Scientific Investigations Report 2006–5274
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2006–5274
A modified version of the distributed-parameter, physically-based PRMS (Leavesley and others, 1983) was used to numerically simulate the hydrologic processes that occur in the Salmon Creek Basin. The model was run inside the MMS (Leavesley and others, 1996), which is a modeling system that allows users to develop application-specific models by selecting or creating a set of modules that each represent specific hydrologic processes. Examples of such processes include interception of precipitation by vegetation, evapotranspiration, and snow accumulation and melt.
In addition to the Salmon Creek Basin, MMS and/or PRMS have been used for many other basins in the western United States, including the Yakima River Basin in Washington (Mastin and Vaccaro, 2002b), small basins in the Oregon Coast Range (Risley, 1994), the Willamette River Basin in Oregon (Laenen and Risley, 1997), the San Juan River Basin in Colorado and New Mexico (Kuhn and others, 1998), the Truckee River Basin in California and Nevada (Jeton, 1999), and the upper Rio Grande Basin in Colorado and New Mexico (Boyle and others, 2004).
The modified version of the PRMS (Leavesley and others, 1983) used in this study simulates hydrologic processes that occur in Salmon Creek Basin (fig. 6). The model was developed from a set of 15 modules given in table 1 that are identical to those used in precipitation-runoff simulations for the neighboring Methow River Basin (Ely and Risley, 2001; Ely, 2003). Each module represents either a hydrologic process or reads model input parameters or time series. Of the 15 modules, 5 are standard PRMS modules (Leavesley and others, 1983 and 1996) and 10 are modified versions developed for either the Yakima River Basin study (Mastin and Vaccaro, 2002a and 2002b) or the Methow River Basin study (Ely and Risley, 2001; Ely, 2003). The functionality of the module that reads model input time series of irrigation diversions, irrigation returns, and irrigation applications (divrt_apply_prms.f, table 1) is not used in this study because the precipitation-runoff model is used to simulate unregulated streamflow. A detailed description of the combination of modules used in this study is provided in the documentation of the precipitation-runoff simulations for the Methow River Basin (Ely and Risley, 2001).
The Methow River Basin model was selected so that calibrated parameters from that model could be used as initial estimates for model parameters in this study. Justifications for this approach are that the Methow River and Salmon Creek Basins are adjacent and share some hydrologic characteristics and that multiple long-term records of measured daily streamflows for the Methow River Basin were available for model calibration while none are available for the Salmon Creek Basin. Both the Methow River and Salmon Creek Basins drain mountainous subbasins on the east side of the Cascade Range (fig. 1) that have similar types of vegetation. The elevations in the basins also are similar and range from 775–8,950 ft in the Methow River Basin and from 820–8,250 ft in the Salmon Creek Basin. Some differences between the basins are that the area of the Methow River Basin is significantly greater than that for the Salmon Creek Basin (1,800 mi2 versus 152 mi2) and that mean annual precipitation for the Methow River Basin is greater than that for the Salmon Creek Basin (32 in. versus 21 in.).
Spatial diversity in the model area is represented by simulating the basin as a set of sub-areas, called Modeling Response Units (MRUs), that have similar hydrologic characteristics. Characteristics assigned to each MRU included such parameters as slope and aspect, elevation, vegetation type and summer and winter density, soil type, and percent pervious. In addition to the spatial characteristics, model inputs included time series of measured or estimated precipitation and air temperatures. The model distributes the data from the locations of the climate stations to each MRU by considering differences in elevations and distances between the MRUs and the stations.
At each MRU, the model simulates a sequence of hydrologic processes (fig. 6) at a user-selected time step, which was 24 hours for this study. At the end of each time step, several model outputs, including surface runoff, subsurface flow, and ground-water flow, are available at each MRU. Those model outputs form the components of streamflow for unregulated conditions, when irrigation diversions and returns do not occur (fig. 6). The time series of simulated streamflow components for each MRU are assigned to user-selected stream nodes that represent different locations of interest in the stream network. At each node, the contributions from the assigned MRUs are accumulated and routed downstream to simulate the daily streamflow at the various stream locations represented by the nodes. Simulated streamflows are compared to estimated streamflows and the model is calibrated by adjusting the model parameters until the fit between the simulated and estimated values is reasonable. For this study, measurements of daily streamflows were not available. Therefore, the model was calibrated by comparing simulated and estimated monthly mean streamflows.
