Está en la página 1de 8

HYDROLOGICAL PROCESSES, VOL.

10, 1147-1 154 (1996)

ESTIMATION OF TEMPORAL CHANGES IN SOIL MOISTURE


USING RESISTIVITY METHOD
V. C. GOYAL
National Institute of Hydrology, Roorkee-247 667, India
P. K. GUPTA
Earth Sciences, University of Roorkee, Roorkee-247 667, India
S . M. SETH
National Institute of Hydrology, Roorkee-247 667, India
AND
V. N. SINGH
Earth Sciences, University of Roorkee, Roorkee-247 667, India

ABSTRACT
The temporal variation in a soil moisture profile can be studied using resistivity sounding data acquired at different
times. The layered earth model based estimation of soil moisture from apparent resistivity data is a two-step non-linear
inversion. Firstly, the apparent resistivity data are inverted to derive the layer resistivity variations and thicknesses and,
secondly, the moisture content is estimated from these layer resistivity variations using a calibration equation. The soil
moisture-resistivity problem was studied using the one-dimensional formulation of resistivity problem. A generalized
geoelectric earth model was considered to simulate the soil moisture distribution and its temporal variation in the un-
saturated zone. An algorithm (RESMOS) for the interpretation of the apparent resistivity data in terms of soil moisture
variations through this two-step inversion process is reported.

KEY WORDS soil moisture; resistivity interpretation; layered earth model

INTRODUCTION
Obtaining reliable in situ estimates of soil moisture is one of the most difficult problems in field hydrology.
Not only are there considerable difficulties in determining the moisture content at different depths in a soil
profile, but also the temporal variability of soil moisture content makes it difficult to obtain reliable esti-
mates of this variation using a few point measurements.
Hydrologists and soil scientists have traditionally used gravimetric sampling, neutron probes, gypsum
blocks and tensiometers to determine the soil moisture content in the field. These methods have been
used for a long time in spite of the fact that they suffer from certain serious operational limitations. In
view of the limitations of these methods, alternative techniques, e.g. the resistivity method, has been
used in many soil investigations. This method has also been used in soil moisture studies (Dolgov,
1937; Robertson, 1949; Keller and Frischknecht, 1966; Nerpin and Chudvovski, 1970; Gupta and Hanks,
1972; Bohn et al., 1982; Reynolds et al., 1987; Chandrasekharan el al., 1989).
In a few studies, surface resistivity soundings have been applied for soil moisture estimation (Kean et al.,
1987; Frohlich and Parke, 1989; Pane, unpublished data; Adam, unpublished data). A literature survey
revealed that the lack of a resistivity interpretation algorithm considering a continuously varying resistivity
in the subsurface limited the studies to qualitative estimations only.
CCC 0885-6087/96/091147-08 510.00 Received 12 January 1995
0 1996 by John Wiley & Sons, Ltd. Accepted 26 April 1995
1148 V. C. GOYAL ET AL.

This paper presents an algorithm (RESMOS) for the estimation of soil moisture, and its temporal varia-
tion, from apparent resistivity data in the case of a layered earth model with transition layers.

MOISTURE ESTIMATION FROM RESISTIVITY DATA: AN INVERSE PROBLEM


The resistivity of soil in the unsaturated zone is a function of three main parameters, namely moisture con-
tent, salinity and temperature. It would be difficult to determine the contribution of each of these para-
meters on the measured resistivity when their effects are considered simultaneously. The monotonic
reciprocal dependence of soil resistivity (reciprocal of electrical conductivity) on moisture content forms
the basis of estimating the soil moisture content from the apparent resistivity measurements.
In applications where resistivity measurements are used to determine the moisture variations, therefore, it
would be desirable to develop a calibration equation between the resistivity and the moisture content at the
measurement site, with the effect of the other two parameters lumped into the constants of the calibration
equation.
The use of a linear relation between moisture content and resistivity in the present study is guided by (i)
its effective use by several workers (e.g. Gupta and Hanks, 1972), (ii) direct translation of moisture-depth
variation into resistivity-depth variation and (iii) simplification of computations involved in translating the
layer resistivities into moisture contents and vice versa. The linear dependence of resistivity (p) on moisture
content (e),for the depth z and time t , can be written in the form

p ( z , t ) = L2 + b#(Z, t ) (1)
where a and b are empirical constants implicitly containing the soil and water characteristics (:.g. porosity,
temperature, salinity) and are assumed to be invariant with time.
The estimation of the moisture variation in a soil profile from the apparent resistivity measurement is
essentially an inverse problem. The whole exercise may be viewed as a two-step inversion; firstly, the
resistivity variation with depth is determined after interpreting the apparent resistivity data, and then
the moisture content is estimated from this resistivity variation using a moisture-resistivity calibration
equation.
The moisture profile in the unsaturated zone can be represented as a one-dimensional model in situations
where the movement of infiltrated water is dominantly vertical. Such a continuous profile can be analysed
using a stratified earth model, with different layers corresponding to different continuous segments of the
moisture profile. The soil moisture variation with depth within any layer can be approximated by a con-
stant, power law (including linear, quadratic) or exponential function. This would, in turn, require that
the resulting resistivity variation is incorporated into the apparent resistivity data inversion algorithm.

TEMPORAL VARIATIONS IN SOIL MOISTURE PROFILE


The linear dependence of resistivity on moisture content given by Equation (1) can be restated, after drop-
ping the arguments z and t for simplification of notations, for observations at different times as
p!) = q + b, &i),
i = 1,2, ...,nt
j = 1,2, ...,nt
where nt is the number of apparent resistivity data sets acquired at different time instants, nl is the number
of layers, subscript j denotes a particular layer and the superscript i denotes the particular data set corre-
sponding to a given time instant.
Equation (2) can be used in three situations: (i) when estimates of the calibration constants (a and b) are
available; (ii) when estimates of the calibration constants are not available, but moisture content profiles
corresponding to two apparent resistivity data sets are available; and (iii) when neither of the above two
sets of information (i and ii) are available, In the first case, Equation (2) can be used directly. In the second
case, assuming that the available moisture content profiles correspond to the first two apparent resistivity
TEMPORAL CHANGES IN SOIL MOISTURE 1149

data sets, the moisture content in thejth layer for the ith data set can be computed, after eliminating the
constants a and b from Equation (2) as

In the last case, when the moisture content profiles are not available to estimate the calibration constants
a and b, scaled layer resistivity values can be used to estimate scaled relative change in moisture content.
The scaled estimate (r$& for the ith data set can be computed from Equation (2) as

The problem of monitoring the temporal variation of moisture content is thus reduced to the determination
of layer resistivity variation from the apparent resistivity data sets observed at different times. Periodic ac-
quisition of resistivity data and use of this interpretation procedure would provide the temporal variation of
moisture content in the soil profile.

RESISTIVITY INVERSION
The resistivity inversion algorithm RESMOS developed for inversion of the apparent resistivity data uses
the Marquardt's ridge-regression operator (Marquardt, 1963). The iterative indirect scheme works with a
current guess earth model and compares its computed response with that of the true model, in the case of
synthetic study, or with the field sounding data, in the case of a field study. It estimates, on the basis of the
difference between the two responses, the parameter correction vector using Marquardt's algorithm. The

Table I. Percentage error in computing the moisture content from 2% noise added two-
electrode array apparent resistivity data of the desaturating model SYNl
~ ~~ ~

