Exploratory Development of Predictive Model to Study the Rice Blast Disease Development at Different Growth Stages Using Machine Learning

Abstract

Rice blast disease caused by Pyricularia grisea Sacc. has become an emerging constraint in Basmati rice cultivation in Punjab for the recent years. A detailed field investigation was conducted during Kharif 2015 and 2016 to study the impacts of different meteorological elements on blast disease development and compute predictive models to predict the disease ahead of its appearance in the field at different growth stages. Correlation analysis showed that maximum air temperature and relative humidity were the key elements to govern the disease in the field among all other meteorological elements. Maximum air temperature around 34^oC and relative humidity above 60% were observed to be favorable for the disease spread in the field. Predictive models were developed for nursery stage (R²=0.71), tillering stage (R²=0.81), panicle stage (R²=0.99) and for whole growth period (R²=0.55) using R programming. A step-wise multilinear regression approach was adopted to identify the most appropriate predictive variables to formulate the model.