Laserfiche WebLink
<br />C) <br /> <br />184 <br /> <br />) <br /> <br />t\J <br />I'" <br />'""'" <br />Ul <br /> <br />In the context of hydrologic simulation, the variability in run- <br /> <br />off sequences is generally high. Simulation length increases with the <br /> <br />size of the reservoirs, or number of possible reservoir states, as shown <br /> <br />by Vicens <br /> <br />and Schaake, (1972). Reservoir size in this sense is <br /> <br />taken relative to average inflow, simulation time increasing as average <br /> <br />inflow decreases. For simulations of the type used in this study the <br /> <br />stochastic inputs are autocorre1ated and the state of the system <br /> <br />(reservoir storage) is dependent upon the state of the system in <br /> <br />previous time periods, both factors contributing to long simulation <br /> <br />times. Finally, convergence of the extremes of distributions requires <br /> <br />longer simulations than for convergence of the first and second moments. <br /> <br />Each of these factors increases the total simulation times required <br /> <br />for stabilization of the response of hydrologic systems. <br /> <br />Other factors limit the practical length or number of simulation <br /> <br />runs. Computer costs may be prohibitive in some cases. A more impor- <br /> <br />tant issue with respect to design decisions has to do with the quality <br /> <br />of information obtained from long simulations. Yevjevich (1972a; p. <br /> <br />144) cautions that generation of long sequences of hydrologic data <br /> <br />does not serve to increase the amount of information contained in the <br /> <br />original data sample. Aside from presentations of statistical techniques <br /> <br />for establishing confidence intervals of parameter estimates, the <br /> <br />writer has found no discussion in the literature on hydrologic simu1a- <br /> <br />tion relating the information content of the original data to the <br /> <br />quality of simulation results. <br />