Beginning in October 2022, BOLD v.4 (Ratnasingham and Hebert 2007) was manually datamined for Platygastroidea BINs with intercontinental distributions. Our search terms for Platygastroidea followed the revised classification of Chen et al. (2021), which recognized Geoscelionidae, Janzenellidae, Neuroscelionidae, Nixoniidae, Platygastridae, Scelionidae and Sparasionidae as the extant families. The BOLD BIN database portal was queried with the following family and genus level terms: “Geoscelionidae”, “Janzenellidae”, “Neuroscelionidae”, “Nixoniidae”, “Platygastridae”, “Scelionidae”, “Sparasionidae”, “Huddlestonium”, “Plaumannion”, “Janzenella”, “Neuroscelio”, “Nixonia”, “Archaeoteleia”, “Mexon”, “Listron”, “Sceliomorpha” and “Sparasion”. The resulting BINs were then assessed for the number of countries in which they occurred. BINs with multiple countries were further scrutinized. If the countries were on different continents or otherwise spanned large distances (e.g., Pacific islands and the mainland), then the BIN was included in the dataset. A set of information was recorded and downloaded for each intercontinental BIN: BOLD BIN, BOLD taxonomy lowest level, maximum intra-BIN p-distance, distance to nearest neighbor BIN, the BIN fasta file of COI sequences and the Darwin Core (Wieczorek et al. 2012) BIN metadata text file (10.5281/zenodo.7930011; 10.5281/zenodo.7930274).

In February 2023, the dataset was expanded using an automated scripting approach (10.5281/zenodo.7930407). The Darwin Core data files for “Insecta” and “Araneae” were downloaded from BOLD. The Insecta files were downloaded on 3 January 2023. The Araneae files were downloaded on 16 March 2023. The Insecta file was examined for all the categories listed under the field “country”. The country fields were assigned to continents or island categories (Suppl. material 1). Continent assignments largely followed the United Nations Statistics Division (2023) designations, except in a few edge cases where changes were made to follow the borders of zoogeographic regions more closely. Edge cases were assigned to groups based upon the closest alignment of political boundaries and biogeographical barriers. To separate the Asian and European continents, Russia was divided along provincial borders that most closely continue the line of the Ural Mountains southward. Turkey, Georgia, Armenia and Azerbaijan were included in Europe to avoid dividing the Caucasus region. Island nations sufficiently distant from continents were placed into four categories based on oceanic region. The Indonesian provinces of Papua and West Papua were included in the Australian region to avoid dividing the island of New Guinea. Solomon Islands, New Caledonia and New Zealand were grouped with Australia.

Each BIN in the intercontinental and island platygastroid dataset was individually analyzed to determine whether they contained exact COI barcode matches across large distances. Each BIN’s COI fasta file was aligned using the default settings of MUSCLE (Edgar 2004) as implemented in MEGA7 (Kumar et al. 2016). In MEGA7, these individual BIN alignments were used to build neighbor-joining trees (when there were more than two terminal taxa) and distance matrices. Analysis settings were identical for both approaches: the data were labelled as protein-coding nucleotide sequences with the invertebrate mitochondrial genetic code, p-distance set as the method/model, substitutions included transitions and transversions, rates among sites were uniform and missing data treatment was set to partial deletion with a 95 percent site coverage cut-off. P-distance calculations involving ≤ 255 terminal taxa were exported as matrices. P-distance calculations involving > 255 terminal taxa were exported as pairwise columns. Neighbor-joining trees were exported as Newick tree files and uploaded into the online annotation portal of Interactive Tree of Life (iTOL) v.5 (Letunic and Bork 2021).

Trees were viewed in iTOL to determine individual sequence membership in haplotype clusters. Apparent haplotype matches were then examined in the p-distance calculation files for confirmation. The geographic distribution of exact matches was evaluated by examining the specimen level metadata present in the BIN’s Darwin Core text file. Putative exact intercontinental matches were then validated in the underlying DNA alignment. This was necessary due to slight variation in the length of COI barcode sequences. If longer sequences in the alignments displayed polymorphisms toward either the 5’ or 3’ ends and the putative intercontinental matches lacked these flanking data, then the matches were considered inconclusive. Ambiguous DNA base pairs were ignored for considering exact matches.

Individual BIN alignments were ultimately combined into one fasta file for tree building and visualization. A species of Periclistus Förster (Hymenoptera, Proctotrupomorpha, Cynipidae) with appropriate data coverage was selected to root subsequent analyses, based on the sister relationship of Cynipoidea to Platygastroidea (Blaimer et al. 2023). Alignment of the combined intercontinental and island COI barcode dataset was performed in the MAFFT online service v.7 (Katoh et al. 2019) with the FFT-NS-1 setting. This alignment (Suppl. material 3) was used for neighbor-joining analysis in MEGA11 (Tamura et al. 2021) using the same tree building parameters described above. This tree topology (Suppl. material 4) was manipulated for viewing ease and annotated in iTOL v.5 (Letunic and Bork 2021) and FigTree v.1.4.4 (Rambaut 2018) (Suppl. material 5).

