Improving volume forecasting tools for snow dominated basins
Statistical volume forecasts are a critical aspect of reservoir operations within Reclamation. Currently, Reclamation offices across the PN and GP regions use a variety of methods and models to produce volumetric forecasts. Many of these forecasts are not evaluated for accuracy or uncertainty, nor have many been reviewed for statistical rigor. The lack of standardization and forecast validation has likely resulted in less-than-ideal reservoir operations in the mountain west.
In 2018, the GP Region developed the PyCast Software application to address some of the aforementioned issues. The software provides an interactive tool that allows users to search-for and download relevant hydrologic datasets and develop well performing statistical runoff forecast equations using a variety of regression techniques. The software uses a set of novel search algorithms to find skillful and predictive forecast equations using from the complete set of hydrologic data, and allows users to analyze groups of forecast equations for similarities and outliers.
This project seeks to: (a) evaluate the capabilities of the PyCast software, (b) develop additional functionality for the program, and (c) determine how remote sensing products developed in other S&T projects can improve forecast skill. The software will be evaluated by generating hindcasts in GP and PN basins, as well as conducting real-time forecasting during the 2019, 2020, and 2021 seasons. Developers from the PN and GP regions will continue to develop the software, train users, gather user input and suggestions, and add additional features and compatibility. New snow water equivalent (SWE) and snow covered area datasets proposed under two concurrent GP and MP S&T projects will be incorporated into the software and the resulting forecasts will be compared to traditional snow products from the NRCS.
Need and Benefit
Reservoir operations are critical to Reclamation's mission of delivering water and power to stakeholders. In basins
where it is important to provide flood protection as well as store water for the coming irrigation season, it is critical to
have accurate forecasts of the runoff volume. Statistical forecasts have been used for decades for this purpose, but
have not been modernized or updated in many cases since their original development. Leveraging the PyCast tool
developed by GP, this project will continue to modernize the statistical forecast process as well as look for additional
improvements to forecast skill, increasing accuracy and improving Reclamation's ability to continue to meet its
mission. This project is submitted in response to the Pacific Northwest Regional DIrector Needs project, Improving
Volume Forecasts using Remote Sensing Data.
Contact the Principal Investigator for information about partners.
Contact the Principal Investigator for information about these documents.