Classification of Natural Disasters

Introduction

The question of how to define a disaster and which criteria should be applied to classify it has been the subject of vigorous debate among practitioners of the field (MunichRE, 2006; Perry and Quarantelli, 2005; Quarantelli, 1998). For example, Berren et al. (1980) offer an independent and comprehensive classification that is not limited to natural disasters and is based on type and duration of disaster, magnitude of impact, potential for occurrence, and ability to control the impact. Other classification schemes consider the differentiation by magnitude of event or consequences, by the different scales (such as individual, family, community, and region), or by speed of onset and predictability. Hence, numerous classification schemes have been proposed, and little would be gained from reviewing them all here.

Despite these reservations, there is broad consensus that a disaster is an event or situation that severely disrupts normal socioeconomic activities and causes damage and possibly casualties. Attempts to quantify the definition, for example, in terms of monetary losses and numbers of people killed (Foster, 1976; Keller et al., 1992; Munich Re, 2006), have not met with universal acceptance. Nonetheless, it is clear those disasters there is a qualitative difference between disasters and lesser events, in that they require extraordinary responses in terms of resources and organization (Kreps, 1983). A common definition of a disaster is that the coping capacities of the affected individual, group or unit (local, regional or national governments, public institutions, social groups, etc.) are exceeded and external support is likely to be required. Hence, it may be appropriate to base the classification of the magnitude of emergencies and contingencies upon ability to cope with and respond to events of a given size (Table 1).

Classification of Natural Disasters, Table 1 A size classification of emergencies and contingencies (Partly after Tierney, 2008)

Three global data sources for natural disasters are available. Two are data catalogs compiled by insurance companies: the Sigma database of SwissRe and NatCatService of MunichRE. However, the most widely used data bank on disasters is the OFDA/CRED International Disasters Database (EM-DAT, refer also to www.em-dat.net), maintained at the Centre for Research on the Epidemiology of Disasters (CRED) of the Catholic University of Louvain in Belgium. Besides temporal information, all entries are arranged by continent, country, and theme, as requested by the UNISDR Secretariat.

It has long been noted that the term “natural disaster” is not particularly apt (O’Keefe et al., 1976). For example, although most earthquakes are entirely natural phenomena, the root cause of seismic disasters could be regarded as poor construction of buildings rather than the occurrence of ground shaking. Hence, there is a motive for regarding earthquakes as human-made disasters. In fact, because so much of the impact of disasters depends upon vulnerability, a predicament that mainly depends on human decision making, “natural” disaster can be regarded as a convenience term which distinguishes one class of phenomena from others. In this case the generating mechanisms stem directly from events in the geosphere, biosphere, atmosphere, and hydrosphere.

“Natural” disasters are caused by extreme events, in the sense of large departures from long-term mean values. For instance, sudden excesses of precipitation can cause floods, whereas long drawn-out shortages can result in drought. In this respect, speed of onset and duration are important criteria in classifying events. Earthquakes and rapid debris avalanches are examples of sudden-impact disasters, whereas drought and desertification or soil erosion are examples of slow-onset events. Most earthquakes have a main shock that will last from a few tens of seconds to a couple of minutes, but the sequence of aftershocks can stretch the emergency period to hours or days. This contrasts with a drought that may be prolonged for months or years and desertification that is essentially a permanent condition, i.e., one that is technically challenging and expensive to reverse. The typology suggested by the U.S. National Research Council’s Committee on Disaster Research in the Social Sciences. (US NRC, 2006) is similar to the discussion presented here, but also includes the scope of impact.

Among extreme natural phenomena there is a wide variety of speeds of onset and duration, and in turn a large variation in predictability and potential for warning. For example, tsunamis are generated abruptly by the sudden displacement of a column of ocean water. Triggers may be earthquake activity, submarine landslides, or meteorite impact. However, the very long distances that tsunamis travel allow monitoring to take place and warnings to be issued to distant coastal areas in their path. This has been successfully applied, for instance, in the tsunami generated by the Chile earthquake on February 27, 2010, and the respective precautionary response along the Western Pacific coasts including Japan and New Zealand. For the Pacific basin warning lead times may exceed, at most, 17 h, but the main problems occur with “near-field” tsunamis (those that are generated locally) in which even instant detection does not allow more than a few minutes’ warning to be issued to local communities. However, one should never forget that the population at risk from a local tsunami can often recognize earthquake shaking as an environmental cue indicating a need to evacuate to higher ground (McAdoo et al., 2009).

