Background & Summary

Ecology is facing a ‘functional revolution’. Researchers are collecting species traits (glossary in Table 1) at unprecedented rates and depositing them in centralized databases focused on all kinds of organism including fungi1,2, plants3,4, and animals5,6,7,8,9,10,11. Traits are becoming an essential currency in ecology, especially to quantify functional diversity12,13. For example, traits allow us to bridge some of the conceptual gaps that exist between ecology and evolution14, to mechanistically explore ecological15 and biogeographical16 processes, as well as to prioritize species, habitats, and ecosystems for conservation17,18,19.

Table 1 Glossary of specialized term used.

Despite all these promising developments, an impediment we are facing in advancing functional ecology is that there is still limited availability of trait databases for most organisms, especially those living in difficult-to-access, and hence generally less studied, ecosystems. Due to their inherent inaccessibility20 and several impediments to research21, caves and other subterranean habitats exemplify well this point. Whereas the community of subterranean biologists largely agrees that a functional perspective is key to better understand the ecology of caves22, trait-based subterranean studies remain unicorns23,24,25. The scarcity of centralized trait database for virtually any subterranean animal group means that we are still far from seeing the ‘dark side’ of the functional revolution as described above.

Spiders (Arachnida: Araneae) are a dominant component of subterranean biological communities, playing an important role as predators and showing various functional adaptations26. Despite their important role in subterranean food webs27, cave-dwelling spiders remain poorly known, especially in the tropics28. For most families of subterranean spiders, there is a general lack of distribution data, phylogenies, information of conservation status and, above all, traits. Europe is probably an exception to this general trend, given that there is a preliminary check list of subterranean spiders29, distributional and community composition data for selected caves and species scattered across the continent30, and growing information on their conservation status31.

Responding to the recent effort of the international arachnological community for compiling spider traits in an open centralized repository10, and building upon the launch of the first online version of the World Spider Trait database (WST)32, we present a functional trait database for all known species of European subterranean spiders. Integrating literature data and direct measures on specimens, we collected 64 morphological and ecological traits for all known subterranean spiders in Europe (Fig. 1). Here, we describe this trait database and make it freely available online. Finally, we discuss examples of trait-based research questions in subterranean biology hoping to promote the maximum use of this database.

Fig. 1: Infographic summarizing the study and the collected traits.
figure 1

See Table 2 for details about specific traits. All spider silhouettes are original drawings by Irene Frigo, except for the Pholcidae silhouette and the silhouettes of insects taken from PhyloPics (all with open licence). Original drawing of caves by Stefano Mammola.

Methods

Acronyms and jargon

A glossary of specialised terms used in this study is given in Table 1. The following acronyms are used: AME, Anterior Median Eyes; ALE, Anterior Lateral Eyes; PLE, Posterior Lateral Eyes; PME, Posterior Median Eyes; AME–ALE and PLE–PME, eyes distances; WST, World Spider Trait database. The notation NA (not available) is used for missing data in the database.

Taxonomic and geographical coverage

We initially updated the checklist of European subterranean spiders provided in ref. 29 to obtain a list of species for which to collect relevant traits. Specifically, we:

  1. I).

    included new species described after 2017;

  2. II).

    included species that we had overlooked in the previous version of the checklist—see, e.g., missing species pointed out in ref. 33; and

  3. III).

    updated taxonomy following recent nomenclatural changes34 and corrected a few mistakes we detected.

Concerning the geographical coverage of the checklist, we considered all European countries as defined in the Spiders of Europe35 database. However, we excluded North Africa, because it was only recently included in the Spiders of Europe, and oceanic islands (Azores, Madeira, and Canary Islands), because their insularity may lead to different processes shaping regional diversity36.

Trait collection

For each species, we collected traits using literature data (mainly taxonomic descriptions) and, whenever specimens were available to us, direct measurements (27%; n = 137 species) (Fig. 1). We retrieved literature primarily from the World Spider Catalog34—i.e., the main repository of bibliography on the taxonomy of spiders37—and secondarily from the Spiders of Europe repository and Google Scholar. We used the latter two sources to retrieve most literature on ecological traits.

