Natural disasters and man-made tragedies have always been a feature of life on Earth, but the world does seem to have become particularly hazardous of late.

A year ago, an explosion on a BP drilling rig in the Gulf of Mexico killed 11 workers and triggered a leak that spewed more than 4 million barrels of oil into the ocean. As we report on page 152, scientists are still struggling to track the possible long-term effects on the region. Meanwhile, halfway across the world, Japan is reeling from the triple crisis that killed tens of thousands, caused at least US$200 billion in direct damages and continues to spread radiation into the skies and sea.

The two disasters are of different magnitude and have very different causes. But they do have one thing in common: overconfidence in the strength of human systems and decisions.

In the case of the petroleum industry, companies pushed into deep-water locations without doing sufficient research or making the investments needed to prevent — and respond to — problems that might arise from drilling in such an extreme environment. And the US government did not have sufficient oversight of the industry. Both the government and the petroleum companies seemed to think that there was little risk of the type of blowout that caused the accident. “The Deepwater Horizon disaster exhibits the costs of a culture of complacency,” concluded the presidential commission charged with investigating the spill.

Critical systems must be made more resilient.

Similar overconfidence ruled parts of the seismological community in Japan. Official maps of the seismic hazard in the country draw heavily on records of past earthquakes. But because no great earthquake had struck off the coast of Sendai in recent centuries, the hazard assessments did not take such a large event into account (see Nature doi:10.1038/nature10105; 2011, and Nature 471, 556–557; 2011). Some researchers were more cautious. They pointed to geodetic data showing that strain was increasing in the region and to geological signs of a tsunami in 869 that was much larger than anything more recent. But such evidence did not undermine seductive faith in the official quake risk-assessment method, which turned out to be fundamentally flawed. In addition, the designers of the Fukushima nuclear power plant failed to adequately prepare for the possibility that their back-up generators would fail, and misjudged how hard it would be to re-establish electrical power after a tsunami.

One lesson that must be taken from these tragic events is that many of our critical systems are simply unable to withstand situations that are entirely possible. Think the oil and gas industry has learned the lessons of Deepwater Horizon? Don't hold your breath. The rush to exploit the resources exposed by dwindling Arctic ice cover seems riddled with the same dismissal of legitimate risks of rare but plausible events with terrible consequences (see page 162).

Such critical systems — and society at large — must be made more resilient: the core elements of society need to function even when disasters strike. And many fields, ranging from seismic engineering to urban planning for the impacts of climate change, are working to build up this resilience.

In practice, this means strengthening crucial buildings and other infrastructure, developing back-up power systems and planning for multiple tiers of failures. It also requires better training for local communities and government officials in how to respond. Earthquake and tsunami drills are common in places such as California and Japan, but other regions, too, need to engage in these and other types of exercises that simulate massive oil spills, nuclear crises, terrorist attacks and hurricanes, to name a few.

On 28 April, almost two million people in the United States will take part in the Great Central US Shakeout by responding as if a large earthquake had taken place in the New Madrid Seismic Zone, which 200 years ago produced the strongest historical quakes in the conterminous United States. This relatively simple public exercise will be followed up a few weeks later by a more detailed drill involving federal, state and local managers.

Such simulations are essential to probe emergency plans for weaknesses before the real hazards come along, which they will.