WSP00029 - Laserfiche WebLink

Home Browse Search

::) ,":) I,;) Iv W W 202 probabilities might provide useful information regarding streamflow patterns (Yevjevich, 1972a, b). Model modification could include the generation of streamf10ws using the fractional noise model, developed specifically with the idea of preserving streamflow pattern characteris- tics (Mande1brot and Wallis, 1968). However, fractional noise models are still in a developmental stage (Jackson, 1975; p. 57). Analyses of runs probabilities and application of any streamflow model suffer from the limited quantity of data available. The data base of 40 years used to calibrate the streamflow model can not provide information for modeling runoff patterns which might be expected to occur over the 300 year lifetimes of the reservoirs in the Colorado River system. In fact, the entire 80 years of runoff data from the Lees Ferry gauge has recently been shown to exhibit only a small portion of the variance which has been observed in long-term, tree-ring correlated runoff data (Stockton, 1975; Jacoby, 1975). Stockton reported that tree-ring reconstructions of annual runoff for the last 300 years reveal that the mean, variance, and serial correlation of Colorado River flows are slowly varying, non-cyclic functions of time. In the absence of additional information on hydrologic processes in the Colorado River Basin, the uncertainty associated with the fre- quency of reservoir failure must be accepted as a property of the system. Extremely long simulations of five to ten thousand years might produce stable extremes of the discharge probability distributions. However, the probabilities so obtained would have no practical value. As men- tioned in the introductory remarks to Section 6.2, extended simulation