High rates of carbon and dinitrogen fixation suggest a critical role of benthic pioneer communities in the energy and nutrient dynamics of coral reefs

were also active sites of N 2 fixation. While these laboratory experiments indicate an important role of diazotrophs on apparently ‘bare’ substrates after coral mortality, in-depth investigations targeting the significance of pioneer communities in their natural environment on the C and N dynamics of a reef are still lacking. Here we investigate C and N dynamics during the succession of pioneer communities in a coral reef in the central Red Sea. For this, rough carbonate tiles were deployed to simulate the occurrence of bare substrates after disturbance events. Primary productivity and N 2 fixation of the colonizer communities were measured


| INTRODUC TI ON
Coral reefs display some of the highest rates of gross primary production (GPP) in the marine environment, despite a low ambient nutrient availability and little sustained exogenous nutrient inputs (Hatcher, 1990). The high rates of productivity are primarily attributed to the efficient retention and recycling of nutrients and substantial trophic transfer of energy by coral reef communities (De Goeij et al., 2013;Muscatine & Porter, 1977;Odum & Odum, 1955;Wild et al., 2004). In addition, diazotrophy, the metabolic ability to fix atmospheric dinitrogen (N 2 ) into a biologically useful form (i.e. ammonia), plays an essential role in replenishing the nitrogen (N) pool and sustaining net ecosystem production and growth (Cardini, Bednarz, Foster, & Wild, 2014;O'Neil & Capone, 2008).
In healthy coral reefs, scleractinian corals comprise a major component of the benthic 'nutrient capacity', mediating the fluxes of carbon (C) and N between benthic and pelagic compartments of the ecosystem (Allgeier, Burkepile, & Layman, 2017;Wild et al., 2004Wild et al., , 2011. In fact, corals contribute 41%-76% to benthic GPP budgets, and 10%-14% to benthic N inputs by N 2 fixation during nutrient-depleted conditions, suggesting a strong biogeochemical coupling between diazotrophy and the reef C cycle . Thereby, the efficient transfer of energy and nutrients is facilitated by corals mainly via the production, release and subsequent remineralization of coral mucus to other reef organisms and trophic levels (Coffroth, 1990;Wild et al., 2004Wild et al., , 2005 but also through corallivory (reviewed in Rotjan & Lewis, 2008), or decay.
Particularly, recurrent bleaching, as a recent phenomenon mainly triggered by global warming, can inflict a widespread and rapid loss of live coral cover (Hughes, Anderson, et al., 2018;Hughes et al., 2017;McClanahan et al., 2019).
The sudden change in the benthic community structure arising from bleaching-induced coral mortality implies the loss of critical ecosystem functions conveyed by corals, including nutrient provisioning and regeneration (Holmes & Johnstone, 2010 Whether coral mortality occurs naturally on a small scale or in the form of mass mortality events after severe (anthropogenic) stress, the surface area of bare hard-bottom substrates, that is, the carbonate structures and skeletons from dead corals, is available for (re)-colonization. These substrates are quickly covered with biofilms and colonized by diverse benthic organisms, including heterogeneous assemblages of filamentous algae, crustose coralline algae (CCA), but also sessile invertebrates and coral larvae (reviewed in Norström, Nyström, Lokrantz, & Folke, 2009). Constant disturbances and small-scale differences in environmental conditions result in a mosaic of community stages at varying stages of succession. Consequently, coral reef community composition can be highly heterogeneous resulting in the co-occurrence of pioneers as well as mature coral communities but their relative abundance may change as a result of coral mortality and colonization. Thus, the existing literature mainly focuses on recruitment and succession patterns of these 'pioneer communities' to predict trajectories of recovery (or shifts to alternate states; e.g. Burt, Bartholomew, Bauman, Saif, & Sale, 2009;Humanes & Bastidas, 2015;Jessen, Voolstra, & Wild, 2014;Roth, Stuhldreier, Sánchez-Noguera, Morales-Ramírez, & Wild, 2015;Stuhldreier, Bastian, Schönig, & Wild, 2015). Yet, a study by Davey, Holmes, and Johnstone (2008) indicates that skeletons of artificially bleached corals were also active sites of N 2 fixation. While these laboratory experiments indicate an important role of diazotrophs on apparently 'bare' substrates after coral mortality, in-depth investigations targeting the significance of pioneer communities in their natural environment on the C and N dynamics of a reef are still lacking.
Here we investigate C and N dynamics during the succession of pioneer communities in a coral reef in the central Red Sea. For this, rough carbonate tiles were deployed to simulate the occurrence of bare substrates after disturbance events. Primary productivity and N 2 fixation of the colonizer communities were measured organic C was exported into the environment, constituting a vital source of energy for the coral reef food web. 5. Our findings suggest that benthic pioneer communities may play a crucial, yet overlooked role in the C and N dynamics of oligotrophic coral reefs by contributing to the input of new C and N after coral mortality. While not substituting other critical ecosystem functions provided by corals (e.g. structural habitat complexity and coastal protection), pioneer communities likely contribute to maintaining coral reef nutrient cycling through the accumulation of biomass and import of macronutrients following coral loss. monthly over 1 year. C and N elemental, as well as stable isotope analyses complemented community-wide flux measurements to (a) quantify rates of primary production and N 2 fixation of benthic pioneer communities, (b) investigate the effects of temporal variability of key environmental variables and (c) gain insights into element acquisition (storage in biomass) and regeneration (export or supply to the environment) of pioneer communities in coral reefs after disturbance.

