Multivariate Diagnosis of Nutrient Imbalance in Potato Crops

Abstract

AbstractCompositional Nutrient Diagnosis (CND) involves a correction procedure for two constraints inherent to foliar analytical data, i.e., the bounded‐sum constraint to 100% of the components and the redundant information among interactive nutrients, which are addressed only partially by the Diagnosis and Recommendation Integrated System (DRIS) and not at all by the Critical Value Approach (CVA). The objective of this study was to extract the correlation structure among nutrients using principal component analysis, to test for correlation between DRIS and CND nutrient indices, and to diagnose independent samples. Yield and foliar analyses (N, P, K, Ca, and Mg) of late potato were collected in the St‐Lawrence Lowlands for 3 yr. The high‐yielding subpopulation (182 observations) produced >40 Mg ha−1. Three varimax‐rotated principal components (PC) explaining 93.5% of total variation in the high‐yielding subpopulation were designated as (N−P−Ca+), (P−Mg+) and (K+). A euclidian distance d computed from the PCs provided a nutrient imbalance index for the crop. DRIS and CND indices were highly correlated to each other (r = 0.956‐0.989) using 36 independent early potato samples. The CVA diagnosed correctly 66.7% of the cases. The CND classified the crops in two groups assuming χ2 distribution for d2 values (P ≤ 0.10). Nitrogen was the most limiting nutrient in the high‐yielding subgroup and K was extremely limiting in the low‐yielding subgroup. Defining critical CND nutrient indices empirically, CND produced correct diagnoses 87.5% of the time.