Flow-Based Method for Stream Generation in a GIS

Abstract

The method presented here provides an objective approach to stream network delineation in a GIS and can be applied anywhere in the conterminous United States. The approach can delineate streams based on any statistical measure of streamflow (mean, median, min, max, etc.) for any month and year, or the streamflow metric can be averaged over months (average Spring flow) or multiple years (30-year average flow). Because the method is based on streamflow, it can provide an estimate of where headwaters begin at a certain flow threshold. This can be very useful not only for modeling purposes but also for mapping areas where streams may not be represented on traditional maps.

Introduction

The U.S. Geological Survey (USGS) has monitored our Nation's streams since the agency's inception in 1879. An important component of that monitoring effort is the ability to map where streams are located. Since the late 19th century, the USGS has been producing topographic quadrangle maps (fig. 1). These detailed maps show the shape and elevation of the land, transportation networks, vegetation, and stream networks. Advances in survey techniques, instrumentation, and computing technologies, as well as the use of aerial photography (fig. 2) and satellite data, have dramatically improved mapping coverage, accuracy, and efficiency. Geographic Information Systems (GIS) are altering the production and use of traditional maps and, subsequently, the representation of streams (fig. 3). Typically, streams are digitized from USGS topographic maps and brought into various software packages for display and modeling purposes. This mapping process can introduce map error because streams are usually determined by visual inspection only, which can lead to missed streams (fig. 4). The definition of a stream is left to each individual mapmaker's visual interpretation, which can introduce inconsistencies in the mapped stream network (fig 5.) A clearly defined and consistent methodology of stream delineation would allow streamflow quantity and quality to be studied in a more meaningful context across broad geographic regions.

Question: What is the definition of a stream and what is the best way to represent them in a Geographic Information System (GIS)?

Figure 1. Traditional USGS 1:24,000 topgraphic quad map
Galena quad, Maryland.

Figure 2. Aerial photo of Galena quad, Maryland.

Figure 3. Aerial photo of Galena quad, Maryland
with digitized streams.

Figure 4. Aerial photo of Galena quad, Maryland
showing digitized and missing streams.

Figure 5. Stream density issues of the digital National
Hydrography Dataset of Indiana and surrounding states
with 1:100,000 USGS topo-quad lines in red.

Definitions

Streams are not static, they are dynamic. The extent of a stream network can change from summer to winter (fig. 6), from dry to wet years, or even from storm to storm (fig. 7). Moreover, there is no one standard definition of a stream. There are hydrologic definitions, regulatory definitions and even a GIS definition.

Figure 6. Extent of Spoon Creek during the summer and winter of 1998. Wigington and others, "Stream Network Expansion: A Riparian Water Quality Factor." Hydrological Processes Scientific Briefing, Hydrol. Process. 19, 1715-1721 (2005).

  
Figure 7. These images of before and after the flooding at the junction of the Mississippi and Ohio Rivers were captured by MODIS on April 25 and May 18, 2002. http://www.gsfc.nasa.gov/indepth/photos_earth2002.html

Hydrologic Definition

"A course of running water usually flowing in a particular direction in a defined channel and discharging into some other stream or body of water. Streams may be classified as follows: In relation to time; ephemeral, intermittent or seasonal, perennial." (Lo, 1992).

Question: How can this hydrologic definition of a stream be represented in a GIS?

Regulatory Definition

As part of the 2000 Regulatory Nationwide Permit Program of the US Army Corps of Engineers (USACE) which authorizes discharges into headwaters and isolated waters, the USACE wrote: "District engineers will utilize the best methods available to identify where the average annual flow of a stream is 1 cubic foot per second (cfs) ...This approach recognizes that streams with highly irregular flows, such as those occurring in the western portion of the United States, could be dry at the 1 cfs point for most of the year and still average, on an annual basis, a flow of 1 cfs because of high volume, flash flood type flows which greatly distort the average. ...The definition allows the district engineer to use approximate means to compute it. The drainage area that will contribute an average annual flow of 1 cfs can be estimated by approximating the proportion of average annual precipitation that is expected to find its way into the stream. Knowing the amount of area that will produce this flow in a particular region, the 1 cfs point can be approximated from drainage area maps. For example, in most areas of the eastern United States (i.e., east of the Mississippi River), one square mile of drainage area produces 1 cfs of stream flow annually." (http://www.wetlands.com/coe/nwp/fr09mar00ps3f.html)

Question: How would one determine where the average annual flow is 1 cfs in a GIS?

GIS Definition

Stream networks can be delineated from a Digital Elevation Model (DEM) using the output from the ARCINFO GRID FLOWDIRECTION and FLOWACCUMULATION functions (fig. 8). FLOWDIRECTION uses a DEM to determine the direction of flow from every cell in the raster. Flow accumulation, in its simplest form, is the number of upslope cells that flow into each cell. By applying a threshold value to the results of FLOWACCUMULATION, a stream network can be delineated. For example, selecting only those cells that have 100 or greater upslope cells produces the stream network in Figure 9. This method is arbitrary, because it is based on the GIS user selecting a threshold value. (ESRI Help documentation, 2006)

Question: Is there a way to include the hydrologic and regulatory definition in a GIS?

  
Figure 8. Schematics of a DEM, FLOWDIRECTION and FLOWACCUMULATION GRIDs in a GIS. (ESRI Help documentation, 2006).

Figure 9. Example of a stream network generated from a flow accumulation grid where cell values are 100 or greater.

