Clearing the path to project snow on a continental scale
SNOW AND PERMAFROST - The Canada1Water project uses the Community Land Model version 5 (CLMv5) to recalculate snow depth and permafrost distribution, calibrated separately from the hydrologic models.
Precipitation is crucial to any climate model — and in Canada that often means snow. While snow can be modelled accurately on small scales with good weather data, projecting its effects continent-wide and far into the future is a whole other story. Senior Climate Scientist Andre Erler and his team are solving the puzzle by coupling innovative bias correction with climate projections to enable Canada1Water’s long-term view.
Precipitation is generally tricky because climate models reproduce time/space patterns well but have systematic temperature and precipitation biases that skew their results. While temperature adjustments are relatively straightforward to make, precipitation — and snow in particular — poses all kinds of challenges.
“A temperature bias can cause a bias in snow,” Erler explains. “Other precipitation biases can bias snow. Depth measurements are easy but not so useful; knowing snow density is better, but it’s hard to gauge and highly variable. What we really, really want to know is snowmelt — how much water is coming out of the snow — and that’s impossible to measure directly.”
Erler and his colleagues knew they wanted to use modelled snowmelt for Canada1Water given the goal of projecting into the future. But they also needed observational estimates for calibration and validation. Those are hard to come by in Canada, which has relatively few weather stations, especially in the North. Satellites can provide observations of snow cover but estimating the water mass of snow is tough, since snow density can vary between sites by a factor of five.
Forcing the issue with land-surface modelling
Given the challenges, the team decided to take a two-pronged approach, with one group focused on better interpolation and correction of available observational data, and another developing a machine-learning technique to correct models without requiring observations at every point.
The real novelty of the Canada1Water approach, according to Erler, comes from land-surface modelling. After the climate and precipitation projections undergo the two-pronged bias correction, the platform will force a land-surface model to recalculate snow and permafrost, calibrated separately from the hydrologic models.
Andre Erler is a Senior Climate Scientist at Aquanty and climate lead on the Canada1Water project, coordinating all activities related to climate, snow and permafrost.
“With the land-surface model as a ‘post-processor’ we will likely have the best possible projections for snow and permafrost across Canada, which should be very valuable for any kind of northern development,” Erler says.
Innovating by looking back
Lakes also have a significant impact on precipitation and snowfall, but modelling lakes in 3D is complex and expensive, and many of Canada’s northern lakes aren’t characterized well enough for 3D treatment. That prompted the Canada1Water team to resurrect an older approach known as column modelling, which is lightweight, fast and affordable.
“Parts of the scientific community believe the column approach has had its day, but when you add in adjustments for ice cover and wind, we’ve determined you can use it with sufficient accuracy,” Erler says.
Now that Canada1Water’s regional climate model is established, the team is turning attention to generating projections and implementing land-surface modelling.
“Our aim is to produce results at many different levels of interest to the public,” says Erler. “We’re contributing to open-source projects like CORDEX, and our climate projections will be usable for any type of climate change impact study in North America.
“With the land-surface model as a ‘post-processor’ we will likely have the best possible projections for snow and permafrost across Canada.”