Annals of Botany 90: 65-76, 2002
© 2002 Annals of Botany Company
A Generic Model to Describe the Dynamics of Nutrient Concentrations within Stemwood across an Age Series of a Eucalyptus Hybrid
1 CIRAD Forêt, Campus international de Baillarguet TA 10/C, 34398 Montpellier, France, 2 CIRAD-UR2PI,BP 1291 Pointe Noire, République du Congo, 3 INRA, Cycles biogéochimiques, Centre de recherche de Nancy,54280 Champenoux, France, 4 UR2PI, BP 1291 Pointe Noire, République du Congo and 5 CNRS-CEFE 34000 Montpellier, France
* For correspondence. Fax 33 (0)4 67 59 37 33, e-mail standre{at}cirad.fr
Received: 23 November 2001; Returned for revision: 5 February 2002; Accepted: 18 March 2002
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
Nutrient concentrations (N, P, K) were determined within stemwood in an age series of eucalyptus stands. Four trees per stand were selected according to their size to represent the whole range of basal areas in 1-, 2-, 3-, 4-, 5-, 6- and 7-year-old stands. Cross-sections were sampled every 4 m from the ground to the top of the tree, and chemical analyses were performed for each annual ring in the cross-sections. We constructed a new and generic model to describe the dynamics of nutrient concentrations within the stemwood. Three main parameters were used: (1) the initial concentration of the ring, Ic; (2) the final concentration of the ring at harvest, Fc; and (3) the rate of change in concentration, k. The model is very flexible and was adapted to describe N, P and K concentrations within the stems, and their dynamics over time. An analysis of the parameters showed that k was constant for a given nutrient. Ic varied with height within the tree for P, whereas for N and K it was a function of: (1) the age of the tree when the ring was initiated; and (2) height within the tree. Fc was constant for N, and dependent on the age of the tree when the ring was initiated for K and P. The final models showed a low Root Mean Square Error for a limited number of parameters (less than seven). When validated on an independent sample, the models were shown to have high predictive quality.
Key words: Nutrient, concentration, dynamics, modelling, Eucalyptus, stemwood.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
Since 1978, 42 000 ha of clonal eucalyptus plantations have been established in the littoral savannas of Congo, mainly for pulpwood production. These plantations are based on two natural hybrids and, increasingly, on the artificial hybrid Eucalyptus urophylla S. T. Blake x Eucalyptus grandis W. Hill ex Maiden. They were established on sandy soils, characterized by low reserves of available nutrients and a low water retention capacity (Laclau et al., 2000). An intensive silviculture is carried out in these plantations, resulting in the export of high amounts of biomass and nutrients every 7 years. To assess the risks of nutrient deficiencies and unsustainable production, studies have focused on the biogeochemical cycle of nutrients in clonal stands during the planted crop rotation (Bouillet et al., 1997).
The production of stemwood represents a major sink of nutrients for eucalyptus hybrids in Congo (Laclau et al., 2000), as in other forests (Switzer and Nelson, 1972; Binkley et al., 1982; Ranger and Turpault, 1999). This component of the trees is largely involved in the biogeochemical cycles in forest ecosystems, through nutrient removals at harvest and nutrient translocations throughout stand rotation (Turner and Lambert, 1983; Miller, 1995; Colin-Belgrand et al., 1996; Ranger et al., 1997). Previous studies dealing with the dynamics of biomass and nutrient accumulation during stand development (Laclau et al., 2000) and the dynamics of nutrient translocations within stemwood (Laclau et al., 2001) were carried out for one Eucalyptus hybrid clone in Congo. In the present paper, we model the dynamics of nutrient concentrations within annual growth sheaths of stemwood for the same clone.
Our objectives were: (1) to describe annual ring nutrient concentrations within and along the tree stems and how they change with age; and (2) to model these variations. An independent validation sample was collected to evaluate the predictive quality of the equations.
Background
Distributions of nutrients within the stemwood have been studied for a wide range of forest species and some general patterns can be found:
Variation among rings, from the periphery to the pith.
Different patterns are found according to the nutrient being studied: for mobile elements (usually N, P and K), concentrations decrease sharply and reach an asymptote, whereas for non-mobile elements (generally Ca) this decrease is less marked and can be followed by a positive linear function (Bamber, 1976; Clément and Janin, 1976; Helmisaari and Siltala, 1989; Dambrine et al., 1991; Colin-Belgrand et al., 1993a; Myre and Camiré, 1994; Rochon et al., 1998; Laclau et al., 2001).
