Variation in both inter- and intra-specific traits affect community dynamics, yet we know little regarding the relative importance of external environmental filters vs internal biotic interactions that shape the functional space of communities along broad-scale environmental gradients, such as latitude, elevation or depth. We examined changes in several key aspects of functional alpha-diversity for marine fishes along depth and latitude gradients by quantifying intra- and inter-specific richness, dispersion and regularity in functional trait space. We derived eight functional traits related to food acquisition and locomotion, and calculated seven complementary indices of functional diversity for 144 species of marine ray-finned fishes along large-scale depth (50 m – 1200 m) and latitudinal gradients (29° – 51° S) in New Zealand waters. Traits were derived from morphological measurements taken directly from footage obtained using Baited Remote Underwater Stereo-Video systems and museum specimens. We partitioned functional variation into intra- and inter-specific components for the first time using a PERMANOVA approach. We also implemented two tree-based diversity metrics in a functional distance-based context for the first time: namely, the variance in pairwise functional distance, and the variance in nearest-neighbour distance. Functional alpha diversity increased with increasing depth, and decreased with increasing latitude. More specifically, the dispersion and mean nearest-neighbour distances among species in trait space, and intra-specific trait variability all increased with depth, whereas functional hypervolume (richness) was stable across depth. In contrast, functional hypervolume, dispersion and regularity indices all decreased with increasing latitude; however, intra-specific trait variation increased with latitude, suggesting that intra-specific trait variability becomes increasingly important at higher latitudes. These results suggest that competition within and among species are key processes shaping functional multi-dimensional space for fishes in the deep sea. Increasing morphological dissimilarity with increasing depth may facilitate niche partitioning to promote coexistence, whereas abiotic filtering may be the dominant process structuring communities with increasing latitude.

We observed 144 species of marine ray-finned fishes (Class Actinopterygii) on Baited Remote Underwater stereo-Video (stereo-BRUV) footage. We analysed data on the basis of the species present (observed in video footage) within each of the 47 depth-by-location cells. There were 144 species recorded, and 509 species-by-cell occurrences (many species naturally occurred in more than one cell). Our original dataset was comprised of a complete set of 15 raw morphological measurements for a total of 722 individuals (136 of these required some random-forest imputation, and missing traits were remeasured for 4 individuals) obtained directly from Stereo-BRUV footage. 

From this original dataset, we calculated all species-level functional metrics (i.e., FHV, MPFD, MNND, VPFD and VNND; see descriptions below). We created 100 tables of 509 unique species-by-cell occurrences (rows) for the 8 traits (columns; eye position, pectoral fin position, caudal peduncle throttling, elongation, eye size, oral gape position, jaw length/head length, and total length) by randomly drawing 1 individual from the list of all complete individuals for each species. To maintain any spatial structures in trait variability as well as possible, we drew an individual for each species within each cell from conspecific individuals that were (in order of preference): a) within that depth-by-location cell, b) at the same depth, or c) from anywhere within the Stereo-BRUV study design (n = 722) or d) from a museum specimen (n = 291). All species-level functional metrics were calculated for each replicate table, and we calculated the mean across all 100 tables for every metric for subsequent analyses.

All functional metrics were calculated using 8 normalised continuous traits. We calculated the following species-level metrics for each depth-by-latitude cell, for each of the 100 species-by-trait (509 x 8) data matrices after calculating Euclidean distances: (i) mean pairwise functional distance (MPFD; (Clarke & Warwick 1998; Somerfield et al. 2008; Swenson 2014), (ii) mean nearest-neighbour distance (MNND; Swenson & Weiser 2014), (iii) variance in pairwise functional distance (VPFD; adapted from Clarke and Warwick (2001) and Somerfield et al. (2008)), and (iv) variance in nearest-neighbour distance (VNND; Swenson (2014). We also performed principal component analysis (PCA) on the normalised traits in order to calculate functional hypervolume (FHV; Blonder et al. 2014; Blonder et al. 2018). FHV was calculated using the first 4 principal component axes (which accounted for 70.2 – 74.4 % of the total variation in the 8D functional trait space across the 100 species-by-trait tables). We did not retain all 8 dimensions due to difficulties associated with calculating FHV when few species were present. FHV has been used as a proxy to estimate niche space, including high-dimensional, irregular spaces (Lamanna et al. 2014; Cooke, Eigenbrod & Bates 2019). 

For metrics focusing on intra-specific trait variability we used data solely from the in-situ stereo-BRUV footage (i.e., the dataset comprising 722 individuals). Due to the inability to measure every species observed on the stereo-BRUVs, this dataset represents a reduced number of species (62 out of 144), and cells (43 out of 47). Within this dataset, we were able to measure intraspecific trait variability for 42 species (2 or more individuals per species per cell). There were, on average, 3.34 species per cell (min = 1, max = 10, sd = 1.86) and 4.32 individuals per species per cell (min = 2, max = 15, sd = 2.5) to measure the intra-specific trait variability. 

We calculated mean pairwise functional distance (MPFD.I) directly, considering only the intra-specific distances. In addition, partitioning was done by performing a PERMANOVA on the Euclidean distances among all complete individuals separately within each cell. Different species were treated as different levels of the factor “Species”, and individuals within each species were treated as replicates in a one-factor design. Prop.I is equal to the proportion of total trait variation (within each cell) attributable to individual-level variation (i.e., Prop.I).
Prop.I and MPFD.I could only be calculated when there were two or more individuals representing the same species within a depth-by-location cell.
 

We provide everything else that is needed to know in the metadata documents for each data table.