bioRxiv Science⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.01.16.697653

COLLEMBOT: AI-Based Counting of Collembola for OECD 232 Tests

Wehrli, M.·Meyer, A. F.·Souza da Silva, E.·van Loon, S.·van Hall, B. G.·van Gestel, C. A. M.·Natal-da-Luz, T.·Doering, M. V. R.·Feldhaar, H.·Mair, M.·Jordan, D.·Langer, M.
View original source ↗Export metadata ↓

Abstract

Ecotoxicological tests with soil organisms, such as the collembolan Folsomia candida, are essential for assessing chemical risks in terrestrial ecosystems. However, the current Organization for Economic Co-operation and Development (OECD) 232 reproduction tests rely on manual counting of juvenile and adult Collembola, a process that is costly, labor-intensive, time-consuming and prone to operator bias. These limitations restrict data availability and hinder robust risk assessments. We therefore developed COLLEMBOT, an automated counting tool based on a YOLOv11 convolutional neural network, designed to integrate seamlessly into OECD workflows without protocol modifications. The model was trained on high-resolution images (n = 3207) from multiple laboratories and validated using 22 independent datasets (n = 1704 images) from Amsterdam (Netherlands), Basel (Switzerland), Bayreuth (Germany), Coimbra (Portugal) and Aarhus (Denmark). Datasets consisted of relevant standard soils (OECD artificial soils with 2.5%, 5% and 10% sphagnum peat; LUFA 2.2) and the springtail Folsomia candida. Automated counts showed strong agreement with manual counts (R{superscript 2} = 0.88-0.99). Dose-response curves derived from automated and manual counts strongly overlapped and effect concentrations (EC and EC) differed minimally (Median %{Delta} 6.2 {+/-} 23 and EC10 - EC90 R2 [≥] 0.977), remaining within acceptable limits for regulatory risk assessment and confirming reliability. Time efficiency improved significantly: a test with [~]300 images and up to 1,500 individuals per image was processed in less than 3 hours, compared to [~]137 hours needed for manual counting, a reduction of approximately 97%. By reducing labor and improving reproducibility, COLLEMBOT enables broader hazard data generation for collembolans, supporting science-based chemical risk assessment. The code and workflow are publicly available to facilitate adoption and community-driven development. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=187 SRC="FIGDIR/small/697653v1_ufig1.gif" ALT="Figure 1"> View larger version (131K): org.highwire.dtl.DTLVardef@1d3ddc5org.highwire.dtl.DTLVardef@84c90aorg.highwire.dtl.DTLVardef@1aaf06aorg.highwire.dtl.DTLVardef@18ddfcd_HPS_FORMAT_FIGEXP M_FIG C_FIG

Source connections

Explore related subjects

ecologyPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Wehrli, M., Meyer, A. F., Souza da Silva, E., van Loon, S., van Hall, B. G., van Gestel, C. A. M., Natal-da-Luz, T., Doering, M. V. R., Feldhaar, H., Mair, M., Jordan, D., Langer, M.. 2026-01-20. COLLEMBOT: AI-Based Counting of Collembola for OECD 232 Tests. https://doi.org/10.64898/2026.01.16.697653

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Species-composition distributions under competition, ecological drift and immigration

Understanding how species interactions, ecological drift, and immigration jointly shape community composition is a central goal of community ecology. These processes determine not only average community composition, but also how often different community states occur. Here, we ask how competition reshapes the neutral distribution of community composition generated by drift and immigration. We analyze a stochastic two-species competition model with immigration and derive the stationary distribution of relative species composition near neutrality. The stationary distribution shows when communities are usually balanced and when they are instead dominated by one species. Species dominance can become common even when the corresponding deterministic model predicts stable coexistence, particularly in smaller communities. Balanced and dominance states can also occur together, producing a trimodal distribution. Competitive asymmetry further biases the distribution toward the competitively favored species. Thus, the interplay between competition, ecological drift, and immigration determines not only how much community composition varies, but also which community states are most likely to be observed. Our results provide an analytical bridge between neutral and competition theory and a framework for interpreting variation in community composition.

ecology↗
bioRxiv2026

Floristic composition, phenology, and conservation value of four peat bogs in Bucovina, with the presence of Betula nana

This paper presents a comparative analysis of the floristic composition and site characteristics of four peat bogs in Bucovina, Romania: Poiana Stampei, Romanesti, Saru Dornei, and Gaina-Lucina. The research was based on phytosociological releves on 25 msq plots and direct field phenological observations, on six field visits from May to August 2026. Vegetation was characterised using the Braun Blanquet method, and floristic similarity between sites was assessed with the Sorensen and Bray Curtis indices. All four plots shared a common core of taxa characteristic of peatland vegetation: Sphagnum spp., Carex rostrata, Drosera rotundifolia, Eriophorum vaginatum, and Vaccinium species. Species richness was 13 taxa at Poiana Stampei, Romanesti, and Saru Dornei, and 12 at Gaina-Lucina. Romanesti and Saru Dornei showed the highest floristic similarity (descriptive values, not statistically tested, given a single releve per site), while Gaina-Lucina differed most markedly, not through species richness, which was similar across sites, but through species identity and through the presence of Betula nana, a glacial relict absent from the other sites. The results provide a descriptive basis for future research on the floristic composition and conservation of these habitats.

ecology↗
bioRxiv2026

Long-Term Surveillance Reveals Establishment of Aedes albopictus in Eastern Nebraska, USA

Aedes albopictus (Skuse), the Asian tiger mosquito, is a highly competent arboviral vector whose range has expanded substantially across the United States over the past four decades. Despite predictive models placing Nebraska within the species' climatically suitable range, its establishment status in the state has remained poorly characterized. Here, we report results from a nine-year mosquito surveillance program (2017-2025) conducted across 44 Nebraska counties in collaboration with the Nebraska Department of Health and Human Services. Ae. albopictus was detected in five counties, with sustained, annually increasing populations documented in Richardson, Douglas, and Lancaster counties. Richardson County recorded continuous detections during 2017-2025, with proportional representation rising to 60.50% of collected mosquitoes by 2025. In Douglas and Lancaster counties, temporal advancement of first seasonal detection in 2024 and 2025 provide evidence consistent with successful overwintering rather than annual reintroduction. A cumulative degree-day model predicted adult emergence in mid-May across all county-year combinations, consistently preceding trap deployment by two to seven weeks and revealing a systematic early-season surveillance gap. Generalized linear mixed-effects models indicated that trap-level detection persistence, rather than urban location, was the primary predictor of yearly Ae. albopictus positivity, suggesting that current invasion dynamics are driven by focal source populations. These findings provide strong evidence for the establishment of Ae. albopictus in eastern Nebraska and highlight the need for earlier seasonal surveillance and standardized criteria to define establishment in northward-expanding vector populations.

ecology↗