Assessing concordance between two methods for sugarcane brown rust evaluation

Concordance evaluations between evaluation methods or evaluators are common in medical disciplines, but not in agricultural sciences. The objective of this study was to detect the concordance between two methods used to determine the percentage of leaf area affected by brown rust in sugarcane: visual estimation and image analysis using ImageJ® 1.43u software. For this, the percentage of area affected by pustules was determined according to each procedure during two trials. Correlation analysis, concordance correlation coefficient calculation and concordance test were performed. Pearson's correlation was not constant within the range of data evaluated, which represents a difficulty in detecting systematic differences. The concordance correlation coefficient was classified as poor, while the Bland-Altman concordance analysis allowed detecting a bias of -1.48% and -0.57% between the methods in both trials, which is considered acceptable. The study showed that both procedures agree, but the values measured visually tend to be higher than those obtained with the ImageJ® 1.43u software, mainly in cultivars with a larger leaf area affected by the disease. The procedure described in this work facilitates decision-making on the opportunity to substitute or use different methods for the evaluation of brown rust in sugarcane, which can also be applied to other pathosystems or disciplines.