Introduction

Aquatic ecosystems are highly contrasting, from marine to freshwater, from deep water to intertidal and temporary freshwater habitats, from coastal to high altitudes, and from tropical to polar, in all-encompassing distinct environments and communities1. These ecosystems are threatened by increasing levels of disturbance caused by several drivers of change, such as overexploitation, water pollution, habitat loss and degradation, flow modifications and the introduction of invasive species1,2,3. Such threats have resulted in the change and loss of aquatic biodiversity at alarming rates4,5, affecting ecosystem functioning and the ecosystem’s capacity to supply ecosystem services2,6,7. Shedding light on the processes and mechanisms shaping community dynamics and how they respond to disturbances is of great importance since it may facilitate the creation of more effective management and conservation strategies1,8. To this end, trait-based and functional diversity approaches have been increasing in recent decades, as they provide information on the effects of species and communities on ecosystem functions and thus on how systems can function and respond to ecological change9. This can represent answers to pivotal questions in ecology, such as why species coexist, what affects this coexistence, what are the different ecological niches and, thus, their distribution.

Community ecology has traditionally relied on species-based approaches10. However, trait-based approaches have been used for over a century and can be more informative when used to complement more traditional approaches11,12,13 or even as an alternative, for instance, when studying the impact of low-to-moderate intensity drivers in populations and communities9. Examples of the added value of using traits include comparing patterns between geographical areas and/or different biogeographical regions that do not share common species to assess the ecological processes driving community composition. Traits can be defined as any morphological, physiological, or phenological feature measurable at the individual level, while functional traits are any trait which influences organismal performance14,15. Traits can be further divided into effect traits, defined as how any given species affects ecosystem properties, and response traits, defined as how it responds to a disturbance or environmental change16. In addition, traits might be indicators of ecosystem functioning and may facilitate the identification of the mechanisms and drivers generating community patterns and changes8,14,17,18. Furthermore, conclusions may be transferable among ecosystems across biogeographic regions, being independent of species identities, thus, allowing for more generalised knowledge of community patterns and changes14,19,20.

There is a large body of work on species’ traits, with several publications and databases providing trait data compilations (e.g., BIOTIC21; freshwaterecology.info22; FishBase23) that aim to organise the large amount of information produced and published during the last century. The increase in data availability24 and computational tools have promoted the rise of trait-based approaches10, which further evidences the need for collaborative work on trait data compilations for a variety of biological groups25,26. Despite considerable efforts (e.g.24), increased transparency and data access, trait data remain scattered and their nomenclature historically unstandardised11,12,27, which hampers their use to carry out traits’ metanalyses studies, comparative studies or understanding the relevance of such approaches across different studies. Different authors may utilise misleading terms and definitions that may be associated with distinct concepts, a situation further aggravated by publications written in different languages. In this context, having access to additional information, such as both trait and modalities’ definitions, along with a proposed synonym, may be crucial to avoid misunderstandings, especially for researchers who are less familiar with working with traits.

The lack of standardisation, clearly present within trait-based studies, can be overcome by employing different approaches. One can discuss the semantics of what constitutes and defines a trait13,28 and their mathematical use28, and one can organise and propose a standard template for future databases to follow11, or one can tackle the multitude of terms and definitions used to define traits and functional traits by developing thesaurus, glossaries or vocabularies (e.g.21,29,30,31). Notably, 29 provided a baseline for the creation of a standard vocabulary for marine species traits, which is currently implemented in the WoRMS database32 and available under a Wikipedia format33. Additionally, it is an important tool for the scientific community to build on towards attaining a unified terminology and reducing nomenclature misuse. However, restricting this comprehensive vocabulary to “marine species” might reflect that many aquatic community ecology studies address their questions from a single-type ecosystem perspective. We believe the next step towards a unified trait terminology is making it applicable across different ecosystems. For instance, 34 acknowledged the importance of cross-compatibility between ecosystems by establishing a correspondence between their marine deep-sea traits database and other databases, highlighting that apparently similar traits may be associated with different terminologies.

Other obstacles regarding trait-based approaches span from as early as the planning of the research itself9 since selecting and defining traits to include in a community ecology study can also be very subjective. Despite directly impacting the results, trait selection remains unstandardised and, along with how traits are measured, depends completely on the researcher’s approach. Species displaying unclear or variable behaviour represent an equally subjective aspect, where expert judgement is frequently sought after to help categorise traits for poorly described species and may result in different categorisations according to who performed them.

