Building A Better Model

As the climate changes, Canada’s freshwater systems – and the economies that depend on it – are also transforming. In the Great Lakes region, for example, periods of extended drought or extreme rainfall can dramatically change water levels. A shift in these levels may mean less water is available for communities, or that communities are at risk of flooding. Shifts can threaten ecosystems that rely on certain conditions to thrive. They can change the predicted output for hydropower, or limit our ability to ship goods by water, or require farmers to cut back on watering their crops.

To help predict and manage these impacts and others, we need tools to understand the weather, and how it might affect the water we currently have available to us.

Bryan Tolson is a Professor of Civil and Environmental Engineering at the University of Waterloo, where he focuses on helping predict floods and water availability on earth’s surface.

The partners Tolson works with – conservation authorities, governments, industry, and others – are the agencies that rely on flood plain and hydrological modelling to make decisions about how to manage water resources and related risk.

Tolson’s research team looks at how we build those models. “We ensure hydrological models are as accurate and easy-to-use as possible,” he says. “We build the tools that help our partners fine tune their models, optimize their parameters, and ultimately make sure the model can accurately predict historical field data like streamflow. We want to give them access to what they need to make the best decisions that will protect people and nature.”

Improving our predictions

Tolson has spent much of his career using the available tools for modelling and finding ways to improve them.

In one of his first publications at Waterloo, he applied what he learned from his PhD research at Cornell University to develop an algorithm called DDS, or Dynamically Dimensioned Search. Today, users including Environment and Climate Change Canada (ECCC) apply DDS to calibrate their hydrological land-surface models.

“Making the models they build and the data they collect work together is paramount,” Tolson says. In his experience, however, the tools available to do this work at the time were clunky and complex. The DDS algorithm improves upon those tools, working quickly and effectively. “For our users, it is a gamechanger,” he explains.

According to Tolson, improving models also means ensuring that improved tools and methods are accessible to potential end users. Together with fellow UW professor Dr. James Craig leading the team, Tolson helped develop Raven, a robust, flexible, and open-source hydrological modelling framework designed for challenging hydrological problems in academia and practice, including flood forecasting and climate change studies.

“Raven is one of the most commonly used streamflow prediction models across Canada. It has been proven to be really useful for flood forecasting,” Tolson says. “We also know that hydropower agencies use it to optimize production by being able to model incoming flows, for instance.”

Tolson says his experience with flood plain and hydrological modelling also opened his eyes to what’s missing from the toolbox. As part of the NSERC Floodnet program, Tolson worked together with Dr. Juliane Mai to develop CaSPAr, the Canadian Surface Prediction Archive.

CaSPAr is a dynamic database of historical numerical weather forecasts that allows users to access historical data to help build and test their own systems. “An archive of this sort has many uses, but until a few years ago, nothing like it existed,” Tolson explains. “For years, people looking for historical forecast data had to piece it together from multiple sources or, in many cases, start archiving real-time forecasts from scratch.”

Today, the database helps hundreds of users find everything they need in one place. “It’s one step, but it greatly improves our users’ ability to model and predict future conditions,” he explains. “It takes a lot of the initial work out of the equation.”

After Floodnet, Tolson worked together with Dr. Mai again through Global Water Futures (GWF), one of the world’s largest university-led freshwater research programs. The colleagues published several studies, including a series of award-winning hydrological model intercomparison studies (summarized here). These studies were instrumental in showing the power of machine learning models in contrast with traditional hydrological models.

Improving what we know

Another challenge for some agencies includes deciding how to break up – or “discretize” – the complex and different parts of a watershed in order to model them. With initial support from GWF, Tolson and his colleagues developed BasinMaker, an open-source software. They then worked with Ontario’s Ministry of Natural Resources and Forestry (MNRF) to build on BasinMaker to create the Ontario Lake-River Routing Product, which is a geospatial dataset accessible via a map-based website that includes 82,000 lakes.

“BasinMaker helps users break up watersheds in intelligent ways, especially where there are thousands of lakes, rivers, and streams like in northern Ontario,” Tolson says. “With MNRF, we’ve developed a tool that is specific to the province and its spatial data. BasinMaker translates spatial data into routing maps that show the connections between waterbodies and how water moves through these complicated surface water networks.”

The tool is compatible with Raven, he adds, and the progress doesn’t stop there. “We’re currently working with ECCC on the Canadian Lake-River Hydrofabric project, which will broaden the geographic scope of BasinMaker’s generated routing networks to capture the full country,” he says. “This tool will be available to anyone who wants to build a hydrological model of their system.”

The hydrofabric will also be instrumental in being able to leverage data from NASA’s brand-new SWOT satellite mission, which will provide the very first comprehensive view of Earth's freshwater bodies from space and allow scientists to determine changing volumes of fresh water across the globe at an unprecedented resolution. 

“BasinMaker is really the only way to make sense of this data from a hydrological modelling perspective,” Tolson says.

Building resilience through environmental data

The GWF funding he received has accelerated many of Tolson’s ideas and projects, and the direct impacts include modernizing outdated methods and approaches to modelling. “The GWF program motivated me to think more about what was missing from the community,” he says.

GWF also provided some core funding that allowed Tolson to hire research professor Dr. Mai. “Julie’s presence – in addition to the students and colleagues on our team – enhanced my program tremendously. As we move onto new roles, we are sharing our experiences and knowledge to support positive changes in modelling communities in Canada and around the world. GWF helped make that happen.”

On a broader scale, Tolson’s participation in GWF has contributed to the ways Canadians can access and use environmental data to effectively build resilient communities and protect freshwater resources. “We have a lot of data available to us, but if we’re going to manage and adapt to the impacts of a rapidly changing climate, we have to improve and evolve the tools we need to make sense of our world,” he says.

Cover image: Toronto Flood by Angelina Earley Flickr. CC BY-NC-ND 2.0 Deed.