Variability of methane content in bottom waters of 46 African lakes

Borges, Alberto V.; Morana, Cédric; Deirmendjian, Loris; Okello, William; Omeja, Patrick; Isumbisho, Mwapu; Descy, Jean-Pierre; Bouillon, Steven

doi:10.5194/egusphere-2026-1976

Preprints

https://doi.org/10.5194/egusphere-2026-1976

Preprints

14 Apr 2026

| 14 Apr 2026

Status: this preprint is open for discussion and under review for Biogeosciences (BG).

Variability of methane content in bottom waters of 46 African lakes

Alberto V. Borges, Cédric Morana, Loris Deirmendjian, William Okello, Patrick Omeja, Mwapu Isumbisho, Jean-Pierre Descy, and Steven Bouillon

Abstract. Methane (CH₄) accumulates in bottom waters of lakes, however, the extent and drivers of inter-lake variation in bottom-water CH₄ concentrations are poorly understood and have been studied mainly in northern lakes. This limits predictions of how bottom-water CH₄ concentrations respond to warming and eutrophication in lakes, and how these changes might influence surface-water CH₄ concentrations and, consequently, CH₄ emissions. We report 168 measurements of paired bottom- and surface-water CH₄ concentrations from 46 African lakes spanning a wide range of surface area (SA; 0.02–67,075 km²) and maximum depth (2–180 m). Bottom-water CH₄ concentrations ranged from 7 to 5,608,382 nmol L⁻¹, spanning six orders of magnitude, and increased with increasing stratification, quantified from vertical density profiles using potential energy anomaly (PEA) and mixed layer depth (MLD), and inferred from NH₄⁺ concentrations or vertical conductivity gradients. Surface-water CH₄ concentrations ranged from 7 to 168,114 nmol L⁻¹ and increased with both bottom-water CH₄ concentrations and vertical stratification (positively related to PEA and negatively to MLD). The most strongly stratified lakes exhibited high bottom-water CH₄ concentrations, resulting in enhanced vertical transfer of CH₄ to surface waters despite lower vertical diffusion coefficients. In addition, these lakes had shallower mixed layers and therefore thinner oxygenated surface layers, likely reducing CH₄ removal via methane oxidation. The positive relationship between both bottom- and surface-water CH₄ concentrations and chlorophyll-a (Chl-a) has previously been interpreted as reflecting enhanced methanogenesis driven by phytoplankton-derived organic matter delivered to sediments. However, such relationships may be indirect and should be interpreted cautiously, as Chl-a was negatively related to MLD in our dataset, and both bottom- and surface-water CH₄ concentrations were also negatively related to MLD. The ratio of surface to bottom CH₄ concentrations (surface:bottom CH₄ ratio) may indicate the relative increase in surface CH₄ in response to increases in bottom CH₄ driven by warming and eutrophication. This ratio was negatively related to bottom depth, PEA, and MLD, and positively related to bottom-water O₂, indicating that the relative increase in surface-water CH₄ with increasing bottom-water CH₄ is greater in shallower, less stratified systems than in deeper, more stratified systems. Diffusive CH₄ emission rates were highest in shallower, less stratified systems, where the response of surface-water CH₄ to increases in bottom-water CH₄ is expected to be greatest, as indicated by high surface:bottom CH₄ ratios. We further tested whether surface-water CH₄ concentrations scale with simple metrics in a dataset including highly stratified, small, and deep crater lakes with elevated hypolimnetic CH₄. A multiple linear regression using SA and Chl-a explained ~51 % of the variance and appears suitable for upscaling dissolved CH₄ concentrations. This approach could enable large-scale extrapolation of diffusive CH₄ emissions using spatial datasets for SA and remotely sensed Chl-a.

Received: 07 Apr 2026 – Discussion started: 14 Apr 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Biogeosciences.

