HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Montagnon, C. Articles by Cilas, C. Search for Related Content PubMed Articles by Montagnon, C. Articles by Cilas, C. Agricola Articles by Montagnon, C. Articles by Cilas, C. Related Collections Other Crops Crop Genetics

COFFEE

A New Method to Assess Competition in Coffee Clonal Trials with Single-Tree Plots in Côte d'Ivoire

^a CIRAD-CP 01 BP 6483 Abidjan 01, Côte d'Ivoire
^b CIRAD-CP, TA 80/03, 34398 Montpellier Cedex 5, France. This work is the result of a cooperation between the Centre National de Recherche Agronomique of Côte d'Ivoire (CNRA) and the Centre International de Recherche Agronomique pour le Développement–Cultures pérennes (CIRAD–CP)

Corresponding author (christian.cilas{at}cirad.fr)

	ABSTRACT

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 
A new method is proposed for the assessment of competition in clonal trials with completely randomized single-tree plots, without needing to choose a covariable a priori linked to competition. This method is applied to a coffee clonal trial planted in a completely randomized single-tree plot design located in Côte d'Ivoire. Microenvironmental effects were first taken into account using the conventional Papadakis method. The competition effect (C_k) of each clone (k) was then estimated as the influence of its presence on the residual value, after removal of the clonal effect, of its neighbors. Residuals were thus modeled as a linear regression of the neighboring clones' C value. Competition effects were shown this way to explain 4% of the residual yield in young trees and 10% in adult trees. Several clones had C values significantly different from 0. Some were identified as aggressive for their neighbors (C < 0), others as stimulating (i.e., promoting yield, C > 0). Architecture and vigor variables that were likely to be related to the C value of clones were then sought. For young trees, vigor, estimated by stem diameter, was best correlated to competition effects: vigorous clones were more aggressive than others. When trees became adult, the length of the orthotropic internodes (L_ort) was the trait that most effectively explained (42%) the competition effect of clones: the shorter the internodes, the less aggressive was the clone. In the future, L_ort may thus be used in a selection index to prevent from selecting aggressive coffee clones that would undergo their own aggressiveness when grown alone in plantations.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 
ROBUSTA coffee (Coffea canephora Pierre ex A. Froehner) is a perennial plant whose improved varieties are either clones distributed in cutting form, or clone hybrids distributed in seed form (Montagnon et al., 1998). Given the strict self-incompatibility of robusta coffee (Berthaud, 1980), varieties are always mixes of clones or mixes of clone hybrids. In Côte d'Ivoire, variety comparative trials are generally planted in a completely randomized single-tree plot design (Charmetant and Leroy, 1990). The effects of competition between genotypes, defined as the influence of a given genotype on the performance of its neighbors, have never been studied in coffee, even though the existence of these effects has sometimes been mentioned (Leroy et al., 1994). Evaluation of competition effects in coffee has two main objectives: (i) improvement of the effective prediction of the genetic value and (ii) selection of low-competition varieties with a high harvest potential enabling an increase in planting density.

In theory, genetic analysis of the effects of competition between genotypes means setting up specific trials based on a comparison of binary varietal mixes, often organized in a diallel type comparison table (Gallais, 1970). Such trials, which are feasible on annual plants that take up little space, would be too laborious to set up on tree crops such as coffee. In designs with elementary plots comprising several trees, it is possible to evaluate a competition index as the ratio between the production of the outer and inner trees of the elementary plot (Glendinning and Vernon, 1965; Nouy et al., 1990).

In the case of completely randomized single-tree plots, competition is usually taken into account by introducing covariables assumed to be correlated to the competition effect of the nearest neighbors (Draper and Guttman, 1980; Kempton, 1982). The problem is then to identify relevant covariable(s). In forestry, the covariable is usually tree height, which also is the target trait for selection (Correll and Anderson, 1983). The angle between the horizontal and the axis passing through the apex of the tree and that of its neighbor also can be measured (Magnussen, 1989). However, although tree height is expected to be correlated to competition in forestry, no variable can be selected a priori as being representative of competition in coffee. The canopy diameter, which reflects tree bulk, can be assumed to be positively correlated to competition. However, coffee trees with a small canopy diameter are often taller than the others, so they could compete for light interception. This phenomenon has been clearly demonstrated for oil palm (Nouy et al., 1990). In cocoa, the area of the trunk cross-section 50 cm from the ground is considered to be correlated to the tree's competition effect (Lotodé and Lachenaud, 1988). For coffee, such a relation between vigor and the degree of competition has not been demonstrated.

