In July 2021 the Ahr valley in Germany flooded. 185 people died. Most of them at night, in their homes, in a valley that had flooded before but never like this.

The flood was not a surprise to the systems that watch such things. The European Flood Awareness System had issued warnings 48 hours earlier. The rainfall models were accurate. The river level data was real-time. The authorities knew.

The people asleep in the valley did not.

The Warning That Did Not Arrive

Forty kilometres north, in the Netherlands, people in the Limburg flood zone received NL-Alert messages on their phones. No app required. No internet needed. The message arrived regardless of whether the phone was on silent. It overrode the quiet. It said: water is coming. Move.

NL-Alert is the Dutch emergency broadcast system. Activated in 2012. Built on a technology called Cell Broadcast that has been in every mobile phone since 1991. Every phone that can receive a call can receive a Cell Broadcast message. The technology is older than smartphones. It works when the network is congested. It works when people are asleep. It costs nothing to receive.

Germany had not activated an equivalent system. Not because the technology was unavailable. Because a decision had been made that the risk of building the infrastructure outweighed the benefit of having it.

The concern was reasonable on its face: a system that can send a message to every phone in a geographic area is also a system that could track every phone in a geographic area. The capability for mass notification is also a capability for mass surveillance. If the government builds it, the government might misuse it.

True. And: a fire alarm can also be used to wake you up unnecessarily at 3am. We still have fire alarms.

The Decision Made on Behalf of the 90%

The people who decided not to activate Cell Broadcast in Germany were not wrong about the dual-use risk. The concern is legitimate. Surveillance states have used emergency infrastructure for control. The history is real. Germany’s history with state surveillance in particular makes this concern more than theoretical.

The Netherlands had a head start — not in wisdom, but in near-misses. The 1953 flood and the 1993/1995 near-misses had already forced the Dutch to weigh the tradeoff and decide: the warning system matters more than the misuse risk. Germany had not yet had that forcing moment. The river had always behaved. The privacy concern prevailed. It was a reasonable decision in the context available at the time.

The context changed in July 2021. The tradeoff looked different after the flood than it had before it. That is not a condemnation — it is how these decisions usually work. The near-miss clarifies the mind. The catastrophe confirms what the near-miss was trying to say.

The warning existed at the institutional level. It did not cross the last metre to the person who needed to act on it. The gap between “the authorities know” and “the person in the valley knows” is exactly the gap that NL-Alert is designed to close.

185 people. The seam between institutional knowledge and individual action was closed. The warning never crossed it.

The 1953 Lesson and the 1993 Near-Miss

The Netherlands did not build NL-Alert because Dutch engineers are cleverer than German engineers. They built it because they had earlier lessons — and lessons learned from near-misses and catastrophes produce a different kind of knowledge than lessons learned from textbooks.

In 1953 the North Sea broke through the dikes in a single night. 1,836 people died in the Netherlands. The response was the Delta Works — the largest flood engineering project in history. But also: a culture of flood awareness that has never faded. The assumption, baked in at the civilisational level, that water is always a near-term threat. That the flood map must be consulted before the building permit is issued. That the warning system must be tested before the flood arrives.

That assumption did not come from being smarter. It came from 1953. From a generation that lived through it, and a generation after that which was raised by people who lived through it. The knowledge is in the culture because the catastrophe was in living memory.

In 1993 and 1995 the rivers rose to within centimetres of the top of the dikes. No breach. But 250,000 people were evacuated in 1995. The cattle were moved. The furniture was loaded onto trucks. The villages emptied.

The dikes held. The people came home. And the near-miss produced Room for the River — the programme that moved dikes back instead of building them higher, that gave rivers the space they were going to take anyway, that accepted the flood plain as what it is.

The Ahr valley had not had that near-miss yet. The river had always behaved. The valley was safe. The warning system could wait.

Then it could not wait anymore.

After the Flood

Germany activated Cell Broadcast in 2022. The learning came after the hurt. That is how it usually works — the signal arrives at maximum volume when the cost of ignoring it is no longer deferrable.

NL-Alert was tested again in the Netherlands in 2021 during the same weather event. It worked. People in the affected areas received the warning. The seam between the forecast and the person was open.

The technology is the same. The phone in the Ahr valley and the phone in Limburg are the same hardware. The Cell Broadcast standard is the same in both countries. The gap was a decision, not a capability.

The Seam Between Knowing and Acting

Every warning system is a seam between two layers: the layer that knows (the forecast, the model, the sensor, the authority) and the layer that acts (the person, the family, the community). The seam is only useful if it is open — if the signal can cross from one side to the other in time for the action to matter.

A warning that reaches the authority and stops there is not a warning. It is a record. The authority knows. The record is filed. The people in the valley sleep on.

The person asleep in the Ahr valley was not wrong to be asleep. It was night. The river had always behaved. There was no reason to expect otherwise. The seam was closed. The signal never crossed.

NL-Alert is the seam made open. Not an app you have to download. Not a website you have to check. Not a system you have to opt into. Just: the warning, arriving on the device in your pocket, at the moment it matters, whether you expected it or not.

The user does not sux. The person in the valley was not wrong. The seam was wrong to be closed.

185 people. The record is straight. The warning was there.

Fixing the seams across the waterline — that is the work. Higher dikes are not always the solution. The 1953 lesson was: fix the dike. The 1993 lesson was: fix the floodplain. The 2021 lesson was: fix the seam. All three correct. None of them sufficient alone.

The river carries what it carries. The rain falls where it falls. The physics does not negotiate with the dike height. But the warning can cross the last metre — from the authority who knows to the person who needs to act — if the seam is open.

Keep the seam open.

This post is part of the Gutenberg/Semantic series — though you do not need the framework to understand this one. Sometimes the waterline is literal. Sometimes the seam between knowing and acting is the difference between a warning and a record. Keep the seam open.