Close Window

Univ. of Virginia: "Probabilistic river stage forecasting"

Final Report

The Weather Service Forecast Office (WSFO) in Pittsburgh and the Systems Engineering Department of the University of Virginia have collaborated under the auspices of COMET since September 1990. The previous Cooperative Project resulted in the development and operational testing of a prototype system for probabilistic quantitative precipitation forecasting (PQPF). The objective of the current project, to develop and test a prototype system for probabilistic river stage forecasting (PRSF), was a logical extension of the previous work. Probabilistic QPFs are now routinely prepared by the Pittsburgh WSFO, and these provided input for the PRSF system that was implemented at the River Forecast Center (RFC) in Wilmington, Ohio. Thus the proposed project: a) fully utilized products from the previous project, b) supplied the necessary means of testing the value of PQPFs in river forecasting, and c) effectively created the first operational prototype of an integrated probabilistic hydrometeorological forecasting system.

In October 1996, the first PRSFs for lead times of one, two, and three days were calculated for the gauge in Eldred, PA, located on the upper Allegheny River. These PRSFs were produced from routine NWS operational data and models using the PQPFs as inputs. Another important requirement that was met was that the PRSFs be produced within the operational constraints on computing and processing time. Although the implementation of the methodology was for headwaters basins only, the underlying theory is general and applies to any basins. The project has demonstrated that it is feasible to produce PRSFs and has provided a prototype methodology that can be implemented within the operational environment of the RFC. Thus the system can be viewed as a methodological enhancement to the NWS modernization efforts.

It is generally agreed that improvements to river stage forecasts should aim at three objectives:

In addition to offering a longer lead time and unbiased estimates of river stages, a PRSF allows users to consider a tradeoff between the forecast uncertainty and the lead time. In effect, risks can be explicitly accounted for in decision making. This additional information is expected to increase economic benefits of forecasts for emergency management agencies, reservoir managers, and waterways and barge operators during significant hydrologic events, particularly floods. Likewise, a PRSF will allow the NWS to cast flood outlooks, watches, and warnings in a quantitative, probabilistic format, which is more informative and effective in conveying predictions of rare and severe weather events.

The PRSF system consists of four subsystems: precipitation forecast processor, river forecasting methodology, river climatological guidance, and river forecast verification. The precipitation forecast processor performs quality control tasks, reconciling any significant discontinuities between PQPFs for adjacent river basins, generating a single PQPF for the entire RFC service area, and computing a time series of basin average precipitation amounts that are input into a hydrologic model. The river forecasting methodology is a hydrologic-statistical forecasting system, the purpose of which is to transform the PQPF for a river basin (or its sub-basins) into a PRSF for a station at the basin outlet. One important feature of the model is that it decomposes the total forecast uncertainty into hydrologic uncertainty and precipitation uncertainty, which are independently estimated and then integrated. The purpose of the river climatological guidance is to provide prior statistics and distributions of river stage time series for each river station and each month of the year. The river forecast verification process consists of operational testing of the prototype PRSF system for gauging stations for which the WSFO prepares PQPFs.

The results of the PQPF and PRSF projects sponsored by the COMET Outreach Program have served as key inputs into national plans for developing an integrated probabilistic hydrometeorological forecast system. The results will be used to assemble a prototype of an integrated probabilistic hydrometeorological forecast system. The system will be operated to demonstrate and test the basic concept of the end-to-end probabilistic forecast process. The ultimate objective of this demonstration is to validate the scientific basis for designing a probabilistic hydrometeorological forecast system for national implementation. This system is envisioned to become a part of the Advanced Weather Interactive Processing System (AWIPS) of the modernized NWS.