Severe Biomass-Burning Aerosol Pollution during the 2019 Amazon Wildfire and Its Direct Radiative-Forcing Impact: A Space Perspective from MODIS Retrievals

Round 1

Reviewer 1 Report

This manuscript investigates biomass burning pollution during a heavy wildfire episode, or period, in 2019 and compares the results with those outside this period, including during the same time in 2018. The investigation is based on updated remote sensing product combined with radiative calculations. The methods and analysis appear scientifically sound, and the paper is well organized. I have a few relatively minors comments that should be considered before I can recommend accepting this paper for publication.

My main comment is related to the discussion of the obtained results. Currently, this discussion is quite technical, mainly focusing on presenting the numbers obtained from either remote sensing or radiative calculations. I would like to see more discussions about the potential implications of these results. For example, what is the meaning of strong atmospheric heating during this episode? Can it affect atmospheric stability or circulation, thereby having potential feedbacks on pollutant transport or dilutions? Furthermore, what is expected in the future? Are similar or even stronger fire episodes likely as a result of future anthropogenic activities in the regions, or as a result of changing climatic conditions?

 

Technical issues:

The authors should make sure that all the quantities in the equations are explained once the appear the first time. For example, now the wavelength is explained after equation 6 although it already appears in equation 1.

Table 1 claims that the meaning of the symbols in that table can be found in equations 1-4, but this seems not to be the case. Also, tables should be understandable by their own, so I would recommend explaining shortly these quantities in the table caption.

Although the text is relatively well written, there are small grammatical issues here and there. I recommend checking out the language of the paper once more after revising it.

Author Response

Response to Reviewer 1 Comments

 

 

This manuscript investigates biomass burning pollution during a heavy wildfire episode, or period, in 2019 and compares the results with those outside this period, including during the same time in 2018. The investigation is based on updated remote sensing product combined with radiative calculations. The methods and analysis appear scientifically sound, and the paper is well organized. I have a few relatively minors comments that should be considered before I can recommend accepting this paper for publication.

My main comment is related to the discussion of the obtained results. Currently, this discussion is quite technical, mainly focusing on presenting the numbers obtained from either remote sensing or radiative calculations. I would like to see more discussions about the potential implications of these results. For example, what is the meaning of strong atmospheric heating during this episode? Can it affect atmospheric stability or circulation, thereby having potential feedbacks on pollutant transport or dilutions? Furthermore, what is expected in the future? Are similar or even stronger fire episodes likely as a result of future anthropogenic activities in the regions, or as a result of changing climatic conditions?

Response:

Thank you very much for your great comment.

Recently, there is an increasing awareness that BC has measurable effects on atmospheric and land surface temperatures, primarily through radiative scattering and absorption in the atmosphere and systematic albedo changes at surfaces. These two effects make BC a strong driver of climate change from local to global scales [51].

For surface albedo, the events in 2019 provide a solid negative forcing at the surface (-38 W/m²) by absorbing and blocking direct solar radiation, which directly leads to the darkening of the surface. This effect may, in turn, reduce the evaporation and rainfall over the region [52], and a drier environment is more likely to lead to frequent fires and trapped in a cycle of deterioration. Atmospheric heating (positive radiative forcing, +26 W/m²) can directly exacerbate the warming process and curb the thermoregulatory capacity of the rainforest itself [53,54], especially for such vital pollution events. In addition, the heating of the atmosphere can increase the thermodynamic stability of atmosphere and the decrease of precipitation, which not conducive to the diffusion of pollutants [55]. It makes it easier for these strongly absorbing particles to be deposited on the land and ocean surfaces, forcing the surface to be in a longer-term influence.

Therefore, these radiation related effects of absorbing aerosol in the long-lasting, widespread wildfire events may have a chance to accelerate the deterioration cycle of drought and fire. Brazil was reported to have suffered the worst drought in the following two years of 2019[56,57]. The feedback effect of excessive anthropogenic burning of trees in 2019 on the extreme drought climate in the Amazon from 2020 to 2021 is worthy to further investigated in the future.

In the revised manuscript, we have included the above discussion in Section 4.3. The corresponding content can be found in lines 659-681.

Technical issues:

Point1: The authors should make sure that all the quantities in the equations are explained once the appear the first time. For example, now the wavelength is explained after equation 6 although it already appears in equation 1.

Response 1:

Thanks for the suggestion, we checked again the descriptions of all the quantities to make sure that they were explained when they first appeared.

