You are on page 1of 3

VOL. 15, NO.

6 WATER RESOURCES RESEARCH DECEMBER 1979


Estimating Soil Water Retention Characteristics From Particle Size
Distribution, Organic Matter Percent, and Bulk Density
S.C. OUPTA AND W. E. LARSON
Science and Education Administration, Agricultural Research, U.S. Department of Agriculture
Soil and Water Management Research Unit, University of Minnesota, St. Paul, Minnesota 55108
Regression models are presented for estimating soil-water-retention curves from particle-size distribu-
tion, percentage of organic matter, and bulk density. Models were developed from the measured soil-wa-
ter-retention curves of artificially packed cores (7.6 x 7.6 cm) of 43 soil materials. These soil materials
included 13 agricultural soils. Curves predicted with these models approximated reasonably well the
measured water retention of 61 Missouri soils. Because conventional methods of obtaining retention
curves are expensive and time consuming, these equations will be valuable for modeling salt and water
ttow in soils and for estimating available water capacities.
INTRODUCTION
Soil water retention characteristics are needed to describe
the availability of soil water to plants and to model movement
of water and solutes in unsaturated soils. Measuring soft-wa-
ter-characteristic curves is expensive and time consuming.
Several efforts have been made to predict water retention
from easily and routinely obtainable textural and structural
soil properties. Soil water content at the -15-bar matric po-
tential has been successfully related to clay [Lund, 1959; Kivi-
saari, 1971] and organic matter [Crupta et al., 1977] contents of
the soil. In other cases [Lund, 1959; $alter et al., 1966; Petersen
et al., 1968], the difference in water retention corresponding to
two matric potentials (available water capacity) was predicted
from textural and structural properties. Using particle-size
analysis, Husz [1967] developed relationships that describe
water retention characteristics between-0.33 and- 15 bar. In
this paper, statistical relationships are presented that can be
used to predict soil water retention over a wide range (-0.04
to -15 bar) of soil matric potentials. These relationships are
based on percent sand, silt, clay, organic matter, and bulk
density (pb).
MATERIALS AND METHODS
The materials used in this study included a dredged sedi-
ment (DS) and a soil sample from each of 10 geographic loca-
tions in eastern and central United States. These samples were
air dried on a plastic sheet and ground in a hammer mill to
pass a 2.0-mm screen. Each sample was then mixed in a ce-
ment mixer to obtain a homogenous sample. Air dried
dredged sediment and soil samples from each location were
further mixed in various proportions for laboratory and
greenhouse plant growth studies [Crupta et al., 1978]. The
treatments were (1) dredged sediment, (2) 2/3 dredged sedi-
ment to 1/3 soft, (3) 1/3 dredged sediment to 2/3 soft, and (4)
soft.
In addition, three productive softs from Minnesota were
also included as controls for the greenhouse study. Each
sample is described by Cruta et al. [1978].
Particle-size distribution was determined by direct sieving
for the sand fraction and by the international pipette method
for the silt and clay fractions [Day, 1965]. Organic matter was
estimated from the carbon content of the samples, as deter-
This paper is not subject to U.S. copyright. Published in 1979 by
the American Geophysical Union.
Paper number 9W 1249.
mined by a Leco carbon analyzer. (The mention of a com-
mercial product does not constitute endorsement by the U.S.
Department of Agriculture.) Water retentions were deter-
mined in the laboratory on artificially prepared cores.
Loosely packed, air dried samples were saturated overnight
on the pressure plate and desorbed to -15-bar matric poten-
tial in the pressure plate apparatus [Richards, 1965]. Three
cylinders (7.6-cm diameter x 7.6 cm high) were taped to-
gether, and a known weight of the soil at -15 bar water con-
tent was compressed at both ends until the desired bulk den-
sities were obtained. These bulk densities corresponded to the
air dried bulk densities obtained in a greenhouse experiment
[Gupta et al., 1978; Crupta and Larson, 1979]. Soil was com-
pressed at a rate of 1.0 cm/min with an Instron universal test-
ing instrument. The middle core was used for water retention
determinations. Bulk density of the middle core was deter-
mined from the weight and volume of the soil core. Variation
in the packing of each soil core was checked with a gamma
probe at 1-cm-depth intervals. Cores showing significant de-
viation between their average reading and readings with
length were discarded.
