Establishing Common Nomenclature, Characterizing the Problem, and Identifying Future Opportunities in Multihazard Design

Strong interest in extending the service life of critical infrastructure, compounded by the severity of damages during major disasters, such as Hurricane Katrina in 2005 (FEMA 2006) and the Tohoku earthquake in 2011 (NILIM and BRI 2011), has triggered a growing interest in design concepts that account for cascading effects and the interaction of multiple hazards. Traditionally, design is focused on the effects of various kinds of individual single hazards. In today’s structural design practice, the impacts of various single hazards are translated into equivalent forces. Modern design codes account for concurrence and combinations of multiple hazards by suggesting load combinations and load factors intended to include uncertainties and significance of different hazards. A structural system that is designed to resist maximum load effects is expected to survive the damaging effects of multiple hazards. In recent years, other concepts such as displacement-based design and performance-based design were developed for hazards such as earthquakes. However, current design philosophies fail to consider the complex and intertwined effects of multiple hazards at system-wide and societal levels. Some of the shortcomings of the current design philosophies in reflecting the complex nature of multiple hazards can be identified as follows:

In designing for the effects of multiple hazards, several other deficiencies may be present in current design practice. The concept of superposition of different hazards cannot always accurately predict the risk of damage. The effects of multiple hazards, acting concurrently or over time, can significantly increase the damaging impact of individual hazards. Therefore, an explicit multihazard design is necessary to achieve robustness and resiliency (as further defined later) at a large scale. Multihazard design requires an in-depth understanding of the nature of various hazards and their interactions. It must also include the effects that the hazards have on one another and on the behavior of structures or physical components of a system.

Design for multihazard mitigation is a multifaceted and complex challenge that may prohibit the development of a unified approach. This forum paper is not intended to provide guidelines and/or recommendations for multihazard design, but aims to present ideas and provoke future discussions on this subject. Ideally, these future discussions will ultimately lead to the development of a general framework for multihazard design. Such a framework should be practical, inclusive of different hazards, applicable to various types of structures and design objectives, easily extensible, and not in stark contrast to current design guidelines. Specifically, this paper discusses (1) the classification of various hazards and their possible effects, (2) potential interactions between individual hazards, (3) the importance of a probabilistic framework for multihazard risk assessment, (4) the need for a multihazard design framework, and (5) the current knowledge gaps in the area of multihazard mitigation. Discussions around this topic may raise awareness in designers and stakeholders that current methods may not satisfy the demands of a multihazard reality.

Establishing a common nomenclature of multihazard design is critical for effective communication in the technical community. The current literature lacks a uniform use of terminology and classification for possible interactions of hazards. These interactions are referred to as cascades, cascading effects, cascading; chains; coincidence of hazards in space and time; coinciding hazards; compound hazards; coupled events; cross-hazards effects; domino effects; follow-on events; interactions; interconnections; interrelations; knock-on effects; multiple hazards; synergic effects; and triggering effects (Kappes et al. 2012). In addition, some studies do not specify if the discussed interactions are between hazards, as with tsunami and earthquake, or are through impacts of hazards on physical components and systems, as with corrosion damage and earthquake. The classification of hazards is not consistent throughout the literature because different criteria are used; this greatly hinders the ability to develop a universal design framework.

In addition to defining hazard interactions, it is important to present a clear definition of the objective of multihazard design. In the current literature, design for multihazard resilience, design for multihazard robustness, and design for multihazard mitigation are terms that are being used interchangeably. Although this terminology is essentially used to communicate the same concept, particular attention should be paid to preventing ambiguity. The term resilience is defined as an ability to recover from or adjust easily to misfortune or change; robustness implies a capability to perform without failure under a broad range of conditions, or a property of allowing the severity of damage to be minimized in other instances; and mitigation means reducing the severity of a negative action or effect. In a robust system, the goal is that damage is minimized or prevented in the first place; however, in a resilient system, some level of damage is anticipated, but an additional objective is that the system should recovers efficiently. Mitigation may be achieved through design for resilience or robustness of a system or by diminishing the damaging effect of the hazards themselves, independent of their impact on a system. Mitigation measures are geared more toward lowering global risks than toward provisions for robustness and resiliency, which may mainly focus on passive improvement of system responses to hazards.

