Bird Population Monitoring Solutions by Audio Identification and AI

Technical articles - Biodiversity

Biodiversity Birds Bats AI Audio

Bird the Virgin

Bioacoustic monitoring consists of recording and automatically identifying sounds emitted by animals (mainly birds or bats, but also insects and frogs), in order to provide information about the species present in a given area.

This article explains why it may be interesting to implement such monitoring on your land, presents some available tools, and describes how to use them if you decide to take the step.

What are the advantages of bioacoustic monitoring for land managers?

Many land managers, curious to learn more about the wildlife on their plots, are interested in it. But there are also more practical reasons why this type of monitoring should concern all managers.

With the evolution of public policies, which put more emphasis on producing environmental benefits from agriculture, and the gradual arrival of private funding in emerging natural capital markets, it becomes essential for land managers to be able to measure biodiversity in a standardized way, in order to establish baselines and track changes. Bioacoustic monitoring, using birds and bats as biodiversity indicators, could be a solution.

Why record animal sounds rather than images?

Animal songs and calls, with their distinctive frequency patterns, lend themselves better to automated identification by machine learning than images, which can lack consistency due to lighting conditions or the posture and behavior of animals.

Unlike other remote monitoring tools such as camera traps, which require the animal to pass in front of the camera, acoustic sensors collect data over a wider area.

What equipment is needed?

Audiomoth

Many know the application Merlin Bird ID, which allows identification of birds from a smartphone microphone, or the ultrasound detectors used to listen to bats. But these methods require the presence of an observer and risk missing some species, especially those that vocalize little or at unusual hours.

Increasingly, passive and autonomous acoustic recorders are used to collect data continuously, day and night, and over long periods.

These specialized devices cost between just over €100 and more than €1000. Among affordable models are the Audiomoth and the Song Meter Micro 2:

• Audiomoth: configurable to record either birds or bats (but not both at the same time).
• Song Meter Micro 2: records only audible sounds (not bat ultrasounds).

Recordings are saved on micro-SD cards and must be analyzed by specialized software. This requires some computer skills, which has led to the emergence of new simplified solutions, presented below.

Simplified options for bird recording

Autonomous recorders like the Audiomoth and Song Meter offer great flexibility but require technical skills. To simplify the user experience, new options have appeared, often more expensive or requiring a subscription.

• Chirple Chorus: a recorder integrating directly the AI for identifying European birds. It requires an Internet connection (Wi-Fi or 2G), which implies a subscription in addition to the initial cost. Solar panels exist for isolated sites.
• Chirrup Nano: offered as a service rather than a product to buy. Recorders are temporarily installed on a farm, then collected. Results are delivered via a comparative dashboard. Currently, this solution targets companies rather than individuals.

How to configure passive recorders?

• Audiomoth: configuration via computer and USB cable.
• Song Meter Micro: configuration via a mobile app over Bluetooth.

You need to define:

• the sampling frequency (48 kHz for birds, 250 kHz or more for bats),
• the frequency and duration of recordings (for example, 1 minute every 10 minutes),
• the times of day (morning for birds, dusk for bats).

The monitoring objective also determines the chosen time of year:

• breeding birds → more active between March and June,
• bats → active throughout the warm season.

How to deploy the recorders?

The choice of location depends on the objectives:

• detect a specific species → place in its habitat,
• establish a complete list → move or multiply recorders,
• monitor biodiversity changes → arrange a regular grid.

Recorders are generally fixed between 1 and 2 meters above the ground, on a tree, stake, or pole.

How to analyze the results?

Recordings can represent several hours of data. The most effective way is to analyze them with AI software.

• BirdNET (developed by Cornell, like Merlin) is one of the most used, but its interface is not very intuitive. Solutions like Chirpity facilitate usage.
• It is important to limit species to the studied area, otherwise errors appear (exotic species).
• BirdNET provides a confidence score (%) for each identification. Results must be filtered and rare species verified by an expert.

Other solutions exist:

• BTO Acoustic Pipeline (free service if data is shared with the BTO, useful especially for nocturnal species or monitoring Curlew).
• For bats: identification possible via BTO, but with free volume limits. For large amounts of data, a paid service like Kaleidoscope is needed.

Limitations of bioacoustic monitoring

• Vocal species are overrepresented: a loud and common bird will be detected more often than a discreet species, even if their populations are similar.
• Weather conditions also influence vocal activity.
• Thus, bioacoustic data mainly serve to compare the abundance of the same species over time or between sites, but not to compare different species with each other.

Long-term monitoring helps reduce biases related to specific conditions.

Further reading

• British Ecological Society – Good Practice Guidelines
• Bird Survey Guidelines for acoustic bird surveys

Sources

• The original version of this article (in English) can be found here: https://www.fas.scot/article/bio-acoustic-monitoring-solutions-for-land-managers/