Big Game Meets big data - University Of Montana · 2020. 5. 27. · Big Game Meets. big data. by Kasey Rahn. eliable olatio estimates are te dametal bildig blos o ildlie maagemet

Innovative technologies, from GPS

collars to cameras, can provide

massive amounts of data about the

local herd. But how do you harness

that information to make for better

wildlife management—and better

hunting—all across elk country?

I n 1970, John and Frank Craighead put the first satellite collar around the neck of an elk they named Monique (Moe, for short) at the National

Elk Refuge in Wyoming, opening up an entirely new way for scientists to study wildlife. The 23-pound, red aluminum collar fed Moe’s location to a Nimbus weather satellite twice a day. The information was recorded, relayed back to a tracking station in Fairbanks, Alaska, and was eventually sent to scientists at NASA’s Goddard Space Flight Center in Maryland, helping to shed light on an important migration route. Today, rapid advances in technology continue to shape the way we learn about and manage wildlife.

“We used to have Very High Frequency (VHF) telemetry, and somebody would go out with a little hand-held antenna and listen for the beeping. It was pretty slow and inaccurate,” says Paul Lukacs, who specializes in quantitative wildlife biology at the University of Montana.

First introduced in the 1960s, it was the standard PH

OTO

: DA

VID

STO

NN

ER

-MD

C

Big Game Meets

big databy Kasey Rahn

Reliable population estimates are the fundamental building blocks of wildlife management. But as any biologist who’s ever spent time in a Bell or Cessna banking 500 feet above a sage flat trying to count a river of galloping elk will tell you, it ain’t easy. Same goes for trying to pick them out of Pennsylvania hardwoods or doghair fir thickets in Idaho.

Last issue in “Counting Our Blessings,” we looked at the current challenges and solutions to measuring populations across elk country. In this second half of the series, we explore the mind-bending ways wildlife managers and computer geeks are teaming up to count elk and other wildlife faster, safer, cheaper and above all more accurately in the future.

The innovations are pairing everything from billion-dollar satellites to $100 trail cams with the massive math that underpins quantitative statistics. Backed by banks of servers massive enough to rival the wall of Clairs, Crowns and Carvers on a Stones tour, a biologist armed with nothing but a laptop may soon estimate a herd with uncanny precision. Plus plug in dozens of variables to predict that herd’s future.

On top of the article that kicks off here, check out “Navy EOD Blows up Wildlife Biology with Big Data” (page 55) and “Trail Cams Meet Lab Coats” (page 64). And for a refreshing reminder that elk hunters make great citizen scientists, don’t miss “Guess How Montana Counts Wolves Now?” (page 81). It’s a cool synthesis of old-school observation and Moneyball-worthy predictive analytics.

Without dumbing down the concepts, we’ve tried to distill some pretty abstract ideas and math-wiz jargon into clear, readable stories. Hope you enjoy.

Big Game Meets

big dataTech Revolution Rocks Elk Country

MAR/APR 2018 • BUGLE • 59

for field biologists on hundreds of species around the globe for the better part of half a century. Using VHF, a field biologist might locate a collared animal once every 30 days.

“Now with GPS data, you could have a fix every 30 seconds on the animal’s locations just pouring into your computer,” Lukacs says.

Combine GPS collars with highly sophisticated trail cameras that link to your phone with real-time photos and videos, Google Earth-style satellite photography, drones and more, and wildlife biology becomes a brave new world.

“Data management is a big part of what I deal with,” says Josh Nowak, Lukacs’s colleague in the statistics-soaked end of wildlife biology at the University of Montana. “You see an elk collar capable of taking a fix every half-hour for better than a year—you put 40 collars out and you’re talking about a million data points. You can’t just pull that up in Excel and figure stuff out, right?”

A recent remote camera project studying elk captured over a million photos in three months. That’s 1.3 terrabytes of information.

“So how do you store it? Then how do you get the right information out of it?” Nowak says. “Then how do you do that for as a few dollars as possible?”

Those are the questions that drove him and Lukacs to create a new way for biologists to harness these massive amounts of data to make for better wildlife management—and better hunting—all across elk country. They call it PopR.

When Lukacs ventured across the country to Montana as a freshman, he did not arrive dreaming of statistics.

“I wanted to go somewhere that was really far from New Jersey,” he says. “I came here to be a fuzzy mammal biologist, and I wanted to work in the field. I took two required calculus and stats classes my freshman year, and said ‘I’m never touching math again.’ Then I had a semester without math and said ‘Huh, I think I actually liked that. I miss that.’ So I filled in all of my spare credits with math and computer science classes.”

