Statistical assessment of maize leaf nutrient and yield relationships in rural agriculture in the Mbizana area, Eastern Cape province, South Africa

Abstract

A medium-term study was conducted under dry land to develop regression models to predict maize yield from leaf analyses for cost-effective fertiliser recommendations. Treatments consisted of lime, organic and inorganic fertiliser applications in a randomised complete block design with three replicates. Measurements consisted of leaf and topsoil sampling, harvesting, as well as chemical and statistical analyses. Linear regression analyses of pooled leaf macronutrient data over four growing seasons indicated highly significant relationships. Linear regressions of leaf macro- vs micronutrients generally revealed inverse relationships. Highly significant synergistic interactions were observed for leaf micronutrients. Cluster analysis of maize grain yields produced four distinct groups. Multiple regression models for the clusters explained 24–90% of the yield variation. Certain leaf macro- and micronutrients appeared in more than one cluster model. Multiple regression analyses on the yield clusters indicate that it would be worthwhile to cluster large yield data sets. The regression model for the entire data set contains the variables P, Zn, Cu and Fe, accounting for 68% of yield variation. These findings may contribute to a reduction in leaf analysis costs and enable cost-effective fertiliser recommendations. It is recommended that similar regression studies be performed on maize leaf and yield data from contrasting agro-ecological zones.