the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Sep 2023
Parameterization and tuning of the Bay of Biscay Atlantis model v1
Abstract. This paper describes the parameterization and calibration of an end-to-end Atlantis model for the Bay of Biscay, characterising spatially the structure and functioning of the ecosystem. The Bay of Biscay is considered rich in terms of ecological diversity and different oceanographic events such as coastal upwelling, coastal run-off and river plumes, and seasonal currents, take place in the area. These features, in addition to the different pressures caused by human activities and management criteria, demand for concurrent modelling of all the characteristics of the Bay of Biscay ecosystem in order to improve our understanding of the system and its functioning. The modelled area is 145 970 km2 and was divided into 36 spatial polygons, each with multiple vertical layers. The model was composed by 54 functional groups, ranging from primary producers to top predators. Our results highlighted the importance of lower trophic levels to the pelagic system and how the trophic interactions among phytoplankton and zooplankton groups impact the structure of the ecosystem. The results also demonstrate the importance of having accurate and precise data for biological processes and showed the need of further study in the age-specific data such as biomass and weight distribution per age and diet interactions between juvenile and adult fish stages. Overall, the Bay of Biscay Atlantis model has been shown to be a tool that has the potential to improve our understanding of the spatial functioning of the Bay of Biscay ecosystem that will help establishing management measures of human activities.
- Preprint
(2797 KB) - Metadata XML
-
Supplement
(4946 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CEC1: 'Comment on egusphere-2023-1368', Juan Antonio Añel, 11 Oct 2023
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.htmlFirst, independently, if you do have or do not have the right to publish the code you use in your work, it is necessary that you archive it permanently. Therefore, please, upload the Atlantis code to a Zenodo private repository, and reply to this comment with the DOI and link to access it.
Also, in your manuscript, you state, "This information is required to filter out fake registrations (bots), and to plan future model development and user support. For further details see https ://github.com/ Atlantis‐Ecosystem‐Model/
Atlantis_example_and_instructions." The statement on the need for information is irrelevant here and even misleading, as the reasons that you mention are debatable. Therefore, in potential future reviewed versions of your manuscript, please, remove it. Regarding the "further details", you have stored them in GitHub. First, the link is not working, so it is not possible to access such information. Second, GitHub is not a suitable repository for scientific publication, our policy makes it clear. Moreover, GitHub itself instructs authors to use other alternatives for long-term archival and publishing, such as Zenodo. Therefore, please, publish your "further details" in one of the appropriate repositories, and reply to this comment with the relevant information (link and DOI) as soon as possible, as it should be available before the Discussions stage.Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2023-1368-CEC1 -
AC1: 'Reply on CEC1', Ane Lopez de Gamiz Zearra, 01 Nov 2023
Dear Juan,
Sorry for our late reply to your comment. The model developer (affiliated with CSIRO, Australia) suggests that the reviewers registert, and that way can get access to the code. The registration is necessary to protect government IP to meet export laws in Australia, and none of the authors of the manuscript will have access to the overview of registered users. If you accept this, we'll create a repositoty to share the remaining code and files used in the development of the manuscript.
Ane
Citation: https://doi.org/10.5194/egusphere-2023-1368-AC1 -
CEC2: 'Reply on AC1', Juan Antonio Añel, 01 Nov 2023
Dear authors,
First, thanks for your reply.
While we can understand that some researchers are not allowed to share the code because of legal reasons imposed by their employers, it is not clear to me what law or regulation forbids you or the developers from storing the model in a long-term repository such as the ones offered by Zenodo (private or not). Therefore, we need more detailed information on the regulations that forbid the authors to share it to judge if we can accept an exception in this case.
I understand that in this work, all the authors are simply users, and none of you is involved in the development of the model used. If this is the case, this would be an extra possibility to clear your manuscript in compliance with our Code and Data Policy, as you would not have a way to influence the decision on the license or sharing of the model. However, we need clarification on this too.
Regards,
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2023-1368-CEC2 -
AC2: 'Reply on CEC2', Ane Lopez de Gamiz Zearra, 03 Nov 2023
Dear Juan,
The Atlantis code was developed by employees of the Australian government and so automatically falls under both Australian export laws and Crown Copyright. This means that while the code can be distributed royalty free from the CSIRO repository it cannot be posted to a public repository long or short term. Anyone accessing the code has to be registered to comply with this ruling by CSIRO’s legal department. The developers are very happy to register applicants or to provide the manual to any interested users - that is able to be shared widely and is already publcially accessible with full documentation of equations used.
We, the authors, are simply users of the model, so none of us is involved in the development or/and improvement of the model used.
Ane
Citation: https://doi.org/10.5194/egusphere-2023-1368-AC2
-
AC2: 'Reply on CEC2', Ane Lopez de Gamiz Zearra, 03 Nov 2023
-
CEC3: 'Reply on AC1', Juan Antonio Añel, 01 Nov 2023
Dear authors,
A clarification on the other comment that I have posted replying to you. In the first paragraph, when I refer to "the authors", I mean the authors of the model, not the authors of the manuscript, as in the second paragraph.
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/egusphere-2023-1368-CEC3
-
CEC2: 'Reply on AC1', Juan Antonio Añel, 01 Nov 2023
-
AC1: 'Reply on CEC1', Ane Lopez de Gamiz Zearra, 01 Nov 2023
-
RC1: 'Reply on CEC1', Vidette McGregor, 12 Dec 2023
Parameterization and tuning of the Bay of Biscay Atlantis model v1
The report provides a clear presentation of the Bay of Biscay Atlantis Model. There is a lot of information to include. Most of my comments and suggested edits are annotated on the pdf.
The model spatial dynamics seemed a bit unclear. In the methods it seemed the spatial distribution of the age-structured groups were specified for each season which means they don’t change. But the spatial distributions were then presented as model outputs and compared to other data. Also, the discussion mentioned environmental, temperature and salinity effects on movement which were not mentioned in the methods, so it is confusing as to whether these were turned on in this model or not.
Biomass pool biomass should be shown spatially to check it is sensible (i.e., we don’t have all the biomass in one box, unless that’s where it’s meant to be).
Realised diets were not presented anywhere, which I feel they should be. The diet interactions are such a key part of an ecosystem model, and while the prey availability matrix is defined, the realised diets often differ considerably from these due to spatial and temporal overlap and gape sizes if these are being used. In the methods, it seemed gap sizes were not used, but in the discussion, these were mentioned, so it would be good to clarify this in the methods.
Bluefin tuna weights are concerning, and it looks like they could be starving. Has this been checked? Perhaps will be apparent in the realised diets if they have the correct food available to them.
Weights have been shown for each age group, but not numbers. It would be good to include these too. Comparing weights-at-age with expected growth rates from the literature can be helpful for checking the realism of growth in the model. Also checking realised natural mortality in the model to values in the literature (this can be done since fishing is not included).
In the supplementary material:
Table S6. Fractions by depth layer are given for each depth layer, but some polygons have a subset of depth layers. Were the proportions scaled up within each polygon, so they still add to one? It would be good to include this in the caption.
Table S7. It would be good to highlight which of these runs were from which calibration phase (NPZD or full model).
Figure S3. Presenting these in alphabetical order would help. Are the figures after the burn-in period? State that in the caption. Same with other figures that have timesteps.
-
AC3: 'Reply on RC1', Ane Lopez de Gamiz Zearra, 24 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1368/egusphere-2023-1368-AC3-supplement.pdf
-
AC3: 'Reply on RC1', Ane Lopez de Gamiz Zearra, 24 Feb 2024
-
RC2: 'Comment on egusphere-2023-1368', Anonymous Referee #2, 16 Jan 2024
General comments
This manuscript presents the development of an Atlantis model applied to the Bay of Biscay ecosystem, from its parameterization to a first calibration. It places the model in the context of spatial ecosystem-based managment, and concludes that the first calibration met the target of stable biomasses and validated spatial distribution of two species, while emphasizing some data limitation and parameters sensitivity. I recognize that the work done to build this ecosystem model is huge, and has required to integrate a lot of information. However, some of the choices that have been made seem wrong to me, and I do not agree with the validation section. Therefore I suggest to reject the manuscript (but see comments below, hopefully to serve as a basis to improve the model and resubmit a manuscript in the future).
Specific comments
- Spatial validation
The model is presented as a good tool to better understand ecosystem functioning in 3D. Based in the introduction, I would have expected some results regarding spatial functioning. This was limited to only two species (the authors could have produced spatialized ecosystem, community or trophic indicators for instance), and the validation was not convincing. Indeed, the model is intialized with the average state of 2000-2003 (cf line 85) with repeated oceanographic forcing of the year 2000 (cf line 160), i.e. no interannual environmental variability. Sardine and anchovy biomass distribution is parameterized from eggs distribution from BIOMAN 2018 (cf appendix). The authors then compared the simulated biomass of 2019 and 2020 (i.e. 20 years after initialization with similar environmental forcing) with eggs distribution observed during BIOMAN 2019 and 2020. The comparison performed is more about how similar are BIOMAN 2018 to BIOMAN 2019 or BIOMAN 2020 than how good the model is at fitting observation, specially as there is no interannual varibility in the forcings. Therefore you cannot use these results to validate the spatial structure and functionning of this Atlantis model.
