Biological sample design, data collection and analysis

2.3.1 Sample design

Video transects were undertaken over both megahabitat features and used to ground-truth the acoustic data. Transect length ranged between 0.5 km and 3.3 km, but were for the most part approximately 500 m long. Variations in transect length was either a result of poor sampling conditions (e.g. rough terrain, strong currents) or the desire to sample complete geomorphological features (e.g. a ridge). For the majority of transects, vessel speed was ~0.5 knots, with most transects lasting between 0.5-1.5 hrs. The drop frame was deployed from the starboard side of the vessel and towed in the water column 1-3 m above the seabed (dependent on substratum type, slope angle and currents) to capture the change in habitats, seabed substratum and larger conspicuous epifauna.

Bathymetry -735 -2077

28 2.3.2 Video/image collection

A Seatronics drop-frame camera system (Fig. 2.4) was used during each survey to enable characterisation of deep-water benthic habitats and seabed substratum. Camera-transects were selected using the multibeam bathymetry and backscatter datasets to capture varying inferred sediment type (inferred from backscatter intensity), geomorphological features and water depth. To ensure comparability between datasets, the same standard setup was used for each survey, although position of lights and camera varied slightly from year to year, thus, highlighting the importance of calibrating the field of view of the camera for each survey.

Fig. 2.4: The Seatronics drop frame camera system being deployed from the MV Franklin. Photograph taken by K. Howell.

The Seatronics drop-frame was fitted with a camera system, four lights set at oblique angles to the seabed to provide optimal illumination, and a flash unit to provide additional light for the collection of still images. The camera system comprised a DTS 6000 digital video telemetry system with a live feed to the surface, and a five megapixel Kongsberg Simrad digital stills camera (containing a Canon Powershot G5). The cameras were mounted opposite each other (with lights either side) at oblique angles to the seabed for optimal seabed coverage and to aid species identification. The frame was

29

also fitted with sensors to record depth, altitude and temperature, and an ultra-short baseline (USBL) beacon to collect accurate positional data for the frame.

Prior to data collection, the fields of view for both the stills and video cameras were calibrated by attaching a gridded quadrat of known dimensions to the camera frame which could be overlaid on stills images to allow quantitative analysis of fauna.

Calibrations were made for ‘on bottom’ (drop frame sitting on the seabed; Fig. 2.5) and at 1 m, 2 m and 3 m elevation off the seabed. The calibration grid allowed measurements to be made of area cover of encrusting, colonial and lobose growth form organisms.

Fig. 2.5: Image of the calibration grid (on bottom) overlaid on a sample image from the 2009 survey. The grid cell size is 4.9 cm (vertical in figure) by 5.5 cm (horizontal in figure).

Following the MESH guidelines for data collection, a 2-5 min camera stabilisation period was undertaken at the beginning of each transect to ensure the camera was moving at a constant speed. Video footage was recorded along the entire transect, and at

30

approximately one minute intervals the drop-frame was landed and a stills image taken (sampling unit) which will be referred to here as a ‘sample’ image. Additional images were also taken to capture abrupt changes in substratum and to aid in species identification.

2.3.3 Data analysis

‘Sample’ images (those taken approx. every minute) and those that captured abrupt changes in substratum were examined to assess their quality for analysis, those which were designated ‘of poor quality’ (i.e. obscured by silt clouds, out of focus, or too high off the seabed to identify organisms) were not included in the analysis. Remaining sample images were quantitatively analysed using the calibration grid as a measure of area.

An inherent problem with working with deep-sea imagery data is that it is difficult and often impossible to identify organisms to species level without the use of physical samples, this is particularly true for poorly-sampled regions. However, observed organisms can be identified as distinct morphospecies [Operational Taxonomic Units (OTU)] which can correspond to species, genus, family or higher taxonomic levels. The use of OTU numbers, rather than taxonomic identifications, adds an extra level of resolution to the data, as functional groups can be used and it also allows the data to be revisited and identifications changed, enabling the dataset to be more readily combined with others.

All visible organisms >1 cm (at their widest point) were identified as distinct morphospecies (morphotypes) and assigned an OTU number (See electronic appendix for species catalogue). OTUs were identified to the lowest possible taxonomic level.

31

This work contributed towards the development of an image catalogue of deep-sea species, which is now available on the web for use by the wider deep-sea community [Howell and Davies (2010) http://www.marlin.ac.uk/deep-sea-species-image-catalogue/]. All individuals were counted, except in the case of encrusting, colonial and lobose forms, where area cover was recorded using the calibration grid overlain on the image as a quadrat (as described by Underwood and Chapman 2005). Image data were standardised to individuals/1 m² and percent cover/1 m² for each taxon.

A measurement scale that has been adopted to define littoral and sublittoral biotopes is the SACFOR scale (Connor and Hiscock 1996) where abundance estimates are based upon individual counts or the percentage cover of organisms. As the scale used is a ranked abundance scale it is only semi-quantitative and a level of resolution is lost in the data. Therefore, it was decided not to use the SACFOR scale for this study. The high resolution of the still images obtained (ability to identify species as small as 1cm) and the full calibration of the stills camera undertaken, meant that high resolution, quantitative biological data was attainable and using the SACFOR scale would not accurately reflect higher abundance, smaller species (Underwood and Chapman 2005).

Additionally, full quantitative data was desired to use to measure diversity of the biotope, to aid in their characterisation.

To allow the combination of the two matrices for the cluster analysis and without losing partial resolution of the data by converting it to a semi-quantitative SACFOR scale, each matrix was standardised to the same scale (Stevens and Connolly 2004; Howell et al. 2010b). Count and cover data were treated independently prior to multivariate analysis, each were standardised to 1 m² (percent/1 m² for cover) and transformed according to the distribution of data. Standardisation per matrix was achieved by

32

dividing the matrix through itself and multiplying it by an appropriate factor to put the count and cover on relative scales (Prof. R. Clarke pers. comm).