A Copula-Based Water Supply Forecasting System for Large Lakes

This research focuses on the application of Copula functions for water supply forecasting in large lakes. Many current forecasting systems use historical water supply data to create deterministic forecasts of water levels. However, we have found that this approach leads to forecasts that underrepresent the actual variability of seasonal water levels. In contrast, we propose a new approach that fits a Copula function to the data rather than directly applying historical data sequences. This approach provides a much more robust set of possibilities for water supply components. The Copula function has the additional benefit of analyzing the relationship, dependence, and distributions between precipitation, evaporation, and runoff to produce thousands of water supply sequences that each have similar dependence and distributions to historical water supply patterns. Thus far, we have tested this technique on the Laurentian Great Lakes, and have been able to reasonably reproduce water levels. Future plans for this research include manipulating the means of the Copula function water supply variables to reflect the effects of climate change, and application to a broader set of large lakes.