the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Jul 2025
Variability of greenhouse gas (CH4 and CO2) emissions in a subtropical hydroelectric reservoir: Nam Theun 2 (Laos PDR)
Abstract. Hydroelectric reservoirs, while supporting renewable energy production, contribute notably to greenhouse gas (GHG) emissions in tropical and subtropical regions. This study presents a 14-year (2009–2022) analysis of CH4 and CO4 emissions from the Nam Theun 2 (NT2) reservoir, using a combination of discrete water sampling, bubbling funnel traps, and high-frequency eddy covariance (EC) measurements. Emission pathways assessed include diffusion, degassing, and ebullition. EC results of CO2 fluxes were consistently higher than estimates from discrete samplings due to the system’s placement in shallow, high-emission areas and its capacity to capture real-time turbulence and diurnal variability. In contrast, CH4 fluxes from eddy covariance were often lower than discrete-based calculations, particularly in later campaigns, due to spatial limitations, wind filtering, and reduced sensitivity to bubbling events. The findings highlighted the importance of integrating multiple techniques to address spatial and temporal variation and reveal the influence of reservoir stratification, carbon cycling, and hydrological operations on greenhouse gas dynamics. Over the study period, total emissions reached 10736 Gg CO2eq, with CH4 contributing 51 % (5468 Gg CO2eq) and CO2 49 % (5268 Gg CO2eq). Emissions peaked in 2010 (1276 Gg CO2eq) and declined by approximately 70 % by 2021, driven by reservoir aging and depletion of labile organic matter. Seasonally, the warm dry season accounted for 3809 Gg CO2eq (35.5 %), the cold dry for 3841 Gg CO2eq (35.8 %), and the warm wet for 3086 Gg CO2eq (28.7 %). CH4 emissions were highest in the WD season due to enhanced ebullition under stratified and anoxic conditions, while CO2 emissions peaked in the CD season due to reservoir overturn and increased respiration. Ebullition dominated CH4 emissions (77 %) and remained stable over time, whereas diffusive CH4 fluxes declined by 97 %. CO2 emissions were majorly diffusive (96 %) and showed consistent decline (-87 %). This long-term dataset improved the understanding of subtropical reservoir emissions and provided insights for global carbon budgets and improving the climate impact assessment of hydropower development.
- Preprint
(1237 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 17 Sep 2025)
-
RC1: 'Comment on egusphere-2025-3295', Alex Zavarsky, 18 Aug 2025
reply
The paper is well within the scope of BG and addresses relevant scientific questions especially concerning CH4 emissions from dams which could also be extended to lakes and even rivers in similar climate zones. Especially interesting is the use of various techniques (EddyCov, point sampling and ebullition) of measuring the sources and sinks of CO2 and CH4. A 14-year dataset is a very nice dataset to do research and especially as they are capable of looking at intra-annual features.
The conclusion is basic and lacks, especially given the long dataset, being put into perspective to other dams, lakes and global CH4 sinks/sources from similar environments.
The language is fluent and precise. However, Part 4 Discussion, is in my opinion, lacking a focus on substantial points, together with graphs supporting them. It is really hard to read and an endless point after point list of every feature found in the dataset. It would have helped to focus on less and support this with graphs. Maybe adding more sections and grouping the discussion would be helpful.
This paper, is in my opinion a continuation of previous work of Mr. Deshmukh which is also an author of the paper and is cited accordingly.
I support this paper because it is a summary and interpretation of 14 years vast dataset at such a hydroelectric reservoir. I would suggest major for section 4, with a better arrangement and partitioning into subsections.Scientific question:
Line 55 and after: There are three mechanisms: ebullition, diffusive fluxes and degassing. One could briefly explain what these three are and how they are measured. Ebullitionà Bubble traps. Diffusive fluxes EddyCov, K*DeltaC, ebullition upstream-downstream.
Line 88: The Dam was impounded and the commissioned. Was happened when it was commissioned? Water through the turbines? Was the water before discharged via the spill-over?
Line 250: What are gross emissions. Is it already source and sink subtracted? Is it influx of OM minus GHG coming out?
Line 505-510: The ebullition effect of atmospheric pressure change. Did you see this also in the EC data?
