Modeling Resilient Water Systems with Machine Learning
A Q&A with Kevin Fries, PhD, Flood Scientist at One Concern
At One Concern, we pride ourselves on uniting deep scientific expertise with a bias toward practical applications that change outcomes. Our data science team combines machine learning, environmental science, structural engineering, and various other disciplines to design solutions that address systems-level resilience, not just individual risks in a vacuum.
Kevin Fries, PhD, is a data scientist at One Concern who focuses on research and development for our flood hazard and impact models. We sat down to discuss his career path and interest in hydrology:
What did you do at work last week?
I’m on the flood team, so my focus is on papers in hydrology and hydraulics, focused around scientific computing. I’m looking at how we can model flooding in a more efficient and effective way so we can generate outputs quicker for our clients.
To do that, I read a bunch of papers. A lot of my job is doing research — both looking at the current research out there, as well as iterating on existing knowledge in the field. I also spent plenty of time debugging code, writing new code, and brainstorming with teammates around our product’s future functionality.
What have you learned since working at One Concern?
Well, on the technical side, I got familiar with Python. Like most PhD students, the only language I ever used was MATLAB, so I got to pick up a new language and learn the software engineering side of things: creating tests for code, using Github effectively, and working in cloud-based environments.
Since we’re a startup, I also adapted to more fast-paced work: getting good results, meeting objectives, and then moving on. In an academic setting, we really drilled deep and focused on the first 80 percent of the research process. But that 80 percent might be the easiest part; you might publish a paper, and nobody ever picks it up again. The hard part, the last 20 percent, is taking that idea and pushing it through to reality — making sure that whatever you published is usable by the field of production.
Can you tell me more about your academic background?
I did my undergrad and PhD at the University of Michigan. My background was in civil engineering with a focus on infrastructure systems and hydrology. I looked at how we can use systems engineering techniques — control theory, signal processing, and machine learning — to improve the way we do civil engineering.
My research projects all tied back to water. I looked at how to use ship-based observations to improve understanding of evaporation on the Great Lakes. Another area was using citizen science, such as stream gauges that aren’t part of the USGS network, to predict flooding more accurately. When combined with large-scale models like the National Water Model, I found that just a couple of sensors could be used to identify leaks or blockages in whole stormwater systems.
How did you get into civil engineering and hydrology?
I started as a structural engineer and realized I didn’t really like it. But what did interest me was my fluid mechanics class and systems engineering class, which was essentially an optimization and linear programming course.
So I went to grad school thinking I would study optimization of water resource systems — figure out how to improve irrigation or something similar. When I actually got to grad school, a whole new world of ideas opened up. I happened to be there around the time that machine learning was becoming a thing, and I latched onto that and started applying it to water resource problems.
I know you’re passionate about interdisciplinary solutions to global problems. Why is that so important?
I took a certificate in Public Policy, and the biggest thing I learned is the importance of framing. No matter how big a problem is, or how awesome the solution is, you have to frame the problem correctly so that the people you’re speaking to understand why it’s important.
A lot of issues we have with communicating about climate change come from not framing our problems effectively. When you go into a problem like that and try to find a solution, you should have a team of people who understand it from multiple perspectives. That way, you can frame the problem in a way that covers all the bases, and the solution you create isn’t only plugging into one part of the issue.
Okay, so let’s say I have an uncle who doesn’t believe in climate change. What’s a tip you have for talking about it?
You have to frame climate change as an existential problem.
One of the big issues climate change had for a while is that all the impacts people talked about were 100 years out. And everyone says, Well, that’s not gonna happen in my lifetime, so who cares? Or, the impacts didn’t seem local. Well, I don’t live near the ocean, so sea level rise doesn’t matter.
So a big thing when it comes to climate change is showing people the impact it has on them personally. And the easiest way to do that is economics — money talks. Unfortunately, that’s why climate change was a bipartisan issue until the 80s, but only became partisan when certain industries realized it would hit their bottom line.
Finally, what does ‘resilience’ mean to you?
Resilience boils down to systems-based approaches. There, it’s defined as the ability for a system to absorb a perturbation and return to a steady state. It’s the idea that we live in a complex world with a complex system, so our systems need to be designed so that when some outside force — natural disaster, climate change, terrorism, anything — tries to shake it up, the system can return to a steady state and it doesn’t just implode.
That’s how I look at things; it gets at both the resilience and the adaptation side of the equation. Some people like to think that resilience and adaptation are different, but I view them in the same line.
If you enjoyed this Q&A, keep an eye out for the rest of our Data Science Showcase in the following weeks!
Want to help us build planetary-scale resilience?
Check out careers at One Concern to see how you can help.
Related articles:
