the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Mar 2026
Coordinated methane flux measurements from northern lakes by the SITES Water program – open data and learning examples
Abstract. Lakes represent one of the main sources of methane (CH4) to the atmosphere, contributing roughly 3–25 % of the total global yearly emissions. While the scientific interest in understanding and modelling these emissions is increasing rapidly, methodologically consistent long-term flux measurement programs of integrated lake CH4 emissions are largely missing. Here we present results from a systematic and comparable spatiotemporal multi-lake CH4 flux program initiated by the Swedish Infrastructure for Ecosystem Science (SITES). Five lakes distributed across Sweden covering a latitudinal gradient from 68° N to 57° N, including the arctic subalpine, boreal and north temperate zones, were monitored during 2016–2022 in ways that captured variability in space and time within lakes and that allowed between-lake comparisons. The data includes 2375 unique CH4 flux measurements (incl. total CH4 fluxes, diffusive fluxes and surface water concentrations) along with other common physical and chemical lake variables. This paper presents these data (Swedish Infrastructure for Ecosystem Science, 2026, https://doi.org/10.23700/05ax-st65), underlying methodological priorities and procedures, and provides examples of what can be learned from such a coordinated sampling program. Briefly, the data show the need for appropriate consideration of ebullition and space-time integration to properly represent whole lake CH4 fluxes. The data further show that both diffusive and total CH4 fluxes have an apparent temperature dependency, which can be modulated by water depth in ways that differ among lakes. Also, yearly open water CH4 flux estimates were consistently shaped by temperature in all lakes, but the strength of this response differed between lakes; increases of whole season mean water temperature between 0.5 and 3.2 °C corresponded to increases in CH4 fluxes ranging from 16 to 74 %. Finally, we propose that temperature-normalized CH4 fluxes should be used in lake emission inter-comparisons.
- Preprint
(2213 KB) - Metadata XML
-
Supplement
(79 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2026-141', Anonymous Referee #1, 05 May 2026
-
RC2: 'Comment on essd-2026-141', Anonymous Referee #2, 02 Jun 2026
This paper provides valuable data and methodology for comparative research on CH4 emission processes and mechanisms in high-latitude lakes, as well as in global lakes. There are some suggestions and comments for the authors' consideration as follows:
- Line 64-65 It is true that diffusive flux is dynamic and continuous. The diffusive flux can show high temporal resolution variations (e.g., diurnal variation). Consequently, if measurements lack sufficient temporal coverage, such variations may be overlooked.
- In Table 1, Can authors provide the information about freeze thaw period or open water period?
- Part 2.3, Eq1. only can be used to calculate diffuse flux. F in Eq.q should be Fdiff.
- Figure 2 left column show the all measurements in several years. Dose the right column show only one year in different lakes? If they show only one year, please provide the year in figure 2 right column.
Citation: https://doi.org/10.5194/essd-2026-141-RC2 -
RC3: 'Comment on essd-2026-141', Anonymous Referee #3, 20 Jul 2026
This manuscript describes a genuinely valuable dataset with 2375 chamber-based CH4 flux measurements from five Swedish lakes spanning 68°N to 57°N, collected over 2016–2022 with a deliberately standardized design. Methodologically consistent multi-lake, multi-year flux programs are rare, and the value of this one will only grow as the time series lengthens. The dataset is openly archived and clearly falls within ESSD’s scope.
My reservations concern the balance and the rigor of the two halves of the paper. As a data paper, the dataset documentation is thinner than ESSD normally expects — there is no variable/file description, no QA/QC section, and no analytical performance metrics, with readers instead referred to an external documentation file. As an analysis paper, the temperature-extrapolation exercise (Sect. 2.4, 3.3) carries more weight in the abstract and conclusions than its statistical treatment can support, and several of its assumptions are either untested or, in a few places, internally inconsistent. I would recommend the authors either strengthen the extrapolation substantially or present it more explicitly as an illustrative example rather than a headline result.
Below I have suggestions and comments. None of these issues are fatal, but several need to be addressed before publication.
