Prediction of maize yields at scale using remote sensing, machine learning and geographically dispersed trials data

The ability to map actual and potential yields at scale would enable much more informed policy interventions and targeted investments. While methods exists, much empirical validation remains to be done for maize systems in sub-Saharan Africa.

Objectives

To evaluate the prediction of actual and potential maize yields from remotely sensed predictors (NDVI, etc.) coupled with ground control data from Ethiopia &/or South Asia (Nepal, Bangladesh, India).

Work expectations

Student would need to have remote sensing & machine learning skills. We would expect to produce a paper from this work. Student would work closely with CIMMYT scientists in the Nairobi office.

Activities

• Assemble data for analysis
• Implement and evaluate alternative yield prediction models
• Write up analysis