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, 1814 KB)

Supplement (396 KB)

Download & links

Alberto V. Borges, Cédric Morana, Loris Deirmendjian, William Okello, Patrick Omeja, Mwapu Isumbisho, Jean-Pierre Descy, and Steven Bouillon

Status: open (until 03 Jun 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-1976', Anonymous Referee #1, 28 Apr 2026 reply
Comments to “Variability of methane content in bottom waters of 46 African lakes” by Borges et al.
General comments
This manuscript presents an extensive dataset of dissolved CH₄ in surface and bottom waters in African lakes. In combination with measurements of various other biogeochemical parameters, the authors identify vertical stratification strength as a dominant driver of CH₄ concentrations using a regression analysis. The assessment of African lakes is especially valuable, since these lakes are underrepresented in the existing literature. The manuscript is well structured and presents the results in a clear manner.
However, there are some inconsistencies regarding the log-transformation that need to be addressed before publication and an improved statistical analysis could strengthen the interpretation of the data. Overall, my concerns can be easily resolved by a revision of the manuscript.
Specific comments
l. 223: It seems to me that the log₁₀ transformation has not been applied in all instances. As large parts of the discussion rely on interpreting the strengths of correlations, it should be clearly stated and justified when data were not log-transformed. My identified issues are:
Table S2 states a log transformation for most relationships but not for all.

It states it was not applied to bottom depth for surface CH₄ bottom depth (Fig 8a). Why was it not applied in this instance but for bottom CH₄ vs. bottom depth (Fig. 6e)?

The regression in Fig. 6d for large lakes seems linear in log vs. non-log space, however, Table S2 states log transformation of both variables.

Bottom depth and MLD were also not log-transformed in Fig. 9a and b, respectively. Both were log-transformed in other instances.

There are several instances where R² values allow for ambiguous causation patterns (e.g., ll. 267-272, ll. 302-308, ll. 327-329). A refined statistical analysis would make the interpretation clearer. For example, partial correlation coefficients can help disentangle direct/indirect relationships and could make the interpretation more straightforward.

It has been shown that that the production of methane under oxic conditions can contribute significant amounts of CH₄ to lake surface water (e.g., Ordóñez et al. 2023, Thottathil et al. 2022). However, the importance of this process is little discussed in the manuscript, even though there is a positive correlation between CH₄ and Chl-a. The authors seem to have submitted another manuscript (not published yet) addressing this topic. Nevertheless, it might be useful to mention the relevant results so it becomes clear to the reader why the process is assumed to be less important for the studied lakes.

The suitability of the proposed model for large-scale upscaling extrapolation of dissolved CH₄ concentrations seems a bit overstated in the abstract. A substantial part of the variance remains unexplained. Furthermore, as the model is based on data from a limited geographical region, it is unclear if the relationship can be spatially extrapolated. Incorporating other published data could make the relationship more transferable. Otherwise, emphasizing regional applicability and discussing transferability limitations would be adequate.

Technical corrections
ll. 294-297: Estimating the apatite saturation index could make this argument stronger

l. 299: Fig. 6e instead of 6d?

Beaulieu et al. (2019) and Cole and Caraco (1998) appear twice in the references

Several figure captions in the SI refer to Table S1 instead of S2

Table S2:
Figures are listed in an inconsistent order

Typo for Figure 6g (66 instead of 6g)

Figure 5b written instead of 5e, f

References
Ordóñez, C., DelSontro, T., Langenegger, T., Donis, D., Suarez, E. L., & McGinnis, D. F. (2023). Evaluation of the methane paradox in four adjacent pre-alpine lakes across a trophic gradient. Nature Communications, 14(1), 2165.
Patel, L., Singh, R., & Thottathil, S. D. (2024). Contribution of photosynthesis-driven oxic methane production to the methane cycling of a tropical river network. ACS ES&T Water, 4(7), 2836-2847.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-1976-RC1
RC2:
'Comment on egusphere-2026-1976', Anonymous Referee #2, 22 May 2026 reply
This manuscript by Borges et al. describes patterns of methane concentrations in the surface and bottom waters of 46 African lakes. The authors observed substantial variability in bottom-water methane concentrations among lakes, and they explain these differences using linear regressions with metrics related to lake morphometry, stratification, and chlorophyll-a.
This dataset and analysis addresses an important gap in our understanding of greenhouse gas concentrations in lakes. Data for bottom-water gas concentrations are needed to develop, calibrate, and improve models of greenhouse gas emission from lakes, and African lakes are underrepresented in existing greenhouse gas concentration datasets. However, I have several concerns described below, primarily regarding the statistical analysis of these data. With revision, this analysis could make important contributions to advancing our understanding of methane cycling in lakes.

