Clonal Variation in Nutrient Content in Woody Biomass of Hybrid Aspen ( Populus tremula L . × P . tremuloides Michx . )

Differences in the nutrient concentrations and nutrient amounts of stems and branches amongst clones of hybrid aspen (Populus tremula L. × P. tremuloides Michx.) were investigated. Seven clones with superior and seven with medium growth rates were selected from a test of 119 clones in southern Sweden. Four trees per clone were randomly identifi ed and harvested in dormant conditions. Sample discs from the stems and branches were collected and analysed for N, K, P, Ca, Mg, and S concentrations, as well as wood density. The analyses revealed signifi cant genetic differences in wood density, K, P, and Mg concentrations in the stems. There were weak (non-signifi cant) and negative genetic correlations between stem volume and concentrations of all the nutrients, except potassium, suggesting that nutrient-effi cient clones could be selected without signifi cantly sacrifi cing genetic gain for growth. In the branches K, Ca, and Mg concentrations differed signifi cantly among clones. After selecting more nutrient effi cient clones, the potential savings of nutrients compared with current hybrid aspen material was estimated to be around 5%, which seems fairly low, at least in a short-term perspective. However, the use of clones with different nutrient storage strategies may be regarded as a possible way in the long run to save nutrients in hybrid aspen ecosystems, or of removing them when sludge is applied.


Introduction
Plantation forestry offers a possible way to increase the access of renewable energy.In Sweden, so far, developments of this kind have mainly focused on energy forests, i.e. cultivation of willow (Salix sp.) clones in rotation periods of up to fi ve years (e.g.Sirén et al. 1987, Christersson et al. 1993, Ledin 1996).However, in the last decade interest in other tree species for biofuel production has intensifi ed (Johansson 1992, 1999, Sonesson et al. 1994, Telenius 1999, Rytter 2002).The attraction of using systems with ordinary forestry species is that the production of merchantable wood and biofuel wood, in the form of tops and branches, can be combined, resulting in a more fl exible and less market-sensitive mode of cultivation.
For this purpose, in Swedish conditions, hybrid aspen (Populus tremula L. × P. tremuloides Michx.) has been found to be highly competitive because of its high productivity and ability to regenerate through root suckers.Hybrid aspen, a cross between the European aspen (P.tremula L.) and its North American counterpart, trembling aspen (P.tremuloides Michx.), has shown mean annual increments of more than 15 m 3 ha -1 yr -1 during approximately 25-year rotation periods in the Nordic countries (Jakobsen 1976, Elfving 1986).These growth rates are anticipated to increase in the near future, since the Swedish breeding programme (Karlsson andDanell 1992, Stener 2002) has generated more productive plant material (Rytter et al. 2000, 2002, Rytter 2002, Stener 2002), that is expected to exceed 20 m 3 ha -1 yr -1 in mean increment over 20-25 year rotation periods.
However, high productivity implies that large quantities of wood and biomass, and thus substantial amounts of nutrients, will be removed at harvest.Estimates derived by Rytter (2002) indicate that whole-tree utilization systems (i.e.harvest of all above-ground tree parts) will cause net losses of N. Similarly, K, P, Ca and Mg is also lost, and must be replaced, via reapplications of ash or weathering, to sustain the productivity of the site.One way to reduce nutrient removal in hardwood plantations is to harvest during the wintertime, thereby avoiding removal of nutri-ent-rich leaves.Another alternative would be to use nutrient-effi cient clones, provided clonal differences occur in nutrient effi ciency in the harvested tree parts.
Recent genetic studies of hybrid aspen have revealed signifi cant clonal differences in several tree characteristics such as growth, phenology, wood density and other wood properties (Ilstedt and Gullberg 1993, Stener 1998, Yu et al. 2001a, 2001b).Therefore, clones are likely to display differences in nutrient concentrations in their stems and branches, especially since such differences have been reported for other Populus species (e.g.Heilman andStettler 1986, Singh 1998).
The objectives of this study were to investigate the extent of clonal differences in nutrient concentrations and amounts in stems and branches of hybrid aspen in dormant conditions, and their practical implications with respect to nutrient removal at harvest.Our hypothesis was that genetically based nutritional differences exist between clones and are worth considering when selecting plant material for future cultivation.

