the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 May 2026
BlueMinerva: An integrated, autonomous platform for carbon flux and environmental monitoring at the aquatic–atmospheric interface in low-flow waterbodies
Abstract. Water-air fluxes of carbon dioxide (CO2) and methane (CH4) in lakes exhibit substantial spatial heterogeneity, often varying across remarkably fine spatial scales. While manual flux chamber measurements offer high spatial resolution and the potential to capture this variability, their application is typically constrained by labor intensity and logistical limitations. In contrast, eddy covariance (EC) measurements integrate fluxes over a larger footprint, effectively averaging over spatial gradients and complicating a process-based interpretation of the data. Bridging this gap requires methods that combine spatial precision with scalable, continuous monitoring – essential for advancing our mechanistic understanding of lake carbon dynamics.
To facilitate highly resolved biogeochemical measurements, we developed BlueMinerva, an autonomous platform to monitor surface carbon fluxes with the static chamber method, and simultaneously determine physicochemical, biological, bathymetric, and meteorological variables at pre-defined locations. The platform can be programmed to navigate a user-defined track across a waterbody, and collect flux and ancillary data both in transit and at fixed locations for several hours to days, depending on the sensor configuration and related battery requirements.
We deployed the setup at Dagow Lake (Germany) and in a small lake in the Stordalen Mire (Sweden). In total, we obtained 485 chamber-derived flux estimates over 72 measurement hours. We compared CO2 flux estimates between measurements with two different gas analyzers that were simultaneously mounted on the BlueMinerva. The lower-cost sensor (CARBOCAP GMP343, Vaisala) performed equally well as the precise, but costlier sensor (LI-7810, LI-COR) as long as measurements were sufficiently long (around 5 min). Furthermore, we compared measured carbon fluxes with those from an eddy covariance tower at Dagow Lake where CH4 fluxes diverged slightly, possibly linked to the usage of different sensors (closed-path versus open-path), while magnitudes of CO2 fluxes matched well. At both lakes, we identified areas of higher emissions, especially for CH4, and were thus able to resolve the spatial variability of carbon fluxes within the waterbodies. Concurrently, we measured differences in meteorological conditions, and critical limnological variables (water temperature, specific conductivity, pH, dissolved oxygen, chlorophyll, phycocyanin, turbidity, and fluorescent dissolved organic matter) – valuable measurements that enable a comprehensive assessment of environmental drivers behind flux variability.
We conclude that platforms like the BlueMinerva have the potential to be adopted widely by scientists and stakeholders to better capture biogeochemical processes in lakes at high spatial and temporal resolution.
- Preprint
(7149 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2262', Anonymous Referee #1, 14 Aug 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-2262/egusphere-2026-2262-RC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2026-2262-RC1 -
AC1: 'Reply on RC1', Judith Vogt, 16 Oct 2026
We would like to thank the reviewer for their thorough comments. Our responses are indicated below the comments in italics.
General comments
This is a nicely presented instrumentation paper describing an autonomous surface
platform that combines chamber flux measurements with a broad ancillary and bathymetric
payload. The study is the most valuable contribution, it gives a concrete, useful
recommendation to the community.
However, my main concern is that, for AMT, the paper currently reads more as a capability
demonstration than as a measurement-technique characterization. Here are my suggestions.Thanks for this feedback. In the revised manuscript, we will follow the suggestions and include uncertainty analyses for the bathymetry and flux maps. We will also add the inherent uncertainty of the chamber setup and the flux detection limits. We believe these additions firmly place the manuscript within the framework of a measurement-technique characterization as expected by AMT.
Major comments
1. Eq. (3) appears to compute blade-tip circumferential speed rather than axial jet velocity;
please justify that this represents the water velocity, or recompute from thrust via
momentum theory.Thanks for pointing that out. We will revise this section entirely and remove the formulas which were a poor approximation of the complex fluid dynamics of the platform.
2. Wind bursts exceeding 30 m s-1. This value (~108 km h-1) is inconsistent with the campaign conditions described and, as far as I can tell, with the wind-speed distributions in Figs. 6 and C1. Please verify against the raw anemometer record. If these are anemometer spikes or a unit issue, plausible given the 1 m sensor height amid the platform superstructure, I'd recommend removing the claim rather than presenting it as evidence of robustness.
Thanks for catching that. The measured values indeed exceeded 30 m s-1, but represented bad measurements, which were filtered out. We will remove the claim.
