Customized Nutrient Standards to Diagnose Nutrient Imbalance in Fertigated ‘Nanica’ Banana Groves

Banana (Musa spp.) is an important fruit production in Brazil, but crop productivity is still too low. The ‘Nanica’ cultivar and fertigation have been introduced, but more accurate guidelines are needed to support fertilization decisions at the orchard scale. This study aimed to develop customized nutrient standards for fertigated ‘Nanica’. A commercial ‘Nanica’ orchard provided 129 observations on yield and foliar nutrient concentrations from 2010 to 2017 in eight groves of 3.26 ha each. Plant density averaged 1479 plants ha−1. The diagnostic leaf was analyzed for 13 elements. Concentration values were transformed into centered log ratios (clr), weighted log ratios (wlr), and isometric log ratios (ilr) to account for nutrient interactions and normalize the data. Yield cutoff between low- and high yielders was set at 27 t ha−1 semester−1. The XGBoost classification models relating yield to tissue composition returned an area under curve averaging 0.715 for log ratio expressions. Nutrient standards were expressed as clr, wlr, and raw concentration means and standard deviations of performing specimens. The clr and wlr diagnoses of a low-yielding and imbalanced specimen against a benchmark specimen (Euclidean distance = 2.5) or the performing subpopulation (Mahalanobis distance = 37.6, p < 0.01) indicated Mn shortage and Na excess. Sufficiency concentration ranges may not agree with log ratio diagnoses, especially for Mn. The clr and wlr nutrient standards were site-specific, supporting precision farming. The concept developed in this paper is applicable to endogenous research conducted by stakeholders in orchards worldwide.