During the above data-mining activity, it was noticed that a large proportion of platygastroid BINs were unidentified below the family level. A list of Platygastroidea BINs was pulled from BOLD as candidates for identification using the provided digital morphology framework. Identifications were made, when possible, by comparison to images of primary type specimens provided by Talamas et al. (2017). Most Platygastroidea BINs were sight-identified using the BOLD BIN images to family, sub-family, tribe or genus by taxonomic specialists Elijah Talamas, Zachary Lahey and Jessica Awad.

Genus level occurrence data (Table 1) were downloaded as Darwin Core Archives from the Global Biodiversity Information Facility (GBIF 2022) for all platygastroid genera identified in the intercontinental BIN dataset. These occurrence data were sorted by species, then country, to determine which species had records from more than one continent (Suppl. material 6). Relevant taxonomic revisions were searched to corroborate intercontinental taxa reported in GBIF and BOLD. For a total evidence analysis of Platygastroidea, the GBIF, literature and BOLD datasets were combined to assess their continental and island connections. Species with records from multiple data types were only counted once and their geographic occurrences were rectified for analysis.

Topotypes of platygastroid species described from Duval County, Florida, were collected between July 2018 and December 2021 in and around the Timucuan Ecological and Historic Preserve near the mouth of the St. Johns River. These collections were largely from Malaise traps that were placed at three different sites in the area. Collecting heads were provisioned with propylene glycol and wrapped with aluminum foil to prevent UV damage to the specimens. 3D printed yellow cylinder traps were experimentally deployed on the ground and suspended from overhanging branches during one sampling period and additional specimens were collected in yellow pan traps and by sweep netting around the trap sites. Bulk samples were returned to the laboratory and sorted under a Zeiss Discovery V8 Stereomicroscope. All platygastroid specimens were transferred to 95% ethanol and screened for matches to species described from that area.

DNA was non-destructively extracted from specimens using the Qiagen DNeasy Blood and Tissue Kit. Molecular voucher specimens were recovered and deposited at the Florida State Collection of Arthropods (Florida Department of Agriculture and Consumer Services – Division of Plant Industry; Gainesville, Florida). PCRs were conducted as 25 µl reactions using the KAPA HiFi HotStart Readymix Kit (Roche Diagnostics) per the manufacturer’s recommended protocol. Oligonucleotide primers used for PCR and direct sequencing were the universal arthropod COI barcoding sets LCO1490/HCO2198 (Folmer et al. 1994) and LEP-F1/LEP-R1 (Hebert et al. 2004). PCR products were visually verified by gel electrophoresis and positive products were prepared for sequencing with the Qiagen QIAquick PCR purification kit. PCR products were bidirectionally sequenced utilizing BigDye Terminator v.3.1 chemistry on the Applied Biosystems SeqStudio platform. Sequence traces were trimmed and assembled into contigs in Sequencher 5.4.8 and Geneious Prime. New COI barcodes were uploaded to BOLD and GenBank (OQ561913–OQ561961) and assessed for their nearest matches.

The German Barcode of Life III: Dark Taxa project (Hausmann et al. 2020) provided 34 platygastrine specimens matching 14 BINs with intercontinental distributions. These were selected for closer morphological examination and taxonomic analysis. Methods for trapping, non-destructive DNA extraction and COI sequencing follow Awad et al. (2021). Species partitioning was performed with ASAP (Puillandre et al. 2021) using the Jukes-Cantor model and default web server settings.

Following non-destructive DNA extraction, voucher specimens were mounted and photographed with a Macropod imaging system consisting of a Canon EOS 6D Mark II camera body, EF 70–200 mm lens and 10× or 20× M Plan APO Mitutoyo objective lenses. Imaging software included Canon EOS Utility 3.14.30.4 and Helicon Focus Pro 7.7.5 for image stacking. Adobe Photoshop 23.2.2 was used for limited post-processing and addition of scale bars. Images were uploaded to BOLD with specimen metadata.

An R (R Core Team 2022) script was developed that created two outputs: 1) a list of BINs with intercontinental or mainland/island distributions and their associated metadata (e.g., taxonomic information) and 2) a list of BINs with intercontinental or mainland/island distributions and an individualized tally of long-distance occurrences. The platygastroid data were compared to our manually-extracted list. The taxonomic spread in the Insecta file was summarized at the family and genus level, with special attention given to known host groups for platygastroids. Summary statistics were completed with R 4.2.2 (R Core Team 2022), using the dplyr (v.1.0.10; Wickham et al. (2022)), data.table (v.1.14.8; Dowle and Srinivasan (2023)), and stringr (v.1.4.1; Wickham (2022)) data manipulation packages. Data was visualized using the package ggplot2 (v.3.4.1; Wickham (2016)). The full reproducible code is available in Supplementary materials (10.5281/zenodo.8380145).

