The following piece was written by OAS Communications Coordinator Ryan Bower for the Snapshot Wisconsin newsletter. To subscribe to the newsletter, visit this link.
A few months ago, the Snapshot team said farewell to someone who has worked with the Snapshot Wisconsin team for several years. John Clare, a former graduate student at the University of Wisconsin-Madison, completed his PhD in December, 2020 and has moved on to a post-doctoral position at the University of California-Berkley.
While completing his doctoral degree, Clare worked in Drs. Ben Zuckerberg and Phil Townsend’s labs, and his research has helped push Snapshot Wisconsin to the next level, expanding the capabilities and reach of Snapshot Wisconsin. Although he played a behind-the-scenes role, one of a few students studying how to use Snapshot data in new and useful ways, his contributions to the team are appreciated, so the Snapshot team decided to share a piece of his research with those of you who follow this newsletter.
Clare first connected with Snapshot Wisconsin when current Snapshot team leader, Jen Stenglein, got in touch. Stenglein, Quantitative Research Scientist at the DNR and a leading member on the data analysis side of Snapshot Wisconsin, was interested in the sampling parameters of Clare’s Masters research, which also dealt with trail cameras. Stenglein hoped to learn from that project and apply lessons to Snapshot Wisconsin.
“I didn’t know about Snapshot Wisconsin until after talking with [Stenglein],” said Clare. “I later saw a posting for a PhD assistantship related to the program, and I applied. That is how I first got involved.”
Clare was one of the first graduate students working with the project. He initially helped get the program up, and later he started to sort through the data and deliver some useful results.
While we don’t have the space to cover everything he worked on for his dissertation, Clare and the Snapshot team wanted to share a small piece of Clare’s research with the volunteer community and showcase part of how Clare contributed to the project.
Leveraging Snapshot Data
A central goal of Clare’s dissertation was to develop strategies to better leverage the spatial and temporal capabilities of the Snapshot database. Clare mentioned that two unique features of Snapshot Wisconsin are that Snapshot operates both statewide and year-round. Many other monitoring programs can’t operate at such wide and long scales because it would be too resource intensive for them. Fortunately, Snapshot has the help of thousands of people across Wisconsin (and the globe) to overcome that resource barrier and operate statewide and year-round.
“[Using data from all over the state and from all times of the year], we can explore questions in ways we didn’t have the ability to before,” said Clare, and Clare investigated a few of these questions in his dissertation. Two of Clare’s research questions were what broad factors drive where species are distributed and how species are active across the year.
To answer both questions, Clare needed to build a special type of model that leveraged both spatial and temporal data at the same time. Not an easy feat.
Setting Up Clare’s Model
Clare needed a unique model that could account for how species are spatially distributed around the state and temporally distributed throughout the year. “I think it’s important to take advantage of both the spatial and temporal components at the same time,” said Clare. “The question isn’t just where are species located, but also how species are distributed at time x, time y and time z.”
Both the spatial and temporal scales were needed because there are components of the environment that vary strongly across space and time. Snow depth, for example, is not fixed over the course of the year. One week, there may be six inches, and the next week there may be twelve. Snow depth also varies spatially. A few miles could be the difference between seeing snow on the ground or not. Many environmental factors are highly dynamic and variable like this, so Clare needed to think of these factors within a model that accounts for both.
It is common for models to use one type of data but incorporating both is a challenge. The main challenge is having enough data (and data of the right types) to run this kind of analysis. Fortunately, Snapshot images have both location and time data attached to each image.
Another unique aspect of Clare’s model is that it considers multiple species at once. “We were pretty sure that individual species are distributed dynamically throughout space and time, but entire communities have not been heavily studied in the same way,” Clare said. “The appeal of using a spatial-temporal structure across the entire community is that we can explore which species are interacting with others at different parts of the state and at different times of the year.”
This concept isn’t new to the realm of modeling, but it is hard to accomplish. Researchers would need separate data for each additional species they added in the model. It can be hard enough getting data for one species, let alone multiple. However, that is where Snapshot shines best. Volunteers can tag up to 50 unique species in their Snapshot photos, so an equal number of species-specific datasets can be pulled and created from the larger Snapshot dataset.
“The advantage of a multispecies approach is that you can take into account the responses of each of those species, as opposed to modeling one species and assuming the results apply uniformly for other species,” said Clare.
Driving Distribution and Activity
Knowing he had the ability to answer his questions, Clare thought about which factors might be most influential across the entire community, in terms of predicting where species were located and how active they are. “We had a couple ideas about what these factors might be,” said Clare. “Some were related to seasonal variation like the amount of snow and the greenness of the vegetation.”
