Arinze, Stella N., Nwajana, Augustine ORCID: https://orcid.org/0000-0001-6591-5269, Egbe, Gregory O. and Ebenuwa, Augustine (2025) Real-time IoT-enabled flood risk prediction and adaptive water redistribution with machine learning: a case study of Nigeria. Modelling and Simulation in Engineering. ISSN 1687-5591 (Print), 1687-5605 (Online) (doi:10.1155/mse/4694316)

Preview

PDF (Open Access Article) 51761 NWAJANA_Real‐Time_IoT‐Enabled_Flood_Risk_Prediction_and_Adaptive_Water_(OA)_2025.pdf - Published Version Available under License Creative Commons Attribution. Download (2MB) | Preview

Abstract

Flooding remains one of the most destructive natural disasters, causing loss of life, property damage, and economic disruption. It is a persistent challenge in Nigeria, with Mokwa Local Government Area (LGA) in Niger State being particularly vulnerable due to its location in the Lower Niger River Basin, influenced by both the River Niger and the Kaduna tributary. This study presents a hybrid convolutional neural network–long short-term memory (CNN–LSTM) model integrated with an automated intervention system for real-time flood forecasting, prediction, and proactive control. The CNN component extracts spatial correlations across a network of 16 hydrometeorological, geospatial, and sensor-based variables (including rainfall, river discharge, soil moisture, and flood risk indices), while the LSTM captures long-term dependencies to enhance time-series prediction accuracy. The model was trained on historical records prior to 2000, with validation and operational testing using daily observations and simulated sequences from 2000 to 2025, totaling 9125 daily records; flood events were labeled when water levels exceeded 3.5 m. The proposed model achieved 96% forecast accuracy, a precision of 0.93, a recall of 0.89, and an F1 score of 0.91. The hybrid framework demonstrated robustness to noise, faster convergence, and improved recall for extreme flood events. Automated intervention simulations showed a 72% reduction in peak water levels, confirming the framework′s effectiveness in mitigating flood severity. Comparative analysis against conventional machine learning models highlighted the system′s superiority in covering the full flood management cycle of data acquisition, forecasting, prediction, and automated response, enabling timely decision-making and reducing delays from human intervention. This CNN–LSTM automation approach provides a scalable, AI-driven solution for proactive flood management in riverine communities such as Mokwa, with potential applicability to other flood-prone regions.

Item Type:	Article
Uncontrolled Keywords:	flood forecasting, CNN–LSTM, IoT-based early warning system, data-driven water management, flood resilience planning, automated water diversion, hydrological time-series modeling
Subjects:	Q Science > Q Science (General) Q Science > QA Mathematics Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Faculty / School / Research Centre / Research Group:	Faculty of Engineering & Science Faculty of Engineering & Science > Natural Resources Institute Faculty of Engineering & Science > School of Engineering (ENG) Faculty of Engineering & Science > School of Science (SCI)
Last Modified:	24 Nov 2025 16:48
URI:	https://gala.gre.ac.uk/id/eprint/51761

Actions (login required)

View Item

Downloads