A randomized 5% of the recovered Insecta and Araneae BINs were validated by manual examination in the BOLD BIN database (Suppl. materials 2, 19). Only the first page of returned search results (which contain order, species and country level distribution) were examined in the validation process. BINs with many different species identifications were scored as having minor taxonomic conflicts. BINs with multiple species or genus identifications were scored as having major taxonomic conflicts, except when it appeared that specific epithet discrepancies were due to varying genus classifications. The list of recovered Insecta and Araneae BINs was summarized at the family and genus level. BINs containing known host groups of Platygastroidea were similarly summarized. The Insecta and Platygastroidea occurrence datasets were transformed into pairwise matrices capturing their geographic distributions. Matrices (Suppl. materials 10, 12–16) were imported to the online version of Circos (Kryzwinski et al. 2009) for visualization. Default settings were used in the online Circos viewer with the exception of the following: 1) labels segment set to large, 2) data filters intra-segment cells hidden, 3) row and column segments order set to col/row ratio, 4) row and column segments with normalized segment size set to remap segments size to 1000, 5) contribution tracks set to hide and no stroke, 6) ribbon caps completely disabled, 7) ratio layout enabled, 8) image format with no strokes and all tick labels hidden.

Two hundred and one platygastrid and scelionid BINs were found to have intercontinental and island distributions (Table 2, Suppl. material 11). No cases were detected in the other platygastroid families. Of these, 140 were identified only to the family or subfamily level in BOLD, sometimes incorrectly. We identified all but one of these 140 BINs to a lower taxonomic level; 130 of them were identified to genus or lower. The dataset contains 27 genera (10 Platygastridae genera; 17 Scelionidae genera) (Table 2, Suppl. material 11). Telenomus (69 BINs) and Platygaster (35 BINs) are the most represented genera in their respective families. Canada (101 points) and Germany (90 points) are the most represented country level occurrence points in the platygastroid BIN dataset (Table 2, Suppl. material 11), likely biased by the existence of nationally directed barcoding initiatives.

The minimum BIN size was two (necessary for a geographic match) and the largest, a Platygaster species, contained 677 COI sequences. Most BINs (180 of 201) contained fewer than 100 COI sequences (Fig. 1, Suppl. material 11). Intra-BIN maximum p-distances reported by BOLD ranged from 0.00% to 5.42% (Suppl. material 11). Most BINs (157 of 201) in the platygastroid dataset had an intra-BIN maximum p-distance of less than 2.50% (Suppl. material 11). P-distance to the nearest-neighbor BIN was reported from all the platygastroid BINs. These p-distances ranged from 0.99% to 17.86% (Suppl. material 11). Twenty-eight BINs had larger intra-BIN p-distances than the p-distances to their nearest-neighbor BIN (Suppl. material 11). Fifty-five of the platygastroid BINs displayed exact COI matches across distant localities. These exact matches were discovered in Amblyaspis (2), Amitus (1), Anteris (1), Aphanomerus (1), Dicroscelio (1), Euxestonotus (2), Fidiobia (1), Gryon (2), Idris (3), Leptacis (2), Platygaster (9), Psix (1), Synopeas (5), Telenomus (18), Trimorus (1), Trissolcus (1), unidentified Platygastrinae (2), unidentified Scelioninae (1) and an unidentified Telenominae (1) (Suppl. material 11). Six BINs also included two exact intercontinental haplotype matches: Euxestonotus (1), Gryon (1), Platygaster (1), Synopeas (2) and Telenomus (1) (Suppl. material 11). An additional 28 BINs contained COI sequences that appeared to be exact matches, but were scored as inconclusive due to sequence length variation, precluding the most meaningful comparisons (Suppl. material 11).

Figure 1.

Pareto chart displaying the number of COI barcodes contained per BIN in the intercontinental and island platygastroid dataset.

In the total evidence Platygastroidea dataset, North America (248 connections), Europe (211 connections) and Asia (198 connections) were the most common connections (Fig. 2, Suppl. material 12). Telenomus and Trissolcus were the most represented scelionid genera across continents and islands. In Platygastridae, Platygaster and Synopeas were the most represented genera. The webs of geographic connection found in Telenomus and Trissolcus are more complex than those of Platygaster and Synopeas (Fig. 3, Suppl. materials 13, 14). For example, Synopeas and Platygaster lack any conclusive geographic data points in the Caribbean and Pacific Ocean and only one each in Africa. Europe and North America are the most connected continents in these genera, except for Trissolcus in which Asia and Europe are the most connected (Fig. 3, Suppl. materials 15, 16).

Figure 2.

Circos plot displaying the relative contribution of continents and oceanic island groups to geographic connections in the combined platygastroid BOLD BIN, GBIF and literature dataset. Numbers in parentheses indicate the total number of connections to that geographic grouping.

Figure 3.

Circos plot displaying the relative contribution of continents and oceanic island groups to geographic connections in the combined Platygaster, Synopeas, Telenomus and Trissolcus BOLD BIN, GBIF and literature datasets. Numbers in parentheses indicate the total number of connections to that geographic grouping.

