From the journal Environmental Science: Atmospheres Peer review history

03-Jan-2023

Dear Mr Anand:

Manuscript ID: EA-ART-11-2022-000166
TITLE: Estimation of hourly black carbon aerosol concentrations from glass fiber filter tapes using image reflectance-based method

Thank you for your submission to Environmental Science: Atmospheres, published by the Royal Society of Chemistry. I sent your manuscript to reviewers and I have now received their reports which are copied below.

I have carefully evaluated your manuscript and the reviewers’ reports, and the reports indicate that major revisions are necessary.

Please submit a revised manuscript which addresses all of the reviewers’ comments. Further peer review of your revised manuscript may be needed. When you submit your revised manuscript please include a point by point response to the reviewers’ comments and highlight the changes you have made. Full details of the files you need to submit are listed at the end of this email.

Please submit your revised manuscript as soon as possible using this link:

*** PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm. ***

https://mc.manuscriptcentral.com/esatmos?link_removed

(This link goes straight to your account, without the need to log on to the system. For your account security you should not share this link with others.)

Alternatively, you can login to your account (https://mc.manuscriptcentral.com/esatmos) where you will need your case-sensitive USER ID and password.

You should submit your revised manuscript as soon as possible; please note you will receive a series of automatic reminders. If your revisions will take a significant length of time, please contact me. If I do not hear from you, I may withdraw your manuscript from consideration and you will have to resubmit. Any resubmission will receive a new submission date.

The Royal Society of Chemistry requires all submitting authors to provide their ORCID iD when they submit a revised manuscript. This is quick and easy to do as part of the revised manuscript submission process. We will publish this information with the article, and you may choose to have your ORCID record updated automatically with details of the publication.

Please also encourage your co-authors to sign up for their own ORCID account and associate it with their account on our manuscript submission system. For further information see: https://www.rsc.org/journals-books-databases/journal-authors-reviewers/processes-policies/#attribution-id

Environmental Science: Atmospheres strongly encourages authors of research articles to include an ‘Author contributions’ section in their manuscript, for publication in the final article. This should appear immediately above the ‘Conflict of interest’ and ‘Acknowledgement’ sections. I strongly recommend you use CRediT (the Contributor Roles Taxonomy, https://credit.niso.org/) for standardised contribution descriptions. All authors should have agreed to their individual contributions ahead of submission and these should accurately reflect contributions to the work. Please refer to our general author guidelines https://www.rsc.org/journals-books-databases/author-and-reviewer-hub/authors-information/responsibilities/ for more information.

I look forward to receiving your revised manuscript.

Yours sincerely,
Dr Nønne Prisle
Associate Editor, Environmental Sciences: Atmospheres

************

Reviewer 1

General comment

The paper reports a study done to investigate the possibility to extract information on black carbon concentrations form image processing of photos of deposit on filters used in beta attenuation monitors. The topic could be interest even if there are some parts not clear and likely not all aspects of the limitations of the method have been explained (see my specific comments). I suggest to consider the paper for publication after a major revision addressing my comments.


Specific comments

There is some work done to evaluate the limit of detection of the method. However, it would be useful to include also the info regarding the uncertainty on 1 hour measurement. For example, Figure 4 shows that the most probably values of BC concentrations are around 0.4-0.6 ug/m3. What is the uncertainty of your method in this range?

Another aspect that is not discussed or evaluated is the amount of material deposited. I suppose that is the sampling on a specific spot is long particles will accumulate in a thick layer and the photo will just look at the upper part of the deposit (i.e. the last sampled aerosol). I suppose that you method not only have a minimum LOD but also a maximum concentration or at least a maximum mass of sampled material per cm2. This also need to be discussed.

Introduction it is discussed the difference among BC and EC and I suggest to make a mention to the work of Contini et al. (2018, Atmosphere 9 (5), 181) to better put in evidence this aspect.

Line 131. The first time it is mentioned R to BC model is not clear. Please specify here that it is the integer scale of the red colour indicated as R.

