Earth Surface Processes and Landforms

Earth Surf. Process. Landforms, 23,1151-1170 (1998)

CALIBRATING STORAGE TANKS FOR SOIL EROSION

MEASUREMENT FROM PLOTS
VINCENZO BAGARELLOI*AND VITO FERRO~
' ~ i ~ a r t i m e nITAF,
r o [Link] Palermo, V i d e delle Scienze, 90128 Palermo, Italy

Received I July 1997; Revised 5 February 1998; Accepted 12 May 1998

ABSTRACT
Many plots for soil loss measurements are equipped with a sequence of tanks for measuring runoff volume and sediment
concentration. The stored water volume is easily determined by a water depth measurement while the sediment
concentration is often measured by collecting samples of the mixed suspension. In this paper, using the lI-theorem of
dimensional analysis, the functional relationship describing the mixing of the suspension in the tank is expressed in a
dimensionless form. The recognized dimensionless groups allow the establishment, for given soil and water depth in the
tank, of the relationship (calibration curve) between the actual and the measured concentration. The calibration curve,
using measured concentration values obtained from sampling taps located at different heights on the vertical of a tank wall,
is shown to be linear. For a given soil, the slope of the calibration curve is related to the water depth. We also show that the
same dimensionless groups allow the deduction of a scale-up procedure and the possibility of using a small tank, similar to
the field one, for investigating additional effects neglected in the theoretical analysis. The effects of the following factors on
the calibration curve are examined: (1) sampling direction along the vertical, i.e. starting the sampling from the upper tap to
the lower one or vice versa; (2) mathematical shape of the concentration profile; (3) representativeness of the calibration
curve; (4) sampling volume; (5) water depth into the tank (i.e. filling level); (6) suspension mixing time; (7) grain size
distribution of the eroded soil particles for given soil type; (8) time between the end of the erosive event and the starting of
the sampling procedure (delay time). Finally, an attempt is made to define a simpler sampling procedure based on a single
measurement with a dipped sampler. O 1998 John Wiley & Sons, Ltd.
KEY WORDS:soil erosion; plot measurements; sediment concentration; calibration curve

INTRODUCTION
Experimental studies on sheet and rill erosion due to overland flow are generally carried out by plot soil loss
measurement. Runoff from a plot bounded by galvanized steel or plastic is collected and carried by a
conveyance system (e.g. H-flume or pipe) to a sampling unit. The sampling unit can directly sample runoff for
measuring runoff volume and sediment concentration (e.g. Coshocton wheel; Fagna-type unit). The Coshocton
wheel (Carter and Parsons, 1967) has a sampling head with a narrow opening cutting across the jet from the
flume and collecting a given portion of the runoff which is transported to a storage tank. In the Fagna-type
hydrological unit (Bazzoffi, 1993), runoff, cleaned of the coarser material by passing through a sedimentation
tank, falls on a revolving pot,. For each pot rotation, a few cubic centimetres of the outgoing jet are intercepted
by a sampling hole and conveyed to a small tank. This sampling unit allows measurement of the hydrograph by
counting the number of rotations and the correspondingtime. Both sampling devices need field tests to calibrate
the measurement system.
A simple method for measuring the sediment concentration is to store all runoff or to divide it into a sequence
of tanks. In each tank, stored water volume is easily determined by a water depth measurement. Sediment
concentration can be measured by either catching the whole sediment amount, after siphoning the supernatant
cleared water, or collecting a sample of the mixed suspension (Pierson et al., 1994). In the first case, the

*Correspondence to: V. Bagarello, Dipartimento ITAF, Sezione Idraulica, Universita degli Studi, Vide delle Scienze,90128 Palermo, Italy, e-mail: begeve e [Link]
Contractlgrantsponsor: CEE
Contractlgrantsponsor: Ministero Universith e Ricerca Scientifica e Tecnologica, Govemo Italiano
Contractlgrant sponsor:ConsiglioNazionale delle Ricerche

