Precipitation-runoff and streamflow-routing models for the Willamette River Basin, Oregon

Abstract Precipitation-runoff and streamflow-routing models were constructed and assessed as part of a water-quality study of theWillamette River Basin. The study was a cooperative effort between the U.S. Geological Survey (USGS) and the Oregon Department of Environmental Quality (ODEQ) and was coordinated with the USGS National Water-Quality Assessment (NAWQA) study of the Willamette River. Routing models are needed to estimate streamflow so that water-quality constituent loads can be calculated from measured concentrations and so that sources, sinks, and downstream changes in those loads can be identified. Runoff models are needed to estimate ungaged-tributary inflows for routing models and to identify flow contributions from different parts of the basin. The runoff and routing models can be run either separately or together to simulate streamflow at various locations and to examine streamflow contributions from overland flow, shallow-subsurface flow, and ground-water flow. The 11,500-square-mile Willamette River Basin was partitioned into 21 major basins and 253 subbasins. For each subbasin, digital data layers of land use, soils, geology, and topography were combined in a geographic information system (GIS) to define hydrologic response units (HRU's), the basic computational unit for the Precipitation-Runoff Modeling System (PRMS). Spatial data layers were also used to calculate noncalibrated PRMS parameter values. Other PRMS parameter values were obtained from 10 nearby calibrated subbasins of representative location and character. About 760 miles of the Willamette River system were partitioned into 4 main-stem networks and 17 major tributary networks for streamflow routing. Data from time-of-travel studies, discharge measurements, and flood analyses were used to develop equations that related stream cross-sectional area to discharge and stream width to discharge. These relations were derived for all 21 stream networks at approximately 3-mile intervals and used in the Diffusion Analogy Flow model (DAFLOW) in streamflow routing. Ten representative runoff models and 11 network-routing models were calibrated for water years 1972-75 and verified for water years 1976-78. These were the periods with the most complete and widespread streamflow record for the Willamette River Basin. Observed and estimated daily precipitation and daily minimum and maximum air temperature were used as input to the runoff models. The resulting coefficient of determination (R2) for the representative runoff models ranged from 0.69 to 0.93 for the calibration period and from 0.63 to 0.92 for the verification period; absolute errors ranged from 18 to 39 percent and from 27 to 51 percent, respectively. Bias error for the runoff modeling ranged from +13 to -32 percent. Observed daily streamflow data were used as input to the network-routing models where available, and simulated streamflows from runoff model results were used for ungaged areas. Absolute error for the network-routing models ranged from about 21 percent for the Molalla River model, for which 70 percent of the subbasin was ungaged, to about 4 percent for the Willamette main-stem model (Albany to Salem), for which only 9 percent of the subbasin was ungaged. With an input of current streamflow, precipitation, and air temperature data the combined runoff and routing models can provide current estimates of streamflow at almost 500 locations on the main stem and major tributaries of the Willamette River with a high degree of accuracy. Relative contributions of surface runoff, subsurface flow, and ground-water flow can be assessed for 1 to 10 HRU classes in each of 253 subbasins identified for precipitation-runoff modeling. Model outputs were used with a waterquality model to simulate the movement of dye in the Pudding River as an example application in the Willamette River Basin.