Chemical characterization and source apportionment of fine particulate matter in Eastern Africa using aerosol mass spectrometry

Habineza, Theobard; Robinson, Allen L.; Dewitt, H. Langley; Gasore, Jimmy; Croteau, Philip L.; A. Presto, Albert

doi:https://doi.org/10.5194/egusphere-2025-1700

Preprints

https://doi.org/10.5194/egusphere-2025-1700

Preprints

09 May 2025

| 09 May 2025

Chemical characterization and source apportionment of fine particulate matter in Eastern Africa using aerosol mass spectrometry

Theobard Habineza, Allen L. Robinson, H. Langley Dewitt, Jimmy Gasore, Philip L. Croteau, and Albert A. Presto

Abstract. Ambient air pollution poses a significant threat to public health, particularly in low and middle-income countries, where detailed data on particulate matter (PM) mass and composition are scarce. We conducted a year-long study on PM composition and sources in Eastern Africa (Kigali, Rwanda). The annual mean concentration of PM₁ was 31 μg/m³, with slightly higher concentrations during the dry season. Organic aerosols (OA) contributed 73 % of the observed PM₁ mass, black carbon (BC) 16 %, nitrate 6 %, sulfate and ammonium 2 % each, and chlorine 1 %. BC is approximately 60 % due to fossil fuel and 40 % from biomass burning emissions. Tracer ions detected by the mass spectrometer suggest that photochemistry plays a significant role in the formation of secondary OA during the daytime (6:00 am to 6:00 pm), while primary OA dominates in the morning and evening due to increased anthropogenic activity and shallower boundary layer height. PM₁ in Kigali is primarily composed of Oxygenated Organic Aerosols (OOA, 45 %), Hydrocarbon-like OA (HOA, 32 %), and Biomass Burning OA (BBOA, 23 %). Secondary organic aerosol (SOA) accounted for 47 % and 41 % of PM₁ during the wet and dry seasons, respectively, while primary OA (POA: BBOA + HOA) contributed 53 % and 59 %. This suggests that seasonal changes in PM₁ mass in Kigali are primarily driven by deposition rather than shifts in emissions, chemical processing, or source strengths.

Received: 09 Apr 2025 – Discussion started: 09 May 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 1363 KB)

Supplement (2170 KB)

Download & links

Theobard Habineza, Allen L. Robinson, H. Langley Dewitt, Jimmy Gasore, Philip L. Croteau, and Albert A. Presto

Status: final response (author comments only)

RC1:
'Comment on egusphere-2025-1700', Anonymous Referee #1, 14 Jun 2025
Review of “Chemical characterization and source apportionment of fine particulate matter in Eastern Africa using aerosol mass spectrometry” by Habineza et al., doi:10.5194/egusphere-2025-1700.
This manuscript reports the chemical composition of particles in the city of Kigali, Rwanda, over the course of a year (April 2023 - May 2024). In addition to data from aerosol mass spectrometry (AMS) (mentioned in the title), black carbon (BC) from an aethalometer and sub-2.5 micron in diameter particulate matter (PM2.5) mass measurements are reported here. The main data analysis includes comparing seasonal variations in composition and variations of species including AMS tracer ions as a function of time of day. The AMS mass spectra were further processed by positive matrix factorization and BC data were attributed to fossil fuel or biomass burning using differences in absorption at two different wavelengths. These processed data were also compared as a function of time of day to determine source contributions to the particulate matter found in this city.
Kigali is the capital of Rwanda with a high population density and located only 2 degrees south of the equator at ~1500 m elevation. The measurements reported here appear to be the first detailed PM composition data for this city and represent a small but growing body of PM data for other locations in Africa.
Readers will find this manuscript useful and informative for understanding the air quality issues in this part of the world. I recommend publication in ACP after addressing the concern/questions noted below.
Main Concerns/Questions:
It was not mentioned until the end of the paper that Kigali has several initiatives to reduce the amount of pollution in the city. One of these is “car-free-days” and data for this paper has not been parsed for car-free Sundays/regular Sundays nor for working day/non-working day trends. There is a missed opportunity here to compare the current data set more directly with the 2017 and 2018 PM2.5 mass and BC values reported in Subramanian et al. (2020) and a newer paper by Kalisa et al. (2025, 10.1080/23748834.2025.2468017) that had additional data covering the time period through COVID-19 partial lockdown days (through June 2020). The data reported in the current paper can provide more recent information on how the car-free-days project is working with respect to presumably more emissions than in 2017-2019 with extra information on how the PM composition changes. Many questions arise relating to these comparisons. For example, has the BC/PM2.5 mass changed appreciably? Are the seasonal differences now greater or not as large as they were in the (pre-COVID-19) past. Are there differences in the relative contributions of BC from fossil fuels compared to BB from biomass burning on the cleaner days (either car-free or non-working day)? What other compositional differences (or similarities) are there for cleaner days? I suggest making an additional figure for the main manuscript, showing speciated changes (HOA, BBOA, OOA, BC_ff, BC_bb, NO3, SO4, and NH4) in different panels as a function of time-of-day for car-free Sundays/regular Sundays. Each plot should have averages with standard deviations, like Figure 6a in Subramanian et al. (2020). These might need to be separated by (combined long and short) wet or dry seasons. I urge the authors to expand on some of these and potentially other comparisons.