CCC 0197-9337/98/131151-20 $17.50

0 1998 John Wiley & Sons, Ltd.
1152 V. BAGARELLO AND V. FERRO

sediment concentration measurement is accurate but it requires the removal of high mud volumes which have to
be oven-dried. Therefore, this technique is more and more cumbersome and time-consuming for increasing
amounts of collected runoff, and number and size of plots. As a matter of fact, the sample suspended sediment
concentration is representative of the whole suspension if it is well mixed, i.e. the suspended concentration Ci
assumes the same value in each i measurement point of the tank. Consequently the sediment amount is
calculated by multiplying the C, sample concentration by runoff volume.
Recently, Lang (1992) carried out an experimental study on a clay soil to test a bottle sampler for sampling
soil-water mixtures containing 1-187 gl-l of solid particles. The analysis showed that the actual suspended
particle concentration was underestimated by a factor of two. The results of the study induced Lang (1992) to
throw doubt on the reliability of soil loss data from plots collected using a runoff sampling technique.
Since in many plots the runoff and sediment concentration are sampled manually, Zobisch et al. (1996) also
verified the accuracy and the repeatability of this sampling procedure by comparing the sampling results of
different field workers. The authors showed that runoff volume was slightly underestimated while an
unacceptably poor accuracy of soil loss measurement was recognized. Zobisch et al. (1996) concluded that the
differences were due to the way the suspended sediment was stirred just before sampling and the way the
sample was retrieved (by plunging a beaker into the suspension). In fact the thoroughness of mixing of the
suspended sediment influences its homogeneity within the tank; moreover, the timing and plunging depth of the
sample beaker determine the concentration of suspended sediment.
In this paper, the results of an investigation carried out to determine an accurate manual sampling procedure
for measuring the suspended sediment concentration in a runoff-soil loss storage tank are reported. First, using
the 11-theoremof dimensional analysis (Barenblatt, 1987),the functional relationship describing the mixing of
the suspension into the tank is expressed in a dimensionless form. The recognized dimensionless groups allow a
scale-up procedure to be established and the possibility of using a small tank, similar to the field one, for
investigating additional effects (mathematical shape of the concentration profile, sampling direction along the
vertical, representativeness of a calibration curve, mixing time, etc.) neglected in the theoretical analysis.
Finally, an attempt to define a simpler sampling procedure based on a single measurement by a dipped sampler
is carried out.

STORAGE TANK CALIBRATING PROCEDURE

When considerable runoff events occur in a soil erosion plot, the most widely applied sampling procedure for
the sediment stored in a tank is to mix the whole suspension for a given time and to collect one or several
'representative7suspension samples. These samples are collected by either plunging a bottle to a given depth
into the suspension or sampling from some taps along a given vertical (frequently the vertical in the middle of a
tank wall, Figure 1).
The ideal condition is that after the mixing a constant sediment concentration C,is observed in each i point of
the suspension (complete mixing) and that Ci can be measured from each tap along a given axial vertical during
the sampling time. In this ideal case, the concentration profile along the vertical is uniform (Figure 1) and the
mean concentration C,, which is equal to the local concentration C , allows direct calculation of the actual
sediment amount stored in the tank. This also implies that a single sample taken at any point of the stored
suspension is representative of the actual sediment amount.
The ideal condition does not occur because the manual sampling procedure determines an incomplete
mixing condition and during the sampling time sedimentation phenomena take place within the suspension. As
a consequence, the concentration profile is not uniform and is characterized by increasing Civalues going from
the surface to the bottom of the suspension (Figure 1). Both factors (incomplete mixing, sampling duration)
suggest that the mean C , concentration,obtained by integrating the vertical concentration profile, could not be
representative of the actual concentration C of the suspension into the tank. The uncertainties obviously
increase if a single sample is used for estimating C .
Calculating the mean concentration C , from the measured Ci values implies the choice of a concentration
profile shape. The simplest one assumes a linear concentration variation between two subsequent measurement
points.

O 1998 John Wiley & Sons, Ltd. Earth Suif Process. Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGE TANKS

Figure 1. Diagram of a storage tank for soil loss measurement and typical concentrationprofiles

Hurst (1929) and Rouse (1938) experimentally investigated the concentration distribution in a vertical cylinder
in which a lattice structure mixed a suspension of water and sand having a diameter ranging from 0.03 mm to
0.9mm. The experiments by both investigators demonstrated that the concentration profile has the following
mathematical shape (Graf, 1984):

C' = exp[bo(yi - a)]

ca

in which Ci is the measured concentration at a distance yi from the tank bottom; Cais the reference concentration
at the distance y,=a chosen nearest to the tank bottom; bo=- w/&,in which w is the particle settling velocity and
E, is the diffusion coefficient of the soil particles. Equation 1 is obtained by integrating the differential equation
of diffusion by assuming a steady-state condition and a constant diffusion coefficient of the soil particles.
Experiments by Hurst (1929) and Rouse (1938) also showed that the agreement was not quite perfect for larger
particles. As a consequence, Equation 1 should be more useful for describing measured concentration profiles
for clay suspensions rather than for sandy ones.
Independently of the choice of the theoretical profile, the incomplete mixing condition due to the manual
sampling procedure suggests the need to determine a relationship, which will be termed the calibration curve,
between the actual concentraction C and the mean concentration C,.
For a tank having a square plane, with a side length L (m), which is filled up to a height h (m) (Figure 1) with a
suspension having an actual concentration C (kg m-3), the mixing phenomenon is expressed by the following
functional relationship:
=0
f(Ci,C,x,y,z,h,L,g,p,y,yS,d) (2)
in whichf is a functional symbol, x (m), y (m) and z (m) are the coordinates (y-z is the plene reported in Figure
I), g (m s-~)is acceleration due to gravity, p (kg s m-2) is water viscosity, y (kg m-3) is water specific weight, y,
(kg m-3) is soil particle specific weight, and d (m) is the particle diameter. If the suspension is sampled on a
given tank wall (x=O, Figure 1) and along its axial vertical (z=L/2), and taking into account that p and d can be
joined into the settling velocity w (ms-I), of particles (Graf, 1984),Equation 2 becomes:
f(C,,C,y,h,L,g,y,y,,w) = 0 (3)
in which f is a functional symbol.
Since the functional relationship (Equation 3) represents a physical phenomenon that does not depend on the
choice of measurement units, according to the II-theorem of the dimensional analysis (Barenblatt, 1987)
Equation 3 can be expressed using six dimensionless groups IIi (i= 1 to 6):
O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
1154 V. BAGARELLO AND V. EERRO

