Duration
Analysis
The Duration Analysis function computes the duration curve for a regular interval
time series data set and stores the results in a new paired data set. Flood
duration curves are useful in assessing the general low flow characteristics of a
stream. For example, if the lower end drops rapidly, the stream has low
ground-water storage and low or no sustained flow. The overall slope of the curve
is an indication of the flow variability in the stream. Refer to EM 1110-2-1415
Chapter 2 (USACE, 1993) for a description of duration analysis. Duration
analysis is usually applied to daily flow or elevation or similar data. This
function is not Volume Duration Frequency Analysis; see the program HEC-SSP for this
capability.
HEC-DSSVue provides two techniques for compute duration curves. The first method develops the curve by ranking all the data and then extracting points along that curve. The second technique segregates values into “bins” and then plots the cumulative amount. This technique was used by the HEC-STATS program and was developed to accommodate the small amount of memory that computers used to have, or when the analysis was done by hand. Because memory limitations have been removed, the first technique is more accurate and is recommended.
The data to be analyzed must have a regular time interval of “1DAY”, “1WEEK”, “TRI-MONTH”, “SEMI-MONTH”, “1MON” or “1YEAR”. Typically the time series data set for the duration analysis spans multiple years. The time series data values are first sorted by time of the year into periods of “Annual”, “Quarterly”, “Monthly” or “Other Defined”. For an “Annual” period type, all data are assigned to a continuous single duration period. For a “Quarterly” duration period type, the data values are sorted by quarter year: 1 January to 31 March, 1 April to 30 June, 1 July to 30 September, and 01 October to 31 December. For the “Monthly” duration period type, data values are sorted by month. The “Other Defined” duration period type allows arbitrary duration periods to be defined and used.
For the standard duration analysis technique, the data values within a duration period are ranked (ordered by descending value). The ordered values form the y-values for a new paired data set. The x-values represent the percent of time exceeded computed by:
E = 100 * [ M /(n+1) ] (Weibull plotting positions), percent of the time the value is equaled or exceeded
M = the rank position of the value
n = number of values.
A curve is computed for each duration period. Thus for a duration period type “Quarterly”, the paired data set has four curves with the curve labels “Jan-Mar”, “Apr-Jun”, “Jul-Sep” and “Oct-Dec”.
The set of x-values generated by the duration analysis are dependent upon the number of values in the time series. Typically all values should be used for plotting the curve. However, this may become unwieldy when presented in a table form, so values may be a standard set of percentages, for example: 1%, 2%, 5%, 10 %, etc. If desired, the Duration Analysis function can sample the analysis results to a standard set of frequency points, or to a set of evenly distributed x-values. Using log spacing provides more clarity of the curve at its ends. User defined points can be specified by entering those points in a table after selecting the User Defined button.
Duration Analysis Screen for Standard Method
To compute the duration analysis for a regular interval time series:




