1. “The method ignores previously described species”

This is not true, not for the Meierotto et al. (2019a) paper and not for the present paper. In the Meierotto et al. (2019a) paper there was a clear statement that coauthor Sharkey had seen all of the relevant types and that in his opinion none of the species treated were conspecific with these except for Zelomorpha arizonensis. The critics suggest that evidence be given to demonstrate that Sharkey had seen the types. Are they asking for letters from curators of the museums that Sharkey visited, or did they simply overlook the statement? Impossible to tell, and way beyond standard taxonomic practice by well-established authorities for a major group of insects. Perhaps they ignored another paper by Meierotto et al. (2019b) in which the presently recognized species of Zelomorpha and Hemichoma were established by creating dozens of new combinations. Clearly, one of the authors (i.e., MJS) of this latter paper must have viewed the type species, as stated in the paper. In this taxonomically broader treatment also, we have reviewed all of the relevant literature and found a number of previously described species that are mentioned where appropriate.

2. “Mitochondrial trees often disagree with nuclear species trees, especially in taxa where Wolbachia may be altering mtDNA introgression” and “congruence between nuclear and mitochondrial signal should be tested to better reinforce the species units identified”

This mostly appears to be pointless because we are not trying to recapitulate phylogeny. BOLD automatically screens out sequence uploads for Wolbachia (and other contaminants). Indeed, when barcode sequence libraries were surveyed for Wolbachia, evidence was found in only 0.16% of cases (Smith et al. 2012a). More independent DNA sequence data is clearly more useful (e.g., Janzen et al. 2017) and there is little doubt that in the years and decades to come the barcode will constitute many markers, if not entire genomes. However, adding more genes is not cost or time effective at present, as the critics themselves point out. Furthermore, part of the first pass taxonomic approach for megadiverse areas/taxa using DNA barcodes includes the creation of a whole genome DNA extract that is available for future phylogenetic work. While Costa Rica has achieved biopolitically public domain status for its CO1 barcodes as presented here (Decree 41767. 2019), the whole genomes are not (yet) full open public access for Costa Rica and most other tropical countries.

3. “purely DNA-based descriptions will… make the identification of millions of historical specimens impossible”

As we have made clear in the Meierotto et al. (2019a) paper and here, we are not proposing that comprehensive revisions that include keys and morphological diagnoses be abandoned. Rather, we view barcode-based descriptions as a first pass in an iterative approach to solve the taxonomic impediment of megadiverse and under-taxonomically resourced groups that standard technical and biopolitical approaches have not been able to tackle. For example, if a taxonomist wishes to integrate these elements to Zelomorpha or any of the taxa that we review below they will have a great starting point (and resources). When a large number of specimens, from a wide geographic range, are barcoded, effective morphological keys may be written and old museum specimens will regain their value to go along with their barcodes (e.g., Janzen et al. 2017). Nonetheless, until this milestone is achieved, and even then, the old museum specimens have little value, because of the enormous cost in time and budgets to examine and identify them morphologically, e.g., imagine trying to identify in a morphological key one of the more than 3,000 species of Neotropical Triraphis 95% of which are currently undescribed.

4. “it will impair this science [taxonomy] in developing countries which house most of the undiscovered portion of biodiversity, due to high costs and lack of staff and technology”

This is a point that one of the reviewers of our present manuscript also mentioned. Firstly, high costs, lack of public interest, and lack of staff, to say nothing of lack of biopolitical enthusiasm or budgets, have caused the taxonomic impediment worldwide, not just in developing countries. In Canada, for example, a Canadian Council of Academics report from a decade ago found that previously deep and respected taxonomic resources suffered from decades of lack of investment that resulted in insufficient capacity to describe Canadian biodiversity (Lovejoy et al. 2010). Costa Rica is a relatively poor country that cannot maintain a cohort of taxonomists to describe its biodiversity and yet we here describe more than 400 new species from the country, with many more to come, all public domain and all available to future users at any level if they have barcode access, which they will. The critics may have meant to write that demanding barcodes for species descriptions, which we do, will disenfranchise those who cannot afford the price of barcoding. At the Canadian Centre for Biodiversity Genomics (CBG) the current rates for barcoding vary between 1.02 and 25 US dollars, depending on the quantity of specimens and their age, e.g., old museum specimens are more expensive. Today most taxonomists would be looking at $10 per specimen for a minimum of 95 specimens. That is almost $1,000 and certainly an impediment to many, although not out of the range of routine expenses like salaries, lab infrastructure, and page charges. DNA evidence is necessary to delimit species in megadiverse taxa, as we demonstrate further in the introduction, and the use of which Zamani et al. (2020) promote elsewhere in their article. Our first pass approach is not more expensive than the current standards in comprehensive revisions.

5. [it will] “drastically affect other related fields of study and, importantly, conservation”

The authors did not expand on this criticism, but the opposite is true. For example, barcoded species will allow for the identification of metabarcoded specimens. This is a tool that is being used to determine the rareness and distribution of a species and potentially allow for their protection. Our method allows for the rapid naming of species and it is very unlikely that unnamed species will ever be protected, and they will have no chance of being integrated into the socioeconomics of a tropical country, a process that is vital for the conservation of tropical biodiversity (Janzen and Hallwachs 2019b).

6. “there must be several photographs available, not a single lateral photograph of a single specimen”

Of course many photographs are desirable as are fully integrated revisions. However, if we are to rid ourselves of the taxonomic impediment, time is a critical factor. Our simple images are meant simply as a voucher for a comparison with newly barcoded specimens, as well as to offer many gestalt traits. For example, we have a number of species in our treatment of Costa Rican fauna below that appear to have conspecific specimens on BOLD that were collected by others in other countries, e.g., Belize and Argentina, i.e., they are in the same BIN. In most of these cases the details on BOLD are “private” (unfortunately owing to other taxonomists’ possessiveness) and so we cannot easily examine the specimens to include them in our revision. However, the proprietors of these specimens can look at our image and conclude either, “yes that looks a lot like my specimen and is probably the same species or not”, and all of our specimens are available for examination from the CBG or from their final deposit in the Canadian National Collection in Ottawa.

7. “some text highlighting important diagnostic features is valuable… [and] the actual description of the species may be written within minutes when material and expertise are already available”

This is false. If morphological descriptions and diagnostics could be written in minutes we would include them, and the taxonomic impediment would be just a pebble in the taxasphere’s road. Even the revision of Alabagrus by Sharkey et al. (2018), which included barcode data and morphology, could not separate morphologically a number of species that, based on host data and COI sequence data, were clearly different. For megadiverse, understudied taxa like those treated herein, a morphological diagnosis separating tens to hundreds of new species from a few previously morphologically described species is a waste of time. This is because the vast majority of species in these taxa are undescribed and distinguishing a new species from the few described species leaves dozens or hundreds or thousands of species that could fit any simple morphological diagnostic. One of the advantages of the barcode approach is that it acts as a diagnostic that is effective in distinguishing newly described species from almost the entirety of both described and undescribed species yet to be captured. It is impossible to guess at all of the different morphological combinations that may be possessed by the plethora of undescribed species for any particular genus, but the COI barcode is diverse enough across taxa that it is an effective diagnostic even for these unknown species, as has been shown by its reliability for already well-known taxa such as tropical Lepidoptera.