The intercontinental and island BIN dataset contains 9,874 platygastroid COI barcode sequences. The MAFFT alignment was 669 base pairs wide after trimming excessive data from the 3’ end of some barcodes. The alignment contains some gap regions due to varying COI amino acid phenotypes present among Platygastroidea (Talamas et al. 2021). All positions with less than 95% site coverage were eliminated in the combined BIN p-distance neighbor-joining analysis; 520 positions were included in the final analysis. Terminal sequence clusters were collapsed and annotated to reflect their BIN assignment, revised or confirmed genus level identification and the number of sequences in that cluster (Fig. 4, Suppl. material 5). These terminal clusters corresponded unambiguously to their BIN assignment (i.e., they formed “monophyletic groups”), except in the case of Telenomus BOLD:AAU4881.

Figure 4.

Circularized p-distance neighbor-joining tree of the intercontinental and island BOLD BIN dataset. The BOLD BINs containing Gryon aetherium, Platygaster sp. (BOLD:ACI8542), Psix striaticeps, Trissolcus basalis, Tr. japonicus and Telenomus sp. (BOLD:ACY0393) are highlighted by an enlarged terminal cluster. Lines emanating from these clusters show generalized geographic localities where the BIN was detected. Solid black lines indicate different COI haplotypes. Solid red or blue lines indicate exact COI haplotype matches across continents. Striped red and blue lines indicate multiple exact COI haplotype matches found at that generalized geographic locality.

A total of 2,565 Platygastroidea BINs were evaluated for their identification accuracy using the specimen images provided by BOLD (Suppl. material 17). Nearly all these BINs were correctly identified to superfamily (12 were misidentified to order or family). Updated subfamily classification was provided for 2,551 platygastroid BINs (Suppl. material 17). A total of 2,209 BINs had images of sufficient quality to add genus level identifications. Sixty-four platygastroid genera were present among these BINs (Suppl. material 17), of which eight represent the first DNA sequence data for the genus in GenBank or BOLD: Euxestonotus, Gastrotrypes, Isocybus, Metaclisis, Parabaeus, Embioctonus, Styloteleia and Xenomerus.

GBIF and literature searches returned 130 intercontinental taxa (37 Platygastridae; 93 Scelionidae), with an overall discrepancy of plus 67 BINs (Table 3). The genera Amblyaspis, Leptacis, Baeoneurella, Baryconus and Xenomerus were present only in the BOLD BIN dataset (Table 3, Suppl. material 11). Conversely, the genera Inostemma, Tetrabaeus, Duta, Platyscelio, Probaryconus and Scelio were present only in the GBIF or literature datasets (Table 3, Suppl. material 6). In the process of analyzing the GBIF data, we encountered errors that were likely to be detected only by those with intimate knowledge of the group: Trissolcus japonicus (Ashmead) was listed as Gryon japonicum (Ashmead); records for Trissolcus cultratus (Mayr) were incomplete; Calliscelio elegans (Perkins) was listed as Caenoteleia elegans (Perkins); and Telenomus dalmanni (Ratzeburg) was misspelled as Telenomus dalmani. Notable name changes are that Trissolcus davatchii (Javahery) is now treated as a junior synonym of Tr. elasmuchae (Watanabe) and Tr. grandis (Thomson) is a junior synonym of Tr. belenus (Walker).

Talamas et al. (2017) provided names and images for species described from Jacksonville with primary types in the National Museum of Natural History. This enabled identifications for taxa that have yet to be revised with modern standards, at least for species with distinctive morphology. For example, Synopeas cynipsiphilum (BOLD record SUPER036-23) has a conspicuous divide between the mesoscutum and mesoscutellum that make it easy to recognize. For many others, diagnostic characters were either unknown or too subtle for us to confidently make a determination of species without studying the taxon in detail. In some cases, we did not make a determination because the specimens we sequenced were not the same sex as the primary type. As these taxa are revised, we are certain that more matches will be made between our vouchers and primary types from the region. Of the specimens that yielded COI barcode data, we matched seven specimens to primary types, totalling six species (Table 4).

Forty-nine specimens were COI barcoded from Timucuan Ecological and Historic Preserve and Buck Island, Jacksonville, Florida (Table 4). Twenty-eight specimens had BOLD identification hits greater than 97% matches (Table 4), representing 25 BINs. Specimen vouchers FSCA 00094179 and FSCA 00094185, both identified as Telenomus sp., were 100% matches to a specimen from San Diego, California, USA in the BIN BOLD:ACY0393 (Table 4). This is an intercontinental Telenomus BIN identified from the Insecta scripting procedure (Fig. 4), now with geographic data points in California, Florida and Bangladesh.

Comparison of BOLD data to platygastrine specimens from the GBOL III barcoding initiative yielded 14 intercontinental BINs, representing 11 species in five genera (Amblyaspis, Euxestonotus, Leptacis, Platygaster and Synopeas). BOLD identified six species (seven BINs) only to family, two species (four BINs) to subfamily and three species (three BINs) with binomials. Of the three species identifications provided by BOLD, we verified two (Pl. demades and Pl. sagana) by comparison of voucher specimens to primary types, while one (Pl. tuberosula) was unverifiable. One more species (E. error) was unidentified in BOLD, but identifiable by our own examination. The remaining seven species were unidentifiable due to the superficial description impediment (Meier et al. 2022) in platygastrine taxonomy.

