Scott N. Miller (1), D. Phillip Guertin (2), David C. Goodrich (3)
Offering a significant savings in time and labor, geographic
information systems (GIS) have improved the efficiency and
repeatability of geomorphologic assessment and hydrologic model
parameterization. The objective of this study was to couple field
measurements, spatial data, and GIS analytical capabilities to improve
our understanding of channel geomorphologic processes. A
high-resolution GIS database was constructed for the USDA - Agricultural Research
Service (ARS) Walnut Gulch Experimental Watershed (148 km2). Field
measurements of channel characteristics (cross-sectional area, width,
and depth) were taken at 222 sample points. To characterize the areas
contributing runoff to each of the sample points a suite of GIS tools
was developed in GRASS and ARC/INFO. A routine capitalizing on the
arc-node topology of stream vector data was created to order the
extensive channel network on Walnut Gulch. Relationships were derived
between channel shape variables and watershed characteristics with
robust predictive capabilities. Channel cross-sectional area and
width were found to be significantly related to channel order,
upstream watershed size, and maximum contributing flow length within a
watershed. The ability to accurately and efficiently model channel
characteristics in the southwestern United States offers the potential
for improving the performance of hydrologic models as well as aiding
the integration of hydrologic models and GIS.
KEY TERMS: GIS, channel morphology, hydrologic modeling, channel cross-section.
Watershed characterization based on geometric and physical
properties was carried out in a GIS on 222 subwatersheds within the
Walnut Gulch Experimental Watershed. At the same time a field
measurement program was completed in which channel shape
characteristics were measured at the outlet of each subwatershed.
Statistical analysis between the two data sets showed a strong
relationship between channel shape and watershed characteristics. It
was shown that the derivation of hydrologic model parameters may be
effectively carried out in a GIS on a large number of data points in a
relatively short amount of time.
With its long history of data collection and observational
data, Walnut Gulch serves as an excellent location on which to conduct
research into geomorphologic and hydrologic processes (Renard et al.,
1993) (link
to SWRC publications). Relatively little work, however, has focused on the
characterization of the entire watershed. This lack of data has
limited the ability to model landscape processes on a basin scale
(Lane et al., 1994). Additionally, most of the research on the
relationship between channel and watershed characteristics has been
conducted on intermittent and perennial streams. A knowledge gap
therefore exists for this type of data on aridland watersheds and the
processes acting on ephemeral channel systems such as exist on Walnut
Gulch (Osterkamp et al., 1983; Lane et al., 1994). Instead of
limiting this work to a small section of the watershed, it was decided
to characterize as much of the watershed as possible. Sample sites
were located randomly across the entire 148 km2 watershed within all
soil types and many hydrologic conditions. Strahler ordering analysis
(Strahler, 1952) and other measures of channel and watershed
characteristics were utilized to describe the watershed as
quantitatively and thoroughly as possible.
Analyses of basin characteristics have been carried out in a
GIS environment for many years (Burrough, 1986; Garbrecht and Martz,
1995), but many of these processes were found to be incompatible or
unworkable for the data collected during this project. Therefore, a
suite of GIS analysis tools and ARC/INFO Macro Language (AML) programs
were developed to facilitate the GIS investigation (trade names are
mentioned solely for the purpose of providing specific information and
do not imply recommendation or endorsement by the U.S. Department of
Agriculture). Channel shape, required for hydraulic routing, cannot
be accurately predicted (or extracted) from DEMs. Therefore a
principle goal of this project was to develop a methodology for
predicting channel shape from watershed characteristics that could be
readily derived from commonly available GIS coverages. During this
process, field research was synthesized with GIS applications and
photogrammetry to more thoroughly describe the channel and
geomorphologic characteristics than had previously been attempted.
Located in southeastern Arizona
(approx. 110oW, 31o45'N) and comprised of rolling hills and some
steep terrain, the elevation of Walnut Gulch Experimental Watershed
ranges between 1190 and 2150 m
A.M.S.L. Some urbanization exists in and around the town of
Tombstone, but cattle grazing and recreational activities are the
major land uses. Vegetation within the watershed is representative of
the transition zone between the Chihuahua and Sonoran deserts, and is
composed primarily of grassland and shrub-steppe rangeland vegetation.
Underlying Walnut Gulch is the geology of a high alluvial fan
contributing to the San Pedro River watershed (Renard et al., 1993).