4 ( ~ 1 , ~ 2 , ~ 3 & & &=j 0) (4)

in which 4 is a functional symbol.
In particular, using the variables C, h and w as dimensionally independent, the II, group is determined by the
following relationship:
n,=[Link] (5)
in which 6,P and a are numerical constants. Substituting in Equation 5 the measurement units of each variable,
the following equation is obtained:
-k 6 -36mam"s-"kgm-3
1- g m (6)
Since II, group is dimensionless, the numerical values of 6,P and a have to be derived solving the system of
three equations deriving from Equation 6. Following the above-mentioned procedure, Equation 4 becomes:

which can be rewritten in the following explicit form:

in which is a functional symbol.

Equation 8 is also valid for C, which is calculated by integrating the Ci values:

in which 4, is a functional symbol. For a tank having a known L value, for given tank water level h and soil,
Equation 9 establishes that ratio b is a constant, i.e. the relationship between the measured mean concentration
C, and the actual one C, which will be termed the calibration curve, is linear.
For a given tank and soil type, Equation 9 becomes:

in which cp is a functional symbol, which establishes that the slope b of the calibration curve depends on the tank
water level.
A phenomenon is defined as self-similar in a given dimensionless group when the functionalrelationship II,
=II/ (112,II,, ......,II,), representing the physical phenomenon, is independent of II,. When the function II/ has a
limit equal to zero or infinity, the phenomenon is expressed by the following functional relationship:
n1= n ~ ~ l ( n 2,.....,
9 n,,)
~3 (I1)
in which 11/1 is a functional symbol and a is a numerical constant. This instance is termed incomplete self-
similarity in the parameter II, (Barenblatt, 1979,1987;Ferro, 1997).Since the II/ function vanishes for h +0, the

O 1998 John Wiley & Sons, Ltd. Earth SUI$ Process. Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGETANKS

phenomenon is incompletely self-similar with respect to ghlw2:

in which k and n are two numerical constants.

According to Equation 10, for each level h one experimental run (a pair (C,, C)) would be sufficient to
determine b; according to Equation 12 two experimental runs (for two different levels) would be necessary to
determine the functional relationship between b and h.
For tanks having different sizes (different L values), for a given soil type, Equation 9 becomes:

in which cp2 is a functional symbol. Since the (p function vanishes for L+O, the phenomenon is incomplete self-
similar also with respect to Llh:

where kl and m are numerical constants which can be estimated by two experimental runs, carried out for two
different levels in a tank of known side length L.
If the values of the two constants k1and m are determined using a small tank with a side length L,, which will
be named model, Equation 14 shows that the following scale-up relationship can be established with a
prototype tank having a side length L,:

where hpand h, are the water levels in the prototype and model tanks, respectively, and b, and b, are the slopes
of the corresponding calibration curves.
In conclusion, the scale-up relationship (Equation 15) allows the experimental investigation of other factors
affecting the calibration curve (sampling direction along the vertical, representativeness of the calibration
curve, sampling volume, mixing time, delay time, mathematical shape of the concentration profile) using a
small model tank.

EXPERIMENTAL INSTALLATION AND TESTING PROCEDURE

In a semi-arid Mediterranean area of Sicily an experimentalinstallation for plot soil erosion measurement was
built (Bagarello et al., 1996). The sampling device is a sequence of three tanks hydraulically connected by
triangular weirs (Figure 2). Each tank has a volume of about 1m3(0.94 m x 0.94 m x 1.16m) and is equipped with
10 taps installed along the axial vertical of a wall.
For calibrating a storage tank, these taps were installed with a constant spacing of 0.10m. A sandy soil (S
soil, Figure 3), a quasi-full (h= l.O6m) tank and five actual concentrations, ranging from 8.5 to 173gl-', were
used to determine the calibration curve. In particular, for each C value, water and soil put into the tank were
O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

Figure 2. Scquencc of htorasc tanks

Figure 3. Particle size distribution of the investigated soils

Figure 4. Diagram of the soil particles suspension method

O 1998 John Wiley & Sons. Ltd. Earth Sltrf Process. Latdforms 23.1151-1170 (1998)
 CALIBRATING STORAGE TANKS

