Where the idea began
Project S.W.A.N. — the Scientific Wildlife Acoustic Network — began with a simple question: could we build a practical, regional network that listens for birds every day and makes the results useful to ordinary people?
I wanted something larger than a single garden recorder, but more accessible than a specialist research project. The aim was to create a network that could be hosted by schools, academy trusts, nature reserves, councils, community groups, businesses and other organisations, with each station contributing to a wider picture of local bird activity.
The motivation was not to replace birdwatchers, ecologists or professional surveys. It was to create another layer of evidence and engagement: a way of listening consistently, over long periods, across many different places.
For anyone interested in Norfolk birds, sound offers a revealing starting point. Birds are heard before they are seen. A song at dawn, an alarm call from a hedge or geese passing overhead can reveal activity that a person might otherwise miss. Passive acoustic monitoring turns that soundscape into data that can be explored, questioned and, where necessary, verified.
That idea became the foundation of Project S.W.A.N.
Why build a network rather than a single station?
A single monitoring station can tell you something about one place. A connected group of stations can begin to reveal patterns.
The value comes from comparison. How does bird activity differ between a school site and a nature reserve? Which species are likely to be detected in built-up areas? How does the soundscape change through the seasons? When does the dawn chorus begin at different sites?
The vision behind Project S.W.A.N. is regional and community-driven. Each station acts as one listening point, while the network allows those observations to become part of something broader.
For schools, that means access to local data that can support lessons in science, computing, geography and environmental studies. For communities and host organisations, it creates an understandable way to notice and discuss nearby wildlife without expecting everyone to become an expert in acoustic ecology or engineering.
From the workshop to the wild
My working background is not in a university laboratory. My day-to-day career is as a heavy vehicle and PSV mechanic.
Modern fleet maintenance is not simply about replacing worn parts. It involves tracing intermittent faults, interpreting diagnostic data, testing circuits, understanding control modules and working out why several individually functioning systems are not behaving properly together.
It also involves electronics, radio-frequency tracking and the practical reality of keeping equipment working in difficult conditions.
Those skills translated naturally into Project S.W.A.N.
An autonomous wildlife station is, in many ways, another piece of field equipment. It has a power supply, a computer, sensors, communications hardware, cabling, protective enclosures and software. Every part must work together, and a failure in one area can affect the whole system.
The workshop teaches you to approach problems methodically. Is the fault caused by power, signal, software, wiring, heat, moisture, configuration or physical damage? Is the problem constant, or does it appear only under load? Can the system recover safely after a failure?
That way of thinking became central to the network. Project S.W.A.N. sits at the intersection of practical, blue-collar engineering and environmental conservation. It is built on the belief that hands-on technical skills can be repurposed for public science.
Starting with real-world constraints
It is easy to imagine a wildlife station working perfectly on a desk. The harder question is whether it will still work after weeks or months outdoors.
A remote monitoring station may have no mains power, no fixed internet connection and no technically trained person nearby. It must cope with rain, temperature changes, condensation, weak mobile signal and periods of limited sunlight.
The station cannot depend on daily visits, continuously upload huge amounts of data through a poor mobile connection or waste limited battery power. It must also treat captured audio carefully, especially near schools or community spaces.
These constraints led to an off-grid, privacy-first design.
The goal is not to build the most complicated station possible. It is to build one that is understandable, maintainable and capable of working reliably with minimal intervention.
Why Raspberry Pis became the core
The Raspberry Pi was a natural choice for the main onboard computer.
It is compact, comparatively power-efficient and flexible enough to run a full operating system. It can manage audio capture, process recordings, store information temporarily and communicate with the Project S.W.A.N. platform.
The station must do more than record sound. It has to decide what to do with the audio it captures.
The Raspberry Pi allows machine learning software to run directly at the station. This is called edge processing because the analysis happens close to the source of the data, rather than sending everything to a distant server first.
Running recognition locally gives the station more control. It can analyse short pieces of audio, identify possible bird sounds and prepare only the relevant information for upload. It can also continue working when the mobile signal is unreliable, with data queued until a connection returns.
Machine learning as a filter, not an oracle
The recognition software is one of the most interesting parts of the system, but its role must be described honestly.
Machine learning compares patterns in a recording with patterns learned from labelled bird sounds. It can then suggest which species may have been present and assign a confidence score.
That output is a likely detection, not automatic proof.
Wind, traffic, machinery, overlapping calls and similar-sounding species can all cause mistakes. Even a high-confidence result should be treated in context, particularly if the species would be unusual or important.
Project S.W.A.N. uses automated recognition as a filter and guide. It turns a continuous soundscape into manageable events and highlights recordings that may deserve closer attention.
Important records should still be reviewed and verified appropriately. The system supports ecological observation; it does not replace professional judgement or formal survey work.
Privacy-first edge filtering
Processing audio on the Raspberry Pi also supports the privacy-first approach.
Rather than sending every recording away from the site, the station can examine audio locally and apply rules before anything is uploaded. This is especially important where people may be nearby. A microphone intended for wildlife should not become a tool for collecting unnecessary human conversation.
Edge-filtering software allows the station to focus on likely bird activity, minimise the amount of audio transferred and apply privacy checks as close to the microphone as possible.
