Project summary

ChirpWatch-NCC is a twelve-month pilot to build a solar-powered acoustic sensor network across the METU Northern Cyprus Campus, using it to track which birds are present, when, and where, without ever touching a bird (toaha/ChirpWatch-NCC: Open solar-powered AI bioacoustic observatory for long-term bird monitoring at METU Northern Cyprus Campus.).

Cyprus sits on a major Mediterranean migration corridor, and the campus itself spans several distinct habitats, which makes it a good natural laboratory for this kind of long-term, passive listening.

Eight field stations, each running on solar and LiFePO4 storage, will record continuously and run BirdNET-style species identification at the edge a LoRaWAN gateway will carry back only the lightweight signals the project actually needs. Battery health, storage headroom, environmental readings, and high-confidence detection summaries. The raw audio stays local until it can be pulled over Wi-Fi, LTE, or a maintenance visit. The result will be an open, reproducible biodiversity baseline for the campus and a hardware/software build that anyone else with a similar site could replicate.

What are this project's goals? How will you achieve them?

The goal is to get a working, unattended sensor network that produces a year's worth of trustworthy species-activity data, alongside documentation good enough that someone else could rebuild the whole thing from the repository. The methodological goal is to prove that a campus-scale acoustic network doesn't need to ship raw audio over a low-bandwidth radio link to be useful, and that the metadata alone, if chosen carefully, can carry most of the decision-relevant information.

Eight stations are to be installed across representative habitat types, each self-sufficient on power and storage for weeks at a stretch. BirdNET gives a baseline inference layer, and a subset of detections gets manually reviewed so we know how much to trust the automated calls before publishing anything.

The LoRaWAN layer is deliberately kept thin to its telemetry and alerting, not a data pipe, which keeps power draw and airtime low enough for solar operation to actually hold up through Cyprus's winter months. Everything gets synced and reconciled centrally when connectivity allows, and the dashboard and final report get built once there's enough season-over-season data to say something real about activity patterns rather than a single snapshot.

How will this funding be used?

The estimated budget is USD 18,000, with a floor of USD 12,000 below which the pilot isn't really viable and a stretch ceiling of USD 20,000 if a forecasting work package gets added. The money is direct-cost, not overhead, as it pays for the field stations themselves (solar panels, batteries, enclosures, microphones, compute), the LoRaWAN gateway and edge-analysis hub, storage media, and the routine costs of getting people and equipment to eight sites over a year for installation and maintenance.

Who is on your team? What's your track record on similar projects?

This is a PI-led project rather than a large team, which is intentional at the pilot stage. The plan is to bring on student help for the labour-intensive parts of installation and manual validation once the deployment phase actually starts. What I can speak to directly is the technical track record behind the approach. My recent work on VISTA-LoRa (IEEE Wireless Communications Letters, 2026) tackled essentially the same design problem in a different domain: running YOLOv8s inference on a Jetson Orin Nano Super at the edge and transmitting only task-relevant semantic summaries over LoRa, rather than raw imagery. That's the same bet ChirpWatch-NCC is making with bird calls instead of images do the intelligence locally, send the decision, not the data.

I'm also midway through a related research program on decision-sufficient and belief-aware variable-rate communication for LoRa networks, currently applied to earthquake early warning, which asks the same underlying question
How much you can compress a sensor's output before you start losing what actually matters. Separately, I've been part of an ISPF-funded project (£80,000) on post-earthquake communication resilience combining UAVs, LoRa, and GDPR-compliant systems, which is where much of my practical experience with LoRa hardware and field deployment logistics comes from. None of that is bird monitoring specifically, but the hard parts, edge inference under power constraints, minimal-bandwidth telemetry design, keeping a distributed sensor network alive in the field, are the same hard parts.

What are the most likely causes and outcomes if this project fails?

The most mundane failure mode is also the most likely one that is solar sizing that doesn't hold up through Cyprus's shorter winter days, leaving stations going dark for stretches during exactly the season when migratory data matters most. The fix there is conservative power budgeting and battery headroom, but if it's underestimated, the outcome is patchy data rather than a clean year-round baseline. Another risk is on the science side, if BirdNET's detection accuracy on Cyprus's specific bird community turns out weaker than expected, the manually validated subset ends up smaller than planned, and the confidence behind the published indicators is correspondingly lower. None of these failure modes is catastrophic in the sense of losing the hardware or the funding outright; the more realistic bad outcome is a partial dataset, a network that runs less reliably than hoped, or the forecasting work package getting quietly dropped so the core monitoring goal can still be met. The hardware, documentation, and pipeline would still exist and be reusable even if the first year's data came back thinner than intended.

How much money have you raised in the last 12 months, and from where?

For ChirpWatch-NCC specifically, the honest answer is nothing yet, the repository is explicit that this is a pre-submission, funding-readiness project, and it hasn't drawn on any external funds because the institutional payment route through METU hasn't been confirmed. I don't want to blur that with unrelated funding on other projects (like the ISPF grant mentioned above) by presenting it as capital behind this proposal, since it isn't earmarked for bird monitoring and I'd be guessing at whether it falls inside your 12-month window. So, zero has been raised for this project to date, and Manifund would be the first funder if the application goes through.