Methodology:

Detailed acoustic surveys for mesopelagic fish were conducted in the Scotia Sea as part of the British Antarctic Survey's Discovery 2010 programme, as described in Collins et al. (2012). Follow up investigations were carried out using the same methods during JR15004 and JR16003. We used an EK60 echosounder operating at 38 kHz to collect acoustic data, pinging at a rate of 0.3 - 1 Hz. Nautical Area Scattering Coefficient (NASC, m2 nmi-2) a linear measure of acoustic backscatter, was integrated between depths of 17 m to 1000 m or maximum available, in 1 km distance sampling units. Acoustic data was coupled with environmental data from weekly or monthly climatologies. All environmental data was extracted corresponding to the same date that the acoustic sample was collected or the nearest prior date (for weekly or monthly climatologies), by latitude and longitude, with the exception of net primary productivity, which was extracted from a mean climatology spanning Jan-Mar 2005-2017, as this was being used to characterise regions of high and low productivity.

The environmental data were compiled from external data sources using r package raster to re-project NPP and Sea Ice data to WGS84 using bilinear interpolation. Geostrophic current speed was calculated from eastward and northward velocity data. Environmental data includes: sea surface temperature (°C), bathymetry (m), geopotential height (m), geostrophic current speeds (ms-1), sea ice concentration (%), sea ice zone, time of day (day/night), sea temperature at 200m (°C), distance to coast (km), net primary productivity (mgC m-2 d-1), and hours of daylight.

Data collection:

Acoustic Data collected using a Simrad EK60 echo sounder

Acoustic Data processed using Echoview version 8.0.95.32073

Analysis in R version 3.5.3

Data quality:

The acoustic data was calibrated at least once per field season using standard calibration spheres and methods (Demer et al. 2015). The data was cleaned in Echoview to remove bad data regions, including false bottom echoes, seabed, surface near-field, intermittent noise and attenuated signal (all set to 'no-data'). Data was integrated in 1km distance by 10m depth cells. Further removal of data in water shallower than 1000m was carried out in R version 3.5.3 and the top 1% of high NASC values were manually scrutinised to check for any residual noise bias. Less than 10% of high NASC values were found to be attributable to noise and these were excluded from further analysis.

For environmental data coupled with NASC, Environmental data was extracted at the location of NASC using r package raster (Hijmans, 2018). The majority of climatology data was already in WGS84 projection, with the exception of Sea Icea and NPP data which were reprojected rasterProject() funtion with bilinear interpolation.