Hydrological Sciences - Journal - des Sciences Hydrologiques, 32, 2, 6/1987
Spatial variability of infiltration rates on a
semiarid seeded rangeland
MOHAMED MERZOUGUI & GERALD F, GIFFORD*
Watershed Science Unit, College
Resources,
Utah State University,
Logan, Utah 84322, USA
of
Natural
UMC 52,
ABSTRACT
Field studies determined the spatial variability patterns of field-measured infiltration rates on
a semiarid seeded rangeland near Eureka, Utah, USA. Two
kinds of instruments, namely a double-ring infiltrometer
and a modular type rainfall simulator, were used in both
a moderately grazed pasture and an ungrazed exclosure
(protected for over 20 years) to measure infiltration
rates at 10 and 30 minutes. There were 104 grid points
per instrument on a 24 m x 24 m grid (2 m spacing) at
each site, with a total of 416 infiltrometer plots.
Dependency of specific infiltration rate measurements on
nearby measurements, as determined from autocorrelograms
and semivariograms, was nonexistent. Percent cover
(vegetation plus litter) explained from 18 to 36% of the
variance associated with measured infiltration rates.
Variabilité
spatiales
des vitesses
d'infiltration
sur un
pâturage artificiel
semi-aride,
RESUME
On a déterminé par des études au champs les
conditions de variabilité spatiale des taux d'infiltration mesurés sur un pâturage artificiel semi-aride prés
de Eureka, Utah, EU. Deux sortes d'instruments, à
savoir une infiltromètre à anneau doublé et un simulateur
de précipitations modulaire, ont été employés dans un
champ pâturé modérément, et aussi dans un terrain clôturé
non-pâturé (protégé pendant plus de 20 ans) pour mesurer
les vitesses d'infiltration en 10 et 30 minutes. Il y
avait 104 points de quadrillage par instrument sur un
quadrillage de 24 m x 24 m (2 m d'espacement) à chaque
site, pour un total de 416 parcelles pour infiltromètre,
Il n'y a pas de relation significative entre les résultats des mesures de la vitesse spécifique d'infiltration
même pour des points voisins. Ceci a été déterminé sur
des autocorrélogrammes et semi-variogrammes. De 18 à 36
pourcent de la variance associée aux vitesses d'infiltration mesurées sont expliqués par le pourcentage
d'involucre (végétation plus la litière),
*Now at:
Department Range, Wildlife
Nevada, Reno, Nevada 89512, USA.
Open for
discussion
until
1 December
and Forestry,
University
of
1987.
243
244 Mohamed Merzougui & Gerald F. Gifford
INTRODUCTION
The key process in semiarid land use hydrological modelling is the
determination of rainfall excess, which in turn requires the
identification of source areas. Realistic modelling of runoff
resulting from precipitation requires distributed models to account
for variations in the production of rainfall excess. A knowledge
of the spatial variability of soil properties, which determines
source areas, should be reflected in the models (Springer & Gifford,
1980).
Most spatial variability studies of infiltration rates have been
conducted on agricultural soils (Sisson & Wierenga, 1981; Vieira
et al., 1981; Wagenet, 1981). A recent study (Achouri & Gifford,
1984) looked at spatial and temporal variability of infiltration
rates on a seeded rangeland site in Utah. Autocorrelograms and
semivariograms revealed a complete lack of variance structure among
infiltration rates. A double-ring infiltrometer was used in that
study and the authors made the assumption that results would be
similar with a rainfall simulator (a closer approximation to natural
rainfall). This present study was designed to test that assumption.
The hypothesis was that the beating action of raindrops might alter
infiltration spatial variability patterns on sparsely vegetated
semiarid rangeland.
The specific objective of this study was to determine infiltration rate variability patterns on a grazed and ungrazed situation
using a rainfall simulator and also a double-ring infiltrometer.
Autocorrelograms and semivariograms were used to identify the degree
of dependence (zone of influence) of infiltration rates on the
distance between pairs of measurements.
METHODS
Study area
The study was conducted approximately 3.2 km
southwest of Eureka, Utah, in the west-central area of the state,
within Utah State University experimental pastures. In the 1950s,
the pastures were ploughed and seeded and then fenced into 24 units
each of area 28 ha. Two experimental areas within a pasture seeded
to crested wheatgrass (Agroypxon desertorum)
were utilized in this
study, one moderately grazed for several years during parts of June,
July or August by cattle (1.5 ha/AUM) and the other ungrazed for
over 20 years.
Average annual precipitation for the area is 200-300 mm, most of
which falls during the winter months. Soils in the area are a
coarse loamy variant of the Juab Series as described by Jensen
(1983). The taxonomic soil classification is a coarse loamy, mixed,
mesic, torrifluventic haploxeroll. Cracks 5 to 15 mm wide, 10 to
20 mm deep, with pentagonal configurations 100 to 150 mm in diameter
are common in relatively undisturbed areas.
Experimental
design
The study utilized a grid 24 m long and
24 m wide (2 m spacing) in both grazed and ungrazed sites, similar
to grids described by Achouri & Gifford (1984). Data were systematically collected in eight columns in the east-west direction (13
Spatial variability of infiltration rates
245
samples per column) with 104 samples in total per grazed and
ungrazed situation and per instrument type (416 total plots).
Procedures
A double-ring infiltrometer was used during the
summer of 1981 to measure infiltration rates. The inner ring
diameter was 0.305 m (0.073 m 2 ) and the outer ring diameter was
0.457 m. The infiltrometer was inserted to a depth of 0.1 m in the
soil with a minimum of disturbance. All plots (both instruments)
were prewet following installation with 51 mm of water and covered
for a minimum of 3 h in order to minimize confounding effects of
antecedent moisture. Infiltration rates were measured using a 76 mm
constant head for 32 minutes and rates were determined at 10 min
(measured over the period 8-12 min) and 30 min (measured over the
period 28-32 min).
Besides the double-ring infiltrometer, a modular type rainfall
simulator as modified by Malekuti & Gifford (1978) was used (Fig.l).
r" \D
Fig. 1
Rainfall simulator,
The simulator provides relatively uniform drops (2.7 to 2.9 mm drop
size) with a kinetic energy of about 30% of a natural storm.
Rainfall application rate over 0.37 m2 was 102 m m h l ± 2.5 m m h - 1
and infiltration rates were calculated as the difference between
246 Mohamed Merzougui & Gerald F. Gifford
rainfall application rate and runoff at the same time intervals as
used with the ring infiltrometer. The single ring used as a plot
under the modular type infiltrometer was geometrically and dimensionally similar (0.073 m 2 ) to the one used in the double-ring
infiltrometer, except that it had a small circular opening to allow
for runoff to flow out of the ring into a short length of plastic
tubing and then into a collection vessel. Water used met drinking
water standards and was obtained from a nearby well. Water temperatures were not measured.
Plot rings for both instruments were inserted about 0.2 m from
the original grid point. Their location around the original point
was randomly chosen among the four major quadrants without superposition of the plots. Disturbance of soil within the rings was
minimal.
Throughout the study, the modular type infiltrometer and plots
were enclosed with canvas curtains as a shield from any wind which
might tend to reduce uniformity of rainfall application.
Spatial variability analyses were made utilizing autocorrelograms
(Rendu, 1978; Davis, 1973) and semivariograms (Rendu, 1978).
RESULTS AND DISCUSSION
Most soil properties reflect long term geological processes that
occur during the genesis of soils. These processes are known to be
sequential in time and not random (Davis, 1973). Although it is
expected that nearby locations will have similar measured infiltration rates, a source of variation often causes place-to-place
differences in that property. This variation can be attributed to
some combination of experimental error, time variation, and spatial
variation.
Figure 2 is a typical diagram of the autocorrelation function
vs. lag distance as found in this study. Examination of autocorrelograms may disclose intervals of space at which the sequence
has a repetitive nature; further, it gives information about how
far apart elements become independent of each other. In this study,
as lag distances are increased, correlations drop rapidly to, and
12.0
16.0
20.0
LUG-METERS
Fig. 2 Typical autocorrelation function versus the lag distance. Data represent 10-min
infiltration rates on the grazed site for the rainfall simulator.
Spatial variability of infiltration rates
247
then fluctuate about, zero which means that the values being
correlated have no interdependent relationship. Therefore no
consistent trend exists in measured infiltration rates for the
adopted 2 m spacing.
An alternate and more powerful tool to analyse the degree of
dependency of spatially distributed data is the semivariogram.
Although the precise nature of the variability of a regionalized
variable (a variable distributed in space or time which exhibits a
specific structure) might be too complex for complete functional
description, the average rate of change over distance can be
estimated by means of the variogram. It is also used to define the
distance over which values are interdependent (Campbell, 1978;
Achouri & Gifford, 1984). Therefore, besides the fact that the
variogram provides information about spacing between samples, it
also reveals important characteristics of the geographical distribution of the regionalized variable over the area, which is often
called spatial structure or variance structure. A typical variogram
from this study is shown in Fig.3. The semivariograms indicated
•
12.0
i—i—i
16.0
20.0
i
i
24.0
i
i
28.0
i
32.0
DISTANCE-METERS
Fig. 3 Semivariogram for the 10 min infiltration rate in the ungrazed area (as
measured with the rainfall simulator).
that for all cases the variance yOi) reaches the range (the distance
beyond which infiltration rate measures are considered to be
independent of one another) at the first multiple of h which
corresponds to the sampling spacing adopted. This means that at
the 2 m interval, the semivariogram y(h) reaches a value approximately equal to the variance (a ) . A horizontal line can be fitted
to the computed semivariogram, which means that the covariance
between values is zero for all distances. It means that for the
sample spacing adopted, the observations are all independent of one
another, i.e. there is a complete lack of variance structure.
Average values for various treatments are given in Table 1.
Many factors and combination of factors contribute to the
variability in infiltration rates. Assuming homogeneous soil
characteristics and conditions, the uneven distribution of vegetation and litter may be a major element influencing infiltration
variability. Information related to the percent of total cover
(live vegetal canopy plus litter) for each 0.073 m plot was
248 Mohamed Merzougui & Gerald F. Gifford
Table 1 Mean values, standard deviations, and coefficients of variation for measured
infiltration rates
Infiltrometer,
site and time
MODULAR TYPE
Grazed site
t = 10 min
t = 30 min
Ungrazed site
t = 10 min
t = 30 min
DOUBLE RING
Grazed site
t = 10 min
t = 30 min
Ungrazed site
t = 10min
t = 30 min
Standard
deviation
(mm h"1)
Coefficient
of variation
26.5
17.3
16.1
19.0
0.61
1.10
49.4
30.6
22.1
22.8
0.45
0.75
53.1
49.2
19.0
18.3
0.36
0.37
134.5
124.4
56.8
54.8
0.42
0.44
Mean
(mm h - 1 )
collected (ocular estimates) and correlation coefficients for the
10 min infiltration rates were calculated (Table 2). Though the
correlation coefficients are significant, the correlation between
infiltration rates and percent cover is not strong. In the best
case, only 36% of the variation in infiltration rates can be
explained or predicted through knowledge of the cover distribution.
The uneven distribution of vegetal and litter cover is not the only
factor responsible for the plot-to-plot variability of infiltration
rates.
Table 2 Correlation between total cover (live vegetal + litter) and field measured
10-minute infiltration rates
Double-ring
(n=104)
Modular type
(n = 104)
R2
Rz
Site
Grazed site
t = 10 min
0.42**
0.18
0.48**
0.23
Ungrazed site
t = 10 min
0.60**
0.36
0.46**
0.22
•Significant at 0.05 level of probability.
The effect of cover on the spatial distribution of infiltration
rates was also investigated. Autocorrelograms and semivariograms
were constructed for all points having approximately the same cover
percentages. No noticeable improvement was observed to affect the
spatial distribution of measured infiltration rates. This, however,
Spatial variability of infiltration rates
249
is not unexpected given the relatively low correlation coefficients
shown in Table 2.
CONCLUSIONS
Results of this study closely parallel results reported by Achouri
& Gifford (1984), Infiltration rates measured with both a doublering infiltrometer and a modular type rainfall simulator were all
considered independent of one another even within a 2 m grid system.
