The Bouraké semi-enclosed lagoon (New Caledonia). A natural laboratory to study the life-long adaptation of a coral reef ecosystem to climate change-like conditions

Abstract. According to current experimental evidence, coral reefs could disappear within the century if CO2 emissions remain unabated. However, recent discoveries of diverse and high cover reefs that already thrive under extreme conditions seem to contradict these projections. Volcanic CO2 vents, semi-enclosed lagoons and mangrove estuaries are unique study sites where one or more ecologically relevant parameters for life in the oceans are close or even worse than currently projected for the year 2100. These natural analogues of future conditions hold new hope for the future of coral reefs and provide unique natural laboratories to explore how reef species could keep pace with climate change. To achieve this, it is essential to characterize their environment as a whole, and accurately consider all possible environmental factors that may differ from what is expected in the future and that may possibly alter the ecosystem response. In this study, we focus on the semi-enclosed lagoon of Bouraké (New Caledonia, SW Pacific Ocean) where a healthy reef ecosystem thrives in warm, acidified and deoxygenated water. We used a multi-scale approach to characterize the main physical-chemical parameters and mapped the benthic community composition (i.e., corals, sponges, and macroalgae). The data revealed that most physical and chemical parameters are regulated by the tide, strongly fluctuate 3 to 4 times a day, and are entirely predictable. The seawater pH and dissolved oxygen decrease during falling tide and reach extreme low values at low tide (7.2 pHT and 1.9 mg O2 L−1 at Bouraké, vs 7.9 pHT and 5.5 mg O2 L−1 at reference reefs). Dissolved oxygen, temperature, and pH fluctuates according to the tide of up to 4.91 mg O2 L−1, 6.50 °C, and 0.69 pHT units on a single day. Furthermore, the concentration of most of the chemical parameters was one- to 5-times higher at the Bouraké lagoon, particularly for organic and inorganic carbon and nitrogen, but also for some nutrients, notably silicates. Surprisingly, despite extreme environmental conditions and altered seawater chemical composition, our results reveal a diverse and high cover community of macroalgae, sponges and corals accounting for 28, 11 and 66 species, respectively. Both environmental variability and nutrient imbalance might contribute to their survival under such extreme environmental conditions. We describe the natural dynamics of the Bouraké ecosystem and its relevance as a natural laboratory to investigate the benthic organism’s adaptive responses to multiple stressors like future climate change conditions.



135
seawater pHT (total scale), and salinity. A summary of the deployments is given in Supplementary Table S1. YSI dissolved 136 oxygen optical sensors were calibrated against zero, and 100 % saturated seawater at 25 °C. Two SeaFETs were calibrated 137 by the manufacturer, while the third was corrected before deployment by measuring its deviation from the two others in the 138 same seawater.   oscillations. To do this, all data were overlaid on a daily period and a tidal phase. First, we calculated a predicted tide for the 159 study area using the Nouméa harbour tide (50 km south of our study site) modified with coefficients from the Naval 160 Hydrographic and Oceanographic Service (SHOM; http://data.shom.fr ). The predicted tide was used to assign a semidiurnal 161 tidal phase (12 h) to each sampling time, and the data were averaged for each of these tidal phases. Similarly, the data were 162 averaged for each hour of the day (24 h). Because tides at sea are a sequence of sinusoidal harmonic components that are 163 different for each location, we performed a harmonic tidal analysis on the DO and pH data. We used the "UTide"-ut_solv() 164 tidal analysis package (Codiga, 2011) using the principal semidiurnal lunar constituent (M2), principal semidiurnal solar 165 constituent (S2), and solar diurnal constituent (S1). For each parameter, the amplitudes of the tidal harmonics M2 (12.4 h), S2 166 (12 h), and S1 (24 h) were calculated with a 95% confidence interval based on the 200 Monte-Carlo simulations.

175
In 2017, during three consecutive days (from May 31 th to June 02 nd ), seawater was sampled six times: twice during 176 both high and low tide, and one time at both rising and falling tide. In total, we sampled one reference station (R2), three

179
In 2019, during three consecutive days (from July 16 th to 18 th ), sampling was carried out every hour from 8 am to 3 180 pm. We sampled B1 and B2 on the first day, R1 on the second and R2 on the third day.

