Decision Support for Extreme Weather Impacts on Critical Infrastructure

Transcription

1 Decision Support for Extreme Weather Impacts on Critical Infrastructure B. W. Bush Energy & Infrastructure Analysis Group Los Alamos National Laboratory Research Applications Laboratory and Computational & Information Systems Laboratory National Center for Atmospheric Research Presentation to the North American THORPEX Societal and Economic Research and Applications Workshop 15 August 2006

2 Critical Infrastructure & Societal/Economic Impacts Historically, weather-related events have created the greatest damage among incidents involving critical infrastructures. Many societal and economic impacts stem from their dependence on critical infrastructure. Critical infrastructure often mediates the impact of extreme weather upon society and the economy: it may either exacerbate or mitigate consequences. Many stakeholders All levels of government, particularly emergency managers Insurers & re-insurers Infrastructure owners/operators Markets / traders / brokers End users (individuals or businesses)

3 Weather/Climate Impacts on Critical Infrastructure THORPEX SERA Sectors Emergency Management Food Production Energy Health Services General Public Private Sector Providers Leisure Tourism Water Management Transportation US DHS Critical Infrastuctures High winds Extreme convection Extreme precipitation Ice storms Hurricanes Flood-inducing storms Fire weather Cold snaps Heat waves Drought high impact moderate impact Climate change Infrastructure Impacts

4 Existing Major DSSs for Multiple Critical Infrastructure Decision Support System User Description Critical Infrastructure Protection Decision Support System U.S. Department of Homeland Security Integrated analysis of interdependencies and cascading effects among infrastructures, and their potential economic and societal impacts. Critical Infrastructure Protection Modeling and Analysis Project Australian Attorney- General s Department Integrated analysis of interdependencies and cascading effect among infrastructures. National Infrastructure Coordination Center National Infrastructure Simulation and Analysis Center Visualization and Modeling Working Group U.S. Department of Homeland Security U.S. Department of Homeland Security U.S. Department of Energy Continuous assessment of the operational status of national critical infrastructures; rapid, real-time analysis of critical infrastructure incidents. Planning and rapid-response analyses of select critical infrastructure sectors, with an emphasis on physical, demographic, and economic impacts. Quick-turnaround studies of events potentially impacting energy production, transmission, and distribution systems.

5 Decision Support Example: VMWG Quick-Response Process Analyses performed 3-5 days prior to estimated landfall. experimental

6 Example Products for Hurricane Katrina (2005) Forecast Wind Damage Electric Power Damage Forecast 2 days 4 days 7 days 14 days 21 days 28 days 45 days 60 days Electric Power Restoration Forecast

7 Categories of Critical Infrastructure Decision Support Pre-event Contingency planning Security & reliability improvement Near event Independent assessment & verification Mid-event Situational awareness Post-event Consequence assessment Event reconstruction Recovery & restoration operations Deployment of protective & assistive forces

8 Some Characteristics of Institutional Decision Makers General styles of decision maker: intuitive vs. quantitative. Intuitive decision makers generally do not want uncertainty information presented directly to them: they rely on analysts to fold uncertainties into their guidance. Quantitative decision makers often want to probe or follow a chain of reasoning to show which effects are important; they value probabilistic information. Infrastructure decision-support product lines have many layers of users. The presentation of results are typically simplified and abstracted for higher levels of authority. Institutional considerations constrain decision support. Decision makers often have a preference for analyses based on officially sanctioned meteorological products, instead of unofficial, experimental, or incidental products. Emphasis on consistency among analyses is often paramount.

9 Risk Attitude, Uncertainty & Utility Multi-attribute utility 100 theory can sometimes 95 aid in characterizing 90 and visualizing complex, 85 multi-dimensional decision problems involving 80 uncertainty. 75 Relative Preference for Alternative (Risk Neutral Profile) Risk Neutral 0.0% 0.5% 1.0% 1.5% 2.0% Annual Likelihood of Incident ~1 in 13 & 5 years Relative Preference for Alternative (Risk Averse Profile) Risk Averse 0.0% 0.5% 1.0% 1.5% 2.0% Annual Likelihood of Incident Relative Preference for Alternative (Risk Prone Profile) Risk Prone ~1 in 50 & 20 years ~1 in 5 & 4 years 0.0% 0.5% 1.0% 1.5% 2.0% Annual Likelihood of Incident

10 Identify high-impact weather forecasts How can we effectively use ad hoc ensembles in forecasting infrastructure impacts of extreme weather? What decision-theoretic framework is best suited for this synthesis? How can infrastructure impact forecast uncertainty based on such ensembles be visualized and communicated in a statistically sound manner? Can ensemble generation be tuned to better support specific infrastructure, economic, and societal impact models? What are the specific NWP model requirements for inputs to particular infrastructure impact models? For example,... At what point do uncertainties in impact models overwhelm the effects of improved forecasts of hurricane landfall location and intensity? Can NWP models represent the time variation of winds (gusts, downbursts, microbursts, vortex breakdown, tornados, etc.) adequately for first principles infrastructure damage models?

11 Develop advanced forecast verification measures Historical event reconstruction to correlate wind, precipitation, flooding to observed damage for various infrastructures has lagged far behind the ever-growing sources of verification data. Multi-way comparisons of meteorological analysis, NWP model outputs, observed damage, and forecast damage would provide valuable guidance for model calibration, algorithm development, and uncertainty quantification. A fuller understanding of available verification data will drive nextgeneration impact model development and define effective limits for algorithm development. Verification measures are specific to individual impact sectors. Data are available for detailed case studies.

12 Develop new user-specific weather products Can impact intensity scales similar to the Saffir-Simpson Scale be developed for infrastructure, economic, or societal metrics? This can be presented geospatially as risk maps. Is it possible to develop a unified framework and methodology for the diversity of user-specific extreme-weather-impact products? Linking infrastructure impact models more closely to extreme weather/climate models, and using the growing body of historical statistics for weather-related infrastructure damage (as correlated to meteorological fields), can result in infrastructure damage forecasts with significantly higher accuracy out further in the future, and such improvement is of keen interest to decision makers. Probabilistic and ensemble information can be propagated directly through the whole modeling process, from the NWP models to the impacts models.

13 Data & Infrastructure Needs Consistent and formal use of established standards (especially for real-time data) will greatly facilitate DSS research and development: metadata access protocols Rich-metadata catalogs of consistent impact verification data set are essential for detailed verification studies: meteorological analysis tuned for impact verification impact surveys (physical damage, economic effects, demographic impacts, evacuation, etc.) in standardized formats consistency between meteorological and impact data sets Modular software components enable rapid development of user-specific DSS products: cross-forecast (weather or infrastructure) comparison and visualization ensemble management/manipulation decision-theoretic data processing