8. “Simply assigning all BINs taxonomic names.. would indeed complete the inventory of life on Earth extremely quickly (at precisely the same pace as the rate of barcoding) – that we do not dispute. But it would also remove the quantitative and qualitative difference between these preliminary identifiers (based on a single DNA marker) and full taxonomic recognition (based on a more comprehensive diagnosis, ideally supported by multiple lines of evidence including genetic data) that lend taxonomy its value”

When the authors and the taxasphere have the budgets, the enthusiasm, the time, and the biopolitical permissions to achieve this for the 20+ million tropical species of terrestrial Eukaryota, it will be wonderful. As mentioned earlier, our method is not meant to be the final treatment for a higher taxon. We admire and we practice the integrative approach, and all authors who are taxonomists in this publication regularly publish integrative revisions; this is our standard. But every integration starts with pieces to integrate. The barcode and its associated administrative accounting, along with the specimen itself, and whatever higher taxon information is available, is what we can offer to tropical conservation through sustainable biodevelopment (e.g., Janzen et al. 2020a). We are now offering an alternative method as a taxonomic first pass for megadiverse understudied taxa, the goal being to overcome the taxonomic impediment, thereby allowing the full power of tropical biodiversity value to become truly part of tropical societies. The critics also misrepresent our method. We do not simply assign taxonomic names to BINs. Firstly, ca. 10% of BINs contain multiple species. More rarely, BINs that are presently on BOLD today may coalesce, resulting in one of the BINS disappearing, just as happens when a species is found to bear several classical scientific names (incidentally, often revealed by barcoding). A careful examination of the content of the BINS and their nearest neighbors is necessary. Working with the specimens reared by Dan Janzen, Winnie Hallwachs et al. in the Costa Rican ACG mega-inventory (Janzen and Hallwachs 2016) has been enlightening to discern species boundaries, and to understand the limits of the BIN-only approach, because of the host-caterpillar data that they have. For example, host data, and morphological evidence allowed us to distinguish seven species of Macrocentrus within the BIN BOLD:ACK7466. This is a low frequency example; for 90% of the species, BINs make a good proxy to species, just as morphological comparisons often do. They are the equivalent of placing look-alikes in a unit tray, later to be sometimes split into a species complex by yet more data, precisely what happens in classical taxonomic practice following naming of museum specimens in a monograph. This in turn generates papers with titles such as “17 new species hiding in 10 long-named gaudy tropical moths (Lepidoptera: Erebidae: Arctiinae)” (Espinoza et al. 2017), each of the 17 having been lumped within the ten classical species in unit trays for decades.

9. “It would supplant taxonomists with technicians, who need to know nothing of the biology of the units with which they are dealing”

We wish this were true and imagine the power it would and perhaps will give the general public. Certainly, more people will play a larger part in a barcoding approach. If the criticism were true, we could overcome the taxonomic impediment much more quickly, to say nothing of exporting bioliteracy possibilities to the majority of the world (Janzen and Hallwachs 2019b). Interestingly, many of the co-authors of the present paper fit the technician category. Nonetheless, due to the necessity to identify specimens to genus as well as DNA contamination, multiple species in BINs, and other minor issues, an expert is necessary for the next level of resolution or confirmation.

1. “real revolutions are undoubtedly coming, especially from the fields of machine learning and integrative species delimitation (Solís-Lemus et al. 2015; Favret and Sieracki 2016), and also that it is possible to produce massive, rapidly assembled taxonomic monographs without compromising on quality (Rakotoarison et al. 2017)”

The ‘good news’ reported by the critics does nothing, and will do nothing, to attack the core of the taxonomic impediment. The visual identification system introduced by Favret and Sieracki (2016) is dependent on preexisting taxonomy for machine training. That is, after a species is described, numerous specimens are investigated, and a machine is trained to recognize the species. BOLD does precisely this by its barcode. The proposed revolution is a fine technology, but unrelated to the taxonomic impediment. Integrative molecular species delimitation proposed by Solis-Lemus et al. (2015) requires far more loci, sequencing, and money than does DNA barcoding and therefore suffers from most of the same criticisms the critics aim at barcoding. The comment that “massive”, rapidly assembled taxonomic monographs (Rakotoarison et al. 2017) can be assembled “without compromising on quality” refers to a revision treating fewer than 50 species, a drop in the bucket compared to our groups. It also is constructed on profuse sequencing, which would be problematic (according to the critics) for labs in developing countries, to say nothing of the general user community.

2. “a true paradigm shift in taxonomy will come only when there is a revolution in the level of financial investment in taxonomy and the natural history museums that house the described and undescribed reference material of life on Earth”

The taxasphere has been hoping for this for many decades and it will likely never happen until we show the funding agencies that we can make a dent in the taxonomic impediment without costing billions of dollars, to say nothing of showing the overall public that there is reason to bother. To our knowledge, in the USA (and we presume in many other parts of the world) there are no taxonomists that are federally funded for alpha taxonomic research unless they are also addressing interesting scientific, commercial, or public questions. Trivial funding is available for higher level phylogenetic research, although on a very limited scale. We believe that NSF (USA) and other agencies may fund large scale revisions if the revisor, employing a barcode approach, is treating thousands of species and attacking the taxonomic impediment. Unfortunately, current economic crises are causing many of the countries with appropriate resources to be more interested in funding research that will largely benefit themselves rather than the tropical countries so rich in rapidly disappearing biodiversity.

We continue now by justifying our diagnostic approach. COI barcodes are indispensable for diagnosing the species treated herein, as well as signaling where seemingly trivial morphological variation actually indicates a species boundary, and they are an invaluable tool for studies of most other hyperdiverse arthropod taxa as well. For our taxa, COI barcodes, though not infallible, are a magnitude more effective than morphology, and the evidence presented below demonstrates this. Dozens of additional examples can be found in the following studies: Hebert et al. (2004); Janzen et al. (2005, 2009, 2011, 2012, 2017, 2019b, 2020b); Hajibabaei et al. (2006); Smith et al. (2006–2008, 2012b); Burns et al. (2007, 2008); Chacon et al. (2013); Bertrand et al. (2014); Brown et al. (2014); Fernandez-Triana et al. (2014, 2020); Fleming (2014); Hansson et al. (2015); Phillips-Rodriguez et al. (2014); and Espinoza et al. (2017). Using a morphology-only approach, Leathers and Sharkey (2003) revised the Costa Rican species of Alabagrus (Braconidae) with an emphasis on specimens from the La Selva Biological Field Station. Later, using a morphology and COI barcode approach, Sharkey et al. (2018) re-revised Costa Rican Alabagrus with an emphasis on specimens reared in the Área de Conservación Guanacaste, only 100‒150 km distant from La Selva, and containing all of the La Selva ecosystems. When the 39 species treated by Leathers and Sharkey were barcoded, 17 were found to have been incorrectly lumped with previously described species that do not even occur in Costa Rica, and five endemic morphology-based Costa Rican species were discovered to be species complexes. The morphology-only approach resulted in an error rate of more than 60%. The following examples illustrate the sources of the errors. The Leathers and Sharkey concept of Alabagrus englishi was found to comprise five distinct species when barcoded. To quote directly from Sharkey et al. (2018). “Many paratypes of A. englishi, sensu Leathers and Sharkey (2003), will now key to the following species: A. johnobryckii, A. sarameierottoae, A. sarahsharkeyae, and A. barbsharanowskiae.” A second example is A. pachamama, which was recorded by Leathers and Sharkey as occurring in Costa Rica. As the name implies, the holotype is from Peru. This “species” turned out to be a complex of five species in Costa Rica, i.e., A. jackiemillerae, A. scottshawi, A. fernandezi, A. fernandodiasi, and A. genemonroei. Their COI barcode sequences are significantly divergent, and their host caterpillars are conspicuously different species. These facts led the authors to detect subtle morphological differences that distinguish the species, and further, to conclude that A. pachamama does not even occur in Costa Rica. These findings suggest that this complexity is common in neotropical ecosystems.

