Adaptability and stability of organic-grown arabica coffee production using the modified centroid method

This study aimed to identify promising arabica coffee genotypes for organic systems. The experiments were arranged in a randomized block design, with 30 genotypes and three replications. The adaptability and stability analysis was carried out using the modified centroid method, considering the mean yield of two biennia (2005/2006 and 2006/2007, 2007/2008 and 2008/2009) in three municipalities (Araponga, Espera Feliz, and Tombos), totaling six environments. Significant genotype x environment interaction was observed for yield, and the municipality of Espera Feliz was the only favorable environment. Genotypes were classified into four of the seven groups proposed by the modified centroid method: maximum general adaptability (I), minimum adaptability (IV), mean general adaptability (V), and mean specific adaptability to favorable environments (VI). Cultivars IBC Palma 1, CatucaíAmarelo24/137, Sabiá 708, and H 518 are widely adapted, stable, productive and suitable for organic farming.


INTRODUCTION
Brazil is the largest coffee producer and exporter in the world, and accounts for about one third of the production and exports worldwide (OECD/FAO 2015).Since its introduction in the country in 1727, coffee has been managed in several ways, always seeking maximum yields, requiring heavy use of chemical inputs, which can negatively impact the environment.This fact, combined with the growing demand by the consumer market, which has increasingly become aware of environmental and social issues involved in the production process, have required the search for more sustainable production systems, such as organic and agroecologically-based ones (Moura et al. 2011).
Organic coffee is the most important category in the segment of certified coffee.Mexico, Honduras, Indonesia, Ethiopia, Colombia and Brazil are the largest organic coffee exporters, accounting for 84% of world trade, and the main destinations are the US, Germany, Belgium, Canada, Sweden and Japan (ICO 2014).
Production and trade of organic coffee have grown worldwide.Coffee export tripled in 2013, when compared with 2005, to approximately one million and two WM Moura et al. hundred bags, while Brazil's share in that period increased from 10,371 to 78,568 bags (ICO 2014).Despite this growth expectation, production in the country is still in its initial stages.This may be associated with unsuitable cultivars, and with the fact that organic farms were established by the conversion of existing conventional farms, keeping the previous cultivars.In fact, crops are formed mostly by the cultivars Catuaí Vermelho, Catuaí Amarelo, and Mundo Novo, which are highly susceptible to rust (Hemileia vastatrix), a major coffee disease (Malta et al. 2008).In this context, investments in research are necessary to provide the information for the recommendation of cultivars for organic production.Since coffee is a perennial and biennial crop, research demands long periods of assessments.In addition, coffee producing regions in Brazil have very different characteristics, whichfavor the genotypes x environments interaction.The recommendation of stable cultivars with wide adaptability is an alternative to mitigate this interaction (Cruz et al. 2012).According to Moura et al. (2014), the estimates of adaptability and stability parameters should be analyzed based on biennia yield, aimed at selecting promising cultivars for organic systems.
Stability and adaptability analysis allowed the identification of cultivars with predictable behavior, and which are responsive to environmental variations, in specific or broad conditions (Cruz et al. 2012).Among several methodologies for this purpose, the centroid method (Rocha et al. 2005) has been widely used in several crops, including soybean (Barros et al. 2010, Pelúzio et al. 2010, Barros et al. 2012), tomato (Pereira et al. 2012), sweet potato (Amorin et al. 2011), and coffee (Rocha et al. 2015).The centroid method is based on the principal components methodology, and takes into account the genotype x environment interaction.Its main advantages, when compared with others methods, are the recommendation of genotypes and their classification into four ideotypes of maximum or minimum adaptability in response to the data set, and the use of the estimated probability as a classification criterion, avoiding duplicity of interpretation, and allowing the analysis of a large number of genotypes simultaneously (Rocha et al. 2005).This method has been recently modified by Nascimento et al. (2009), who added three other ideotypes of average performance, providing a greater scope for the genetic characterization and ensuring greater biological sense to the method.
The recommendation of coffee cultivars for organic systems has presented varieties adapted to different environments, according to the methodology proposed by Eberhart and Russell (Moura et al. 2014).However, this method has some limitations, since they group genotypes in a few classes of adaptability and stability, which requires the use of more modern methods that allow greater flexibility in the classification and selection of genotypes, such as the modified centroid method.
Therefore, the aim of this study was to identify promising coffee genotypes for organic systems, using adaptability and stability analysis with the modified centroid method in different environmental conditions.

