Spatial and temporal trends in dung beetle research

Literature search

We chose to limit our search to peer-reviewed articles. We acknowledge that there is a large amount of grey literature and other publications, such as books on the subject of dung beetles. However, much of this literature is not accessible to a global audience and is not peer reviewed. As a result, we targeted the most scientifically robust and readily available publications, an approach which fits well within our aim to explore the temporal and spatial trends of dung beetle research.

We searched the Scopus and Web of Science literature indexers using the following Boolean search terms: dung & beetle*; scarabaeinae; aphodiidinae; geotrupinae; coprophag* & beetle*; coprophag* & scarab*; coprophag* & coleop*. These results were combined with another Endnote library compiled by the authors, which consisted of articles published from 1933 to 2017. This library consisted of articles retrieved from Scopus and Google Scholar using the search terms: dung beetle; scarabaeinae; aphodiinae; geotrupinae. Web of Science returned a total of 3,179 publications, Scopus returned a total of 3,140 publications, and the existing library contained 1,448 publications. After removing duplicates, the final corpus contained 4,145 articles published between 1933 and January 2024. The data file used here can be found on Figshare (Hemmings, Evans & Andrew, 2024a).

Topic modelling

To prepare the corpus for topic modelling, we used the ‘textProcessor’ function in the ‘stm’ package (Roberts, Stewart & Tingley, 2019) in R (R_Core_Team, 2023), to remove punctuation, stop words, numbers, and words with fewer than three characters. We also stemmed words to their root form (e.g., walk = walked, walking, walker) and removed the most rare (in <1%) and common (in >85%) words in each abstract and title.

We then fitted a structural topic model (STM) using the ‘stm’ package, analysing abstracts and titles of our corpus. After trialling a number of topics, we chose 20 topics as a number large enough to provide sufficient detail to analyse the topic landscape of our corpus, but not too large as to be overly complex (Andrew et al., 2022; Evans et al., 2023b; Westgate et al., 2015). We used spectral initialisation for model fitting (Roberts, Stewart & Tingley, 2016). We then gave our topics a short summary title by examining the 20 highest-weighted words of each topic (Westgate et al., 2015).

Post-hoc topic analyses

To examine the temporal prevalence of topics over time, we used the ‘estimateEffect’ function in the ‘stm’ package, to fit topic prevalence through time, treating year as a linear term. We used the ‘hot’ and ‘cold’ topic nomenclature to describe topics as either increasing or decreasing in prevalence over time, respectively (Evans et al., 2021; Westgate et al., 2015). We examined topic similarity by undertaking a hierarchical cluster analysis using Ward’s minimum variance method on the dissimilarities of the model-derived probabilities of word occurrence matrix (Ward, 1963). We then grouped these topics into six groups of closely related articles based on this clustering (Evans et al., 2023b). We also explored how each topic was distributed through the corpus by examining the specificities and generalities of each topic within the whole corpus (Westgate et al., 2015). Topics that are considered general would have topic weights that span multiple topics, whereas topics that are considered specific would have topic weights heavily biased towards one topic. To calculate this, we assigned each article to its highest-weighted topic and calculated mean weights for each topic of those articles selected, vs those that were not selected (Westgate et al., 2015).

Taxanomic entity extraction

We used the Global Names Finder v1.1.3 (https://finder.globalnames.org/) to scan for taxonomic mentions in all abstracts and titles in the corpus. We then fetched genera, orders, classes, and kingdoms of all taxonomic names, using the National Center for Biotechnology Information database Application Programming Interface (API) through the ‘taxize’ package (Chamberlain & Szöcs, 2013) in R.

