Prior to the development of data sets described in this report, the most detailed digital basin maps available on a statewide basis were the nationally-standardized cataloging units that are designated with 8-digit hydrologic-unit codes derived from 1:250,000-scale maps (Steeves and Nebert, 1994). The cataloging units were developed as part of a nationally uniform hierarchical system organized by the U.S. Water Resources Council in the mid-1970's. The system divides the country into regions, subregions, accounting units, and cataloging units. A hierarchical code consisting of two digits for each level is used to identify units. Eight-digit cataloging units average 450,000 acres in size (Seaber et al., 1987). The U.S. Department of Agriculture, Natural Resources Conservation Service (NRCS), formerly known as the Soil Conservation Service (SCS), further divided cataloging units into subunits with 11-digit hydrologic-unit codes in Oklahoma. Unlike the hydrologic units of higher levels, many of the NRCS subunits were delineated along project or administrative boundaries rather than hydrologic divides.
The availability of Geographic Information Systems (GIS) and
digital data sets made it possible to automate the
watershed-delineation process using Digital Elevation Models
(DEM's). However, many of the existing DEM's were derived from old,
inaccurate topographic data using outdated techniques of data
conversion. Many DEM's were created using older methods that
introduced systematic errors into the DEM's. Additionally,
1:100,000-scale hypsography (land-surface point elevation and
contours) and hydrography data in Digital Line Graph (DLG) format
became available for the entire state in early 1995. New GIS
algorithms were available, allowing the use of the DLG data to produce
a statewide DEM and watershed map of better quality than previously
available.
Approach
The ANUDEM software package, version 4.4, developed by Hutchinson
(1989) at Australian National University was used to make a statewide
DEM with a horizontal grid-cell resolution of 60 meters. Four types of
input data were used for the production of the DEM: both contour-line
and point hypsography, hydrography, and depressions extracted from the
hypsography data. After processing with ANUDEM, further processing was
done to remove all depressions except a few large depressions.
Watershed outlet points were selected and watersheds were delineated
using an automated process, followed by interactive editing of the
watershed boundaries.
The U.S. Geological Survey (USGS) 1:100,000-scale Digital Line Graph (DLG) files were used for input hypsography data. The DLG files were converted into ARC/INFO coverage format and elevations were associated with the contours and points.
ANUDEM retains depressions specified by the user. ANUDEM requires that a depression be represented by a point within the depression. Small depressions--those less than 3 cells wide at the widest--are likely to be smoothed over by the gridding algorithm, so only depressions larger than this were specified for retention. The elevations associated with these points were set to the elevation of the surrounding depression contours minus half of the contour interval.
The 1:100,000-scale hydrography data were acquired in the form of ARC/INFO data sets. When acquired, these data had been separated into hydrologic cataloging units but were later appended into one data set. These data were an early release of the River-Reach File (RF-3) distributed by the U.S. Environmental Protection Agency (EPA). Cataloging units that included any part of Oklahoma were processed.
Four significant problems in the RF-3 were corrected before use in ANUDEM. 1. Many small waterbodies and streams that were not connected to the main stream network were eliminated. 2. Centerlines were generated for all large lakes, wide streams, and other waterbodies. The polygons forming the waterbodies were removed. 3. ANUDEM requires that all hydrographic lines point downstream, so all lines pointing upstream were flipped. 4. The RF-3 data were incorrect in several places. Large parts of several rivers and lakes were missing and two streams were incorrectly connected at their headwaters. Also, the stream segments in an area covering one 7.5-minute quadrangle were shifted approximately one kilometer to the west. Corrections were made using data extracted from the USGS 1:100,000-scale hydrography DLG's.
The ANUDEM software is based on an algorithm that produces a hydrologically-conditioned DEM by interpolating elevations using hypsography and hydrography data. It uses a method of drainage enforcement to remove erroneous depressions from the DEM. The drainage enforcement algorithm "significantly increase[s] the accuracy, especially in terms of their drainage properties, of digital elevation models" (Hutchinson, 1989). This algorithm removes depressions only when drainage conditions contradict input elevation data by less than a user-specified tolerance. The interpolation method is implemented by fitting a thin-plate spline to the data, conditioned by a surface-specific roughness penalty (Hutchinson, 1989). Four user-specified tolerances are used to control how the data are interpolated. The tolerances were set as suggested in the software documentation, based on the contour interval of the USGS quadrangle maps being processed.
The data for the state of Oklahoma could not be processed at one time because of computer storage limitations and contour interval differences. When available, data from quadrangles adjacent to the state boundary were used. Different tolerances were used for each processing block according to the contour interval. The processing blocks overlapped, in most cases, by 12 kilometers on each side. All input hypsography and hydrography data were appended and trimmed to cover the areas for each processing block. After ANUDEM processing, 6 kilometers were trimmed from the overlapping edges of each processing block, to avoid problems introduced by interpolations near the edges of the input data sets. The elevations in the remaining overlapping areas were averaged together using a distance-weighted method. Using the ARC/INFO GRID function MOSAIC (ESRI, 1994), the processing blocks were combined to create two DEMs: one for the Oklahoma panhandle and the other for the rest of the state.