MATERIAL AND METHODS
The experiments were arranged in a randomized complete block design, with 30 genotypes (cultivars and lines) and three replications, in the municipalities of Araponga, Espera Feliz and Tombos.The plots consisted of 10 plants, spaced 0.5 m within rows and 4.0 m between rows for short cultivars, and 0.8 m within rows and 4.0 m between rows for tall cultivars.The experimental sites were selected based on the tradition of family farming and organic coffee production, and on different soil and climatic conditions.The municipality of Araponga (lat 20° 40´ S, long 42° 31´ W and alt 1040 m asl), presents mesothermal humid subtropical climate, with minimum and maximum annual average temperature of 14.8 and 26 ºC, respectively; the soils is classified as Oxisol, A moderate, with clayey texture and high potential acidity.Espera Feliz (lat 20° 39´ S, long 41° 54´ W and alt 772 m asl), presents tropical climate and average annual temperatures ranging from 12.8 to 25.3 ºC; the soil is classified as Oxisol, A moderate, with clayey texture.Finally, Tombos (lat 20° 54´ S, long 42° 01´ W and alt 620 m asl), presents warm tropical climate, with seven months of drought, minimum and maximum annual average temperature of 12.6 and 30.8 ºC, respectively; the soil is classified as Paleudult-Yellow soil, A weak, with very clayey texture.
Liming fertilization at planting and top dressing of the experimental areas were based on both soil analysis and on the Lime and Fertilizer Recommendations for coffee crop in Minas Gerais (Ribeiro et al. 1999), using dolomitic limestone, cattle manure, thermophosphate, and double of potassium and magnesium sulfate, which are allowed for organic farming.
Throughout the experiment, soil liming, topdressings, and foliar fertilizations were carried out according to the crop requirement.Nitrogen sources included castor cake, complemented with green manures, such as the legumes Crotalaria juncea and Arachis pintoi, grown between the coffee rows, and cut at the beginning of flowering.The other nutrients were provided by the same sources used at planting.Foliar fertilization was carried out with the Supermagro biofertilizer.Weeds management included hoeing and regular mowing, and residues were used as mulch.
Data were examined by the joint analysis of variance, based on the plot means, and the sources of variation were analyzed as block/environments, genotypes, environments and genotype x environment interaction.
Adaptability was analyzed by the centroid method (Rocha et al. 2005), modified by Nascimento et al. (2009).The centroid method compares the Cartesian distances between the genotypes and the seven ideotypes (control genotypes).The control genotypes were established based on the experimental data, in order to represent genotypes of maximum general adaptability (ideotype I), maximum specific adaptability to favorable environments (ideotype II), maximum specific adaptability to unfavorable environments (ideotype III), minimum adaptability (ideotype IV), mean general adaptability (ideotype V), meanspecific adaptability to favorable environment (ideotype VI), and mean specific adaptability to unfavorable environment (ideotype VII).To use this method, the environments were classified into favorable and unfavorable, according to the environmental index proposed by Finlay and Wilkinson (1963): in which Y ij = mean of genotype iin the environment j; Y...= total observations; a = number of environments; and g = number of genotypes.After the formation of the environments and the determination of the reference points (ideotypes), the Cartesian distances between the genotypes and the seven ideotypes were compared.
A measure of spatial probability was calculated using the inverse of the distance between one treatment and the seven ideotypes: , in which: P d(i,k) is the probability of having stability similar to the k-th centroid, and d ik is the distance of the i-th genotype to the k-th centroid in the plane formed by the principal component analysis.All statistical analyses were performed using the software Genes (Cruz 2013).

