Multivariate Nutrient Diagnosis of the Carrot Crop

Abstract

Compositional nutrient diagnosis (CND) provides undistorted (linearized) variates amenable to principal component analysis (PCA) using a row-centered logratio transformation of foliar nutrient data. Our objectives were to carry PCA on raw or transformed nutrient data for carrot (Daucus carota L.) crops and to compare the critical value approach (CVA), diagnosis and recommendation integrated system (DRIS), and CND diagnoses using independent data. PCA conducted on percentage or log-transformed data produced similar multivariate structures difficult to interpret. PCA conducted on DRIS indexes and on row-centered logratios produced PCs (K-Ca+), (N+K-Ca-Mg+), and (P-Mg+) and PCs (K-Mg+), (N-Ca+), and (P-), respectively. Nutrient contrasts were easiest to interpret with CND and reflected either K-Mg antagonism or N dilution and Ca accumulation over time. CVA diagnosis of independent samples was generally not in line with DRIS or CND. DRIS and CND diagnostic indexes were highly correlated (r = 0.98 to 0.99). By summing bivariate DRIS functions, the DRIS index calculation procedure effectively row-centered the nutrient values for carrots. DRIS and CND index diagnosis indicated treatment-dependent Ca shortage. In contrast, CND PC diagnosis indicated overall stationary values for PC (N-Ca+) whatever treatment was applied. CND PC diagnosis is a multivariate (PCA) approach providing simplified computational effort and a theoretical basis for further improvements in foliar diagnosis.