Major comments:
One of the strengths of this dataset is that it contains lakes with widely varying size and shape. However, the authors bin the lakes into two size classes (small and large) for most analysis, which does not fully take advantage of this diversity of lakes. For analyses in Figure 4, it would be easy to switch to a regression approach, treating lake size as a continuous variable. In analyses of greenhouse gas concentrations (e.g., Figure 5, 6, 8, 9), lake size could be added as a continuous variable using a multiple regression approach, which the authors already do in one regression (Table 1). Alternatively, or in addition, it may be useful to analyze the relationship between gas concentrations and a metric such as the dynamic ratio, which integrates both lake depth and surface area rather than treating these two variables separately.
Most lakes were sampled multiple times, but the authors do not currently account for repeated measurements in their analysis. Using mixed effect models with a random effect for lake ID would better address these relationships. More broadly, bottom-water methane concentrations are expected to vary dramatically throughout the year due to changes in stratification, biogeochemical processing, and other factors. The manuscript is currently lacking any discussion or analysis of intraannual variability in gas concentrations between the rainy and dry period sampling.
The authors calculate diffusive methane fluxes using surface concentrations and monthly mean wind speeds sourced from WorldClim. However, wind speeds vary substantially across fine spatial and temporal gradients, so these are likely very poor estimates of the actual wind speed at the time of sampling, and diffusive flux estimates are highly sensitive to k600. I do not think diffusive fluxes are necessary in this manuscript, which is primarily focused on bottom-water gas concentrations, and I would recommend removing this analysis.

Minor comments, by line:
Lines 45-47: I am confused by this sentence—is this a separate dataset?
Line 48: An R² of 0.51 is still fairly low, and it is not clear what metric the authors are using to conclude that this linear regression “appears suitable for upscaling dissolved CH4 concentrations”
Lines 61-63, and throughout: Rabaey et al. 2026 may be a useful reference for contextualizing the concentrations in this study
Rabaey, J. S., Lewis, A. S. L., Attermeyer, K., Aurich, P., Bansal, S., Bartosiewicz, M., et al. (2026). Depth-resolved carbon dioxide and methane concentrations in 522 lakes, ponds, and reservoirs worldwide. Scientific Data, 13(1), 483. https://doi.org/10.1038/s41597-026-06751-0

Line 139: This sentence is missing the word “water” in “surface water”
Lines 170–181: I am not convinced that the use of 0.2 ºC as a threshold for thermal stratification is appropriate. Density gradients would be a more robust way of classifying whether lakes were stratified (e.g., the threshold of 0.1 kg/m3, as the authors reference on line 77). That tropical lakes tend to have weaker vertical density gradients (lines 179-181) does not justify the use of a different threshold for stratification in these systems. At a minimum, it would be helpful to have a figure illustrating the relationship between thermal gradients and oxygen gradients in the supplemental information to justify this choice of threshold.
Line 218-219: Why not use secchi depth directly? Also, it would be helpful for readers if this relationship was restated here rather than having to refer to Morana et al. (2022).
Line 223: Was the distribution normal after log-transformation?
Line 231: I would not consider MLD to be a metric of stratification strength. Please explain or rephrase.
Line 231 and throughout: I am not convinced that PEA is the relevant metric of thermal stratification here. PEA describes the energy required to mix the entire water column, but gases could exchange between surface and bottom waters without mixing the entire water column. Buoyancy frequency at the thermocline might be more relevant here?
Line 239: It is not clear what is meant by “at the level of individual measurements”
Lines 242-244: Again, this is complicated by the use of PEA, which is likely to be larger in deeper lakes. Is the same true for surface-bottom density differences or buoyancy frequency at the thermocline? More broadly it would also be helpful to have an explanation of why these lakes would show contrasting patterns compared to northern lakes.
Line 255: I agree that this is generally true, but it contrasts with the point the authors made earlier about surface area and maximum depth being poorly correlated in this dataset (e.g., Figure 2).
256: Please clarify that these are negative correlations.
Lines 286-288: This sentence implies that anoxia is simply an indicator of stratification, but methane is not expected to accumulate even under strongly-stratified conditions until oxygen levels approach anoxia (thereby increasing NH₄ and conductivity).
Line 284: How was anoxia defined in this study?
Line 292: I would assume that this is due to anoxia in small lakes, which in turn leads to PO₄^3- release from sediments.
Lines 303-308: This does not necessarily indicate an indirect relationship. Chlorophyll-a is temporally variable and an imperfect proxy for phytoplankton biomass. Even if this relationship does not show up across lakes, there is a direct mechanistic connection between autochthonous productivity and methanogenesis (as the authors describe in the introduction).
Lines 309-321: This paragraph is great. Very interesting, and highlights the importance of the work well.
Line 321: What specific “functional and climatic differences”?
Line 327-329: This relationship could also result from many other processes including oxic methanogenesis and inputs of CH₄ from bottom waters
Lines 353-356: I am not clear on what subset of lakes was used for this analysis
Lines 361-366: This is a useful point that could be made more forcefully- the empirical relationships developed elsewhere do not hold for African crater lakes
Line 406: Is there a reason you do not count Kraemer et al. 2022 here?
Kraemer, B. M., Kakouei, K., Munteanu, C., Thayne, M. W., & Adrian, R. (2022). Worldwide moderate-resolution mapping of lake surface chl-a reveals variable responses to global change (1997–2020). PLOS Water, 1(10), e0000051. https://doi.org/10.1371/journal.pwat.0000051

