The main objective of this project is to develop reliable and reusable ML-driven crop yield prediction models using historical meteorological data, satellite data, and proximal sensor data to enhance crop yield predictions in Tanzania. Specifically, the project:

  • To develop a model that utilizes historical multi-source data to predict maize and sorghum yield at the district level using climate reanalysis data, satellite imagery data, weather data, soil data, and proximal IoT data to train machine learning models to predict district-level yield data.
  • To deploy a small-scale smart farming system using low-cost Internet of Agro Things (IoAT) sensors and interactive cloud-based big data analytics to monitor and evaluate crops’ performance in real-time.
  • To pilot a big data model to predict farm-level yield using low-cost agricultural Internet of Agro Things (IoT) sensor data by enhancing district-level resolution yield data to farm-level resolution yield data using Generative Adversarial Networks (GANs).
  • To conduct the economic feasibility of using agricultural IoT and big data for small-scale farm monitoring and yield prediction.
  • To formulate a data-driven policy brief on crop prediction using multi-source big data. The research will identify and reach potential farmers to use agricultural IoT.