Science1 publisher2 min readPublished
Sea surface temperature bias pushes the North Pacific jet south in 34 of 35 CMIP6 models
A Geophysical Research Letters paper compares coupled CMIP6 runs against atmosphere-only runs fed observed ocean temperatures, then adds the bias back and watches the jet move south again. Northwestern Pacific bias was particularly important.
The Scientist · Science desk
What happened
- In 34 of the 35 CMIP6 models compared, the North Pacific westerly jet sat farther south when the model generated its own sea surface temperatures than when observed temperatures were prescribed.
- Separate atmospheric experiments that added the models' sea surface temperature biases onto the observed ocean field reproduced the same southward shift of the jet.
- Imposing those biases region by region showed that errors in the northwestern Pacific were particularly important in moving the jet south.
- The jet in question is the band of strong west-to-east midlatitude winds in the upper atmosphere, and the work is published in Geophysical Research Letters by Yusuke Ushijima and co-authors.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- constraint Averaging across the CMIP6 ensemble does not clear the jet error, because almost every member leans the same way, so a multi-model mean wind field over the North Pacific inherits the offset.
- decision Anyone downscaling North Pacific circulation now has a configuration choice to justify: coupled output, atmosphere-only output forced with observed SSTs, or a bias-adjusted ocean field.
- capability Ocean model development gets a narrow target to test against, since the northwestern Pacific can be corrected in isolation and the jet response measured.
The coupled-versus-prescribed comparison is an association. The two configurations differ by more than the ocean bias: one computes its own sea surface temperatures, the other is handed observed ones and cannot respond to the atmosphere at all [2]. So the authors ran the cleaner test. They took the observed ocean field, added the biases their models produce, and ran the atmosphere again [4]. Only the SST field changed between those runs. The southward shift can therefore be attributed to the ocean error.
Thirty-four out of thirty-five is about 97 percent of the models compared [10], and it leaves one exception [11]. An error with the same sign in almost every member does not average out. The multi-model mean carries it, and the spread across members will understate how far off that mean sits.
The subject throughout is where models put the jet. Every figure comes from simulations, including the observed-SST runs, which are still model atmospheres fed a real ocean [2]. Present-day bias is worth chasing anyway, because the model characteristics that produce it also act on simulations of the future [9]. Ushijima and co-authors argue that better representation of SSTs could make simulated circulation over Japan and East Asia more accurate and improve the reliability of future projections [7].
The paper does not report the shift in degrees of latitude, or what it does to storm tracks and rainfall [12]. It stops short of a correction factor you could apply to a regional wind field, and the link it draws to East Asian weather is qualitative [6]. What transfers is the sign of the error and the part of the ocean to look at first. Until a given model's northwestern Pacific SSTs are shown to be close to observations, I would treat the latitude of its North Pacific jet as too far south, and check any regional product built on that wind field against an observed-SST run before trusting the position.
What to watch
- Whether the same SST bias pattern also skews the models' projected future change in jet position, not only its present-day latitude.
- Whether the one model of 35 that did not shift south has a cleaner northwestern Pacific ocean field, which would make it a useful test case.
- Whether modelling centres target northwestern Pacific SST error in the next round of coupled models, and whether the jet latitude follows.