Land use changes, morpho-dynamics, and future projections of Char Kukri Mukri Island in Bangladesh using remote sensing and machine learning

• Published in: Regional studies in marine science 2025-01, Vol.81, p.103959, Article 103959
• Main Authors: Karmaker, Karabi, Hossain, Md. Imam Sohel, Hamid, Taspiya, Rana, Md. Shohel, Bhuiyan, Md Mesbah Uddin, Samad, Md Abdus
• Format: Article
• Language: English
• Subjects: Char Kukri Mukri; Machine Learning; Morphodynamics; Prediction; Random Forest
• ISSN: 2352-4855; 2352-4855
• DOI: 10.1016/j.rsma.2024.103959

Coastal chars are vulnerable to climate hazards, including rising sea levels, storm surges, flooding, and erosion. These landmasses, inhabited by thousands, require an understanding of their spatio-temporal changes to assess stability. This study provides a comprehensive analysis of land use and land cover (LULC) changes and morpho-dynamics of Char Kukri Mukri, a coastal island at the mouth of the Bay of Bengal. Using a Random Forest (RF) machine learning algorithm and ArcGIS, we analyzed Landsat images over a 22-year period (2000–2022). Our results indicate significant increases in forest, barren, and urban areas, with growth rates of 22.07 %, 6.73 %, and 1.69 %, respectively, driven by afforestation and urbanization efforts. In contrast, vegetation decreased by 6.02 %, and water bodies declined by 11 %, mainly due to land expansion. Notably, accretion surpassed erosion, resulting in a net increase of 26.44 km², primarily in the western and northern areas, attributed to sediment from the Ganges-Brahmaputra-Meghna river system combined with climate variability and anthropogenic activities. Despite a 5.70 km² loss in the eastern region due to tidal effects, 56.70 % of the total area (68.52 km²) in the central portion of the island remained stable. Cross-validation and field visits confirmed our findings with high accuracy, exceeding 88 %. Projections for the next five years indicate that the Random Forest algorithm is the strongest predictor, outperforming other machine learning methods. Based on predictions from the RF model, we anticipate continued stability in forest, urban, and barren areas, suggesting effective conservation efforts and sound urban planning. Additionally, agricultural land is expected to expand due to ongoing development projects and population growth. In conclusion, this study offers crucial insights into LULC changes, erosion-accretion dynamics, and future stability, with significant implications for global coastal management. It supports the development of sustainable land use strategies in the face of climate change.