IMPACT 4 November 20044th IMPACT Workshop - Zaragoza1 Investigation of extreme flood Processes and uncertainty Model uncertainty How uncertain are your.

Slides:



Advertisements
Similar presentations
INCREASING THE TRANSPARENCY OF CEA MODELING ASSUMPTIONS: A SENSITIVITY ANALYSIS BASED ON STRENGTH OF EVIDENCE RS Braithwaite MS Roberts AC Justice.
Advertisements

Bayesian tools for analysing and reducing uncertainty Tony OHagan University of Sheffield.
Assessing Uncertainty when Predicting Extreme Flood Processes.
IMPACT 6-7th November 20033rd IMPACT Project Workshop Louvain-la-Neuve 1 Investigation of extreme flood Processes and uncertainty IMPACT Investigation.
Design of Experiments Lecture I
Desktop Business Analytics -- Decision Intelligence l Time Series Forecasting l Risk Analysis l Optimization.
FAO assessment of global undernourishment. Current practice and possible improvements Carlo Cafiero, ESS Rome, September CFS Round Table on.
Sensitivity Analysis In deterministic analysis, single fixed values (typically, mean values) of representative samples or strength parameters or slope.
EPSRC Grant: EP/FP202511/1 Predicting Breach Formation Mark Morris HR Wallingford The Science of Asset Management London – 9 th December.
PHYSICAL MODELING OF BREACH FORMATION Large scale field tests Kjetil Arne Vaskinn, Sweco Gröner Norway.
Mitigating Risk of Out-of-Specification Results During Stability Testing of Biopharmaceutical Products Jeff Gardner Principal Consultant 36 th Annual Midwest.
Propagation of Error Ch En 475 Unit Operations. Quantifying variables (i.e. answering a question with a number) 1. Directly measure the variable. - referred.
Kinematic Routing Model and its Parameters Definition.
Sediment Movement after Dam Removal
L Berkley Davis Copyright 2009 MER301: Engineering Reliability Lecture 14 1 MER301: Engineering Reliability LECTURE 14: Chapter 7: Design of Engineering.
DETAILED TURBULENCE CALCULATIONS FOR OPEN CHANNEL FLOW
The Calibration Process
Decision analysis and Risk Management course in Kuopio
Determining Sample Size
Gaussian process modelling
Mote Carlo Method for Uncertainty The objective is to introduce a simple (almost trivial) example so that you can Perform.
1 Flood Hazard Analysis Session 1 Dr. Heiko Apel Risk Analysis Flood Hazard Assessment.
VTT-STUK assessment method for safety evaluation of safety-critical computer based systems - application in BE-SECBS project.
WinTR-20 SensitivityMarch WinTR-20 Sensitivity to Input Parameters.
Final Presentation UNSA, Nice HydroEurope 05 th March, 2010.
Fundamentals of Data Analysis Lecture 9 Management of data sets and improving the precision of measurement.
BREACH FORMATION A Review of State-of-the-Art Mark Morris HR Wallingford IMPACT Project Workshop Wallingford, 16/17 May 2002.
Fundamentals of Data Analysis Lecture 10 Management of data sets and improving the precision of measurement pt. 2.
Industry & Research: Issues, needs and conclusions from the US IMPACT Investigation of Extreme Flood Processes & Uncertainty Workshop at HR Wallingford,
A. Betâmio de Almeida Assessing Modelling Uncertainty A. Betâmio de Almeida Instituto Superior Técnico November 2004 Zaragoza, Spain 4th IMPACT Workshop.
TOUS CASE STUDY. MODELLER OVERVIEW REVIEW. COMPARATION AND ANALYSIS OF MODELLING RESULTS 4th IMPACT WORKSHOP 3-5 NOVEMBER 2004.
Formulation of a New Breach Model for Embankments IMPACT Project Workshop, Wallingford 2002 Breach Formation Theme.
Extreme values and risk Adam Butler Biomathematics & Statistics Scotland CCTC meeting, September 2007.
Sensitivity analysis of hydraulic model to morphological changes and changes in flood inundation extent J.S. Wong 1, J. Freer 1, P.D. Bates 1, & D.A. Sear.
Propagation of Error Ch En 475 Unit Operations. Quantifying variables (i.e. answering a question with a number) 1. Directly measure the variable. - referred.
Sensitivity and Importance Analysis Risk Analysis for Water Resources Planning and Management Institute for Water Resources 2008.
Channel Routing Simulate the movement of water through a channel
Channel Routing Simulate the movement of water through a channel
5-1 ANSYS, Inc. Proprietary © 2009 ANSYS, Inc. All rights reserved. May 28, 2009 Inventory # Chapter 5 Six Sigma.
Uncertainty Management in Rule-based Expert Systems
EPSRC Grant: EP/FP202511/1 WP4.4 A Rapid Embankment Breach Assessment (AREBA) Myron van Damme Sept 2011.
IMPACT 3-5th November 20044th IMPACT Project Workshop Zaragoza 1 Investigation of extreme flood Processes and uncertainty IMPACT Investigation of Extreme.
1 Blend Times in Stirred Tanks Reacting Flows - Lecture 9 Instructor: André Bakker © André Bakker (2006)
IMPACT 3-5th November 20044th IMPACT Project Workshop Zaragoza 1 Investigation of extreme flood Processes and uncertainty IMPACT Investigation of Extreme.
IMPACT 5 November 20044th IMPACT Workshop - Zaragoza1 Investigation of extreme flood Processes and uncertainty UniBwM Breach Modelling Dr.-Ing. Karl Broich.
Probabilistic Design Systems (PDS) Chapter Seven.
IMPACT 3-5th November 20044th IMPACT Project Workshop Zaragoza 1 Investigation of extreme flood Processes and uncertainty IMPACT Investigation of Extreme.
ONE-DIMENSIONAL ANALYSIS ON BEDEVOLUTION ACCOMPANING BANK EROSION Satoru Nakanishi Hokkaido University Graduate School Kazuyoshi Hasegawa Hokkaido University.
Statistics Presentation Ch En 475 Unit Operations.
FLOOD PROPAGATION UNCERTAINTY J. Mulet – F. Alcrudo Area de Mecánica de Fluidos, CPS-Universidad de Zaragoza.
Workshop on Disproportionate Costs, 10./ Copenhagen Summary and draft conclusions 11 April 2008.
IMPACT 13 May Investigation of extreme flood Processes and uncertainty IMPACT Investigation of Extreme Flood Processes and Uncertainty.
Flood propagation Simulation of the IMPACT case study on the Tous dam-break flow Université catholique de Louvain Sandra Soares Frazão and Yves Zech.
Rick Walker Evaluation of Out-of-Tolerance Risk 1 Evaluation of Out-of-Tolerance Risk in Measuring and Test Equipment Rick Walker Fluke - Hart Scientific.
Probabilistic Slope Stability Analysis with the
RASP - Risk Assessment of flood and coastal defence for Strategic Planning A High Level Methodology Project partners and co-authors Paul Sayers / Corina.
Beach Modelling: Lessons Learnt from Past Scheme Performance Project: SC110004/S Specific Modelling Tools & Techniques.
Uncertainty Analysis in Emission Inventories
Channel Routing Simulate the movement of water through a channel
The Calibration Process
Channel Routing Simulate the movement of water through a channel
Chapter 11 Simple Regression
RAM XI Training Summit October 2018
Hydrotechnical Design Guidelines for Stream Crossings
Instituto Superior Técnico instituto superior técnico
Numerical Analysis of slopes
IMPACT ~ Breach Formation (WP2)
WLTP CoP Procedure for CO2/FC
Rupro, breach model used by Cemagref during Impact project
WLTP CoP Procedure for CO2/FC
Presentation transcript:

IMPACT 4 November 20044th IMPACT Workshop - Zaragoza1 Investigation of extreme flood Processes and uncertainty Model uncertainty How uncertain are your modelling results?

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza2 Session 19 divides into two sections: This Session (#19): Overview of uncertainty analysis aims & objectives Analysis of uncertainty in breach modelling –Findings and conclusions for breach modelling Mark Morris ~ 25 mins Analysis of uncertainty in flood propagation –Findings and conclusions for propagation modelling Overall conclusions & observations Q&A Francisco Alcrudo ~ mins

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza3 IMPACT: Modelling uncertainty: Sediment Movement [WP4] Flood Propagation [WP3] Breach Formation [WP2] Assessment of modelling uncertainty Implications for End User Applications [WP5] Geophysics / Field Data [WP6]

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza4 Why do this? Dambreak modelling requires a high degree of modelling expertise / experience. Uncertainties in the modelling process are higher than for ‘normal’ river modelling work. Flood risk management may be performed more effectively if additional information is given supporting ‘best estimates’.

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza5 Aims & objectives: Investigate uncertainty within modelling predictions for breach, flood propagation and sediment transport Demonstrate how uncertainty within each of these modelling domains contributes towards overall uncertainty in predicting final conditions (e.g. water levels at specific locations) Consider the implications of the magnitude of uncertainty found in terms of use by ‘end users’

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza6 Scope of work under IMPACT The scope of work under IMPACT: does not allow for an investigation of uncertainty in the impact of flooding or the assessment / management of flood risk does not allow for development of extensive, statistical analysis techniques [Simple, practical assessment of likely order of magnitude of uncertainty]

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza7 Progress to date Understanding the problem Development of approach Recognition of limitations for sediment modelling Selection of case studies Implementation for breach modelling Implementation for propagation modelling Analysis and conclusions  (!)  ?   .. ..

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza8 Development of approach... A slow process that ended with a simple practicable but perhaps non rigorous approach balancing the need to analyse the problem with limitations in time and budget (the real world!) The main arguments: –Fully rigorous statistically based analysis model by model and also linking of models –Removal of subjective - expert judgement type stages –Simple, practicable approach allowing indicative answer within acceptable time / budgets Natural 

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza9 Approach taken... Sediments: Recognition that current ability to model sediment movement under dambreak / extreme flood conditions was limited Progress has been made (Yves Zech / Sandra Soares) but not sufficient to allow assessment of uncertainty

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza10 (Simple) Overview of approach 1Adopt the concept of using upper, mid and lower estimates of modelling results parameter (e.g. flood hydrograph, water level) 2Assess uncertainty in breach model leading to upper / mid / lower flood hydrographs 3Assess uncertainty in propagation models using 3 flood hydrographs from breach modelling as input conditions (leading to nine propagation predictions) 4Select upper, mid, lower estimates for presentation of results to end user

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza11 Some issues Need to balance expert judgement (subjectivity): –What does upper, mid, lower mean? –What model parameters do you vary to create these? –Different models use different parameters - how do you compare these? Run time and flexibility of models dictates approach: –Differences between similar model types (rel small) –Differences between breach and propagation (significant)

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza12 Selection of case studies: Very difficult to find extensive, reliable data sets… –Ongoing problem…forensics team? Two case studies selected: –Tous Dam Failure (breach & propagation) –Lake Ha!Ha! (sediments)

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza13 Uncertainty in breach modelling Basic process: Sensitivity analysis to modelling parameters Selection of top 3-5 parameters; identification of realistic parameter range and distribution Monte Carlo analysis Review and selection of upper, mid & lower hydrographs Modelling undertaken by 3 organisations / 3 different models Selection of overall upper, mid, lower

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza14 Modelling with HR BREACH –1D flow model - piping and overtopping –Soil erosion through shear stress / sediment transport; no predefinition of breach growth –Integrated soil mechanics for lateral and longitudinal discreet slope failure (undercutting etc) –Surface protection simulation –Composite structure and associated failure mechanisms –Variable sediment equations, adjustable probability distribution for slope failure (uncertainty in soils), –Monte Carlo facility

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza15 Sensitivity analysis

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza16 Monte Carlo simulations 5 parameters varied for MC analysis 1600 MC runs undertaken –< than statistical requirement –little difference in results from 3 parameter MC analysis No failure - only overtopping Failure distribution

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza17 All upper-mid-lower results

