A few years ago, an epidemiologist studying the spread of West Nile virus in Ontario needed a reliable way to pull together years of water-level data. Mosquitoes, which carry the virus, breed in standing water. How much of that water is around depends on rainfall and the seasons. Making sense of that pattern meant digging up records reaching back further than any single summer. That search led to a small, decade-old software tool built for an entirely different purpose, written by a software developer named Sam Albers to make Water Survey of Canada data easier to access.
Nobody designed the tool with mosquitoes in mind.
That's largely the point.
In the summer of 2025, the Fraser River ran low, with flows below normal in June, July, August, and September. By itself, one dry summer isn't remarkable. Rivers fluctuate. But at Hope, British Columbia, home to the river's longest-running gauge, the picture goes back further: mean monthly discharge in July, August, and September has been on a statistically significant downward trend across 113 years of records, from 1912 through 2025.
That distinction is what ecosystem scientist Ian Giesbrecht has spent the past year thinking about. River flow affects more than many people realize. It impacts the salmon that move between fresh water and ocean over the course of their lives, decisions about water withdrawal and flood planning, and judgments about what restoration should look like. All of it depends on knowing whether conditions fall within a normal range or are shifting somewhere new. Answering that well requires more than a single measurement; it requires being able to place this year alongside the past decades.
"We're seeing change in the water cycle, change in the climate, change in rivers," Giesbrecht says. "Their averages and their extremes."
It's also, Giesbrecht thinks, a moment when a few things are converging. The change itself has become harder to miss: the same Hope gauge record that shows summer flows declining also shows a statistically significant rise in winter flows, from January through April. Giesbrecht says the pattern matches what climate studies across western North America have been documenting: warmer winters pushing snowmelt earlier and shifting rivers toward higher winter flows and lower summer ones. At the same time, software for combining data from many different sources has improved enough that a national-scale analysis, once a multi-year undertaking, can now be assembled far more efficiently. Funding and policy have caught up, too: the Canada Water Agency, which helped fund this work, released a national freshwater data strategy and a freshwater science agenda earlier this year identifying water data sharing as a national priority.
By the time Albers joined the Hakai Institute, he'd already spent close to a decade building and maintaining tidyhydat, the package the West Nile epidemiologists later used. It gave hydrologists a direct route to Water Survey of Canada data through R, the statistical programming language many of them already used for analysis.
At a recent conference, Albers was surprised by how widely the package had been adopted across the Canadian hydrology community.
Free Flow Canada, where he leads the technical development at Hakai, asks a related but distinct question: how can we access all the freshwater data that is available outside the Water Survey of Canada network?
In many cases, the information people need already exists. It's collected by provincial governments, hydroelectric utilities, conservation groups, Indigenous water programs, and other groups across the country, each publishing in its own format, on its own systems. Accessing that data from those sources often meant writing a unique piece of code for each source.
hydrocan, the package Albers built in 2026 through Free Flow Canada, aims to reduce some of that friction. It reaches into the open data services that organizations already maintain and hands back a single, usable dataset with code that is consistent across data sources.
"Through hydrocan and tidyhydat, we connect R users to hydrometric data," Albers says.
The agencies that collect the information remain responsible for it: hydrocan doesn't store or alter anything, just provides a consistent way to access it. That was a deliberate choice.
Albers is blunt about the alternative: persuading every province, utility, and monitoring organization in the country to defragment the data landscape is "certainly probably a worthy goal," he says, but not one he's interested in chasing. Free Flow Canada works with the fragmentation as it exists. Today the package connects four data providers, including the governments of British Columbia and Quebec, Hydro-Québec, and Hakai's own river-gauging network on the central coast. Because hydrocan is open source, other agencies and developers can add their own data sources over time. New contributors are already in conversation about doing exactly that.
Albers sees hydrocan's narrow scope as one of its strengths: a small, working tool, deliberately not trying to be everything to everyone, that other people are now free to build on if they choose to.
The wider Free Flow Canada project, meanwhile, is also being developed with partners across the country. On the central coast, the Central Coast Indigenous Resource Alliance is helping shape the project so its tools better support First Nations water stewardship and decision-making. That work ultimately reaches the Guardians and technicians out on the water taking measurements, and the decision-makers relying on what they find.
On the opposite side of the country, scientists at Memorial University in Newfoundland and Labrador and their collaborators are applying the same approach in Nunatsiavut, Labrador, a vast stretch of coastline where many individual watersheds have little or no flow monitoring of their own. Communities there are building datasets for the specific rivers they rely on, but in the meantime, tidyhydat and hydrocan offer an easier way to draw on data collected across that larger surrounding region. Once local records exist, the same tools make it easier to share that information back out, for others to draw on in turn.
Hakai's interest in any of this, Giesbrecht points out, didn't begin with rivers. It began with the ocean. Watersheds deliver fresh water, nutrients, sediment, and carbon to the coast, making rivers an essential part of understanding what happens in the coastal ocean. That coastal perspective, paired with Albers's decade of hydrology software development, gave Hakai both a reason to take on the work and the technical capacity to do it: techniques developed to understand one stretch of the British Columbia coast turn out to be just as useful for watersheds draining into the Arctic, the Great Lakes, or the Atlantic.
The next phase aims to bring the results of that work to a much broader audience. hydrocan was built for analysts working in R; it handles the data acquisition. The next step is a web application that takes what Giesbrecht's analysis produces and makes it accessible to a broader audience: watershed managers, decision-makers, and members of the public who want to understand what their local rivers are doing, no programming required.
"I think that'll be the hardest part," Albers says. Building software for hydrologists is relatively straightforward, because they already understand the data. Building something useful for watershed managers, Indigenous Guardians, and the broader public means answering a much harder question about what information people need in order to make decisions. For now, the scope is deliberately narrowing rather than expanding: the first web application will focus on British Columbia, while Giesbrecht develops its vision and user experience.
It's a problem Albers doesn't think AI will solve out from under them. He's watched an AI model write hundreds of lines of working code to pull data from an open API on the first attempt. What that code can't supply is judgment about what the result should look like, or whether it's asking the right question. That, in his view, is still Giesbrecht's job: knowing what a watershed manager, an ocean scientist, or an Indigenous Guardian needs to see in order to act on it. If anything, Albers thinks tools like hydrocan matter more as AI use grows, not less. A stable, human-maintained interface gives people and AI systems alike a consistent, trustworthy place to get the same answer, rather than each one generating a slightly different script and getting a slightly different result.
Giesbrecht sees that need building from a few directions at once. Stream drought, barely part of the conversation on this coast a couple decades ago, now comes up regularly, even in places that get some of the heaviest rainfall in the country. There's more data available to catch long-term trends, and more confidence that what shows up in it reflects something real rather than noise. Water management carries economic and policy pressure that didn't exist a generation ago. And government data, increasingly, is published openly by default rather than released on request. Taken together, these changes are why people here now care as much about what a river is becoming as what it's doing today.
In some cases, the data has existed all along. It was scattered across systems that weren't always built with integration in mind. Free Flow Canada is an attempt to make it easier to find, so that communities, decision-makers, and scientists across the country can better understand what Canada's rivers are telling us.