Lines 224-225. Why it should be expected an exponential relationship? I do not see any reason explained in the paper. In addition, looking at Table 1, the hybrid and the polynomial regression work substantially with the same performances. So please discuss better this aspect and why choosing an exponential form.

Line 258. Actually 1 ug/m3 mentioned here is much higher than the others so the word similar is inappropriate.

Figure 5 and related discussion. It seems that the performance of this approach is very different at the two sites especially for the highest values that are measured in EC but not in BC for Liberty site. I suggest to interpret better this differences and to show also the scatter plot of EC and BC data at the two sites to understand the scattering of the results.

Lines 349-354. This sentence is an exaggeration and should be modified. Results clearly show that the method cannot separate wood burning from other sources. This is clearly shown in your “doped” measurements. The data of 2004 are out of trend but it could be for artefacts considering that you analyse them 14 years after collection. It is really necessary to include them? If yes you should discuss the possible changes occurring after such a long time and if they can bring to artefacts in the measurements. I would be surprised if this simple method could discern wood burning from other source.

Line 262 and also in other instances in the paper. No PM composition because this approach does not give any info on composition, you should mention only BC content.

Reviewer 2

Review of the paper titled: “Estimation of hourly black carbon aerosol concentrations from glass fiber filter tapes using image reflectance-based method” by Anand et al.

The paper discussed a low-cost method to estimate black carbon concentrations from reflectance measurements using cellphone cameras and a simple setup. The paper is very well written and clear, and the technique seems to be promising and it might to measure B concentrations in locations where research grade or monitoring commercial instrumentation is not available. Overall, I think the work is sound and it deserves publication. I have only a few comments that I would like the authors to address before acceptance.

Comments
- Why use cellphone cameras? There are several other more cost-effective available on the market. It is true that cell phones are widespread, but I wonder if it could be worth also checking the possibility to use even lower-tech cameras (this could be a task for future developments, not necessarily for this paper, but maybe the authors could comment on this issue)
- The authors use the aethalometer as a reference. While that’s reasonable, it is well known that filter-based attenuation techniques such as that used in the paper (AE31) can be prone to significant biases especially in the presence of BC internally mixed with other materials. I am not asking to add any additional experiments, but it would indeed be nice to have some other instrument such as an SP2 or a photoacoustic for some of the tests at least. As I mentioned, I am not asking the authors to perform additional measurements, but the authors could at least acknowledge the limitations of filer-based attenuation methods as well established in the literature (see for example the work by Subramanian or by Lack)
- The method measured reflected light using reference printed colors. The method is ingenious, but it would have been interesting to see some clear evidence of how the lighting, the source, as well as the surroundings, might (or hopefully not) affect the results. It is reasonable to think that if the method is adopted by other researchers, the lighting might not be the same. Related to this topic, it would have been nice to have some more information about the illumination system and setup.
- Are the reference card colors stable over time? It might be good to discuss this potential issue. Also, it would be very useful to provide detailed information (in the SI would be sufficient) on how to generate these cards (maybe provide a file) and how to print them (provide the exact paper, printer, and ink types and manufacturers), so others can use an identical protocol to enhance consistency.
- It would be good to see some more explanation on the selection of the various test models, why those models exactly, is it just a random search or is there some logic?
- Section 3.2 assumes that BC = EC, but the two quantities are operationally different (much work on this topic is available in the literature). Therefore, the last sentence on page 19 (lines 298, and 299) might be somewhat misleading in this context. For example, figure S2 show a good correlation between EC and BC up to 2.5 um/cm2, but then a bias seems to emerge for higher concentrations with the EC overestimating the BCAeth.

Response to reviewer 1:

Response to specific comments
• There is some work done to evaluate the limit of detection of the method. However, it would be useful to include also the info regarding the uncertainty on 1 hour measurement. For example, Figure 4 shows that the most probably values of BC concentrations are around 0.4-0.6 ug/m3. What is the uncertainty of your method in this range?