0 field tank

0 smalltank

Figure 5. Calibration curve of the field and the small tank

mixed for a given time t, (t, = 15 min) by a flat scoop, in order to suspend the soil particles from the tank bottom
to the water surface (Figure 4). After the mixing, 10 suspension samples, each having a volume of 1.5 1, were
taken starting from the upper tap. For each actual concentration, a linear concentration variation between two
subsequent measurement points were assumed to calculate C,. Figure 5 confirms that the relationship between
the mean (C,) and the actual (C) concentration is linear with a slope b equal to 5.59. As a matter of fact,
coefficient b is very different from unity, confirming that the conditions of a poorly mixed suspension, the
sampling duration, and the use of samples collected along a single vertical produced a calculated mean
concentration less than the actual one. As a consequence, a calibration curve has to be determined to use a
storage tank as a soil loss measurement device.
To investigate the influence of factors affecting the slope of the calibration curve, we first tested a small-scale
model of the field storage tank. The laboratory tank has a volume of about 0.128 m3 (0.47 mx0.47mx0.58 m)
and is equipped with six taps having a spacing equal to 0.09m and installed along the axial vertical of a wall.
The same sandy soil S, a quasi-full (h=0.56m) tank and eight actual concentrations, ranging from 5 to 173gl-l,
were used to determine the calibration curve. For each C value, a mixing time equal to 15 min and a sampling
volume of 0.5 1 were used. Figure 5 also shows the comparison between the calibration curve, having a least
squares estimate of the slope b equal to 2.79, of the small tank and the experimental points. For actual
concentration values less than 35 Figure 5 also shows that the experimental points (C,, C) of the small tank
are near to experimental pairs (C,, C) determined for the field tank. Taking into account that the small tank and
the field tank are characterized by the same ratio Llh, this result agrees with the scale-up relationship (Equation
15) which would provide the same slope b of the two calibration curves. For C values greater than 35 gl-', the
slope of the two calibration curves is noticeably different. This result can be justified taking into account that the
mixing efficiency, the sampling procedure, the sedimentation phenomena, the tank volume, the sampling
volume and other factors can strongly affect the measured C, values.
Figure 5 shows that the experimental (C,, C) data pairs have a noticeable scattering around the calibration
curve, proving the need to use more than a single pair (C,, C), as Equation 10 establishes, to estimate coefficient
b.
To verify that the calibration curve is a straight line for given water depth and soil, for each of the four soil
types, ranging from a sandy soil to a clayey one (Figure 3) which will be used in the following experimental
investigation, the relationship between concentrations C and C, was determined. Each run was carried out by a
full tank (water depth h equal to 0-56m),a mixing time, t,, of 15 min, and a sampling volume V=0.51. For each
soil, Figure 6 shows the comparison between the experimental pairs (C,, C) and the linear calibration curve
whose b values are listed in Table I. For each calibration curve, the determination coefficient ? and mean square
error MSE are also listed in Table I. A linear calibration curve is generally adequate and useful for practical
application even if the scattering of the experimental data, which is more appreciable for the sandy soil (Figure
6a), clearly confirms the need to perform a series of (C,, C) measurements for establishing the calibration
curve.
O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

Figure 6. Calibration curves for the investigated soils

Table I. Statistical parameters of the calibration curves

for different soil types
Soil b r2 MSE

In the following, the results of an experimental investigation, carried out for the different soil types of Figure
3, will be reported. The laboratory runs were carried out to investigate the influence on the calibration curve of
the following factors: (1) sampling direction along the vertical, i.e. starting the sampling from the upper tap to
the lower one or vice versa; (2) mathematical shape of the concentration profile; (3) representativeness of the
calibration curve; (4) sampling volume; (5) water depth in the tank (i.e. filling level); (6) suspension mixing
time; "(7)grain size distribution of the eroded soil particles for given soil type; (8) time between the end of the
erosive event and the start of the sampling procedure (delay time). Finally, the results of an attempt to define a
simpler sampling procedure will be presented.

FACTORS AFFECTING THE CALIBRATION CURVE

Sampling direction along the vertical
To test the influence of the sampling direction along the vertical on the measured soil particle concentration
profile, four runs were carried out for the clay soil (Figure 3) by using actual concentration C equal to 15 and
40gl-'. For each C value, after mixing the suspension, the sampling was performed starting from the upper tap
to the lower one (up-down sampling) and vice versa (down-up sampling). Figure 7 compares the concentration
profiles measured for each sampling direction. The up-down sampling procedure is characterized by

O 1998 John Wiley & Sons, Ltd. Earth S u Process.

~ Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGE TANKS

Figure 7. Comparison between the concentration profiles measured for the up-down and down-up sampling directions

concentration gradients lower than the ones corresponding to the down-up sampling. The concentration
measured in the first tap is higher for the up-down sampling direction than for the opposite one because in the
first case the sampling occurs just after mixing; in the second case the measured concentration is reduced for the
suspension sedimentation processes which occur during the time spent for the sampling from the other taps. On
the contrary, the bottom concentration is higher for the down-up procedure because the sampling is carried out
from a well mixed suspension, while for the up-down sampling during the time between the end of mixing and
sampling from the last tap sedimentation of coarser particles occurs.
In conclusion,for establishing the slope b of the calibration curve, a sampling direction has to be selected. In
the following, the up-down procedure will be used because this sampling follows the sedimentation direction
and produces a more uniform concentration profile.
Influence of the mathematical shape of the concentrationprofile
A theoretical mean concentration, C,, is deduced integrating the concentration profile (Equation 1):

