MODELOS MATEMÁTICOS E COMPUTACIONAIS PARA AVALIAÇÃO DO ESTADO NUTRICIONAL DA SOJA POR MEIO DE DIAGNOSE EM FOLHAS E GRÃOS

Abstract

The models based on the method of Chance Mathematics (ChM), Diagnosis and Recommendation Integrated System (DRIS) and Compositional Nutrient Diagnosis (CND) have been used to assess the nutritional status of plants through foliar diagnosis. However, some factors may interfere on the leaves mineral composition, as the variety of species or cultivar, plant age, soil type and management, climate conditions, and the attack of pests and diseases. We hypothesized that the assessment of soybean nutritional status can be performed through diagnosis both on leaves and grains. The objectives were (i) to evaluate the efficiency of the ChM, DRIS, and CND models in assessing the soybean nutritional status through diagnosis in leaves and grains, and (ii) using the computational methods Support Vector Machine (SVM) and Artificial Neural Networks (ANN) as an alternative to evaluating the soybean nutritional status. The study was carried out from a database consisting of 212 leaf samples and 216 grain samples of soybean collected in 2013–2014 from five field experiments under no-till systems installed in the Central-South region of Paraná State, Brazil. The leaf samples were collected at flowering period and the grain samples were collected after harvest. The levels of N, P, K, Ca, Mg, S, Cu, Fe, Mn, and Zn in leaves and grains were determined, and grain yields were evaluated. The ChM, DRIS, and CND methods were satisfactory when applied to the nutrient content in leaves and grains. However, the DRIS model showed greater consistency, especially when applied to the nutrient content in grains. Computational methods SVM and ANN showed values close to those extracted by DRIS and CND. The results suggest that the DRIS method could be used to assess the soybean nutritional status through diagnosis in grains, and the computational methods SVM and ANN would be an alternative to DRIS and CND models in evaluating the soybean nutritional status through diagnosis in leaves and grains.