The distributions of two species (Pl. demades and putative Pl. tuberosula) can be explained by deliberate introductions for pest control on apple/pear and wheat, respectively. One species (E. error) is probably an unintentional introduction, moving with its host, the wheat midge Sitodiplosis mosellana (Géhin) (Echegaray et al. 2016). The biogeographic history of the remaining eight species could not be determined. It is possible that one of the unidentified Platygaster species (either BOLD:ACP1536 or BOLD:AAZ3286) is Platygaster hiemalis Forbes, deliberately introduced to New Zealand for control of the Hessian fly, Mayetiola destructor (Say) (Ferguson et al. 2007). However, the type material of Pl. hiemalis is unknown and the historical literature provides conflicting diagnoses.

The R script recovered 15,391 Insecta BOLD BINs with intercontinental and island distributions (Suppl. material 7). Members of 23 Insecta orders were present in the dataset, with only Mecoptera, Notoptera (Mantophasmatodea and Grylloblattodea) and Zoraptera absent. Due to conflicting taxonomies present in BOLD, the family and genus representation numbers are close estimates of higher-level diversity. Orders with the most family- and genus-group diversity were Coleoptera (66 families; 560 genera), Lepidoptera (77 families; 1,984 genera), Diptera (82 families; 899 genera), Hymenoptera (62 families; 736 genera) and Hemiptera (64 families; 454 genera) (Table 5). Lepidoptera and Diptera contained the most BINs in the dataset (Table 5). The R script recovered 499 Araneae BOLD BINs with intercontinental and island distributions (Suppl. material 18). Members of 42 Araneae families were present, comprising 224 genera (Suppl. material 18).

We randomly selected 769 BINs from the Insecta dataset for cursory validation of the scripting process (Suppl. material 2). Validated BINs were examined in Blattodea (1), Coleoptera (50), Diptera (166), Ephemeroptera (5), Hemiptera (38), Hymenoptera (96), Lepidoptera (391), Neuroptera (1), Orthoptera (6), Plecoptera (1), Psocodea (1), Thysanoptera (6) and Trichoptera (7). No validation BINs were discovered to be geographic false positives upon the initial pass. However, four BINs (BOLD:AAC6546, BOLD:AAP8198, BOLD:AAE7880 and BOLD:AAD4954) were recalculated in BOLD during the intervening time period of data gathering, analysis and validation. These four BINs had the appearance of being geographic false positives, but they were confirmed to be accurate by comparison to the records present in the analyzed data files. Major taxonomic conflicts were present in Blattodea (1/1; 100%), Coleoptera (12/50; 24%), Diptera (44/166; 26%), Hemiptera (15/38; 39%), Hymenoptera (31/96; 32%), Lepidoptera (97/391; 25%), Neuroptera (1/1; 100%), Orthoptera (3/6; 50%) and Trichoptera (4/7; 57%) (Suppl. material 2). There were far fewer minor taxonomic conflicts (Lepidoptera = 19, Hymenoptera = 3, Hemiptera = 1, Diptera = 4) (Suppl. material 2). Twenty-five BINs from the Araneae dataset were randomly selected for validation of the scripting process (Suppl. material 19). Validated BINs were examined in Araneidae (1), Clubionidae (1), Gnaphosidae (1), Linyphiidae (10), Lycosidae (4), Pholcidae (1), Tetragnathidae (1), Theridiidae (4) and Thomisidae (2) (Suppl. material 19). No validation BINs were discovered to be geographic false positives upon the initial pass. Major taxonomic conflicts were present in Araneidae (1/1; 100%), Clubionidae (1/1; 100%), Linyphiidae (3/10; 30%), Lycosidae (1/4; 25%) and Thomisidae (1/2; 50%) (Suppl. material 19).

Greater than 80% of recovered Insecta BINs were present on two continents or islands (Fig. 5, Suppl. material 8), totalling 254 unique combinations of continent/island points (Suppl. material 9). Europe–North America (4,207 points), Asia–Europe (2,324 points), North America–South America (2,248 points), Asia–Australia (827 points) and Africa–Asia (665 points) were the five most common geographic combinations captured within Insecta BINs (Suppl. material 9). In total, North America (12,612 connections), Europe (11,402 connections) and Asia (8,728 connections) were the most connected geographic categories (Fig. 6, Suppl. material 10). Europe–North America (5,563 connections), Europe–Asia (3,549 connections), North America–South America (2,718 connections), North America–Asia (1,520 connections), Africa–Asia (1,458 connections) and Australia–Asia (1,345 connections) were the most common connections in Insecta BINs (Fig. 6, Suppl. material 10).

Figure 5.

Histogram displaying the percentage of Insecta BINs with a given number of continent and island data points.

Figure 6.

Circos plot displaying the relative contribution of continents and oceanic island groups to geographic connections in the Insecta BOLD BIN dataset. The single data point for Antarctica was eliminated from this visualization. Numbers in parentheses indicate the total number of connections to that geographic grouping.