- Model setting
I sometimes wondered what was the rationale behind the grouping of some species into functional groups (e.g. the distinction between « deep sea fishes » (including alfonsino, red bandfish, roughsnout grenadier…) and small demersal fishes (including softhead grenadier) is not always clear to me), but there is no perfect grouping, and I can work with that. The main difficulty I have with the grouping relates to the « representative species » used for parameterization. For instance conger represente large demersal fishes, but even if most abundant in the survey, it is quite different from the other species. Using the most abundant species is a correct way of proceeding, but similarities between species should be considered as well (for instance, if a group counts several gurnard species, maybe the abundance of all the gurnards together should be computed to assess what is the main species). I mostly wonder what are the ecological meaning of using the growth parameter of one species (e.g. grenadier) with maturity and mortalities from another species (e.g. alfonsino) as these species can be quite different. Databases of funtional traits exist now, and maybe could be used to ensure the ecological coherence of grouping some parameters of one species with parameters of another. If no other solution exist, this should be at least discussed.
Finally, a lot (if not all) recruitment parameters come from the SE-Autralian Atlantis model, with no indication about possible similarities between ecoysystem functioning and/or taxonomic group. Some details about the generality of this procedure should be provided (it is the same for most Atlantis models – and therefore could be considered as default values – or is it specific to the BoB application, and if so why ?) and the impacts of the results should be discussed as well.
- Biomass estimates and targets
Biomass estimates from survey : how did you translate biomass index from survey to absolute biomass estimates ? Did you consider that species are homogeneously distributed and use the CPUE from the survey scaled to the entire BoB ? Which catchability coefficient did you use ? When estimating the biomass of a functional group with several species, did you use the biomass index of this representative species or of all species ? How did you couple information from EVHOE and DEMERSALES (for exemple regarding spatial distribution, eg bivalves and polychaetes, are you sure that both survey have the same catchalibilites or that both survey provide the same level of taxonomic details for benthic invertebrates?).
Dealing throughout the manuscript both with total biomass (tons) and biomass concentrations (tons/km²) is confusing, and I would suggest to use only one if they are equivalent (and if they are not, specify how to go from one to the other). Some initialization values seems odd in regards to the reference (e.g. other large pelagic fishes initialized with biomass of 0.14 t/km² supposedly from EwE ; but in Corrales et al, this group has a biomass of 0.27 t/km² (idem mesolpelagic fish : 1.80 t/km² in Atlantis – 1.05 in EwE).
I think that more biomass estimates could be obtained from stock assessment (e.g. for sole, which is not well caught by GOV trawls), and could even be associated with smaller uncertainty. Why haven’t you use that source of information ?
- Seasonality
I couldn’t find any information about the seasonality of the parameters (it is mentionned e.g. Line 228 but distribution from once-a-year survey cannot be used for that), so what is variable accross seasons and what’s is not ? How come we can see some seasonality in the simulated biomass of some vertebrates groups but not in the plankton biomass ? It seems to be one of the target of the calibration (cf line 351) but doesn’t seem to be well represented.
- Calibration steps and targets
The calibration steps are great, and should be better emphasized I think. I didn’t understand if when calibrating the NPZD part of the model the authors used the current structure of the model (with several phytoplankton and zooplankton boxes) and a really simple NPZD model. It would have been intersting to plot the initial simulation, the simulation #14 (end of the first step) and the final simulation to see the evolution of this low trophic levels groups. The targets of each calibration step are not specified, so we wonder what model properties is assessed at each step.
Finally I also wonder if one can obtained stable biomass around an observed averaged biomass (with fishing mortality affecting most groups in real life) with no fishing activity/mortality considered in the model and an objective of keeping all mortalities as low as possible (ie not even accounting for the non-represented fishing mortality). Is such a state conceptually reachable ?
Technical corrections
Line 82 : why the model starts from Ribadeo and not Cabo de Finisterre as it is the case for Ecopath ?