Point 2: Table 1 claims that the meaning of the symbols in that table can be found in equations 1-4, but this seems not to be the case. Also, tables should be understandable by their own, so I would recommend explaining shortly these quantities in the table caption.

Response 2:

Thanks for the suggestion, we have added a description of symbols in Table 1. See lines 179-181 for the specific changes in the revision.

Point 3: Although the text is relatively well written, there are small grammatical issues here and there. I recommend checking out the language of the paper once more after revising it.

Response 3:

Thanks for the suggestion. In the revised manuscript, we rechecked the language and requested the service to make grammatical changes. We hope the revised manuscript will meet the needs for publication.

 

 

Author Response File: Author Response.docx

Reviewer 2 Report

This manuscript analyses an extreme biomass burning event occurred in the Amazonian from Jul to Sept 2019. To this end, the authors made use of data acquired by the Moderate-resolution Imaging Spectroradiometer (MODIS) to study the aerosol properties (e.g. single scattering albedo or absorbing optical depth) of the biomass burning event. The authors also give an estimation of the radiative forcing produced by the event. Although the results of this work may be of interest, the research shows important flaws that are listed below:

-1. Model retrieval products: A large part of the manuscript is about the model used to estimate the aerosol properties from MODIS data. However, almost not information is given about the model (only a reference is provided). As this model hasn’t been used widely in other works, I recommend the authors to add a section fully describing the model (inversion procedure, errors…). Note that without a full description, the reader can only trust the retrieval products given in the manuscript.

-2. Comparison with AERONET: one of my main concerns is the comparison with AERONET. Indeed, the comparison is important to show the capability of the algorithm to derive the biomass-burning-aerosol products. However, the authors do not sufficiently show the capability of the algorithm to achieve the research goal:

Selection of the data: the authors do not explain how the data in figure 3 was selected. Does it represent the data given in figure 2?. This is not clear in the manuscript. If that is the case, how is it possible to have a few points in the comparison of e.g. the single scattering albedo (right upper panel of figure 3). Also, why the number of points in the AOD is different to those for the single scattering albedo? Again, this is not explained in the manuscript. The authors also say nothing about if they used the AERONET daily average values or the instantaneous ones.

Errors: The errors in the estimation of the aerosol products are key to evaluate the potential of the model to perform this kid of analysis. However, an exhaustive error analysis of the aerosol products is not given in the manuscript. How can we know the quality of the aerosol products if the errors are not given? This is also very important to derive the conclusions about how the aerosol properties vary along the time.  

-3. Time variation of the aerosol properties: Figure 6 and table 3 provide the aerosol-properties time variation for a case of study. Again, the authors do not provide any information about what the errors are, and thus any conclusions about this time evolution of the aerosol properties cannot be derived. In figure 6, I can see temporal variations in the aerosol products but it is not clear to me if those are real variations or artefacts produced by the inversion? It needs further analysis.

-4. Radiative forcing: In the same way, without information about the errors in the aerosol products it is hard to believe the results given in section 3.5. The authors also do not provide enough information about the radiative transfer computations.

For the reasons listed above, I suggest the authors to re-do their analysis, extend the comparison with AERONET and perform a much deeper error analyses of the retrievals. As this new analysis may change the main conclusions of the research, I recommend to reject the manuscript in the present form.

Author Response

Response to Reviewer 1 Comments

This manuscript analyses an extreme biomass burning event occurred in the Amazonian from Jul to Sept 2019. To this end, the authors made use of data acquired by the Moderate-resolution Imaging Spectroradiometer (MODIS) to study the aerosol properties (e.g. single scattering albedo or absorbing optical depth) of the biomass burning event. The authors also give an estimation of the radiative forcing produced by the event. Although the results of this work may be of interest, the research shows important flaws that are listed below:

Point 1:Model retrieval products: A large part of the manuscript is about the model used to estimate the aerosol properties from MODIS data. However, almost not information is given about the model (only a reference is provided). As this model hasn’t been used widely in other works, I recommend the authors to add a section fully describing the model (inversion procedure, errors…). Note that without a full description, the reader can only trust the retrieval products given in the manuscript.

Response 1:

Thanks for the great suggestion. Since the related work has been described in detail in our previous study, we therefore keep the brief description in the main text considering the length and the focus of the article, but add a more detailed description, including the algorithm procedure and errors, in the Supplementary Materials. We hope the additional information will make the algorithm more convincing to the readers. The relevant contents can be seen in the Supplementary Materials of the revised manuscript (Line 42-110).

