The Risk Assessment Process: The Role of Catastrophe Modeling in Dealing with Natural Hazards

Abstract

Probabilistic risk analysis has long played an important role in engineering design for natural hazards. For example, the lateral loads imposed by hurricanes or earthquakes, and characterized by a specified probability of exceedance, are used by structural engineers to design buildings that minimize injuries and fatalities. More recently, these techniques have been extended to estimate the damage to existing building inventories and, ultimately, to estimate the economic and insured losses that result from the occurrence of natural catastrophes. Catastrophe loss estimation techniques, known collectively as catastrophe modeling, have gained widespread acceptance by the insurance and risk management industries and are now heavily relied upon to support a wide range of financial decisions. A probabilistic approach to catastrophe loss analysis is the most appropriate way to handle the abundant sources of uncertainty inherent in all natural hazard related phenomena. As pointed out in Chapter 2, the relative infrequency of catastrophe events results in a scarcity of historical loss data. Hence statistical techniques used by actuaries for estimating future losses stemming from automobile or fire insurance policies, for example — techniques that rely on a wealth of available claims data — are not appropriate for estimating future losses from natural catastrophes. Furthermore, the usefulness of the limited historical loss data that do exist cannot be easily extrapolated to estimate the economic impact of disasters because of the ever-changing landscape of properties. Property values change, as do the costs of repair and replacement. Building materials, design and practice change along with building codes. Therefore new structures may be more or less vulnerable to catastrophe events than existing ones. 46 While it is generally agreed that the probabilistic approach is the most appropriate, it is highly complex and multifaceted. It requires modeling complex physical phenomena in time and space, compiling detailed databases of building inventories, estimating physical damage to various types of structures and their contents, translating physical damage to monetary loss and, finally, summing over entire portfolios of buildings. From the modeler's perspective, the task is to simulate, realistically and adequately, the most important aspects of this very complex system. Risk managers need to familiarize themselves with the underlying assumptions of the models and understand the implications and limitations of their output in order to utilize the results effectively. Briefly, the hazard component of catastrophe models estimates the probability that the physical parameters that define the hazard will exceed various levels. In the case of earthquakes, for example, the model estimates the probability that parameters such as peak ground acceleration or spectral acceleration (defined as the maximum acceleration experienced by a simple oscillator, used as a representation for building response) will exceed various levels at a particular site. The model's vulnerability component deals with the potential for the hazard to damage structures and their contents. It estimates the probability that building damage will exceed various levels as a result of ground motion. The loss module translates physical damage into monetary loss and estimates the probability of exceeding various levels of loss. Together, the hazard and vulnerability modules comprise what is traditionally known as probabilistic risk analysis. This approach to modeling earthquake risk is based on the pioneering work of Cornell (1968) and is now well established in the literature. Catastrophe loss models can be thought of as one application of probabilistic risk analysis, characterized by their refinement of the financial loss estimation component. The final result of the catastrophe model, commonly used in financial analysis, is the exceedance probability, or EP, curve introduced in the preceding chapter. At each stage in the process, the model takes into consideration the uncertainty in the various parameters that describe the model. All catastrophe models require substantial amounts of data for model construction and validation. In addition, the reliability of such models depends heavily on our understanding of the underlying physical mechanisms that control the occurrence and behavior of natural hazards. While no one would claim to have a complete understanding of all of the intricacies of these physical systems, scientists and engineers, aided by increasingly sophisticated instrumentation and computing capabilities, have accumulated vast amounts of information and knowledge in these areas. By incorporating this information and knowledge, the sophisticated theoretical and empirical models currently being developed can reasonably simulate these complex phenomena.