Integrating machine learning algorithms and fuzzy AHP for assessing livelihood vulnerability in Southwestern Coastal Bangladesh

Abstract

The people living in the coastal regions of Bangladesh rely significantly on nature for their livelihoods, which renders them sensitive to climate change. The aim of this study is to determine the key indicators which contribute more to livelihood vulnerability of disaster-prone Gabura union in southwestern coastal Bangladesh. To achieve this goal, three machine learning algorithms are employed for determining the key indicators of livelihood vulnerability. Subsequently, a livelihood vulnerability index (LVI) is constructed using these key indicators with the weighting of indicators facilitated by Fuzzy AHP method. And finally, a livelihood vulnerability map (LVM) is generated to visualize the spatial distribution and analysis of livelihood vulnerability within the union. The study employs a mixed-methods approach, including questionnaire surveys, focus group discussions, key informant interviews, and remote sensing image analysis. A household survey of 950 households tracked livelihood vulnerability using 25 indicators across three domains of vulnerability, e.g., exposure, sensitivity, and adaptive capacity. ArcGIS and Google Earth Engine facilitated spatial data analysis. The Gabura Union exposed high vulnerability (LVI 0.63), driven by elevated exposure (0.61) and sensitivity (0.59) and lower adaptive capacity (0.30). The Livelihood Vulnerability Map (LVM) illustrates vulnerability across wards of the union, emphasizing high vulnerability zones on the periphery of the union, along the Kopothakho and Kholpetua rivers while central part of the union shows a moderate vulnerability level. The study's novelty lies in effectively integrating multiple methods for livelihood vulnerability assessment. Policymakers should target interventions focusing on areas along the Kopothakho River, to enhance community resilience.