Due to the enormous thickness and extent of the alluvial fill, the
groundwater reserves are substantial, and can be found at depths
ranging from 50 to 145 m (Libby et al., 1970). Some geologic control
over the hydrology exists in the western regions of the watershed
where metamorphic and orogenic activity has resulted in the fracturing
and faulting of the bedrock. In 1994 the USDA Soil Conservation
Service completed a detailed soil survey, finding that the watershed
is dominated by 30 principle soil types (Breckenfield et al., 1995).
Major soil units are Elgin-Stronghold (Ustollic Paleargid, Ustollic
Calciorthid), Luckyhills-McNeal (Ustochreptic Calciorthid),
McAllister-Stronghold (Ustollic Haplargid, Ustollic
Calciorthid), and Tombstone (Ustollic Calciorthid).
The climate of Walnut Gulch can be classified as semiarid or
steppe. Mean annual temperature in the city of Tombstone is 17.6
deg. C, with a mean annual precipitation of 324 mm. Annual
precipitation is highly variable in both timing and total depth. Rain
occurs mainly during two seasons: summer rains are
the product of monsoonal, highly localized, convective storms; winter
rains are generally low-intensity events that cover a larger
proportion of the watershed. The majority of runoff occurring on
Walnut Gulch is the product of summer storms, and is therefore
episodic and of relatively high intensity (Renard et al., 1993).
A field measurement program was undertaken whereby 222 channel
cross-sections were surveyed for morphometric assessment. To account
for basin-scale variability, a large number of randomly selected
points were used, and multiple measurements were taken at each site to
account for local variability in channel shape. These randomly located
sample point locations were pre-stratified by soil type using a GIS
procedure: each major soil type was assigned a weighted proportion of
the sample points based on the areal extent of the soil coverage. At
each site 3 cross-section surveys were taken to characterize the
channel section just above the outlet of the subwatershed. Width and
depth were measured at breakpoints (changes in slope). The three
surveys were then combined to determine the average width and depth
for the channel segment, and these results were combined to derive
average cross-sectional area.
A strict protocol was followed at each sample location in
order to ensure proper measurement and consistency between sites.
Upon arrival at a site, an inspection of the bank morphology,
vegetation, and soil characteristics along the entire reach was
completed to ensure that cross-sections were located where they would
be most representative of the channel section. A site description was
recorded in a log book for future analysis, and potential problems
related to channel complexity and morphology were noted where
applicable. Bankfull indicators, including slope breaks, changes in
bed or bank materials, a shift in vegetative type, debris lines, and
bank staining were noted in order to determine the bankfull depth
(Osterkamp et al., 1983; Gordon et al., 1992; Harrelson et al.,
1994). Wherever possible, evidence indicative of a constructive,
rather than destructive process, was used to determine bankfull depth.
In the southwestern United States channel processes are governed by
rapid and violent runoff events, and many of the channels on Walnut
Gulch are actively degrading. Channels that were clearly degrading
and out of equilibrium were not subjected to channel measurement since
an adequate determination of bankfull depth was not possible.
At each of the site locations a minimum of three
cross-sections were surveyed. If the channel reach was complex, up to
five cross-sections were measured to ensure adequate representation.
At each of the cross-sections a light line was pulled level across the
channel top at the bankfull depth. The line was leveled and pulled
taught to reduce sag. Measurements of channel depth and distance from
the left bank (looking upstream) were taken at each break in slope
across the cross-section.
Channel width was more easily measured with precision than
channel depth. Although determining the stage to which floodwaters
rise proved difficult, the possibility for error was greater when
measuring depth. This is due to a number of factors. First, depth
was only measured at break points, which are to some degree
subjective. Second, there was always a slight amount of sag in the
line when it was stretched across a channel, lending a source of
imprecision to the depth measurements. Third, more random deposition
or scour of the stream channels tends to impact local channel depth
measurements to a greater degree than width measurements.
Given
Walnut Gulch's history as a research site into various
aspects of hydrology and natural resource management and its extensive
rainfall and runoff database, it was decided that the GIS database
would be created at a resolution not ordinarily attempted. Throughout
the database development, an answer to a basic question was sought:
what are the highest levels of precision and accuracy that could be
achieved? There can be a tendency by GIS developers to overestimate
the level to which data may be discretized. By attempting to create
maps with a higher resolution than is allowable by the data, errors
may be introduced, and a false level of analysis can be attempted
(Burrough, 1986). Fortunately, data available for Walnut Gulch were
of a quality that allowed for a very high level of resolution and
positional accuracy.
Of particular relevance was the stream channel
coverage. In many GIS studies, the channel network is derived from a DEM in a
raster environment and then translated into vector data.
Alternatively, channels may be digitized from USGS topographic maps,
but channels drawn on these maps are partly based on DEMs.
Traditional GIS technique dictates that the majority of channels be
digitized as single vectors bisecting the channel position, with a few
of the larger drainages characterized with two lines to illustrate
relative width. Since a correlation was to be made in this study
between channel shape and watershed variables, a channel network map
was constructed whereby only the smallest channels were digitized as
single vectors. Channels wider than approximately 1.5 meters were
drawn as polygonal features. This highly detailed procedure relied on
the 1:5000 orthophotographs as the base from which the stream
positions and characteristics could be extracted. Most of the channels
on Walnut Gulch were thus characterized in the GIS database as
polygons, with associated width characteristics. In addition, where
channel islands and bars were visible on the orthophotographs they
were digitized. Thus, the channel network theme layer provides a
detailed record of the channel system and its hydrologic
characteristics as existed at the time (April, 1988) the aerial
photographs were taken.
An important variable for the understanding of geomorphologic
relationships is stream order. In this case the intensive channel
network map was a drawback: because most of the channels were
digitized as polygonal features it was not possible to automatically
order the streams. To take advantage of GIS arc-node topology, the
stream channels were vectorized. First, the map was translated into
GRASS and rasterized with a one-meter resolution. The GRASS module
"r.thin", which draws a parallel bisector through polygons, was
executed on the stream map (Geographic Resources Analysis Support
System, 1991). Upon completion of the vectorizing process, the maps
were appended together and edited to remove spurious vectors created
as a byproduct of the thinning process.
The vector stream channel map was then re-exported into
ARC/INFO, which supports both vector and routing functions. An ordering routine was created that takes
advantage of the "from" and "to" node data structure that ARC/INFO
imposes on vector maps. All the streams first had to be oriented in
the downward direction (i.e. pointing downstream). Once the streams
were all pointing in the downstream direction, the ordering program
was initiated. By assigning all vectors that had an open-ended "from"
node an order value of one, it was possible to stimulate a cascading
effect, whereby all vectors were assigned a stream order based on
their relationship and connectivity to other channels.
A 10m resolution DEM provided the basis for the articulation
of subwatersheds and the creation of many theme layers important to
the statistical analysis of field data. Created from a large number
of spot elevation points, contour data, and a thinned version of the
channel network using the ARC/INFO tool "topogrid" (Environmental
Systems Research Institute, 1994), the DEM was resolved to a 10 by 10
meter gridded surface. Using the "selectpoint" option within the
"watershed" command in GRID, subwatersheds were delineated above each
of the 222 channels surveyed in the field. From the DEM theme layers
for flow direction and flowlength were created for each
watershed. Watershed characteristics that were derived with the GIS
included: watershed area; maximum flow length; average slope;
elevation characteristics; and watershed shape variables.
For the purposes of evaluating the relationship between
channel morphology and the contributing area, the relationships
describing the channel cross-sectional area were of primary interest,
and deterministic models were derived using regression
analysis. Channel cross-sectional area is a function of both channel
width and average depth and thus reflects the total channel response
to its hydrologic regime. Channel width can be extracted from a high
resolution GIS such as exists for Walnut Gulch. Therefore, given a
strong statistical relationship between cross-sectional area and
watershed parameters, it would be possible to fully articulate channel
geometry (width, depth, cross-sectional area) for all channels
throughout Walnut Gulch. This ability to model channel shape
accurately when a minimum of field data is available may benefit the
application of a host of hydrologic models that incorporate hydraulic
channel routing (i.e., HEC-1, Army Corps of Engineers- Feldman, 1995;
the USGS DR3M model-Alley and Smith, 1982; KINEROS-Woolhiser et al.,
1990).
Horton (1945) investigated the role of stream order on channel
shape and hydrologic processes. He found that stream order was highly
correlated to many watershed and channel variables, and that stream
order could be used as a predictive tool for these variables. Strong
relationships between stream order and channel shape were also found
to exist on Walnut Gulch (Table 1). In this project, statistically
significant differences were found to exist between the means of
channel width, depth, and cross-sectional area for each step in stream
order. Stream order, which is closely related to contributing area,
was found to exert a strong effect on channel shape, and was used to
stratify the data into subcategories for further analysis.
Table 1: Relationship of channel morphology variables to stream order.