Snow depth can change substantially from day to day, even during the winter. Snow depth could dramatically impact how species move around and where food is available. Snow can even correlate to which species are even seen during parts of the year. For example, black bear behavior is often related to the winter, and thus with snow.
Wisconsin’s black bears sleep through the winter. Since winter is also associated with snow, black bear activity inversely correlates well (in Clare’s model) with snow. When there’s more snow on the ground, we are (most likely) deep into winter and see the least activity from black bears.
Another environmental variable that Clare was interested in was vegetative greenness. Vegetative greenness is, from space, how green the landscape looks. In the spring, trees will start to bud burst, and the grass will grow. The landscape itself will just be greener than the previous months, and more nutritional energy will flood into the food web. Vegetative greenness varies throughout the year and can impact how animals use the land, depending on when and where food is available.
For example, a black bear’s seasonal activity could reflect the cycle of vegetative greenness. Black bears maximize their activity at times of the year when there are more food resources around, either plants or prey. These times of the year may strongly correlate to peak greenness of the landscape, or so Clare theorized.
But you might be thinking, “Wait, can the Snapshot cameras measure vegetative greenness and snow depth from the trail camera photos?” The answer is possible, but more research is needed before we can use the cameras that way. Instead, Clare used daily satellite images of the state to calculate vegetative greenness and snow depth.
Linking Satellites with Snapshot Wisconsin
Clare used satellite images from NASA to measure snow depth and vegetative greenness. Part of Clare’s assistantship position was funded by a NASA grant whose purpose was to figure out ways to integrate a continuous stream of animal observations with a continuous stream of Earth observations coming from space. Between the trail camera data and the satellite data, Clare aimed to find connections that were meaningful to wildlife management.
Consider winter severity in deer population modeling, for example. Winter severity is already used by the DNR to predict the impact of winter on deer populations and plays a partial role in making harvest decisions for the subsequent fall. One hope of the NASA collaboration was to develop more integrated measures like winter severity for deer overwintering, especially ones that impact multiple species in similar ways.
Using the images from satellites passing over the state, Clare derived data on land use, surface temperature, vegetative greenness and snow depth. All of these variables were tested across spatial and temporal scales for all classifiable species.
Confirmation and Surprise
Clare wanted to share two results from his dissertation with the Snapshot community. One of these results was a confirmation of what he expected, but the other was surprising and took longer for him to wrap his head around.
“I wasn’t surprised that snow depth was a major negative driver of species activity,” Clare said. “We expected that because snow provides a refuge for some species [and a signal for other behavioral changes like hibernation].” These behavior changes cause sightings of these species to drop off during the winter and strengthens the negative correlation between snow depth and species activity. Snow is also associated with winter, when species tend to be less active to conserve energy resources.
What was more surprising was that the peak period of species activity was not associated with the peak of the growing season, or when the land was at its greenest. Clare expected these two peaks to match because there would be a maximum amount of food on the landscape. However, after some rethinking, Clare came up with a new theory about why peak activity wasn’t at peak greenness. “What we [now] think is happening is that animals don’t have to move around as much during the peak of the growing season. They don’t have to go as far to find food. It is all in one
aisle,” Clare said.
As for linking satellite data with wildlife data, snow depth and vegetative greenness both were the best predictors of species distribution and activity out of the environmental variables Clare tested. Even though vegetative greenness didn’t function how he predicted it would, it still was a good predictor of community activity and distribution. Both of these variables showed promise as potential satellite-based metrics that NASA and the DNR can use to better predict how the environment is impacting the greater wildlife community.
Now that Snapshot Wisconsin has a few years of data across most of the state, Snapshot will start looking into broader trends like year-to-year weather variation and how species habitat associations may vary from year-to-year.
“As we anticipate global changes, including more extreme events like polar vortexes, heavy rain and droughts, there is a need to understand how species react to different weather phenomenon. By looking at how species are distributed at finer time scales, we can start to address those types of questions. That wasn’t the exact focus of my research, but my research can help us start to quantify what [counts] as an extreme event for different species,” explained Clare. However, that work will be done by someone else, since Clare has graduated and moved to California.
Clare took a moment to reflect on his years working with Snapshot Wisconsin. Clare said, “My favorite part has been seeing the broader project move from a concept to an operating system. It has been really exciting to see that dream come to fruition. Most of that credit is due to the folks on the Snapshot team like Jen Stenglein and Christine Anhalt-Depies.”
Clare was also appreciative of the community of volunteers that sustain Snapshot Wisconsin. “It has been rewarding to see so many Wisconsin residents get involved,” said Clare. “I’ve been blown away with how smoothly and effectively it all has worked.”
With Clare moving on to the next step of his career, the Snapshot team wishes him the best and thanks him for helping the program get set up and running, as well as his contributions on the research side of Snapshot.
Thanks John Clare, and good luck!