Cm,t = C2 [exp[bo(h - a)]- 1]

bo(h - a )

For each actual concentration C, Equation 16 allows the corresponding measured concentration C,, to be
calculated and the slope b, of the calibration curve to be determined for given water depth and soil. Figure 8
shows, as an example for CL soil, a good agreement between the pairs (em,,,C) and a linear calibration curve.
To test the influence of the concentration profile shape (linear, theoretical Equation 1) on the calibration
curve, the experimental runs corresponding to SL1 and CL soil and to actual concentrations ranging from 5 to
172.8g1-1 were examined. Figure 9a-d, as an example for two actual concentration values (5 and 30glV1),
compares the pairs (yi-a, Ci/C,) with the theoretical concentration profile. The agreement between
experimental pairs and Equation 1 is better for the CL soil than for the SLl one. This result can be explained
taking into account that sedimentation processes during sampling are less important for finer particles (CL soil)
than for coarser ones (SLl soil). For increasing Ci values, which are measured in the points nearest to the tank
bottom, the sedimentation and impact particle processes become more and more relevant. In other words, near
O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

Figure 8. Calibration curve for the clay soil and assuming the theoretical concentration profile (Equation 1)

0.8

c43 Oh
" 0.4

0.2

0
0 10 20 30 40 50 0 10 20 30 40 50
-
yi a [cm] yi - a [cm]

0.2 -
0 10 20 30 40 50 0 -1 %

-
yi a [cm]
0 10 20
-
30
yi a [cm]
40 50

Figure 9. Comparison between the theoretical concentration profile (Equation 1) and the measured concentration values

O 1998 John Wiley & Sons, Ltd. Earth Su$. Process. Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGE TANKS

200

-
Fi loo
U

50

0
0 20 40 60 80 100
Cm [dl

0 20 40 60 80 100
Cm,t [dl

Figure 10. Comparison between the calibration curves corresponding to the linear (a) and the theoretical (b) mathematical shape of the
concentration profile

Table 11. Statistical parameters of the calibration

curves of four different series of runs
Run b r' MSE
I 2.15 0.993 42.7
I1 2.08 [Link] 1.5
I11 2.11 0.994 35.6
IV 2.07 0.996 22.0

the bottom of the tank the agreement between (yi-a, CjC,) and Equation 1is worse than near the water surface
because the concentration values are so high that the diffusion scheme becomes inapplicable.
In conclusion, the theoretical-based profile is less adaptable for the higher concentration values measured
near the bottom and its choice does not improve the accuracy of the tank calibration curve (Figure 10). On the
O 1998 John Wiley & Sons, Ltd. Earth Su$ Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

Figure 11. Comparison among the calibration curves of different series of runs

other hand, the simpler linear profile reproduces the measured Ci values and allows a calibration curve to be
obtained which is as accurate as the previous one.
Representativeness of the calibration curve
To establish the number of run series useful for obtaining a representative calibration curve, the calibration
curve was determined for the SL2 soil four times by using a homogeneous soil mass taken from a single site at
the same time. Stones and crop residuals were removed from the soil mass. After oven-drying, the soil
aggregates were broken into small particles. Each linear calibration curve was experimentally determined by
four actual concentrations (C= 5,30,105.4 and 172.8 gl-l). Each run was carried out by a full tank, a mixing time
of 15 min, and a sampling volume V=0.51.
Table I1 shows, for each series of runs, (I, 11,111and IV), the coefficient b, the determination coefficient r2 and
mean square error MSE. Figure 11 compares the calibration curves corresponding to each series of runs. Both
Table I1 and Figure 11 show negligible differences among the four replicates. To verify that the four calibration
curves were not significantly different, the homogeneity of the residual variance of the four calibration curves
was at first verified by Bartlett test; an F-test was then applied for establishing the parallelism among the four
regression curves (Snedecor and Cochran, 1989). A 0.05 probability level was chosen for each test. The
statistical analysis confirmed that the four series of runs gave calibration curves that were not statistically
different and that a single b value equal to 2.10 can be used for the investigated soil. As a consequence, the
analysis showed that a calibration curve obtained from a single series of runs adequately describes the
relationship between the mean and the actual concentration. In other words, the same operator repeating the
same experimental procedure a given number of times is able to determine the same calibration curve
(repeatability of the calibration procedure).
Injuence of the sampling volume
For studying the influence of the sampling volume on the calibration procedure, two series of tests were
carried out for two soils (SLl and CL of Figure 3) and two sampling volumes V (V=0.5 and 1.51). For each test,
h=0.56m and tm=15 min were used. Each calibration curve was determined for 13 actual concentrations
ranging from 5 to 172.8gl-'. Figure 12 shows, for each investigated soil, the calibration curves corresponding to
each sampling volume. For each actual concentration C, C, values increase for decreasing Vvalues. This result
likely depends on the relationship between the sampling time and the used sampling volume; in fact, higher V
values cause higher sampling times during which the suspended particle sedimentation is more relevant. Since
the V influence on the b coefficient is noticeable, the sampling volume has to be considered as a constant
parameter of the whole calibration procedure.