That led him further and further into quantitative wildlife biology—the synthesis of mathematics and biology that transforms big game counts into population estimates and ultimately hunting regulations. It’s a blend of old-school biology and the cutting-edge of “Big Data.“

PopR’s origin story begins a dozen years ago, when Lukacs worked as a biometrician for Colorado Division of Wildlife where he built population modeling software tailored to individual species. It planted the seed that tools like these held untapped potential, and when he returned to UM in 2011, the seed traveled with him.

Meanwhile, Nowak had metamorphosed from

a maniacal hunter to a Navy Explosive Ordnance Disposal expert to a self-professed computer and statistics geek…all while still hunting with intensity. (For the great backstory, see “Navy EOD Blows up Wildlife Biology” on page 55). After earning his Ph.D. at Universite Laval in Quebec, he ended up in Lukacs’ lab at the University of Montana designing mountain lion population modeling software for Montana Fish, Wildlife and Parks.

The circles started to overlap. They met a biologist working on sage grouse population models. A colleague from Idaho saw Nowak’s mountain lion software and suggested doing something similar for elk. South Dakota expressed interest in creating a population model for deer.

“We got to thinking that everybody’s doing the same thing, just with different species,” Lukacs says. ”We started thinking about the synthesis of the underlying population model and started building that as an interactive tool for biologists.”

The first few versions were promising but had an Achilles’ heel: they ran on individual computers. That meant there were multiple versions of both the software and the data floating around, and they might or might not work depending on each computer’s hardware and who had or hadn’t installed updates.

“You’d have a biologist who would have this clunky old laptop,” Lukacs says. “Nobody could really talk to each other because there was no consistency.”

To solve that problem, they turned to the cloud—a collective term for a digital information bank and programs that run over the internet, just like iTunes or Gmail, instead of through software downloaded directly to your computer. PopR software runs on a server housed at the University of Montana. Biologists can access it from anywhere in the world. All they have to do is pull up a web browser. It works the same way your Mac, iPhone and iPad seamlessly share music or the receipt of your most recent Amazon binge between devices.

“Having it on the web and connecting to one centralized data repository makes that so much better, because everyone’s getting literally the exact same data,” Nowak says. “All of the areas are reproducible, so we can fix things quickly when they break. And our models are the same for everybody. The same updates go to everyone, instantly.”

For those who haven’t taken a stats class in a few years, a model is simply a mathematical representation of a system.

“We all use models all the time, we just don’t know it,” says Nowak. “When you go to the checkout at the grocery store, you have to pick which lane to use. You’re going to have some preconceived notions of which lane is going to move faster because of how

60 • BUGLE • MAR/APR 2018

many and what kinds of items are in a cart. That’s a model. And we can keep going with that idea, and it can become more and more complex.

“When you use Google to find something on the internet, you’re using a model. It’s just been abstracted to a single input and a button. That’s sort of our metaphor for what we’d like to do, in terms of the information. We’d like for this to be as efficient as possible for biologists so all they really have to do is Google what’s going on with their population, and they can find out.”

PopR (based in a statistics package called R, hence the name) creates integrated population models in the same way, using fancy math to mash together data from different sources to create a population trajectory—an estimate of how the population will ebb and flow in the future.

“The gist of it is, ‘If I give this many tags out next year, what’s going to happen?’” Nowak says.

Every winter, state wildlife agencies survey elk, deer, bighorns, pronghorns and more from airplanes and helicopters to figure out how many there are and how the population breaks down between males and females, mature animals and young of the year.

“Historically, what you would do is you would go out, you’d count and you would say, ‘Okay, there’s 10,000 elk,’” Nowak says. “Then you could put collars on some, and you’d say, ‘Well survival is .89 in adult females.’ You have all these pieces of information, but they’re not synthesized in any formal way. It’s left up to each individual to mentally go, ‘All right. Well, if they survive at .89, there are 10,000, they reproduce at this rate, and we harvest this many, how many will be there next year?’”

Finding answers to those questions is expensive, time-consuming and, perhaps worst of all, inconsistent from one biologist and one hunt unit to the next. Lukacs and Nowak collaborate with state agencies to do the modeling with the data biologists

collect and put it into terms people can understand and apply more evenly on the landscape.

