Benthic primary production in an upwelling-influenced coral reef, Colombian Caribbean

Study site and sampling seasons

This study was conducted in Gayraca Bay (11.33°N, 74.11°W), one of several smaller bays in TNNP, located near the city of Santa Marta (Fig. 1). The continental shelf in the area is relatively narrow due to the proximity to the Sierra Nevada de Santa Marta–the world’s highest coastal mountain range. The TNNP contains small fringing coral reefs reaching to a water depth of ∼30 m (Garzón-Ferreira, 1998; Garzón-Ferreira & Cano, 1991). The region is subjected to strong seasonality caused by the Caribbean Low-Level Jet of northeast (NE) trade winds (Andrade & Barton, 2005; Salzwedel & Müller, 1983), resulting in two major seasons; a dry season from December to April and a rainy season from May to November (Garzón-Ferreira, 1998; Salzwedel & Müller, 1983). Whereas the rainy season (non-upwelling) is characterized by low wind velocities (mean 1.5 m s⁻¹) (Garzón-Ferreira, 1998) and high precipitation (>80% of the annual rainfall) (Salzwedel & Müller, 1983), during the dry season (upwelling), strong winds prevail (mean 3.5 m s⁻¹, max 30 m s⁻¹) (Herrmann, 1970; Salzwedel & Müller, 1983) resulting in a seasonal coastal upwelling. The upwelling-related changes in key water parameters are well characterized by the comprehensive study of Bayraktarov, Pizarro & Wild (2014). During upwelling, water temperature can decrease to 20 °C while salinity and nitrate availability increase up to 39 and 3.59 µmol L⁻¹, respectively (Table 1). Water currents triggered by prevailing winds predominantly move from NE to SW, and a clear gradient in wave exposure between the exposed western (EXP) and -sheltered northeastern (SHE) sides of the bay can be observed (Bayraktarov, Bastidas-Salamanca & Wild, 2014; Werding & Sánchez, 1989).The study was carried out during non-upwelling in 2011 (1st November–2nd December 2011) and during the consecutive upwelling event (20th March–29th March 2012), allowing for the investigation of the influence of seasonality on benthic primary production.

Benthic assessment

For the assessment of benthic community structure, the dominant groups of benthic primary producers and the percentage of benthic cover were identified at EXP and SHE prior to primary production measurements using line point intercept transects at a water depth of 10 m (50 m length, n = 3), modified from Hodgson et al. (2004). Benthic cover was monitored at 0.5 m intervals directly below the measurement points (101 data points per transect). The dominant benthic autotrophs at the study sites consisted of scleractinian corals, frondose macroalgae, algal turfs (multispecific assemblage of primarily filamentous algae of up to 1 cm height, sensu Steneck (1988)), crustose coralline algae (CCA), and sand potentially associated with microphytobenthos. These categories represented 97 ± 1% of the total seafloor coverage at SHE and 91 ± 2% at EXP and were therefore selected as representative primary producers for the subsequent incubation experiments. During benthic community assessment, rugosity was determined at both sites using the chain method described by Risk (1972). Rugosity was quantified along three 10 m sub-transects within each of the 50 m transects and were used to calculate the rugosity factor for each study site as described by McCormick (1994) (SHE: 1.53 ± 0.12, EXP: 1.32 ± 0.13).

Sampling of organisms

Specimens of scleractinian corals, macroalgae, algal turfs, and CCA as well as sand samples, from 10 ± 1 m water depth were used for quantification of O₂ fluxes (see Table S1 for number of replicates). All samples were brought to the water surface in Ziploc^® bags and transported directly to the field laboratory. Scleractinian corals of the genera Montastraea (including the species M. faveolata, M. franksi and M. annularis, currently belonging to the genus Orbicella; Budd et al., 2012) and Diploria (including D. strigosa, currently belonging to the genus Pseudodiploria Budd et al., 2012) accounted for more than 80% of the total coral cover at the study sites and were therefore used as representative corals in our study. Coral specimens were obtained from the reef using hammer and chisel, fragmented with a multifunction rotary tool (8,200–2/45; mean fragment surface area: 13.16 ± 7.96 cm², Dremel Corp.), and fixed on ceramic tiles using epoxy glue (Giesemann GmbH, Aquascape). After fragmentation, specimens were returned to their natural habitat and left to heal for one week prior to the incubation experiments. Algae of the genus Dictyota (mainly D. bartayresiana) amounted to nearly 100% of macroalgal cover. Therefore small bushes of Dictyota spp. (surface area 1.86 ± 0.88 cm²) were used as representatives for macroalgae. Macroalgae were transferred to a storage tank (volume: 500 L in which water was exchanged manually 3–5 times per day and water temperature was within the ranges of incubation experiments; Table 2) one day before incubation experiments and left to heal. All other functional groups were incubated immediately after sampling. Rubble overgrown by algal turfs and CCA served as samples for the respective functional group (surface area covered by the organisms: 15.63 ± 10.80 cm² and 7.48 ± 3.60 cm², respectively). For sand samples, custom-made mini corers with defined surface area (1.20 cm²) and sediment core depth (1.0 cm) were used. All necessary permits (DGI-SCI-BEM-00488) were obtained by Instituto de Investigaciones Marinas y Costeras (INVEMAR) in Santa Marta, Colombia which complied with all relevant regulations.

Surface area quantification

Digital photographs of coral specimens were used to quantify planar projected surface areas of samples by image-processing software (ImageJ, V. 1.46r, National Institute of Health). The 3D surface area of the samples was estimated via multiplication of the planar projected surface areas by the genera-specific 2D to 3D surface area conversion factors derived from computer tomography measurements of Diploria and Montastraea skeletons (2.28 ± 0.16 and 1.34 ± 0.56, respectively), as described by Naumann et al. (2009). Planar leaf area of spread out macroalgal specimens was likewise quantified by digital image analysis and multiplied by the factor 2 to obtain 3D surface area of the samples. Image analysis of in situ photographs and whole spread out macroalgal thalli were used to obtain covered substrate areas (2D surface) as well as 3D surface areas in order to calculate the 2D to 3D conversion factor for macroalgae (4.29 ± 0.82). This conversion factor was used to correct for the overlap of macroalgal tissue. The 2D surface area of algal turf samples was determined by image analysis of digital photographs. For CCA, the simple geometry method described by Naumann et al. (2009) was used to estimate the surface area of overgrown pieces of rubble. The obtained surface areas were related to the planar projected surface area of the samples to generate 2D to 3D conversion factors for CCA (2.10 ± 0.89). Specimen surface area for sand samples was defined by the size of the utilized mini corer (1.20 cm²).

Data analyses and statistics

To quantify net O₂ production (P_n) and respiration of functional groups, O₂ concentration before incubations was subtracted from concentration after incubations and blank control values were subtracted from the measured O₂ fluxes. Individual gross O₂ production (P_g) of investigated functional groups was calculated by adding values of P_n and respiration; individual O₂ fluxes were expressed as mmol O₂ m⁻² specimen surface area h⁻¹.

The contribution of each functional group to total reef production (given as: mmol O₂ m⁻² seafloor area h⁻¹) was estimated as follows: ${\displaystyle c_i=p_i\hspace{0.17em}{s}_i\hspace{0.17em}{b}_i\hspace{0.17em}r}$ taking into account the individual production rates (p_i), the respective mean 2D to 3D surface conversion factor (s_i), group-specific benthic coverage (b_i) as well as the rugosity factor (r). Estimation of total daily benthic productivity was furthermore calculated by summing up the contribution of the investigated groups and extrapolating the incubation periods to a 12 h light and 12 h dark cycle.

After testing for normal distribution (Kolmogorov-Smirnoff test) and homogeneity of variances (Levene test), benthic coverage of functional groups were analyzed using two-way ANOVA and Bonferroni’s post hoc tests to detect possible effects of season (upwelling vs. non-upwelling) and site (EXP vs. SHE) and their interaction on benthic cover.

We tested the influence of benthic groups, season, wave exposure, and their interactions on O₂ productivity by generalized linear models (GLMs) for individual P_n and P_g of the investigated groups, their contribution to reef metabolism as well as total benthic productivity. We used Markov-chain Monte Carlo (MCMC) estimations of GLM regression coefficients. In traditional Frequentist statistics, the parameters of interest (i.e., the O₂ productivity describing regression coefficients) are estimated just once (e.g., using Maximum-Likelihood) and their significance is inferred indirectly based on a test-statistic. In contrast, Bayesian methods reallocate the coefficients across a set of possible candidates during each MCMC generation (Kruschke, 2011). If the bulk of these values, that is the 95% highest posterior density (HPD), does not include zero, one can directly conclude that the regression coefficient is credible different than zero and an effect on O₂ productivity exists. Moreover, we here performed pair-wise comparisons between benthic groups at different sites and seasons, traditionally being performed by post-hoc testing with P-value correction for preventing false positive results. A Bayesian GLM does not suffer this drawback because difference of groups can be directly estimated by the posterior (Kruschke, 2011). Again, there is credible evidence in non-equal group-means, if the posterior-based 95% HPD interval of the group differences does not include zero. Model performance for all 19 possible combinations of the three independent variables and their interactions was assessed by the deviance information criterion (DIC), a Bayesian measure of model fit that penalizes complexity (Spiegelhalter et al., 2002). In this information theory based model selection, often there is not a single best model describing the data. Therefore, averaging of regression coefficients for all models within ΔDIC < 2 of the best one (Johnson & Omland, 2004) was performed according to DIC weights (i.e., support for the respective regression model).

Here, Bayesian GLMs using the MCMCglmm package (Hadfield, 2010) for the R 3.0.3 environment for statistical computing (R Core Team, 2014) with a Gaussian error distribution were applied. Prior to the analyses, the mean–variance relationship of measured O₂ flux was stabilized by power transformation (Yeo & Johnson, 2000). Visual inspection of preliminary GLMs with default weakly informative priors showed high autocorrelation in their posterior distribution. Thus to infer the posterior distribution of the final analyses, we ignored the first 50,000 estimates as burn-in and sampled every 5th out of 650,000 MCMC generations.

All values are represented as mean ± standard deviation (SD) if not noted otherwise.

Benthic community composition

At EXP, scleractinian corals dominated the benthic community during non-upwelling and upwelling (41 ± 12 and 39 ± 12%, respectively; Fig. 2). At SHE, corals, algal turf, and sand cover was similar during non-upwelling (24 ± 3%, 26 ± 6%, and 25 ± 13%, respectively), while during upwelling, macroalgae exhibited highest benthic cover (47 ± 3%, Fig. 2). During the entire study period, coral and CCA cover was significantly higher at EXP than at SHE, whereas sand showed a contrary pattern with significantly more coverage at SHE (Fig. 2). Macroalgae was the only group where interaction between sites and seasons occurred with significantly higher cover at SHE and higher abundances during upwelling at both sites (Fig. 2). CCA cover also differed between the seasons, showing a significant decrease during the upwelling event (Fig. 2).

O₂ fluxes of organisms

More complex Bayesian GLMs, including interactions among the three independent variables season, benthic group, and site, described individual O₂ fluxes better than simple models (For details see Table S2).

Of all investigated functional groups, scleractinian corals had highest individual net (P_n) and gross production (P_g), followed by algal turfs, macroalgae, CCA, and microphytobenthos (Fig. 3; see also Table S3 for detailed results of all pair-wise comparisons). Regarding spatial differences in individual productivity, significant differences were detected for algal turfs and CCA. During upwelling, P_n of algal turfs and P_g of CCA was higher at SHE than EXP. On the contrary, during non-upwelling, P_n and P_g of CCA was higher at EXP (Fig. 3).

Temporal differences in O₂ production were detected for corals, algal turfs, and CCA (Fig. 3). Whereas P_n of scleractinian corals on both sites and P_g of CCA at EXP were higher during non-upwelling, P_g of CCA at SHE as well as P_n and P_g of algal turfs at both sites showed an opposite pattern with higher productivity rates during upwelling (Fig. 3).

Contribution of organism-induced O₂ fluxes to total reef O₂production

As in the case of individual O₂ fluxes, contribution and total reef production were better explained by GLMs of higher complexity (Table S2).

Contribution of functional groups to benthic productivity exhibited similar pattern than individual productivity with corals contributing generally most to total reef P_n and P_g, but macroalgae contributed most to benthic P_n and P_g at SHE at the end of upwelling (Fig. 4; see also Table S3 for detailed results of all pair-wise comparisons).

Significant spatial differences in contribution to total benthic P_n within functional groups were detected for corals, algal turf, and macroalgae, and spatial differences for P_g were present in all investigated groups except CCA (Fig. 4). At EXP, Corals contributed more to total P_n and P_g during non-upwelling and upwelling (Fig. 4). At SHE, contributions of macroalgae (P_n and P_g) and microphytobenthos (P_g) were higher only during upwelling, and algal turfs contributed more to P_g at SHE during non-upwelling (Fig. 4).

Temporal differences in contribution to total benthic productivity within the investigated groups were present for corals, macroalgae, CCA (for P_n and P_g), and for algal turfs (only P_g) (Fig. 5). During non-upwelling, corals contributed more to the total productivity at SHE and CCA at EXP, whereas during upwelling, macroalgae contributed more to the total productivity at SHE and algal turf at EXP (Fig. 4).

Regarding the total daily benthic O₂ fluxes (Fig. 4), no spatial differences between EXP and SHE were detected, neither during non-upwelling nor during upwelling (see also Table S3 for detailed results of all pair-wise comparisons). During the study, significant temporal differences were only present for P_g at the exposed site with higher O₂ fluxes during the upwelling in 2011/2012 compared to non-upwelling (Fig. 5). Comparing total benthic productivity during the upwelling event in 2010/2011 with the subsequent non-upwelling and upwelling, P_n and P_g were significantly higher during the upwelling 2010/2011 for all comparisons (Fig. 5).

O₂ fluxes of organisms

Individual mean P_n and P_g were generally highest for corals at both sites during the study periods (P_n: 11.2–16.1 and P_g: 17.4–20.8 mmol O₂ m⁻² specimen area h⁻¹). These high productivity rates of corals compared to other investigated primary producers (see Fig. 3) may be attributed to the mutualistic relationship between zooxanthellae and coral host leading to enhanced photosynthetic efficiency under high CO₂ and nutrient availability (D’Elia & Wiebe, 1990; Muscatine, 1990). Estimated daily P_g per m²seafloor for the investigated coral genera, (441–610 mmol O₂ m⁻² seafloor d⁻¹), is within the range of other Caribbean corals (67–850 mmol O₂ m⁻² seafloor d⁻¹, Fig. 2, Kanwisher & Wainwright, 1967), and O₂ fluxes of all investigated organism groups are comparable to values reported in the literature (Fig. 2).

Significant spatial differences during non-upwelling were found for CCA with higher productivity at EXP compared to SHE (Fig. 3). These differences may be attributed to the prevailing water current regime in the bay together with high water temperatures during non-upwelling (Tables 1 and 2). An increase in water temperature typically intensifies metabolic activity in CCA (Hatcher, 1990; Littler & Doty, 1975). However, the lower water flow at SHE (Bayraktarov, Bastidas-Salamanca & Wild, 2014) may have prevented the required gas exchange and nutrient uptake, resulting in lower individual CCA productivity at this site. In contrast, the higher rates in individual productivity of algal turfs and CCA at SHE during upwelling (Fig. 3) are potentially a result of the differences in species composition (sensu Littler, 1973; Chisholm, 2003; Copertino, Cheshire & Kildea, 2009; Ferrari et al., 2012).

Temporal differences in individual O₂ production within the investigated organism groups generally showed two contrary patterns: whereas scleractinian corals on both sites and CCA at EXP produced less O₂ during upwelling, algal turfs and CCA at SHE produced more O₂. The decreased productivity rates of corals and CCA at EXP during upwelling indicate that low water temperature has an adverse effect on the productivity of these groups. This argument is supported by studies showing that low water temperatures lead to a decrease in photosynthetic performance of primary producers in coral reefs (Hatcher, 1990; Kinsey, 1985). In contrast, the two-fold higher photosynthetic performance of algal turfs during upwelling may be due to higher nutrient concentrations together with higher water currents during this season (Bayraktarov, Bastidas-Salamanca & Wild, 2014; Bayraktarov, Pizarro & Wild, 2014), facilitating gas exchange and nutrient uptake. Our findings are supported by Carpenter & Williams (2007), showing that photosynthesis of algal turfs in coral reefs is mainly limited by nutrient uptake, which in turn depends on nutrient availability and water current speed. Whereas productivity of CCA at EXP seems to be temperature-limited, our findings indicate that their productivity at SHE is limited by nutrient availability as previously suggested for benthic algal communities in water current-sheltered coral reef locations (Carpenter & Williams, 2007; Hatcher, 1990).

Contribution of organism-induced O₂ fluxes to total benthic O₂ production

Our results indicate that the spatial differences in contribution to total benthic O₂ production for scleractinian corals, macroalgae, CCA, and microphytobenthos are directly linked to spatial differences in their benthic coverage. For instance, the major contribution of corals (Fig. 4) can be explained by their comparably high benthic coverage (ranging from 24 to 39%; Fig. 2) and highest quantified individual O₂ production rates among all investigated groups (Fig. 3). This finding is supported by the estimates of Wanders (1976a), showing that corals accounted for about two-thirds of the total benthic primary production in a Southern Caribbean fringing reef.

Although individual macroalgal production rates were rather low as compared to coral productivity (Fig. 3), the extremely high cover of macroalgae at SHE during upwelling (47 ± 3%, Fig. 2) resulted in macroalgae being the main contributors to total benthic production. Macroalgal cover (incl. the dominant genus Dictyota) has previously been found to be particularly high during upwelling (Bula-Meyer, 1990; Cronin & Hay, 1996; Diaz-Pulido & Garzón-Ferreira, 2002) probably due to elevated nutrient concentrations and low water temperatures (Bayraktarov, Pizarro & Wild, 2014).

The elevated contributions of corals and CCA at EXP as well as macroalgae and microphytobenthos at SHE during upwelling (Fig. 4) might be due to site-specific differences in abundances (Fig. 2), which in turn are likely caused by site-specific differences in water current regimes (Bayraktarov, Bastidas-Salamanca & Wild, 2014; Werding & Sánchez, 1989).

Corals, macroalgae, algal turfs, and CCA also exhibited distinct temporal differences in contribution to total benthic productivity. At SHE, corals contributed more to the benthic O₂ production during non-upwelling and macroalgae and algal turfs during upwelling, whereas contribution of CCA at EXP was higher during non-upwelling. These differences can be explained with seasonal growth patterns, temperature-dependent changes in individual O₂ productivity and temporal shifts in abundances (Figs. 2 and 3). Opposite abundance patterns of CCA and macroalgae are, for example, in agreement with previous studies showing that macroalgae can shade CCA, usually leading to negative correlated abundances of these groups (Belliveau & Paul, 2002; Lirman & Biber, 2000).

Total benthic O₂ fluxes and ecological perspective

The estimated total daily benthic O₂ production at both sites during non-upwelling and upwelling (Fig. 5) are, although comparable, on average slightly lower than the values previously reported for other fore reefs communities (Table 3). These differences might be due to a methodological bias. Whereas previous studies utilized flow respirometry techniques, the current study used incubation methodology, which accounts for production values in the target groups only.

Despite the high spatial and temporal differences in benthic coverage and group-specific O₂ fluxes of investigated benthic primary producers as well as their contribution to total benthic productivity, no spatial differences in total benthic O₂ fluxes were detected between EXP and SHE. These results were consistent during both non-upwelling and upwelling (Fig. 5). Our findings are supported by Hatcher (1990), showing that the relative coverage of benthic photoautotrophs in a reef community may have little effect on its areal production rate. In TNNP, seasonal differences were only present for P_g at EXP with higher rates during upwelling compared to non-upwelling. These differences are mainly related to individual productivity of algal turfs, being generally two-fold higher during upwelling compared to non-upwelling (Fig. 3), and to the absence of macroalgae at EXP during non-upwelling (Fig. 2). This is in agreement with studies by Kinsey (1985) and Hatcher (1990), reporting that algae, as one of the most seasonal component in coral reefs, account for seasonal shifts in benthic reef productivity.

The lack of seasonality of P_n and P_g regarding communities at SHE as well as P_g at EXP stands in contrast to earlier studies (Eidens et al., 2012; Kinsey, 1977; Kinsey, 1985; Smith, 1981), which found an approximately two-fold difference in benthic primary production between seasons. This lack of seasonality in P_n and partly in P_g might be related to seasonal changes of abiotic factors in TNNP that compensate for each other (Table 1). The observed similarity in productivity rates during different seasons suggest that coral reefs in TNNP can cope with pronounced seasonal variations in light availability, water temperature, and nutrient availability. Nevertheless, total P_n and P_g during the upwelling in 2010/2011 (P_n: 244–272 and P_g: 476–483 mmol O₂ m⁻² seafloor d⁻¹) were not only higher compared to non-upwelling (Eidens et al., 2012) but also higher than during the subsequent upwelling in 2011/2012 (Fig. 5). These findings suggest that interannual variations affect the productivity of TNNP coral reefs. Dramatic ENSO-related water temperature increases and high precipitation in the study area (Bayraktarov et al., 2013; Hoyos et al., 2013) led to coral bleaching at the end of 2010 (Bayraktarov et al., 2013). Surprisingly, bleached corals in the bay recovered quickly in the course of the following upwelling event (Bayraktarov et al., 2013) and exhibited similar O₂ production rates during all study periods (Eidens et al., 2012), indicating a high resilience of TNNP corals. Moreover, macroalgae and algal turf seemed to benefit from the environmental conditions during the upwelling following the ENSO-related disturbance events, resulting in higher group-specific productivity during the upwelling in 2010/2011 compared to subsequent study periods (Eidens et al., 2012). The elevated production rates of macroalgae and algal turfs together with the quick recovery of corals from bleaching likely accounted for a higher benthic productivity during the upwelling in 2011/2011 compared to non-upwelling (Eidens et al., 2012) and the upwelling in 2011/2012 (Fig. 5). These findings indicate that extreme ENSO-related disturbances do not have long-lasting effects on the functioning of local benthic communities in TNNP.

In conclusion, the present study showed that total benthic productivity in TNNP is relatively constant despite high variations in key environmental parameters. This stable benthic productivity suggests a relatively high resilience of local benthic communities against natural environmental fluctuations and anthropogenic disturbances. We therefore recommend that TNNP should be considered as a conservation priority area.