3. For a methods paper, each flux from Eq. (5) should carry a propagated uncertainty
(regression slope error, sensor noise, T, p, and chamber geometry). Please also state the
minimum detectable flux for CO2 (both sensors) and CH4 at your 5-min duration and
chamber V/A. You discuss the EC analyzer's detection limit (Sect. 3.2) but never your own
chamber's, which is precisely the number needed to support the claim that the lower-cost
GMP343 is “sufficient.”We would like to make clear that the focus of the manuscript is to present a comprehensive measurement platform suitable for spatially resolved, autonomous surface measurements in lakes in the first place, and does not aim to prove the chamber method or the use of floating chambers. Nonetheless, in the revised manuscript, we will add a propagated uncertainty of our chamber setup using available tools (Cowan et al, 2025; https://doi.org/10.1111/ejss.70104), the minimum detectable fluxes for CO2 and CH4, and the volume-to-area ratio.
4. The chamber is 25 cm tall with ~6 cm submerged, leaving a headspace of ~19 cm. Please
state the numerical volume and area used, and confirm the effective headspace volume (not
the full cylinder) was used, this is a ~30% effect on every reported flux.We can confirm that the effective headspace volume, i.e. a height of 19 cm was used and will clarify in the text.
5. CH4 fluxes are discarded for R2< 0.8 (n = 22), but “all other CO2 fluxes were kept.” Please report the R2 distribution for CO2 and either apply a consistent linearity criterion or justify why one is unnecessary for CO2. Given an opaque chamber and a 5-min enclosure, please also confirm whether linear vs. non-linear fits were tested for the higher CO2 fluxes.
The LI7810 performs well for measurements of CH4, but comparatively worse for CO2. Therefore, it can be justified that the applied criteria differ for each gas. Especially for low CO2 mole fractions, the level of noise increases, leading to lower R2 values for CO2 flux estimates. However, neglecting low CO2 fluxes based on a strict R2 criterion would bias the measurements towards higher CO2 fluxes. We will add a clarification in the text. Linear and exponential fits were tested, but did not yield substantial differences. Therefore, exclusively linear fits were chosen here for simplicity, but it would be up to the user to decide which data processing approach is most suitable in the end.
6. The consequences of using an opaque chamber are not discussed. This is most relevant to the CO2 uptake reported at Inre Harrsjön (Fig. 7C), measured in daylight with a dark
chamber. Please justify the opacity choice and discuss how it affects the representativeness
of the mapped CO2 fluxes.Across studies investigating lake carbon fluxes with floating chambers, opaque chambers are widely used. We acknowledge that opaque chambers exclude incident light and therefore reduce photosynthetic activity. The chamber encloses a volume of water containing organisms (plankton, algae, etc.), and the light availability is determined by the chamber material and the optical properties of the water and the organisms present. Consequently, CO2 uptake may still occur during a chamber closure with an opaque chamber.
We therefore do not interpret the measurements with opaque chambers as a direct measure of dark respiration alone. Rather, they represent the net CO2 flux resulting from the processes occurring within the enclosed water-air system. While chamber opacity may alter the contribution of photosynthesis relative to ambient conditions, the resulting flux can nevertheless provide an estimate of lake-atmosphere CO2 exchange. We will add a discussion in the text. For the BlueMinerva, a transparent chamber would not have been feasible given that other parts of the structure (e.g. the hull of the boat itself) would have cast shadows on the chamber anyways.7. Ebullition was detected once and excluded, yet at Inre Harrsjön ebullition is the dominant
CH4 pathway (your cited Wik et al., 2013: up to 567 nmol m-2 s-1). A 5-min chamber rarely
captures bubble events, so the CH4 maps (Figs. 7-8) represent the diffusive component
only. This limitation should be stated plainly in the abstract and conclusions, not only in
passing.Correct, we did not include ebullition given that we only captured it once. Therefore, the CH4 fluxes refer to diffusive fluxes, and we will clarify this in the manuscript. We would also like to note though, that the setup of the BlueMinerva and the associated data processing could be adapted to quantify ebullition fluxes: since the release of bubbles through ebullition can be considered a stochastic process, the total emissions should be best constrained through a large number of spatially distributed measurements, which is exactly what the BlueMinerva platform makes possible.
8. Within the matched footprint the EC median is -0.27 (uptake) and the chamber median
+0.29 (emission), opposite signs, while you conclude “no substantial systematic
difference” from a non-significant Mann-Whitney (p = 0.52, n = 17 vs. 38). With EC scatter
(SD 1.31) an order of magnitude larger than the chamber (0.11), the test is underpowered;
non-significance does not demonstrate agreement. Please report the sign disagreement in
the medians, temper “CO2 fluxes matched well,” and comment on why the chamber
distribution is so much narrower than EC.We agree that the previous wording overstated the implication of the non-significant Mann-Whitney test. The test does not demonstrate agreement between the two methods, and we will therefore revise the text accordingly. We will explicitly report the opposite signs of the median fluxes and the substantially greater variability observed for EC measurements compared with chamber measurements. Instead of “CO2 fluxes matched well”, we will modify the text to “CO2 flux distributions overlapped substantially”.
The greater variability in EC fluxes may partly reflect uncertainty in the footprint estimates. As discussed for the CH4 fluxes, the heterogeneous surrounding terrain and differences in surface roughness between the lake and forest may have caused the actual EC footprint to differ from the modelled footprint, potentially resulting in a variable contribution of terrestrial signals (predominantly photosynthetic uptake during the day, and respiration during the night) to the EC measurements. This may have contributed to the broader EC CO2 flux distribution compared with the chamber measurements. We will add this in the text.
9. Please clarify whether EC half-hours and chamber points were compared as pooled
distributions over May 13-16 using a maximum-extent footprint, or matched to concurrent
half-hourly footprints. Since fluxes have diel structure and Dagow was sampled day and
night, distributional comparison risks mismatched times of day. If feasible, restrict to
concurrent half-hours.We will clarify that the EC footprint was determined based on Kljun et al. (2004) separately for each half-hourly EC flux estimate. Chamber measurements were then pooled if their locations fell within the corresponding EC footprint area. Thus, the comparison accounted for the spatial extent of the EC measurements, but the chamber and EC observations were not temporally matched to concurrent half-hours since that would have left us with only 2 comparable observations.
We acknowledge that this approach does not account for diel variability in CO2 fluxes. The comparison should therefore be interpreted as an assessment of the distributions of EC and chamber fluxes under broadly comparable spatial conditions over the study period, rather than as a direct comparison of concurrent flux measurements. We will clarify this in the manuscript.10. The two sampling blocks a week apart, with bimodal water temperature, mean the mapped spatial patterns may partly reflect between-block differences rather than true spatial
structure, most relevant to the headline CH4 Moran's I = 0.33. Please test this directly, e.g.
by colouring Fig. 7C/D by measurement day or reporting Moran's I within each block.Thanks for raising this concern. We will add different markers in Figure 7 C and D and report Moran’s I for both blocks. This yields that during measurement days with warmer water, the spatial autocorrelation is significant, while during measurement days with cooler water, no statistically significant spatial autocorrelation is found. This further expands our analysis and shows that spatial variability in lake carbon fluxes is temporally dynamic, and that spatial representativeness of point measurements may depend on environmental conditions and therefore changes over relatively short periods.
11. The eastern shore of Inre Harrsjön was deliberately unsampled yet Fig. 8C/D show filled
fields there, and σ = 30 m at Dagow fills sizeable gaps. Please mask low-data-density cells
or add a sampling-density/uncertainty layer so interpolated fill is not read as measured flux.We will modify our analysis to consider the uncertainty in undersampled areas.
Minor comments
12. Abstract: “CH4 fluxes diverged slightly” contradicts the >3× difference with large effect
size. Please align.We will revise the text.
13. The 5-min vs. 1.5-min comparison is confounded, the 1.5-min data are from one mission
that also maximized travel distance and motor use. A cleaner test with existing data:
truncate the 5-min records to their first 1.5 min and recompute, isolating the duration effect.We will redo the comparison considering truncation of the data.
14. Bland-Altman: the near-zero mean difference at 1.5 min reflects cancellation of a
magnitude-dependent bias; please emphasize the limits of agreement/RMSE instead.We mentioned the RMSEs in the text, but will emphasize them rather than the mean absolute difference between sensors.
15. Please state the calibration/zero-span protocol and any drift checks for both analyzers, and confirm both fluxes are on a consistent dry-mole-fraction basis.
The Vaisala GMP343 was manufacturer-calibrated and the LI7810 zero-spanned in November 2023. We did not check for analyzer drifts. We can confirm that we used dry mole fractions for the LI7810, and realized that for the GMP343, the applied T-RH-P-O2 correction did not yield dry mole fractions yet. Hence, we will convert to dry mole fractions and update the manuscript accordingly.
