Introduction

Arthropods play a dominant role in the dynamics of practically all natural and human-modified terrestrial ecosystems1,2,3 and have significant economic and health effects. For example, certain insects provide significant economic benefits (e.g. pollination) exceeding $57 billion a year to the United States alone4. Invasive insects, however, cost an estimated $70 billion dollars per year globally5 and insect pests may reduce agricultural harvests by up to 16%, with an equal amount of further losses of harvested goods6. Particularly noteworthy is a subset of arthropods that are disease vectors, transmitting pathogens to and between animals as well as plants. Vector borne diseases cause billions of dollars in crop and livestock losses, every year7,8,9. In humans, vector borne diseases account for more than 17% of all infectious diseases (e.g. malaria, Chagas, dengue, and leishmaniasis, Zika, West Nile, Lyme disease, and sleeping sickness), with hundreds of thousands of deaths, hundreds of millions of cases, and billions of people at risk, annually10,11.

The current economic and health burden of arthropod pests, exacerbated by invasive species, and uncertain effects of climate change12,13, has driven significant research programs and data collection efforts. These include crop pest, mosquito, and tick survey and reporting initiatives14,15,16,17,18, citizen science projects19,20,21, and digitization of museum specimen data22,23, all yielding a rich and growing trove of field-based data spanning multiple spatial and temporal scales. The monitoring of arthropod abundance (e.g. Fig. 1) in different disciplines (e.g., biodiversity research, pest-control assessment, vector borne disease monitoring, and pollination research) has similar objectives — to quantify abundance, phenology and geographical ranges of target arthropod species — and entails similar techniques. However, the data produced by these various efforts are often not reusable, or comparable to similar data, as they are typically not recorded in a standard format (e.g. Darwin Core), or do not provide adequate metadata. In contrast, the advent of journal-mandated deposition of data from high-throughput technologies (e.g. NCBI and GenBank), data and code sharing, and other practices to improve transparency and reusability of research results are increasing rapidly across the sciences24,25,26,27,28,29. Furthering these advances through standardization and public archiving of arthropod abundance data can bring significant benefits, including (1) supporting empirical parameterization and validation of mathematical models (e.g. of pest or disease emergence and spread), (2) validation of model predictions, (3) reduction in the duplication of expensive empirical research, and (4) revealing new patterns and questions through meta-analyses11,30,31,32,33. This will also lead to substantial public benefit through improved human, animal, plant, and ecosystem health, and reduced economic costs.

Fig. 1
figure 1

Population abundance time-series example. From New Jersey light trap mosquito surveillance performed by the Iowa State University Medical Entomology Laboratory from 1977-201761,62. Data available for download on VectorBase50.

Full size image

One of the key impediments to the re-use of these data is the lack of adequate metadata or data descriptors (i.e. data about the data)34,35,36,37. In general, for data to be most valuable to the scientific community, they should meet the FAIR Principles – they should be Findable, Accessible, Interoperable and Reusable – and delineate the key components of good data management and stewardship practices38,39. Data are Findable and Accessible when they are archived and freely downloadable from an online public data repository that is indexed and easily searchable. Interoperability and reusability describe the ease with which humans or computer programs can understand the data (e.g. via metadata) and explore/re-use them across a variety of non-proprietary platforms. Even when data are available, metadata for arthropod abundance data are often absent or not readily interpretable, limiting their reusability at a fundamental level.

Results

A minimum information standard for arthropod abundance data

Here, we present a Minimum Information for Reusable Arthropod Abundance Data (MIReAD) standard for reporting primarily longitudinal (repeated, temporally explicit) field-based collections of arthropods. ‘MIReAD’ also evokes ‘Myriad,’ a countless or extremely great number. Abundance is measured and reported in different ways, and MIReAD fields have been designed to allow researchers to capture this complexity. Examples 1–4 (which can be found in Figshare40), provide examples for how to report such different types of abundances. However, we do not encourage the reporting of relative population abundances since these are not raw data as such, but derived values. One might argue that this could lead to (and it probably will) loss of information, but we argue that the reporting of raw abundance or occurrence data is non-negotiable if these data are to be reused. For example, incorrect statistical methods to aggregate data, such as taking the arithmetic mean for skewed abundance data across samples or replicates are not uncommon, and we wish to discourage such practices. In the same manner as has been developed in other biological disciplines41,42,43,44,45,46, this standard is “minimum” because it defines the necessary minimal information required to understand and reuse a dataset without consulting any further persons, text, materials, or methods47. MIReAD is designed to facilitate data archiving efforts of publishers and field researchers. It is not a data model [the explicit definition of data field names, data formats (e.g., for dates and GPS locations)] and therefore does not define controlled vocabularies, or specific field titles, but should be easy to understand and interpret by the wider scientific community47.

