The resilience test

This is article 1 of 3 in the Resilience Test series.

In five weeks in summer 2026, earthquakes hit Venezuela, Japan and the Naples area. Magnitudes ranged from catastrophic to modest. Death tolls did not track the seismographs. They tracked buildings, institutions and decisions. The world already has disaster frameworks and a UN promise of early warning for everyone by 2027. This investigation asks how that system performed when the ground actually moved.

Three shocks, one examination

On 24 June 2026, two large earthquakes struck northern Venezuela about 39 seconds apart: magnitude 7.2, then magnitude 7.5. By early August the official death toll had risen above 6,000. On 28 July a magnitude 6.8 earthquake (Japan Meteorological Agency magnitude 7.1) hit Kumamoto Prefecture. By 8 August the prefecture's toll stood at 39 dead, including cases still being assessed as disaster-related. About an hour after the shaking, an explosion tore through Aeon Mall Kumamoto after reports of a gas leak. On 31 July a magnitude 4.7 tremor west of Naples, at about three kilometres depth, injured about twenty people, damaged buildings in Pozzuoli and put hundreds outdoors overnight. It was the strongest quake recorded in Campi Flegrei in about forty years.

The energy released in the three events was not comparable. The outcomes still refuse to line up with the physics. The weakest shaking of the three raises the hardest question about readiness.

This is not an argument that earthquakes are dangerous. It is a test of the system built, on paper, for exactly these nights: frameworks, warning networks, evacuation plans, and a United Nations initiative with a deadline.

What exists on paper

After the December 2004 Indian Ocean earthquake and tsunami killed roughly 228,000 people, many of them without any warning, warning centres and coastal sirens spread. In 2015 almost every country adopted the Sendai Framework for Disaster Risk Reduction, a fifteen-year plan to 2030 with targets on deaths, people affected, economic losses and early warning coverage. It is a framework, not a treaty with sanctions. It is still a shared, measurable commitment.

In 2022 the UN Secretary-General launched Early Warnings for All: every person protected by an early warning system by the end of 2027. The UNDRR/WMO status report for 2025 found that 119 countries (60%) now say they have a multi-hazard early warning system. Two years before the deadline, four countries in ten still report none. Small island developing states sit at 43%. The Americas and the Caribbean, Venezuela's region, sit at 51%, the lowest regional coverage in that report.

A warning is the cheapest layer of readiness: sensors, software, a message. The expensive layers are buildings that stand, institutions that function, and decisions taken before the ground moves. No global initiative has promised those by 2027. The Sendai midterm review in 2023 already said the world was off track on most targets.

That is the scoreboard. Then came the examination.

Venezuela: the structural layer

On the evening of 24 June the ground broke first near Yaracuy, then off the coast toward La Guaira. The US Geological Survey lists the pair as a shallow doublet about 39 seconds apart. La Guaira took the worst of the structural damage. Official tallies kept rising as debris was cleared. By early August National Assembly figures put deaths above 6,000, with hospital treatments in the tens of thousands.

Nobody died for lack of a warning that could have arrived between those two ruptures. No message helps when the mainshock follows in under a minute and the building was never built to survive either shock. Venezuela is a Sendai signatory. The signature was real. Decades of economic stress, weak institutions and informal construction on steep coastal ground meant readiness had existed mainly as documents. The earthquake did not invent that gap. It measured it.

In a prepared economy, disaster loss is spread across insurers, budgets and years. In an unprepared one, the loss is subtracted from the future: homes, savings, another push toward migration. The same shaking becomes a repair bill in one place and permanent poverty in another. That is the structural layer. It is where most of the world's earthquake deaths still happen. There is no 2027 target for it.

Japan: the ceiling of preparation

Japan is what long implementation looks like: building codes revised after disasters, school drills, dense seismometer networks that can brake trains and stop elevators before strong shaking arrives.

On 28 July, Kumamoto Prefecture took a shallow quake near dense urban areas. Warnings fired. Buildings, overwhelmingly, stood. By 8 August the prefectural toll was 39, of which 36 were counted as direct earthquake deaths, with further cases assessed as disaster-related, including heat. That is a fraction of what comparable physics produced in Venezuela. Preparation showed.

It did not end there. About an hour after the shaking, an explosion tore through Aeon Mall Kumamoto. Customers had been evacuated after the quake; some staff were still inside or had returned. Authorities investigated a gas-related blast. The secondary hazard that kills after buildings stand opened inside a prepared system.

