Longitudinal And genotaype By Environment Interaction Analysis Of Arabica Coffee Bean Yield In South West Ethiopia: Application Of Linear Mixed Model

Abstract

Background: Arabica coffee (Coffea arabica L.) is the fine flavored, aromatic type makes up 60- 65% of total production and usually fetches the highest prices. Arabica Coffee is the most important and backbone of Ethiopian economy, which accounts for an average 60% of export earnings. Coffee is a perennial crop which can be harvested multiple times of years, and it is known to be affected with a characteristic biennial, which is more pronounced in the species Arabica coffee. The immediate objective of this study was to analyze Arabica coffee bean yield longitudinally by using Linear Mixed Model (LMM), and to assess its Genotype by Environment interaction (GEI). Coffee Bean Yield (CBY), Coffee Yield, and Yield are used interchangeably in this document. Methods: The data for this study came from coffee variety field trials conducted by Jimma Agricultural Research Center (JARC) over several years. The trial was conducted in south west Ethiopia across coffee growing areas (Jimma, Agaro, and Metu). The experimental design of the trial was RCBD with 4 replications and 17 Arabica coffee genotypes. A complete CBY data set of these coffee growing areas which had been collected during 2005-2011 was considered in this study. Exploratory Data Analysis (EDA) and LMM were employed for longitudinal analysis, whereas combined ANOVA and AMMI model were used for GEI analysis. All analyses were done with the help of R statistical package. Results: The LMM results revealed that the heterogeneous variance function (varIdent(t)) and autoregressive order three (AR3) were, respectively, found to give better fit to the variance and correlation structure among measurements of CBY. Biennial interacts significantly with location and genotype. The estimated variance of random effect of block associated with intercept and biennial were (b0j) = (221.81)2 and (b3j) = 145.242, respectively. The result also showed significant location by linear and quadratic time effect interactions. Estimates of quadratic time effects for Jimma, Agaro, and Mutu were, respectively, -151.51, -66.05, and -4, whereas estimates of linear time effects for these locations were 158.92, 158.92, and 31.08, respectively. The combined analysis of variance revealed that the genotype, environment, and GEI effects are highly significant (Pvalues<0.001). GEI accounted for 16.2% of the total sum of squares and was about 2 times larger than that of genotypes. The AMMI procedure revealed that AMMI-5 was the best truncated AMMI model that can sufficiently explain the information contained in GEI. The first three interaction principal components (IPC1, IPC2 and IPC3) retained by Gollob’s F-test for graphical display accounted for 64.2% of GEI. Conclusion: The measurements of CBY that are obtained from Arabica coffee tree over time induce an autocorrelation which is known as serial correlation. There is initially an increasing and gradually a decreasing trend in Arabica CBY over time years with linear rate of growth. There is also a differential response of genotypes and environments in the presence and absence of biennially. The major factor that influence yield performance of Arabica coffee in Ethiopia is the environment, and among 17 Arabica coffee genotypes, G1, G2, G3, G7, G8, G9 and G12 have the best performance with G1, G2, G3, G8 and G12 being relatively stable across the test environments. It was recommended to use information from longitudinal and GEI analysis to investigate the effect of time and biennial and the association between genotype and environment in Arabica CBY.