Return to Task 1. Create a New HEC-SSP Project and Import Data.

Background Information

The Bulletin 17C guidance brings a change to the computation of peak flow frequency within the United States.  This guidance incorporates changes motivated by four of the items listed as future work within Bulletin 17B and more than 30 years of post-Bulletin 17B research on flood processes and statistical methods (England, et al., 2015).  As part of the Bulletin 17C methodology, the moments/parameters of the Log Pearson Type III distribution are estimated using the Expected Moments Algorithm (EMA).  Like Bulletin 17B, the Bulletin 17C/EMA (17C EMA) methodology also estimates distribution parameters based on sample moments, but does so in a more integrated manner that incorporates non-standard, censored, or historical data at once, rather than as a series of adjustment procedures (Cohn, Lane, & Baier, 1997).  The use of Bulletin 17C procedures also provides improved confidence intervals for the resulting frequency curve that incorporates skew uncertainty and diverse information appropriately, as historical data and censored values impact the uncertainty in the estimated frequency curve (Cohn, Lane, & Stedinger, 2001).  Within the 17C EMA methodology, every annual peak flow in the analysis period, whether observed or not, is represented by a flow range.  That range might simply be limited to the gaged value when one exists.  However it could also reflect an uncertain flow estimate.

Flow ranges can also be used to account for the events that are missing within the historical record.  For instance, we do not know what the peak flow rate at the Big Sandy River at Bruceton, TN gaging station was outside of the systematic record (except for the three noted historic events in 1897, 1919, and 1926).  However, we do know that some event occurred (i.e. the Big Sandy River still existed).  We can also infer from physical constraints and past experience within the contributing watershed that had an event over a certain “threshold” occurred, it would have been recorded and documented.  This places an upper and lower “perception threshold” and corresponding flow range on the possible events that occurred during those missing years.

Bulletin 17C also allows a more effective description to account for missing data.  When using Bulletin 17B procedures, the broken period of record data set is simply “lumped” together into a period that leaves out the missing years.  With Bulletin 17C procedures, the missing years are explicitly accounted for with perception thresholds and flow ranges, allowing more information to be used.

Create a New Bulletin 17 Analysis

  • To begin, select Analysis | New | Bulletin 17 Analysis.
  • Name the new analysis “BigSandyRiver_Systematic_B17C” and add an adequate description.
  • Select the “BigSandyRiver” data set.
  • Select “17C EMA” within the Method for Computing Statistics and Confidence Limits panel.
  • Utilize the default selections of Use Station Skew, Multiple Grubbs-Beck low outlier test, and Hirsch/Stedinger plotting position.

Input EMA Data

  • Move to the EMA Data tab.

The EMA procedure introduces the new concept of Perception Thresholds.  Perception thresholds define the range of streamflow for which a flood event could have been observed.  The inherent assumption and consequence is that any year for which an event was not observed and recorded must have had a peak flow rate outside of (usually below) the perception range.

The first row within the Perception Thresholds table will automatically be created to span the entire period of record of the selected Flow Data Set.  The start and end year of this first perception threshold can be modified to alter the analysis time frame.  This first perception threshold must have a low value of 0 and a high value of infinity.  Additional rows within the perception threshold table supersede the rows above, for the specified time frame.

It is recommended that additional perception threshold time frames not overlap one another within HEC-SSP.

For any missing years in the analysis period, perception thresholds other than zero to infinity must be entered after the first row.  The reason for this requirement is that a perception threshold of zero to infinity presumes any flow that occurred could have been observed, implying that unobserved years would not be possible.  Therefore, unobserved years must have a perception threshold with either a lower bound greater than zero or an upper bound less than infinity.  Most commonly, since very large flows do tend to be observed in some way (as historic events are estimated based on some evidence in the watershed), a lower bound greater than zero is chosen.

  • Add information to the second row of the Perception Thresholds table for the missing period between 1988 and 2002. Enter “1988”, “2002”, and “18000”, for the Start Year, End Year, and Low Threshold in this second row. The High Threshold should have a default value of inf (infinity). If it does not, double left-click in the cell, right-click and select Set as INF.  (I.e., do not just type "inf" in the cell.)


Question: What does the use of this perception threshold imply about the missing years (i.e., 1988 - 2002)?

The use of this perception threshold assumes that, had a peak flow rate occurred in excess of 18000 cfs from 1988 - 2002, it would have been observed and recorded.  Thus, all annual maximum flows within this time period must have been less than or equal to 1800 cfs.


  • Use the Comments column to provide an adequate descriptive note.
  • The Perception Thresholds table should resemble the following figure.

BigSandyRiver_Systematic_B17C Perception Thresholds Table

  • Once all of the information has been entered to the Perception Thresholds table, click the Apply Thresholds button. This will fill out the Flow Ranges table with the complementary information.  For instance, since a perception threshold from 1988 – 2002 of [18000 – inf] was entered, the corresponding flow range is [0 – 18000].  This implies that the annual peak flows for these years must have been within this range, as no observation was made of some larger flow.
  • Ensure that the Flow Ranges table contains a low and high value for every single year in the analysis period. The completed EMA Data tab should resemble the following figure.

BigSandyRiver_Systematic_B17C EMA Data Tab

Compute and Analyze Results

  • Click Compute.
  • Click Plot Curve. This will result in the computed curve, 5- and 95-percent confidence limits, and observed events being plotted.
  • Close the computed curve window.
  • Move to the Tabular Results tab.  Note the computed curve, 5-, and 95-percent confidence limits for all of the desired frequency ordinates, the moments/parameters of the Log Pearson Type III fit to the data, and other data related to the analysis.


Question: What is the Equivalent Record Length (ERL) of this analysis?  What does this imply about the information content of the missing period from 1988 - 2002?

The ERL for this analysis (when using a perception threshold of 18000 cfs for the missing period) is approximately 73 years.  This implies that the missing period from 1988 - 2002 added approximately 8 years' worth of information content to this analysis (i.e., ~73 year ERL - 65 exactly known observations = ~8 years added by the missing period).


Question: How would a perception threshold with a larger magnitude for the missing period affect the results of this analysis?  How would a perception threshold with a smaller magnitude for the missing period affect the results?

Using a perception threshold larger than 18000 cfs would result in complimentary flow ranges larger than [0 - 18000].  The mean, standard deviation, and skew magnitudes are all larger when a perception threshold larger than 18000 cfs is used.  When using a perception threshold smaller than 18000 cfs, the mean, standard deviation, and skew magnitudes are all smaller than when using an 18000 cfs perception threshold.  When a perception threshold larger than 18000 cfs is used for the missing time period, a smaller ERL would be realized. This translates to an increase in uncertainty in these results and wider confidence limits.  The opposite effect on ERL would be realized if a perception threshold with a smaller magnitude were used. This translates to a reduction in uncertainty in these results and narrower confidence limits.


Question: What results are realized when using a perception threshold of [inf - inf] for the missing period?

A perception threshold of [inf - inf] (and complimentary flow ranges of [0 - inf]) is used to imply no knowledge of flood magnitudes during missing periods.  Using a perception threshold of [inf - inf] results in wider confidence limits, indicating more uncertainty in the results.  In addition, the mean, standard deviation, and skew magnitudes are all larger than if a perception threshold other than inf was used.  This results in an ERL of exactly 65 years.  However, this is typically too conservative for most analyses.  It is likely that some information could be found to infer a threshold that was not exceeded during that time period from interviews with residents, newspapers, and/or other publications.


Continue to Task 3. Bulletin 17C | Systematic Record + Historical Information.