the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Enhancing Climate Model Performance through Improving Volcanic Aerosol Representation
Abstract. An accurate representation of Earth's surface temperature is crucial for simulating climate change. Yet many climate models struggle to reproduce the evolution of historical temperature records, especially after the major 1963 Mt. Agung volcano eruption. This study investigates whether the method of specifying the volcanic forcing could be contributing to this bias using the Energy Exascale Earth System Model (E3SM). The CMIP6 protocol represents volcanic eruptions through simplified radiative forcing, neglecting the interaction between volcanic aerosols and clouds. Here we adopt a new approach based on an updated volcanic eruption inventory, which includes volcanic sulfur dioxide emissions and hence allows for a more realistic representation of subsequent physical processes that involve volcanic aerosols. With this new approach, E3SM simulates slightly warmer surface temperatures and improved interannual variability during years 1940–1980 compared to the standard CMIP6 approach. The improvements mainly stem from two factors: 1) the inclusion of volcanic aerosol-cloud interactions, which reduces aerosol indirect effect by volcanic quiescent warming effect, and 2) the more accurate representation of volcanic eruptions after 1963, which leads to less volcanic aerosol cooling. Overall, this study highlights the importance of more accurate volcanic forcing in improving climate simulation and is strongly in favor of an emission-based volcanic forcing treatment.
- Preprint
(6225 KB) - Metadata XML
-
Supplement
(128 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1612', Anonymous Referee #1, 04 Aug 2024
This study assesses the role of volcanic aerosols on 20th century climate, in particular whether simulating volcanic aerosol interactions with clouds (and a few added small explosive eruptions) leads to improved historical climate replication. Global climate model (GCM) treatments of volcanic aerosols are typically highly simplified, with impacts of interactive volcanic aerosol processes being rarely accessed for more than a decade. There has been a lot of recent effort with E3SM to assess potential improvements in climate simulations resulting from interactive aerosol and cloud process additions, and these volcanic aerosol impacts are a worthwhile candidate. The experiments in this study are a fitting and worthwhile effort, though there are considerable issues with how the results are presented. Most critically, the manuscript claims substantial improvement but does not clearly show this to be the case. The vast majority of the targeted historical temperature bias remains despite the altered volcanic scenario, so it looks clear that improvements are small. These results, though mostly negative, are of interest to other modeling centers and worth ultimately being published. But I must ask the authors to please accurately communicate the results, to add figures to show whether the hypotheses hold any water (i.e. volcanic impacts on CCN, net all-sky and cloud forcings), and to choose a more accurate title.
Major Comments
The title is very vague and optimistic to an extent not supported by the results. This work is about the role of specific volcanic aerosol processes in historical climate simulations, so should be titled along these lines. E.g. ‘Impact of explosive volcanic eruptions on mid-20th century surface temperature development: the role of aerosol-cloud interactions’ or something similarly explicit. Climate model performance is only minimally shown to be “enhanced”, as the lion’s share of historical temperature bias remains, and “performance” is unclear.
While the simulation changes are appropriate, the key claim is that in Fig. 10 the altered experiment (red line) is more similar to observations (black and grey) than the original experiment (dark blue line), and that this is due to more accurate volcanic representation. But I do not find the improvement or its cause clear, so must ask the authors to address this in two ways. First, I want to see that the net TOA forcing evolution is clearly different between the model versions, as a substantial net forcing difference is requisite for a substantial forcing-driven surface temperature anomaly, and the first is less susceptible to the very heavy noise in Fig. 10. So I request that net forcings be shown in Figure 7, along with appropriate uncertainty bounds of the authors choosing. Second, because one of the main hypotheses is that the new setup improves climate by accurately representing aerosol-cloud interactions, I want to see evidence of this: First, some depiction of CCN number density differences between experiments, and second a depiction of net cloud radiative effect/forcing (CRE) development over time.
The results appear quite negative yet the manuscript is extremely upbeat about large improvement, which I find deceptive. I think the results being somewhat negative is interesting, and that this is overall a worthwhile effort, so the strongly upbeat tone is confusing and unnecessary. The model changes do not come close to fixing the targeted bias, with the revised model’s global mean surface temperatures veering off into practically the same bias as the original model around the time of the Agung eruption. Yet the manuscript is in total denial of this, cf Lines 505-6: “V2-CMIP6 simulates a prolonged temperature drop after the Mt. Agung eruption in 1963 compared to observed temperature trends, while the V2-IVA simulation mitigates this temperature drop.” If the requested figure panels work out as anticipated, the authors can tout a small-to-moderate amount of improvement, but the results and conclusions can’t grossly misrepresent the data. Further, this denial prevents the authors from explaining why the results are not stronger, which would be interesting and beneficial to know.
