egusphere-2025-1253

Breathing Storms: Enhanced Ecosystem Respiration During Storms in a Heterotrophic Headwater Stream

Jativa et al.

Jativa et al. present an elegant study on stream metabolic rates during storm events from continuous data collection in a non-perennial Mediterranean stream. The framing of the story is logical and methods clearly address the narrative throughout the manuscript. Indeed, this contributes to a small but intriguing literature on resistance and resilience of ecosystem function in rivers. I have no large comments but raise a handful of additions to improve the clarity of the methods in the specific comments below and a comment on addressing temporal variability in metabolic patterns in rivers that could be expanded on in the introduction.

L30: ‘regulates’

L50: ‘triggers’ and ‘stoppers’ seem like unnecessary potential jargon. Is there another schema or metaphor that could be used?

L71: I have no disagreement with any of the introduction to this point, but I think the strong temporal variability in GPP and ER need to be emphasized as potential variability to deal with in identifying resistance or resilience. A wide range of recent literature have shown within year and across year variability in GPP that are influenced by river size, hydrologic variability, and light availability (e.g., Savoy et al. 2019; Marzolf et al. 2024). I would also recommend citing Lowman et al. 2024 as an example of identifying recovery of GPP in response to storm events across large scales.

L116: reviewer preference for ‘concentration’ instead of ‘levels’

L122: odd wording. Maybe change to ‘we installed a monitoring station in the stream with upstream area of 9.9 km2’.

L125: what is average depth in this case? In a stilling well or staff gauge? Or is this hydraulic depth of the 200 m upstream reach? Are the pools located in areas that may alter or disrupt advective flow and create longitudinal heterogeneity in DO patterns (Rexroade et al. 2025)?

L129: how was lux converted to PPFD? This is an increasingly common practice in the literature and readers would benefit from specifics on how this was done for use in their own studies.

L150: What value of Q during the storm event was used in calculating RC? Or is it the total water flux during the storm (i.e., the integral of stream flow/total precipitation)? A few more details would be welcome as this is a potentially useful metric for others to use.

L155: this is a great presentation of metabolism data collection and modeling. One addition I would like to see is how mean depth was determined. Mean depth is the average cross-sectional depth of the upstream contributing reach, as is defined in this study as the 200 m upstream of their sensor installation. Mean depth is often the most difficult measure to obtain from a stream reach and across flow conditions but can be estimated in similar ways with rating curves and presumably available with the data collected for the propane injections. I would like to see 1-2 sentences added to this section describing how mean depth was determined. And another sentence on QAQC approaches to the continuous data and how DO.sat was calculated too (basically address how each of the inputs to streamMetabolizer were prepared).

L166: this is a great way of constraining K in the model inputs and a great example for future researchers to approach single-site evaluations. How well does the coverage of propane injections cover the hydroperiod in the stream? These injections are often biased towards lower flows for logistical reasons, but I wonder how well empirical measures were obtained at higher flows? And as you say in L174, getting metabolism estimates during highest flows is difficult or impossible based on data and/or the models failing to converge on days with high flows.

L206: subscript PImax as is in L194

L280: might nit-pick on the ‘biota’ part of the response. Yes, organisms from bacteria to macro-fauna contribute to ecosystem metabolism, particularly ER, but with that statement, I would anticipate some measure of re-colonization of organisms post-storm events, whereas the response variable in this study is integrative ecosystem-scale metabolic function.

Figure 1) should the caption for the orange dot also include ‘ER’?

Figure 4) It maybe my computer screen but it’s difficult to see the non-filled circles against the filled circles. Might recommend a different, contrasting color. Also, purely aesthetic, but can the x-axis be extended to 1000? An additional component that may help the reader discern the relationship with flow: could a vertical line be added where the typical storm flow begins? Or where is the typical baseflow? This would create a part of the graph with baseflow or losing flow metabolism could be easily compared with gaining or stormflow metabolism. If there is not a single or narrow range of flows that separate base from storm flows, disregard this final comment.

Figure 5) Just to be sure, the lines of best fit are coming from the methods text L193-199? What model comparison or evaluation was done to determine linear, logarithmic, or exponential was the ‘best’ fit to the data? Could all the evaluations be compiled into a supplementary table, perhaps with AIC and AICw values?

References

Lowman HE, Shriver RK, Hall RO, et al (2024) Macroscale controls determine the recovery of river ecosystem productivity following flood disturbances. Proceedings of the National Academy of Sciences 121:e2307065121. https://doi.org/10.1073/pnas.2307065121

Marzolf NS, Vlah MJ, Lowman HE, et al (2024) Phenology of gross primary productivity in rivers displays high variability within years but stability across years. Limnology and Oceanography Letters 9:524–531. https://doi.org/10.1002/lol2.10407

Rexroade AT, Wallin MB, Duvert C (2025) Measuring Gas Transfer Velocity in a Steep Tropical Stream: Method Evaluation and Implications for Upscaling. Journal of Geophysical Research: Biogeosciences 130:e2024JG008420. https://doi.org/10.1029/2024JG008420

