You are on page 1of 9

EARTH SURFACE PROCESSES AND LANDFORMS

Earth Surf. Process. Landforms 34, 18581866 (2009)


Copyright 2009 John Wiley & Sons, Ltd.
Published online in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/esp.1869

Mechanisms of ephemeral gully erosion caused


by constant ow through a continuous soil-pipe
G. V. Wilson*
USDA-ARS National Sedimentation Laboratory, Oxford, MS, USA

Received 18 June 2008; Accepted 6 November 2008

*Correspondence to: G. V. Wilson, USDA-ARS National Sedimentation Laboratory, 598 McElroy Dr. Oxford, MS, USA. E-mail: Glenn.Wilson@ars.usda.gov

This article is a U.S. Government work and is in the public domain in the U.S.A.

ABSTRACT: Ephemeral gully erosion is considered to be driven by convergent surface ow while the role of subsurface ow
is often overlooked. This study sought to characterize and quantify the soil erosion mechanisms associated with ephemeral gully
erosion by pipe ow. A soil pipe (1 cm o.d.) was formed in a 10 cm soil bed immediately above a 5 cm water restricting layer.
Flow into the soil pipe was established at steady-state rates of 190 and 284 l h1. Experiments were performed for pipe ow alone
and with rainfall. Despite a constant ow rate into soil pipes, pipe ow was highly unstable due to internal mass wasting clogging
soil pipes until pressure increases forced the debris plug out of the pipe. Short (1020 Seconds) periods of negligible ow were
followed by surges in ow with high sediment concentrations that included a high proportion of aggregates. Increases in soil
water pressures associated with these debris ows were observed but were likely not representative of the pressures inside the
soil pipes due to hydraulic non-equilibrium between the soil pipe and soil matrix. Hydraulic non-equilibrium resulted in hydraulic
gradients in the opposite direction of ow through the soil pipe during early stages of pipe ow. Pipe ow rates and sediment
concentrations during debris ow periods were likely more extreme than observed due to integration over the three minute sample
collection interval. The Slot Erosion Test (SET) was extended to conditions of constant ow rate through an internal soil pipe that
was observed at the front face of a soil bed. The modied SET provided estimates of shear stress between 22 to 53 kg m1 s2,
however, the technique did not prove effective for estimation of the soil erodibility coefcient. Published in 2009 by John Wiley
& Sons, Ltd.

KEYWORDS: ephemeral gully; internal erosion; soil pipe; preferential ow; shear stress

Introduction contributes to gully erosion by seepage ow (Huang and


Laen, 1996; Darboux and Huang, 2005) and/or by preferen-
Wilson et al. (2008a) estimated that 54% of the sediment yield tial ow through soil pipes (Wilson et al., 2008b). The term
from a 11 000 ha watershed in Mississippi was from gully piping sometimes refers collectively to both mechanisms of
erosion. This is consistent with ndings by the USDA-Natural subsurface ow erosion (Dunne, 1990; Bryan and Jones,
Resource Conservation Service (NRCS) (1997) for 17 States in 1997), but these processes are distinguished in this paper by
the USA of ephemeral gully erosion accounting for 35% on referring to piping as strictly erosion resulting from ow
average of the total erosion. Poesen et al. (2003) did an exten- through a discrete macropore or soil pipe.
sive review on the contribution of gully erosion worldwide Macropore ow has been an important topic of soil physics
and found it to range from 10% to 94% with a median value research for almost 40 years following the, at that time, con-
of 44% for all conditions, whereas Capra and Scicolone troversial work of Thomas and Phillips (1979). Jarvis (2007)
(2002) found a range of 19% to 80% for agricultural water- noted in his review paper on macropore research that interest
sheds. The large range in contribution was explained in part has increased exponentially with over 50 papers currently
due to the scale-effect with the role of gully erosion increasing published per year on the subject. Despite the wealth of
as the size of the study area increases. Concentrated ow is research on preferential ow and transport of solutes and
generally considered the controlling process and subsurface microorganisms/colloids associated with macropores, apart
ow is often overlooked. However, subsurface ow contrib- from the recognition of the importance of macropore ow to
utes indirectly through the relationship of soil water pressure transport of particulate bound solutes, research on the role of
increase on soil cohesion decrease and can also directly result macropores in erosion is negligible.
in gully erosion independent of the impact on cohesion Flow through macropores or soil pipes has been recognized
(Chu-Agor et al., 2008; Fox et al., 2008). Subsurface ow as an important gully erosion mechanism (Faulkner, 2006) but
EROSION BY FLOW THROUGH SOIL-PIPES 1859

