Comment on bg-2021-64 Anonymous Referee # 2 Referee comment on " Functional consequences of Caribbean coral reef habitat degradation "

Webb et al., measured the community metabolism of small areas of a degraded Caribbean coral reef through in-situ incubations of benthic communities. Five incubation tents were deployed over coral, algae, and sand dominated benthos, representative of different states of coral reef degradation. Biogeochemical parameters were measured over 4-hour incubations at night and day. An inverse modelling approach was applied to the collected data. The key results were interpreted in the context of ecological function. Calcification and productivity were low and night-time respiration outweighs daytime productivity. The manuscript presents a unique and interesting approach to quantifying differences in biogeochemical processes on degraded coral reefs, however, there are some limitations to the study which should be addressed, and the inferences/conclusions that the authors make may need to be re-framed accordingly. The experimental design had low replication, and the measurements were made at one single location over just a few days / nights. The tents were leaking during the incubations, which would also have impacted the measurements. The logistics of such in-situ incubations are very challenging, and it was a good idea deploy the tents in duplicates / triplicates, but there is some variability within substrate replicates (in terms of composition and biogeochemical activity) to suggest that they could be evaluated individually. I think that the authors could provide some more information about the inverse modelling approach they use, and the advantages of using such an approach.


Introduction
Community composition and biodiversity across all kinds of ecosystems are responding to escalating anthropogenic activities 35 (McGill et al. 2015). In both terrestrial and aquatic systems, climate change, pollution and habitant fragmentation have promoted the expansion of opportunistic and tolerant species and the elimination of more sensitive yet key specialists (Clavel et al. 2011). The latter has resulted in the increased similarity of biological communities within ecosystems and across spatial scales (Burman et al. 2012;Cramer et al. 2021). This is worrisome as it may lead to a decrease in functional diversity therefore limiting services provided by biological communities (Matsuzaki et al. 2013;White et al. 2018). Furthermore, this functional 40 homogenisation may synchronise the biological response to new or intensified anthropogenic pressures across local communities thus reducing resilience of metacommunities (Tobias and Monika, 2012;Sonnier et al. 2014;Petsch et al. 2020).
Coral reefs support immense biodiversity and provide important ecosystem services to millions of people (Moberg and Folke, 1999). They are however in global decline as they are experiencing major loss in coral abundance and shifts in species composition in response to increasing human pressures and accelerating rates of environmental and climate change (Koop et 45 al. 2001;Langdon and Atkinson, 2005;Andersson and Gledhill, 2013;De'ath et al. 2012;Chen et al. 2015). Returning degraded reefs to their original state is, in many cases, no longer an option (Hughes et al. 2017). Instead, today's challenge is to guide coral reefs through this transition while identifying and securing the ecosystem services that highly altered reef assemblages can provide to people in the future (Oliver et al. 2015). Therefore, it is essential to understand and quantify the functional consequences of community changes on increasingly degraded coral reefs. 50 The depauperate reef systems in the Caribbean, have, since the early 1970s, undergone considerable reorganisation with regards to community composition and structural appearance (Gardner et al. 2003;Jackson et al. 2014). The communities encountered on these reefs bear little resemblance to the systems once dominated by reef building Acropora spp. and Orbicella spp. (van Duyl, 1985;Alvarez-Filip et al. 2009). Major declines in the abundance of species such as Acropora palmata and Acropora cervicornis severely compromised reef function as these species was the main representatives of critical functions 55 including carbonate accretion, productivity and structural complexity. Low functional redundancy on Caribbean reefs, i.e. the reduced capacity of one or more species to functionally compensate for the loss of another, makes them particularly vulnerable to functional homogenisation (McWilliam et al., 2018). Instead, algal turf assemblages and macroalgae, excavating sponges, cyanobacteria, rubble and sand have increased, thereby mirroring the decrease in stony corals (Aronson et al. 2005;Burman et al. 2012;Cramer et al. 2021). Although changes in community composition are well documented as they can be followed 60 by monitoring the coverage of the various benthic taxa over time (Barott et al. 2012;de Bakker et al. 2016;de Bakker et al. 2017), assessing the impact of these shifts on the community ecophysiology in situ has proven more challenging.
The keystones of coral reef functioning include provision of a structural habitat through carbonate deposition, production and assimilation of biomass produced through photosynthesis and efficient cycling of nutrients within the ecosystem (Brandl et al. net budgets of these processes will provide insight into how reef degradation and community reorganisation affect reef functioning (Brandl et al. 2019). Presently, efforts to quantify community functions have focused on individual functional groups (Brocke et al. 2015(Brocke et al. , 2018Webb et al. 2017;de Bakker et al. 2018) and the limited amount of studies that incubated whole communities in situ have so far not accounted for the complexity of interactions between 70 biogeochemical processes (Yates and Halley, 2003, Kline et al. 2012, Roth et al. 2020).
Here, biogeochemical processes underlying key reef functions were quantified in situ across five different benthic assemblages found on the fringing reef of Curaçao, with functional groups that currently characterise many degraded shallow reef habitats throughout the wider Caribbean. To this end, a custom-made tent was placed over substrates dominated by either 1) coral, 2) turf and macroalgae, 3) bioeroding sponges, 4) benthic cyanobacteria mats or, 5) sand. Chemical fluxes between water column 75 and reef were then determined by monitoring nutrients, inorganic carbon chemistry and oxygen. This was done both during the day and night to estimate overall net metabolism of these communities. To account for the multidimensionality of processes interacting on the measured variables, the change in their concentrations is related to the responsible metabolic processes by solving a system of ordinary differential equations that describe the contribution of each process to the measured chemical fluxes. With this approach, we aim to provide accurate estimates of biogeochemical processes that underlie functions of the 80 newly configured shallow Caribbean reefs.