Fig 1 : ICES divisions are not easy to see on the map ;
Section 2.2 : there is a lot of information on the Atlantis framework, but that it is not used in the BoB application. This is misleading, I would suggest to focus on key aspects used in the BoB application and/or emphasize on the modular aspect of Atlantis.
Line 124 : what is the difference between age-strucutred biomass pools and age-structured groups ?
Line 188 : what does « nitrogen created as spawn is temporally removed from the model and returned as recruits » mean ?
Line 197 : what are the conversion factors used between nitrogen and wet weigth for vertebrates groups ?
Section 2.4 : I suggest to present model details as they are used for the BoB application (i.e. if a process is optional and not used – do not present it here).
Ei in equation 2 in not detailed/explained, nor is CX line 278
Equation 3 : if you have twice mL.Bp, it becomes a quadratic mortality. I don’t understand this equation
Migration of some species : what’s happening outside the domain ?
Figure 4. years on the x-axis would be better than days. What is really represented here as you remove the spin up period (15 years) but only have 15years long simulations ?
Lines 369-371 : reducing cannibalism was alreday mentionned lines 363-364. The reduction is huge, what does it mean to keep it and not to put it to 0 ?
Figure 6 : not clear, I think several types of link should be used, and that the figure could be presented as plankton food web (one initial, one calibrated), including all the links (ideally with several levels of intensity – not just 2).
- Appendix
Throughout the document : specie should be species
Section 1
- In the first table listing the species, please indicate (in bold or using *) which species is used to set the parameters of the group. Why no species/groups are listed for the functional groups at the end of the table (bivalves, echinoderms zooplankton…)
- I couldn’t find in Audzijonyte et al. 2017 explanations or guidelines about how to convert the annual mortality values found in the literature into mortality value per day for juveniles and adults to be used in Atlantis – could you explain it?
- Check the entire appendix, there are some small mistakes (I didn’t list them all) – for ex : page 19, line 373 should be 32cm and not 32m, page 31, line 710 « maturity of % years », table S7 p90, line 9 and 11 are both labelled BFT…. There is a reference to Wikipedia 2018 for Munida sarsi, but there is no page on Munida sarsi on wikipedia, not information about bathymetry on the Munida spp page, thus remove or update the reference (NB : wikipedia provides references, which are probably more constant through time (and sould be preferably refered to) than wikipedia text itself, always evolving.)
- Table S2 and S3 : for biomass there is no need to provide decimals to the numbers. However, for mortality term, it would be better to use the scientific format (10^X) to provide more significatives numbers and less 0.
- Table S2 : benthic primary producers instead of primary producers
- Table S3 – sardine : what are the implications of having > 99% of 4 age classes mature (age 3 to 6) instead of 100% ; Isn’t it too refined compared to the model structure and level of details of other species?
- Table S6 : why bivalves and suprabenthos are not distributed in L1 (closest to sediment) ? Why zooplankton is mostly distributed into L1 (closest to sedmient) ? why primary producers (should be benthic primary producers) are only located in the surface layer (should be in the bottom layer) ? I understand than in some cases a functional group is not present everywhere in the map, and therefore has not access to the 5 layers [and for those cases I suggest to replace 0 by – in Table S6 to better represent that], but when looking at the benthic primary producers map, it is present in several boxes with several layers, but by definition this group should be on the sea floor.
- Table S8 : this table is impressive, both because it reflects the amount of work done but also the rigor followed when calibrating the model. I am sure it can be further exploited in different ways. I was surprised though that when a parameter is calibrated, you don’t come back to adjust its value. Does it mean that there is no retroaction effects from the upper trophic levels components on the dynamics of a group ?
- Is there a difference between figure S2 and figure 3 from the main text ?
Citation: https://doi.org/10.5194/egusphere-2023-1368-RC2 -
AC4: 'Reply on RC2', Ane Lopez de Gamiz Zearra, 24 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1368/egusphere-2023-1368-AC4-supplement.pdf
-
EC1: 'Comment on egusphere-2023-1368', Heather Hyewon Kim, 24 Feb 2024
Dear authors,
Please submit your revised manuscript.
Best,Heather
Citation: https://doi.org/10.5194/egusphere-2023-1368-EC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 567 | 196 | 50 | 813 | 109 | 27 | 30 |
- HTML: 567
- PDF: 196
- XML: 50
- Total: 813
- Supplement: 109
- BibTeX: 27
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 275 | 34 |
| Spain | 2 | 84 | 10 |
| France | 3 | 59 | 7 |
| Germany | 4 | 52 | 6 |
| Netherlands | 5 | 42 | 5 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 275