No. Depth (m) Data set

1 2 3 4 5 6
Observed moisture content (Vol. Yo)
1 0.40 45.00 40.00 34.00 30.00 27.00 20.00
2 1.oo 43.16 37.81 32.25 27.76 24.28 18.65
3 1.50 41.11 35.62 30.65 25.70 21.77 17.48
4 2.00 38.45 33.04 28.90 23.46 19.02 16.26
5 2.50 35.00 30.00 27.00 21.00 16.00 15.00
6 3.00 39.60 34.13 30.05 23.55 21.51 23.61
7 3.50 42.79 37.40 32.70 25.88 26.13 30.10
8 3.75 44.00 38.78 33.89 26.96 28.15 32.72
9 4.00 45.00 40.00 35.00 28.00 30.00 35.00
Percentage error in moisture content
1 0.40 0.18 0.59 1.29 2.07 2.22 2.50
2 1.oo 0.08 0.27 0.52 0.83 1a00 0.97
3 1.50 -0.08 -0.18 -0.36 -0.67 -0.61 -0.63
4 2.00 -0.39 -0.89 -1.55 -2.83 -3.10 -2.63
5 2.50 -0.94 -2.03 -3.15 -5.95 -7.13 -5.13
6 3-00 -0.32 -0.83 -1.32 -2.87 -2.61 -1.38
7 3.50 -0.05 -0.17 -0.17 -0.78 -0.55 -0.19
8 3.75 0.03 0.03 0.24 0.02 0.06 0.10
9 4.00 0.08 0.18 0.57 0.68 0.50 0.29
1150 V. C. GOYAL ET AL.

Table 11. Percentage error in computing the moisture contents from 2% noise added two-
electrode array apparent resistivity data of the saturating model SYN2

NO. Depth (m) Data set


1 2 3 4 5 6
Observed moisture content (Vol. %)
1 0.40 20-00 22.00 26.00 32.00 40.00 4540
2 1-00 18.65 20.40 24.42 29.79 37.47 43.16
3 1.so 17.48 19.00 23.02 27,74 34.89 41.11
4 2.00 16.26 17.54 21.55 25.48 31.77 38.45
5 2.50 15.00 16.00 20.00 23.00 28.00 35-00
6 3-00 20.96 21.99 24.70 27.80 32.02 39-60
7 3.50 25.90 26.93 28.66 31.75 35.31 42.79
8 3.75 28.04 29.06 30.40 33.46 36.72 44.00
9 400 30.00 31.00 32.00 35.00 38.00 45.00
Percentage error in moisture contact
1 0.40 3.65 3.00 2.31 1.34 0.66 0.18
2 1.oo 1*47 1.22 0.91 0.57 0.30 0.08
3 15 0 -0.84 -0.72 -0.60 -0.38 -0.24 -0.08
4 2.00 -3.71 -3.23 -2.53 -1.73 -1.17 -0.39
5 2.50 -7.33 -6.50 -5.00 -3.70 -2.79 -0.94
6 3.00 -2.53 -2.26 -1.99 -1.47 -1.18 -0.32
7 3.50 -0.49 -0.44 -0.43 -0.31 -0.28 -0.05
8 3.75 0.09 0.08 0.07 0.05 0.02 0.03
9 4.00 0.50 0.45 0.44 0.3 1 0.24 0.08

parameter correction vector is then used to update the guess model for next iteration. The iterative proce-
dure ends when either the convergence is achieved or the number of iterations exceeds a given limit (Goyal,
unpublished data).

RESULTS AND DISCUSSION


The subsurface variation of moisture content can be broadly studied by simulating the saturating and de-
saturating modes of the moisture profiles. Two moisture profiles were synthesized by assigning plausible
realistic values of the moisture content up to a depth of 4 m. By increasing/decreasingthe moisture content
in the two original profiles, five profiles each were additionally generated. The set of six profiles where the
moisture content decreases in the successive profiles was termed as the ‘desaturating model’ (SYNI),
whereas the other set with increasing moisture content was termed as the ‘saturating model’ (SYN2). Dif-
ferent rates of change of moisture content were used at different depths in both the models.
Based on the inflexion points, these two sets of moisture content profiles were then approximated by two
sets of four-layer resistivity models, with the layer interfaces in both the models being at the depths of 0.40,
2-50 and 4.00 m. The two intermediate layers in these models were assumed to be the transition layers with
an exponential resistivity variation. The following calibration equation (Goyal, unpublished data) was used
to convert the moisture content values (6) into the corresponding resistivity values (p)
p = 500 - 10 6 (5)
A value of 2% random noise was added to the two-electrode array apparent resistivity response of each
model. These noise-added responses were inverted for the layer resistivities using the algorithm RESMOS.
These computed resistivity values were then used, in conjunction with the calibration equation, to obtain
the computed moisture content profiles.
The inversion results for different profiles in the two sets are summarized in Tables 1 and 11. In these
TEMPORAL CHANGES IN SOIL MOISTURE 1151