181
At each station and sampling time, pH and temperature were measured at the surface (0.5 m deep) using a portable pH-meter 182 (913, Metrohm) calibrated with TRIS buffer (Dickson lab, batch #T28). A subsample (50 mL) was filtered through 0.45-μm 183 Whatman TM Puradisc CA filters using a syringe and poisoned with 20 µL saturated HgCl2 to further measure AT. Two 20 mL 184 subsamples were analysed using an auto titrator (EcoTitrator, Metrohm), and AT was calculated from the Gran function.

185
Results were corrected against AT standards (A. Dickson, batch #155, Scripps, USA). The seawater carbonate parameters 186 pCO2, CO3 2-, and aragonite saturation state (Ωara) were then calculated from the pHT, AT, temperature, and mean salinity (35) 187 using the free-access CO2SYS package (Pierrot et al., 2006). Ammonium concentration was determined on a 40 mL subsample of unfiltered seawater, collected using a 60 mL 189 Schott bottle and stored in the dark. Samples were processed using a fluorimeter (Turner Designs) between six and 18 h after 190 two mL of OPA reagent (o-phthaldialdehyde) was added (Holmes et al., 1999).

191
The sampling of nutrients was performed using two replicate 20 mL polypropylene vials, rinsed three times using

195
Seawater samples for DIC were collected in two replicate glass vials (20 mL), filled with unfiltered water and 196 poisoned with 10 µL saturated HgCl2. The vials were immediately closed, the absence of bubbles was visually checked, and 197 the samples stored in the dark at room temperature for later analysis on a Shimadzu TOC-L analyser (Non-Dispersive Infrared, 198 NDIR). Typical analytical precision was less than ± 2 μmol kg −1 . The accuracy was verified using regular measurements of 199 reference material (CRM) from A. Dickson's laboratory.

200
Seawater samples for DOC were collected in two pre-combusted (4h at 450 °C) glass ampoules filled with water 201 filtered using a glass syringe filtration system (SGE TM ) with two pre-combusted 25-mm GF/F Whatman TM filters. Samples 202 were then acidified with ultrapure orthophosphoric acid (H3PO4), sealed, and stored in the dark at room temperature for later 203 analysis by high-temperature catalytic oxidation (HTCO) (Sugimura and Suzuki, 1988;Cauwet, 1994) on a Shimadzu TOC-204 L analyser. Typical analytical precision was ± 0.1-0.5 μM C (SD). Consensus reference materials 205 (http://www.rsmas.miami.edu/groups/biogeochem/CRM.html) were injected every 12 to 17 samples to ensure stable 206 operating conditions. DOC concentrations are only available for the 2017 sampling because of a sample's pollution in 2019.

207
Finally, one litre of unfiltered seawater was collected in a borosilicate glass bottle and stored on ice during sampling 208 for later measurement of POC and PON contents. In the lab, particulate matter was collected on pre-combusted (4h at 450 209 °C) Whatman TM GF/F filters using a Nalgene® vacuum system. The filters were dried at 60°C in the oven for 24 h and stored 210 in airtight glass vials at 4 °C in the dark until analysis on a CHN Perkin Elmer 2400.

211
All glass bottles and vials used were pre-combusted, washed with HCl solutions (10 %) and rinsed using milliQ water.

212
Seawater chemistry data were pooled by sampling area (R1, R2, Outer, Middle, Inner), and differences were tested using the 213 Kruskal-Wallis test followed the Conover Multiple comparison test (Benjamini-Hochberg-adjusted). We focused on the effect 214 of the tidal phases (i.e., falling and rising tide) on the seawater chemical composition in the Bouraké lagoon only, by attributing 215 each sample a tidal phase between 0 (high tide) and 6 h (low tide), and between 6 and 12 h (high tide). Multiple linear 216 regression was used to assess the adjusted R 2 and significance (p < 0.05) of the data from 0 to 6 h (falling tide) and from 6 to 217 12 h (rising tide) separately. Statistical analyses were performed using either Statistica® or R (version 3.2.4, R Core Team, 218 2018), the latter using the "FSA", "stats", and "Conover.test" packages. The benthic community and bottom substrate of the Bouraké lagoon, referred hereafter as biotic and abiotic descriptors, 221 respectively, were assessed in April 2018. Twenty-four 30 m-long geo-referenced transects (T1-T24) were laid in the lagoon 222 along the terraces' edge at similar depths (i.e. ~ 1 m), targeting coral dominated benthic assemblages. On each transect, a 0.5 223 x 0.5 m PVC quadrat was placed every meter, and a picture was taken with a waterproof photo-camera (Nikon AW 130) of the most common and identifiable sessile species. For each of the 835 pictures collected, we estimated the cover of abiotic 226 (i.e., mud, sand, rock, rubble, dead corals and unreadable) and biotic descriptors (i.e., branching, massive and soft corals, 227 sponges, macroalgae and "others") with photoQuad software both by automatic multi-scale image segmentation regions and 228 manual grid cell counts when necessary.

