This study aims to evaluate the degree of association between fishes and coral by conducting a meta-analysis on peer-reviewed published literature extracted from two databases. We extracted 7863 articles from the two databases (Web of science and scopus), which after removing for duplicates, yielded 4653 unique articles. Articles were first screened for their title and abstract, and subsequently for their full text, both of which utilised a deliberate decision tree (provided in methods/supplemental). Within this decision tree, the inclusion criteria was formulated such that the final set of studies only contained original research articles that were 1) conducted on tropical, shallow (<30 metres) coral reefs where corals and fishes (unbaited) were synchronously surveyed in-situ, 2) contained appropriate metrics (percent coral cover, fish abundance, biomass, and species richness) from which fish-coral associations can be quantified, and 3) provided clear documentation of the statistical analyses (e.g. sample size, statistical test used etc.) used to obtain the result. First, a scoping/preliminary search for key literature (including existing reviews on fish-coral associations) was performed to familiarise on typically used nomenclature, and to subsequently formulate specific keywords to be used in the systematic search string. Keywords were chosen (see decision tree included in methods/supplemental) to allow for a broad, yet relevant collection of studies to be gathered. These keywords were typical terms used in empirical coral reef studies in which corals and fishes were quantified synchronously for the metrics of interest (i.e. coral cover, fish abundance, biomass, and species richness). As we only wanted studies that had documented species richness along with surveys of abundance/denisty and/or biomass, richness was not included as an initial search term. Through this process, we found that certain studies (particularly those pre-dating 1990s) did not have their abstract recorded in the database. Thus, another search string was formulated to account for this (see decision tree). This search string basically was to account for studies that had no abstract and hence we used used AND NOT with terms that would commonly be in an abstract (e.g. the OR a OR and). Our final dataset includes 723 effect sizes extracted from 66 papers (from 2 databases: web of science and Scopus) that document a statistical relationship between percent coral cover and fish metrics (abundance, biomass and richness), with articles published between 1984 and 2022.

Software/equipment used to create/collect the data: Microsoft excel

Software/equipment used to manipulate/analyse the data: Rstudio