This section aims to classify various hazards, both natural and man-made, in the context of their effects at a physical component-, system-, and societal-level. This multilevel approach reduces the complexity of multihazard design and facilitates understanding of possible hazard interactions. Infrastructure may be defined as systems of integrated and interrelated physical components enabling function and growth of a society. Examples of infrastructure systems include roads and bridges of transportation systems; electric power generation units and distribution grids of power systems; and water treatment and distribution facilities of water supply systems. Realizing the impacts of different hazards on physical components of a system and the system itself is necessary for evaluating multihazard risk and developing mitigation methods.

In recent years, several studies have attempted to develop a framework for natural interaction of a broad range of hazards such as earthquake and tsunami with hurricane and storm surge (Gill and Malamud 2014). In contrast, this forum paper characterizes hazards in terms of their impacts and not their nature—that is, intensity and spatial and temporal scales. This paper does not address direct impacts of hazardous events on human health such as toxic fumes during a fire, which are outside the scope of the current discussion. Different hazards may cause the following four inherently different, but possibly interacting, effects.

Site effects essentially define the hazard and represent the actions produced by the hazard at a given location. They are to be understood in the context of causality of the physical impact. The existence of these effects does not depend on the presence of a physical component or a system of physical components. For example, the site effects of an earthquake include ground vibrations and liquefaction/ground settlement, which are independent of the existence of the physical component. Similarly, site effects of a tornado are excessive loads, uplift loads, and flying debris. Hazards that are caused or triggered by another hazard are not listed as site effects. For instance, a tsunami resulting from an earthquake is not considered as a site effect of the earthquake hazard; instead, it is regarded as an additional hazard.

Physical impacts are modifications of the behavior and/or function of a physical component and are assumed to be directly caused by one or multiple site effects associated with hazards. They are not necessarily independent or mutually exclusive. For example, the physical impacts of a flood hazard include foundation/support damage and change in material properties of a structure’s elements (physical components) because of the hazard’s site effects—that is, water overflow/accumulation and corrosive chemicals such as saltwater.

Network and system disruptions are defined as interrupting or impairing effects on the function of a system or network at large scales. In the case of a significant storm surge, inundation of subway tunnels may cause a significant interruption of the public transportation system.

Social and economic consequences recognize the role that affected structures and infrastructure systems play in societal functioning and human behavior. In the case of a tsunami, extensive damage to infrastructure may cause a mass migration of a local population.

Fig. 1 shows different hazards in a way that aids visualization of their effects and promotes understanding of their possible interactions through site effects and physical impacts. It presents possible main and subsidiary site effects and physical impacts of hazards (dark and light gray, respectively). The main effects/impacts are defined as those that are more probable and more damaging, whereas the subsidiary effects/impacts are those that are less probable or less severe. This differentiation between main and subsidiary effects/impacts is based mainly on the engineering judgment of the authors on the basis of their review of past damage reports. The site effects are categorized as transient or perpetual in terms of temporal scales. The physical impacts are categorized as a change in characteristics or as impaired function. In Fig. 1, changes in characteristics are essentially the possible impacts of a hazard on a structure. Impaired functionality is an occurrence that limits or interrupts a structure’s intended function and performance. Changes in characteristics may directly result in impaired functionality, or they may alter a structure such that its response to other hazards becomes different from that of the structure when intact. It should be noted that hazards that are induced by other hazards are not considered effects for the purpose of Fig. 1—for example, high wind and blizzard. The hazards causing potentially significant network- or system-level disruptions and social and economic consequences are identified in the figure footnote.