The prospect of short-term prior warning of earthquake main shocks has long been a goal for seismologists, but has proved consistently elusive, mainly because each earthquake involves some degree of complex uniqueness. Hence, most seismic disasters occur without prior warning, other than the long-term identification of areas at risk and recurrence intervals of earthquakes of a particular maximum size.

The predictability and warning potential of volcanic eruptions is highly variable. Heat fluxes, harmonic tremor, and gas emissions all indicate the rise of molten magma close to the Earth’s surface, but the exact timing of eruptions tends to defy prediction. In the mid-1980s volcanic emergencies occurred in the Caribbean and southern Italy that lasted months without any actual eruptions. In contrast, many extreme events of an atmospheric or hydrological origin have a higher degree of predictability. The preparatory phenomena can be observed by direct measurement (e.g., rain gauges and streamflow monitoring) or remote sensing (synoptic views of storms), and numerical modeling can give forecasts of pending events.

Drought is the archetype of a slow-onset, or “creeping” disaster, a category that includes desertification (the degradation of land productivity) and accelerated soil erosion. This sort of phenomenon tends to be insidious. It may go undetected until the impact is chronic, and thus the state of disaster is defined by the cumulative sum of effects.

Recurrence interval and regularity are two further elements in the classification of natural disasters. Although many anthropogenic phenomena are nonrecurrent (for example, transportation crashes and catastrophic pollution episodes), most extreme natural events are repetitive. The degree of regularity, or cyclicity, depends on the type and origin of the phenomenon. Meteorological and hydrological events tend to be the most cyclical on account of seasonality. In South Asia, monsoon-induced rains cause summer flooding; in the eastern central Atlantic Ocean the general circulation of the atmosphere leads to a hurricane season that extends from May to November; in the Pacific basin the El Niño-Southern Oscillation (ENSO) causes a 4-year cycle of floods and storms; and in the European Alps, large magnitude snowfalls often associated with temperature changes cause snow avalanches that tend to have their peak occurrence in particular months of the year (e.g., slab avalanches in deepest winter and slush avalanches in spring).

Many extreme atmospheric phenomena recur on complex cycles. Annual seasonality provides the shortest of these, whereas sunspot occurrence, fluctuations in the Earth’s ionosphere, and trends in climate change provide others. Earthquakes tend to have definable cycles based on the gradual accumulation and sudden release of strain in the Earth’s crust. On the San Andreas Fault the occurrence of high-magnitude earthquakes has been established by carbon dating of exposed faults, and other indications, as averaging once in 160 years. However, variations in the recurrence of seismic events can require confidence intervals that may be more than 50% as large as the cycle itself. The situation is even more indeterminate for volcanic eruptions, where the intervals between events may be very much longer than human timespans, in some cases as much as 10,000 years.

Regarding earthquakes, the picture may be complicated by other contributory factors. For instance, some research indicates that precipitation events may trigger large earthquakes (see Huang et al., 1979). The large amount of water made suddenly available adds weight to the earth’s surface over a relatively short-time span. The stress–strain field in the Earth’s crust can thus change very rapidly. If there are stable conditions, precipitation will have no influence on earthquakes, but if a region is already weakened then additional loading through precipitation may very well have an influence. The same is true of marine tides and “Earth tides” the pull of the Sun-moon system on the Earth’s crust.

The case of river flooding illustrates the difficulty of using recurrence intervals in planning and preparedness. Following the practice in the USA, many countries use the 100-year flood as a benchmark. This is an event that has a probability of occurrence of 100% in a century and 1% in any single year. It usually corresponds to a defined floodable area and a set of depths of inundation, both of which can be expressed on maps. Convenient as the 100-year flood is, there is no guarantee that it will be the most significant or most disastrous event. Neither will it necessarily occur after a 100-year interval without major floods. The dilemma for land-use and emergency planners is what size of event should be used for preparation.

Probability distributions of diverse kinds of natural events were given by Hewitt (1970). Most of them were deemed to follow the magnitude-frequency rule, in which the larger the event, the smaller its probability of occurrence during any given interval of time. With normalization using logarithms, this can be reduced to a straight-line on a graph. However, the degree of predictability tends to fall with larger events for which there are limited or no data because the time spans are longer than those of existing measurements. Current thinking (Blöschl and Zehe, 2005) suggests that larger, less frequent events may be more significant than recognized in the past. This is the so-called “fat-tailed distribution” problem, in which large events are overrepresented in probability distributions, inducing major disasters to be more common than expected. This may be reinforced by the future tendency of climate change to increase the intensity, if not the frequency, of extreme meteorological events.