Savoy P, Appling AP, Heffernan JB, et al (2019) Metabolic rhythms in flowing waters: An approach for classifying river productivity regimes. Limnology and Oceanography. https://doi.org/10.1002/lno.11154

General comments:  The authors investigate how storm events influence stream metabolism, GPP and ER, in a headwater stream, by using high-frequency DO, hydrological, and environmental measurements to analyze 35 storm events, applying Bayesian modeling. A key strength of this study lies in its robust, high-resolution dataset, which allows for a detailed examination of metabolic dynamics. The clear finding is that most analyzed storms (those with Q < 100 L/s) act as "metabolic triggers" significantly stimulating ER and demonstrating a positive relationship between ER stimulation (ΔER) and storm magnitude (ΔQ). The second finding is also very nice information about the quantification of metabolic resilience, particularly the finding that ER recovery time increases with storm magnitude but appears to saturate around 6 days.

Despite these strengths, the manuscript requires major revisions to address the conceptual framework established in the Introduction fully and to enhance the robustness and transparency of its interpretations. Specifically, revisions should focus on (1) evaluating the concept of metabolic saturation introduced in the Introduction section, (2) addressing the implications of excluding high-flow data (Q>100 L/s) for testing the "stopper" hypothesis and the overall representativeness of the findings, and (3-optional) acknowledging uncertainty related to gas exchange estimation during dynamic conditions. 

Specific comments:

Lines 58: introduces an interesting question about River Network Saturation concept. However, the Results section, the authors only focus on the positive linear relationship found between ΔER and ΔQ, and the Discussion does not revisit whether the data showed signs of approaching or reaching this saturation/asymptote. 

Was the ecosystem's processing capacity likely exceeded in the largest analyzed storms, or was the range insufficient to observe this? The authors may explore more Figure 5b, such as whether the observed range of storm magnitudes was likely sufficient or insufficient to induce metabolic saturation in this system. It seems that in Figure 5d, there is a visual saturation, but this is not the concept the authors introduced in the Introduction. Please clearly differentiate the observed saturation in recovery time from the lack of observed saturation in the magnitude of the ER response.

Line 170: All the estimates with Q>100L/s  were excluded due to failed QC checks. I agree that the exclusion of high-flow data (>100 L/s) is based on the reported QC failures, but I am not sure if this action may prevent an empirical test of the "stopper" hypothesis. In the Introduction, lines 60-65, "Finally, during large storm events, […] decreasing mean water residence time, scouring the benthic biomass, […] reduce in-stream processing". These sentences refer to the "stopper" for the  large storm events, but most valid estimates were skipped to check it. Therefore, the inability to assess larger events means the full spectrum proposed in Figure 1 cannot be validated. Here are some suggestions that only use the current dataset:

1)    Report the frequency/duration of flows > 100 L/s to know the unanalyzed portion.

2)    Table S2 does not explicitly link these failures to discharge levels. Please report more details on the QC-Failed Outputs in Supplementary Information to know which QC criteria failed. 

3)    Figure S1 about Q-K600 relationship is very informative, but the highest discharge measured during these injections appears to be only around 32 L/s. Applying the derived Q-K600 relationship via extrapolation beyond the measured range (~32 L/s) during dynamic storm flows (up to 100 L/s) introduces uncertainty. Is it a reason for the model failing at high discharge? I recommend the SI provide a discussion about why the model likely failed QC at high flows in this system while contrasting with successful high-flow modeling in larger systems (e.g., Diamond et al., 2025a, 2025b)

4)    I would like to see output distributions (credible intervals/ranges) for GPP/ER/K600 for all these high-flow runs (Q>100 L/s). Even though the median values failed for QC, using the credible intervals may give us some helpful information, such as the system is more "stoppers" or more "triggers" behavior at these high flows.

--> I suggest that authors may explicitly state the "stopper" hypothesis remains empirically untested by reliable data from this study for higher flows if using credible intervals output still does not give any further information.

Line 174: "did not passed" -> did not pass

Line 155 and 175: Add a brief example of ΔER calculation with negative ER. This is quite confusing when saying to increase or decrease ER while the ER is negative.

Line 270-275: Consistent adding ΔMET/ΔGPP/ΔER definition. Better clarifying axis labels in Figs 5 & 6.

Line 55/88: Clarify "recovery time" vs. River Network Saturation.

Line 352: "ER recovery times at ca. 6 days (Fig. 5a)" -> it should be Fig. 5d

References: 

Diamond, J. S., Truong, A. N., Abril, G., Bertuzzo, E., Chanudet, V., Lamouroux, R., & Moatar, F. (2025a). Inorganic carbon dynamics and their relation to autotrophic community regime shift over three decades in a large, alkaline river. Limnology and Oceanography.

Diamond, J. S., & Bertuzzo, E. (2025b). A coupled O2‐CO2 model for joint estimation of stream metabolism, O‐C stoichiometry, and inorganic carbon fluxes. Journal of Geophysical Research: Biogeosciences, 130(4), e2024JG008401.