Scholarly Article
Machine Learning and Morphometric Analysis for Runoff Dynamics: Enhancing Flood Management and Catchment Prioritization in Bayelsa, Nigeria
Jonathan, Lisa Erebi, Winston , Ayebawanaemi Geraldine, Chukwuemeka, Prince
2025-06-09 · Journal of Computational Systems and Applications · Cultech Publishing Sdn. Bhd.
Abstract
Flooding is a recurring environmental hazard with devastating socio-economic and ecological impacts, especially in vulnerable regions like Bayelsa State, Nigeria. The state's low-lying terrain, dense river networks, and poor drainage infrastructure exacerbate its flood susceptibility. This study employs morphometric analysis to assess flood-prone areas across major river basins using Shuttle Radar Topographic Mission (SRTM) data, Geographic Information Systems (GIS), and remote sensing techniques. Key morphometric parameters stream order, drainage density (2.41-3.57 km/km²), bifurcation ratio (1.84-2.84), relief ratio (0.03-0.15), stream frequency (5.00-11.71 streams/km²), infiltration number, and form factor (0.64-1.04) were extracted and analyzed using ArcGIS 10.5, Arc Hydro tools, and Python. Results indicate significant spatial heterogeneity in flood susceptibility. The Forcados River catchment recorded the highest flood risk, with a priority score of 3.4/5, a relief ratio of 0.15, drainage density of 3.57 km/km², and stream frequency of 11.71 streams/km². This aligns with 78% of historical flood event locations. Conversely, the Ekole and Seibri catchments exhibited the lowest susceptibility, with priority scores of 2.0-2.1, relief ratios below 0.05, and drainage densities under 0.9 km/km². The Nun River catchment showed moderate risk (priority score: 2.4), with a stream frequency of 3.2/km² and elongation ratio of 0.6. To enhance predictive capacity, machine learning models were employed. The Random Forest classifier achieved 89% accuracy and an AUC-ROC of 0.93, outperforming the Support Vector Machine model. This study offers a scalable flood assessment framework for data-scarce regions and recommends targeted structural interventions and nature-based solutions tailored to each catchment's flood profile.
Keywords
Flood risk management, Machine learning, Catchments, Geospatial morphometry, SRTM, Remote sensing
Citation Details
Journal of Computational Systems and Applications, pp. 1-16