Soil water retention values were obtained between -0.02
and -15.0 bar with a pressure plate apparatus [Richards, 1965]
by desorbing the saturated cores at several pressure steps. The
cores were initially saturated overnight from the bottom under
a small head of water. At least three cores were used to obtain
water retention values over the entire range studied; i.e.,
-0.02 to -2 bar (millipore filter), -0.1 to -1.0 bar (1-bar ce-
ramic plate), and -3.0 to -15.0 bar (15-bar ceramic plate).
Water content at each pressure step was calculated from the
volume of outflow between pressure steps, the final water con-
tent, and the weight of oven-dried soil.
The cubic spline method was used to smooth the experi-
mental data points. Since experimental data were not deter-
mined at exactly the same pressures for all samples, water
content values used in the regression analyses (equation (1))
were taken from the smoothed curve:
0, = a x sand (%) + b x silt (%) + c x clay (%) + d x
organic matter (%) + e x bulk density (g/cm 3)
where 0 s is the predicted water content (cm3/cm 3) for a given
matric potential and a, b, c, d, and e are regression coeffi-
cients.
1633
1634 GUPTA AND LARSON: SOIL WATER RETENTION CHARACTERISTICS
]o0 i ! i i i i i i i , i [ : iil i i i i i iii i i i i i i r
.8
z
i-
z
,2
SOUTH CAROLI NA
2/3DS + ]./3 SOIL
.'.__ , . "' 1/3DS + 2/3 SOl L
.' ' ' 1.32 . - SOIL
I I 1 I IIII I I I I [ ill 1 I I I I IIIII I I I I I I I 01 -, 1 - - 10 - 100
MATRI C POTENTIAL , BARS
Fig. 1. Soil water retention characteristics of dredged sediment (DS), soil, and their mixtures from South Carolina.
Smooth curves were fitted using the cubic spline method.
TABLE 1. Regression and Correlation Coefficients for Prediction of Soil Water Content at Specific
Matric Potentials
Matric Regression Coefficients Correlation
Potential, Coefficient,
bars ax 103 bx 103 cx 103 dx 103 ex 102 R
-0.04 7.053 10.242 10.070 6.333 -32.120 0.950
-0.07 5.678 9.228 9.135 6.103 -26.960 0.959
-0.10 5.018 8.548 8.833 4.966 -24.230 0.961
-0.20 3.890 7.066 8.408 2.817 -18.780 0.962
-0.33 3.075 5.886 8.039 2.208 -14.340 0.962
-0.60 2.181 4.557 7.557 2.191 -9.276 0.964
- 1.0 1.563 3.620 7.154 2.388 - 5.759 0.966
-2.0 0.932 2.643 6.636 2.717 -2.214 0.967
-4.0 0.483 1.943 6.128 2.925 -0.204 0.962
- 7.0 0.214 1.538 5.908 2.855 1.530 0.954
- 10.0 0.076 1.334 5.802 2.653 2.145 0.951
- 15.0 -0.059 1.142 5.766 2.228 2.671 0.947
Sand (%) + silt (%) + clay (%) = 100. Sand = 2.0 - 0.05 mm. Silt = 0.05 - 0.002 mm. Clay = < 0.002
mm.
RESULTS AND DISCUSSION
Basic data used in developing these regression equations in-
cluded soils with a wide range of sand (5-98%), silt (1-72%),
clay (0-65%), organic matter (0-23%), and bulk density (0.74-
AAATRIC POTENTIAL R /
TBAR- -- -
8 o. o4 0.00
o O. 33 O. 962 . /o
7 o 7.0 O. 954 ..
ep - a SAND + b SILT + c CLAY + d OM + e BD, /.
6 #.. .
.
z .5 :$ z
I 3 Oo ,,,p o
o
ee %o
.1
0 i i i i i i i i
0 .1 .2 .3 .4 .5 .6 .7 .8 .9
MEASURED WATER CONTENT O m, cm3/cm3
Fig. 2. Predicted versus measured water content at -0.04, -0.33,
and -7.0 bar matric potential.
1.74 g/cm 3) values. Soil water contents measured between
-0.02 and -15 bar by volume ranged from 0.03 to 0.82 cm3/
cm 3. These ranges include most agricultural soils. The clay
fraction of these sediments and soils included both expanding
(montmorillonite) and nonexpanding (kaolinite, i!!ite, chlo-
rite, and vermiculite) type clay minerals. Surface area mea-
surements and X ray analysis indicated the presence of
amorphous clay-sized materials in some sediment samples.
An example of the shape of the water retention curves and
scatter of the experimental data for dredged sediment, soil,
and their mixture from South Carolina is shown in Figure 1.
Table 1 gives the regression and correlation coefficients be-
tween soil water retention and five soil properties at 12 de-
TABLE 2. Regression Analysis (y = a I -- fll x) of Predicted (y) and
Measured (x) Water Contents at Four Mattic Potentials for 61 Mis-
souri Soils
Matrix
Potential,
bars a I +_ se* [1 +-- se* s t
-0.1 0.0494 +_ 0.0244 0.9934 +_ 0.0677 0.0604
-0.33 0.0456 +_ 0.0151 0.9489 +_ 0.0456 0.0434
-1.0 0.0478 + 0.0116 0.9173 + 0.0386 0.0364
-15.0 0.0555 + 0.0090 0.9336 + 0.0412 0.0346
*Standard error.
Standard error of the regression.
GUPTA AND LARSON: SOIL WATER RETENTION CHARACTERISTICS 1635
' I I i i Illl[ [ I I I Ill[I ] [ [ [ [ Ill[ [ I I I 11[[I
o
-O.Ol
SHARKEY 1 ICLAY)
( Vertic Haplaquepts)
RICHLAND B 12 (SILT LOAM,)
(Typic Hapludalfs)
SHELBY All (SANDY CLAY LOAM,)
(Typic Argiudolls)
FREELAND B1 (SANDY LOAM,)
( G lossic F ragiudal fs )
BEULAH Cl (LOAM,Y SAND)
(Typ/c Dystrochrepts)
i I I lllll i I I lilii i I I IIIIII I I I IIIIII
-. 1 -1 -10 -100
M,ATRIC POTENTIAL, BARS '
Fig. 3. Comparison between the measured (points) and predicted (smooth curve) soil water retention characteristic
curves of five Missouri soils.
creasing matric potentials. At all matric potentials, correlation
coefficients were 0.94 or greater. High correlation coefficients
obtained for these regression equations are reflected in the
limited data scatter between the predicted and measured wa-
ter content values at three typical matric potentials (Figure 2).
The regression models in Table I were tested on data from
61 Missouri soils [Jamison and Kroth, 1958; Kroth et al., 1960].
Regression analysis (Table 2) suggests intercept (a) to be sig-
nificantly different than zero and the slope of the line (fi) not
significantly different than 1.0 at all matric potentials. The
presence of a nonzero intercept and a slope of 1.0 suggests a
constant bias between the predicted and measured water con-
tents. This bias could be due to the differences in the experi-
mental procedures used by Kroth et al. [1960] and by us in the
present study: (1) 1- to 3-cm-thick versus 7.6-cm-thick cores
and (2) estimation of particle-size distribution by the hydrom-
eter method as compared with the pipette method (present
study). Low values for measured water retention by Kroth et
al. [1960] may also have been caused by the presence of coat-
set (>2 ram) fragments in the undisturbed core samples. Fig-
ure 3 shows the difference between the predicted and mea-
sured water retention characteristics of five Missouri softs of
different texture. The agreement between the measured and
predicted value is good.
