Laserfiche WebLink
<br />I\) <br />,... <br />O':l <br />0-) <br /> <br />drawback 'is that a fixed system operation with a given set of hydrology <br />will alw~ysgive one number to a quantity which is actually random in <br />nature. ;With a single historic trace, the question (for example) of <br />what size of reservoir would assure a specified supply has a single <br />answer. 'In reality, a more severe drought than the historic extreme <br />can occu~. Thus, the required storage to assure a given flow for a <br />specific 1period is a random variable, Unfortunately, use of a single <br />historic sequence yields only one estimate and does not lead to <br />probabil!stic inferences about the quantity being studied. This simple <br />example !llustrates the most important attribute of stochastic simula- <br />tion, the ability to arrive at probabilistic inferences concerning <br />important variables. Thus, to transcend the limited span of historic <br />data, syqthetic hydrology can often be used. Synthetic hydrology or <br />data gen~ration is nothing more than a method of manufacturing numbers <br />which ba~e properties similar to the historic data. Usually used in <br />conjunction with a computer program which has access to a virtually <br />infinite;number of random numbers, the technique of data generation <br />permits the user to create any number of hydrologic sequences. These <br />sequenceS are constructed to maintain statistical properties that the <br />historic ,data showed. This insures that the synthetic traces will be <br />acceptable for studying the behavior of a basin opera~ion just as the <br />historic idata set may be acceptable, <br /> <br />When a synthetic trace is used to estimate a property of the basin's <br />behaviori one measurement is produced. However, each new trace pro- <br />duces an~ther estimate.. When a sufficient number of traces are used, <br />the reSulting batch of estimates can be ranked and statistically <br />analyzed: the data generation method can then produce enough values <br />to estab~ish probability distributions of quantities that would have <br />had only 'one value if the historic data were used alone. <br /> <br />Before g~neration of synthetic hydrologic traces can proceed, the <br />properti~s and estimates of coefficients used to define them must be <br />obtained ,from historic data. These data could be either raw measure- <br />ments or:data adjusted to reflect known historic events which have <br />affected ,it. A data analysis program is used to extract all the <br />properti~s needed to determine the structure of a time series and <br />to make dstimates of all necessary coefficients. These numbers along <br />with a description of their relative position in the basin are used <br />to gener~te the synthetic streamflows. These synthetic flows are <br />then used in the basin operation model to produce many estimates of <br />how the ~ater-controlling operation performs. It should be noted <br />that many assumptions are involved in the basic structuring of the <br />river basin and in the choice of quantities to be generated. All <br />of these assumptions greatly influence the quality of the results. <br /> <br />3 <br /> <br />