Our BOLD data-mining approach confirmed several well-characterized cases of parasitoid range expansion in Tr. basalis, Tr. japonicus, Tr. hyalinipennis, Te. remus, G. aetherium, Ps. striaticeps and Pl. demades. Therefore, the geographic distribution patterns found in the platygastroid dataset are generally considered credible even when lacking any directionality. The effort of COI barcoding topotypical specimens can be helpful for determining directionality, particularly if the specimens match primary types described a long time ago. Antiquity of specimens does not preclude the possibility that they were adventive at the time of collection, but it gives, at minimum, a historical perspective.

Exact COI barcode matches from geographically disparate populations might be the most conclusive evidence of a new adventive population within our framework. However, even among the validated cases mentioned above, only Te. remus, G. aetherium and Ps. striaticeps were scored as exact long-distance matches. Trissolcus japonicus and Tr. basalis both have the appearance of exact intercontinental matches, but were scored inconclusive due to sequence length variation. Trissolcus hyalinipennis and Pl. demades have no evidence for exact intercontinental matches in our dataset. Specimens identified as Pl. demades are divided into five BOLD BINS, two of which contain populations introduced into Canada and the USA (Mason et al. 2017). These Nearctic Pl. demades BINs were not captured by our method and show that the expectation of exact matches may be unnecessary when inferring range expansion, although it does lend additional confidence. The BIN calculations still captured relevant information in these cases even though additional geographic data points were present in neighboring BINs.

A majority of the intercontinental platygastroid BINs were unidentified below the family level at the beginning of this study, affording our group a “clean” taxonomic slate on which to analyze and interpret these results. Alternatively, the identifications for many of the validated intercontinental Insecta BINs were conflicted, which confounds interpretation of the data. Conflicting taxonomies present in BOLD BINs complicated our ability to extract and summarize higher-level taxonomic data from the Insecta dataset. Even if these taxonomic conflicts can be rectified or understood on a case-by-case basis by expert systematists, that database problem is likely to persist in future analyses. We advocate that experts use our baseline dataset to closely examine the intercontinental BINs in their group of interest and make informed judgements about their veracity.

Many of the BOLD BIN criticisms levelled by Meier et al. (2022) are applicable to our approach. We found that BOLD BIN recalculations affected our study, once again demonstrating their instability. A few of our validation BINs were missing when they were queried back to BOLD for checking, having been recalculated between the time of database mining and data analysis. This feature of BOLD BINs will certainly complicate repetition of our analyses in the future. For better or worse, we must rely on BINs calculated database-wide for our proposed method because they are tied to the specimen metadata in an efficient way for mining. To further verify our relatively small (nearly 10,000 COI barcodes) platygastroid dataset, the sequences could be aligned, phylogenetically analyzed and those results used to provide de novo mOTUs by varying methods with more public documentation than RESL (Ratnasingham and Hebert 2013). However, this smaller sampling of barcodes would influence mOTU calculations. The most rigorous results would be obtained by analyzing all available Platygastroidea data in BOLD, comprising well over 140,000 COI barcodes and growing. Such large analyses would firmly move our method into the realm of bioinformaticians and supercomputing, rather than being repeatable by scientists with less capability for big data. We consider it a major strength that our approach can be performed on an internet-connected, standard desktop computer with freely-available statistical software.

Our case study of Platygastrinae using GBOL specimens indicated that BINs overestimated “species richness” and we reiterate that the BINs in the Insecta and Platygastroidea dataset do not necessarily equate to species. Results for poorly-known or hyperdiverse insect groups must be interpreted with caution.

Platygastroid wasps are one of the most dominant flying insect groups worldwide, with a high rate of geographically structured community turnover and high taxonomic neglect, further complicating faunistic studies (Srivathsan et al. 2023). Srivathsan et al. (2023) discovered 44 platygastroid mOTUs in multiple sampling sites. They also reported that usually less than 3% (1–9% within a given insect family) of their mOTUs in the top 20 most dominant insect families were found at multiple sites. About five members of our platygastroid BIN dataset display geographic ranges which may be “false positives” based on our intentions. For example, BINs shared between the Republic of Georgia and Iran or Egypt and Saudi Arabia were scored as intercontinental even though these countries are in close geographic proximity. Conservatively, we detected 195 platygastroid BINs comprising specimens with vast geographic ranges. As of February 2023, BOLD analytics has calculated 11,468 Platygastroidea BINs. From our study, approximately 1.7% of available platygastroid BINs display this striking distribution pattern.

Determining whether these distributions are natural or adventive for most of the platygastroid BINs is difficult pending taxonomic revisions and follow-up research. Gilligan et al. (2020) suggested that a species association with Beringia supports hypothetically natural Holarctic distributions in Tortricidae (Lepidoptera), while also providing a good list of expectations for Holarctic species versus recent introductions. For truly Holarctic organisms, expectations include: 1) no direct evidence of introduction, 2) association with native hosts, 3) initial reports from inland areas, 4) lack of recent range expansion and 5) presence in the Arctic biogeographic region (Gilligan et al. 2020). These expectations are a useful framework for assessing many of the intercontinental platygastroid BINs present in Canada, Europe and northeast Asia.