In this article, we propose a way of evaluating competition effects that does not require the choice of a variable that is assumed a priori to be correlated to competition. On the contrary, evaluation of the competition effect for each variety will make it possible to identify a posteriori the vigor or architecture variables that most effectively explain this effect. This method was applied to a robusta coffee clonal trial. The relative importance of the competition effects and environmental micro-heterogeneity in the trial, estimated by the Papadakis method (Papadakis, 1937; Brownie et al., 1993), is discussed.

	Materials and methods

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 
Planting Material
The study involved a comparative trial with 26 robusta coffee clones planted in 1991 in the experimental station of CNRA in Divo (Côte d'Ivoire). The number of replicates per clone varied from 24 to 30. There was a total of 693 trees in the trial. The trial was planted in a completely randomized single-tree plot design. The density was 1961 plants per hectare. The spacing was 3 m between rows and 1.70 m between trees along the row, in a rectangle. The plants were cultivated on three stems with free growth.

Observations
The target trait studied was fresh berry weight per tree expressed in hectogram. The harvests were weighed per tree over the first 6 yr of production. Variable Y₁₃ represents cumulated yields over the first 3 yr of production (early period) and Y₄₆ represents cumulated yields from Year 4 to Year 6 (adult period). Y₁₆ corresponds to cumulated yield from Year 1 to Year 6, that is, over the 6 yr of the study. In an open growth system, coffee trees are cut back after the sixth year of production. Another production cycle then begins under similar conditions to the adult period of the first cycle.

The following traits also were measured for each tree at the age of 2 yr. Some are linked to vegetative vigor and others to architecture (Snoeck and de Reffye, 1980):

Traits linked to vegetative vigor:

• D_stem = stem diameter
  
• H_tree = height of the tree
  
• N_ort = number of orthotropic internodes
  
• L_bra = length of the longest plagiotropic branch
  
• N_pla = number of plagiotropic internodes on the longest primary branch
  
• N_tot = N_ort x N_pla = estimation of the total number of internodes
  

Architecture descriptor traits:

• L_ort = H_tree/N_ort = average length of an orthotropic internode
  
• L_pla = L_bra/N_pla = average length of a plagiotropic internode
  
• S_tree = H_tree/L_bra = estimation of tree shape
  

Analyses
For a clonal trial with a randomized single-tree plot design, K clones are compared, with r_k (k [1,K]) being the number of replications (= tree) of clone k.

Let X be the trait studied, for which the competition effects are evaluated. The position of each tree is identified by the number of the planting row (l) and the number of the tree (t) along that row.

Let the function v be such that , where k is the clone to which the tree in position (l,t) belongs. The statistical model for the analysis of variance (ANOVA) of trait X based on the varietal factor can be written as follows:

(1)

where X_lt = performance of the tree whose position is (l,t); m = total mean of the trial; a_v(l,t) = effect of clone v(l,t); and E_lt = residual value of the tree.

The rest of the analysis is based on the postulate that the variance of the residual values (E_lt) of this basic ANOVA is due to the quality of the environment, particularly the soil, in the vicinity of each tree; competition linked to neighboring trees; and a random, uncontrollable effect. Therefore, this study proposed to (i) correct the data by the Papadakis analysis of covariance (Papadakis, 1937) in order to limit the environmental effect in the vicinity of each tree as far as possible, and (ii) study competition between trees using the corrected data.

Taking Account of Micro-Heterogeneity in the Trial (Papadakis)
For each trait X considered and for each tree (l,t), the Papadakis covariable P_lt was calculated as the mean on the neighbors of (l,t) of the X values corrected for the clone effect:

(2)

where n_lt = number of neighbors of (l,t) (generally eight, fewer if some had died) and N_lt = set of couples (,) identifying the neighbors of (l,t).