| Study site and environmental conditions
The experiments were carried out at Abu Shosha reef located in the central Red Sea on the west coast of Saudi Arabia (22°18ʹ16.3ʺN; 39°02ʹ57.7ʺE). All experiments described below were repeated monthly from January 2017 until December 2017 (duration = 52 weeks).
Prior to the start of the experiments, the relative benthic cover was assessed by line-point-intercept surveys along three consecutively placed 25 m line transects, with a 5-m inter-transect distance (Hill & Wilkinson, 2004). The benthic composition was determined at 50 cm intervals for a total of 50 data points per transect. The relative benthic cover of the main functional groups was as follows (mean ± SE): Filamentous turf algae = 36.8 ± 4.8%; hard corals = 28.7 ± 6.6%; coral rubble = 10.2 ± 3.1%; biogenic rock = 8.7 ± 3.3%; soft corals = 8.5 ± 0.2%; reef sediments = 6.0 ± 2.4%; Tridacna sp. = 0.7 ± 0.3%; and macroalgae 0.4 ± 0.1%. Reef rugosity (as a measure of the structural complexity) was assessed by the chain method (Frost, Burrows, Johnson, Hanley, & Hawkins, 2005) of the same transects described above and averaged 1.90 ± 0.26. Abu Shosha reef shows a high standing biomass of herbivorous fish (>150 g/m 2 ), as surveyed by Roth et al. (2018). We performed additional fish surveys around experimental aluminium frames to explore whether the introduced structures attracted fish to the area. No differences in both the abundance and community structure were observed around experimental frames as compared to fish surveys in the wider reef (see Table S1).
The area is characterized by strong seasonal fluctuations of environmental variables throughout the year. Key environmental variables were routinely monitored at the sampling site. Water temperature (T) was measured continuously (logging interval = 30 min) for 1 year with data loggers (Onset HOBO Water Temperature Pro v2 Data Logger-U22-001; accuracy: ±0.21°C). Light availability was measured continuously (logging interval = 1 min) for three full days per month with a data logger (Onset HOBO Pendant UA-002-64; spectral detection range 150-1200 nm). Light readings in lux were converted to photosynthetically active radiation (PAR; μmol quanta m −2 s −1 ; 400-700 nm wavelengths) using the following approximation: 1 μmol quanta m −2 s −1 = 51.8 lux. This conversion factor was obtained by inter-calibrating the lux readings with data obtained from a parallel deployment of a PAR sensor (LI-COR LI-192S quantum sensor) for 4 hr of daylight. Both readings showed a linear correlation (r 2 = 0.92), and the obtained conversion factor of 51.8 was similar to 52.0 reported by Valiela (1984). Salinity was measured during each day of sampling with a conductivity measuring cell (TetraCon ® , 925, WTW, accuracy: ±0.5% of value).
Seawater samples for the determination of dissolved nitrate (NO − 3 ), nitrite (NO − 2 ) and phosphate (PO 3− 4 ) were taken in triplicates each month from 1 m above the reef substrate with 60 ml syringes.
On the boat, samples were filtered immediately through syringe filters (Isopore TM membrane filters, 0.2 μm GTTP) into acid-washed 15 ml centrifuge tubes and stored dark and cool for transportation.

| Experimental design
Eight anodized aluminium frames (50 × 50 cm) were haphazardly deployed onto the reef at a depth of 5 m ( Figure 1a). The frames were secured to the substrate with weights, keeping a distance of 1.5-2.5 m between frames. Each of the frames was equipped with 12 rough, untreated, carbonate tiles. Each tile was about 8 × 8 cm and had a mean (±SE) planar surface area of 64 ± 0.8 cm 2 . Rough carbonate tiles were used, as their surface mimics coral rock and enhances natural species richness and biomass compared to other artificial substrates (Kennedy, Ordonez, Lewis, & Diaz-Pulido, 2017;Mallela, Milne, & Martinez-Escobar, 2017). In addition, previous studies assessed whether artificial substrates affect the abundance and composition of recruits on tiles as compared to natural reef substrata. As spatial patterns of variation and the relative abundance of organisms on artificial and natural substrata do not significantly vary when using rock-like tiles (e.g. Penin et al., 2010) such as the ones used in the present experiments, we conclude that the use of carbonate settlement tiles did not bias the results presented here.
Tiles were installed at a 45-degree angle relative to the substrate on a tough plastic net fixed between the vertical poles to reduce excess sedimentation. Monthly, randomly selected settlement tiles were collected from each frame for further analysis. In total, 96 settlement tiles were analysed.

