Videos

PhD Defence: "Mountain glaciers as modifiers of streamflow in Western Canada"

How much can we learn about how glaciers modify streamflow, simply by analyzing historical streamflow datasets?  Who is vulnerable to the loss of glacier ice in Western Canada?  By how much -- and in what ways -- do glaciers control the streamflow response to heatwaves?  All that and more in my PhD defence presentation, titled: "Mountain glaciers as modifiers of streamflow in Western Canada: Insights from data analysis and machine learning".

Resilient rivers: Glacier controls of post-heatwave river flows

The unprecedented June 2021 heat wave revealed strikingly different behaviours in glacier-fed rivers as compared to non-glacier-fed rivers.  Motivated by these observations, we ask: when, where, and by how much do glaciers control post-heatwave streamflow in British Columbia?

Originally presented to the Water and Environment Student Talks 2022 conference (June 2022).

Using glaciers to interpret the decision-making of deep learning hydrological models

Deep learning models have been found to be excellent at streamflow prediction tasks, but their success begs question: why?  What is it that these models are learning?  Here we use glacier runoff to help interpret the decision-making of deep machine learning models in Western Canada.

Originally presented to the Canadian Geophysical Union on June 1, 2022.

Machine learning for modelling regional streamflow: What, where, when, and why do machines learn?

Deep machine learning models might be good for predicting river flows, but the question is: why?  What is it that they are actually learning?  Here we investigate models trained in southwestern Canada to see where they focus, on which variables they focus, and how these features vary through time.

This talk was presented to the Department of Earth, Ocean, and Atmospheric Sciences at the University of British Columbia on May 19, 2021.

Deglaciation in Alberta: Lessons from the climate crisis in Oil Country

I reflect on lessons I've learned by studying the impacts of climate change on glaciers and water supply in Alberta. I discuss: how does messaging from fossil fuel companies shape the science questions I've asked and the communications strategies I've chosen?

(Talk begins at 38: 52)

Deep learning for streamflow prediction in Western Canada (AGU 2020)

In this study, we design a spatio-temporal deep learning model which both: 1) performs well according to traditional hydrological modelling metrics, and 2) learns physically realistic features. 

We find that different streamflow regimes are simulated with varying degrees of success, with mountain-runoff dominated streams having the best performance.  We show that the model tends to be most sensitive in the areas near the stream gauge stations being predicted for each streamflow regime, providing evidence that the model is automatically learning the areas which are physically relevant for streamflow prediction.  This work marks a step towards the development of interpretable deep learning hydrological models.