MIReAD is separated into two components, metadata and data. For each component, we provide a description of the information that should be included, recommendations for how to make that information as useful as possible, and examples. The metadata component (Online-only Table 1) includes information for the origin of the data set (e.g. study information and licensing for usage). The second component (Online-only Table 2) lists and describes specific data fields that should be included in data collection sheets. We also provide recommendations and examples to demonstrate how these recommendations can be implemented. MIReAD was designed to match the data that are generally collected by academic researchers and surveillance initiatives, and can serve as a checklist for important information that needs to be recorded but is often unintentionally omitted (e.g. Fig. 2a). By adhering to MIReAD standards, omissions and ambiguity can be avoided even if the data are shared in different formats (Fig. 2b,c). Finally, we identify common problems likely to be encountered across all the MIReAD metadata and data fields, and data quality standards that can be employed to avoid confusion (Box 1).

Fig. 2
figure 2

MIReAD reduces data ambiguity. (a) Seemingly clean data can still lack key information or have ambiguous metadata, hindering data reuse. (b) MIReAD compliant data includes the metadata necessary for data reuse and removes ambiguity. (c) Note data can be formatted differently, but still be MIReAD compliant such as by presenting data in a wide format.

Full size image

Examples

Below we provide three examples to illustrate MIReAD-compliant data (linked to examples 1–4 in Figshare40, respectively). Researchers can use these data sheets as a basis for formatting their own data. In these examples, note that all data meet the data quality standards of Box 1; are adequately described, have columns labeled, etc. to eliminate ambiguity (even if the data appear repetitive; for example, the sex and life stage are repeated in every row). Examples 1 and 2 should be sufficient for most data generators. Examples 3–4 demonstrate more complex data collection scenarios.

Long-format trapping data

Each row captures count data for a single species’ occurrence in a given sampling event. This illustrates an example of the most common mosquito collection protocol (MIReAD_example_1.csv40). Also see Fig. 2b.

Wide format trapping data

Each row captures count data from a given sampling event. Each identified taxonomic group is identified in a separate column. An ‘additional sample information’ field, ‘sub-location,’ has been added to describe the various locations around the village where collections were made (MIReAD_example_2.csv40). This illustrates an example of adult mosquito populations that have been tracked over time and in specific locations. Also see Fig. 2c.

Complex trapping data scenarios

Tick surveillance performed using tick drags and flags and collections of ectoparasites on trapped mice. The tick drags/flags report three life stages independently (adult, larvae, and nymph) (MIReAD_example_3.csv40). Larvae are only identified to the genus, while adults and nymphs are identified to the species. Observations of different life stages and sexes are preferably documented in separate records. A Sample Name is used to help link these records (but would not be necessary.) The mouse survey uses an additional sample information field to record the sex of the trapped mouse from which the parasites were collected (MIReAD_example_4.csv40).

Discussion

MIReAD as the path to FAIR data principles

We designed MIReAD to achieve a balance between standards that are too onerous for data generators with guidelines that are sufficient to ensure at least minimal reusability31,41. It balances a perfectly formatted and reusable dataset with all necessary metadata in a consistent format (but comes with a high burden on the dataset generator) versus a dataset that is unusable or re-usable due to missing or incomplete metadata. MIReAD allows for a relatively easy standardization, as it ensures all necessary collection metadata is present in an unambiguous manner. By not mandating any particular field name, field order, or controlled vocabulary terms, we will in fact gain traction from other more rigorous (and thus more onerous) data models, for which lack of minimum standards in data are often a first and major hurdle. In striking this balance, we note that MIReAD focuses on capturing information on ‘what’ was done, rather than ‘why’. We acknowledge that for some use cases, this may hinder reusability48 but for the majority of cases, where the results of the original data can be interpreted without understanding the rationale, providing data in MIReAD format will be sufficient for data reuse.

Like all minimum standards, MIReAD only aims at ensuring data ‘Reusability’. However, ultimately this will promote the implementation of data models, and controlled vocabularies (e.g., the Darwin Core49). Data models enable ‘Interoperability’, and in turn facilitate structured databases, public repositories, and development of data analysis tools47,50. Deposition in open databases make data ‘Findable’ and ‘Accessible’51,52,53. MIReAD compliant data contain sufficient information for established aggregators/databases such as VectorBase and SCAN (Symbiota Collections of Arthropods Network54) to process and store the data in a standardized data model [e.g., Darwin Core, a widely used universal data standard that supports opportunistic observation and collection data (occurrence core) as well as presence/absence and abundance data collected using strict and documented methodology (event core)49], and ultimately facilitate data transfer to even more comprehensive biodiversity databases. For example, GBIF contains over one billion species occurrence records, from thousands of environmental, ecological, and natural resource investigations, including research on Arthropoda in numerous ecological and monitoring projects, allowing for study of changes and trends in populations53. Indeed, in Tables 1, 2, we provide an example of the mapping of data fields from MIReAD to DarwinCore and GBIF. In this way, MIReAD opens the door to FAIR data and more sophisticated methods to integrate data across many scales.