Point 2:. Comparison with AERONET: one of my main concerns is the comparison with AERONET. Indeed, the comparison is important to show the capability of the algorithm to derive the biomass-burning-aerosol products. However, the authors do not sufficiently show the capability of the algorithm to achieve the research goal:

Selection of the data: the authors do not explain how the data in figure 3 was selected. Does it represent the data given in figure 2?. This is not clear in the manuscript. If that is the case, how is it possible to have a few points in the comparison of e.g. the single scattering albedo (right upper panel of figure 3). Also, why the number of points in the AOD is different to those for the single scattering albedo? Again, this is not explained in the manuscript. The authors also say nothing about if they used the AERONET daily average values or the instantaneous ones.

Response 2:

We acknowledge that the table in Fig. 2 is misleading. The data in Fig. 2 represent the days that this pollution event can be observed by the ground based AERONET (using instantaneous measurements, even if only one record is counted). However, the points in Fig. 3 are the matched records that 30 min before and after the overpass time of MODIS, corresponding to the 5 × 5 pixels over the stations. The matchup criteria are at least 2/3 pixels of valid MODIS retrievals within a window surrounding the AERONET site. Therefore, the data in Fig. 3 are necessarily less than records in Fig. 2. In the revised manuscript we removed the table for Fig. 2 and added the matching procedure between satellite and AERONET in section 3.1. See line 358-361 for details.

For the doubt about the different quantities of AOD and other products, the main reason is that AERONET provides AOD and SSA from two different products (AOD from Aerosol Optical Depth Datasets and SSA from Aerosol Inversion Datasets), and the records of AOD are much more than those of SSA. This is attributed to the limitations of the AERONET inversion algorithm. Only those results that can converge in the model will be published while those bad inversions are excluded in the datasets. Furthermore, such inversion data reaches Level 2.0 only when the level of 440 nm AOD is greater than 0.4 and other almucantar and hybrid quality check is satisfied. Thus, it is inevitable that the number of points in the AOD is different to those for the SSA. The relevant content has been mentioned in the revised manuscript Section 3.1 (line 363-375).

Errors: The errors in the estimation of the aerosol products are key to evaluate the potential of the model to perform this kid of analysis. However, an exhaustive error analysis of the aerosol products is not given in the manuscript. How can we know the quality of the aerosol products if the errors are not given? This is also very important to derive the conclusions about how the aerosol properties vary along the time. 

Response:

Factors contributing to  (volume fraction of BC, which is the primary retrieval in our algorithm) errors have been adequately tested in our previous studies. Based on the procedure of our inversion algorithm, the model errors are mainly related by the input AOD, surface reflectance assumptions, and background aerosol properties. Table A1 provide a simple summary of the expected uncertainty of the algorithm.

The error of the MODIS AOD at 550 nm over land is expected to be 0.05+0.15×AOD. Significant impacts contributing to the retrieval bias are found under different surface backgrounds and varying AOD levels in the simulations. The retrievals under higher aerosol loadings are less affected by the AOD than those in clear-sky conditions. The uncertainty ranges from -54% to 72% in extremely clear-sky conditions (AOD=0.1) but significantly decreases to ±15% under higher aerosol loadings (AOD≥0.5). Additionally, the uncertainties over dark targets (vegetation) are much lower than those over bright targets (sand), likely due to the stronger contribution of the AOD to the TOA reflectance over bright surfaces.

To represent most land covers, the surface reflectance error for the MODIS red channel (645 nm) is expected to be ±0.01, and the bias of the MODIS red (645 nm) vs blue (469 nm) surface relationship is expected to be ±0.2. These uncertainties demonstrate significant biases in retrievals that are excessively high when the AOD is lower than 1.0. Fortunately, overestimating or underestimating the surface reflectance at the red channel will produce the same surface reflectance change at the blue channel, which influences final retrievals to the opposite extent. This finding indicates a possible bias lower than 40% for AOD≥0.5 and lower than 20% for AOD≥1.0.

The variance of the clustering models for the background aerosols also contributes to the uncertainties in the retrieval because the microphysical parameters averaged from these clusters cannot reflect realistic conditions especially on a daily scale. These AOD-independent uncertainties ranging from -24% to 9% (-15% - 9% when AOD ≥3.0).

From these analyses we can see that although the error may be higher when the AOD is low (which is unavoidable because the satellite can only receive very low atmospheric signals at low AOD), it has a very good performance for the pollution case. For Amazonia in this paper, the inversion uncertainties are lower because the AOD performance is better (see Fig.3a), the uncertainties of the surface reflectance assumption is lower (because our surface model is more applicable to vegetation and there is a large amount of vegetation cover in the Amazonia), the background aerosol is less complex (the background aerosol changes are not dramatic in the same season, and the clustering results are closer to the real situation).