The variable P_lt is designed to assess the relative microenvironment effect at location (l,t) and is thus expected to explain a part of the yield variation. The Papadakis-corrected performance of each tree (l,t), called XCORR, is the residual of the simple linear regression of X on P, that is, the part of X that cannot be explained by the variable P:

(3)

where b and c are respectively the estimated slope and intercept of the simple linear regression of X on P.

After Papadakis correction, the new ANOVA model is:

(4)

where XCORR_lt = corrected performance of the tree whose position is (l,t); µ = total mean of XCORR; _v(l,t) = effect of clone v(l,t) on XCORR; and _lt = residual value of the tree (l,t).

Study of Competition Effects
The proposed method made it possible to evaluate any effect of immediate neighboring varieties in the same row as tree (l,t) on the residual value (_lt) of that tree. For each clone k, we postulate the existence of an additive competition effect (C_k), which modifies the performance of the two neighbors of a tree belonging to the clone k. The residual (_lt) is then modeled as:

(5)

where C_v(l,t-1) and C_v(l,t+1) are the additive competition effects of the clones to which the previous and the next tree on the same row belong, and _lt = an error term (random effect explained neither by environmental quality in the vicinity of the tree nor by competition from the neighboring trees).

For a tree (l,t) surrounded by trees of two distinct clones (k and k'), this model can be written:

(6)

For a tree (l,t) whose two neighbors belong to the same clone k, the model becomes:

(7)

The K parameters C₁, ... , C_K can then be estimated as the coefficients of the multiple linear regression of on K variables, each indicating the number of times where one clone occurs in the neighborhood of a tree. To ensure a unique solution, the constraint is added to the model. The significant nature of the difference between each C_k and 0 is tested by a t-test on the regression coefficients.

The main idea underlying the computation of the C_k coefficients is to check wether the values for all trees vary independently of their neighbors' variety, or if some varieties are systematically in the neighborhood of trees whose are high or low.

The variable C_k is the competition effect of variety k, or partner effect as defined, for example, by Gallais (1975). This means that the predicted residual value of a tree of any variety surrounded by two trees of varieties k and k' is C_k + C_k'. If variety k is aggressive (unfavorable to its neighbors), then C_k is negative. On the other hand, if variety k is stimulating (favorable), C_k is positive.

Where there were empty positions in the study trial (following the death of certain individuals), it was possible with this method to estimate the competition effect of the empty plots on the performance of neighbors, by considering them as a variety. However, care had to be taken to ensure that the empty plots were randomly distributed within the plot. For example, if death was due to fertility factors corresponding to certain zones in the trial, the competition effect of the empty plots would be biased.

Once competition effects of all clones were estimated, the architecture or vigor traits that most explained them were sought. For this purpose, a multiple linear regression using the clonal mean of each trait to explain the competition effect (C) was run. The Stepwise method was used and variables were allowed for entry into the model below the 0.15 significance level. Analyses were performed using SAS software (SAS Institute, 1989).

	Results

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 
Taking Account of Micro-Heterogeneity in the Trial
The regression of variables Y₁₃, Y₄₆, and Y₁₆ on the Papadakis covariable indicated that micro-heterogeneity within the trial had a highly significant effect (P < 0.001). Indeed, the variances explained by the Papadakis covariable (R²) were equal to 34, 16, and 30% of total variance for Y₁₃, Y₄₆, and Y₁₆ respectively.

Detection of Competition Effects
During the early production period (Y₁₃), the competition effect only explained 4% of total variance (Table 1). Only Clone 202 revealed a competition effect that significantly differed from 0 at the 5% level.

For the production period as a whole (Y₁₆), competition effects explained 8% of variance. The results were comparable with the adult period for which competition effects were greatest ( as opposed to 4% for the early period).

Relations between Competition Effects and the Other Variables
The competition effect of the clones during the early period was primarily explained by the stem diameter, a vigor variable, with a partial R² of 34% (Table 2). The less vigorous the clone, the more it stimulated its neighbors.

Regressions for the whole period of production were comparable with those of the adult period.

Influence of Competition Effects on Clonal Selection Quality
A competition covariable was defined for each tree as the mean partner effect of the clones represented by its neighboring trees. Over the adult production period, during which competition effects were the greatest, correction of the clonal mean yield in line with this competition covariable did not lead to any significant changes in clone ranking.

