Abstract:
【Objective】The aim of this study was to analyze the relationship between agronomic trait of coffee(
Coffea arabica)and yield,and to provide theoretical references for breeding superior coffee varieties.【Method】Based on the observed data of national germplasm resources garden of Ruili coffee from 1999 to 2020,used correlation(r),gray correlation(GRA)and projected importance(VIP)methods,the linear regression equation was established by partial least squares regression(PLS)method to analyze the main agronomic traits and yield per plant of coffee.【Result】There was extremely significant correlation between agronomic traits and yield per plant(
P<0.01). Gray correlation analysis showed that there was a strong correlation between each agronomic trait and yield per plant,and the relational coefficient ranged from 0.750 to 0.769. VIP showed that all agronomic traits had good explanatory ability to yield per plant,in which number of primary branch and stem diameter had the highest explanatory ability to yield per plant,and VIP value were 1.290 and 1.140,respectively. The optimal PLS model identified six relevant traits(plant height,main stem diameter,number of primary branch,canopy diameter,length of the first primary branch and number of the first primary branch node). Path analysis showed that cumulative indirect effects of agronomic traits on yield per plant (1.472)was greater than the cumulative direct effects(0.471). The length of the first primary branch had the greatest indirect effect on yield per plant through other agronomic traits,and the value was 0.333. The cumulative indirect effect of other agronomic traits on yield per plant through length of the first primary branch was the greatest(0.576). In addition,there was also a greater cumulative indirect effect of other agronomic traits on yield per plant through the number of primary branch(0.562).【Conclu-sion】Based on the analysis,all the six traits in this study significantly influence coffee yield. Thus,those agronomic traits are the main observation objects in the breeding process. PLS linear regression equation established by using six agronomic traits has the highest accuracy,which can be used as an auxiliary tool to predict the yield per plant.