The manuscript is often confusing to read and repeats itself, so can be shortened and better organized. Section 2.3 is largely a repeat of the latter half of the Introduction that seems unnecessary. Similarly, the Discussion is mostly a repeat of the Results rather than a discussion of caveats, lessons for future efforts, etc. The section on forcings from aerosol-cloud interactions (3.3) would more logically go before the temperature anomalies these are presumed to cause (3.2). I also feel the preindustrial control simulations are too disjoint from the central results about temperature changes to warrant their own results section (3.4), but that some of this material can be used elsewhere. Further, several aspects of this study, which I’ve laid out in the specific comments, make it confusing to understand and should to be clearer.
Specific CommentsLine 13: The abstract’s first sentence is odd for two reasons: 1) The phrase “representation of Earth’s surface temperature” is vague, as accurately simulating historical temperature is what presumably lends credence to future projections, rather than simply representing present-day climatological temperatures, and 2) The revised simulation setup still has considerable bias in replicating historic temperature, so if this would be an argument *against* using E3SM for climate projections.
Line 23: It’s pretty impossible to know what the “volcanic quiescent warming effect” is before it is properly introduced, so I hope the authors can describe this in simpler language here, e.g. removal of artifacts that stem from a simplified volcanic background aerosol state.
Line 25: “strongly in favor” would indicate the influence on climate is strong, which is not the case. The wording in the abstract is too positive given the largely negative results.
Lines 34-6: I feel the first paragraph of the Intro would be stronger if it kept to the focus of this study, rather than alluding to much stronger eruptions. The focus here is on mid-20th century eruption impacts, of which Agung’s 0.1-0.2 C cooling is the most consequential. I hence find the Tambora and Pinatubo references distracting, as they’re multiple times stronger than any event in the 1940-80 assessed window. Likewise, the line that eruptions play “a crucial role in modulating climate changes” most evokes impacts of greenhouse gases from Large Igneous Provinces, as “crucial” is debatable for pretty much any other eruption and certainly the relatively modest ones assessed in this study. I hope the authors can make this study’s focus interesting rather than emulate standard intros of volcano-climate literature. The authors are free to do as they want here, though one path could be to start by mentioning that climate models have trouble replicating 20th century surface temperature evolution, that this reduces confidence in projections of future climate, the entangled roles of anthropogenic aerosol, greenhouse gas, and natural aerosol changes, introducing why poor representation of volcanic eruptions is a primary suspect, and briefly describing Agung’s cooling as the largest volcanic impact of this period.
Lines 34: It seems odd to start with “natural radiative forcing” before saying that volcanic eruptions result in sunlight-blocking aerosols. I’d suggest holding off on this term until it is defined.
Line 45: The current placement of “Water vapor is scarce in the stratosphere” after a line saying that “eruptions emit a variety of gases […] into the stratosphere” suggests this study will involve water vapor emissions into the stratosphere, but this gets no further mention. I would weave this into the following line to remove the perceived focus on water vapor, ie “lack of wet removal in the water-scarce stratosphere”.
Line 53: Because this study’s main focus is on aerosol-cloud interactions, it would be fitting for there to be a paragraph here on whether past research has indicated a strong effect or not. For instance there are the two studies I cite below, one claiming ‘substantial cooling’, the other a weak effect (oddly with largely the same authors). I suspect there are other relevant studies. It would also be beneficial to have a little overview of arguments for why the effect could be weak and why it could be strong, e.g. whether volcanic sulfate falls to the troposphere over wide areas or mostly near the poles, the global abundance of CCN, scalings of cloud optical depth and radiative forcing with CCN number density). I also hope the authors can clarify if this study’s focus is solely on explosive eruptions, or if effusive eruptions – which also release sulfate and hence affect cloud properties – are also a focus.
Chen, Y., Haywood, J., Wang, Y., Malavelle, F., Jordan, G., Peace, A., ... & Lohmann, U. (2024). Substantial cooling effect from aerosol-induced increase in tropical marine cloud cover. Nature Geoscience, 1-7.
Malavelle, F. F., Haywood, J. M., Jones, A., Gettelman, A., Clarisse, L., Bauduin, S., ... & Thordarson, T. (2017). Strong constraints on aerosol–cloud interactions from volcanic eruptions. Nature, 546(7659), 485-491.