Data gathered from everything from hunter harvest reports to GPS collars are fed into the program. Biologists specify what they’d like to know or what question they’re trying to answer. Using computer coding and statistics, PopR determines which model is needed and which data is best for it, then spits out an answer, usually in less than a minute. It can analyze the abundance and composition of the herd, pregnancy rates, hunter mortality, predation and more to help wildlife managers look at and think about hunt units across a large landscape. And, like Google, it can do all this in real-time, instantly analyzing the probabilities. There’s even a new feature that shows real-time weather on top of current snow conditions to help biologists plan their game counting flights.

“All of these numbers come in and most of them are not useable in raw form, so we have to do something to them,” Nowak says. “We have to apply some sort of statistical model to get at the number that’s actually of interest. Those are the types of models that I write.”

Lukacs says models help avoid the mental mistakes or inconsistencies we make as people—to help quantify things accurately and to make those processes repeatable.

“Our overall goal is to provide consistency and transparency for the agencies so that they can do their job well and keep their biologists thinking about biology instead of being overwhelmed with copying spreadsheets,” he says.

In the past four years, Lukacs and Nowak have worked with half a dozen state agencies across the West, creating models for elk, bison, mule deer, white-tailed deer, mountain lions and sage grouse. They’re currently working on models for bobcats and black P

HO

TOS

CLO

CK

WIS

E F

RO

M T

OP

LE

FT:

RM

EF

/ JU

STI

N G

UD

E-M

TFW

P

PopR uses mathematical modeling to turn the data collected by boots-on-the-ground biologists—from GPS collar points to hunter harvest survey answers—into easy-to-use information to help state agencies better manage wildlife populations.


bears, and there’s the potential for grizzly bears in the future. Plus, they’re trying to build effective models to harness very sparse and widely spaced data for rare and elusive species like snow leopards and jaguars.

But there’s a lot more than just math that goes into it. Nowak and Lukacs start by helping agencies determine what data their biologists need to collect and the best ways to get it.

“We work out sampling schemes so it takes the fewest man hours and the least amount of money for them to get a number that’s as good as they need it to be,” Nowak says. “In other words, how do we make it more efficient logistically and financially?”

Nowak says the process is collaborative and requires buy-in from all levels of agency personnel to be effective. They meet with boots-on-the-ground biologists to review current sampling techniques and set goals, and they discuss species’ biology and local knowledge—like where elk are causing crop damage or where migration routes run—to glean information that will eventually become part of the model.

“That local knowledge is pretty important to the whole process. Then we try to figure out where we need to make compromises, because there’s always compromises,” Nowak says. “There are a lot of conversations. After the conversations, we’ll go away and sort of build this all up. Then we’ll come back to that core group again and we’ll go through a few more variations of, ‘Why is this good? Why is this bad? What meets your needs? What doesn’t?’ In that way, it’s never done. If people ask when it’s done, I say never.”

The data collected by agency biologists is housed in state databases, usually in the state capital.

“We work with their state IT people and build an API—an application program interface—which is the way computers talk to each other basically. Our server at UM [University of Montana] sends a request over to their web server, which then provides sort of a gate between just free flowing information,” Lukacs says. “It gets our request, goes into the database, grabs the data, sends it back over the internet to the UM server, we quick do the analysis. We don’t house the data.”

Nowak adds, “Everything is password protected, and all data is sent through encrypted channels. With that being said, the data that we’re moving is publicly accessible stuff, but we take the security seriously because that’s hard-earned data.”

Some of the longer analyses can take up to six hours, but most can be completed in under a minute. Nowak has even seen biologists use the program on their phone to answer questions about proposed tag numbers during state game commission meetings.

To maximize efficiency, the program uses as many little chucks of prewritten code as possible. Nowak even wrote a code that writes code, an idea that Lukacs calls “functional programming.” Biologists click a few buttons to tell the program what P

HO

TO: C

OU

RTE

SY

OF

JOS

H N

OW

AK

Paul Lukacs (left) and Josh Nowak savor a successful pronghorn stalk last fall. The two University of Montana researchers are the brains behind PopR, a cloud-based computer program that creates integrated population models for state wildlife agencies.

information they’re looking for, and the computer algorithm automatically figures out how to structure the model and applies the right data.

Biologists then use those numbers to help determine management strategies.

State agencies are already putting PopR to good use. In southern Idaho, biologists take to the skies to get a population estimate for elk, but that doesn’t work so well in the thick cover of the Idaho Panhandle.

