How a fieldwork challenge sparked a new way to think about occupancy models – Methods Blog

Post provided by Vitek Jirinec

In 2013, I found myself knee-deep in the marshes and forests of southeastern Virginia, USA, embarking on what would become an unpredictable journey in my career. Fresh into my graduate thesis at the College of William & Mary under the guidance of Matthias Leu, the plan was studying the habitat use of Wood Thrushes—a species that is often researched, but as I would come to learn, doesn’t always cooperate with research plans.

The start of something unexpected

To earn my keep with Matthias, I also conducted his annual point count surveys, but the heart of my thesis was tracking Wood Thrushes using radio telemetry (Figure 1). Besides focusing on habitat use, we had grand ideas of analyzing how bird movements compare with detections at point count stations, so we tagged territorial males near stations and planned to overlap tracking and counts.

Centered around Williamsburg, our study area was a mosaic of parks, urban sprawl, rural pockets, and even restricted military bases like Camp Peary that hosts a covert CIA training facility known as “The Farm”. It was an exciting challenge, and my initial focus was on parks, where I had to navigate a maze of permissions just to get research access.

Then the birds started moving—far more than I had anticipated. Contrary to the general idea that birds have static breeding territories, Wood Thrushes we tagged would often disappear from one spot, only to pop up several hundred meters, sometimes kilometers away. Given the limited detection radius of the radio tags (a few hundred meters), this was a nightmare. Tracking them across such a fragmented landscape felt like chasing ghosts. But this was necessary—no birds, no habitat data, no thesis.

The great bird hunt begins

So began a wild thrush chase. Most often, tracking involved countless hours roaming the woods with a hand-held directional antenna (called a Yagi) connected to a receiver that looked suspiciously like a blue beeping lunchbox. For longer forays, I rigged our lab truck with radio antennas (Figure 2), biked around with a whip antenna stuffed in my backpack (the “Sputnik method”), and even canoed through marshes, often in the dead of night (Figure 3). Why at night, you ask? Because Wood Thrushes tend to roost higher up in trees after dark, which elevates the tags just enough for a stronger signal. Those late-night paddling expeditions felt more like searching for UFOs than birds, but they often worked.

One of my most memorable moments came when an undergraduate assistant, armed with a Yagi antenna, inadvertently spooked a suburban homeowner, who was convinced we were trying to hack his WiFi. It all ended well, though—it turned out he was a William & Mary history professor who, in a former life, had been an IT guy. We laughed it off, but moments like this highlighted just how bizarre fieldwork could get. In fact, with all our unusual setups and nighttime tracking, it’s surprising we only had three run-ins with the police—each of which, thankfully, ended without any trouble.

But not all places were as easy to access as suburban backyards. The real problem was tracking birds that decided to move into more restricted areas—especially those military bases. That’s where things got really interesting.

Enter captain Fuzzzo

At some point, after exhausting the ground options, we decided we needed to go airborne. Enter Captain Fuzzzo, a Vietnam War Air Force pilot with a small Cessna and a flair for adventure. I got his business card (a blank paper that only had a phone number, email, and the words “Fuzzzo. Oracle, Soothsayer”) from the local Center for Conservation Biology that used his pilot services for decades to survey Bald Eagles in the Chesapeake Bay. But Fuzzzo had never done aerial telemetry before—and neither had we—so we did what any resourceful field biologists would do: we strapped antennas to his plane with duct tape and bike inner tubes (Figure 4). Fuzzzo seemed confident it would hold, so off we went.

Flying with Captain Fuzzzo was an experience in itself. Fuzzzo had no qualms about buzzing military and CIA bases to help us find the elusive Wood Thrushes. And thanks to him, we located several missing birds, including one that had flown about 5 kilometers and settled within the Naval Weapons Station Yorktown. To my surprise, after a background check, the Navy granted me clearance to drive the antenna-rigged truck into the Weapons Station to continue collecting habitat data on the bird. I even received a restricted area badge, which I proudly flashed at the bewildered soldiers guarding the entrance and patrolling inside.

