Early within-season foliar diseases detection and yield prediction using Sentinel-2 satellite constellation imageries and machine learning technologies in biomass sorghum
Sorghum (S. bicolor (L) Moench) is grown for several purposes including biomass for producing energy and fodder, and grain for producing health-promoting foods . Sorghum is a C4 and drought resistant cereal crop with low input requirements, making it one of the most promising crops under the world’s tropics and higher latitudes. Sorghum grain is a staple food in Africa and Asia, it is used in North American food industry, and is an interesting raw material in Europe for the formulation of gluten-free foods, due to its high contents of phytochemicals expressing antioxidant activities, cholesterol-lowering properties and other health benefiting properties.
The importance of sorghum is greatly rising under climate change scenarios characterized by drier and hotter conditions. Crop monitoring, one of the leading activities in smart farming, can help cut production costs and more so under climate change.
In this study, Sentinel 2A and 2B-derived fAPAR biophysical data and NDVI index were used to monitor sorghum phenology, foliar diseases (anthracnose and bacterial stripe), and to predict aboveground biomass yields months before harvest using machine learning approaches (Gradient Boosting, R-CNN, etc.).
The results obtained in this work were preliminary yet encouraging. In the first run, we were able to predict biomass yields up to 6 months before harvest with MAE (%) < 19, while diseases were detected with accuracy up to 90%. These results were achieved at a Pilot level (project DataBio) and the technology showed industrial scale implementation potentials with tremendous benefits for the farmer and several other parties at interest.
Full session publication from September 18 on the web site “lifesciences.knect365.com”