After calibration, the precipitation-runoff model can be used to forecast daily unregulated streamflows based on near-real-time known or simulated initial hydrologic conditions in the basin and an assumption of future climate conditions. Streamflows can be forecast using the Extended Streamflow Prediction (ESP) technique in MMS, which is based on a modified version of the National Weather Service ESP program (Day, 1985; Leavesley and others, 1996). In the precipitation-runoff model for this study, ESP can be used to forecast the probability of streamflows for as far as 1 year in the future by assuming that historical records of daily precipitation and air temperatures will recur with the same probability as in the past. For example, if near-real-time initial conditions have been simulated for March 31 of a particular year and streamflows need to be predicted for the following April, the model will be run multiple times in the ESP mode for March 31 through April 30. Each run will have a different set of historical climate data for April but the same initial conditions for March 31. The resulting simulated daily streamflows for April can be analyzed further to determine probabilistic forecasts of variables of interest, such as the probability distribution of peak flow and total flow volume. An alternative use of ESP includes constraining forecasts by using only historical climate data that represent specific conditions, such as El Niño conditions. Use of the calibrated model is facilitated by a user-friendly Object User Interface (OUI) similar to that described for the Yakima River Basin precipitation-runoff model by Mastin and Vaccaro (2002b).
The precipitation-runoff model requires input time series of measured or estimated daily precipitation and daily minimum and maximum air temperatures. In addition, time series of measured or estimated streamflow are needed to calibrate the model.
Historical records of daily precipitation and minimum and maximum air temperatures were obtained from different sources, including the National Weather Service (Hydrosphere Data Products, 2005), Natural Resources Conservation Service (U.S. Department of Agriculture, 2006), and Reclamation (Bureau of Reclamation, 2006a and 2006c). The climate stations used in this study are listed in table 2 and shown in figure 1.
Real-time inputs of precipitation and air temperature are required by the precipitation-runoff model to simulate real-time initial conditions for use in the ESP mode. Of the stations used in this study, only the Conconully CCR Hydromet, Omak OMAW AgriMet, and Salmon Meadows SNOTEL stations provide real-time data. The first water year for which complete sets of daily air temperature and precipitation data are available for two of the real-time stations, Omak OMAW AgriMet and Salmon Meadows SNOTEL, is 1990 (table 2). The first water year for which complete sets of daily air temperature and precipitation data are available for all three of the real-time stations is 2000. For the Conconully CCR Hydromet station, real-time daily precipitation and minimum and maximum air temperatures for water years 2000–04 were used as input to the model. For the Omak OMAW AgriMet station, real-time daily precipitation for water years 1990–96 and minimum and maximum air temperatures for water years 1990–2004 were used as input to the model. For the Salmon Meadows SNOTEL station, real-time daily precipitation data for water years 1990–96 were used as input to the model. The daily minimum and maximum air temperatures were not used in the model because those data were questionable for April 21, 1999, through August 16, 2005 (S. Strachan, Natural Resources Conservation Service, written commun., 2006). The February 1989 through September 1998 data were used to compute minimum and maximum air temperature lapse rates between the Omak OMAW AgriMet and Salmon Meadows SNOTEL stations.
Model simulations of historical conditions used input time series based on climate data from non-real-time stations for water years 1949–89, from non-real-time and real-time stations for water years 1990–99, and from real-time stations for water years 2000–04. For the Conconully station, daily precipitation and daily minimum and maximum air temperatures for water years 1949–99 were used as input to the model. The daily precipitation record was missing 8 percent of the data for water years 1949–96 and 23 percent of the data for water years 1997–99. The daily minimum and maximum air-temperature records were missing 8 percent of the data for water years 1949–96 and 21 percent of the data for water years 1997–99. Water years 1997 through at least 2003 were missing data for most of the winter months (generally for November through February). The missing daily precipitation and minimum and maximum air temperatures were estimated for water years 1949–99 on the basis of all possible combinations of single and multiple linear regressions (with intercept zero for regressions of daily precipitation) between the Conconully station and either one or both of the other long-term stations, the Omak 4 N and the Winthrop 1 WSW stations (fig. 1). The regression that had the highest coefficient of determination (r2) for which all required independent variables were available was used to estimate the missing data.
