The objective of this work was to test the application of accuracy measures to determine the usefulness of nutritional diagnosis to compare different DRIS formulas, and to evaluate the use of these measures in the DRIS function modeling process. Three DRIS formulas were evaluated: bivariate relations (simplified); multivariate relations; and DRIS function modeling. Monitoring data from commercial mango orchards grown at São Francisco River Valley, Brazil, were used. Formula performances were evaluated based on the agreement of their diagnoses with the one provided by the sufficiency range method. Global accuracy, and the ones for insufficiency, equilibrium, excess, deficiency, sufficiency and nutritional toxicity were determined. The simplified and multivariate formulas showed similar results for insufficiency and deficiency accuracies, but they had a lower performance than the DRIS function modeling formula. This latter, for allowing of the adjustment of each DRIS function sensibility coefficient, showed to be promising in reducing the risks of false diagnosis for deficiency (as for not recommend the nutrient when it is necessary) and for insufficiency (as for recommend the nutrient when it is not necessary). However, the DRIS formula for function modeling does not show a satisfactory performance for other accuracy measures.
Mangifera indica; sufficiency ranges; foliar diagnosis; response potential