has received less attention in the Unied States where it is often observing pipe erosion processes. Neither method has been
assumed that gully erosion is exclusively a convergent surface used sufciently to describe ephemeral gully erosion.
ow process (Foster, 2005). Pipe ow has been attributed to Wilson et al. (2008b) simulated the effect of ow under a
about 60% of the gully erosion under agronomic conditions constant head through a non-erosive (i.e. no internal erosion)
in European elds (Bocco, 1991). Internal erosion by prefer- 3 cm diameter soil pipe that had been cut off at the head of
ential ow through soil pipes or cracks is considered the a gully and made discontinuous by tillage inlling the gully.
leading cause of failure of hydraulic works such as dams and That work did not address the role of ow through a continu-
dykes (Bonelli et al., 2006). Faulkner (2006) noted that internal ous soil pipe on ephemeral gully erosion. The objectives of
erosion of soil pipes may be occurring undetected at the this study were to simulate the conditions under which soil
surface until the soil above can no longer be supported and pipes that are continuous along a soil slope contribute to the
tunnel collapse results in a sudden development of a gully at development of an ephemeral gully by internal erosion. The
an advanced stage of development. In agronomic settings, purpose was to gain a better understanding of the controlling
such a gully would be lled in by tillage and the soil pipe mechanisms and evaluate the application of SET and HET
initiating the gully would be cut-off at the head of the previous techniques to the characterization of the soil erosion proper-
gully. Wilson et al. (2008b) studied the impact of discontinu- ties associated with pipe erosion for ephemeral gully condi-
ous soil pipes on gully erosion through a series of soil-bed tions. In contrast to the study of Wilson et al. (2008b), this
experiments with pipe ow under different constant heads study involved steady-state ow though an continuous, ini-
with and without rainfall on the surface. They found that ow tially 1 cm diameter, soil pipe that is allowed to enlarge by
through the soil pipe alone resulted in negligible erosion, internal erosion.
however, pipe-ow with rainfall produced a synergistic effect The difculty with this type of research is deciding what the
that resulted in rapid re-establishment of gullies with large soil most realistic boundary condition is to apply to a soil pipe for
losses. This work partly explains why ephemeral gullies re- ephemeral gully erosion. In natural settings of ow through a
establish in the same locations despite attempts to rell the macropore, if the pore is stable, i.e. no internal erosion, and
gullies and many times even redirect the surface ow such the pore is water-lled and in equilibrium with the water pres-
that convergent ow does not occur at that location. Poesen sure in the adjacent soil then a constant head will be appropri-
et al. (2003) noted that gully growth rates decline exponen- ate for the limited period in which the perched water table is
tially after formation, however, ephemeral gullies are by de- stable. However, perched water tables are highly dynamic,
nition lled in and re-established thereby producing an rising during early stages of a storm event and quickly dissipat-
intermittent step increase in erosion (Gordon et al., 2008). ing after rain cessation (Wilson et al., 1990; Wilson et al.,
Ephemeral gullies typically erode down to an erosion resis- 1991). For erosive soil with shallow soil pipes fed by a perched
tant layer such as an argillic horizon, plow pan, fragipan, water table, which is the case that this paper addresses, the
petrocalcic horizon or bedrock (Poesen et al., 2003). Such macropore would enlarge as preferential ow occurs and the
horizons may also be restrictive to root penetration and/or water table would not remain stable for any signicant time.
water percolation and result in a lateral spreading of roots and The rate of ow convergence from matrix pores into the mac-
perched water above the interface. Botschek et al. (2002) ropore would control the ow rate. In this paper it was
noted that such duplex soils are conducive to pipe-ow and assumed that as the macropore enlarges into a soil pipe, the
Wilson et al. (2006) reported observations of ow out of a ow into the pipe would remain stable at a rate fed by the
3 cm diameter soil pipe at the head of an ephemeral gully that adjacent pores. Application of a constant head may be appro-
formed down to the depth of a fragipan. priate for soil piping associated with levees or dams since the
Poesen et al. (2003) noted that the interactions of gully water reservoir is directly connected to the pipe and provides
development and the hydrologic processes, such as seepage essentially an innite source with minor drop in head as the
forces producing pipe erosion, need further research. Even less pipe enlarges. However, for shallow soil pipes associated with
work has been done on the development of models to describe ephemeral gully erosion a constant ow rate was deemed
these processes (Bonelli et al., 2006). Before models can be more appropriate.
developed for pipe erosion, the concepts behind the mecha-
nisms driving this process need to be quantied yet experi-
mental work is seriously lacking. Two simple methods, the
Slot Erosion Test (SET) and the Hole Erosion Test (HET), have Materials and Methods
been used to quantify the soils erodibility associated with
pipe erosion (Wan and Fell, 2004). The SET involves a 100 cm Experiments were conducted on soil beds in a 100 cm wide
long soil bed with 15 cm depth of soil packed in lifts to a by 150 cm long ume (Figure 1) at a 15% slope. Bulk soil was
desired bulk density. One side of the soil bed has a 10 mm collected from a depth of 0 to 10 cm from a Providence silt
by 22 mm preformed slot in contact with the transparent side loam (ne-silty, mixed, active, thermic Oxyaquic Fragiudalfs)
of the test ume. The upstream end has a 25 m head and the soil on the Holly Springs Experiment Station (HSES) in north
downstream side has a 03 m head. The growth of the slot Mississippi. The soil contains 15%, 69%, and 16% sand, silt,
width is monitored with time by video imaging and related to and clay, respectively. Soil was sieved to less than 2 cm and
the hydraulic shear stress. The HET uses a cylinder (115 cm maintained in eld-moist conditions for packing in 25 cm
long and 105 cm diameter) compacted to a desired bulk lifts. The bottom 5 cm of the soil bed mimicked a water restric-
density with a 6 mm diameter hole drilled through the center. tive layer by packing silty clay loam material to the average
The cylinder is oriented horizontally and a constant head bulk density (157 g cm3) of fragipans in this area (Rhoton and
established on the inow and outow ends. Growth of the Tyler, 1990). Providence soil was packed above the restrictive
internal diameter of the pipe is not measurable so the ow layer to a bulk density of 13 g cm3, typical of surface condi-
rate is used as an indirect estimate of the pipe diameter. Wan tions. The upper end of the ume had a port for hydraulic
and Fell (2004) found a close correlation in the erosion rate connection of a water reservoir to the articial soil-pipe
indices between the two methods but recommended that the created immediately above the water-restrictive layer. Prior to
HET be used for characterizing soil properties because it is soil packing, a 1 cm o.d. rod that extended the length of the
easier and less expensive, whereas, the SET is preferable for soil bed was connected to the port on the upper side. After