Study Site
Reef incubations were carried out on the leeward side of Curaçao (Piscadera Bay; 12°07'16.3"N 68°58'13.2"W) between February 12th and March 22th 2018, at depths ranging from 5 to 7 meters. The water at the study site is characterised by 85 episodes of high turbidity and is periodically eutrophied due to terrestrial runoff and ineffective waste-water treatment.
Sediment plumes transporting high concentrations of nitrate, ammonium and phosphate into the shore's fringing reef are commonly encountered after a period of heavy rainfall . The shallow reef flat nearby the entrance of the bay in which we conducted our incubations is characterised by rubble and patchy distribution of small coral heads making this location particularly suitable for the deployment of tent incubations. 90

Tent Incubations
The incubation enclosure consists of a custom made, tetrahedron-shaped "tent" (Fig. 1). It has transparent, vinyl-and-butanyl walls with rigid pole edges of 1 m, resembling the cBIT described by Haas et al. (2013). It also includes 0.5 m long flaps extending outward from each of the tent's three sides, allowing for proper sealing of the tent to the substrate by placing weights on the flaps. It covers a 0.43 m 2 planar surface, and encloses a 118 L volume. All three sides of the tent contained an opening 95 to allow flushing of the enclosed volume between incubations: during incubations these openings were sealed. Water enclosed in the incubation tent was homogenised during the experiment by means of a continuously running brushless submergible https://doi.org/10.5194/bg-2021-64 Preprint. Discussion started: 23 March 2021 c Author(s) 2021. CC BY 4.0 License.
water pump (BLDC pump Co., Ltd.). This pump was attached to one of the tent poles, at half the height of the tent, generating a vertical circulating turbulence, while minimising stirring up of sediment. Effectiveness of the stirring was demonstrated by rapid and even dispersal of a small dose of injected fluorescein prior to the incubation. Surge movement was retained due to 100 the non-rigid texture of the tent walls. Five different types of substrate dominated either by turf and macroalgae, sand, bioeroding sponges, benthic cyanobacteria mats or coral (Fig. 2) were incubated both during daytime (in triplicates) and at night-time (in duplicates) for 4 hours each.

