| 21 Aug 2026
Extratropical Cyclone and Storm track Responses to Stratospheric Aerosol Injection in High-Resolution CESM Simulations
Abstract. Extratropical cyclones (ETCs) drive weather variability and dominate extreme wind and precipitation events in storm track regions. This study investigates ETC responses to stratospheric aerosol injection (SAI) compared to a high-emission (RCP8.5) and a reference scenario, using high-resolution CESM simulations with 0.25°/0.1° horizontal resolution in the atmosphere/ocean. Storm track activity is quantified using 2–6-day bandpass-filtered mean sea level pressure (MSLP), and individual cyclones are tracked using CyTRACK. SAI produces regionally asymmetric responses: the North Pacific storm track shifts equatorward, linked to SST changes, while the North Atlantic storm track shifts poleward, linked to an SAI-induced positive North Atlantic Oscillation (NAO) phase. The RCP8.5 scenario shows the fewest cyclones in both hemispheres, whereas cyclone frequency partially recovers toward reference levels under SAI. A risk analysis linking extreme events to ETCs through combined distance and vorticity masking shows that extreme precipitation events (>50 mm/day) increase by 18 % in both hemispheres under RCP8.5 but decrease by 11 % in the Northern Hemisphere and 6 % in the Southern Hemisphere under SAI. Near-surface wind extremes >25 m/s) decrease by 11 % under RCP8.5 and 5 % under SAI in the Northern Hemisphere, but with different spatial patterns. In the Southern Hemisphere, extreme wind events decrease by 4 % under RCP8.5 and 5 % under SAI. These findings demonstrate that SAI does not simply reverse warming-induced changes but produces new circulation patterns, highlighting the need for regional assessments of geoengineering impacts on midlatitude weather extremes.
I cannot recommend acceptance of this paper. The premise is that high resolution modeling is needed for these features, but the authors then show that not only is that premise not true, but the authors didn’t even really use high resolution modeling. There are also numerous flaws with the methods, mechanisms, analysis, and writing.
The introductory section needs a massive rewrite. It has random sentences that don’t seem to tie together, and a lot of the discussion has nothing to do with extratropical cyclones. There are also weird conflations of tropical and extratropical features without any tying of mechanisms, i.e., the authors seem to be ignoring synoptic meteorology. Also, the authors don’t seem to have great motivation for their study other than the fact that they used a higher resolution model than many people use. Also, on line 58, this is not what storm tracks are.
One of your main points is that you conduct risk analysis. I want to say this emphatically: no you don’t. You are looking at changes in physical variables and making inferences about what those mean. Risk is much more than that.
Your resolution of 0.25 degrees is not _that_ different from what many people use. It still doesn’t resolve storms very well, although it does get some features better than coarser resolution models. You’re making a very big deal about using a high resolution model, but you’re not really using a high resolution model. Also, you aggregate all of your results to 5 degree resolution, so why did you need a high resolution model in the first place if you’re just going to smooth everything out?
In Section 2.1 you talk about how you’re not actually using a resolved stratosphere. There are known stratospheric interactions with midlatitude storm tracks, which you’re studying. You need some kind of evidence that your approach doesn’t negate your conclusions.
Figure 1: It looks like you’ve got some pretty important biases between low resolution and high resolution. More specifically, a colder model (which HR is) will have fewer storms. This doesn’t seem to be mentioned in the manuscript at all. If you were just computing differences, this wouldn’t matter. But your metrics involve exceedances, where it will definitely matter. Relatedly, why did you pick the thresholds that you did, and how do those choices affect your results?
Applying a bandpass filter with a 2-day cutoff to daily data is going to give you really noisy results. This is at the Nyquist limit.
Lines 181-186: These results are generally consistent with those of past studies (Ferraro et al., 2013; Gertler et al., 2020; Simpson et al., 2019; Richter et al., 2018). And also, these are zonal mean results, which you said were a limitation in the introduction. I don’t really see how your results in Figure 4 differ all that much from what others have found using coarser models and simpler approaches.
Section 3.2 is very terse and doesn’t really have any explanations in it.
Lines 219-220: This sounds like your definition of the storm track is flawed, not that a different mechanism is required.
Line 226: Why even plot distributions if all you care about is the maximum?