To understand how morphologically similar the species of the aforementioned complex are, compare the images of A. scottshawi and A. genemonroei (Figs 1, 2). A casual look reveals that the propodeal sculpture is different, the basal black bands in the forewings are in different positions, and the hind coxae have different colors. Without molecular data, how can one decide if these features reflect intraspecific variation or are evidence of distinct species, especially when there are only a few specimens (and sometimes only one) in hand? To add to this morphological ambiguity, there are three more congeneric species with similar, but slightly different combinations of these character states. It was only after molecular and host data signaled different species that morphological characters were confirmed to distinguish the species of this complex. Even after incorporating COI sequence data, not all species of Alabagrus treated in Sharkey et al. (2018) could be separated based on morphological characters; for example, A. jennyphillipsae, A. isidrochaconi, and A. jeanfrancoislandryi are inseparable morphologically (they end in the same terminus in the morphological key) but clearly distinguished by COI data.

In the Alabagrus revision (Sharkey et al. 2018) and in the species treatments that follow here, a 2% genetic distance, the conventional threshold for species delimitation using COI barcodes (Jones et al. 2011), was used to cluster putative species (Smith et al. 2012b). Morphology and host data were then employed to test and refine these approximations to species. The 2% divergence results in putative species being coded with Barcode Index Numbers (BINs) (Ratnasingham and Hebert 2013) produced by BOLD (http://www.barcodinglife.org/index.php/Public_BarcodeIndexNumber_Home). The BINs were used to sort specimens just as morphology is traditionally employed to sort specimens into museum unit trays. Morphological data, ecological data, and clustering of barcodes within a BIN were then considered to make final decisions when genetic distances between putative molecular species were near or below the 2% threshold, and commonly, two or more clusters within a BIN were each represented by numerous specimens. In one of the first published examples of “hidden diversity,” that of Astraptes fulgerator (Hebert et al. 2004), six species easily distinguished by caterpillar coloration, food plant, and ecology, fell in the same BIN (Janzen et al. 2011) because they were each less than 2% different in their barcodes, yet distinctly differently clustered within the BIN. In the case of Alabagrus, COI barcodes were 100% consistent with morphological and host data. By this, we mean that the COI data never lumped specimens that were markedly different morphologically in the same BIN; nor did it lump specimens with divergent host data. Furthermore, “Binning” did not split species that were clearly demarcated by host and morphological data.

COI data are not always diagnostic. In the revision of Lytopylus (Braconidae, Agathidinae) (Kang et al. 2017), two of the 32 reared species from Costa Rica did not achieve a 2% COI divergence, yet they had distinct morphologies and host data. The error rate, if COI had been used exclusively for the Alabagrus and Lytopylus revisions, would be ~ 1%, far better than the grossly underestimated 15% error rate assumed for predominantly morphological revisions of braconids as demonstrated in the database of Yu et al. (2016). Clearly, relying solely on COI barcode sequences will result in some overlooked cryptic species, and we do not promote COI-only species delimitation except where mass samples are being species-counted and therefore small amounts of species lumping is irrelevant (e.g., Janzen et al. 2000b), e.g., it may not matter if a year of Malaise trapping records 8,500 species or 8,516 species if it is being compared with a trap-year recording 3,456‒3,602 species. In the treatment of Macrocentrus that follows, seven species fall into the same BIN (BOLD:ACK7466). However, because they are so dissimilar morphologically and separated on the NJ tree, we treat them as distinct species. In this case, because the BIN data are not diagnostic, and will not be until the sample size is large enough to show clustering within the BIN, we provide a morphological key to the seven species. Janzen and Hallwachs’ rearing studies have revealed numerous BINs with multiple species, and they estimate a 10% rate of lumping using BOLD’s 2% genetic distance threshold. For example, in a reared sample of ca. 17,139 reared and Malaise-trapped barcoded specimens of Microgastrinae (Braconidae) from Área de Conservación de Guanacaste (ACG), there are 1,231 BINs, but more detailed examination of morphology and ecology reveals that there are at least 1,346 species in that sample (unpublished). Given the same rate for other major taxa in the inventory (Tachinidae, Ichneumonidae, Hesperiidae), for large samples that have been only BINned (as in a Malaise trap sample), we tentatively add 10% more species to large BIN numbers for a more accurate estimation of the number of species that will be defined with more intense scrutiny by specialists, coupled with more detailed ecological information.

As a final illustration of how ineffective morphology is to differentiate braconid species, we will paraphrase a report published by Kang et al. (2017). Ilgoo Kang (at that time a new graduate student) and Michael Sharkey (with decades of experience in species-level morphological alpha taxonomy) tested their ability to differentiate the species of Lytopylus before including COI and host data. They had error rates of 62% and 54%, respectively. All possible types of errors were discovered, i.e., lumping, splitting, and both lumping and splitting (mixing some of the members of two or more species into two or more incorrect species). In contrast, the molecular species concepts matched with the final species delimitations, which were also based on host data, at 96.6%. In light of the evidence presented above, and much more like it in the literature, it seems absurd to us that many in the taxasphere complain about barcode-based species concepts, yet readily accept delineations based solely on morphology. This is not to say that COI-based species discrimination will work for all taxa, but the method undoubtedly is highly superior to morphology for the Braconidae treated here and all indications are that they will be the same for most other species-rich tropical taxa (e.g., Microgastrinae Braconidae, Ichneumonidae, Tachinidae).

Hopefully, we have demonstrated the utility of COI barcode data for species circumscription. Now we address the hurdle of species descriptions. Critics claim that the description of a new taxon must include more than an image and COI sequence data, insisting that keys, morphological diagnoses, detailed figures, and the examination of the holotypes of all congeneric species known from the same realm are necessary. Our answer to these criticisms is that there are too many species and too little time. For many species-rich higher taxa, detailed place-based neotropical inventories such as that of ACG (e.g., Janzen et al. 2009, Janzen and Hallwachs 2016) deal with only a fraction of the total local diversity. In these cases, is seems inane to present a morphological key where it is obvious that no key or array of diagnoses will discriminate the described from the hundreds of species yet to be described (e.g., Fernandez-Triana et al. 2014, 2020; Fleming et al. 2014; Hansson et al. 2015). Due to humankind’s assault on the planet, there is an urgency to recognize the multitude of species that are yet to be described, quite simply because by not knowing what they are, where they are, and what they do, humanity has no compulsion to allow them and their wild ecosystems to co-exist (Janzen and Hallwachs 2019b). For years, this urgency of dealing with the taxonomic impediment has been recognized. The predominant solution proposed has been “give us more money and more taxonomists”, a laudable strategy but one doomed to fail as the last three or four decades have demonstrated.

In the following paragraphs we will demonstrate the magnitude of the task of describing the world fauna of one hymenopteran superfamily and demonstrate that the present approach to taxonomy is woefully inadequate regardless of the number of taxonomists employed. The Ichneumonoidea contains the two most species-rich families of Hymenoptera, i.e., Braconidae and Ichneumonidae. As of 2016, the superfamily comprised ca. 44,000 valid species (Yu et al. 2016). The true number of species can only be crudely estimated, but it is conceivable that there are 1,000,000 or more. Estimates of total species richness for this group are quite variable and have increased greatly over recent decades. Rodriguez et al. (2013) used the ratio of described wasp species to Lepidoptera hosts from relatively well studied sites to estimate the total number of species in the braconid subfamily Microgastrinae, which currently has ca. 3,000 described species (Fernandez-Triana et al. 2020). They estimated between 17,000 and 46,000 species of Microgastrinae, but they noted this is likely an under-estimate due to the many undescribed species of Microgastrinae from well-studied sites that were used to make the extrapolations. Extrapolating from Yu et al. (2016), five of every one hundred described ichneumonoid species are microgastrines (Yu et al. 2016). Assuming that this ratio holds true for undescribed species, and that the estimates made by Rodriguez et al. (2013) are sound, there are between 300,000 and at least 900,000 species of Ichneumonoidea.

Extrapolating again from Yu et al. (2016), from 2000 to 2011, an average of 468 species of ichneumonoids were described per year (Fig. 3). Given the current rate of species description and these estimates of species diversity, all ichneumonoids, that manage to remain extant, would be described somewhere between the years 2561 and 3843, if they were ever actually captured and curated.

Given the number of the species treatments needed to tackle even a small portion of this diversity, classical taxonomic monographic treatments are not the answer. Most of the criticisms offered against our approach stifle or skate over the fundamental purpose of the method. More images, checking types of dubious likelihood of being conspecific with any of the species treated, and producing morphological keys and diagnoses all take time, too much time, and far greater financial resources are required. Worse, they can only be used by the ever-dwindling number of people that society will support to be “experts” in the field, and even when such experts do exist, they and their literature are not available to the global society at large that needs to have reason to know and understand. But there are other more practical reasons not to include some of these elements which we will get to shortly.

With a COI barcode approach, it will be relatively easy to check the identity of a new specimen of braconid from Costa Rica, and in these new stages of method evolution, we are strategically fortunate that > 95% are undescribed and will remain so unless treated differently than in a traditional manner. To identify any specimen, one simply needs to obtain the COI barcode from the specimen and check for a match in BOLD with a barcode or BIN, then compare the images and other information in the BOLD database to confirm or refute the identification. If there is no match, that specimen is then a starting point for further identification or description, and it has its collateral and barcode data already databased with its specimen voucher code. This may not yet be an effective approach today for most braconid taxa from Costa Rica, but it will be effective for all of the genera that are given a preliminary treatment, such as those presented in the chapters that follow. If a Costa Rican specimen is absent from BOLD after the publication of this and further preliminary revisions, it is straightforward for anyone to describe the species and make it available to all through the web. Processing at the Centre for Biodiversity Genomics (CBG) and the output in BOLD allows its placement in a taxonomic hierarchy; e.g., family, subfamily, and usually genus, the latter almost 95% accurate in braconids. Such description can simply follow an example from any of the following chapters. This is why some refer to COI barcoding as the democratization of taxonomy, and perhaps why it frightens some of our more conservative taxonomic practitioners, and especially those who normally work with just a few (today) important species far from the hyperdiverse tropics.

Compare this to the standard approach of identifying a braconid from Costa Rica. One would start by identifying the specimen to subfamily and genus, impossible tasks for most people, even with good keys, and a difficult task for specialists. In the case of Costa Rican braconids, this is also a rather tedious process for a specialist due to the number of new genera being encountered and generated annually. To make this example a little more understandable to the reader, we will not make a reference to the publications required to do this, rather we will let the reader verify the difficulty for themselves by trying to find the references themselves to attempt the subfamilial and generic identification of any braconid specimen from any Neotropical country. Once a generic name is in hand, the identifier will need to proceed to the species level with the knowledge that fewer than 10% of neotropical species are described. Typically, this means obtaining numerous publications, many of which are protected by copyright and expensive to purchase, even if available on the web, and then comparing the specimen in hand with diverse keys and descriptions, in assorted languages (e.g., German for Costa Rican Opiinae). Furthermore, many of the abundant English texts are based on a vocabulary that often does not have unambiguous translations to Spanish or Portuguese. When a tentative identification is made, an examination of the type specimen will almost certainly be needed for confirmation. Most readers understand the impossible barrier that this presents, especially for non-taxonomists from developing countries. Even if a morphological match with a holotype is obtained, the likelihood of cryptic species will make this morphological comparison far from certain.

Morphological keys for masses of hyperdiverse, poorly known species are all but useless, and their exclusion from the following chapters is not an oversight. This is why. For the genera treated herein, we estimate that we have barcoded at most 10–20% of the Costa Rican fauna. Most of the specimens come from caterpillar rearings in Área de Conservación Guanacaste (the size of the footprint of New York City and its suburbs, www.acguanacaste.ac.cr), and for practical reasons, fruit-feeding, stem-boring, and leaf-mining caterpillars have not yet been reared, nor those from canopy-restricted caterpillars. Added to this are specimens from approximately 16 Malaise trap-years of samples from six localities scattered across dry forest, rain forest, and cloud forest (e.g., Janzen and Hallwachs 2020b), a very small number given 120,000 terrestrial hectares of all ages of succession in ACG. In other words, many more species of the genera treated below are yet to be discovered/collected, and the probability of a specimen in hand during further sampling being in the key or other keys (e.g., Apanteles by Fernandez-Triana et al. 2014) would be considerably less than 50%. It is our intention to continue with a treatment of all species of Costa Rican braconids as new Malaise trap samples and more reared specimens become available through BioAlfa (Janzen and Hallwachs 2019b), which is an effort to DNA barcode the entire national Eukaryota. If keys were included in these preliminary revisions, addenda would need to be added to the keys on an ongoing basis, creating an even more impossible taxonomic morass for anyone other than a fanatically dedicated taxonomist specialized in Costa Rican Braconidae. There may eventually be some utility in morphological keys for these taxa, but only after the survey nears completion. Even then, the audience will be extremely limited because only Costa Rican species will be surveyed, and all genera are far more widespread, many of these with thousands of undescribed species worldwide, e.g., Chelonus, Triraphis, and Aleiodes. In any case, the effort of making a key at any time is of dubious value because the COI barcode alternative is so much more efficient and accurate, and the arcane vocabulary of entomological keys are impossible for non-specialists to interpret. The above arguments apply equally to morphological diagnoses. Perhaps one day there will be a large ecotourist economy attracting braconid aficionados to Costa Rica, certainly then a morphological key would be warranted, but by then each person will likely have their own personal pocket barcode recorder and can obtain their own identifications through the internet, with the barcode of the specimen in hand as the first couplet in the “key”.