The combined DEM's resulting from ANUDEM processing contained numerous depressions that had not been removed by the drainage-enforcement algorithm using the specified tolerances. Because the presence of many small depressions would complicate the process of watershed delineation and because most depressions in DEM's are errors resulting from the representation of the surface in raster form (Jenson and Domingue, 1988 and Hutchinson, 1989), the DEM's were processed using the ARC/INFO command FILL in the GRID module, an implementation of the approach outlined by Jenson and Domingue (1988). The FILL command fills depressions to their pour points that are the minimum elevations along the drainage basin boundaries of the depressions. The identification and removal of depressions is an iterative process. Filling a depression may create new depressions along its boundaries that will be filled in the next iteration (ESRI, 1994). In order to retain large depressions such as playa lakes, cells with values of nodata were entered at the centers of depressions larger than 3 cells wide shown by depression contours on the 1:100,000-scale USGS topographic quadrangles. Areas draining into cells with a value of nodata were not filled by this procedure. After the FILL procedure, the original elevations were replaced into the cells that had been set to nodata.
The two filled DEM's were trimmed to have a 12-kilometer overlap. They were combined using the MOSAIC function to create a seamless statewide DEM with floating-point elevations in meters. The direction of steepest descent for each cell (flow direction) was computed from the statewide floating-point DEM. To save disk space, the DEM elevations were rounded to the nearest meter after calculation of flow directions.
A statewide grid of accumulated flow was generated from the flow-direction grid. Accumulated flow is the number of cells flowing into each cell in the output grid. Cells with high accumulated flow values may be used to identify stream channels. Cells with accumulated flow values of zero are local topographic highs and can be used to identify ridges and drainage basin boundaries. To save disk space, the accumulated flow data were reclassified to five categories.
Watershed boundaries were derived from the flow-direction data set using automated procedures (Jenson and Domingue, 1988). The flow-accumulation data set was used in the selection of watershed pour points or outlets. Outlets were selected at stream confluences in cells with high accumulated flow. Some errors were observed when the boundaries derived from the DEM were compared with the 1:100,000-scale contours and streams. The boundaries were revised so that the watershed boundaries would be consistent with the contours and streams from 1:100,000-scale quadrangle maps. Errors were corrected interactively using the ARCEDIT module of ARC/INFO. Watershed delineation was done using the two separately filled DEM's, one for the panhandle, the other for the rest of the state. The two data sets of watershed boundaries were combined after the necessary revisions were made, then hydrologic-unit codes were added. Eleven-digit watershed codes were assigned following guidelines established in USDA National Instruction 170-304 (USDA, 1995). The first 8 digits of the 11-digit codes match the nationally-standardized system of hydrologic cataloging units. The last three digits generally begin at 010 and increase by ten for each watershed downstream within a cataloging unit. The watershed boundaries were trimmed to the state boundary and inserted into 1:250,000-scale 8-digit cataloging-unit boundaries for the rest of the Arkansas, Red, and White River basins.
Several closed basins not draining into the main stream network
resulted from the retention of large depressions. Boundaries of these
closed basins, known as noncontributing drainage areas, are provided
as a separate data set.
Results
The project described in this report resulted in the production of a
seamless hydrologically-conditioned statewide DEM of Oklahoma with a
horizontal grid-cell resolution of 60 meters. The DEM is well suited
for automated watershed delineation. Because the centerlined stream
network was used in the creation of the DEM rather than water-body
polygons, the DEM is not flat in the areas covered by water. In some
cases, contours of the land surface before construction of reservoirs
were included in the DLGs and were used with ANUDEM. Because these
input data were used, DEM elevations in areas covered by water are not
reliable. Four grid data sets and three vector data sets resulted from
this project. The four grid data sets prepared are the hydrologically
conditioned DEM of Oklahoma, the flow-direction data set, the
reclassified flow-accumulation data set, and the shaded-relief data
set derived from the statewide DEM. The three vector data sets
prepared are the watershed boundaries for Oklahoma with hydrologic
cataloging units outside Oklahoma, noncontributing drainage areas, and
the downstream-directed stream centerline network used in the
generation of the DEM. The statewide data set of watershed boundaries
consists of 11-digit subdivisions of the nationally-standardized
8-digit cataloging units. The watershed map has been edited to be
consistent with the contours and streams shown on USGS 1:100,000-scale
quadrangle maps. A shaded-relief data set was created by using the
HILLSHADE function of the GRID module of ARC/INFO Version 7.0.2 (ESRI,
1994) on the statewide floating-point DEM.
Any use of trade names in this publication is for descriptive
purposes only and does not imply endorsement by the U.S. Government.
References
Environmental Systems Research Institute, Inc. (ESRI). (1994) GRID
Command References, ARC/INFO Version 7.0.2 ArcDoc, Redlands,
CA. [On-line documentation]
Hutchinson, M. F. (1989) A new procedure for gridding elevation and stream data with automatic removal of spurious pits: Journal of Hydrology, v.106, p. 211-232.
Jenson, S.K. and J.O. Domingue (1988) Software tools to extract topographic structure from digital elevation data for geographic information system analysis: Photogrammetric Engineering and Remote Sensing, v. 54, no. 11, p. 1593-1600.
Seaber, P.R., F.P. Kapinos and G.L. Knapp (1987) Hydrologic unit maps: U.S. Geological Survey Water-Supply Paper 2294.
Steeves P. and D. Nebert (1994) Hydrologic-unit maps of the conterminous United States: U.S. Geological Survey.
USDA Natural Resources Conservation Service. (1995) Mapping and digitizing watershed and subwatershed hydrologic unit boundaries (Working Draft, June 1995), National Instruction 170-304.
2.
Alan Rea
U.S. Geological Survey
202 NW 66th Street, Building 7
Oklahoma City, Oklahoma 73116
Email: ahrea@usgs.gov
home page