The authors have taken care when differentiating extreme events and hazards. Extreme events essentially consist of several hazards. For example, a hurricane is an extreme event that consists of high winds, heavy rain, and storm surge, among several other hazards. Another example of an extreme event is a thunderstorm that consists of heavy rain or hail and lightning hazards. Therefore, in Fig. 1 extreme events are not listed as hazards; instead, only the constituent hazards are listed. In the figure, flood/flashflood hazard may include river, coastal, or inland flooding; a high wind hazard may include both synoptic and nonsynoptic winds and hurricane-type winds. The hazards themselves are listed alphabetically and not ordered by significance or prevalence. It is acknowledged that consideration of different hazards in design may vary according to the structural or infrastructure system of interest.

This section identifies possible modes of interaction between various hazards. Understanding interactions between hazards and their effects is necessary to properly evaluate the risk associated with multihazard environments. The multihazard risk may not be properly assessed by superimposing the risks of the individual hazards. For example, for a mainshock-aftershock situation in an earthquake event, the probability of damage due to the combination of the two hazards is not equivalent to the summation of the probabilities calculated for two independent earthquake events. The reason is that the mainshock causes damage and changes a structure’s dynamic characteristics. The aftershock then acts on the previously damaged structure. Thus, combining the level of damage obtained from independent analyses of the undamaged structure under the mainshock and the aftershock does not yield the correct result. In this situation, the structure can be analyzed under a ground acceleration that includes both mainshock and aftershock in a single time series or by subjecting the structure, already damaged by the mainshock, to a suite of seismic records appropriately representing the aftershocks. For a realistic assessment of multihazard effects, the possible interaction of different hazards should be realized at two levels: (1) the nature of the hazards and (2) the effects of the hazards, named here Level-I and Level-II interactions, respectively.

Based on the terminology proposed by the Performance-Based Earthquake Engineering (PBEE) framework developed by the Pacific Earthquake Engineering Research Center (PEER), for a single hazard it is customary to express the risk associated with a specific damage level in terms of mean annual rate of exceedance of the specified damage level, ${\lambda }_D$. This means that the annual rate of exceedance can be obtained by the convolution integral given by (Porter 2016; Jalayer and Cornell 2003)where $im$ = specific realization of a scalar intensity measure $IM$ (i.e., a parameter measuring the intensity of the hazard action); $F(im)$ = cumulative distribution function of the specified damage state conditional to $IM=im$ (also known as fragility function); and $G(im)$ = mean annual frequency of the intensity measure $IM$ exceeding $im$ (also known as hazard curve). Therefore, $-dG(im)/dim$ represents the mean number of events per year corresponding to $IM=im$. It is noteworthy that the fragility and hazard functions are traditionally developed for intact structures and single hazards, respectively.

Because of the complexity of interactions between different hazards, the risk of damage due to multihazards cannot be obtained by simply superimposing or combining the risk of damage due to several hazards considered independently. In a multihazard context, hazard functions can be multidimensional, i.e., hazard surfaces, because of the natural interaction of multiple hazards—that is, Level-I interactions. These hazard surfaces should be developed to account for possible interactions of hazards as joint probability density functions or joint cumulative distribution functions of multiple intensity measures describing different hazards (Wang and Rosowsky 2013). In addition, fragility functions can change as a result of interaction among the physical impacts of different hazards—that is, Level-II interactions. For example, to correctly capture the risk of a specified level of seismic damage to a bridge structure in a corrosive coastal environment, the effects of gradually deteriorating concrete and steel material in columns need to be considered (Zhong et al. 2012). This goal may be achieved by analyzing structures with different levels of corrosion damage under seismic excitation to find the corresponding fragility functions. In practice, the parameters of fragility functions need to be adjusted to incorporate the impact of corrosion. In more complex cases, the joint hazard model is required. An example for such cases is a stability analysis of retaining structures during a hurricane, where the interaction of flooding, high wind, and storm surge should be considered.