Substrate Compositions
Substrates dominated by either coral, turf and macroalgae (TMA), bioeroding sponges (BES), cyanobacteria mats (BCM), or 105 sand were incubated (Fig. 2). Three reef patches of each reef assemblage were chosen depending on their dominant benthic component (see Table S1 for detailed species composition and cover). In some cases, to fit adequate incubation location and the tent capacity, pieces of rubble infested with sponge or covered in turf were added or retrieved from the community to be incubated. Incubated substrate included colonised hard substrate surrounded by bare hard substrate covered in a fine layer of sand for better enclosure deployment (except for sand incubations). The incubated coral species are characteristic of degraded 110 Caribbean reefs and include some of the most prominent tolerant and opportunistic species found on modern reefs (Darling et al. 2012;de Bakker et al. 2016;Cramer et al. 2021) (see Table S1). Turf here refers to epilithic algal matrix defined by Clements et al. (2016) as 'a conglomeration of short, turf-forming filamentous algae (< 1 cm high), macroalgal spores, microalgae, sediment, detritus and associated fauna'. The benthic cyanobacterial mats in all three tent replicates were thick brown/reddish in colour and in line with the description in Brocke et al. (2018) for mats found between 3 and 7 meters dominated by the 115 species Oscillatoria bonnemaisonii. Percentage cover was measured in situ after removal of the tent. For substrates dominated by coral, its cover ranged from 34 to 36 %. Turf and macroalgae cover ranged between 72 to 83 %, bioeroding sponge cover varied from 38 to 40 %, and cyanobacterial mats cover ranged from 83 to 91 % (Fig. 2).

In Situ Measurements
Measurements of salinity (S), temperature (T), dissolved oxygen (O2) and photosynthetically active radiation (PAR) within 120 the tent were recorded at 1 min intervals throughout the duration of the incubations. S and T were measured using a Star-Oddi DST CTD, O2 was recorded using a HOBO U26 dissolved oxygen sensor and data logger and PAR was assessed by an Odyssey light logger (Dataflow Systems PTY Ltd., Christchurch, NZ), calibrated in air against Walz instrument (Walz ULM500, Walz GmbH, Effeltrich, Germany). In addition, S, T and PAR were measured for the duration of the incubations outside the tent using the same sampling frequency. All instruments within the tent were attached to the three ridges except the Odyssey logger 125 which was placed on the substrate facing upwards (covering approximately 150 cm 2 of the substrate).

Discrete Sampling
During each incubation, discrete samples both inside and outside the tent were collected at T0, after 2 (T2) and after 4 hours (T4) by scuba diving for the analyses of total alkalinity (AT), total inorganic carbon (CT) and nutrients. pH was calculated from the former two parameters using the package Seacarb (Lavigne et al. 2009). Sampling of the tent interior was carried out from 130 the outside by drawing seawater through 150 ml plastic syringes connected to a 1.5 m gas-impermeable tube (Tygon; Fig. 1).
Syringes were flushed three times with the sampling water before collecting an actual sample. The tubing was fixed around a rigid edge of the tent in such way that the seawater was sampled from the centre of the tent incubation. The tube end located inside the tent was equipped with a Whatman ® filter (G/F 0.47 µm).
Analyses for AT were performed within 2 hours upon sampling using spectrophotometrically guided single-step acid titration 135 (Liu et al. 2015) and samples for CT were run on an autoanalyser Traacs 800 spectrophotometric system (Stoll et al. 2001).
Accuracy of both instruments was set using certified reference material supplied by Scripps Institute of Oceanography (Dickson et al. 2007). Precision of replicates was 2.7 µmol kg -1 for CT and 0.9 µmol kg -1 for AT. Samples for dissolved inorganic macronutrients (NO2 + NO3, NO2, PO4 and NH4) were prepared by dispensing sampled water through 0.8/0.2 μm Acrodisk filters into 5 ml pony vials, and subsequently stored at -20 ⁰C until analysis at NIOZ on a QuAAtro continuous flow 140 analyser (SEAL Analytical, GmbH, Norderstedt, Germany) following GO-SHIP protocol (Hydes et al. 2010).

Rates of Water Exchange
After sampling water at T0 for AT, CT and nutrients, 450ml of water saturated in salt was injected into the tent. The rate at which the elevated interior salinity equilibrates with ambient salinity during the incubation is used to estimate the rate of water exchange with the surrounding sea water for each incubation. 145 The rate of change of salinity within the incubation can be solved by the differential equation below: Where is the rate at which salinity changes within the tent, Sout is the exterior salinity, S is the interior salinity and K is the water exchange rate.
The equation is solved using the function 'ode', within the package deSolve , the R routine that solves 150 the differential equations. Function 'modfit' from the package FME ) was used to perform iterative minimisation (based on least squares) on residuals to find best fit within lower and upper bounds.