To some extent the complexity of extreme natural events defies classification. A good example of this is furnished by landslides (mass movements). Classifications (e.g., Cruden and Varnes, 1996) have commonly been based on the mechanism and speed of movement, with particular attention to flowing, sliding, falling, toppling, gliding, and creeping, and to a range of speeds that extends from infinitesimally slow to hundreds of kilometers per hour. Other classifications have considered the morphology of the phenomenon or its lithological setting, although perhaps with less success, and yet others have taken into account whether movement is primary, dormant, reactivated, or relict. Clearly, a perfect classification, if such were possible, would have to utilize sets of information on a wide variety of geological, mechanical, kinematic, and environmental factors. The first difficulty is to define boundaries on continuous phenomena, for instance, in speed of movement, which in most cases can only be arbitrarily divided up. Additionally, speed may vary within one large landslide body. The second, and overwhelming, problem is that the majority of mass movements in natural slopes are composite events, for example, flow-slides or avalanche-slide-falls. Slumps, in particular, may be reactivated paleolandslides, which further adds to the complexity. Moreover, many mass movements result from the underlying lithological complexity, particularly the juxtaposition of permeable and impermeable strata and the inclination of the respective layers. These factors tend to make classification somewhat artificial and to prevent it from being definitive.

In classification perhaps the broadest distinction is between natural disasters such as earthquakes and floods, technological disasters such as transportation crashes and toxic spills, social disasters (e.g., riots and crowd crushes), and intentional disasters (conventional and CBRN terrorism – see Showalter and Myers, 1994 and Steinberg et al., 2008). There is, of course, plenty of opportunity for overlap, as in the so-called “NaTech” disasters, which have natural origins and technological effects. For instance, reservoir dams can be affected by floods, earthquakes, landslides, avalanches, or siltation. Indeed, as human impact is a prerequisite for an event to become a disaster, all natural catastrophes are to some extent NaTech events.

Like other forms of disaster, recurrent natural events can be considered in terms of the “disaster cycle” (Figure 1), with its phases of mitigation, preparation, emergency response, recovery, and reconstruction. The cycle has been criticized on various grounds. Not all events are cyclical, not all cycles are regular, and the phases may overlap or be superimposed, rather than be sequential (Neal, 1997). Three major issues are involved when dealing with the cycle. Firstly, mitigation and preparedness should be concurrent, not sequential as indicated in the figure. Secondly, some parts of the community may enter the recovery phase while others are still in response, leading to geographical discrepancies in the application of the cycle. Thirdly, mitigation is more likely to be undertaken if it is integrated into the recovery rather than constrained to follow it (Lindell et al., 2006). But despite these criticisms, the cycle has proved to be a robust and useful model, both in training and in the organization of disaster risk reduction (DRR) activities. Clearly, the duration of the cycle, and also the relative length of its phases, depends on the duration, geographical extent, and seriousness of the disaster impact. A major catastrophe may affect tens of thousands of square kilometers and hundreds of thousands of people. Recovery from it may take decades.

Classification of Natural Disasters, Figure 1

The disaster cycle.

The “disaster cycle” has been lightly revised by Dikau and Weichselgartner (2005 – see Figure 2). In the new version countermeasures taken either in terms of supporting structures or land-use planning may alter or change the effects of the disaster impact.

Classification of Natural Disasters, Figure 2

The revised disaster cycle (Revised by Dikau and Weichselgartner, 2005).

Summary

A natural disaster is an extreme event, caused by a natural phenomenon that has severe adverse impacts on human lives and livelihoods. Such events result from natural processes in the atmosphere, hydrosphere, biosphere, or geosphere. They can be characterized by their type of process, speed of onset (from instantaneous to long drawn-out), duration (from seconds to years), predictability, potential for warning, and scope of impact. A further element in classification is cyclicity and recurrence: events vary from unique to highly cyclic depending on their origin and on factors such as seasonality or the buildup of crustal strain.

The global sequence of natural disasters is irregular, but it shows a trend toward increases in the number and size of events of all kinds. It is dominated by natural processes such as floods, storms, and earthquakes, while others such as tsunamis or landsliding are increasing in importance as the vulnerability to them of human settlements and activities increases. Clearly, the impact of a natural event cannot be a disaster unless humans are affected. If a high-magnitude natural process develops into a disaster, this is not only an expression of its event characteristics, but also a reflection of the social context. Hence, any classification scheme for natural disasters should consider both the natural environmental and social dimensions.

Cross-references

Casualties Following Natural Hazards

Civil Protection and Crisis Management

Disaster Risk Management

Disaster Risk Reduction (DRR)

Disaster Relief

Historical Events

Integrated Emergency Management System

Resilience

Risk Assessment

Risk Governance