RC1: 'Comment on bg-2023-161', Anonymous Referee #1, 10 Nov 2023 

The paper by Tung et al. presents a very valuable dataset of observations and modeling results of the benthic ecosystem of a submarine canyon and the continental slope of Taiwan/ South China Sea where such as study has not been performed before. This research therefore is of high interest to the broad marine community, after several major concerns have been addressed.

Author response: Thanks for the positive comments.

Major concerns:

1) The structure of the linear inverse model lacks mortality as part of physiological constraints. At the moment, no metazoan organism nor prokaryote deceases at the moment. The model compensates this by increased fluxes to the sediment which include both, feces and mortality, but it would be more accurate to include actual equations for mortality and properly constrain them.

Author response: Thanks for the suggestion. We will add carbon flows constrained by the mortality upper and low bounds for metazoan (to detritus, Hendriks,1999; Tenore, 1982, and van Oevelen et al., 2012) and prokaryote (to DOC through viral lysis, Middelboe and Glud, 2006; Danovaro et al., 2008) to estimate the carbon flows to mortality. The final model outputs (in Figure 4) will be shown as the net fluxes to or from the detritus or DOC pool for simplicity.

2) The authors lacked parts of the microbial loop. Like mentioned above, the bacteria do not 'die' which would be caused by virus-induced lysis and it is highly unlikely/ impossible, that they directly feed on sediment. Instead an degradation step of detritus to DOC which then is taken up by bacteria should be included. Dying bacteria also contribute to the DOC pool. Implementing the microbial loop (check papers about LIM in the deep sea by e.g. van Oevelen, Dunlop, Durden, Stratmann, etc.) will not only improve the model quality/ present the natural environment more appropriately, it also allows to compare modeling results of this study with other LIM studies.

Author response: Thanks for the suggestion. Since we do not have data on DOC stock, we will use the equation that calculates prokaryotic growth efficiency (van Oevelen et al., 2011) to back-calculate the DOC stock. In addition, we will rename the sediment stock as detritus stock and add detritus degradation to DOC stock in Figure 4. We will implement the microbial loop as bacteria take up the DOC stock and return organic carbon to DOC stock through viral lysis.

3) Why did the authors pool all metazoans in two size classes without splitting them into larger groups or feeding types? This should either be done or be discussed as a major limitation of study in the corresponding section of the discussion. At the moment, it is very much focused on the sampling and the report of the limitations of the study is underdeveloped.

Author response: Thanks for the suggestion; however, we do not have the sufficient taxonomic resolution to split the data into functional groups. Consequently, we cannot constrain the metazoan feeding preferences by their functional groups, and thus, the associated carbon flows are solely estimated by mass balancing. Lumping the metazoan stocks by size groups oversimplified the model. It may introduce bias (e.g., misrepresentation of the relative contribution of deposit-feeding from sediment and predation on the smaller size class). We will further acknowledge this issue as a study limitation.

4) Going in the same direction: Why did the authors use only one detritus pool instead of splitting it up in several pools like presented in studies by van Oevelen, Dunlop, Durden, or Stratmann? This should be discussed in the model limitation section.

Author response: Thanks for the suggestion. Since we do not have the data on different components of detrital organic carbon (i.e., labile, semi-labile, refractory), we followed other benthic food web model studies (e.g., Rowe et al. 2008) and aggregated all detritus into a singular compartment. The degradation of organic matter occurs across an extensive spectrum of time scales, ranging from minutes for biochemical breakdown in animal guts to 10⁶ years for organic carbon mineralization in deep-sea sediments (Middelburg et al., 1993). This vast dynamic range implies that the characteristic time scale of experimental observations significantly influences the measurable degradation rates (Hedges and Keil, 1995). Our current model for quantifying carbon oxidation relies on oxygen consumption, but it does not enable the direct measurement of the degradation of semi-labile and refractory organic carbon due to their extended time scales. Moreover, Middelburg and Meysman (2007) proposed three primary factors governing organic matter degradation in sediments: the chemistry of the organic matter itself, the physical characteristics of the sediment environment, and the biological agents responsible for breaking down organic matter. Unfortunately, the lack of directly measured degradation rates in the Gaoping area hinders our ability to partition detritus into distinct pools, as observed in previous studies. Nevertheless, we will acknowledge this issue as a study limitation.

5) The authors should also discuss how the exclusion of megabenthos affects the results of the study, especially because van Oevelen et al. (2011) showed the importance of megabenthos at parts of the Nazare canyon.