“The sightability model would come up with these fairly variable estimates year to year with really wide confidence intervals. If you get an estimate of 3,000, plus or minus 3,000, it doesn’t do a lot of good. The variance was so big,” says Wayne Wakkinen, Idaho Department of Fish and Game (IDFG) regional wildlife manager for the Panhandle.

They’ve had to find other ways to glean information about their herd—from hunter feedback to GPS collars to the trail cameras countless hunters use (see “Trail Cams Meet Lab Coats” on page 64).

“We don’t have population estimates like the rest of the state does with their helicopter sightablity surveys,” says Wakkinen “So we take all this various input and try to make sense out of it.”

Enter PopR.IDFG already has an integrated population

model—accessible through PopR—for mule deer in southern Idaho, and biologists have been working closely with Nowak and Mark Hurley of the IDFG research department to create one for elk in the Panhandle.

“My first thought is, there’s this black-box IPM thing. You put numbers in, and numbers come out. But I want to know what happens inside the black box—how does it weight the variables, what goes


into that? From a management standpoint, that’s been enlightening,” Wakkinen says. “It’s been an interesting process to work with the researchers to figure out what’s behind the curtain.”

The program will take all kinds of input data and then use statistical measures to determine and rank the quality of that data, giving more weight to the highest quality stuff.

“When we have a bunch of radio collared mule deer, obviously your survival estimate is quite precise,” Wakkinen says. “The model interprets that as, ‘Okay, this is high-quality data and therefore we’re going to weight it quite a bit when we come up with our population estimate.’ It depends on the survey, too. If you have a not-so-good survey, and you’ve got an estimate of a fawn/doe ratio that has wide variances, then the IPM is going to say, ‘Yeah, we’re not going to put a whole lot of stock into that.’ It uses it, but based on the variance, it doesn’t weight it as heavily.”

The elk model for the Panhandle is still in the early stages, but Wakkinen says they plan to incorporate a host of different data sources—from cow and calf survival rates gleaned from GPS collar information to harvest numbers reported by hunters. It’s all information they’re already collecting, Wakkinen says. This is just another way to utilize it.

“What we’re hoping is that we can take all this variable input and develop an IPM for elk that can inform us on population trend and what’s affecting the population,” Wakkinen says. “If it’s predation related, we can address that. If it’s habitat issues, we can try to work on that. That’s how we’re approaching elk management up here.”

And, about Nowak and Hurley and the rest of the stats team, Wakkinen adds, “Those guys are way smarter than I am! We let them figure out all that high-

powered statistics stuff, and then we use it!”

It’s unlikely that you’re thinking over mathematical formulas as you chase elk (unless of course, you’re Nowak, who says he started this whole thing because he wanted to know more about how and why elk move). So why bother to care about statistics?

“Hunters are paying a lot in license fees and Pittman-Robertson tax dollars to fund conservation. We want to use that money efficiently,” Lukacs says.

He and Nowak both say that the biggest challenge to precise population estimates is money.

“It’s the balance between costs and effort. If we had everything we needed, if money weren’t limited, we could have a really good estimate,” Lukacs says. “But you’d be talking millions and millions of dollars. It’s just completely unrealistic. And in many cases not even necessary.”

As a hunter, Nowak says that models are important because, when applied properly, they can help mangers stabilize populations by shifting to a more proactive management approach, where when a decision is made it’s understood what is most likely to happen in the future. When it comes to wildlife, management is just as often about people as it is about biology. Lukacs says that accurate data and statistics help dial in on those important, but often difficult conversations.

“In a mountain lion management workshop that was being held here in Montana, there was a whole bunch of argument going about two licenses here, there. If you can run the model and show, okay, two licenses one way or the other doesn’t matter to the lion population, then you can redirect the important part of the conversation which is what are your personal interests that are driving this management? What are your goals? On a bigger scale, say it’s elk or deer, we’re talking hundreds or thousands of licenses,” he says. “Ten years down the line, how has this management structure changed the population? That’s where the mental synthesis part starts to break down. There are so many things in play.”

It’s often up to people to tell managers what they want in terms of elk or deer populations, Nowak says, and it’s a state agency’s challenge to get there in a way that makes enough people happy. Clear numbers can help in the decision process—and then help explain those decisions to the public.

That’s part of the motivation for PopR. “I like being involved in the decision-making

process of how do we get from A to B, and I like the defensibility of it,” Nowak says. “All of the managers I know spend a lot of time on the ground. They worry about the data, the details, the people in their area. Whatever I can do to help them out and make their lives easier or make the decisions they make a little more defensible, that would be wonderful.”


by Paul Queneau

Using the same trail cams found at your local sporting goods store,

Idaho big game managers are breaking new ground for how they tally

elk and other wildlife.