Many Platygastroidea distributions in the BOLD dataset were from the Southern Hemisphere, between the Northern and Southern Tropics or island localities, precluding the need to consider a naturally Holarctic distribution as an explanation for the pattern. We think those cases are best considered introductions mediated by human activity. However, in some Platygaster and Telenomus BINs, the locality data imply enormous geographic ranges across the entirety of Canada, northern and central Europe, northeastern Asia and other spurious localities. For example, Telenomus sp. (BOLD:AAV1142; 461 public barcodes) has occurrences across Canada including Nunavut in the north, south to desert regions of eastern California and eastern Europe. One unidentified platygastrine (BOLD:ABW3192; 37 public barcodes with several exact matches) has occurrences in eastern and western Canada, Germany, the Russian Far East, South Africa and California. These geographic ranges encompass several climates and biomes, showing the apparent ability of the wasps to tolerate dramatically different environmental conditions for at least a short period, considering that species that fail to establish would have to persist long enough to be collected.

Historically, the Palearctic and Nearctic Regions have been regarded as separate by platygastroid taxonomists. Early European authors rarely made comparisons to the fauna of neighboring countries, let alone distant continents. Likewise, early American hymenopterists often treated the Nearctic fauna as unique. This approach contrasts with that of early American lepidopterists, who tended to misidentify Nearctic species as similar-looking European species (Gilligan et al. 2020). The discrepancy may be explained by the availability of detailed keys and illustrations for Lepidoptera in the 19^th century, while descriptions of Platygastroidea were generally short and vague.

Contrary to the assumptions of the past, the results of our study suggest that some platygastroid genera, such as Platygaster, include many naturally Holarctic species as well as human-mediated introductions. Landry et al. (2013) came to a similar conclusion for Lepidoptera, with the added benefit of host plant data to help distinguish natural from anthropogenic distributions. It is difficult to make such determinations for Platygastrinae, as little is known of their host repertoire and Cecidomyiidae is itself a dark taxon. However, our results also indicate a high number of intercontinental cecidomyiid BINs, which matches well with the parasitoid distributions.

Constraints due to a lack of data are a consistent theme throughout platygastroid taxonomy, especially when compared with better-studied groups of insects. Our method offers a path to gather and interpret the available data, albeit with limitations. For example, the Nearctic and western Palearctic were remarkably well sampled, allowing for more detailed examination of distribution patterns. On the other hand, the Pacific Islands and Caribbean yielded no Platygaster or Synopeas records. This likely reflects reality in the Pacific Islands but is a result of under sampling in the Caribbean, a distinction which cannot be made by our method alone. Ashmead (1900) recorded 29 species of platygastrine wasps from St. Vincent and Grenada, while the only platygastrine in Hawaii is considered an accidental introduction (Drake 1969).

Specimen images in BOLD allowed us to provide a list of genus level identifications for about one fifth of all platygastroid BOLD BINs. This has just begun the process of overcoming taxonomic impediments in the group and the database more broadly, as many thousands of BINs remain to be examined and identified. Platygastroid wasps are generally small insects (0.5 to 10 mm), making species level characters difficult to assess without proper microscopy and high-resolution images. Regardless, the image quality and habitus views in BOLD were generally sufficient to identify BINs to genus. We encourage systematists to examine BIN images in their group of expertise to see if this process can be repeated for other under-studied insect families. Given that we discovered several BINs which contained genera lacking any other DNA sequence data, is it likely that more such cases remain to be found.

Data presented here suggest that taxonomic revisions of platygastroids should be targeted at the global level as much as possible, even if there are feasibility concerns. A consideration of data from across the world will likely be necessary for the accurate and precise description of some sections of platygastrid and scelionid biodiversity. The BIN database should be preemptively searched to quickly quantify diversity present in an area and inform the taxonomic approach. Our genus level identification of BINS has already facilitated one such study. Melotto et al. (2023) described a new species, Synopeas maximum Awad & Talamas, which parasitizes an emerging pest of soybean. Given that soybean is a cosmopolitan crop, it was a possibility that this parasitoid was adventive. A phylogenetic analysis of Synopeas BINs enabled them to determine that the closest relatives of S. maximum are from North America, suggesting that it was not adventive. Furthermore, specimen images in BOLD allowed them to assess if certain morphological characters corresponded to a monophyletic species group.

On another research track, the molecular evolution of COI barcodes across Animalia was evaluated by Pentisaari et al. (2016), demonstrating that parasitic groups had many convergent amino acid variations. Several COI barcode amino acid phenotypes have been noted in Platygastroidea and some appear informative at the genus level (Talamas et al. 2021). Increased BOLD resolution for Platygastroidea genera could allow for new analyses of these patterns for the superfamily (Chen et al. 2021).

These analyses highlight the urgent need for more detailed and comprehensive approaches (Talamas et al. 2021) when identifying biological control agents, whether classical or augmentative. This is especially the case for platygastroid egg parasitoids; they are small, cryptic and their biodiversity dwarfs what has been adequately described. They also attack small, cryptic and poorly-described insect life stages. COI barcoding of proposed classical biological control agents must occur immediately at program outset. New data can then be quickly assessed as distinct or not present in BINs calculated from other non-native areas. Putatively “native” communities of parasitoids being evaluated for use as augmentative biological control agents should be similarly treated. Our results and those of Srivathsan et al. (2023) demonstrate the significant likelihood of unintended and undocumented platygastroid introductions.

