-
- H12E-06: Deep Learning Advances Arctic River Water Temperature Predictions
-
NOLA CC
Author(s):Generic 'disconnected' Message
Shuyu Chang, Franklin & Marshall College (First Author, Presenting Author)
Jonathon Schwenk, Los Alamos National Laboratory
Kurt Solander, Los Alamos National Laboratory
The Arctic is warming rapidly, altering river flow and river water temperatures, which poses risks to local ecosystems. Despite the significance, limited data and methods have made it difficult to predict river water temperatures in this region. The newly released AKTEMP data set and advancements in machine learning offer a chance to improve these predictions. We developed a Long Short-Term Memory model to predict river water temperatures in Alaska, a region heavily impacted by snow and permafrost. Our model performed better than existing models for similar high-latitude regions. We used advanced techniques to understand what the model had learned and found that air and soil temperature, sunlight, soil moisture, and snow are the most important processes for predicting river water temperatures. Glaciers and permafrost were also found to be important factors, especially when predicting river water temperatures in spring and summer. Our findings suggest that river temperature is shaped by a mix of weather and environmental factors that operate over different timescales, from days to months. As the Arctic undergoes rapid environmental transformation, our model offers reliable predictions that could help support sustainable fisheries.
Scientific DisciplineSuggested ItinerariesNeighborhoodType
Enter Note
Go to previous page in this tab
Session