Manufacturers in the region paused plants while they checked damage and suppliers. The shaking map was local. The consequence map ran through supply chains on other continents.

Summer heat sat on households without power and on people living in cars and shelters. Japan has learned, including after the 2024 Noto peninsula earthquake, that in a prepared country many deaths are not caused by falling buildings. They arrive later. Japan has a category for it: disaster-related death. Preparation moved dying from the first minute toward the following weeks. It has not abolished it.

That is the ceiling. Funded readiness converts thousands of deaths into dozens. It does not convert them into zero. The remaining failures sit in the hour after, the week after, and the systems inside the systems.

Naples: the decision layer

On 31 July at 19:46 Italian time, Italy's National Institute of Geophysics and Volcanology recorded magnitude Md 4.7 at about three kilometres depth in Campi Flegrei. In Japanese terms, a footnote. Locally it was the strongest tremor in about forty years. About twenty people were injured. Buildings and streets in Pozzuoli were declared unsafe. Landslides fell. Transport paused. Aftershocks kept people outdoors.

Campi Flegrei is not a simple volcano cone. It is a caldera about thirteen kilometres across, with western districts of Naples inside it. The ground there rises and falls (bradisismo). Italian Civil Protection states that uplift since 2005 exceeds 150 centimetres, with stronger seismicity in recent years. The volcanic alert level was raised to yellow in 2012 and has been kept there since.

The national emergency plan exists. The red zone covers about 500,000 people. Evacuation on volcanic alarm is designed to finish in 72 hours. Italy's IT-alert cell broadcast system has been tested for that scenario.

On the night of 31 July no volcanic alarm went out. Under doctrine, that system is reserved for transition to the volcanic alarm phase. For the earthquake itself there was nothing useful to send. At three kilometres depth under the town, shaking arrives before any message. Japan's warning physics does not apply.

So the sensors work. The plan is printed. The missing piece is the signature. Who declares the alarm, on what evidence, when eruption science deals in probabilities rather than dates?

Italian memory holds both mistakes. In 1983–84, during the last major bradyseism crisis, tens of thousands were evacuated from Pozzuoli. No eruption came. Many remember the evacuation as trauma. In 2009, after a swarm near L'Aquila, public reassurance preceded a magnitude 6.3 that killed 309 people. Italy prosecuted seven experts for manslaughter; six scientists were later acquitted, while Bernardo De Bernardinis, then deputy head of civil protection, kept a conviction. Evacuate and be wrong: 1984. Reassure and be wrong: L'Aquila. Every official who might one day fill in that blank date knows both files.

The last eruption of this caldera was in 1538. It built Monte Nuovo in days. The years before it were years of rising ground.

What the three cases show

Magnitude did not decide who died. Venezuela's 7.5 sequence killed thousands. Japan's 6.8 killed under forty on official tallies. Deaths tracked systems, not seismographs. That means the dying is not only fate. Where control has been built, outcomes change.

Each case failed or held at a different layer. Venezuela failed at structure: buildings and institutions, the expensive layer with no global deadline. Japan held at almost every layer and hit the ceiling: secondary hazards, heat, supply chains, deaths that arrive in week three. Italy has sensors and plans and is stuck at the decision layer, the one that costs little in hardware and everything in responsibility.

The warning layer the UN promised by 2027 sits at 60% of countries two years out. Meanwhile the largest recent expansion of earthquake warning came from elsewhere. Research published in Science in 2025 showed that Android phones now form a detection and alert network covering about 2.5 billion people across 98 countries, increasing access to earthquake warning about tenfold in three years. The cheapest layer scaled through a software update. The expensive layers barely moved.

Prevention still returns more than it costs, on evidence that has been available for decades. Most disaster money still arrives after the event, as rescue and reconstruction.

The world is not simply unready. It is ready in patches: ready where someone paid for expensive readiness anyway, unready where the cheap promise was signed and the concrete was not poured, and stuck where the sensors work but the signature is missing.

The ground moved under Venezuela in June, under Kumamoto in July, and under Pozzuoli three days later. Frameworks do not negotiate with that schedule. Venezuela shows readiness that exists on paper. Japan shows what readiness can buy, and where it stops. Naples shows that after the sensors and the plans, readiness still comes down to a person, a probability and a decision that cannot be fully pre-written.

illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.

The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.