I daho’s Panhandle isn’t your postcard western elk country trimmed in aspen and freckled with grassy parks. Instead its mountains are cloaked in conifer

thick as malamute hair—more than 1.4 billion trees on public lands alone fueled by precipitation topping 90 inches per year in some places.

“We tried counting elk from the sky in the Panhandle, but it didn’t really work because of the cover,” says Wayne Wakkinen, regional wildlife manager for Idaho Department of Game and Fish Department (IDFG). “As a result, we can’t use that same ruler to measure elk herds here.”

Biologists use winter flights to count elk in most of Idaho, but the Panhandle’s canopy makes flying as effective as hunting elk there with a 24-power scope. Sightlines are limited and latticed with branches.

Yet biologists know it’s a wapiti mecca. IDFG ranks the Panhandle as the state’s top elk zone, and 2,400 hunters filled freezers there in 2016. The agency bases tag allocations on data from hunter check stations, post-hunt surveys and most recently from

more than 150 GPS collars on cow and calf elk there. But seeing is believing, and IDFG is busy

researching novel new approaches to assess elk and other wildlife in the Panhandle and across the state using a network of the same trail cameras hunters use.

Synchronicity After Anna Moeller graduated magna cum laude

from the University of Puget Sound with a bachelor’s in biology, she landed a job as a wildlife research technician for the state of Idaho. She soon became fascinated by an IDFG pilot project that was testing if game cameras could be used to estimate wildlife populations. They hold several advantages over counting from planes and helicopters.

“First off, flying is really expensive,” Moeller says. “And second, it’s inherently dangerous. It goes without saying that we’ve lost too many biologists to aerial survey crashes. Also, it’s just not very effective over dense forests.”

With that as inspiration, she applied to the

Trail Cams Meet Lab Coats

PH

OTO

: LIA

M R

OS

SIE

R


NETWORK ANALYSIS—Wildlife technician Gus Geldersma checks one of 162 trail cameras across Idaho’s Panhandle and Beaverhead, testing a revolutionary new way to count elk.

renowned wildlife biology program at the University of Montana, just across Idaho’s border in Missoula. Moeller landed a spot working with professor Paul Lukacs at UM’s Quantitative Wildlife Ecology Lab (see “Big Game Meets Data” on page 58), where she launched a master’s thesis project to see if she and fellow IDFG staffers could employ some of the same statistical analytics favored by insurance wonks and baseball managers to count elk using trail cameras.

Photos have long been used to count “marked” animals—i.e. ones that can be recognized individually by their unique coat patterns.

“It works really well on things with stripes like tigers,” Moeller says. “But I was interested in coming up with a way to use cameras on unmarked animals. That’s a much harder problem.”

To test new theories and with the help of IDFG staff and funding, Moeller designed two networks using more than 160 trail cameras across a pair of unique landscapes. The first was Idaho’s Panhandle, which due to the heavy timber had never before achieved a formal count of elk abundance. The second was its polar opposite—the wide-open, windswept

ELK SUDOKU—Anna Moeller’s team drew a grid across elk winter range in Idaho’s Panhandle and Beaverhead areas (shown here). A computer then randomly picked nine cells. She split each into nine sub-cells, where she placed 81 trail cameras synchronized to take photos simultaneously as part of her research.

Beaverhead Mountains, a high-desert world ruled by grass and sagebrush, where finding a tree to even mount a camera can be a challenge. She used metal T-posts instead.

The Beaverhead acted as her control. IDFG biologists fly it every few winters to estimate elk populations using the long-proven method of aerial surveying, allowing Moeller to test the validity of her results using trail cams instead.

“We knew these would be two really different landscapes on which to test the model,” Moeller says. “If it could work in both, that would be really useful knowledge,” says Moeller.

She took maps of elk winter ranges and drew sprawling grids of 1.5 kilometer squares across them. Then she used a computer to randomly pick nine cells from those grids, which she further split into nine sub-cells.

Then Moeller and her crew placed a camera within each sub-cell, pointing them north to limit direct sunlight and clearing any vegetation obstructing their views.

Using these two trail cam networks, she experimented with three different ways to gauge elk population. The first she called the Instantaneous Sampling Estimator. She set all the cameras to shoot photos simultaneously at a given frequency over the course of four months. That ensured no elk could be double-counted since they can only be in one place at one time.