Response: Thank you for the comment. I have added the following explanation in the main text.
Lines 303-309: The median for the BC histogram is 0.67 µg m-3. The measured BC at this site is above the EDL for > 98% of all hours. BC concentrations in many developing nations are higher (Figure S8), indicating that this approach will be able to determine hourly BC concentrations from BAM tapes in many locations worldwide.
We use the RMSE of BC to R model for low BC levels (below 1.66 µg/m3, or R > 224) to estimate the uncertainty of the method to be around 0.1 µg/m3 at the typical ambient BC concentrations shown in Figure 4.
Additionally, I have added the following information to provide evidence-based explanation on the robustness of the method.
Lines 272-281: A cell phone is used in our experiments due to its easy availability. A lower resolution camera such as a webcam is equally suited for the task as this method requires an average of only a few pixels of evenly deposited particles on a filter to quantify BC concentrations.
We investigated the effects of lighting conditions on estimation of BC concentration with the image reflectance method for a filter sample (Section S1). BC concentrations for two filters, with BC loadings of 1.725 µg cm-2 and 8.089 µg cm-2, were calculated with the BC to R model at five light intensities from very dim (level 6) to very bright (level 10) lighting during image capture (Table S1). We observed a maximum increase of only 1.2% in R and 3.8% decline in BC compared to those in the reference light settings. Thus, the same model can work for quantifying BC concentrations for an unknown sample in a wide range of lighting conditions.


(SI: Lines 33-55)
S2. Effect of lighting conditions on estimation of BC concentrations
A uniform lighting is critical for image capturing to ensure same effect of color correction is applied for the entire reference card. Point sources such as LEDs might introduced localized errors during color correction. Therefore, we used two 6000 K ring lights to establish a uniform diffused lighting with capabilities with 10 different light intensities.

Figure S5: Image capturing setup that includes a pair of 6000K ring lights placed equidistant (D = 8 inches) from the reference cards to ensure a uniform diffused lighting environment. The lights were set at second maximum intensity (level 9). Ring lights were the only source of light in the room to avoid optical interference.

We selected two filter samples in different BC concentration ranges and varied the intensities of the lights simultaneously to study the effects of lighting variations on the performance of image reflectance method. The minimum light intensity was limited at level 6 as the lower levels were too diminished compared to usual working environments. Level 9 was used as a standard to compare the red scale change in filter samples. We observed a percentage red scale change of only 0.4% and 3.7% change in BC estimation for lower concentration sample (1.725 µg cm-2), whereas only 1% change in R and a maximum change in BC of 3.8% for the higher BC sample (8.089 µg cm-2).

Table S1: Effect of varying light intensities on RGB color channel and BC estimations of filter samples. The light intensities varied from level 6 to 10 for two samples with BC concentrations (CA) of 1.725 µg cm-2 (Sample ID: 25) and 8.089 µg cm-2 (Sample ID: 21). The changes in R and BC are calculated with those in level 9 as a standard.

Sample ID Light Intensity R G B %ΔR/R %ΔBC/BC
25 6 224 220 203 0.4 -3.7
7 224 220 203 0.4 -3.7
8 224 220 204 0.4 -3.7
9 (reference) 223 220 205 - -
10 223 220 206 0.0 0.0
21 6 172 167 148 1.2 -3.8
7 171 167 149 0.6 -1.9
8 171 167 149 0.6 -1.9
9 (reference) 170 166 149 - -
10 171 167 152 0.6 -1.9

--------
• Another aspect that is not discussed or evaluated is the amount of material deposited. I suppose that is the sampling on a specific spot is long particles will accumulate in a thick layer and the photo will just look at the upper part of the deposit (i.e. the last sampled aerosol). I suppose that you method not only have a minimum LOD but also a maximum concentration or at least a maximum mass of sampled material per cm2. This also need to be discussed.

Response: Thank you for the comment. I have included the following text to the manuscript in lines 293-300.

