After an earthquake destroyed much of Lisbon in 1755, Jean-Jacques Rousseau wrote to Voltaire with a pointed observation: nature had not built twenty thousand houses of six and seven stories there, and if the city's people had been spread out and housed more lightly, the damage would have been far smaller. More than two and a half centuries later, specialists in disaster risk say much the same thing. The ground shakes, rivers rise and storms make landfall whether or not anyone is there. Whether that becomes a catastrophe depends largely on people.

Hazard, exposure and vulnerability

Researchers separate the event from its effects. A hazard is a natural event that could cause harm; it becomes a disaster when it strikes people and places that are exposed and vulnerable. A magnitude 7 earthquake in an empty desert is a hazard; the same quake beneath a crowded city of weak buildings is a disaster. Disaster risk is usually described as a combination of three factors: hazard, exposure and vulnerability. Hazard is how likely and how strong the event is. Exposure is what lies in its path: people, homes, roads, crops. Vulnerability is how easily those things are harmed, which depends on building quality, wealth, health, age, disability, information and whether help can arrive. In 1976 three geographers wrote an influential paper arguing that we should take the "natural" out of natural disasters, since so much of the harm comes from social choices.

Who bears the risk

Because vulnerability tracks poverty, the burden is deeply unequal. Disasters kill far more people in poor countries, while the largest losses measured in money occur in rich ones, where there is more costly property to destroy and better protection for life. Within any country, the poorest, the elderly, people with disabilities and those excluded from information tend to suffer most. Seen this way, reducing disaster risk is part of protecting equal human worth.

Forecasting, not prediction

People often ask why scientists cannot simply tell us when the next big earthquake will hit. No one can predict the exact time, place and size of a particular earthquake, but scientists can forecast the probability of strong shaking in a region over the coming decades, and map which areas will shake hardest. This matters for building codes and land use even if it cannot set a date. Communicating such uncertainty is hard. Before the 2009 earthquake in L'Aquila, Italy, which killed 309 people, an official gave the public a falsely calming message after a meeting of experts. Several scientists were convicted of manslaughter in 2012, then acquitted on appeal in 2014; the case became a warning about how risk is explained.

Floods are described in a similar language of chances. A "100-year flood" is a flood with a 1 percent chance of happening in any given year, not one that comes on a fixed schedule. Two such floods can strike in consecutive years.

Warning signs and memory

A tsunami is a series of huge waves set off by a sudden shift of the seafloor, usually in a large undersea earthquake. When the sea suddenly pulls back far from the shore, it can be a warning that a tsunami is about to arrive. In 2004, a ten-year-old British girl named Tilly Smith, on a beach in Thailand, recognized this sign from a geography lesson; her warning helped clear the beach before the wave struck. Elsewhere that day the story was grim. The 2004 Indian Ocean tsunami killed around 230,000 people, many far from the earthquake, largely because the region had no tsunami warning system. One was set up in the years that followed. In Japan, old stone markers on some hillsides record where past tsunamis reached and urge later generations not to build below them.

Reducing risk, and its traps

The tools are familiar: land-use planning that keeps homes off floodplains and fault lines, building codes, early warnings, insurance, evacuation drills and natural buffers such as wetlands and mangroves. Some protections bring hidden dangers. The American geographer Gilbert F. White, who shaped modern floodplain management, argued in 1945 that while floods are natural, flood losses are largely made by human choices. He also noted what is now called the levee effect. Building levees and flood walls can encourage more building behind them, so when a rare large flood overtops them, the losses are greater.

Worked example: cyclones in Bangladesh

Bangladesh sits on a low, crowded delta facing the Bay of Bengal, where tropical cyclones push storm surges inland. In 1970 the Bhola cyclone killed somewhere between 300,000 and 500,000 people. A cyclone in 1991 killed about 140,000. Cyclone Sidr in 2007, a storm of similar strength, killed a few thousand. The hazard had not changed. What changed was vulnerability: tens of thousands of volunteers in a national preparedness program now carry warnings village to village with megaphones and flags, and thousands of raised concrete shelters give people somewhere to go. Applying this level's ideas, risk fell because exposure and vulnerability fell, even though the storms are as fierce as ever. Much remains to be done, especially as sea levels rise, but few examples show more clearly that a disaster's toll is shaped by human effort.