Computational protocol: Functional, size and taxonomic diversity of fish along a depth gradient in the deep sea

Similar protocols

Protocol publication

[…] During the survey, catch was identified to the finest taxonomic resolution possible, which was species level for 99.9% individuals caught. The full classification of taxa was determined using the World Register of Marine Species (). Photographs taken on board in 2013 were used for subsequent morphological measurements using the measuring software ImageJ (). Morphological data were collected for 31 species () that together account for 84% of all biomass caught over the duration of the survey. These species were selected for their abundance and in order to include species that span all depths. Data for additional species could not be collected due to time constraints in photographing individuals on the survey. Measurements were replicated by using photographs of multiple individuals; the number of replicates differed among species ().The morphological measurements taken using photographs were total length, head length, tail height, tail surface area, eye size, eye position, angle of mouth in relation to lateral line and surface area of mouth protrusion if present (; –). Mouth height and mouth width were measured on board due to the difficulty of photographing the mouth. Gape size was then calculated as the area of an oval with mouth height and mouth width as the diameters (; ). The tail measurements were used to calculate the aspect ratio of the fish, which can be used to deduce activity levels (; ; ). Head, eye, surface area of mouth protrusion and gape size were divided by total length in order to calculate relative trait values (). Relative traits were used in all analyses because body size varies substantially within species. By controlling for body size, the relative trait value can be assumed to be constant throughout an individual’s life because it represents an inherent body plan. Relative traits represent differences in function between species regardless of body size ( and references therein). The individual correlations between each of the continuous traits can be found in .Total length was measured on board the survey for all individuals caught, in addition to the measures taken using the photographs of subsets of individuals. For 12 (39%) of the 31 species for which morphological measurements were taken (hence the species on which most analyses presented here focus), it was inappropriate to measure total length due to tails commonly breaking off in the net, so alternative measurements were taken and converted to total length using conversion factors calculated from a subset of the data ().Subsets of the survey data were used to calculate conversion factors for translating the total length measurements to weight. Predicted weights were then standardised by controlling for the duration of time spent trawling. This measure of biomass caught per hour of trawling was used in all further analyses as the measure of abundance.The species-level measure of body size to be included in the calculation of functional diversity was the maximum recorded length of a species, or Lmax. Lmax was set as the maximum length listed on FishBase () or the maximum length recorded on the survey, whichever was the greater (). Of the 31 species for which morphological data were available, one (Apristurus aphyodes) did not have an Lmax listed on FishBase. Therefore, its Lmax was set as that of the largest species of that genus caught on the survey (Apristurus manis). Standard Lengths on FishBase were converted to total length using conversion factors calculated from the survey data where possible (). If there was no survey-derived conversion factor available (due to total length being measured on the survey and Standard Length being provided by FishBase) then the conversion factor listed on FishBase was used. Where both conversion factors were missing, we used an average conversion factor that was calculated across all species caught on the survey for which there was a Standard Length conversion available.Stable isotope data were available for 21 of the species for which morphological data were collected. The stable isotope analyses are described in ; data are available at Dryad Digital Respository doi: 10.5061/dryad.n576n). The isotopic dataset was compared to a meta-dataset of diet studies based on stomach content analyses (). Where species were present in both datasets, stable isotope compositions clearly distinguished between species categorised as feeding on either benthic (seabed) or pelagic (water column) prey (). Stable isotope compositions were subsequently used to assign feeding guild to species and individuals lacking reliable stomach content data (). The distinction between benthic and pelagic feeders was less pronounced in the assemblage at 500 m, as the diets of the two guilds are similar at this depth. However, species could still be assigned to a feeding guild based on their relative isotope signatures throughout the rest of their depth range. Specialised signatures within these two feeding guilds could be established in some cases: if the smallest individual sampled for that species was in the upper half of stable isotope space for that category, the species was defined as high trophic level; if the largest individual sampled was in the lower half of stable isotope space, the species was defined as low