Case Study: Crop Disease and Irrigation

A client needed to monitor crops for risk of disease and drought/poor irrigation on a national scale; the client needed weekly updates on the risks to crops and production.  The scale of the fields and the land overall meant it was not possible to check all of the fields every day manually. Even with drones, it would have been cost-prohibitive, costing tens of millions.  For this reason, we built a solution that looked at every 10m of land every week, across the entire country and used machine learning, based on billions of data points, to detect disease.

We use a wide variety of satellite data; in this case, the satellite used was  Sentinel-2. We gathered data every week for every 10m of the entire country for years of historical data. The data was also obtained at a high temporal frequency – every week.

We also built a mobile app that allowed the farmers to report on the disease in their field and used this to build a machine learning model – this data is known as “ground truth“.   By combing satellite data with weather data and ground truth, we were able to build a highly effective model disease prediction model.