the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
What controls the historical timeseries of shortwave fluxes in the North Atlantic?
Daniel Peter Grosvenor
Kenneth S. Carslaw
Abstract. Both aerosol radiative forcing and cloud-climate feedbacks have large effects on climate, mainly through modification of solar shortwave radiative fluxes. Here we determine what causes the long-term trends in the shortwave (SW) top-of-the-atmosphere (TOA) fluxes (FSW) over the North Atlantic region. The UK Earth System Model (UKESM1) and the Hadley Centre General Environment Model (HadGEM) simulate a positive FSW trend between 1850 and 1970 (increasing SW reflection) and a negative trend between 1970 and 2014. We find that the pre-1970 positive FSW trend is mainly driven by an increase in cloud droplet number concentrations due to increases in aerosol and the 1970–2014 trend is mainly driven by a decrease in cloud fraction, which we attribute mainly to cloud feedbacks caused by greenhouse gas-induced warming.
Using nudged simulations where the meteorology can be controlled we show that in the pre-1970 period aerosol-induced cooling and greenhouse gas warming in coupled atmosphere-ocean simulations roughly counteract each other so that aerosol forcing is the dominant effect on FSW, with only a weak temperature-driven cloud feedback effect. However, in the post-1970 period the warming from greenhouse gases intensifies and aerosol radiative forcing falls, leading to a large overall warming and a reduction in FSW that is mainly driven by cloud feedbacks. Our results show that it is difficult to use satellite observations in the post-1970 period to evaluate and constrain the magnitude of the aerosol-cloud interaction forcing, but that cloud feedbacks might be evaluated.
Comparisons to observations between 1985 and 2014 show that the simulated reduction in FSW and the increase in temperature are too strong. However, analysis shows that this temperature discrepancy can account for only part of the FSW discrepancy given the estimated model feedback strength (λ = ∂FSW/∂T). This suggests a model bias in either λ or in the strength of the aerosol forcing (aerosols are reducing during this time period) is required to explain the too-strong decrease in FSW. Both of these biases would also tend to cause temperature increases over the 1985–2014 period that are too large, which would be consistent with the sign of the model temperature bias reported here. Either of these model biases would have important implications for future climate projections using these models.
Daniel Peter Grosvenor and Kenneth S. Carslaw
Status: closed
-
RC1: 'Review of Grosvenor and Carslaw', Anonymous Referee #1, 26 Sep 2022
In this study, the authors used the HadGEM3/UKESM1 climate models to understand how and why shortwave fluxes have changed over the North Atlantic Ocean over the CMIP6 historical period (1850-2014). They identify two periods where the trend in outgoing shortwave flux, and the causes for that trend, differ. The first period is 1850-1970, characterized by an increase in outgoing shortwave flux, which the authors link in the model to an aerosol-driven increase in cloud droplet number. The second period, 1970-2014, sees a decrease in outgoing shortwave flux, explained by a feedback of greenhouse gas warming on cloud fraction. The analysis also contains a comparison of the models to relevant observations, and comparisons between different simulations.
The analysis is very thorough and proceed in well-defined steps. Figures illustrate the discussion well, and Table give detailed numbers. The paper is very well written, although is a challenging read because of the high density of information that is discussed. The findings have interesting implications on the use of observations to constrain aerosol forcing, given that concurrent, non-aerosol cloud feedbacks are also present.
Given that HadGEM3 and UKESM1 are tightly related, the study is not far from being a single-model study, but that is justified because the depth of analysis and the need for additional simulations make the work difficult to replicate in a multi-model context. The methodology is interesting, with a complicated gymnastic of double differences between simulations, especially in section 3.5.
The main weakness of the study, which is implicitly acknowledged by the authors, is that the implications for the real world are difficult to identify. The comparison to observations arrives late and potentially invalidates a lot of what the manuscript said up to that point. We know that observations of trends are unfortunately insufficient to constrain aerosol radiative forcing and climate sensitivity, so the authors cannot satisfactorily unravel (lines 619-623) what the comparison to observations means for the preceding analysis. Similarly, the differences between the AerChemMIP and DAMIP simulations discussed in Appendix A, which are not really understood, make the findings fragile.
I am not sure how to mitigate that weakness. The paper could be built the other way around, perhaps, dealing with model-observations and model-model differences first. That could force the discussion to account for the implications more explicitly. But it will always be the case that results and discussions in sections 3.5 and 4 leave many questions open. Perhaps simply be more upfront in acknowledging the issue in the abstract and conclusion?