trophic level; fish that feed on benthic suspension feeding prey have a noticeably enriched isotope signature for a given body size, depth and feeding guild, so were categorised separately. [...] Diversity was calculated in four ways: 1) functional richness, 2) functional divergence, 3) size diversity, and 4) species richness. The two measures of functional diversity are described by and were calculated using the R () package FD (). Functional richness is an estimate of the degree to which the assemblage fills functional space (; ) and functional divergence measures how abundance is distributed within the volume of functional trait space occupied by species (; ). The traits included in the calculation of functional diversity were relative head size, aspect ratio of the caudal fin, relative eye size, eye position, angle of mouth in relation to lateral line, relative surface area of mouth protrusion if present, relative gape size, and Lmax (). A species-level mean was calculated from the relative trait values for all continuous traits (). Functional richness does not include species abundances in its calculation; functional divergence includes a weighting of traits by species abundance, which in this case was biomass caught per hour of trawling. Due to only having trait data for a maximum of 31 species, functional diversity was only calculated using those species and their biomasses, and the rarer species were not considered. As these 31 species accounted for 84% of all biomass caught on the survey and spanned the entirety of the depth range studied, they were considered to be a good representation of the study system.Size diversity was calculated using the generalised measure of diversity proposed by . In this index, abundance of biologically meaningful groups and similarities between them are accounted for. Here the groups were size classes each of 10 cm in width, and abundance was calculated as the proportional biomass per hour that each size class accounts for in each station, when only the species for which morphological data were known were included. The Euclidean distance matrix (d) between the mid-points of size classes was converted to similarities using the formula suggested by : Similarity=elog(2)*d The final input for the Leinster-Cobbold measure of diversity is the sensitivity parameter, q, which determines how much emphasis is given to rare species (or in this case, size classes; ). Here a value of q = 1.1 was used in order to balance the richness (lower q) and evenness (higher q) components of diversity, and to be comparable to the widely used Shannon index (; ).Species richness was calculated using only hauls that were of 120 ± 5 min in duration in order to control for sampling effort. For this subset of hauls, the number of species present was averaged across hauls in each station. All species were included in the calculation of species richness, not just those with morphological data available. This is because calculating species richness using only the morphological subset would merely be a count of the number of species for which morphological data were available and not be meaningful in a diversity context.The four diversity measures were calculated for each station and then analysed with respect to the depth of that station with Generalised Additive Models (GAMs) using the R () package mgcv (). A smoother function of depth was used as the predictor, and the values for the test statistic, significance, R-squared, and effective degrees of freedom (e.d.f.; the flexibility of the fitted model; ) were extracted from the model summary.Abundance-weighted station means were calculated for each continuous morphological trait included in the functional diversity metric and analysed with respect to the depth of the station using GAMs. The weighted mean was said to be the mean value across species, where values were weighted by the biomass caught per hour of trawling for each species. The mean observed size of individuals, irrespective of species identity, was also calculated for comparison. This value was not included in the functional diversity metric because a species-level measure of body size (Lmax) was needed. The station mean body size was therefore calculated as the average length across individuals in a station, when only individuals of species for which morphological data were obtained were included, in order to be comparable to the measures of functional and size diversity. The standard deviation of each continuous trait at each station was also calculated and analysed with respect to depth using GAMs in order to relate variation in traits to patterns seen in functional diversity. The Pearson’s product-moment correlation coefficient was calculated for the relationships between the means and standard deviations of each of the traits, and each measure of diversity.The isotopic feeding guild data () were interpreted using the percentage of biomass that each guild accounted for in depth bands of 200 m in width. The percentage was calculated as a proportion of the biomass accounted for by the species for which there were morphological data.All data manipulation and analysis was performed using R version 3.1.2 () and figures were produced using the packages ggplot2 (), gridExtra () and marmap (). […]

Pipeline specifications

Software tools ImageJ, Ggplot2
Applications Miscellaneous, Microscopic phenotype analysis
Diseases Pulmonary Fibrosis