We have often heard the criticism that a barcoding approach to species circumscription will result in duel taxonomic systems, one morphological and the other DNA-based. There is some truth to this statement if an effort is not made to blend the old with the new, and our suggestion is to discontinue revisions of complex faunas based only on morphology, as well as stop imagining that a visual comparison with ancient holotypes allows accurate decisions as to which of multiple species in a cryptic complex actually matches the holotype. Furthermore, once the species have been tagged with interim names, be they codes or human-readable scientific names, they can be studied and grouped morphologically at will.

Our arguments follow. In the generic treatments below every effort has been made to search the literature for species described by specialists who have spent a lifetime doing braconid taxonomy. In some cases, this has resulted in the discovery of published names for our barcoded specimens, e.g., Adelius gauldi and Adelius janzeni. A recent publication by Shimbori et al. (2019) treated the New World species of Adeliini and included color photos, facilitating identification. Furthermore, the two aforementioned species were both captured in the same areas of Guanacaste Province that are the source of many of our specimens, and where many years ago Janzen used the same kind of Malaise trap to capture them under the guidance of Ian Gauld. The same quality photos are in a revision of Leptodrepana by Dadelahi et al. (2018) in which they described 24 species from Costa Rica. Of these, only one species convincingly matched one of our barcoded species, and it too was collected in an area heavily sampled by Janzen and Hallwachs and their team. Papp (2016) described 25 new species of the several thousand undescribed species of Chelonus (as Microchelonus) from Costa Rica. From his descriptions, and the lack of color photos, it is very difficult to tell if any match our barcoded specimens. A number of his holotypes are from Guanacaste and Alajuela, where the bulk of our specimens were collected, but none are from the Área de Conservación de Guanacaste and it is doubtful that any match those treated here. This lack of matching is all the more probable when you consider that only one of the 24 species of Leptodrepana matched. When the Papp holotypes are barcoded, as they eventually will be, some synonyms may be revealed, and our names synonymized. This is a better situation than guessing that one of our species is the same as one of Papp’s. Information is not lost in the long run if species are split. However, if species are lumped, the literature becomes confounded with false data, in this case, host and distributional data being attributed to one species when there are actually two or more species. The Costa Rican braconid fauna is in an ideal position for our approach since few morphology-based revisions have been conducted on it. For areas such as Costa Rica, the solution to the perceived problem of duel taxonomic systems lies in barcoding the holotypes, not stasis until the myriad of new species has been treated at the traditional pace, if ever. These comments apply to any other neotropical country or large region as well.

As a final comment, the strategy employed herein may not be optimal for all taxa and in all regions of the globe. For example, even in Costa Rica, there are taxa for which there are already many described species (e.g., some families of Lepidoptera), and in these groups our approach may result in the description of an unacceptable number of synonyms, so caution has been employed. For example, the tortricid fauna of Costa Rica includes ca. 250 described species, the vast majority of which have not been barcoded. In this group, describing the 300‒400 BINs that currently have not been identified, would almost certainly result in an unacceptable number of synonyms. Also, because the genitalia of Lepidoptera harbor many of the most compelling morphological features for species discrimination and circumscription, considerably greater effort would be required to examine the morphology in these families than in Braconidae. On the other hand, the comparison of barcodes at $10/specimen is vastly cheaper in specialist’s time than is genitalic study of thousands of specimens and every effort should be made to barcode holotypes.

If there is any hope of overcoming the taxonomic impediment it will be achieved with the approach outlined here or something similar to it. The shortcomings of the method are obvious in that it only provides a first pass at the comprehensiveness that we all desire. First passes, such as those presented in the following chapters, will provide the foundations that can be augmented by subsequent generations of revisors as time, money, and determination allow.

Approximately half of the species newly described here, and most of the specimens, were collected by rearing wild-caught host caterpillars in ACG in northwestern Costa Rica (Janzen and Hallwachs 2016). Caterpillars were collected by a team of Costa Rican parataxonomists (Janzen and Hallwachs 2011, 2016) as part of the ongoing project to document all ACG non-leaf-mining Lepidoptera larvae, their food plants, and their parasitoids. These caterpillars were databased with collection information, food plant information, and often a photograph, and they were reared to adults. When an adult moth, butterfly, or parasitoid emerged, the specimen was preserved oven-dried or frozen in 95% ethanol, and a leg was sent to the CBG at the University of Guelph, Canada (http://biodiversitygenomics.net; http://ibol.org) for DNA barcoding at a cost ranging from $3–$10 USD per specimen. Analyzed barcodes are deposited with their collateral specimen voucher data in the BOLD public database (www.boldsystems.org). The barcodes and specimens work their way towards whoever is eager to do morphology-based taxonomy, with the associated vouchers’ collateral data. Although the DNA barcodes are public domain (Costa Rican government decree #41767, Janzen and Hallwachs 2020b), as in a public dictionary, the genome associated with that barcoded specimen is currently only available through a permit-based contractual relationship with the government of Costa Rica, just as in any biodiversity prospecting relationship, be it for academic or commercial purposes.

Four sites in ACG are the subject of most of the Malaise-trapped specimens described here. Bosque San Emilio (BSE) is a 100-year old, 10‒20 m tall, secondary successional Pacific dry forest at ca. 300 m elevation in Sector Santa Rosa. Estación San Gerardo (ESG) is 30‒80-year old mid-elevation Caribbean rain forest at ca. 600 m elevation. Derrumbe (Cloud Forest) is a somewhat fragmented old-growth cloud forest at ca. 1400 m elevation near the top of 1500 m Volcán Cacao, a member of Cordillera Guanacaste that separates BSE from ESG. These three sites had one trap apiece that were operated continually, emptied once a week, for a year, and positioned beneath a broken portion of the canopy rather than in full sun or full heavy shade. The fourth site is a complex of seven Malaise traps in an area of ca. 5 km² of forest with a 1.5 km² bare-earth-rock geothermal drilling platform (trap numbers PL12-1-7) in the middle of the area; this forest is mostly composed of old growth at ca. 800 m elevation. The year of sampling at this site ranged from traps nearly insolated at the platform edge to 150 m distant into the shady forest understory. This forest lies at the intersection of ACG dry forest and rainforest at 800 m but slightly below the lower margin of cloud forest that starts at ca. 1000 m. All wild-caught caterpillars from which braconids were reared were found in some variant of the four ecosystems (dry, rain, cloud and interface) and within 20 km of one or more of the Malaise traps. In their first year (sometime between 2011 and 2014, these traps collected ca. 28,000 species of arthropods (excluding spiders, collembola and mites) as based on their BINs (Janzen et al. 2020).

