Abstract
Abstract
In recent decades, global rapid urbanization has exacerbated the impacts of natural hazards due to changes in Southeast Asia’s environmental, hydrological, and socio-economic conditions. Confounding non-stationary processes of climate change and global warming and their negative impacts can make hazards more complex and severe, particularly in Vietnam. Such complexity necessitates a study that can synthesize multi-dimensional natural-human factors in disaster risk assessments. This synthesis study aims to assess and monitor climate change and land-cover/land-use change impacts on flood and landslide hazards in Vietnam’s Gianh River basin. Three Deep Neural Network (DNN) and optimization algorithms, including the Adam, Tunicate Swarm Algorithm (TSA), and Dwarf Mongoose Optimization (DMOA) were used to determine the regions with the probability of the occurrence of flood and landslide and their combination. All efficiently evaluated hazard susceptibility based on a synthesis analysis encompassing 14 natural and anthropogenic conditioning factors. Of the three, the Deep Neural Network (DNN)-DMOA model performed the best for both flood and landslide susceptibility, with area-under-curve values of 0.99 and 0.97, respectively, followed by DNN-TSA (0.97 for flood, 0.92 for landslide), and DNN-Adam (0.96 for flood, 0.89 for landslide). Although the area affected by flooding is predicted to decrease, the overall trend for total hazard-prone areas increases over 2005–2050 due to the more extensive area affected by landslides. This study develop and demonstrate a robust framework to monitor multi-hazard susceptibility, taking into account the changes in climate and land-use influence the occurrence of multiple hazards. Based on the quantitative assessment, these findings can help policymakers understand and identify confounding hazard issues to develop proactive land-management approaches in effective mitigation or adaptation strategies that are spatially and temporally appropriate.