Because of the time and degree of technical skill required for
the completion of geomorphology studies, individual projects have
historically been limited in size and scope. With the advent of
geographic information systems (GIS), these technical problems have
been assuaged. The GIS capability of storing large and diverse
quantities of spatial data allows for the complex analysis of many
sites to be carried out quickly, efficiently, and with a high degree
of repeatability (Burrough, 1986). However, GIS-based projects often
fail to integrate field-collected data with GIS spatial data. This
project was designed do relate the GIS characterization of spatially
distributed watershed characteristics with field measurements of
point-attribute data (channel cross-section surveys). These data sets
were related using statistical analysis to derive relationships
between watershed characteristics and channel shape.
Descriptive statistics implied that stream order was
significantly related to channel shape variables. An analysis of
variance showed that significant differences exist for channel width,
depth, and cross-sectional area between each stream order. Multiple
analysis of variance proved average watershed soil clay content to
have no influence on channel shape. Relationships between channel
shape variables and watershed parameters were investigated using
simple linear regression analysis. Having found strong relationships
between these variable sets, multiple regression analysis was employed
to further refine these relationships.
| Order / N | Average Width (cm) | Average Depth (cm) | Average Cross-Sectional Area (m2) | |
| 1 / 58 | 279.65 | 26.32 | 0.802 | |
| 2 / 65 | 404.32 | 34.57 | 1.47 | |
| 3 / 40 | 563.03 | 40.10 | 2.54 | |
| 4 / 26 | 960.39 | 54.94 | 5.63 | |
| 5 / 20 | 1967.42 | 52.58 | 10.58 | |
| 6 / 13 | 3329.99 | 79.69 | 26.21 |
Channel characteristics were related strongly and in a semi-log fashion to stream order. Average channel width, depth, and cross-sectional area were all directly related to order, with a break in the trend occurring between the fourth and fifth order channels, but only for channel width and depth; cross-sectional area maintains a semi-log relationship throughout each step in order. The average value for channel depth shows a decrease between channel orders four and five, which is out of trend for every other increase in order (Figure 1). However, there is a significant increase in channel width between the fourth and fifth order channels, effectively counteracting the decrease in depth so that the relationship between cross-sectional area and stream order remains consistent across each order. The overall effect on channel shape is an increase in the channel width:depth ratio, while the relationship of cross-sectional area to order (and, hence, upstream watershed area) remains consistent (Figure 1).
Figure 1: Semi-log plots of channel shape variables per stream order.
Channel width appears to be more sensitive to the influence of
watershed parameters than channel depth. Measured values of width
have a large spread in their data, while the values for depth show a
more central tendency with a lower variation. Without exception
channel width proved to have a higher coefficient of determination
than depth (e.g. r2 = 0.33 for depth, and 0.72 for width when related
on a log-log basis to watershed area) when regression analysis was
performed. In fact, depth proved to be resistant to any deterministic
model based on the variables used in this study. Some of this
resistance to forming a deterministic relationship may be a function
of the difficulties assocaited with precisely measuring depth in the
field. Fluvial characteristics are undoubtedly important to this
tendency: as flow energy increases in a channel, the channel will
adjust its shape to accommodate the increased level of power and
erosive energy. This can be accomplished through the widening, and/or
deepening of the channel. In the loosely consolidated soils of Walnut
Gulch, the channels appear to respond to elevated flow energy by
increasing their channel width proportionally more than depth.
Responding to the runoff they receive from uplands, stream
channels constantly adjust their shape to achieve equilibrium with the
flow volume. Changes in channel morphology may result in either
degradation or aggradation, with a resultant change in the width:depth
ratio, but the net effect is a change in the channel cross-sectional
area. As such, the measurement and analysis of channel
cross-sectional area is an effective method of illustrating the manner
in which channels are responding to watershed characteristics.
A strong relationship exists between channel area and the
maximum flow length within a watershed (r2 = 0.79). Table 2 shows
the results of regression models involving channel area. Long flow
lengths within a watershed have been directly related to discharge
(Leopold et al., 1964). With higher flows, the channel will become
enlarged, either through bed scour or bank erosion, to accommodate the
larger flows, resulting in an increased channel cross-sectional area.
