the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Jun 2024
Isolating aerosol-climate interactions in global kilometre-scale simulations
Abstract. Anthropogenic aerosols are a primary source of uncertainty in future climate projections. Changes to aerosol concentrations modify cloud radiative properties, radiative fluxes and precipitation from the micro to the global scale. Due to computational constraints, we have been unable to explicitly simulate cloud dynamics, leaving key processes, such as convective updrafts parameterized. This has significantly limited our understanding of aerosol impacts on convective clouds and climate. However, new state-of-the-art climate models running on exascale supercomputers are capable of representing these scales. In this study, we use the kilometre-scale earth system model ICON to explore, for the first time, the global response of clouds and precipitation to anthropogenic aerosol via aerosol-cloud-interactions (ACI) and aerosol-radiation-interactions (ARI). In our month-long simulations, we find that the aerosol impact on clouds and precipitation exhibits strong regional dependence, highlighting the complex interplay with atmospheric dynamics. The impact of ARI and ACI on clouds in isolation shows some consistent behaviour, but the magnitude and additive nature of the effects are regionally dependent. This behaviour suggests that the findings of isolated case studies from regional simulations may not be representative, and that ARI and ACI processes should both be accounted for in modelling studies. The simulations also highlight some limitations to be considered in future studies. Differences in internal variability between the simulations makes large-scale comparison difficult after the initial 10 – 15 days. Longer averaging periods or ensemble simulations will be beneficial for perturbation experiments in future kilometre-scale model simulations.
- Preprint
(53814 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1689', Anonymous Referee #1, 15 Jul 2024
This study investigates the impacts of anthropogenic aerosols on radiation and cloud properties using a global convection-permitting model. Four 40-day simulations are conducted to isolate the aerosol impacts via the aerosol-cloud interaction and the aerosol-radiation interaction. The methodology and analysis are straightforward. However, several severe flaws in the manuscript prevent its publication in its current form.
Major comments:
- Sections 2.2 and 2.3 are unclear. I don’t understand what the authors changed for the three PD simulations. Only aerosol concentrations? Please provide more details.
- The authors mentioned deep convection throughout the manuscript but never provided any quantitative analysis about how much of the responses are for convection or large-scale environments. Even if the authors select several regions with frequent deep convection, I don’t think convection is the only precipitating process in those regions. For example, how do you know how much precipitation is from deep convection?
- The discussion of the results is verbose, while the reasoning is oversimplified throughout the manuscript. The latter severely degrades the quality of the study. Please explain your results more logically, not just provide some simple assumptions. Please find further details below.
Minor comments:
Line 7: Please provide the full name of “ICON” for its first appearance.
Line 39: What do you mean by lack of memory?
Line 43: “Limited area” to “Regional”?
Line 44-45: What do you mean by large-scale controls on the availability of energy and water vapor?
Line 62: What is the resolution of NICAM?
Line 66: “or” to “nor”?
Lines 74 and 75: The full names of ICON and MACv2-SP?
Line 112: add “separately” after “cloud water.”
Line 115: What is SLEVE?
Lines 140-141: Any references for such an adjustment?
Line 152: Figure 1 shows an overestimation of Nd. Do you have any statistical calculations validating the changes of aN and bN?
Section 2.2: The description of aerosols in ICON needs more detailed clarification. Which type of aerosol variables does ICON need? How do you consider pre-industrial and present-day aerosol conditions? The current description is confusing. Did you only add fN in the model? Where are Nd,cld and Nd,sfc from? Does ICON contain any aerosol microphysical processes (processes converting emissions to aerosol concentrations), or are aerosols entirely prescribed in the model (input aerosol concentrations, AOD, etc.)?
Lines 172-178: I am confused with what MACv2-SP provided to ICON. Anthropogenic aerosol concentrations?
Lines 179-180: Please rewrite this sentence.
Lines 179-185: Both PI and PD simulations contain biomass-burning sources. What are their differences? In Line 132, you mentioned that in MACv2-SP, biomass-burning emissions are anthropogenic sources. Could you please provide more details about that?
Lines 212-213: Can internal variability be eliminated by only focusing on regional responses?
Line 229-230: How did you know it is due to internal variability but not actual model differences? Internal variability refers to variability due to natural internal processes within the climate system. But here, you talked about the differences between the two simulations.
Line 234: Again, how did you know it is due to internal variability?
Line 235: add “outgoing” before “shortwave radiation.”
Line 253: “internal variability” to “spatial heterogeneity.” Please check the whole manuscript to ensure the correct terms are used.
Lines 269-271: Would applying the tool to Section 3 (original model data but not differences) be better also to remove your so-called “internal variability”?
Lines 279-283: There is an assumption here: the long-term component is constant and independent of time. Is it correct? At least Figure 7f doesn’t show that!
Line 286: Delete “with the SE Atlantic region”.
Line 305: Do you have an explanation for the nighttime enhancement?
Line 319: Does ACI increase cloud cover? Did you refer to Figure 4h?
Line 345: Please provide more details about how you constrain ascending air masses. Using hourly or daily data?
Line 349: Why not decompose them given that you did it in Section 4.1?
Line 357: Does the profile of potential temperature changes align with the anthropogenic aerosol loading profile? Do you have any model output showing the suppression of boundary layer mixing and drying aloft? How do you define the boundary layer? 1 km? 2 km?
Line 360-362: Did you mean the daily mean or afternoon mean? If you meant afternoon mean, Figures 11a and 11d showed that Amazon has weaker w effects than Congo. If you meant daily mean, Figure 8 shows the ARI effect is the most significant in the afternoon. Please provide more reasonable descriptions and explanations for the results.
Line 367: How does the delayed release of CAPE increase high-altitude IWC?
Line 367-370: How did you know convection occurred certainly given increased CAPE aloft? More evidence is necessary to support such types of conclusions.
Lines 380-381: Which variables did you refer to?
Lines 381-383: Please provide more details about the reasoning.
Line 391: What do you mean by delayed CAPE? In addition, how can they explain the non-linear response?
Line 400: Is 1 mm day-1 consistent with Figure 9?
Line 415: Do you mean the red line? It is over zero at all hours.
Lines 416-417: ARI is positive in Figure 9h.
Lines 417-418: The ACI effect on LWP is always positive, while its effect on Mflux is negative only between 12:00 and 15:00 LST. How can you explain it?
Lines 422-428: Please provide more detailed reasoning!
Line 440-441: ACI and ARI are comparable in the ice-phase region!
Citation: https://doi.org/10.5194/egusphere-2024-1689-RC1 -
RC2: 'Comment on egusphere-2024-1689', Anonymous Referee #2, 15 Jul 2024
The authors take a crude aerosol parameterization (MACv2) and proceeded to parameterize it further for use in km-scale aerosol sensitivity studies. It’s still crude, though it seems to be more intelligently refined. Overall, the authors try to present these one-month-long km-scale simulation as a way to study generic features of aerosol radiative response. I don’t think they made a convincing case. As Reviewer 1 indicated, there are “several severe flaws” in this work. However, I would go further than Reviewer 1 to say the methodology is also flawed in that it is not fit for the purpose at hand.
I don’t think this manuscript should be published in ACP in its current form, and I do not think it is likely any revision along the lines of the submitted manuscript will make it closer to acceptable form. Besides the points raised by Referee 1, here are additional points that will likely prevent this manuscript from proceeding further.
- The authors should release the code for the model itself, the code edits required for their specific MACv2 setup, and the resulting MACv2 files (Nd, optics, etc.) as part of this manuscript. The analysis code released is not sufficient for reproducibility (anyone can make up a few netcdf files). I understand there may be limitation at some European centers regarding data/code transparency, but there are workarounds. If not possible, detailed explanation must be given why it is not possible. Note that this is not a request to release model outputs (though that can be nice; and the authors should definitely consider releasing those too), it is simply the underlying code/files that should be made public (as much as possible).
