Improving watershed models for northern Canada
Water is essential to Manitoba, from agriculture to drinking water to hydroelectric generation, but it can also be dangerous; floods and droughts are both threats to our province.
We can reduce the negative impacts from both too much and too little water with good water management and appropriate infrastructure, both of which depend on accurate predictions of future water supply.
With a changing climate, we should no longer assume that our future river flows will continue as they were in the past. Instead, water managers are turning to computer models of watersheds to predict floods and plan for long-term hydroelectric development.
However, modeling rivers accurately is a significant challenge in regions where streamflow and weather observations are sparse, and/or have limited data record lengths.
Data scarcity is particularly acute in northern Canada, where the region’s remoteness and limited accessibility restricts the expansion of data networks. We can use models to simulate flows in these northern watersheds, and reproduce flows from the historical record, but will these models predict future flows accurately? We need models that can get the right answer for the right reasons, accurately representing the physical processes generating streamflow, in order to predict flows under different climatic conditions. Information on individual processes like evaporation, however, is even less common than weather or flow data—and is even more costly and less feasible to obtain across large, remote regions.
To get more information on the sources of river flow, we’ve turned to stable water isotopes as tracers. These naturally occurring water molecules are slightly heavier than normal water, but that small mass difference leads to different isotope concentrations in rain and snow, and lets us identify how much water has been lost to evaporation. It’s also relatively easy to collect stable water isotope data in remote regions, with just a sealed grab sample needed from the field.
By simulating not only water in a watershed model, but also stable water isotope concentrations, I can identify models that have the wrong amount of snowmelt coming into the river, or lose too much or too little water to evaporation.
Using both flow data and stable water isotope data to evaluate if a model can accurately simulate historical flows improves our confidence that the model is giving the right answer for the right reason. Now I’m trying to find out if this increased confidence will reduce the uncertainty in simulated river flows. Predicting the future is easy, getting your prediction right is hard; I hope to make predicting our future water a bit easier.
Story originally published in ResearchLIFE Summer 2018 Edition. Read the full magazine online.
Research at the University of Manitoba is partially supported by funding from the Government of Canada Research Support Fund.
What if you guys extract data from AMSR-E Snow water equivalent and compare it with in-Situ dat for minimum 15 years and see if they have a minimum difference of around 5-10mm then we will be able to pridict flooding somehow by simulation of making a model.