Other comments:
Line 166: “as realistic as possible” is an optimistic view of the process of creating historical emission datasets. It should be noted that CEDS emissions have gone through many revisions, which show sizeable changes even in sulfur dioxide emissions compared to the version documented in Hoesly et al (2018). The simulations presented here may not use the latest version.
Line 228: Nit-pick, but cloud fraction does not affect cloud albedo. It affects planetary albedo.
Table 1: Which are the HadGEM simulations?
Captions of Figures 1 and 3: Why “intermodel standard deviation”? There is only one model in that calculation.
Lines 301-304 and Table 2: Any idea why the offline calculations do a poorer job after 1970? Offline calculations must miss an input that becomes important then. Perhaps something to do with water vapour or gaseous absorption? And offline calculations seem to have less variability – I suppose that is to be expected?
Line 380: The lack of an impact from natural aerosols is partly by construction. Sea salt is the natural aerosol most likely to affect liquid cloud albedo. Its historical trend in the North Atlantic region is probably limited to locations where sea ice becomes open oceans, but those regions are excluded from the analysis.
Lines 409-413: The definition of climate feedbacks introduced here is unclear. Does that include rapid adjustments? The remainder of the work suggests adjustments are excluded, so I suggest rephrasing to make that clearer. IPCC practice is to see feedbacks as the climate response mediated by a large-scale change in temperature – that is, excluding rapid adjustments by definition.
Line 416: I suggest updating that discussion by basing it on the AR6 assessment, for example Table 7.4 on CO2 rapid adjustments. That will not change the conclusion that rapid adjustments to aerosol forcing are larger than those to greenhouse gases.
Caption of Figure 9: I do not think that this caption is the place for explaining how feedbacks are estimated. I would suggest adding a diagram or a table showing how forcing, feedbacks etc. are separated from double differences between pairs of simulations.
Line 534: Could cite Andrews et al. (2019) https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019MS001866 here. Mid-latitude cloud feedbacks are indeed listed as a possible source of the large ECS in HadGEM3/UKESM1.
Line 578: It should be noted here that the strength aerosol forcing has been decreased during model development (Mulcahy et al. 2019). It does not follow that aerosol forcing is still too strong, but it shows that an excessive strength is a long-standing concern of the developers of the model.
Technical comments:
Line 135: Typo cooilng
Line 234: I do not think that tau_c and r_e have been defined, but I may have missed them.
Line 249: Parenthesis is not closed.
Caption of Figure 13: Typo atmosphere
Line 665: Typo simlar
Line 712: Missing words “lack reduction”
Line 782: Typo tempeatures
Citation: https://doi.org/10.5194/acp-2022-583-RC1 -
RC2: 'Comment on acp-2022-583', Anonymous Referee #2, 12 Oct 2022
The authors present a very thorough investigation into the behaviour of shortwave radiation fluxes above the North Atlantic, over the historical era, in two versions of the UK climate model (UKESM, HadGEM3). They find two regimes with markedly different behaviour; before and after 1970; and attribute them (primarily) to an increase in aerosol concentrations and cloud responses to sufrace temperature change, respectively.
Overall, this is an impressively detailed study, with well described reasoning and broad ranging but established methods. It reads almost like a textbook at times, taking the reader through all main factors thought to be able to influence F_SW and disentangling their various influences. The analysis and the manuscript are clearly very well worked through, and I therefore have very little to offer in terms of deeper feedback. This paper could well be published as-is, and should certainly not require more than a minimal revision.
Some minor questions and comments:
* If I have one concern with the paper, it is that it is very long and at times quite wordy. There is a risk that the nice and highly instructve results get lost because the community doesn't have time to read through it all. Therefore: Would it be an idea to include a process level schematic of the factors contributing before and after 1970? I.e. an annotated version of Figure 1a, with some arrows and icons, to show what is changing and why?
* Unless I'm missing it, I don't think you discuss the temperature fedback on sea salt aerosols as a potential contributor to F_SW trends? This effect should be there for the North Atlantic, at least for the post-1970 period, somewhat counteracting the reduction in anthropogenic CCN. (You have the DAMIP natural forcer experiments, and show that it can be ignored in this context, but that simulation will not have the natural aerosol feedbacks.)
* In section 2.3, I can't quite see that you've quantified the impact of excluding grid boxes with sea ice formation. Presumably the effect is small, but could it introduce some biases? (Domination of southern grid boxes, or spurious seasonality?)