- The design of the simulations (namely 40-day DYAMOND cases) is not really appropriate for the aerosol response being targeted here. The authors make note of this serious issue several times in the manuscript (see below), but they somehow overcome it without much explanation, or maybe I missed it. Why do you think you say so much about ARI and ACI in 30-day runs? That doesn’t seem quite right to me. There’s simply too much noise (internal variability, etc.) in this for any result to be meaningful.
- More clarity about the one-way coupling here will be beneficial, and in so doing, it is important to highlight how limiting the setup is. By one-way coupling, I mean that MACv2 prescription affects the optics, radiation droplet number, and cloud droplet number, but nothing in the model affects the MACv2 prescription. Is that correct? If so, I think it should be highlighted more prominently — sections where we cannot say much definitively about what’s going on should be deleted (e.g., S 3). As a corollary, do you think the one-way coupling will make the response overestimated or underestimated?
Potentially, the authors can consider submitting aspects of this work to GMD (e.g., focusing on the modifications to MACv2 and maybe some aspects of remapping, etc.). For ACP, I think a refocused, less verbose manuscript can potentially be useful for the community. However, the manuscript should narrowly tackle what is possible and avoid less certain topics.
Below are some comments I wrote down while (re)reading the manuscript:
- L 23 and elsewhere: I would talk about ARI before ACI (because that’s the logical progression, think direct vs indirect).
- L 29–24: Hmm, is that really “a primary” one?
- L 73: What’s "well defined" here? And why do you feel the need to say so? Are you trying to say people before this study used poorly defined treatments?
- L 76–80: the manuscript is already tortuously long; these types of meaningless lines can be deleted.
- L 86: This is not a classic “AMIP” experiment by any stretch of definition, or am I confused? (Cf. L 121 and thereabouts.)
- Fig 3 and associated text: What’s the deal with the “upscaled” panel? Are you simply remapping the middle panel to 2-degree resolution? If so, and if you really want to discuss this, you will have to give more details about the remapping algorithm and all sorts of things associated with this. I don’t quite see the point of all of this though, so more motivation may be needed to begin with…
- L 210–215: I don’t think this is enough to circumvent these serious issues. Can you give more reasoning why you think you’d adequately address these challenges by doing something different?
- L 234–235: Yeah, or at least, we simply don’t know…
- S 4: Why do we need a global model to study regional responses? Maybe some regionally refined setup will be more useful (much cheaper) than we have here?
- L 450: I cannot find this reference; looks like it has not been published yet? If so, maybe this is an improper citation…
- L 454: I don’t think we can say that (“link”), due to all sorts of challenges (internal variability, etc.) related to the design.
- L 466: Future direction for…?
- L 466–467: I am not following — perhaps elaborate further here and elsewhere?
Citation: https://doi.org/10.5194/egusphere-2024-1689-RC2 -
RC3: 'Comment on egusphere-2024-1689', Anonymous Referee #3, 25 Jul 2024
Review of egusphere-2024-1689: Isolating aerosol-climate interactions in global kilometre-scale simulations
Summary: In this study, the authors utilized month-long, global simulations with 5 km horizontal grid spacings to assess the aerosol-radiative impact, the aerosol-cloud impact, and their combined effects, using current day and pre-industrial aerosol estimates. Aerosols are introduced into their model by affecting the number of cloud droplets, the cloud droplet effective radius, and through AOD impacts on radiation. While I commend the authors on running these computationally expensive, global simulations, I question whether their framework and model can be used to come to the process-level conclusions that are being drawn in this study. Furthermore, some of the methods and model set-up can be made clearer. Throughout their results, the authors provide plausible explanations for their results, but they often seemed to be speculative, as opposed to rooted in analysis from the model data. As such, while global, long-term, simulations of aerosol effects do provide unique science opportunities (such as impacts on the larger-scale features that can be better resolved with 5 km grid spacing), this manuscript seems to focus on more uncertain and less justifiable aspects of their simulations.
