Development of an empirical tomato crop disease model: a case study on gray leaf spot

This paper deals with the development and evaluation of a disease model to forecast the risk and incidence associated with tomato gray leaf spot ( Ascochyta lycopersici Brun ) - one of the reasons for significant tomato yield loss in the Mediterranean area and in other countries. It comprises a leaf wetness model, a disease occurrence warning model and a disease incidence model. The methodology followed was based on studying plant disease epidemiology to clearly understand how the disease progresses, analyzing input parameters used in the published literature and selecting the most suitable methods for calibrating the model thresholds and evaluating its performance. The developed sub-models were evaluated according to the following performance indexes: (1) the area under the receiver operating characteristic curve for choosing the threshold of the leaf wetness model using three methods (a classification tree, support vector machines and the Naive Bayes method); (2) the root mean square error; and (3) the mean absolute error, both for evaluating the curve fitting based on disease incidence models (using Power, Exponential, Polynomial, Gaussian, Logistic and Gompertz approximations). The obtained results provided a calibrated relative humidity threshold of 84.5% from the classification tree method, while the best fitting function was the Logistic equation providing a root mean square error of 3.17 and a mean absolute error of 2.54; the evaluation results of two plant seasons in 2017 and 2018 proved that the Logistic equation can simulate gray leaf spot incidence good, with an R 2 of 0.97 and 0.92, and a RMSE of 2.7 and 1.8. This work contributes to tomato gray leaf spot management by providing a basis for decision support to help growers make timely and precise decisions and thus avoid major economic losses.