Comparison of GGE biplot and AMMI analysis of multi-environment trial ( MET ) data to assess adaptability and stability of rice genotypes

Genotype × Environment (G×E) interaction and stability performance were investigated on paddy yield of eighteen rice genotypes and twelve locations using two well renowned statistical models; genotype main effect and G×E Biplot analysis (GGE) and additive main effects and multiplicative interaction (AMMI) analysis. The aim of this study was to elucidate the performance of some advance rice lines/genotypes at multiple locations in multi environment trials (METs) using GGE biplot and AMMI analyses. The results of GGE biplot and AMMI analyses performed over the data of paddy yield at multiple locations of two years 2014 and 2015 indicated that G×E interaction plays a crucial role in determining the performance of genetic material in METs. The results declare that GGE and AMMI not only provide easy and affective evaluation of genotypes into environment interactions in a number of locations but also a comprehensive understanding of the variability of the target locations. AMMI analyses for data of both years indicated that RRI 7 was the highest priority selected genotype for six locations, NIAB 1175 for four and RRI 3 for three locations. Dhokri and Kala Shah Kaku were the highest yielding, while Faisalabad and Dhokri were the most stable environments in 2014. Likewise, Faisalabad and PARC Islamabad were the highest yielding as well as most stable environments in 2015. Basmati 515 and PS 2 were the most favorable genotypes in 2014 and 2015, respectively for their high paddy yield and stability at all locations. The results further suggested that both models were useful and presented similar interpretations about MET data.


INTRODUCTION
Rice is being used as staple food by more than three billion of world population which represents 50 to 80% of their daily calorie intake (Khush, 2005;Amirjani, 2011).This population will increase to over 4.6 billion by 2050 (Honarnejad et al., 2000) which will demand more than 50% of rice needs to be produced to cope with the growing population (Ashikari et al., 2005;Srividya et al., 2010).Rice is the second most important staple food *Corresponding author.E-mail: z.haider.breeder@gmail.com.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License crop in Pakistan not only in local consumption but also in view of large exports (Anonymous, 2006).In Pakistan, rice was grown on approximately 2.89 million hectares with a total production of 7.01 million tonnes and earned a foreign exchange of worth US$ 1.53 billion.It accounts for 3.2% in value added in agriculture and 0.7% of Gross Domestic Product (Pakistan Economic Survey, 2015).
Rice production in Pakistan is confined in four, more or less distinct agro-ecological zones.Each zone represents diverse edaphic, hydrological and climatic conditions.Genotypes with wide adaptability, which can perform consistently well over a range of environments is required.Crop stability, is the ability of a crop to exhibit minimum interaction with both predictable and unpredictable environments (Qayyum et al., 2000).Rice yield in Pakistan is very low as compared to other rice growing countries of the world.With the passage of time, due to drastic climatic changes, there are developing vast differences in agro-climatic conditions among different locations of Pakistan.These phenomena make new developing rice varieties more unstable when grown under diverse environmental conditions and results in poor yield (Duncan, 1955).
Genotype × Environment (G × E) interaction is important to plant breeder for examining yield stability of genotypes (Davis and Rutgar, 1976).Linnemann et al. (1995) stated that it is important to understand crop development in relation to biophysical conditions and changes in season when selecting well-adapted genotypes and correct planting date.Varieties that show low G × E interaction and have high stable yields are desirable for crop breeders and farmers, because due to their genetic composition their yields are larger that indicates that the environments have less effect on the performance of genotypes.Therefore, evolution of rice varieties that have high yield and stability in performance over a wide range of environments will remain an important criterion in rice breeding (Tai, 1971).
To achieve this goal, innovative statistical techniques like Genotype main effect and G × E biplot analysis (GGE) and additive main effects and multiplicative interaction (AMMI) analysis will be needed to compare the lines and testing environments in order to separate rice genotypes with high adaption capability to diverse environments for accomplishing specific socio-economic and agricultural needs (Prasad and Singh, 1990;McLaren and Chaudhary, 1994;Kueneman, 2006;Khush et al., 1979).The objective of the present study is to present the idea how to interpret results of GGE biplot and AMMI analyses to study G × E interactions and stability in performance of advance fine rice genotypes grown at different locations of Pakistan.
One month old seedlings were transplanted at plant to plant and row to row distance of 20 cm.The experiments were sown in randomized complete block design (RCBD) in a triplicate fashion at each site.To avoid the chance of mortality, two to three seedlings per hill were placed.The fertilizer was applied at the rate of 120 N + 60 P2O5 + 0 K2O kg/ha, whereas in aromatic group at the rate of 80 N + 40 P2O5 + 0 K2O kg/ha.During crop growth stages, standard agronomic practices were followed.The quality characters such as size (length, length/width, length/width × thickness) were carried out according (Kush et al., 1979;Sagar et al., 1988;Kawai and Amano, 1991;Singh, 1997).Paddy yield was recorded after the harvest of the crop.At maturity, yield was recorded and subjected to analysis of variance (ANOVA) as suggested by Steel and Torrie (1980) to determine the significance of genotypes, environments and genotype environment interactions.
GGE biplot is being rigorously used for analyzing multienvironment trail (MET) data, which provides a working guide for breeders on biplot analysis and interpretation of results (Domitruk et al., 2001).Earlier, many studies have been interpreted using simple statistical techniques.Therefore, advance statistical techniques were used for this MET data in order to interpret three important results necessarily required in multi-location trials such as identifying high yielding and most stable genotype, best test environment and best performing genotype in specific environment.These questions become very hard in such METs that involve a large number of genotypes.

