Investigating recent decadal trends in the Pacific westerly jet in response to various atmospheric forcings using CMIP6 model results and reanalysis data
Abstract. The strength and location of the North Pacific westerly jet (NPWJ) affects weather and trans-Pacific pollution transport as it triggers and directs atmospheric river events toward North America. We used four reanalysis datasets and eight Coupled Model Inter-comparison Project Phase 6 (CMIP6) models to investigate the characteristics and changes of the NPWJ during 1980&nash;2019. The NPWJ climatologic core seasonally swings between north and south, being most southerward (~33° N) in winter and most northward (~45° N) in summer, as shown by the observation-based reanalysis data. All data provide strong evidence for the weakening (up to -0.45 and -0.68 ms-1decade-1) and northward shift (0.2° and 1.0°) of the NPWJ in summer and autumn during the study period. Various atmospheric forcing experiments performed by the CMIP6 models further reveal aerosol forcing being the main driver, which can be traced back to the spatially inhomogeneous anthropogenic aerosol emission changes that increase in Asia and decrease in Europe. When we apply Earth system climate models to investigate the feedback between atmospheric forcings and atmospheric dynamical fields on decadal scales, two points should be noted. First, there is a need to include interactive chemistry in the CMIP6 model simulations to bring the dynamical fields closer to those based on observational data. Second, in addition to the well-mixed greenhouse gases, anthropogenic aerosols, and natural forcings proposed in the Detection and Attribution Model Intercomparison Project (DAMIP) single-forcing simulations, time-varying ozone radiative forcing is also important to climate change.
This study examines trends in the North Pacific westerly jet over the period 1980-2019 using four reanalysis data sets and eight CMIP6 models. Over this time period, the jet is found to be shifting northward and weakening during summer and fall months. Consistent with prior literature, these trends are primarily due to aerosol forcing, mainly an increase in Asia and decrease in Europe.
I find this study to be very similar to that conducted by Kang et al. (2024), who also used DAMIP models and concluded that aerosols were responsible for a weakening of the North Pacific storm track during summer. Perhaps the authors were unaware of this work. While slightly different methods were used here, I’m not sure this study is distinct enough from that one to warrant publication. If a revised version of this manuscript were to be submitted, the authors must clearly distinguish their work from Kang et al. (2024) and motivate why their results are novel.
In its current form, I also find the quality of the methods in this manuscript to not meet the standards of ACP. As I detail in the suggested revisions below, among other things, the authors do not take full advantage of the CMIP archive to answer their questions and fail to properly account for internal variability in their analysis.
Major revisions
1. The authors use a very small number of ensemble members per model from the DAMIP runs (see Table 2). As part of the LESFMIP project (Smith et al. 2022), most of these models should now have at least 10 ensemble members each for the single forcing runs, which are already available on the CMIP ESGF online archive (https://aims2.llnl.gov/search). Using a larger set of models and ensemble members would help to distinguish this work from earlier work on the subject.
2. If I understand their analysis correctly, the authors do not adequately account for internal variability in their analysis.
a. First, the authors compare the model trends from one 35-year segment in the piControl run (Figs. 3c-d) with the forced trends. The trends from the last 35 years of the piControl run represent only one realization of internal variability. The forced trends should be compared against ALL 35-year segments in the piControl run to see if any 35-year period in the piControl runs has a trend as large as that associated with the forcing.
b. Second, trends from individual model ensemble members (not the multi-model mean or single model mean over all ensemble members as in Fig. 4) should be compared with observations. The observed trends include both internal variability and any role of forcing, so a fair comparison would be to compare the observed trends to the distribution of trends from individual model ensemble members. Then, it can be assessed how likely the observed trends would occur in the PDF of trends from the control climate versus a PDF of trends from a climate with one or more of the anthropogenic forcings.
3. Some measure of statistical significance should be provided in all figures, such as stippling in all regions where p < 0.05. Currently, a p < 0.10 threshold is used in the figures where statistical significance is assessed. In my opinion, this is not conservative enough, as there is a 10% chance that the trend could be deemed significant by chance. Additionally, for the CMIP figures (Figure 3 and 7), it’s unclear how significance is assessed. Is significance assessed on the multi-model mean trend? What’s more important for the CMIP models is the level of agreement among models. Stippling where all models (or 7 out of 8 models) agree on the sign of the trends would be more helpful than assessing the trend significance of the multi-model mean trend.
