Keywords

flood inundation modeling, artificial neural network, meta model, surrogate model

Location

Colorado State University

Start Date

26-6-2018 5:00 PM

End Date

26-6-2018 7:00 PM

Abstract

Flood inundation models are important tools for flood risk estimation, flood protection infrastructure design and river system management. Very often, one-dimensional (1D) or two dimensional (2D) hydrodynamic models are used to simulate key variables, such as velocity and water depth. However, hydrodynamic models are computational expensive, which prevents their application in uncertainty and probability related analyses, where a large number of model runs are required. To significantly improve the efficiency of flood inundation modeling process, artificial neural network (ANN) based surrogate models can be used. This study investigates the suitability of ANN models as surrogate models for modeling flood inundation. Key issues in developing ANN-based inundation models, including sampling locations and sampling methods and their impact on derived flood information in a flood plain, are also discussed. This study provides insights into the best practice in developing ANN-based surrogate models for flood inundation modeling.

Stream and Session

C11: Integrated Methods and Tools for Flood Risk and Water Supply Management

Share

COinS
 
Jun 26th, 5:00 PM Jun 26th, 7:00 PM

ANN-based surrogate models for flood inundation modeling

Colorado State University

Flood inundation models are important tools for flood risk estimation, flood protection infrastructure design and river system management. Very often, one-dimensional (1D) or two dimensional (2D) hydrodynamic models are used to simulate key variables, such as velocity and water depth. However, hydrodynamic models are computational expensive, which prevents their application in uncertainty and probability related analyses, where a large number of model runs are required. To significantly improve the efficiency of flood inundation modeling process, artificial neural network (ANN) based surrogate models can be used. This study investigates the suitability of ANN models as surrogate models for modeling flood inundation. Key issues in developing ANN-based inundation models, including sampling locations and sampling methods and their impact on derived flood information in a flood plain, are also discussed. This study provides insights into the best practice in developing ANN-based surrogate models for flood inundation modeling.