Our method relies on red light reflected from the sample spot. It is possible that under very high filter loadings, BC could accumulate into such a dark spot that R = 0 and additional material does not change the absorbance. Such a situation is unlikely under ambient conditions but could be encountered if this method is used to quantify BC for direct emissions testing from combustion sources. We illustrate our performance up to an hourly BC concentration of 15 µg m-3 (and a filter loading of 15.8 µg cm-2 for BAM 1020). This is an extremely high hourly BC concentration for an ambient environment, even for highly polluted ambient environments. Therefore, we do not expect to encounter a ‘maximum’ detection limit for typical ambient conditions.

--------
• Introduction it is discussed the difference among BC and EC and I suggest to make a mention to the work of Contini et al. (2018, Atmosphere 9 (5), 181) to better put in evidence this aspect.

Response: Thank you for the comment. I have included a revision in the manuscript (lines 55-59).

EC (elemental carbon) is another measure of carbon soot in the air and is quantified operationally as carbonaceous aerosols measured via thermal-optical methods. Contini et al. mentions anthropogenic and natural combustion sources to be the main contributors to BC or EC in the atmosphere.23 The correlation between BC and EC as well as the BC-to-EC mass ratio can vary with sampling environments due to differences in sources.24

--------
• Line 131. The first time it is mentioned R to BC model is not clear. Please specify here that it is the integer scale of the red colour indicated as R.

Response: Thank you for the comment. The revision is included in lines 134-137.

The 51 glass-fiber filter samples were used for training the R to BC model, where R is red scale values for a filter sample that takes integer values between 0 to 255, and BC represents area loading of black carbon (µg cm-2) in the filter sample.

--------
• Lines 224-225. Why it should be expected an exponential relationship? I do not see any reason explained in the paper. In addition, looking at Table 1, the hybrid and the polynomial regression work substantially with the same performances. So please discuss better this aspect and why choosing an exponential form.

Response: Thanks for the comment. I have included following lines for a better understanding in the model selection.

Lines 247-259: Linear, polynomial, and logarithmic models are commonly used regression models. Additionally, gradient boosting, random forest, ridge, and support vector machine are machine learning models widely used in air quality research recently. Ensemble methods try to improve accuracy by combining predictive performance of different machine learning models. We also explored a hybrid model to exploit the fairly linear correlation of BC with R on lower BC concentrations. This hybrid model consisted of a linear part for low BC values and an exponential curve to explain dependence of high BC range on R.

We used R2, mean absolute error (MAE) and root mean square error (RMSE) to assess model performance. Table 1 summarizes the performance of all trained models, and the training data are shown in Figure 3 and Table S2. All models in Table 1 were evaluated with random 4-fold cross-validation. While many of the models showed similar performance, the exponential model seems to be the simplest model with the best performance. Thus, we used this model; the model fit and parameters are shown in Figure 3.

--------
• Line 258. Actually 1 ug/m3 mentioned here is much higher than the others so the word similar is inappropriate.

Response: The line is revised as:

Lines 291-292: The image-based approach can therefore detect BC concentrations of 0.07 µg m-3 for the Met One BAM and 0.15 µg m-3 for the Thermo Fisher BAM at an hourly resolution.

--------
• Figure 5 and related discussion. It seems that the performance of this approach is very different at the two sites especially for the highest values that are measured in EC but not in BC for Liberty site. I suggest to interpret better this differences and to show also the scatter plot of EC and BC data at the two sites to understand the scattering of the results.

Response: I included the following explanation in the main text (line 343-353) and Figure S7 in the supplementary information.
Figure S7 shows that while BC and EC are correlated at Liberty, the image reflectance-based BC is consistently lower when EC is above 0.5 µg m-3. This disparity in EC and BC-OPT is reflected in our quartz training filters (Figure S2). In those filters, EC was systematically higher than BC-OPT, with larger differences at high concentration.

