Review: Implementing a coral reef CaCO3 production module in the iLOVECLIM climate model

Bouttes et al.

**

GENERAL COMMENTS

**

This paper presents a new model of coral reef carbonate production, designed to be coupled to earth system model frameworks and allowing for the simulation of long-term feedbacks between coral calcification and the larger carbon cycle.  I am not aware of any other models of coral calcification that are designed to be run in a fully-coupled manner with ESMs, and as such agree that this model fills a previously unfilled niche in climate system modeling.  While the impacts of changing carbonate system properties on biology is often simulated, the feedback effects of that biology on the oceanic and atmospheric carbon are much less often addressed, and may become increasingly important as research into novel fields like carbon dioxide removal increases.  Overall, this was a clearly-written paper.  While the results are somewhat preliminary, covering the initial setup and some short simulations to verify its ability to replicate current conditions, I think that even as a proof of concept this will be a valuable contribution to the field.  There are a few places where the model details and the thought processes leading to those details could be expanded and clarified, which I include in my specific comments below.

My one broader comment for this paper is in regard to how iCORAL is alternatively framed as either 1) a coral module specific to the iLOVECLIM framework vs 2) a module more generically available for use in ESMs as a whole.  Several of the model implementation choices (e.g. subgridding the bathymetry to compensate for low horizontal resolution, adding temperature variability in the tropics) are clearly tailored to iLOVECLIM.  Also, the source code is not written in a standalone, modular manner (it loads many of its variables through external fortran modules from the wider iLOVECLIM code*, many parameters are hard-coded, etc.); porting to a new ESM would be a non-trivial task.  I suggest revising the text where relevant to clarify that the algorithms and concepts behind the model may be applicable more widely, but significant work would be necessary to port, test, and validate the model under a new ESM coupling.

*Regarding the source code, the author comments (CEC1) imply that they made iLOVECLIM source code available to referees, but I did not receive any documents apart from those available for public review.  Therefore, I have not been able to check whether the coral_mod_paper.f90 code (as linked in the Code availability section) compiles and runs in that context.  However, the coral_mod module itself is clearly written -- well-commented, cleanly organized and formatted -- and seems mathematically sound.

**

SPECIFIC COMMENTS

**

Section 2.1: The iLOVECLIM setup description could use a bit more detail.  First, I wasn't entirely clear what the relationship between iLOVECLIM and LOVECLIM was.  Based on the iLOVECLIM GMD special issue text (https://gmd.copernicus.org/articles/special_issue30.html), I gather that iLOVECLIM derived from LOVECLIM, minus ice (and possibly minus ocean carbon/bio?), then developed independently?  A quick summary of this history in the paper would be useful to orient readers.  Also, is the ocean carbon cycle model described in this paper a standard part of the iLOVECLIM configuration?  I believe you're using the HAMOCC model for ocean biogeochemistry; can you specify which version?  Are there any other parts of your iLOVECLIM configuration (other than the addition of iCORAL) that differs from other published implementations? 

Line 91-92: "Photosynthesis takes place in the euphotic zone in the upper 100 meters."  Can you clarify whether this depth is prescribed or an emergent property of the model?

Line 95: Given that iCORAL is an extension of ReefHab, a quick description of that model would be useful.  It sounds like ReefHab is a classic coral production model, and iCORAL uses the same basic equations but adapted to use ESM-derived input rather than just observational datasets (and with some expanded constraints like the bleaching addition.)  Saying this explicitly would help orient readers unfamiliar with ReefHab.

Figure 1.  I didn't find this figure particularly useful; the graphical elements don't provide any new information.  I suggest either turning this into a more information-dense schematic (perhaps diagraming the input from coupled model components vs. offline datasets, and the additional pre-preprocessing steps applied before feeding that data into iCORAL, etc.) or eliminating it.

Line 110: Perhaps clarify that the production depth is defined as the depth at which light is at the Imin level?  Readers unfamiliar with the typical light attenuation equation may not immediately parse this detail.  