Morphological measures

We collected several traits that different authors have considered proxies for body size, food specialization, and subterranean adaptation38,39,40,41 (Table 2). All length measures are given in millimetres.

Table 2 List of traits collected in this study. See section “Acronym and jargon” for a definition of all acronyms used.

For measured species, we used averaged values whenever multiple specimens were available to us. If total body length was not reported in the original description, we approximated it as the sum of prosoma + opisthosoma. Also, in a few descriptions, authors reported tibia length as the sum tibia + patella. In such cases, we approximated tibia length as a fraction of the value based on the ratio tibia/patella in congeneric species.

In six-eyed spider families (Dysderidae, Leptonetidae, Sicariidae, Symphytognathidae, and Telemidae), for which a pair of eyes is missing due to ontological reasons unrelated to subterranean adaptation, we assumed the missing pair of eyes to be AME42. If a paper reported the missing pair of eyes to be ALE, we ignored the information and assumed AME to be the missing one. We assigned a value of 0 to the missing pair of AME in six-eyed families. However, to distinguish these from missing AME due to cave adaptation in eight-eyed spider families, which were also scored as zeros, we included the ontology of eyes as a variable in the dataset (categorical variable ‘AME_type’ made up of three levels: “Present”, “Absent_Ontology”, and “Absent_Adaptation”). This variable can be included in some analyses to ensure that, in functional diversity calculations, the ontologically missing AME do not mix with missing AME due to subterranean adaptation.

Habitat and ecological traits

We classified ecological traits (Table 2) based on ref. 43. We included functional guild, foraging strategy (type of web and method of active hunting), and prey range (specialist or generalist). Conversely, we excluded vertical stratification (ground or vegetation) and circadian activity (diurnal or nocturnal), as these are not relevant for subterranean ecosystems. Instead, we classified vertical stratification in a cave (ground, wall, or both) and potential for long-range dispersal outside subterranean habitats (e.g., ballooning in Meta spiders44 or active dispersal on the ground in Pimoa45).

Ecological classification

In the previous version of the checklist of European subterranean spiders, we also reported an indication of the level of affinity of each species to the subterranean medium46,47. This was assessed by the group expert involved in this paper or taken from literature (whenever this information was mentioned in the original description), and included two categories: i) Troglophile, for species able to maintain stable subterranean populations or inclined to inhabit subterranean habitats, being, however, associated with surface habitats for some biological functions or able to maintain surface populations too; and ii) Troglobiont, for species strictly bound to subterranean habitats. For consistency, we included this ecological classification in this update of the checklist and in the trait database. However, since definition of ecological categories is traditionally a stumbling stone of biospeleology48,49, and sparkled some debate in the form of personal communications, we would like to clarify its real meaning. The attribution of some species to one category or another may be problematic as this is not a strictly categorical trait but often can be seen as a continuum from troglobionts to surface dwellers—including the intermediate troglophiles. As we see it, this is just a practical tool that allows one to roughly subdivide groups of species in broad macro-categories. The proper way for assessing the species affinity for subterranean or surface habitats would be a systemic survey including extensive sampling primarily in the surface habitats, population studies, and a robust phylogenetic framework47, all of which are practically non-existent for most subterranean spiders. There are, however, alternative ways to do so depending on the research questions of interest50. For example, one can by-pass this classification and simply use morphological traits such as eye regression, leg elongation, and pigmentation as a proxy for the subterranean specialization of each species.

Data visualisation

To illustrate the usage of the dataset, we plotted the distribution of key traits as density plot with the R library ‘ggplot2’51. We also generated a representation of the trait space for European subterranean spiders showing its general organisation and the position of each spider family within it. To this end, we selected a subset of traits from the whole trait matrix, representing:

  1. I).

    General morphology of species (Average body size, Sexual size dimorphism, and Prosoma shape);

  2. II).

    Morphological adaptation to subterranean condition, including both categorical (Pigmentation, Presence/absence of Eyes, Eye regression, Leg elongation, AME, ALE, PME, and PLE, and AME type) and continuous (Femur elongation and Profile reduction) traits;

  3. III).

    Hunting strategy (all binary variables referring to hunting strategy and diet, as well as the continuous variable Fang length);

  4. IV).

    Dispersal behaviour (Dispersal); and

  5. V).

    Microhabitat occupation (Verticality).