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
1860 G. V. WILSON

Flow Rate
Control
Tensiometers

8 cm
4 cm
25 cm
12cm
50 cm

10 cm 75 cm
5 cm

Runoff
collection flume
Figure 1. Illustration of the 10 cm thick soil bed in a 100 cm wide by 150 cm long ume at a 15% slope with a 150 cm long soil-pipe extend-
ing from the upper end to the open face lower end just above the 5 cm thick water restrictive layer. Tensiometer locations are indicated by solid
circles at distances of 25, 50, and 75 cm from the upper end. A pump maintains a steady-state ow rate through the soil-pipe. This gure is
available in colour online at www.interscience.wiley.com/journal/espl

packing, the rod was removed from the lower end thereby simple boundary condition of steady-state ow was chosen
creating a 1 cm i.d. soil pipe. to simulate the pipe erosion process using ow rates equiva-
Tensiometers were inserted vertically into the soil bed lent to the pressure heads used by Wilson et al. (2008b) but
(Figure 1) at three distances along the soil pipe. Each location for a soil pipe at its initial 1 cm diameter condition prior to
included four tensiometers with their porous cups for two enlargement. Constant heads of 15 and 30 cm were estab-
tensiometers positioned 1 cm above the water restrictive layer lished on a 1 cm i.d. PVC pipe and the steady state ow rates
and 6 cm on either side of the soil pipe center, and two determined. From these measurements, steady-state ow rates
directly above the soil pipe at distances of 2 and 6 cm (i.e. 2 of 190 and 284 l h1, respectively, were used on the pre-
and 6 cm depth) from the top of the soil pipe. Tensiometers formed soil pipe.
were monitored on one minute cycles by a data logger. Experiments included combinations of pipe-ow with and
A rainfall simulator consistent with the design by Meyer and without rainfall and each treatment combination was con-
Harmon (1979) was located 3 m above the soil surface. It ducted in duplicate experiments. A total of 10 experiments are
consisted of two overlapping Veejet nozzles (80150). Nozzles reported here but additional exploratory experiments were
oscillated back and forth across the soil bed in order to apply also conducted. The following four treatments were con-
uniform rainfall at a rate of 65 mm h1 with an impact energy ducted in which the antecedent soil water content was near
of 211 kJ ha1 mm1. Rainfall was applied for one hour dura- eld capacity (20% gravimetric water content): pipe ow
tion, followed 05 hour later by a 05 hour duration rainfall, alone under steady-state ow rate of 190 l h1, pipe ow alone
and 05 hour later by a nal 05 hour duration rainfall. at steady-state rate of 284 l h1, pipe ow at steady-state ow
The pressure head in situ on soil pipes is controlled by their rate of 190 l h1 simultaneous with rainfall, and pipe ow at
depth as pieziometric observations on these soils commonly 284 l h1 simultaneous with rainfall. An additional treatment
indicate perched water tables that reach the soil surface consisted of pipe ow under 284 l h1 simultaneous with rain-
during high precipitation events. Soil pipes typically develop fall but with the soil bed pre-wet by simulated rainfall to near
immediately above the water-restrictive layer, thus the depth saturation prior to the experiment.
to the fragipan governs the pressure head. Fragipan depths Total runoff volume was collected over three minute inter-
on the loess soils in this region vary due to past erosion but vals and a sample taken for sediment analysis. Sediment
typically range from 15 to 112 cm at the HSES (Rhoton and content was determined by decanting excess water and then
Tyler, 1990). Therefore, Wilson et al. (2008b) used subsurface evaporating to oven-dryness (105 C). If mass wasting occurred,
ow at constant head pressures of 15 and 30 cm in the study the slumped material was collected manually by cleaning out
of discontinuous soil pipes for this same soil. For this study the weir section of the ume immediately after mass failure.
on soil pipes that are continuous through the soil bed, a Care was taken to only remove the slumped material in the
constant head approach was deemed inappropriate due to weir section while leaving soil deposited from internal erosion
pore enlargement by internal erosion would lead to ow rates that was not yet sampled. The slumped material was weighed,
that would not be supplied under natural conditions by con- and sampled to determine water content. The dry mass of
vergent ow from matrix pores into the soil pipe. To facilitate sediment loss by mass wasting was calculated after correcting
future modeling and be consistent with the SET method, the for the water content.

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
EROSION BY FLOW THROUGH SOIL-PIPES 1861

