The checklist of species-level names, published by Andrews et al. (1979), was used as the starting point for this study, with nomenclatural updates enacted to reflect the taxonomic concept labels of Bousquet et al. (2018). We then surveyed the appropriate subsequent taxonomic literature to add species-level names authoritatively reported from the Algodones; specifically: Papp (1981), Doyen (1984), Doyen (1987), MacLachlan and Olson (1990), Aalbu (2005). The checklist was completed by surveying darkling beetle specimens from the authors' personal collections, particularly the Rolf L. Aalbu Collection (henceforth: RLAC; located in California, USA), which has extensive holdings of Algodones tenebrionid material. In other words, the expert-generated checklist includes a combination of (1) literature records where no individual occurrences are explicitly recognized and (2) under-mobilized RLAC vouchers.

Excluded sources. In our assessment, the RLAC and the California State Collection of Arthropods (CSCA; located at the California Department of Food and Agriculture in Sacramento, California) are the two research collections with the most comprehensive holdings of Algodones darkling beetles. Neither of these collections currently serves occurrence data to aggregators. Meanwhile, the R.M. Bohart Museum of Entomology (UCDC; University of California, Davis), which houses the Kimsey et al. (2017) material, presently serves up data only through their institutional website: http://museums.ucdavis.edu/bohart.aspx. A total of 308 focal records were available through this website as of January 10, 2018 (Suppl. material 1). These records are not Darwin Core-compliant, however, typically lacking information on the date of collection, collector, identifier and georeference data. Therefore, they were not included in the occurrence data-based checklist. The California Terrestrial Arthropods Database (CalBug; see Hill et al. 2012; available at http://calbug.berkeley.edu/index.html) had no focal records as of January 10, 2018. Lastly, after carefully inspecting non-vouchered occurrences (observations) in select citizen science/social networks (e.g.,https://www.inaturalist.org), we were unable to confidently identify many of the photo-vouchers ourselves and judged many more non-expert identifications too doubtful to be included.

Included sources. Three major biodiversity data aggregators were queried for darkling beetle occurrence records from the Algodones: (1) the Symbiota Collections of Arthropod Network portal (SCAN), (2) the Integrated Digitized Biocollections portal (iDigBio) and (3) the Global Biodiversity Information Facility portal (GBIF). Records from each aggregator were downloaded on January 02, 2018. The occurrence records were sorted by the Darwin Core term "dwc:scientificName", yielding a list of unique taxonomic names and a count of the total number of records for each. All original scientific names were manually remapped to the classification of Bousquet et al. (2018). Species-level names not included in our expert-generated checklist were evaluated at the individual record level and are discussed below.

A total of 54 darkling beetle species-level names are included in the expert-generated Algodones checklist (Table 1). Of these, 34 were previously documented in the literature; the remaining 20 are formally published here for the first time. This increase in recognized species relative to the study of Andrews et al. (1979) (23 recorded species) is remarkable, as the new total amounts to nearly half of the 113 species-level entities reported for the entire Sonoran Desert region of California by Aalbu and Smith (2014).

Not surprisingly, access to reliable taxonomic identifications of vouchered specimens was the greatest challenge to creating the checklist, given also the scarcity of modern systematic treatments for many of the recognized species. Several groups - e.g. Edrotes LeConte, 1851 sec. Bousquet et al. 2018 and Ulus Horn, 1870 sec. Bousquet et al. 2018 - have revisions in progress, whereas others such as Helops Fabricius, 1775 sec. Bousquet et al. 2018 and Hymenorus Mulsant, 1852 sec. Bousquet et al. 2018 are in great need of revision. Hence, future studies could drastically change the species-level names and concepts employed here. Indeed, the genera Hylocrinus Casey, 1907 sec. Bousquet et al. 2018 and Metoponium Casey, 1907 sec. Bousquet et al. 2018 were last revised by Casey (1907) - a treatment that entails so many poorly differentiated species-level concepts that we know of no subsequent specialist who would confidently identify new specimens to these concepts. We similarly refrain from this task in the expert-generated checklist.

The results of all three aggregated occurrence data-based checklists for the Algodones darkling beetles are summarized in Table 2. The underlying raw portal data and steps taken to process and interpret them in relation to the expert-generated checklist, are provided in Suppl. material 6. Accordingly (Section I of Table 2), the SCAN portal contains 559 valid occurrences corresponding to 31 species-level concepts as recognized in Table 1 (with 108 ~ 19.3% records needing nomenclatural adjustments); the iDigBio portal serves up 386 such occurrences representing 25 species-level concepts (with 175 ~ 45.3% records needing nomenclatural adjustments; and GBIF offers 100 valid occurrences of 15 species-level concepts (with 34 ~ 34.0% needing nomenclatural adjustments).