We performed data exploration following recommendations in ref. 52, checking variable distribution, multicollinearity among continuous traits via Pearson’s r correlations, and presence of missing data. As a result of data exploration, we excluded Fang length and Profile reduction because they contained more than 80% of missing values. To homogenize variable distribution, we log-transformed all continuous variables that do not assume negative values. We also standardized all continuous traits to mean = 0 standard deviation = 1 to ensure comparable ranges among traits.

Since the trait matrix contains both continuous, binary, and categorical variables, we used a Gower distance to estimate trait dissimilarity among species53. Because different traits span different functional roles, we used an optimisation method to attribute weight to traits within groups54. To this end, we assigned traits to the five groups of variables as defined above.

We visualized the trait space as the first two axes of a principal coordinate analysis using the trait dissimilarity matrix as input data, using the R package ‘ape’ version 5.5.055. For graphical visualisation, we estimated density of species onto the ordination diagram using a kernel density. Furthermore, we visualized the centroid of each family to get an overview of the spatial relationships among families within the trait space. To relate traits to ordination axes, we used the function envfit from the R package ‘vegan’ version 2.5.756. This function calculates a multiple linear regression of the traits (dependent variable) and species scores on ordination axes (independent variables). The normalized regression coefficients multiplied by the square root of the coefficient of determination are used to position the trait onto the ordination diagram. Note that this analysis was only possible for complete cases—i.e., species without missing traits (N = 154). We performed all analyses in R version 4.1.057.

Data Records

The trait database is available in Figshare58 as a tab-delimited file (.csv) and in Excel (.xlsx) format. Traits missing from the World Spider Trait database were also deposited therein—accessible directly in R environment using the function traits in the R package ‘arakno’59.

Detailed explanation of traits, including their hypothesized functional meaning, is given in Table 2. The dataset consists of 64 traits (some examples of trait distributions are given in Fig. 2) for 520 species belonging to 20 families associated with caves (Table 3)—34 species more than in the previous checklist29. The family comprising most species is Linyphiidae (224 species, almost half of them belonging to a single genus Troglohyphantes), followed by Dysderidae (62 species), Leptonetidae (60), Nesticidae (56), and Agelenidae (43). All these families comprise several specialized species only found in subterranean habitats and showing traits such as full eye regression and complete depigmentation, but also generalist species exhibiting a low degree of morphological specialisation to subterranean life (Fig. 3). The remaining families are all represented by up to 30 species and encompass spiders with diverse levels of subterranean specialisation. We refer the reader to ref. 29. for an in-depth taxonomic and biogeographical account.

Fig. 2: Variation in selected traits values of subterranean spiders split by whether a species occurs in deep caves (obscure zone) or not.
figure 2

Jittered points are the actual values, boxplots summarize median and quantiles, and density plots summarize data distribution. All length measures are in millimetres. (A) Average Body length (male and female are averaged). (B) Average size of the Anterior Median Eyes. (C) Average size of the Posterior Median Eyes. (D) Femur elongation, calculated as the ratio between the Femur I length and the average Body length. (E) Average size of the Anterior Later Eyes. (F) Average size of the Anterior Lateral Eyes.

Table 3 Spider families with number of genera and species occurring in subterranean habitats across Europe—updated from ref. 29. Total number of genera for each family are derived from Spiders of Europe35.
Fig. 3: Variation in selected traits values of dominant families of spiders in European caves split by Ecological classification.
figure 3

Density plots summarize data distribution, dots are median values, and bars summarize quantiles. All length measures are in millimetres. (A) Average Body length. (B) Average size of the Anterior Median Eyes; no data are shown for Leptonetidae and Dysderidae, both of which are six-eyed spider families. (C) Average size of the Posterior Median Eyes. (D) Femur elongation, calculated as the ratio between the Femur I length and the average Body length. (E) Average size of the Anterior Later Eyes. (F) Average size of the Anterior Lateral Eyes.