But this method also stacked up more than 1.3 million digital images on Moeller’s hard drive after she retrieved the media-cards. Luckily the statistical wizardry of this model only requires that 4 percent of the photos actually be analyzed. But even at that reduced number, it took four researchers several months to get through all the photos, counting any and all elk that appeared in each frame within 50 meters of the camera. For cameras in spots with long, sprawling vistas, she placed flagging to act as a distance guide.

Using this model, Moeller estimated 1,258 elk wintering within the Panhandle survey area—the first-ever scientific glimpse into the size of its elk population. In the Beaverhead, meanwhile, it estimated 1,613 elk—close enough to the last aerial count of 2,272 to demonstrate it could be a useful tool.

“You see a fraction of the elk when you take a picture all at the same time,” Moeller says. “If you just did that once, it wouldn’t be very accurate. But if you do that over and over and over again, you can average it out, and it turns out to be a fairly accurate estimate of how many animals you’ve got roaming across a given area.”

Moeller later tested two additional methods she dubbed “Space-to-Event” and “Time-to-Event.” The former mimicks her previous research in that it


synchronizes each camera using its time-lapse mode, but doesn’t depend on counting every elk in every photo. Instead it estimates the number of elk that would need to be wandering around in order to detect them on a certain number of cameras across the network.

Time-to-event, meanwhile, relies on the motion-trigger of each camera to test how long it takes before an elk walks in front of any of the 160 camera traps, tripping their sensors to snap an image. Using that elapsed time between sightings as the primary measure, the model can extrapolate how many elk would need to be roaming an area for them to keep appearing at a given interval. Once again it doesn’t matter how many elk there are in each photo, just how often one or more shows up.

If that makes your brain hurt, consider this third model also depends on an approximation of just how fast your average wapiti cruises the landscape come winter.

“We calculated a median elk speed using 122 GPS collars in the Beaverhead and Panhandle in January 2015, which put that number at approximately 30 meters per hour,” Moeller says.

Both models also require those wanderings to be at least somewhat random. That seemed to work in the Beaverhead where elk are more prone to paint an arbitrary path across the rolling sage. The time-to-event model estimated 2,217 elk there—just 55 shy of the last aerial count—while the space-to-event method estimated 1,405.

But in the Panhandle, elk seemed to violate the random-roaming requirement by following more redictable trails through the heavy timber. The space-to-event estimated 1,368 elk, while the time-to-event model estimated a somewhat eyebrow-raising 5,670 elk.

“There’s some more work that needs to be done looking into how to allow a fast-moving population of animals to re-randomize,” Moeller says. “But for a field estimate, it’s a good start. In the Panhandle, based on harvest and what we know about the area, the true abundance [number of elk] is most likely somewhere between the low and the high estimates. But again, we don’t have any previous estimates to compare this to, so this is a really big step forward just getting an estimate at all.”

Wakkinen agrees.“It holds promise for sure,” he says. “If we

can use her technique and repeat it here in the Panhandle, we should at the very least be able to get an estimate of population trend, which by itself would be hugely beneficial.”

IDFG senior wildlife research biologist Jon Horne led the pilot project that inspired Moeller’s work. He recently laid another network grid of 83 trail cameras to begin stockpiling photos with hopes

POSTMASTERS—Unlike the Panhandle, Idaho’s Beaverhead provided few trees on which to mount cameras. Moeller and crew substituted with T-posts.

of eventually being able to estimate not only elk numbers but also moose, whitetails, black bears and mountain lions at a busy crossroads for wildlife in the Clearwater National Forest east of Lewiston.

“This is the next frontier,” he says, “and we’re going to see what’s possible.”

Yet these camera networks are tiny compared to what IDFG research biologist David Ausband has put together to size up Idaho’s wolf packs.

Border to Border with 200 CamerasMontana and Idaho recently adopted the Patch

Occupancy Model (POM) as their primary tool to estimate statewide wolf populations (see “Guess How Montana Counts Wolves Now?” on page 81). In a nutshell, POM breaks both states into giant grids of 600-kilometer squares—aka patches—reflecting the average pack-territory size of wolves in the Northern Rockies based on decades of radio-collar data. Biologists determine the size of the population based on whether wolves are sighted within each patch within a given time span.

Yet where Montana relies on hunter reports to fuel its POM estimates, Idaho uses trail cams.