While some of the differences between BC-OPT and EC at Liberty might be due to methodological differences, the source of BC may also play a role. High BC (EC) concentrations both at Liberty and at the CMU site are often associated with industrial emissions from a metallurgical coke works located 3 km south from the Liberty.52 If these industrial emissions have a different EC-to-BC ratio than typical traffic-dominated urban emissions, that could explain the poorer agreement at Liberty than at Lawrenceville or CMU.

--------
• Lines 349-354. This sentence is an exaggeration and should be modified. Results clearly show that the method cannot separate wood burning from other sources. This is clearly shown in your “doped” measurements. The data of 2004 are out of trend but it could be for artefacts considering that you analyse them 14 years after collection. It is really necessary to include them? If yes you should discuss the possible changes occurring after such a long time and if they can bring to artefacts in the measurements. I would be surprised if this simple method could discern wood burning from other source.

Response: Thank you for the comments. I have revised the text in lines 401-414.

These samples have excess absorption of both Blue and Green light. This excess absorption may be due to the presence of light-absorbing brown carbon formed during smoldering combustion.57 The wood smoke samples from 2022 were dominated by flaming combustion and therefore had minimal emissions of brown carbon. The excess absorption of Blue and Green light is not observed in the freshly collected ‘Wood smoke 2022’ samples nor in the ambient samples doped with wood smoke. The excess Blue and Green absorption for the 2004 samples might also be due to artefacts introduced to these samples due to aging of particle deposits during storage.

Our results suggest that excess absorption in the Blue and Green channels might be useful as a qualitative indicator for the presence of BC from biomass burning. However, since flaming combustion produces minimal brown carbon, samples dominated by flaming biomass combustion will not have excess absorption in the Blue and Green channels. Better separation of wood smoke BC from other sources, such as diesel, will require further investigation under a variety of biomass burning conditions.

--------
• Line 262 and also in other instances in the paper. No PM composition because this approach does not give any info on composition, you should mention only BC content.

Response: Thank you for the comment. I have revised the term as following in line 422.

Post analysis of the BAM filter tapes using cell phone camera images can provide valuable information on BC concentrations with hourly time resolution.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 
Response to reviewer 2:

Response to specific comments
• Why use cellphone cameras? There are several other more cost-effective available on the market. It is true that cell phones are widespread, but I wonder if it could be worth also checking the possibility to use even lower-tech cameras (this could be a task for future developments, not necessarily for this paper, but maybe the authors could comment on this issue)

Response: I have included the following explanation in the main text (lines 272-274).

A cell phone is used in our experiments due to its easy availability. A lower resolution camera such as a webcam is equally suited for the task as this method requires an average of only a few pixels of evenly deposited particles on a filter to quantify BC concentrations.

--------
• The authors use the aethalometer as a reference. While that’s reasonable, it is well known that filter-based attenuation techniques such as that used in the paper (AE31) can be prone to significant biases especially in the presence of BC internally mixed with other materials. I am not asking to add any additional experiments, but it would indeed be nice to have some other instrument such as an SP2 or a photoacoustic for some of the tests at least. As I mentioned, I am not asking the authors to perform additional measurements, but the authors could at least acknowledge the limitations of filer-based attenuation methods as well established in the literature (see for example the work by Subramanian or by Lack)

Response: Thank you for the comment. I have added text acknowledging limitations of filter-based optical techniques and cited the work from Subramanian (lines 138-153).
Filter-based light attenuation techniques face challenges due to continuous particle loading on the same spot and multiple scattering of light rays. The AE-31 tends to underestimate BC concentrations as the filter tape gradually becomes loaded with particles, a phenomenon referred to as “shadowing effect”, that is prominently observed in experiments with high concentrations of freshly emitted soot.39 We applied appropriate loading corrections to the raw BC concentrations from AE-31 and the corrected BC is referred to as BCAeth in this article.40
Subramanian et al. reported that filter-based optical BC measurements can experience errors due to aerosol emissions from smoldering biomass burning or other sources of liquid organic matter.41 Therefore, we compared BCAeth with EC for 7 quartz fiber filter samples to evaluate the performance of the Aethalometer (Figure S2). EC was quantified using the Interagency Monitoring of PROtected Visual Environments (IMPROVE)-A protocol. Figure S2 shows a high correlation between BC and EC, as expected, but that EC concentrations were 13% higher than BC.
BCAeth data was used as validation for all 51 training filters to ensure readily available data at high time resolution and to avoid always relying on filters for reference measurement. The aethalometer is used as ground truth for our models due to its wide applicability in continuous monitoring of BC in the ambient environments.