O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGETANKS

Figure 12. Calibration curves corresponding to different sampling volumes

InJEuenceof the tank water level

Since in field conditions the complete filling of the tank is not always assured, we also verified whether the
calibration curve is affected by the water depth h in the tank according to Equation 12. From a practical point of
view, the water depth affects the number of taps usable for sampling operations and, as a consequence, the
measured concentration profile used to calculate C,. Three soils (S, SL2 and CL), a mixing time of 15 min and
seven h values ranging from 0.14 to 0.56m were used for this investigation. For each water depth, the
calibration curve was obtained by experimental runs carried out for seven actual concentration values ranging
from 5 to 172.8gl-l. For given soil type and water depth, the measurements confirmed that coefficient b is
constant (linear calibration curve). Figure 13 shows that for each soil the b(h) relationship assumes the
mathematical form established by Equation 12. For each soil, Table I11 lists the k and n coefficients.
It should be noted that the C, concentration is calculated by a concentration profile measured by a decreasing
number of taps for decreasing water depth in the tank, and moreover a different mixing effectiveness
corresponds to different water depths. In other words, two different effects, which always happen in the field,
are lumped together.
InJEuenceof the mixing time
The influence of the mixing time t, was studied by determining the calibration curve for the CL soil, in the
range 51~5172-8gl-', corresponding to three t, values (tm=5,10 and 15 min). In each experimental run the
tank was full. Figure 14 shows, for the CL soil and a mixing time equal to 5 min, that the linear calibration curve
can still be applied. Since the same result was also obtained for the experimental pairs (C,, C) corresponding to
0 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

Figure 13. Relationship between coefficient b and the water depth for different soils

Table 111. Coefficients k and m of

equation 12 for each investigated soil
Soil k n
S 251.6 0.5203
SL2 531.8 0.3646
CL 10558319 0.5111

tm=10min, we can conclude that the shape of the calibration curve is independent of the mixing time. For
verifying whether coefficient b is also independent oft,, the three regression lines were statistically compared.
The analysis showed that the slope of the calibration curve corresponding to tm=5min is significantly different
from the other ones; on the contrary, b coefficients were not significantly different for mixing times equal to 10
and 15 min (Figure 15). The experimental results show that the suspension mixing efficiency increases when t,
increases from 5 to 10 min (decreasing b values) while the efficiency can be considered practically invariable
when the mixing time increases up to 15 min. This result can probably be explained taking into account that for
high tmvalues the efficiency increase is balanced by the soil particle settling processes.
Influence of the input particle size distribution
According to Cogo et al. (1983), several factors affect the size distribution of primary particles and
aggregates of an eroded soil (type of erosion processes, soil type, surface conditions, slope length and
steepness, rainfall pattern and upslope deposition). These authors showed, for example, that the size
distribution of eroded aggregates is strongly dependent on the tillage system under the absence of residue cover.
For a given soil type, to test the dependence of the calibration curve on the characteristics (size distribution,
preliminary treatment) of the sediment entering the tank, proper experimental runs were carried out.
In particular, for the SL2 soil the tests were carried out using the following treatments: no altered soil (as
taken in the field) (NAS); oven-dried soil passed through a 4.76mm sieve (OD476); oven-dried soil passed
through a 2.00mm sieve (OD200); soil taken in the field with manual removal of rock fragments and crop
residuals (MD). For the CL soil, the tests were carried out with the same soil taken from two neighbouring
experimental sites (Sl, S2) and the eroded sediment stored in a field tank (ES). A full tank, a mixing time equal
to 15 min and concentrations up to 172.8gl-l were used.
Figure 16 shows, for the SL2 soil, the comparison between the calibration curves corresponding to the
treatments NAS, 0D476, OD200 and MD. The analysis showed that the calibration curves corresponding to

O 1998 John Wiley & Sons, Ltd. Earth Surf Process. Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGE TANKS

Figure 14. Calibration curve for the CL soil corresponding to a mixing time equal to 5 min

Figure 15. Comparison of calibration curves corresponding to different mixing times

different treatments coincide statistically. Figure 17 and the results of the statistical analysis show that, for the
CL soil also, the calibration curves corresponding to the different sediment sources are not significantly
different.
InJEuence of the delay time
All experiments previously described were carried out by putting the soil into the tank and sampling
immediately after the soil suspension mixing. For field applications, this procedure corresponds to a sampling
performed immediately after the end of the erosive event.
To test the effect of the delay time d,, i.e. the time between the end of the erosive event and the start of the
sampling procedure, the CL soil and an actual concentration equal to 25 gl-l were used. First, a cylindrical test
tube was used to visualize the sedimentation process of the selected soil. The test showed that most soil particles
settled after one hour. Figure 18 shows the measured C, concentrations for each selected delay time (d,=O, 1,2,
3,4,5,6,12,24 and 48 h). Each C, value was measured after filling the tank with clear water, putting the soil
into the tank, and mixing for 5 min after the selected delay time. With the exception of the concentration
measured at dt=O, which is affected by the input process of the soil into the tank, the C, values are practically
unaffected by the delay time; as a consequence, the scattering around the mean value of the concentration
values measured for dt>O (16.2~1-')is attributable to experimental errors. This result agrees with the observed
O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