“We have around 200 out statewide distributed MA

P &

PH

OTO

: CO

UR

TES

Y O

F A

NN

A M

OE

LLE

R


from Canada to darn near Utah,” says Ausband. “We are primarily placing the cameras within 500 meters of predicted wolf rendezvous sites using a specialized habitat map.”

Rendezvous sites are one of the few dependable places to find and view wolves, a spot where they post their pups after denning ends but before young are ready to hunt with the pack. Wolves return again and again to these sites with food, and Ausband helped create prediction software back when he was a University of Montana grad student. It’s proven its prowess at keying in on those habitat characteristics wolves seek out for rendezvous sites such as areas of grass or sedge adjacent to wetlands.

He says cameras have so far worked well to estimate Idaho’s wolves using the POM method. The last collar-based count in 2015 placed the state at 108 packs, and Ausband says he anticipated a similar estimate for 2016.

“We figured POM would put it somewhere between 80-100 packs. Our results are still preliminary, but patch occupancy is putting us at 90 packs statewide in Idaho.”

While Moeller’s work in the Beaverhead was occasionally challenged by elk using cameras as scratching posts, Ausband deals with bears that want to chew on them. He says the greatest challenge, though, has been data management.

“We need a human to look at every picture and say whether it was triggered by a squirrel, a wolf or just a blowing branch. We have the camera out for three months, and in a way, it’s less effort [than traditional methods]. But the real work comes later when you have two terabytes of images for some poor soul to go cross-eyed sitting there categorizing everything.”

To ensure consistent quality control, Ausband had a fellow IDFG staffer comb through every image.

“It wasn’t too bad,” he says. “We went through about 200 cameras in probably three months, all day every day, for 40 hours a week.”

Looking forward, IDFG Research Supervisor Mark Hurley says his agency is expanding its efforts to experiment with trail cameras to assess wildlife. The agency is conducting additional pilot projects to see how well trail cam networks might be used to estimate populations of moose and mountain lions as well, and working again with Moeller as a researcher to see how cameras might also gauge sex and age ratios for ungulates. While POM is now Idaho’s go-to for wolves, other trail-cam survey methods remain in the test stages. IDFG biologists feel they hold great promise.

“This really is a huge breakthrough,” says Hurley. “To be able to estimate populations of unmarked animals in this way is very exciting. We’ve simply never been able to do it before.”

WILDLIFE R&D—With hundreds of trail cameras taking millions of photos, Idaho Fish and Game is counting its wolf packs border to border and expanding its research into camera grid networks with hopes of mastering the mathematical and statistical challenges that could save lives and money if they are able to replace game count flights.

2017-06-16 10:17:37 AM 50° F M 2/3

2016-09-22 10:27:20 AM 43° F M 2/3

2016-06-23 10:59:32 AM 74° F M 2/3

2016-03-18 07:53:56 AM -9° C M 1/5

PH

OTO

S: C

OU

RTE

SY

IDFG


I n the early 2000s, the Oregon Department of Fish and Wildlife (ODFW) faced an informational black

hole when it came to estimating wildlife populations and other herd statistics in some of its smaller Roosevelt’s elk units in the state’s western half.

The dense forests of this region limit the options for any aerial or ground-based elk counts, so for decades the department instead relied on hunter feedback to gauge the health of the herds. The more elk are killed in a unit for a given number of hunters, the better the species in that area is likely faring overall. It’s a method used by wildlife departments nationwide, especially for animals that lurk in heavy cover such as whitetails, bears and moose.

But budget cuts in the 1990s forced ODFW to shrink the quantity of

post-hunt phone surveys it made to hunters. At the same time, fewer hunters were answering the remaining calls as cell phones and caller ID became more prevalent. By 2012, response rates to ODFW’s survey plunged to less than 30 percent.

For special draw units, the department could seek feedback from specific hunters who drew those tags, but over-the-counter units open to anyone with an elk tag left no such paper trail. It became a stroke of luck for the phone surveys to snag a hunter who had in fact gone to the smaller OTC Roosevelt’s elk units, and gradually feedback sank too low to provide valid herd data.

“It was like finding a needle in a haystack,” says ODFW director Curt Melcher.

So in 2008 ODFW replaced the phone

surveys it had depended on for more than 40 years with a mandatory hunter reporting system. It required that all deer, elk, pronghorn, cougar, bear and turkey hunters chime in post-hunt either online or by phone about whether or not they filled their tags. But it also lacked any teeth for non-compliance. After three years under the new system, only 41 percent of hunters reported back after the season.

