9625 - Laserfiche WebLink

Home Browse Search

20 Probable Effects of the Proposed Sulphur Gulch Reservoir on Colorado River Quantity and Quality near Grand Junction, Colorado Table 6. Summary of statistical parameters for streamflow, salinity, and evaporation residuals. [ft, feet; mg/L, milligrams per liter; ft, feet; The respective terms shape and mean refer pazameters used in the Weibull and Logistic probability distribution func- tions] Variable residual Probability distribution Minimum Maximum Shape, or mean Scale streamflow at Plateau Creek, Logistic X84.7 511.6 13.4 83.6 cubic feet per second Evaporation, ft Logistic -36.322 35.280 -0.52100 6.007000 Colorado River salinity at Cameo, Logistic -0.2079 0.2079 -0.00380 0.034594 mg/L Plateau Creek salinity near Cameo, Logistic -0.27 0.28 0.00 0.05 mg/L Colorado River salinity at Palisade, Logistic -155 393 6 23 mg/L Runoff salinity to Sulphur Gulch, Weibull -7149 1267 15 0.034594 mg/L mixing of both natural and anthropogenic system concentra- tions just upstream from the point of interest. For example, the primary contributions to background historical (ambient) salin- ity over the study reach include the Colorado River basin upstream from Cameo, runoff from Sulphur Gulch, Plateau Creek, and the Orchard Mesa Irrigation District check structure. Changes to the background concentrations are directly related to reservoir releases during peak and(or) low-flow periods. Res- ervoir related considerations that may affect the dissolved-sol- ids concentration of water released to the Colorado River are associated with when water is pumped into the reservoir, runoff to the reservoir, and reservoir evaporation. A general expres- sion that describes the mixing of various water types is given by MODEL VALIDATION Validation of the stochastic mixing model involved three primary steps: (1) test the reliability (stability and convergence) of a representative Monte Carlo forecast, (2) compare statistics for selected forecast simulations to field measurements, and (3) evaluate the overall mass balance. In general, model valida- tion is a subset of scenario modeling because the user is required to define one (or more) forecast(s) and enter appropri- ate decision variables before starting the simulation process. The following sections describe the bootstrap approach to test reliability, forecast comparisons, and availability of pumpable water. M ~J Cl -t C~ = M ~Qi~ -~ (1) where Q is the streamflow discharge (negative values indicate losing and positive values indicate gaining), Cis the dissolved-solids concentration, M is the total number of mixed components, j is an index representing a location along the study reach, and i is an index representing each type of water. Because streamflow is required when computing concen- trations with nonlinear regression equations, a direct link to the hydrology model passes streamflow values to points where computations are being conducted. A flow chart describing the stochastic water-quality model operations is provided (fig. 16), and a descrip- tion of the Excel cell-based equations is included as Appendix 2. Stabilifiy and Convergence To test the reliability (stability and convergence) of mixing model forecasts, the so-called bootstrap approach is used (Wer- ckman and others, 2001). In using the bootstrap approach, sam- ple statistics (estimated mean, standard error, and confidence intervals) are computed from 200 independent (repeated) fore- casts of annual streamflow at the Colorado River gage near Cameo for a fixed number of Monte Carlo trials and constant decision variables (hydrologic constants include reservoir pump, 0 ft3/s, Grand Valley Irrigation Canal senior water right, 520 ft3/s; Grand Valley Irrigation Canal junior water right, 120 ft3/s; minimum flow at the Colorado River gage near Pali- sade, 85 ft3/s, and maximum return flow at the Orchard Mesa Irrigation District check structure, 400 ft3/s; whereas the water- quality constants include: initial storage = 0, initial concentra- tion = 0, no reservoir release, maximum reservoir storage 16,000 acre-ft, and a no flag = 0 for reservoir releases). The bootstrap approach is repeated for an increasing number of Monte Carlo trials (500, 1,000, 1,500, 2,000) until the percent change in upper and lower confidence intervals is less than 1 percent.