GGE biplot analysis
Biplots graphs (Figure 1a and b) are plotted where aspects of all the Basmati rice genotypes and environments are on the same axes so that the interrelationship can be easily depicted.Therefore, biplot 1a explains 65.64% of the total variations while Figure 1b explains 77.61% of the total variations.In both biplots, all the genotype and environment vectors are drawn showing the specific interactions between genotypes and environments.
Based on Figure 1a and b, it is possible to assess both mean yields and stability performance through biplot environments with shorter/acute angle between their vectors are more correlated while those with larger/obtuse angle are negatively correlated.Shorter the angle, higher the value of positive correlation between two locations.Presence of close associations among the environments suggests that the same information about the behavior of studied genotypes was obtained from other locations and performance of all the genotypes was same at these locations.The location vectors are lines that connect the biplot origin and the markers of test locations and the angle between them is related to the correlation coefficient.Therefore, the results suggest to drop one of these locations in order to secure resources and any other location may be added having different environmental conditions.For example, in the year 2014, locations Dhokri, Faisalabad, D.I. Khan and NARC Islamabad have shorter angles and thus show higher correlation among themselves.Therefore, it will be wiser to consider any one of these locations next year that will be representative of all these locations in order to secure inputs.GGE biplot can also be used to interpret genotypes with similar response to the target environments.Two genotypes can be easily compared by joining them with straight lines passing through them and biplot origin.Based on Figure 1a and b, it can be visualized that genotypes that have shorter angle or are in close proximities on biplots indicate their similar response to the environments.For example, Rachna-1080 and PK 8431-1-2-1-2-4 showed similar response to environments tested in 2014, while response of NIAB-1175 was very different in comparison with these two.NIAB-1175 and NIAB-1114 have shorter or acute angle between them indicating that these two responded similarly and the differences between them was proportional in all target locations.Genotypes located at more distance from center of biplot are less stable and near to origin genotypes are more stable (Jones, 1926).RRI-7 are more stable than NIAB-1114 in 2015.
An average tester coordinate (ATC) horizontal and vertical axes is drawn such that it passes through the biplot origin and the average location and show the stability and mean yield of each genotype (Yan and Nicholas, 2006).Genotypes and environments on the same parallel line or ordinate have similar yield and a genotype or environment on the right side of the midpoint of ATC vertical axis has higher yield than those of left hand side.Figure 2a and b shows that PK BB-4 was the least yielding genotype among all the studied locations as the angle of its vector is obtuse with respect to all the environment vectors, in both years.While the check variety Basmati 515 was above average in 2014 and PS 2 check variety in 2015, in all the locations as indicated by the longest vector and acute angle with respect to all the other location vectors.
The smaller the absolute length of projection of a genotype, the more stable it is (Fehr, 1987).The biplots in Figure 2a and b further showed that NIAB 1147, Rachna 1080 and PK 8431-1-2-1-2-4 were the least stable genotypes and showed unstable yield performance at all the locations in 2014 as indicated by longer vectors/distance from ATC horizontal.During 2015, NIAB 1175, NIAB 1114 and RRI 3 were the least stable genotypes, as indicated by the longer distances from ATC horizontal.Super Basmati and RRI 7 were the most stable genotypes in 2014, while RRI 7, PK BB-4 and then PS 2 were the most stable in year 2015 as indicated by shorter distances from ATC horizontal axis (Figure 2a  and b).
According to the biplot shown in Figure 3a and b, the corner genotypes that are the most responsive ones, can be visually determined.These corner genotypes were PK 8685-5-1-1-1-1, Rachna

AMMI analysis
The mean paddy yield (t/ha) value (

Figure 1 .
Figure 1.Vector view of GGE biplot for relationships among locations/environments during 2014 and 2015.

Figure 2 .
Figure 2. GGE biplot of ideal genotype and comparison of the genotypes with the ideal genotype (center of the concentric circles) for year 2014 and 2015.

Figure 3 .
Figure 3. GGE biplot identification of winning genotypes and their related mega-environments for year 2014 and 2015.

Table 1 .
AMMI analysis showing genotype means and scores for years 2014 and 2015.

Table 1a
in each environment can be visualized.Table3aand b shows the ranking of the genotypes according to their performances in respective environments for the year 2014 and 2015, respectively.The candidate variety RRI-7 performed best in two locations (Dhokri and Faisalabad) in terms of yield and four locations (Thatta, Swat, Quetta and DI Khan) in 2014

Table 2 .
AMMI analysis showing environment (Location) means and scores for years 2014 and 2015.

Table 3 .
First four AMMI selections per environment for years 2014 and 2015.