Table 1 The MIReAD data fields and their mapping to the GBIF metadata profile.
Full size table
Table 2 The MIReAD data fields and their mapping to the Darwin Core data standard.
Full size table

Benefits to field researchers

It is essential that the benefits of a minimal data standard extend not just to data re-users, but also to the researchers who collect and generate data in the first place. MIReAD provides a framework for data preparation that can help scientists achieve recognized professional merit for sharing data such as increased citation rates, academic recognition, opportunities for co-authorship, and new collaborations [sensu Roche et al.31]. Large, deposited data sets can now themselves be standalone, citable “data papers” (e.g.55,56,57) or even depositions without any traditional manuscript (but as an authored ‘digital product,’ with persistent identifiers, such as a DOI number), if desired. Data sets are increasingly recognized as valuable research outputs that count towards academic recognition and professional advancement (e.g. grants, interviews, and tenure). For example, several funders (e.g. United States National Science Foundation and Swiss National Science Foundation) have adopted or are in the process of adopting the Declaration on Research Assessments (DORA)58, offering further opportunities for data generators to gain recognition and publication credit for their work59. Also, an increasing number of funders are mandating public data access, and detailed data management plans are often required even at the grant proposal stage. Therefore, reporting data according to MIReAD will provide a basis for stipulating archival formats. We also note that by storing data in MIReAD format, data generators can assure that their data contains all the necessary metadata for their own internal use. As time passes, research staff, sampling protocols, and sampling locations change, and thus the recording of minimal information ensures long-term reusability of data.

Furthermore, many data generators are also data users. Developing analyses that rely on standardized fields can facilitate the development of generalized analytical tools that can be easily extended to datasets beyond those that were collected by a single individual or lab. In this way, they can enable extensions of work that would otherwise not happen, such as comparisons of population dynamics in different locations or assessments of interspecies interactions. Adopting MIReAD can, therefore, both help data generators reap the benefits of sharing data they have collected and enable them to more readily leverage data collected by others.

Further MIReAD applications and extensions

The creation of minimum information standards for these types of databases facilitates analyses of data at scales that cannot be attained by a single individual or lab group. Linking records to additional information also extends the utility of these data to address population level questions. For example, a well-populated database presents opportunities to investigate interactions between populations of different species of arthropod that overlap in geography but may be of interest individually to different realms of research. As a case in point, in the northeastern USA, Agrilus plannipennis, the Emerald Ash Borer, is a highly destructive invasive insect, monitored closely by both state and federal agencies for management60. Interestingly, Emerald Ash Borers are creating new habitats for carpenter bees, a species interaction that can be tracked and anticipated using large scale arthropod data.

Another example of the utility of linked data is for disease vectors. Data on insecticide resistance linked with time and place would be valuable for coordinating control strategies within and between nations and communities. Presence/absence data on infection levels would be helpful for tracking and investigating disease outbreaks and dynamics. Standardization of these data would be particularly useful for pathogens that infect multiple vectors and hosts and would facilitate a “One Health” approach. Other important vector phenotypes that contribute to control and transmission such as pathogen susceptibility, biting preferences, and breeding behaviours could be measured over time and space.

Indeed, MIReAD would be useable for any arthropod abundance data collection effort, not just medical, veterinary, and agricultural pests or  invasive arthropods. We note this standard is applicable not only to abundance measurements, but could be easily extended to any other kind of routinely sampled time-series field data. For example, in addition to aphid abundance, plant pathogen (such as mosaic virus) infection and insecticide resistance statuses of the aphids could be reported in MIReAD format. Note that MIReAD can also be used for cross-sectional data (i.e. a non-continuous, one-time sampling effort) by simply reporting data from the single collection period by utilizing a single Start Time and End Time (Online-only Table 2).

Disseminating MIReAD

Many data generators are already storing or sharing data in a manner that would be consistent with MIReAD (e.g. on MosquitoNet or NEON), but we call on data generators, authors, reviewers, editors, journals, research infrastructures (e.g. data repositories) and funders to embrace MIReAD as a standard to facilitate FAIR data use and compliance for arthropod abundance data. We propose that workshops, outreach at conferences and meetings, and interfacing with data repositories, societies and organizations (e.g. SpeciesLink, the American Mosquito Control Association, MosquitoNet, Symbiota, VectorNet, and VectorBiTE), and journal editors will be the best way to spread the adoption of this standard.

Conclusion

We present MIReAD as a minimum information standard for representing arthropod abundance data. MIReAD will facilitate collation and analyses of data at scales that cannot be attained by a single individual or lab in order to address key questions across temporal and spatial scales, such as within and across-year phenology of abundance of target arthropod taxa over large geographical areas. This is particularly important given the pressing need to understand and predict the population dynamics of harmful (e.g., disease vectors and pests) as well as beneficial (e.g., pollinators, bio-control agents) arthropods in natural and human modified landscapes. This is the first step for achieving the broad benefits of FAIR data for arthropod abundance.