Project

Automated monitoring of ground dwelling arthropods

A biodiversity-focused field project that tested automated pitfall-style monitoring and image-based recognition as a more scalable way to observe soil arthropods in agricultural systems.

Overview

Soil dwelling arthropods such as ground beetles, rove beetles, and spiders are relevant for biodiversity assessment and pest management in arable fields. They are commonly monitored with pitfall traps, but that process is labor intensive and depends on specialist identification.

This project explored how a prototype automated pitfall trap could capture useful images before the organism fell into the trap and how those images could later support automated recognition workflows.

  • Duration: 12 months - January 2023 till December 2023
  • Project number: KVW - 00616
  • Funding Agency: Kansen voor West - European Regional Development Fund
  • Partners: Space Value, Unip. Lda and Stichting Wageningen Research

Practical goal: reduce the operational burden of biodiversity monitoring by combining improved field hardware with automated image recognition.

Focus

The project tested whether automated capture and recognition could make biodiversity monitoring more practical for researchers and farmers.

  • Approach Test the hardware in the field, improve the device where needed, collect annotated image data, and compare the system with traditional methods.
  • Recognition performance 97% overall accuracy on the available dataset.
  • Outcome Strong validation of the concept, with clear directions for future refinement.
Back to Projects
Space Value
Wageningen
Co-funded by the European Union
Kansen voor West
Work Packages

Core Areas Of Work

Optimization and testing

Several alternative designs were tested to determine which version was most suitable for attracting and recording soil dwelling arthropods.

Recognition model training

Hundreds of annotated images were used to train a model capable of estimating catch frequency and species richness with less manual work.

Monitoring implications

The new system was compared with ordinary pitfall traps to understand how well it records activity-density and species diversity in the field.

Key findings

The i2MS device produced qualitatively good pictures, especially in daylight, where both small and large arthropods could be recognized. However, the number of recorded pictures was lower than expected even when arthropods were being caught.

For several groups, including rove beetles, spiders, myriapods, springtails, and woodlice, the device captured a diversity comparable to a regular pitfall trap.

Conclusion

Ground beetle comparisons were less conclusive due to seasonality and activity levels, which means further trials are still valuable.

Even so, the model trained on the available dataset achieved an overall accuracy of 97%, demonstrating strong potential for practical biodiversity monitoring applications.

Why it matters

  • Monitoring quality Useful image capture already supports real scientific value.
  • Operational learning The project highlighted exactly where hardware capture still needs refinement.
  • AI potential The recognition results confirm strong promise for future biodiversity workflows.