henhaobio
Sign in

Proceedings of the Royal Society B Biological SciencesZoologyPeer-reviewed journal article

High-throughput biodiversity surveying sheds new light on the brightest of insect taxa

Ela Iwaszkiewicz-EggebrechtiD, Robert M. GoodselliD, Bengt-Åke Bengsson, Marko MutaneniD, Mårten KlinthiD, Laura J. A. van DijkiD, Piotr ŁukasikiD, Andreia MiraldoiD, Anders F. AnderssoniD, Ayco J. M. TackiD, Tomas V. RosliniD, Fredrik RonquistiD

PDFDOIPreprintCodeData

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (25 Oct 2024) · CC BY 4.0

One year of COI metabarcoding of 197 Malaise traps across Sweden detected 1,535 of 2,990 known Lepidoptera species, plus specimen-validated new country records, range extensions and putative undescribed species.

Malaise trapping with DNA metabarcoding is increasingly used for insect inventories, but its accuracy is hard to judge because it is mostly applied to poorly known groups and regions where the true species list is unknown. Here it is benchmarked on Swedish Lepidoptera, which have a long-curated national checklist, dense naturalist records and nearly complete barcode reference libraries, and form a conservative test because Malaise traps are often assumed to catch Lepidoptera poorly.

Key results

  1. ShownA single year of trapping recovered just over half of all Lepidoptera species ever recorded in Sweden. Small moths were recovered somewhat better than large moths, butterflies were recovered least often, and the species that were missed tend to be rarer in naturalist records.

    1,535 of 2,990 known species (51%); 1,532 of 2,697 natural residents (57%). Macrolepidoptera 44% vs Microlepidoptera 56%; butterflies 25%. 9 of 76 families entirely missing, all species-poor. Missed species had significantly fewer GBIF records than detected ones (two-sided t-test, p<0.001). · Species coverage

  2. ShownBarcoding of individual specimens from selected bulk samples confirmed three species new to Sweden, two range extensions and eight distinct COI lineages that may be undescribed species. One of these lineages matches Acrocercops andreneli, described in 2023 and first reported from Sweden in 2024.

    20 candidate clusters tested: 12 possibly new to science (6 flagged as noise), 4 new country records, 4 range expansions. Validated: 3 of 4 new to Sweden (Eupithecia groenblomi, Phyllonorycter mespilella, Agriopis budashkini), 2 of 4 range expansions (a third was a pipeline lumping error), and 8 putative new species (3 of 6 noise-flagged, 5 of 6 unflagged), each at least 3% from known COI. · Validation of new discoveries; Individual specimen barcoding; Validation of new discoveries; Discussion

  3. ShownMetabarcoding clusters corresponded closely to traditional species identifications, and errors tended toward oversplitting rather than lumping.

    544,263 amplicon sequence variants (ASVs) in 33,888 clusters, 1,705 of them Lepidoptera. After noise filtering: precision 0.997, recall 0.980, homogeneity 0.997, completeness 0.991. The default pipeline assigned 1,437 clusters to species; this rose to 1,636 after name conflicts and BOLD reference errors were fixed by hand. · Results

  4. ShownCompared with all naturalist records for 2019, the trap survey produced far fewer occurrence records but about two-thirds as many species, and per-species record counts from the two sources were barely correlated.

    Naturalists: 272,380 Lepidoptera records (61% of 444,078 Swedish insect records) covering 2,379 species. IBA: 42,885 records (16% of naturalist Lepidoptera records) covering 1,535 known species; 1,228 species (52% of the naturalist list) shared. Per-species record counts: Pearson r (log scale) 0.11. · Comparison to data from naturalists; Discussion

  5. SuggestedNaturalist data are skewed toward butterflies, other large moths and populous regions. The trap survey gave more even coverage, matching or exceeding naturalists for many small moth species and in sparsely populated areas.

    Average IBA records per known species equalled or exceeded naturalist records for about half of Microlepidoptera families. Naturalist data concentrate around Stockholm, Mälardalen, Gothenburg and Skåne. Across 125 roughly equal polygons, IBA recovered a larger share of local species in sparsely populated areas, most clearly for Microlepidoptera (described as a visual heuristic). · Comparing regional coverage of naturalist and…; Comparison to data from naturalists

Methods, data and reproducibility

Malaise traps at 197 systematic sites across Sweden were emptied weekly by citizen scientists in 2019; ~4,749 bulk samples were metabarcoded for a 418-bp cytochrome c oxidase subunit I (COI) fragment, and clusters were matched to the DynTaxa checklist and 2019 naturalist records from the Global Biodiversity Information Facility (GBIF). Discoveries were checked by 658-bp MinION barcoding of single specimens.