Technical Approach
SETs and HETs are typically performed to characterize the
susceptibility of embankments, e.g. dams and levees, to soil
Soil bed
piping. When a soil pipe develops that extends through an
embankment, the ow rate will increase as internal erosion
enlarges the pipe. According to Poiseuilles Law, the ow rate at
will be proportional to the fourth power of the pipe radius.
Despite the potentially enormous ow rates, the soil pipe will
have essentially an invariant head for conditions of an embank- bt Area
ment. However, for ephemeral gully erosion conditions, ow Di
eroded
through the soil pipe does not have a semi-innite water
source. The preferential ow rate through the soil pipe is
controlled by the rate of convergent ow into the pipe. While
the soil pipe may enlarge by internal erosion, the ow rate Restrictive layer
does not necessarily increase. Thus, to simulate ephemeral
gully erosion processes the experiments were performed
under a constant ow rate that was equivalent to a constant Figure 2. Circular soil-pipe with initial diameter, Di, of 1 cm in the
head established on a 1 cm diameter soil pipe. soil bed immediately above the water restrictive layer. At time t, the
The analytical methods used to describe erosion by pipe soil-pipe eroded into half ellipse with dimension of major axis at and
ow for HET and SET methods (Wan and Fell, 2004) were minor axis bt.
modied to describe pipe ow for ephemeral gully erosion as
simulated in these experiments. That is for conditions of ow
through a soil pipe in the center of a soil bed, immediately
above a water restricting layer, under steady-state inow, with t = w g ( h L )( Dt 4 ) (4)
the erosion of the soil pipe observed at the exposed outow
face, Figure 1. For these conditions, the internal erosion within For an elliptical soil pipe, the area of the pipe, Ap, is simply
soil pipes has been described by the same excess shear stress the product ab and the wetted circumference, Cw, can be
equations as commonly applied to open channels: approximated by

qs = K ( c )

(1) { (
Cw = ( a + b ) 1+ 3r 10 ( 4 3r ) 2
1

)} (5)

where the sediment ux, qs (in kg s1 m2), is related to the soil where the dimensionless shape parameter, r, is
erodibility coefcient, K (s m1), and the hydraulic shear stress,
r = ( a b ) ( a + b )
2
(in kg m1 s2) above a critical shear stress, c (in kg m1 s2), (6)
necessary for initiation of erosion with the assumption that the
exponent, , is unity. The hydraulic shear stress on the surface Based upon knowledge of the mass of soil loss, ow rate and
of the soil pipe at time t, t is pipe dimensions (D for circular or a and b for elliptical) at
selected times, the shear stress can be calculated. The shear
t = w g ( h L ) ( Ap Cw )
stress can be plotted as a function of the sediment ux, qs, and
(2)
t the erodibility determined from Equation 1 as the slope of the
linear regression. Additionally, the critical shear stress can be
where w is the density of the uid (in kg m3), g is the accel-
estimated from the intercept, although Wan and Fell (2004)
eration constant (98 m s2), h is the hydraulic head differ-
suggested that a more appropriate method for the critical shear
ence across the soil pipe (in meters), L is the length of the soil
stress would be to conduct successive HET runs at different
pipe (in meters), and (Ap/Cw)t is the ratio of the cross-sectional
heads in which a minimum shear stress is determined from
area of the pipe, Ap, and the wetted circumference, Cw, around
the minimum head at which pipe erosion is initiated.
the pipe at time t.
The sediment ux can be calculated by

qs = (1 Aw ) Mt t (3)
Results and Discussion
Differences among treatments
where Aw is the wetted area which equals CwL, if the mass of
soil loss, M (in kilograms), during time interval t is measured. The ow rates into the continuous soil pipe (PF) at the equiva-
Equation 3 assumes that the soil pipe has a uniform cross- lent head of 15 cm (190 l h1) were signicantly lower than
sectional area along its length that is represented by the the rates associated with the 30 cm head (284 l h1) (Table I).
change in cross-sectional area observed at the open face. Due These PF rates were also about two-orders of magnitude
to the water restricting horizon below the soil pipe, it was higher than observed by Wilson et al. (2008b) for 15 and
assumed that internal erosion will produce an elliptical soil- 30 cm constant heads maintained on discontinuous soil pipes
pipe in which the major axis width of the soil-pipe, a, is larger of 3 cm diameter that were blocked by lling of the ephemeral
than the minor axis height, b (Figure 2). The dimensions of the gully.
soil pipe are estimated from digital images at the bed face at For tests with pipe ow alone (no rainfall), advective transfer
each sampling time. through the periphery of the pipe into the soil matrix resulted
Assuming that the shear stress along the pipe length is con- in the runoff rate (Ro) being lower than the PF inow rate. For
stant, Equations 2 and 3 can relate t to changes in the pipe tests with rainfall, runoff from rainfall was in addition to the
dimensions observed at the front face (Figure 2). For a circular soil pipe runoff and thus runoff rates were higher than PF rates.
soil-pipe of diameter, Dt, at time t, Despite these factors, the differences between treatments

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
1862 G. V. WILSON

Table I. Mean response to ow through a continuous soil pipe, at ow rates of 190 and
284 l h1, which are equivalent to constant heads of 15 and 30 cm, respectively, on a 1 cm
diameter soil pipe, with and without rainfal

Head (cm) Rain ASW Ro (l/h1) PF (l/h1) SC (g/l1) SL (kg) MW (kg) TSL (kg)

15 No FC 1563a 1883a 450a 210a 230ab 440a


30 No FC 2665b 2843b 418a 342a 000a 342a
15 Yes FC 2164a 1890a 643a 454ab 293ab 747ab
30 Yes FC 3105b 2840b 843a 817bc 712b 1530b
30 Yes Sat 3260b 2841b 717a 746bc 667b 1413b

Note: Mean values for runoff rate (Ro), pipe-ow (PF) rate, sediment concentration (SC), soil
loss (SL) by internal pipe and/or sheet erosion, mass wasting (MW), and total soil loss (TSL).
The antecedent soil water condition (ASW) was either at eld capacity (FC) or pre-wet to near
saturation (Sat).
Values within a column separated by letters indicate signicant differences at 005 level.

