Evidence on the impacts of chemicals arising from human activity on tropical reef-building corals; a systematic map

Tropical coral reefs cover ca. 0.1% of the Earth’s surface but host an outstanding biodiversity and provide important ecosystem services to millions of people living nearby. They are currently threatened by local stressors (e.g. nutrient enrichment and chemical pollution arising from poor land management, sewage effluents, agriculture, industry) and global stressors (mainly seawater warming and acidification, i.e. climate change). Global and local stressors interact in different ways, but the presence of one stressor often reduces the tolerance to additional stress. While global stressors cannot be mitigated solely by local actions, local stressors can be reduced through ecosystem management, therefore minimizing the impact of climate change on coral reefs. We systematically mapped the evidence of impacts of chemicals arising from anthropogenic activities on tropical reef-building corals, which are the main engineer species of reef ecosystems, to inform decision-makers on the available evidence on this topic. We searched the relevant literature using English terms combined in a tested search string in two publication databases (Scopus and Web Of Science Core Collection). The search string combined terms describing the population (tropical reef-building corals) and the exposure (chemicals). We searched for additional literature through three search engines, three dissertations repositories, 11 specialist websites, and through a call to local stakeholders. Titles, abstracts, and full-texts were successively screened using pre-defined eligibility criteria. A database of all studies included in the map with coded metadata was produced. The evidence was described and knowledge clusters and gaps were identified through the distribution and frequency of studies into types of exposure and/or types of outcomes and/or types of study. The initial searches identified 23,403 articles which resulted in 15,177 articles after duplicate removal. Among them, 908 articles were retained after screening process, corresponding to 7937 studies (a study being the combination of a taxon, an exposure, and an outcome). Among these studies, 30.5% dealt with the impact of nutrient enrichment on corals while 25% concerned the impact of human activities without reference to a chemical. The most measured outcomes were those related to the chemical concentration in corals (bioaccumulation, 25.8%), to coral physiology (16.9%), cover (14%), and mortality (9%). Half of the studies (48.4%) were experimental—the exposure was controlled by the researchers—and were conducted in laboratory conditions (39.4%) and in situ (9%). The most studied taxa, exposure, and outcomes were different between experimental and observational studies. We identified four well-represented subtopics that may be amenable to relevant full syntheses via systematic reviews: (1) evidence on bioaccumulation of chemicals by corals; (2) evidence on the effects of nutrient enrichment on corals; (3) evidence on the effects of human activities on corals; and (4) evidence on the ecotoxicological effects of chemicals on corals (except nutrient enrichment). The systematic map shows that corals in their natural environment can be exposed to many categories of chemicals, and that there is a complete gap in experimental research on the combined effects of more than two categories of chemicals. We therefore encourage research on this topic.

Background Tropical coral reefs cover ca. 0.1% of the Earth's surface but they host an outstanding biodiversity [1] and provide important ecosystem services to millions of people living nearby [2,3]. Despite their biological and economic importance, 75% of the world's tropical coral reefs are currently threatened by both global and local stressors [2,4,5]. While the most prominent global threats are represented by seawater warming and acidification [6], local threats are mainly unsustainable and destructive development of coastal areas, excess sedimentation, overfishing, as well as nutrient and chemical pollution arising from poor land management, agriculture and industry [7,8]. Global and local stressors interact in different ways, but the presence of one stressor often reduces the physiological tolerance of individuals to additional stress. For example, corals are more sensitive to seawater warming if they are already physiologically stressed by poor water quality [9]. While global stressors cannot be mitigated solely by local actions, local stressors can be reduced through ecosystem management, therefore avoiding the exacerbation of climate change effects by the interaction of multiple stressors [10].
The health of reef ecosystems is largely based on the health of their main engineer species, the reef building corals, which are key organisms responsible for reef accretion, but also form the three-dimensional structures serving as habitat, food and nursery for thousands of other reef organisms. The vast majority of such corals (Hermatypic corals, sensu [11]) are colonial scleractinian corals (Cnidaria Hexacorallia) living in association with endosymbiotic dinoflagellate algae belonging to the Symbiodiniaceae family [12]. Symbionts are key to the success of corals in oligotrophic reef waters as they transfer most of the photosynthetically-acquired nutrients to the coral host for its own use [13,14]. This association is however fragile. Many reviews have now made clear that elevation in seawater temperature above a certain threshold is the main factor responsible for the breakdown of the coral-algal symbiosis also called coral bleaching (see for example [15]). As symbionts are the main nutritional source for corals, prolonged bleaching condition may ultimately lead to coral death, and affect the overall functioning of coral reef ecosystems. Coral symbiosis is also largely impacted in coastal reefs by water pollution, which is a major threat per se [16], but also reduces coral resistance to thermal stress and acidification [17,18]. According to the type of pollution, the host, the symbionts, or both partners can be impacted through reduced calcification or photosynthesis, enhanced bleaching or cellular damage, and reduced fecundity among other damages [19][20][21]. The effect of water pollution on corals is a complex subject, due to the vast array of pollutants present in the surrounding environment, and interactions among pollutants, or with other environmental stressors. While several reviews have focused on the subject (e.g. [22,23]), they often addressed only one source of pollution or class of chemicals such as nutrients [24], herbicides [25], oil [26], or sunscreen ingredients [27]. Furthermore, none of these reviews mention the method used to collect the studies, so they are not reproducible and the risk of bias due to the selection of particular studies cannot be assessed.
In this paper, we thus systematically map the evidence related to the impacts of chemicals arising from human activities on tropical reef-building corals. Such knowledge is vital for an effective ecosystem management and coral reef protection.

Topic identification and stakeholder input
In the French Overseas Territories, coral reefs cover 14,280 km 2 corresponding to 5% of the world's total coral reef area [28,29]. France is hence the country with the 4 th largest coral reef area in the world, after Indonesia (18% of world total area), Australia (17%) and the Philippines (9%) [29], and therefore has substantial responsibility towards coral reef protection. The French Ministry of Ecology has launched an assignment for a project aiming experimental-the exposure was controlled by the researchers-and were conducted in laboratory conditions (39.4%) and in situ (9%). The most studied taxa, exposure, and outcomes were different between experimental and observational studies. Conclusions: We identified four well-represented subtopics that may be amenable to relevant full syntheses via systematic reviews: (1) evidence on bioaccumulation of chemicals by corals; (2) evidence on the effects of nutrient enrichment on corals; (3) evidence on the effects of human activities on corals; and (4) evidence on the ecotoxicological effects of chemicals on corals (except nutrient enrichment). The systematic map shows that corals in their natural environment can be exposed to many categories of chemicals, and that there is a complete gap in experimental research on the combined effects of more than two categories of chemicals. We therefore encourage research on this topic.
Keywords: Contamination, Hermatypic, Nutrients, Pollution, Scleractinian to assess the impacts of chemicals and nutrients on coral reefs and to find ways to improve coral reef protection and management at the national scale. The project includes a systematic review in order to gather and analyse the existing knowledge on the impacts of chemicals and nutrients on coral reefs. Because the topic is very broad (all chemicals and all types of coral response should be considered) the first step was to produce a systematic map of evidence, in order to identify relevant knowledge clusters for which evidence can be further analysed in systematic reviews. The review team formulated the primary question of the map and its components, focusing on reef-building corals that are the main engineer species of reef ecosystems, and this was then approved by the French Ministry of Ecology. The French Ministry of Ecology, as well as the French Ministry for Overseas are part of the steering committee of the overall project, and therefore regularly followed the progress of the map.

Primary question
The primary question of this systematic map is: What evidence exists on the impacts of chemicals on tropical reef-building corals?

Components of the primary question
The above primary question has the following key elements: Population: All tropical reef-building coral species (hermatypic scleractinian species, Millepora species, Heliopora species and Tubipora species).
Comparator: Population not exposed to chemicals; Population prior to chemical exposure; Population exposed to a different concentration of chemicals.
Outcome: All outcomes related to tropical reef-building corals, from the molecular (e.g. gene expression, enzyme activities) to the community level (e.g. coral cover, species richness).

