A Physically-Motivated Regression Approach to Forecasting Lake Powell Inflow
AuthorPotteiger IV, Samuel Edwin
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractThe Colorado River provides water for a growing population in seven U.S. states and Mexico, making it undoubtedly one of the most important rivers in the entire world. Lake Powell is the second largest reservoir on the river and is responsible for much of the water which enters Lake Mead. Forecasts of the water supply to Lake Powell are relied upon immensely by water users for planning purposes. The UA model was developed in this study to take a physically-motivated regression approach to modelling Lake Powell Inflow. The UA model has two parts: the water year estimation method and the inflow model. The water year estimation method uses physical observations to make a binary forecast indicating if a given water year is a high-flow year or low-flow year based on the water year’s peak monthly inflow or April-July volumetric flow. The water year prediction is first made on December 1st and provides information for a seven-month forecast. This method was able to predict all high-flow years throughout the 1982-2016, whereas the official forecasting center’s January 1st forecast predicted less than half. Most flow comes to Lake Powell in the April-July period, and the operational forecast center releases a forecast for this period on April 1st. The UA model forecast made on April 1st has a 29% lower RMSE for the 1982-2016 period. Sensitivity tests indicate that the use of snow water equivalent in our model is an important reason for the UA model’s good performance. The UA model and the operational model do not have a statistically significant difference in performance in the 2013-2016 period at all lead times except for the 4-month lead time. The operational model has statistically significant better performance at the 4-month lead time during this period.
Degree ProgramGraduate College