Table 111. Sensitivity of apparent resistivity response to dif-


ferent percentage changes in moisture content
Change in RMS change in apparent resistivity (%)
0 (Yo)
Saturating model Desaturating model
1 0.2 3.1
2 0.5 6.2
5 1.2 15.2
10 2.5 29.6

profiles, the maximum error between the observed and the computed moisture contents is of the order of
7%, whereas the average error was less then 1%.
Sensitivity of apparent resistivity to changes in moisture content
To study the sensitivity of apparent resistivity to changes in moisture content, various percentage changes
were made in moisture profiles of the two sets SYNl and SYN2. The resulting apparent resistivity curves
were computed and compared with the original curves in terms of the per cent RMS changes using the
equation

here the superscript k is the model number, subscript i is the data point and nd and nm are the number of
data points and number of models, respectively.
The changes in the apparent resistivity of the first profile of each set for 1, 2, 5 and 10% changes in the
moisture content are presented in Table 111. As seen in the table, the same percentage change in the moist-

0.9-

1.88-

1.50-

2.66-

v
E
2.50-
5
a
0 3.88-

Figure 1. Variation of O,(vol. %) and 4pwith depth in a synthetic desaturating model. Exponential depth variation considered for p in
different layers. po computed for two-electrode array, with 2% random noise added at all electrode spacings
1152 V. C. GOYAL ET AL.

0.mj
I ,
'
?
'
* V
' '
10

'
n
I
rn
I
. .
I 1

0.58 -
1.m-

1.58-

2.68-
n

v
E
2.58-
.c
c
a
0
0 3.68-

-,

4.86 l!
Figure 2. Variation of O,(vol. %) and $ J ~with depth in a synthetic saturating model. Exponential depth variation considered for p in
different layers. pa computed for two-electrode array, with 2% random noise added at all electrode spacings

ure content caused differentpercentage RMS changes in the response of the two models. This suggests that
the sensitivity of apparent resistivity is higher for low moisture content profiles. However, this study sug-
gests that even a small change in moisture content causes a perceptible change in the apparent resistivity
curve, which can, in turn, be utilized to infer the moisture content profile.

Temporal variation in absence of a calibration equation


The developed algorithm was used to monitor temporal variation in soil moisture even when the cali-
bration equation is not available. For this purpose, the two sets of models SYNl and SYN2 were used
again. Using the linear relation between moisture content and resistivity given by Equation ( 5 ) and the
defining Equation (4), the scaled moisture parameter q5p was computed for the six profiles each of SYNl
and SYN2. The variations of the observed moisture content (0) and c $ ~ with the depth for selected pro-
files of these sets are plotted in Figures 1 and 2. It is observed that the q5p value follows the same trend as
that of 8, although in the opposite direction. This implies that the movement of moisture with depth can
be monitored using resistivity data alone, even in the absence of a calibration equation. In this case,
however, moisture movement corresponding to the third and subsequent resistivity soundings only
can be studied.

Validation with field apparent resistivity data


Resistivity sounding data for four different dates at a site in Western Uttar Pradesh (India) were utilized
to ascertain the viability of the developed algorithm. Moisture content profiles for the same dates were also
available (Goyal, unpublished data).
The regression analysis between resistivity and soil moisture values of each data set yielded high values of
the coefficient of correlation ( ~ 0 . 9 9 and
) the goodness of fit (-0.98). In general, the fit is of high quality
(Figure 3). The error > 10% was observed only at 10% points. At all other points, the error was much
TEMPORAL CHANGES IN SOIL MOISTURE 1153

Moisture Content (s mi)


19.0 16.1 a.l X1 IIB.0 1 . 1 1 1 .1
1 6.1 68.1
l.m I ~ ~ " ' " ' ' ~ ' ' ~ I ~~ ~ "~ ' ~' *' ~ ~ ' 1' " ' ~I ' ~

3
1.z-
20.1285
L6m-

*.=- m.76-

!.a- 1.m-

!.a- 1.a-
h

1.m- v
E 1.m-
1.6s

fa 5(1
1.X- a 1.75j
1.75-
n n
2.-

ze 1

1.76-

1.68-

CI
b.76-

1.-

1.S
h
.-I
1.m

E
Y 1.e- E
v

f f
z
n
I.%- 3
0

zm

za

Figure 3. Observed and computed (from pa data) soil moisture contents for a site in Western Uttar Pradesh (India). Calibration equa-
tion used p = 228.95 - 4.550

less, mostly less than 1%. The linear relationship between resistivity and moisture content was found to be
valid for moisture contents up to 40-42vol. %, which was the saturation limit of the soil.