Inverse Modelling and Model-Data Comparison
We can describe the mathematical "state" of the incubation's dynamic system based on the mass balance of the measured variables. The five differential equations depicted below relate the change in their concentrations to the responsible processes, 155 which are assumed to have remained constant over time. Since the involved processes affect the different chemical components simultaneously, the combination of these differential equations can be used to solve the contribution of the processes to the observed changes. The processes of interest include aerobic mineralisation (O2 consumption related to mineralisation), primary production (PP), calcification, dissolution, nitrification and denitrification. 160 With mineralisation describing the degradation of an organic compound to its mineral components, i.e. carbon dioxide and inorganic nutrients. PP is the primary production and calcification is the deposition of calcium carbonate. Pnh4 is the part of N uptake as NH4 for primary production. Dissolution results in an increase of calcium and carbonate ions by degradation of 170 calcium carbonate shells and/or skeletons and K is the water exchange rate. Nitrification is the process by which ammonium (NH4 + ) is converted into nitrate (NO3 -); two moles of oxygen are needed to oxidize one mole of ammonium during nitrification.
pDeni is the fraction of mineralisation that respires nitrate (i.e. denitrification). The OCratio is the ratio between the concentrations of oxygen and CT. The NCratio is the ratio between N and CT. The 0.8 constant refers to the denitrification redox reaction (Soetaert et al. 2007). 175 We start by determining the parameters that can be fitted, based on parameter collinearity. After producing a best-fit set of the selected parameters, we quantify parameter uncertainty, and produce sensitivity ranges around the modelled variables.
The OCratio, NCratio and K parameters are always fixed and estimated from data prior to running the model. Others vary between fixed and free (to be fitted) depending on collinearity and light. For instance, primary production is fixed at 0 during night incubation, however during the day, only the dominant process can be estimated. Some parameters are highly correlated with 180 each other such as primary production and remineralisation or calcification and dissolution and therefore cannot be estimated simultaneously. In general, when the collinearity index exceeds 20, the linear dependence is assumed to be critical (i.e. it will be impossible or difficult to estimate all the parameters in the combination together).
Collinearity of the parameter sets is measured using function 'collin' within the FME package (Soetaert and Petzold, 2010). (remineralisation, calcification, etc.). The differential equation model is solved using function 'ode', within the package deSolve, the R routine that solves the differential equations.
The discrepancy of the model solution with observed changes within the tents is calculated using function 'modCost' still in the FME package which estimates the residuals and the variable and model costs (sum of squared residuals). 190 Function 'modfit' was then used to perform iterative minimisation (based on least squares) on residuals to find the parameter giving the best fit within lower and upper bounds. Estimated parameters are the unknown fluxes (mineralisation, PP, calcification, etc.).

Conversion to fluxes
The best-fit parameters, i.e. the input rates R (in µmol kg -1 min -1 ), in the tent are converted to fluxes from the water-substrate 195 interface (mmol m -2 h -1 ), assuming an enclosed mass of water of 108 ± 10 kg (tent encloses approximately 118 litres of volume; of which substrate volume is ~10 L; seawater density ~1022 kg m -3 ) and an incubated planar surface of 0.43 m 2 .
Net community calcification (NCC) fluxes were determined from the sum of predicted calcification and dissolution. The model captures the dominant net flux and does not distinguish the relative contributions of gross calcification and dissolution to the integrated NCC rate. Net community production (NCP) is the difference between remineralisation and primary production. Denitrification is estimated by multiplying pDeni and mineralisation parameters together.

Statistics
To assess the differences between the effects of community composition on biogeochemical processes, we ran a nonparametric permutational multivariate analysis of variance (PERMANOVA) with pairwise contrasts. The vegan Package in R (Oksanen et al. 2007) was used to calculate a dissimilarity matrix using Euclidian distance. The Holm-Bonferroni method was 205 used to adjust the rejection criteria for each of the individual hypotheses and therefore reduce the higher probability of obtaining Type I errors (false positives) when performing multiple comparisons (Holm, 1979).

