부산대학교 도서관

Hydrodynamic models are an essential tool for studying the movement of water and other materials across the Earth surface. However, models remain limited by practical constraints on size and resolution, particularly in coastal environments containing topologically complex and multi‐scale channel/wetland networks. Unstructured meshes have helped address this problem by allowing resolution to vary spatially, and many models support local mesh refinement using breaklines or internal regions‐of‐interest. However, there remains no standardized, objective, or easily reproducible method to implement internal features between different users. The present study aims to address whether remote sensing can fill in that gap, by embedding information about landscape structure and connectivity directly into model meshes. We present an automated image processing methodology for preserving dynamically active connected features in the unstructured shallow‐water model ANUGA, while reducing computational demand in less active areas of the domain. The Unstructured Mesh Refinement Method (UMRM) converts a binary input raster into a collection of closed, simple polygons which can be used to refine the mesh, meanwhile preserving connectivity and enforcing model‐related constraints. We apply this workflow to a large‐scale model of two coastal river deltas, and refine our mesh using time‐series of optical Planet imagery, InSAR measurements of water level change, and topographic data. We compare the results of the connectivity‐preserving mesh to results from a mesh using a uniform mesh resolution, and find that the UMRM decreased the computational demand by a third, without any discernible loss in accuracy when compared to in‐situ and remotely sensed water level measurements. Plain Language Summary: One of the central challenges when modeling flows of water on the Earth surface is achieving high model accuracy with limited computational resources. In order to capture small channels in our models, we need our model cell sizes to be similarly small, which can lead to prohibitively large simulation times and file sizes. Ideally, we would prioritize computations in the locations that are most important or active, which we can do in some models by changing the mesh resolution spatially. Satellite and airborne imagery can show us which locations in channel/wetland networks are most active using optical or radar sensors, but there is no standardized method to embed that remotely sensed information into a model mesh. The present study aims to address this by introducing a new open‐source method with which any remotely sensed data can be used to optimize the structure of a model mesh, in a way that requires minimal user input and can be easily replicated by others. We construct two models of coastal river deltas–one that was refined using multiple sources of remote sensing data, and one that was not–and show that using this method led to significant reductions in computational demand, without losing any accuracy. Key Points: We propose an open‐source, fully automated method to integrate remote sensing data into the structure of hydrodynamic model meshesWe demonstrate this method using optical, InSAR, and topographic data to optimize the mesh in a model of coastal river deltasWe observe a one‐third reduction in model computational demand in our test application due to the proposed method [ABSTRACT FROM AUTHOR]