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza18 Modelling with Rupro (Cemagref) –Aim: providing an outflow hydrograph from piping or overtopping –Simplified assumptions: average breach cross- section, sediment transport using Meyer Peter Müller equation with sediment described by only one representative diameter and one friction coefficient –Advantages: very rapid calculation, few parameters to test can be integrated in 1-D and 2-D propagation models thus cascade breaching or dike breaching easy to calculate –Disadvantages: difficulty to assess parameters in case of complex structure of the dam

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza19 Modelling with Rupro (Cemagref) Uncertainty calculations –Step 1: what are the parameters to which outflow hydrograph is sensitive? –Step 2: select the three main parameters and 5 values for each (very low, low, average, high, very high) to which a probability is associated 18 : 45 /19 : / 13600not limited / 36 m Maximum breach width 21 : 51 / 25 : / / 0.08sediment diameter (mm) 24 : 19 / 21 : / / 20Breach Strickler friction coefficient 21 : 36 / 20 : / / 70Breach bottom elevation (m) Time of peak outflow Peak outflow (m3/s) Values of parameter Parameter considered

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza20 Modelling with Rupro (Cemagref) Uncertainty calculations –Step 3: calculation for all combinations (125) and ranking hydrographs by peak outflows to obtain 5, 10, 50, 90 and 95 % occurrences.

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza21 Breach Model DEICH_P Breach Formation

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza22 Breach Model DEICH_P Breach Formation without core

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza23 Breach Model DEICH_P Breach Formation with core

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza24 Selected upper-mid-lower results

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza25 Some observations… Q p varies between +50% (~22,500m 3 /s) and -17% (~12,500m 3 /s) –BUT, what don’t forget that non failed data has been removed Timing influenced by inflow data, but upper-lower still shows ~2.75hr difference

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza26 Some observations… How do our best estimate results compare to mid values? –Mid  Best estimate

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza27 Some observations… Choice of sediment transport equation worth a closer look… –analysis was prompted by a bug in our MC routine - looking for way to model in homogeneous rather than composite form –considered 2 sediment transport equations (cohesive / non cohesive); core and outer layer material parameters –highlights the importance of selecting the right sediment relationship –highlights the effect of assuming a homogeneous bank

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza28 Some conclusions (breach uncertainty) Uncertainty range for Qp for Tous study was ~ +50% -17% –Note - this includes uncertainty from 3 models / modellers. Range from 1 model / modeller will appear less. Modeller best estimate was better than ‘mid’ estimate (for this case). –Skill of modeller will play a significant part –Routine use of mid value not necessarily the best

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza29 Some conclusions (breach uncertainty) Understanding the sensitivity of models to different parameters is essential –different models will include different parameters –model sensitivity to parameters will vary from model to model All breach models have one or more parameters relating distribution of shear stress / sediment / erosion / breach growth. –This parameter is usually hidden but has a major influence on results –Care is required to validate particular parameter values for each application

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza30 Some conclusions (breach uncertainty) Application of Monte Carlo approach is ‘better’ than sensitivity analysis, but the process requires a quick running model for practicality –Current breach models are on the limit of practicability for MC analysis Current breach models are not well designed to predict the timing of breach initiation –current models are very sensitive to boundary conditions (ie defined flow, water level etc) rather than actually predicting the onset of breach

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza31 Some conclusions (breach uncertainty) Breach model results are highly dependent upon choice of sediment transport equation –No existing equations are ideal (steady state, long term…) –selection of most appropriate equation for case application should be done (look at conditions that equation was developed for)

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza32 Some conclusions (breach uncertainty) Simulation of composite or homogeneous structures and assumptions of averaged soil properties can significantly affect results –Variations in Qp of >300% can be seen when different assumptions / sed equations are made (compared to +- 50% for earlier uncertainty range) –Therefore use appropriate model (composite or homogeneous) & appropriate equation. Be wary of assuming homogeneous for a composite structure...

IMPACT Investigation of extreme flood Processes and uncertainty 4 November 20044th IMPACT Workshop - Zaragoza33 Beyond IMPACT… Awaiting DSIG work with interest FLOODsite - Task 20 “Development of framework for the influence and impact of uncertainty” “…develop an approach and prototype software for propagating uncertainty through integrated flood risk models…” –University Bristol (UK) & IHE (Delft) under theme lead from HR - > 60 mm research –