In this context, we build on the previously published trait vocabularies by29, compiling trait databases available from the literature for three taxonomic groups—fish, aquatic invertebrates and zooplankton, and across aquatic ecosystems (i.e., freshwater, estuarine and marine). Identifying the traits currently described in scientific publications and online datasets allowed us to determine the terms and designations used to classify them, and to provide information on the availability of traits information based on the number of classified taxa (important to further select the traits to be used) as well as to detect inconsistencies regarding terminology. These data were then utilised to propose a new holistic trait hierarchy/vocabulary compatible across different aquatic ecosystems, ensuring a clearer and more coherent use for future studies in aquatic functional ecology. Thus, we aim to harmonise the concepts used to describe traits, in general, updating an existing and used nomenclature, and providing guidance on trait selection based on their informativeness versus data availability in the literature. For this, we also included a brief section on trait selection along with a roadmap exemplifying the main steps necessary to go through the abovementioned processes.

Materials and methods

Literature search

A systematic literature search was performed in May 2020 to identify public databases compiling information on traits of fish, aquatic invertebrates and zooplankton. We utilised Web of Science, running the following Boolean combination in the “topic” field tag: (fish* OR macroinvertebrate* OR zooplankton OR fauna* OR marine OR freshwater OR brackish water OR saltwater) AND (ecological OR biological OR functional) AND trait* AND (database* OR dataset*). The search was conducted in all Web of Science databases and for an all-years timespan up to May 2020.

The results were then manually screened for eligibility. Peer-reviewed studies meeting the following criteria were selected for inclusion in the present work: (i) taxa must belong to our target taxonomic groups (fish, aquatic invertebrates, and zooplankton), (ii) trait data must be readily available (data file available via a database dedicated website, in-text table, supplementary material or accessible via R package) and (iii) more than 50 taxa must be covered, to limit the number of entries that might be too context-dependent and avoid data replication. If the study failed one of these criteria, we excluded the study.

Since we aim to report on the traits that are available and easily accessible in the literature, we have deliberately excluded publications when data were only available on request or where the authors did not indicate, for example, the meaning of the abbreviations for taxa or traits. We also excluded studies focused exclusively on paleobiology or toxicology-related traits. Furthermore, in many cases, multiple studies use the same previously published database for their analyses, such as22,23,35, resulting in the pseudoreplication of information. To avoid this, we eliminated these subsets, maintaining only the original complete database. Thereby, we removed studies which made use of (i) an integral copy of other databases, (ii) other databases as the main source, further completing its gaps with information from the literature, or (iii) subsets of other databases. We understand the information in these subsets might appear rearranged, better organised or easier to process, but we removed them as we aim to compile the original sources of traits information (e.g.36,37). Finally, if a given database was published and subsequently updated, we kept only the most recent version.

Trait categorisation, definition and standardisation

Each study's taxa x trait tables were retrieved and all the traits described in each database were characterised by collecting several descriptors, such as name and definition according to the author, geographical region and environment of the study, name in the original dataset file, the number of taxa described, measurement type, trait modalities or measurement unit, data accessibility, among others (complete list and examples in Table 1). Many studies only provide an estimate of the number of taxa for the entire traits dataset, without providing the number of taxa for each trait individually (i.e., there might be a mismatch between trait data completeness per species and the number of taxa described in the complete dataset). Thus, we calculated the number of different taxa with information at any taxonomical level for every trait from the datasets we compiled to allow the comparison of traits’ availability. Most of the entries described were for species-level (more than half), followed by genus in a lower proportion or for specific datasets (e.g.38), and occasional entries for family or order. Our estimates varied a bit when compared with the counts provided by some of the authors, which is most likely due to assessing slightly different dataset versions in case it was updated recently, or by employing different counting criteria regarding the multiple taxonomical levels described in some of the studies. Nevertheless, we do not consider these differences relevant, as our taxa counts are only meant to indicate a broad idea of the distribution and availability of trait information in the literature. The measurement type refers to how the trait is quantified and can take a numerical (for continuous numerical traits), or a categorical value. We also add the possibility of text for when the trait is further discriminated as text (e.g., all food items from a diet). The categorical traits include (1) categorical, for multi-categorical traits (> 2 modalities), ordered (e.g., small, medium, large) or unordered (e.g., detritivore, herbivore, omnivore); (2) binary, when restricted to 2 modalities (e.g., yes/no or true/false traits), and (3) fuzzy coding when species can be assigned to more than one modality with an affinity score.