In addition (Section II), each portal includes occurrences not considered valid for the focal taxonomic entities, mostly due to erroneous or uncertain identification (in our judgment), as follows: SCAN includes 133 occurrences corresponding to 21 taxonomic concepts; iDigBio contains 59 occurrences representing 21 taxonomic concepts; and GBIF serves up 34 records pertaining to 11 taxonomic concepts.

The patterns of occurrence-level overlap amongst the three data portals tell a potentially interesting story about biodiversity data meta-aggregation and signal propagation (or loss), as well as the relationship between regionally and/or taxonomically constrained portals and data quality (Mesibov 2013, Gries et al. 2014, Franz and Sterner 2018, Mesibov 2018). However, these topics reside somewhat outside of our current focus. Similarly, with the exception of the select occurrences discussed below, we will not dissect in detail the various apparent instances of nomenclatural adjustments and incorrect or uncertain identifications that the portal data represent.

The nearly contiguous Algodones Dune formation and the large sand sea of the Gran Desierto de Altar are both derived from sediments from the Colorado River (Lancaster et al. 1987, Muhs et al. 1995) and are narrowly separated by the river's current course. The Colorado River begain draining into this region around 4 mya (Winker and Kidwell 1986, Derickson et al. 2008), depositing sediments that formed the Colorado River Delta, which now marks the northern limit of the Gulf of California (Waters 1983). The presently dry Salton Trough, the low-lying region north of the Colorado River Delta, has seen periodic flooding during the Holocene - by the Colorado River changing course westward and draining into the prehistoric Lake Cahuilla - at least three times in the past two thousand years (Waters 1983). Sediments from these sequential fillings of Lake Cahuilla are thought to have formed the Algodones Dunes (Norris and Norris 1961, Derickson et al. 2008). As a biogeographic factor, the Colorado River could present a barrier to gene flow and dispersal for sand-dune restricted lineages, particularly if these are flightless and thus dispersal-limited. Nevertheless, it is unclear whether any species-level entities of darkling beetles are unique to either the Algodones Dunes or the Gran Desierto de Altar. Moreover, historical shifts in the placement and volume of the Colorado River may have facilitated the homogenization of faunal distributions. Thus we consider the Colorado River-derived dunes - spanning both the Algodones Dunes and the Gran Desierto de Altar - as a single cohesive biogeographic region and we refer to it simply as the Gran Desierto.

As shown in Table 3, the following five entities are seemingly restricted to the Gran Desierto. Araeoschizus andrewsi sec. Papp 1981 and Araeoschizus wasbauerorum sec. Papp 1981 are both known from the Algodones and the Gran Desierto de Altar. Batuliodes wasbaueri sec. Doyen 1987 is known from the Algdones as well as from a small remnant sand dune area, located approximately 20 miles southeast of Mexicali, Mexico, near the Colorado River. The congruent distributions of these three flightless species reinforce the notion of a single biogeographic subregion. Batuliomorpha imperialis sec. Doyen 1987 and Lepidocnemeplatia sp. (nov.) sec. Aalbu et al. (in prep.) are both small species (~ 3 mm in length) collected mainly by sifting sand. They are currently only recorded from the Algodones, though we may expect them to be more widespread but uncollected throughout the Gran Desierto.

Kimsey et al. (2017) considered the following four species as "only recorded from the [Algodones] dunes": Edrotes arens sec. Doyen 1968, Eusattus dilatatus sec. Doyen 1984, Nocibiotes crassipes sec. Casey 1895 and Tonibius sulcatus sec. Casey 1895. We hereby refute all of these assessments of Algodones-constrained precinction. Edrotes arens sec. La Rivers 1947 was originally described based on three specimens from the Yuma Dunes in Arizona, with subsequent literature reports from many sand dune localities throughout California (Andrews et al. 1979). SCAN and iDigBio hold multiple occurrences of Edrotes arens sec. Doyen 1968 from Arizona and California localities. Specimens of Eusattus dilatatus sec. Doyen 1984 have been reported in literature from deep sands throughout the Lower Colorado River Valley, ranging from Puerto Peñasco, Mexico, to Blythe, California (Doyen 1984). Again, SCAN and iDigBio serve up the corresponding non-Algodones occurrences. Nocibiotes Casey, 1895 sec. Bousquet et al. 2018 is in need of revision, with many specimens in research collections currently not identified to the species level. However, specimens of Nocibiotes crassipes sec. Casey 1895 are known to occur in Baja California and throughout southern California (RLA, unpublished data). Tonibius Casey, 1895 sec. Bousquet et al. 2018 is presently monotypic, containing only Tonibius sulcatus (LeConte, 1851) sec. Casey 1895, which is the entity presumably referred to in Kimsey et al. 2017, with misattributed name authorship ("Casey"). The type locality for Tonibius sulcatus sec. Casey 1895 is "San Diego" (LeConte 1851) and additional occurrences are recorded in literature from Baja California (Blaisdell 1943) and Nevada (Thomas 1983). Again, SCAN and iDigBio contain respective occurrences from non-Algodones localities.

