Laserfiche WebLink
<br />TECHNIQUES OF WATER-RESOURCES INVESTIGATIONS <br /> <br /> <br />8 <br /> 20,000 <br /> 10,000 <br />0 <br />Z <br />0 <br />u <br />w <br />OIl <br />'" <br />w <br />'" <br />.... <br />w 1000 <br />w <br />~ <br />U <br />'" <br />::l <br />U <br />Z <br />0 <br />0 <br />0 100 <br />~ <br />~ <br />'" <br /><( 016 <br />w <br />,. <br />I <br />0 <br /> <br />20X <br /> <br />20x <br /> <br />0" <br /> <br />\70 <br /> <br />EXPLANATION <br /> <br />40 <br /> <br />Parameter i$ mean annual <br />runoff, in inches <br />)( Based on 9 or more years <br />of record <br />. Based on less than 9 <br />yeon of record <br /> <br />.,' <br /> <br />10 · <br />1 <br /> <br />. <br /> <br />\6x <br /> <br />:;: <br />~ <br />~ <br /> <br />010 <br /> <br />o <br />~ <br />w <br />> <br />.. <br />:> <br />u <br />~ <br />o <br />::: 2 <br /> <br /> <br />s <br /> <br />x <br /> <br />~ <br />Z <br />o <br />~ <br /> <br />Q <br />~ <br />:: 0.5 <br />9 <br />~ <br /> <br />o <br /> <br />, <br /> <br />~ <br />o <br />z 0.2 <br />o S <br />~ <br />.. <br />> <br />w <br />Q <br /> <br />10 20 <br />MEAN ANNUAL <br />RUNOF F <br />IN INCHES <br /> <br />10 100 <br />DRAINAGE AREA, IN SQUARE MILES <br /> <br />1000 <br /> <br />. <br /> <br />4. Graphical regression of lO-year floods defined from short records, Kootenai River bosin, Montana. <br /> <br />than about 0.8. However, under certain condi- <br />tions improvement can be obtained at smaller <br />values of the correlation coefficient (Matalas <br />and Jacobs, 1964). <br />The improvement of a frequency curve by <br />correlation with a longer record is obviously <br />desirable if one is concerned only with charac- <br />teristics at that site. But use of such a modified <br />frequency curve in a regional analysis may <br />not improve the result over that obtained by <br />use of the unmodified curve, the reason being <br />that the flood experience at the two sites may <br />not be independent. Because the regionaliza- <br />tion process attempts to average that part of <br />the variability due to random occurrences of <br />weather, a substantial extension of a fre- <br />quency curve on the basis of another one al- <br />ready included in the analysis will tend to <br />duplicate the experience at the site of the <br />longer record. This duplication may bias the <br />result of the regional analysis. As an extreme <br /> <br />example, consider a region containing 20 rec- <br />ords of 10 years and one record of 50 years, <br />and assume that each of the short records is <br />adequately correlated with the concurrent <br />part of the long record to justify extension to <br />50 years. If the 20 records were all for the <br />same 10-year period, the upper part of the <br />regionalized curve would be nearly identical <br />to the individual curve for the 50-year period <br />of record (given the same basin characteris- <br />tics). Now suppose that given the same data, <br />a regionalization based on unextended fre- <br />quency curves was made. This would still lean <br />heavily on the long record for the higher re- <br />currence intervals and might not be too much <br />different at 20 years and below because of <br />the high correlation among all records. So <br />which result is the better? There is little basis <br />for a decision, but because the difference in <br />results likely would be small, we probably <br />would select the method usiug the unextended <br /> <br />. <br />