Developing a forecasting model for the Great Lakes – UROP Spring Symposium 2021

Developing a forecasting model for the Great Lakes

Jenna Sherwin


Pronouns: She/her

Research Mentor(s): Andrew Gronewold, Associate Professor
Research Mentor School/College/Department: School for Environment and Sustainability,
Presentation Date: Thursday, April 22, 2021
Session: Session 2 (11am – 11:50am)
Breakout Room: Room 17
Presenter: 7

Event Link


The Laurentian Great Lakes constitute the world’s largest freshwater lake system, and support hundreds of different animal and plant species, as well economic activity for many Americans and Canadians living in the Great Lakes basin. However, fluctuations in observed water levels over the past decade indicate a possible shift in regime, leading to uncertainty regarding the future of water levels and water resources in the region. This research sought to develop and expand upon a rudimentary simulation model for forecasting Great Lakes water levels across different time horizons to explore climate scenarios. In R, I used the copula package to generate stochastic series of precipitation, evaporation, and runoff based upon historical observed water supplies records. I also fit a log-linear regression model of water levels on each lake to outflow on each lake. Using the copula produced net basin supply and the log-linear outflow model, I generated a number of possible 12 month forecasts for water levels on Lake Superior, Lake Michigan-Huron, and Lake Erie given their respective beginning of month January water levels. Then the 95% confidence interval was found to create a range of likely water level values for a given future month. This model can be used to explore different long-term and short-term plausible scenarios for future water levels on the Great Lakes, due to perturbations like climate change or the implementation of a diversion of freshwater to other states.

Authors: Jenna Sherwin, Andrew Gronewold
Research Method: Data Collection and Analysis

lsa logoum logo