Submarine groundwater springs are characterized by distinct fish communities

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Marine Ecology published by Wiley-VCH GmbH. 1Faculty 2 Biology/Chemistry, University of Bremen, Bremen, Germany 2Leibniz Centre for Tropical Marine Research (ZMT), Bremen, Germany 3Mauritius Oceanography Institute, Albion, Mauritius 4School of Science, Auckland University of Technology, Auckland, New Zealand 5Institute of Geosciences, University of Kiel, Kiel, Germany


| INTRODUC TI ON
Nutrient input to the coastal ocean is attributed mainly to terrestrial runoff and river discharge (Fabricius, 2005;Jickells, 1998).
However, the discharge of fresh terrestrial groundwater into the ocean (fresh submarine groundwater discharge: fresh SGD) was shown at several locations to be a major local source of nutrients to coastal waters (Burnett, Bokuniewicz, Huettel, Moore, & Taniguchi, 2003;Knee & Paytan, 2011;Slomp & Cappellen, 2004).
Historical ethnographic studies have often suggested that fish seem to thrive around submarine springs (Moosdorf & Oehler, 2017).
Past studies also propose that fresh SGD may promote fisheries production in temperate coastal systems (Burnett, Wada, Taniguchi, Sugimoto, & Tahara, 2018;Shoji & Tominaga, 2018) and suggest higher fish abundances near a submarine spring in the tropics (Starke, Ekau, & Moosdorf, 2020). As elevated growth will over time translate into elevated population sizes in marine teleosts (Retzel, Hansen, & Grønkjaer, 2007), one might argue that higher food abundance and physiologically beneficial environmental conditions around sources of fresh SGD allow fish to prosper (Boeuf & Payan, 2001;Di Franco et al., 2019;Jarrold & Munday, 2019). Two recent studies showed a positive effect of fresh SGD on fish growth: Fujita et al. (2019) explained a positive relationship between fish growth and fresh SGDderived nutrient loadings with elevated primary producer and prey abundances, whereas Lilkendey et al. (2019) proposed that a combination of physiological effects caused augmented growth. In any case, all past results imply a link between acidic nutrient-rich fresh SGD and elevated abundances of marine fishes.
The present study aimed to assess the impact of submarine groundwater springs on the productivity of a tropical lagoonal ecosystem, primarily focusing on the abundance and community structure of coral reef-associated fish. To investigate the relationship between groundwater springs and the abundance as well as the diversity of fishes, we used remote underwater video surveys within a coral reef lagoon in Mauritius, Indian Ocean. We also collected baseline data on water salinity, pH, and nutrient concentrations, benthos composition, and total suspended solids to determine the effects of fresh SGD on lagoonal hydrography.

| Sampling site
Mauritius is an island (area: 108 km 2 ) of volcanic origin situated ca. 900 km east of Madagascar and ca. 2,000 km from the east African coast in the southwestern Indian Ocean (Figure 1a). Mauritius is inhabited by a population of 1.2 Mio (2016) and visited by 1.4 Mio tourists yearly (Martial & Boolakey, 2016;Martial & Fatha Mahomed 2019). The sugar industry is still a major agricultural sector in Mauritius, with sugarcane plantations occupying 85% of cultivated area, representing ca. 40% of the country's total area (Kwong, 2005).
A high amount of sugar production, increasing tourist numbers, poorly developed wastewater management, and sewer systems are thought to be the major sources of anthropogenic nutrients to coastal waters (Ramessur, 2002). Trou aux Biches is a small town on the northwestern coast of Mauritius, where six distinct groundwater springs discharge in the southern part of the lagoon (Figure 1b,c). The groundwater springs in Trou aux Biches are connected to the aquifer of the Northern Basin, covering an undulating area of 200 km 2 generally below 100 m of elevation (Saddul, 2002). Highly permeable recent lavas underlie the greater part of the basin, and the surface is mainly comprised of sugarcane plantations inland (Proag, 1995) and hotels, resorts, and bungalows in the nearshore area.

| Environmental parameters
The field survey was conducted from October to December 2017, during the early summer months in Mauritius. Salinity was measured in situ using a handheld probe (WTW Cond 3310, TetraCon 325).
Water samples were taken directly from the spring area (n = 6), the spring-influenced part in the south (n = 12), and the control (n = 18) in the north part of the lagoon. Additionally, water samples were obtained from two oceanic stations offshore the SGD-influenced and control part of the lagoon (not on map). Water pH was determined in the laboratory using a stationary pH probe (Ohaus Starter 2100). Water samples for nutrient analysis were taken from a depth of 50 cm using a peristaltic pump, filtered, and stored frozen until measurement.
Nutrient analysis of water samples was conducted in the laboratory at the Mauritius Oceanography Institute in Albion, Mauritius.
Nitrite, nitrate, phosphate, and silicate were determined using standard methods with a discrete analyzer (SYSTEA Easychem Plus) equipped with a 5 cm absorbance reading unit. The analytical procedures were adapted from Grasshoff, and Koroleff (1983).
For nitrite (NO − 2 ) determination, sulfanilamide was added to 700 µl of seawater sample to obtain a diazo compound. Under the addition of N-(1-naphthyl) ethylenediamine, the compound formed a pink complex. Colorimetric measurements were taken at 543 nm wavelength against a reference of artificial seawater prepared in deionized water (Rider & Mellon, 1946).

For nitrate (NO −
3 ) determination, the nitrite content of the sample was determined as described above, followed by a nitrate reduction and a subsequent second determination of nitrites. Starting nitrite concentrations were subsequently subtracted from the NOx concentrations to obtain the nitrate concentration of the sample. Nitrate reduction was done manually on a column filled with copper-coated cadmium pellets activated by a 10 M solution of sodium nitrate prepared in ammonium chloride buffer. The samples were poured into the burette and let pass through the cadmium pellets. The nitrite content of the captured sample was measured in the discrete analyzer. After ca. 5 ml passed through the pellets, 1.5 ml of the sample was captured in 1.5-ml microcentrifuge tubes for subsequent analysis in the discrete analyzer.
For the determination of phosphate (PO 2− 4 ) concentrations, ammonium molybdate and potassium antimony tartrate were added to 700 µl of seawater sample to form an antimony-phospho-molybdate complex. This complex was reduced to an intensely blue-colored molybdenum complex by the addition of freshly prepared ascorbic acid. The absorbance was measured at 880 nm wavelength against a reference of artificial seawater prepared in deionized water.
Silicate (SiO 4− 4 ) was determined by adding ammonium molybdate to 700 µl of seawater sample. Oxalic acid was added to eliminate phosphate interference. The sensitivity of the analysis was increased further by a reduction in the silicomolybdic acid using ascorbic acid to procure "molybdenum blue." The absorbance was measured at wavelengths of 810 and 660 nm against a reference of deionized water.
Benthic coverage was determined using 50-m point intersect transects (Hill & Wilkinson, 2004). All transects were video-recorded (Canon Powershot G16) for subsequent analysis. Transect F I G U R E 1 Mauritius in the southwestern Indian ocean (a), and the hollow star marks the location of Trou aux Biches lagoon (b). The lagoon was split into the sampling sites "springs," "springinfluenced," and "control" (boxes). White stars indicate the groundwater springs. Nutrient sampling stations are indicated by dotted circles, benthic transects by dashed lines, and camera positions by hollow circles (c) videos were evaluated in the laboratory, recording benthic makeup every 50 cm (n = 100) using classifications by English, Wilkinson, and Baker (1997). Total suspended solids (TSS) were sampled in triplicates along 100 m horizontal transects at the springs, the spring-influenced part of the lagoon, and at the control site using a 20-µm plankton net. Per transect, a total of 4.91 m 3 water (net diameter 25 cm) were filtered. All samples were kept cold in a portable cooling box and frozen at −20°C in the laboratory for subsequent analyses. Frozen TSS samples were defrosted, filtered on a pre-weighted microfiber filter (Whatman GF/F), dried, and weighed again.

| Fish community
To evaluate fish abundances, Go-Pro Hero 4 video cameras were weighted with 1 kg lead and placed at two stations close to the springs and at three stations at the spring-influenced site as well as at the control site ( Figure 1c). We have refrained from placing cameras close to shore at the control site to avoid their theft through tourists. Videos were taken between 9:00 and 16:00 to ensure the capture of diurnally active fishes (English et al., 1997).
After each placement, the first 15 min of video was disregarded to allow the fish community to recover to a more natural state after placing the cameras. The first 130 min of video footage was split into three 20-min segments (15 min between segments).
Each segment was further divided into 40 single frames (one each 30 s) and evaluated using the MaxN/MIN count method (Cappo et al., 2003;Wells & Cowan, 2007). This method records the maximum number of each species observed at any one time in each of the single frames and yielded nine observations (three stations × three replicates) per sampling site. Videos were used to (a) compose a list of all fish species occurring in the lagoon and (b) determine abundances.
Diversity was assessed using three commonly used metrics: species richness, Shannon's diversity index H′, and Pielou evenness J′. Functional groups were used to evaluate changes in the fish community structuring. This method is increasingly used in coral reef studies to understand the effects of disturbances on ecosystem functioning (Graham et al., 2006). Groupings are based on similar ecosystem functioning disregarding taxonomic relationships (Bellwood, Hughes, Folke, & Nyström, 2004). Fish species were classified into feeding groups indicating feeding behavior and dietary composition (Pratchett, Hoey, Wilson, & Messmer, 2011).
Herbivores and corallivores were further classified using definitions by Green and Bellwood (2009) and Cole, Pratchett, and Jones (2008), respectively.

| Statistical analyses
Statistical analyses were carried out using JMP (Pro 14.3, SAS Institute Inc., www.jmp.com). Nutrient concentrations and benthic coverage data were log + 1-transformed and normalized before a principal component analysis (PCA) based on Euclidian distance was performed to test for significant differences between lagoons and sites. To account for pseudoreplication in our dataset, we used one factorial repeated-measures analyses of variances (RM-ANOVAs) to identify differences in TSS, fish abundance, and diversity indices between sampling sites (Millar & Anderson, 2004).
Depending on skewness, non-normal data were either log-or square root-transformed (Underwood, 1997). Subsequently, a Bartlett test was applied to test for homogeneity of variances. We tested the response variables against site (springs, spring-influenced, control) as the main effect and replicate as a random effect nested within sampling station. Significant model results were further analysed using Tukey's honest significance difference (HSD) post hoc tests.
To visualize differences among fish communities across sites, non-metric multidimensional scaling (NMDS) was applied based on the rank order of the Bray-Curtis similarities using PRIMER-e v7 (Clarke & Gorley, 2015). Fish abundance data were fourth root-transformed to reduce the values and influence of highly abundant species while also allowing midrange to rare species to influence the analysis (Clarke, 1993;Clarke & Green, 1988;Field, Clarke, & Warwick, 1982). The aptness of the NMDS was determined through the stress factor by which low values express a higher variance (Warwick & Clarke, 1991). Relative grouping of sampling stations was validated using cluster analysis. A similarity percentage (SIMPER) analysis with pooled replicates was used to examine which secondary feeding groups and fish species contributed to similarities within sites. Differences between sites were examined using twoway analysis of similarities (ANOSIM) with station nested within site (Clarke & Warwick, 1994).

| RE SULTS
Water salinity and pH values were lower at the springs than at the spring-influenced site and again lower at the spring-influenced than at the control site. Average nutrient concentrations were generally higher at the springs than at the spring-influenced and control sites.
The offshore station near the spring-influenced part of the lagoon exhibited lower salinity and pH values and higher nutrient concentrations when compared to the station offshore the control site (Table 1).
Principal component analyses were used to investigate spatially resolved differences in hydrography and benthic cover composition. Benthos composition showed only little clustering between sampled transects at the springs, the spring-influenced site, and the control site (Figure 2b). The first principal axis (PC1) explained 37.1% of the observed variability and mainly separated the sampled transects by their percentage cover in sand, coral, rubble, and turf algae.
Benthic composition of transects in the spring-influenced part of the lagoon was mainly described by positive PC1 loadings and thus by elevated coral, rubble, and turf algae cover. Mean (±SD) coral cover was consequently markedly lower at the springs (0.3 ± 0.6%) and the control site (1.7 ± 2.1%) compared with the spring-influenced part of the lagoon (17.6 ± 15.3%). Mean turf algae cover was lower both at the control site (5.5 ± 4.0%) and the springs (12.0 ± 13.6%) compared with spring-influenced part of the lagoon (20.6 ± 18.1%).
Mean macroalgae cover, on the other hand, was markedly lower throughout transects at the spring-influenced part of the lagoon (2.4 ± 3.7%) when compared to the springs (18.8 ± 19.5) and the control site (14.5 ± 10.2%). The second axis (PC2) separated transects according to their cover in macroalgae, rock, and seagrass cover and explained 24.5% of the observed variances.
In total, 1,709 individual fishes, representing 95 species from 28 families ( Table 2), were recorded in Trou aux Biches lagoon. Out of these, 84 species from 26 families were included in sampled frames.
Cluster analysis and NMDS showed apparent clustering of fish community structure composition at the sampling stations for each site. In respect to secondary fish feeding groups, the control site was characterized by generalists, the spring-influenced site by obligate corallivores, territorial farmers, facultative corallivores, omnivores, and scrapers, and the springs by a high abundance of grazer/detritivores, piscivores, planktivores, and invertivores ( Figure 4). TA B L E 1 Mean (±SD) salinity, pH, and nutrient concentrations (µM) for all three sites, control, spring-influenced, and springs within the Trou aux Biches lagoon and two offshore stations
These between-sites differences in fish community structures were significant with regard to both feeding groups (two-way ANOSIM, global R = .33, p = .029) and species (two-way ANOSIM, global R = .34, p = .046).

| D ISCUSS I ON
Although globally, fresh SGD amounts to only a small percentage of river discharge (Luijendijk, Gleeson, & Moosdorf, 2020;Taniguchi, Burnett, Cable, & Turner, 2002) tropical coasts export more than 56% of all fresh SGD (Zhou, Sawyer, David, & Famiglietti, 2019). In coastal ecosystems nutrient-rich fresh SGD increases primary production and sustains higher primary and secondary consumer biomass (Dale & Miller, 2008;Encarnação et al., 2015;Hata et al., 2016;Lecher & Mackey, 2018;Piló et al., 2018;Utsunomiya et al., 2017;Waska & Kim, 2011). Further, physiologically beneficial environmental conditions brought about by the submarine influx of terrestrial nutrientrich cold acidic freshwater elevate the fitness of reef fish, potentially resulting in increased population sizes (Fujita et al., 2019;Lilkendey et al., 2019). The assessment of factors influencing the abundance of consumers, in particular fish, is of ever-growing concern as this information is vital to predicting consequences of anthropogenic actions on ecosystem functioning and productivity (Burnett et al., 2018;Shoji & Tominaga, 2018). Our results suggest a fresh SGD-driven positive relationship between altered hydrography and enhanced secondary consumer abundances around groundwater springs in a coral reef lagoon.
Because of a time-delayed sampling between the spring-influenced part of the Trou aux Biches lagoon and the control site, we were markedly lower at the springs and throughout the spring-influenced part of the lagoon, as was made visually evident via PCA.
The influx of fresh groundwater is a locally important pathway for inland-derived nutrients, especially to tropical coastal marine environments (Luijendijk et al., 2020;Zhou et al., 2019). Recorded nitrate and phosphate concentrations were higher than previously reported for Trou aux Biches lagoon (Hurbungs, Jayabalan, & Chineah, 2002).
Higher phosphate levels at the spring-influenced site when compared to the springs may be explained by temporal variability in phosphate influx at differing sampling dates. However, the observed seaward increase in phosphate concentrations may also be facilitated by rapid phosphate uptake of algae found at the springs and stations closest to shore: The spring area was mainly devoid of macroalgae except for large aggregations of Dictyota sp. directly at the springs. Members of this genus reportedly prefer waters with high nutrient loadings (den Haan et al., 2016;Lapointe, 1997) and low pH (Cornwall et al., 2017), conditions which were predominant at the springs. As Dictyota sp. can rapidly take up large quantities of phosphate, high abundances of these macroalgae at the springs could explain the low phosphate concentrations observed in the surface waters (den Haan et al., 2016). Also, as high silicate concentrations are a tracer for SGD (Oehler, Tamborski, et al., 2019), and because of the lagoonal tidal dynamics and currents (Y. Neehaul personal observation), we do not assume any nutrient enrichment process apart from the influx of groundwater through the springs. Our results, therefore, showed a significant impact of nutrient-rich fresh SGD on lagoonal hydrography well beyond the nearshore spring area. Our control site in the north, on the other hand, clustered with the oceanic sampling stations in the PCA, and therefore, we assumed it to be unaffected by any kind of nutrient enrichment.
The occurrence of fresh SGD in coral reef ecosystems is often associated with adverse effects on coral cover, diversity, and growth to impact coral physiology negatively (Fabricius, 2005). Besides organic matter, TSS encompass all suspended particles such as sand and other inorganic materials. High plankton abundances due to nutrient enrichment processes through groundwater discharge may, therefore, be only one explanation for increased TSS loads among others (Lecher et al., 2015;Sugimoto et al., 2017). Still, elevated planktonic food biomass at the spring-influenced part of the lagoon may sustain high coral cover (Anthony, 1999) and high abundances of the planktivorous damselfish D. abudafur.
Besides many advantages, remote underwater video surveys underestimate species richness in comparison to traditional diver-based methods (Caldwell, Zgliczynski, Williams, & Sandin, 2016), mainly because of a smaller survey area. The confined spring area, however, restricted us to a point sampling approach. By choosing unbaited over baited remote underwater video surveys, we avoided a sampling bias toward generalists, carnivores, large predators, and mobile species (Mallet & Pelletier, 2014). Also, ubiquitous coral loss caused by climatic changes may lead to a decrease in reef fish biodiversity (Jones, McCormick, Srinivasan, & Eagle, 2004), and the disappearance of just a single species of coral can already affect overall diversity in fish communities (Komyakova, Jones, & Munday, 2018). Effects of habitat type on fish assemblages on reef flats in the Indian Ocean are consequently regarded as much stronger than temporal factors (e.g., seasonality) (Graham et al., 2007;Letourneur, 1996a). Therefore, we consider the bias on fish assemblage structure introduced by Spring influenced temporal differences in camera deployments as minor. We cannot, however, completely rule out seasonal effects such as spawning migrations and recruitment processes on the recorded fish abundances (Harmelin-Vivien, 1989;Letourneur, 1996b).
Nevertheless, in agreement with past findings (Hata et al., 2016;Shoji & Tominaga, 2018;Starke et al., 2020;Utsunomiya et al., 2017) our observations showed that fish abundances were higher at the groundwater springs when compared to the spring-influenced site and the control. Especially, structural complexity is often positively correlated with the abundance and diversity of fishes in tropical ecosystems (Darling et al., 2017). Groundwater springs also exhibit a certain degree of rugosity, such as crater-like depressions and coneshaped fissures (Oehler, Bakti, et al., 2019). Pronounced structural complexity may thus serve to explain the elevated fish abundances at the springs when compared to the rest of the lagoon (Brown et al., 2017). Still, diversity, evenness, and species richness were not significantly different between sites, suggesting that the three-dimensional structure of the springs does not drive fish abundances through provision of diverse ecological niches (Rogers, Blanchard, & Mumby, 2014).
Macroalgae cover was higher at the springs and the control site than throughout the spring-influenced part of the lagoon, where nutrient concentrations were consistently higher when compared to the control site. This suggests either only a small influence of nutrient enrichment on macroalgae growth or robust top-down mechanisms limiting macroalgal proliferation at the spring-influenced site (Burkepile & Hay, 2006;Heck & Valentine, 2007). As in tropical regions fresh SGD has been shown to contribute significantly to reef productivity (Greenwood et al., 2013;Oehler, Bakti, et al., 2019), high macroalgae may point toward increased primary productivity at the springs, potentially supporting the observed higher herbivorous fish abundances. The fish may preferentially graze upon the phosphorus enriched algae directly at the springs (Peterson, Stubler, Wall, & Gobler, 2012). Accordingly, high macroalgal cover at the springs coincides with a high abundance of juvenile Siganidae spp. and high contribution thereof toward community structuring. Some Siganid species are known to recruit in numbers so high that they consume almost all available macroalgae (Paul, Nelson, & Sanger, 1990). The only macroalgae genera that occurs at the springs, Dictyota sp., is also a predominant part of the diet of Siganids (Stergiou, 1988) as the family is less repelled by the algae's deterrent metabolites (Paul et al., 1990;Wylie & Paul, 1988). Therefore, the springs may act as a nursery for Siganids while changing from a planktonic to an algal-based diet (Duray, 1998;Kami & Ikehara, 1976).
In contrast to fishes, invertebrates are among the best-studied animal taxa concerning the impact of SGD on marine biota (Encarnação et al., 2013(Encarnação et al., , 2015 Observations in temperate systems lead to the hypothesis that fresh SGD contributes to coastal fishery resource biomass through nutrient enrichment as well as via alterations to coastal hydrography (Shoji & Tominaga, 2018;Utsunomiya et al., 2017). This study provides further evidence of a positive relationship between submarine groundwater springs and increased fish abundance in a tropical coral reef ecosystem. Elevated fish abundance, as well as high contributions of herbivore and invertivore fish species to the fish community at the springs, suggests a positive effect of nutrient-rich fresh SGD on food availability and secondary consumer biomass. Especially, the observed elevated fish abundances at the springs could have implications for the management of small-scale fisheries in tropical lagoons: Namely that altered groundwater fluxes on land may cause differences in fish biomass available to fisheries. We acknowledge that this study is just a first step in determining the influence of groundwater springs on tropical lagoonal fisheries. Nonetheless, we recommend that processes such as fresh SGD should be incorporated into management approaches throughout tropical coral reef environments.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.