Line 115: Please provide more detail about how the K490 coefficients were chosen and calculated.  What specific MODIS data was used to construct the binned data?  Did you use the entire mission composite, and if so, through what dates?  At what horizontal resolution (4km or 9km)?  How were these data matched to the iLOVECLIM horizontal grid?  Other satellite-based algorithms are available to calculate this attenuation coefficient near turbid water (for example, NOAA's version based on Tomlinson et al., 2019: https://doi.org/10.1080/2150704X.2018.1536301); were these considered?

Also, I'm curious about the decision to use a satellite product to derive these coefficients, rather than deriving the attenuation directly from the model itself (i.e. as a function of chlorophyll + clear-water attenuation, with the latter tied to the simulated phytoplankton group).  Given the intent to use this model for paleo-scale simulations, I'm concerned that this choice may decouple the model's simulated chl from the prescribed attenuation coefficients that are at least in part a function of satellite-era chl concentration.  I recognize that using a prognostic base for this calculation comes with its own set of difficulties (requiring skillful phytoplankton simulation and some estimate of the CDOM/sediment/etc. distribution), but I think this choice warrants some additional discussion in the paper.  You do later mention this as a caveat in the Discussion section 4.1 (Model caveats), but I'd like to see it addressed more fully, either in the Methods or in that caveats section.

On a related note, is the iLOVECLIM configuration capable of responding to long-term changes in wetting/drying?  I.e. if sea level rises, can grid cells previously designated as land become inundated, or vice versa?  If so, how does the iCORAL model handle parameterization of grid cells without satellite-era data to use?  Given that an inundation-related scenario is presented as a driving theory for the potential importance of coral reefs in the carbon cycle, this is a key technical point in the potential usefulness of this model setup.

Line 155-160: The description of the 1-m vertical subgridding could be clearer; it took me a few readings to figure out what was being done.  Using terminology like "vertical resolution" and "depth interval" was a little misleading to me, since initially I interpreted this as a sub-gridding of the water column, rather than a pairing of each 3-deg grid cell with its subgrid-scale bathymetry distribution.

Line 243: Criv and Ariv are never explicitly defined; I assume those are the river input flux of DIC and Alkalinity, respectively?

Line 259: "AR(p)" may be too jargon-y for a non-statistical-modeling audience; please provide a quick definition of this model type and what the model order (p) refers to.  Also, you mention supplementary material here, but I could not locate any supplementary material associated with this manuscript.

Line 263: Where did the values of 0.90 and 0.28 come from?  Are they standard values or were they derived from the fitting process?

Figure 4: Anomaly plots would be helpful to allow quicker direct comparison between column 1 and 2.

Lines 319-322: You mention that coral overproduction in the model might be due to lack of riverine nutrient input leading to lack of competition from macroalgae.  While that is one possible source, lower-resolution models (even 1-deg) tend to underestimate coastal pelagic production due to poor resolution of shelf dynamics, which could alter the conditions in these regions.  And it doesn't appear that this model simulates macroalgae at all, so even with improved riverine input that competitive pressure would be missing in the simulations.  A bit more discussion of the possible drivers of oversimulation of coral could be useful here.

Line 323: "might not be present in the observed data."  Are you saying that these types of features might exist in the real world but not be captured by this particular dataset?  Or that the model is simulating a type of isolated coral reef that probably doesn't exist in the real world?  

Lines 432-434: This statement implies that perhaps the answer to my inundation question above is no, the model cannot handle wetting/drying.  It's not clear to me how one would reparameterize coral growth parameters to be dependent on sea level without explicitly allowing for this; can you expand on this idea?

**

TECHNICAL COMMENTS

**

Equations: Equation numbers would be nice (possibly not a requirement for this journal?).

Line 266, 277, etc: Use of present perfect rather than past tense read a little strangely to me.

Figure 2: Is this just presence/absence data?  Please clarify in the caption or add an appropriate legend.

Fig. 7a: I suggest adding the blue/purple highlights to this panel for consistency with Fig 7b, 7c, 8, and 9. 

[see PDF]