  • Insect Biome Atlas (IBA) Malaise trap bulk samples, collected weekly across Sweden from January to December 2019 and metabarcoded for a 418-bp COI fragment · insects (Lepidoptera analysed here) · n = ~4,749 bulk samples; 197 traps
  • DynTaxa checklist of Lepidoptera recorded from Sweden, with occurrence status (Artdatabanken) · Lepidoptera · n = 2,990 species in 76 families (2,697 natural residents)
  • GBIF occurrence records of Swedish Lepidoptera up to and including 2019 (human observations, preserved specimens, Artportalen) · Lepidoptera · n = 272,380 Lepidoptera records in 2019
  • Custom COI reference database used for SINTAX taxonomic annotation · 10.17044/scilifelab.20514192.v4
  • Individual specimens sorted from selected bulk samples and barcoded for the 658-bp COI Folmer region · Lepidoptera · n = 20 bulk samples; 20 candidate clusters
  • Independent PCR-free metagenomic analysis of IBA samples at genus level (cited analysis, López) · Lepidoptera · n = 40 samples
Methods & tools
Townes-type Malaise traps, FAVIS metabarcoding protocol, Illumina NovaSeq 6000, SINTAX, SWARM, IBA noise/NUMT filtering algorithm v1, simple features (R package), MinION, ONTbarcoder, BOLD, GenBank BLAST
Reproducibility gaps
The GitHub URL is printed as 'githttps.com/...' and looks malformed. The Figshare link is an account-area address that may not resolve publicly. No raw-read accession is given; sequence processing is referenced to Miraldo et al. (2024).
Limitations (6)
  • Accuracy likely benefits from the near-complete Lepidoptera reference library and the long 418-bp amplicon. Groups poorly represented in databases (Diptera, Hymenoptera) or richer in nuclear copies of mitochondrial DNA (NUMTs; Orthoptera, Isoptera) may give less precise, harder-to-validate data. · stated by the authors
  • NUMTs that lack indels or premature stop codons are hard to tell apart from genuine COI, so unknown haplotypes warrant caution. Contamination, PCR chimeras and sequencing errors can also produce spurious clusters. · stated by the authors
  • Variation at primer binding sites could cause missed detections. The authors consider this a minor problem for Lepidoptera, based on a cited PCR-free metagenomic analysis of 40 IBA samples at genus level. · stated by the authors
  • Some discoveries have other possible explanations. The Agriopis budashkini signal could reflect mitochondrial introgression into A. aurantiaria, and the far-northern Aleimma loeflingiana record could be a stray vagrant. · stated by the authors
  • Counts differ between sections. Methods report 197 trap sites, but the Discussion says 197 traps at 100 sites. Results give eight validated putative new species and three new country records; the Discussion gives seven and four. · noted in this summary
  • Specimen-level validation covered only 20 hand-selected candidate clusters. The species status of the putative new taxa rests mainly on COI distance (at least 3%), and the regional comparison with naturalist data is a visual heuristic with no statistical test. · noted in this summary

Read this if you run or evaluate Malaise-trap metabarcoding for insect monitoring and want a benchmark against a well-documented national fauna and naturalist records.

Citations
11
FWCI
5.04
Field percentile
top 4%

Abstract

Sampling of species-rich taxa followed by DNA metabarcoding is quickly becoming a popular high-throughput method for biodiversity inventories. Unfortunately, we know little about its accuracy and efficiency, as the results mostly pertain to poorly-known organism groups in underexplored environments or regions of the world. Here we ask what an extensive sampling effort based on Malaise trapping and metabarcoding can tell us about the lepidopteran fauna of Sweden - one of the best-understood insect taxa in one of the most-surveyed countries of the world. Specifically, we deployed 197 Malaise traps for a single year across Sweden in a systematic sampling design, then metabarcoded the resulting 4,749 bulk samples, and compared the results to existing data sources. We detected more than half (1,535) of the 2,990 lepidopteran species ever recorded as occurring in Sweden, and 323 species not reported during the sampling period by other data providers. Full-length barcoding of individual specimens confirmed three new species for the country and extensive range extensions for two species. It also corroborated eight genetically distinct COI variants that may represent new species to science, one of which has since been described. Most of the new records are for small and inconspicuous species and poorly surveyed regions, suggesting that they represent previously overlooked components of the fauna. Our findings, corroborated by independent metagenomic analyses, show that DNA metabarcoding can be a highly efficient and accurate method of biodiversity sampling, to the extent that it can generate significant new discoveries even for the most well-known of insect faunas.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISBuchner '23Miraldo '25Iwaszkiewic '26Ronquist '20Koutsovoulo '26Mata '20
75% semantic matchEcology
D. Buchner, J. Sinclair, M. Ayasse+23 more·bioRxiv·May 4, 2023