Geoparsing

We scanned all article abstracts and titles for geographic mentions using the CLIFF-CLAVIN geoparser (D’Ignazio et al., 2014; Millard, Freeman & Newbold, 2020) in Python using a Docker container (Merkel, 2014) hosted on the GitHub repository (<https://github.com/havlicek/CLIFF-docker>). CLIFF-CLAVIN is able to resolve mentions to the most likely physical coordinates. We categorized the mentions into ‘minor’ mentions (specific locations within countries) and ‘major’ (countries) (Millard, Freeman & Newbold, 2020). We then assigned World Wide Fund for Nature (WWF) biomes for all the minor mention locations (Olson & Dinerstein, 2002).

Q1. What are the key research topics that have been carried out using dung beetles?

We based our assessment on 20 topics (Table 1) across the entire dung beetle corpus based on the word clouds generated (Appendix 1). The 20 topics split into six broad dendrogram clusters (Fig. 1): three large clusters and three single-topic clusters. The largest topic cluster (Cluster #1) was agricultural and biologically focused. It included ‘Nesting behaviour’, ‘Food preference and diet’, ‘Navigation’, ‘Statistical analysis’, ‘Biomimetics’, ‘Soil health and plant growth, ‘Agricultural associations’, ‘Conservation & biodiversity’, and ‘Abiotic response variables’. This is a broad and diverse cluster primarily driven by a high overlap of keyword usage. Four subgroups with more ecological alignment come out within the group: Subgroup 1: ‘Nesting behaviour’, and ‘Food preference and diet’ relate to biological interactions the beetles have with dung, and their preference for offspring food provision. Subgroup 2: ‘Navigation’, ‘Statistical analysis’ and ‘Biomimetics’ are aligned with data analysis and statistics (da Silva, Mota Souza & Neves, 2022), methods of assessing dung beetle movement (Dacke et al., 2013), and the use of dung beetles in the development of new technologies (e.g., Tong et al., 2005; Wang et al., 2018). Subgroup 3 ‘Soil health and plant growth’ and ‘Agricultural associations’ directly aligns with the roles that dung beetles play in farm productivity (Beynon, Wainwright & Christie, 2015; Doube, 2018). Subgroup 4: ‘Conservation and Biodiversity’ and ‘Abiotic response variables’ aligns with research in native environments and ecological research relating to aspects such as climate change (Maldaner et al., 2021).

The second largest topic cluster (Cluster #6) was ecologically focused. It included ‘Ecosystem function’, ‘Landscape ecology’, ‘Species distributions’, ‘Sampling’, and ‘Assemblage structure’. This was a clear cluster around the ecological functions dung beetles provide that are key to the continued existence of many habitats (Noriega et al., 2023). The provision of these functions also provides a number of services that directly benefit humans (deCastro-Arrazola et al., 2023; Nichols et al., 2008), as well as the way the dung beetles are collected (Heddle, Hemmings & Andrew, 2023) and the associated assemblage structures (Noriega et al., 2021).

The third largest topic cluster (Cluster #2) was taxonomically focused. It included ‘Taxonomy’, ‘Phylogeny’, and ‘Scarabaeidae’ clustered together. This was a clear taxonomic grouping based on species naming and descriptions. One of the key elements to this grouping was the taxonomic name change of Onthophagus to Digitonthopohagus in 2017 (Génier & Moretto, 2017). Additionally, species descriptions and taxonomic changes have been especially active (Cupello, Silva & Vaz-de-Mello, 2023).

The topics ‘Seed dispersal’, ‘Sexual selection & sexual traits’, and ‘Effect of veterinary chemicals’ formed unique topic clusters (respectively Cluster #3, #4, and #5). All three are unique and distinctive topics. ‘Seed dispersal’ is an explicit ecosystem service that dung beetles provide (Manns et al., 2020; Midgley et al., 2015). ‘Sexual selection and sexual traits’ has focused on male selection for reproductive success at both the adult (Kotiaho et al., 2003) and sperm level (House & Simmons, 2006). ‘Effect of veterinary chemicals’ is aligned with the loss of ecosystem service provisions provided by dung beetles in agricultural ecosystems based on the use of antiparasitic drugs on cattle, sheep and other livestock (Mackenzie et al., 2021; Verdú et al., 2020; Weaving, Sands & Wall, 2020).

