View article

11 ing standard best agronomic management practices. Soil sampling began at the start of the trials, before application of fertilizers and amendments. Soil samples were analyzed for major soil characteristics including organic C, total N, available P, S, B, Mn, Cu, Zn, K, Ca, Mg, Na, Fe, exchangeable Al, and pH. Crops were harvested at maturity in a net plot of 6.75 m2 and grain yield was expressed on dry weight basis (12.5% moisture content). Cluster analysis was conducted using K-Means clustering on the differences between the grain yield from a given treatment and the control treatment, to identify various classes of nutrient response patterns. A multinomial logit regression model was developed and used to identify the possible soil factors influencing the identified response clusters. Four main clusters were identified as appropriate for categorizing observed nutrient responses. These clusters explained 60% of the variation in the yield data. Yields from the various treatments in each cluster were plotted (Figure 1), and the clusters were interpreted as followed: Cluster 1: Fields where maize was not responsive to any nutrient application or soil amendments. The cluster was further disaggregated according to fertile soils (Cluster 1b in Figure 1 referred to as fertile non-responsive fields) with high yields (attainable yield level between 4 to 5 t/ha) and infertile fields with low yields (Cluster 1a referred to poor non-responsive fields, attainable yield level remains below 2 t/ha) and have major limitations that need to be addressed before any nutrients or amendments can have any significant effect. 25% of the fields considered in this study were in this …