Variation along the bole.
Lemoine et al. (1988, 1990) found a decrease in concentrations of N, P and K in stemwood, from the top of the stem to a height of 2 m, for maritime pine (Pinus pinaster Ait.). Concentrations then increased slightly down to the bottom of the bole. At each sampling height, the whole cross-section was considered and no information about the individual tree ring distributions was available. Rochon et al. (1998) also found variations in nutrient concentrations along the bole, but the differences were not significant at the 1 % probability level regardless of the nutrient (N, P, K, Ca, Mg) or species studied [aspen, Populus tremuloides Michx.; birch, Betula papyrifera March.; white spruce, Picea glauca (Moench) Voss; or balsam fir, Abies balsamea (L.) Mill.]. Again, no information was available about the individual tree ring distribution: the samples of wood were taken from the periphery of the bole only and ring ages were not recorded. The study of Myre and Camiré (1994) focused on two 18-year-old plantations of larch [Larix decidua (Mill.)] and tamarack {Larix laricina [K. Koch (Du Roi)]}. Cross-sections were collected every 20 cm along the bole, and individual ring data were recorded. Except for the apical zone where higher concentrations of mobile elements were found, no particular vertical trend was identified by the authors. Lastly, Colin-Belgrand et al. (1993a) identified individual groups of rings and found a clear effect of height position within the tree on nutrient concentrations for Castanea sativa (Miller) coppices. A decrease from the upper part of the tree to the bottom was observed, except close to the stump where concentrations increased slightly. Similar trends were reported by Helmisaari and Siltala (1989) for Pinus sylvestris L. and by Monestier (1993) for Pseudotsuga menziesii Mirb.
Variation with ring age.
Most studies that deal with the dynamics of nutrient concentrations within stemwood throughout stand rotation have not distinguished individual rings. For example, Helmisaari and Siltala (1989) found an effect of stand age on nutrient concentrations, but they considered the whole cross-section and concentrations in the rings were not determined independently. However, Colin-Belgrand et al. (1993a) analysed nutrient concentration in individual groups of rings across an age series of chestnut trees. A clear decrease in concentration with ring ageing was seen for all nutrients, regardless of the height position within the bole. Monestier (1993) carried out a similar analysis on Douglas fir (Pseudotsuga menziesii Mirb.) and found the same trends. In contrast, an increase in concentration of Ca during ageing of the rings was found by Dambrine et al. (1991) in an age series of Picea abies (Karst.).
Some models have been designed to describe these nutrient patterns and their dynamics: Lemoine et al. (1990) proposed a model with two components (two opposite gradients) to describe changes in N, P and K concentrations along the bole. Both gradients were modelled as a negative (or positive) exponential function of the height position within the tree.
Rochon et al. (1998) used the model proposed by Myre and Camiré (1994) to predict nutrient distributions within the bole of four species. The context of their study was a colonizing forest following a fire, and the age of the trees was unknown. For each species, nutrient concentrations within the bole were a function of the distance from the periphery of stemwood and not a function of the ring characteristics. The equation was integrated (assuming that the bole was a cone) to evaluate the whole stem nutrient concentration. Such a model, coupled with a stem taper equation, could be used to assess nutrient removals at harvest, for example.
Colin-Belgrand et al. (1993b, 1996) and Monestier (1993) developed another approach, based upon an age series of chestnut tree coppices and Douglas fir stands, respectively. To our knowledge, this was the first time that a model had been constructed to describe the dynamics of nutrients in individual growth sheaths within stemwood throughout stand rotation. Colin-Belgrand et al. and Monestier used the same methodology: in each stand of the age series, nutrient distribution within stemwood was a function of the height position within the bole, the physiological age of the ring (time since ring initiation), the cambial age (time since cambium initiation at the considered height) and ring width. In a second step, they studied variations in parameters with stand age to obtain a global model. Results were satisfactory, showing low Root Mean Square Errors (RMSE), and parameters that were relatively uncorrelated. Such a model could be combined with a growth model to quantify the impact of various management practices on the removal of nutrients at harvest. However, using such an approach (modelling within-stand variation and thereafter between-stand variation), nutrients dynamics were not explicitly modelled: tree age was introduced in the equation through the parameters, rather than being the principal component of the model. The authors finally obtained some global equations that had a great number of parameters (from ten to 15).
