Review of “The Response of the Amazon Ecosystem to the Photosynthetically Active Radiation Fields: Integrating Impacts of Biomass Burning Aerosol and Clouds in the NASA GEOS ESM”

by Bian et al.

Summary:

There has been a growing number of studies looking at the potential impact on land carbon uptake from increased availability of diffused radiation associated with aerosol particles. Although conceptually simple, this effect is hard to quantify accurately as complex couplings between different components of the Earth system are at play. An Earth System Modelling (ESM) approach appears as a natural framework for this kind of problem, yet only a handful of studies using ESMs has been published so far. This submission from Bian et al. is therefore timely as they used the results from simulations performed with the NASA GEOS-ESM to analyze the impact of biomass burning aerosols on the Amazon rainforest gross primary productivity. The diffuse light fertilization effect from aerosols is not only uncertain, but it is also buffered by clouds as those compete with aerosols for radiation. This is in a way similar to pre-industrial natural aerosols controlling the amplitude of the anthropogenic aerosol radiative forcing. The role of clouds on the aerosol diffuse light fertilization effect has not been properly explored before and Bian et al. provide a novel quantification for this modulating effect.

It’s an interesting paper, easy to read, the structure is sound, and the supporting material is adequate. The work perfectly fits within the topics covered by ACP so I strongly support its publication after addressing those minor few points.

General Comments:

1. How was the 7 years period selected? Surely the observational datasets used for the evaluation cover a longer range so it could have been an opportunity to extend the statistics.
2. I would be tempted to rename the “fertilizer effect” as the “fertilizing effect” as fertilizer and effect are both nouns and it sounds odd to me. I’m not a native English speaker thought.
3. How is the seasonality (not just August – September) of clouds in your simulations - e.g. ICTZ temporal/spatial position which would affect moisture availability - when compared against observations? Same question for the seasonality of GPP/NPP. Have you evaluated the simulated seasonal cycle against the two products mentioned in the manuscript (i.e. FluxCOM and FluxSat)? Those could be potentially be added to the supplementary material.
4. The description of the representation of light interception by vegetated canopies (Line 235-237) could benefit from being described in a bit more details as this part of the model is central to the present study. Is it the same parameterization as the one in CLM4?
5. A bit off topic but I am curious as you mentioned photolysis and brown carbon developments in section 2. Are these two coupled? If so, is there any detectable impact on photolysis rates and consequently on ozone formation?
6. If I understand the experimental design correctly, the radiation driving the atmospheric model in callaer and cnobbaer for pair 2 (respectively allaer and nobbaer for pair 1) is the same, namely R1. Does that imply that there is potentially an implicit accounting of BB aerosol semi-direct effects in cnobbaer (respectively nobbaer), meaning that any potential change in cloudiness when doing a Diff between callaer and cnobbaer won’t be captured? Given that 2010 was a drought year with high BB emissions (see figure 10), could this lack of change in CF be meaningful and affect the slope of ddGPP/dAOD?
7. Another aspect of the experiment design that is not clear to me is whether the inputs (not just radiation) from the atmospheric model to Catchement-CN in nobbaer and cnobbaer are from the atmosphere that has experienced R1 or from the atmosphere that has experienced R2. If the former, as I understand it, it would mean that the vegetation only ‘feels’ the aerosols via the change in dir/diff radiation but not via other changes in the energy balance that aerosols also introduce. Can you clarify?

Specific Comments:

1. Line 29, change “the impact” to “this impact”.
2. Line 32, change “call” to “called”. I would however argue that the light fertilization effect is only one of the impacts (plural, as in the manuscripts) resulting from aerosol radiative effects (e.g. less surface radiation affects the energy balance hence has an impact on vegetation productivity too). Maybe this sentence could be slightly reworded.
3. Line 38, Replace “lost” by something like “average loss” otherwise it becomes slightly ambiguous (i.e. lost would correspond to the total loss over 7 years, so ~7 times 250-300 TgC/yr).
4. Line 39, replace “is highest for” by either “is higher in” or “is at the highest in”
5. Line 50, remove “dioxide”, carbon has been sequestrated in different molecular forms by the vegetation.
6. Line 59, “It is in the dry season, when light becomes a key-controlling”. Could be reworded. Light is a key controlling factor outside of that period as well, but it is probably less of a bottleneck.
7. Line 100, technically they accounted for the variability in cloudiness as those were free running simulations. However, the authors have not explicitly quantified the impact of this variably on the aerosol light fertilisation effect.
8. Line 108, The definition of CI is not clear. Does it correspond to the ratio of total (i.e. dif + dir) light at surface over the total light at ToA ? Please clarify.
9. Line 129, “sunlight … drenches the trees due to reduced rain”. If by “drenches” you mean “floods the canopy with light”, I would use a different verb as “to drench” means “to wet thoroughly” and this is in contradiction with rest of the sentence (e. “having lesser rain”).
10. Line 184, Replace “augmentation” with development.
11. Line 259, can probably remove “site-level” as in situ literally means on site.
12. Line 294, can replace “a regional and a time evolution” with “both a spatial and temporal view”
13. Line 296, remove “s” from observations-based
14. Line 296, put name of products in bracket after mentioning there are two.
15. Line 296, reword end of sentence for clarity g. “ecosystem productivity in the GEOS simulations”
16. Line 297, move “Through upscaling using machine learning methods (Jung et al., 2020)” to the end of the
17. Line 312-31 It’s a nudged run basically isn’t it?
18. Line 321, maybe I missed it in the model description (section 2), are aerosols externally mixed in GEOS-GOCART?
19. Line 330 to 344, I hope I got this correctly, in nobbaer, R2 is only used by the physiology part of Catchment-CN, is that correct? Meaning that the rest of the land surface energy budget is calculated assuming R1 as in allaer.
20. Fig 1b is the same as Fig 1a. State in the legend what the error bars on 1c represent.
21. Legend for Fig 2a and 2c specify the wavelength used for those AOD.
22. Line 407, it would help the reader if the box defined here was depicted on some (all) of the contour Are all spatial averages quoted in the manuscript calculated over the same area?
23. Line 421, maybe a reference here would be useful. Is vegetation not dark enough for MODIS dark target algorithm to perform well?
24. Line 480 to 492. To avoid confusion, it would be better have full indexing for the radiative quantities, g. Rtot@toa instead of Rtop, Rdiff@srf instead of Rdiff …
25. Line 480 to 492, As the notation has changed from the rest of the manuscript, I believe the radiation quantities here are integrated over the full SW spectrum not just PAR, is that correct? Please clarify in text.
26. Line 480 to 492, how were the direct and diffused components calculated here? Was any delta-rescaling of the aerosol optical properties still applied?
27. Fig 7 legend, Should the unit for GPP be kg/m2/s?
28. Fig 7, A fifth timeseries representing the fractional / absolute change in GPP could be useful here.
29. Fig 7, Although the GPP for the site at 54W 15S is relatively high in GEOS (Fig 5C), it does not seem to be a very productive pixel in both FluxSat and FluxCom (Fig 5a and 5b) which makes sense as this is probably in the arc of deforestation. Is the tile in the model mostly covered by tall canopy PFTs or is dominated by lower grass?
30. Line 630, can be more specific and replace “quantities” by “GPP”.
31. Fig 10 legend, BBAOD is labelled as the brown carbon part of the biomass burning aerosol, is that correct?
32. Fig 10 legend, remove “for the ecosystem”. Replace “dot-lines” by “dashed lines”.
33. Fig 10 legend, replace “occurrence frequency” by “frequency of occurrence”.
34. Fig 10 legend, “dGPP is 119.5% (201008) and 92.6%”, hard to tell without seeing it plotted but from a quick and dirty “visual” interpolation, these numbers seems high. Can you confirm them?
35. Line 628-29, sentence is already in Fig 10 legend.
36. Line 635, Isn’t this strong correlation to be expected from the experiment design? The atmospheric state should be pretty similar in allaer and nobbaer. The main ESM feedback allowed in nobbaer is from the vegetation that has received R2 instead of R1 isn’t it? (see general comment #7 and specific comment #19). Anyway, such feedback probably has a negligeable impact on cloud fields (see g. Pedruzo-Bagazgoitia et al. 2017).
37. Line 632-643, In this discussion, it would be useful to quote the GPP changes (both relative and absolute) averaged over the analysis period, maybe excluding 2010 as it is an unusual year. Additionally, these period mean quantities for GPP, DFPAR, DRPAR, AOD, CLDFRC could be incorporated in a table in the main manuscript (table S1/2 a,b,c,d are useful but clearly too detailed for the main manuscript).
38. Line 647-652, This sentence is confusing. If cloudiness is similar in 2011 and 2010, how can you conclude from this that its effects are of second order compared to aerosol effects. Surely it won’t have an impact if it doesn’t change.
39. Line 653, a followed up to comment #36. This conclusion is based on the simulation results. Other environmental controls might have affected the vegetation productivity in the real world. Could you further support your conclusion by using the observational proxies (i.e. FluxCOM and FluxSat)? One way could be to compute a GPP anomaly as GPP for 2010 minus the mean of GPPs for the other years (excluding 2010) for FluxCOM, FluxSat and the allaer simulation. Then repeat the same calculation for 2011. Seven years might be a bit short for getting a representative mean, but I believe these two datasets have records longer than the analysed period.
40. Line 657 to 664, This is the part of the paper where the most novel results are introduced. This paragraph could benefit from having some further explanation on the physical meaning of the anomalies calculated here. I sort of understand ddX/dAOD as something similar to a radiative forcing (i.e. a difference in net radiative fluxes between a pair of simulations). Maybe it could be called susceptibility of the diffuse fertilization effect to BB aerosols. The double “dd” notation is a bit confusing and could be improved. It would be useful to use the same notation for the denominator (AOD) although I appreciate that the background aerosols in allaer and callaer cancel each other.
41. Fig 11, Instead of (or in addition to) the mean values for 2013/15, it could be useful to have a +/- one standard deviation marks around each means at each bin points. I expect much more variability in ddGPP/dAOD at low CF than at high CF. If so, there might be a larger range of CF where the aerosol effect could be detrimental to plant growth.