The Stand
The study was performed in a 14-year-old fi eld trial of 119 hybrid aspen clones (Svalöv,Sweden;lat. 55º57'N;long. 13º06'E;alt. 80 m.a.s.l.).Oneyear-old container plants were established at a density of 1600 plants ha -1 (2.5 m × 2.5 m) as single tree plots in a randomised block design with a total of 15 ramets per clone.No thinning was carried out before the study.The site was on fertile former agricultural land with clayish moraine, and the stand consisted of 1360 stems ha -1 .The average height of all clones was 12.8 m and the average breast height diameter of the stem of mean basal area was 12.5 cm.Growth in the four years preceding the study averaged 21.6 m 3 stemwood ha -1 yr -1 .

Sampling
All 119 clones in the trial had been selected as plus-trees in trials and commercial stands from the 1950's and 1960's.Based on assessments from the autumn of 1999, all the clones were divided into three categories: 1) superior, 2) medium and 3) poor, depending on their volume growth.For this study, 14 clones (Table 1) were randomly selected from amongst the fi rst two of these groups, i.e. seven clones of superior growth and seven of medium growth.For each clone four individuals, free from visible external stem damage, were randomly selected and harvested in January 2001.
Breast height diameter, green crown height, and total tree height were recorded for each selected stem.Stem discs of 3-5 cm thickness were cut at every third metre along the stem, starting at stump height and continuing to the top.Depending on tree height, fi ve or six discs were collected from each tree.
Immediately after cutting, the discs were transported to the nearby Skogforsk laboratory at Ekebo research station.The diameter of each disc was recorded by cross measurement and the disc volume over bark was estimated by the water displacement method, thereby allowing individual density estimates to be made.All discs were dried at 85ºC to constant weight (~ one week) before dry weight determinations.
From each sample tree, three representative living branches were collected from the lower, middle, and upper thirds of the green crown.The fresh weights of these branches were determined.Three subsamples of exactly 2 cm length were collected from each sampled branch.The subsamples were cut at a distance of 10 cm from the stem, exactly in the middle of the branch, and 10 cm from the tip.Fresh and dry weights (85ºC) were recorded for each subsample.The total fresh weight was recorded for the rest of the living branches of each tree.
Dead branches were also collected from each tree.However, they were few in number and their weight was low in comparison to that of the living branches (data not shown), so they were not included in further investigations.

Nutrient Analyses
In order to produce a representative sample for nutrient analysis from each stem disc, a cross section of even thickness was removed by a circular saw blade, when sawing each disc into two horizontal halves.The sawdust thus produced was carefully collected and mixed and a sample of it was used to analyse concentrations of nitrogen, phosphorus, potassium, calcium, magnesium, and sulphur.The branch subsamples from each tree were combined, ground and thoroughly mixed.A sample was taken to analyse the same nutrients as for the stem discs.The sampling procedure described above was assumed to give representative estimates of weight-based averages of the nutrient concentrations.
Nitrogen was analysed by gas chromatography in an elemental analyser (CarloErba NA 1500, Rodano, Italy; Kirsten and Hesselius 1983).The other macronutrients (K, P, Ca, Mg, and S) were analysed with an inductively coupled plasma with

