Zeng et al. 2019 NCC; Monerie et al. 2026 Clim Dyn (wind stilling)
Finding
ASOS sensor steps explain most of the North American decline and about 30% of the global 1978–2010 decline
| Verdict | Real error (via data) |
|---|---|
| Significance | High (Size 2 · Reach 2 · Consequence 2 = 6 of 6) |
| Outcome | Published |
| Where reported | Via the ASOS wind preprint Our write-up on Zenodo (latest version): doi:10.5281/zenodo.24357792 |
| Date | 2026-11-18 |
Summary
Zeng et al. (2019, Nature Climate Change 9:979) reported a global decline in land wind speeds ("stilling") that reversed around 2010. Two documented US equipment changes (see the ASOS wind sensors entry) sit inside their station record. Removing only these steps erases most of the North American decline and about 30% of the global 1978–2010 decline. The post-2010 rise is not caused by the steps, though it is smaller over a longer window.
What we found
We replicated the paper's method (GSOD stations complete for 1978–2017, annual station means, one-break piecewise regression). The replication reproduces the published numbers closely: 1,257 stations, turning point 2010, −0.088 m/s per decade over 1978–2010.
| Series | Turning point | Before | After |
|---|---|---|---|
| North America, clean subset, raw | 2012 | −3.6 %/decade | +7.9 %/decade |
| Same, both US steps removed | 2012 | +0.3 %/decade | +8.2 %/decade |
| Global, raw | 2010 | −2.7 %/decade | |
| Global, US steps removed | 2006 | −1.9 %/decade |
- A natural control supports the result: 41 US stations outside the NWS/FAA programme (mostly military) show −0.5% per decade for 1978–2009, against −4.6% for the NWS/FAA ASOS stations.
- Extended to 2025, the North American rise after the turning point is about +2% per decade rather than +8%: much of the published reversal slope comes from a short window starting at the trough years 2010–12.
- The paper's homogeneity checks (an automation flag and a single Pettitt test per station) cannot remove two staggered steps of 6–9%, and the automation flag misses the 2005–09 sensor swap at stations that were already automated.
- Monerie et al. (2026, Climate Dynamics) use HadISD. In a HadISD-based hemispheric design like theirs, removing the two steps cuts the North American 1980–2010 decline by about 64% and the Northern Hemisphere decline by about 24% (a lower bound). Their main conclusion, a hemispheric decline, survives.
What it does not show
Stilling outside North America is untouched by this analysis, and other networks may have their own equipment changes. The global decline remains, smaller.
Reproduce
Preprint (Sections 4.6–4.7), code and derived data: doi:10.5281/zenodo.24357793.
Discussion
Questions and discussion are welcome: ask other readers, the authors or the maintainer. Authors and developers are especially welcome to reply. Comments use GitHub Discussions (sign in with GitHub). No GitHub account? Write to us.