Title : Integrating artificial intelligence, IoT-ready monitoring and technology-enhanced learning for aquatic pollution assessment in the Gongola–Benue River System, Nigeria
Abstract:
Aquatic ecosystems in Nigeria are increasingly exposed to diffuse and localized anthropogenic pressures associated with mining, urbanization, agriculture and other land-use activities. Detecting the resulting contamination footprint is difficult where environmental monitoring remains spatially sparse and largely dependent on episodic field sampling. This study investigated the feasibility of integrating field-derived contamination measurements, artificial intelligence (AI), an Internet of Things (IoT)-ready monitoring architecture and Technology-Enhanced Learning (TEL) to transform sparse aquatic pollution observations into interpretable and decision-ready environmental intelligence. The study was conducted as a pilot proof-of-concept investigation along an approximately 10-km section of the Gongola–Benue river system near Numan, Adamawa State, Nigeria. Six strategically selected stations were sampled for water and sediment, with concentrations of Pb, Cd, Zn, Cu, Cr and Fe determined by atomic absorption spectrometry. The environmental measurements provided the empirical basis for examining the spatial distribution of contaminants and for testing a small-data AI analytical workflow. Rather than treating the six stations as sufficient for generalisable spatial prediction, the AI component was designed as an exploratory proof of concept for identifying contamination patterns and generating Eco-Status information from sparse observations. A conceptual IoT architecture was developed to demonstrate how future continuous sensing could feed environmental data into the analytical layer; no field-deployed real-time IoT sensor network was used in the present study. The TEL component was implemented as a human-centred knowledge-translation intervention linking AI-derived environmental outputs to an Eco-Status dashboard, an interpretation module, stakeholder learning activities and structured decision scenarios. Stakeholder evaluation examined usability, comprehension, confidence, trust and decision readiness across participating technical and non-technical groups. The environmental results indicated spatial variation in metal concentrations and greater accumulation of several metals in sediment than in water, although the limited number of stations constrains inference about the broader spatial behaviour of contamination. The integrated workflow demonstrated the feasibility of converting sparse environmental observations into interpretable pollution intelligence and of using TEL to support stakeholder engagement with AI-derived information. The study therefore provides a pilot demonstration of an IoT-ready, human-centred AI-TEL approach for aquatic pollution assessment under data-constrained conditions in Nigeria. Its principal contribution is not a claim of operational real-time monitoring or nationally transferable prediction, but a methodological pathway linking environmental evidence, computational analysis, knowledge translation and decision readiness. Multi-season, multi-site validation with larger independent datasets and field-deployed sensors is required before the approach can support broader predictive or operational monitoring applications.
Keywords: aquatic pollution; artificial intelligence; Technology-Enhanced Learning; IoT-ready monitoring; heavy metals; environmental intelligence; Gongola–Benue River; stakeholder decision-making; Nigeria