16. Please state whether an initial mixing deadband is excluded from the dC/dt fit, and discuss chamber-induced disturbance of the near-surface layer (the ~6 cm submersion) as a
possible contributor to the high chamber CH4.We excluded the first 10 s of gas measurements after the chamber was lowered into the water. This was mainly done to account for the offset associated with the tubing length and the travel time the gas needed to arrive at the analyzer. Beyond that, we did not exclude any CO2 or CH4 data, since the first seconds are most representative of the gas exchange rate prior to chamber placement. The ~6 cm submersion of the chamber (as also in Barbosa et al., 2016; https://doi.org/10.1002/lno.10358) was necessary to guarantee no entry of ambient air into the chamber headspace even under windy/choppy conditions. Besides, we can confirm that, upon visual inspection, the initial increase of mole fractions for the filtered measurements was smooth. Therefore, we believe that disturbance by lowering the chamber was not a contributing factor for higher CH4 fluxes than measured by the EC tower at Dagow Lake.
17. Given CEP up to 1.0 m and propeller station-keeping during loitering, please quantify how often propellers fired during active measurements and whether this correlates with flux R2 or magnitude.
The frequency of propeller activation during loitering depends on environmental conditions, particularly wind and water currents. Propeller activation was not recorded during the measurements; therefore, we cannot retrospectively quantify the frequency of propeller activation during individual flux measurements, nor can we directly test its correlation with flux R2 or flux magnitude.
Nevertheless, we consider the potential influence of propeller-induced disturbance on the chamber measurements to be limited. The propeller was positioned at the rear of the platform, whereas the chamber was located at the front, thereby maximizing the distance between the propeller and the chamber and minimizing the direct influence of propeller-generated turbulence on the chamber. In addition, the upper ~4 m of Dagow Lake were well mixed during the measurement period, based on our observations. Therefore, propeller-induced vertical mixing would not be expected to introduce water with substantially different carbon concentrations into the measurement volume. We could therefore not identify evidence that propeller activity systematically affected the measured fluxes.18. Bathymetry RMSE, clarify whether this is a fit residual or a cross-validation error, and
justify the virtual-0.1 m near-shore depth assumption.The reported bathymetry RMSE represented an in-sample interpolation residual. The linear interpolation was fitted using all measured bathymetric points together with the virtual near-shore points, and the RMSE was calculated from the differences between the interpolated surface and the measured depths. However, because the measured points were also used to fit the interpolation, this is not an independent validation error. Therefore, we will revise this approach.
For the lake boundary, we introduced virtual points with a depth of 0.1 m along the mapped shoreline. This value was used as a near-zero boundary condition because water depth is expected to approach zero at the shoreline. The 0.1 m value was therefore not treated as a measured depth, but was used to constrain the interpolation to transition toward shallow water at the lake margin. We will also make that clearer.19. Justify the different k (5 vs. 8) in the Moran's I analysis and show the conclusions are not
sensitive to it.We will justify the choice of k as follows: The number of neighbours (k) was selected based on the spatial configuration of the sampling locations and the connectivity of the resulting spatial weights matrix. A minimum of k = 5 was required for Dagow Lake, whereas k = 8 was used for Inre Harrsjön to ensure a fully connected matrix for both CO2 and CH4.
When performing a sensitivity analysis across k = 3-10, we obtain consistent evidence of positive spatial autocorrelation for CO2 and CH4 at Dagow Lake, and for CH4 at Inre Harrsjön. For CO2 at Inre Harrsjön, positive spatial autocorrelation was detected for k ≥ 6 and remained consistent once the network was fully connected at k ≥ 8.
20. Eq. (5): write R as 8.314 J mol-1 K-1 (Pa m³ mol-1 K-1).
We will add this.
21. Kinematic viscosity is taken at 20 °C while water was 5-16 °C; use a representative value
or note the approximation.We will remove the formula in the revised manuscript.
22. State the flux sign convention (positive = emission) in Sect. 2.6.
We will state the flux sign convention.
23. Label the “~170 per day / >1,000 per week” figures clearly as a theoretical ceiling and
reconcile with the 6-8 h battery/swap constraint.We will make clear that this is a theoretical ceiling considering batteries are continuously swapped.
24. Make the novelty relative to Zug et al. (2025) and Pose et al. (2023) explicit early in the
Introduction.We will do so.
