a multi site approach to risk assessment for the
play

A multi-site approach to risk assessment for the insurance industry - PowerPoint PPT Presentation

A multi-site approach to risk assessment for the insurance industry Linda Speight Supervisors: Jim Hall, Chris Kilsby, Paul Kershaw linda.speight@ncl.ac.uk HydroPredict, Prague 2010 Outline Introduction Methodology Modelling


  1. A multi-site approach to risk assessment for the insurance industry Linda Speight Supervisors: Jim Hall, Chris Kilsby, Paul Kershaw linda.speight@ncl.ac.uk HydroPredict, Prague 2010

  2. Outline • Introduction • Methodology – Modelling extreme events – Flood defences • Calculating Risk • Conclusions

  3. Catastrophe Models Premium = AAL + Risk Load + Expense Load AIR model output

  4. Aims and Objectives To develop a method for multi-site concurrent damages due to weather related extremes Spatial dependencies at multiple scales FLOODsite 2007 Flood (BBC News)

  5. Methodology Process Statistical Statistical based Multiple modelling of Deterministic modelling of modelling of sites nested flood calculation of extreme water level & in national defence damage events floodplain framework failure inundation

  6. Simulate extreme For each event event across network Simulate crest Estimate peak For each site height and defence flows and reliability hydrographs Hydraulic modelling Hydraulic modelling of of water level and water level assuming flow through breach no breaching Sample sequence of Sample defence failure. breach widths Repeat until Hydraulic modelling for P(breach breach sequence Floodplain sequence) Sample inundation → 0 breach modelling Estimate probability of failure state breach sequences Damage per site Calculate conditional on event damage

  7. Simulate extreme For each event event across network Simulate crest Estimate peak For each site height and defence flows and reliability hydrographs Hydraulic modelling Hydraulic modelling of of water level and water level assuming flow through breach no breaching Sample sequence of Sample defence failure. breach widths Repeat until Hydraulic modelling for P(breach breach sequence Floodplain sequence) Sample inundation → 0 breach modelling Estimate probability of failure state breach sequences Damage per site Calculate conditional on event damage

  8. Modelling extreme events • Aim Y |X, x > u x – Spatial dependency between events Y = set of gauges – Large scale model for UK X = conditional gauge – Good representation of x = daily mean flow extremes u x = threshold • Method Y = a ( x ) + b ( x )Z – Conditional dependence model of Heffernan and a = strength of dependences Tawn (2004) applied by b = changing dependence Keef et al (2009) Z = residuals

  9. Example application 28046 28018 54022 55008 54001 54019 55026 54002 55023 · = observed below threshold o = observed o = simulated

  10. Simulate extreme For each event event across network Simulate crest Estimate peak For each site height and defence flows and reliability hydrographs Hydraulic modelling Hydraulic modelling of of water level and water level assuming flow through breach no breaching Sample sequence of Sample defence failure. breach widths Repeat until Hydraulic modelling for P(breach breach sequence Floodplain sequence) Sample inundation → 0 breach modelling Estimate probability of failure state breach sequences Damage per site Calculate conditional on event damage

  11. Estimating peak flows at site 40 35 30 Flow(m 3 /s) 25 20 15 10 5 0 0 2 4 6 8 10 12 14 16 18 20 22 24 Time (h) Daily mean flow Hydrograph Flood Peak Peak flow conversion Ungauged site transfer methods methods • Shape of hydrograph • Analogue sites • Catchment characteristics • Weighted by distance and catchment characteristics

  12. Simulate extreme For each event event across network Simulate crest Estimate peak For each site height and defence flows and reliability hydrographs Hydraulic modelling Hydraulic modelling of of water level and water level assuming flow through breach no breaching Sample sequence of Sample defence failure. breach widths Repeat until Hydraulic modelling for P(breach breach sequence Floodplain sequence) Sample inundation → 0 breach modelling Estimate probability of failure state breach sequences Damage per site Calculate conditional on event damage

  13. Flood defence system Characterised by • Design standard • Construction High ground type Floodplain Novel aspects • No restriction on d1 d2 d3 length • Consideration of Embankment Defence wall upstream Embankment SOP = 5 years SOP = 100 years SOP = 25 years breaches

  14. Sampling crest heights 7 6,9 6,8 Crest height (m) 6,7 6,6 6,5 6,4 6,3 6,2 6,1 6 0 10 20 30 40 50 60 70 80 90 100 Chainage (m) design crest level defence A - old embankment Defence B - sea wall Autocorrelation function for crest height = f (defence type, age, condition, data quality)

