Typhoon season opens with a faster alarm

The first named typhoon of the western Pacific season formed on 2 May east of the Philippines, and the forecasters who tracked it did something that would have been unthinkable a decade ago: they issued a reliable landfall forecast six days ahead, with a track cone half the size of the 2020-era products and a storm-surge model accurate enough to guide evacuations. The typhoon, which eventually brushed the coast of Luzon before turning north, was not a severe one, which is precisely why it was a useful test. The machinery of prediction worked, and the region had the rare luxury of practising with a storm that did not demand heroics.

The improvement is real and worth naming precisely. The gains come from three places: denser satellite coverage, better ocean data from profiling floats, and the machine-learning trackers that now assimilate all of it into forecasts that beat the physics models on speed and skill at the same time. The six-day lead time is not a fluke; it is the new baseline. The last two seasons in the western Pacific have both shown this, and the lesson has been absorbed by the agencies that issue the warnings.

What has not been absorbed, and this is the part I keep coming back to, is what a better forecast is actually worth on the ground. A six-day warning is only as good as the evacuation system that uses it, and the evacuation system depends on roads, shelters, public trust and the willingness of people to move twice in a season when the first storm does not strike. In the Philippines and Vietnam, where these storms do most of their damage, the bottleneck has moved from prediction to preparation, and the forecasts have exposed that bottleneck rather than removed it. Every improvement in lead time shifts the burden onto the parts of the system that are hardest to improve: the last-mile alert, the school closure decision, the fishing boats that must decide whether to ride it out.

This is the right kind of problem to have. A decade ago the argument was about whether the forecast could be trusted; today it is about whether the warning reaches the right person at the right hour, which is an argument that institutions can win. The Asian Disaster Reduction Centre has been running a pilot that sends the same forecast through eleven channels, from cell broadcast to village loudspeaker, and measuring which ones actually change behaviour; that kind of unglamorous research is where the next decade of life-saving is going to happen.

There is also a colder arithmetic underneath the good news. The western Pacific is warming faster than the global average, and the storms that form in it are becoming more intense and more likely to intensify rapidly near the coast. The forecasters who issued this season’s first six-day track did not invent the extra margin; they found it in the data, and the data is also telling them that the storms are getting harder. Better alarms, warmer seas, faster storms: the race between the forecast and the ocean is not being won by the forecast, only kept even, and keeping even in a race that is accelerating is a kind of victory that no one should mistake for the finish line.