(Table I) were the same for Ro rate and PF rate in that the
190 l h1 rates were signicantly lower than the 284 l h1 Ro
rates. A
For pipe ow alone, internal erosion resulted in sediment
concentrations ranging from 01 to 45 g l1 in the runoff but
mass wasting of the face generally did not occur. This is in
contrast to the sediment concentrations observed by Wilson
et al. (2008b) for discontinuous soil pipes which were essen-
tially zero but similar in that pipe ow alone rarely caused
mass wasting. The sediment concentrations were higher than
observed in situ for a continuous soil pipe exposed at a gully
face (Wilson et al., 2006) in which concentrations ranged from
02 to 85 g l1. The sediment concentrations for pipe ow
alone were the result of internal erosion within the soil pipe
caused by the high pore-water velocity exceeding the critical
shear stress of the pipe walls. When rainfall was combined,
sediment concentrations ranged from 03 to 77 g l1 and mass
wasting at the bed face always occurred. Wilson et al. (2008b)
observed sediment concentrations for rainfall alone under the
B
same rainfall regimes and soil conditions of this study to
average 225 g l1 for a 5% slope. The sediment concentra-
tions for pipe ow with rainfall were a result of internal erosion
of the pipe plus sheet erosion, i.e. soil detachment by raindrop
impact and surface ow exceeding the soil surfaces critical
shear stress. Sediment concentrations were generally 45% to
100% higher for the pipe ow plus rainfall test than the pipe
ow alone conditions, however, differences were not signi-
cant (Table I). There were signicant differences in soil loss
by internal erosion and/or sheet erosion, Table I. The soil loss
for the 190 l h1 test increased 100% when rainfall was
involved, however differences were not signicant. In con-
trast, soil losses were signicantly higher for the 284 l h1 test
when rainfall was included.
Wilson et al. (2008b) found for discontinuous soil pipes that Figure 3. (A) The soil bed prior to establishment of pipe ow.
pipe ow alone did not result in re-establishment of ephem- (B) The soil bed after pipe ow at 190 l h1 ow rate. The wetting of
eral gullies but that it was the synergistic combination of the soil along the pipe ow path is evident along with mass wasting
at the bed face.
rainfall with pipe ow that resulted in mass wasting that served
to re-establish ephemeral gullies. For continuous soil pipes,
only one of the four pipe ow alone runs resulted in measur-
able mass wasting, Figure 3. That occurred for the rst run
tested, a 190 l h1 run, in which the water restricting clay loam tinuous soil pipes but the mechanisms and amounts were
layer below the soil pipe eroded at the front face. This resulted considerably different. Wilson et al. (2008b) found mass
in the front face being undercut which led to sudden mass wasting of 162 to 629 kg for the 15 and 30 cm head tests on
wasting of the front face. This is probably not representative discontinuous soil pipes, respectively, whereas for continuous
of the natural response to pipe ow alone. The six tests with soil pipes the mass wasting was only 29 and 71 kg for steady-
rainfall combined with pipe ow all resulted in mass wasting. state ow rates associated with these respective heads. The
For two of these tests the mass wasting exceeded the soil loss discontinuous soil pipes exhibited sudden pop-out failures of
by internal erosion and/or sheet erosion. The increased pro- large blocks of soil while mass wasting for the continuous soil
pensity for failure with pipe ow combined with rainfall for pipes occurred as sloughing of the bed face around the soil
continuous soil pipes may be similar to the nding for discon- pipe.

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
EROSION BY FLOW THROUGH SOIL-PIPES 1863

Ro rate T2 T6 T10 Ro rate T2 T6 T10

168 0 270 0

SW Pressure (cm)
162 -15 240 -15

SW Pressure (cm)
DF
Ro Rate (L h )

Ro Rate (L h )
-1

-1
A DF
156 -30 210 -30
A
150 -45 180 -45

144 -60 150 -60


15 cm Head, Pipe flow only DF
138 -75 15 cm Head, with rainfall
120 -75
0 30 60 90 120 150 180 0 30 60 90 120 150 180
Time (min) Time (min)
Ro rate T2 T6 T10
Ro rate T2 T6 T10
300 0
410 15

SW Pressure (cm)
288 -15 375 0
Ro Rate (L h )

SW Pressure (cm)
-1

Ro Rate (L h )
-1
276 -30 340 -15

305 -30
264 -45
DF
B DF
270 -45
252 -60
30 cm Head, Pipe flow only 235 DF DF -60
240 -75 30 cm Head, with rainfall
0 30 60 90 120 150 180 200 -75
0 30 60 90 120 150 180
Time (min)
Time (min)
Figure 4. Temporal response of soil water pressure and runoff rate
Figure 5. Temporal response of soil water pressure and runoff rate
to steady ow into a soil pipe at (A) 190 l h1 rate and (B) 284 l h1
to steady ow into a soil pipe at (A) 190 l h1 rate and (B) 284 l h1
rate. Tensiometers were 2 cm directly above the soil pipe at distances
rate simultaneous with three intermittent rainfall applications.
from the upper end of 25 cm (T2), 75 cm (T6) and 125 cm (T10). The
Tensiometers were 2 cm directly above the soil pipe at distances from
arrows indicate periods of debris ow (DF) surges.
the upper end of 25 cm (T2), 75 cm (T6) and 125 cm (T10). The
arrows indicate periods of debris ow (DF) surges.