Table 1 Information collected in each dataset compiled from the systematic review of literature trait data sources, with a description of the data collected, their description and an example of the data collected for each of the variable (see Trait_Sources&Measurements dataset in Supplementary Information).
Full size table

Traits were then sorted along different hierarchical levels from the Marine Species Traits Vocabulary (www.marinespecies.org/traits/wiki33) that follows the nomenclature initially proposed by29. We opted to use this Wiki’s structure as the baseline for constructing our new nomenclature, not only due to its well thought structure but also because it is already a vocabulary familiar within the scientific community, as it is currently implemented within the WoRMS platform32. Additionally, the fact that WoRMS information is continually being refined and updated may also allow for a simpler and faster implementation of new trait nomenclature standardisation proposals published in the future. Adjustments were made to the original Wiki structure available online (1) considering some sections are incomplete (e.g., Physiology which had no information assigned) and (2) to accommodate all newly collected traits. Establishing a correspondence between traits and Wiki implied adding entries to account for all new traits which had not been previously included and removing previously existing traits if there was not any data available for them. Some of the original traits proposed in29 classification may serve as proxies or reflect a combination of the traits that would actually allow species to exist, tolerate or even thrive under certain environmental conditions (e.g. biogeographic distribution, salinity preference, among others). For example, biogeographic distribution reflects the interplay between physiology, movement/dispersal ability and environmental conditions, whereas salinity preference lies between physiology and environmental conditions. We have also reviewed these proxies of traits in this work, as part of the original trait nomenclature29 and because authors still use them as reflecting the species performance in the environment39,40, as their information is easier to collect9. Additionally, our trait datasets focused on fauna, thereby we did not consider any traits specific to flora (plants, seaweeds or microalgae). Nevertheless, there is some work on standardising plant traits and protocols for their measurements15,41,42 and datasets (e.g. TRYdatabase: www.try-db.org), while for seaweeds recent work with datasets (www.algaetraits.org43) has compiled information on their traits and how to measure them (e.g., seaweedtraits.github.io). While some traits are specific to flora (e.g. photosynthesis, light uptake), others are common, such as size, growth rate, respiration and others.

The present work focuses on three trait descriptors proposed by29: “Biological descriptors”, describing a species' body size, its life history characteristics or physiology aspects; “Distribution descriptors”, describing the “environment and habitat in which a species lives, and its spatial distribution by geography, depth, and time”; and “Ecological descriptors”, describing the species habit (including external appearance or form and behaviour) and feeding. In this classification, the “Species Importance to Society” main group was excluded. All traits were sorted along the hierarchical levels of the vocabulary, starting by placing them in one of these three trait descriptors, followed by the Traits classification and Trait standardised name. A key step for sorting all traits accordingly was to identify and group similar traits that are likely to translate the same information but may be referred to with distinct names. Given the overall lack of guidelines for employing cohesive trait terminology, different authors commonly refer to the same trait with several names (e.g. life span, life duration, longevity). Naturally, this also means there will be cases in which authors refer to a trait using a name more typically associated with other trait modalities or trait information, resulting in some trait designations becoming misleading. To overcome this, we first assessed each trait’s definition, measurements/set of modalities and their definitions when available, guaranteeing we understand clearly which information is truly being conveyed. The synonyms (here referring to all different terms that translate the same trait information) were retrieved from the compiled databases and correspond to the trait name used by the author in their taxa x trait tables. Once grouped, synonyms were then associated with one Trait standardised name. We further labelled synonyms as “Misleading” if they were ambiguous (e.g., “feeding ecology” for being too general, “feeding type or habit”, which suggests a feeding method rather than a food type), and as “Specificities” in case they represent a very specific aspect from the corresponding Trait standardised name. The complete list of traits and classifications, their definitions, synonyms, measurement options based on the datasets and literature references for the definitions are available in the Standardized_traits dataset (Supplementary material). This list was based on the information available in the literature, from the selected databases that matched our criteria, all compiled in the Trait_Sources&Measurements dataset (Supplementary Information).

Alluvial diagrams and flowchart

To aid the visualisation of the compiled datasets, we plotted three alluvial diagrams, each corresponding to one of the three trait descriptors: Biological descriptors, Distribution descriptors, or Ecological descriptors. The maximum number of taxa for each standardised name trait was sorted along three levels of information: Descriptors, Traits classification, and Trait standardised name (Table 1). A dendrogram was additionally used to plot our entire dataset, displaying the distribution of the number of traits along the hierarchy and allowing for easier navigation of our hierarchy’s different levels. Finally, we constructed a flowchart describing the main steps involved in the process of choosing which traits should be analysed in each study, simultaneously working as a roadmap for any user interested in using this study’s data compilation (Trait_Sources&Measurements dataset). The alluvial and dendrogram plots were created with the data visualisation tool Raw Graphs (www.rawgraphs.io) and the flowchart with Draw.io (www.drawio.com). Figures were then joined with Adobe Illustrator.

