-
- H11M-1044: Integrating Diverse Meteorological Inputs for Machine Learning-Based SWE Estimation in Complex Catchments
-
Board 1044‚ Hall EFG (Poster Hall)NOLA CC
Author(s):Generic 'disconnected' Message
Emily Golitzin, University of Utah (First Author, Presenting Author)
Ryan Johnson, University of Utah
Dane Liljestrand, University of Utah
Ethan Ritchie, Colorado School of Mines
Kaitlin Meyer, University of Utah
Snow in the mountains acts as a seasonal reservoir, storing water in the form of snow in the winter and releasing water downstream as the snow melts in the summer. For areas that rely on mountain snow for water supply, it is important to know how much water, called snow-water equivalent or SWE, is stored in the snowpack. Complex mountain terrain makes it challenging to observe or compute this quantity directly. Machine learning can simplify the modeling process by using a large amount of data to learn statistical relationships between quantities that are easier to observe and then making predictions based on that information. We use a machine learning model trained on multiple large-scale datasets to estimate SWE across mountainous watersheds. We find that training the model with two different sources of precipitation data, one that is less detailed and one that is more detailed, does not significantly change model performance in areas where there is already a lot of information. However, having more detailed data helps the model make better predictions in areas where there is less training data overall, so that dataset may help expand the model to entirely new areas.
Scientific DisciplineNeighborhoodType
Enter Note
Go to previous page in this tab
Session