General vs specific topics

General topics (Fig. 2) are found in the bottom right-hand corner and specific topics are found in the top left corner. ‘Agricultural associations’ and ‘Assemblage structure’ were defined as a general topics. Specific topics included ‘Seed dispersal’, ‘Biomimetics’, ‘Sexual selection & sexual traits’ and ‘Taxonomy’; ‘Effect of veterinary chemicals’ was weakly defined as a specific topic.

Q2. How has topic prevalence changed over time?

Articles published on dung beetles have exhibited an exponential increase over time (Fig. 3A), with a rapid rise after the year 2000: this trend is consistent with other insect-related publications (e.g., Andrew et al., 2013). Topics showed a range of temporal trends over the period of research assessed (1933 to 2024) (Fig. 4). ‘Ecosystem function’ had the largest increase in prevalence over the time period. Modest increases in prevalence were found in topics including ‘Landscape ecology’, Conservation & biodiverstiy’, ‘Abiotic response variables’, ‘Statistical analysis’, and ‘Species distributions’. No change in prevalence was found in ‘Taxonomy’, ‘Phylogeny’, ‘Sampling’, ‘Biomimetics’, ‘Soil health & plant growth’, ‘Seed dispersal’, ‘Food preference and diet’, and ‘Sexual selection & sexual traits’. Topics that reduced in prevalence include ‘Agricultural associations’, ‘Assemblage structure’, Scarabaeidae’ and ‘Navigation’. The largest reduction in prevalence during this period included ‘Nesting behaviour’ and ‘Effect of veterinary chemicals’. ‘Nesting behaviour’ had been a popular topic (top 4) up until the 1990’s and ‘Veterinary chemicals’ had been the most popular topic in the 1970s–1990’s (Fig. 3). This was likely caused by i) research for the Australian dung beetle project (e.g., Bornemissza, 1976), and ii) the necessity for fundamental research into the behaviour and biology of common species (e.g., Ridsdill-Smith, 1988, 1993; Ridsdill-Smith, Hall & Craig, 1982), which once undertaken, have paved the way for more complex research on the interactions between functional groups and their effect on the environment. ‘Ecosystem function’ only emerged as a topic in the 2000’s and had a relatively meteoric rise over the following two decades (Fig. 5). While ecosystem services were studied prior to 2000 (Bornemissza, 1976; Hughes, 1975), it was not until post 2000 that the term ‘ecosystem function’ saw common usage in the literature, but the terminology used was different.

Q3. How are these topics distributed globally?

Most of the topics were dominated by research emerging from North America and Europe (Figs. 3B, 6). This is an example of the Global North domination in research published in English across a range of biological disciplines (Ballari et al., 2020; Piguet, Kaenzig & Guélat, 2018), but more broadly across a range of research diciplines (Collyer, 2018; Oztig, 2022), as well as issues of publishing using English language (Haelewaters, Hofmann & Romero-Olivares, 2021; Zenni & Andrew, 2023). The most ‘global’ topics include ‘Nesting behaviour’, ‘Taxonomy’, ‘Landscape Ecology’, ‘Species Distributions’, and ‘Sampling’. All topics are represented in North America, Western Europe and Africa. Research topics not covered in Eastern Europe include ‘Phylogeny’, ‘Seed dispersal’ ‘Sexual selection & sexual traits’, ‘Effect of veterinary chemicals’, ‘Ecosystem function’, ‘Conservation & biodiversity’, ‘Species distributions’, ‘Statistical analysis’, ‘Agricultural associations’, ‘Assemblage structure’, and ‘Biomimetics’. Topics not covered in Asia, excluding Japan is ‘Sexual selection and sexual traits’; and including Japan is ‘Statistical analysis’. Not included in Australasia include ‘Seed dispersal’ and ‘Navigation’. ‘Biomimetics’—the study of nature that may benefit science, engineering, and medicine more broadly (Fayemi et al., 2017)—was not a topic covered in South America.

