The file "Zarco-Perello et al Temperate Reef Fish Trophic Guilds Complete Diet Dataset.xlsx" contains several spreadsheet tabs related to the analyses carried out in the paper: Refining the trophic diversity, ecological network structure, and bottom-up importance of prey groups for temperate reef fishes: https://doi.org/10.32942/X2CC97 All analyses, with the exception of the network calculations, of the study were carried out in the computer software R. The code is contained in the file "Zarco-Perello et al Temperate Reef Fish Trophic Ecology.R". For trophic network analyses we used the computer program Gephi v0.1 (Bastian et al. 2009). DATASET DESCRIPTION Region of Study The region of study encompasses all the temperate reefs of south-western Australia (SWA). Extending along ~1600 km of coast, from Jurien Bay Marine Park (30° 18.6 S, 115° 0.1 E) to the Recherche Archipelago Nature Research (33° 53.7 S, 123° 52.3 E; supplementary Fig. S1), the temperate reefs of SWA are distributed across the Leeuwin and Houtman biogeographical ecoregions (Spalding et al. 2007), conforming approximately ⅓ of the total distribution of temperate Australia, known as the Great Southern Reef (Bennett et al. 2016). Species Composition The species composition of the metacommunity of temperate reef fishes of the region was obtained from a total of 4589 underwater visual surveys conducted across 206 reefs in 12 locations by the Reef Life Survey (RLS) citizen science program, and the Australian Temperate Reef Collaboration (ATRC, with support from the Department of Biodiversity Conservation and Attractions; https://www.atrc.au) from 1997 to 2021. Trophic Information All fish species listed in the RLS-ATRC database were classified in trophic guilds based on collected diet information from studies of gut content analyses in SWA, or other Australian and international regions in the absence of local information. A total of 298 fish species composed the metacommunity. For every species, we obtained diet information from the scientific literature reported on Fishbase (Froese and Pauly 2019) and through the search engine Scopus using the search terms: TS = (name of species* OR *common name of species*) AND TS = (diet OR *stomach content* OR *gut content* OR consump* OR herbi* OR predat* OR feeding). Diet information consisted of the average proportions of food items represented as the number of items (%N), percent volume (%V), or biomass (%W) in a population of each species. Preference was given to diet studies conducted in the region of study and those presenting biomass proportions. Species that lacked diet information globally were assigned diet proportions based on phylogenetically related species with similar size and habitat preferences based on the Fish Tree of Life (Chang et al. 2019). > Trophic guilds classification To quantify the diversity of trophic guilds and identify important fish consumers of specific groups of prey, we classified the fish species into trophic guilds performing a multi-step cluster analysis. Firstly, species were grouped into main trophic guilds using the mutually exclusive major categories of prey items. The diet proportions in these categories were used to create a dissimilarity matrix among species based on the Bray-Curtis linkage method using the function vegdist of the R package Vegan (Oksanen et al. 2022), which was used to run a sequential divisive hierarchical cluster analysis using the function diana (divisive analysis) of the R package Cluster (Maechler et al. 2022). Subsequently, because there are mismatches in the resolution of diet identification between species belonging to different trophic levels (e.g. the diets of herbivorous fish tend to have higher resolution on macrophytes, while carnivorous species tend to have higher resolution on animal prey), species within each identified main trophic guild were subject to a cluster analysis with higher definition of prey items to identify groups of species with diet specializations using sequential agglomerative hierarchical cluster analysis based on Ward's Method and Bray-Curtis or Euclidean dissimilarity matrix (Pineda‑Munoz and Alroy 2014). The stomach content of most scarid species (parrotfish; Labridae: Scarinae) is very difficult to identify due to their pharyngeal mill, which grinds all food items to indiscernible particles. However, they are well identified as a special group that ingest detritus and algae by scraping the reef substrate with their specialized fused teeth. Thus, for the sake of differentiating their trophic guild, the proportions of diet for species of parrotfish was arbitrarily defined based on field observations as sediment and detritus (90%) and short filamentous algae (10%) (Bonaldo et al. 2014). Additionally, cleaner fish and false cleaners are a special group of fishes that are difficult to group by diet given that they feed on prey that could be identified as zooplankton or zoobenthos, while in fact true cleaners forage, at least in part, on parasitic invertebrates attached to bigger fish, in addition to fish skin and scales (Grutter 1997); thus, given their particular trophic ecology these labrid and blenny species were arbitrarily grouped in the major trophic group "fish cleaners" for the subsequent specialized trophic group classifications. Visual analysis of the differences in multidimensional space between trophic guilds was done with Non-metric Multidimensional Scaling based on the dissimilarity matrix calculated for clustering using the function metaMDS of the R package vegan (reported in supplementary materials; (Oksanen et al. 2022). Statistical significance in dietary differences among major and specialized trophic guilds (diet proportions ~ trophic guilds) was tested with permutational analysis of variance (PERMANOVA) using the function adonis2 of the R package vegan (Oksanen et al. 2022), followed by pairwise comparisons using the function pairwise.adonis2 of the R package pairwiseAdonis (Martinez 2017). > Trophic Network Links Between Specialized Guilds The trophic links between fishes and their invertebrate and macrophyte prey groups were identified by our trophic guild classification (Other Guilds Links tab in dataset); however, the trophic role of piscivores is faced with what here we called a "matrioshka paradox", because to know their links with other guilds, we must first know the trophic links of their prey. Moreover, this is not straightforward because the highest taxonomic identification of piscivorous prey is usually limited to family level, which could belong to multiple trophic guilds. This paradox is usually not explicitly stated in the literature, and it is unclear how trophic links have been drawn in previous studies without performing detailed quantitative trophic classifications. Here we estimated the trophic links between piscivorous guilds and the rest of fish guilds by (i) assigning each fish family identified in the diets of piscivorous fishes into their respective specialized guilds based in our trophic classification, (ii) pooling their diet proportions into each specialized trophic guilds they could belong to, (iii) standardizing values by number of species in each piscivorous guild, and (iv) dividing by the total sum of diet proportions to estimate their potential predation (0-100%) on other trophic guilds in the trophic network. Trophic links that had pooled diet proportions with values > >