Citation: https://doi.org/10.5194/egusphere-2026-2262-AC1
-
AC1: 'Reply on RC1', Judith Vogt, 16 Oct 2026
-
RC2: 'Comment on egusphere-2026-2262', Anonymous Referee #2, 26 Aug 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-2262/egusphere-2026-2262-RC2-supplement.pdf
-
AC2: 'Reply on RC2', Judith Vogt, 16 Oct 2026
We would like to thank the reviewer for their thorough comments. Our responses are indicated below the comments in italics.
GENERAL COMMENTS
This manuscript presents a novel and well-documented autonomous platform for high-
resolution monitoring of aquatic carbon fluxes and environmental variables and it fits well
within the scope of biogeoscience/environmental instrumentation journals. The study
represents a valuable technological contribution that fits well within the scope of AMT. The
platform design, field demonstrations, sensor intercomparison, and open-access
documentation are major strengths. However, before publication, the manuscript would
benefit from a more thorough discussion of flux uncertainty, stronger justification for the use
of linear chamber flux calculations, a more cautious interpretation of comparisons with eddy
covariance observations, and additional clarification regarding the robustness of spatial
interpolation procedures. These concerns do not undermine the overall quality and novelty of
the work but should be addressed prior to acceptance.We thank the reviewer for this feedback. In the revised manuscript, we will add discussion about uncertainties and accordingly revise our interpretation and interpolation approaches further detailed within our individual responses.
Overall assessment
The BlueMinerva platform fills an important gap between point-scale chamber measurements and ecosystem-scale eddy covariance observations. The combination of autonomous navigation, chamber flux measurements, water-quality monitoring, meteorological sensing, and bathymetry in a single platform is innovative. The authors provide unusually detailed information on hardware, software, communication architecture, power management, payload, and operational constraints. This greatly enhances reproducibility and community uptake. The evaluation in two contrasting environments (a temperate German lake and a sub-Arctic Swedish lake) strengthens the manuscript and the dataset is reasonably extensive (485 flux measurements over 72 h).
Main concerns (Major revisions)
1. Insufficient validation of flux accuracy
The paper validates agreement between two gas analyzers and compares measurements with eddy covariance, but it does not fully validate the accuracy of flux estimates themselves.
The authors essentially compare:
• sensor A vs sensor B,
• chamber measurements vs EC measurements.
However, neither comparison establishes whether the measured fluxes are correct.
The manuscript provides a convincing evaluation of instrument consistency but less evidence regarding the absolute accuracy of the chamber-derived fluxes. I encourage the authors to discuss in more detail possible biases associated with chamber deployment, chamber-induced turbulence, pressure perturbations, and non-linearity of concentration
accumulation. A quantitative uncertainty analysis of flux estimates would substantially
strengthen the study.We thank the reviewer for raising this concern. Within the presented study, we aim to present a new measurement platform that is able to conduct spatially resolved measurements autonomously. The goal is not to show which sensor or method measures correct fluxes since this would require an environmentally controlled lab experiment which we did not conduct. Besides, both sensors come with their own advantages and disadvantages, and based on the setup and goal of a potential user, it may be more important to have a lower-cost sensor over a higher-cost sensor, a lighter-weight over a heavier sensor, or a sensor that measures CH4 or not. Therefore, we merely present and compare these options.
The comparison of chamber versus EC measurements has been repeatedly conducted on a similar basis. Again, both methods come with their own caveats, and the user ultimately needs to decide which method fulfills their goal. In this study, we show that under the given conditions, the flux estimates differed between the two methods. It also shows that the footprint of the eddy covariance tower only covers a small section of the lake and likely neglects sections with potentially different flux signals that can be captured when using the BlueMinerva.
2. Exclusive use of linear flux estimation
The authors calculate all fluxes through linear fits (Eq. 5). This is potentially problematic
because chamber accumulation often deviates from linearity. Usually also exponential fit is
used. The manuscript would benefit from a justification for the exclusive use of linear
regressions for flux estimation. Why were non-linear chamber models (e.g., Hutchinson-
Mosier type approaches or alternative accumulation models) not considered? Please discuss
potential biases resulting from assuming linear accumulation during the 5-minute enclosure
period.We can confirm that the vast majority of flux estimates was based on linearly increasing mole fractions. Rather than mixing model approaches depending on whether a linear, exponential, or other fit would be most suited, we decided to stick to one approach – the linear fit – throughout for simplicity and consistency. This could be adapted by the user, if need be, but we do not consider this to be a crucial point of discussion for this study since we do not aim to compare among different model approaches.