Results
In-tent light and temperature were only slightly impacted by the tent enclosure compared to the exterior (Fig. 3). Light was on average 17 % lower inside than outside the tent and changes in temperature were dampened within the tent. In-tent temperatures 210 were on average 0.2 ºC higher than those outside the tent.
Application of Equation (1) to salinity data collected during all incubations yields leak rates K of the enclosure f ranging between 0.004 and 0.044 min-1. This indicates that 0.5 to 4.8 kg of seawater (i.e., K × 108 kg) is exchanged every minute between the incubation enclosure and the environment. These rates correspond to the intensity of the water movement observed and recorded visually at the time of each incubation. returns to ambient concentrations after ~1 and ~2 hours respectively. The dilution rate K = 0.0192 and K= 0.0751 indicate that these particular tents leaked at a rate of 2.1 to 8.1 kg of seawater per minute.
The model shows a relatively good fit to the observations, indicating that the interactions between processes and their effects on chemical fluxes were considered correctly (Fig. S1). Overall fit is usually better on night data, which is mostly due to the 220 inability of the model to predict irregular oxygen evolution during the day-time (Table S2). Graphic output of two incubations for the model employed to estimate reef processes are presented in Fig. 5. Output for all incubations, are listed in the supplement (Fig. S1, as well as all parameter predictions and their significance in Table S2).
As the process estimates are limited to net increase or decrease, fluxes for PP and mineralisation are presented as net community production (NCP) and calcification and dissolution are combined into net community calcification (NCC; Fig. 6; 225 Table 1).
NCP shows a clear diurnal pattern (Fig. 6). While all NCP values are modestly skewed towards net autotrophy during the day (except for sand), the strongest signal is found for substrates dominated by BCMs with an average daily NCP of 5.6 mmol m -2 h -1 . Night values indicate net respiration ranging from an average of -2.64 mmol m -2 h -1 on substrates dominated by TMA to -16.28 mmol m -2 h -1 on substrates dominated by bioeroding sponges. 230 A clear diurnal signal also resides in NCC fluxes for all substrates involved (Fig. 6). Most NCC fluxes recorded during daytime (with the exception of sand incubations) indicate net CaCO3 precipitation. At night, most NCC fluxes indicate net CaCO3 dissolution, especially on substrates dominated by BES and BCMs. The absence of change in AT for coral dominated substrates during the night indicates that dissolution equals calcification during these incubations and hence, the average NCC is close to 0. Substrates dominated by coral generated the strongest decrease in AT (net precipitation) during daytime, yielding an average 235 NCC rates of 0.45 mmol CaCO3 m -2 h -1 . Highest net dissolution was found at night-time for incubations of substrates dominated by bioeroding sponges and cyanobacterial mats, with a similar average of 0.56 and 0.63 mmol CaCO3 m -2 h -1 respectively. All average fluxes and respective 95% confidence intervals are listed in Table 1.
Nitrification was found to occur predominantly at night with higher fluxes in incubations of substrates dominated by bioeroding sponges, cyanobacterial mats and corals. Denitrification also occurred mostly at night except on sand where daytime and night-240 time fluxes were small but relatively similar (Fig. 6).
The pairwise perMANOVA results revealed that there were no significant differences in processes measured between any communities with distinct composition (Table S3). R-squared values varied from 0.05 to 0.17 indicating that 5 to 17 % of the variation in distances is explained by the grouping being tested (here for BCM vs BES (p= 0.691 and BCM vs sand (p=0.304 respectively). The differences between day and night processes were significant (R 2 = 0.43, p=0.001). 245

Discussion
The biogeochemical flux assessment has enabled us to identify and quantify the biological functions that are currently at play of this degraded Curaçaon reef. Although the present study only investigated the shallow part of one single reef, it gives us insight into the dire effects that shifts in coral species and functional groups has had on the overall functioning of these communities. Comparison with coral reefs in different biogeographical context is needed to establish whether the rates 250 obtained here are site-specific or representative of degrading Caribbean coral reefs in general.
The shallow reef communities investigated here barely support reef functions that are usually ascribed to a healthy coral reef.
Overall, net community calcification and production on these substrates are low compared to reef flats worldwide (Atkinson, 2011). No significant gain in primary habitat is recorded with very low or negative net community calcification rates on all substrates. Net production and therefore accumulation of biomass produced through photosynthesis is also low, while 255 heterotrophic processes are prominent. Recycling processes, nitrification and denitrification, are high but do not prevent net nutrient release from aerobic mineralisation, rendering all substrates, sources of nitrogen. Although processes recorded on substrates dominated by coral, bioeroding sponges and cyanobacterial mats show some variation between types of substrates, the overall performance of complementary processes for each of these assemblages is relatively comparable. In addition to the modest biological functions of these benthic communities -resulting from low net community calcification and production -260 results suggest that some degree of functional homogenisation is occurring between substrates dominated by different functional groups.

