Probablistic Operational Inflow Trace Development
* Can scaled historical datasets of reservoir inflows be used in a probabilistic framework to create quantitative uncertainty estimates associated with reservoir outflows?
Reclamation reservoir managers are faced with the question of how much water to release from storage facilities that will meet specific criteria. The criteria are often to maximize the potential of filling the reservoir by a certain date and minimizing the potential of spilling water. This requires some knowledge of how much and when water will be entering the reservoir. The current state of practice, within the Pacific Northwest (PN) Region specifically, is to use historical years that are "similar" to the current situation as traces into an operational model. Given the length of the historical record, this approach often leaves a couple of possible scenarios for analysis. It is possible to enhance the number of scenarios that can be analyzed using historical years in a probabilistic framework.
Need and Benefit
Inflow traces are, simply stated, time series of discharges that could potentially enter a reservoir. The existing capabilities within Reclamation, the PN Region specifically, is the use of an operational spreadsheet in which historic flows are entered as possible realizations of inflows in a year similar to the historic year. All work is done, comparing snowtell gages, rainfall, and flows to historic values by hand and visual estimation and selection of possible realizations is limited to a couple of manageable years. In reality, while the historic years represent possible realizations of what is to come they are by no means exhaustive realizations, nor can they provide a probabilistic estimate of uncertainty associated with outflow selection.
There exist ensemble operational traces developed by the National Weather Service (NWS) available for use. The River and Reservoir Operations Group in Boise, Idaho have expressed concern over the validity of these ensemble traces due to inconsistencies with volumetric forecasts both within Reclamation and the NWS. In short, there is a lack of confidence in these ensemble traces and, therefore, they are not used. The River and Reservoir Operations Group has also expressed concerns with developing ensemble traces using purely stochastic techniques because of fears of stakeholders' comprehension and utilization. Reclamation needs the ability to make probabilistic estimates of the uncertainty of spilling and filling a storage facility given an outflow selection. The method for making these probabilistic estimates must be accepted by the user and stakeholders. The proposed research is a step towards pure stochastic ensemble traces and will give Reclamation the ability to make probabilistic estimates of uncertainty associated with water releases from storage facilities.
Contact the Principal Investigator for information about these documents.