229
We used the photos of quadrats, the many other pictures collected during fieldworks, and laboratory morphological  (Veron, and Wallace, 1984;Wallace, 1999;Veron, 2000). Sponges were identified either based on their spicules' 241 morphological characteristics (i.e., length and width), or using a series of morphological descriptors (e.g., shape, size, colour, 242 texture, surface ornamentations, fibres) for species without spicules. In the lab, a subsample of the collected sponges were

372
At stations R1 and R2, pH was closely linked to tidal oscillations. It changed on average by about 0.1 pHT units and was 373 mostly dependent on the 24-hours cycle (S1 = 0.058 ± 0.004 and S1= 0.049 ± 0.007 pHT units for R1 and R2, respectively).

374
Simultaneous short-term pH measurements showed significant spatial differences (

376
There were also significant depth-related differences between shallow reefs and bottom water at stations R and S. In both 377 spatial and depth-related analyses, differences were approximately 0.05 to 0.1 pHT units, and we found the lowest values at  respectively. We overlaid all data on a single tidal phase of 12.4 h (Fig. 4c) and a 24-hours cycle (Fig. 4d). As with pH, the 388 mean diel DO was higher at the reference reefs than in the Bouraké lagoon. Mean DO values were 4.89 ± 1.18 and 5.23 ± 389 0.89 mg L -1 at B1 and B2, respectively, and 6.45 ± 0.95 and 6.48 ± 1.05 mg L -1 at R1 and R2, respectively.

390
At stations B1 and B2, DO was strongly correlated with the tidal cycle (82 and 72 % of the total DO variance were explained 391 by the tidal harmonic analysis, respectively). DO oscillations were mainly semidiurnal (M2 = 1.17 ± 0.08 mg L -1 , and M2 = 392 0.83 ± 0.09 mg L -1 for B1 and B2, respectively) with a substantial diurnal component (S1 = 1.12 ± 0.04 mg L -1 , and S1 = 0.681 393 ± 0.04 mg L -1 for B1 and B2, respectively). During a semidiurnal tidal cycle, DO was lower at low tide (3.7 mg L -1 and 4.6 394 mg L -1 at B1 and B2, respectively). During a 24-hours cycle, DO was lower in the early morning (4.0 and 4.3 mg L -1 at B1 395 and B2, respectively) and higher in the middle of the day (5.8 and 5.5 mg L -1 at B1 and B2, respectively). In a single day, we 396 recorded DO fluctuations of up to 6.37 mg L -1 at R1 and 4.91 mg L -1 at B2. The minimum DO value, 1.89 mg L -1 , was 397 measured during low tide, and the maximum DO value, 7.24 mg L -1 , was measured at B1during high tide (data not shown).

399
with lower values during the night (5.5 mg L -1 at both stations), and higher values in the middle of the day (7.8 mg L -1 , both 400 stations). Simultaneous short-term DO measurements ( Fig. 6; Table 1)

474
The reference reefs R1 and R2 had higher pH, CO3 2and Ωara, and consequently, lower DIC and pCO2 than the outer, middle