There are clearly AI-generated images and some associated text in the paper and SI that do not make sense. Per the Copernicus publication manuscript preparation webpage, “Should you have used AI tools to generate (parts of) your manuscript, please describe the usage either in the Methods section or the Acknowledgements.” Furthermore, ALL AI-generated content should be checked for clarity and errors before submitting manuscripts with it.

Minor Concerns/Questions:
Because more PM research is being published on Eastern Africa, consider changing title to say “Kigali, Rwanda” instead of “Eastern Africa”

Correct “slop” to “slope” everywhere.

Abstract: Include standard deviations of the averages mentioned here and throughout.

L 65-71: Add here that while there is some PM composition data from the Rwanda Climate Observatory (Kigaro et al., 2022), a remote area outside of Kigali, and measurement of BC (Subramanian et al., 2020; Kalisa et al., 2018, 2025), and PAHs, and NPAHs (Kalisa et al., 2018), no detailed PM composition data have been reported for the city itself.

L 99-101: What defines “short” and “long” seasons? The number of months for each are the same.

L 101-103: The met values described here are somewhat inconsistent with the values shown in Figure S2a. This figure appears to be AI-generated and needs to be corrected. I’m a bit skeptical of the small bands for the standard deviations of the measurements shown. In particular, the relative humidity standard deviation of the mean and rainfall amounts should indicate that the relative humidity extends up to 100% at times during the wet seasons. I suggest breaking up Figure S2a into averages for each season, and refining the seasons by examining the weekly precipitation averages. Furthermore, the solar radiation averages in Figure S2a are unrealistically too low for this location and the legend for solar radiation has the incorrect units. Since its importance is mentioned in interpreting time-of-day patterns later, is there any met data on boundary layer height that could be added, even if it’s from a different location like the airport?

L 104-109: How do motorcycles and construction activities contribute to vehicle (fossil fuel) emission sources?

L 132: Correct the wavelength listed for the Magee AE33 from one value (880 nm) to the value range for this instrument. Later in that paragraph, two wavelengths (950 and 470 nm) are used to estimate the fraction of biomass burning black carbon. Are these the two endpoints of the wavelength range for this instrument? Please clarify.

L 158: Figure S3b shows SMPS data. This instrument was not mentioned earlier. Why is the ACSM+BC mass so much higher than the SMPS total?

L 174: add “the speciated” in front of PM1

L 178-182: Clarify where these measurements were made (urban road or urban background?).

L 183-189: Are the absolute values of the various species measured in the remote area different from the absolute values measured in this study?

L 196: add “see Figure S4”. Interesting that the biggest change in the relative composition is for the long-rainy season where there is relatively more black carbon and less organic carbon. Might want to add an investigation of the BC sources as a function of season to the source apportionment section.

L 208 (and Figure 2): consider replacing Figure 2 (annually-averaged data) with the four seasonally-averaged plots shown in Figure S4.

L 229-231: If Figure 2 is replaced with Figure S4, these numbers might need to be adjusted.

L 243: Isn’t there sulfur in the diesel and gasoline fuel (L 107)?

