Risk Assessment, Modeling and Decision Support

The first years of the 21^st century brought horrific loss of life and property from earthquakes and tsunamis worldwide. Briefly, the world focused on international disaster prevention, response and recovery. Terrorism loomed large as well, after 9–11, leading to the creation of the Department of Homeland Security in the United States, and a plethora of related efforts globally. Many of these focus on the built environment.

Seismology and earthquake engineering have rich histories, perhaps due to the fact that earthquakes tend to affect almost everything around us — after all, one can take shelter from a storm, but not from an earthquake. Great minds — Aristotle, Plato, Da Vinci and Kant to name a few — have grappled with the problems of earthquakes. When issues of risk are added to those of earthquakes, the field and history are further enriched (think of adding Pascal, Fermat, Bernoulli, Keynes, etc.), and the challenge of writing a history increased commensurately. The task is daunting — Housner (1984) observed:

“Earthquake engineering is a 20th Century development, so recent that it is yet premature to attempt to write its history ... Although 1984 is too soon to write a definitive history, it is an appropriate time for an historical view of earthquake engineering development to see where we were, where we now are, and where we are going...”

This writer agreed entirely when he heard those words in 1984, but two more decades later the perspective is greatly improved, at least as regards seismic risk assessment. In fact, as we shall see, in many ways we were just on the verge of seismic risk analysis and assessment in 1984, whereas today we have very significant capabilities.

Each of the following authors brings his own distinguished viewpoint on the history and future directions for [seismic] risk assessment. Robert Whitman focuses on the lessons learned from regional seismic risk assessments, including the need for clearly defined objectives for successful outcomes; accounting for user needs in loss estimates; assessing uncertainty; including lifelines in analyses; and developing a meaningful inventory prior to analysis. Amr Elnashai extends the history of scientific inquiry into earthquakes back to the 1600’s and fills in some important events, while emphasizing how Europe, Japan, and the United States interacted in the development of seismic research and building codes. Finally, Dennis Mileti urges practitioners to take sociological perspectives into account, as losses that extend beyond traditionally assessed risk are harder to quantify, but no less important.

In the early 1990s, the Federal Emergency Management Agency (FEMA) and the National Institute of Building Sciences (NIBS) embarked on an initiative to develop a consistent, standardized approach to estimating losses from earthquakes. In 1997, FEMA released the first HAZUS^® (Hazards U.S.) model for earthquakes to “provide state, local, and regional officials with the tools necessary to plan and stimulate the efforts to reduce risk from earthquakes and to prepare for emergency response and recovery from an earthquake” (FEMA and NIBS 2003).

The commentaries in this chapter represent three important perspectives: academic research in earthquake engineering; the federal agency that sponsors and oversees HAZUS^®; and expert users. Rachel Davidson fills in some gaps in the background and suggests directions for the development of HAZUS^®, including improving its mitigation planning capability and structural inventory data, as well as providing for web-based user overrides based on post-catastrophe observations. Phil Schneider presents several examples of user-driven developments in HAZUS^® and the role of users in pushing forward improvements and enhancements that are currently in the works, by members of the self-established user/developer consortia. Finally, Subrahmanyam Muthukumar, an expert HAZUS^® user, confirms Schneider’s proposition that user input based on experience with applications of the software provide significant pull for its development. Muthukumar suggests several improvements to HAZUS^®, including more dynamic inventory data, better cost-benefit analysis capability, and lifeline/network models integrated into the overall system.

Earthquake loss estimation — although around since the early 1970’s — has not been considered a decision-making tool until recently. Because early models were based on expert opinion with limited hand calculations, they lacked the ability to perform sensitivity studies (which could provide important insight into drivers of loss potential) or the ability to produce results in a timely fashion. The results of these early studies were generally displayed on maps and/or in the form of tables that listed potential losses to a limited number of exposure categories, e.g., residential, commercial or industrial.

Several broad themes emerge in this chapter to point the way for strategic directions in the development of HAZUS^® and other tools for estimating losses from disasters. These include expanding and enhancing data types, data sources and data collection schedules. New empirical data collection efforts — including network, engineering and building data, and measures of social and economic changes — have the potential to strengthen loss estimation models and metrics. Network data, data on non-economic losses, and additional input from sociologists and social models are seen as essential elements, currently missing from existing methodology.

Decisions about the level of resources appropriate for seismic risk mitigation must be made in the context of uncertainties about seismic risk. This is true if uncertainties are quantified and is also true if they are ignored, for ignoring uncertainties does not make them go away. Critical uncertainties are related to the level of seismic hazard (the amplitudes of ground motion and their frequencies of occurrence) and to the vulnerabilities of facilities to seismic shaking. With a complete description of these uncertainties, complete seismic risk curves (loss vs. annual frequency of occurrence) can be developed. These seismic risk curves can form the basis for informed decisions about seismic risk mitigation.

This chapter provides a dialogue between the risk engineering and risk analysis communities, particularly as it pertains to communicating risks to decision-makers. Clearly, there is a conflict between the ability of engineers and risk analysts to perform necessary quantitative analyses, and the ability or willingness of decision-makers and construction professionals to use these analyses in practice. Society faces significant tradeoffs when allocating scarce resources to mitigating risks, and the information required to generate well-informed risk analyses is costly.

Over the past two decades, as the cost of natural disasters has skyrocketed in the United States, much emphasis has been placed on mitigating these hazards. From earthquakes to floods to hurricanes, agencies such as the Federal Emergency Management Agency (FEMA), the Institute for Business and Home Safety (IBHS), and the Association of Bay Area Governments (ABAG) have tried to increase public awareness of natural hazard risk and encourage those at risk to mitigate for the hazard. At the same time, advanced software tools have emerged to allow stakeholders to more accurately assess their hazard risk exposures. These include the proprietary catastrophe models used by the insurance and reinsurance industries and the federal government’s model, HAZUS^®.

In this chapter, four commentaries from experts in planning, engineering and sociology who specialize in natural hazard planning and recovery provide additional depth and breadth to the topic of modeling seismic mitigation strategies. Laurie Johnson draws on her experience in the catastrophe modeling field to advocate for a more holistic approach to risk management, in which estimating risks and losses are part of a larger process of ongoing risk management and reduction with shared responsibility between public and private entities. Mary Beth Hueste brings a structural engineering perspective to mitigation techniques and points out the nonmonetary aspects of natural hazard losses. Rob Olshansky provides existential support for mitigation modeling as a tool to help decision-makers and the public to understand the benefits of investing in sometimes costly mitigation efforts. Finally, Yang Zhang develops a matrix of the potential interactions between technical analysis and stakeholder participation in making mitigation decisions and discusses the consequences of neglecting either aspect of these components of model development.

“Presenting data without error rate is misleading”. This is a quote from the O.J. Simpson defense team regarding the presentation of DNA evidence without valid error rate statistics. Taken more generally, this practice is a prevalent shortcoming in the scientific and information visualization communities where data are visualized without any indication of their associated uncertainties. While it is widely acknowledged that incorporating auxiliary information about data, i.e. data quality or uncertainty, is important, the relative amount of work in this area is small. On the other hand, developments by the geographic, cartographic, and GIS communities in this regard is much more concerted. Some of the early efforts were spearheaded by the participating members of the National Center for Geographic Information and Analysis (NCGIA) initiatives (Beard et al. 1991; Goodchild et al. 1994), where different methods of displaying and animating data with uncertainty were proposed. An excellent summary of this body of work can be found in (MacEachren et al. 2005). Combining these works with those from the information and scientific visualization communities, a typology for uncertainty visualization was presented which tries to map data, uncertainty, and tasks with the appropriate visual presentation (Thomson et al. 2005). Specifically, the typology for uncertainty visualization would give the user some guidance about the visual representations for the different types of uncertainty.

This chapter focuses on the perspective of consumers of visual information about natural hazards — particularly decision makers and individuals who must make difficult personal choices in the face of vast quantities of data that have been produced for other purposes. Ellen Peters draws on empirical studies to show that “Less is More” in the presentation of information. She also argues that visualizations should be tailored to specific decision needs by choosing precision levels and visual cues appropriate to the evaluation context. Ann Bostrom expands the discussion on the definitions of risk and uncertainty and discusses how spatial information in particular is perceived by users. She emphasizes the particular visual tools that have been shown to be effective in affecting individuals’ judgments about certain risks. Finally Susan Cutter discusses the strengths and weaknesses of various visual approaches to representing uncertainty, and recommends keeping the overall presentation simple.

The analysis and mitigation of the risk posed by natural and other hazards has been developed over several hundred years. The last 20 years have seen a significant leap forward in our ability to model the effects of earthquakes on the built environment. This rapid progress in risk modeling is based upon significant advances in engineering, and information and sensor technologies that have resulted from significant research investments by the National Science Foundation in the United States and sister agencies in Japan and Europe. Currently available risk models enable planning and mitigation of natural hazards in ways that were not previously possible. The U.S. federally funded multi-hazard HAZUS^® model is the most widely known example of this history. HAZUS^® and similar risk modeling tools are poised to play an increasingly important role in catastrophe management and risk mitigation. The papers and discussions at the Boulder workshop highlighted several strategic directions that are emerging in the field of risk analysis and modeling. How these directions develop will determine the shape of risk modeling over the next decade. Strategic investment in specific areas can reinforce and accelerate one or more of these directions.