Net Dissolving Reef
Although net calcification was recorded during the day on all substrate types (except sand), it did not compensate for higher dissolution rates at night except on substrates dominated by coral. Diel shifts between net calcification and net dissolution are 265 not uncommon and have been recorded on healthier reefs than the one studied here (Yates and Halley, 2003;Albright et al. 2013;Albright et al. 2015;Koweek et al. 2015) with instances of net dissolution mainly taking place at night, coinciding with net respiration (and most likely with low gross calcification) (Cyronak et al. 2018). However, the community calcification budget over 24 hours resulting from these shifts in the present study results in a modest average accretion rate of 5.7 mmol CaCO3 m -2 day -1 (95% ci = 1.2; 3.1) on coral-dominated substrates. This is low compared to rates reported for reef flats 270 worldwide (with an average around 130 mmol CaCO3 m -2 day -1 and ranging from 20 to 250 mmol CaCO3 m -2 day -1 ; Atkinson, 2011). Overall, the limited number of in situ flux-based experiments carried out in the wider Caribbean (Yates and Halley 2003;Muehllehner et al. 2016;van Heuven et al. 2018) suggest they are among the lowest NCC rates recorded worldwide (Albright et al. 2015;Shaw et al. 2015;Silverman, 2007) an particularly low compared to those in the Indo-Pacific region (Koweet et al. 2015;Takeshita et al. 2016). Surveys using a census-based approach (Perry et al. 2013;de Bakker et al. 2019) 275 also showed that some Caribbean reefs are net eroding. The recorded low rates of net carbonate production in the wider Caribbean may be expected simply due to the region-wide decrease in coral coverage since the 1970s. However, the decrease in net calcification relative to historical values is likely related to more than just coral cover loss. Indeed, the latter has subsequently left surfaces available for colonisation by turf, macroalgae (Hughes, 1994) and more recently, cyanobacterial mats (de Bakker et al. 2017). Shallow reefs around Curaçao (<10m) are covered by filamentous algal turf canopies that 280 presently represent the most dominant benthic component on these reefs (Vermeij et al. 2010). Given their abundance and high https://doi.org/10.5194/bg-2021-64 Preprint. Discussion started: 23 March 2021 c Author(s) 2021. CC BY 4.0 License. release rates of dissolved organic carbon (Mueller et al. 2016), heterotrophic activity is likely to be stimulated. Furthermore, cyanobacterial mats release part of their photosynthetically fixed carbon as DOC into the water column at a higher rate than turf and macroalgae (Mueller et al. 2014;Brocke et al. 2015). They have been shown to be responsible for 79% of the total DOM release over a 24 h diel cycle at this same study site (Brocke et al. 2015). Considering their proliferation around the 285 islands of Curaçao and Bonaire since 2003 (De Bakker et al. 2017) and the prevalence of turf algae in the area, an accumulation of organic matter may have resulted in a reduction of pH due to oxidation of organic matter, i.e. stimulated heterotrophic activity, resulting in reduced calcification (Bates et al. 2010). Muehllehner et al. (2016) suggested that the seasonal character of reef dissolution they recorded on the Florida Reef Tract coincided with an accumulation of organic matter, following the die-off of annual sea grasses in the area. In the present study, incubations took place in Feb-March when water temperature is 290 lower (~26 °C) than Aug-October (~29 °C) for instance. Average ambient pH was 7.9 which did not alter significantly between day and night. This is lower than average 'summer' pH which is usually between 8.1 and 8.2 (den Haan et al. 2016) within this area indicating a potential seasonality component to reef dissolution in the Piscadera Bay.