Additional reference:

Pedruzo-Bagazgoitia, X., Ouwersloot, H. G., Sikma, M., vanHeerwaarden, C. C., Jacobs, C. M., and Vilà-Guerau deArellano, J.: Direct and Diffuse Radiation in the Shal-low Cumulus–Vegetation System: Enhanced and Decreased Evapotranspiration Regimes, J. Hydrometeorol., 18, 1731–1748, https://doi.org/10.1175/JHM-D-16-0279.1, 2017.

Review of “The Response of the Amazon Ecosystem to the Photosynthetically Active  Radiation Fields: Integrating Impacts of Biomass Burning Aerosol and Clouds in the NASA GEOS ESM” by Huisheng Bian et al.

General comments:

This study uses the NASA GEOS Earth System Model framework to investigate the impact of biomass burning aerosols and cloud cover on the Amazon region ecosystem productivity. This is a very interesting topic and the paper is clearly structured and generally well written.

However, while this work could certainly bring an important contribution to existing published studies on this topic, I think in its current form it still needs major revisions. I really hope this is something the authors can and will address in a revised manuscript.

Major comments:

1. A key question that needs to be addressed is whether the simulated response of GPP to changes in diffuse radiation fraction is realistic. More specifically, does the model accurately simulate observed GPP response to changes in diffuse, direct, and total surface radiation? And how does this simulated GPP response compare with other existing model estimates?

2. Why is the role of other climatic feedbacks associated with biomass burning aerosol emissions (e.g. reduction in leaf temperature) completely ignored, despite the fact that an ESM is used? While, the authors do acknowledge at the end of the paper (lines 716-719) that the aerosol induced changes in meteorological fields can also affect plant growth, this seems to be a huge missed opportunity here. Malavelle et al. (2019) showed that the overall impact of biomass burning aerosols on NPP is the net result of multiple competing effects and it would be interesting to see if similar responses are simulated with the NASA GEOS ESM system.

3. The second research objective (and the way it is addressed) is a bit unclear and it should be formulated and addressed much more clearly.

• 3a. It is evident that clouds have a substantial impact on the efficiency of the aerosol diffuse radiation effect, as they have a strong effect on diffuse radiation fraction. In a similar way it can be said that the aerosols have an impact on the efficiency of the diffuse radiation effect caused by clouds. So this in itself is not necessarily a research question.
• 3b. Is there a difference in the model between the simulated GPP response to changes in diffuse radiation fraction caused by aerosol changes and those caused by cloud cover changes?
• 3c. The fact that during the investigated period (lines 649-654), the interannual variation in regional cloudiness is small and therefore plays only a secondary role on the diffuse radiation fertilisation effect (compared to the dominant role played by the variation in biomass burning aerosol) is not surprising and does not really address the second research objective.
• 3d. Lines 657-682: The cause for the difference between the 2013 and 2015 lines in Figure 11 is suggested to be the difference in cloud cover between the two years. I wonder whether this is indeed the case, since the results illustrated in Figure 11 are in fact for binned cloud fractions anyway? I would speculate they are in fact caused by the difference in (i) biomass burning emissions (they do matter in your calculated ddX/dBBAOD, which is defined in terms of both Pair1 and Pair2) and (ii) temperature and precipitation. This needs to be investigated and clarified.

