Machine Learning Model for Mapping and Monitoring Water Hyacinth In Lake Victoria, Kenya

The proliferation of Eichhornia crassipes (water hyacinth) continues to threaten fisheries, aquatic biodiversity, transportation, and livelihoods within Lake Victoria. Effective management of this invasive species requires timely, accurate, and scalable monitoring approaches. This study explores the application of Artificial Intelligence (AI) and remote sensing technologies to develop a machine learning-based framework for mapping and monitoring water hyacinth distribution in the Kenyan waters of Lake Victoria. Multispectral imagery acquired from Sentinel-2 and Landsat-8 satellites was analyzed using unsupervised machine learning techniques, including K-Means Clustering, Autoencoders integrated with K-Means, Self-Organizing Maps (SOM), and Hierarchical Clustering. The models were evaluated using the Silhouette Score, Calinski-Harabasz Index, and Davies-Bouldin Index to determine clustering performance and spatial discrimination capability. Results demonstrated that SOM achieved the highest clustering performance, recording a Silhouette Score of 0.753 and the lowest Davies-Bouldin Index of 0.471, indicating superior cluster cohesion and separation. Autoencoders integrated with K-Means also produced strong results, while conventional K-Means and Hierarchical Clustering showed comparatively lower performance. The findings demonstrate the potential of AI-driven remote sensing approaches to provide accurate and cost-effective monitoring of invasive aquatic vegetation over large geographic areas.

The study contributes to the growing field of digital environmental monitoring by demonstrating how machine learning and satellite data can support evidence-based decision-making for aquatic ecosystem management. The proposed framework offers a scalable solution for continuous monitoring of water hyacinth and can be adapted to other environmental surveillance applications within the Blue Economy and natural resource management sectors.