Major Comments and Concerns:
- The authors state that “the novel aspect of this study is the globally resolved deep convection” (L262). However, 5 km grid spacing does not resolve most deep convection (i.e., isolated and scattered deep convective clouds), so the authors should reconsider framing their study in this way. This is particularly concerning given that the authors choose their regions of analysis due to having deep convection. This may require a shift to better resolved features in their simulations.
- Many studies have shown that aerosols have a clear diurnal cycle. Given the focus of this study’s analysis on the diurnal cycle of aerosol effects, can the authors comment on and justify their use of a time-independent aerosol perturbation that does not vary diurnally? This simplification seems especially concerning, given the focus on the diurnal cycle of aerosol effects in this study.
- L211-215: The authors state that they will only discuss the global scale briefly due to internal variability but can mitigate internal variability on a regional scale by compositing over multiple diurnal cycles. It is unclear why this same method cannot be applied globally, and whether this compositing can be used to reduce internal variability in these simulations. Internal variability comes up several more times in the manuscript, so being clearer about internal variability, and its impact on this study up front would be helpful.
- Aerosol-Radiation details. It seems that plumes of biomass burning emissions and industrial emissions are used. Do these aerosols have different radiative properties that are interacting with the radiative scheme? Where do the aerosol particles live in the vertical? Are they advected over the course of the simulation? The authors state that the nine plumes are configured to reproduce the AOD for the year 2005 (L134), but how are they configured? I think this section should have more details, which would improve the clarity of the experimental set-up.
- There have been countless limited-aera modeling studies of aerosol effects, specifically ones that focus on the regions highlighted in this study. However, there were only a few limited-area studies included in the authors’ introduction or through their results section. Process-level insights from additional limited area studies may help the authors disentangle and contextualize their results.
Minor comments:
- L4: “we have been unable to explicitly simulate cloud dynamics.” We have been able to in limited-area model studies, so can you please make this more explicit about global models.
- L44: What does “regional drivers” mean here?
- L97: It is unclear to me why the same number concentration cannot be used for both radiation and microphysics? This seems like an unnecessary source of inconsistency.
- L134: Unclear how the extrapolation is done here. Can you make clearer?
- L197-202: The authors state several important features are well-simulated in their simulation, but it is difficult to see from Figure 2 whether these features are being simulated properly.
- L215-225: Most of this section focuses on results from the spin-up phase. Why do the authors focus so much on the spin-up phase, as this phase is used to spin up their model?
- L267-272: The authors separate their time series into short-term and long-term components, which reduces contributions from internal variability but introduces “non-local impacts to the region.” More details are needed here in terms of how this reduces internal variability, and what is meant by non-local impacts.
- L275: The “second application of the decomposition tool with a large prescribed periodicity.” What is the prescribed periodicity, and how is this used to provide new, helpful information? More details would be helpful here.
- L349: Given the importance of removing internal variability, why aren’t Figures 10 and 11 decomposed with similar methods as the other figures?
- L391: The authors mentioned delayed CAPE as a potential process? Did the authors look at this in the simulations or is this speculation?
- Figures 8-11: Given that the author’s describe the results for each region at a time, it may be easier for the reader if the authors combined these figures based on region. I found it challenging to jump around from each figure to understand how these different variables come together to tell a consistent story.
Citation: https://doi.org/10.5194/egusphere-2024-1689-RC3
Data sets
Dataset for manuscript "Isolating aerosol-climate interactions in global kilometre-scale simulations" Ross J. Herbert https://zenodo.org/doi/10.5281/zenodo.11470778
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 501 | 128 | 26 | 655 | 16 | 17 |
- HTML: 501
- PDF: 128
- XML: 26
- Total: 655
- BibTeX: 16
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1