5C: Exploring Bear Lake’s future through AI

Water levels in Utah lakes and reservoirs are affected by complex cycles and interactions between weather and land use. While traditional computer models can explain some of these complexities, implementing scenarios for decades-long forecasts of future climate conditions can be challenging and time consuming. One promising alternative is a process called machine learning emulation. This technique uses advanced data-driven models such as artificial intelligence to predict different variables within a complex environmental process by training the artificial intelligence using past data. These emulators can then make predictions in a fraction of the time compared to more traditional computer models. 

Machine learning emulation was used to predict Bear Lake’s annual maximum and minimum water levels through the end of the century under different future climate scenarios. First, researchers “taught” the model using climate measurements from the last 40 years. Then, they used the models to project possible climate outcomes. The emulation results are available in a publicly accessible data dashboard.

Climate models that only accounted for annual snowpack in their analysis predicted lower average lake elevations in the future, as observed in (Figure 5.C.1). Conversely, models that used total precipitation (i.e., rain and snow combined) predicted higher average lake levels. Further research is needed to understand how lake levels might be affected by more rain and less snow. 

The public can access these data models and become a “citizen climate data analyst.” Allowing people to see what happens when temperatures become warmer, snowpack diminishes, and more rain and evaporation occurs helps them better understand the complexity of the Bear Lake watershed.