Laserfiche WebLink
<br />Fitted lines based on the assumption that the logarithms of the data <br />are described by the log-Pearson Type III distribution are plotted in <br />figure 4. It is impractical to develop a horizontal axis for this distribu- <br />tion that is distorted such that data would plot as a straight line, as was <br />the case with the Gumbel distribution, because a different distortion would <br />have to be developed for each value of skewness. Instead, data points are <br />connected by a smooth curve. <br /> <br />The prediction of the magnitude of any event with a recurrence interval <br />of 10 years or less is relatively insensitive to the distribution used--the <br />predictions made using the three different distributions agree within <br />10 percent. This insensitivity is due to the fact that the period of record <br />is long compared to the recurrence interval being considered (Frank <br />Richards, National Weather Service, oral cornmun., 1987). In contrast, the <br />predicted magnitude of events with recurrence intervals of 100 years or <br />longer is quite sensitive to the distribution used. For example, the pre- <br />dicted magnitude of the 5-minute, 100-year event is more than 50 percent <br />larger if the log-Gumbel distribution is used to describe the data than if <br />either the arithmetic Gumbel or the log-Pearson Type III distribution is <br />used. <br /> <br />For durations where the August 1, 1985, storm was an outlier <br />(5 minutes, 1 hour, 2 hours, and 24 hours), the predicted magnitudes of <br />events with long recurrence intervals are sensitive to whether or not data <br />from 1985 and 1986 were used in the analysis. This sensitivity is an <br />indication of how much weight the various distributions give to the outlying <br />point that represents the 1985 storm. Because the 1985 storm was not an <br />outlier for the 10- and IS-minute durations, the calculated magnitudes for <br />those durations is almost the same regardless of whether the previously <br />defined or \lpdated distribution is used, even for events with very long <br />recurrence intervals. <br /> <br />INTENSITY-DURATION CURVES <br /> <br />Intensity-duration curves for storms of selected recurrence intervals <br />were developed using regional equations, and the previously defined and <br />updated Gumbel, log-Gumbel, and log-Fearson Type III analyses. The <br />calculated curves were then compared to the observed intensity-duration <br />curve for the storm of August I, 1985. <br /> <br />Method of Analysis <br /> <br />Steel and McGhee (1979, p. 326-327) divide the United States into seven <br />areas and present regionalized equations of intensity-duration curves for <br />selected recurrence intervals. The authors advise that these equations <br />should be used with caution since wide variations in precipitation patterns <br />are to be expected over such large sections of the country. However, the <br />curves can ,serve as useful preliminary estimates. The 100-year-storm <br />intensity-duration curve calculated from the Steel and McGhee equations for <br />area 5 (that includes southeastern Wyoming) is plotted on figure 5. <br /> <br />29 <br />