Standardisation of the trait nomenclature

As previously mentioned, despite several attempts to reach a general consensus for traits’ nomenclature (e.g.21,29,33), current traits designations remain full of misunderstandings with an overall lack of organisation27. By reviewing the terms utilised in multiple databases concerning different aquatic ecosystems, we were able to uncover traits with conflicting nomenclature. For instance, a trait commonly poorly defined is the Food type/Diet (e.g., carnivore, herbivore, omnivore), which can be found in the literature referred to as “diet”44, “trophic mode”34, “feeding habit”45 or “feeding diet”22. The lack of consistency across different studies proves itself as a powerful obstacle, making it extremely difficult for anyone to properly search and find traits for their analyses in a time–cost effective manner. In addition, a single nomenclature is the first step towards the standardised use of traits and to the development of a more comprehensive global trait database for aquatic ecosystems, both of which constitute fundamental steps to improve the comparison between different studies, spatial and temporal scales, as well as biological groups.

The present study tackles these issues by using two different approaches. First, we built on the hierarchical classification started by29 by adapting its current Wikipedia version33 to render it suitable not only for marine traits but also for freshwater and transition ecosystems. Secondly, to reduce the multitude of terms in the literature, we organised and grouped similar terms, proposing a single standardised name for each trait, according to its significance and modalities, and stating its different designations found in the literature (i.e., synonyms, as described above). New traits were added when there was no correspondence with the traits already described by33, and definitions were adapted to broaden their meaning. Despite all modifications, we aimed to make the lowest number of possible changes, which we understand will facilitate its implementation in future studies.

Trait proposed nomenclature

The revision and nomenclature adaptation of Biological, Distribution and Ecological descriptors proposed by29 resulted in 18 trait classifications for each, further divided into 59 trait-standardised names. Of the latter, 23 traits were newly added to fill in the information gaps from the previous Marine Species Traits Wiki (Table 2). These new traits are based on our review as they were not discriminated in the original dataset.

Table 2 Description of the different synonyms compiled from the literature, distributed on three hierarchical levels: (1) Descriptor, (2) Traits classification and (3) Trait standardised name.
Full size table

More specifically, the four trait classifications defined by the33 for the Biological descriptors and 11 of the trait standardised names remained unaltered (Table 2). Then, four entries combined two previously distinct traits: (1) “Fecundity” combined “Fecundity and Clutch size”, as the latter also refers to the number of laid eggs (thus fecundity) for organisms that produce batches; (2) “Larval juvenile development” combined “Larval juvenile development”, the existence of “Parental care” and the “Brooding behaviour”, as parental care and brooding refer to particularities of the eggs/larval development; (3) “Larval settlement period and stage duration” merged “Larval settlement period” and “Duration of the larval stage” since both refer to time-based information over which larval settlement occurs; and (4) “Fertilised egg/larval settlement location” merged “Egg, egg mass or clutch location” and “Larval settlement substrate” as these are often mentioned simultaneously. In the Biological descriptors group, 17 new Standardised trait names were added, from which seven new traits were placed in “Life history” and six in “Physiology” (which had no information assigned in33). “Life history—miscellaneous” is one of the new additions and compiles trait information on life history which did not fall into the previously defined standardised names, such as “Spawning aggregation” or “Metabolic type”, respectively (Table 2). “Dormancy/diapause stage” and “Larval or juvenile dispersal potential”, two traits previously assigned as Biological—Life history on33 were moved to the Ecological group since they represent specificities of Ecophysiology and Movement (namely “Resistance form” and “Dispersion”, Table 2). “Gamete type” and “Life cycle” were removed as we did not find any information available on these from the retrieved datasets. All definitions found in the Biological group were either newly added or adapted from33 to better represent the information provided by traits, except “Generation time” and “Life stage” which remained unchanged.

We reduced the Distribution’s trait classifications to six, removing “Habitat preferences” and “Province”, while adding “Geography”. The trait “Province” was then merged with “Vertical Biological zone (or zonation)” as their information is often combined in the literature. “Biogeographical distribution” and “River longitudinal profile” were the only standardised trait names further added to the Distribution group, and eight definitions were either altered or newly defined.

The third and final trait group, the Ecological descriptors, saw the addition of “Body pigmentation” and “Morphometrics” to the six trait classifications previously listed in the33. These two new classifications bared each one a new standardised name, respectively, “Colour and Body parts size”. Regarding the “Body form” classification, we renamed Supporting structures and enclosures to Body structures, allowing for a broader definition, and further included Reproduction-related body form. We further propose a new organisation for the Ecophysiology classification. First, we included here “Environmental parameters”, which were created to combine traits such as “Climate change affected species”, “Metabolic type”, “pH”, “Oxygen and Temperature optimal, preferable, tolerable, or lethal values” (Standardized_traits and Trait_Sources&Measurements datasets). The last two were previously described as “De-oxygenation tolerance” and “Temperature range tolerated”. Secondly, we moved “Growth rate” from Ecophysiology to Biological—Physiology, which was renamed as “Life history rates”, as it reflects an intrinsic physiologic characteristic of the individual, despite the possibility of being affected by the local environment or the conditions to which it was measured (if experimental). Finally, we added the “Resistance form” trait (previously named “Dormancy/diapause stage” and placed it in the Biological group-Life history classification) as it allows a broader definition and inclusion of other resistance traits. “Dependency”, classified under “Mode of life”, was renamed as “Dependency/Association”. Within the Movement classification, the new “Dispersion” trait combines the former Biological trait “Larval or juvenile dispersal potential” and the Ecological trait “Dispersal potential (adult)”, and “Migration (Migratory)” was renamed as “Migration”. Furthermore, three traits were removed, “Growth form (or type)” and “Height (above substratum)”—both previously part of “Body form”—and “Support” from “Mode of life”. Lastly, 13 trait definitions included in this group were either added new or adapted from the original33 definitions.

As previously mentioned, using different terms when referring to a single trait is common. Grouping the synonyms collected during the literature review exposes this practice very clearly and highlights the importance of addressing it with new nomenclature proposals (Table 2). Most trait standardised names are associated with more than one term, it being a regular synonym, a misleading term, or a trait specificity, with some traits having up to 15 different synonyms (e.g., “Reproductive frequency”, “Food type/Diet”). Additionally, we identified 14 different trait standardised names affected by misleading designations (Table 2, Standardized_traits dataset). From these, half were classified under Life history, associated with Reproduction-associated traits, such as “Age at maturity”, “Fertilisation” or “Larval and juvenile development”, the rest being Habitat-related traits, Feeding-related traits, and traits from “Mode of life” and “Movement”.

Availability of trait data for the proposed standardisation

Our systematic literature review yielded 607 results, from which only 37 taxa x trait datasets matched our search criteria (Fig. 1). These can be separated into two groups, one for trait data focused on a single taxonomic group and a second for datasets with two or more taxonomic groups. The former gathered mostly data on fish (11 datasets) and invertebrates (14 datasets), with only four focused on zooplankton (we are here considering22 as two different sets as they have their data separated by taxa groups). The remaining multi-taxa databases divided into “Fish and Invertebrates” (3 datasets) and “Fish, Invertebrates and Zooplankton” (6 datasets).

Figure 1
figure 1

Distribution of the number of traits for aquatic fauna (fish, invertebrates and zooplankton), at the three hierarchical levels: Trait descriptors, Traits classification and Trait standardised name. The size of the circles at each node is proportional to the number of studies for which databases with information on the trait are available.

Full size image

As aforementioned, aquatic ecology tends to separate studies on freshwater and marine ecosystems, a trend which is reflected in the proportion of freshwater, marine and multi-environment databases. Only a third of the results provide information for more than a single environment, this being a combination of freshwater, marine and/or transitional ecosystems (30%). Exclusively freshwater or marine environments follow a similar proportion, accounting for respectively 32% and 38% each.

Due to the considerable variability associated with the region assessed in the databases, both in terms of their geographical location as well as the magnitude of their scale, we opted for a simple way of grouping the different regions. Fourteen databases compile information at a worldwide level (37.8%), twenty-one databases assess smaller regions, here designated as non-worldwide (56.7%), and which may vary from countries such as Canada or Japan, to the Mediterranean Sea or Europe. The two remaining databases do not clearly describe the region from which the traits were compiled.

Regarding the number of taxa described in each database, sixteen databases had between 50 and 250 taxa and only six databases described 251–500 taxa, which together accounted for 59.4% of all studies. Five databases had between 501 and 1000 taxa, four between 1001 and 2500 and seven with more than 2500. We divided the Freshwater Ecology traits database22 in two subgroups to calculate these taxa number intervals because their data for fish and invertebrates is isolated from each other, allowing for a clearer insight on trait data availability regarding the taxa groups.

