New diagrammatic scale and rate progress analysis of coffee rust

Abstract

Brazil is the world's largest producer and exporter of coffee (Coffeea sp.). Minas Gerais is the central producing state and Arabica coffee production represents more than 70% of the state production. Coffee production can be harmed through losses due to several factors, but the main one is the presence of diseases. Coffee rust (Hemileia vastatrix) is the main disease causing production losses in both arabica and conilon coffee. The progress of the disease must be monitored to choose the best management and time of fungicides application. One of the methods used to practically evaluate the quantification of diseases in the field is the use of diagrammatic scales. Thus, the first experiment aimed to develop a new diagrammatic scale for arabica coffee with color photographs and seven disease levels, based on three other existing scales. First, a new diagrammatic scale was created, from 221 leaves collected at random and later the distribution of the frequency of the disease was obtained. Subsequently, three evaluations were performed with 10 different evaluators for the validation of the new scale and comparison with the existing diagrammatic scales. Validation was performed using linear regression and Lin concordant correlation analysis performed in the R software. The evaluators using the proposed scale improved the accuracy, precision and reproducibility of the evaluations and reduced the residues distribution. The aim of the second experiment was to evaluate the progress rate of rust in arabica coffee with different fungicides and fungicide mixtures. The analyzed data were obtained between 2014 and 2021, from 23 experiments of Arabica coffee crops, cultivar Catuaí Vermelho IAC 99, susceptible to coffee leaf rust. The evaluations were carried out in crops aged between five and eight years, with a spacing of 3.8 x 0.6 m between plants and carried out according to the technical recommendations for the culture. The experimental design was in randomized blocks, with four replications and the experimental plot consisted of 10 plants. Rust incidence, disease progress rate and linearization rate data were obtained, then linear and nonlinear rate models were fitted for all treatment data obtained using R software. The best adjusted model was chosen from the criteria of highest coefficient of determination (R2) and lowest Information Criterion of Akaike (AIC). The exponential nonlinear integral model was the best fit to describe the rate progress of rust as a function of time.