AI-supported biodiversity monitoring: Automated recording of pollinators for comprehensive biodiversity monitoring in Saxony-Anhalt’s agricultural landscape (BiodivAgrar)

Pollinating insects are crucial for the stability of agricultural ecosystems and food security. The BiodivAgrar project uses innovative camera and deep learning systems to quickly and efficiently monitor wild bees, hoverflies, and other pollinators affected by decline on a large spatial scale. The project investigates how various agri-environmental measures to promote biodiversity affect pollinators and compares AI-based recording with traditional entomological methods. The quality of the results obtained from different technical systems varies depending on the species groups and the technology used. Available systems will therefore be evaluated and further developed in a large-scale field experiment (2025/2026). The accuracy of insect identification will be tested and the effectiveness of AI-based data collection will be compared with traditional recording methods. The results will advance the automated, comprehensive monitoring of biological diversity and improve measures to promote biodiversity and food security in agriculture.

Wild bees and hoverflies (left: Lasioglossum leucozonium – white-banded bee; right: Merodon equestris – common narcissus hoverfly) are among the most important pollinators of wild and cultivated plants, but they have different requirements in terms of their habitats and food plants. Both pollinator groups are severely affected by species extinction and are to be recorded in the BiodivAgrar project using automated camera systems.

Project priorities

Data collection

Establishment of a comprehensive field experiment in the agricultural landscape
Comparison of various commercially available camera systems

Analysis

Evaluation of the field data obtained using different image recognition models
Comparison with conventional entomological recording methods

Suggested applications: possibilities and limitations

Nationwide automated biodiversity monitoring
Improvement of automated systems

Project region

Saxony-Anhalt

Further project details

Commercially available camera systems are being tested on trial plots at Anhalt University of Applied Sciences (farmland with marginal flower strips and hedges) and on farmland scattered across Saxony-Anhalt, and their recordings are being analyzed using various AI platforms. Traditional trapping and identification methods (landing nets, colored bowls) are used as controls. The data analysis takes into account the availability of pollen and nectar sources in order to evaluate the performance of the systems in the context of increasing insect diversity and the landscape. The cameras are used at different locations with and without flower strips and hedge plantings.

Automated insect detection in the field using camera traps (left, center: cameras in the field, EcoEye type) and classic entomological identification methods (right: prepared wild bees in the laboratory) are being compared in the BiodivAgrar project in order to assess the performance of modern systems in the field. Images: Lucas Beseler and Simon Dietzel