It was nevertheless possible to theoretically predict what the best clones in the trial would produce in a plantation. Indeed, they should undergo their own aggressiveness or benefit from their own stimulation. Based on the compensation effect usually observed for numerous plants (Gallais, 1975), aggressive clones are themselves more susceptible to competition than stimulating clones. For instance, out of the highest yielding clones in the trial, it is likely that Clones 696 and 701, which were significantly aggressive (Table 1), would obtain lower genetic gains in a plantation than expected, without taking competition effects into account. On the other hand, Clones 685 and 686, which were good yielders without competition effects, and Clone 461, which was a high yielder and significantly stimulating, would probably perform as expected in a plantation.

	Discussion

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 
Most methods applied to perennial crops for studying competition phenomena are based on the a priori choice of a variable assumed to be correlated to competition (Correll and Anderson, 1983; Magnussen, 1989). However, the original method that we propose makes it possible to quantify competition first, then to determine which variables are correlated to it. Given the rectangular planting design, with virtually twice the distance between rows as between the trees (3 m as opposed to 1.7 m), only the two neighboring trees in the same row were taken into account.

In the case of this coffee clonal trial, competition phenomena seemed low (4%) in the early stage of the trial. Over the same period, the effect of micro-heterogeneity in the trial, revealed by the Papadakis analysis, was substantial (34%). The situation was different during the adult period of the trial. The effect of micro-heterogeneity in the trial decreased by more than one-half: the percentage of explanation fell from 34 to 16%. On the other hand, the percentage of variation explained by competition more than doubled from 4 to 10%.

For a perennial plant such as coffee, the rise in competition effects between trees to the detriment of the effects of micro-heterogeneity in the soil depending on the age of the trial was an expected but rarely quantified result. Nouy et al. (1990) clearly showed the increase in competition effects with age in oil palm, but did not cover the effect of the microenvironment without competition. It can be deduced from Correl and Anderson's data (1983) that competition explained barely more than 2% of the residual variance of tree height in a Monterey pine (Pinus radiata D. Don) clonal trial planted in a Fisher block design. In the same trial, soil heterogeneity taken into account by the Papadakis method and linear trends within blocks explained less than 13% of residual variance. According to our results, the effects of micro-heterogeneity within the trial and competition can be considered as strong for coffee.

Competition effects were linked to stem diameter during the young period. That corresponded to the generally acknowledged relation between vigor and competition, which leads in most cases to choosing covariables linked to vigor, in order to take competition into account (Correll and Anderson, 1983; Lotodé and Lachenaud, 1988; Magnussen, 1989). However, the competition effect of adult coffee clones was very largely dependent upon the length of orthotropic internodes. The least aggressive clones had short internodes, which generally corresponded to a bushy and branched growth habit. Clones that were not particularly bushy or branched were aggressive for their neighbors.

Adding other morphological traits at different ages would undoubtedly provide a more precise explanation of competition effects. In that respect, coffee tree architecture modeling studies (Cilas et al., 1998) should shed additional light on the subject. Even so, the study described here already shows that estimating competition effects linked to coffee trees on the basis of their vigor alone does not reflect the reality of interactions between trees in a plot. Moreover, such an approach may lead to vigorous clones, which guarantee good establishment, being pointlessly ruled out. Indeed, the decisive factor connected with competition in coffee is tree architecture, estimated from the length of the orthotropic internodes, rather than vigor.

Even if clone ranking did not change after competition effects of neighboring trees were taken into account through an adequate covariable, genetic progress achieved after selection may be different from expected. Indeed, selected clones, used alone in plantations, might undergo their own aggressiveness (progress lower than expected) or benefit from their own stimulation (progress higher than expected). This type of result tallies with those established on perennial plants such as oil palm (Nouy et al., 1990) and cocoa (Glendinning and Vernon, 1965) or on forest trees (Correll and Anderson, 1983; Magnussen, 1989). It is very likely that genetic gains obtained through the identification of individual trees that perform well in hybrid progeny trials for clonal selection also are highly dependent upon the effects of competition from the immediate neighbors of the chosen trees. This aspect will be covered in future studies.

	Conclusions

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 
One of the aims of robusta coffee varietal improvement is making progress toward high-yielding trees that do not particularly compete with each other by increasing their harvest index (Leroy et al., 1994). With the method described here, it is possible for the first time to identify and quantify competition effects in robusta coffee. The length of the orthotropic internodes has been identified as the main factor that explains clonal competition effects. This variable might therefore be included in the selection index when choosing clonal varieties or clone hybrids.

