Abstract:
Flood-frequency relations that are developed
by fitting the logarithms of annual peak discharges
to a Pearson Type-III distribution are sensitive to
skew coefficients. Estimates of population skew
for a site are improved when computed from the
weighted average of (1) the sample (station) skew,
and (2) an unbiased, generalized skew estimate. A
weighting technique based on the number of years
of record at each of 226 sites was used to develop
a contour map of unbiased, generalized skew
coefficients for New York. An attempt was made
to group (regionalize) the station skew
coefficients into five hydrologically similar areas
of New York, but the statewide version proved to
be as accurate as the regionalized version and
therefore was adopted as the final generalized
skew-coefficient map for New York. An error
analysis showed the statewide contour map to
have lower MSE’s (mean square errors) than those
computed from (1) the five regional skew coefficient
contour maps, (2) a previously used
(1982) nationwide skew coefficient map, and (3)
the weighted mean of skew coefficients for sites
within each of five hydrologically uniform, but
distinct areas of New York.
Purpose:
The effective management of flood-prone areas
and the design of structures along rivers and streams
requires knowledge of the magnitude and frequency of
floods. Discharge-frequency relations for streamflow-gaging
stations on rural, unregulated streams in New
York were developed by Lumia (1991), who fitted the
logarithms of the annual peak discharges to a Pearson
Type-III distribution according to guidelines
recommended by the Interagency Advisory
Committee on Water Data (IACWD) (1982), and
updated these frequency curves by including data
collected through 1996. The Pearson Type-III
distribution requires computation of the mean, the
standard deviation, and the skew of the station’s
annual peak discharges.
The skew coefficient of a station’s peak-discharge
record is sensitive to extreme values; therefore,
accurate values are difficult to obtain for sites with
short records. The accuracy of a station’s skew
coefficient can be improved by weighting that station
skew coefficient with a “generalized” skew value that
represents pooled skew-coefficient data from nearby
stations with long records; generalized skew
coefficients can be estimated through regression,
mapping, or averaging methods (IACWD, 1982). A
nationwide map by the IACWD (1982) provides skew coefficient
contours that represent generalized
estimates, but the IAWCD suggests that separate,
regional skew analyses be made to obtain more
accurate values for local flood-frequency analyses.
In 1998, the U.S. Geological Survey (USGS), in
cooperation with New York State Department of
Transportation (NYSDOT), began a study to develop a
generalized skew-coefficient map of New York,
excluding Long Island. Long Island streams were
excluded from this study primarily because of the
varying degrees of development and urbanization
2 Development of a Contour Map Showing Generalized Skew Coefficients of Annual Peak Discharges of Rural, Unregulated
Streams in New York, Excluding Long Island
within the gaged basins. An attempt was made to
regionalize skew coefficients into five hydrologically
uniform, but distinct areas of New York. The resulting
five contour maps provided little or no improvement
over the statewide contour map, however; therefore
the statewide map (excluding Long Island) was
adopted as the final source for determining
generalized skew coefficients. The statewide map is
based on skew coefficients from 226 rural, unregulated
streamflow-gaging stations with at least 20 years of
annual peak discharges; 194 of these sites have at least
25 years of record.
An error analysis showed the statewide map to
have a lower MSE (mean square error, an estimate of
sampling variance) than either (1) the IACWD’s
nationwide map of 1982, (2) the five regional maps, or
(3) the weighted mean skew coefficients for each of
the five regions. The MSE from the statewide
generalized skew-coefficient map can be used in future
flood-frequency computations.