--------
• The method measured reflected light using reference printed colors. The method is ingenious, but it would have been interesting to see some clear evidence of how the lighting, the source, as well as the surroundings, might (or hopefully not) affect the results. It is reasonable to think that if the method is adopted by other researchers, the lighting might not be the same. Related to this topic, it would have been nice to have some more information about the illumination system and setup.

Response: I added lines 275-281 and introduced section S2 in the SI to address the comment which also contains a revised version of previously included Figure S5.

Lines 275-281: We investigated the effects of lighting conditions on estimation of BC concentration with the image reflectance method for a filter sample (Section S1). BC concentrations for two filters, with BC loadings of 1.725 µg cm-2 and 8.089 µg cm-2, were calculated with the BC to R model at five light intensities from very dim (level 6) to very bright (level 10) lighting during image capture (Table S1). We observed a maximum increase of only 1.2% in R and 3.8% decline in BC compared to those in the reference light settings. Thus, the same model can work for quantifying BC concentrations for an unknown sample in a wide range of lighting conditions.

Section S2 in the SI (Lines 33-55)

S2. Effect of lighting conditions on estimation of BC concentrations
A uniform lighting is critical for image capturing to ensure same effect of color correction is applied for the entire reference card. Point sources such as LEDs might introduced localized errors during color correction. Therefore, we used two 6000 K ring lights to establish a uniform diffused lighting with capabilities with 10 different light intensities.

Figure S5: Image capturing setup that includes a pair of 6000K ring lights placed equidistant (D = 8 inches) from the reference cards to ensure a uniform diffused lighting environment. The lights were set at second maximum intensity (level 9). Ring lights were the only source of light in the room to avoid optical interference.

We selected two filter samples in different BC concentration ranges and varied the intensities of the lights simultaneously to study the effects of lighting variations on the performance of image reflectance method. The minimum light intensity was limited at level 6 as the lower levels were too diminished compared to usual working environments. Level 9 was used as a standard to compare the red scale change in filter samples. We observed a percentage red scale change of only 0.4% and 3.7% change in BC estimation for lower concentration sample (1.725 µg/cm2), whereas only 1% change in R and a maximum change in BC of 3.8% for the higher BC sample (8.089 µg/cm2).

Table S1: Effect of varying light intensities on RGB color channel and BC estimations of filter samples. The light intensities varied from level 6 to 10 for two samples with BC concentrations (CA) of 1.725 µg/cm2 (Sample ID: 25) and 8.089 µg/cm2 (Sample ID: 21). The changes in R and BC are calculated with those in level 9 as a standard.

Sample ID Light Intensity R G B %ΔR/R %ΔBC/BC
25 6 224 220 203 0.4 -3.7
7 224 220 203 0.4 -3.7
8 224 220 204 0.4 -3.7
9 (reference) 223 220 205 - -
10 223 220 206 0.0 0.0
21 6 172 167 148 1.2 -3.8
7 171 167 149 0.6 -1.9
8 171 167 149 0.6 -1.9
9 (reference) 170 166 149 - -
10 171 167 152 0.6 -1.9

--------
• Are the reference card colors stable over time? It might be good to discuss this potential issue. Also, it would be very useful to provide detailed information (in the SI would be sufficient) on how to generate these cards (maybe provide a file) and how to print them (provide the exact paper, printer, and ink types and manufacturers), so others can use an identical protocol to enhance consistency.

