EPIRUS: An Integrated Clouds-to- Coast Ensemble Modelling Framework Of Coastal Flood Risk

EPIRUS: An Integrated Clouds-to- Coast Ensemble Modelling Framework Of Coastal Flood Risk Qingping Zou, Yongping Chen, Ian Cluckie, Richard Hewston, Xin Lv, Shunqi Pan, Zhong Peng, Dominic Reeve,

Capital Value of Assets at Flood Risk ( billions) Region Anglian Midland North East North West South West Southern Thames Total Property 10.7 14.1 7.8 2.9 3.9 4.6 28.6 72.5 Fluvial Agriculture 2.9 0.6 0.4 0.2 0.4 0.1 0.3 4.9 Total 13.6 14.7 8.2 3.1 4.3 4.7 28.9 77.4 Property 9.7 1.8 10.4 7.8 2.9 13.8 81.3 127.8 Sea/ Tidal Agriculture 0.7 0.3 0.3 0.2 0.1 0.3 0.0 1.9 Total 10.4 2.1 10.7 8.0 3.0 14.1 81.3 129.7 Property 20.4 15.9 18.2 10.7 6.9 18.4 109.8 200.3 Total Agriculture 3.6 0.9 0.7 0.4 0.5 0.4 0.3 6.8 Grand Total 24.0 16.8 18.9 11.1 7.4 19.8 110.1 207.1 After MAFF (2001)

Coastal Flood and Erosion 132 billion assets at coastal flood risk 7.8 billion assets at coastal erosion risk 4 million people and properties in England and Wales at coastal flooding and erosion risk Coastal Flood at Brighton (Jan 2007) Coastal Erosion at Happisburgh (July 2006)

Coastal Flood Defence Failure Functional failure: Conditions exceed what the defence was designed for Structural failure: Element or components of defence fail to perform as expected Wave overtopping Erosion of the back and crest of defence Damage to armour layers Toe scour Leads to beach lowering Wave overtopping increase water depth larger waves more beach lowering Undermining defence Toe scour

Objectives To improve the capacity for predicting coastal flood risk due to extreme events and estimate the associated uncertainty To assess the propagation of uncertainty from meteorological forecasts to coastal flood risk predictions Approach a clouds-to-coast, integrated modelling framework for ensemble prediction of coastal flood risk arising from overtopping and scour

Clouds-to-coast Ensemble Modelling Framework of Coastal Flood Risk 1. Weather forecasting model 2. Tide, surge & wave models 3. Surf zone models

Ensemble Weather Forecasting Model Dynamicaldownscaling approach from global to regional to local scale Compromise between domain size and resolution Hi-res wind and pressure forecast using WRF

EPIRUS Meteorological component Example: The Great Storm 16 th October 1987 ERA40 boundary conditions Downscaling high temporal (hourly) and spatial (~1km) wind and pressure fields for input in oceanographic models. Wind and pressure fields at 03:00 on 16 th Oct 1987, in three domains with increasing spatial resolution Ensemble Prediction of Inundation Risk and Uncertainty arising from Scour (EPIRUS)

Tide, Surge & Wave Modelling POLCOMS & ProWAM Models COAST2D Model WAM Model

Herne Bay Deal Pier Model Validations - Waves Boscombe Lymington HaylingIsland Pevensey Bay Folkestone Milford Sandown Bay Rustington Wave meters 3.5 3 Boscombe OBS Control 2.5 Hs (m) 2 1.5 1 0.5 0 16/10/04 18/10/04 20/10/04 22/10/04 24/10/04 26/10/04 28/10/04 30/10/04 01/11/04 t ime 3.5 3 Folkstone OBS Control 2.5 Hs (m) 2 1.5 1 0.5 0 16/10/04 18/10/04 20/10/04 22/10/04 24/10/04 26/10/04 28/10/04 30/10/04 01/11/04 t ime

