9625 - Laserfiche WebLink

Home Browse Search

Probable Effects of the Proposed Sulphur Gulch Reservoir on Colorado River Quantity and Quality near Grand Junction, Colorado reach between the proposed reservoir-release point 25 mi east of Grand Junction (Palisade) and De Beque. The assumption of instantaneous routing is reasonable because the daily time step used in the model is less than the time it takes a parcel of water to travel through the study reach (Don Carlson, written commun., Northern Colorado Water Conservancy District, 2002). 5. 6. Variability in runoff and salinity observed at the USGS streamflow-gaging station 09095300 Dry Fork at Upper Station near De Beque, Colorado, is similar to Sulphur Gulch. The assumption that Dry Fork runoff and associated concentrations can be used as a surrogate for runoff and concentration to the Sulphur Gulch Reservoir may be reasonable because the Dry Fork watershed is located adjacent to Sulphur Gulch watershed and has similar geology and similar precipitation. Limited sampling was conducted to verify this assumption. The reservoir is completely mixed. It is important to note that the assumption of complete mixing neglects reservoir stratification. Because seasonal stratification of the reservoir may cause increased salinity with depth, the actual nature of how and when reservoir releases are managed (from top, bottom, or mixture) may cause variability in downstream salinity that is not considered in this study. Dissolved solids (salinity) are conservative. The assumption that dissolved solids are conservative implies that all of the sources or sinks of water within the study reach are represented. Because dissolved solids and streamflow are highly correlated (as determined by nonlinear regression), the transport of dissolved solids is assumed to be advective with no dispersion. 8. 9. No ground-water seepage (baseflow) to Sulphur Gulch Reservoir. No baseflow assumes that ground-water seepage carrying dissolved solids, as evidenced by evaporative concentration along the canyon walls at the Sulphur Gulch Reservoir site, will be controlled by maintaining a reservoir level that exceeds the hydraulic head governing ground-water seepage. No evaporative concentration residue exists on reservoir canyon walls during the initial filling of the reservoir. Whereas evaporative salts are observed on the canyon walls of Sulphur Gulch at the proposed reservoir site during spring and summer, these salts are periodically flushed following thunderstorms. For this reason, the likelihood for anomalously high initial salinity concentrations caused by dissolution of residue on the canyon walls is considered negligible. Parameterization Parameterization of the stochastic mixing model involves defining random variables, decision variables, and prediction variables (forecasts). Random variables do not have a fixed value at a particular point in space and time and are described in the mixing model by probability distributions that account for a range of possible values. The various daily random variables in the stochastic model include streamflow, diversions, return flows, and salinity in the Colorado River at De Beque, Cameo, and Palisade and in Plateau Creek, a tributary to the Colorado River downstream from Sulphur Gulch but near Cameo; runoff and salinity concentration in the Sulphur Gulch watershed; res- ervoir evaporation; reservoir surface area; and reservoir releases from Sulphur Gulch. Whereas the random variability in Colorado River streamflow and salinity at De Beque, Cameo, and Palisade reflect natural variability, the parameters used in describing streamflow and salinity at Sulphur Gulch are uncer- tain. The use of random variables as input to the stochastic mix- ing model requires identification of relevant probability distri- butions and related statistical summaries. Statistical summaries of random variables are derived from records that incorporate a balanced mix of wet, dry, and average hydrologic periods to avoid model bias. The actual length of record is based on the need to minimize model time step yet provide enough measurements so that a statistically valid probability distribution can be fit for each parameter. One test used to evaluate the goodness-of-fit between measurement frequency and fitted probability distribution is the Chi-square goodness-of-fit criteria (Helsel and Hirsch, 1995). The Chi-square criteria evaluates the goodness-of-fit by breaking the distribution into areas of equal probability and comparing the data points within each area to the number of expected data points. In addition to describing variability and uncertainty using probability distributions, the correlation in time between random variables is incorporated into the mixing model by using a single lag function. In addition to fitting random vari- ables to probability distributions, decision variables also were defined. Decision variables are those variables that can be con- trolled in the stochastic mixing model. Examples of some important user-defined decision variables incorporated into the stochastic mixing model include initial reservoir storage vol- ume, initial reservoir salinity concentration, total reservoir volume, reservoir pumping rate from the Colorado River, total amount of supplemental flow, and amount and timing of daily reservoir releases. Whereas many of these decision variables have fixed (constant) values, the infinite number of possible combinations of magnitude and timing of reservoir release is likely to affect the change in prediction variables, such as the Colorado River salinity (change or amount). Prediction vari- ables represent distributional outcomes of model calculations based on some combination of decision and random variables. For this reason, prediction variables are summarized as either probability distribution or cumulative distribution functions. model are as follows: Assumptions invoked in parameterization of the stochastic 1. Fitted probability distributions accurately represent vari- ability and (or) uncertainty associated with model