“So we tried to give hunters a carrot out there,” says Don Whittaker, ODFW ungulate coordinator.

A carrot—and a bit of a stick. In 2012, ODFW began issuing a $25 penalty to non-reporting hunters, tacked onto tag fees the next time they purchased a license. But ODFW also entered hunters who reported their harvest data before the February deadline into a special drawing. Lucky P

HO

TO: K

IRA

VE

RC

RU

YS

SE

N

Killer StatsHunters help balance the herds when they take to the field—and when they report their results.


winners could choose to hunt deer, elk or pronghorn in an expanded area and for an extended season.

The fines weren’t without controversy for the first few years, but the system has become generally accepted as a necessary step to support sound game management.

It’s also bumped feedback up to more than 80 percent from deer and elk hunters, allowing wildlife biologists to more accurately estimate populations for the following hunting season. The smaller Roosevelt’s elk units now get data that allows biologists to better gauge herd productivity.

The reporting system also eliminated the costly $300,000-per-year telephone surveys and now actually raises about $200,000 a year through the $25 fees from hunters that don’t report back.

“We now have the best harvest survey data the state has ever had,” Whittaker says.

The New Mexico Department of Game and Fish (NMDGF) also launched a mandatory hunter harvest reporting system in 2006 to increase feedback, which before then had been voluntary.

“Reporting rates were low, with a 28 percent average return rate statewide for three years immediately prior to the 2006 hunting season,” says James Pitman, NMDGF Elk Program Manager.

New Mexico’s mandatory system tacks on an $8 fee for hunters who don’t report on time, but goes a step further by blocking non-reporting hunters from entering elk drawings for the following hunting season. This is a big stick indeed in a state where elk licenses are issued by drawing only.

In 2016, New Mexico sold 36,936 elk licenses and received feedback from 29,575 hunters—an 80 percent response rate.

Checks and Balances Another way wildlife biologists

collect data across the nation is through hunter check stations, also known as game or biological check stations. When successful hunters come through, biologists not only keep tabs on how many male and female animals they see, but also estimate the age of each animal by examining its teeth and taking standardized measurements of its antlers or horns. That data helps them gauge age structure of males and females in each hunt unit as well as the survival, growth and longevity of various age classes.

In 2017, the Idaho Department of Fish and Game (IDFG) set up 17 weekend check stations, which they manned for over two months at major funnel-points for sportsmen returning from their hunts. Wildlife Biometrician Bruce Ackerman says these check stations are invaluable for biologists working to chart herd trends and to assemble more specific data than gathered by online reporting—statistics that can reflect the quality of habitat and the health of the herd, and be used to spot any worrisome shifts.

“This might include weight of

yearling deer, measurements of length and width of antlers, and tissue samples to test for chronic wasting disease in deer and elk,” says Ackerman.

Like Oregon, Idaho asks its hunters to also provide harvest data through a mandatory hunter harvest reporting system the state launched in 2001. For the first year it worked wonderfully, with 95 percent of hunters filing a report. Idaho, though, doesn’t have a penalty for not reporting, and its response rate has gradually declined to just 60 percent in 2016.

“In the early years, we issued a small paper report form and a pre-addressed envelope to hunters when they purchased their tags,” says Ackerman. “But it was much too time consuming and expensive.”

However, the harvest data they do get is still extremely useful, benefitting wildlife and hunters alike.

“We calculate the number of hunters and the number of animals harvested in 1,400 different ways,” says Ackerman. That summation incorporates every combination of elk, deer and pronghorn throughout 99 different Game Management Units (GMU) and 29 elk zones, using three weapon types and including over 400 different controlled hunts.

The IDFG then posts the results for all of the 1,400 combinations on its website by May as an information tool, one that can reveal which units are most productive, oversubscribed and other stats to help hunters planning for next fall choose which tags to apply for or new units to try.

Wildlife departments throughout America rely on harvest data and surveys to make sound management decisions. By simply stopping at check stations on the way home from a hunt or filling out harvest reports each season, hunters support a proud tradition of being at the front line of wildlife science and conservation.

Whether motivated by the carrot or the stick, every hunter who does their part to support these efforts helps ensure America’s wildlife populations remain balanced, healthy and sustainable.

— Suzanne Downing


Big Game Meets big data - University Of Montana · 2020. 5. 27. · Big Game Meets. big data. by Kasey Rahn. eliable olatio estimates are te dametal bildig blos o ildlie maagemet

Documents