L 247: Why is this eruption mentioned when it occurred 2 years before these measurements? It’s unlikely that the high amounts of SO2 emitted from the main eruption remained in the atmosphere over this period of time. Is there evidence of continuing SO2 emissions from the area?

L 252 and following (Section 3.3): I suggest reorganizing this section significantly. An average mass spectrum covering an entire year is not particularly useful to discuss in the main part of the paper (nor SI) and could be omitted. However, mentioning the significance of the tracer ions is useful and should probably be moved up to the data analysis section 2.3 (right before or after PMF there). The nitrate attribution calculation (Eqn. 1) could also be moved up to the data analysis section and the discussion of it could be moved up to section 3.2. Since PMF analysis was performed on this dataset, the tracer ions do not provide additional information and I suggest consolidating the relevant parts of the tracer ion discussion with the PMF results/discussion. The current Figure 3 could be moved into the SI.

L 280: Include the values of R(NH4NO3) from this dataset and the calculated R(OrgNO3) based on the ratio-of-ratios. One other thing to mention in the data analysis section about “NO3” with unit mass resolution (UMR) aerosol mass spectrometers there is potentially organic (non-nitrate) contribution to m/z 30 from the CH2O+ ion. Contributions of CH2O+ at m/z 30 could inflate both the UMR “NO3” mass concentrations and the estimation of fOrgNO3 from either approach discussed in this paper. It could potentially be especially problematic when the organic signals from other ions in the rest of spectra are relatively large. Unfortunately, the contribution of CH2O+ ion at m/z 30 cannot be corrected for the UMR data here, but it could be discussed as a potential positive bias for both “nitrate” and fOrgNO3.

L 380: The correlation plots are shown in both Figure S10 and Figure 4a (insets). Suggest removing them from Figure 4 because they are too small and already shown in Figure S10.

L 385-387 (and last line of the abstract): Is this statement valid? There seems to be a significant change in the contributions for each of the factors over the different seasons, with relatively more OOA and less HOA during the dry seasons than during the wet seasons. This result coupled with the relative composition changes shown in Figure S4 seems to indicate that there are source/sink differences between the seasons. Maybe wet deposition removes more OOA than HOA and more OA than BC? Or maybe there is more relative production/emission of OOA (and OA) than HOA (or BC) during the dry season?

L 393 and following: Like the swap between Figure 2 and Figure S4, the time-of-day trends should be shown for each season rather than an annual average shown in Figure 5. It would be good to have the relative contributions for each season as a function of time-of-day presented as well, similar to Figure 3b, but again separated for each season rather than an annual average. Separating them out by season would provide additional information on the seasonal differences.

L 393 and following: This is also the main section that I think could be expanded to a comparison of car-free Sundays/regular Sundays or working days/non-working days. There might be something interesting in the sources of BC and HOA/BBOA/OOA (and perhaps the other aerosol components) in these comparisons. I suggest an additional figure for the main manuscript, showing speciated changes (HOA, BBOA, OOA, BC_ff, BC_bb, NO3, SO4, and NH4) in different panels as a function of time-of-day for car-free Sundays/regular Sundays. Each plot should have averages with standard deviations, like Figure 6a in Subramanian et al. (2020). These might need to be separated by (combined long and short) wet or dry seasons.

L 426-430 (and Figures S9-S11): These sentences need clarification. Anywhere the text says “This versus That”, “This” is expected to be on the y-axis and “That” is expected to be on the x-axis. Note that the information shown in Figure S10 is the factor correlations with the reference spectra, already mentioned in L 380. Figure S11 is essentially showing the fraction of those ions in the factor spectra. It is not surprising that they are so well-correlated because the factors are defined with a specific ratio for each of those ions. These ratios are the “fractions of ion signals in the organic spectra” aka f44, f57, and f60. From the factor spectra in Figure 4, they appear to be around 0.095 for f57 in HOA, 0.035 for f60 in BBOA, and 0.24 for f44 in OOA. It would be good to note what those values are when first discussing the factors (L 368-374) and maybe discuss their significance. Figure S9 is a bit troublesome because the r-squared values make the correlations seem to be better than they appear in the plots. Glancing at the plots without looking at the r-squared values makes it appear that BC_bb is linearly associated with all three factors, whereas BB_ff does not appear to be linearly associated with BBOA or OOA and only loosely associated with HOA. Perhaps they would look better if the marker sizes were smaller? The words in the text do not seem to effectively convey what these correlation plots seem to be showing.