4. If the authors think that ozone forcing is relevant, they should be able to test this using the hist-stratO3 or hist-totalO3 runs in the CMIP archive. Without evidence from these runs, there isn’t a convincing case that ozone would matter, especially because stratospheric ozone depletion dominates in the Southern Hemisphere. In my opinion, the non-linearity among forcings seems to be a more likely explanation for the non-additivity of the trends from the various forcings. Perhaps the authors are correct that the ozone forcing is relevant, but they need to show this using the ozone single forcing runs.
Minor revisions
Lines 41-49: This theory, originally put forward by Francis and Vavrus (2012), has largely been disproven, despite numerous media outlets claiming otherwise. The paper that the authors cite (Blackport and Screen 2020) here actually argues against what the authors are claiming … mainly that the Arctic does NOT cause a wavier jet stream at midlatitudes. Please remove this text and avoid propagating this error.
Line 106: I would suggest removing the NCEP reanalysis. The NCEP reanalysis is an older generation reanalysis that is much more subject to spurious trends. As shown in Table 1, it also has a much coarser spatial resolution compared to the other reanalyses.
Line 129: hist-nat simulations include changes in solar irradiance and stratospheric (volcanic) aerosols, not land use. See Gillett et al. (2016).
Tables 1 and 2: Please improve the clarity of the captions and labels in these tables. For example, the notes column in Table 2 is largely not understandable. If all runs are supposed to have three ensemble members (line 133), then why are exceptions listed in this column?
Lines 172-173: Is the wind maximum found only at grid points? Or, is some interpolation used to locate maxima in between grid points?
Lines 193-198: These trends are very difficult to see in the figure. For example, in JJA, the black and red Xs are basically on top of one another. It might be nice to provide the actual values in a table. From Fig. S2, it is apparent that only the trends in SON are significant at a p < 0.05 level. It should be clearly noted in this paragraph that the SON trends are significant, whereas the trends are not significant (or much less significant) in the other seasons.
Figures 2-7: The observed trends are clearly more significant during SON, so the focus on JJA in the main text is hard to justify. I would suggest moving the SON results to the main text and put the JJA results in the supplement.
Lines 287-288: This could also indicate that the forcings are not additive and that their effects are nonlinear (as noted later on lines 306-308).
Line 336: Technically, thermal wind describes the vertical shear of the geostrophic wind, not the total wind (although the upper-tropospheric wind is very close to geostrophic).
Lines 366-369: This statement is written as if this were observations, but if I understand correctly, the AOD is taken from model simulations (not observations). How confident are we that the models are correctly capturing observed aerosol changes over this period?
Lines 377-378: Good to clarify that the enhanced convection and latent heat release is in the tropical upper troposphere.
Lines 396-418, 478-481: How confident are you that the interactive chemistry is responsible here? There are likely many other differences between these two subsets of four models (e.g., differences in climate sensitivity, etc.).
Lines 458-459: Again, it’s important to note here that the observed trends are much more significant in autumn.
Figure quality does not meet the standards for ACP and should be improved. For example:
Typos
Line 22: southward
Line 124: which are time varying and have forcings
Line 255: tell us about the NPWJ trends
Line 289: Eyring
Line 326: thermal wind balance
Line 374: the climate system
Line 381: meridional
Line 393: interactive chemistry
Line 402: GISS-E2-1-G
References
Francis, J. A., and S. J. Vavrus (2012), Evidence linking Arctic amplification to extreme weather in mid-latitudes, Geophys. Res. Lett., 39, L06801, doi:10.1029/2012GL051000.
Kang, J. M., Shaw, T. A., & Sun, L. (2024). Anthropogenic aerosols have significantly weakened the regional summertime circulation in the Northern Hemisphere during the satellite era. AGU Advances, 5, e2024AV001318. https://doi.org/10.1029/2024AV001318
Smith DM, Gillett NP, Simpson IR, Athanasiadis PJ, Baehr J, Bethke I, Bilge TA, Bonnet R, Boucher O, Findell KL, Gastineau G, Gualdi S, Hermanson L, Leung LR, Mignot J, Müller WA, Osprey S, Otterå OH, Persad GG, Scaife AA, Schmidt GA, Shiogama H, Sutton RT, Swingedouw D, Yang S, Zhou T and Ziehn T (2022) Attribution of multi-annual to decadal changes in the climate system: The Large Ensemble Single Forcing Model Intercomparison Project (LESFMIP). Front. Clim. 4:955414. doi: 10.3389/fclim.2022.955414