Bayesian network based decision support system for insect pest management in tomato crop

Abstract

Tomato (Lycopersicon esculentum L.) is one of the most important vegetable crops grown in the northern plains of India. Timely availability of decision support to the farmers on whether and what management option is required is imperative for effective pest management. In the absence of knowledge and expertise, farmers are over dependent on pesticide dealers for support on decision-making in pest management in the country, which results in excessive, injudicious, and irrational use of chemicals for the pest control. This not only degrades the environment, but also affects human health. For decades, the pest economic threshold level (ETL) has been the basis of selecting appropriate pest management option, used in existing decision support systems (DSSs) of agriculture in the country. This process requires quantitative information about pest activity which needs to be scientifically observed in the farmers fields. However, a large section of the farming community is not able to scientifically obtain this kind of information rather can observe qualitative information consisting of uncertainties. Moreover, in current pest management, decision-making depends upon a large range of pest relevant agro-ecological information, beside pest activity. But there are no methods available to deal with uncertain pest relevant agro-ecological information provided by farmers for decision-making in management option. Fruit borer (H. armigera) and Leaf miner (Liriomyza trifolii) are the key insect-pests severely hampering tomato production in the region as revealed by the descriptive statistical analysis of weekly pest data records of 2012-13 to 2015-16. Significant activity of these pests has been found in the region i.e. mean borer damage was 2.65% damaged fruits/plant with variance 14.62 and leaf miner caused avg. 0.88 mines/5 leaves/plant with variance 1.40. As such, variety of agro-ecological factors affect the pest activity within the field but correlation analysis of data records and the knowledge elicited by domain experts