In 2020, Kossin et al. reported that the proportion of major tropical cyclones (Category 3-5) relative to all tropical cyclones (Category 1-5) had increased from 1979 to 2017. The finding was cited as evidence that warming oceans were producing more intense hurricanes. The headline was alarming. The statistic was real.
Ivo Welch (arXiv:2603.07849, March 2026) decomposed the statistic and found that the trend was driven primarily by fewer Category 1 storms being observed, not by more Category 3-5 storms occurring. Weak storms declined by 17 percent between the first and second halves of the record. The ratio of major to total storms rose because the denominator shrank, not because the numerator grew. The intensification signal was an artifact of declining detection of weak storms — likely a change in how weather agencies classified marginal tropical cyclones over the decades.
Then Welch extended the dataset through 2023. The picture changed. In the extended record, the trend is no longer driven only by fewer weak storms. There are now genuinely more strong tropical cyclone observations as well. The original conclusion — that tropical cyclones are intensifying — turns out to be correct, but the original evidence for it was contaminated by a measurement artifact. The right answer was supported by the wrong data. Correcting the measurement problem and extending the record reveals that the real signal was there all along, hidden beneath the artifact that coincidentally pointed in the same direction.
The structural lesson is about what happens when an artifact and a genuine trend point the same way. They are indistinguishable in the original dataset. Decomposition reveals the artifact. Extension reveals the trend. Neither operation alone — decomposing without extending, or extending without decomposing — would have clarified both. The corrected conclusion is stronger than the original precisely because it survived the correction.
Welch, "Genuine Increases in Tropical Cyclone Intensities," arXiv:2603.07849 (March 2026).