Fire moves through a Sierra Nevada forest in a matter of hours. Working out what happens to the birds living there takes a lot longer, and normally relies on small teams of surveyors walking transects and counting calls by ear, one patch of forest at a time. Across a mountain range that runs for more than 600 kilometres, that approach can only ever sample a sliver of the landscape. According to researchers at the Cornell Lab of Ornithology, the fix is to stop sending people into the forest to listen, and instead leave thousands of microphones behind to do it continuously.
A forest wired for sound
The project, run through Cornell’s K. Lisa Yang Center for Conservation Bioacoustics with the US Forest Service and university partners in Wisconsin, Oregon and California, placed recording units at more than 1,600 sites spread across roughly six million acres of the Sierra Nevada. That is a meaningful share of the whole bioregion, not a fenced-off experimental plot, spanning mixed-conifer forest down through areas shaped by decades of fire suppression and, more recently, by prescribed burning and mechanical thinning.
Those units more than 700,000 hours generated of audio. Laid end to end, that is roughly 80 years of continuous listening, though nobody sat through it. The recordings were run through BirdNET, a machine-learning tool built by the Yang Center and Germany’s Chemnitz University of Technology that can pick out individual bird species from a soundscape full of wind, insects and other animals. It is that automation, not the microphones themselves, that made a project of this size possible. A human crew could never process 700,000 hours of tape in useful time, let alone repeat the exercise across an entire mountain range.
Lead researcher Kristin Brunk, a postdoctoral associate at the Yang Center, has described the scale of coverage as allowing much stronger inferences about where species actually are, at a fraction of the cost of conventional field surveys.
What the recordings actually showed
The published study, led by Brunk with Cornell’s Connor Wood and a group of co-authors from the Forest Service and several universities, appeared in Frontiers in Ecology and the Environment in March 2025. It focused on ten bird species chosen as indicators of forest condition, including the California spotted owl and several woodpeckers whose presence or absence tends to track particular forest structures, things like canopy cover, canopy height and how densely packed the trees are.
Here the brief we were given needs a correction. This was not a before-and-after study of who kept singing once a specific prescribed burn had gone through. What the researchers actually built, using all that acoustic data, was a set of bioregion-wide maps predicting where each of the ten species is likely to occur based on current forest structure. Those maps are meant to be used going forward, to help land managers weigh where thinning or prescribed fire might help or harm a given species before they light anything or send in a mechanical harvester.
Spotted owls, woodpeckers and mountain quail are confirmed as three of the ten species, the last named in the journal’s own volume summary. The remaining seven have not turned up in any source we could verify, so we are not going to guess at them here.
What this does and doesn’t tell us about recovery after fire
A habitat-association model like this can flag, with unusual statistical confidence, that a species such as the California spotted owl tends to favour denser, taller canopy, while certain woodpeckers show up more often in stands that have already burned or thinned. That’s useful for planning. It is a different thing from tracking one patch of forest before a prescribed burn and again afterwards to see which birds stayed.
Separate research gives a partial answer to that second question, and it is worth mentioning because it is sometimes confused with the Cornell project. A study of point-count surveys in Yosemite and Sequoia and Kings Canyon national parks, published in Fire Ecology in October 2025 by Chris Ray, Rodney Siegel and colleagues, tracked 42 bird species across two decades of survey data layered against burn history going back 35 years. It found that, in aggregate, bird populations rose quickly after fire and stayed higher in burned areas for decades, with 11 species showing a clearly positive and durable response. That study used a different method, in a different set of protected areas, and it was not built on the acoustic network at all. The two projects point in a similar direction, that Sierra Nevada birdlife is not simply a casualty of fire, but they are not the same evidence.
What the Cornell bioacoustic network adds is coverage. It cannot yet tell a manager exactly what will happen to spotted owls in one specific stand after one specific burn. What it can do is describe, across six million acres and with a level of detail no survey crew could match, which forest conditions these ten species are actually found in right now.
A large, useful, unfinished picture
That distinction matters for how the project should be read. Think of it as a very large baseline map, built cheaply and repeatedly by machines that never get tired of listening, against which future changes in the forest, burned or not, can eventually be measured — not a verdict on prescribed fire.
The 1,600 microphones are still out there. The next several years of recordings are what will show whether these maps hold up as more of the Sierra Nevada gets thinned and burned on purpose.