Following the same reasoning, a strong relationship between channel
cross-sectional area and watershed area would also be expected. Data
collected in this research support that logical extension. A log-log
relationship (r2 = 0.68) exists between channel cross-sectional area
and watershed area. A strong relationship (r2 = 0.77) exists between
channel cross-sectional area and the watershed area:perimeter ratio, a
measurement of the rotundity of a basin, and hence an indicator of
basin response. Neither average watershed slope nor the relief ratio
correlated strongly with channel cross-sectional area. The log of
cumulative drainage length (total length of all channels in a
subwatershed) had a moderate relationship to the log of channel
cross-sectional area (r2 = 0.62).
In order to improve on the relationships derived using simple
linear regression, channel variables were related to watershed
characteristics using multiple linear regression. Multiple regression
analysis of channel cross-sectional area revealed the relatively
strong role that channel order played in the determination of channel
cross-sectional area. Systematic exploration of the watershed data,
using both stepwise forward and backward regression analysis, showed
that channel area was heavily dependent on stream order and the area
of and maximum flow length within the contributing watershed (Table
3). Depending on the subset of parameters investigated, it was
possible to extract a significant regression model with a number of
different independent variables. To avoid collinearity, multiple
pools of data were used during the regression analysis. For example,
the relief ratio, a product of the maximum flow length and maximum
elevation change, was considered separately from those two variables.
The same separation was used for basin shape variables and watershed
size. Note that a constant was not used in the analysis, and the
equations were driven through the origin.
Table 2: Results of linear regression analysis between channel area
and watershed variables
| Variable | Watershed Characteristic | r2 | Coefficient | Constant | Seyx |
| channel area | maximum flowlength | 0.79 | 0.001 | 1.83 | 3.46 |
| log channel area | log watershed area | 0.68 | 0.49 | -2.44 | 0.34 |
| channel area | area:perimeter ratio | 0.77 | 0.03 | 0.17 | 3.60 |
| log channel area | log cumulative channel length | 0.62 | 0.51 | -1.38 | 0.40 |
| Case | Regression Model | r2 | Seyx
|
| 1 | Ca = 0.686(So) + 0.065(Aw) + 0.909(Lm) - 0.006(h) | 0.849 | 3.36 |
| 2 | Ca = 0.40(So) + 0.009(Aw) + 0.821(Lm) - 0.006(h) | 0.851 | 3.35 |
| 3 | Ca = 0.72(So) + 0.095(Aw) +0.001(Lm) - 0.007(h) - 0.001(Dl) | 0.851 | 3.34 |
| 4 | Ca = 0.616(So) + 0.001(Lm) + 0.001(S) | 0.849 | 3.42 |
where: Ca = channel cross-sectional area (m2); So = stream order; Aw = subwatershed area (m2); Lm = maximum flow length (m); h = relief (m); Dl = sum of drainage lengths (m); S = basin slope.
Strong statistical relationships were derived between channel
variables measured in the field, such as width, depth, and
cross-sectional area, and a host of watershed parameters, including
channel order, watershed area, shape, drainage properties, and
elevation characteristics that were defined using a GIS. Channel
cross-sectional area was related in a deterministic manner to multiple
watershed variables, yielding models with strong coefficients of
determination (r2 > 0.84). Channel shape (and, hence, bankfull
stage) may thus be predicted from watershed characteristics readily
extracted from common GIS coverages.
Field data was successfully integrated with GIS-derived
results. Channel cross-sectional area and other field-measured
channel morphometric parameters were found to be strongly related to
watershed characteristics extracted from a high-resolution GIS. It is
preferable to collect field data when developing parameters for
application in hydraulic routing models, but field collection can be
costly and time consuming. The channel coverage created for Walnut
Gulch contains information on channel width. Using the values for
width that can be extracted for the GIS, in conjunction with the
developed regression models, values for channel depth and
cross-sectional area may be calculated for all channel segments within
the watershed. Relaionships developed upon verification outside Walnut
Gulch have the potential to overcome the inability of DEMs to
parameterize channel cross-section properties. In this fashion
hydrologic models can be parameterized using a GIS to aid in the
understanding of hydrologic processes in the southwetern United
States.
Alley, W. M., and P. E. Smith, 1982. Distributed Routing
Rainfall-Runoff Model - Version II. Computer Program Documentation
User's Manual. USGS-WRD open-file report 82-344. Gulf Coast
Hydroscience Center, NSTL Station, Mississippi.
Breckenfield, D. J., W. A. Svetlik, and C. E. McGuire, 1995. Soil
Survey of Walnut Gulch Experimental Watershed. United States
Department of Agriculture, Soil Conservation Service.