Q4. How are key topics aligned with different biomes globally?

‘Scarabaedae’ and ‘Taxonomy’ topics had a strong association with the Temperate Broadleaf & Mixed Forests, and an association with Tropical & Subtropical Moist Broadleaf Forests and Deserts & Xeric Shrublands WWF biomes (Fig. 7). ‘Landscape ecology’ was strongly associated with Tropical & Subtropical Moist Broadleaf Forests. Mangroves, Boreal Forests/Taiga, Tundra and Flooded Grasslands & Savannas were not represented. The other topics had few, if any, mentions/studies with the remaining WWF biomes.

Q5. What are the key taxa associated with dung beetle research?

Onthophagus was the most studied genera (483 mentions) in the articles that we analysed (Fig. 8A). This reflects Onthophagus’ diversity and global distribution: it is among the most speciose genera in the animal kingdom, with ca. 2,300 species (Breeschoten et al., 2016). Additionally, Onthophagus is a model organism for the study of the evolution of sexual dimorphism and the development of male horns (Kijimoto et al., 2013; Moczek, 2011). Aphodius, Copris, and Canthon were a distinct cluster of second-ranked studies (157, 147 and 146 mentions respectively). Dichotomius, Euoniticellus, Scarabaeus, Digitonthophagus, and Phanaeus had between 88 and 55 mentions. As mentioned previously, taxonomic reclassifications will influence the assessment of species and genera names (Cupello, Silva & Vaz-de-Mello, 2023; Génier & Moretto, 2017).

The two dominant dung beetle families ‘Scarabaeidae and Geotrupidae’ had the most mentions (2,229 and 129 respectively, Fig. 8B). Muscidae were the next common group (88 mentions). At the order level, Coleoptera were the most dominant (2,622 mentions; Fig. 8C), followed by Diptera (157 mentions), Mesostigmata, and Primates (84 and 69 respectively). Due to the strong commensal relationship between dung beetles and vertebrates, there is often extensive co-evolution between the two groups, with dung beetle communities inextricably linked to the historic and contemporary structure of the associated vertebrate community (Bogoni et al., 2016). The availability of multiple types of dung in an ecosystem is strongly correlated with the tribal diversity of the associated dung beetle community, with abiotic processes having a stronger influence on diversity at the generic and species levels (Davis & Scholtz, 2001). The decline of large-bodied dung beetles in Europe has been proximally tied to the extinction of local megafauna and the subsequent loss of large, wet dung as resource (Schweiger & Svenning, 2018), with a similar situation thought to have occurred in Australia (Doube, 2018). Changes in the composition of the vertebrate community can have strong effects on dung beetles of a particular functional group. The density of deer was associated with an increase in the abundance of small bodied dung beetle species (<10 mm), whereas the abundance of large bodied species (>10 mm) was unaffected (Iida, Soga & Koike, 2018). In Panama areas of forest, dung beetle communities differed between areas with no hunting and areas where monkeys were hunted; hunted fragments showing decreased species diversity, with the abundance of nocturnal beetles negatively correlated with the abundance of mammals (Andresen & Laurance, 2007). Flow-on effects from this can further influence the community, as the presence/proportion of larger, more dominant species of dung beetle influences the structure of smaller species via size-asymmetric competition (Horgan & Fuentes, 2005).

With family associations with topics, Scarabaeidae had a high or medium prevalence with all topics, (Fig. 9), Geotrupidae had a weak association with all topics, and Muscidae had a medium association with ‘Effect of veterinary chemicals’. Dung beetles are known to control muscid fly abundance in agricultural dung (Kirk, 1992; Smith & Matthiessen, 1984), and the influence of agro-chemicals can reduce the survival, growth and development of dung beetles (Mackenzie et al., 2021). All other families had a weak or no association with the 20 topics identified.