Volume and Weight Calculations
Stem volume over bark was estimated for each tree by summing the volumes of the 3 m stem sections and the top, using the formulas for the frustum of a cone (V = (πH/3)(R 2 + Rr + r 2 )), and a circular cone (V = (πR 2 H)/3), respectively, where V = volume, H = height of the stem section, R = lower radius and r = upper radius.The tapering of trees, including European aspen according to Opdahl (1989), normally follows a sigmoidal pattern.This deviation from an ideal cone shape might introduce some error into the stem volume estimations based on 3 m sections, but the potential effects were regarded as marginal and were not therefore further considered.Fresh stem volume estimates were converted to stem dry weights using the recorded disc density of each sample tree.An average stem density value was estimated for each stem section by taking the mean of the densities from the two discs at the section ends.For the top of the stem the density of the uppermost disc was used.
The dry weight of living branches was estimated for each sample tree by using the average ratio between the dry weight and fresh weight of the subsamples, and multiplying the resulting fi gure by the total fresh weight of all living branches.

Nutrient Calculations
When calculating the amounts of the different elements in each 3 m stem section, average fi gures for the nutrient concentrations in the discs at the bottom and top of the section were used.For the rest of the top, the nutrient concentrations of the uppermost sampled disc were used.The nutrient amounts in the whole stem were obtained by summing the amounts found in all the stem sections, including the top.
The nutrient amounts in living branches per sample tree were calculated by multiplying the concentration of each nutrient obtained from the combined subsamples with the estimated total dry weight of branches of each tree.

Statistical Analyses
The statistical analysis was based on individual tree observations according to the model: where y ijk = observation k, in block i for clone j, µ = trial mean, b i = fi xed effect of block i, c j = random effect of clone j, NID(0,σ c 2 ), and e ijk = random error term for observation ijk, NID(0,σ e 2 ).The variances σ c 2 and σ e 2 were estimated for different traits according to Henderson's method 3 (Henderson 1953) as performed in LSMLMW software (Harvey 1990).Values of BLUP (Best Linear Unbiased Predictors) for each clone were obtained using Proc Mixed (SAS 1997).Genetic parameters were interpreted as σ G 2 = σ c 2 and σ E 2 = σ e 2 , where σ G 2 = the genotypic variance among clones and σ E 2 = environmental variance.The individual broad sense heritability (H 2 ) was calculated as: The genotypic coeffi cient of variation (CV G ) was calculated as: The genotypic correlations (r G ) between traits were estimated by LSMLMW software (Harvey 1990) from: where σ G1G2 is the genotypic covariance between traits 1 and 2, while σ G1 and σ G2 are the genotypic standard deviations for traits 1 and 2, respectively.Most traits displayed normal distributions.However, stem and branch concentrations of K and Ca and the total amounts of N, K, P, Ca and Mg were only approximately normally distributed.Since both normal-score (Gianola and Norton 1981) and arcsin transformation prior to statistical analysis changed the results only marginally, all traits were left untransformed.
The signifi cance level used for testing the null hypothesis of no difference was equal to or less than 5% (p ≤ 0.05)

