This paper presents the results of a nationwide riverbed erosion risk assessment study, using a potential scour depth model, developed by Williams et al (1992). The nationwide application of the scour depth model, using data captured at different scales, provides a indicator of scour potential rather than an estimate of absolute scour depths.
Geographic Information Systems (GIS) technology is used to perform the analysis. Potential scour depth is computed using the flow with an 1% annual chance of being exceeded (also known as the 100-year flood), sediment size, and a factor describing stream shape characteristics. The scour depth is converted to a continuous surface raster file with a one-square kilometer resolution. This is overlaid with another raster file that contains streams and three classes of annual flooding probability. The results show riverbed erosion hazard defined as potential scour depth within streams and flood plains, associated with annual flooding probability. By overlaying the erosion hazard with the pipelines, pipeline segments that present a high erosion risk are identified.
Introduction
A pipeline failure in 1994 in the San Jacinto River
Valley, near Houston, Texas, was caused by riverbed
erosion. Four pipelines broke and oil and gas were
spilled, causing pollution and fire. As a result
the Research and Special Programs Administration,
Office of Pipeline Safety, USDOT, initiated a study
on the effects of natural hazards on pipelines. As
part of this study, FEMA and its contractor,
Michael Baker Jr., Inc., provide research services,
data, and facilities. Risk assessment of natural
hazards will enable USDOT and FEMA to direct their
resources to those parts of the national pipeline
infrastructure where they will have the greatest
impact. This paper presents the results of a
nationwide riverbed erosion risk assessment study,
using a potential scour depth model, developed by
Williams et al (1992).
Scour
Scour occurs in streams particularly during high
flow periods. Particles are detached from the
riverbed and transported downstream. The amount of
particles detached, resulting in a certain scour
depth, depends on the velocity and volume of the
flow, and on the grain size of the particles. The
processes of scouring and filling, by detaching,
transporting and depositing sediment, can alter the
riverbed (McKnight, 1992). A shape characteristic
of a stream, defined in this study, is used as an
indicator of how much scour (and filling) occurs in
that particular stream.
Data sources
The following data are used for assessing riverbed
erosion hazard:
Streams, and the mean flow associated with those
streams, are extracted from the United States
Environmental Protection Agency's (USEPA) Reach
File 1 (RF1). The original scale of the data is
1:500,000. This file consists of 64,902 reaches,
of which 60,126 have the mean flow as an
attribute. The more detailed version, Reach File
3 (RF3), with more reaches and more attributes
associated with the reaches, is scheduled for
release in the fall of 1996.
Particle sizes and the annual flooding
probability are derived from the States Soil
Geographic (STATSGO) Data Base, published by the
U.S. Department of Agriculture's (USDA) Natural
Resources Conservation Service (NRCS), formerly
known as the Soil Conservation Service (SCS).
The original scale of the data is 1: 250,000.
The particle sizes of riverbed sediments differ
from the particle sizes of the surrounding
areas. However, the soil grain sizes are used
as a relative indicator.
The annual chance of flooding is expressed as
Rare (0-5%), Occasional (5-50%), and Frequent
(50-100%). The flood areas from the STATSGO
database at the one square kilometer resolution
reasonably match the FEMA Q3 flood maps (FEMA,
1995, 1996) for a test area around the San
Jacinto River Valley pipeline failure site.
Since the Q3 flood maps are not yet available
for the entire country, the flood areas from the
STATSGO database are used in approximating the
area of floodplains.
The 100-year peak flow, the mean flow and the
drainage area are extracted from U.S. Geological
Survey (USGS) Streamflow and Basin
Characteristics (SBC) point file (Dempster,
1983). This file contains gauging station data.
Some data points were not used because either
the mean flow was very small, that is smaller
than 0.00001 ft3/s, or the ratio of the 100-year
peak flow and the mean flow was unlikely high,
that is, higher than 15,000. This left 9,364
data points.
The scour depth estimation model
The following equations (1), (2), and (3) are
extracted from Williams et al. (1992). The
equations are used for estimating potential scour
depth:
ds = Z * dm (1)
Where:
ds = potential scour depth (feet)
Z = stream characteristics factor
dm = mean water depth (feet)
The Z factor describes the shape characteristics of
the stream. Simply, the straighter the reach, the
lower the factor, and the sharper the bends, the
higher the factor. This factor is derived from the
ratio of the actual length of the stream to the
Euclidean distance between the nodes.
