Enhancing Stress Prediction Models for Rice with Ambient and Field Data in Rabi 2019-20

Abstract

Every crop needs optimum environment conditions to grow and provide good yield. Similarly, pests and diseases for a given crop require conducive ambient conditions to start proliferating thereby interfering with the yield and increasing management costs for the crop. For farmers to timely respond to crop damage, precisely predicting the likelihood of pest and disease stress conditions for a given region of interest is paramount to take early action. Rice is major crop in India grown over Kharif and Rabi seasons. For a key state Andhra Pradesh and associated major stress conditions in Rabi 2019-20, we present localised prediction models that use ambient micro-climatic conditions and ground reported data to forecast trends for major pests and diseases such as Stem Borer, Leaf Folder, Leaf Blast, and Bacterial Blight. For effective localisation, images reported from the fields for these conditions are validated with AI based detection models before getting processed further. A major contribution of the work is to realise an integrated system that continuously adapts to pest and disease stress conditions on the ground and offers precise risk prediction advisory to the farmer ecosystem for effective management. The approach has been demonstrated to work well with in-season farm surveillance activities at scale.