Line 55 – Flynn & Mauritsen 2020 is cited 4x in this manuscript, and its Fig. 12 is quite relevant here. But I hope the authors appreciate that the aerosol-cloud interactions that study raises as the cause of CMIP6 historical climate bias are influences of ice nuclei on mixed-phase cloud phase, so extremely different than the interactions assessed here. Hence I find the wording here a little misleading, though it’s vague enough I find it okay if the authors keep it.
Lines 62-67: Moving the first 4 sentences of this paragraph into the previous paragraph would allow this one to focus specifically on explaining the two mechanisms. As these are tricky to understand, keeping the paragraph focused would help the reader.
Lines 67-76: I find this critically important paragraph confusing so request the authors please rewrite this. First, the “volcanic quiescent warming” name is confusing, as there is no physical volcanic warming process here and this seems to be more of a “reduced volcanic cooling during quiescent periods” scenario due to removal of a simplified process representation (constant background volc aerosol). This should at minimum be explained more clearly. Second, the discussion of anthropogenic aerosol changes here is confusing to me, as the two mechanisms both involve altered volcanic aerosols in simulations with the same anthropogenic aerosol changes. Interactions between volcanic and anthropogenic aerosol processes can certainly be an influence, but I don’t see this as an explanation for either mechanism, so think it makes more sense to focus on just the volcanic aerosol changes.
Line 81: “version 3” is stated here but everywhere else in the manuscript discusses “version 2”. Please reconcile.
Fig. 1: The SO2 cloud in the stratosphere is just too unrealistic and evokes a polar stratospheric cloud rather than a gas. Perhaps this would be better depicted as just the word ‘SO2’ with no cloud, as with H2SO4.
Lines 100-116: I feel this paragraph is separate from “volcanic forcing representation” and may be better spun off into a separate sub-section on E3SMv2 itself as compared to volcanic experiments. Also, because aerosol-cloud interactions are imperative for this study, it would be beneficial to add 1-2 lines on E3SM’s CCN activation.
Lines 118-135: This paragraph feels longer and more technical than it needs to. I would cite the dataset itself if possible (or a paper that describes it) and give a simpler description, rather than mentioning many instruments it incorporates. It would seem to be the pre-SAGE period that overwhelmingly matters most here. Is this using the standard CMIP6 volcanic emissions dataset? If so, there’s no proper paper to cite, but I suggest one way to cite the data webpage below. But I’m not familiar with the AER-2-D use here, so wonder if this is E3SM’s own method instead. If this is not a standard CMIP6 volcanic aerosol prescription, this should be stated.
IACETH, 2017: CMIP6 Stratospheric Aerosol Dataset (SAD) v3. Institute for Atmosphere and Climate, ETH Zurich, Earth System Grid Federation, accessed 3 August 2024, https://doi.org/10.22033/ESGF/input4MIPs.1681.
Line 138: Are there really only two volcanic events? In Fig 2 there are two unexplained peaks between the two mentioned eruptions in the V2-CMIP6 simulations that presumably are using this data method.
Line 137-144: Somewhere here, as well as elsewhere in the study, should explain that – if I understand right – the focus is particularly on explosive eruptions that emit (primarily?) into the stratosphere. Effusive eruptions more directly bring sulfur into the troposphere but do not seem to be included in either of datasets for prescribed aerosol forcing or prescribed SO2, so this should be clarified.
Lines 161-2: There have been efforts to correct volcanic aerosol impacts in E3SMv2/MAM4, e.g. Brown et al., 2024. Maybe this particular version wasn’t used here, but hopefully there don’t wind up with a bunch of separate E3SM versions for representing the same processes.
Brown, H. Y., Wagman, B., Bull, D., Peterson, K., Hillman, B., Liu, X., ... & Lin, L. (2024). Validating a microphysical prognostic stratospheric aerosol implementation in E3SMv2 using observations after the Mount Pinatubo eruption. Geoscientific Model Development, 17(13), 5087-5121.
Lines 180-1: For the SAGE period the original methods feeds into the model observed aerosol extinctions, so it would be highly possible this performs more accurately than the new SO2-emission method, at least if stratospheric aerosol falling into the troposphere does not cause a prominent forcing. But for the mid-century period both methods rely on (presumably quite poorly constrained) SO2 estimates. For this period we expect the new version is improved specifically because of 1) the added explosive SO2 injection events and 2) the addition of aerosol-cloud interactions caused by CCN resulting from explosive eruptions. ARI after Fuego and Agung will be different but for this this isn’t a reason to expect improvement. Could the authors please state if my understanding here is correct? Potentially the text should clarify some of this.
