Development of computational method for decision-making in sustainable coastal development

Abstract

The destruction of marine environments from coastal development is inevitable. For sustainable coastal development, a method or tool is needed to understand, assess, and predict marine environments and their changes in a comprehensive and quantitative manner. Marine environments can be largely explained through four study areas: ocean physics, water quality, marine ecosystems, and marine geology. Each area can be described by a number of observational items. For comprehensive study, it is necessary to choose appropriate observational items in each study area and it is important to understand their interrelationships. In general, oceanographic research, hydrodynamic ocean models have been established, and several items on the physical states of sea waters have been computed. For the areas of water quality, marine ecosystems, and marine geology, the coupling of a eutrophication water quality model and an ecological model with a hydrodynamic model has been attempted and adjusted through observational data. However, it is challenging to get a full picture when unnatural coastal development is involved. In this thesis, we propose a computational method using the derivation of a mathematical model with observational data and a probabilistic inference. From the structure of the relationships and standardized estimates determined through model estimation, marine environments and their changes can be understood comprehensively and assessed quantitatively. Furthermore, changes in marine environments can be predicted through a probabilistic inference by establishing a probabilistic graphical model using the derived model. The proposed computational method was applied to the Saemangeum coast using observational ocean data collected for sustainable coastal development from a land reclamation project. For an optimal mathematical model, a multilevel structural equation model consisting of four unobserved latent variables, fifteen observed variables, and their causal relati...