A total of 1127 traits were collected and assigned to three main groups, 475 Biological (42.2%), 162 Distribution (14.4%) and 490 Ecological (43.4%) (Fig. 1). The three largest trait classifications were Life history (n = 253, 22.45%), Body form (n = 115, 10.20%) and Physiology (n = 109, 9.67%). Overall, the three most common traits are Body parts size (n = 99, 8.78%), Body length (n = 82, 7.28%) and Body structures (n = 80, 7.10%). Within the Biological descriptors, Body length was the most common trait (n = 82, 7.28%), followed by Fecundity (n = 36, 3.19%) and Elemental composition (n = 32, 2.84%). The most abundant Distribution traits were Substratum (n = 22, 1.95%), Depth range (n = 21, 1.86%) and Biogeographical distribution (n = 20, 1.77%). Finally, Body parts size (n = 99, 8.78%) was the largest classification under Ecological traits, followed by Body structures (n = 80, 7.10%) and Environmental parameters (n = 68, 6.03%) (Fig. 2).

Figure 2
figure 2

Distribution of the maximum number of taxa registered and described for each Trait standardised name over the three hierarchical levels from our unified nomenclature proposal: Trait descriptors, Traits classification and Trait standardised name. For clarity, data were divided into three alluvial diagrams corresponding to the three Trait descriptors: Biological (a), Distribution (b) and Ecological (c). Colours are the same as in Fig. 1 and correspond to the different Traits classifications. The size of the bars is proportional to the number of studies for which databases containing information on each hierarchical level of traits are available.

Full size image

Specifically for the fish group, 507 fish traits were gathered from datasets exclusively focused on fish taxa, from which roughly half were categorised as Ecological (n = 281, 55.4%), followed by Biological (n = 171, 33.73%) and Distribution (n = 56, 11.05%). The largest trait classification belonged to the Biological group, where 107 traits were classified as Life history (21.10%), followed by 95 Morphometrics traits (18.74%) and 78 Body form (15.38%). Body parts size (n = 95), Body structures (n = 65) and Body length (n = 45) were the most common traits, and Physiography (n taxa = 31,369), Body shape (n taxa = 17,556), and Body structures (n taxa = 14,173) were the traits with the highest number of taxa.

We found 390 traits accessible in invertebrates’ datasets, the majority of which equally distributed between Ecological (n = 157, 40.26%) and Biological (n = 154, 39.49%), with the remaining traits falling within the Distribution group. Similarly, to the fish traits described above, Life history was the classification with the largest number of traits (n = 98, 25.13%), followed by Ecophysiology (n = 46, 11.79%) and Habitat (n = 45, 11.54%). Despite being the main group with the lowest trait number, Distribution presented the highest number of taxa described for invertebrates (Habitat—3606; Geography—3339). The third and fourth traits with the highest number of taxa were Feeding (2891) and Body form (2473), both belonging to the Ecological descriptors.

Zooplankton datasets described 116 traits which present a different distribution when compared with the other taxa groups. The vast majority was grouped as Biological (n = 103, 88.79%), where Physiology (n = 61, 52.59%), Body size (n = 23, 19.83%) and Life history (n = 19, 16.38%) were the largest trait classifications. The rest of the traits were distributed along the Ecological (n = 8, 6.9%) and Distribution descriptors (n = 5, 4.31%). Body size, more specifically Body length (1850), had the highest number of taxa described, followed by two Ecological traits, Environmental position and Biogeographical distribution, both described for 701 different taxa.

Lastly, 114 traits were described in datasets which combined more than one taxonomical group (more specifically, datasets of Fish, Invertebrates and Fish, Invertebrates, Zooplankton). With a trait distribution along the three groups similar to the one seen in invertebrates, where the majority of traits were grouped under Biological (n = 47, 41.23%) and Ecological (n = 44, 38.60%). The three largest trait classifications were Life history (n = 29, 25.44%), Habitat (n = 14, 12.28%) and Ecophysiology (n = 12, 10.53%), with Environmental parameters (n = 12, 10.53%) being the most common trait, followed by Food Type/Diet (n = 8, 7.02%) and Elemental composition (n = 6, 5.26%). Overall, there were no discrepancies in standardised trait numbers, as they displayed similar frequencies within each trait classification. However, the traits with data for the largest taxa number were Environmental position (28741), Body structures (20518) and Food Type/Diet (17743).