Author response: Thanks for the suggestion. We will acknowledge the limitation of not having megafauna in our carbon food web. Currently, the megafauna predation is only constrained by the net growth efficiency of meiofauna and macrofauna, not by the available megafauna stocks; therefore, we assume all the meiofauna and macrofauna growths are consumed by megafauna. This is because we did not sample megafauna, and no prior megafauna data is available in the GPSC due to the high risk of losing bottom trawl in the treacherous terrain. van Oevelen (2011) also shows that megafauna deposit feeding dominated the mid-section of Nazare Canyon, which is not considered in our current model. Consequently, in our model, the sediment detrital carbon consumed by deposit-feeding megafauna would be channeled to export, burial, or metazoan, not to megafauna predation. 

-> Summaryzing these comments, I believe that the authors should re-develop the linear inverse models for the two stations to implement my concerns 1 and 2 in the model. The other aspects should preferrably also be included in the new versions of the model, but if that is not possible, they have to be elaborated in the discussion section.

 Author response: Thanks for the suggestion. We will implement suggestions 1 and 2 in the models and study limitations 3, 4, and 5 in the discussion section.

Minor/ Technical comments:

l 80: Avoid abbreviations in headlines if not strictly necessary. Hence, please write the full name of the canyons here.

Author response: Thanks for the suggestion. We will revise the headlines accordingly.

l 111: Here and in other parts of the manuscript, the authors refer to Fig. B1. I think they refer to Fig. 4, but I'm not sure. Either way, they should correct this and refer to the correct figure.

Author response: Here, we indeed refer to Fig. B1. We intend to use Fig. B1 to explain the food web structure with a table of figure legends describing the definition of each flow.

l 112: "export processes" What export processes do the authors have in mind here? Please explain in a short half sentence.

Author response: We refer to “export processes” as the organic carbon leaving the system. We will add a few words in the sentence to better explain.

l 145: Which "taxonomic group" did the authors include here? Please be more specific (which taxonomic rank, or list the taxonomic groups).

Author response: They comprise 20 and 15 classes or orders of macrofauna and meiofauna, respectively. Since including the list in the taxa may read tedious, we will add the taxonomic ranks in the sentence.

Figures: I don't know how it will look like in the published manuscript, but at the moment the letter size of the figures is relative/ very small. The authors should increase the letter size in the figures a bit.

Author response: Thanks for the suggestion. We will increase the letter size in the figure.

Citation: https://doi.org/10.5194/bg-2023-161-RC1

RC2: 'Comment on bg-2023-161', Anonymous Referee #2, 01 Dec 2023 

Tung et al., “Contrasting carbon cycling in the benthic food webs between river-fed, high-energy canyon and upper continental slope.”

Tung et al. present a nice study investigating organic carbon cycling across the sediment-water interface in Gaoping Submarine Canyon off Southwest Taiwan. This study is notable in how they combined sampling of the benthic food web over multiple seasons with a linear inverse model to investigate the fate of organic carbon and the role of benthic communities in modulating carbon flows in two distinct sites, representing a high energy, high disturbance site with low faunal biomass vs a more stable site with higher faunal biomass. The paper is overall well written, with a nice discussion, particularly on limitations of the LIM model in estimating the food web consumption of oxygen.

Author response: Thanks for the encouraging words.

I have two main comments:

1) Since the wet and dry seasons are so different, and the study only sampled once during the summer, what would the results of the LIM be if you separated out the seasons? E.g., one model result for spring, and one for autumn? Or just dry season only (grouping together spring and autumn and excluding summer)? I’m not sure it makes sense to include the summer data for an “annual average,” as the data become quite skewed towards the dry season. Further, the TOU and DOU measurements are only available for the spring and autumn, and thus wouldn’t be representative of an annual average anyway.

Author response: Thanks for the suggestion. We will re-run the models using the average of dry season-only data. We think this is better justified as we did not find statistical variation between the autumn and spring stock and oxygen consumption.

2) I personally found the network characteristics hard to understand, as someone who works on food web dynamics but not on deep sea food webs. There was no definitions or equations describing the main terms used, and so the whole set of analyses on the food web network was lost to me. If possible, please include in the supplemental a set of definitions and/or equations describing the terms that are used. It would just help broaden the readership of this manuscript.

 Author response: Thanks for the suggestion. We will include a more detailed description and equations of each network index. If the new addition exceeds the text limit of Biogeoscience, we will opt to provide them in the supplemental information.

Minor comments:

Fig 4. I found the overlapping numbers and arrows, particularly in the POC -> Sediment arrow hard to read. Also, would it be possible to include in the table or the supplement the fraction (or percentage) of flows going to each sink? That would help readers be able to more easily interpret how the ecosystem functions differently in the two sites.

Author response: Thanks for the suggestion. We will revise the figure accordingly. We will also expand Table A6 to include the relative contribution of each flow to major sinks in the food web.

Section 3.7, line 368. Total system throughflow is marked as T.. and total system throughput is marked as TST. I believe this should be flipped.

 Author response: Thanks for the correction. These are typos. We will correct them accordingly.

Citation: https://doi.org/10.5194/bg-2023-161-RC2