Study Sites and Sampling Design
Surveys were conducted around five islands in the Bahamian archipelago: Andros, Bimini, Exumas, San Salvador and South Caicos in the Turks and Caicos Islands (figure 1) between July 2002 and November 2003 and during October 2004 in Exumas.
Benthic composition was analysed by documenting the species identified from 30 - 40 randomly placed 1 m2 quadrats at each site. The content was filmed using a high resolution digital camera for later analysis and identification to the highest taxonomic level possible. Areas of high relief or those containing cryptic organisms were filmed in greater detail. Depth and rugosity (maximum vertical relief assigned to one of five categories: 0-10 cm, 10-50 cm, 50-100 cm, 100-200 cm and >200 cm) were also noted at time of filming. Taxonomic data were later converted to frequency of occurrences at each site (number of times observed divided by the number of quadrats).
Montastraea forereef was sampled at two or three sites (~150 x ~150 m) about 1 km apart in three separate areas around the islands (about 5-10 km apart). The percentage coral and macroalgal cover were quantified at each sample site from a mean of 14.1 of the 1 m2 quadrats used for benthic characterisation; percentage cover was documented as the mean of five randomly sampled 20 x 20 cm (0.04 m2) sub sections within each 1 m2 quadrat.
Fish were divided into three groups by size class and surveyed using discrete group visual fish censuses (Green & Alevizon 1989). Fish size (to the nearest centimetre) and density were estimated along belt transects at each site around the five island systems. The number and size of each transect were optimised using previous data from equivalent surveys of the Caribbean (Mumby et al 2004). Small demersal fish such as the Pomacentridae were surveyed on four 30 x 2m transects; Scaridae and other mid-sized demersal fish were surveyed on ten 30 x 4m transects, and large demersal and pelagic fish were surveyed on five 50 x 4m transects.
To assess the abundance of both Scaridae and Pomacentridae the density of each species and phase (adult, intermediate, juvenile) was calculated by dividing the number of individuals observed at a site by the number of transect surveys and then standardising the result to an area of 200 m2. For predator species where size is a more important factor than abundance, the biomass was calculated by converting fish lengths to mass values using Bohnsack and Harper's (1988) allometric scaling relationship:
log(M) = log(a) + b[log(L)]
where M is mass in grams, L is fish length in millimetres and a and b are constants.
In terms of this study into Scaridae recruitment, recruits were identified as juveniles post settlement from the plankton. The four most abundant species were chosen for analysis: the princess parrotfish (Sparisoma taeniopterus), redband parrotfish (Sparisoma aurofrenatum), stoplight parrotfish (Sparisoma viride) and the striped parrotfish (Scarus iserti), due to their significantly higher juvenile abundance at the sites surveyed.
All Pomacentridae adults encountered were included in the analysis as territoriality is well known within this group. Five species of damselfish adults were encountered: the beaugregory damselfish (Stegastes leucostictus), bicolour damselfish (Stegastes partitus), cocoa damselfish (Stegastes variabilis), dusky damselfish (Stegastes adustus) and the threespot damselfish (Stegastes planifrons). Juveniles of other species were present in the surveys but are not included in this analysis.
Potential predators were defined by trophic group. All piscivorous organisms were included in the predator biomass count, irrespective of whether fish only made up part of their diet or they were exclusively piscivorous.
Initially, to determine which underlying mechanisms might have an effect on recruitment, different variables were plotted against juvenile parrotfish density for each species individually as it is well documented that post-settlement processes tend to be species specific (Tolimieri et al 1998, Gust 2002, Valls et al 2008), making it important to differentiate between parrotfish species in these statistical analyses. In order to assess the strength of the post-settlement processes, it is important to note that none of the processes affecting recruitment are acting in isolation and all potential effects should be analysed together using multivariate methods.
In order to determine which factors were affecting the abundance, and therefore recruitment, of juvenile parrotfish in the Bahamas I performed a linear mixed effects model in R (version 2.10.1), specifying islands and reef position within islands as random effects. The linear mixed effects model was the most appropriate due to the hierarchical nesting of sites within the dataset. I initially included all possible fixed factors then simplified the model by removing the least significant factor, and used analysis of variance (ANOVA) to determine whether simplification had significantly altered the model outcome.
The variables I included in the model were: (a) coral cover, as previous studies have shown a correlation with recruitment (Tolimieri 1998a,b, Gust 2002); (b) macroalgal cover, as parrotfish are herbivores it represents a potential food source; (c) rugosity, as highly rugose benthos has been shown to increase levels of recruitment at small spatial scales (Valls et al 2008) and could be an indicator of structural complexity which Tolimieri (1998a) found to increase the abundance and diversity of demersal fishes; (d) predator biomass, as a high abundance of predators is likely to increase recruit mortality; (e) damselfish density, as aggression from damselfish has been shown to reduce recruitment in the US Virgin Islands (Tolimieri 1998b);(f) conspecific density, as Tolimieri (1998b) found that presence of conspecifics increased persistence of recruits, and (g) an interaction between rugosity and predator biomass as rugose environments provide more potential shelter so I would expect rugosity to be an important factor when predator biomass is high.
Percentage cover data for coral and macroalgal cover were arcsine-transformed to ensure a normal distribution for statistical analysis.