Methods
The systematic map followed the Collaboration for Environmental Evidence Guidelines and Standards for Evidence Synthesis in Environmental Management [30] and the protocol has been published in Environmental Evidence [31]. A small deviation to the protocol occurred during the review process: because the searches for dissertations gave relatively few records we extracted all search records instead of the first 100 hits. The systematic map conforms to ROSES reporting standards [32] (see Additional file 1).

Search for articles Search terms and languages
Searches were performed using search terms exclusively in English language. This search however retrieved articles written in languages other than English, and articles written in English and French were included (see section "Eligibility criteria"). The list of search terms is presented in the next section (see section "Search string").

Search string
The best combination of search terms obtained after a scoping exercise (i.e. that gave the highest comprehensiveness and specificity, see Additional file 2 in [31]) was (Web Of Science format): TS = (coral$ AND (contamin* OR pollut* OR toxicant$ OR chemical$ OR "industrial discharge$" OR runoff OR run-off OR sewage OR eutrophication OR effluent$ OR waste$water OR waste-water OR "shipping" OR biocide$ OR "industrial product$" OR "consumer product$" OR "household product$" OR "biocidal product$" OR disinfect* OR nutrient$ OR oil OR metal$ OR pesticide$ OR herbicide$ OR insecticide$ OR fungicide$ OR antifoul* OR anti-foul* OR organochlorine$ OR "flame retardant$" OR detergent$ OR "perfluorinated compound$" OR pharmaceutical$ OR "personal care product$" OR cosmetic$ OR PAH$ OR petroleum OR hydrocarbon$ OR microplastic$ OR nanoparticle$ OR nano-particle$ OR "endocrine disrupt*" OR "organic compound$" OR dispersant$ OR metalloid$ OR solvent$ OR petrochemical$ OR additive$ OR preservative$ OR plasticizer$ OR hormone$ OR "transformation product$" OR "degradation product$" OR byproduct$ OR by-product$ OR sunscreen$ OR "UV filter$" OR "ultraviolet filter$" OR antibiotic$ OR phthalate$ OR PCB$ OR cyanide$ OR chlordecone OR nickel OR copper OR zinc OR cadmium OR mercury OR iron)).

Estimating the comprehensiveness of the search
To assess the comprehensiveness of the search string, we used a test list of 58 articles considered by the review team as relevant to answer our question and spanning a wide range of chemicals (see Additional file 3 in [31]). The search string retrieved 56 of the 57 articles of the test list indexed in the WOS CC and/or Scopus databases. The article of the test list that was not retrieved on either WOS CC or Scopus was a review about the impact of UV filters on aquatic biota that does not contain the term "coral" in its title, abstract, or keywords. This review reviewed only one article about corals that was nevertheless retrieved by the search string (Additional file 2 in [31]).

Bibliographic databases
We performed searches on two online multidisciplinary publication databases: Scopus (Elsevier) and WOS CC (Clarivate Analytics) that we can access through a CNRS (the French National Centre for Scientific Research) subscription. Searches were performed on March 19th 2020. The abovementioned search string was adapted to fit the search facilities of the Scopus database (Additional file 2). All search strings used for the different sources are provided in Additional file 2.
We had access to the following WOS CC Citation Indexes: We had access to all Scopus database (1788-present). No time restriction was applied during searches.
Searches were performed on July 7th 2020 for CORE, and July 8th 2020 for Google Scholar and GreenFile, and the search string developed during the scoping exercise on WOS CC database was adapted to fit the search facilities of these search engines (Additional file 2). In particular, the search string had to be split into six search strings for Google Scholar. Searches were performed on titles, then the results were sorted by relevance and the first 400 hits were extracted. Extraction of results from CORE was done one by one into Zotero using the Zotero connector for web browser. Results from Google Scholar were extracted using the software Publish or Perish (version 7.15.2643.7260, https:// harzi ng. com/ resou rces/ publi sh-or-perish, version accessed 16 March 2020). Results from GreenFile were extracted using the offered export facilities (results can be sent by email in various bibliographic formats e.g. RIS format).
Additionally, we also searched on September 15th 2020 for dissertations in ProQuest Dissertations and Theses (https:// search. proqu est. com/, Publicly Available Content Database), Open Access Theses and Dissertations (https:// oatd. org/) and the French thesis repository (https:// www. theses. fr/). The search string was adapted to fit the specificities of each repository (Additional file 2). Searches were performed on titles and all hits were extracted.