Mean water depth can be estimated by:
dm = 0.47 * (Q/f) 1/3 (2)
Where:
Q = discharge (CFS)
f = Lacey's silt factor
Lacey's silt factor is defined as:
f = 1.76 * (Dm) �(3)
Where:
Dm = mean bed sediment size (mm)
Methodology
ArcInfo, version 7.03, is the software used for
this application. Most of the analyses are
performed in its GRID module, on a one square
kilometer resolution. The map projection used for
this study is the Albers Conic Equal-Area
projection.
The 100-year peak flow is not available in RF1 and
is computed in several steps, using the SBC gauging
stations data. First, the specific discharge values
(ft3/s/mi2) are calculated for all the gauging
stations by dividing the flow (ft3/s) by the
drainage area (mi2). Specific discharge is
calculated for both the mean and the 100-year peak
flow. Stations with a mean flow of zero are
deleted. Second, two Triangulated Irregular
Networks (TIN's) are created from the data points;
one of the specific mean annual discharge and one
of the specific 100-year peak discharge. Third,
from these TIN's, raster files are derived with a
one square kilometer resolution. Fourth, a ratio is
computed for each grid cell of the specific 100-year peak discharge and the specific mean annual
discharge. Fifth, data points with a ratio of
15,000 or higher are deleted, since it is unlikely
that the specific 100-year peak discharge is 15,000
(or more) times higher than the specific mean
annual discharge. Sixth, the mean flow associated
with (almost) each reach, provided by the RF1 file,
is multiplied by the ratio derived from the SBC,
resulting in an approximated 100-year peak flow.
For reaches that are not associated with a mean
annual flow value, this value is interpolated. The
resulting 100-year peak flow is entered in equation
(2).
Lacey's silt factor (f) is derived from the STATSGO
soils database, as described in equation (3). The
soils description is cross referenced with a
particle size table to provide a particle size
polygon coverage. This polygon coverage is
converted to a grid with a one square kilometer
resolution.
The variables are substituted into the scour
equations. For the areas that have no data for
particle size, no scour depth is calculated. The
results from equation (1), potential scour depth,
are added to the midpoint of each reach. These
points are first converted to a TIN, then converted
to a raster file to provide a continuous surface of
stream scour for the entire country. This stream
scour data was applied only to the areas within
streams and flood plains. Areas outside the flood
plains and streams are assigned NODATA.
Results of the scour model
The results of equation (1) show scour depth
ranging from zero to 88 feet, with a mean of 3.5
feet. Every part of a stream that has a potential
scour depth of six feet or more is considered most
hazardous, since pipelines are buried about five
feet deep at river crossings where scour hazard is
assumed. The results show that scour hazard is
lower upstream and higher downstream, where
velocity and volume of the flow are higher.
Erosion hazard
The erosion hazard layer combines the scour data
and the flood areas. Both the scour data and the
flood areas are normalized from zero to 100 and
then summed with equal weight:
Hazard = 0.5 * scour + 0.5 * flood(4)
Thus, areas with the highest rank in scour depth
(six feet or greater) and the highest rank in
annual chance of flooding (50-100%), which includes
streams, have a erosion hazard of 100. Areas
outside a stream and without flooding probability
have a erosion hazard of zero.
Figure 1 shows the erosion hazard in the area
around the San Jacinto River Valley pipeline
failure site. The erosion hazard value at the
actual pipeline failure site is 97 and is
surrounded by values 95-100.
Conclusions
Estimating scour depth, using the methodology
described in this paper, can be used for
prioritizing areas for mitigation purposes. Once
priority areas are selected, finer studies with
more detailed data can be performed on regional
scale. This nationwide study, combining data from
different sources and with different original
scales, came up with quite reasonable results. More
detailed input data, such as EPA's Reach File 3
(RF3) and FEMA's Q3 flood data will most likely
give more detailed results. Also, the more data
fields are populated, the less values need to be
interpolated.
Acknowledgments
The authors wish to thank David T. Williams, of
WEST Consultants, Inc., for providing the particle
size reference table.
References
Dempster, G.R., Jr., "Streamflow/Basin
Characteristics Retrieval (Program E796)", U.S.
Geological Survey WATSTORE User's Guide, 1983.
Federal Emergency Management Agency, "Q3 Flood Data
Specifications", 1996.
Federal Emergency Management Agency, "Q3 Flood Data
Users Guide", 1995.
McKnight, Tom L., "Essentials of Physical
Geography", Prentice-Hall, Inc., 1992.
Williams, David T., Samuel Carreon, Jr., and
Jeffrey B. Bradley, "Evaluation of Erosion
Potential at Pipeline Crossings", Hydraulic
Engineering 1992, pp. 689-694.