Line 448: Data are not currently available- access is restricted
Figure 3: Panels d and h do not add much additional information to this figure, and sample sizes are very different between the two classes. I recommend treating lake size as a continuous variable throughout.
Figure 4: This figure would be much more informative if lake size was treated as a continuous variable rather than binning the data into size classes (i.e., creating scatterplots with regression lines rather than boxplots)
Figure 6: Are these all fundamentally oxygen relationships?
Figure 7: I do not think this figure is currently helpful. It is not clear where the numbers come from (e.g., how they were calculated using the data in this study, or which numbers come from Holgerson et al.). “Very stratified” is also unclear and subjective.
I hope these comments are helpful. This is a great dataset and has the potential to be an impactful contribution.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-1976-RC2
RC3: 'Comment on egusphere-2026-1976', Anonymous Referee #3, 22 May 2026 reply

The authors present an impressive dataset covering >40 lakes spanning a wide range of environmental conditions in a region where methane concentrations remain poorly constrained. The potential value of this dataset is however reduced by the relatively superficial analysis. The predictors used in the numerous linear regression relationships appear to have been selected rather arbitrarily, and the causal relationships are often unclear. Proposed mechanisms are repeatedly described as “likely,” while the analysis rarely goes beyond simple linear regressions, some of which are statistically insignificant. In several cases, regressions are constructed between variables that are not independent a priori, such as mixed-layer depth and stratification strength.

The studied lakes differ by orders of magnitude in both environmental characteristics and methane concentrations, yet the analysis relies on a simple binary classification into “shallow” and “deep” lakes. Could the lakes instead be grouped into more meaningful categories, either based on their environmental characteristics or using unsupervised statistical clustering? I noticed a similar remark by another reviewer and fully agree with it. In general, the dataset appears to offer possibilities for a substantially more sophisticated analysis than a collection of linear regressions applied to the entire lake ensemble. A thorough reconsideration of the analytical framework is therefore strongly recommended, ideally based on a more process-oriented selection of predictive variables.
The presentation of the results is overloaded with repetitive figures and difficult-to-read tables (many of which moved to the Supplementary). A more synthetic and concise presentation would substantially improve readability.
The manuscript itself is difficult to follow. It contains too many unnecessary abbreviations and lacks a clear structure. A stricter separation between the Results and Discussion sections is strongly recommended and would help the reader follow the main message. Statements containing the word “likely” should generally be avoided outside the Discussion section.
The attempt to quantify the vertical methane fluxes based on rough estimates of Kz and climatic data on wind is highly uncertain and, in my opinion, could be omitted altogether. Here I also agree with the concerns raised by the previous referee.

Specific remarks:

- 90-92 “In stratified lakes, MOX at the pycnocline and in the mixed layer reduces surface-water CH4, whereas in well-mixed systems, CH4 concentrations tend to be vertically homogeneous.” What is meant here? Does MOX not reduce concentrations in well-mixed lakes? The statement should be clarified. Cf. Lines 281-283.

- 139 Typo: Two liters of surface were collected for ancillary measurements. ‘Water’ is missing.