Specialist searches
We searched for links or references to relevant articles on the following eleven specialist websites (English-or French-written websites): Searches were performed between April 21st and May 29th 2020 (Additional file 2).

Call for literature
A call for literature was addressed on July 13th 2020, mainly to the French overseas local authorities (especially the local French Coral Reef Initiative (IFRECOR) committees), for a total of 18 people contacted.

Assembling and managing search results
The results of all searches were collated and duplicates were automatically removed using the package revtools in the R software [33]. Additional removing of duplicates was done manually with the Microsoft Excel software (duplicate conditional formatting and visual identification). The retrieved records from the searches were processed with the R and Microsoft Excel softwares, and reference management softwares (EndNote and Zotero) were specifically used for searching for full-texts.

Article screening and study eligibility criteria Screening process
Articles were screened for eligibility in two successive stages: first on titles and abstracts, and second on fulltexts. Articles with unclear eligibility status during title/ abstract screening were included for full text screening. The list of articles with unclear eligibility status after completion of full-text screening is provided in Additional file 3 with explanation of why they could not be classified. Articles without abstract and retained on the basis of title screening were directly screened on their full-text.
Screening on titles and abstracts was distributed among four reviewers (DYO, LH, RS, YR) and full-text screening among four reviewers as well (DYO, MC, MD, OP). Before the actual screening, all reviewers independently screened a subset of randomly sampled references, and we assessed the consistency between reviewers' decisions by computing the Randolph's Kappa coefficient (Additional file 4). We performed the process in two steps. A first test was done on a small number of references, all disagreements were discussed and the definition of eligibility criteria was further clarified where necessary. Then a second test was performed and all disagreements were again discussed and solved. For titles and abstracts screening a total of 2148/15,177 references (14.2%) were independently screened by all reviewers, and the Randolph's Kappa coefficient was 0.735 for the first test (499 references) and 0.82 for the second test (1649 references). For full-text screening a total of 180/2,700 full-texts (6.7%) were independently screened by all reviewers, and the Randolph's Kappa coefficient was 0.467 for the first test (30 full-texts) and 0.789 for the second test (150 full-texts). During all screening process, we ensured that reviewers never had to screen their own articles.

Eligibility criteria
The eligibility of articles was assessed using the criteria displayed in Table 1. The list of articles rejected at full-text screening is provided with the reasons for their exclusion in Additional file 3. Reviews and meta-analyses were excluded but those eligible according to the Population-Exposure-Outcome criteria are listed in Additional file 5 to make them easily accessible for possible further use.

Study validity assessment
No critical appraisal of study was performed for the systematic map. Studies were however classified according to whether the exposure was controlled by the researchers (experimental studies) or not (observational studies).

Data coding strategy
All articles included in the map were split into studies, i.e. the combination of a taxon, an exposure, and an outcome, and the following information was recorded in Microsoft Excel sheet from full-texts (details are given in Additional file 6): -Bibliographic information (unique identifier assigned by the review team, source, title, authors, journal, year, DOI, language and publication type); -General description of the study (type of study, ISO 3166 country or territory name, latitude and longitude or location); -Description of the population (taxon and taxon level); -Description of the exposure (as described by the authors and as an exposure category defined by the review team); -Description of the type of outcome(s) (as described by the authors and as an outcome category defined by the review team).
The categories of exposure, outcome and type of study defined by the review team are described in Table 2. When corals were simultaneously exposed to several categories of exposure, all of them were indicated (e.g. "Metal | Pesticide").
Data coding was performed by six reviewers (DYO, IDC, KB, MD, MG, RM). Before the actual coding, all reviewers independently coded a random selection of 20/908 articles (2% of all articles, except one of them who coded 10 of the 20 selected articles), all disagreements were discussed, and the coding book (see the first sheet in Additional file 6) was improved where necessary. In case of missing or unclear information, it was coded as such. After completing all the coding, two additional variables were coded from already coded variables. First, the region was determined from the countries following the classification of [34]. Second, the category of chemicals bioaccumulated was determined from the raw exposure variable. Finally, an attempt was made to identify articles linked to the same experiments (linked articles in the database) and species that have been studied under different names (e.g. Montastraea annularis and Orbicella annularis are synonyms).

Data mapping method
We produced a database (Microsoft Excel sheet) of all included studies and their coded data (Additional file 6). We mapped the evidence at two levels. First, we described the source, document type, and chronological and geographical distribution of the articles (article level). Then, we described the distribution of studies in taxa, types of exposure and outcome (study level). The description was made separately for experimental and observational studies because whether the exposure was controlled by the researchers (experimental studies) or not (observational studies) is an important criterion for the subsequent exploitation of the studies. Finally, through the distribution and frequency of studies by types of exposure and/or types of outcomes and/or types of study, we presented four clusters for which a full synthesis of evidence (systematic review) should be possible and relevant for stakeholders, and knowledge gaps.

Review descriptive statistics
Searches returned 11,342 records from Scopus, and 9,472 records from Web of Science Core Collection. The additional searches gave 400 records from CORE, 1,344 from Google Scholar, 172 from GreenFile, 274 from dissertations repositories, 341 from specialist websites, and 58 from the call for literature (Additional file 2). The whole search gave a total of 23,403 records which resulted in 15,177 articles after duplicate removal (Fig. 1). Among Table 1 Eligibility criteria

Include Exclude
Population: -All tropical reef-building coral species (hermatypic scleractinian species, Millepora species, Heliopora species and Tubipora species) living in the shallow and the mesophotic zones -Cold-water or deep-water corals -Ahermatypic corals -Free-living dinoflagellates (not as symbionts in corals) -Studies conducted in coral reefs but not about corals (e.g. about coral reef fishes) Exposure: -All natural (e.g. nitrate), geogenic (e.g. nickel) and synthetic (e.g. diuron) chemicals coming from human activities -Studies assessing the impact of human activities (e.g. river discharge, distance to a dump or to an industrial effluent source, tourism) on corals without reference to a chemical -Studies assessing the impact of chemicals coming from natural sources (e.g. nutrients from guano) -Studies assessing the impact of organic carbon -Studies assessing the impact of sedimentation per se or impact of physical disturbances on coral -Marine debris, macro-plastics Comparator: -Studies comparing population exposed to chemicals and population unexposed to chemicals -Studies comparing population exposed to chemicals and population prior to exposure to chemicals -Studies comparing population exposed to a range of concentrations/ levels of chemicals Outcome: -All outcomes related to tropical reef-building corals, from molecular to community level -Studies reporting evidence of ingestion, concentration or accumulation/ uptake of chemicals in the population studied without reporting health consequences -Studies assessing impacts on coral microbiome/symbionts Language: All articles written in English or French (in case a title or an abstract could not be found in English or French, it was directly screened on full-text) Type of document: Journal article, book chapter, report, conference proceeding, PhD or MSc thesis Presentation, editorial material, letter or news item, conference or meeting abstract, poster Type of content: In-situ or ex-situ studies Reviews and meta-analyses, modelling studies without experimental data them, 2,938 remained after titles and abstracts screening. We could not retrieve 238 full-texts (8%), leaving 2,700 full-text to screen. At full-text screening, articles were mostly excluded because they were reviews/meta-analyses or only synthetizing some findings (29.6%), because of irrelevant exposure (24.3%), irrelevant type of document (15.8%) or irrelevant population (14.7%, Fig. 1). All articles excluded or marked unclear at full-text screening are listed with corresponding reasons in Additional file 3, as well as articles for which we could not retrieve full-text. Among the excluded reviews/meta-analyses, more than a half (285) were however eligible according to the Population, Exposure, and Outcome criteria; to make them easily accessible for possible further use, they are listed in Additional file 5. At the end of the screening process, a total of 908 articles were found to answer the review question and coded as studies; i.e. the combination of a taxon, an exposure, and an outcome, one article often comprising several studies. Then, the systematic map database comprised 7,937 studies.

