A Geo-Spatial Application for Brown Plant Hopper Pest Risk Prediction Over Rice Growing Areas of India

Annually 10%–30% of yield losses in rice fields are reported in India due to damage from Brown Plant Hopper (BPH). It is highly imperative to develop an advanced digital solution for BPH infestation risk prediction so that framers can take effective measurements against the pest. We have developed a geo-spatial application for BPH risk prediction by integrating multi-sensor data: Sentinel – 1 satellite images, weather data and ground incidence reporting from Corteva Agriscience field team. A 10m crop land map was produced using Support Vector Machine model with cross-validation accuracy of 92.34% followed by 10m rice map with 89.22% cross-validation accuracy. These maps have set up a strong base for geo-spatial BPH risk prediction. Weather-based decision rule system was able to predict the BPH risk with 87.16% accuracy when validated with the ground incidence reporting data. It is observed that BPH infestation is highly sensitive to rainfall and wind speed. The location-specific rainfall pattern is very dynamic and unpredictable, therefore, it is crucial to have the timely information about BPH risk. Results obtained have proved the advantage of our model for large scale adoption. This model is particularly suitable for cloud prone regions where detecting rice growing areas is challenging with optical images. Our geo-spatial application will support both big and small holder farmers to plan effective mitigation strategies and prevent yield losses.