Soil pipes were observed to enlarge signicantly from the


initial 1 cm i.d. to over 5 cm. However, in no case tested did
internal erosion from pipe ow result in tunnel collapse of the lic gradients in the opposite direction of ow through the soil
bed surface as previously reported in the literature (Faulkner, pipe. Thus, lateral gradients during the stages of hydraulic
2006). An additional treatment was included to foster tunnel non-equilibrium can give erroneous hydraulic gradients
collapse by pre-wetting the soil bed to near saturated anteced- and directions of ow. De Vries and Chow (1978) observed
ent soil water conditions. This was accomplished by raining hydraulic gradients in the opposite direction of ow on for-
on the soil bed multiple times over a 24 hour prior to the test. ested hillslopes due to preferential ow through macropores.
Despite the higher antecedent soil water pressures (35 cm), Simon and Wells (2006) observed seepage from streambanks
the two tests behaved almost identical to the tests at eld while tensiometers suggested hydraulic gradients away from
capacity (75 cm) for the same hydrologic controls (284 l h1 the bank face. They reasoned that tensiometers measured
pipe ow with rainfall). tension in the soil matrix instead of the macropores and that
macropore ow resulted in the edge of the headcut to wet up
faster than indicated by tensiometers further back from the
Pipe ow hydrodynamics headcut. Thus, hydraulic gradients were in the opposite direc-
tion of the measured seepage. Wilson et al. (1997) used the
Typical tensiometer responses are shown for the tensiometers modeling results of Gwo et al. (1996) on hydraulic and physi-
positioned 2 cm directly above the soil pipe at distances of cal non-equilibrium during preferential ow through macro-
25 cm (T2), 75 cm (T6), and 125 cm (T10) from the upper pores to demonstrate how heterogeneity in the advective
inlet for the test with pipe ow alone (Figure 4) and pipe ow transfer coefcient, which controls the exchange between the
combined with intermittent rainfall (Figure 5). Soil water pres- soil pipe and soil matrix, can result in hydraulic gradients
sures exhibited an initial stage (<60 minutes) of hydraulic being observed by tensiometers in the soil matrix not being
non-equilibrium in response to pipe ow before an equilib- representative of water pressures in a macropore or soil pipe.
rium soil water pressure was reached. While the 284 l h1 Soil water pressures required up to 60 minutes to reach a
provided quicker tensiometric response than the 190 l h1 semi-equilibrium condition. Once reached, tensiometers
ow rate for all positions and tests, there was not a consistent 2 cm above the soil pipe generally did not exhibit positive soil
response pattern with slope position. There were instances in water pressures despite hydraulic heads on the pipe in excess
which the downslope position responded quicker than an of the 2 cm distance. However, tensiometers immediately
upslope position. This is clearly seen by comparing responses above the water restricting layer on either side of the soil pipe
for tensiometers T6 versus T2 in Figure 4(A), T10 versus T2 (not shown) did exhibit positive soil water pressures of up to
and T6 in Figure 4(B), T10 versus T2 and T6 in Figure 5(A) 6 cm of water. The exception was the rainfall tests at tensiom-
and 5(B). Each of these instances exhibit periods with hydrau- eter T2, and for a short time T6 at the end of one of the two

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
1864 G. V. WILSON

284 l h1 tests, in which pressures up to 5 cm of water were 0.6


temporarily observed. y = -0.14x + 0.84

-2
A

Erosion rate kg s m
Interestingly, the T2 tensiometer showed the marked R2 = 0.95

-1
increase in pressure during the period of no rain in which it 0.4
would be expected that pressures would decrease. An increase
in pressure during the period between rainfalls was also
observed for the 190 l h1 with rainfall test at position T6 but 0.2
the pressures did not reach positive values. The reason for the
unexpected rises in pressures is likely related to the observa-
tion of non-steady ow through the soil pipes. While the ow 0
rate into the soil pipe was controlled at a steady rate, ow out 0 1 2 3 4 5 6 7
of the soil pipe would commonly exhibit periods of blockage -1 -2
Shear stress, kg m s
that would temporarily stop the ow until sufcient pressures
built up to clear the debris plug. These debris ows would 0.6
subsequently exhibit surges with high sediment concentra-

Erosion rate kg s -1 m-2


B y = 0.12x - 0.47
tions. While the size distribution of the sediment was not R2 = 0.31
measured, these surges clearly had a high proportion of aggre- 0.4
gated material suggesting that this was caused by internal mass
wasting. During these short periods of pipe blockage, soil
water pressures for some tensiometers were seen to increase.
0.2
The actual increase was likely more dramatic than observed
due to the fact that tensiometer readings were integrated over
a one minute period.
0
Surges from negligible ow to very high ow typically
0 1 2 3 4 5 6 7
occurred over reciprocating intervals of 1020 seconds. These Shear stress kg m -1 s-2
surges in ow can be seen in the unstable ow rates in Figure
4 and, despite the pipe ow being combined with the surface Figure 6. Erosion rate as a function of shear stress for two of the
ow during rainfall periods, surges can be seen in Figure 5. 284 l h1 rate with rainfall tests. The slope of the linear regression
However, due to the ow sampling occurring over three equals the erodibility coefcient.
minute intervals, the measured runoff rates do not fully reect
the unstable nature of the pipe erosion regardless of whether
there was surface ow from rainfall excess. In fact, while the
Shear Stress Erodibility
runoff rate at time 110 minutes in Figure 4(B) and time 57 6 0.12
minutes in Figure 5(B) indicate jumps in runoff rate, these tests

Erodibility Coefficient (s m -1)


