The inspiration for DEWSClim stemmed from observing the devastating effects of climate change on farmers in Nigeria. Unpredictable weather patterns, crop failures, and pest outbreaks were worsening their already precarious livelihoods. We recognized the need for an early warning system that could not only predict weather changes but also provide actionable insights tailored to individual farmers' needs. This project aims to help farmers make data-driven decisions that could significantly reduce crop losses and improve productivity.

Throughout this project, we gained a deeper understanding of how machine learning models can be applied to real-world problems, particularly in agriculture. From wrangling climate, soil, and crop data to building models that forecast future conditions, we learned the critical role accurate data plays in making predictions. Additionally, integrating the Gemini API showcased how APIs can simplify complex calculations, such as yield estimation and sustainability assessments, making it easier to deliver insights to end-users.

The project involved creating a mobile application, DEWSClim, which serves as a digital early warning system. The workflow included:

Data Collection: We gathered climate, soil, and production data relevant to Nigerian agroecological zones.

Model Development: Machine learning models were built to predict weather patterns, soil conditions, and crop health.

Gemini API Integration: This API was integrated to provide yield predictions and climate-smart practices.

Mobile App Development: We developed the app for both Android, ensuring farmers across the country could access it. The app features a user-friendly interface that displays real-time weather updates, risk assessments, and recommendations tailored to the user's location.

Building DEWSClim came with several challenges:

Data Quality: Acquiring clean, reliable data on climate and soil was difficult. We spent significant time cleaning and validating the data to ensure accurate predictions.

Localization: Ensuring predictions and recommendations were hyper-localized to farmers' individual needs posed a challenge. Our models had to account for the wide variety of agroecological zones in Nigeria.

Usability: Designing an intuitive interface that could easily be used by farmers, many of whom are not tech-savvy, was a critical challenge. We had to prioritize simplicity without sacrificing the depth of insights.

DEWSClim is still in its design phase, but we plan to expand it by incorporating more advanced forecasting techniques and gathering more granular data. Farmer feedback will play a crucial role in refining the system and making it more effective in mitigating climate-related agricultural challenges.