L 431-440: It is unclear what current results are intended to be compared with these other studies, which indicate that there was a fraction of the total PM2.5 attributed to traffic. Where is that result for the current manuscript? 32% of OA is HOA (annually averaged, L 382) and 59% of BC is BC-ff (annually averaged, L 409). Depending on how the contributions from traffic were determined from the other studies, the first sentence of these paragraphs should mention how the fraction of PM2.5 was determed to be the traffic contribution (is it HOA+BC_ff divided by PM2.5 mass?) and provide comparable seasonal values for the current study. What about the SOA from traffic?

L 441-466: How many of these policies are currently in effect for Kigali? As mentioned in the main concerns, it would be useful to have comparisons with prior studies in Kigali and assess any progress with these policies. Is the reduction in PM2.5 better/worse from those policies now than presented in the Subramanian et al. (2020) paper? Emissions everywhere have presumably increased since then.

L 496-499: This is different from what is said at the end of the abstract, which should probably be revised.

L 501-510: This is somewhat inconsistent with first paragraph in the conclusion (L 469-472), because the results of this study indicate that pollution in Kigali is local (primary from vehicles and cooking) and is not related to transboundary issues. A discussion should be added to the end describing how well the current pollution control policies are working for the city.

Figure 1 and others: The font for the legends are too small.

Figure 2: Suggested above to replace it with Figure S4a.

Figure 3: Suggested above to move it to the SI.

Figure 4: Suggested above to delete the correlation plots in part a. Label the important ions in part a. The shading in part b makes it look like there isn’t any HOA during the rainy season. Consider another way to distinguish the different seasons. The legend obscures a large part of the plot.

Figure 5: Use consistent colors for Figures 4 & 5. Since data like this from the study are used to estimate source apportionment, it would be interesting to see if there are seasonal contrasts (or not) as a function of time-of-day. Suggest replacing the current figure with a new 4-panel plot of that. Suggest adding a corresponding figure to the SI examining the mass fractions of the (OA+BC) from HOA, BBOA, OOA, BC-ff, and BC-bb for each season as a function of time-of-day.

Section 9 (References): Check for multiple listings of the same paper (e.g. Andersson et al. 2020 and Subramanian et al. 2020). Note that the Subramanian et al. 2020 doi number is incorrect. The correct number is 10.17159/caj/2020/30/2.8023.

Figure S1: AI-generated. Remove labels for businesses on map and in legend. What is the significance of the pin? It would be more appropriate to show locations of the major highways along with the other locations mentioned in the main manuscript, such as the location of the met data (Kivugiza meteorological station) and the PM2.5 data (US Embassy). How far out does the metropolitan area extend? Could have a second map to show that.

Organic Nitrate Section in SI needs to be re-written more clearly. For example, Eqn. 1 is for the organic nitrate mass (preceding sentence says fraction) and Eqn. 2 is for the organic nitrate fraction of the total nitrate mass (preceding sentence says particulate organic nitrate). Also, please use the same abbreviations/definitions for the variables as used in the main paper, so that Eqn. 2 is identical to Eqn. 1 in the main paper. Because the variables were ambiguous, I was unable to determine what was being plotted in Figure S5.

Other sections of the SI also need to be clarified and corrected for errors in the AI-generated content.
Citation: https://doi.org/10.5194/egusphere-2025-1700-RC1
- AC1: 'Response to reviewers(1 and 2) comments', Theobard Habineza, 12 Sep 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1700/egusphere-2025-1700-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-1700-AC1
RC2:
'Comment on egusphere-2025-1700', Anonymous Referee #2, 17 Jun 2025

Review of Habineza et al
The authors present aerosol chemical composition data in Kigali Rwanda using a Q-ACSM and a BC aethalometer over a 12 month period. This is an impressive dataset in a critically understudied location. The scientific results are robust, though the writing, formatting, and proofreading could use some improvement. I recommend publication subject to what are largely minor revisions.
My only major comment is that the paper aims to be representative of all of East Africa; however, then spends a lot of time focusing in specifically on one location in Kigali, and attributing results to local behaviors in Kigali. East Africa is a large, heterogenous place, with larger cities than Kigali. The authors should spend some effort justifying why they think these results are representative of broader East Africa. A few sentences to a paragraph should be sufficient.