Shear Stress (kg m -1s-2)

each had eight or more other periods of debris ow surges 5 0.10

that do not show up because the runoff rate was integrated 4


y = 7.47x + 3.84
0.08
over a three minute sampling period. Despite the internal mass R2 = 0.21

wasting that produced these debris ow surges, tunnel col- 3 0.06

lapse was not observed. 2 0.04

1 0.02

Shear stress calculations 0 0.00


0 0.05 0.1 0.15 0.2
Internal erosion of the soil pipe was uniform around the cir- Sediment flux (kg s-1m -2)
cumference until pipe enlargement reached the water restric-
tive layer. After this point, the erosion elongated the pipe at Figure 7. Mean shear stress and erodibility coefcient from the 10
the base and produced a semi-elliptical pipe. Thus, for early experiments as a function of the sediment ux and the best t linear
times (less than six minutes) the shear stress was computed regression for shear stress relationship.
with Equation 4 and for later times using Ap = ab and Cw
being half of Equation 5. Computing shear stress based upon
dimensions of the soil pipe observed at the front face is depen- early measurements in these tests did exhibit the highest soil
dent upon the accuracy of the interpretation of the video losses however soil losses tended to continue to decline for
images for the values of a and b. This analysis became prob- long periods into the experiments which combined with fairly
lematic after around 21 minutes due to irregular erosion at the stable pipe ow rates resulted in declining sediment uxes
pipe face. Therefore values of shear stress were only computed with time. As a result, the calculated shear stress for eight of
for less than 30 minutes into the tests. Additionally, Wan and the 10 experiments showed a negative trend (Figure 6A)
Fell (2004) recommended that the early period of sampling between erosion rate and shear stress, i.e. unrealistic erod-
shortly after ow initiation (less than six minutes) not be ibility coefcients. Only two of the experiments showed a
included in the analysis due to loose materials being eroded positive relationship (Figure 6B) and thus an estimate of the
around the pipe that are not reective of erosion due to excess erodibility. This is likely a result of the ow rate not being
shear stress. allowed to increase as the soil pipe enlarged by internal
The shear stress is proportional to the mean ow velocity erosion as would be realistic for soil pipe through an embank-
for laminar ow and to the velocity squared for turbulent ow. ment that is supplied by a semi-innite reservoir.
Because these tests were for steady-state ow into a soil pipe The mean value of shear stress for each test did show a
and not a constant head in which the pipe ow rate would correlation to the mean sediment ux, Figure 7. However, the
increase as the pipe enlarged, the ow rate out of the soil pipe shear stress values cover a narrow range in ow velocities and
typically approached a near steady-state rate rapidly. The thus had a narrow range from 30 to 55 kg m1 s2.

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
EROSION BY FLOW THROUGH SOIL-PIPES 1865

