Watershed characterization in a temporal-geomorphic context
Dave Galbraith
CEE 6440-GIS in water resources
Introduction
This study represents an effort to compile physical and spatial information, both descriptive and quantitative, to illustrate the characteristics of several mountainous watersheds. The motivating goals of this work were to highlight similarities and differences between watersheds, supplement existing hydrologic analyses with a geographic framework, visually depict recent flow regimes, and develop a physically-based template to interpret observed channel change. Input data for the GIS project came freely from several various government and educational entities. The six watersheds contain long-term USGS gaging stations near their outlets, and also geomorphic study reaches whose downstream boundary defined the watersheds' outlets. The study reaches, established in the mid 1990s by previous USU graduate students, span 20-25 bankfull channel widths in length. They contain several monumented cross-sections, longitudinal profiles, detailed geomorphic maps, and substrate data gathered by Wolman pebble counts in addition to repeat photography.
The study area lies in the Uinta Mountains of Utah. The watershed areas are largely designated wilderness areas, and almost entirely on Forest Service land. Stream channels are fairly steep (from 2.5%-4% gradient) and often consist of complex, multi-threaded channel patterns. Their beds consist of gravel and cobbles, and channel interactions with hillslopes coincide with fresh landslides and resultant coarse sediment supply sources.
Study area with watershed shapefiles
(using
no outlines in symbology avoided displaying the sub-basins at this scale): 
Procedure
Detailed parameters and figures in this exercise focus on Whiterocks River, the easternmost watershed. This site, established in 1994 and surveyed in 1994, 1995, 1996, and 2003 coincides with a 96-year gaging record and exists as a free-flowing stream. The first steps taken were to locate on the internet and download several data sets for the entire study area: the NED site provided 30 m elevation grids and landcover data, the NHD provided preliminary drainage networks and the locations of gaging stations, springs, and various other features, while Oregon State's PRISM data provided a coverage with mean annual precipitation estimates.
Following data compilation, all coverages, grids, and shapefiles were projected into a UTM zone 11, NAD83 projection. They all originated in geographic (GCS_NA_83) coordinate systems, so this was a routine operation carried out with ArcTools. Hades Creek's gaging station was discontinued in the mid-1970s, so the NHD data did not contain this feature. A text file containing a single record of the gage's latitude and longitude in NAD27 was created and displayed in a separate map document; it was then exported as a shapefile and reprojected into the UTM projection. A new feature was then added to the larger NHD dataset to represent this gage. The establishment of the UTM projection allowed simpler area and distance calculations.
To facilitate further data processing, base layers were clipped to generate separate layers for each study watershed. Clip coverages, created in Arc and ArcEdit, provided the boundaries used when implementing the latticeclip and clip functions in Arc. Initial clip coverages were too small, and resulted in artificial stream piracy once grid processing was implemented. Boundaries were adjusted, and the NED clipped a second time.
The TauDEM module established drainage networks, discreet watershed boundaries, subbasin delineation, and flow accumulation calculations. Prior to TauDEM implementation, watershed outlets and reach boundaries were created based on the gage locations and reach lengths as established by using the measure tool in a map of 2003 survey data. Survey data was collected using an EDM, and the gages were all located within the study reaches; the distance from the gage to the top and bottom of the reach was defined by following the thalweg trace, so channel sinuosity effectively expanded the resulting reach if we assume the NHD networks fail to depict reach-scale meanders. The outlets were moved to lie on the appropriate NEDSRC grid cell, and TauDEM worked successfully.
Flow data from the NWIS system
identified the forcing mechanism for channel change. Using the “get NWIS data” tool in conjunction
with the tracking analyst generated a temporal theme by joining the flow table
with the TauDEM-delineated stream network.
Exporting the joined temporal table allowed the addition of a new field
that calculated a weighted flow estimate for each day (from
Qw = (downstream flow accumulation) * (mean daily flow) / (downstream
flow accumulation at outlet)
The tracking analyst then joined this new temporal table with the channel network to reestablish the spatial definition. The resulting temporal theme allowed the creation of an animation using the animation tool located in the tracking analyst, but due to its size became quite unwieldy to use within the GIS.
Channel cross-sections existed in a larger dataset of 2003 survey data; this information, while in an arbitrary coordinate system, can be brought into a separate map. This allowed the separation of cross-section survey points and their export to an appropriate text file. Following this, tabular data was processed in order to adopt the distance convention for 2-dimensional display that previous researchers used.
Results
General watershed characteristic variables for precipitation and topography are presented in the table below and were taken from the statistics box of the raster data properties, and calculated as area-weighted average precipitation in the case of the PRISM data. This involved the intersect command’s use in Arc. The channel network delineated by TauDEM failed to delineate a small perennial tributary that lies in the monitoring reach; its drainage density was far less that that of the NHD dataset. However, the NHD network appears to attempt to identify braided segments, but this level of detail seems inappropriate for the spatial scale of the NHD data; braided segments sometimes appear far from the actual riparian corridor and due to the dynamic nature of such channels are rarely accurately depicted. For these reasons the network derived from elevation data seems more appropriate.
Spatial analyst generated aspect and
slope grids to characterize the topography of the
From these plots we see that much of the watershed lies at intermediate elevations at relatively moderate slopes. The map of slopes indicates that bands of steep slopes exist near the valley bottoms in the lower watershed and at the basin divides in the upper watershed. The watershed is predominately south and east facing, with the dominate landcovers being evergreen forest and shrublands. Grasslands, bare rock, and deciduous forests occupy moderate portions of the landscape, with all other categories negligible.
USGS flow records indicate a 2-year flood event is 1,125 cfs; flows between 1996 (the year of the latest previous survey) and 2003 exceeded this threshold twice with magnitudes of 1,490 and 2,090 cfs recorded in 1998 and 1999 (see animation). This record suggests the possibility for some channel adjustment since 1996, but does not indicate the likelihood of massive erosion or catastrophic channel avulsion. Below are the channel cross-sections; note the vast amount of change following the 1995 flow of 2,240 cfs as compared to the relatively moderate change in the intervening 7 years.
Conclusions and future study
GIS is an excellent tool for interpreting documented geomorphic channel change and developing a geographic context for a variety of studies. Furthermore, many of its functions lend themselves directly to survey data processing; once georeferenced according to an arbitrary benchmark network, new survey points can be overlain on old map images to highlight areas of planview channel change.
Much of the readily accessible data I found did not apply well to a reach scale, although it did an excellent job at the watershed scale. This does not mean GIS is inappropriate at the reach scale, only that its application is different. Much of this fine-scale data must be compiled from field work or from the painstaking digitizing process from old analog maps or air photos. Further GIS work for this area will involve digitizing old geomorphic maps and possibly georeferencing them into real-world coordinates using high-definition DOQs. I would also like to investigate the application of survey data’s elevation values to digitized arcs in order to generate a more robust reach-scale surface that could help develop some flow modeling efforts. These would ultimately culminate in raster data that estimates boundary shear stress values in past years’ high flows. During the completion of current mapping exercises we hope to establish some plots of painted rocks that may help to narrow down the threshold flow for coarse sediment transport.
Acknowledgements
David Tarboton (USU, CEE)
Mark Smelser (Dept of Environmental Quality, CA)
Melissa Stamp
(BioWest, Inc.)