Fuzzy Classification in Mapping the Nutritional Status of Coffea Canephora

• Published in: Communications in Soil Science and Plant Analysis Vol. 52; no. 19; pp. 2304 - 2317
• Main Authors: Lima, Julião Soares Souza, Barreto Soares, Cássia, Silva, Samuel de Assis, Fonseca, Abel Souza, Fraga Pajehu, Levi, Medauar, Caique Carvalho
• Format: Journal Article
• Language: English
• Published: Philadelphia Taylor & Francis 28.10.2021; Taylor & Francis Ltd
• Subjects: Classification; Coefficients; Confidence intervals; Correlation coefficient; Correlation coefficients; Distribution; Fuzzy control; index of agreement; Kriging interpolation; Leaf analysis; macronutrients; Nutritional status; Spatial distribution; standardization of units; Statistical methods
• ISSN: 0010-3624; 1532-2416; 1532-4133
• DOI: 10.1080/00103624.2021.1924187

Knowing the spatial distribution of nutritional status allows us to understand plants' metabolic requirements and identify zones for differentiated management. Thus, the objective of this work was to use the fuzzy classification to standardize the values of macronutrients (N, P, K, Ca, Mg, and S) to construct the map of the average spatial distribution of nutritional status for Coffea canephora. A sample mesh of 80 georeferenced points was constructed to collect the leaves. A fuzzy controller, the Mamdani method, was used as linguistic variables the ranges of nutritional sufficiency: low, adequate, and high and the rules of inference. Geostatistical analysis was used to define semivariograms and perform interpolation by kriging and cokriging, having as covariates the fuzzy indexes for each macronutrient. The percentage of agreement between the maps was determined by correlation coefficients, confidence indexes, and the RMSE. The estimated maps for the macronutrients constructed by cokriging compared with the observed maps constructed by kriging presented spatial correlation coefficients (r co ) from 0.81 to 0.97, concordance indexes from 0.84 to 0.97 and confidence from 0.68 to 0.91 and RMSE from 0.01 to 0.23, showing high percentage of agreement between the maps in the use of fuzzy indexes as covariate of cokriging.