According to the above considerations, three constraints were laid down in constructing our model: (1) the dynamics of nutrients within individual rings had to be explicit in the equation; (2) there had to be a small number of parameters and little correlation among them; and (3) the equation had to be generic enough to be applicable to different nutrients and species.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
Plant material
Calibration sample.
The ecological situation, the characteristics of the age series and the methodology used to determine nutrient concentrations within the stemwood were presented in a previous paper (Laclau et al., 2001). Briefly, 48–180 trees were inventoried in an age series of eucalyptus stands (1-, 2-, 3-, 4-, 5-, 6- and 7-year-old stands were sampled) in the coastal plains of Congo. The soil properties of this Ferralic Arenosol (FAO classification) and the silvicultural practices (planted crop rotation, stocking density about 700 stems per ha) were similar in all the seven stands. The climate is equatorial with a mean annual rainfall of 1200 mm and a mean annual temperature of 25 °C.
In each stand, four classes of basal areas were defined from an inventory. One tree per class was felled for further analysis. Within the 28 sampled trees (seven stands x four trees), discs of wood were cut every 4 m from the bottom to the top of the stem. A particular methodology was used to retrieve annual rings from all the sampled discs of wood despite the continuous growth of eucalyptus in the Congo (for details, see Laclau et al., 2001). Chemical analyses were then performed individually for all annual rings. All samples were scanned with a near infrared reflectance spectrophotometer (NIRSystem 6500) to check the consistency of the determinations.
Validation sample.
The validation sample was collected in the same area. A complete description of the age series can be found in Laclau et al. (2000). Three trees were sampled at ages 13, 50 and 88 months in plantations with similar characteristics to the calibration sample (same clone, identical silviculture, similar growth). Trees were selected according to tree dominance (in each stand a suppressed, an average and a dominant tree). Discs of wood were sampled every metre from the stump to the tip of 1-year-old trees and every 3 m in 4- and 7-year-old trees. A total of 53 cross-sections was sampled. Individual rings were not recorded in these samples and chemical analyses were performed on the whole cross-sections.
Definitions
Variables were defined to take into account explicitly the nutrient dynamics (Table 1 and Fig. 1). We distinguished variables that locate the ring within the tree (Rh, tc) from variables that describe the ring status (tin, t). Any symbols not defined in the text are defined in Table 1.
|
|
Model fitting
Fits of non-linear equations were performed with the Proc NLIN of SAS software (SAS, 1989) using the Marquardt convergence procedure. Parameters were removed from the equation when their asymptotic interval of confidence included zero. We took into account neither the auto-correlations between measurements of nutrient concentrations in the same ring along the bole, nor the auto-correlations between successive values during ageing in a given ring at a given height. However, the latest ones were limited owing to the method of sampling (age series).
Four main criteria were used to evaluate model performance: (1) the RMSE; (2) the distribution of errors; (3) the number of parameters; and (4) the correlation between parameters.
Special attention was paid to the RMSE, which reflects the spread of errors. The model’s suitability to describe the dynamics of nutrients decreases with an increase in RMSE. However, when the number of parameters was different between two models, the comparison was not so direct and we used the following test F[p1 – p2,n – p1]:
where p1 is the number of parameters for model 1, p2 is the number of parameters for model 2 (p1 > p2),
Model evaluation
The final model of each nutrient was evaluated using the validation sample (Fig. 5). First, we estimated the past height and diameter growth of each tree through a simple growth model (monomolecular function) that was fitted on the age series data. Secondly, using a taper equation and the past growth of the tree, we were able to assess the internal ring width distribution. Thirdly, the nutrient concentration of each ring was calculated using the models developed in the present study. Finally, the average nutrient concentration of each cross-section (Ch) was calculated using the equation:
|
where Atc,h and Ctc,h are, respectively, the area and the nutrient concentration of the rings within the cross-section (Table 1). The estimated concentrations for each cross-section were then compared with measured values. Model performance was evaluated from the distribution of errors to check the spread of errors and to detect any bias.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
Between-tree and within-tree variation in nutrient concentrations
Figure 2 shows changes in concentrations of N, P and K within stemwood along the bole for the suppressed and the dominant tree at the age of 7 years. Whatever the height within the tree, the concentrations of N, P and K tended to decrease from the periphery to the pith, and from the top of the tree to the bottom. This overall pattern was found whatever the age of the trees, but some differences among elements were noted.
|
The highest concentration of K was observed down to 8 m from the apex (which corresponds approximately to the living crown; Nouvellon et al., 2001), irrespective of tree age. The concentration then decreased sharply down to the bottom of the bole. The lowest concentrations were found in the inner rings close to the stump.