Nine entities present in the Algodones appear to have distributions wider than the Gran Desierto yet are still restricted to the Lower Colorado River Valley (Table 3). Two of these, Eupsophulus horni sec. Spilman 1959 and Mycotrogus angustus sec. Spilman 1963, are poorly known both in terms of their natural history and distributions. The remaining seven recognized species are typically found in areas with sandy soils. Some are only found in deeper sand dune habitats - e.g. Edrotes arens sec. Doyen 1968 and Eusattus dilatatus sec. Doyen 1984 - whereas others inhabit sandy washes and alluvial flats (e.g. Hymenorus thoracicus sec. Fall 1931). A total of 259 occurrences are available for these nine species in SCAN, of which 234 are considered valid in our assessment (Suppl. material 5). These occurrences are also mapped in Fig. 1 and suggest the presence of a shared distributional pattern: both towards the north, along the Colorado River and east, throughout the low desert regions of the Yuma Desert in south-western Arizona and north-western Sonora. The pattern is tentative, though plausible given similarities in habitat temperatures, rainfall and soil type. More than half of the specimens (125 occurrences) are from the well-sampled Algodones, thus offering little data regarding broader distributions of the respective species. We predict that further sampling and taxonomic identification efforts will reveal more extensive distributions for many of these.

Figure 1.  

Lower Colorado River Valley Restricted Species Distributions. 239 digitized records from SCAN for 9 species. Map generated using www.simplemappr.net with background colors indicating ecoregions. The bright pink region encompassing the occurrence records roughly corresponds to the Lower Colorado River Valley subregion of the Sonoran Desert.

The Algodones and surrounding desert environs of southern California, though usually classified as part of the Sonoran Desert (Brown 1994), have strong floristic ties to both the Mohave Desert to the north and the Vizcaíno Region in the center of the Baja California peninsula (Shreve 1942). The tenebrionid fauna of the Algodones also has strong biogeographic ties to these regions (Table 3). The strongest faunal overlap is with the Mohave Desert, which shares 29 herein recognized species with the Algodones. In contrast, only 17 species extend their distributions into non-Lower Colorado River Valley regions in Baja California. Only 28 out of the 52 examined species have ranges that extend into other biogeographic areas, which typically included either coastal California or other subregions of the Sonoran desert. This rich tenebrionid fauna of the Algodones may owe its diversity in part to the blending of psammophilic faunas from the surrounding regions.

Regional checklists are published to be used, corrected, expanded and inevitably become outdated - the sooner the better. In that sense and only for the subcomponent of the Tenebrionidae sec. Bousquet et al. 2018, the checklist of Kimsey et al. (2017) has already served its purpose. At the same time, we have shown that these authors (and the reviewers, presumably) could have worked more thoroughly on their checklist product (see also Suppl. material 7). In addition to significant literature record omissions (e.g. Andrews et al. 1979) and nomenclatural errors, we may consider the institutional-only, non-Darwin Core database to be inadequate in the context of global biodiversity data aggregation (Maddison et al. 2012, Page et al. 2015). Moreover, occurrences of as many as 31 focal recognized species of Tenebrionidae sec. Bousquet et al. 2018 in the Algodones could have been discovered and included just by querying the SCAN portal. Indeed, every species recognized in Table 1 has at least one occurrence record in SCAN, though not necessarily from the Algodones. Thirteen species reported on SCAN from the Algodones were not listed in Andrews et al. (1979), including five which have never been reported from the region in published literature until now. In our view and considering the presence of nearly 7 million North American occurrences in SCAN currently (see Introduction), this suggests that any author, aspiring to generate a comprehensive and reliable checklist of North American insects, is well advised to explore and selectively include aggregated, occurrence data to their product. At a minimum, we would expect an explanation why such data were discarded, following their exploration (see also Ferro and Flick 2015, Sikes et al. 2016).

Of course, the flipside of the above message is this: a very considerable subsection of the Table 1 checklist depends solely on our access to and reliance on, specimen material from the Rolf L. Aalbu Collection. This collection has no on-line presence at the moment, nor foreseeable support to digitize these data moving forward. The RLAC data are both invaluable in their content and unsuited in their current form for a strictly Darwin Core-based checklist approach.

Aggregated occurrence data typically come with a combination of data formatting and quality insufficiencies that are justly attributed to the digitizing source collection, plus other shortcomings newly generated in the process of aggregation (Mesibov 2013, Mesibov 2018, Franz and Sterner 2018). Rather than reviewing these issues (once more) in the context of our particular checklist update, we limit our discussion to a few pragmatic as well as more future-oriented solutions to enhancing occurrence data-based checklists.