By comparison, Wagenet (1981) and Vieira et al. (1981) found a zone
of influence for individual infiltration measures of 5.3 m and 50 m,
respectively, on agricultural soils. Since cover (live and litter)
explained only 18-36% of the variance associated with measured
infiltration rates, the complex nature of rangeland soils and
trampling patterns of cattle (on the grazed pasture) must account
for the differences. For example, on these same crested wheatgrass
pastures, Balph & Malechek (1985) found that cattle attempt to
avoid stepping on crested wheatgrass tussocks, because the tussocks
provide an uneven surface upon which to walk. They conclude that
stepping in the interspaces compacts the soil, makes the soil
surface more susceptible to erosion, and increases rate of tussock
growth in elevation. The result is a community comprised of pockets
of biological activity above and below the ground in an environment
of barren, often eroded, ground. Within the ungrazed area, the
release from even moderate grazing pressure results in a maximum
expression of biological activity and freeze-thaw processes in terms
of optimizing the surface infiltration process. Though exact
mechanisms for the variable but enhanced infiltration under ungrazed
conditions are not known, the phenomenon has been noted by others
(Rhoades et al., 1964; Dortignac & Love, 1961; Rauzi, 1963; Reed
& Peterson, 1961; Gifford, 1982).
ACKNOWLEDGEMENTS
This project was sponsored in part by the US
Agency for International Development and in part by the Utah
Agricultural Experiment Station (Projects 771 and 749), Utah State
University, Logan, 84322. Journal Paper 3023, Utah Agricultural
Experiment Station.
REFERENCES
Achouri, M. & Gifford, G.F. (1984) Spatial and seasonal variability of field measured infiltration rates on a rangeland site
in Utah. J. Range Manage. 37, 451-455.
Balph, D.F. & Malechek, J.C. (1985) Cattle trampling of crested
wheatgrass under short duration grazing. J. Range Manage. 38,
226-227.
Campbell, J.B. (1978) Spatial variation of sand content and pH
within single contiguous delineations of two soil mapping units,
Soil Sci. Soc. Am. J. 42, 460-464.
Davis, J.C. (1973) Statistics
and Data Analysis
in Geology.
John
Wiley, New York.
250 Mohamed Merzougui & Gerald F. Gifford
Dortignac, E.J. & Love, L.D. (1961) Infiltration studies on
ponderosa pine ranges of Colorado. US Forest Serv., Rocky Mt
Forest Range Exp. Sta., Sta. Paper 59.
Gifford, G.F. (1982) A long-term infiltrometer study in southern
Idaho, USA. J. Hydrol. 58, 367-374.
Jensen, S. (1983) Soil Survey of the USA Range Experiment Area in
Tintic Valley, Utah. White Horse Assoc, Smithfield, Utah.
Malekuti, A. & Gifford, G.F. (1978) Impact of plants as a source of
diffuse salt within the Colorado River Basin. Nat. Resour.
Bull.
14, 195-205.
Reed, M.J. & Peterson, R.A. (1961) Vegetation, soil and cattle
responses to grazing on northern Great Plains range. USDA Forest
Serv. Tech. Bull.
1252.
Rendu, J.M. (1978) An introduction to geostatistical methods of
mineral evaluation. S. African. Inst. Mining and Metallurgy,
Johannesburg.
Rhoades, E.D., Locke, L.F., Taylor, H.M. & Mcllvain, E.H. (1964)
Water intake on a sandy range as affected by 20 years of differential cattle stocking rates. J. Range Manage. 17, 185-190.
Sisson, J.B. & Wierenga, P.J. (1981) Spatial variability of steady
state infiltration rates as a stochastic process. Soil Sci.
Soc.
Am. J. 45, 699-704.
Springer, E.P. & Gifford, G.F. (1980) Spatial variability of
rangeland infiltration rates. Wat. Resour. Bull. 16, 550-552.
Vieira, S.R., Nielsen, D.R. & Biggar, J.W. (1981) Spatial variability of field-measured infiltration rate. Soil Sci. 126,
342-349.
Wagenet, D.W. (1981) Variability of field-measured infiltration
rates. MS Thesis, Utah State Univ., Logan.
Received
21 August
1986;
accepted
5 January
1987.