The diffusive/ebullitive partitioning rests on an unvalidated threshold (L163–169). The assignment of chambers to “diffusion-dominated” using the minimum apparent k and a 2-fold cutoff is the single most consequential methodological choice in the paper — it determines the reported 62% (ALM) vs 19–36% (other lakes) diffusive fractions and, downstream, every Fdiff result. Yet the threshold is inherited from earlier work without testing here, and the qualifier “provided that k was in a realistic range given the wind speed” is left undefined: what range, based on which wind data, and how many chamber-deployments were rejected on this basis? Please also state whether the minimum k is identified per lake-deployment (across all 12 chambers) or per depth-zone. Given the very different wind exposure of ERK versus the small forest lakes, a sensitivity analysis using alternative cutoffs (e.g. 1.5×, 2×, 3×) would considerably strengthen confidence in the partitioned data product.
The Monte Carlo treatment of uncertainty is not defensible as implemented (L213–221, L334–337, Table A3). The confidence intervals in Table A3 appear to be intervals on the means of 1000 draws, not the 2.5–97.5 percentiles of the simulated flux distribution. For ALM 2016 the SD is 0.0012 while the CI spans 0.0032–0.0033. Readers need the predictive interval; as presented, the CIs give a misleading impression of precision. Consequently, the among-year ANOVA and the significance letters in Fig. 6 are not interpretable. With N = 1000 arbitrarily chosen simulation draws, p-values can be driven to any desired level by increasing N. These are simulated replicates, not independent observations.
The “open water period” definition produces implausible integration windows (L216–218, Table A3). Defining the open-water season as all days with modelled Tw above the lowest observed Tw gives ERS 266–361 days per year — essentially the entire year in 2020. The threshold may include winter/ice-covered periods unless checked against ice records, and the air-temperature regression (Eq. 3) has no mechanism to represent ice cover. Fluxes are then extrapolated across ~300 days from roughly a dozen summer-weighted campaigns. Please either impose an ice-phenology constraint (do SITES stations have ice records?) or restrict integration to the observed sampling window, and report how sensitive the annual totals are to this choice.
Annual fluxes are reported for years in which some lakes were not sampled. Table 2 gives 4 sampling years for ALM, 3 for FER, 5 for ERK — but Table A3 and Fig. 6 present 2016–2022 estimates for every lake. For FER in particular, four of the seven annual estimates appear to rest on no observations from that lake in those years, and Fig. 2 suggests ERK was not sampled early in the period either. Since the 2018 warm-summer signal is a headline finding, it matters whether 2018 was actually observed at each lake. Please state the sampled years explicitly per lake and mark purely extrapolated years in Fig. 6.
Internal inconsistency in the ALM temperature parameters (Table A2). For all three ALM depth-zones, Ftot20 and Fdiff20 are identical (0.10), with θdiff> θtot in every case. Taken literally, this parameterization implies that ebullition falls to zero at 20°C and becomes negative above it, which is mechanistically backwards and inconsistent with the reported 38% ebullition contribution in ALM. Please check whether this reflects rounding, a lower bound imposed during fitting (the recurrence of exactly 0.10 across eight entries is striking), or a genuine fitting problem and constrain the fits so that Ftot ≥ Fdiff across the applied temperature range.
Dataset documentation is insufficient for an ESSD paper (Sect. 2.2, 2.3, 5). The paper should be self-contained enough that a user can work with the archive without reconstructing the protocol from a separate file. Currently missing: a table of file structure, variable names, units and flags; GC calibration standards, analytical precision and detection limits; blank handling; the effect of vial storage duration on measured concentrations; how failed or lost chambers were treated. On the last point, ERS reports 887 fluxes from 82 campaigns (~10.8 per campaign against a nominal 12), implying routine losses that are nowhere described. A short QA/QC subsection plus a variable table would address most of this.
No code availability statement. Processing spanned Python (curve_fit), a supplemental spreadsheet, and GraphPad Prism. Given that the flux calculations, the k-threshold partitioning and the Monte Carlo routine all involve non-obvious choices, depositing the processing code alongside the data would substantially raise the reuse value of the dataset and is consistent with ESSD practice.
Minor comments
L20, L33–34 (Abstract) and L343–344: The claim that “increases of whole season mean water temperature between 0.5 and 3.2°C corresponded to increases in CH4 fluxes ranging from 16 to 74%” pairs two quantities from Table A3 that are computed over different year pairs. Δmax Tw is the range in mean Tw (e.g. ERK 2018 vs 2020), while Δmax Ftot is the ratio of maximum to minimum flux (ERK 2018 vs 2017). The two should not be presented as a cause-and-effect pairing.
