Lucy Myrol
HS Hydrological Sciences
The 2026 Outstanding Student and PhD candidate Presentation (OSPP) Award is awarded to Lucy Myrol for the poster/PICO entitled:
Causal non-stationarity does not explain seasonal LSTM forecast degradation (Myrol, L.; Adamowski, J.)
Click here to download the poster/PICO file.
Lucy Myrol is an MSc student in Bioresource Engineering at McGill University in Québec, Canada. Her research focuses on the causal interpretation of deep learning models for water-level forecasting in the Laurentian Great Lakes.
At the EGU General Assembly in 2026, she presented research investigating why a deep learning model used to forecast water levels in Lake Erie performed much worse in winter and autumn than in spring and summer. Her results indicate that non-stationarity in either the marginal distributions or the conditional distribution between hydrometeorological variables and water level does not explain this degradation, and that non-stationarity is highest in the best-performing season (spring) and lowest in the worst-performing season (winter). This counterintuitive result suggests that some degree of non-stationarity in the predictor–response relationship is either tolerable to, or in some cases potentially beneficial to, deep learning models.