Laserfiche WebLink
<br />j <br /> <br />{ <br /> <br />in which: <br /> <br />k = G/6 <br /> <br />4.3. ANALYSIS OF SPECIAL CASES. In hydrologic engineering <br />applications, frequency analysis of annual maximum discharge is complicated by special <br />cases. These include broken records, incomplete records, zero-flow years, outliers, <br />historical data, and small samples. Bulletin 17B provides guidance for dealing with <br />these cases. <br /> <br />If one or more years of data are missing from a time series of annual <br />maximum discharge due to reasons not related to flood magnitude, the record is broken. <br />For analysis, the record segments are combined, and the combined record is analyzed <br />as previously described. <br /> <br />If data are missing because the events were too large to record, too small to <br />record, or the gage was destroyed by a large event, the record is incomplete. Any <br />missing large events should be estimated, and the estimates included in the time series. <br />Missing small events are treated with the conditional probability adjustment <br />recommended for zero-flow years. <br /> <br />The log-Pearson type III distribution is not suited to analysis of series which <br />include zero-flow years. If the sample contains zero-flow years, the record is analyzed <br />using the conditional probability procedure. With this procedure, the sub-series of non- <br />zero peaks is analyzed as described previously. The resulting frequency curve is a <br />conditional frequency curve. The exceedance frequencies from this curve are scaled <br />by the relative frequency of non-zero flow years. The log-Pearson type III model <br />parameters are estimated for the upper portion of the curve. With these parameters, <br />a synthetic frequency curve is developed. Section 3.6 of EM 1110-2-1415 describes <br />the procedure. <br /> <br />An outlier is an observation that departs significantly from the trend of the <br />remaining data. Procedures for treating outliers require hydrologic and mathematical <br />judgement. Bulletin 17B describes one procedure for identifying high and low outliers <br />and for censoring the data set. High outliers are treated as historical data ff sufficient <br />information is available. Low outliers are treated as zero-flow years. <br /> <br />Large floods outside the systematically-recorded time series may be used to <br />extend that record. The procedure recommended for analysis of these historical flows <br />is as follows: <br /> <br />. Assemble known historic peaks and determine the historic record length. <br /> <br />7-99 <br />