The concentration of N decreased more smoothly from the top to the bottom of the bole, whatever the age of the tree. Nevertheless, a clear increase in concentration was observed close to the stump in suppressed trees in the 7-year-old stand. This trend was less marked for dominant trees.
The concentration of P within stemwood decreased smoothly from the top of the bole to a height of 4 m, as observed for N. A sharp decrease in P concentration was observed close to the stump in all rings.
Variation over time
Marked changes in nutrient concentration during ageing of annual growth sheaths were observed at various heights for the suppressed tree (Fig. 3A) and the dominant one (Fig. 3B). We focused our analysis on three points: (1) the initial concentration of the ring – Ic; (2) the final concentration of the ring in 7-year-old trees – Fc; and (3) the rate of change in concentration – k (shape of the curve).
|
An overall decrease in nutrient concentration with time was observed. For potassium, the shape of the curve could be fitted by a negative exponential function followed by a null linear function (constant value). The negative exponential function was generally followed by a positive linear function for nitrogen and phosphorus (which corresponds to an increase in concentration at the end of rotation). A similar pattern was found regardless of the height within the bole and the dominance of the tree.
The initial concentration (Ic; Fig. 3A and B) of potassium decreased sharply with the age of the tree at ring initiation (tin). By contrast, Ic was almost constant for phosphorus for all values of tin; nitrogen followed an intermediate pattern. For all nutrients, Ic varied slightly according to the height position within the stem. The highest values were observed at the top of the tree.
The final concentration (Fc; Fig. 3A and B) of P was clearly tin-dependent: the earlier the ring was initiated, the higher the P concentration. The same pattern was found for K, but it was much less pronounced. No obvious trend was noted for N; the final concentration of the ring initiated in the first year of growth (tin = 1) was not systematically lower than that of other rings (Fig. 3A and B).
Model description
Considering the above observations, the following model was fitted ring by ring:
where the first component of the model corresponds to the decrease in nutrient concentration with time and the second component is introduced to separate Ic from Fc clearly. The main aim of this approach was to reduce the possible correlation between Ic and Fc, and to take the rotation length (tfin) explicitly into account in the equation. Results of individual fits were satisfactory: the model converged quickly, all parameters were significantly different from zero and the R2-value was rarely lower than 0·7. Therefore, we analysed how the parameters varied with height, ring and tree characteristics to give a complete model for each nutrient (Table 2). k was constant for a given nutrient. Ic varied with height within the tree for N, P, K, and was also a function of tin for N and K. Fc was constant for N and tin-dependent for K and P. For each nutrient, Table 3 gives the global model that was finally fitted on the whole calibration sample.
|
|
Model fitting
All model parameters were significantly different from zero (Table 3). The RMSE was very low compared with the average values for N and K. The number of parameters (less than seven) and correlation among them were quite low. The best results were obtained for N (ten out of 15 correlations had R-values lower than 0·1) and the least satisfactory for P (four out of ten correlations had R-values lower than 0·1). The asymptotic intervals of confidence were quite low compared with the parameter values, indicating relative stability of the models. Residuals were well distributed around zero whatever the variable used in the model (Fig. 4). However, a slight underestimation was observed at t = 6 for N, at Rh > 0·7 for P and at tin = 6 and 7 for K.
|
An F-test was performed to compare the global model with a tree-by-tree model and thus test the effect of tree dominance on the dynamics of nutrients. For N and P, the observed value of F indicated that the global model was sufficient to describe the dynamics of these two nutrients at the 95 % level of confidence. By contrast, the test was significant at the 95 % level of confidence for K. However, owing to the low RMSE of the global model, which was close to the accuracy of the chemical analysis, it was decided to use the global model to describe the dynamics of this nutrient.
Model evaluation
Figure 5 illustrates the stem reconstruction using: (1) an individual height and diameter growth model (monomolecular function); and (2) a stem profile equation for evaluating the ring increment at any height within the tree. Qualitatively, the estimated profile was quite realistic: rings were larger at the top of the tree than at the bottom; for a given height, the decrease in ring width with the cambial age was correctly predicted. Unfortunately, it was not possible to make a quantitative evaluation of such a reconstruction with the validation sample because rings were not recorded on the cross-sections.
The nutrient concentration within each cross-section was then estimated using eqn (2). The average and standard deviation of residuals (measured concentration minus the estimated one) were calculated for each nutrient and for the whole validation sample (Table 4). Results were satisfactory: for all nutrients, the average of the residuals was close to zero, and with the exception of K, the standard deviation was similar to the RMSE found for the calibration sample.
|
For each nutrient, Fig. 6 shows the average of residuals according to tree age and tree dominance. For K and N, a slight effect of tree size was detected at 13 and 88 months, respectively. For P, a slight underestimation at age 13 months was observed. However, errors were quite well distributed for all nutrients.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
Nutrient distribution in stemwood
The dynamics of nutrient concentrations in annual growth sheaths during stand development have already been discussed for this age series (Laclau et al., 2001). However, the distribution of nutrient concentrations along the bole and the variability of concentrations according to the dominance of the tree were not tackled.
The decrease in concentration of N, P and K observed for this eucalyptus clone within stemwood from the top of the bole to the bottom, and from the outer rings to the inner ones, is a general feature of many temperate forest species (Lemoine et al., 1988; Helmisaari and Siltala, 1989; Colin-Belgrand et al., 1993a; Monestier, 1993). This behaviour is characteristic of mobile elements and may be interpreted as internal translocation from degenerating cells to cells with high nutrient requirements located in the living crown and close to the cambium. A slight increase in nutrient concentrations within stemwood close to the stump has also been reported for maritime pine (Lemoine et al., 1988) and chestnut trees (Colin-Belgrand et al., 1993a) but not for European larch and tamarack (Myre and Camiré, 1994). In our case, this phenomenon was observed only at 7 years of age for nitrogen, suggesting that this increase in concentration might be the result of storage in the stump at the end of stand rotation. The function of nitrogen storage of roots was demonstrated by 15N-labelling for Picea sitchensis (Millard and Proe, 1992). However, dynamics of concentrations of P and K in the bole showed that storage in the stump did not occur concomitantly for all mobile elements.
The low between-tree variability in concentrations of N and P within stemwood led to non-significant differences (P < 0·05) between tree-by-tree models and the global one. In contrast, higher concentrations of K within stemwood of suppressed trees, irrespective of stand age, was involved in the significant loss of accuracy of the global model compared with tree-by-tree models. Discrepancies in the concentration of K among trees were most marked in the outer growth sheaths. This feature might be interpreted as a dilution effect of K in the current biomass of wood produced. Strong variations in concentrations of N, P and K were also found in the outer rings of chestnut trees according to the dominance of the tree, whatever the stand age (Colin-Belgrand et al., 1993a). At each age, however, concentrations in the smaller and the larger tree sampled were higher than those in trees of intermediate size, indicating that dilution of nutrients in the biomass of wood produced is not the only process involved.
Modelling nutrient dynamics: interpretation of the parameters
The proposed model was quite simple, with only three main parameters [eqn (1)]. It was generic enough to fit well with the three nutrients studied, although their concentration dynamics within the stemwood were different.
The initial concentration of the rings varied with height within the bole (Rh) and the age of the tree when the ring was initiated (tin). These two variables express an ageing of the cambium, which is younger at the top of the tree than at the bottom (tc = 0 at Rh = 1 and tc = Tage at Rh = 0) and matures as and when the tree grows. Therefore, ageing of the cambium is likely to explain a large part of the variability in initial concentration in the rings of this clone throughout stand rotation. Such a pattern is consistent with the results of Colin-Belgrand et al. (1993a, b) who found an effect of the relative height on the initial ring concentration. However, they did not introduce tin in their models. Dambrine et al. (1991) found more contrasting results concerning initial nutrient concentrations in rings of Picea abies (the highest concentrations were not observed for the youngest trees for K and N); these authors did not study variation in concentrations along the bole.