Approximately 84% (12,876/15,391) of the recovered intercontinental and island Insecta BINs were detected across two geographic categories. When these geographic categories are contiguous or adjacent, they are the most likely to be ‘false positives’ based on our data-mining approach. For example, BINs present in lowland rainforest habitats of both Costa Rica and Colombia were scored as intercontinental even though these are probably natural distributions (North America/South America; 2,248 BINs). Indeed, there are many such cases in the dataset, especially among Neotropical Lepidoptera BINs. In the Palearctic, Europe and Asia (2,324 BINs) were connected at a similar magnitude. BINs detected across contiguous landmasses warrant additional scrutiny by taxonomic experts to determine their status as adventive or natural distributions. Just over 2,500 Insecta BINs were detected from three or more continents or island chains. This set of BINs contains insect species which appear to be truly successful global invaders. Our results were intuitive for insect BOLD BINs with the most widespread geographic data; these included species long associated with human activity. For example, Ctenocephalides felis (cat flea; BOLD:AAY6332), Aedes aegypti (yellow fever mosquito; BOLD:AEI9358), Culex quinquefasciatus (southern house mosquito; BOLD:AAA4751) and Aphis gossypii (cotton aphid; BOLD:AAA3070) were recovered in the analysis with 9 or 10 intercontinental and island data points.

The highly structured global sampling of flying insect communities undertaken by Srivathsan et al. (2023) provides some meaningful context for our results. In that study, greater than 97% of mOTUs from the most dominant flying insect families were sampled only at a single site, suggesting massive community turnover between sites (Srivathsan et al. 2023). Our BIN dataset, which notably was not designed for statistical testing a priori nor sorted by biological traits, suggests that about 3.15% of the BOLD Insecta BINs are distributed across continents and islands (489,156 Insecta BINs as of September 2023). Simpson et al. (2021) considered 3,747 insects and 58 spiders to be adventive in the continental United States. That dataset serves as a valuable baseline of verified introductions for comparison to the BOLD BINs provided herein. We found 12,612 intercontinental Insecta BINs with a distribution that included North America, indicating a large discrepancy between these methods. Our larger number may be an artifact of the data-mining approach, as mentioned previously, but may also reflect undetected introductions. Rectifying these datasets is a herculean task, given their size and the taxonomic expertise needed. However, given the magnitude of the agricultural, environmental and economic consequences of invasive species, this should be a research priority for biosecurity experts, bioinformaticians and systematists.

Biosecurity practices entail anticipatory risk assessment and preparedness, methods of surveillance, emergency management and policy enforcement (Barker and Francis 2021). This dataset has potential applications for risk assessment and surveillance. Our approach sought to use large geographic spans in BINs with the goal of identifying putatively adventive platygastroids. Small adjustments to the data-mining and analysis methods could be made to focus on administrative borders instead. BOLD data have already been used to inventory fauna at the country level (Geiger et al. 2016; Hebert et al. 2016), making new BIN additions to Canada and Germany straightforward to detect.

Emerging infectious disease surveillance programs have implemented several automated, internet-based data-mining approaches for intelligence gathering (Stevens 2021). Similarly, automation and scheduled repetition of a modified version of our analysis could provide information supporting risk assessment and invasion preparedness as BINs accumulate in new geographic areas. Detection at the earliest stages of invasion, followed by rapid response, provides the best economic and environmental outcomes. This proposed method of biosecurity surveillance will probably not be timely enough to meet early detection criteria; once new data points are detected in an area, the proverbial cat is already out of the bag. Agencies also rely on in-hand specimens for diagnosis of regulated species and are unlikely to take a decentralized BIN record as definitive evidence of new introductions. However, this may be a situation analogous to several cases of verified arthropod introductions that were first noticed or tracked as records in iNaturalist and BugGuide (Iwane 2018; Halbert et al. 2020; USDA 2021; Hayden et al. 2022; Chuang et al. 2023; iNaturalist 2023; VanDyk 2023).

Further scrutinizing the Insecta BINs with consideration for specimen spatiotemporal data might also inform pathway analyses for species moving via global trade, weather phenomena or animal migration events. Data from the Jacksonville area uncovered an intercontinental Telenomus species, with matching specimens collected in 2014 and 2015 from coastal areas of Bangladesh and California. Relevant agencies could have been alerted to the risk of this new introduction, possibly centered around international seaports-of-entry, eight years prior if the available data were scanned systematically. Thus, nearly a decade of potential research progress on this unidentified, globally-mobile Telenomus went unrealized. Cases like these might be valuable input for pathway models or agent-based models of long-distance insect dispersal. Detailed examination of the Insecta BIN dataset should reveal similar cases for groups with a widely-variable range of life history traits, dispersal potentials and introduction histories or modalities.