482
Most of chemical parameters were in general more concentrated (up to 5-times) in the Bouraké lagoon than on the reference 483 reefs R1 and R2, and increased from the outer to the inner reef ( Fig. 9a-d; Supplementary Fig. S2; Tables 1, 2).    limits. The abiotic substrate of cluster C1 was characterized by 66% mud, 12% rocks and 10% sand (see Supplementary Table   520 S4 for detailed cover data per transect). Only a few branching corals (<10%) were found, but species richness was relatively 521 high (6 at T9 and 10 at T10). Cluster C2 was distinct, which is not surprising due to its location in a relatively shallow 522 convergence zone that divides the lagoon into two parts. There, the substrate is made of coarse sand (13%) and rocks (12%) 523 and is mainly colonized by macroalgae and sponges (31% and 32%, respectively). Species richness in the area was 524 heterogeneous and ranged from 4 to 12. Dictyota spp. and Halimeda discoidea were the main macroalgal species, while 525 Rhabdastrella globostellata was the dominant sponge species in the area. Cluster B1, located on the outer reef, is characterized 526 by an abundance of soft corals (48%) and rubbles (21%), and high biological richness (Fig. 10a). Cluster C3 is characterized 527 by coarse sand (49%), rocks (17%) and a few benthic organisms such as macroalgae (10%) and soft corals (8%).

558
providing an opportunity to assess the resilience of organisms and to study their adaptive mechanisms in a natural 559 environment. Coral reefs, that are exposed to seawater pH and temperature values close to or even worse than those expected 560 for the future, have likely developed physiological trade-offs and expressed molecular changes that allow them to survive 561 sub-optimal, climate change-like conditions. When using these natural laboratories to predict species responses to future 562 environmental conditions, it is essential to assess a multi-scale approach that incorporates the spatial and temporal variability 563 of the key physical and chemical parameters characterizing the study site. In this study, we mapped the spatial and temporal

580
We found several marked differences in the environmental conditions between the Bouraké lagoon and the reference reefs.

581
First, the seawater temperature is higher in summer in the Bouraké lagoon (Fig. 3)  with what should be expected from reef metabolic activities and daily cycles but, in the Bouraké lagoon, these parameters,

601
including salinity (Fig. 7), are entirely driven by tides. Here, seawater pH and DO varied between extremely low values at 602 low tide and close-to-normal values during high tide (see also Fig. 8a-c for pH). Finally, we found that the timing of the tide 603 was out of phase between sites, with a delay of about 45 minutes at high tide and 1.5 h at low tide in the Bouraké lagoon ( Fig.   604 2).

605
The unique environmental conditions measured in the Bouraké lagoon are mainly due to the topographical and 606 geomorphological characteristics of this unique site, the resulting water circulation and the direction of the tide. At each rising 607 tide, new water enters through the channel, flows into the semi-enclosed lagoon towards the large mangrove area behind it, 608 and mixes with the acidic, warm and deoxygenated water that was already in the system and the mangrove area. There, we 609 hypothesize that the water chemistry changes due to the metabolic reactions in the sediments, coral reefs and mangrove roots.

610
Conversely, on a falling tide, the seawater becomes gradually more acidic, hot and oxygen-depleted because the water that 611 resided in the mangrove area gradually drains out of the system. This takes about 6 hours, during which the vast reservoir of 612 shallow mangrove water continues to be chemically altered, becoming increasingly acidic, oxygen-depleted and hot. As a 613 result, we measured a significant spatial differences in pH between the outer reef (the entry of the lagoon) and the inner reef 614 (near the mangrove forest), as well as a considerable delay in the synchronization of the tidal shift (Fig. 5b). Interestingly, 615 because the volume of seawater discharged in 6 hours is so large, it affects also the area outside the system where we measured

620
It could be argued that because the fluctuations are linked to tidal phases, organisms living in the Bouraké lagoon may benefit 621 from periods of normal conditions at high tide during which they can recover from the stress they have experienced at low 622 tide. While this could be partially the case for species living on the outer reef, close to the main lagoon, the environmental 623 conditions inside the Bouraké lagoon rarely reach normal values (Fig. 8a-c), and also persist longer since the low tide is 624 delayed by 1.5 h compared to the reference reef (Fig. 2).