Response: I have included the reference card image as Figure S4 of SI and have provided a separate JPG file for it. Additional details are in Section S1 of SI (lines 24-31, SI).

S1. Reference card details

Figure S4: Reference card template (JPG file available)
The reference card is designed in Adobe Illustrator with RGB inputs for each of the squares on the template. The card is letter sized (8.5 x 11 inches) and is printed on a matte finish photo paper to avoid light reflections from the paper during photo capture. It is cost effective to use Photo Printing Services in FedEx or similar outlets, which also provides high quality prints. Any other color printer that can print a uniformly distributed color for the squares would work too.

--------
• It would be good to see some more explanation on the selection of the various test models, why those models exactly, is it just a random search or is there some logic?

Response: Thank you for the comment. I have included the following explanation in lines 247-253.

Linear, polynomial, and logarithmic models are commonly used regression models. Additionally, gradient boosting, random forest, ridge, and support vector machine are machine learning models widely used in air quality research recently. Ensemble methods try to improve accuracy by combining predictive performance of different machine learning models. We also explored a hybrid model to exploit the fairly linear correlation of BC with R on lower BC concentrations. This hybrid model consisted of a linear part for low BC values and an exponential curve to explain dependence of high BC range on R.

--------
• Section 3.2 assumes that BC = EC, but the two quantities are operationally different (much work on this topic is available in the literature). Therefore, the last sentence on page 19 (lines 298, and 299) might be somewhat misleading in this context. For example, figure S2 show a good correlation between EC and BC up to 2.5 um/cm2, but then a bias seems to emerge for higher concentrations with the EC overestimating the BCAeth.

Response: I included the following explanation in the main text (line 343-353) and Figure S7 in the supplementary information.
Figure S7 shows that while BC and EC are correlated at Liberty, the image reflectance-based BC is consistently lower when EC is above 0.5 µg m-3. This disparity in EC and BC-OPT is reflected in our quartz training filters (Figure S2). In those filters, EC was systematically higher than BC-OPT, with larger differences at high concentrations.

While some of the differences between BC-OPT and EC at Liberty might be due to methodological differences, the source of BC may also play a role. High BC (EC) concentrations both at Liberty and at the CMU site are often associated with industrial emissions from a metallurgical coke works located 3 km south from the Liberty.52 If these industrial emissions have a different EC-to-BC ratio than typical traffic-dominated urban emissions, that could explain the poorer agreement at Liberty than at Lawrenceville or CMU.

05-Mar-2023

Dear Dr Presto:

Manuscript ID: EA-ART-11-2022-000166.R1
TITLE: Estimation of hourly black carbon aerosol concentrations from glass fiber filter tapes using image reflectance-based method

Thank you for submitting your revised manuscript to Environmental Science: Atmospheres. I am pleased to accept your manuscript for publication in its current form. I have copied any final comments from the reviewer(s) below.

You will shortly receive a separate email from us requesting you to submit a licence to publish for your article, so that we can proceed with the preparation and publication of your manuscript.

You can highlight your article and the work of your group on the back cover of Environmental Science: Atmospheres. If you are interested in this opportunity please contact the editorial office for more information.

Promote your research, accelerate its impact – find out more about our article promotion services here: https://rsc.li/promoteyourresearch.

We will publicise your paper on our Twitter account @EnvSciRSC – to aid our publicity of your work please fill out this form: https://form.jotform.com/211263048265047

How was your experience with us? Let us know your feedback by completing our short 5 minute survey: https://www.smartsurvey.co.uk/s/RSC-author-satisfaction-energyenvironment/

By publishing your article in Environmental Science: Atmospheres, you are supporting the Royal Society of Chemistry to help the chemical science community make the world a better place.

With best wishes,

Dr Nønne Prisle
Associate Editor, Environmental Sciences: Atmospheres

Reviewer 1

Authors revised the paper and answered in a reasonable way to my questions. I suggest to accept the paper in the current form.