P = 0.148) by 23% for P. nigra, and more than doubled for P.×euramericana. Differences were considerably smaller, or even opposite in sign, for sylleptic branches produced in the second growing season (sylleptic branches on HGI 2, Table 1, Fig. 3). Fewer proleptic branches were present in the FACE treatment at the end of the second growing season for P. nigra and P. × euramericana (only significant for P.×euramericana) (Table 1, Fig. 3). The number of sylleptic branches on HGI 1 at the end of the second growing season (Fig. 3) was slightly different from the first growing season because of branch mortality and abscission, and because of different sample size. There were significantly more sylleptic branches in the FACE treatment for both P. nigra and P.×euramericana. The number of second-order branches per unit branch length (sylleptic branches of HGI 1) was not significantly different between the FACE and control treatments (Table 1).

Figure 3.

Schematic representation of the architecture of three Populus species at the end of the second growing season in the control and FACE treatments. Phenograms are drawn true-to-scale on the basis of mean tree height, mean branch characteristics for each branch category (length, angles of origin and termination, and insertion heights on the tree). The number of branches represented is half the actual number for proleptic branches for all species, and for sylleptic branches of P. nigra. Proleptic branches are indicated by thick lines. The small horizontal line on the stem represents the mean tree height at the end of the first growing season, i.e. it divides height growth increment 1 and 2. The horizontal 1 m line indicates the real spacing between trees in the field.

Branch dimension and internodal length

For all three species the FACE treatment increased average branch dimensions (diameter and length) in the first growing season (P < 0.1, Table 1). Branch length was stimulated 10% for P. alba, 19% for P.×euramericana, and 41% for P. nigra. In the second growing season, only the length of BGI 1 of sylleptic branches on HGI 1 was significantly affected, which was a carry-over effect from the first growing season (Table 1). Sylleptic branches of HGI 1 were characterized by a significantly larger internodal length in the first growing season, and this was also true for proleptic branches in the second growing season (Fig. 2b, Table 1). By contrast, internodal length on sylleptic branches, produced in the second growing season, was not or only slightly affected by FACE (Fig. 2c, Table 1).

Branch angles of origin and termination

Mean branch angles were calculated as a vector with a certain length and inclination. The vector length was typically high (0.98), and this indicates a strong concentration of branch angles around the mean (Batschelet, 1981).

Over the two growing seasons, significant treatment effects on mean branch angles were only observed in four cases, without a clear overall pattern (Table 2). When angles of origin for different branch categories were compared, the branch angles became steeper when going upwards through the canopy (Table 2, Fig. 3). The angle of termination of sylleptic branches was larger in the first growing season, as compared to the second growing season, and this was due to larger branch dimensions. Branch angles were negatively correlated with branch length, that is, longer branches were generally more vertically orientated. Pearson correlation coefficients between branch length and angles of termination for sylleptic branches on HGI 1, for proleptic branches, and for sylleptic branches on HGI 2 ranged from −0.45 to −0.68 (P < 0.0001).

Table 2.  Branch angles of origin and termination of three Populus species in the control and FACE treatments
  P. albaP. nigraP. × euramericana
  1. Values are the average (± SE) for the first and the second growing season of all branches on 12 trees per treatment, and a sample, n= [35–343]  of branches per treatment, respectively. Averages were calculated as described in Materials and Methods. Values within the same row are significantly different (P < 0.05) if followed by different letters. HGI, height growth increment. All values are expressed in degrees.

1st growing season
 Angle of originSylleptic65 (1)b63 (1)b53 (1)a62 (1)b51 (2)a52 (1)a
 Angle of termination 46 (1)b44 (1)ab42 (1)a42 (1)a40 (2)a45 (1)b
2nd growing season
 Angle of originSylleptic HGI 161 (1)b59 (1)b59 (1)b60 (1)b48 (1)a49 (1)a
 Proleptic56 (1)b57 (1)b48 (0.5)a48 (1)a60 (1)c56 (1)b
 Sylleptic HGI 249 (1)c48 (1)bc45 (2)ab45 (1)b43 (1)ab41 (1)a
 Angle of terminationSylleptic HGI 129 (1)b25 (1)a28 (1)b30 (1)b28 (1)b27 (1)ab
 Proleptic30 (2)b27 (2)ab27 (2)ab24 (2)a29 (2)ab24 (2)a
 Sylleptic HGI 229 (1)ab27 (1)a28 (1)ab30 (2)ab31 (2)b28 (1)ab