| Quantification of primary production and N 2 fixation
Every month, eight settlement tiles were incubated with surrounding seawater to quantify rates of net primary production (NPP), dark respiration (R) and N 2 fixation under natural water and light conditions. Incubations for NPP and R took place immediately after collection inside separate 1,000 ml gas-tight glass jars that were placed on the seafloor in the direct vicinity to the sampling location. Incubations always started at 10:30 a.m. One additional incubation jar containing only seawater served as a control for planktonic background activity.
Another jar was closed with the others and immediately brought to the surface to measure the initial concentration of dissolved oxygen (DO) using an optical DO probe (FDO ® , 925, WTW, temperature and salinity corrected; accuracy: ±0.5% of value). The incubation jars for R were covered with thick aluminium foil to measure O 2 fluxes in the absence of photosynthesis. After 60-70 min incubation time, all glass chambers were transported to the boat, and the concentration of DO was measured in the incubation medium of each chamber. Readings were recorded after gently stirring the seawater to ensure homogeneous DO distribution. After DO measurements, tiles were left in the jars and transported to the laboratory, where they were kept in an incubator at in situ temperature and light conditions until the beginning of the acetylene (C 2 H 2 ) reduction assay (ARA) starting in the evening of the same day.
Acetylene reduction assay was used to measure the activity of the unspecific N-fixing enzyme nitrogenase (Stewart, Fitzgerald, & Burris, 1967) to approximate N 2 fixation rates (Capone, Taylor, & Taylor, 1977;Shashar, Feldstein, Cohen, & Loya, 1994). For the assay, the eight settlement tiles (i.e. originating from the NPP and R incubations in the field) were transferred to 1 L gas-tight glass chambers for subsequent incubations. In addition, duplicate seawater control chambers were incubated to correct for planktonic N 2 fixation. Each chamber contained 720 ml of seawater collected from the field on the same day and 80 ml of C 2 H 2 -enriched seawater. Of the 200 ml air headspace, 10% was replaced with C 2 H 2 gas. About 2.5 ml gas samples were collected at the beginning from the gas headspace with a glass syringe and injected into vacuum glass tubes (COVIDEN, Monoject F I G U R E 1 Experimental setup at Abu Shosha reef (a), environmental background variables during the study period (b), exemplary pictures of the settlement tiles 1, 4, 8 and 12 months after deployment (c) and the respective proportional cover of functional groups on settlement tiles (d). Aluminium frames with settlement tiles were haphazardly deployed directly onto the reef at 5 m water depth (a). Communities developing on settlement tiles showed a gradual development from filamentous turf algae towards more encrusting algae and sessile invertebrates (details in c and d). In (b), photosynthetically active radiation (PAR, in µmol photons m −2 s −1 ) values are presented as daytime averages of three separate days per month ± SE. Dissolved inorganic nitrogen (DIN, in µmol/L) values are averages calculated from three discrete water samples per month ± SE. Seawater temperature was measured continuously, and values represent the monthly mean ± SE. Lines in (b) connect the means. Abbreviations: mo, months. Picture in (a) adapted from Roth et al. (2018). Scalebar in (c) = 0.5 cm Blood collection tubes). Subsequently, glass chambers were placed on inductive drive stirrers (500 rpm) in laboratory incubators set to in situ temperature and light conditions. Conditions in the incubators followed a 12 hr light/dark cycle. Additional gas samples were drawn after 12 hr to quantify ethylene (C 2 H 4 ) production during night-time and after 24 hr, for the full 24 hr period respectively. C 2 H 4 concentrations in the gas samples were determined using a gas chromatograph with flame ionization detector (Agilent Technologies 7890A) and a GS-GasPro column (Agilent J&W GC Columns, length 30 m, diam. 0.320 mm, Agilent Technologies) with an LOQ = 7.41 ppm.

| C and N elemental and stable isotope analysis
After the incubation experiments, all tiles were rinsed with freshwater to remove mobile invertebrates, sediments and salt. All sessile organisms were scraped off with razor blades and collected on precombusted, pre-weighed tinfoil. Tinfoil packages with content were dried at 40°C for 48 hr and subsequently weighed again to derive the total dry mass of material from each settlement tile. Dried samples were homogenized with an agate mortar and pestle. Subsamples were weighed and transferred into tin capsules for bulk C and N isotope analysis. For 13 C org analysis, dried subsamples were weighed into silver capsules and decalcified by dripping hydrochloric acid (2N HCl) directly onto the sample. After completion of the decalcification process, samples were re-dried at 40°C for 24 hr. Isotopic ratios for 13 C org / 12 C org and 15 N/ 14 N signatures, and total C org and N contents were measured using an elemental analyser (Thermo Flash EA 1112) coupled to a stable isotope mass spectrometer (IRMS, DELTA V Advantage). C org and total N contents (given in weight %) from the subsamples were scaled back up to the entire tile by multiplying the percentage to the total dry mass of all the material on the tile. Afterwards, values were converted from grams to moles and normalized to the surface area of each tile (i.e. expressed as μmol C/cm 2 and μmol N/cm 2 ) to achieve standardization with the C and N 2 fixation measurements. C and N stable isotope ratios were expressed by the delta (δ) notation in units per mil (‰) and calculated as: δ 13 C or δ 15 N = (R sample /R ref − 1) × 1,000, where R sample is the ratio of 13 C/ 12 C or 15 N/ 14 N in the sample, and R ref is the heavy/light isotope ratio of the reference material (C/R ref = 0.01118, Vienna Pee Dee Belemnite; N/R ref = 0.00368, atmospheric N 2 (Mariotti, 1983). The standard deviation for repeated measurements of laboratory standard material (peptone) was <0.15‰ for N and C respectively. Standard deviations of concentration measurements of replicates of the laboratory standard were <3% of the concentration analysed.

| Data exploration and analyses
Daytime averages were calculated for seawater temperature and light availability from continuous measurements. NPP and R for each incubated tile were derived from DO concentration differences in the incubation medium calculated by subtracting the start from the end concentration. Results were corrected for DO concentration differences measured in controls and incubation water volume, and normalized to incubation time and 2D surface area of the settlement tile. GPP was calculated by adding positive R rates to their corresponding NPP rates (GPP = NPP + |R|). The biological C fixation was calculated by the amount of oxygen released/taken up during photosynthesis/respiration as a proportion of the volume of carbon dioxide (CO 2 ) used in that process with a photosynthetic quotient of 1.04 and respiratory quotient of 0.96, as previously established for turf algae and coral rock (Carpenter & Williams, 2007). Hourly rates from each incubation were extrapolated to daily rates (12 hr for NPP, 24 hr for R, and GPP = NPP + |R|), resulting in daily NPP, R and GPP rates in µmol C cm −2 day −1 . We acknowledge a small overestimation of NPP for the first and last months of the experiment and a small underestimation during the middle part of it by not incorporating the seasonal change of the day length. However, N 2 fixation experiments were run in the laboratory in incubators set to 12:12 hr (dark:light) throughout the study.
Thus, N 2 fixation rates cannot be re-calculated. For a better comparability between the two processes (C and N fixation) and the subsequent calculation of the photo-metabolic nitrogen demand (PND; see details further below), we refrain from only adjusting the C fluxes.
We explored the relationships of GPP and N 2 fixation with environmental variables. Details for the model selection process are in the Appendices S1 and S2; Figures S1-S5; Tables S3-S6.
The photosynthetic C assimilation is known to vary with environmental temperature due to the temperature-dependent activation and deactivation of the carbon-fixing enzyme Rubisco (Farquar, Von Caemmerer, & Berry, 1980). We, therefore, used a modified Arrhenius equation (Leuning, 2002) to fit the measured temperature dependence of GPP.
where H a represents the activation energy, H d is the energy for deactivation, S v is a constant related to entropy, T is the ambient temperature, T ref is a reference temperature of 27°C, R is the gas constant, and f is a scaling factor to convert to units of µmol cm −2 day −1 . For parameter values see Table S5. N 2 fixation rates were calculated based on C 2 H 4 concentration differences in the headspace between the start and endpoint of the incubation. The amount of C 2 H 4 absorbed in the incubation water was accounted for by Bunsen gas solubility coefficients of 0.072-0.082 of ethylene, depending on the temperature and salinity during the incubations (Breitbarth, Mills, Friedrichs, & LaRoche, 2004). To convert the temporal evolution of C 2 H 4 to N 2 fixation rates, a conservative theoretical ratio of 4:1 (C 2 H 4 :N 2 ) was used, which assumes that 4 mol of C 2 H 4 are reduced per 1 mol of N 2 (Mulholland, Bronk, & Capone, 2004;Stal, 1988). As appropriate conversion factor may differ depending on substrate and environmental conditions (Capone, 1988;Capone et al., 2005;Graham, Hamilton, & Campbell, 1980;Larkum, Kennedy, & Muller, 1988), we applied a conservative theoretical ratio to allow direct comparisons to other studies in similar environments Cardini et al., 2016). All N 2 fixation rates were corrected for the respective seawater controls and normalized to incubation time and 2D surface area of the settlement tile, and are expressed as nmol N cm −2 day −1 .
We estimated the importance of N 2 fixation derived N to the PND Cardini et al., 2014;. PND is a function of the atomic C org /N ratio in phototrophs, reflecting the amount of N assimilated relative to C org fixed (e.g. Raven, Handley, & Andrews, 2004). Thus, we calculated PND using the C org /N ratios from the C and N elemental analyses of the scraped off material from settlement tiles, as: with PND = the theoretical amount of N required to supplement fixed carbon from net photosynthesis for the incorporation into biomass (in μmol N cm −2 day −1 ); NCP = net community production (in μmol C cm −2 day −1 ); and C org /N = the atomic ratio of organic carbon to nitrogen in the biomass on settlement tiles. As seasonal variations resulted in corresponding variations in C org /N ratios, the PND was calculated for each month and tile individually. By considering the actual biomass elemental properties of the communities on our settlement tiles, we avoided using theoretical C org /N ratios (such as the Redfield ratio of 106 C org :16 N) that can vary in time, space and among species (Redfield, 1958). Subsequently, we calculated the contribution (in %) of N 2 fixation derived N to the PND of each tile and timepoint, as: where N 2 fix is the amount of N fixed (in μmol N cm −2 day −1 ) by diazotrophy on settlement tiles.
Finally, we explored how much of the C org fixed by NCP was exported/lost from the settlement tiles to the environment. This export can be due to, among other processes, grazing (Fong & Paul, 2011), organic matter exudates (Haas et al., 2010;Quinlan et al., 2018) or by the decay of biomass (Duarte & Cebrián, 1996). As we did not measure these 'losses' directly and as individual processes, we subtracted the absolute gains of C org from one month to the next (i.e. calculated as the differences between C org contents of tiles between two consecutive months) from the theoretical gains of C org that could have been accumulated by NCP in the respective timeframe (i.e. if all C org fixed by NCP would have remained on the settlement tiles). We accounted for the losses of C by respiration using only daytime NCP rates, and subtracting night-time R. The calculation can be summarized as: with NCP = hourly net community production (μmol C cm −2 hr −1 ); R = hourly respiration (in μmol C cm −2 hr −1 ); d = the number of days from one to the following sampling event; and C org(a) and C org(b) are the organic carbon contents of biomass of one month (i.e. a) and the following month (i.e. b) respectively. where V max is the maximal rate of dinitrogen fixation and K i is the half-saturation concentration for the inhibition by NO − 3 . For parameter values see Table S6.
In addition, we explored whether the cover of benthic functional groups influenced GPP or N 2 fixation rates using multiple regression analysis with GPP or N 2 fixation as the response variables, respectively, and the % cover of all functional groups on settlement tiles individually as predictor variables.

