FLOOD07843

• County: Statewide
• Basin: Statewide
• Title: Nationwide Summary of US Geological Survey Regional Regression Equations for Estimating Magnitude and Frequency of Floods or Ungaged Ssites 1993
• Date: 1/1/1993
• Prepared By: USGS
• Floodplain - Doc Type: Educational/Technical/Reference Information

Hydrologic Flood Regions

'de flood-frequency reports, the
I most statewI I .
n divided their States into separate hydro. O~IC
analysts , f h ogeneous flood charactensttcs
. Regions 0 om hed
regions. d termined by using major wat~rs
were ge?erally e al sis of the areal distnbutton of
danes and an an y
boun , 'd I (differences between regres-
sSlOn resl ua s
the reg~e (on (observed) T-year estimates), In some
sion an sta I h drolo ic regions were also defined by
instances, the y, : the watershed or by statistical
the mean elevat~~c~xon signed-rank test. Regression
tests s~ch as ~:edefined for 210 hydrologic regions
equatt~n~tthe Nation, indicating that, on average,
throug 0 C . ns per State. Some areas of the
e about .our reglO fi od
the~ ar h ever have inadequate data to defin~ 0 _
Natton, ow " For example, there are regIOns of
frequencx ~e:~~uency in Rorida, Ge01g~a, Texas,
undefine OF th State of Hawaii, regressIOn equa-
d Nevada. or e R
an . ded for the Island of Oahu, egres-
tions are only provI, , flood- ak discharges for
. n equations for estlmatmg pe "d
SIO . were computed as part of a nattonwl e
the other Islands 1972) but are not
ork analysis (Yamanaga, .
netw 'NFF' that study was not specdically
included m smce .
oriented to flood-frequency analysIs.

Measures of Accuracy

Every USGS regional flood report prov~des some
f accuracy of the regression equations. The
measure 0 f Y is the stan-
most frequently used measure 0 accu~ rcent This
dard error of estimate, usually report~ ~n pe . th
standard error is a measure of the vanatlOn between e

regression estimates and !be Station data for those sta-
tions used in deriving the regression equations. More
recently, analysts have begun reporting the standard
error of prediction, which is a measure of the accuracy
of the regression equations when predicting values for
watersheds not used in the analysis. The standard error
of prediction is generally slighdy larger than the stan-
dard error of estimate. The standard error reported in
the individual statewide report is the standard error
given in NFF because that is the only estimate of error
that was available. Often, the standard errors of esti-
mate or prediction are converted to equivalent years of
record. The equivalent years of record are defined as
the number of yean; of actual streamflow record needed
to achieve the same accuracy as the regional regression
equations.

The standard errors of estimate or prediction
generally range from 30-60 percent, with 2 I States and
the Commonwealth of Puerto Rico having standard
errors in this range. There are 14 States where there is
at least one hydrologic region Within the State with a
standard error less than 30 percent. The remaining 15
States have at least one hydrologic region where the
standard error is in excess of 60 percent. The largest
standard errors generally are for equations developed
for the western portion of the Nation where the at-site
variability of the llood records is greater, where the net-
work of unregulated gaging stations is less dense and
there are more diffiCUlties in regionalizing llood char-
acteristics, and the flood records are generally shorter.
The smallest standard errors are generally for equations
developed for the eastern portion of the Nation where
the converse of the above conditions are generally true.

RURAl R 000 --;l'BlCr ES'TWATlMGTECHNIQUE$ 7