Ecological relevance of the traits

In addition to the main set of traits available based on their abundance and number of taxa described, we briefly assessed their availability according to their ecological meaning. We adapted13 unified typology of aquatic functional traits, in which they established a link between trait type (Morphological, Life history, Physiological and Behavioral) and an ecological function (Resource acquisition, Growth, Reproduction and Survival). We can also analyse it considering the trait reflects the fitness of the species under given environmental conditions (i.e., response trait) or influences an ecological process (i.e., effect trait)9. In order to transpose trait data availability to a potential ecological meaning, we determined the traits with the most taxa descriptions for each group and established a correspondence adapting13 diagram. Nevertheless, it is worth noting that these traits do not necessarily reflect an ecosystem function when analysed alone9. Frequently, a combination of different traits or even their diversity is key for assessing an ecosystem function, which in turn is highly context dependent9,20,46.

First, the most abundant Biological traits were Body length, Fertilisation, Reproductive frequency and type, larval and juvenile development and life history rates (Fig. 2a), which successfully correspond to the four ecological functions mentioned above (Fig. 3).

Figure 3
figure 3

Overview of available traits to measure ecological functions by adapting the unified typology on aquatic functional traits proposed by13 to the identified standardised trait names with most data described available according to this study.

Full size image

Body size, measured as body length, is one of the most common and measurable traits across biological groups, often considered a key trait influencing the dynamics and structure of aquatic communities. It influences energy requirements, food-web and biotic interactions, metabolic rates and dispersal29,40, thus relating to all functions (Fig 3). The life-history rates from the physiology group, collectively combining rates from various processes/functions, including growth itself, also relate to all functions (Fig. 3), by influencing species' performance or fitness against an environmental background. Fertilisation and reproductive frequency relate to reproduction but may also influence dispersal abilities, whereas larval and juvenile development provides information on both growth and survival.

Secondly, the most common Distribution traits were Environmental position, Physiography, Substratum and Vertical Biological Zone (Fig. 2b), all matching ecological functions (Fig. 3). These traits are usually considered as proxies of the actual trait that allows a species to distribute in an environment or habitat, as mentioned above. Yet, those proxies are often used since it is easier to get information. Physiography and the Substrate relation may be used to infer conclusions for all four ecological functions as broadly reflecting the environment where the species occurs and develop, including habitat preferences (Table 2). Information on the Environmental position and Vertical Biological Zone is often mixed in literature, as they reflect the species position either relative to the water column or substratum (i.e., environmental position) or wider zonation patterns (details in Standardized_traits dataset), which are potentially related to resource acquisition or survival (Fig. 3).

Finally, the five most abundant Ecological traits are as follows: Body parts size, Body structures, Environmental parameters, Food Type/Diet and Mobility (Fig. 2c). These correspond with multiple sections of the diagram, exhibiting potential links to all four ecological functions, with a higher number of overlaps for resource acquisition and lower for reproduction (Fig. 3). There can be also different interactions between these traits. For instance, mobility can be expressed through morphometric measurements or body shapes, while mobility itself is related to all functions and also influences dispersal abilities.

How to select traits for a given study & good practices

The trait selection process necessary to navigate when starting a study is equally complex as it analyses25 and of utmost importance, as it will directly influence the study’s results. Here, we provide a roadmap (Fig. 4) that aims to facilitate this process by guiding the user through a series of steps and answering a few crucial questions. The roadmap selection was built considering the data availability under this review, where available information can be accessed using the Trait_Sources&Measurements dataset filter selection.

Figure 4
figure 4

Roadmap describing the main steps and questions when choosing a set of traits for a functional diversity study. At least four main points should be addressed: deciding the target ecosystem, target taxonomic groups, study’s scale and testing for trait redundancy.

Full size image

Briefly, the first step should be defining the research questions, objectives, and testing hypotheses. An initial framework for the steps involved in setting the hypotheses based on effect or response traits is available in9,47. Then, one should consider whether it is a large- or local-scale study by selecting the ecosystem type and whether the focus is on a particular taxa group of interest, as both strongly influence the data availability. For high diversity studies (e.g., large scale, diversity hotspots), comprising multi-trophic taxa groups, the traits selection should focus on those with the most information available (Fig. 2) that can still inform the researcher's questions (Fig. 3). Still, the trait range values may change for each of these trophic groups or even for different taxa within each trophic group. Specifically, multi-trophic studies may consider these groups as compartments to calculate their functional properties, or to consider a multitrophic response-effect framework to include their interactions (more details in9).

For other studies, researchers can select other particular traits fitter for specific questions or even consider measuring them directly (realised traits13). Once traits suitable to answer the research questions are chosen, one must check for trait redundancy, as different traits serve as a proxy for the same functions and analyses will benefit from a lower number of traits with low levels of redundancy (e.g., increasing the number of traits in the analysis can lead to an artificial increase of differences in functional space or mask functional trait patterns and response to changes). If present, redundant traits should be removed, simplifying the set of chosen traits. The one to choose should rely on traits with precise and interpretable functional relationships for the research question and data availability for the species' dataset and region, with priority given to standardised measurements and to the easiest to interpret. Selection could also rely on adding traits only when they provide new information or applying model selection approaches or dimensionality reduction techniques. However, if relevant to the aim of the study, redundancy relationships can be taken into account when interpreting the results, but should be avoided to run the analyses (e.g. estimation of functional diversity indices, to modelling trait response).