Regression models presented here may be used to estimate
with reasonable accuracy water retention characteristics from
particle-size distribution, percentage of organic matter, and
bulk density. This will be of particular help in modeling salt
and water flow in softs and in estimating available water ca-
pacities. Water retention curves obtained from these regres-
sion equations may also be used to approximate hydraulic
conductivity-water content relationships [Green and Corey,
1971; Campbell, 1974].
Acknowledgement. Appreciation is expressed for financial support
of this research in part by the U.S. Army Corps of Engineers.
Day, P. R., Particle fractionation and particle-size analysis, in Meth-
ods of Soil Analysis Part I, Agronomy, vol. 9, edited by C. A. Black
et al., pp. 545-567, American Society of Agronomy, Madison, Wis.,
1965.
Green, R. E., and J. C. Corey, Calculation of hydraulic conductivities:
A further evaluation of some predictive models, Soil Sci. Soc.
Amer. J., 35, 3-8, 1971.
Gupta, S.C., and W. E. Larson, A model for predicting packing of
soils from particle size distribution, Soil Sci. Soc. Amer. J., 32, 758-
764, 1979.
Gupta, S.C., R. H. Dowdy, and W. E. Larson, Hydraulic and thermal
properties of a sandy soil as influenced by incorporation of sewage
sludge, Soil Sci. Soc. Amer. J., 41, 601-605, 1977.
Gupta, S.C., W. E. Larson, R. G. Gast, S. M. Combs, and R. H.
Dowdy, The agricultural value of dredged material, Tech. Rep., D-
78-36, U.S. Army Eng. Waterways Exp. Sta., Vicksburg, Miss.,
1978.
Husz, G., Determination of the pF curve from texture, using multiple
regressions, Z. Pfianzenernaehr. Bodenk., 116, 115-125, 1967.
Jamison, V. C., and E. M. Kroth, Available moisture storage capacity
in relation to textural composition and organic matter content of
several Missouri soils, Soil Sci. Soc. Amer. Proc., 22, 189-192, 1958.
Kivisaari, S., Influence of texture on some moisture constants, Acta
Agr. Femica, 123, 217-222, 1971.
Kroth, E. M., V. C. Jamison, and H. E. Grogger, Soil moisture survey
of some representative Missouri soil types, ARS-41-34, pp. 8, 9, 11,
12, 14, 27, 29, 35-37, 43, 47, 48, 52, 56, U.S. Dep. of Agr., Wash-
ington, D.C., 1960.
Lund, Z. F., Available water-holding capacity of alluvial soils in Lou-
isiana, Soil Sci. Soc. Amer. Proc., 23, 1-3, 1959.
Petersen, G. W., R. L. Cunningham, and R. P. Matelski, Moisture
characteristics of Pennsylvania soils, II, Soil factors affecting mois-
ture retention within a textural classysilt loam, Soil Sci. Soc. Amer.
Proc., 32, 866-870, 1968.
Richards, L. A., Physical condition of water in soil, in Method of Soil
Analysis Part 1; Agronomy, vol. 9, edited by C. A. Black et al., pp.
128-151, American Society of Agronomy, Madison, Wis., 1965.
Salter, P. J., G. Berry, and J. B. Williams, The influence of texture on
the moisture characteristics of soils, III, Quantitative relationships
between particle size, compaction and available water capacity, J.
Soil Sci., 17, 93-98, 1966.
REFERENCES
Campbell, G. S., A simple method for determining unsaturated con-
ductivity from moisture retention data, Soil Sci., 117, 311-314,
1974.
(Received April 25, 1979;
revised August 13, 1979
accepted August 23, 1979.)

You might also like