Keywords
2D Latin Hypercube sampling; Monte Carlo simulation; Uncertainty Analysis; Hydraulic modelling; Flood Hazard
Start Date
27-6-2018 10:40 AM
End Date
27-6-2018 12:00 PM
Abstract
Floods are the most common natural risk to life and property worldwide, causing over £6B worth of damage to the UK since 2000. Climate projections would result in an increase of UK properties at risk from flooding. It thus becomes urgent to assess the possible impacts of these changes on extreme floods, while evaluating the uncertainties related to these projections and assessing the predominant sources in an impact framework. This paper aims to assess the changes to flood extent for the 1:100 year return period event of the River Don in Scotland (UK) as a result of climate change. It is based on the analysis of the Future Flow dataset for the Parkhill gauge station (Collet et al., 2017). Extreme value (EV) distributions are fitted for the 11 climate-change ensembles over the baseline (1961-1990) and the 2080s (2069-2098) to account for climate non-stationarity. Monte Carlo (Random sampling) and Latin Hypercube (LH) are undertaken and compared using a 1D-2D hydraulic model (LisFLOOD) on a 5km stretch of the Don, with LH reducing the computational cost with no loss of accuracy. To investigate how the uncertainties cascade into the modelling framework, values were sampled from two different variables: extreme peak flow (Climate Model and EV distribution parameter uncertainty) and Manning’s n coefficient (hydraulic model parameter uncertainty). Results show the change in extent from the baseline to the future, capturing the uncertainty associated to each source, indicating that quantifying these uncertainties is essential when planning engineering interventions.
Quantification of Uncertainty Sources in Hydraulic Modelling in a Climate Change Impact Framework
Floods are the most common natural risk to life and property worldwide, causing over £6B worth of damage to the UK since 2000. Climate projections would result in an increase of UK properties at risk from flooding. It thus becomes urgent to assess the possible impacts of these changes on extreme floods, while evaluating the uncertainties related to these projections and assessing the predominant sources in an impact framework. This paper aims to assess the changes to flood extent for the 1:100 year return period event of the River Don in Scotland (UK) as a result of climate change. It is based on the analysis of the Future Flow dataset for the Parkhill gauge station (Collet et al., 2017). Extreme value (EV) distributions are fitted for the 11 climate-change ensembles over the baseline (1961-1990) and the 2080s (2069-2098) to account for climate non-stationarity. Monte Carlo (Random sampling) and Latin Hypercube (LH) are undertaken and compared using a 1D-2D hydraulic model (LisFLOOD) on a 5km stretch of the Don, with LH reducing the computational cost with no loss of accuracy. To investigate how the uncertainties cascade into the modelling framework, values were sampled from two different variables: extreme peak flow (Climate Model and EV distribution parameter uncertainty) and Manning’s n coefficient (hydraulic model parameter uncertainty). Results show the change in extent from the baseline to the future, capturing the uncertainty associated to each source, indicating that quantifying these uncertainties is essential when planning engineering interventions.
Stream and Session
Stream E: Modeling for Planetary Health and Environmental Sustainability
Session E3: Complexity, Sensitivity, and Uncertainty Issues in Integrated Environmental Models