Laserfiche WebLink
<br />2-04. <br /> <br />that occur within given rBDgesof IIlagI11tude. Then the ~ower ~imit <br />of magD.1tude in 8a.ch nmge is p~otted 'lgl'-1n"t the percentage of <br />dqs of record that mean-<1aily nOll'll exceed that magnitude. A <br />typical 1'1ow-duration curve is shown on exhibit 9. - <br /> <br />2-05. <br /> <br />TBmNICAL APPROACH m ESTDlA.fiNG FBEQ,UEIlCY l,;U1tVS; <br /> <br />a. There are two basic approaches to estiJDating frequency <br />curves - graphical and analytical. Each of these approaches has <br />severlu variations in current practice,but the discussion herein <br />will be limited to selected methods. <br /> <br />b. Graphica.lly, frequencies are evaluated sim;ply by arrang- <br />ing obserred. Talues in the order of magnitude and considering <br />that a smooth curve suggested by that arr~ of values is represen- <br />tative of future possibilities. Each value represents a fraction <br />of the future possiMlities and, when plotting the frequency curve, <br />it is gl.ven a ":plotting position" that 1s calculated to give it <br />the proper weight (see paragraph 3-05). <br /> <br />~. In the application of analytical (statistical) procedures, <br />the concept of theoretical populations or distributions is employed, <br />as discussed in paragraph 2-03. A distribution is a set of values <br />that would occur under fixed conditions in an infinite amount of <br />tiJDe. Tbose that have occurred are presumed to constitute a random <br />sam;ple and accordingly are used to make particular inferences . <br />regarding their "parent populat:lon" (Le. the distribution from <br />which they were dert ved) . Such inferences are necessarily attended <br />by considerable uncertainty, because a given set of observaticns <br />could result fl'Olll any of III8lIY sets of physical conditions (fl'Olll <br />any one 01lllLlly s ons. owever, by the use of statis- <br />tical processes, the .!DOst probable nature of the distribution from <br />which the data were derived can be estiJDated. Since in all. proba- <br />bility this is not the true parent population, the relative chance <br />that ~ations :from this "_Tlmnm likelihood" distribution might <br />be true must be evaluated. Each range of possible parent popula- <br />tion is then weighted in proportion to its likelihood to obtain a <br />weighted average as demonstrated in reference 4. A probability <br />obtained. from this weighted average is herein referred to as the <br />expected probability, PH' CODq>Utation procedures are given in <br />section 4. <br /> <br />g,. Because of the shortness of hydrologic records, fre- <br />quency determinations are relatively unreliable where based on a <br /> <br />- 7 - <br />