Multihazard risk and loss assessment requires the development of appropriate fully probabilistic frameworks that can account for the complex interaction among hazards. This interaction can affect structural and infrastructural behavior not only because of the correlation of different intensity measures describing different hazards but also through hazard chains and degradation/changes in structural system properties when subjected to multiple hazards. It is noteworthy that a first attempt to develop such a framework wascarried out in hurricane engineering, in which a performance-based hurricane engineering (PBHE) framework was proposed that takes into account the interaction among different concurrent subhazards—wind, wind-borne debris, storm surge, and rain (Unnikrishnan and Barbato 2016; Barbato et al. 2013). As the understanding of hazard interaction and multihazard impacts on fragility and hazard functions expands, it will become apparent that there is a need for more advanced probabilistic models (e.g., multiparameter marginal probability density functions, high-dimensional joint probability density functions, and more informative statistical data such as high-order statistical moments) that will better describe both marginal distributions of intensity measures and correlation coefficients between different intensity measures. Until more studies are performed on multihazard risk and loss assessment and until more data are collected to probabilistically describe the interaction among different hazards, expert judgment will play a critical role in practical assessments and in the derivation of fragility and hazard functions.

Multihazard mitigation can be approached from different perspectives. For the purpose of this forum paper, the focus is on mitigation strategies for physical components such as infrastructures. Traditionally, added strength and ductility are known as effective mitigation measures for single hazards. Given the varied and complex nature of impacts resulting from hazards, this traditional approach may not be an effective mitigation strategy. On the other hand, guidelines that overcomplicate or transform current design methodologies may not be easily adopted or accepted by the engineering community. In addition, the present limited understanding of multihazard effects hinders the development of design guidelines that rigorously account for these effects. In general, the physical response to multiple hazards varies from structure to structure; thus, adequate design adaptations should consider the diverse nature of structures and/or materials used.

One innovative approach that might effectively mitigate multihazard effects is reducing the level of hazard interaction, possibly by decoupling the physical impacts of different hazards. This strategy could be used to reduce a complex, multihazard environment that explicitly considers interactions among multiple hazards down to a simplified, single-hazard environment that indirectly considers interaction among different hazards through the use of load combination coefficients and factors. This approach might allow designers to continue using existing design provisions that are essentially developed for multiple hazards acting independently. Decoupling could be achieved in different ways depending on the type of interaction—for example, the Level-I and Level-II interactions discussed earlier. When hazards naturally interact, actual decoupling of their effects may be infeasible. For example, the interaction of mainshock and aftershock cannot be eliminated. However, designing a more robust structure can lessen the damage magnification effects typical of aftershocks and thus serve as a mitigation measure for mainshock-aftershock interaction effects. Where hazards interact through their effects, the decoupling of these effects could effectively minimize multihazard damages and losses. A large number of strategies may be devised to achieve this goal which could be as simple as protecting a structure against corrosive environments, thus preserving its capacity to resist any other hazard. Minimizing the interaction of multiple hazards may require creative solutions, which may not yet be developed or not used routinely in structural/infrastructural design.

Multihazard design (as defined here, in contrast to simpledesign for multiple hazards) is a relatively new topic in engineering, leaving considerable room for future research. An underlying knowledge of the interrelations among multiple hazards is key to the development of multihazard design guidelines and next-generation multihazard robust structural materials and systems. Although not exhaustive, the following list of the most pressing needs in the field of multihazard design is proposed:

This forum paper introduced the concept of multihazard design, differentiating it from the more traditional concept of design for multiple independent or noninteracting hazards. The main goal was to stimulate future discussions to advance this newly proposed concept. Shortcomings of current methodologies were acknowledged to identify the potential focus points of future multihazard design guidelines. The need for common terminology was established to enable effective communication across the multihazard technical community. Multihazard risk assessment and design should consider four types of effect: (1) site effects, (2) physical impacts, (3) network and system disruptions, and (4) social and economic consequences. Hazards were classified in the context of their effects. Hazard interactions, through their nature or through their effects, were identified and presented to further understanding of multihazard complexity. Possible risk and loss assessment concepts and physical component mitigation strategies were suggested to promote further discussion. Finally, significant knowledge gaps in multihazard assessment, mitigation, and design, along with future research needs, were briefly discussed.

The authors would like to thank the contribution of the members and friends of the Multi-hazard Mitigation Committee of ASCE’s Structural Engineering Institute for their critical contribution to the development of this forum paper.