Long term relationship between farming damselfish, predators, competitors and benthic habitat on coral reefs of Moorea Island

Understanding the processes that shape biodiversity is essential for effective environmental management. Across the world’s coral reefs, algal farming damselfish (Stegastes sp.) modify the surrounding benthic community through their creation of algae “farms”. Using a long-term monitoring dataset (2005–2019) from Moorea Island, French Polynesia, we investigated whether the density of dusky damselfish (Stegastes nigricans) is associated with benthic habitat composition, the density of predators and/or competitors, and whether the survey area was inside or outside of a Marine Protected Area (MPA). We found no evidence that benthic cover or number of competitors were associated with dusky damselfish densities, both inside and outside MPAs. In contrast, fluctuations in dusky damselfish densities were negatively associated with the density of predators (e.g. Serranidae, Muraenidae and Scorpaenidae) in the preceding year in non-MPA areas, and both within and outside of MPAs when predator densities were high (2005–2010). These results suggest that healthy predator populations may be important for regulating the abundances of keystone species, such as algal farming damselfish, especially when predator densities are high.

. Across the world's coral reefs, some herbivorous fish are keystone species and the diverse array of species within this guild provides various ecosystem services, including those that may aid in the recovery of live hard coral 11,[14][15][16][17] . By contrast, other keystone species, such as farming damselfish, can favor the development of algal turfs and defend them against other herbivorous fish [18][19][20] .
Damselfish of the genus Stegastes (Pomacentridae, herein 'Stegastes') are highly territorial species that develop dense turf algae patches (i.e. farms) which act as a food source 18,19 . Reliance on algae for sustenance varies across Stegastes species, spanning from facultative to obligate [21][22][23] . They actively control the algal species composition within their farms, and defend them against other herbivorous fish, invertebrate micro-herbivorous grazers and sea-urchins 17,19,20 . In order to provide substratum for their algal field, they can actively removed scleractinian corals, such as by biting the living tissue and cultivating dense algal lawns on the coral skeletons 24 . Correspondingly, the farms of several Stegastes species (e.g. S. apicalis and S. nigricans) can act as reservoirs of pathogens that cause coral diseases 25 . While these species can modify habitats around them, other species can regulate their density and distribution. For example, Precht et al. 24 found that Stegastes densities were strongly regulated by predation risk, and Randazzo-Eisemann et al. 26 found that predators are likely to play an important role in regulating the distribution of Stegastes and reducing the stress that they impose on the coral system. Overall, research conducted on farming damselfish over the last four decades has highlighted the influence of Stegastes population dynamics on the benthic cover of coral reefs and thus their potential usefulness as an ecological indicator, but also the potential effects some competitors and predators have on their density 18,19 .
While long-term studies have found that the density of Stegastes-associated turf can increase significantly as the damselfish take advantage of coral mortality [26][27][28][29] , few monitoring studies have tracked whether their abundance is associated with, and affected by, predator and competitor populations as well as benthic habitat composition 30 . Notably, Naim et al. 28 showed that algal turf abundance increased significantly between 1993 and 2002 due to the expansion of Stegastes territories over time in some coral reefs at Reunion Island. During their monitoring across 64 Mesoamerican reefs, Randazzo-Eisemann et al. 26 showed that the density of algal-gardening damselfish increased from 1.5 individuals per 100m 2 in 2006 to 4.7 individuals per 100m 2 in 2016, and that their density was strongly correlated with fleshy macroalgae cover. It is also recognised that some herbivorous fish and invertebrate micro-herbivorous grazers are competitors of farming damselfish by regulating the height of the algal canopy and inhibiting algal dominance on the reef 17,30 . However, sediment-rich algal turfs, that farming damselfish cultivate, may inhibit herbivory due to their high carbonate content as that may interfere with herbivorous fish digestion 31 . Thus, a reduction in herbivory will favor first denser turf algae patches, and subsequently the farming damselfish population. Conversely, if predator populations are reduced due to overfishing, the farming damselfish population may increase 26 .
In the present study, we utilized an extensive long-term monitoring dataset (2005-2019) to investigate how the dusky gregory (Stegastes nigricans, Tahitian name: 'atoti) abundances, the abundance of predators and algal competitors, and benthic cover vary through time inside and outside Marine Protected Areas at Moorea Island (French Polynesia). Given this species' close association with turf algae, we predicted that variations in the density of S. nigricans would correlate with variations in the proportion of living coral and algal covers. We also predicted a negative association between predators and/or competitors populations and S. nigricans density. From 2005 to 2019, the density of S. nigricans increased significantly inside and outside MPAs (Mann-Kendall tests, τ14 = 0.52, P = 0.008) ( Fig. 2A and B). In 2005, the density was 2.64 ± 0.25 fish per 50m 2 (mean ± SE) on reef flats throughout all sites. In 2019, the density was 12.06 ± 2.19 fish per 50m 2 . A significant increase in fish

Changes in fish predators and competitors since 2005.
In 2005, the predator density inside the MPAs was 2.00 ± 1.00 fish per 50m 2 . Following this, there was an increase to 6.67 ± 2.19 fish per 50m 2 in 2006, before stabilizing between 3.00 ± 1.00 and 3.67 ± 1.20 fish per 50m 2 from 2007 to 2010. From 2011 to 2019, predator density was similar to 2005 levels, ranging between 1.00 ± 0.00 and 2.00 ± 0.82 fish per 50m 2 ( Fig. 2A). The density of predators inside the MPAs showed a significant decreasing trend (Mann-Kendall test, τ 14 = -0.43, P = 0.008). The same pattern was observed outside the MPAs; however, this trend was not significant (

Discussion
Our long-term monitoring data add to the ongoing debate regarding the role that Stegastes play as macro-scale ecosystem engineers on coral reefs, and provides insights into the factors that regulate their abundances. Over the 15 year study period (2005-2019) the density of Stegastes increased from 0.05 to 0.24 fish per m 2 . This increase was not associated with changes in the cover of live hard corals, macroalgae, or algal turf inside or outside the MPAs (Fig. 2). However, while we found no relationship between predator density and S. nigricans densities in the same year, a significant negative relationship was found across the whole study period outside of MPAs when considering predator density from the previous year, and this negative relationship was significant both inside and outside of MPAs during periods where predator density was relatively high (e.g. 2005-2010) (Fig. 3). These data suggest that predators may play a role in regulating the density of this species on coral reefs, especially when predator densities are high. While correlative, our results provide insights into some of the ecological factors that may regulate the abundance of S. nigricans. Most notably, our data suggest that high predator densities may suppress the abundances of S. nigricans in the following year. This result was expected and is consistent with the findings of several comparable studies 29,34 ; however, the density of predators was only high for part of our study period (2005-2011). Predator densities were consistently low between the years 2012-2015, after which it rose again slightly towards the final years of the study. While it is difficult to discern whether predator densities were unusually low between 2012 and 2015, or whether they were unusually high between 2005 and 2011, it may be worth noting that Moorea Island was hit by Cyclone Oli in 2010 and the observed decrease in predator abundance appears to align with that. Furthermore, while predator abundances decreased and stabilized at low levels in the years following Cyclone Oli, the abundance of S. nigricans increased both within and outside of the MPA areas. In addition to predator densities being generally low during this period 35 , the increase in abundance of S. nigricans is also likely a product of increased coral damage by the cyclone, which can support the development of Stegastes' farms 25 .
In contrast to past studies 19,36,37 , we observed no significant relationships between habitat substrates (hard corals, algal turf and macroalgae) and S. nigricans abundances. Interestingly, as similar to our finding that only high predation density affected S. nigricans density, the lack of an effect of S. nigricans on substrate composition may be due to their generally low densities around Moorea Island. For instance, we found that S. nigricans abundance increased to 0.24 fish per m 2 over the course of this study, which is substantially less than the 4.2 fish per m 2 reported by the study by Wilkes et al. 37 in Florida, which found that Stegastes may have substantial effects on benthic community dynamics. Additionally, Ceccarelli et al. 18 reviewed the role of territorial damselfishes as determinants of benthic communities, and concluded that most observations and experiments have been undertaken at the scale of individual territories, and found a strong influence of Stegastes on benthic communities 24,26,37,38 . As our work is a long-term study over a large geographic area, it differs from most of these past studies. We are cautious about drawing major conclusions on the role that S. nigricans play as a habitat modifier, given our results and their relatively low density around Moorea Island.
Overall, our long-term study conducted over a large spatial scale suggests that the density of S. nigricans around Moorea Island was not associated with variations in substrate composition or the density of their competitors. Instead, we found a negatively association with the density of their predators, but only when predator densities were relatively high. While correlative, our results complement the results of past work on this topic, which have tended to be much more targeted in their investigations. As a correlative study, we were unable to investigate the direct/indirect effects of additional factors such as the bleaching events that occurred in 2016 and 2019, ocean acidification, or temperature rise. Nonetheless, our results would suggest that predators may play a role in regulating the populations of keystone species, such as algae-farming damselfishes, in order to avoid "algal phase-shifts" in coral reef ecosystem. Stegastes nigricans are mainly found on shallow reef flats (i.e., reefs contiguous to the coast-less than 400 m from the coastline) around Moorea 32,[39][40][41] . Therefore, within each site (eight MPA and five non-MPA sites), only shallow reef flats were considered this study (Fig. 1). The surveys were done between 8:00 and 11:00am, once per year (in February-warm and wet season).

Methods
Fish surveys. For the fish surveys, a belt transect with an underwater visual censuses sampling technique (snorkeling) was used. A 25 m linear transect was placed at the center of a 2 m-wide belt. All fish seen along the transect were identified to the species level. On the first transect pass, the observer recorded highly mobile fish that entered the transect but usually fled as a snorkeler approached. On the second pass, less mobile, cryptic, and site-attached species were targeted with more detailed examinations of crevices. For substrate surveys, the cover proportions of live hard corals, dead corals with algal turf, macroalgae, sand, rubble, and others (e.g. anemones, shells, soft corals) were sampled using the Point Intercept Transect method every 50 cm over 25 m of the belt transect used for the fish survey. Sampling was conducted along three transects (three replicates) separated by 25 m at each site. The three belt transects were set up in the middle of each shallow reef flat (depth: 1 m).
A total of 221 fish species were identified on the fringing reef of all MPA or non-MPA sites at Moorea 35 . The classification of these species as predators or competitors of S. nigricans was based on direct observations made during previous research at these locations 11,[39][40][41][42][43] . When research on a particular species was not available in the literature, we relied on expertise provided by Prof. Galzin, who has worked on coral reef fishes at these locations continuously since 1974 [39][40][41][42][43]  Statistical analyses. The normality of the distributions of the density of S. nigricans, predators, and competitors was tested with Shapiro-Wilk tests (W = 0.53 -0.61; P < 10 −3 ), like the distributions of the proportion of hard live corals, algal turf, and macroalgae (W = 0.64 -0.87; P < 10 −3 ). Temporal trends for these variables were then evaluated from 2005 to 2019 both inside and outside MPAs with modified Mann-Kendall tests for serially correlated data using the approach proposed by Hamed and Rao 45 with variance correction to address potential autocorrelation, using the R package "modifiedmk" 46 .
The annual average density of S. nigricans was correlated to the annual density of its predators and competitors with Spearman's correlations tests. The annual average density of S. nigricans was correlated to the annual proportions of live coral, algal turf, and macroalgae with Spearman's or Pearson's correlations tests. As a latency in the response may occur, the annual average density of S. nigricans in year n + 1 was also correlated to the annual density of predators, competitors and the proportions of live coral, algal turf, and macroalgae in year n. In our survey protocol, S. nigricans juveniles become adult fish after one year on the reef 40 . Moreover, predators (Serranidae, Muraenidae and Scorpaenidae) may eat juvenile S. nigricans 32 . Therefore, we correlated the annual average density of S. nigricans in year n and n + 1 to the annual density of predators in year n. The statistical analysis was conducted using R-Studio and R version 3.5.1 47,48 at the significance level α = 0.05. Ethical approval. This study did not involve endangered or protected species and was carried out in accordance with the guidelines of the French Polynesia Code de l'Environnement for animal ethics and scientific research (https:// www. servi ce-public. pf/ diren/ parta ger/ code/). Moreover, the visual surveys (no experiment conducted on fish) were approved by SNO CORAIL (licensing committee: PGEM 2004-http:// obser vatoi re. criobe. pf/ wiki/ tiki-index. php). Lastly, the study was carried out in compliance with the ARRIVE guidelines (http:// www. nc3rs. org. uk/ page. asp? id= 1357) to improve the reporting of research involving animals.

Data availability
All data generated and analysed during this study are available upon reasonable request to the corresponding author (DL).