Guidelines for Identifying and Evaluating Peak Discharge Errors




In Reply Refer To:                                    July 2, 1992
Mail Stop 415



OFFICE OF SURFACE WATER TECHNICAL MEMORANDUM NO. 92.10

SUBJECT:  Guidelines for Identifying and Evaluating Peak
          Discharge Errors

Recent examinations of U.S. Geological Survey (USGS) files, 
especially peak-flow files, by District personnel have identified 
significant errors.  Some of these errors have resulted from 
entering incorrect station numbers, entering information under the 
wrong station number, failing to correct for digital recorder 
errors, and data entry (keypunch) errors.  In addition, questions 
have been raised about the accuracy of peak-flow determinations in 
steep streams with highly movable beds.  There is concern that 
some peaks that have been computed as water flows were actually 
debris flows.  Also, questions are being asked about the accuracy 
of assigned roughness values in channels with coarse bed material.  
Some investigators suggest that actual Mannings "n" values are 
higher than those typically assigned to such channels.  This 
memorandum has been assembled to call attention to the importance 
of our peak-flow data base and to describe procedures to ensure 
the integrity of that data base.

The USGS data bases are used for a wide variety of purposes and by 
many public and private entities.  It is critical that data we 
provide to the nation be as accurate as current technology allows.  
It is imperative that keypunch or other forms of mechanically 
based errors be identified in an expeditious manner and correct 
values entered into the data base.  It is also important that 
peak-discharge events that may actually have been debris flows be 
critically examined.  

The burden of maintaining the accuracy of our data files lies with 
the Districts.  The California District has been involved in 
searching for anomalies in their peak-flow files and correcting 
erroneous values.  We encourage all Districts to review and 
revise, if necessary, their data bases using similar procedures.  
The following suggested procedures are based, in large part, on 
the California District's experience.

1.  Transfer data from the WATSTORE peak-flow file to a local 
relational data base.  Use of a relational data base will simplify 
the retrieval and display of the data.  All data elements in the 
peak-flow file should be transferred, including county and 
hydrologic-unit codes, latitude, longitude, dates, discharges, 
gage heights, historical information, and qualification codes, for 
both annual and secondary peaks above the threshold.  

2.  Retain a copy of all data as retrieved from WATSTORE for 
future use in quality assurance of the review and updating 
processes (step 13).

3.  Proofread the data against the published record.  This step is 
important because there is concern that an unknown fraction of the 
data have been corrupted by clerical errors such as entering data 
under the wrong station number.  Most of this work can be done by 
clerical staff, concurrently with technical review of the data by 
the hydrologic staff.  The clerical work can be eased by using the 
data base to print out old data by station number and date for 
comparison with the mid-60's flood compilations (USGS Water-Supply 
Papers 1671-1689) and to print out the newer data by year and 
station for comparison with the annual data reports.  Check all 
data elements, including station numbers, dates, gage heights, and 
qualification codes.  Keep a permanent copy of the data listing 
showing a check mark for each item checked.

4.  If any discrepancies are found, keep a permanent record of the 
correction and the basis for the correction.  Prepare necessary 
input and images for updating the WATSTORE peak-flow file, but do 
not change either the local data base or the WATSTORE file until 
the review is complete.

5.  Concurrently, with the clerical review, conduct a technical 
review of the data with emphasis on identifying and critically 
examining apparent outliers.  Use data-base graphical facilities 
to plot discharge versus drainage area by State, county, or 
hydrologic unit code.  Draw smooth envelope curves that enclose 
about 95 percent of the data, and identify apparent outliers (low 
as well as high).

6.  Where appropriate, compare recorded peaks with values computed 
by the Rational Formula or by appropriate regional flood-frequency 
equations, or with data from the daily-values and measurement 
files.  Identify apparent outliers.

7.  When apparent outliers are found, remember that just because a 
peak is anomalous does not mean that it is incorrect; this is just 
a screening criterion.  When an apparent outlier is found:

a.  Compare the peak with the published values and with original 
data; look for transposed numbers and values repeated from 
previous years, and ensure that the station ID is correct.


b.  Determine the basis of the peak.

(1)  If it was an estimate, check that it was coded as such.

(2)  If it was an indirect measurement, review the measurement; 
give special scrutiny to measurements rated poor or based on one 
or two cross sections.

(3)  If the peak was from a rating extension, look for other 
evidence to verify the extension and the peak.

(4)  Use data-base graphical facilities to plot discharge against 
gage height to check for datum changes and possible misapplication 
of shifts.

(5)  Make a permanent record of the review, including the 
conclusion reached, the basis for the conclusion, and any 
necessary corrections or revisions.

(6)  Prepare any necessary input cards for updating the WATSTORE 
peak-flow file, but do not change either the local data base or 
the WATSTORE file.

8.  Determine if the site is susceptible to debris flows and, 
therefore, if the peak could have been caused by a debris flow 
rather than by a clear-water flood.  Review photographs taken 
during indirect discharge measurements for evidence of debris 
flows.

9.  Determine if the site is susceptible to scour and fill.  
Review photographs taken during indirect discharge measurements 
for evidence of channel scour.

10.  Review the location and drainage area for the site in 
question.  This has been found to be especially important for 
sites smaller than 5 square miles and for sites at which drainage 
areas were determined on 15-minute maps.  This has required some 
field verification.