625
Preliminary results from a hydrodynamic model of the study site also suggest that tide-associated water mass movements are 626 spatially heterogeneous and likely to play an essential role in shaping coral resilience to extreme conditions (see discussion 627 below). Indeed, one can imagine a single water mass moving with the same physical characteristics from the mangrove area 628 towards the outer reefs or in the opposite direction depending on the tide. However, the complex geomorphology of the probably change the seawater physical and chemical properties. We measured significant spatial differences in pH within 631 each reef area (inner, middle and outer reefs; Fig. 5a, b), as well as throughout the water column (i.e., between the surface 632 and the bottom; Fig. 5d,e). In general, bottom seawater was 0.1-0.2 pHT units lower than the surface, probably due to a 633 pumping mechanism by the water mass of more acidic pore water from the sediments. The pH also differed spatially within

639
We can assume that, throughout the Bouraké lagoon, organisms are exposed to extreme and fluctuating suboptimal physical 640 parameters, such as pH, and DO which are more pronounced on the bottom and last longer, and with more extreme values,

641
on the inner reefs. It is also true for the seawater chemistry with higher concentrations in the Bouraké lagoon than on the 642 reference reefs (see Table 2). For instance, we found that orthosilicic acid, phosphate, dissolved and particulate organic

676
Remarkably, the two coral species currently considered New Caledonia endemics thrive in the innermost benthic assemblages 677 of the Bouraké lagoon, making it not only a unique natural laboratory but also a potential conservation priority site in light of 678 its overall uniqueness characterized in this study. In the Bouraké lagoon, benthos species richness was very high throughout . However, our study shows that although the Bouraké system can reach conditions close to hypoxia (< 3 mg L -1 ; Fig.   734 4), species seem to have promoted compensation mechanisms that allow them to survive in these conditions. The natural 735 laboratory of Bouraké, where DO fluctuates with the tide, in combination with other environmental stressors, offers a perfect 736 setting to test the practically unknown effects of deoxygenation and hypoxia thresholds in reef-building corals exposed to acid

738
Besides the hypothesis that environmental variability improves the metabolism of organisms, particularly their resilience to 739 extreme conditions, a series of other physical and chemical parameters in the Bouraké lagoon may work in combination to 740 offset or enhance these effects. Mangrove habitats are highly productive ecosystems and are sites of intense carbon processing, In the Bouraké lagoon, benthic communities might have access to a range of heterotrophic inputs, nutrients, carbon and 743 nitrogen sources. These sources can be metabolized by the species to increase their energy budget and cope with the 744 suboptimal parameters, but they can also become toxic, if too concentrated, or depleted, leading to functional limitations. We 745 measured particularly high concentrations of organic and inorganic carbon and nitrogen, but also of some nutrients, notably 746 silicates and phosphorus, and we confirmed the potential contribution of the mangrove in those inputs, especially during the 747 falling tide (Fig. 8 and 9; Supplementary Fig. S1 and S2;

753
We also found that nutrients could partially explain the distribution of organisms throughout the Bouraké lagoon (Fig. 11).

773
Although these data should be considered with caution, they suggest that the Bouraké lagoon seawater is not limited in

803
We used a multi-scale approach to characterize the physical and chemical environmental parameters of one of the most 804 suitable natural analogue for future environmental conditions, the semi-enclosed lagoon of Bouraké (New Caledonia), and 805 accurately map its benthic community for the first time. We studied several physico-chemical parameters such as pH, 806 dissolved oxygen, temperature, and salinity, but also nutrients and organic matter and we found that: i) they fluctuate between 807 low and high tides, ranging from suboptimal-extreme to near normal values; ii) although predictable according to the tide, 808 they differed spatially, and iii) suboptimal values persisted longer and were more acute on the inner reef. Our data clearly 809 show that fluctuations are: i) predictable, at least for some (pH, DO, temperature and salinity) of the physical parameters for 810 which we have enough data; ii) mainly driven by the tide; and iii) that seawater nutrient imbalance and organic inputs increase 811 during the falling tide and originate from the mangrove forest and associated sediments. Although several studies suggest that 812 ocean acidification, warming and, to some extent, deoxygenation will lead to a reduction in biodiversity, increase in bleaching 813 and reef dissolution, in the Bouraké lagoon, we found a rich and healthy reef with high coral cover and species richness, but 814 also sponges and macroalgae (including CCA). It was beyond the scope of this already multidisciplinary study to assess the 815 contribution of environmental variability and nutrient imbalance to the organism' stress tolerance under extreme conditions.

816
However, both coexist in the Bouraké lagoon, and we believe there is evidence of their contribution to the survival of 817 organisms to future-like environmental conditions. Our study provides evidence that this is possible in nature, giving a