Society For Risk Analysis Annual Meeting 2017
Session Schedule & Abstracts
* Disclaimer: All presentations represent the views of the authors, and not the organizations that support their research. Please apply the standard disclaimer that any opinions, findings, and conclusions or recommendations in abstracts, posters, and presentations at the meeting are those of the authors and do not necessarily reflect the views of any other organization or agency. Meeting attendees and authors should be aware that this disclaimer is intended to apply to all abstracts contained in this document. Authors who wish to emphasize this disclaimer should do so in their presentation or poster. In an effort to make the abstracts as concise as possible and easy for meeting participants to read, the abstracts have been formatted such that they exclude references to papers, affiliations, and/or funding sources. Authors who wish to provide attendees with this information should do so in their presentation or poster.
|Chair(s): Cameron MacKenzie firstname.lastname@example.org
Sponsored by Applied Risk Managment Specialty Group
W2-G.2 10:50 am Storm surge-based flood risk in coastal Louisiana: impacts of Louisianaâ€™s 2017 coastal Master Plan and methods for uncertainty propagation. Johnson DR*, Fischbach JR, Kuhn K; Purdue University and RAND Corporation email@example.com
Abstract: Louisiana’s 2017 Comprehensive Master Plan for a Sustainable Coast is a $50-billion set of 124 projects intended to reduce flood risk and land loss across the coastal zone over the next 50 years. The plan allocates $19 billion to structural flood protection projects and $6 billion for nonstructural measures such as home elevations and floodproofing. A series of systems models was used to evaluate flood risk in terms of the expected annual damage (EAD) with and without projects, under current conditions and at three future time periods (10, 25, and 50 years into the future). Future time periods were evaluated in multiple scenarios with varying assumptions about factors such as sea level rise, future storm characteristics, population growth, and system fragility. The Coastal Louisiana Risk Assessment model (CLARA) uses a variety of methods for uncertainty propagation to generate integrated confidence bounds around estimates of flood depth and damage exceedances, as well as EAD. Sources of uncertainty addressed include the small observed sample of historic storms, variability in levee and floodwall overtopping rates, response surface predictions of surge and wave behavior, noise in hydrodynamic models and digital elevation models, and geospatial correlations in surge and wave heights. CLARA strikes a balance between tackling the complexity of the physical system and its attendant uncertainties, while maintaining computational efficiency that allows for exploration of risk in the wide range of future scenarios needed to inform policy-making and project selection and prioritization. We discuss these new, state-of-the-art methods for modeling uncertainty in flood risk and place them in the context of top-line results regarding future flood risk in coastal Louisiana and the benefits of the 2017 Master Plan.
W2-G.3 11:10 am Quantitative Risk Analysis in a Multirisk Scenario of Natural Hazards. Bronfman NC, Cisternas PC*, Gonzalez D; Universidad Andres Bello and National Research Center for Integrated Natural Disaster Management firstname.lastname@example.org
Abstract: The frequency and magnitude of hydrometeorological events worldwide has been increasing. According to the Global Climate Risk Index for 2017, Chile ranks 10th in the list of the most affected countries by climate change. In addition, Chile’s location in the Pacific Ring of Fire derives in a high seismic activity. In the past 60 years, 33 earthquakes of magnitude over 7.0 in the Richter scale have occurred, seven of which have generated tsunamis. The main goal of this work is to develop a quantitative risk analysis of natural events under a multi-risk scenario to support public decision-making about land-use planning and preparedness measures to strengthen the resilience of the potentially affected communities. The coastal city of Chañaral in northern Chile was selected as case of study due to the massive landslide by rainfall event in March 2015 and its exposure to seismic activity. For the multi-hazard scenario, earthquakes, tsunamis and landslides by heavy rainfall were assessed. The results showed that a large part of the population lives at the landslides flow and near the coast. This means that those inhabitants have the highest level of total risk. Although landslides have caused more deaths in the study area in recent years, individual risk is higher for tsunamis than for landslides. The foregoing highlights the relevance of considering multi-risks scenarios in the design of preparedness and mitigation strategies. These results also provide valuable information for the development of acceptability and tolerability criteria for natural hazards.
W2-G.4 11:30 am learning from imbalanced data sets for estimating power outages. Kabir E*, Guikema S; University of Michigan email@example.com
Abstract: Storms regularly have substantial, wide-spread impacts on power systems, posing many risks and inconveniences to the public. While these impacts cannot generally be prevented, utilities can prepare for the post-storm restoration process to more quickly recover from the event. Having accurate prediction of power outages based on the weather forecast before the storm event can help utility companies to make better decisions regarding their restoration procedures, particularly the number of external crews to request. In many situations the power outage data are highly zero-inflated; that is, the number of zeroes are significantly higher than one would expect based on standard probability distributions. With zero-inflated data, standard statistical models struggle to appropriately model the data and make accurate predictions. In our study, we examine several statistical learning theory models for imbalanced power outage data. We show that modern statistical learning theory methods, particularly two-stage methods, can offer strong predictions with zero-inflated data. Such zero-inflated data is common in risk analysis, where (thankfully) there are far more non-events than events.
[back to schedule]