Results
Tree growth, stem wood density, and dry weight of stem and branches all showed signifi cant and high broad sense heritabilities (H 2 ), indicating that there were signifi cant differences among clones (Table 2, Fig. 1).The genetic coeffi cients of variation (CV G ) were high for volume and dry weight of stems.
The genetic correlations among growth traits (H, Dbh, V S , W S ) were all signifi cant (Table 3).Signifi cant genetic correlations between stem volume and stem/branch nutrient concentrations were only found for K. Thus, stem and branch concentrations of P and Mg were independent of stem size.This relationship is also indicated in Figs. 2 and 3, where clone values of K, P, and Mg are presented in order of increasing stem volume.
The range of clone differences in nutrient concentrations can be illustrated by selecting clones with the lowest and highest overall BLUP-values for several nutrients in stems.This was done for the superior clones by summing the relative BLUP-values per clone for the elements showing signifi cant between-clone differences in stem concentrations (K, P and Mg).The pairs of clones with the lowest (clones 14 and 59) and highest (clones 12 and 119) levels of nutrients were selected (Fig. 2, Table 4) in order to stabilize the values and avoid individual variations affecting the overall nutritional patterns (Table 3).The differences in stem concentrations between these two groups of clones were 18%, 25% and 49%, for K, P, and Mg, respectively, corresponding to amounts of 32, 12 and 24 g m -3 of stem wood (Table 4).If the stem and branch nutrients on a weight basis were combined, a total tree concentration of 1.81 g K kg -1 was found for the low-nutrient group of clones.The corresponding value for the high-nutrient clone group was 2.08 g K kg -1 (Table 4).Thus, the difference was 0.27 g kg -1 , equivalent to a 15% difference in K levels between the K-rich and low-K clones.The corresponding differences for P and Mg were 14% (not signifi cant) and 25%, respectively.Wood density showed a weak and negative genetic correlation (r G = -0.41) to stem volume.Thus, the r G correlation between volume and dry weight of the stems was strong (0.94).Branch weight was positively correlated with stem volume and stem weight (Table 3), although individual deviations from this pattern were seen (Fig. 1).
Signifi cant H 2 values for nutrient concentration traits in stems were found for K, P and Mg, while for branches H 2 was signifi cant for K, Ca and Mg (Table 2).Thus, H 2 was not signifi cant for N and S concentrations, which is why they were not included in further analysis.The CV G was high for total amounts of nutrients in stems and branches, while for nutrient concentration traits, in both stems and branches, it was more modest but still around the same magnitude as for diameter and height.Apart from P there was a positive genetic correlation for the concentration of each element between stems and branches, although the correlations were only signifi cant for K and Mg (Table 3).
To assess the signifi cance of clonal differences in nutrient concentrations on the whole tree level an additional analysis was carried out, in which the nutrients in stem and branches were summed (data not presented in tables).This resulted in approximately the same genetic parameters as in the previous analysis, where stems and branches were evaluated separately (Table 2), with two exceptions.If compared to the H 2 of the stem nutrient concentration the H 2 of Ca increased to 0.36 (signifi cant) and H 2 of P decreased to 0.32 (not signifi cant).These results were expected since the genetic correlations between stem and branches were positive for Ca and negative for P (Table 3).
Intensive cultivation with the removal of high amounts of biomass during harvest operations increases the need to replace nutrients at a site (cf.Ericsson et al. 1992, Jug et al. 1999, Rytter 2002).The need to replace nutrients can be reduced by harvesting during the winter, since no leaves are removed from the stand.Another possibility could be to use plant material for which the nutrient concentration in aboveground woody biomass is low during dormancy, and in this sense could be regarded as nutrient effi cient.This study showed that clonal differences exist in stem and branch nutrient concentrations.Knowledge of genotypic coeffi cients of variation (CV G ) and heritability (H 2 ) values are essential for evaluating how traits will respond to selection.In the present study the CV G and H 2 values for K, P and Mg concentrations of the stem were of roughly the same magnitude as for diameter and height (Table 2), which are usually the primary selection traits in tree breeding.It should be pointed out that the stratifi ed sampling procedure, i.e. division of all the material into three growth categories for which the best two thirds were used in the study, probably infl uenced the results.In addition, some of the clones were related, i.e. clone 1 and 79 were selected within the same full sib family and clones 12 and 18 within the same half sib family (Table 1).It is likely that the overall genetic variation for at least the growth and dry weight traits was reduced by this selection process, resulting in underestimations of H 2 and CV G , and a poorer precision for r G .On the other hand H 2 was relatively high and CV G was similar to previous estimates (Karlsson and Danell 1992, Stener unpublished data).
In practical hybrid aspen forestry only clones belonging to the superior and possibly the medium growth groups will be of commercial interest.Thus, the sampling strategy was consistent with likely future commercial clone selection.We also chose to consider the nutrient concentrations and amounts of stems and branches separately, because the future use and assortments of stems and branches are diffi cult to predict.For example, it has long been expected that tops and branches will continue to be the main tree parts used for biofuel purposes.Recently, however, new economic estimates have shown that stems may be the most valuable energy component of the trees (Alriksson 2002a(Alriksson , 2002b)).