Tables 1 and 2: All these eruptions appear to be explosive rather than effusive. Can the authors please make it clear what the focus is in the captions?
Table 1 specifically: Since this says “(CMIP6)”, can the authors please confirm this is the method used in the CMIP6 protocol and not just the CMIP6 E3SM simulations, and if not, specify?
Table 2 specifically: Is it confidently known that these eruptions emitted into the stratosphere, or are these very crude estimates of what occurred? The stratospheric fraction is expected to be the most important, as it is the longest lived and most globally distributed – is there no stratospheric amount available?
Lines 250-90: Most of the material here is redundant with material in the Introduction. For instance VARI, VACI, “volcanic quiescent warming effect”, and “volcanic surplus cooling effect” are all being introduced as if for the first time, but this is not the case. I’d recommend omitting this section and moving unique information into the Introduction’s version, as this isn’t really methodology.
Lines 270-277: I find this confusing, as most experiments and nearly all results are about volcanic aerosol impacts on historical climate but this largely is about anthropogenic aerosols and PI. I’m not convinced it’s worth focusing on interactions between volcanic and anthropogenic aerosols, and there aren’t experiments here to isolate these from volcanic aerosol impacts on their own, which would entail also varying anthropogenic aerosol emissions. I feel the manuscript would be stronger if it kept a clearer focus on how explosive volcanic aerosols affect mid-20th century temperature development. The PI background state is worth testing, but I feel reporting this can be more briefly folded into the results only.
Figure 3: As above, I find this figure distracting from the main experiments and unnecessary. I also don’t understand what it depicts. Adding background volcanic aerosols would, if interactive, mean more CCN and more cloud scattering, not less. The PI background state is certainly worth testing but is not a focus in any other figure. If this authors want to keep the figure, I hope they can make it and its importance clearer. The capitalization should also be consistent.
Table 3: Since “V2-IVA-NPI” isn’t used in any Figures, I’m not convinced it’s worth adding here or describing in the methods, as the paper would be easier to digest this way. I do, however, think this would be suitable for a paragraph in Results, e.g. “We repeated the V2-IVA experiment but branched off from a preindustrial control simulation having no volcanic emissions. This altered the results such that […]”, etc.
Lines 340-60: These first two paragraphs seem largely more suited to Methods, as they’re mostly just saying what’s in the volcanEESM datasets. Contrarily, the Methods has Figure 2 and lines describing it (183-9) that are instead results in that they show sulfate properties calculated by the model rather than SO2 inputs.
Lines 401-26: I recommend splitting off the radiative forcing results into its own subsection, as this section is very long right now.
Figure 7: I feel strongly that the net forcing should be shown rather than its shortwave and longwave components, as it’s otherwise very hard to gauge whether we should expect these forcings to cause different surface temperature developments. The SW and LW would be appropriate for the supplement if the authors want to cite them as showing clearer differences than the net effect, but do not seem strictly necessary. Second, because these results are critical for the next section and the paper’s conclusions, there should be uncertainty bars. Third, because the paper focused on aerosol-cloud interactions it’s important to show the net cloud radiative effects (all sky – clear-sky), as currently it’s impossible to much this varies among experiments. And fourth, the LW, if shown, should be positive, as it is offsetting rather than augmenting the SW forcing.
Line 492: Does “V2” mean “V2-CMIP6”?
Lines 501-10: First, it can’t seriously be ignored that there’s a huge remaining bias in surface temperature development. Following Agung the revised model experiment nevertheless descends into a heavy bias just like the original experiment. Given the stated goal was to fix this bias, the result is largely negative and this really must be fairly explained. Second, shouldn’t the surface temperature differences relate to the volcanic quiescent warming and active cooling processes the authors have been discussing? Perhaps the authors can put these results into that context? Third, it doesn’t seem like Table S1 needs to exist, as most of the values in it are already stated in this paragraph. Maybe just bring in the rest?
Line 502: Is the V2-IVA line slope here statistically significantly distinct from zero? Should it be?
Line 506: “V2-IVA” mitigates this temperature drop”? The temperature drop is nearly exactly the same.
Line 507: Which observations are used in the comparison? Fig 10 shows 3 different datasets. Is this one or particular or an average?
Fig. 10: There doesn’t seem to be any explanation (or citation) of the observations in the text. Also, the colors in the legend don’t seem to match – possibly NOAA NCDC is one of the grey lines, despite being blue in the legend?
