Header for Extended Forecast Description Page

Description of Single Trace Extended Forecast Procedure

Background
The Northwest River Forecast Center (NWRFC) produces long-range hydrologic forecasts critical for planning and regulating the Columbia River Basin. The importance of these forecasts has increased during the past decade along with the complexity of issues surrounding water resource management.

The following description will describe the development of the long-range single hydrologic trace used by Columbia Basin Technical Management Team to negotiate water utilization in the Columbia River basin. In the past, this trace was produced by the Streamflow Synthesis and Reservoir Regulation (SSARR) modeling system. The NWRFC, in the fall of 2001, completed the transition to the National Weather Service River Forecast System (NWSRFS). This new technology will be used in the production of an extended 45 to 120 days single trace forecast. This is an interim procedure and will be replaced by the Ensemble Streamflow Prediction (ESP) component of NWSRFS.

Development
The Extended Forecast Procedure uses historical temperature and precipitation information combined with current hydrologic model states, reservoir regulation, and short-term meteorological forecasts to produce a single deterministic hydrologic trace out 45 to 120 days.

The first step in the development of the Extended Single Trace procedure was to build a historical trace which represents normal conditions at a forecast point. An analysis was made of the historical Mean Areal Precipitation (MAP) and Mean Aerial Temperature (MAT) for each forecast point using the Extended Streamflow Prediction Analysis and Display Program (ESPADP).

Figure 1. Sample Display of ESPADP
Figure 1. Sample Display of ESPADP

Figure 1. is a sample display from this program. It represents all the MAP traces for water years 1949 to 1993 at Dworshak on the Clearwater River. MAPs and MATs were fit to a normal distribution. This distribution was found to best represent the mean historical trace with the least amount of annual bias. The table below shows this comparison for Dworshak Reservoir:

CALIBRATION HEADER INFORMATION 
Calibration File Name: dwri1iu.map 
Output Variable: Sum 
Data Type: 
Units: IN 
Analysis Period: 10/1/2002 6 - 9/30/2003 24 Z 
Interval: 10/1/2002 - 9/30/2003 Z 

Statistics based on all years. 

EXCEEDANCE PROBABILITY ESTIMATES 

     dwri1iu.map: Mean: 52.43, StdDev: 0.00 
     dwri1il.map: Mean: 42.93, StdDev: 0.00 

SAMPLE POINTS 
                   dwri1iu.map                dwri1il.map 
Trace    Year       Data Exceed.    Year       Data Exceed.  
 year  Weight      Point   Prob.  Weight      Point   Prob.  
 ------------------------------------------------------------
  1956   0.500     52.430   0.667   0.500     42.930   0.667 
  1957   0.500     52.430   0.333   0.500     42.930   0.333 

(Calibration) 
basin MAP          52.44                      42.91 

Figure 2. Sample of 50 Percent Exceedence Trace
Figure 2. Sample of 50 Percent Exceedence Trace

Figure 2. is an example of the 50% exceedence trace for the upper zone MAP at Dworshak. Note the trace labels of 1956 and 1957, these are required to run the ESPADP program and do not represent a specific trace or year.

Figure 3. Represents a Distribution for one 6 hour Period of Precipitation
Figure 3. Represents a Distribution for one 6 hour Period of Precipitation

Figure 3. represents a distribution for one 6 hour period on 12-01 between 0000-0600 hours. The diamond markers represents the empirical distribution and the straight line the normal. The 50% exceedence value for the empirical distribution is ZERO inches while the normal 50% exceedence is about 0.04 inches..

Figure 4. Sample of 6 Hour Exceedence Histogram
Figure 4. Sample of 6 Hour Exceedence Histogram

Figure 4. is a 6 hourly exceedence histogram plot for Dworshak. You can see from this plot that using the 50% exceedence trace, precipitation will be applied almost daily.

To aid in the quality control or validation process of forecast generation, the historical observed runoff was analyzed to produce daily summary hydrographs. These were data based and are used to relate one forecast point with another.

The final step of the development was to piece together the mechanical processes to generate, review, regulate, and distribute these forecasts.


Operations
The forecast process begins with the short-range Operational Forecast System (OFS) of the NWSRFS. During this process, the forecaster reviews and makes modifications to the runs insuring all forecast points are tracking reasonably. Between January 1st and June 1st, the snow updating system will check and adjust the snow states of the snow simulation model of OFS. A copy of the current model states is saved. Model states reflect the current channel conditions, soil moisture, and snow accounting. During this phase, short-range reservoir regulation is set by the Corps of Engineers (COE) and the Bureau of Reclamation (BOR).

The ESP component of NWSRFS is then used to combine the short-term temperature and precipitation forecasts (10 days) and the historical 50% MAP and MAT. After a review and validation process by the NWRFC, the COE and BOR set regulation for the longer term. The final step is to run single trace procedure and disseminate forecast products.

Several products are available for the web. 1) a text forecast with detailed values for each day at a specific forecast point, 2) a comma delimitated forecast file to facilitate use in spread sheet programs, 3) plots of the long-range forecasts with their summary hydrographs, and 4) a text summary of data found on the plots.


Summary
1. - The Extended Forecast Procedure is a reasonable forecast product within the forecast time scales. See samples forecasts generated during the development stage, figures 5 and 6.

Figure 5. Sample of Extended Trace Simulation
Figure 5. Sample of Extended Trace Simulation

Figure 6. Sample of Extended Trace Simulation
Figure 6. Sample of Extended Trace Simulation

2. - There is an apparent long-term bias when forecasting beyond the 120 day forecast. Weekly model updates and the inclusion of 10 day meteorological forecast minimize any effects throughout the forecast time scale.

3. - Produces consistency between forecasts.

4. - Improved product suite. The extended forecast products include a plot of forecast, comma delimitated forecast for use in spreadsheet software, and the previous forecast table provided under SSARR.