Table 1: TN is used in Fig. A3 but is absent from Table 1. Please add it. Mean annual air temperature and typical ice-cover duration per site would also be useful.
Table A3 caption: Related, the caption defines Δmax Ftot as “the change in emission from the year with the lowest Tw to the year with the highest Tw expressed as percent,” but the tabulated values (121, 138, 174, 120, 116) are max/min flux ratios × 100. Either relabel as a ratio or recompute as a percentage change. Also: “lower and upper bond” –> “bound.”
L96: “measurements of to seven years” –> “of up to seven years.”
L109–110 vs Table 1: Asa is given as 57.18°N in the text and 57.17°N in the table.
L134–139: With only two sampling points (t = 0 and 24–48 h), the non-linear flux model cannot be verified against the observed accumulation curve. Please note this limitation, and clarify whether fluxes are normalized to the actual deployment duration, which appears to vary by a factor of two.
L193–194: “Hydroplogical” –> “Hydrological.”
L196-200: The text states that regressions were developed “between the observed surface water CH4 concentrations and three MESAN temperatures,” but Eq. 3 predicts Tw. This appears to be an error — please correct to “observed surface water temperatures.”
L232–233 and L259–261: Several of the quoted fold-ranges don’t match Table 2. Ftot varies ~2100-fold in ERK (not “100- to 1000-fold”); Fdiff varies ~30-fold in STO (below the stated “50- to 200-fold”). Please recheck.
L252: “becomes important is both measurement design and data processing” –> “in both.”
L262–266 (Fig. 3): The Ftot regression has a negative intercept (−0.24 mmol m-2 d-1), which is physically impossible. Given the strong heteroscedasticity evident in the panel, consider a log-log fit or a zero-intercept model, and note that the reported R2 depends on this choice.
L286: “Natichimuthu” –> “Natchimuthu.”
L293 vs L346: “Bastviken and Johnson 2025” vs “Bastviken and Johnsson 2025” — inconsistent.
L318–323 (Fig. 5 caption): The caption states that significant differences were found “except for F20 for diffusive flux and for Caq”, but the figure shows p = 0.015 and p = 0.010 for precisely those two panels, with “ns” on the other three. The caption reverses the result and contradicts the text at L313–314.
L316–317: “provides support for a lateral transport of dissolved CH4 from shores towards central parts of lakes" overstates what a near-shore-to-offshore concentration gradient can show; the same gradient is equally consistent with local shallow-sediment sources plus bubble dissolution. Please soften.
L365–367: “annual whole lake Ftot” is misleading given that ice-cover, ice-out and overturn fluxes are excluded (L218–219). Please use “open-water season” consistently, and note that the omission of ice-out likely makes these lower bounds on true annual emission.
L409: "conduced" –> "conducted."
L559 (Table A2 caption): “F20 and θ in Eq 1” should refer to Eq. 2.
Figs. A2–A5 captions: "blown crosses" –> "brown crosses"
Title/scope: The program is described as a Greenhouse Gas Flux Program but only CH4 is presented. If CO2 data exist in the archive or are forthcoming, a sentence in Sect. 5 would help users.
Citation: https://doi.org/10.5194/essd-2026-141-RC3
Data sets
SITES Water Layer 6, Greenhouse Gas Flux Program - Lake methane flux Data Collection Swedish Infrastructure for Ecosystem Science https://doi.org/10.23700/05ax-st65
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 331 | 140 | 24 | 495 | 25 | 21 | 25 |
- HTML: 331
- PDF: 140
- XML: 24
- Total: 495
- Supplement: 25
- BibTeX: 21
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript makes a valuable contribution to understanding methane emissions in inland waters. The authors describe a coordinated, multi-lake methane flux dataset collected across a significant latitudinal gradient in Sweden, which helps address an existing gap in consistent, long-term, and comparable lake methane observations. The work is timely, given the growing emphasis on aquatic methane emissions and their climate feedbacks, even though the study does not yield particularly novel conclusions. The manuscript is clear and concise, and it effectively sets out the scientific rationale, methodological novelty, and key findings. I have some minor comments that I hope will help improve this manuscript.
Here are my suggestions: