Study site
The study was carried out randomly on three dates (October 2005, February 2006 and May 2006) in three offshore gas platforms (Luna A, Luna B and Hera Lacinia) located in the southern Ionian Sea (central Mediterranean Sea), respectively, 5.3, 6.2 and 2.6 km offshore (Fig. 1). Two of them (Luna A and Luna B) are 8-leg platforms while the third one (H. Lacinia) is a 4-leg platform. All these platforms lie on a sandy seabed in 75, 90 and 30 m of water, respectively, and are fixed to the sea floor by concrete or steel legs, which are connected by an assemblage of cross beams.
The platforms were colonized by several invertebrates providing shelter and food to cryptic and nekto-benthic fish species. The most represented sessile organisms were Mytilus galloprovincialis, Balanidae indet., Ostrea sp., and Arbacia lixula (personal observation).
Sampling methods
Fish species and their abundance were recorded by UVC and a ROV (“Hyball” of Hydrovision Ltd) equipped with a high resolution video camera, compass, rate gyro and two fixed 100 W quartz halogen lights aimed forward and two 75 W lights mounted on the camera chassis.
At each sampling month, ROV and UVC surveys were performed randomly on three out of 8 pillars at Luna A and B platforms (8-leg) and out of 4 at Hera Lacinia platform (4-leg). Two depth ranges were chosen: 0–6 and 12–18 m thus dividing pillars into 6 m-high units.
During both UVC and ROV surveys, the underwater “Mobile Point Count” technique was adopted in order to study fish assemblages. This technique, specifically designed for offshore platforms by Rilov and Benayahu (2000) and applied, for the first time, by Consoli et al. (2007) and Andaloro et al. (2011) in the Mediterranean Sea, was chosen because it is particularly suitable for the analysis of species strictly associated with the pillars and also for detecting benthic and cryptic species (Consoli et al. 2007; Andaloro et al. 2011).
The diver (in the case of UVCs) or the ROV, turning around each unit and looking toward the pillar, censused all fishes occurring up to 3 m from it. In particular, they began, firstly, to record the more conspicuous and easily identifiable fishes from a maximum distance of 3 m from the pillar (so that to have an entire view of the census unit) and straight after, approached the pillar and counted the benthic and crypto-benthic species. Since the pillars had a radius of 1 m, the total censused volume corresponded to a cylinder of 7 m of diameter and 6 m high, from which the volume of the pillar (1 m of diameter, 6 m high) was subtracted. The resulting volume was 225.7 m3.
A total of 108 observation units were carried out in study period. Moreover, the ROV was used to survey three deeper depth layers: 22–28 m at H. Lacinia, Luna A and Luna B, 50–56 m at Luna A and Luna B and 70–76 m at Luna B. ROV and UVC trials were carried out in different moments of the same days by exploring the same horizontal units within a maximum time of 5 min.
Fish abundance was estimated by counting single specimens to a maximum of 10 individuals, and using abundance classes (11–30, 31–50, 51–100, 101–200, 201–500, >500) for schools. This recording system leads to a similar degree of error over a wide range of abundances, ensuring homogeneity of variance after log-transformation of the data (Frontier 1986; Guidetti et al. 2003). ROV trials were recorded on video tapes and four different researchers contributed to both underwater and video data recordings thus reducing any systematic error between methods.
Fish taxa were pooled into ecological categories on the basis of their spatial organization and behavior in relation to the platform: PLA, planktivorous fishes (e.g., Chromis chromis, Anthias anthias and Boops boops) which form schools and are strictly associated to the artificial structures mostly for shelter and reproduction issues; NEC-BEN, nekto-benthic species (mostly Sparids) feeding on sessile organisms; CRY, crypto-benthic fish (e.g., Blennids and Scorpaenids) which are visually and/or behaviorally cryptic and keep them in close association with the substratum (La Mesa et al. 2006); PEL, pelagic fish (e.g. Scombrids and Carangids) attracted by the artificial structures.
Data analyses
The fish community was described by means of fish abundance (N), species richness (S) and frequency of occurrence (%O) data.
A three-way permutational multivariate analysis of variance (PERMANOVA) (PERMANOVA; Anderson 2001; McArdle and Anderson 2001) was used to test for differences between fish assemblages with regard to the factors method (2 levels, fixed), platform (3 levels, fixed) and period (3 levels, random) at each depth layer. The analysis was based on Gower distances calculated on log(x + 1) transformed data, and each term of the analysis was tested using 4,999 random permutations of appropriate units (Anderson and ter Braa 2003). This permutation method is generally thought to be best suited because it provides the best statistical power and the most accurate Type I error (Anderson and Legendre 1999).
Moreover, to test whether ROV and UVC recorded different average abundances of each of the four ecological categories at each depth layer, a three-way permutational univariate analysis of variance (permutational ANOVA) was performed considering the same factors used in the multivariate analyses. This univariate analysis was also used to detect significant differences between the mean values of species richness recorded by both methods. Unlike multivariate analyses described above, we used a Euclidean distance in this univariate model. Data were transformed at the log(x + 1) in order to reduce the weighting of abundant categories and increase that of the rarer ones.
Both univariate and multivariate analyses were employed using the software package PRIMER 6 with PERMANOVA+add-on (Anderson et al. 2008).
Whereas ANOVA/MANOVA assumes normal distributions, PERMANOVA works with any distance measure that is appropriate to the data and uses permutations to make it distribution free. Thus, the same F-statistics were calculated, but p values were obtained by permutation.
A two-dimensional non-metric multidimensional scaling (nMDS) plot was generated on the basis of Gower similarity matrix of abundance data averaged per platform.
Moreover, the similarity percentage procedure SIMPER (Clarke and Warwick 1994) was used to identify the fish species mostly contributing to the differences between the methods. In this case, crypto-benthic fish were pooled into a single group.
Finally, a rough cost estimate for both methods was provided considering the rent of the boat, ROV and diving equipment, the manpower needed for the surveys, video analysis and data elaboration.