Species differences

Although the FACE treatment only significantly affected canopy depth, internodal length, and number of branches (Table 1), nearly all crown characteristics significantly differed among species (Figs 1, 2 and 3, Tables 1 and 2). P.×euramericana, for example, was characterized by a small number of branches and large internodal lengths. Both P. nigra and P. alba were very branched, although P. nigra to a larger extent, especially in the first growing season. Branches of P. alba were on average the longest and more horizontally orientated at the base (larger angles of origin, Table 2, Fig. 3), except for proleptic branch initiation that was flatter for P.×euramericana (Table 2). Termination angles of branches did not significantly differ among species (Table 2).


For all three species the canopy was deeper in the FACE stand at the end of the second growing season as a result of the first-year stimulation of sylleptic branching. The FACE treatment significantly increased stem internodal length because sylleptic branching was unaffected in the second growing season. This agrees with most studies that were reviewed recently; these showed increased stem or branch elongation without elevated CO2 effects on the number of nodes, and suggested a greater growth stimulation as compared to the formation of organ primordia at the shoot tip (Pritchard et al., 1999). Nevertheless, many studies have reported stimulated branch initiation even without increased plant height (Pritchard et al., 1999), or showed no response, as also recently reported for Betula pendula (Centritto, 2000). Earlier CO2 experiments with Populus in controlled- and open-top-chambers showed higher absolute number of branches, and increased stem elongation (Radoglou & Jarvis, 1990; Ceulemans et al., 1995). However, Kubiske et al. (1997) showed that internodal distance of P. tremuloides was not affected by elevated CO2.

Our observations of deeper crowns, fewer proleptic branches, and longer internodes in the FACE treatment confirm our analysis of LAI (Gielen et al., 2001). Increased crown depth confirms our previous suggestion that the loss of a significant CO2-stimulation of LAI after canopy closure was caused by increased shading. Although we observed deeper crowns in response to FACE, the possible positive effect on growth may be restricted to the initial period of CO2 enrichment, and may disappear once leaf-fall starts in the lower part of the canopy.

In general, our results showed evidence for increased branch dimensions under elevated CO2 concentrations, in agreement with previous results on Pinus sylvestris and other tree species (Jach & Ceulemans, 1999; Pritchard et al., 1999), but in contradiction to the results of a model ecosystem with Picea abies (Hättenschwiler & Körner, 1998). Based on our observations, branch growth and stem growth were similarly affected in the first growing season, with branches having longer internodes. Nevertheless, this was not true for sylleptic branches produced in the second growing season, indicating a decline in the response to the FACE treatment.

Large increases of higher-order branch biomass in response to elevated CO2 have been observed previously (Sionit et al., 1985; Idso et al., 1991; Wilkins et al., 1994; Jach & Ceulemans, 1999), but also a lack of response in second-order branching (Barton & Jarvis, 1999). Our study showed a reduced number of second-order branches, but future characterization of higher-order branching is needed to verify these preliminary observations. Based on these findings and evidences, we conclude that woody crown architecture is affected by FACE through CO2-induced stimulation of growth, and of initial branching.

Because CO2 effects on branch insertion height and branch dimension were observed, apical dominance and apical control were examined in response to the FACE treatment. In the first case, the position of the top branch was relatively higher in the FACE than in control treatments at the end of the first growing season. This suggests a partial loss of apical dominance, since this dominance determines whether the meristem initially forms a sylleptic branch, or a bud (Brown et al., 1967; Ford, 1985; Wilson, 2000). In the case of apical control (the process that regulates the amount of elongation, and diameter growth of proleptic branches) (Brown et al., 1967; Ford, 1985; Wilson, 2000) we found that proleptic branch length appeared to be stimulated by the FACE treatment for P. nigra (+22%) and P.×euramericana (+6%). Nevertheless, these observations were not statistically significant and may be solely caused by an initial stimulation of branching and growth under elevated CO2. This prevents us from concluding that either apical dominance or apical control was reduced in an elevated CO2 atmosphere. Reduced apical dominance under elevated CO2 has been suggested in earlier studies (Pritchard et al., 1999), but also contrary results have been observed, for example, under a combination of elevated temperatures and increased CO2 (Martin et al., 1995).