Net heterotrophic reef
Low net community production rates in the current study indicate that autotrophic processes dominate modestly during the 295 day. Integrating the NCP values over 24 hours (day + night) yielded rates skewed towards net respiration, indicating heterotrophy in all incubations. Although net community production of reef flats has been reported to vary notably over the course of the day (Koweek et al. 2015), with values ranging from -220 to +310 mmol m -2 day -1 (Atkinson, 2011), large amplitude shifts between net autotrophy and net heterotrophy are usually recorded between day and night (Yates and Halley, 2003;Albright et al. 2013;Albright et al. 2015;Koweek et al. 2015). Here, the amplitude of this shift between day and night 300 is modest. It should be noted however, that the reduction in light intensity by 17 % on average may have resulted in a slight underestimation of NCP measurements. This would hold especially true for BCM incubations. Reductions in light would intensify down the steep vertical physiochemical gradients present in these microbial mats, and could interfere with lightcontrolled circadian regulation of photosynthesis and respiration in these cooperative communities (Hörnlein et al. 2018), favouring respiration and decreasing net community productivity. 305 The reduction in the amplitude of the diel shift in net production and calcification recorded in the present study may have severe implications. For instance, metabolic fluctuations from reef biota cause strong temporal fluxes in compounds which affect the oscillatory behaviour of reef seawater microbial communities (Kelly et al. 2019;Weber et al. 2020) leading to less distinct populations and more redundancy in microbial specialists' functions, i.e. a shift to a dominance in catabolic pathways.
Organic material supplied to the ecosystem by benthic primary producers as exudates is also thought to play a pivotal role on 310 microbial growth (Haas et al. 2011) and diversity Haas et al. 2016) depending on its origin. Studies on the effect of exudates of macroalgae and turf on microbial metabolism demonstrated that the composition of exudates stimulated rapid growth of less diverse microbe communities compared with coral derived exudates. Microbial communities shift towards copiotrophic populations that have the potential to remineralise available organic nutrients at a high rate and encode greater https://doi.org/10.5194/bg-2021-64 Preprint. Discussion started: 23 March 2021 c Author(s) 2021. CC BY 4.0 License. numbers of potential virulence factor genes, ultimately harming corals and maintaining algal dominance 315 Dinsdale and Rohwer, 2011). We infer that the amount and type of organic matter provided by abundant algal turfs mats on this reef, likely enhances heterotroph activity and stimulates the proliferation of less diverse copiotrophic microbial populations, rendering the studied reef net heterotrophic regardless of substrate type.

Nitrogen cycling
Nitrogen pathways support high primary productivity in oligotrophic environments by supplying nutrients while 320 simultaneously preventing the build-up of excess nutrients that may favour opportunistic primary producers such as algal turfs (e.g. O'Neil and Capone, 2008;Karcher et al. 2020). The abundance of non-coral primary producers on these reefs suggest that nitrogen is not a limiting factor for growth. Results showed that all substrate types acted as NH4 + and NO3sources during the day and the night, with the exception of sand and turf substrates which acted as sinks for NO3 -. This is to be expected from overall net heterotrophic communities, however, even in instances of net autotrophy during the day, substrates still acted as 325 DIN sources. This is comparable to recent results from in situ incubations carried out in the central Red Sea on net autotrophic coral and algae-dominated communities (Roth et al. 2020). The concurrent community-wide processes such as the consumption and transformation of organic matter by microbial populations (e.g. Pfister and Altabet, 2019) are expected to mask the assimilation of DIN by primary producers.
Nitrification and denitrification rates measured in the present study generally fall within the published range of in situ 330 measurements in tidal pools dominated by algae and corals (Webb and Wieber, 1975), in cavities covered in encrusting sponges (Scheffers et al. 2004), on cyanobacterial mats (Bonin and Michotey, 2006) and on carbonate sand (Capone et al. 1992;Eyre et al. 2013b). However, there was no nitrification during the day (except for the community dominated by sponges), which may be explained by light causing a reduced activity of nitrifiers (Kwon et al. 2020). Owing to the rather shallow depths of our experiment, nitrifiers may have been negatively affected by the light. As mentioned above, microbial communities are 335 impacted by organic matter composition and temporal fluctuations in biogeochemicals. Shifts in diversity and abundance of the microbial communities inhabiting the reef substrate may also lead to diel shifts in nitrogen-cycling capacity (Rädecker et al. 2015). Further research investigating how alterations in diversity and abundance of these microbial functional groups relate to changes in the nitrogen-cycling capacity of reef assemblages is needed at this point.