Synthetically generated temperatures for the Omak OMAW AgriMet station were used as inputs to simulate historical conditions for water years 1949–89. Synthesized temperatures for the Omak OMAW AgriMet station were used in addition to measured temperatures for the Conconully station to assure continuous daily minimum and maximum air-temperature time series at one location for water years 1949–2004. The synthetic daily minimum and maximum air temperatures for the Omak OMAW Agrimet station were generated on the basis of all possible combinations of single and multiple linear regressions between that station and one to three long-term climate stations, the Conconully, Omak 4 N, and Winthrop 1 WSW stations, in the study area (fig. 1). Separate sets of regressions were performed for each month to account for seasonal shifts in temperature relations. For each month, the regression that had the highest coefficient of determination for which all required independent variables were available was used to estimate the air temperatures for the Omak OMAW AgriMet station.
Daily precipitation and minimum and maximum air temperatures for the real-time Conconully CCR Hydromet station were used as inputs to simulate historical conditions for water years 2000–04. For the purposes of this study, it was assumed that the daily precipitation and air temperatures measured for the Conconully CCR Hydromet station were equivalent to the daily precipitation and air temperatures measured for the Conconully station. This assumption was based on a comparison of monthly mean minimum and maximum air temperatures and monthly total precipitation when measurements were available for both stations, August 1999 through September 2003. During this period, however, measurements were missing for the Conconully station from November through February. In 2003, the missing record extended through April. Linear regressions and plots of the data demonstrated that the monthly mean air temperatures and total precipitation compared well between the stations. The coefficient of determination was 0.99 for monthly mean minimum temperatures, 0.99 for monthly mean maximum temperatures, and 0.84 for monthly total precipitation for a linear regression with intercept zero. Because future measurements for the Conconully CCR Hydromet station will be used for simulating near-real-time initial hydrologic conditions for ESP forecasting, the precipitation-runoff model could be more reliable if the assumption of equivalency of air temperatures and precipitation between Conconully CCR Hydromet and Conconully climate stations could be confirmed using future, year-round data for a multiyear period for both stations.
Measured historical streamflows for Salmon Creek are limited to records for two USGS gages that were discontinued in the early 1900s, records for a Washington State Department of Ecology (DOE) gage that has been in operation since 2002, and records for a continuously recording gage operated by the OID since 2003 (T. Sullivan, Okanogan Irrigation District, oral commun., 2006). Prior to becoming a continuously recording gage in 2003, the OID gage was a staff gage that was read daily. The discontinued USGS gages are located just downstream of Conconully Reservoir (fig. 2; USGS station no. 12446500) and 6 mi upstream of the mouth of Salmon Creek (USGS station no. 12447000). The periods of record for the gages are water years 1912 through 1922 and 1904 through part of water year 1910, respectively. The DOE gage (DOE station no. 49M100) is located in North Fork Salmon Creek, about 2 mi upstream of the North Fork Diversion (fig. 2). The first complete April–July streamflow record was measured in water year 2004 and data for water year 2005 is currently (2006) provisional (Washington State Department of Ecology, 2006a; J. Shedd, Washington State Department of Ecology, written commun., 2006). The OID gage is located just upstream of the OID diversion in lower Salmon Creek (fig. 2). During periods of high snowmelt runoff in the spring, the OID gage is submerged and does not accurately measure streamflow for Salmon Creek. The OID also has measured diversions from Salmon Creek at the OID diversion since 1996 (Dames and Moore, 1999).
To expand the limited historical data set, Dames and Moore (1999) estimated a monthly time series of unregulated streamflow for Salmon Creek at Conconully Dam for 1904 through 1998 based on reservoir outflows and storage changes. The U.S. Department of Energy (2004) extended the estimates through 2002, and in this study, the estimates were extended through March 2006 (appendix 1) using storage and outflow data provided by the OID (T. Sullivan, Okanogan Irrigation District, written commun., 2006). The estimates represent unregulated streamflows for the upper Salmon Creek Basin (defined as the drainage area upstream of Conconully Dam), which encompasses 78 percent of the total drainage area of the basin.
The method used to estimate unregulated streamflow from reservoir outflows and storage changes has several potential sources of error. The first source of error is that the estimate does not include a correction for evaporative losses from Conconully Reservoir and the larger surface area of Salmon Lake Reservoir compared to the natural lake that was present prior to the construction of Salmon Lake Dam. Dames and Moore (1999) estimated that those evaporative losses are about 2.2 ft3/s or about 1,600 acre-ft/yr. A second source of error is measurement error in the estimation of reservoir storage. Reservoir storage is estimated by measuring the stage in a reservoir and then converting the stage to a storage volume by using a stage-storage relation. Even small errors in measuring the stage can lead to significant errors in estimating the storage volume and, thus, to significant errors in estimating the monthly storage changes. Such errors may help explain why some estimates of monthly runoff are negative (Dames and Moore, 1999). For example, Dames and Moore (1999) suggested that stage-reading errors that resulted from wave run-up caused by wind may amount to storage-volume estimate errors of several hundred acre-ft. During a period of months, however, the errors in the estimates of monthly storage change are expected to cancel each other.