Source, document type, chronological and geographical distribution of the articles
The 908 articles included in the systematic map were mainly retrieved from publication databases (85.8%) but specialist websites gave a substantial number of additional articles (6.5%, Table 3). Articles were mostly written in English (97.8%), and a large portion were journal articles (87.8%), then conference proceedings (5%), PhD, MSc or BSc thesis (3.6%) and reports (3%). The two oldest articles dated back to 1971, and half (49.6%)  [32] reporting the screening process of the articles of the systematic map of the articles have been produced since 2010 (Fig. 2). The corals studied were mainly from Southeast Asia (22% of the articles), Australia (13.2%) and Middle Eastern Seas (13.2%, Fig. 3a Fig. 3b). Corals from the French Overseas were studied in 4.4% of the articles, mainly from French Polynesia (10 articles), Réunion (10 articles) and New Caledonia (9 articles).

Description of the studies
Among the 7,937 studies included in the systematic map, roughly half (48.4%) were experimental studies (i.e. the exposure was controlled by the researchers), the experiments being conducted most often in laboratory conditions (39.4%) but also in situ (9%). The remaining studies (51.6%) were observational ones.

Taxa studied
The information on the taxa studied was generally available at the species level (76.1%) or at the genus level (12.6%). A total of 317 taxonomic units (+ the group "reef-building corals") were recorded, with the most studied species being Pocillopora damicornis (9.1% of the studies) and Stylophora pistillata (7.6%, Table 4). The taxa studied were different according to the type of study (Table 4). For experimental studies, a total of 148 taxa (+ the group "reef-building corals") were recorded, mostly at species level (94.7% of the studies), the most studied species being Pocillopora damicornis (14.2%), Stylophora pistillata (14%), Acropora muricata (4%), Acropora tenuis (3.8%), Acropora cervicornis (3.8%) and Acropora millepora (3.6%). For observational studies, a total of 277 taxa (+ the group "reef-building corals") were recorded but information was available at the species level for only 58.7% of the studies, the group "reef-building corals" being the most studied (12.8%). The taxa studied were reported in articles spanning almost 50 years (1971-2020). Thus we took into account the latest and on-going revisions of the taxonomy of scleractinians as revealed by molecular phylogeny to update taxon names where needed and possible. For instance, Madracis auretenra the shallow-water species name replaced the one of Madracis mirabilis (a synonym of Madracis myriaster) which develops in deeper water [35]. We made the correction when we could get enough information in the study to distinguish between both names. For the most studied coral Pocillopora damicornis which is now recognized as a species complex split into several species including the resurrected Pocillopora acuta [36], we could not however distinguish within it.

Exposure
A third of the studies included in the map dealt with the impact of nutrient enrichment (28.4% and 2.1% of the studies in the nutrient and eutrophication categories, respectively), and a quarter (25%) concerned the general impact of human activities on corals, without reference to one or several specific chemical compounds ("Undefined pollutants" category, Table 5). Then, the most studied exposure categories were metal (11.3%), hydrocarbon (7.7%), and pesticide (5.2%). Nutrient enrichment was the first and the second most studied exposure for experimental (43%) and observational (18.8%) studies, respectively. Nearly half of the observational studies (47.5%) belonged to the "Undefined pollutants" category. Combinations of three or four categories of exposure could be found in observational studies but never in experimental ones.

Knowledge clusters Evidence on bioaccumulation of chemicals by corals
A first cluster gathering more than a quarter (25.8%) of the studies in the systematic map measured the concentration (or uptake) of chemicals in corals (2,050 studies for bioaccumulation, Table 6). These studies were mostly observational studies (88.6%) and reported bioaccumulation of metals (74%, Fig. 5). They can thus be the focus of a specific systematic review as they are very numerous, and evidence of bioaccumulation of chemicals is important to know as the entry of contaminants into the cells of organisms is the first step for potential subsequent toxic effects.

