Helping find Walleye (and friends)

The Wetland Ecosystems Research Lab at McMaster University is attempting to modernize fish surveying in Great Lakes coastal wetlands, specifically Georgian Bay, Ontario. We need your sleuthing skills to comb through video footage and spot the fish hidden in the scene! Every one of you who decides to help with this project will be an invaluable member of the team and to our fish friends.

Why is this project important?

Wetlands are biodiversity hotspots, and the home for many fish species, including gamefish, such as bass, perch, and muskie/pike babies! As a volunteer your help in classifying this project's footage is two-fold; you will aid our lab in assessing the fish community of coastal wetlands in Georgian Bay now, and further validate the use of a new fish survey method that is hopefully cheaper, more accessible, faster, and less harmful to the fish researchers want to study.

Why Georgian Bay?

Georgian Bay, the northeastern arm of Lake Huron/Michigan, is the largest freshwater archipelago in the world, and home to a variety of wildlife, from turtles, to birds, and of course, fish. For us humans, it is the place to go for the true Canadian cottage experience. As a part of the Great Lakes, Georgian Bay is relatively the least impacted by direct human development, making it the perfect place to study the effects of environmental changes, such as the indirect consequences of human-driven climate change.

But wait! What is a coastal wetland?

A wetland is very simply where land meets water and so it is wet. Most wetlands also need to be wet long enough for a period of time in the year - year after year - for it become a dependable habitat for animals to use. A coastal wetland is a wetland located on the coast of a larger permanent water-body, like a Great Lake (sometimes biologists do give things names that make sense). For fish, coastal wetlands can be their home for life, their summer home, or a hotel, providing them places to eat, have their babies, be safe from predators while they grow bigger, or just be a resting spot from the waves and predators out in the big open-water of the lake.

Water levels and why they matter for coastal wetlands

Did you know that Great Lakes' water levels change? Lake Huron/Michigan, and so Georgian Bay too, is even less of an exception, since its water levels are not controlled by dams like other Great Lakes. Even if we humans don't like it much, change is good for a coastal wetland. Changing water-levels means not one individual ends up dominating the wetland; in other words, change encourages biodiversity! However, in the past two decades, water-levels first didn't change for 14 years (and were low), and then they perhaps changed too much, and rose over a meter (that is 3.3 feet for our American volunteers)! Such dramatic changes are not "normal" for Georgian Bay coastal wetlands, and so our lab is interested to know how wetlands changed because of these water-levels.

What will we learn?

In the short-term, the data will tell us how the changing water-levels in Georgian Bay impacted fish communities in the past. In particular, you may see trees, roots, or bushes in some of the footage you watch. Those aren't usually there in a coastal wetland, and not for as long as our lab has seen them (over 5 years). We are interested to know if this "novel" vegetation does actually impact the fish that use the wetlands. Long-term, this research project will not only help us predict the impacts of climate change, but potentially how we sample fish communities in the future. If researchers can place a camera instead of a net, that is a lot less work for the researcher to get important data, but is also less intrusive and stressful for the fish!

Our Machine Learning Journey and Beyond

You may be thinking, it's the 21st century, surely a computer could do this work? The short answer is...it's complicated. 😕 There are machine learning models out there that have been used to classify underwater footage, either flagging fish occurrences or even going as far as to identify them to species. Machine Learning (ML) is the process of "teaching" a computer to recognize patterns and run automated processes such as classification by using statistical models and training data. Think of how a toddler learns which animal is which by being shown a set of flash cards.

We ourselves tried to use open-source models, "out-of-the-box" so to speak, to process this project's footage. Unfortunately, these attempts were not successful, since these ML models were trained using different environments. Just like you couldn't ask a toddler to recognize African savanna animals if they have only been taught farm animals, the same is true for these models and our footage. Many ML models were developed using coral reef habitats, which are much more open and clear than our sites, and don't have dense canopies of wavy underwater plants that a computer sometimes "thinks" is a fish moving. Rather, a ML model would have to be coded and trained from scratch with our footage, using different mathematical formulas than other models. This requires time, expertise, and funding not available to our lab at this point in time.

HOWEVER, in the future, it is our lab's hope to take the outputs volunteers so graciously helped us create here to then train a computer to do the work for us moving forward. At the moment, our lab is in discussions with Dan Morris from Stanford University, who is currently developing a new ML model that appears promising for our specific environment type!

How "Where's Walleye" started

In the beginning, Where's Walleye started simply because our lab couldn't put our nets in wetlands, because there were trees there! So we tried looking for an alternative, and thought cameras may be the best solution. After searching for a commercial solution, nothing quite fit our needs (or our budget), so we decided to make our own underwater camera system. With easily-available material, we made a weighted-frame from PVC pipes embedded in a bucket of cement, attached action cameras with bicycle mounts, tied on a rope and a buoy, and voila! Now we could place cameras wherever we want, and process the footage after. And boy, did we have a footage (over 2000 hours to be precise). With this footage, we now had a window into the world of fish, and we could show this world to citizen scientists to help us with the task of seeing what goes in this world.