Figure 16. Comparison of calibration curves corresponding to different treatments of the SL2 soil

Figure 17. Comparison of calibration curves corresponding to different sediment sources of the CL soil

settling process in the cylindrical test tube; all the delay times greater than 1 h determine the same initial
condition for mixing.
Simplified sampling procedure
According to Equation 8, a single suspension sample can be used to calibrate a storage tank. A similar
sampling procedure was tested by Lang (1992) who observed a linear relationship between the bottom
concentration measured by a bottle sampler and the actual concentration.
Two different sampling procedures were used in this work: (1) the calibration curve was established by
relating the actual concentration to the measured one Cb by the tap nearest to the tank bottom only; (2) the
ralationship between C and the measured bottom concentration C,, was determined by a single sample carried
out by a bottle dipped into the tank.

O 1998 John Wiley & Sons, Ltd. Earth Suq. Process. Landforms 23,1151-1170 (1998)
 CALIBRATING STORAGE TANKS

0 10 20 30 40 50
delay time [hours]
Figure 18. Relationship between the measured mean concentration values and the delay time

For testing the first procedure, 10 experimental runs were carried out in the full tank by using the CL soil and
actual concentration values ranging from 2 to 40~[Link] each actual concentration, after mixing, the sampling
from the lowest tap and the up-down procedure were started at the same time. The comparison between the
calibration curves C-C, and C-C, is plotted in Figure 19. Although both calibration curves are linear, as
expected according to Equation 8, a more noticeable scattering characterizes the C-C, relationship. In other
words, the calibration curve C-C, gives an actual concentration estimate which is less accurate than the
calibration curve deduced by using a mean concentration calculated by a measured profile.
The sampling procedure using a bottle was carried out by dipping a bottle having a volume equal to 0-51into
the mixed suspension. The bottle was kept closed until reaching the tank bottom and was then opened to allow
sampling (Figure 20)
Figure 21 shows that the calibration curve can be represented by a straight line, further confirming the
applicability of Equation 8, and that the slope b depends on the water depth in the tank, as established by
Equation 10:

It should be noted that the scattering of the experimental data pairs (C,,, C) around the regression line is
comparable to the scattering ofthe couples (C,, C) around the corresponding calibration curve (Equation 9). In
other words, the accuracy of estimating C by Equation 17 is comparable to that achievable by Equation 9.

CONCLUSIONS
In this paper, an investigation was carried out to examine different factors affecting the manual sampling
procedure for measuring the suspended sediment concentration in a runoff-soil loss storage tank.
At first, using the 11-theoremof the dimensional analysis, the functional relationship describing the mixing
of the suspension into the tank was expressed in a dimensionless form. The recognized dimensionless groups
allowed to establish that for a given soil and water depth in the tank, the relationship between the actual and the
measured concentration (calibration curve) is linear. For a given soil, the slope of the calibration curve is related
to the water depth. The same dimensionless groups allowed scale-up procedure to be deduced and enabled the
use of a small tank, similar to the field one, for investigating additional effects (mathematical shape of the
concentration profile, sampling direction along the vertical, representativeness of a calibration curve, mixing
time, etc.) neglected in the theoretical analysis.
In particular, the effects of the following factors were examined: (1) sampling direction along the vertical; (2)
mathematical shape of the concentration profile; (3) representativeness of the calibration curve; (4) sampling
volume; (5) water depth into the tank; (6) suspension mixing time; (7) grain size distribution of the eroded soil
O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
 V. BAGARELLO AND V. FERRO

0 10 20 30
Cm [a1
Figure 19. Comparison between the C-Cb (a) and C C , (b) relationship for the CL soil

Figure 20. Operating scheme of the bottle dipping into the tank to collect the sample

Figure 21. Calibration curve for the CL soil obtained by using the bottom concentration measured by a bottle dipped into the tank

particles for given soil type; (8) time between the end of the erosive event and the start of the sampling
procedure.
The experimental runs showed that the up-down sampling procedure was characterized by measured
concentration gradients lower than the ones corresponding to the down-up sampling because an interference
between the sedimentation processes during the sampling time and the chosen sampling sequence occurs.