Evidence on the effects of nutrient enrichment on corals
A second important cluster of studies dealt with coral exposure to nutrient enrichment (2,496 studies about exposure to nutrient or eutrophication). This cluster includes exposure to nutrients in combination with other exposure categories, but excludes studies measuring bioaccumulation. Coral exposure to nutrients was highly studied, through both experiments (64.6% of the studies) and field observations (35.4%, Fig. 6). Coral physiology (31.5% of the studies) and cover (18.3%) were the most studied outcomes, mainly through experiments and observational studies, respectively. The effects of nutrient exposure on corals can be the focus of a large systematic review, but given the high number of studies, focusing on some specific outcomes and/or type of study may be relevant.

Evidence on the effects of human activities on corals without reference to chemicals
A third cluster of studies (1,127 studies) dealt with the impact of human activities on corals without reference to a chemical ("Undefined pollutants" category). The cluster defined here excludes studies measuring bioaccumulation. These studies were mainly observational (96.9%). Among these studies, the outcomes most often studied were coral cover (48.2%), mortality (10.2%), and disease (10.1%, Fig. 7). Because these studies provided no information on the chemicals that may explain the observed effects on corals, they should be synthetized separately, and can therefore be the focus of a systematic review indicating how various human activities (e.g. urbanisation, tourism, agriculture, industries) impact corals. The review could also focus on the impact of human activities on coral cover as this represents a large cluster of 543 studies.

Evidence on the ecotoxicological effects of chemicals on corals
A fourth large cluster of studies (2,007 studies) gathered evidence of experimental studies on the effects of chemicals on corals. The cluster defined here excludes studies measuring bioaccumulation, exposure to unknown chemicals ("Undefined pollutants" category) and exposure to nutrient or eutrophication. Exposure to metals (21%), hydrocarbons (19.8%), and pesticides (19.6%) were the most studied (Fig. 8). Some exposure categories were only recently studied: since 2008 for UV filters, 2014 for nanoparticles and detergents, and 2017 for microplastics. Outcomes related to coral physiology, mortality, microbiome, reproduction, recruitment and growth were studied for nearly all exposure categories (Fig. 8). A systematic review comparing the relative effects of the different exposure categories on these outcomes is therefore possible.

Geographical regions
The central (124 studies) and western Indian Ocean (215), Micronesia (96) and Melanesia (222) where the less studied areas, and most of the studies in these areas were observational (71.8%).

Population
Among the 317 taxonomic units recorded in the map, 147 were represented by less than five studies. At family level, the two least studied families (with less than five studies) were the Oulastreidae and the Rhizangiidae families.

Exposure
Exposure to nanoparticles (15 studies) and detergents (21) were the least studied (Table 5). Some combinations of categories of exposure were also hardly studied (see Table 5) and combinations of three or four categories of exposure were never studied experimentally.

Outcomes
The least studied outcomes were those related to the distribution (occurrence) and genetics of corals (Table 6).
For coral distribution, this information may however be extracted from coral cover, which is the second most studied outcome in observational studies, thus we may consider that this is not entirely a knowledge gap. For

Limitations of the synthesis method
Firstly, we found a high number of syntheses/reviews/ meta-analyses that met our eligibility criteria (285 articles listed in Additional file 5). They may contain a substantial number of references that were not retrieved by our literature searches. For instance, 19% of the articles in the systematic map of Sordello et al. [37] came  Extracting references from these reviews and screening them was not possible here due to time constraints, but this could be done by others in the future. Secondly, the search was conducted in English only, and the evidence was limited to English and French literature (the languages understood by the review team), although it likely exists in other languages [38]. This may have resulted in less geographical coverage of the map and an under-representation of countries where English is not widely spoken or where the scientific literature is also widely published in non-English language (e.g. China, Japan). During the screening process, we excluded 64 articles due to language which are listed in Additional file 3 and could be screened for eligibility and used by others.
Thirdly, some of the outcome categories defined by the review team were very broad and gathered very different outcomes (e.g. the outcome "physiology" may refer to either photosynthetic efficiency, enzyme activity, or mucus production). For these broad categories, defining more specific outcome categories may be necessary before conducting syntheses; this could be done directly from the map database using the detailed raw description of the outcomes.
Finally, considering exposure to nutrients and according to our eligibility criteria we included nutrient enrichment arising from human activities but we excluded natural nutrient enrichment (stemming from guano, or upwelling). Although effects are likely similar, exposure to natural nutrient enrichment is thus not considered in this systematic map.