CONCLUSIONS
A major lacuna in the use of the resistivity method for soil moisture studies has been the lack of quantitative
interpretation of acquired data. The developed algorithm is the first resistivity interpretation algorithm based
on generalized layered earth geoelectric model comprising the combination of transition and uniform layers.
Moisture content estimates can be obtained from the apparent resistivity data using a calibration equation
1154 V. C. GOYAL ET AL.

which provides a pointwise relationship between the moisture content and true resistivity value in the soil
profile. The accuracy of moisture estimation from apparent resistivity data is highly dependent, as in other
field methods, on the reliability of the calibration equation used.
The temporal variation in a soil moisture profile can be studied using resistivity sounding data acquired
at different times. The temporal variation in soil moisture can be directly estimated when either the calibra-
tion equation itself or at least two moisture content profiles corresponding to any two apparent resistivity
data sets are available. If neither of these two pieces of informations is available, then the trend in the tem-
poral variation of soil moisture can be estimated from the third and subsequent apparent resistivity data
sets. This feature may prove to be of immense use, for example, in monitoring wetting front movement.

ACKNOWLEDGEMENTS

V.C. Goyal is grateful to the Director, National Institute of Hydrology, Roorkee, for allowing publication
of this paper.

REFERENCES
Bohn, H. L., Ben-Asher, J., Tabbara, H. S.,and Manvan, M.1982. ‘Theories and tests of electrical conductivity in soils’, Soil Sci. SOC.
Am. J., 46, 1143-1146.
Chandrasekharan, H., Yadav, B. R., Tyagi, S. K., and Sarma, P. B. S. 1989. ‘Soil moisture studies in soils of Delhi territory using the
resistivity meter’ in Proc. Ini. Workshop on Appropriate Methodologiesfor Development and Management of Groundwater Resources
in Developing Countries. NGRI, Hyderabad, 385-393.
Dolgov. S. I. 1937. Trudy Vseso.yuznogo- Nauchno-lssledovatel’skogoInsrituta Udobrenii. Agrotekhniki i Agropochvovedeniya,
- _ No. 18,
I&. VASKhNIL.
Frohlich. R. K. and Parke. C. D. 1989. ‘The electrical resistivitv of the vadose zone - field survev.’ Ground Water. 27. 524-530.
Gupta, S. C. and Hanks, R. J. 1972. Influence of water content i n electrical conductivity of the soil‘,-&oifSci. SUC.Am. Pioc., 36,855-
857.
Kean, W. F., Waller, M. J., and Layson, H. R. 1987. ‘Monitoring moisture migration in the vadose zone with resistivity’, Ground
Water, 25, 562-271.
Keller, G. V . and Frischknecht, F. L. 1966. Electrical Method in Geophysical Prospecting. Pergamon Press, New York.
Marquardt, D. W. 1963. ‘An algorithm for least-squares estimation of nonlinear parameters’, SOC.Ind. Appl. Math., 11,431-441.
Nerpin, S.V. and Chudvovski, A. F. 1970. Physics ofrhe Soil. [Translated from Russian by Israeli Program of Scientific Translation,
Jerusalem.]
Reynolds, T. D., Shepard, R. B., Laundre, J. W., and Winter, C. L. 1987. ‘Calibrating resistance type soil moisture units in a high clay
content soil’, Soil Sci., 144.
Robertson, E. I. 1949. ‘Measurement of soil moisture at Winchmore by resistivity methods’, Report No. 94, Geophysics Division, Dept.
of Scientific and Industrial Research, New Zealand.

También podría gustarte