3. Eddy covariance comparison is based on a small sample size
The EC comparison uses only:
• 17 chamber observations,
• 29-38 EC observations after footprint filtering
The conclusions are somewhat stronger than warranted. The comparison with eddy
covariance measurements is useful but based on a limited number of overlapping
observations (TABLE 3). The authors should moderate statements regarding agreement
between methods and explicitly acknowledge the limited statistical power of this comparison
because a non-significance difference does not demonstrate the agreement. In my opinion a
scatter plot between the two measurements, with RMSE and determination coefficient is a
better way to compare two data sets.We agree that the non-significance difference did not demonstrate the agreement between chamber and EC measurements. We will address this as mentioned in response to comment 8 of reviewer 1. A scatterplot is a nice suggestion, but is unfortunately not feasible for this test, since we present the pooled data from within the same footprint area during the test period for each method, which means they are not paired. If we pair the measurements, we are left with only two half-hourly data points, and a comparison is not feasible with such a small number of measurements.
4. Potential temporal confounding at Inre Harrsjön
The authors correctly note that measurements were acquired in two different time period (one week apart).Because measurements at Inre Harrsjön were conducted on different dates with substantially different environmental conditions, some of the observed spatial patterns may partly reflect temporal variability. Please discuss this limitation more prominently when
interpreting hotspot distributions and smoothed flux maps.As stated in the response to comment 10 by reviewer 1, we will visualize the different measurement periods (first half was colder, second half was warmer) in Figure 7. Besides, we will split up the analysis of Moran’s I and the smoothed flux maps into the two measurement periods at Inre Harrsjön.
5. Interpolated flux maps may overstate spatial patterns
The generated maps (Fig. 8) are visually attractive but could give the impression of greater
certainty than justified.
No sensitivity analysis is presented for:
• grid size,
• kernel bandwidth (σ),
• measurement density.
The interpolation approach requires further justification. Please provide a sensitivity analysis
demonstrating whether the identified hotspots remain robust under different kernel
bandwidths and interpolation parameters.We will modify our analysis and include uncertainty maps to clarify limitations of the flux maps.
Minor comments
Methods
• Explain why k = 5 and k = 8 were selected for Moran's I.We will do that. Please see our response to comment 19 of reviewer 1.
• Clarify how measurement duration of 5 min was chosen initially.
We chose a measurement duration of 5 min based on rigorous testing under field conditions of the setup. 5 min were chosen to yield reasonably good flux estimates for the GMP343, which is considerably noisier than the LI7810.
• Report uncertainty on bathymetric interpolation in the main text, not only as RMSE
values.We will revise our analysis to better describe the uncertainty of the interpolation and add explanation in the text.
Results
• Include confidence intervals around median fluxes.We will include confidence intervals in the revised version where applicable.
• Add effect sizes consistently when discussing all statistical comparisons.
We will revise the manuscript accordingly.
Figures
• Figure 8 is attractive but uncertainty is not shown.We will add uncertainty maps in the appendix.
• Consider adding observation density layers beneath interpolated maps.
We will clarify the observation density in Figure 8.
• Several figures are information-dense and fonts appear small.
We will increase the fonts where needed.
Citation: https://doi.org/10.5194/egusphere-2026-2262-AC2
-
AC2: 'Reply on RC2', Judith Vogt, 16 Oct 2026
Data sets
Carbon flux, bathymetry and vertical profile data collected with the BlueMinerva at Dagow Lake, Germany Judith Vogt, Tarek S. El-Madany, Elliot Pratt, Mathias Göckede https://doi.org/10.5281/zenodo.19365068
BlueMinerva: 3D view, datalogger program, wiring diagram, and mission file example Judith Vogt, Tarek S. El-Madany, István Héjja, Christian Burgold, Abdullah Bolek, Elliot Pratt, Mathias Göckede https://doi.org/10.5281/zenodo.19136563
Video abstract
BlueMinerva 3D View Judith Vogt, Tarek S. El-Madany, István Héjja, Christian Burgold, Abdullah Bolek, Elliot Pratt, Mathias Göckede https://doi.org/10.5281/zenodo.19136563
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 323 | 159 | 65 | 547 | 53 | 49 |
- HTML: 323
- PDF: 159
- XML: 65
- Total: 547
- BibTeX: 53
- EndNote: 49
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
