Spatiotemporal analysis of land cover change, projection, and fragmentation: an application of Google earth engine and machine learning approach on Baraitali forest, Bangladesh

Abstract

ABSTRACT Understanding the spatiotemporal dynamics of land cover/use change and forest fragmentation from 1981 to 2021 across the Baraitali Forest was very crucial for modeling future land cover, formulation of sustainable and robust forest/land-management strategies, and policy. The study adopted a mixed modern tool of Google Earth Engine platform, Machine Learning Algorithms, ArcGIS software, Land Change Modeller of TerrSet software, and FRAGSTATS programs. The study revealed that total land cover from 1981 to 2021 found a negative change in vegetative coverage, and a positive change in settlement which was more visible within the 6 land cover classes. Projections of future land cover for the next 10 years anticipate that vegetation coverage will be reduced by 65.96 ha, while settlement will be enhanced by 41.91 ha in 2031. The study also found conversions/transition of land cover throughout the research area. Overall change of spatiotemporal patterns was leading to substantial forest fragmentation. The overall accuracy and Kappa statistics for all the supervised land cover classifications were satisfactory. The study will fill an information gap by providing the revealed information at national and regional scales and may contribute to understanding of global change.