Project Goals

 

1. A Reusable Blueprint for Listening to Wildlife (Methodological Innovation: an AI + Ecology Framework) We're not just studying wolves — we're building a method other scientists can reuse. Our goal is a generalizable pipeline: a repeatable, step-by-step system that carries field recordings all the way to AI analysis, and that researchers can point at entirely different animals, from songbirds to whales to elephants. Wolves are our proving ground, but the real aim is a framework that advances bioacoustics across the tree of life (other taxa), not a one-off case study.

2. Does Every Wolf Have Its Own Voice — and Is It Inherited? (Individual Identity, Kinship, and the Evolution of Vocal Signatures) Just as you can recognize a friend the instant they speak on the phone, each wolf may carry its own vocal signature — an acoustic fingerprint. Because Yellowstone's wolves have decades of documented family trees (pedigrees), we can ask deeper questions: Do related wolves sound alike? Is a wolf's "voice" shaped by its genes, its upbringing, or its experience? By linking who a wolf is, and who it's related to, with how it sounds, we can study how individuality in communication evolves.

3. Counting and Mapping Wolves by Ear (Population Modeling and Occupancy) How many wolves are out there, and where are they? Those answers have traditionally required sightings, collars, or capture. We're testing whether sound can help deliver them. Using population modeling and occupancy analysis — statistical tools that estimate how many animals there are and which areas they actually use — we can turn thousands of hours of recordings into reliable estimates of wolf numbers and territory, in some cases without having to see or handle an animal.

4. Mapping the Pack's Social Web from Sound (Social Network Dynamics from Sound) Wolves are intensely social, and their calls are conversations: who howls, who answers, and when. By tracking these exchanges across our network of recorders, we can reconstruct the relationships within and between packs — essentially a social network rebuilt from sound. This shows us how packs are organized, how they hold territory, and how those bonds shift over time.

5. What Do Howls Actually "Mean"? (Behavioral Context and Meaning) A howl is never just a howl. Much as a word's meaning depends on its context, the same call might signal a reunion, a warning, a territorial claim, or the start of a hunt. By pairing our recordings with direct observations of what wolves are doing — and where, when, and with whom — we can begin to decode the function behind specific calls. And because we work at scale, across thousands of events rather than a handful of anecdotes, the patterns we find can be tested rather than just guessed at.

6. The Soundscape as Shared Territory (Information Landscapes and Acoustic Hypervolumes) Animals don't only compete over space and food — they also share an information landscape made of sound. This goal extends a classic idea in ecology, Hutchinson's niche theory, which pictures a species' place in nature as a multi-dimensional "space" defined by everything it needs to survive. We add sound as one of those dimensions. By studying how predators like wolves and coyotes use — and eavesdrop on — this acoustic space, we aim to understand how rivals share one landscape and still manage to coexist.

​​

Pursuing questions like these across an entire ecosystem takes tools that can listen everywhere, all the time. Wildlife has long been monitored with trail cameras, but increasingly sound is proving to be an efficient approach. Passive Acoustic Monitoring (PAM) — surveying wildlife by continuously recording the sounds of a landscape — is a cost-effective, noninvasive method that often outperforms camera traps in detection probability. Wolves make the case nicely: they're loud (think "car horn" loud) and can be heard from miles away, so a single recorder can cover ground that would take many cameras. Bioacoustic monitoring isn't new, but modern hardware and AI-driven software have sharply reduced its costs and boosted productivity in telemetry work.

 

To that end, our team developed GrizCam, an AI-powered multisensory recording unit — a single rugged device that captures panoramic video, spatial audio, and species detection in some of the most remote and demanding terrain on earth.

​

Just as important, GrizCam is built to take the most expensive, tedious parts of wildlife monitoring off people's plates. Traditional field monitoring is labor-intensive: technicians often hike to remote sites to swap batteries and pull memory cards, then spend hundreds of hours manually scrubbing through footage and audio — most of it empty frames or false triggers. GrizCam automates that work. Its onboard AI filters false positives like snow or waving grass, filtering out the false positives that normally flood these systems like snow or waving grass; long-range HaLow private networks or Satellite networks send compressed data and monitor device health, cutting the number of trips into the backcountry and speeding up data processing.

 

Our software tool then leans on AI classifiers to handle the first-pass sorting and labeling — and that software isn't locked to our hardware: it's built to work with acoustic recorders and trail cameras from other manufacturers, too. For wildlife managers, that adds up to lower costs and far more time spent making decisions instead of managing data.

​

Those capabilities now serve four interconnected conservation goals:

1. Advancing field science by giving biologists unprecedented tools to study elusive wildlife

2. Incenting private landowners to steward habitat through automated, AI-verified wildlife monitoring — and helping reduce predator-livestock conflicts through science-backed acoustic deterrence

3. Combating wildlife crime by giving enforcement agencies real-time detection and geolocation of illegal activity in the field

4. Building the digital wild by creating immersive connections between people and the natural world

​

Throughout, we follow the ethical framework outlined by the NYU MOTH Program.