From Field Observations to Early Warning: A Lightweight Green AI Proposal for Climate-Resilient Coffee Farming in Rwanda
This working paper investigates how lightweight, energy-conscious artificial intelligence could support climate resilience and regenerative coffee farming among smallholder farmers in Rwanda. Drawing on a review of existing agricultural AI applications, including disease recognition, soil monitoring, precision agriculture and satellite-based landscape analysis, the study identifies a gap between technological observation and practical farmer-level decision-making. The paper proposes an accessible early-warning and decision-support system combining farmer observations with environmental information, including rainfall patterns, crop stress, flowering irregularities and seasonal changes. It develops a conceptual AI architecture, stakeholder analysis, implementation pathway, pilot design, evaluation indicators and safeguards for farmer data governance. The proposed approach prioritises explainability, low-resource deployment, human oversight and the integration of local ecological knowledge. The research contributes a practical framework for future testing of responsible Green AI in smallholder agricultural systems. The proposed application has not yet undergone field validation. This paper builds on research completed as part of the Green AI module at Tomorrow University of Applied Sciences in May 2026.
Authors
- Elizabeth Iember Muehlich
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24410458
- Primary Topic
- Smart Agriculture and AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00