A third source of error in the estimation of unregulated streamflow is measurement error in reservoir outflow. Dames and Moore (1999) reported that prior to 1997, reservoir outflow was measured periodically by a weir located in Salmon Creek a few hundred feet downstream of Conconully Dam. This measurement included water that seeped below the dam, scheduled releases from the reservoir, and uncontrolled spills during periods of high runoff. Measurements of uncontrolled spills, however, are considered approximate (Dames and Moore, 1999; U.S. Department of Energy, 2004). Starting in 1997, the weir no longer was used and reservoir releases were measured only through the outlet tunnel. As a result, the measurements no longer include seepage below the dam [estimated to be about 1.6 ft3/s or 100 acre-ft per month by Dames and Moore (1999)] and uncontrolled spills.
Dames and Moore (1999) indicated significant, unexplained discrepancies existed between the measured outflow from Conconully Reservoir and streamflow measured at the OID diversion. They attributed the discrepancies to probable errors in the measured outflows from Conconully Reservoir. Starting in 1997, the reliability of the estimates of runoff from upper Salmon Creek Basin becomes less certain than in previous years. Therefore, only data through water year 1996 were used in this study to calibrate and test the precipitation-runoff model.
A fourth source of error in the estimation of unregulated streamflow is that Conconully Reservoir overlies relatively permeable sediments through which ground-water recharge may occur. Some of the recharged ground water may leave Salmon Creek Basin through subsurface flow because the surface-water drainage boundary to the southeast of Conconully Reservoir in the Graveyard Flat area (fig. 2) is unlikely to be a ground-water divide during current (2006) hydrologic conditions. Therefore, some ground water is expected to flow southeasterly towards Scotch Creek. This interpretation is based on water-level altitude data for surface-water features adjacent to the Graveyard Flat area and wells in the Graveyard Flat area. The water-level altitude data may indicate the raised water level of Conconully Reservoir above the natural creek bed may have created a hydraulic connection between the reservoir, the water-table aquifer below Graveyard Flat, and Scotch Creek. Too few data currently are available to determine whether a similar hydraulic connection may have existed prior to the creation of Conconully Reservoir. However, even if a hydraulic connection did exist, ground-water losses would have been significantly smaller during unregulated conditions than during current conditions.
In summary, the estimated streamflows for the upper Salmon Creek Basin, including estimates prior to 1997, have several potential sources of error. However, the estimates are the best available and, therefore, were used in this study as a surrogate for the missing long-term unregulated streamflows for Salmon Creek at the current location of Conconully Dam. The time series of unregulated streamflows is referred to as uncorrected unregulated streamflow (UUS) in this report.
A second time series of unregulated streamflow for Salmon Creek at Conconully Dam was generated by adding estimates of ground-water losses and evaporative losses to the uncorrected time series. The corrected time series is expressed as
CUS = UUS + GW + RES_EVAP, (1)
where
| CUS | is corrected unregulated streamflow, in cubic feet per second; |
| UUS | is uncorrected unregulated streamflow, in cubic feet per second; |
| GW | is the ground-water flux, in cubic feet per second, from Salmon Creek Basin; and |
| RES_EVAP | is reservoir evaporation, in cubic feet per second. |
As described in the sections “Corrections for Ground-Water Losses” and “Corrections for Evaporative Losses,” the ground‑water flux from the basin (GW) is estimated to be about 2 ft3/s and reservoir evaporation (RES_EVAP) is estimated to be about 1 ft3/s. The total of the corrections is about 9 percent of the long-term mean UUS, which is 32.2 ft3/s (23,300 acre-ft/yr) for water years 1949–96. On an annual basis, the sum of the estimated ground-water flux and reservoir evaporation also represents about 9 percent of the total maximum active storage capacity of 23,500 acre-ft for Conconully and Salmon Lake Reservoirs.
