10th U6+ Consortium of African Universities International Conference

Flood risk forecasting stands at a critical juncture, as the escalating frequency of extreme weather events, driven by climate change, outpaces the capabilities of conventional hydrological models, particularly in data-sparse regions. There is also an observable gap in the uptake of meteorological predictions by local stakeholders, partly due to prediction unreliability and the process’s exclusivity. This paper examines the emerging paradigm of integrating Machine Learning (ML) with Indigenous Knowledge Systems (IKS) for flood risk forecasting by moving beyond techno-centric approaches for meteorological data production. The paper argues that Indigenous Knowledge, rooted in multigenerational place-based ecological observation, offers not merely supplementary data but a fundamentally different epistemological framework for understanding hydrological uncertainty in Africa. The paper discusses how ML architectures can be redesigned to operationalize traditional ecological knowledge using AI-driven climate resilience initiatives and mobile-based knowledge quantification platforms without violating principles of data sovereignty and epistemic pluralism. The paper further addresses critical challenges, including the risk of “data colonialism,” the incommensurability of knowledge systems, and ethical protocols for Indigenous data governance. It concludes by proposing a convergence framework that positions ML as a “boundary object” capable of translating between scientific and Indigenous knowledge systems without reducing either to the terms of the other, thereby advancing a decolonized vision of climate science.