Functional homogenisation 340
Although processes recorded on distinct community assemblages show some variation between substrate types, a multivariate pairwise analysis revealed that there is no significant difference between overall processes occurring on any of the assemblages.
Expectedly, significant variation is found between day and night for all processes. Considering the differences in benthic composition, these results suggest that some degree of functional homogenisation (Clavel et al. 2011) exists between substrates with distinctly different community compositions. Results indicate that even on substrates with coral cover ranging from 34% 345 to 36%, which is high relative to the Caribbean region (de Bakker et al. 2016  low. In fact, daily rates are in a similar range to those for substrates dominated by bioeroding sponges where coral cover ranged from 1 to 9 % and to substrates covered by cyanobacterial mats where no live coral was recorded. This suggests that cementation/lithification processes carried out by coralline calcifying algae, micro-calcifiers (e.g. foraminifera and juvenile shells) and benthic microbial communities, resulting in the trapping and binding of rubble and sediment in cryptic habitats and 350 within/on the rubble, may play an important role on such impacted coral reefs, counteracting some of the dissolution in these communities and stabilising coral rubble into a consolidated reef framework.
The main differences between coral-dominated substrates and others, in terms of NCC, is that they can balance out dissolution at night, whereas other communities cannot. Primary production is barely compensating heterotrophic processes during the day on all substrates. Although substrates incubated in the present study are distinct in taxa dominance, they do share some 355 similarities that may be drawing biological function differences closer to each other between distinct benthic assemblages.
Indeed, turf covers any part of hard substrate available, the sand and rubble harbour a variety of potentially similar cryptic organisms and the microbial community within and above each of these substrates may be shifting similarly towards generalist copiotrophic populations. These shifts in community composition have resulted in the impairment of key reef functions that are usually attributed to conventional reefs, leading to functional homogenisation among these communities. It is noteworthy 360 that this functional homogenisation on degraded reefs might be the result of seasonality. Roth et al. (2020) recently found that summer temperatures amplified functional differences between coral-and algae-dominated communities in the central Red Sea. Higher temperatures benefit algae-dominated communities in terms of primary production and growth while coraldominated communities shifted towards a more heterotrophic state with depressed net community calcification rates. The fact that coral-dominated substrates studied here are already in a heterotrophic state with very low NCC values in winter 365 temperatures attests to the differences in the studied systems and provides an opportunity for comparison between a relatively healthy system and a degraded one (Roff and Mumby, 2012). However, further research looking at the seasonal component of the functionality of these degraded systems is essential to understand how they will respond to higher temperatures.

Conclusions
The combination of in situ incubations and inverse modelling taking into account the complexity of interactions between 370 processes, has proven to be an effective tool to provide quantitative data on the functional state of coral reef patches. Results acquired on this shallow Curaçaoan reef provide insight into the impact of habitat degradation and species composition shifts on reef functions. Remaining corals, although resilient, calcify at a slower pace than more specialist species (Acropora spp.) and cannot balance out heterotrophic processes from other functional groups. Coral presence does however contribute to counteracting dissolution processes at night, therefore acting as buffers to reef deconstruction. In the context of ongoing global 375 change, the environmental resilience of generalist species could be a determining factor of ecosystem stability (Clavel et al. 2011 (Pseudo)Diploria spp., Porites astreoides and Siderastrea sidereal) (de Bakker et al. 2019). Surprisingly, these are often found near areas which have locally suffered chronic stress from terrestrial sources (i.e. inflow, intense coastal development, factory 380 outflow) but often limited to areas covered by hard substrate and relatively less sand. Data on the processes underlying such developments however is virtually absent, but this could indicate that even the most severely degraded reefs could slowly regain essential functions when a critical adaptive capacity is reached.
Data and code availability Data and R code will be made available on request. 385 Author's contributions AW, DdB, SvH and LdN conceived the ideas and designed methodology; AW, DdB and TdC collected the data; AW and KS analysed the data; AW and LdN led the writing of the manuscript in consultation with DdB, SvH, FvD, KS and GR. All authors contributed critically to the drafts and gave final approval for publication.

390
Competing Interests The authors declare that they have no conflict of interest.    Fig. 2, replicate 2). Circles and crosses represent measured values of the state variables inside and outside the tent respectively. Fixed and fitted parameters used to run models are presented in bottom right boxes. Parameters escorted by the star sign * relate to significant parameter estimation (with standard error). The † symbol refers to non-significant parameter estimation. The remaining parameters are fixed. Note that the fixed mineralisation parameter in the left panel was estimated from the data of the same incubation at night Figure 6: Average process rates with respective 95% confidence interval inferred from observed concentration changes and model output in the tent enclosure on every substrate and during all incubation periods. 620 Table 1: Average net community production, net community calcification, nitrification, and denitrification fluxes in benthic communities with respective 95% confidence interval inferred from observed concentration changes in the tent enclosure on every substrate and during all incubation periods. White cells show day fluxes, while grey cells depict night fluxes. Crosses relate to 625 instances where no nitrification or denitrification was estimated by the model and the parameter was therefore fixed at 0.