Improving Food Security and Crop Yield in Senegal Through Machine Learning

Background

Senegal faces significant challenges in agriculture due to climate change, including recurrent floods and droughts, threatening food security. With populations largely dependent on agriculture, maximizing crop yields is critical. Leveraging AI and data-driven technologies, this project aimed to support farmers by optimizing resource usage, such as water and fertilizers, and understanding crop losses. The initiative focused on providing precise recommendations using soil data, weather patterns, and satellite imagery to improve agricultural outcomes.

Objective

The project’s main goal was to leverage machine learning to:

• Improve crop yield prediction in Senegal.
• Help farmers identify when and where to add water or fertilizer.
• Support data-driven agricultural investments.
• Strengthen the understanding of crop losses and promote a sustainable agricultural system.

Approach

The team adopted a structured and data-intensive methodology:

1. Data Collection:
   • Remote sensing data from Google Earth Engine (GEE).
   • Ground truth yield data for maize, rice, and millet provided by IPAR.
   • Additional datasets from ANSD, DAPSA, CSE, ANACIM, and CSA.
2. Data Engineering and Pre-Processing:
   • Satellite imagery datasets were cleaned and processed.
   • Advanced pre-processing techniques prepared data for deep neural networks.
3. Machine Learning Techniques:
   • Convolutional Neural Networks (CNNs) were used for a transfer learning process on satellite imagery.
   • A Deep Learning model was implemented to predict crop yields accurately.
4. Application Development:
   • An open-source application was built to identify crops and estimate yields for specific areas using satellite imagery.

Results and Impact

The project achieved significant milestones:

• Created a deep learning model for accurate crop yield prediction.
• Enabled data-driven decisions for resource allocation, reducing losses and increasing revenue for farmers.
• Promoted a sustainable agricultural system in Senegal, addressing food security amidst climate challenges.
• Facilitated insights on where to invest and how to optimize agricultural inputs like water and fertilizers.

The application developed provided farmers and stakeholders with actionable insights to ensure better planning and productivity.

A demo of the application is attached below:

Future Implications

The project’s findings can influence future agricultural policies, emphasizing the adoption of AI for climate-resilient farming. Further research could extend these methodologies to other regions facing similar challenges, scaling the impact on global food security. These innovations pave the way for a more sustainable and technology-driven approach to agriculture.