Lines 546-602: I feel this section should precede the section on historical temperature, as the cloud forcing is a contributor to temperatures and the temperature anomalies are really the culminating results of this study. I also feel this section would benefit from a figure on cloud changes. Perhaps CCN density and/or net CRE could be plotted as a function of latitude only? Okay if not, as long as the net CRE is worked into Fig. 7 and the reader can somehow appreciate whether CCN are different dissimilar between experiments or not. One potential issue worth checking is that if the explosive eruption impact on CCN is substantial but only in high clouds, the targeted clouds could have balanced shortwave and longwave effects (unlike low clouds which predominantly cause cooling via SW, but already have many CCN).
Line 602 and also Lines 694-5: “Low bias spreading” and “low bias spread”? The temperatures are too cool (a low bias) but I don’t understand why these instances refer to a ‘spread’.
Table 4: Can the authors please show some info on CCN number density? As is I don’t see clear evidence of aerosol-cloud interactions being directly affected by falling sulfate, which seems to be one of the main motivations for this study. Cloud properties will be radiatively affected by the aerosols in the stratosphere even with no CCN change. Even better would be a figure on this. Also, there are two different reds in this table.
Lines 626-652: I don’t feel this is worth its own section, given it’s quite independent from the main results on 20th century impacts. I think it would make more sense to add the most pertinent results to the preceding cloud forcing section. I also think it makes the most sense to show these results as in Table 4, but comparing V2-IVA-NPI (which I don’t think needs its own name and mentions elsewhere) to V2-CMIP6 (rather than V2-IVA-NPA vs V2-IVA). Then Table 4 could simply show both comparisons and the whole supplement can be omitted if the other reviewer(s) don’t request new plots there.
Lines 656-709: The Discussion section largely repeats the Results. There should be some discussion of caveats and also reasons why the model improvement is not larger (see major comments). Clearly the improvement is not as large as would have been hoped, as the surface temperature bias grows after Agung nearly as much as previously. I’d really like the authors to discuss this and why the effect is not larger. E.g. Are there simply far too many CCN for slowly falling stratospheric aerosols to matter? Are the cloud forcing changes too balanced between SW and LW? Are the volcanic aerosol being drawn down over narrow polar regions rather than broad cloud-rich areas?
Citation: https://doi.org/10.5194/egusphere-2024-1612-RC1 -
RC2: 'Comment on egusphere-2024-1612', Anonymous Referee #2, 09 Aug 2024
General comments:
The paper is devoted to reducing the bias in climate simulations by improving the representation of volcanic aerosols in CMIP6 models. This objective failed, as no notable improvement was obtained. However, the research has a promise. The authors implemented a sulfur cycle and the formation of volcanic aerosols. This is advantageous compared to the prescription of monthly aerosol fields, as in most CMIP6 simulations. This approach is not new and has been widely implemented in different studies in the last 20 years. The second mechanism employed to reduce the bias is the indirect effect of volcanic aerosols on tropospheric clouds. This is good but not a new idea. Ulrike Lohman worked on detecting this effect after major volcanic eruptions. In addition, this effect is poorly described by the models, so the first thing the authors have to show is that their model can reproduce this effect. It, by itself, is a complex task. The authors claimed that volcanic aerosols affected low-level clouds and said nothing about the impact on upper-level clouds. This is hardly believable. Volcanic aerosols will first affect cirruses and upper-tropospheric clouds. I doubt any volcanic aerosols could reach the lower troposphere, and, in any case, their contribution will be negligible in comparison with tropospheric (natural and anthropogenic) aerosols. The figures are reasonably well prepared, although showing the globally averaged tropopause height is useless. The authors should more clearly define the statistical significance of the results and make ensemble calculations to reach statistical significance. The text is poorly written, with repetitions, incorrect terminology, and poor English.
Specific comments:
L105: macrophysics > microphysics
L148: processes > microphysics
L262: Not only LW
L279-285: Improve language
L287-290: Volcanic aerosols penetrate into the troposphere mostly through tropopause folds and in high-latitudes
L362: optical properties of the atmosphere in the stratosphere
L388: light extinction > aerosols extinction
Helka > Hekla in multiple places
L393: response > extinctionon
L408: mentioned > showed
L412: Panel c > Figure 7c
L417: panel d > Figure 7d
L422-426: Clarify the text
L488: dimmer volcanic eruptions?
L491 anomalies are warmer > anomalies are greater
Citation: https://doi.org/10.5194/egusphere-2024-1612-RC2
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
423 | 97 | 120 | 640 | 33 | 14 | 15 |
- HTML: 423
- PDF: 97
- XML: 120
- Total: 640
- Supplement: 33
- BibTeX: 14
- EndNote: 15
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1