The results from the present study should be interpreted with caution since it is based on a single trial with 14 clones (Figs. 1, 2, 3).The results may refl ect the specifi c conditions at the study site in the restricted period examined.To our knowledge there are no reports about genotype -environment interactions on nutrient contents in trees, and further investigations are needed to confi rm the results obtained in this study.The existence of nutritional differences is, however, consistent with clonal differences found for other characteristics in aspen and hybrid aspen (Stener 1998, Yu et al. 2001a, 2001b, Fernandez et al. 2002).The tested material also contained similar concentrations of nutrients in stems and branches (Table 2), as reported for fast-growing hardwood trees (Alkan 1985, Heilman and Stettler 1986, Ericsson et al. 1992, Jug et al. 1999).Thus, the nutrient concentrations of the selected hybrid aspen clones seem to have been within the normal ranges expected for the species at such a site.
The differences in stem and branch nutrient concentrations among clones may refl ect differences in various physiological processes in the trees.It is possible that the nutrient use effi ciency varies among the clones, i.e. that there is a difference in the amounts of nutrients they require to produce a certain amount of biomass.Such differences in nutrient use effi ciency have been reported for short rotation forest species (e.g.Heilman and Stettler 1986, Simon 1992, Heilman and Xie 1993).Another possibility is that the amount of nutrients withdrawn from the leaves before leaf fall varies between clones.The amounts of nutrients removed from the leaves can be substantial in fast-growing hardwood species (e.g.Bernier, 1984, Alkan 1985, Pregitzer et al. 1990, von Fircks et al. 2001).Delays in dormancy may result in less nutrients being transported back to the woody tissues (Ericsson et al. 1992).Differences in phenology, i.e. the timing of leaf-colour change and leaf fall, have been reported among hybrid aspen clones (Yu et al. 2001b).A third possibility is that the nutrient storage allocation pattern may differ between clones.Pregitzer et al. (1990) showed that large diameter structural roots constituted a major site of N storage in young hybrid poplars, and that differences existed among the clones they studied.In our study we only observed the overall effect of the processes mentioned above.This may be enough to draw conclusions about practical implications for hybrid aspen forestry, but in order to improve nutritional effi ciency in future breeding programmes it would be desirable to have more precise knowledge about nutrient use and circulation.
The study detected signifi cant clonal differences in stem concentrations of K, P, and Mg (Table 2).Except for K, the genetic correlations between nutrient concentrations and stem volume were negative but not signifi cant (Table 3).This suggests that selection for nutrient effi cient clones would generally be possible without any signifi -cant reduction in genetic gain for growth.Moreover, the correlations between stem and branch nutrient concentrations were positive, except for P (Table 3).This indicates that when selecting clones with high or low nutrient concentrations in stems, the corresponding nutrients in the branches will be changed in the same direction.
The practical implications of the identifi ed clonal differences for the amounts of nutrients removed during harvest operations were evaluated after making some assumptions.The fi rst was that the hybrid aspen would grow with a mean annual increment of 20 m 3 over bark volume ha -1 yr -1 during a 25-year rotation period (Rytter 2002, Rytter et al. 2002).Therefore, only the superior clones were used.The second assumption was that clones with the lowest stem concentrations of the elements would be used instead of those with the highest stem concentrations (Table 4).The annual differences per hectare due to this clonal selection would be 0.64 kg K, 0.24 kg P, and 0.47 kg Mg for stemwood, and 1.02 kg K, and 0.44 kg Mg if branches were included.Phosphorus was not included in the whole-tree estimates due to a lack of signifi cant differences at the whole-tree level (see Results).It also seems reasonable to assume that nutrient concentrations in existing Swedish hybrid aspen material are on average close to the mean of the 14 clones included in the present study.Consequently, a third assumption was that half of the difference between the "extreme" clones (Table 4) could be expected when our nutrient-effi cient clones are compared with the average existing hybrid material in Sweden.This indicates, according to the results presented here, that it would be possible to save the removal of about 0.3 kg of K, 0.1 kg of P, and 0.2 kg of Mg ha -1 yr -1 from stem harvests using nutrient-effi cient clones.For a whole-tree system the fi gures would be 0.5 kg of K, and 0.2 kg of Mg ha -1 yr -1 , which correspond to 12.5 kg of K and 5 kg of Mg during a 25 year rotation period.According to estimates from Rytter (2002) for a whole-tree hybrid aspen system, around 18, 4, and 4 kg ha -1 yr -1 of K, P, and Mg, respectively, is likely to be removed by harvest.The potential gain indicated here that could be obtained using high-productivity clones with low nutrient concentrations in woody biomass during the dormancy period would be in the order of 5% for a whole-tree system.This gain seems fairly low in a short-term perspective, but in the long run cultivation of nutrient-effi cient trees may be regarded as a small but possible way of saving nutrients and decreasing the reliance on reapplication of ash and weathering.
Low concentrations of nutrients in stems and branches are generally desirable in plantation forests.However, if the objective is to remove nutrients from the ecosystem, as in sludge application systems, clones with high concentrations (i.e.those showing a high capacity to store nutrients in woody tissues) should be preferred.The results of this study could therefore also be exploited for the opposite purpose, i.e. to identify potentially valuable clones that store high amounts of nutrients in woody tissues during dormancy.