Statistical analyses were conducted in R v3.2.2 (R Development
Core Team, 2015) and JMP © Pro14 (SAS Institute) statistic software. Models for GPP and N 2 fixation were programmed in Python, with details on packages and algorithms in the Supporting Information.

| Environmental conditions
All monitored environmental variables exhibited a strong temporal variability (details in Table S2; Figure 1b). Mean monthly water temperature ranged from 24.6 ± 0.1°C in February to a maximum of 32.2 ± 0.1°C in August. DIN was lowest in April (0.29 ± 0.01 μmol N/L) and highest in September (1.35 ± 0.12 μmol N/L), while PO 3− 4 was lowest in January and highest in July (0.02 ± 0.01 and 0.18 ± 0.02 μmol PO 3− 4 L −1 respectively).

| Succession of benthic pioneer communities on settlement tiles
Bare tiles were gradually inhabited by various benthic organisms ( Figure 1d). Within 4 weeks, over 50% of the tile surfaces were covered by filamentous turf algae (29%), green and brown crustose algae (18%), as well as CCA (7%). Filamentous turf algae proliferated for the following 2-4 months, covering more than 50% of each tile, while other biotic functional groups did not increase in their occurrence.
After 4-6 months from the beginning of the experiment, the availability of bare space on the settlement tiles halved and remained at 20%, while crustose algae and CCA progressively replaced filamentous turf algae. In addition, calcifying invertebrates (1%-5%), such as barnacles, polychaetes, but also coral spats and recruits were observed more frequently. Representative pictures of the settlement tiles 1, 4, 8 and 12 months after deployment are shown in Figure 1c.

| Biomass accumulation of C and N
Total C org and N accumulated gradually on settlement tiles plateauing after 6-8 months (Figure 2a,b). Total C org increased by 32% monthly, reaching a maximum of 52.4 ± 8.1 μmol C/cm 2 in September. Likewise, total N increased by 30% per month, until reaching a maximum of 4.0 ± 0.7 μmol N/cm 2 in August. The concomitant increase in elemental C org and N resulted in stable C org /N ratios of biomass (11.6 ± 0.4) throughout the study period. A slight but significant drop of the C org /N ratio to 7.7 ± 0.5 was observed in June (Figure 2c).

| Primary production and N 2 fixation activity
Within 4-8 weeks after deployment, benthic pioneer communities reached a state in which they exhibited high rates of C and N 2 fixation, with strong temporal variations (Figure 3a,b).
Gross primary production averaged 17.6 ± 0.5 μmol C cm −2 day −1 , with highest rates in June and lowest rates in February (24.7 ± 0.9 and 12.9 ± 2.1 μmol C cm −2 day −1 , respectively). Overall, productivity was driven by NPP; daily community R contributed 20%-30% to GPP only, resulting in a stable GPP/R ratio of 3.9 ± 0.2 throughout the study period. GPP was temperature-dependent and increased with increasing temperature until it reached an optimum temperature at 30.5°C. Above 30.5°C, GPP declined with rising temperature (Figure 4a). Multiple regression analysis revealed a slight correlation between the cover of functional groups on settlement tiles and GPP (r 2 = 0.15 and p = 0.0485). However, this significant correlation was only attributed to the relative contribution of bare substrates on tiles (p = 0.0176), while the per cent cover of all other functional groups were no significant predictors of GPP (Table S7).
N 2 fixation rates were highly variable, ranging from 8 to 54 nmol N cm −2 day −1 (Figure 3b). NO − 3 concentration had a strong inhibitory effect on the rates of N 2 fixation (Figure 4b (Table S8).

| Estimating regeneration rates of carbon from benthic pioneer communities
The export of C org (C org produced, but not consumed or incorporated by the community itself) from the settlement tiles was derived by subtracting the C org acquisition rates (elemental analysis) by the C fixation rates (incubations) by primary production. On an average, 10.2 ± 0.5 μmol C cm −2 day −1 , comprising almost 60% of the photosynthetically fixed C, was not incorporated into biomass of the communities, implying that it was released into the seawater or consumed by higher trophic levels outside these pioneer communities.
The rates of this export/loss to outside of the pioneer communities fluctuated between months ( Figure 5). Interestingly, however, already at the first measurement 4 weeks after deployment, 57% (8.0 ± 0.9 μmol C cm −2 day −1 ) of the fixed C was exported/lost into the environment, hence, only limited biomass accumulation took place even during the initial settlement phase. Oct Dec F I G U R E 4 Nonlinear relationships of gross primary production (GPP) with temperature (a), and dinitrogen fixation with environmental nitrate concentrations (b). The temperature dependence of GPP is simulated with a modified Arrhenius curve (Equation 1) for the activation and inactivation of the carbon assimilating enzyme Rubisco (Farquar et al., 1980;Leuning, 2002). The nitrate dependence of N 2 fixation is simulated with an inverse Michaelis-Menten equation (Equation 5) 1.0

| D ISCUSS I ON
The data presented here demonstrate that benthic pioneer communities play a critical role as a trophodynamic component of coral reefs.
Organisms establishing on bare substrates can produce considerable quantities of C org and N within only weeks after the initial settlement.
While the biomass accumulation of pioneer communities is limited, the export of organic material to the ecosystem is substantial and consistent over time, despite significant variations in the rates of assimilation that are influenced by environmental conditions. Ultimately, pioneer communities may serve as an important source of C and N through which the wider coral reef ecosystem may acquire (limiting) elements. Under conditions where pioneer communities colonize available space previously occupied by corals (e.g. after mass coral mortality following prolonged bleaching events), C and N by pioneer communities may help during ecosystem reorganization and recovery.

| High rates of primary productivity and N 2 fixation in benthic pioneer communities
Pioneering functional groups on settlement tiles were, as commonly observed, mainly short (<2 mm) filamentous turf algae, crustose algae and CCA (e.g. Adjeroud et al., 2009;Jessen et al., 2014;Mellin, Aaron Macneil, Cheal, Emslie, & Julian Caley, 2016). Usually, these consortia of epibenthic algae belong to the most productive functional groups of benthic producers (Hatcher, 1988). Already 4 weeks after the introduction of bare substrates, epilithic autotrophs on settlement tiles displayed high rates of GPP. Rates of GPP by pioneer communities were slightly higher than those of common coral reef framework substrates, such as 'biogenic coral rock' (i.e. 13-25 μmol C cm −2 day −1 in this study, compared to 4-9 μmol C cm −2 day −1 in van Hoytema et al., 2016) and CCA (i.e. 515 μmol C cm −2 day −1 in Anthony, Kline, Diaz-Pulido, Dove, & Hoegh-Guldberg, 2008), and lower than GPP rates reported for dense turf algae (i.e. >30 μmol C cm −2 day −1 in Adey &Goertemiller, 1987 andRix et al., 2015). The cover of individual functional groups (i.e. the 'makeup' of the community) was not a significant predictor of GPP, further demonstrating the taxonomic heterogeneity of pioneer communities on settlement tiles. Importantly, GPP of pioneer communities was similar to community-wide budgets for coral reefs in the Northern Red Sea (i.e. 13-25 μmol C cm −2 day −1 in this study, where (Eidens et al., 2014;Hatcher, 1988).
N 2 fixation of pioneer communities slightly exceed rates of coral-dominated communities shortly after colonization Shashar, Cohen, Loya, & Sar, 1994). Thus, our study corroborates the importance of pioneer communities for coral reef nitrogen budgets and highlights the ubiquity of diazotrophs in benthic reef environments (Cardini et al., 2014).