In the revised manuscript, relevant content has been added to the Supplementary (line 42-110), and we hope it is clear to show the product quality in our study.

Table A1. Uncertainties of the algorithm under different AODs ( )
Factors	=0.1	=0.5	=1.0	=3.0
AOD error (Expected to be 0.05+0.15×AOD)	-54%~72%	-15%~15%	-6%~6%	-2%~3%
surface reflectance error (Expected to be ±0.01 for MODIS Band1/3)	79%~110%	25%~37%	7%~20%	1%~3%
Background Error (Expected to be the variance of the clustering models)	-24%~8%	-21%~7%	-19%~7%	-15%~9%

Point 3:. Time variation of the aerosol properties: Figure 6 and table 3 provide the aerosol-properties time variation for a case of study. Again, the authors do not provide any information about what the errors are, and thus any conclusions about this time evolution of the aerosol properties cannot be derived. In figure 6, I can see temporal variations in the aerosol products, but it is not clear to me if those are real variations or artefacts produced by the inversion? It needs further analysis.

Response 3:

Figure 6 shows the trends of aerosol properties in the three sub-study regions, but all three sub-regions lack real value references, like AERONET. The above error sources analyses shows that the satellite algorithm we used has a very reliable inversion accuracy in the case of high pollution scenarios. Given the presence of large pollution processes in all three regions, we believe that the results in Fig. 6 can demonstrate the regional pollution characteristics. As for the results corresponding to low AOD before the event, which may be suffered higher uncertainties in retrieval, they are not significant in absolute order of magnitude.

Point 4:. Radiative forcing: In the same way, without information about the errors in the aerosol products it is hard to believe the results given in section 3.5. The authors also do not provide enough information about the radiative transfer computations.

Response 4:

As we mentioned above, the satellite inversion results are reliable for the high pollution scenarios. Although the uncertainty is relatively high at lower AOD, this does not affect the later conclusions on radiative forcing, since the effect of very low aerosol concentrations on the radiation is almost negligible, regardless of the scattering and absorption strength of aerosol particles themselves. This point is also mentioned in our revised manuscript (Section 2.3, line 342-346).

As for the calculation of the radiative forcing, we have further refined it in the revised manuscript (Section 2.3).

For the reasons listed above, I suggest the authors to re-do their analysis, extend the comparison with AERONET and perform a much deeper error analyses of the retrievals. As this new analysis may change the main conclusions of the research, I recommend to reject the manuscript in the present form.

Author Response File: Author Response.docx

Reviewer 3 Report

Overall, the paper is well written and well organized. I recommend it to be published with revisions to address below comments and suggestions.

1. Line 129, please add a reference here.
2. What is RII in Figure 1? typo? In addition, it seems the authors only described how to calculate the mixed optical properties of the RT model input, but didn't provide how to retrieval the AADO and fbc.  I know  some of the retrieval work have been published before, however, it is better to make a description about the retrieval algorithm after 2.1.2 (3). 
3. How did the authors do this re-cluster work? From my point of view, your background aerosol still has a great portion of BC. 
4. Where did the authors get the coefficients of eq(4), what are n and k in eq(4) and table 1?
5. What is meaning of the number in figure 2? day? or data simple? 
6. A comparison between your updated algorithm and the original one should be given, to see the improvement and your contributions.

Author Response

Response to Reviewer 3 Comments

Specific comments:

Overall, the paper is well written and well organized. I recommend it to be published with revisions to address below comments and suggestions.

Point1. Line 129, please add a reference here.

Response 1:

Thanks. We have added the reference [32] on line 154 in the revised manuscript.

Point2. What is RII in Figure 1? typo? In addition, it seems the authors only described how to calculate the mixed optical properties of the RT model input, but didn't provide how to retrieval the AADO and fbc.  I know  some of the retrieval work have been published before, however, it is better to make a description about the retrieval algorithm after 2.1.2 (3). 

Response 2:

Thanks for the suggestion. RII in Figure 1 represents the imaginary part of refractive index, which is highly corresponds to the volume fractions of BC. The related explanations have been added in the revised description of Fig. 1.

The detailed description of the algorithm and relevant contents about the existence of error sources can be seen in the Supplementary of the revised manuscript (Line 42-110).

Point3. How did the authors do this re-cluster work? From my point of view, your background aerosol still has a great portion of BC. 

