AI-Powered Climate Risk and Crop Yield Prediction System for Food Security and Disaster Resilience

Climate change continues to intensify the vulnerabilities of rural communities across Sub-Saharan Africa, disrupting food systems, livelihoods, and disaster preparedness. Extreme weather events prolonged droughts, floods, and unpredictable rainfall are becoming more frequent and destructive, threatening already fragile ecosystems and deepening poverty. Smallholder farmers for example my mother, who form the backbone of African agriculture, often lack timely, localized, and accessible information to make informed decisions that could safeguard their crops, income, and food security. This innovation emerged from direct engagement with youth, students, farmers, and educators in Western Kenya. Many shared how weather shifts left them guessing about when to plant, harvest, or seek alternative food sources. Others, particularly women and marginalized communities, reported having little or no access to climate advisory services. Most lacked digital literacy or access to formal forecasting systems, especially in rural and semi-urban areas. This was a clear gap where global forecasting tools were too generalized or inaccessible, and local decision-makers operated on uncertainty. Our AI-powered solution addresses these challenges by combining a locally developed natural language parser and a trained Random Forest ML model to forecast climate risks and crop yield probabilities based on user queries. Unlike static dashboards, users simply type or speak questions such as ?Will there be hunger in Bungoma next season?? and receive real-time predictions grounded in historical data patterns and localized variables. This removes the barrier of technical jargon, enabling community-friendly access. The model is trained on curated datasets reflecting African climate conditions, food production trends, and socio-economic factors. A Streamlit web app currently serves as the front-end, making the solution instantly usable via browser or mobile, with Swahili input planned to further increase reach. The MVP is functional and in early deployment within a small user community, with feedback guiding continuous improvements. What makes this innovation distinct is its ability to combine AI prediction with inclusive design, ensuring that women, youth, and low-literacy users are not left behind in the shift to AI-powered agriculture. Our design principles emphasize ethical use, explainable predictions, and user empowerment. With support, we envision this tool being adopted by farmer cooperatives, extension officers, NGOs, and even regional governments to proactively plan for food shortages, trigger early response strategies, and reduce climate-driven inequalities.

Sub-Saharan Africa faces a critical challenge: climate-induced disasters and hunger are rising, yet most vulnerable communities lack access to accurate, localized risk forecasts. Farmers, aid organizations, and local governments often make key decisions planting dates, food stockpiling, emergency response based on guesswork or delayed information. This information gap leads to preventable crop losses, unplanned food insecurity, and poor disaster preparedness. Conventional forecasting tools are often too generalized, require internet or weather literacy, and fail to deliver timely, user-specific insights. Additionally, marginalized groups including women farmers, persons with disabilities, and rural youth face barriers to accessing existing systems, further deepening inequality. Our innovation addresses this gap by offering a low-barrier, AI-powered tool that enables users to ask natural questions and receive meaningful predictions. It brings together machine learning, natural language processing, and inclusive design to democratize access to climate intelligence turning uncertainty into preparedness.

The primary objective of this innovation is to provide underserved communities in Africa with a user-friendly, AI-powered tool that can predict climate risks such as droughts, floods, hunger, and crop yield fluctuations based on natural language input. By leveraging artificial intelligence, the goal is to reduce uncertainty in decision-making for smallholder farmers, policymakers, and disaster relief organizations. Sub-objectives include simplifying access to climate forecasting through a natural language interface; empowering vulnerable populations, especially women, youth, and low-literacy users; and improving agricultural and emergency planning outcomes in regions most affected by climate variability. The project also aims to promote responsible AI development by embedding transparency, inclusivity, and community feedback in both the design and scaling process. Ultimately, this innovation seeks to increase climate resilience and food security across Africa.

The innovation is an AI-powered climate risk and crop yield prediction tool that enables users to ask plain-language questions and receive localized forecasts about weather hazards and food security. It consists of two key technical components: a custom-built Natural Language Processing (NLP) parser and a machine learning model based on Random Forest. The NLP parser, developed using Python and SpaCy, extracts relevant information from user queries such as the hazard type, location, and timeframe. The structured query is then passed to the prediction model, which has been trained on climate, crop yield, and socio-economic datasets from sources like the World Bank, FAO, and Kaggle. Currently deployed as a web app via Streamlit, the system allows users to ask questions like ?Will there be hunger in Bungoma in October?? and receive interpretable predictions with confidence scores. The interface is intentionally simple, and expansion plans include voice input, Swahili language support, and WhatsApp/mobile integration to accommodate users with limited literacy or bandwidth. The solution is designed with modularity in mind enabling easy retraining for new regions using local datasets. It is hosted on Render and can be rapidly scaled through cloud services and decentralized partnerships with NGOs and agricultural agencies. The innovation brings immense value by bridging the gap between complex climate models and real-world accessibility for at-risk populations. It allows both individuals and institutions to plan proactively for climate shocks and food shortages. From an ethical standpoint, the project prioritizes responsible AI practices, including transparency in predictions, non-biased data sourcing, and clear user consent for any data inputs. No personally identifiable data is collected during predictions, and the system is structured to incorporate community feedback for continuous improvement. Plans are also in place to include fairness indicators and localized explainability features. The innovation?s strength lies in its balance of technical sophistication and grassroots accessibility?empowering users to act on AI-driven insights without needing to be tech experts.