Using Machine Learning and GIS to Monitor Sandbars Along the River Niger in the Niger Delta, Nigeria

Abstract

The use of GIS and machine learning techniques to map sand bars along the Niger River in the Niger Delta, Nigeria, spanning the period from 1974 to 2024. It integrates DEM, Landsat series satellite imagery obtained from the USGS. Rainfall data from 1983 to 2023, sourced from the Center for Hydrometeorology and Remote Sensing, supplements the analysis. Object-Based Image Analysis is employed to identify and map sand bars, while Support Vector Machines automate classification to ensure precision and recall metrics. ArcGIS 10.5 tracks temporal changes, revealing significant morphological shifts influenced by both natural processes and human activities. Results show a significant reduction in sandbar length from 1.6502 km in 1974 to a low of 0.7437 km in 2004, while sandbar area decreased by 68% to 0.0587 km² by 2004 before partially recovering to 0.1271 km² in 2024. Key parameters such as Aspect Ratio (AR) and Elongation Ratio (ER) demonstrate relative stability, indicating consistent directional flow influence on sandbar shape. Spatial autocorrelation analysis (Moran's Index of 0.138562) links sandbar dynamics to elevation, with a significant correlation between rainfall and sandbar area fluctuations (R² = 0.7576). Regression analysis reveals strong associations among sandbar length, width, and area (R² values up to 0.9737), indicating predictable morphometric responses to environmental changes. Additionally, grain size impacts sandbar stability, with medium to coarse sands forming more stable structures. Comparative global analyses reinforce the broader implications of these findings for sustainable river management, stressing the need for balanced policies in response to climate change, sediment transport, and anthropogenic activities. The study underscores the importance of advanced monitoring technologies for effective riverine ecosystem management and sediment regulation.