O 1998 John Wiley & Sons, Ltd. Earth [Link]. Landforms 23,1151-1170 (1998)
 CALIBRATINGSTORAGE TANKS 1169

Therefore for establishing the linear calibration curve, a sampling direction has to be selected. The updown
procedure which follows the sedimentation direction and produces a more uniform concentration profile
appears to be preferable.
The analysis also showed that the theoretically based profile (Equation I), obtained by integrating the
differential equation of diffusion, was less adaptable for the higher concentration values measured near the
bottom and its choice did not improve the accuracy of the tank calibration curve, compared to that obtained by
assuming a linear variation between two measurement points.
A calibration curve obtained from a single series of runs adequately described the relationship between the
mean and the actual concentration. In other words, the same operator repeating the same experimental
procedure a given number of times should be able to determine the same calibration curve.
A noticeable influence of the sampling volume on the slope of the calibration curve C-C, was recognized; as
a consequence the sampling volume has to be considered as a constant parameter of the whole calibration
procedure.
Since in field conditions the complete filling of the tank is not always assured, the relationship between the
degree of filling and the slope of the calibration curve was studied. A significant influence of the water depth in
the tank on the calibration curve, as established by theoretical analysis, was also recognised experimentally.
The experimental results showed that the suspension mixing efficiency increases when the mixing time tm
increases from 5 to 10 min while the efficiency can be considered practically invariable when the mixing time
increases up to 15 min. Therefore specific calibration curves have to be deduced for short durations (tm<10 min)
of mixing time.
The dependence of the calibration curve on the characteristics (size distribution, preliminary treatment) of
the sediment entering the tank was examined. In all examined cases the analysis showed that the calibration
curves corresponding to different treatments were not statistically different.
The effect of the time between the end of the erosive event and the start of the sampling procedure (delay
time) was tested. All delay times greater than 1h determined the same initial condition for mixing.
Finally, an attempt to define a simpler sampling procedure was carried out. The calibration curve was first
established by relating the actual concentration to the measured one Cb by the tap nearest to the tank bottom
only. This calibration curve C-Cb was less accurate than the one deduced by using a mean concentration
calculated by a measured profile. A further sequence of runs showed that for improving the accuracy of the
calibration curve based on a single measurement, a different sampling procedure using a dipped sampler should
be used.

ACKNOWLEDGEMENTS

This research is supported by a CEE grant 'Research Project POP - Sicilia' and by grants from Ministero
Universith e Ricerca Scientifica e Tecnologica, Govemo Italiano, quota 40%, and Consiglio Nazionale delle
Ricerche (CNR). The authors are indebted to Drs A. Li Puma and C. Corrao who carried out some experimental
runs for their degree thesis.

REFERENCES
Bagarello, V., Ferro, V., Giordano, G. and Minacapilli, M. 1996. 'Indagini sperimentali sui deflussi e sull'erosione idrica nei piccoli
bacini', Atti della Giornata di Studio POP Sicilia - Cartografia Tematica, Morfologia delle Coste, Erosione del Suolo e Gestione dei
Sedimenti, Palermo, Marzo.
Barenblatt, G. I. 1979. Similarity, Self-similarity and Intermediate Asymptotics, Consultants Bureau, New York.
Barenblatt, G. I. [Link] Analysis, Gordon & Breach, Amsterdam.
Bazzoffi, P. 1993. 'Fagna-type hydrological unit for runoff measurement and sampling in experimental plot trials', Soil Technology, 6,
251-259.
Carter, C. E. and Parsons, D. A. 1967. 'Field tests on the Coshocton-type wheel runoff sampler', Transactions of ASAE, 10(1), 133-135.
Cogo, N. P., Moldenhauer, W. C. and Foster, G. R. 1983. 'Effect of crop residue, tillage-induced roughness, and runoff velocity on size
distribution of eroded soil aggregates', Soil Science Society of America Journal, 47,1005-1008.
Ferro, V. 1997. 'Applying hypothesis of self-similarity for flow-resistance law of small-diameter plastic pipes', Proceedings of ASCE,
Journal of Irrigation and Drainage Engineering, 123(3),175-179.
Graf, W. H. 1984. Hydraulics of Sediment Transport, Water Resource Publications, Littleton, Colorado.

O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)
1170 V. BAGARELLO AND V. FERRO

Hurst, H. E. 1929. 'The suspension of sand in water', Royal Society of London, Series A, 157.
Lang, R. D. 1992. 'Accuracy of two sampling methods used to estimate sediment concentrations in runoff from soil-loss plots', Earth
Surface Processes and Landforms, 17,841-844.
Pierson, F. B. Jr, Van Vactor, S. S., Blacbum, W. H. and Wood, J. C. 1994. Incorporating Small Scale Spatial Variability into Predictions
of Hydrologic Response on Sagebrush Rangelands, Soil Science Society of America, Variability of Rangeland Water Erosion
Processes, SSSA Special Publication 38.
Rouse, H. 1938. 'Experiments on the mechanics of sediment suspension', Proceedings of 5th International Congress on Applied
Mechanics, Cambridge, Mass.
Snedecor, G. W. and Cochran, W. G. 1989. Statistical Methods, 8th edn, Iowa State University Press, Ames, Iowa.
Zobisch, M. A., Klingspor, P. and Oduor, R. A. 1996. 'The accuracy of manual runoff and sediment sampling from erosion plots,' Journal
of Soil and Water Conservation, 51(3), 231-233.

O 1998 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 23,1151-1170 (1998)