Technical Validation

There are some limitations that one must be aware of when using the dataset:

  1. I).

    Given the low availability of specimens—many of these species have been collected at the time of their description and never recorded thereafter—the dataset is a mixture of literature data and direct measures. It includes families for which we have been able to measure all species (Pholcidae) and others for which over half of species traits are derived from original description and other literature sources (e.g., Linyphiidae and Leptonetidae). Unfortunately, many original descriptions, both recent and old, contain poor information, limiting the possibility to extract traits.

  2. II).

    For the same reason, there is a high frequency of missing data for some traits and species. This means that one may want to focus on traits that are well sampled and use a reduced matrix of only well-sampled traits in community-level analyses. There are statistical ways to partly remedy these problems. Different imputation methods can be used to infer missing trait values60,61. Most of these imputation tools are implemented in the function fill in the R package ‘BAT’62,63. Also, in community-level analyses, one can use functional distance measures able to accommodate missing data, especially Gower distances54. The latter method is the one we used to generate the trait space in Fig. 4.

    Fig. 4: Example of a trait space representation for European subterranean spiders.
    figure 4

    (A) Distribution of European subterranean spiders along the first two axes of a principal coordinate analysis describing the trait similarity among species. Gradient of colour denote density of species—higher density in darker areas. (B) Same as the representation of the trait space in A, but mapping the centroid of each family onto the ordination diagram. (C) Same as the representation of the trait space in A, but mapping the position the trait onto the ordination diagram. (D) Distribution of values of the four traits that contribute the most to determining the distribution of species in the trait space (untransformed values are shown). Original spider silhouettes by Irene Frigo, except for the Pholcidae silhouette by Sean McCann via PhyloPics (Public Domain Dedication 1.0 license) and Symphytognathidae by CG-R.

  3. III).

    Given the scale of the dataset and the lack of multiple specimens for most species considered, this dataset do not contain information on intraspecific variability—one exception being the minimum-maximum range for body size. It is well known that intraspecific trait variability is an important aspect of community ecology64,65, which can be pronounced in many taxa66,67. This is seemingly true also for subterranean spiders. For example, in the well-studied case of Western Alpine Troglohyphantes (Linyphiidae), intraspecific variability has been reported for morphological traits relating to subterranean adaptation38,39,68, but also in individual thermal tolerance68. Likewise, individuals of Kryptonesticus eremita (Simon, 1880) (Nesticidae) may show different levels of pigmentation depending on how far from the cave entrance they are collected69. Accordingly, any analysis based on this database must be taken as an average representation of the process under study, and the information relativized accordingly70.

Usage Notes

Complementarity with other databases

The database is fully compatible with community composition data for European cave spiders available in Figshare71 and cave spider occurrence records deposited in the Global Biodiversity Information Facility (GBIF)30. Traits can be univocally linked to each species included in the latter datasets (or others) using the species Latin name. However, one may need to check and update the taxonomy of some species depending on the latest nomenclature changes. This can be done automatically using the function checknames in the R package ‘arakno’59, which checks for nomenclature changes, synonyms, and spelling errors dialoguing with the most up-to-date version of the World Spider Catalog34.

Example of trait-based research questions in subterranean biology

As old as the recognition of the bizarre morphology of cave species is the search for ecological and evolutionary explanations for these unique adaptations72. Trait-based ecology is a critical framework to this end73. By focusing on how traits interact mechanistically with environments across spatial scales and levels of organization, we can use geographically ubiquitous and ecologically diverse cave spiders to test hypotheses in subterranean biology and beyond22. Here, we provide some examples of avenues of research, hoping to both stimulate re-use of the dataset and the quest for developing similar databases for both spiders outside Europe and for other subterranean taxa.

Quantifying functional redundancy and subterranean specialization

Different spider species and families occupy distinct regions of the trait space (Fig. 4). The position of the species in the trait space can be mapped to obtain a quantification of their functional redundancy (e.g., if multiple species fall within highly sampled areas of the trait space). Within a given group (e.g., family or genus), one could also rank species according to their degree of adaptation by using traits relating to subterranean adaptation and calculating the functional distance of each troglobiont species from the average troglophile species or the closest surface species50—following the saying “nothing [makes] sense in speleobiology without a comparison of cave animals with the ‘normal’ epigean ones74. Ultimately, the quantification of the level of subterranean specialization of species on a continuous scale allows us to explore the degree to which the specialization of a given community relates to local environmental conditions, interspecific interactions, and more.