* Figure 1, and others: Would it be worth also showing NA AOD? Simply because there are so many other studies that use AOD, and therefore it becomes easier to compare your results to theirs?
* Figures 4, 7, 8, ...: In many of these, one dot is the net of others. Could this be highlighted more clearly, with colors, symbols or similar? (You do this for aerosols in places, but I still struggled a bit to understand how all the factors summed up - or not - in the various figures.)
* Your convention is that F_SW is upwelling; I got this after reading a bit, but I don't think you explicitly define it? Maybe have it already in the abstract, line 5? ("positive upwelling F_SW trend"?)
* The last reference (Zhou et al. 2016) comes twice in the ref.list.Thanks for a very interesting paper.
Citation: https://doi.org/10.5194/acp-2022-583-RC2 - AC1: 'Response to reviewers (acp-2022-583)', Daniel Grosvenor, 24 Feb 2023
Status: closed
-
RC1: 'Review of Grosvenor and Carslaw', Anonymous Referee #1, 26 Sep 2022
In this study, the authors used the HadGEM3/UKESM1 climate models to understand how and why shortwave fluxes have changed over the North Atlantic Ocean over the CMIP6 historical period (1850-2014). They identify two periods where the trend in outgoing shortwave flux, and the causes for that trend, differ. The first period is 1850-1970, characterized by an increase in outgoing shortwave flux, which the authors link in the model to an aerosol-driven increase in cloud droplet number. The second period, 1970-2014, sees a decrease in outgoing shortwave flux, explained by a feedback of greenhouse gas warming on cloud fraction. The analysis also contains a comparison of the models to relevant observations, and comparisons between different simulations.
The analysis is very thorough and proceed in well-defined steps. Figures illustrate the discussion well, and Table give detailed numbers. The paper is very well written, although is a challenging read because of the high density of information that is discussed. The findings have interesting implications on the use of observations to constrain aerosol forcing, given that concurrent, non-aerosol cloud feedbacks are also present.
Given that HadGEM3 and UKESM1 are tightly related, the study is not far from being a single-model study, but that is justified because the depth of analysis and the need for additional simulations make the work difficult to replicate in a multi-model context. The methodology is interesting, with a complicated gymnastic of double differences between simulations, especially in section 3.5.
The main weakness of the study, which is implicitly acknowledged by the authors, is that the implications for the real world are difficult to identify. The comparison to observations arrives late and potentially invalidates a lot of what the manuscript said up to that point. We know that observations of trends are unfortunately insufficient to constrain aerosol radiative forcing and climate sensitivity, so the authors cannot satisfactorily unravel (lines 619-623) what the comparison to observations means for the preceding analysis. Similarly, the differences between the AerChemMIP and DAMIP simulations discussed in Appendix A, which are not really understood, make the findings fragile.
I am not sure how to mitigate that weakness. The paper could be built the other way around, perhaps, dealing with model-observations and model-model differences first. That could force the discussion to account for the implications more explicitly. But it will always be the case that results and discussions in sections 3.5 and 4 leave many questions open. Perhaps simply be more upfront in acknowledging the issue in the abstract and conclusion?
Other comments:
Line 166: “as realistic as possible” is an optimistic view of the process of creating historical emission datasets. It should be noted that CEDS emissions have gone through many revisions, which show sizeable changes even in sulfur dioxide emissions compared to the version documented in Hoesly et al (2018). The simulations presented here may not use the latest version.
Line 228: Nit-pick, but cloud fraction does not affect cloud albedo. It affects planetary albedo.
Table 1: Which are the HadGEM simulations?
Captions of Figures 1 and 3: Why “intermodel standard deviation”? There is only one model in that calculation.
Lines 301-304 and Table 2: Any idea why the offline calculations do a poorer job after 1970? Offline calculations must miss an input that becomes important then. Perhaps something to do with water vapour or gaseous absorption? And offline calculations seem to have less variability – I suppose that is to be expected?
Line 380: The lack of an impact from natural aerosols is partly by construction. Sea salt is the natural aerosol most likely to affect liquid cloud albedo. Its historical trend in the North Atlantic region is probably limited to locations where sea ice becomes open oceans, but those regions are excluded from the analysis.