Burrough, P. A., 1986. Principles of Geographical Information Systems
for Land Use Assessment. Oxford University Press. New York, 194 pp.
Environmental Systems Research Institute (ESRI), 1994. ARC/INFO
ver. 7.0 manual (on-line documentation). Environmental Systems
Research Institute Corp. Redlands, CA.
Feldman, A. D., 1995. HEC-1 flood hydrograph package. Chapter 4 of
Computer Models of Watershed Hydrology. Water Resources
Pub. Highlands Ranch, CO, pp. 119-150.
Garbrecht, J., and L. Martz, 1995. TOPAZ: An Automated Digital
Landscape Analysis Tool for Topographic Evaluation, Drainage
Identification, Watershed Segmentation, and Subcatchment
Parameterization; TOPAZ Overview. USDA-ARS Publication NAWQL 95-1.
Geographic Resources Analysis Support System (GRASS), 1991. GRASS
ver. 4.0 Manual. U.S. Army CERL. Champaign, IL.
Gordon, N. D., T. A. McMahon, and B. L. Finlayson, 1992. Stream
Hydrology an Introduction for Ecologists. John Wiley and Sons. New
York, 526 pp.
Harrelson, C. C., C. L. Rawlins, and J. P. Potyondy, 1994. Stream
Channel Reference Sites: an Illustrated Guide to Field Technique. USDA
Forest Service General Technical Report RM-245. 61 pp.
Horton, R. E., 1945. Erosional Development of Streams and their
Drainage Basins: Hydrophysical Approach to Quantitative
Morphology. Geological Society of America Bulletin 56:275-370.
Lane, L. J., M. H. Nichols, M. Hernandez, C. Manetsch, and
W. R. Osterkamp, 1994. Variability in Discharge, Stream Power, and
Particle-Size Distributions in Ephemeral-Stream Channel
Systems. In: Variability in Stream Erosion and Sediment
Transport; Proceedings of the Canberra Symposium, December 1994. IAHS
Publication 224: 335-342.
Leopold, L. B., M. G. Wolman, and J. P. Miller, 1964. Fluvial
Processes in Geomorphology. W.H. Freeman and Co. San
Francisco, 522 pp.
Libby, F. J., D. E. Wallace, and D. P. Spangler, 1970. Seismic
Refraction Studies of the Subsurface Geology of the Walnut Gulch
Experimental Watershed, AZ. USDA-ARS 41-164, 14 pp.
Osterkamp, W.R., L.J. Lane, and G.R. Foster, 1983. An Analytical
Treatment of Channel-Morphology Relations. USGS Professional Paper
1288, 21 pp.
Renard, K. G., L. J. Lane, J. R. Simanton, W. E. Emmerich,
J. J. Stone, M. A. Weltz, D. C. Goodrich, and D. S. Yakowitz,
1993. Agricultural Impacts in an Arid Environment: Walnut Gulch
Studies. Hydrologic Science and Technology 9(1-4): 145-190.
Strahler, A. N., 1952. Dynamic Basis of Geomorphology. Geological
Society of America Bulletin 63:923-938.
Woolhiser, D. A., R. E. Smith, and D. C. Goodrich, 1990. KINEROS- A
Kinematic Runoff and Erosion Model; Documentation and User Manual.
USDA-ARS Pub. ARS-77, 130 pp.
2.
D. Phillip Guertin
3.
David C. Goodrich
Support for this project was provided by the USDA
Agricultural Research Service, Southwest Watershed Research Center,
Tucson, Arizona. The authors would also like to thank the
Advanced Resource Technology
Group (ART), University of Arizona, for the use of equipment,
space, and technical advice. Authors
1.
Scott N. Miller
USDA ARS - Southwest Watershed Research Center
2000 E. Allen Rd.
Tucson, Arizona 85719
tel: 520-670-6380 ext. 150
fax: 520-670-5550
miller@tucson.ars.ag.gov
University of Arizona
Advanced Resource Technology Group
325 Biological Sciences East
Tucson, Arizona 85721
tel: 520-621-1723
fax: 520-621-8801
phil@nexus.srnr.arizona.edu
USDA ARS - Southwest Watershed Research Center
2000 E. Allen Rd.
Tucson, Arizona 85719
tel: 520-670-6380 ext. 162
fax: 520-670-5550
goodrich@tucson.ars.ag.gov
home page
Maintainer:
AWRA Webserver Team
Last modified: 24Nov99 gaw
Copyright(c) 1996, American Water Resources Association