Considerations of the models
The final number of parameters was relatively small (maximum of seven parameters for K) compared with the number of parameters used by Colin-Belgrand et al. (1993b) and Monestier (1993) for a similar RMSE and an equivalent range of nutrient concentrations (with the exception of P). Furthermore, the confidence interval of the parameters and the correlation among them were both very low, indicating stability of the estimations.
However, for proper use of these models, some limitations have to be imposed. The validation step showed a slight effect of tree dominance on estimates of nitrogen concentration at age 88 months (Fig. 6). An accurate evaluation of nitrogen content in the trees at harvest is essential because this element is the main factor limiting tree growth in the Congo (Laclau et al., 2000). This behaviour of the model will have to be confirmed and, if necessary, the equation will have to be adapted to take into account the dominance of the trees at the end of stand rotation. A similar influence of tree dominance was observed for K at age 13 months. This result was expected because a significant difference between the global model and the tree-by-tree models was found for this element. However, at age 88 months the evaluation was correct, and the effect of tree dominance was much less pronounced. One possible way of improving these models could be use of non-linear mixed models to simulate the between-tree variability that is not taken into account in the fixed effects (Hervé et al., 1999; Fang and Bailey, 2001).
Furthermore, we have to keep in mind the calibration domain of these models. First, they were constructed and validated using 1- to 7-year-old trees. The predictive quality of the model is satisfactory for the current rotation age (7 years) when wood production is devoted to the pulp-wood industries. However, if rotation length is increased for timber production, then use of these models for older trees is clearly questionable. Indeed, we do not know whether the curves are simply expanded or if more complex phenomena occur after 8 years of age. Secondly, concentrations within the stems were determined for the silviculture carried out in the Congo. If a different silviculture is applied, and particularly if heavy fertilization is used, the distribution of nutrient concentrations within stemwood is likely to change greatly. A field experiment has shown a strong impact of the availability of nutrients at planting on concentrations within stemwood at 1 year of age (Bouillet et al., 2001).
These two limiting points (stand age and silviculture) need to be verified to extend the validity of the models. However, the equations have been formulated to avoid outliers in the estimation: the ring nutrient concentration is somehow constrained to vary between the initial concentration and the final one, both being fixed (Table 3). Therefore, estimations done outside the calibration range should remain consistent with the actual values if the initial and final concentrations are suitable.
Another advantage of the model formulation is that the parameters could be roughly estimated for other clones through simple measurements. We illustrate the method for nitrogen, but it can be applied to the other nutrients (Table 2). The parameter a3 could be assessed by measuring Ic at the top and at the bottom of 1-year-old trees of another clone (Ic = at the top of the tree and Ic = at the bottom). Similarly, sampling the first ring (near the pith) and the outer one in 7-year-old trees makes it possible to evaluate Fc, a1 and a2. By assuming that the shape of the curve (parameter k) and parameter a4 do not vary greatly among clones, we obtain a model that would be valid for a new clone (to a first approximation). Such a property of these equations is beneficial and avoids the need for many measurements if very precise estimates of nutrient content are not necessary. Of course, to obtain more accurate models for other clones, it is necessary to sample trees in an age series, as was done in this study, and fit the equations again.
In addition, this model could easily be adapted to other nutrients (e.g. Ca or Mg) or extended to other species. Such equations involving an initial property and its changes over time in reaching a final value could also be used to describe the dynamics of other wood properties that depend on tissue ageing. For example, the equation could be tested on the wood colour distribution which results from complex chemical transformations during the maturation of wood (Klumpers, 1992; Burtin, 1999).
We shall include these new equations in a growth model that is under construction for eucalyptus plantations in Congo. The process can be summarized in five steps: (1) at a given age, growth models estimate increments in height and diameter along the bole; (2) taper functions fitted for this clone calculate the stem profiles and the volume of annual growth sheaths; (3) nutrient concentrations of the outer sheath and the changes in nutrient concentrations of the inner ones are estimated using the models presented in this study; (4) the biomass increment is calculated from the volume increment and a wood density model (for this clone, changes in wood infradensity are negligible during heartwood formation processes; H. Bailleres, pers. comm.); and (5) nutrient contents are therefore calculated for the whole stemwood. These five steps are then repeated every year throughout stand rotation.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
The chronosequence approach used in this study provided useful information to describe the dynamics of nutrient concentrations within stemwood over the whole stand rotation. Simple models were built to predict concentrations of N, P and K in the stemwood of a eucalyptus clone, whatever the age of the tree. The equations were constructed to take into account explicitly the ageing of the rings and to limit the predictions of the model within a range of possible values. The biological significance of the parameters and the formulation of the equation distinguishing two successive stages (ring initiation and maturation) were probably responsible for the low correlation among parameters, accountable for the stability of the predictions. Evaluation of these models on an independent data set showed that their predictive quality was satisfactory.