Line556: GE measurements (TBL and bubbling) is this the calculation method for Gross Emissions? This should be explained before.Line 591: In the methods section there should be a clear definition of EC and TBL(GE) method. Then just use one abbreviation TBL or GE. I think that the way of calculating GE is through TBL. That should not be mixed up.
Line 608: kt values are often highly discussed and vary regarding which parametrization you use. This could be mentioned earlier when you compare the fluxes.
Line 617: “Temporal dynamics of CH4 emissions from the reservoir water surface.” Why is the abstract called “from the water surface” you mention diffusive fluxes, ebullition and degassing and water discharged at the pill-over. What is so significant to the water surface now? You describe EC before and now the other pathways. I would choose a different subtitle.
Line 631: Do you mean the water is discharge from the reservoir or coming from the surroundings into the reservoir?
Line 780: Degassing + Feature. It should be written that the relative minor role of degassing due to the features at the damn, suggests that future projects….
Line 790: This is a very interesting paragraph putting your measurements in relation to others.
Line 824: Can you just briefly remind us what the design features are: intake depth, ventilation, … just one or two catchwords.
General question: Did you take a deeper look at water-level influence. You mention the hydrostatic pressure changes influencing ebullition but only cite Deshmukh 2014. Have you seen any influence from falling dry and resubmerged banks?
Technical remarks:
Abstract Line 31: I would spell it out in the abstract "warm dry". The same goes for cold dry. They are not too long. Its perfectly fine in the main body
Line 43: At the first mentioning I would write greenhouse gas (GHG)
Line 470/ Figure 5: You use ebullition and bubbling. I would recommend only use ebullition. This also would be appropriate for figures.
Citation: https://doi.org/10.5194/egusphere-2025-3295-RC1 -
RC2: 'Comment on egusphere-2025-3295', Anonymous Referee #2, 03 Sep 2025
reply
Hoang and authors present a detailed narrative of emissions from a well-studied hydroelectric reservoir over multiple temporal scales. The breadth of the data and variability in emissions from this reservoir is a great contribution, especially as few hydroelectric reservoirs have long-term and/or seasonally resolved data to understand temporal patterns at these scales. However, I struggled to clearly understand what the most important findings of this study were, and how it was a distinct and novel contribution relative to previous work in this reservoir. My foremost suggestion is majorly trimming down much of the Results and Discussion to focus on the most important and new findings and comparisons, while better integrating quantitative statistical comparisons and clear mechanistic linkages. I detail a few overarching suggestions first, followed by several minor comments for your consideration.
Overarching suggestions:
- Presentation of results and discussion – Much of the Results and Discussion are very long and detailed, and distract from the most important findings of the paper. For example, Section 3.2 in the Results takes up a lot of space for being only ancillary to the main findings of this paper. I suggest significantly reducing this section and perhaps putting it at the beginning of the Results (details could go in supplementary materials). The Results could then begin with a very brief overview of these vertical chemical dynamics, and follow with emphasis on the emissions comparisons between methods and over time, which are the real highlights of this paper. Similarly, the Discussion is very long winded, with many sections reading like a laundry list of similar studies without clearly linking to key findings of this study. I suggest focusing on linking specific results with mechanisms and relevant papers, and highlighting the specific, novel contributions from this study. Some closing paragraphs for long sections (i.e., paragraphs beginning on line 699 and 766) could succinctly summarize much of the section without needing multiple pages of text.
- Integrating statistical tests – Much of the Results section uses visual or tabular comparisons to make statements about differences in emissions rates, when statistical tests should explicitly be used. For example, comparisons of GE versus EC emissions are only presented using mean and uncertainty, when t-tests could easily compare these rates across all years. The authors write in line 349 emissions rate from EC were significantly higher than GE in some campaigns, including in March 2011 as 1.31 ± 0.10 Gg CH4 month-1 (EC) versus 1.25 ± 0.23 Gg CH4 month-1 (GE). However, considering the uncertainty values, these are not really different than each other. Please incorporate appropriate statistical tests for these and other comparisons (i.e., statistical trend tests for results in lines 486-492; mechanistic statement in line 620) rather than only qualitatively or visually describing results.