Fig. 1 .
Fig. 1.Genetic values (BLUP) for stem volume, stem wood density, stem dry weight, and branch dry weight, expressed as relative values for each of the 14 clones included in the study.The mean relative value for all clones is equal to 100.The clones are presented in ascending order of stem volume, where the seven clones to the left belong to the medium and the seven to the right to the superior growth category.Average standard errors of BLUP values for volume, density, stem weight, and branch weight, were 12, 3, 11, and 18%, respectively.

Fig. 3 .Fig. 2 .
Fig. 3. Relative genetic values (BLUP) for concentrations of K, P, and Mg in the branches, with bark included.The clones are presented in ascending order of stem volume (see Fig. 1).Standard errors for the BLUP values of K, P, and Mg concentrations, were 19, 19, and 18%, respectively.More details are given in Fig. 1.Fig. 2. Relative genetic values (BLUP) for concentrations of K, P, and Mg in the stems with bark included.The clones are presented in ascending order of stem volume (see Fig. 1).Standard errors of BLUP values for K, P, and Mg concentrations, were 6, 6, and 6%, respectively.For more details, see Fig. 1.

Table 1 .
Origin of the parent trees for the hybrid aspen clones included in the study.Shelton, CT, USA).Milled samples were prepared by digestion in a solution of concentrated HNO 3 and HClO 4 , 2 M HNO 3 was added to the digest and the solution was fi nally diluted with distilled water(cf.vonFircks et al. 2001).

Table 2 .
Mean values with standard deviation and range based on individual observations (n = 56), heritabilities (H 2 ) with standard error (S.e.H 2 ), and genetic coeffi cients of variance (CV G ) for different traits of 14 hybrid aspen clones.Bold fi gures for H 2 and CV G are signifi cantly different from zero (p ≤ 0.05).

Table 3 .
Genetic correlations (r G ) between growth, wood density, and nutrient concentration traits in the hybrid aspen stems ( S ) and branches ( B ). Bold fi gures indicate signifi cant correlations (p ≤ 0.05).Abbreviations: H = tree height; Dbh = diameter of breast height; V = stem volume; Dens = wood density; W = dry weight.

Table 4 .
Concentration and amount of nutrients with signifi cant stem differences among clones (K, P and Mg) in stems, branches and entire trees.The clone selection was based on a process in which clones with the lowest and highest relative stem concentrations (BLUP-values) of all three nutrients combined were selected.The clones were selected from within category 1, i.e. from clones with superior growth.