Owing to their simple formulation, these models are generic enough to describe changes in nutrient concentrations in individual rings of other species or to study other wood properties that change during ageing in a similar manner: initial exponential l decrease followed by smaller linear variation. These models will be embedded within a growth and yield model (EUCALYPT) that is under construction for the plantations in Congo. We shall first model the dynamics of nutrients in annual growth sheaths throughout stand rotation, coupling the equations developed here with models describing volume increments and distribution of infradensity in the stems. Later, we intend to drive the growth model from nutrient and water availability in the soil.
|
ACKNOWLEDGEMENTS |
|---|
We acknowledge Jean-Claude Mazoumbou for collecting all the field samples and for stem ring analysis. We thank Laurent Veysseyre (IRD/Pointe-Noire) and Gisèle Heral-Llimous (CIRAD-Amis) for performing mineral analysis. We also thank Jean-Pierre Bouillet for his careful review of the manuscript. The founders of UR2PI, République du Congo, CIRAD-Forêt and ECO s. a. supported this study financially.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSLITERATURE CITED |
|---|
-
Bamber RK. 1976. Heartwood, its function and formation. Wood Science and Technology 10: 1–8.
Binkley D, Kimmins JP, Feller MC. 1982. Water chemistry profiles in an early- and mid-successional forest in coastal British Columbia. Canadian Journal of Forest Research 12: 240–248.
Bouillet JP, Nizinski G, Nzila JD, Ranger J. 1997. The sustainability of Eucalyptus commercial plantations: the Congolese approach. In: Proceedings of IUFRO Conference 4, Salvador, Brazil, 24–29 August, 1997, 232–237.
Bouillet JP, Nzila JD, Laclau JP, Ranger J. 2001. Slash management effects on tree growth and nutrient cycling in young Eucalyptus replanted sites in the Congo. Proceedings of the 4th CIFOR Workshop ‘Site Management and Productivity in Tropical Plantation Forests’, Pointe-Noire, Congo, 11–14 July 2001.
Brown D, Rothery P. 1994. Models in biology: mathematics, statistics and computing. Chichester, UK: John Wiley & Sons Ltd.
Burtin P. 1999. Origine biochimique de la couleur du Bois de Noyer (Juglans sp.). PhD Thesis, Nancy I, France.
Clément A, Janin G. 1976. Etude comparée de la répartition des principaux cations et du phosphore dans une tige de peuplier ‘Fritzi-Pauley’. Plant and Soil 45: 543–554. [CrossRef][Web of Science]
Colin-Belgrand M, Ranger J, Bouchon J. 1996. Internal nutrient translocation in chestnut tree stemwood: III. Dynamics across an age series of Castanea sativa (Miller). Annals of Botany 78: 729–740.
Colin-Belgrand M, Ranger J, Bouchon J, d’Argouges S. 1993a. Transferts internes d’éléments nutritifs dans le bois de chataignier (Castanea sativa Miller): approche dynamique sur une chrono séquence de peuplements: I. Distribution des éléments minéraux. Acta Oecologica 14: 653–680.
Colin-Belgrand M, Ranger J, Bouchon J, d’Argouges S. 1993b. Transferts internes d’éléments nutritifs dans le bois de chataignier (Castanea sativa Miller): approche dynamique sur une chrono séquence de peuplements: II. Modélisation de la distribution des éléments minéraux. Acta Oecologica 14: 747–769.
Dambrine E, Le Goaster S, Ranger J. 1991. Croissance et nutrition minérale d’un peuplement d’épicéa sur sol pauvre. II – Prélèvement racinaire et translocation d’éléments minéraux au cours de la croissance. Acta Oecologica 12: 791–808.