Holotypes of all newly described species are deposited in the insect collection of the Canadian National Collection of Insects, Ottawa (CNC). Paratypes and all other specimens are deposited currently for the most part in the Biodiversity Institute of Ontario, at the University of Guelph, 579 Gordon St., Guelph, Ontario, N1G 1Y2, Canada. Others are divided between the CNC and the Hymenoptera Institute Collection (HIC), 116 Franklin Ave., Redlands, California, 92373, USA. Specimens deposited in the HIC will eventually be transferred to the CNC, and eventually representative specimens will be donated to the Museo Nacional de Costa Rica, San Jose, Costa Rica.

Identification of specimens to the subfamily level can be achieved using the key by Sharkey (1997a). Exceptions are the Hormiinae and Rhysipolinae, which were included in a large polyphyletic concept of Hormiinae s. l. in that publication. Also, the Ichneutinae and Proteropinae were included under Ichneutinae s. l. in Sharkey (1997a), and here they are separated following Chen and Achterberg (2019). For this reason, a diagnosis is given for these subfamilies in their respective chapters.

Morphological terms largely follow Sharkey and Wharton (1997), and definitions in English may be found at the Hymenoptera Anatomy Ontology Portal (http://portal.hymao.org/projects/32/public/ontology/).

Some host species are still awaiting full identification and are given interim names. For example, Antaeotricha Janzen233 is identified to the genus Antaeotricha by classical morphology-based criteria and to Janzen233 by barcode and ecological information. However, no formal scientific species name is available until a barcode-match is obtained with an existing holotype or until it is described as new, which may be decades. Equally, Antaeotricha radicalisEPR03 is also an interim name based on what the species LOOKS like, but is not, by its barcode, and note that the species epithet is not italicized and contains a number: it is not a scientific name, but temporarily retains the information that this species was called simply its look-alike A. radicalis before barcoding and associating it with other ecological data. Also, a name such as gelJanzen01 Janzen407 signifies a caterpillar in the family Gelechiidae for which even a generic name is not obtainable at present. In the future, the barcode, the temporary name, name, and BIN will remain searchable in the Janzen and Hallwachs database as well as in BOLD, GenBank, and any other public sequence repository. Continually updated copies of the Janzen and Hallwachs master database are deposited in the National Museum of Natural History, Washington, DC (USNM) and iCLOUD through Amazon, as well as in the Janzen lab and the BioAlfa project in parallel with the Museo Nacional de Costa Rica.

Focus-stacked images of specimens were taken using a JVC digital camera mounted on a Leica microscope and compiled with the program Automontage. Non-distortional image post-photography was done in Adobe Photoshop.

Molecular work was carried out at the CBG using their standard protocols. A leg of each specimen was destructively sampled for DNA extraction using a glass fiber protocol (Ivanova et al. 2006). Extracted DNA was amplified for a 658-bp region near the 5’ terminus of the cytochrome c oxidase subunit I (COI) gene using standard insect primers LepF1 (5’-ATTCAACCAATCATAAAGATATTGG-3’) and LepR1 (5’-TAAACTTCTGGATGTCCAAAAAATCA-3’) (Ivanova and Grainger 2007). If initial amplification failed, additional amplifications were conducted following the established protocols using internal primer pairs, LepF1-C113R (130 bp) or LepF1-C_ANTMR1D (307 bp) and MLepF1-LepR1 (407 bp) to generate shorter overlapping sequences. Amplified products were sequenced using Sanger technology though the most recent were sequenced by SEQUEL II. Specimens that “failed” barcoding are not included here unless otherwise indicated. When included, they are usually identified by unambiguous morphological and ecological information equally possessed by others from ACG in that species.

Voucher codes are presented for all holotype specimens (and other barcoded individuals) treated herein; and all host caterpillars are individually vouchered to their individual records (yy-SRNP-xxxxx). Codes beginning with DHPARxxxxx are for the parasite (or hyperparasite) specimens reared from the caterpillar. The SRNP voucher codes are from the Janzen and Hallwachs’ database (http://janzen.sas.upenn.edu/caterpillars/database.lasso). Specimen voucher codes beginning with BIOUG are from the BOLD database (http://www.boldsystems.org), and are specimens obtained from ACG Malaise traps. The abundant collateral information obtainable from these two databases complements the species treatments in important ways. A brief introduction to what to look for and how the databases supplement the species treatments follows.

As mentioned above, the braconid specimens reared in the ACG have a DHJPAR code and their host caterpillar has a SRNP code. These codes can be entered on the home page of the Janzen and Hallwachs database (Fig. 4). As an example, the information associated with holotype of Chelonus michellevanderbankae, DHJPAR0029134, appears as in Fig. 5, which is cropped to conserve space. The information includes details on the collection locality, host caterpillar and food plant information, dates of caterpillar collection and wasp eclosion, number of specimens emerging from the host (in the case of gregarious species), and notes on any interesting observations. In the case of a gregarious species, each wasp selected from the clutch to be DNA barcoded was given its own unique DHJPARxxxxx code, while its sibs remain tagged with the host yy-SRNP-xxxxxx code on their pin or in their alcohol preservative, a tube that is later deposited in the CNC. The written data can be in Spanish or English, or both. All dates are mm/dd/yyyy. Photos of the host caterpillar are obtained by clicking on the caterpillar image in the upper left of the image in Fig. 5. One of the five images of this parasitized caterpillar is shown in Fig. 6. Finally, if one wishes to see images of the adult of the host caterpillar, these can be found by returning to the home page, clicking on “Adult photographs” and typing the name of the host. In this example, some of the images generated are in Fig. 7.

The BOLD website has a wealth of information related to the species that are treated in the chapters below. For example, selecting the link associated with the alphanumeric voucher code of the holotype of Aleiodes adrianaradulovae, BIOUG28804-F01, (http://boldsystems.org/index.php/Public_RecordView?processid=JICFP038-16) brings one to the specimen’s page on BOLD. Some of the contents of that page are shown in Fig. 8, including an image of the specimen (Fig. 8A). These images, taken by the Centre for Biodiversity Genomics Photography Group, are produced under significantly different lighting conditions than those in the following chapters and compliment them by showing wing venation more clearly and color more dramatically, even if not as realistically. For example, the image in Fig. 8D is our image of the same specimen as that of Fig. 8A which is from BOLD. Also included in the specimen’s home page is a locality map (Fig. 8B) and a link to the BIN page (Fig. 8C) to which the specimen belongs. Other information included in the specimen page, but not shown in Fig. 8 are locality data, collection data, and the specimen deposition locality. If the link to the specimen’s BIN (indicated with red arrow in Fig. 8) is selected, the BIN page is loaded, the contents of which are shown in Figs 9, 10. This information will increase in kind and bulk as the ACG local and BioAlfa national inventories grow over time; equally, name updates will be added as further ecological information and taxonomic revisions progress. In other words, these are dynamic databases. Over time they will link outward to other databases, such as GenBank and GBIF. While data for a particular specimen will remain tied to that specimen, these databases will contribute to species-based aggregators that will pool this information with that of other specimens from other places.

