A continental-scale dataset of ground beetles with high-resolution images and validated morphological trait measurements

TL;DR

This work delivers a continental-scale, multimodal dataset of ground beetles (Carabidae) by digitizing over 13,200 NEON specimens from 30 sites with high-resolution images and validated morphometric measurements of elytral length and width. It introduces standardized imaging protocols for pinned and vial specimens, combines TORAS-based automated trait extraction with human validation, and employs Grounding-DINO and CVAT for robust specimen segmentation. The authors demonstrate sub-millimeter precision for digital measurements on pinned beetles and provide comprehensive metadata, cross-linking to NEON environmental data to enable trait–environment analyses. Publicly available on HuggingFace and integrated with NEON infrastructure, the dataset aims to close the invertebrate trait gap and enable AI-driven species identification and large-scale biodiversity monitoring.

Abstract

Despite the ecological significance of invertebrates, global trait databases remain heavily biased toward vertebrates and plants, limiting comprehensive ecological analyses of high-diversity groups like ground beetles. Ground beetles (Coleoptera: Carabidae) serve as critical bioindicators of ecosystem health, providing valuable insights into biodiversity shifts driven by environmental changes. While the National Ecological Observatory Network (NEON) maintains an extensive collection of carabid specimens from across the United States, these primarily exist as physical collections, restricting widespread research access and large-scale analysis. To address these gaps, we present a multimodal dataset digitizing over 13,200 NEON carabids from 30 sites spanning the continental US and Hawaii through high-resolution imaging, enabling broader access and computational analysis. The dataset includes digitally measured elytra length and width of each specimen, establishing a foundation for automated trait extraction using AI. Validated against manual measurements, our digital trait extraction achieves sub-millimeter precision, ensuring reliability for ecological and computational studies. By addressing invertebrate under-representation in trait databases, this work supports AI-driven tools for automated species identification and trait-based research, fostering advancements in biodiversity monitoring and conservation.

Figures (10)

• Figure 1: Map of NEON terrestrial sites across the United States of America. Each circle marks a NEON terrestrial site: black fill indicates sites where we imaged carabid samples. The inset shows the Bartlett Experimental Forest BART site in detail, illustrating NEON’s spatial sampling design across plots designated for specific observation types (e.g., birds, mammals, phenology), carabids are collected at (green) Base Plot. The "Base Plot" diagram (upper middle) outlines the standardized pitfall trap configuration used for ground beetle sampling, with three traps positioned at the south, east, and west edges respectively of a 40 x 40 plot.
• Figure 2: Workflow for standardized ground beetle collection and processing at NEON sites. The process begins with pitfall trap setup GroundBeetles and proceeds through the collection of captured organisms. Specimens are filtered to separate carabids from bycatch. If identified as a beetle, the specimen is further curated based on identification confidence and quantity: specimens with uncertain ID or fewer than ten per species are pinned; otherwise, they are preserved in ethanol (vial).
• Figure 3: Workflow for standardized data collection protocol for beetles across two specimen types: pinned (top) and vial (bottom). Top (Pinned Workflow): Pinned specimens, first repositioned and imaged, are being processed using TORAS for precise trait measurements of elytral length (red line), basal pronotum width (purple), maximum elytral width (green), and scale bar (blue). Bottom (Vial Workflow): Ethanol-preserved beetles, arranged on gridded boards and imaged, are uploaded to the Notes from Nature project, where individual beetles are annotated and labeled for measuring elytral length (green) and width (blue). Enlarged depictions of trait annotations on fictional individual beetles are provided in \ref{['fig:Fictional-Traits']} to illustrate the measurement locations more clearly.
• Figure 4: Artificial depiction of trait measurement "recipes" for pinned and vial specimens. (a) Pinned specimens: left: Fictional group image of pinned beetles with scalebar, middle: one individual specimen with three traits measured (elytral length in red, basal pronotum width in blue, and maximum elytral width in green), right: measurement of the centimeter scalebar; (b) Vial specimens: left: Fictional group image of vial beetles with checkbox, middle: one individual specimen with two traits measured (elytral length in red, and elytral width in green), right: measurement of the centimeter checkerbox. Images shown in this figure were generated using Microsoft CoPilot.
• Figure 5: Overview of the individual specimen segmentation workflow. Left: the original group image is processed with Grounding DINO using the prompt "a beetle." Middle: detected beetles are labeled with bounding boxes. Right: each beetle is cropped and saved as an individual image after manual review and (potential) correction of the bounding boxes.