Fang Z, Bailey R L. 2001. Nonlinear mixed effects modeling for Slash Pine dominant height growth following intensive silvicultural treatments. Forest Science 47: 287–300. [Web of Science]
Helmisaari HS, Siltala T. 1989. Variation in nutrient concentrations of Pinus sylvestris stems. Scandinavian Journal of Forest Research 4: 443–451. [CrossRef]
Hervé JC. 1999. Mixed-effects modelling of between-tree and within-tree variations. Application to wood basic density in the stem. In Leban JM, Hervé JC, eds. Final report sub-task A2·1 Modelling stem properties distribution. European FAIR CT-96-1915 project.
Klumpers J. 1992. Influence of age and annual ring width on the wood colour of Oak. Holz als Roh-und Werkstoff 50: 167–171. [CrossRef]
Laclau JP, Bouillet JP, Ranger J. 2000. Dynamics of biomass and nutrient accumulation in a clonal plantation of Eucalyptus in the Congo. Forest Ecology and Management 128: 181–196. [CrossRef][Web of Science]
Laclau JP, Bouillet JP, Ranger J, Joffre R, Gouma R, Saya A. 2001. Dynamics of nutrient translocation in the stemwood across an age series of a eucalyptus hybrid. Annals of Botany 88: 1079–1092.
Lemoine B, Ranger J, Gelpe J. 1988. Distributions qualitative et quantitative des éléments nutritifs dans un jeune peuplement de Pin maritime (Pinus pinaster Ait). Annales des Sciences Forestières 45: 95–116.
Lemoine B, Ranger J, Gelpe J. 1990. Interprétation et modélisation de la distribution des éléments nutritifs dans les différents compartiments de la biomasse d’un jeune peuplement de Pin maritime. Annales des Sciences Forestières 47: 101–115. [CrossRef][Web of Science]
Millard P, Proe MF. 1992. Storage and internal cycling of nitrogen in relation to seasonal growth of Sitka spruce. Tree Physiology 10: 33–43. [Abstract]
Miller HG. 1995. The influence of stand development on nutrient demand, growth and allocation. Plant and Soil 168–169: 225–232. [CrossRef]
Monestier M. 1993. Modélisation de la distribution des éléments nutritifs majeurs dans le bois de tronc de Douglas (Pseudotsuga menziesii Franco). Approche dynamique sur une chronoséquence de peuplements. D. E. S. S. d’ingénierie Mathématique et d’Outils Informatiques, Université Nancy I.
Myre R, Camiré C. 1994. The establishment of stem nutrient distribution zones of European larch and tamarack using principal component analysis. Trees 9: 26–34.
Nouvellon Y, Saint-André L, Dauzat J, Roupsard O, Caraglio Y, Hamel O, Berbigier P, Bonnefond J-M, Irvine M, Mabiala A, Epron D, Mouvondy W, Laclau J-P, Bouillet J-P, Jourdan C. 2001. Linking tree architecture and function for continuous simulations of carbon, water, and energy fluxes from Eucalyptus stands in Congo. Third International Workshop on Functional-Structure Tree and Stand Models, September 27–29, Val-Morin, Quebec, Canada.
Ranger J, Turpault MP. 1999. Input-output nutrient budgets as a diagnostic tool for sustainable forest management. Forest Ecology and Management 122: 139–154. [CrossRef][Web of Science]
Ranger J, Marques R, Colin-Belgrand M. 1997. Nutrient dynamics during the development of a Douglas-fir (Pseudotsuga menziesii Mirb.) stand. Acta Oecologica 18: 73–90. [CrossRef]
Rochon P, Paré D, Messier C. 1998. Development of an improved model estimating the nutrient content of the bole for four boreal tree species. Canadian Journal of Forest Research 28: 37–43.
SAS Institute Inc. 1989. SAS/STAT user’s guide, version 6, 4th edn, volume 1. Cary, NC: SAS Institute Inc.
Switzer GL, Nelson LE. 1972. Nutrient accumulation and cycling in Loblolly Pine (Pinus taeda L.) plantation ecosystems: the first twenty years. Soil Science Society of America Proceedings 36: 143–147. [Web of Science]
Turner J, Lambert MJ. 1983. Nutrient cycling within a 27-year-old Eucalyptus grandis plantation in New South Wales. Forest Ecology and Management 6: 155–168.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  L. Augusto and D. Bert Estimating stemwood nutrient concentration with an increment borer: a potential source of error Forestry, October 1, 2005; 78(4): 451 - 455. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||