In case the scale of the study/target region is relevant to the research questions, one should filter for the region when checking for data availability of the already selected traits. The final step of this process would be downloading the available trait data. However, these questions and line of flow should be repeated if data for selected traits are unfortunately unavailable in order to find a new suitable trait. It is also important to note that most of the information available in the literature reflects an average value for the species in its environment unless it is clearly assumed that the trait measurements also include intraspecific variability. Indeed, most studies of functional diversity assume that interspecific variability is greater than intraspecific variability. Nowadays, however, the importance of this intraspecific or within-species variability is increasingly recognised, including methods to quantify it9,48. If one is interested in quantifying this aspect as well, we primarily advise to either specifically look for all information available for the species (e.g. phenotypic plasticity or trophic plasticity) or actually measure a number of individuals for each species (e.g. size), which in some cases may be the only viable option.

Finally, it is difficult to compare functional diversity (whether indices or trait identity responses) between studies when different traits and different number of traits are selected, not to mention monitoring schemes. This is exacerbated when the same traits are labelled differently. Standardising the nomenclature, as proposed here, and this roadmap for selecting non-redundant traits would thus be a first step towards a common framework to guide the selection of the most relevant traits for the specific aim of the study. This is particularly relevant for meta-analyses, as having the synonyms and definitions would increase the number of potential studies that could be included in meta-analyses. To support this functional diversity studies comparability, we also advise new trait studies to always include the data source used to classify each trait (rather than generic information) and the traits data as an appendix (for aggregated traits, the information of each individual trait should be presented) so that other authors can compare the trait response.

Future directions

Trait-based studies are rising in popularity, thanks to the overall perception of their capability to inform on the relationship between individuals and the environment. Advances in computational techniques and better knowledge of species ecology9 have led to a large number of researchers currently utilising traits in their studies. The number of published trait data is increasing and new databases are frequently released, some of which were not available at the beginning of this work and, thus, have been not included (e.g.,49,50). We emphasize the need to build on this work, which clarifies the traits' possible measurements, definitions and ecological relevance, for comparable trait studies. However, data collection will certainly benefit from being revisited in a few years. Since 2020, the databases available may have already been updated, resulting in changes in the array of traits accessible and on taxa number counts (e.g.23,33). Although our proposed nomenclature considers the current Wiki version of the33, their website now includes a list of links with existing initiatives (e.g.21,23,51) whose information is now also linked under their dataset to avoid online data duplication. We maintain these databases in our data compilation as they provide a valuable and significant source of information that was still being implemented within the33 when this work started and because they will be useful for past published studies using these datasets. Also, other authors may still be using them in the future, as they are widely known in the scientific community for specific biological targets (e.g., fish23). Nevertheless, we acknowledge how essential the efforts of these data compilations are and must emphasise that retrieving data from singular datasets may soon become obsolete.

Another important aspect is the availability of multiple versions of the same dataset and how they tend to be utilised by the community. To avoid the pseudoreplication of data, we only included complete datasets in their most up-to-date version, simultaneously acknowledging this would result in failing to include a small amount of published data. However, not every study will prefer to use the original/complete dataset. Researchers will often analyse subsets of larger databases, sometimes manually adding trait data from the literature to fill in information gaps on their target taxa, thus making available smaller subsets that include additional data originally unavailable. 52 are one of these cases, where the FishBase subset they published is often preferred and analysed instead of the original data source because the data is now reorganised, potentially more complete and easier to access/use. The effort necessary to compare and isolate the differences between an original dataset and its subsets is extremely elevated, and projects seldom can do this type of maintenance and update regularly. This results in increased inaccessibility of small datasets freshly compiled for small-scaled compilations such as this work, and further evidences the urgent need for creating collaborative efforts in the compilation and processing of data.

Finally, this study has clearly expressed the need for unifying traits’ names and definitions, as many terms retrieved directly from the literature were ambiguous in their meaning and use. Being able to search, find and correctly employ traits thus relies on having a scientific community in agreement regarding trait nomenclature. Moreover, in a field of study where the array of traits assessed directly affects the outcome of our research, a clear understanding of the information traits carry is essential to obtain effective answers for our problems and hypotheses.