Artificial Intelligence vs Corvid Intelligence

Will AI outsmart corvids to protect crops? Agroscope is currently testing a system for detecting and scaring away these birds. The aim is to make the system intelligent and adaptive via an interdisciplinary project to be supported by the FOAG from 2026.

Carrion crows and rooks can cause extensive damage in crops, particularly in freshly sown fields of maize and sunflower. Known for their sharp intelligence, these corvids quickly become habituated to currently available deterrence measures such as balloons and acoustic or visual signals. Even control shootings, which are often backed by producers, have not proven effective.

Using AI to outsmart corvids

Agroscope is currently testing a new, artificial intelligence-based approach. To prevent the birds becoming habituated to the deterrent measures, the number of acoustic signals must be reduced to a minimum and be used only if essential. This, however, is only possible with a smart system. "Crops are only at risk for a few weeks, but corvids learn very quickly… so we need to find a way to be quicker than their ability to adapt to the measures" explains Thomas Anken, Head of Agroscope's 'Smart Technologies' Experimental Station.

Monitoring camera + horn combo

Deterrence prototypes developed by Agroscope that combine automatic bird detection with acoustic signals are currently being tested on various Swiss crop fields with a view to their improvement.

Development of a truly 'smart' system - i.e. one capable of adaptation - will begin in 2026 as part of the 'ETHIC' project (described below).

Interdisciplinary project to be funded by the FOAG from 2026

Slated for launch in 2026, the 'ETHIC' project (Exploration of Techniques, study of Habituation mechanisms and Artificial Intelligence Contributions to optimise corvid deterrence in an agricultural setting) will run for five years. The project was submitted by the University of Neuchâtel and is funded by the Federal Office of Agriculture and coordinated by Agroscope (Alice Baux).

The 'ETHIC' project will involve different research fields and institutions beyond Swiss borders. Artificial intelligence techniques will be used to adjust the deterrence signals according to the species present and their relative frequency.

A comparative ethological study will identify corvid warning calls according to species and geographic zone (local 'accents') and analyse their habituation mechanisms.

In addition to smart deterrence, dissuasive techniques will be evaluated: seed treatment with denaturants, undersowings to camouflage the seed rows and the use of different types of scarecrows. By analysing bird behaviour in response to these measures it will be possible to identify the best crop-protection system.

Participating institutions and cantons

Agroscope, the University of Neuchâtel, National Museum of Natural History (Paris), Proconseil, AGRIDEA and the Cantons of VD, GE, FR, AG, SH, TG, BE.