Mangrove Bay
Mangrove Bay (25.8723° N, 34.4179° E), a resort located 29 km South of Quesier, Egypt, attracts snorkelers and divers from all over the world, owing to its proximity with the Red Sea. The diving station offers several dive sites with a great diversity of landscapes and organisms. The house reef is located right in front of the station. It is a vast bay, sheltered from currents and winds with many dive possibilities and great landscape diversity, going from mildly sloping sandy areas mottled with coral blocks and coral gardens, to impressive drop off areas. The other dive sites are located further from the coast and range from charming underwater coral gardens (Erg Fugani) to more sloppy and sandy areas with sparse coral blocks (Shaab Afrid). With its diversity of dive sites, landscapes and depths, the area provides a great study site for reef surveys.
Field Methods
Field data were collected in Mangrove bay between the 14th and the 19th of October 2021. The present bay is one of the largest in Quesier which ensured for the data to be collected at different sites, preventing oversampling of specific areas and ensuring the representation of the entire bay in our data. Most of these sites were reachable by swimming, whereas others had to be reached by means of a small boat. As for the conditions, Mangrove bay is protected from winds and waves, reason why no underwater currents were ever encountered while collecting the data.
Our team consisted of four divers divided into two buddy teams, each diving on average twice per day. During a dive each buddy team would perform between two to four transects in randomly selected locations and no transect area was ever replicated. Furthermore, transects covered three different depth ranges being the shallow (6.5-7.5m), the medium (9-10m) and the deep one (14.5-15.5m). In total we performed 26 transects, of which 23 were then included in the data analyses.
To address the relationship between fish families' abundance and coral reef complexity, we decided to use a transect methodology. The transect line was placed parallel to the reef crest and it consisted of a 30x2m belt of which width was visually estimated by the divers. Once the site was reached, the fish surveyor would attach the extremity of the underwater measuring tape to a dead coral and start slowly swimming either towards the right or the left of the reef while keeping a constant depth and releasing the measuring tape. Meanwhile, they would count roving fish of the different families considered and report the numbers on an underwater slate. Once the 30m were reached, the fish surveyor would swim back and report the territorial fish counts instead. Throughout this process, the coral surveyor would swim a few meters behind the other diver to avoid stressing and scaring fish away. In order to assess coral reef complexity of the site, we opted for a point transect methodology in which the surveyor would determine percentages of each topographic category and total number of coral growth forms for 1m2 quadrats estimated every 5m for a total of five quadrats per transect. For each transect divers would record the duration of the survey as well. Once both buddy teams had finished surveying their respective transect, divers would swim towards the next location.
CORAL COMPLEXITY
Scientists have used different variables to describe coral reef complexity such as topographic rugosity of corals, substratum diversity, varieties, numbers of refuges and so on. We decided to simplify Gratwicke and Speight 2005 methodology and to create a score system that allows us to describe coral reef complexity based on two of these variables being the rugosity/topography of corals (indirectly linked to shelter availability) and the number of growth forms. First of all we decided to distinguish between the following growth forms : Encrusting, Massive, Meandering, Solitary/Mushroom, Foliose, Laminar, Tabulate, Branching, Corymbose/Caespitose and Columnar/Digitate. To score the number of growth forms, we decided to give a a score of 1 to an area inhabiting one single coral form. With every additional growth form we then decided to add 0.5 points to the score (see table below). For the topographic variable we then created three categories to which we assigned a different score of complexity (see image below). A score of one was given to the simple category which included those forms that can host fish only above and around them. The medium category, which was instead given a score of two, consisted of those forms allowing fish to shelter below the coral as well. Finally, a score of three was given to the complex category which included growth forms allowing fish to shelter also within the coral. Furthermore, all other substrates such as algaes, sand and rocks were grouped under a fourth category worth zero.
Statistical Methods
We initially aimed at determining how total abundance of fish and family diversity would vary according to both coral complexity and depth of the water column. However, given the small and unbalanced sample size we couldn’t include depth as an explanatory variable. Therefore, we decided to focus on coral complexity and treat depth as a random factor.
We created two separate main models, one for total fish abundance and the second for number of fish families. For the first model given the nature of the data, we used a generalized linear mixed effect model (GLMER) following the Poisson distribution in which coral complexity is a continuous explanatory variable and depth is a random factor on the intercept. Given a problem of overdispersion we had to use the negative binomial family instead, which allowed us to check for the assumptions of the model. For the second model, depth turned out to not act as a random factor, therefore we fitted a General linear model (GLM) following the quasipoisson distribution to check for the assumptions.
Additionally, we were interested in exploring a potential correlation between coral complexity and each of our two response variables. Therefore, since ours are count data and they represent a small sample size, we fitted a non parametric correlation test being the Kendall's rank one.
We then further investigated trends in individual fish families more closely. Only families with sufficient data were considered for this analysis, that is angelfish, butterflyfish, damselfish, groupers, parrotfish, and surgeonfish. To predict abundances of the single families using coral complexity as the explanatory variable, we fitted either a GLMER or GLM model based on whether or not depth turned out to play an important role as a random factor. In some occasions, because of overdispersion problems, instead of the Poisson family we had to use the negative binomial or the quasipoisson one with the GLMER and GLM models respectively. Nevertheless, we are aware of the small dataset available for each family, therefore we will not consider these analyses as being statistically relevant, rather we will only use them to predict trend and develop some general ideas about their differences.
​
A visual representation of the effect of depth on total fish abundance and number of families was performed as well.
​
The R code of the above mentioned analyses is available on the Appendix page.