RESULTS AND DISCUSSION
The joint analysis of variance for yield showed significant effects between genotypes, environments, and between the genotype x environment interaction, by the F test at 1% probability (Table 1).The significant genotype x environment interaction shows that yield was influenced by both the genotype and the crop environment, this is a basic premise for adaptability analysis and phenotypic stability of genetic material.
Espera Feliz was the only municipality with favorable environment for yield, represented by the positive index in Table 2, with a mean of 38.83 bags ha -1 of processed coffee.Possibly, soil and climatic conditions of this municipality were favorable to the growth and development of genotypes in organic farming.On the other hand, the lowest mean yields were recorded in the municipalities of Araponga (25.62 bags ha -1 of processed coffee) and Tombos (20.55 bags ha -1 of processed coffee), which had negative environmental indices (Table 2) and were classified as unfavorable environments.In Tombos, this mean yield may be due to the high water deficits (116.7mm) in the region (Calderano Filho et al. 2014), higher temperatures and less favorable physical characteristics of the soil; conversely, the municipality of Araponga has low temperatures and high soil potential acidity.
In the adaptability and stability analysis, the first two principal components accounted for over 86% of the total variation (Table 3), which was above the values for alfalfa by the methods of modified centroid (Nascimento et al. 2009) and multiple centroid (Nascimento et al. 2015), generating a two-dimensional plot of genotype dispersion (Figure 1).distribution of genotypes was observed for yield; however, some points (genotypes) were very close to four of the seven proposed centroids, allowing the classification of genotypes as maximum general adaptability (ideotype I), minimum adaptability (ideotype IV), mean general adaptability (ideotype V), and mean specific adaptability to favorable environments (ideotype VI).These results allow the breeder to recommend genotypes that are widely adapted to a number of environments or to a specific environment.Nevertheless, a large number of points (genotypes) dispersed in the central region of the graphic was noticed, which makes genotypes grouping difficult (Figure 1).A similar trend has been observed by several authors (Rocha et al. 2005, Barros et al. 2010, Pelúzio et al. 2010, Amorin et al. 2011, Barros et al. 2012, Pereira et al. 2012).Thus, the estimate of the probability associated    with genotype classification provides amore reliable recommendation, according to the degree of adaptability to the different environments studied (Rocha et al. 2005), as shown in Table 4.
Most genotypes presented mean general adaptability (ideotype V) with mean yield of 26.84 bags ha -1 of processed coffee; however, Obatã IAC 1669-20, Rubi MG 1192, and Catucaí 785/15 presented the highest probabilities (above 40%) of belonging to this group (Table 4).Moura et al. (2013) reported variability among genotypes belonging to this group for rust resistance in organic farming, which should be considered when recommending these genotypes, since the use of agrochemicals to control diseases is not allowed in organic agriculture.Among the genotypes that make up this group, Obatã IAC 1669-20, Tupi, and IAPAR 59 stood out for presenting higher yields in the conventional systems; the first (Obatã IAC 1669-20) was classified as tolerant genotype (Mendonça et al. 2016), and the last two (Tupi and IAPAR 59) were classified as coffee with resistance to rust (Shigueoka et al. 2014).
The genotypes IBC Palma 1, Catucaí Amarelo 24/137, Sabiá 708, and H 518had maximum general adaptability, and stood out from the others (ideotype I).These genetic materials were highly productive (up to 40 bags ha -1 of processed coffee), regardless of the assessed environment, and presented high potential for organic farming.All genotypes belonging to this group are genetically resistant to rust and are also promising when evaluated by Moura et al. (2014), using the method of Eberhart and Russel, which can be attributed to the association between this method and the method used in this study, as reported by Nascimento et al. (2013).

Figure 1 .
Figure1.Scores scatter plot of the first two principal components of the yield analysis (bags ha -1 of processed coffee) of 30 genotypes in six organic farming environments in the Zona da Mata region of Minas Gerais.The seven points numbered with Roman numerals represent the ideotypes: I -maximum general adaptability; II -maximum specific adaptability to favorable environment; III -maximum specific adaptability to unfavorable environments; IV -Minimum adaptability; V -mean general adaptability; VI -mean specific adaptability to favorable environment, and VII -mean specific adaptability to unfavorable environments.PC1, principal component 1; PC2, principle component 2.

Table 1 .
Summary of the analysis of variance of yield (bags ha -1 of processed coffee) of 30 coffee genotypes in six organic farming environments in the Zona da Mata region of Minas Gerais

Table 2 .
Estimates of means yield (bags ha -1 of processed coffee) of coffee genotypes and environmental index (Ij) for the six organic farming environments in the Zona da Mata region of Minas Gerais, according to the modified centroid method

Table 3 .
Estimates of the eigenvalues and cumulative percentage of variance explained by the principal components

Table 4 .
Estimates of means yield (bags ha -1 of processed coffee), classification of coffee genotypes into one of the seven groups proposed by the modified centroid method, and the probability associated with the classification of each genotype