Development of weather-based forewarning model for tomato leaf curl infestation

Abstract

Tomato (Solanum lycopersicum L.) is commercially important crop grown worldwide in wide range of climatic conditions in field or under protected condition and it is highly accepted as fresh salad, cooked and processed food (Peralta et al., 2006). Among the different biotic stresses reported on this crop, Tomato leaf curl virus (ToLCV), a gemini virus is the most important and destructive viral pathogen in many parts of India (Saikia and Muniyappa, 1989; Harrison et al., 1991) including West Bengal. The management of leaf curl disease in tomatoes relies mainly on the application of insecticides at frequent intervals. Excessive use of insecticides not only cause environmental hazards but also triggers the development of resistance in the insect pest and also enhances the production cost of the crop. A reduction in the number of pesticide applications could be achieved by applying pesticides only at times when weather conditions are favourable for disease development. The cause and effect relationship are generally studied using simple regression model or nonlinear regression models. But in situations where data are not in ratio scale, then regression may not be appropriate as it violates the assumption of normality and constant variance. A very well used recommendation is to apply data transformation and proceed with ordinary linear regression but there may be problem in interpretability. In such situations, beta regression model may be employed as it takes into consideration all the properties of the data (Maier, 2014; Cribari‐Neto and Zeileis, 2010; Hijazi and Jernigan, 2009; Ferrari and Cribari‐Neto, 2004; Kieschnick and McCullough, 2003). In Beta regression, it is assumed that the dependent variable is distributed with beta function and the mean of the beta distribution is dependent on a set of independent variables by means of linear predictor with a link function and unknown parameters. A dispersion parameter is also included in the function