Received for publication March 15, 2000.

	REFERENCES

TOP ABSTRACT INTRODUCTION Materials and methods Results Discussion Conclusions REFERENCES

 

• Berthaud, J. 1980. L'incompatibilité chez Coffea canephora: Méthode de test et déterminisme génétique. Café Cacao Thé 24:267–274.
• Brownie, C., D.T. Bowman, and J.W. Burton. 1993. Estimating spacial variation in analysis of data from yield trials: A comparison of methods. Agron. J. 85:1244–1253.[Abstract/Free Full Text]
• Charmetant, P., and T. Leroy. 1990. Méthodologie des essais de sélection caféière en Côte d'Ivoire. p. 496–501. In 13th Colloque Scientifique International sur le Café, Paipa, Columbia. 21–25 Aug. 1989. ASIC, Paris.
• Cilas, C., Y. Guédon, C. Montagnon, and P. De Reffye. 1998. Analysis of growth monitoring of a coffee (Coffea canephora Pierre) cultivar in Côte d'Ivoire. p. 46–55. In INRA, CTIFL, CIRAD, ENSAM (ed.) Architecture and modelling in fruit tree culture. INRA, Montpellier, France.
• Correll, R.L., and R.B. Anderson. 1983. Removal of intervarietal competition effects in forestry varietal trials. Silvae Genet. 32:162–165.
• Draper, R.A., and I. Guttman. 1980. Incorporating overlap effects from neighboring units into response surface models. Appl. Stat. 29:128–134.
• Gallais, A. 1970. Modèle pour l'analyse des relations d'associations binaires. Biométrie-Praximétrie 2:51–80.
• Gallais, A. 1975. Sur quelques aspects de la compétition en amélioration des plantes. Ann. Amélior. Plant. 25:51–64.
• Glendinning, D.R., and A.J. Vernon. 1965. Inter-varietal competition in cocoa trials. J. Hortic. Sci. 40:317–319.
• Kempton, R.A. 1982. Adjustment for competition between varieties in plant breeding trials. J. Agric. Sci. 98:599–611.
• Leroy, T., C. Montagnon, C. Cilas, A. Charrier, and A.B. Eskes. 1994. Reciprocal recurrent selection applied to Coffea canephora Pierre. II. Estimation of genetic parameters. Euphytica 74:121–128.
• Lotodé, R., and P. Lachenaud. 1988. Méthodologie destinée aux essais de sélection du cacaoyer. Café Cacao Thé 32:275–292.
• Magnussen, S. 1989. Effects and adjustments of competition bias in progeny trials with single-tree plots. For. Sci. 35:532–547.
• Montagnon, C., T. Leroy, and A.B. Eskes. 1998. Amélioration variétale de Coffea canephora. I. Critères et méthodes de sélection. Plantation Recherche Développement 5:18–33.
• Nouy, B., Asmady, and R. Lubis. 1990. Effets de compétition à Nord-Sumatra dans les essais génétiques sur palmier à huile. Conséquences sur l'évaluation du matériel végétal. Oléagineux 45:245–255.
• Papadakis, J.S. 1937. Méthode statistique pour les expériences en champ. Bull. scientifique 23. Thessalonique Institut d'amélioration des plantes, Thessalonique, Greece.
• SAS Institute. 1989. SAS/STAT user's guide. Version 6, 4th ed. SAS Inst., Cary, NC.
• Snoeck, J., and P. de Reffye. 1980. Influence des engrais sur l'architecture et la croissance du caféier Robusta. Café Cacao Thé 24:259–266.

This article has been cited by other articles:

				C. CILAS, A. BAR-HEN, C. MONTAGNON, and C. GODIN Definition of Architectural Ideotypes for Good Yield Capacity in Coffea canephora Ann. Bot., March 1, 2006; 97(3): 405 - 411. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Montagnon, C. Articles by Cilas, C. Search for Related Content PubMed Articles by Montagnon, C. Articles by Cilas, C. Agricola Articles by Montagnon, C. Articles by Cilas, C. Related Collections Other Crops Crop Genetics