Despite our best efforts, however, some birds were never found, and the birds that we did manage to find often crossed the detection radius of more than one point count station, including one bird that was detectable at 3 separate stations. These territory shifts raised an important question: can we accurately assess a population of birds that refuse to stay put? Specifically, the problem materializes when using occupancy models, which are built on the assumption that the study site remains “closed”—in other words, the birds don’t move in and out during the study period. If closure is violated, occupancy models yield skewed estimates.

AOS 2018

Fast forward to the American Ornithological Society (AOS) meeting in 2018. During my talk, I complained about this very issue, showing that closure was clearly violated in the Virginia study. One slide read: “Survey methods are often blindly applied without considering the ecology of target species and its effects on corresponding estimates.”

During the Q&A session, someone stood up and asked a question that, while I don’t recall the exact words, was something along the lines of: “Okay, so instead of just pointing out flaws in these popular occupancy models, what can we actually do to make them work for us?” It was a push to think beyond criticism and consider how we could fix the issue. That someone turned out to be Jonathon Valente, a sharp quantitative ecologist who had just finished his PhD at Oregon State.

From problem to paper

That question sparked a years-long collaboration with Matthias and me, culminating in our 2024 paper, recently published in Methods in Ecology and Evolution. Initially, we set out to see how occupancy models could be adjusted to account for the shifting territories of mobile species like the Wood Thrush. But as we dug deeper, our focus broadened. We realized the problem wasn’t just limited to Wood Thrushes or mobile animals—it touched on fundamental issues with how occupancy models are applied across many species and studies. We began examining how different survey protocols, spatial scales, and even the definitions of occupancy itself could drastically alter the accuracy of model estimates. Ultimately, we sought to provide guidance for designing surveys that yield biologically meaningful results, regardless of whether the closure assumption holds.

What we discovered

Our research showed that occupancy estimates are highly dependent on the details of the survey protocol, particularly when dealing with animals that move around a lot. We used simulations (an individual-based model parametrized with the Virginia data; Figure 5) to test 162 different survey protocols and found that the results varied widely based on factors like survey radius, the number of surveys, and the time between them. Here are some key takeaways:

• Not all occupancy estimates are created equal: You can’t compare occupancy estimates between studies unless the survey protocols are very similar. Just because one study reports a higher or lower occupancy rate doesn’t mean much unless the methods are aligned.
• Occupancy has many definitions: It matters whether you’re talking about instantaneous, daily, seasonal, or some other type of occupancy. Researchers need to define what they’re measuring upfront.
• Match the survey to the species: For mobile animals like Wood Thrushes, survey designs need to take their behavior into account. A one-size-fits-all approach just doesn’t work.

Why it matters

This research is more than just an academic exercise. It has real implications for how we monitor and conserve species. Occupancy models are widely used in conservation to assess species distributions, abundance, or habitat use. But if the model assumptions don’t match the biology of the species being studied, the results can be misleading.

For Wood Thrushes—and likely many other species—we showed that survey designs need to be thought through. The biology of the species should dictate the survey design, not the other way around. By making these adjustments, we can get closer to the biological truth and make better-informed decisions about conservation and management.

Final thoughts

Ten years later, the odyssey of tracking Wood Thrushes across parks, backyards, and military bases wasn’t just about following birds. It was about uncovering a deeper truth: that wildlife research needs to be as dynamic as the animals we study. Captain Fuzzzo and his trusty Cessna may have helped track down a few missing birds, but it will take a broader willingness to rethink our methods that can ultimately lead to better science.

Vitek Jirinec
Present address:
Integral Ecology Research Center
Blue Lake, CA, USA

Associated publication:
Valente, J. J., Jirinec, V., & Leu, M. (2024). Thinking beyond the closure assumption: Designing surveys for estimating biological truth with occupancy models. Methods in Ecology and Evolution, https://doi.org/10.1111/2041-210X.14439

Post Edited by Lydia Morley