| Temporal variability and contribution of N 2 fixation to primary productivity
Both primary productivity and N 2 fixation displayed a temporal variability in relation to changing environmental conditions.
The observed fluctuations in primary production throughout the year were, primarily, attributed to temperature (Figure 4a) rather than the successional stage or 'makeup' of the pioneer communities on settlement tiles (Table S7) or other environmental variables ( Figure S1). As a temperature-sensitive process (Yamori, Hikosaka, & Way, 2014), primary production followed a typical combination of two Arrhenius curves (Equation 1) for the activation and deactivation of the carbon assimilating processes (Farquar et al., 1980;Leuning, 2002) that exhibits a maximum at 30.5°C. While primary production maxima are typically observed during summer at many reef locations world-wide, photosynthesis was highest in spring (i.e. May), indicating that summer temperatures can exceed the metabolic optima for some reef organisms in the central Red Sea (Roik, Roder, Röthig, & Voolstra, 2016;Sawall et al., 2015).
Temperature-coded scatterplots of productivity versus light availability ( Figure S3) helped to identify the dominance of the effect of temperature over that of light on photosynthesis in the current experiment (details in Appendix S1). Interestingly, high rates of photosynthesis were sustained throughout the study period, despite low inorganic nutrient availability during some months. This suggests that either primary production was not N limited or that additional nutrient sources contributed to supporting photosynthesis.
Concomitantly, N 2 fixation rates were mainly influenced by seawater NO − 3 concentrations (Figure 4b). This NO − 3 -dependence F I G U R E 5 Export of organic carbon from pioneer communities. The export was calculated by subtracting the organic carbon (C org ) acquisition rates (i.e. the differences between C org contents between two consecutive months) from the maximum amount of carbon that could have been accumulated by primary production in the respective time frame. Losses of carbon by respiration were accounted for by using daytime net primary production rates only, and subtracting night-time respiration. Bar charts show the mean ± SE  closely fitted a classic saturation curve for the NO − 3 -inhibition of the N-fixing enzyme nitrogenase (Holl & Montoya, 2005;Meeks et al., 1983), which corresponds to the inverse of a Michaelis-Menten model (Equation 5). Despite the higher energetic costs of N 2 fixation compared to N assimilation (Gallon, 2001), N 2 fixation was likely a seasonally relevant process to compensate the scarcity of N. We estimated that N 2 fixation supplied up to 13% of the PND during nutrient-depleted conditions, as was previously indicated by a positive relationship between the productivity and N 2 fixation for both turf algae and coral rock  and other reef substrates Cardini et al., 2015;Tilstra et al., 2017). At the same time, N 2 fixation still occurred even when NO − 3 concentrations were high, suggesting that productivity remained N limited throughout the year. The higher rates of N 2 fixation were also reflected in a depleted δ 15 N isotopic signature of biomass that approximated atmospheric N 2 (∼0‰) during periods of high N 2 fixation activity (Ryabenko, 2013;Stuyck, 1902). As an important source of N supply for photosynthesis, N 2 fixation likely also supported the stable C org /N ratios of biomass throughout the year.

| Accumulation and regeneration of C and N by pioneer communities
We observed low rates of biomass accumulation, despite high rates of NPP and little community-wide R. The relative constancy of biomass (i.e. limited net accumulation) after the first succession of pioneer communities suggests a tight coupling between production and loss on settlement tiles. Thereby, the export/loss of C org to the environment from the settlement tiles comprised on average 60% (at times >80%) of the photosynthetically fixed C that was not lost through metabolic activity by the community (R by algae, heterotrophs and decomposers). In absolute numbers, C org exports from pioneer communities (ranging from 2-15 μmol C cm −2 day -1 ) were higher compared to those of corals from the Northern Red Sea (2.5 μmol C cm −2 day −1 , annual mean ;Naumann et al., 2012) or the Pacific Ocean (5 μmol C cm −2 day −1 , Hata, Kudo, Yamano, Kurano, & Kayanne, 2002). Particularly grazing keeps epilithic algae in a cropped, early successional, and highly productive state in which they contribute most to the efficient transfer of energy to higher trophic levels (Fong & Paul, 2011). Concordantly, a high standing biomass of herbivorous fish (>150 g/m 2 ) in the studied reef likely maintained a substantial top-down control over organisms settling on bare substrates (Roth et al., 2018). In addition, C and N from benthic autotrophs are also released as dissolved organic matter exudates (Haas et al., 2010;Quinlan et al., 2018) or by the decay of biomass (Duarte & Cebrián, 1996), constituting energy and nutrient sources for other trophic levels.

