Creating Scalable, Quantitative Smallholder Farmer Data for Model Training through Conversational Farm Record Keeping

Abstract:

Agriculture development strategies will require new, better, and more cost-effective tools and methodologies for data collection and analysis of agriculture statistics to drive investment and decision making. Optical and radar sensors, combined with analytical models, can provide an accurate picture of crop health and deliver detailed recommendations for better crop management. Annually updated, representative, comparable field observations are needed to transform satellite observations into cropped area and yield maps that can be used for decision making. However, to capture the variability of agricultural management across diverse landscapes, strategies require sustained and expensive household surveys across large areas, sampling small to large field sizes, different crops and seeds, fertilizer use, and agroforestry and intercropping. This talk focuses on a project, recently funded by the Bill and Melinda Gates Foundation, that combines multiple mobile phone applications to gather high quality field data and digital boundaries and deliver value to farmers at a radically lower cost per datapoint than is possible with standard household survey techniques. Using innovative mobile chatbot technologies we can train a large base of smartphone owning farmers to enter ground data including field boundaries, crop type, crop variety and previous year production data in exchange for high quality farm record display and remote management of farms via digital tools. Geospatial modeling and analytical tools will then turn this ground data into high quality, crop management recommendations which will be delivered direct to the farmer and help us transform maize cropped area maps into yield estimate maps. This solution capitalizes on the rapidly growing base of smartphone owning farmers. We will present a pilot implemented in Kenya which will use a conversational interface (chatbot) to train farmers on how to collect data and interpret digital farming recommendations, with a focus on a first class user experience to drive engagement.