Gdansk Conference October th, Regional assessments of sea level rise and river floods by computer based expert systems: Dealing with uncertainty J. Kropp, M. Kallache, H. Rust, K. Eisenack Potsdam Institute for Climate Impact Research Structure 1.How to deal with uncertainty in the adaptation discussion? 2.Adaptation to sea level rise: Regional assessments via DIVA 3.River floods assessment, limitations and Chances: The Vistula example 4.Consequences for local adaptation policies 5.Conclusion
Gdansk Conference October th, 2006 Where are our „Achilles heels“: in the economic, natural, and social sense?
Gdansk Conference October th, 2006 Coast Lines Lower Saxony AD Source: Behre 1999
Gdansk Conference October th, 2006 „Miserable Waterflood in Lower-Germany 1717“ At the North sea coast dyke construction since 1100AD Reasons: Maladaptation! mainly landuse
Gdansk Conference October th, 2006 Back to Reality: River Elbe Flood 2002/Pärnu Storm Surge 2005
Gdansk Conference October th, 2006 How to assess “protection level” Wave overflow: Return level right Weser bank for current dyke heights (3900yr) Climate change scenario: average high tide + 70cm +3.8% increase of wind speed (return level 1000yrs). (after Liedermann & Zimmermann 2003)COSTS? Secondary effects?…..
Gdansk Conference October th, 2006 The DIVA Expert System
Gdansk Conference October th, 2006 Initial Settings for DIVA Runs Protection level: 1000yr/return level, storm surge/river flood Dike failure (breach) mode: wave overflow Tidal basin, nourishment: CBA Migration allowed due to changing env. conditions: yes Time steps of calculation: 5 yrs Simulation time: Input SRES scenarios: A1FI („worst case“), B2 („best case“); regionalized SLR scenarios based on PIK‘s CLIMBER model (for each SRES family, low/medium/high-uniform/regionalized)
Gdansk Conference October th, 2006 Regional Sealevel Rise: SRES-A1FI/B2 How large the adaptation costs will be? Best case: B2 Worst case: A1FI
Gdansk Conference October th, 2006 Adaptation Costs: Sea Level Rise (dike construction & preservation, beach nourishment, etc.) Year 2000 Year: 2100
Gdansk Conference October th, 2006 Total Adaptation Costs (Mio US$) A1FI: Most relevant for Estonia due to sandy beaches and no dikes Start-up investments to guarantee 1000yr protection level needed....
Gdansk Conference October th, 2006 River impact length (1000yr flood) A1FI B2
Gdansk Conference October th, 2006 Other Possible Things..... Salinity intrusion costs Sea dike costs River dike costs People actually flooded per storm surge Sand loss Loss of flats Beach nourishment costs Area influenced by salinization due to slr Tidal basin demand for sand nourishment.... Typical expert system which means that usage by stakeholders Needs involvement of experts for simulation runs and interpretation
Gdansk Conference October th, 2006 Improved Flood Risk Assessment Retrospective on river run-offs: assumptions needed, e.g. climate change signal can be found in run-off data (trend = nonstationarity) Main results: No uniform behaviour for rivers worldwide Standard statistics is unsuitable for assessment tasks Adequate analytical procedures can confine uncertainty Examples (annual – stationary, implies no trends!): Odra/Gozdowice ( km 2, Poland) Vistula/Tczew ( km 2, Poland) Daugava/Daugavpils (64500 km 2 Latvia) Nemunas/Smalininkai (81200 km 2 Lithuania) But is this the end of the story?
Gdansk Conference October th, 2006 Trends and Extremes in Time Series Extremes? Definition: A trend is a long-term movement which can be distinguished from oscillation and noise. distinguished from oscillation and noise. x(t) = Trend(t) + Oscillations(t) + Noise(t)
Gdansk Conference October th, 2006 PRUDENCE Comparison 1961/90 – 2075/2100, A2 Hadley Boundary T (°C) P (%) BSR Countries* (model mean) TaTaTaTa T (DJF) T(JJA) PaPaPaPaP(DJF)P(JJA) Denmark Estonia Finland Germany North Latvia Lithuania Poland Sweden North Sweden South
Gdansk Conference October th, 2006 Gauge: Daugava/Daugavapils (JJA) Linear trend in mean and variance (1,1,0 - obtained via model fit routines) Design flood values differ significantly! More torrential rain in summer: regime shift! Kallache/Rust/Kropp 2005: Nonlinear Processes in Geophysics
Gdansk Conference October th, 2006 Bootstrapping for Confining Uncertainty Results: Red: theory, asymptotic fit Grey: bootstrap ensemble 100yr return level9 Estimates for „design flood values“ are too small (6-15% difference!) Rust/Kallache/Kropp 2006: Advances in Water Resources Res., under review Gauge: Vistula/Tczew: Catchment: ~ km 2 Length: Problem: data series too short! 1.Huge model library (more than 50) 2.Define model selection criteria 3.Select best fitting model 4.Generate bootstrap ensemble 5.Perform statistics
Gdansk Conference October th, 2006 My demand: Inclusion of these procedure into the daily practice
Gdansk Conference October th, Improved technique integrated and views can reduce adverse impacts 2. Communities can adapt autonomously only partly, they need help of scientists 4. Planned (anticipated )adaptation measures usually have immediate benefits 6. Adaptive capacity varies considerably among countries, regions and socio-economic groups 8. Enhancement of adaptive capacity is necessary to reduce vulnerability, especially for the most vulnerable (people, regions…) 9. Current knowledge of adaptation & adaptive capacity is insufficient 10. Technical progress is essential for suitable adaptation Conclusion - Main Findings with Respect to Adaptation
Gdansk Conference October th, 2006 Climate Disruptions, Heart Attacks and Market Crashes We need a new science and planning for disasters.... Bunde/Kropp/Schellnhuber, Springer 2002