Model Validations Tides & Surge Boumemouth Elevation (m) Devonport 4 3 OBS Control 2 1 0-1 -2-3 -4 16/10/04 18/10/04 20/10/04 22/10/04 24/10/04 26/10/04 28/10/04 30/10/04 01/11/04 t ime 1 0.8 Devonport OBS Control 0.6 Surge (m) 0.4 0.2 0-0.2-0.4 16/10/04 18/10/04 20/10/04 22/10/04 24/10/04 26/10/04 28/10/04 30/10/04 01/11/04 t ime

Ensemble Modelling Storm Tracks

Ensemble Modelling - Surges Ensemble 06 Ensemble 08 Control Ensemble 39 Ensemble 46

Ensemble Modelling - Waves Ensemble 06 Ensemble 08 Control Ensemble 39 Ensemble 46

Surf Zone Model 1. Fluids: 3-D LES-VOF model-unstructure grid Aberystwyth, Wales 2-D RANS-VOF model-structure grid Overtopping of coastal structures 2. Sediments: Advection-diffusion equation Bedload transport model 3. Morphology: Teignmouth, England Sediment conservation equation over the whole water column - Scour

Numerical Model (2D RANS-VOF) Governing equations: Reynolds-Averaged Navier-Stokes equations (RANS) Turbulence: Algebraic nonlinear k ε turbulence model Surface capture scheme: Volume of Fluid (VOF) VOF function, f : ρ ρ f : fluid density, ρ: averaged density f = ρ f f 0 empty cell = 1 full cell other surface cell The transport equation for f: f + ( uf ) + ( vf ) = 0 t x y

Numerical simulation setup Free surface profile at t=1000s 15 Seawall SWL Y (m) 10 Sponge Layer Source Region Free surface 5 Beach 0 0 50 100 150 200 250 300 X (m) Free surface profile at t=1000 s and setup of numerical simulation.

Waves at the toe of seawall η (m) S(f) (m 2 /Hz) 13 12 11 4 2 0 200 400 600 800 1000 1200 t (s) 0 0 0.1 0.2 0.3 0.4 0.5 f (Hz) Wave conditions at the toe of seawall (x=320 m): H 1/3 =1.68 m; T m = 9.9747 s; T p = 12.8 s.

Velocity vector at t=1025s 12 10 Free surface and velocity vector at t=1025s y (m) 8 6 4 200 220 240 260 280 300 320 x (m) Velocity vector distribution at t=1025s around the seawall. Blue line is the free surface and yellow layer represents the seawall and foreshore with slope of 1:30

Results of wave overtopping Cumulative overtopping volume (m 3 /m) 15 10 5 0 0 200 400 600 800 1000 1200 t (s) Discharge q (m 3 /m/s) 20 x 10-3 15 10 5 0-5 Numerical results Prediction by TAW (2002) 200 400 600 800 1000 1200 t (s) Predicted time series of accumulative overtopping volume Overtopping discharge q calculated roughly after 12 waves. Red dotted line is the prediction by the empirical formula of TAW (2002) and blue solid line is simulation results

Overtopping volume per wave Overtopping volume (m 3 /m) 1 0.8 0.6 0.4 0.2 0 0 20 40 60 80 100 t/t Wave overtopping volume per wave cycle. t is the simulation time and T equals the peak period here.

Comparisons of discharge Q Comparison of dimensionless overtopping discharge, Q, between numerical results and predictions of design formulae by Van der Meer and Janssen (1995) and TAW (2002)

3-D Simulation of Suspended Sediment Under breaking waves (1) (5) (2) (6) (3) (7) (4) (8)

3-D Wave Overtopping and run up over a steep Seawall 25

3-D Wave Overtopping over a Seawall

Random Wave Overtopping a Seawall ---Overtopping Discharge Nondimensional mean overtopping discharge Q 1.0E-2 1.0E-3 1.0E-4 1.0E-5 1.0E-6 1.0E-7 1.0E-8 Van deer Meer (scalling coefficient=0.5) Van deer Meer (scalling coefficient=1.0) Owen (1980) Hedges & Reis (1998) Reeve et al. (2007) Numerical regresion Current model results Numerical regresion 1.0E-9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 Nondimensional Freeboard R 27