Minor comments:
There are a number of typos throughout the manuscript, be sure to proofread carefully.

Abstract: The abstract ends kind of abruptly after a detailed result about seasonality. Consider adding some kind of broader concluding statement. I also don’t quite follow how the pretty small seasonal variation in the SOA/POA split (47/53 wet and 59/41 dry) suggests that (wet) deposition is driving PM mass changes? The differences between dry and wet aren’t very large. Maybe this is clarified further in the paper but the abstract should be internally consistent as well.

Intro: There appear to be some citation issues (“n.d.” for no date after a date is given).
There is also some repetition in the intro. For instance, line 65 says “There is limited information on PM composition and sources in East Africa.” But line 48-49 already makes this point.

Methods: the total PM2.5 mass data being 10 km away from the ACSM and BC sounds like a significant uncertainty. Especially in a heterogeneous city like many growing cities in Africa. How does this impact the comparisons? Also, clarify in the text if the Rwanda Meto station is at a third location and how far that is from everything else?
Does the aethalometer AE-33 have a 1 µm cut point PM cyclone as well? If the NR-PM1 is simply added to the BC from the AE-33 without a PM1 cyclone, the assumption would be that all BC is 1 µm or less. Not necessarily a bad assumption, but it should be cited/documented if so.
Results: Line 217 and several other parts of section 3.2 mentions a PBL height impact. Is this a result? If so, provide the data that supports it. If it is a general assumption add a citation.
Line 245: Regarding the discussion of SO2 or sulfate sources. What about vehicle fuel? Any diesel being used in Kigali?
Line 385. Related to the final line of the abstract, I think the reasoning for why the lack of seasonal variation implies that rain (wet deposition) is driving composition changes. As written it is a bit unclear.

Citation: https://doi.org/10.5194/egusphere-2025-1700-RC2
- AC1: 'Response to reviewers(1 and 2) comments', Theobard Habineza, 12 Sep 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1700/egusphere-2025-1700-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-1700-AC1

Theobard Habineza, Allen L. Robinson, H. Langley Dewitt, Jimmy Gasore, Philip L. Croteau, and Albert A. Presto

Supplement

https://doi.org/10.5194/egusphere-2025-1700-supplement

Theobard Habineza, Allen L. Robinson, H. Langley Dewitt, Jimmy Gasore, Philip L. Croteau, and Albert A. Presto

Viewed

Total article views: 787 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
667	105	15	787	33	20	31

HTML: 667
PDF: 105
XML: 15
Total: 787
Supplement: 33
BibTeX: 20
EndNote: 31

Views and downloads (calculated since 09 May 2025)

Month	HTML	PDF	XML	Total
May 2025	95	33	5	133
Jun 2025	49	16	5	70
Jul 2025	33	13	0	46
Aug 2025	81	12	0	93
Sep 2025	375	14	3	392
Oct 2025	34	17	2	53

Cumulative views and downloads (calculated since 09 May 2025)

Month	HTML	PDF	XML	Total
May 2025	95	33	5	133
Jun 2025	49	16	5	70
Jul 2025	33	13	0	46
Aug 2025	81	12	0	93
Sep 2025	375	14	3	392
Oct 2025	34	17	2	53

Viewed (geographical distribution)

Total article views: 791 (including HTML, PDF, and XML) Thereof 791 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 23 Oct 2025

Short summary

This study reports year-long PM₁ chemical composition in Eastern Africa using aerosol mass spectrometry. Results show PM is dominated by organic aerosol (73 %), black carbon (16 %), and inorganics (11 %), with BC largely from fossil fuel (59 %) and biomass burning (41 %). Findings highlight the impact of solid fuels and aging vehicles and stress the need for regional mitigation strategies to reduce air pollution-related health risks.


Total:	0
HTML:	0
PDF:	0
XML:	0