Machine Learning Analysis of Land Use Change Dynamics in Malawi’s Chia Lagoon

Biodiversity loss across sub-Saharan Africa is increasingly driven by unsustainable land-use practices and climatic variability, yet data-driven evidence from ecologically sensitive freshwater systems remains limited. This study applied advanced machine-learning (ML) techniques to quantify and interpret two decades (2004–2024) of land-use and land-cover (LULC) change in Malawi’s Chia Lagoon catchment—a semi-closed wetland ecosystem of high ecological and socio-economic value. Using multi-temporal Landsat imagery processed through Random Forest classification integrated with Normalized Difference Vegetation Index (NDVI) and Modified Normalized Difference Water Index (MNDWI) analyses, the study achieved robust discrimination of vegetation, bare land, and water classes. Results show extensive vegetation decline (–5.61%) underscoring biodiversity loss and water contraction (–0.57%), accompanied by a 6.18% expansion of bare land. The most pronounced degradation occurred between 2004 and 2014, followed by localized vegetation recovery after 2014, suggesting limited restoration potential under community-based management. These spatial transformations reflect intensifying agricultural expansion, deforestation, and sedimentation processes, exacerbated by climatic stress. The findings demonstrate the power of ML for high-accuracy environmental monitoring in data-scarce contexts and underscore the need for integrated, evidence-driven conservation planning. Protecting Chia Lagoon requires restoring degraded buffer zones, strengthening catchment management, and embedding ML-based monitoring into national biodiversity frameworks to support Malawi’s commitments under the Convention on Biological Diversity (CBD) and Sustainable Development Goal 15 (“Life on Land”).