Limitations of the evidence base
The systematic map revealed some mismatch between the amount of literature available on the impact of chemicals on corals from a country and the coral reef area of that country. For example, corals from the United States of America (mainly Hawaii and the Florida Keys) were the most studied (13.7% of the articles) while the United States are only 16th in the world in terms of reef area (1.3% of the world's total coral reef area, [29]). In contrast, the Philippines is ranked third in the world in terms of reef area (9% of the world's total coral reef area, [29]) Fig. 5 Frequency of studies reporting bioaccumulation of chemicals by chemical category. Whether studies were experimental (dark blue) or observational (grey) is also indicated Fig. 6 Frequency of studies about coral exposure to nutrients or eutrophication by outcome category. Whether studies were experimental (dark blue) or observational (grey) is also indicated but only 2.5% of the articles were about corals from this country. Moreover, several of the top 24 countries with the largest coral reef area are not represented in the systematic map at all (Solomon Islands, Vanuatu, Eritrea, Sudan, [29]). This could be due to a lack of studies for these countries, to studies being available mainly in

Conclusions
This systematic map gathered evidence on the impact of chemicals arising from human activities on tropical reefbuilding corals. The topic is very large as demonstrated by the abundant scientific literature found (908 articles, 7,937 studies). We identified four well-represented subtopics that may be amenable to relevant full syntheses via systematic reviews ( Fig. 9): (1) evidence on bioaccumulation of chemicals by corals; (2) evidence on the effects of nutrient enrichment on corals; (3) evidence on the effects of human activities on corals; and (4) evidence on the ecotoxicological effects of chemicals (except nutrients) on corals.

Implication for policy/management
This structured compilation of all available literature will help guide decision-makers on which clusters to focus on, in a regulatory context, and which priority research topics to support in the future. In addition, one way for decision-makers to take action for coral reef protection from chemicals is to assess risk, which is the result of chemical toxicity and exposure. From the map, it will be possible to extract evidence of toxicity from experimental studies, and exposure data from observational studies. Finally, this systematic map can help local stakeholders identify the body of literature that is relevant to their particular concern. For instance, if local stakeholders are concerned with the impact of nickel mining on corals, they will be able to easily search the map database under the metal exposure category and select studies relevant to their issue.

Implication for research
From the results of our systematic map, we were able to distinguish a group of well-studied pollutants (nutrient, metal, hydrocarbon and pesticide) and another group that has been more recently studied (nanoparticles, detergents, microplastics and UV filters) which thus shows fewer available studies at this time. We can expect that the research effort on these pollutants will continue to increase, and that much more evidence will be available in the coming years. In addition, we identified experimental studies assessing the combined effect of different categories of pollutants (e.g. metal and nutrient) but never more than two categories at a time. We also identified observational studies reporting exposure to three or four different categories of pollutants (e.g. nutrient, pesticide, detergent and hydrocarbon). This confirms that corals in their natural environment are exposed to many different categories of chemicals, possibly interacting with one another (i.e. mixture effects) and that there is a lack of experimental evidence for the combined effects of more than two categories of chemicals. We therefore encourage research on this topic. Fig. 9 Summary of the four well-represented subtopics that may be amenable to relevant full syntheses via systematic reviews (square size is function of the number of studies, "exp" and "obs" stand for experimental and observational studies, respectively). Studies reporting exposure to nutrient in combination with other chemical categories were both counted in clusters 2 and 4