What passive acoustic monitoring is
Passive acoustic monitoring (PAM) means deploying continuously-recording acoustic devices in a habitat, then analyzing the resulting audio for species presence, abundance, behavioral patterns, or community structure. 'Passive' means the device records without active intervention from researchers; the bird (or frog, or insect, or whale) vocalizes whatever it would have vocalized, and the device captures it. The device doesn't play anything back, doesn't move, doesn't disturb. PAM is the methodological inverse of playback experiments: instead of asking what happens when we send signals into the environment, we ask what signals are already in the environment that we haven't been observing.
Until roughly 2015, acoustic monitoring at any scale required expensive specialist equipment, large batteries, and frequent site visits to swap storage.
The hardware story
Until roughly 2015, acoustic monitoring at any scale required expensive specialist equipment, large batteries, and frequent site visits to swap storage. The release of the AudioMoth open-source acoustic logger in 2018 — designed by Open Acoustic Devices and produced at hardware-cost — dropped the per-unit cost of deployable acoustic monitoring by more than an order of magnitude. AudioMoth and successor devices now ship for under one hundred dollars per unit, run for weeks on AA batteries, store hundreds of hours on microSD, and weatherproof reliably. The hardware curve transformed what was possible at the network scale: deploying ten devices across a forest is now an undergraduate field-methods exercise, not a doctoral budget.
The software story
Recording hundreds of hours of audio per device per deployment is useless without analysis. BirdNET[1] — the same Cornell model that powers the Merlin phone app — runs in BatchMode on a laptop and processes a deployment's audio overnight, emitting species-confidence detections at sub-second resolution. For most deployments in most regions, 's species coverage and accuracy are sufficient to surface the questions ecologists want to ask: did this species occur here this season? How does its dawn-chorus phenology compare to last year? Are migration patterns shifting? Custom-trained models from or species-specific classifiers extend the toolkit for rarer species and finer-grained questions.
What PAM is good at
Presence and absence over time, at fine temporal resolution. Population trends across multi-year deployments. Dawn-chorus phenology — the daily start time and duration of acoustic activity, which is itself an integrated indicator of ecosystem health. Detection of rare or cryptic species that visual surveys would miss. Community-level acoustic indices (alpha diversity, beta diversity, acoustic complexity) that can be tracked over time. Continuous monitoring of remote sites without ongoing site visits. The shared characteristic: PAM is at its best when the question is 'what's there' or 'how does what's there change.'
What PAM is not good at
Individual identification at scale (still hard with single-microphone PAM data). Behavioral context (a recording shows the call but not what the animal was doing). Counting individuals reliably (multiple calls from the same bird and single calls from multiple birds are hard to disambiguate). Surveying species that vocalize rarely or quietly. Replacing ground-truth field surveys for population estimates in conservation triage. The shared characteristic: PAM is not at its best when the question requires per-individual attribution or per-utterance behavioral interpretation. For those questions, wearable loggers or focal observation is needed.
How conservation programs use it now
By 2025, PAM has become standard methodology for some categories of conservation work. Forest songbird monitoring in North America — the breeding bird survey tradition — is being supplemented or replaced by PAM grids in many study sites. Endangered-species monitoring (Kirtland's warbler, ivory-billed woodpecker rediscovery efforts, multiple cetacean populations) relies heavily on PAM. Pre-development biodiversity assessments for environmental impact reviews increasingly include PAM deployments. The shift hasn't been complete or uncontested — there are real methodological questions about how to integrate PAM data with traditional point-count surveys — but the trajectory is clear and accelerating.
Why this matters for crow research
American crows, being abundant, conspicuous, and well-studied, haven't been a focus of PAM-driven conservation work — they're not threatened. But the PAM deployment story matters for crow research indirectly: it's the largest deployment of audio foundation model technology to date, and the resulting datasets, methodological refinements, and infrastructure flow back into the corvid research community as available tooling. Every BirdNET[1] update that improves PAM detection accuracy on threatened songbirds also improves crow detection in the long-term recordings researchers are starting to accumulate for behavioral context. Conservation needs drove the deployment; the deployment lifted the broader bioacoustic research ecosystem.