Lines 409-413: The definition of climate feedbacks introduced here is unclear. Does that include rapid adjustments? The remainder of the work suggests adjustments are excluded, so I suggest rephrasing to make that clearer. IPCC practice is to see feedbacks as the climate response mediated by a large-scale change in temperature – that is, excluding rapid adjustments by definition.
Line 416: I suggest updating that discussion by basing it on the AR6 assessment, for example Table 7.4 on CO2 rapid adjustments. That will not change the conclusion that rapid adjustments to aerosol forcing are larger than those to greenhouse gases.
Caption of Figure 9: I do not think that this caption is the place for explaining how feedbacks are estimated. I would suggest adding a diagram or a table showing how forcing, feedbacks etc. are separated from double differences between pairs of simulations.
Line 534: Could cite Andrews et al. (2019) https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019MS001866 here. Mid-latitude cloud feedbacks are indeed listed as a possible source of the large ECS in HadGEM3/UKESM1.
Line 578: It should be noted here that the strength aerosol forcing has been decreased during model development (Mulcahy et al. 2019). It does not follow that aerosol forcing is still too strong, but it shows that an excessive strength is a long-standing concern of the developers of the model.
Technical comments:
Line 135: Typo cooilng
Line 234: I do not think that tau_c and r_e have been defined, but I may have missed them.
Line 249: Parenthesis is not closed.
Caption of Figure 13: Typo atmosphere
Line 665: Typo simlar
Line 712: Missing words “lack reduction”
Line 782: Typo tempeatures
Citation: https://doi.org/10.5194/acp-2022-583-RC1 -
RC2: 'Comment on acp-2022-583', Anonymous Referee #2, 12 Oct 2022
The authors present a very thorough investigation into the behaviour of shortwave radiation fluxes above the North Atlantic, over the historical era, in two versions of the UK climate model (UKESM, HadGEM3). They find two regimes with markedly different behaviour; before and after 1970; and attribute them (primarily) to an increase in aerosol concentrations and cloud responses to sufrace temperature change, respectively.
Overall, this is an impressively detailed study, with well described reasoning and broad ranging but established methods. It reads almost like a textbook at times, taking the reader through all main factors thought to be able to influence F_SW and disentangling their various influences. The analysis and the manuscript are clearly very well worked through, and I therefore have very little to offer in terms of deeper feedback. This paper could well be published as-is, and should certainly not require more than a minimal revision.
Some minor questions and comments:
* If I have one concern with the paper, it is that it is very long and at times quite wordy. There is a risk that the nice and highly instructve results get lost because the community doesn't have time to read through it all. Therefore: Would it be an idea to include a process level schematic of the factors contributing before and after 1970? I.e. an annotated version of Figure 1a, with some arrows and icons, to show what is changing and why?
* Unless I'm missing it, I don't think you discuss the temperature fedback on sea salt aerosols as a potential contributor to F_SW trends? This effect should be there for the North Atlantic, at least for the post-1970 period, somewhat counteracting the reduction in anthropogenic CCN. (You have the DAMIP natural forcer experiments, and show that it can be ignored in this context, but that simulation will not have the natural aerosol feedbacks.)
* In section 2.3, I can't quite see that you've quantified the impact of excluding grid boxes with sea ice formation. Presumably the effect is small, but could it introduce some biases? (Domination of southern grid boxes, or spurious seasonality?)
* Figure 1, and others: Would it be worth also showing NA AOD? Simply because there are so many other studies that use AOD, and therefore it becomes easier to compare your results to theirs?
* Figures 4, 7, 8, ...: In many of these, one dot is the net of others. Could this be highlighted more clearly, with colors, symbols or similar? (You do this for aerosols in places, but I still struggled a bit to understand how all the factors summed up - or not - in the various figures.)
* Your convention is that F_SW is upwelling; I got this after reading a bit, but I don't think you explicitly define it? Maybe have it already in the abstract, line 5? ("positive upwelling F_SW trend"?)
* The last reference (Zhou et al. 2016) comes twice in the ref.list.Thanks for a very interesting paper.
Citation: https://doi.org/10.5194/acp-2022-583-RC2 - AC1: 'Response to reviewers (acp-2022-583)', Daniel Grosvenor, 24 Feb 2023
Daniel Peter Grosvenor and Kenneth S. Carslaw
Daniel Peter Grosvenor and Kenneth S. Carslaw
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 330 | 127 | 8 | 465 | 1 | 2 |
- HTML: 330
- PDF: 127
- XML: 8
- Total: 465
- BibTeX: 1
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1