On the BIN page (Fig. 9) there is an image of a representative specimen from the BIN, and information on the COI barcode data of all members. The “BIN Details” section outlines a basic summary of records included in the BIN, including the number of members, the average genetic distance among them, the maximum genetic distance, and the distance to the closest member of the closest outgroup (Nearest Neighbor) which is normally another BIN. In the example in Fig. 9, BIN “BOLD:AAG1413” has the abnormally large maximum distance of 3.82% among its members and the Nearest Neighbor is almost 11% divergent. The BIN Details section also provides a “BIN DOI” which is a persistent URL that can be used in publications to access the BIN. If the BIN has not yet been assigned a DOI, a button is provided allowing a user to request a DOI.

The “Nearest Neighbor Details” section provides the details on the closest BIN, including its member count and the mean and maximum genetic distances among the members. It also includes the “Process ID” used in the CBG (not to be confused with the specimen voucher code) for the closest member of that BIN as well as the taxonomy for that BIN. In the ‘Specimen Images” section, there are four members of the BIN that have images. Clicking on the thumbnail images will bring them into the large image field above them.

Also included in the BIN page (Fig. 10) is a distribution map of all specimens included in the BIN, a haplotype network, a neighbor joining tree that includes the nearest neighbor (outgroup) to the BIN, and a graph of the pairwise distances between the specimens in the BIN. Considering the six specimens from Costa Rica in that BIN, there are no doubts about species integrity; however, adding in the specimens from other countries leads immediately to the strong suspicion that there are one or two more species in that BIN. We have not been able to examine those specimens. Finally, one can download a PDF of the same tree by selecting the link indicted with the red arrow in Fig. 9. The NJ tree generated is shown in Fig. 11, which has been modified to conserve space. These trees have been very informative for the revisions of Costa Rican braconids that follow. In this example (Fig. 11) there are four barcoded specimens outside of our collections, one from Mexico, two from Belize and another from Kentucky. In this work, we describe only the specimens from ACG or Costa Rica as a whole, and, as would be the case for morphological identification and larger series, leave the unexamined individuals with a question mark hovering over them. Unless there are images on BOLD that clearly suggest that non-Costa Rican specimens are different species (we have several cases of this), we remain agnostic. Otherwise, for the present, we know they are either conspecific or potential cryptic species to the Costa Rican specimens. Just as in morphological approaches, many more specimens between these disparate localities need to be collected and barcoded to reach informed conclusions. Experience to date within ACG has been that partitioning of a BIN such as that in Fig. 11 will often lead to three clearly different species as ecological information accumulates and sample sizes for the BIN increase. One will be the holotype and there will be two more to describe, precisely as occurred with the originally named Udranomia kikkawai (Hesperiidae) (Janzen et al. 2017). If an older, morphology-based holotype is located that appears to be this species, when that holotype is DNA barcoded it will match just one or none of the Costa Rican BINs. In the case of U. kikkawai, the 1906 holotype from Venezuelan rain forest was found to match the specimens from Costa Rican rain forest, and two others, one from dry forest and the other from the intergrade, were described as two new species.

In Fig. 9 there is a red arrow pointing to the link “Public records in this BIN”. When this is selected, all of the specimens in the BIN are listed (Fig. 12) with links to public specimen pages with information such as that in Fig. 8.

From the list of the public records as shown in Fig. 12, users have access to download “Specimen Data,” “Sequences,” and “Trace Files” and generate an occurrence map. These options are in the upper right corner and provide various common formats. For example, “Specimen Data” can be downloaded in Darwin Core Format (DWC), XML, or TSV, and there is also the option to download the “Specimen Data” and “Sequence Data” in a combined file with either XML or TSV. Users may wish to download the sequences to use for further analysis such as Maximum Likelihood or Maximum Parsimony trees. We caution, however, that at present, BOLD does not store all of the collateral information for a particular ACG vouchered specimen. For that, the master database at http://janzen.upenn.edu needs to be consulted.

A focus of BOLD Systems is to support collaboration and research networks, and as such, the support team at BOLD helps to connect interested parties with the owners of the specimens with data in BOLD so that they might inquire for further details or gain access to physical specimens, samples, or DNA extracts. Requests for these particular physical objects can be directed to the CBG but for ACG the request should go to DHJ and WH, who will distribute accordingly.

Sequences (barcodes) of the Costa Rican specimens were assigned to multi-specimen operational taxonomic units called Barcode Index Numbers (BINs) (Ratnasingham and Hebert 2013) that were generated by BOLD (as in mentioned above, like assigning look-alikes to unit trays visually). For each subfamily treated in the following chapters we produced a NJ tree except for Ichneutinae and Proteropinae which are combined in one tree (Suppl. materials 1–10) containing all records from the New World contained in the BOLD database. This revealed many BINs containing extra-Costa Rican specimens such as displayed in Fig. 11. We comment on each of these in the individual species accounts with respect to their likelihood of being conspecific with their Costa Rican BIN partners. The NJ trees are also useful to identify sister-species and closest neighbors. Link to the BOLD treatments of the holotype and the “species”, as represented by the BIN, are provided for each species. The paratypes are listed as BOLD codes where detailed information and often images can be found for each specimen.

Morphology and host information were compared to the BIN assignments that “package” the members of the NJ tree into putative species. Most specimen groups (most BINs) were supported by all data sources and treated as one species. However, as mentioned previously, in the treatment of Macrocentrus that follows here, seven species are lumped into the same BIN (BOLD:ACK7466). This is commonplace in other taxonomic groups; there are many cases where a single BIN has been shown unambiguously to be two or more species by their biology and a deeper genome probe (e.g., Hebert et al. 2004; Janzen et al. 2017). There are several reasons for this phenomenon. First, the BIN algorithm separates groups at ca. a 2% COI sequence divergence (Ratnasingham and Hebert 2013). However, while in the early stages of evolutionary separation, sympatric species can easily have their own distinctive biology and evolutionary trajectories long before their barcodes have attained 2% difference through cumulative mutations. Second, even after long periods of separation, the barcode region, by chance, may not have differentiated to the 2% level. These “shallow splits” in an NJ tree are very difficult to interpret with small sample sizes, but with either large samples and/or ecological data, they become evident flags for the presence of two or more species in the BIN, species that are in turn defined and diagnosed by non-BIN traits as well as the trait of being in that BIN (e.g., Burns et al. 2007, 2008; Bertrand et al. 2014) or by doing a deep genome probe (Janzen et al. 2017). In hyperdiverse tropical taxa, such as many of those treated here, the only practical and reliable method of species-level identification will be by their BIN code (and by a more comprehensive DNA barcode when there are multiple species in a BIN), rather than by attempting species-level identification of specimens one-by-one based on their morphology. This process renders the binominal scientific name and its higher classification, along with all the other more traditional kinds of specimen-based information, to be extremely useful because of its inferential value. However, for a specimen that has been identified by its DNA, whether by a barcode or a more profound examination of its genome, the scientific name is collateral to be curated and used just as are other units of collateral.

Our species treatments are comprised of a link to the BIN on BOLD, a consensus COI barcode, and a lateral or dorsal image of the specimen. Added to this are certain details such as the type depository and holotype locality information. There are a few exceptions to this generality. In several cases more than one species fall in the same BIN and in these cases a morphological key or morphological diagnoses are given to separate these. To facilitate an ongoing revision of the Neotropical Macrocentrinae we have also included morphological diagnoses and more comprehensive imaging for this taxon.