Trait-based (macro)ecology of subterranean spiders

Traits are a useful aid for answering a range of questions in community ecology and macroecology12,15,16,75. To what extent is there functional convergence in the functional space of subterranean spider communities in a given region? Do different microhabitats within a cave select for functionally unique spiders? What is the maximum degree of functional similarity that allows two or more species to occupy the same environment? How does the functional space of a given subterranean community change after a perturbation event (e.g., the extinction of some species, the invasion by a non-native species)?

Similar questions can be answered using metrics such as community weighted trait means76 or more advanced ways to calculate the functional richness, dispersion, and regularity of the trait space occupied by a given community (e.g., functional dendrograms77 or probabilistic hypervolumes78). We refer the reader to recent accounts on functional diversity analyses for operation details about similar analyses12,13. Also, all these questions can be explored at different scales, from local communities inhabiting a single cave or cave system up to entire karst areas and even continents. This latter possibility is enhanced by the availability of broad-scale distribution and community composition data for European cave spiders30. For example, a recent study demonstrated that there is a quick turnover in the taxonomic diversity of subterranean spiders across Europe, mediated primarily by the geographic distance among caves and secondarily by the climatic conditions and availability of karst habitat, ensuring cave connectivity79. The usage of traits enables us to test whether the same distance decay occurs with respect to functional diversity, or if taxonomically distinct communities in caves can fulfil similar functional roles thereby determining a lower turnover in functional diversity in Europe. Whereas it is well-known that taxonomic and functional diversity decays at different rates along geographical and environmental gradients across different terrestrial and marine habitats and organisms80, similar patterns have never been explored in subterranean habitats.

Trait-based conservation of subterranean spiders

Species traits can be useful in conservation science, for example to assess species extinction risk, to prioritize species and habitat for conservation, and ultimately to define long-term conservation strategies81.

At the individual level, there has been recent interest in understanding the relationship between species traits and extinction risk, namely whether species possessing specific traits (e.g. larger body size, greater longevity) are more prone to extinction82,83. To the best of our knowledge, similar considerations have never been applied to subterranean species, let alone spiders.

At the community level, one can identify species with unique and original traits (‘functional outliers’ sensu ref. 84) versus species falling within densely populated regions of the trait space. This enables the possibility to map the extinction risk across a given global functional spectrum (e.g., the functional space of European cave spiders in Fig. 4) and ultimately to provide general guidance of where to focus in the search for priority species for conservation85. The rationale behind this possibility is that functionally unique species are often irreplaceable, whereas the ecological role of functionally redundant species can be performed by functionally analogous species in the community (‘biological insurance’ sensu ref. 86).

Historically, subterranean ecosystems have been overlooked in global biodiversity conservation agendas87. In recent years, as the conservation importance of subterranean ecosystems is being reaffirmed, there is a growing need to develop objective ways to prioritize subterranean species and regions to protect. There are several examples of studies proposing operational indexes targeting top-priority caves or subterranean sites for protection given a scenario of limited resources invested in conservation88,89,90. These high-priority sites usually end up corresponding with so-called “hotspots of subterranean diversity91”. However, in our view, all these prioritization attempts fail short on one key aspect: they only consider number of species and/or endemism in their protocol to design protected areas or conservation priorities—but see, e.g., ref. 92,93. A modern take on this subject would be to not only consider taxonomic diversity and relative measures, but also to maximize phylogenetic and functional diversity within a given protected area94,95, and even the extent to which species niches are accounted for96. A trait dataset such as the one released in this work is a first, necessary step towards the goal of obtaining a multi-pronged prioritization that accounts for multiple biodiversity facets97. This is of the utmost importance given the current threats on subterranean ecosystems, and the unique conservation challenges associated with these biota98,99,100.