Water Supply Forecasts based on Multiple Regression

INTRODUCTION - Water in the West almost never flows unobstructed from where it fell and melted (since it is often snow) to where it will rest in an ocean or lake. Rather it is rerouted through tunnels and ditches, diverted onto croplands and into treatment plants, and temporarily impounded in reservoirs many times along the way. To keep the science of hydrologic forecasting and the engineering of water management separate, the River Forecast Center (RFC) and its associates forecast adjusted or unimpaired runoff, not observed or regulated runoff. The relationship between observable hydrologic parameters (e.g., precipitation, snowpack, terrain, etc.) and adjusted runoff is predictable and well-defined; the same cannot be said for regulated (or managed) runoff. In the Pacific Northwestadjustments are included only if they exceed 10% of the volume at theforecasted site.

The Northwest River Forecast Center (NWRFC) currently uses a statistical regression model to make water supply forecasts for the Columbia River Basin; the coastal streams of Washington and Oregon; and the Great Basin of Oregon. Multi-regression is a functional relationship between two or more correlated variables. Seasonal volumes are related to snow water equivalent, monthly precipitation and in some cases previous streamflow volumes.

From January through June, there are three forecasts produced every month. Each forecast has a varied amount of observed data available for model input and some diminution of forecast results may occur. Each process is described below:

MONTHLY FINAL FORECAST - At the beginning of each month starting in January, the NWRFC in conjunction with the Natural Resources Conservation Service and other cooperating agencies put together seasonal volumetric water supply forecasts for the Pacific Northwest. These forecasts are based on precipitation reports from over 400 sites. Also included in these procedures are snow water equivalent and observed runoff values from all available sites in Oregon; Washington; Idaho; Western Montana; Western Wyoming; Northern Nevada and British Columbia, Canada.

MID-MONTH FORECAST - At approximately the mid point of each month, the NWRFC makes updated water supply forecasts. These forecasts are at the request of participating government agencies in the Columbia Basin. The agencies are monitoring for significant departures especially as the melt season approaches. For these forecasts only about half the precipitation reports used in the monthly final are available. A short term assessment of future precipitation is made with the aide of meteorological models. There are no updated snow or runoff values used directly in the model input but all available snow water equivalent reports are used subjectively by the forecaster during the review of forecast results. These forecasts are intended to show trends in water supply volumes and caution should be used when applying these forecasts to any specific site.

EARLY BIRD FORECAST - Because the Columbia Basin is an international river system, treaty requirements specify that coordination between Canada and the United States takes place early in the month. Since the monthly final is a fairly lengthy process, NWRFC makes what has been termed as the 'Early Bird' forecast toward the end of month to provide a best estimate of what next months final will look like. For these forecasts only about half the precipitation reports used in the monthly final are available. A short term assessment of future precipitation is made with the aide of meteorological models to complete the monthly total. All available snow water equivalent values and observed runoff reports are used and from these, monthly values are estimated. These values are then input and used in the forecast model results.

CONTINGENCY FORECAST - These forecasts are developed assuming future precipitation will be adjusted by a specific percentage (50%, 75%, or 125%). This percentage is then apply uniform ally to the precipitation inputs from the forecast date through the end of the water year or September 30th. Temperature is not taken into consideration.





Water Supply Forecasts based on Ensemble Prediction System

METHODOLOGY - Ensemble Prediction System (ESP) is a modeling component of the National Weather Service River Forecast System (NWSRFS). ESP produces long-range probabilistic forecasts of hydrologic variables. ESP utilizes a physical based modeling system to simulate soil moisture, snow pack, regulation, and streamflow. ESP then accesses the current hydrologic model states, and uses historical meteorological data to create equally likely sequences of future hydrological conditions, each starting with the current hydrological conditions. Statistical analysis is performed on these sequences to generate probabilistic forecasts of seasonal water supply.

NWSRFS is a continual simulating model. ESP can take advantage of this constant updating to provide water supply estimation weekly and through out the year. Future, implementations will make use of the information available in the Climate Prediction Center estimation of precipitation and temperature to shift the forecasts based on our understanding of future meteorological conditions.

The ESP water supply forecast have been constructed to predict as close as possible the same adjusted volumes forecasted using regression techniques. These forecasts do not always represent true natural streamflow conditions. Adjustments for diversions and regulations are only partly taken into account.





ESP Natural

METHODOLOGY - Produced the same as ESP water supply, ESP natural represents natural volume forecasts. "Natural" assumes no regulation or diversions take place above the forecast point. This method takes into account all adjustments above a forecast point. In many cases, ESP natural and ESP water supply will be the same.

ESP natural does not combine the observed data with forecasted data to forecast a period which may extend into the past. It always looks into the future. ESP natural is run weekly and through out the year.