We believe that the emergence of aggregated occurrence data should not only enrich the types of information sources and data formats that contribute to checklists, but should increasingly obviate altogether the notion of static, closed, print or digital checklist publications. Indeed, from a technical and perhaps also scientific point of view, the interaction between the Kimsey et al. (2017) checklist and our update need not take the form of two structurally unconnected information packages, each wholly attributed to either one or the other author team. Instead, we can envision the two respective contributions, or checklist versions, to develop as finely attributed bundles of annotations (Morris et al. 2013), managed on top of an underlying, unified Darwin Core-based occurrence data network. Similarities and differences between each version could then be expressed - almost entirely via automated services - as a differential ("delta" - Δ) between two Darwin Core-compatible sets of occurrence records. Subsequent authors would receive credit mainly for occurrences added, or reviewed and newly annotated, in relation to previously published records sets.

For such incremental, wholly Darwin Core-based published checklist versions to become reality, however, several aspects of authoring checklists need to receive careful attention. In particular, authors should express clearly which data sources of the current checklist version are also traceable to aggregated occurrences, or are solely reliant on expert assessment of non-mobilized records (compare Table 1 and Table 2). Our update shows that the latter category remains essential. At the same time, moving most or all occurrence records into the former category is highly desirable and a pre-requisite for fully Darwin Core standard-based checklists.

Likely, this also means that the biodiversity data community should strive to lower or remove technical and social barriers to mobilizing occurrences from private or institutional collections that currently lack the resources to accomplish aggregation. In other words, we believe that data mobilization by outsiders should become more frequent.

From a technical point of view, it is possible to set up a portal collection where any checklist author can mobilize and annotate any occurrence they are able to process as part of their research and data filtering effort - even and especially if the specimens in question belong to other individuals or institutions. We have done so, on an exploratory scale, with the "ARTSYS" collection (Externally Processed Specimens - Arthropod Systematics Research) in SCAN: http://scan-bugs.org/portal/collections/misc/collprofiles.php?collid=114. However, the prevalent culture for North American insect collections is that decisions regarding formal specimen digitization are strongly tied to the constraints of specimen ownership. This position is not well aligned with checklist author motivations to produce open, reusable data packages. An increased decoupling between the physical specimen repository and the ability to mobilize the associated occurrence data is needed.

Lastly, the notion of open, dynamic data checklists requires additional efforts to contextualize each version's - and indeed each occurrence record's - taxonomic concept usages and concept-referencing identification assertions. Too often the tradition of publishing static biodiversity data products is tied to an underlying assumption that readers will reliably understand the authors' name usages in context (though see Franz et al. 2016, Remsen 2016, Franz and Sterner 2018, Packer et al. 2018, Senderov et al. 2018).

Our use of taxonomic concept labels is one component of making checklists version-ready, by connecting the name usages in the above table to particular systematic treatments in which the corresponding evolutionary entities are circumscribed. Yet we should also note that, at the level of occurrences, our data are not fully there yet (see Suppl. materials 2, 3, 4). Of the 693 occurrences taken from SCAN, maximally 229 records (33.0%) entail some information regarding the terms dwc:identifiedBy and/or dwc:dateIdentified. Only five occurrences (0.7%) have the term dwc:identificationReference filled with data. These ratios are unsatisfactory; and yet this low degree of concept/identification reference annotation is still better in relation to the data served up by the other two aggregators. iDigBio offers 454 occurrences, which detail no identification data at all. Meanwhile GBIF has 133 records, of which 92 (69.2%) show identification data. However, these data are very frequently altered - i.e.,"elevated" to the higher-ranked taxonomic name that the GBIF taxonomy recognizes - while (falsely) retaining the original identifier attribution (see also Franz and Sterner 2018). We note in passing that only the Symbiota portal allows us to directly (via username/password log in) contribute occurrence-level identifications and taxonomic concept information.

For regional, occurrence data-based checklists to become fully open and versioning-ready, the first version should set a high bar of decoupling both taxonomic name usages and the identifications of occurrences from under-contextualized taxonomic names. We have attempted this for our tabular Tenebrionidae sec. Bousquet et al. 2018 of the Algodones checklist update, but are falling short regarding the underlying occurrence-level data. Moving forward, we need to treat every occurrence like a prospective micropublication that can stand on its own (see also Packer et al. 2018), by carrying sufficient taxonomic and identification-related information to be re-aggregated and re-published in updated checklist versions while retaining the provenance of its taxonomic identity and expert work effort. Only then can we assign proper credit to these experts and their work of enhancing the quality of regional checklists.