Apart from the FACE effects, our observations confirm the importance of crown architecture to discriminate among poplar species (Ceulemans et al., 1990). For example P. alba, was characterized by longer, more horizontally orientated branches as compared to both other species. Such differences could influence future responses to CO2 enrichment through light interception, and through other canopy-atmosphere exchange processes. Moreover, within-stand competition could also be influenced by these canopy architectural differences, which, in turn, could alter growth performance in response to FACE. The simplified schematic architecture representations of the species reflect this competition aspect when the planting distance between trees (represented by the horizontal 1 m line), is compared with tree crown width. Contacting and interacting branches are more abundant within the stands of P. alba (Fig. 3). This may further affect crown architecture in the forthcoming years as it is well known that competitive processes modify the structure of a crown (Kellomäki, 1986). Competition also reduced branch dimension in the middle and lower canopy of P. sylvestris stands, although this was likely caused by reduced overall tree growth (Mäkinen, 1996). The competition issue is related to the effect of plant density on branch inclination angles, for example, widely spaced poplars develop wider crowns (Nelson et al., 1981). Optimal performance under closed spacing is, of course, partly genetically determined. We speculate that competition could play a more important role in our P. alba stands, as compared to P. nigra because P. alba has much wider crowns.

The importance of an elevated CO2 effect on canopy architecture was also demonstrated by Reekie & Bazzaz (1989) in a simplified tropical community. Contrary to elevated CO2 effects on individually grown tree seedlings, biomass was significantly affected by CO2 when the trees were grown under competitive conditions. The authors concluded that CO2 affected competitive outcome through its effect on canopy architecture (Reekie & Bazzaz, 1989). Similarly, differences in crown architecture of the three Populus species in our study, and their different relative response to CO2 enrichment, might influence their competitive ability when grown in multispecies stands.

Following two growing seasons of FACE, we found no evidence for altered crown architecture due to CO2 effects on within-stand competition. FACE caused slightly sharper branch termination angles in the second growing season, but this was correlated with increased branch length, that is, long branches grew vertically to the light source. Such a negative correlation between branch length and termination angle was found previously (Bozzuto & Wilson, 1988), although also no correlation has been reported (Cluzeau et al., 1994). Our observations of wider branch angles in the lower part of the canopy has been attributed to self-weight, and to a way to avoid self-shading (Bozzuto & Wilson, 1988; Deleuze et al., 1996; Goulet et al., 2000).

We examined branch abscission of both sylleptic and proleptic branches (data not shown), but only species differences were observed, showing that FACE did not indirectly affect branch abscission through competition, or light (deeper canopy).

In conclusion, crown architecture of two Populus species and one hybrid was affected by the FACE treatment mainly through stimulated growth, that is, increased canopy depth, internodal length, and number and dimension of branches. Nevertheless important species differences were observed, which may further influence responses to elevated CO2 in the long term. Additionally, canopy closure was reached only in the summer of the second growing season, hence competition may become more important in the coming years and could further influence crown architecture.


Funding was provided by the EC Fourth Framework Programme, Environment and Climate RTD Programme, research contract ENV4-CT97-0657 within the Terrestrial Ecosystems Research Initiative (TERI). The authors acknowledge F. Kockelbergh and B. Assissi for development of the inclinometer; Dr T. Crowe (University of Southampton) and Dr S. Van Dongen (University of Antwerp) for statistical advice, G. Cortignani for field assistance, and Dr D. Salvador for editorial help. We are also very grateful to Dr P. Curtis and to two anonymous reviewers for their useful comments and suggestions on an earlier version of this manuscript. B. Gielen is a Research Assistant of the Fund for Scientific Research-Flanders (Belgium, F.W.O.-Vlaanderen). This study contributes to the GCTE (Gobal Change & Terrestrial Ecosystems) Elevated CO2 Consortium of the IGBP (International Geosphere Biosphere Programme).