11.  Review historic-peak (and "highest-since") information.  
Review the definition of items "historic" and "systematic" on 
pages C-44 and 45 of the WATSTORE manual, vol. 4, Ch. I, and 
ensure that the historic-peak qualification code (see pages A-18 
and A-19 of WATSTORE manual, vol. 4, Ch. I) is used properly.  
Determine the specific technical basis and exact source of all 
historical information and make a permanent record of it; if the 
specific basis and exact source cannot be determined, acknowledge 
that fact in the record and provide whatever information is 
available.  If any specific information is available for a 
historic peak (e.g., gage height or "highest since" year), ensure 
that it is stored in the peak-flow file, with a historic-peak 
qualification code, even if the discharge cannot be determined.  
Ensure that values subject to high uncertainty (greater than 25 
percent) are coded as estimates.  Discharges subject to very high 
uncertainty (greater than 50 percent) either should not be stored 
in the file or should be stored as lower-limit estimates with a 
"greater-than" qualification code; both of these cases are treated 
the same by the J407 flood-frequency program.

12.  If any revisions are necessary, make a permanent record of 
the revision and the basis for it.  Prepare input cards for 
updating the WATSTORE peak-flow file.  After all reviews have been 
completed and all corrections and revisions have been determined, 
run a job to update the WATSTORE data base.  Do not update the 
local data base as the data as originally retrieved will serve as 
documentation for revisions to WATSTORE.

13.  Retrieve the data from the (updated) WATSTORE peak-flow file 
and compare it with the data previously retrieved (Step 2).  The 
PRIMOS CMPF command, or the UNIX DIF command, can be used to print 
only the differences between the old and new data sets; review the 
differences to ensure that all updates were made correctly and 
that no inadvertent changes were made.  Keep a permanent record of 
this review for quality assurance.

14.  Keep a permanent machine-readable copy of the updated 
WATSTORE peak-flow data.  This file can be used as the basis for 
future comparisons with the WATSTORE peak-flow file, for quality 
assurance of the annual peak-flow records-production process.

The procedures noted above are aimed at examining extreme high 
(and low) peaks.  These procedures should not be taken to imply 
that other values, such as those within the envelope curves of 
plots of discharge versus drainage area (step 5), are error free.  
Such errors could be found using procedures outlined above, but it 
becomes more difficult to decide which peaks to choose for such an 
analysis.

As noted, simple and obvious errors should be corrected without 
delay.  Such problems require little hydrologic judgment to 
ascertain that corrections are needed.  However, anomalies that 
result from possible mischaracterization of debris flows as water 
flows, or from possible errors in assigned roughness values 
require careful scrutiny and sound judgment before data values are 
revised.  Experts usually carefully reviewed high discharges and 
most indirect measurements at the time the information was 
obtained.  Our understanding of some processes has, however, 
advanced since the time of the collected information.  The 
reviewer should be careful to make technical revisions only where 
there are technical reasons to do so.  Revisions, especially those 
based on roughness characteristics, should not be based simply on 
"second guessing" by different personnel.

Anomalous indirect measurements or anomalous peaks derived from 
ratings based on indirect measurements should be critically 
examined by hydrologists with debris flow knowledge and with 
knowledge of movable bed dynamics.  It is not possible to provide 
specific guidelines and criteria on which to base judgments as to 
the character of any specific flow or conclusions that any 
indirect measurement is erroneous because of improperly assigned 
"n" values.  Some general guidelines regarding recommended 
indirect measurement technique include obtaining a minimum of 
three cross sections, limiting conveyance ratios to between 0.7 
and 1.4, avoiding sections that result in flow moving in and out 
of critical flow, and other general criteria as given in 
"Measurement of Peak Discharge by the Slope-Area Method," 
Techniques of Water-Resources Investigations Book 3, Chapter A2, 
(Dalrymple and Benson, 1968) and in the "Users Manual for WSPRO - 
A Computer Model for Water Surface Profile Computations," Report 
No. FHWA-IP-89-027, 187p (Shearman, 1989).

Violation of any one of the guidelines and criteria given in the 
references mentioned above does not necessarily invalidate any 
indirect measurement.  The decision to accept or reject a 
measurement requires the hydrologic judgment of qualified persons 
in each District.  Measurements that are questionable and 
difficult to assess should be reviewed by specialists outside the 
District, such as the Regional Surface-Water Specialists or others 
known to be authorities in surface-water hydraulics or debris 
flows.

It is recognized that a thorough review of the entire flood-peak 
data file requires substantial investments of time, money, and 
efforts of clerical, computer, and hydrologic personnel.  However, 
the flood data that we collect and disseminate serves important 
national safety-related purposes.  There are growing national 
concerns for safety and for quality-assurance of basic data and 
scientific analyses of safety-related issues.  The effort of 
reviewing our flood data base and making necessary corrections and 
revisions, therefore, will be well worthwhile to the USGS and to 
our cooperators.  Such a review might be included in plans for 
statewide or regional flood-frequency studies; reviews for 
subsequent studies would be much simplified by preservation of 
machine-readable files of the updated data base, as suggested in 
step 14.  It would be desirable to have the review completed 
before the transfer of the peak-flow file from WATSTORE to NWIS-
II.




                                 Charles W. Boning
                                 Chief, Office of Surface Water

WRD DISTRIBUTION:  A, B, FO, PO