Conconully Reservoir possibly recharges the underlying ground-water system and part of the recharged water may leave Salmon Creek Basin through subsurface flow in the general direction of Scotch Creek (fig. 2). The surficial geology between Conconully Reservoir and the headwaters of Scotch Creek is mapped as Continental Glacial Drift deposits (Stoffel, 1990). Driller’s logs obtained from the Washington State Department of Ecology (2006b) for two wells in the Graveyard Flat area show a relatively thick sequence of sedimentary deposits that is likely underlain by a former bedrock valley (field-verification of well locations was beyond the scope of this study). The saturated sediments, which start at about 115 to 130 ft below land surface, consist of sand, gravels, and some clay. Based on the lithologic and water-level information from these logs, it is estimated that the Graveyard Flat area is underlain by at least 70 ft of saturated sediments. Assuming horizontal flow, Darcy’s law states that ground‑water flux can be estimated as follows:
Q = K A i, (2)
where
| Q | is ground-water discharge, in cubic feet per day; |
| K | is horizontal hydraulic conductivity of the sediments, in feet per day; |
| A | is cross-sectional area perpendicular to the direction of flow, in square feet; and |
| i | is hydraulic gradient (dimensionless). |
The hydraulic conductivity (K) of the saturated glacial-drift materials is estimated to be 40 ft/d, which is a conservatively low estimate based on hydraulic conductivities calculated for permeable glacial sediments elsewhere in Washington State (for example, Kahle, 1998; and Kahle and others, 2003). [In a summary of hydraulic-conductivity estimates for glacial sediments from multiple sources, Fetter (1994) reported that glacial-outwash deposits can have hydraulic conductivities that range from about 3 ft/d to about 3,000 ft/d depending on the degree of sorting of the sediments.] Assuming the cross-sectional area through which the southeasterly-flowing ground water discharges is 2,000 ft wide and 70 ft thick, the cross-sectional area perpendicular to the direction of flow (A) is 140,000 ft2. A mean hydraulic gradient (i) of 0.03 was computed by assuming a water level of 2,282 ft for Conconully Reservoir and by assuming the water table intersects the land surface at 2,110 ft in the headwaters of Scotch Creek, where the creek first becomes a perennial stream according to the 1:24,000-scale topographic map for the area. Using these estimates in equation 2, about 2 ft3/s (about 1,400 acre-ft/yr) of ground water discharges from upper Salmon Creek Basin in a southeasterly direction. Some of this ground water likely discharges to Scotch Creek while the remainder is likely to travel along longer and deeper flowpaths and become incorporated in the regional ground-water flow system.
Currently available data are insufficient to determine whether ground-water losses occurred prior to the creation of Conconully Reservoir or, if losses did occur, to determine the rate of the losses. If ground-water losses did occur, however, they would have been significantly smaller than the losses estimated for current conditions. Therefore, for this study, out‑of-basin ground-water losses were assumed to be negligible during unregulated conditions.
The construction of Conconully and Salmon Lake Reservoirs introduced evaporative losses from the reservoir surfaces as an additional source of water loss from the basin. For the purpose of this study, the increase in evaporative losses from Salmon Lake Reservoir was ignored because Salmon Lake Reservoir is located in a narrow canyon and the conversion of Salmon Lake to Salmon Lake Reservoir was assumed to result in a relatively small increase in water-surface area. The construction of Conconully Reservoir, however, created a relatively shallow water body that has a maximum size of about 400 acres where previously only a stream and natural vegetation were present.
Farnsworth and others (1982) estimated, on the basis of regional data collected from 1956 through 1970, an annual free-water-surface evaporation of about 30 in. at the location of Conconully Reservoir. They also estimated that about 23 in., or 77 percent of the annual total, evaporates from May through October. Monthly estimates of reference evapotranspiration for alfalfa at the Omak OMAW AgriMet station (Bureau of Reclamation, 2006a) for water years 1990–2004 range from 0.6 percent of the mean annual total in December to 19.9 percent in July (fig. 7). From May through October, the reference evapotranspiration for alfalfa is 36.5 in., or 81.5 percent of the mean annual total of 44.8 in.
Assuming that the reservoir is at a maximum size of 400 acres during May through July, 200 acres during April and August, and 100 acres during the remainder of the year, and that the annual evaporation is 30 in. distributed according to the monthly reference evapotranspiration at the Omak OMAW AgriMet station, the annual evaporative loss from Conconully Reservoir is estimated to be about 1 ft3/s (about 700 acre-ft) (table 3). Different assumptions for the values of annual free-water-surface evaporation and monthly reservoir surface area could alter the estimate significantly.
U.S.
Department of the Interior | U.S. Geological
Survey
Persistent URL: https://pubs.water.usgs.gov/sir20065274
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 19:27:42 EST