  15. Defence reliability • Initiation – Fragility curves – Condition – Type – Sequencing Fragility curve from Hall et al (2003) Photos from FLOODsite

  16. Sampling breaches 7 1 6,9 0,9 6,8 0,8 6,7 0,7 Elevation (m) P(failure) 6,6 0,6 6,5 0,5 6,4 0,4 6,3 0,3 6,2 0,2 6,1 0,1 6 0 0 20 40 60 80 100 Chainage (m) P(varying failure) crest height water level breaches Autocorrelation function for strength = f (defence type, age, condition)

  17. Breach widths • Growth Rate • Max width – Materials Observed Assumed Case study Source – Floodplain type values values Elbe 20m to 200m Log normal De Kok and – Amount of water Median 20m width mean of Grossmann 64m 2010 Lower Rhine Width 100 – Apel et al – Sequencing 400m 2004 Lower Rhine Width 50 - Kamrath et al 150m 2006 UK RASP Function of load Hall et all method and defence 2003 length Netherlands Largest 520m Muir-Wood wide and 36m and Bateman deep 2005 River Po Normal Width: Govi and 100m - 300m Turitto 2000 Depth: FLOODsite 0.5m - 4m

  18. Simulate extreme For each event event across network Simulate crest Estimate peak For each site height and defence flows and reliability hydrographs Hydraulic modelling Hydraulic modelling of of water level and water level assuming flow through breach no breaching Sample sequence of Sample defence failure. breach widths Repeat until Hydraulic modelling for P(breach breach sequence Floodplain sequence) Sample inundation → 0 breach modelling Estimate probability of failure state breach sequences Damage per site Calculate conditional on event damage

  19. Simulate extreme For each event event across network Simulate crest Estimate peak For each site height and defence flows and reliability hydrographs Hydraulic modelling Hydraulic modelling of of water level and water level assuming flow through breach no breaching Sample sequence of Sample defence failure. breach widths Repeat until Hydraulic modelling for P(breach breach sequence Floodplain sequence) Sample inundation → 0 breach modelling Estimate probability of failure state breach sequences Damage per site Calculate conditional on event damage

  20. Damage • Raster based floodplain inundation model depths • Depth-damage curves damage 3 2,75 2,5 2,25 2 1,75 Depth Metres 1,5 1,25 1 0,75 0,5 0,25 0 -0,25 -0,5 -0,75 -1 0 50 100 150 200 250 300 350 400 450 500 Damage £/m 2 BBC News Penning-Roswell et al (2006)

  21. Risk Risk = Probability x Consequence P(I max,i,j |F i,j |OT i,j ,B i,j , BW i,j |L j,i ,C i,j ,R i,j |Q i,j | X i ) I max,i,j = maximum inundation depth across site j for event i

  22. Risk Risk = Probability x Consequence P(I max,i,j |F i,j |OT i,j ,B i,j , BW i,j |L j,i ,C i,j ,R i,j |Q i,j | X i ) F i,j = Flow over or through P(BW i,j ) = f (type, P(OT i,j ) = f (C i,j , L i,j ) defence material, L i,j ) P(B i,j ) = f (C i,j , L i,j , R i,j , OT i,j )

  23. Risk Risk = Probability x Consequence P(I max,i,j |F i,j |OT i,j ,B i,j , BW i,j |L j,i ,C i,j ,R i,j |Q i,j | X i ) C i, R i = f (type, age, condition) Q i,j = Inflow to hydraulic model X i = Large scale spatial event P(L i ) = f (upstream breaches, Q i,j )

  24. Conclusions • Integrated system model • Detail nested in national scale • Consider risk load – If modelling of flood defences is poor how much impact does this have? – Which areas could flood at the same time / where are we over exposed? Informed • Future... decision! – Sensitivity testing – Resilience measurers

  25. Thank you Linda Speight Civil Engineering & Geosciences Newcastle University, UK Linda.speight@ncl.ac.uk

  26. Modelling extreme events (2) 1. Extreme data at conditional gauge Fit Generalised Pareto • Shape (β) • Scale (ε) • threshold 2. Extreme dependence between gauges Fit dependence model Solid lines: parametric Dashed lines: nonparametric (residuals)

Recommend


More recommend