To comply with the Code of Zoological Nomenclature, a word-based diagnosis or definition is necessary for each new species proposed. In the past this was usually in the form of a morphological key or a suite of morphological characters that distinguish the species from others being treated. In our case consensus barcodes of the COI barcode region were employed. These can be seen as a suite of character states that are universally shared by all members of a species and unique to the species. Consensus barcodes were generated using custom software written by one of us (BVB) implemented at phorid.net/DNAbarcode. Fasta files of barcodes for all specimens of a species were downloaded from BOLD and aligned in AliView (Larsson 2014). A consensus sequence was generated by running the first script on the DNAbarcode website, uploading the aligned fasta files. Consensus sequences for multiple species generated in this way were pasted into a new AliView file and realigned. The third script on DNAbarcode was then run, using the aligned consensus sequences.

Braconid specimens from the following New World countries appear to be relatively well-sampled in BOLD: Canada, USA, Belize, Argentina, French Guiana, and Mexico. There is a small number of cases where specimens from these countries fall in the same BIN as one of our Costa Rican species. More sampling between these disparate localities, and more genomic and/or morphological and behavioral data will help resolve these species-level cases, which are beyond the scope of this paper. Only rarely are there images of these extra-Costa Rican specimens that suggest that they are not conspecific. These are discussed in more detail under their respective species treatments. To reiterate and confirm: we fully understand that barcodes are not a panacea that will solve all species-limit questions, we simply suggest that they are the optimal available first step.

We attach 10 supplementary chapters for the 11 subfamilies treated here representing ten COI NJ trees, all downloaded from BOLD in November 2020. Each is a snapshot of the dynamic project database at http://djanzen.sas.upenn.edu, select fields of which are routinely updated to BOLD for public use and taxonomic treatment as in this work. The project database is forever dynamic owing to the need for constant updating of the names of species in the ever-shifting taxonomy of plants, hosts and parasites, as new publication and biological relationships emerge. We use the NJ trees as quick visual summaries of sample sizes per species, degrees of specialization of the parasites, flagging of species new to the project, indications of taxonomic puzzles, variability in barcode length, BIN composition (the equivalent of sorting morphological look-alikes into unit trays), and data-checking. Typically, what we consider to be conspecifics are grouped together in their own terminal clade. However, specimens with less than about 550 base pairs are potentially suspect as to their placement in a NJ tree. The shorter the barcode, the more likely it is to be misplaced on the tree, requiring taxonomic placement by other traits, such as morphology or host data or both. It is not uncommon for unrelated specimens with short sequences to group together; this is because the sequences that would group them with their conspecifics are not present and only conserved base pairs, common to many species, are present. Defectively short barcodes are not honored with a BIN code, even though they may group with BINed conspecifics. The arrangement of presumed conspecifics within a BIN into shallowly separated monophyletic groups has high potential for suggesting the presence of a species complex within the BIN, a moderately frequent occurrence (10%). Resolution of such species complexes requires ecological correlates, morphological correlates, and/or a deep genomic probe (e.g., Janzen et al 2017). The appendices contain only the specimens that were successfully barcoded, while all specimens obtained by the inventory are recorded in the project database, irrespective of success in naming.

Aerophilus is worldwide in distribution with hundreds (described plus undescribed) of species in the New World. Members are parasitoids of a wide range of caterpillar families. The genus is well represented in both the Nearctic and Neotropical realms. Sharkey et al. (2016) revised the eastern Nearctic species, and Sharkey et al. (2011a) revised and keyed the Costa Rican species. The latter revision employed the name Lytopylus in error, and Sharkey et al. (2016) subsequently transferred all of the species described under Lytopylus in that paper to Aerophilus Szépligeti, 1902.

Members are widespread but rarely collected in the neotropics. The extremely reduced venation of Mesocoelus is rare among agathidines, shared only with the Old World genus Aneurobracon (Agathidinae). The two described species are both from the Caribbean (holotypes from St. Vincent and Cuba). Reported hosts are Chilocampyla psidiella and Acrocercops sp., leaf-miners in the family Gracillariidae. No new host records for the genus are reported here.

Neothlipsis is widespread in the Nearctic and northern Neotropical region. There may be ca. 70 species including undescribed species. The sole host record is Samea multiplicalis (Guenée) (Crambidae). Sharkey et al. (2011b) included eleven species but omitted the following: Neothlipsis smithi (Ashmead), new combination for Microdus smithi Ashmead, 1894; Neothlipsis pygmaeus (Enderlein), new combination for Microdus pygmaeus Enderlein, 1920; and Neothlipsis unicinctus (Ashmead), new combination for Microdus unicinctus Ashmead, 1894. All of the holotypes of these species were examined by MJS, and all are found in Mesoamerica including the Caribbean.

Members are widespread but rarely collected in the neotropics. It is unlikely that there are more than ten species including undescribed species The holotype of the only described species from the New World, P. bassiformis, is from Ecuador. van Achterberg (1990a) also included specimens from Colombia and Honduras under that name. Long and van Achterberg (2016) described a new species from Vietnam. We report the first host records for the genus, Brenthia spp. (Lepidoptera: Choreutidae).

Bortoni and Penteado-Dias (2015) described two new species under Plesiocoelus, however, they both belong to other genera: Therophilus anomalus (Bortoni and Penteado-Dias) new combination for Plesiocoelus anomalus Bortoni & Penteado-Dias, 2015; Aerophilus areolatus (Bortoni and Penteado-Dias) new combination for Plesiocoelus areolatus Bortoni & Penteado-Dias, 2015. See the key to genera above that incorporates these aberrant species with reduced wing venation.

Members of Pneumagathis are restricted to North and Central America. There appear to be only two species, P. erythrogastra (Neotropical) and P. spiracularis (Nearctic), both of which are parasitoids of Hesperiidae.

Therophilus is a cosmopolitan genus with many hundreds of species (mostly undescribed). It is relatively rare in the neotropics. Published hosts include caterpillars of a wide range of lepidopteran families (Yu et al. 2016). No species have previously been recorded from Costa Rica or neighboring countries.

Members are widespread in the neotropics. Sharkey (1990) revised the species, and one new species is reported here. There are no previously reported hosts, but here we document four species of Zacremnops on crambid caterpillars (Crambidae).

Members of this genus are distributed widely across the neotropics. The two described species are from South America. The first host record (Hesperiidae) for the genus is reported here. The genus has not been revised, though there are likely to be fewer than ten species.

The Costa Rican species of Zelomorpha were revised by Meierotto et al. (2019a) employing barcodes as diagnostic characters. The sole gregarious species of Agathidinae, Zelomorpha gregaria (Sarmiento & Sharkey, 2004) (originally named Coccygidium gregarium), was inadvertently omitted from that publication. The inclusion of Z. sarahmeierottoae brings the total number of Costa Rican species to 17. Hosts include caterpillars in the following families: Erebidae, Euteliidae, Geometridae, Nolidae, Notodontidae, and Noctuidae.

Bracon is an enormous, polyphyletic, cosmopolitan genus with thousands of undescribed species. They are idiobiont ectoparasitoids, attacking hosts in many insect orders, but primarily Coleoptera and Lepidoptera.