The principle is straightforward: collect what is useful for wildlife monitoring, avoid retaining what is not needed and make the system's limitations clear.
Privacy is not an extra feature added at the end. It has influenced the engineering decisions from the beginning.
Reaching remote sites with GSM
A regional network cannot assume every host location has reliable Wi-Fi.
Stations may be positioned away from buildings, at the edge of a field, in school grounds or on community land. GSM modules provide the link between those stations and the central Project S.W.A.N. platform by communicating through the mobile network.
This connection can send likely detection data, selected audio where appropriate and basic station health information. It can also support remote configuration and software updates.
Signals weaken, networks drop and rural locations may have limited coverage. The station therefore tolerates interruption: data can be held locally, retried later and uploaded selectively when a connection returns.
This is another area where experience with fleet tracking and radio-frequency systems becomes useful. Real communications systems are rarely perfect. Good engineering is about designing around that reality.
Why the microphone matters
The computer and software receive a great deal of attention, but the quality of the data begins with the microphone.
Project S.W.A.N. uses external shielded microphones. Placing the microphone outside the main enclosure allows it to listen more clearly while the computer, battery and communications equipment remain protected.
Shielded cabling helps reduce electrical interference that could introduce unwanted hum or noise into the recording.
Microphone placement affects everything that follows. A station close to a road will hear a different soundscape from one beside dense vegetation. Height, nearby walls, wind exposure, water, machinery and human activity can all influence the recordings.
Station data must therefore be interpreted with context. A lack of detections does not necessarily mean a lack of birds, and a busy acoustic environment can make recognition more difficult.
Designing for solar power
To make the network useful in remote and off-grid locations, stations need to operate without depending on mains electricity.
Solar power makes that possible, but it also introduces engineering decisions.
The panel must generate enough energy for recording, processing and communications. The battery must store enough power for night-time and poor weather. The computer must use energy responsibly.
This is not simply a case of attaching a panel and hoping for the best.
Power use changes depending on what the station is doing. Recording, machine learning analysis and mobile data transmission each place different demands on the battery. Winter is different from summer, while shading, panel angle and location all matter.
The design must balance performance with efficiency. The station should do enough processing to produce useful data without wasting power on unnecessary tasks.
That is grassroots engineering in practical form: measuring, testing, adjusting and improving until the system works in the environment for which it was built.
Building something maintainable
One of the biggest lessons from vehicle maintenance is that equipment must be serviceable.
A system can be technically impressive but still fail as a real-world project if it is difficult to diagnose, repair or update.
Project S.W.A.N. stations are therefore being designed for maintenance. Software should report its status clearly, components should be replaceable where possible, and updates should be delivered remotely and safely.
The network also needs to know the difference between a quiet wildlife period and a technical problem.
If a station stops reporting detections, that could mean fewer birds are calling. It could also mean the microphone is damaged, the battery is low, the storage is full or the mobile connection has failed.
The intention is to make the stations hands-off for hosts, not invisible to the people responsible for maintaining them.
Citizen science without pretending certainty
Project S.W.A.N. is a citizen science initiative because communities are not simply an audience for the results. They are part of the network.
Host sites make the monitoring possible. Teachers can use the data with pupils. Local groups can explore seasonal patterns. Volunteers and knowledgeable observers may help review notable detections. Members of the public can learn to question how evidence is produced.
Good science is not just a collection of answers. It is a process of asking whether the method was suitable, whether the equipment was working and whether another explanation is possible.
Automated acoustic monitoring is especially useful for teaching this. A likely detection can lead to questions: Was the recording clear? Could another species sound similar? Was there background noise? Has the result been reviewed?
By publishing understandable information while being open about uncertainty, Project S.W.A.N. can help people engage with science as it actually works.
Why this project matters to me
Project S.W.A.N. brings together practical problem-solving, wildlife, engineering and the belief that useful science should be accessible.
I wanted to build something that could move beyond one person's hobby and become shared infrastructure for schools and communities.
There is something powerful about taking skills developed in a workshop — diagnostics, electronics, fault-finding, communications and system reliability — and applying them to a network that listens for birds.
The work is different, but the process is familiar. Understand the problem. Break it into systems. Test each part. Learn from failures. Improve the design. Keep the final purpose in view.
For Project S.W.A.N., that purpose is to help more people notice, explore and understand the avian life around them.
The road ahead
Building a regional acoustic network is not a single finished task. It is an ongoing engineering and community project.
Stations will need to be tested in different conditions. Recognition settings will need review. Power systems, connectivity and enclosures will need to prove themselves over time. The way results are explained will matter just as much as the hardware collecting them.
The network must also remain cautious about what the data can and cannot show.
Passive acoustic monitoring can provide valuable indicators of bird activity. It can support education, reveal patterns and help direct attention. It cannot, by itself, establish the complete ecological condition of a site or replace the work of trained professionals.
That honesty is part of building something trustworthy.
Project S.W.A.N. began with a desire to create a large-scale, accessible way of capturing local avian data. It is being built through practical engineering, shaped by privacy and designed for participation.
At its heart, it is a simple idea: place reliable listening stations in the places where people live, learn, work and care for the environment, then use those stations to help communities hear their local wildlife more clearly.
The technology makes the network possible. The people who host it, question it and learn from it are what will give it meaning.