Approach

Figure 10. The USGS' WaterWatch web page
provides runoff (streamflow per unit area)
estimates, in units of millimeters (mm)
per month and in mm per year, for every 8-digit HUC.

The method outlined here provides a hydrologic basis for the threshold value outlined in the GIS definition above by taking advantage of available USGS streamflow data. Estimated runoff (in units of mm/time) for each hydrologic cataloging unit (HUC) in the conterminous United States exists in tabular format (fig. 10) for every month and water-year between 1900-2002. (http://water.usgs.gov/waterwatch/?m=romap&r=us&w=real%2Cmap).

These runoff data, along with the flow accumulation GRID, can be used to generate stream networks in a GIS based on a 1 cfs flow threshold. The method, applied to the Chesterville Branch watershed in Maryland, consists of five steps described in more detail below.

1. The estimated runoff from HUC 02060002 was mapped in a raster format at a 30-meter resolution for the Chesterville Branch Watershed (fig. 11) and converted into units of meters/time.

2. The flow accumulation GRID, created from a 30-meter DEM, was converted into an accumulated area GRID (fig. 12). This was done by multiplying the flow accumulation GRID (the number of upslope cells) by 900 (the area, in square meters, of one 30-meter cell).

3. Using map algebra, the new "accumulated area" GRID (square meters) was multiplied by the HUC runoff GRID (m / time) (fig. 13). This results in a GRID of accumulated flow in units of volume (cubic meters / time).

4. Using simple calculations, the units of the accumulated flow (cubic meters / time) are converted into units of cubic feet per second (cfs) (fig 14).

5. One cfs is selected as the threshold for initiation of a stream as outlined in the GIS definition above (fig. 15).

Methods

1. Convert HUC runoff data into a GRID

Figure 11. Data for the average annual runoff for 1999 was downloaded from Water-Watch and converted into a GRID. Every cell within an 8-digit HUC has the same runoff value as reported from WaterWatch for that year and for that HUC.

2. Convert Flow Accumulation into Area

Figure 12. The flow accumulation grid is a count of the number of upslope cells. In this instance, the raster has a 30-meter grid cell resolution. Multiplying the flow accumulation grid by 900 (the area, in square meters, of one cell) results in a grid of accumulated area in units of square meters.

3. Multiply the New Flow Accumulation GRID by the HUC Runoff GRID

Figure 13. The new flow accumulation grid measures upstream accumulated area; multiplying this grid by the HUC runoff grid (and correcting for measurement units) results in a measure of accumulated runoff in each cell in units of cubic millimeters per year.

4. Use Map Algebra to Convert to Flow

Figure 14. The HUC runoff GRID is in units of volume divided by time (cubic millimeters per year); further calculations correct the units to cubic feet per second (cfs).

5. Select a Threshold Based on Flow

Figure 15. Using a conditional statement outlined in the GIS definition, the threshold metric is now in units of cfs. The above map shows streams delineated from a GIS based on a threshold of 1 cfs or greater.

6. Estimated Streams Shown on Topographic Maps

Figure 16. The delineated stream network includes more streams than what is indicated on the 1:24,000 topographic maps. One possible reason for the differences is that topographic maps are static, whereas streams generated by the GIS method can reflect current hydrologic conditions.

Results

Gage	Estimated flow (cfs)	NNWIS reported flow (cfs)	% Difference	Basin Area (sq miles)
Chesterville Branch, MD	7.1	7.26	-2.20	6.12
Little Buck Creek, IN	19.07	22.9	-16.72	17
Great Egg Harbor River, NJ	26.07	21.6	20.69	15.1
Salado Creek, TX	39.6	36.4	8.79	189
Little Cobb River, MN	81.7	82.7	-1.21	130
Tulpehocken Creek, PA	102.6	102.4	0.20	66.5
Maple Creek, NE	140.1	142.1	-1.41	368
Tucsawhatchee Creek, GA	233.05	242.6	-3.94	163
Sugar Creek, IL	299.02	276	8.34	446

The delineated stream network for the Chesterville watershed seems reasonable when overlaid on a USGS topographic map (fig. 16). The flow estimation method was applied to 10 USGS gages across the country, and the results were compared to reported annual flow. These results were also encouraging (fig. 17).

The GIS method for stream delineation has several useful characteristics. It provides a consistent approach to stream delineation which can be applied anywhere DEM and runoff data are available. The method also allows for stream delineation to be solely based on hydrologic characteristics, such as mean-annual flow. Streams can be defined based on any statistical measure (mean, median, min, max, etc.) of flow for any month and year, or can be averaged over months (average Spring flow) or years (30-year average flow). The results of this methodology have many implications for mapping and modeling. For instance, as the spatial resolution of DEMs becomes finer and more accurate with LIDAR and other sources, stream networks can be defined for areas where traditional maps do not provide adequate representation. Also, hydrologic simulation models such as MODFLOW, a 'raster-based' ground-water model which uses GIS to input streams for use in their stream-routing package, can take advantage of generating accurate streams to match the same cell resolution as the model itself.

References

Lo, Shuh-shiam, Glossary of Hydrology, Water Resources Publications, 1992.

ESRI Data & Maps [CD-ROM]. (2002). Redlands, CA: Environmental Systems Research Institute.

Regulatory Program of the US Army Corps of Engineers, Final Notice of Issuance and Modification of Nationwide Permits, 330 Federal Register March 9, 2000.