Separability Analysis for Wetland Mapping using Segmented Principal Component Analysis in Pulicat Lake, India

Abstract

Lakes and coastal lagoons are ecologically sensitive systems that require continuous monitoring to understand the impacts of environmental change and human activities. Land cover classification plays a critical role in understanding the spatial distribution and temporal dynamics of key landscape elements, thereby aiding in effective management and conservation strategies for these sensitive regions. This study analyzes land cover dynamics in the Pulicat Lake region using multi-temporal Landsat 8 imagery from 2014, 2019, and 2024. Principal Component Analysis (PCA) and Segmented PCA (SPCA) were integrated with Support Vector Machine (SVM) and Random Forest (RF) classifiers to improve land cover mapping. Four major land cover types, namely Water, Vegetation, Bareland, and Wetland, were mapped and analysed. Accuracy assessment confirmed that SPCA-based classifications consistently outperformed PCA, achieving the highest overall accuracy of 95.83%, 94.47% and 96.79% with RF for the years 2014, 2019 and 2024 respectively. The land cover change assessment revealed notable shifts in water, vegetation, bareland, and wetland extents over the study period, indicating ongoing ecological transitions in the lagoon system. The findings highlights the effectiveness of combining SPCA with Random Forest algorithm for accurate and efficient land cover monitoring and supports informed decision-making in the sustainable management of lake ecosystems.