1.7 - Disaster trends and multiple hazards

Reading long-term disaster records

A disaster record measures an event's interaction with society, so a change in recorded disasters is not automatically a change in tectonic activity.

Three impact series answer different questions. Deaths indicate mortality, although definitions may include or omit indirect deaths. Numbers affected include people needing assistance, injured or displaced under the database's rules. Economic damage measures monetary loss under a stated accounting convention. They can move in different directions: warning may reduce deaths while population and asset growth increase the number exposed and the rebuilding bill.

From 1960 onwards, expanding communications, monitoring and international databases increased the visibility of disasters. A raw chart often shows more recorded events, but older small or remote events were more likely to be missed. The underlying frequency of earthquakes cannot be inferred from that rise. Climate change can influence interacting rainfall hazards and exposure, but a weather-disaster trend is not evidence that plate movement has accelerated.

Mortality is highly uneven through time. The 1960 Chile earthquake generated an ocean-wide tsunami; later peaks include the 2004 Indian Ocean tsunami, the 2010 Haiti earthquake and the 2011 Japanese earthquake–tsunami. Moving the endpoints of a decade can move a mega-disaster into or out of it and reverse the apparent direction of change. A short quiet interval is not proof that risk has been removed. NOAA's 1960 account documents fatalities even across the Pacific in Japan.

Overall disaster statistics also require disaggregation. The CRED–UNDRR report for 2000–2019 found that floods and storms dominated the recorded event count. Geophysical disasters represented about 9% of events but 59% of recorded deaths; earthquakes, including tsunamis in that grouping, carried much of the mortality. Its economic-loss grouping was led by meteorological disasters. Thus the most frequent hazard category was not the most deadly. These are the report's historical database shares, not predictions for every future period. CRED–UNDRR, 2020, printed pp9 and 19.

The same report's printed p6 gives this overall historical comparison, excluding biological and technological disasters:

Recorded deaths changed relatively little while affected totals and real monetary losses rose. This describes those database totals, not a controlled measure of physical hazard change. The pre-2000 reporting caution below limits inference, and overall totals cannot stand in for a tectonic-only series.

Recorded economic losses have grown over long periods, influenced by inflation, greater asset values, urban growth and major costly events. Compare constant-price values and loss relative to GDP before concluding that physical hazards intensified. More people affected can reflect expanding settlement, repeated counting of the same people or better reporting. Neither series is a direct measure of individual vulnerability.

Interrogating data reliability

Interrogating a large dataset means checking how records were produced before treating its totals as evidence.

EM-DAT records disasters at country level under inclusion criteria. An ocean-wide tsunami can generate several country records; counting those rows as separate physical tsunamis exaggerates event frequency. Its documentation identifies time, accounting and geographical biases, and strongly recommends excluding pre-2000 records from statistical trend analysis. Older records remain useful historical evidence when their limitations are explicit. EM-DAT specific biases, checked 2 October 2026.

Use a repeatable procedure:

1. State the hazard classes, region, date window and impact definitions. Decide whether an earthquake and its tsunami are one event or linked records.
2. Inspect missing values, duplicates, revised estimates and country boundaries. Blank damage is unreported damage, not zero.
3. Separate hazard occurrence from exposure and impact. For comparable years, calculate population-adjusted mortality and constant-price loss where possible.
4. Examine annual variation, median and mean, then longer windows. Report the influence of extreme years rather than deleting inconvenient disasters.
5. Test whether the conclusion changes with another time window, classification or reliable source. State uncertainty and what remains unknown.

Here is an original illustrative five-year mortality series, not real global data: 1,000; 1,500; 2,000; 2,500; 43,000 deaths. The mean is $50,000/5=10,000$ deaths per year; the median is 2,000. The extreme final year supplies 86% of all deaths and pulls the mean far above a typical year. Removing it would conceal a real part of risk; reporting both measures communicates the distribution more honestly. A moving average smooths annual noise but can spread one catastrophic spike across several years.

Economic data are often less complete than death data. Insured assets are more visible than uninsured homes or informal livelihoods, and estimates can change after surveys. A total with more complete recent reporting can appear to rise even if like-for-like loss has not. Triangulate database records with scientific event catalogues, national assessments and dated local accounts; each source answers a different part of the question.

Mega-disasters and interdependence

A tectonic mega-disaster can transmit consequences far beyond the area physically struck through connected populations, supply chains and policy decisions.

The 26 December 2004 Indian Ocean tsunami followed a magnitude 9.1 earthquake near Sumatra. The waves affected many countries around an ocean basin. WHO reports more than 227,000 deaths. Coastal housing, fishing equipment, water supplies and tourism were damaged; international rescue and reconstruction linked donors, governments and communities. The hazard crossed borders directly, while lost livelihoods and interrupted travel extended the economic effects. WHO tsunami evidence and NOAA significant events.

Eyjafjallajökull's April 2010 eruption demonstrates a different mechanism. Winds dispersed ash into European air routes; ash could damage jet engines, so authorities closed affected airspace. EUROCONTROL's retrospective reports around 104,000 cancelled flights. Travellers, freight, tourism and businesses dependent on timely air transport were affected far from Iceland. This was large network disruption without comparable direct mortality. The scale of economic impact depended on atmospheric transport, safety decisions and the organisation of international aviation, as well as the eruption. EUROCONTROL, 2021 retrospective.

Japan's March 2011 tsunami flooded Fukushima Daiichi, disabling essential power and cooling and leading to a nuclear accident. The resulting displacement and contamination concerns prolonged impacts. Safety reassessment and reactor shutdowns changed electricity supply; greater reliance on imported fossil fuels created energy-security and emissions implications. This describes the aftermath, not an assertion that Japan permanently abandoned nuclear power. IAEA accident report, 2015 and METI energy-security analysis, 2024.

Multiple hazards in the Philippines

A multiple-hazard zone contains several potential hazards, and some combinations create effects greater than their separate impacts.

The Philippines lies in an active tectonic setting and a tropical-cyclone region. Earthquakes, volcanic eruptions and tsunamis can overlap with intense rainfall, river flooding and slope failure. Coexistence alone does not prove that one hazard causes another. Identify whether hazards are independent, share exposed infrastructure or interact physically.

At Pinatubo in June 1991, Yunya's rain wetted ash during the eruptive crisis. Wet ash imposed a larger load on roofs; roofs that could not support that load collapsed. This combined an eruption product, weather and construction vulnerability. Rain also mobilised loose volcanic deposits into lahars, fast flows of water and debris channelled down valleys. Their pathways differ from the ashfall footprint, so a settlement outside a pyroclastic-flow zone can still be threatened downstream. Later wet seasons could remobilise deposits after the climactic eruption ended. USGS ash and rain evidence and PHIVOLCS retrospective, 2021.

The disaster can also cascade through services: a lahar cuts a bridge, supplies cannot reach shelters, and illness or lost income extends harm. Planners therefore need rainfall monitoring, valley-risk maps and alternative routes alongside volcanic alerts. A shelter selected to avoid a lava flow still needs a safe roof, water supply and access during a storm.