Response 3:

In the proposed algorithm, we assumed that aerosols particles consist of strongly absorbing BC and scattering background aerosols. Therefore, it is important to remove the effect of BC in the clustering of background aerosol types. In our strategy, we perform a two-step operation. In the first step, since BC is a fine particle with strong absorption, we first remove all the strong absorption fine particles that may be affected by BC from the AERONET dataset to be clustered. The threshold value set here is SSA675<0.9 and FMF>0.4. In the second step, the processed dataset is clustered by k-means model. We set 4 clusters here and the similar types should be combined. Finally, the most strongly scattering type is selected as the background aerosol.  In this way we believe we can remove the effect of BC from the clusters. In fact, unlike China (we tested in previous study), the background aerosol type in the Amazonia is very homogeneous, so only one background aerosol can be obtained.

In the revised manuscript, we have added a more detailed description of the re-cluster work, which can be found in Section 2.1.3(line 187-198).

Point4. Where did the authors get the coefficients of eq(4), what are n and k in eq(4) and table 1?

Response 4:

The coefficients of Eq (4) are proposed by the study of Chang, H. (Determination of the wavelength dependence of refractive indices of flame soot). n and k are the real part and imaginary part of BC refractive index, respectively. The explanation of the relevant parameters has been added to the revised version (line 171-172).

Point5. What is meaning of the number in figure 2? day? or data simple? 

Response 5:

The numbers in Figure 2 represent the days that AERONET can observe. However, since this may cause unnecessary misunderstanding for the later validation, the relevant content was removed in the revised manuscript.

Point6. A comparison between your updated algorithm and the original one should be given, to see the improvement and your contributions.

Response 6:

In this paper we do not improve more on the algorithm strategy itself, but only apply the previous algorithm from China to the Amazonia. Therefore, the adaptation from the algorithm itself was only done for the Amazonia (background aerosol type) to prove the ability and application of the algorithm for monitoring strong biomass burning activities in other regions of the world. In addition, two additional output products, AAOD and SSA, were added to this study, which were not calculated in the previous paper. These two new parameters allow us to assess the radiative forcing of pollution events and discuss the potential drivers of pollution events on climate more easily. The relevant statements can be seen in the conclusion of the revised manuscript.

Author Response File: Author Response.docx

Reviewer 4 Report

Summary

The manuscript titled "Severe Biomass Burning Aerosol Pollution during the 2019 Amazon Wildfire and its Direct Radiative Forcing Impact: A Space Perspective from MODIS Retrievals" discusses the Biomass burning event that happened in Amazon in 2019. The authors used MODIS data to retrieve aerosol properties and then calculated the aerosol radiative forcing (ARF) for the year 2019. Comparison with the ARF of 2018 shows an elevated level of aerosol in 2019 compared to 2018. 

 

The manuscript is well written and easy to read and follow. However, I have some comments, concerns, and suggestions. Please see the major and specific comments sections below,

 

Major Comments

1. Check the equation numbering throughout the manuscript. The number in the text doesn't matches appropriately.
2. What are the uncertainties in the retrieved SSA and f(BC) using the technique mentioned in this study?. E.g.: What is the propagated uncertainty in f(BC) if 10% uncertainty in SSA derived? You should discuss the uncertainty in the retrieved parameter. 
3. You should add a description of how ARF is calculated. How do you get a range of values, as seen in Fig 9?. Is it calculated for each pixel?. or for a region?. The distribution is based on the ARF of all pixels?. Please elaborate on what the methodology is.

Specific comments

L#109: What kind of mixing model are we talking about? Can you provide some details?

L#142-145: Does the background aerosol include BC?. Please discuss it

L#148: It is Eq. 4, not 5

L#150: It will be better to mention that wavelength should be in um.

L#151-153: The r_{BA} reported by AERONET is the median radius, and the definition of r in Eq.2 is the mean radius. Did you account for this difference?. If not, you have to use a consistent definition. If yes, please make sure that you discuss it here

 

L#156: Check the spelling of MODIS

L#187: Explain why lambda ranges from 1 to 4 in the first component and 1 to 3 in the second component of the equations

L#191: It should be Eq. 1

L#193: Why 'i' range from 1 to 3, not 4?

L#257-259: There are clouds over the pristine forest, and you should cross-check this statement. There is no fire anomaly detection when the cloud is present, so this statement may not be right.

L#262: Make the font in each plot bigger. For e.g., latitude longitude values are very tiny and not visible.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The authors have reasonably addressed all my concerns.