Specific comments:

1. Terminology: Why is the term “aerosol light fertilizer effect“ being used instead of other already established terminology, e.g. diffuse radiation fertilization effect, Mercado et al (2009). I suggest the use of existing terminology to better integrate the work with other studies, but if the authors feel strongly about introducing this new terminology, a clear rationale for this should be provided.
2. Why was the this particular period (i.e. 2010-2016) chosen? Can this be extended?
3. Lines 100-101 and 140-142: It is not quite true that Malavelle et al (2019) did not consider the effect of clouds altering the diffuse radiation fertilisation effect. They do in fact discuss this and mention in their paper that “despite cloudiness affecting how much aerosols can interact with radiation, we notice that NPP is enhanced in the central part of the Amazon when BBA emissions are increased (Fig. 5).” So this needs to be reformulated and clarified in this paper to avoid confusions. This points also relates to my major comment 3, i.e. the need to better define and address the second research objective.
4. Lines 140-142: The authors seem to have missed other relevant studies on this topic, such as Strada and Unger (2016) and Unger et al. (2017). Results presented in this work should also be compared and integrated with those from these other studies.
5. Lines 403-406: It would be good to investigate a bit more the cause of the difference in observed and simulated SSA in August at Alta Floresta. What about other periods and other sites?
6. Lines 458-461: Only comparing averages over Aug-Oct 2010-2016 for simulated SW radiation and CERES measurements can potentially mask important differences. Please include an assessment and discussion of model vs measurements agreement for the full time series (e.g. 2010-2016 time series of monthly means).
7. Lines 470-477: Similarly to the evaluation of SW radiation, the evaluation of simulated GPP should be investigated in more detail, i.e. time series rather than just averages.
8. Figure 6: I would suggest the add another line corresponding to total radiation (i.e. the sum of the blue and red lines). This should help the discussion and better illustrate the point.
9. Lines 479-516: This simulated response of total and diffuse surface radiation to different aerosol concentrations and cloud conditions needs to be evaluated against some observations. This is a key process to get right for this study and is not currently addressed in the paper. This relates to my major comment 1.
10. Figure 8 and lines 580-586: An Amazon regional average GPP increase of +9.9% resulting from an increase in DFPAR of 10% is substantially larger than other existing estimates, e.g. Rap et al. (2015), Malavelle et al. (2019). However, the corresponding percentage change in NPP (lines 597-601) seems closer to estimates from other studies. It is important to investigate this further and include a discussion on why this is the case (e.g. to what extent the GPP change is driven by changes in respiration and NPP, respectively). This point also relates to my major comment 1, regarding the need to validate the simulated GPP response to changes in diffuse/total radiation against observations and/or other existing model estimates.
11. Lines 605-611: The comparison with Rap et al. (2015) is incorrect and misleading. Firstly, it is incorrect because the 0.5-4.2% range of NPP change in this study is an interannual range, while the 1.4-2.8% range from Rap et al. (2015) is an uncertainty range for the 1998-2007 average due to biomass burning emissions uncertainty. The actual interannual range from Rap et al. (2015) can be inferred from their Fig. 4 and Fig. S5. Secondly, it is misleading as the two periods are different (2010-2016 vs 1998-2007), so any comparison of interannual ranges should also include a discussion on the interannual variability in biomass burning emissions during 1998-2016.
12. Lines 702-703: “The cloud fraction at which BB aerosol switches from stimulating to inhibiting plant growth occurs at ~0.8.” I think this is a potentially confusing statement as it only applies to the biomass burning aerosol loadings recorded during the period investigated here. In reality, as both cloud cover and aerosol concentrations affect the diffuse radiation fraction, this threshold does also depend on the aerosol loading. A more useful threshold would be one defined in terms of diffuse radiation fraction.

Technical corrections:

1. Line 32: “call here” should be “called here”.
2. Line 124: missing supporting citation for the 40% value.
3. Line 354: typo “metrological”
4. Lines 625-631: Description of figure is best included in the figure caption, with manuscript text dedicated to discussion of results.
5. Lines 641-643: Please reformulate to avoid using “presumably” which is a bit too vague. A more precise statement would read much better.
6. Lines 666-673: Why is a different font used in this paragraph?
7. Lines 701-702: “Curiously, BB aerosols stimulate plant growth under clear-sky conditions but suppress it under full cloudiness conditions”. I suggest removing the word “curiously”? This is in fact to be expected.

References:

Strada, S. and Unger, N.: Potential sensitivity of photosynthesis and isoprene emission to direct radiative effects of atmospheric aerosol pollution, Atmos. Chem. Phys., 16, 4213–4234, https://doi.org/10.5194/acp-16-4213-2016, 2016. 

Unger, N., Yue, X., and Harper, K. L.: Aerosol climate change effects on land ecosystem services, Faraday Discuss., 200, 121–142, https://doi.org/10.1039/c7fd00033b, 2017.