Conclusions/Summary Weather forecasting, wave/surge and surf zone models have been identified and tested individually Good agreement between wave, tidel and surge prediction with observations Large uncertaity in ensemble predictions of surge and wave for 2004 storm Overtopping predictions in good agreement with published data Test sites and events have been selected Morphological model at testing stage

Future Work Ensemble run of surf zone model to create ensemble forecasts of overtopping and scour Big question: how best to use ensembles is mean necessarily the most useful? How Uncertainty propagate from meteorological forecasts to overtopping and coastal flooding risk predictions?

EPIRUS Meteorological component Aim: Dynamically downscale coarse climate data for severe storms high resolution wind and pressure fields Methodology: Weather Research and Forecasting model (WRF) Next-generation NWP and data assimilation system WRF run in parallel across several nodes on high performance computer Blue Ice Initial focus on historic severe storms, before exploring impact of climate change Generating ensembles will allow estimate of uncertainty Ref: Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, M.G. Duda, X-Y. Huang, W. Wang, and J. G. Powers, 2005: A description of the advanced research WRF version 2. NCAR Tech. Note NCAR/TN-475+STR. 125pp Ensemble Prediction of Inundation Risk and Uncertainty arising from Scour (EPIRUS)

Thank you for your attention!

Model Scales Grid Size = 1/20 x 1/20 Grid Size = 50 m x 50 m Grid Size = 1/10 x 1/10

Numerical Model (3D LES Coupled LS & VOF) Governing equations: Filtered Navier-Stokes equations (LES) W W C + K + Fc = Fv + S τ t Turbulence: Dynamic Smagorinsky-Lilly SGS model. Only large turbulent eddies (equal to or larger than grid size) are simulated, the rest are modelled. Surface capture scheme: Coupled Level Set (LS) and Volume of Fluid (VOF) VOF function, F : LS function, φ : F + u F = t ϕ + u ϕ = t 0 0

3D Navier Stokes solver Basic features: Turbulence: Unstructured grid based, higher-order Finite-Volume scheme (FV) Parallel computations (MPI) Multigrid, Dual-time stepping & implicit residual smoothing Fully implicit schemes available Large-Eddy Simulation Novel Dynamic SGS model for free-surface simulations Fluid-Structure Interaction: Structural dynamics solver integrated Parallel Immersed Boundary Method (IBM) 34

Free Surface capturing Model scheme Newly developed interface capturing scheme for flows with violent free-surface motion Coupling of VOF and Level Set method renders its capabilities that: Can capture the interface in a very sharp manner Can simulate the surface tension effects Can conserve the liquid mass very accurately Both single-phase and two-phase flow can be simulated Using Unstructured Meshes, surf zone simulations with complex geometry can now be considered without any difficulty Benefited from IBM, sea water interacting with offshore structures and the morphologic change of sea bed can be predicted 35

.0 3D Dam Breaking Wave over an Obstacle --- Wave Impact 0.7. 8. 8 14 8 0.161 Obstacle 0.161 P 8 P 7 P 6 P 5 P 3 P 2 P 1 0.04 0.04 0.021 0.04 0.021 0.176 P 4 0.41 x y z Pressure (Nondimensional) 12 7 4 6 10 3 5 8 4 2 6 3 4 21 Simulation Experiment, [24] P5 pressure P3 pressure P7 pressure P1 pressure 2 1 0 0 0 2 4 6 Time (S)

Ensemble Prediction System (EPS) Why use ensemble predictions? In weather forecasts, tiny errors in the initial state will be amplified such that after a period of time the forecast becomes useless. Instead of running just a single forecast, the model is run a number of times from slightly different starting conditions. Global Models: a. Bred vector (breeding) method used by NCEP b. Singular Vector applied by European Centre for Medium-Range Weather Forecasts (ECMWF). Local Area Models (LAM): a. Ensemble derived from global models accounting for uncertainties of large scale b. Regional perturbations accounting for local sensitivity c. Multi-models to represent model errors