- Upscaling from few stations to entire reservoir – The 3-7 sites used for upscaling emissions rates to the whole-reservoir scale seem very few to precisely capture the full range in spatial variability in emissions rates. How does this number of sites compare to recommendations in Beaulieu et al. 2016 (https://doi.org/10.1002/lno.10284, see Fig. 5) or to the area covered by each sampling site in previous work (i.e., Jager et al. 2022, https://doi.org/10.1016/j.rser.2022.112408, see Fig. 6). The few number of sites is a major caveat for this scale of whole-reservoir upscaling and needs more attention and/or Discussion.
- Linking proposed mechanisms to observed patterns – There are several instances in the Discussion where mechanisms are briefly proposed, but have limited statistical backing or additional explanations that are omitted. For example, in line 711, the measured CO2 concentrations represent the net remaining between CO2 consumption and uptake/emissions, not specifically the total CO2 Hence, the production of CO2 in the surface waters is likely more readily respired or emitted resulting in lower CO2 concentrations, which is not specifically a reflection of the gross CO2 production rates at depth. Another example in line 781 omits the accumulation of gases in deep water during the stratified period as a direct connection with higher degassing emissions during the warm stratified period. In general, more thorough mechanistic linkages would be warranted while removing much of the text that is not as relevant to the key findings of this study.
Minor comments:
- Line 56 – Disregarding degassing emissions can reduce emissions by a large portion (see, for example, Harrison et al. 2021, https://doi.org/10.1029/2020GB006888). Since you calculate and present degassing emissions, I suggest incorporating this into this rough breakdown of emissions by pathway.
- Line 58 – Consider here (and elsewhere as relevant) to discuss potential CO2 uptake (i.e., CO2 influx into the reservoir rather than emissions) that can be common in productive systems, as well as the variability in uptake versus emissions at diurnal to seasonal scales. This is particularly important given the negative CO2 rates presented in Table 1.
- Line 75 – An eddy covariance flux tower is unlikely to be particularly useful for capturing spatial variability, as it integrates signal across a specific area without distinguishing variability in those signals within that area. Many would need to be set up around a reservoir to really get at spatial variability effectively. Given the scope of this paper with one EC tower set up, I suggest reducing the emphasis on capturing spatial variability via this method.
- Line 146 – While ebullition does generally occur in shallower areas, it is possible to have ebullition occurring deeper than 16 m. I suggest rephrasing this and incorporating a couple of references that show ebullition more dominant – but not exclusive to – shallower areas in reservoirs (i.e., DelSontro et al. 2011, https://dx.doi.org/10.1021/es2005545; Beaulieu et al. 2016, https://doi.org/10.1002/lno.10284).
- Line 243 – Can you add more details on the ANN approach for the bubbling emissions estimates? This is a highly variable pathway over space and time, so additional details here (or in supplementary material) would be useful to describe model fitting and performance.
- Line 277 – Can you elaborate on this “spike removal process” for EC data processing? Given the often pulsed emissions nature of ebullition, it might be expected to find occasional to frequent “spikes” in CH4 emissions as a result. Does this data processing remove those and hence artificially reduce the accuracy of ebullitive emissions variability?
- Line 320 – What specific results is this calculation of SD and SE used for? If used for whole-reservoir upscaled results (i.e., from the 3-7 measurement sites), consider utilizing propagated error based on individual measurement uncertainty and portion of the reservoir each represents.
- Line 338 – How comparable are the EC results from 2009-2011 versus 2019 and 2022, given the EC flux tower was (1) located in a very different part of the reservoir between these time periods, and (2) data span different months/seasons?
- Line 341 – Suggest adding a figure (in the main text or supplementary) showing the differences between daytime and nighttime emissions patterns, as it is a major emphasis in the manuscript.
- Figure 3 – Suggest adding error bars to each estimate to visualize uncertainty (i.e., as in Figure 5), which would pair with explicit statistical tests as suggested above.
- Figure 4 – This figure is quite difficult to fully grasp. Consider reducing the number of panels/variables to include only those that are most different or relevant to the emissions patterns (moving other to supplemental material), and/or adding color/different line types to more clearly distinguish the different years. At its current size, the different symbols used are impossible to distinguish.
Citation: https://doi.org/10.5194/egusphere-2025-3295-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 331 | 44 | 14 | 389 | 8 | 15 |
- HTML: 331
- PDF: 44
- XML: 14
- Total: 389
- BibTeX: 8
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1