Conclusions Capra A, Scicolone B. 2002. Ephemeral gully erosion in a wheat-


cultivated area in Sicily (Italy). Biosystems Engineering 83:
Pipe erosion experiments were conducted on a soil bed with 119126.
a preformed soil pipe that exited the bed face in which a Chu-Agor ML, Wilson GV, Fox GA. 2008. Numerical modeling of
bank instability by seepage erosion. Journal of Hydrologic Engineer-
steady-state ow rate of either 190 or 284 l h1 was estab-
ing 13: 11331145.
lished into the pipe. These ow rates were determined to Darboux F, Huang C. 2005. Does soil surface roughness increase or
represent preferential ow through a 1 cm diameter soil pipe decrease water and particle transfer. Soil Science Society of America
under 15 and 30 cm pressure heads, respectively. The runoff Journal 69: 748756.
rates, i.e. ow out of the pipe when no rainfall was applied, De Vries J, Chow TL. 1978. Hydrologic behavior of a forested moun-
quickly reached a near steady-state ow rate. Soil water pres- tain soil in Coastal British Columbia. Water Resource Research 14:
sures were slower to reach an equilibrium condition and until 935942.
they did, hydraulic gradients were occasionally observed to Dunne T. 1990. Hydrology, mechanics, and geomorphic implications
be in the opposite direction of the pipe ow. or erosion by subsurface ow. In Groundwater Geomorphology:
Sediment concentrations for pipe ow alone were fairly The Role of Subsurface Water in Earth-surface Processes and Land-
forms, Higgins CG, Coates DR (eds), Geological Society of America
high and were in the range observed in situ by Wilson et al.
Special Paper 252. Gelogical Society of America: Boulder, CO.
(2006) for a naturally occurring 3 cm diameter pipe. While Faulkner H. 2006. Piping hazard on collapsible and dispersive soils
differences in sediment concentrations were not signicantly in Europe. In Soil Erosion in Europe, Boardman J, Poesen J (eds).
different between pipe ow alone and pipe ow with rainfall, John Wiley & Sons: Chichester.
total soil losses were signicantly higher with rainfall. Internal Foster, GR. 2005. Modeling ephemeral gully erosion for conservation
erosion of soil pipes increased the diameter of pipes from 1 planning. International Journal Sediment Research 20(3):
to 5 cm; however, ow rates did not increase as a result of 157175.
pipe enlargement due to the imposed boundary condition of Fox GA, Chu-Agor M, Wilson GV. 2008. Erosion of noncohesive
a constant ow rate into the pipe. However, ow through the sediment by groundwater seepage: lysimeter experiments and mod-
soil pipes were not constant due to internal mass wasting eling. Soil Science Society of America Journal 71(6): 18221830.
Gordon LM, Bennett SJ, Alonso CV, Bingner RL. 2008. Modeling
resulting in stoppage of ow followed by surges in ow with
long-term soil losses on agricultural elds due to ephemeral gully
high sediment concentrations. These debris ow periods of erosion. Journal of Soil and Water Conservation 63(4): 173181.
negligible to high ows and high sediment concentrations Gwo JP, Jardine PM, Wilson GV, Yeh GT. 1996. Using a multiregion
occurred over 1020 second intervals. Occasionally, soil model to study the effects of advective and diffusive mass transfer
water pressures were observed to increase due to the pipe on local physical nonequilibrium and solute mobility in a structured
blockage. Despite this unstable internal erosion, continuous soil. Water Resource Research 32: 561570.
soil pipes, under the conditions tested of constant ow rate, Huang CH, Laen JM. 1996. Seepage and soil erosion for a clay loam
did not exhibit sudden development of mature ephemeral soil. Soil Science Society of America Journal 60: 408416.
gullies by tunnel collapse as suggested by Faulkner (2006). Jarvis NJ. 2007. A review of non-equilibrium water ow and solute
Given these constant ow conditions, the shear stress values transport in soil macropores: principles, controlling factors and
consequences for water quality. European Journal of Soil Science
computed for eight of the 10 tests produced erroneous values
58: 523546.
of soil erodibility when plotted against sediment ux. Meyer LD, Harmon WC. 1979. Multiple-intensity rainfall simulator
These results provide valuable insight into the soil pipe for erosion research on row sideslopes. Transactions of ASAE 22(1):
erosion processes. However, the application of SET and HET 100103.
methods for internal erosion of soil pipes under a constant Poesen J, Nachtergaele J, Verstraeten G, Valentin C. 2003. Gully
ow rate had limited success. Better methods of measuring erosion and environmental change: importance and research needs.
enlargement of the internal dimensions of the soil pipe with Catena 50: 91133.
time are needed to accurately compute the shear stress and Rhoton FE, Tyler DD. 1990. Erosion-induced changes in the properties
erodibility. While providing an innite water supply for a of a fragipan soil. Soil Science Society of America Journal 54:
constant head was questioned as an appropriate boundary 223228.
Simon A, Wells RR. 2006. Study of the effects of lateral seepage forces
condition for soil pipe ow related to ephemeral gully erosion,
on tension crack development, bank failure dimensions and
such a condition may be necessary to accurately estimate the migration of edge-of-eld gullies. In Proceeding of the 8th Federal
soil erodibility. Interagency Sedimentation Conference, 26, April, Reno, Nevada.
Conference proceedings on CD.
AcknowledgementsThe author wishes to express sincere apprecia- Thomas GW, Phillips RE. 1979. Consequences of water movement in
tion to Mr Alan Hudspeth and Mr Allen Gregory for there dedicated macropores. Journal of Environment Quality 8: 149152.
technical assistance that made this work possible. The use of the Mis- USDA-NRCS. 1997. Americas Private Land: A Geography of Hope.
sissippi Agricultural and Forestry Experiment Station facilities at Holly USDA-Natural Resources Conservation Service Washington, DC,
Springs was also appreciated. http://www.nrcs.usda.gov/news/pub/GHopeHit.html
Wan CF, Fell R. 2004. Laboratory test on the rate of piping erosion
of soils in embankment dams. Geotechnical Testing Journal 27(3):
References 295303.
Wilson GV, Jardine PM, Luxmoore RJ, Jones JR. 1990. Hydrology of
Bocco G. 1991. Gully erosion, processes, and models. Progress in a forested hillslope during storm events. Geoderma 46: 119138.
Physical Geography 15: 392406. Wilson GV, Jardine PM, Luxmoore RJ, Zelazny LW, Lietzke DA, Todd
Bonelli S, Brivois O, Borghi R, Banahmed N. 2006. On the modeling DE. 1991. Hydrogeochemical processes controlling subsurface
of piping erosion. Science Direct. C.R. Mecanique 334: 555559. transport from an upper subcatchment of Walker Branch Watershed
DOI.10.1016/j.crme.2006.07.003 during storm events: 1. Hydrologic transport process. Journal of
Botschek J, Krause S, Abel T, Skowronek A. 2002. Hydrological Hydrology 123: 297316.
parameterization of piping in loess-rich soils in the Bergisches Land, Wilson GV, Gwo JP, Jardine PM, Luxmoore RJ. 1997. Hydraulic and
Nordrhein-Westfalen, Germany. Journal of Plant Nutrition and Soil physical nonequilibrium effects on multiregion ow and transport.
Science 165: 506510. In Physical Nonequilibrium in Soils: Modeling and Application,
Bryan RB, Jones JAA. 1997. The signicance of soil piping processes: Selim HM, Ma L (ed.). Ann Arbor Press: Ann Arbor, M chapter 1,
inventory and prospect. Geomorphology 20: 209218. 3759.

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp
1866 G. V. WILSON

Wilson GV, Cullum RF, Romkens MJM. 2006. Pipe ow impacts on Watershed, Mississippi. Journal of Soil and Water Conservation 63:
ephemeral gully erosion. In Proceeding of the 8th Federal Inter- 420429.
agency Sedimentation Conference, 26 April, Reno, Nevada. Wilson GV, Cullum RF, Rmkens MJM. 2008b. Preferential ow
Wilson GV, Reid-Rhoades P, Bingner RL, DiCarlo DA, Dabney SM. through a discontinuous soil-pipe results in mass wasting and rees-
2008a. Conservation Practices in Little Topashaw Creek CEAP tablishment of ephemeral gullies. Catena in press.

Published in 2009 by John Wiley & Sons, Ltd. Earth Surf. Process. Landforms, Vol. 34, 18581866 (2009)
DOI: 10.1002/esp

You might also like