-122-123 What was the vertical resolution of the YSI EXO-II profiling?

- Line 189-203 (Eqs. 2-4): How the thickness of the thermocline was determined (see also the comment on Lines 122-123)? The concentration difference between the surface and 1 m above the lake bottom is not representative for the concentration jump at the thermocline unless the thermocline is very close to the bottom because the methane content quickly decreases with height due to oxidation and reduced vertical transport due to stratification. The approximation Kz = 3x10-10(N2)-1 suggested by Katzev et al. (2007), when reformulated in terms of the Osborn (1980) expression for diapycnal mixing, means that the TKE dissipation rate ε = 1.5e-9 m2s-3, which is comparably low. Overall, its application to the variety of very contrasting lakes and different seasons is questionable. Note that PEA (Eq. 1) and Kz (Eq. 2) are inherently connected making the correlation in Fig. S6c weakly informative. The thermocline flux estimates are uncertain and better be removed completely.

-Line 235 “MEL normalized by Zmax explained more variance of PEA (r2=0.53, Fig. 3g) than MEL alone (r2=0.25, Fig. 3e)”: MEL/Zmax is a measure of the lake aspect ratio. Logically, deeper lakes with a shorter fetch are stronger stratified. Is Zmax the same as h in Eq. 1?

- 241: “MLD decreased in more strongly stratified lakes, as indicated by its negative relationship with PEA (Fig. S3b).” The statement is confusing. A thinner MLD means a stronger stratification, so the casualty is not clear here. To reduce the hidden correlation between MLD and stratification, PEA can be replaced by a more robust quantity, like the bottom-surface difference of density (temperature).

- 371-373 “To try accounting for intra-lake depth effects in addition to inter-lake SA effects, we tested the relation between surface-water CH4 concentration and SA:bottom-depth that was significant but described less variance (r2=0.420) than SA alone (r2=0.470)”. What kind of intra-lake effects is meant here? The ratio of surface area to lake depth is more or less the same as the characteristic lake length (fetch).

- 383 “MLD and PEA cannot be used as predictors for upscaling purposes as they are not readily available from global or regional spatial datasets.” The statement sounds strange. There are a number of methods and models for estimation of lake stratification from atmospheric forcing with different degrees of certainty, and the atmospheric data are available globally. The lake sector of ISIMIP mentioned in Section 2.2 is a good reference for that.

- 384-387 “The fact that MEL explained more variance (r2=0.483) than SA (r2=0.470) could be linked to the role of stratification/mixing in regulating surface CH4 concentrations. MEL is a better predictor than SA of fetch that determines the vertical mixing and MLD in lakes at equivalent wind speed”. The difference in correlation with SA and MEL is too small for such a conclusion. Probably, the difference will vanish completely if the square root of the surface area (mean fetch) is used as a predictor instead of SA.

Reply

Citation: https://doi.org/10.5194/egusphere-2026-1976-RC3

Alberto V. Borges, Cédric Morana, Loris Deirmendjian, William Okello, Patrick Omeja, Mwapu Isumbisho, Jean-Pierre Descy, and Steven Bouillon

Supplement

https://doi.org/10.5194/egusphere-2026-1976-supplement

Alberto V. Borges, Cédric Morana, Loris Deirmendjian, William Okello, Patrick Omeja, Mwapu Isumbisho, Jean-Pierre Descy, and Steven Bouillon

Viewed

Total article views: 314 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
210	81	23	314	30	16	17

HTML: 210
PDF: 81
XML: 23
Total: 314
Supplement: 30
BibTeX: 16
EndNote: 17

Views and downloads (calculated since 14 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	159	63	20	242
May 2026	51	18	3	72

Cumulative views and downloads (calculated since 14 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	159	63	20	242
May 2026	51	18	3	72

Viewed (geographical distribution)

Total article views: 314 (including HTML, PDF, and XML) Thereof 314 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 25 May 2026

Short summary

We report 168 measurements of paired bottom- and surface-water CH₄ concentrations from 46 African lakes spanning a wide range of surface area and maximum depth. Bottom-water and surface-water CH₄ concentrations increased with stratification. The ratio of surface to bottom CH₄ concentrations was negatively related to bottom depth and stratification, indicating that the relative increase in surface-water CH₄ might be greater in shallower and less stratified systems.


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