| Ecological implications
Based on the processes outlined above, pioneer communities demonstrate features that indicate their critical role as a trophodynamic component of coral reefs, a role that is substantially enhanced after extensive coral mortality. These features are as follows: (a) pioneer communities are naturally abundant on coral reefs, even in coraldominated states. Following coral mortality, pioneer communities develop rapidly, often colonizing all available space previously occupied by corals; (b) pioneer communities show high assimilation rates of C and N. These rates are in the same order of magnitude as observed in coral-dominated communities. Likewise, there is a strong coupling between the fixation of C and N, as indicated by the seasonal changes and (c) there is a high trophodynamic connectivity with higher trophic levels. A large fraction of the accumulated C and N may be exported to higher trophic levels in the ecosystem via the release of organic matter or grazing, normally similar or even exceeding that of coral-dominated communities.
Thus, independent of whether pioneer communities coexist within a healthy reef (e.g. after loss of individual coral colonies) or increase in abundance after widespread coral mortality events, the high associated turnover of organic material from pioneer communities may temporarily serve as an alternative and direct source of C and N to compensate for the loss of functions previously provided by corals. While these pioneer communities cannot substitute important ecosystem functions that characterize scleractinian corals as reef ecosystem engineers (Wild et al., 2011), our data highlight that pioneer communities play an important role in the functioning of coral reef ecosystems via maintaining the nutrient provisioning and regeneration, which, in turn, may trigger a cascade of subsequent processes. First, by providing an alternative source of C and nutrients following coral mortality, pioneer communities aid sustaining higher trophic level consumers and prevent a complete change in the food web architecture. Particulate or dissolved organic compounds released from pioneer communities may be consumed directly by grazers (Crossman, Choat, Clements, Hardy, & McConochie, 2001), plankton (Nakajima et al., 2017) or are quickly remineralized to inorganic forms via detrital and microbial pathways (Haas et al., 2013), helping to transfer energy and nutrients through multiple trophic levels. Fish directly graze on pioneer communities, thereby transferring the acquired energy efficiently to higher trophic levels (Bellwood et al., 2018). The extensive turnover can, thereby, sustain a larger abundance of consumers than would be predicted by their biomass (Russ, 2003).
Second, pioneer communities serve as trophic conduits in a much more direct (and efficient) way than coral production, which, locally, may attract herbivorous fish (Francini-Filho, Ferreira, Coni, De Moura, & Kaufman, 2010). In turn, a high grazing activity ensures the availability of suitable settlement substrates for pioneer communities that include CCA and coral larvae, and the consumption of organic material promotes the remobilization of energy and nutrients within and across the borders of the ecosystem via excretion and egestion (Allgeier et al., 2017;Allgeier, Layman, Mumby, & Rosemond, 2014;Burkepile et al., 2013). Lastly, in cases in which coral mortality is patchy and coexists with bleaching, the increased availability of nutrients and heterotrophic sources may support the recovery of bleached corals and the recruitment of new coral colonies, as it may enhance coral thermo-tolerance, growth and calcification (Anthony, Hoogenboom, Maynard, Grottoli, & Middlebrook, 2009;Cox, 2007;Ferrier-Pagès, Witting, Tambutté, & Sebens, 2003;Fox et al., 2018;Hanafy, Aamer, Habib, Rouphael, & Baird, 2010).
However, the successional development of benthic pioneer communities, and their impact on the C and N dynamics herein, directly depends on the initial habitat characteristics (Roth et al., 2018), rates of grazing (top-down control) and nutrient availability (bottom-up control) within the reef (e.g. Burkepile & Hay, 2006;Smith, Hunter, & Smith, 2010). Under conditions of limited habitat complexity, low grazing pressure or nutrient pollution, pioneer communities may shift to a dominance of dense turf and macroalgae that are considered less palatable, resistant to grazing, and/or allelopathic (Hixon & Brostoff, 1996;Smith et al., 2010). Counterintuitively, taller and denser algae are less productive, transfer less energy through the food web, can trap more sediment (Miller & Barimo, 2001), leak more labile dissolved organic carbon (DOC; Haas et al., 2013) and promote higher microbial loads (Haas et al., 2016)-all of which could negatively impact the trajectories of recovery within the reef system.
Early pioneer communities may, thus, only assist in the recovery of coral reefs as long as a healthy consumer structure is present and the water quality remains adequate (Barkley et al., 2018).
Management and conservation efforts should, therefore, aim at preserving the structural integrity of a reef, reduce the fishing pressure on important functional groups (e.g. herbivores), and limit anthropogenic nutrient inputs. Under these conditions, pioneer communities may support the initial energy and nutrient demands of coral reef communities following coral mortality and large-scale disturbance events.

ACK N OWLED G EM ENTS
We are grateful to the personnel from the King Abdullah University