431

Chapter 22

PEDOTRANSFER FUNCTIONS FOR EUROPE

J.H.M. Wösten1,* and A. Nemes2

1
Alterra, Droevendaalsesteeg 3, 6700 AA Wageningen, The Netherlands
2
USDA-ARS Hydrology & Remote Sensing Lab, Bldg. 007, Rm. 104, BARC-W,
Beltsville, MD 20705-2350, USA
p
Corresponding author: Tel. þ 31-317-474287

For the development of pedotransfer functions (PTFs) large, good quality data sets are
required comprising measured hydraulic characteristics of a wide variety of soils. An
example of such a large data set is the database of hydraulic properties of European soils
(HYPRES). The HYPRES database contains information on a total of 5521 soil horizons
(Wösten et al., 1999). Of these, 4030 horizons have soil hydraulic data that can be used for
the development of PTFs. The soil information is donated by 20 institutions from 12
European countries.
Problems in constructing this large international database were two-fold: (i) various
countries use different soil classification systems with the consequence that soil texture
classes have different meanings in different countries, and (ii) the number of measured
individual points along the various hydraulic characteristics varies considerably due to the
application of different measurement techniques. Both problems had to be resolved to
arrive at an international database that holds compatible, good quality data. It was decided
to adhere to the FAO (1990) and the USDA (1951) particle-size class intervals. As a
consequence, a new interpolation technique had to be developed to get an accurate
estimation of missing particle-size fractions (Nemes et al., 1999). This interpolation
technique was used to arrive at a standardized database. Even more important was the
marked imbalance in the number of measured soil hydraulic data pairs for the different
soil horizons due to use of different measurement techniques by the different
national institutions. To avoid statistical bias, this imbalance was eliminated by approxi-
mating individual hydraulic characteristics with van Genuchten – Mualem equations
(van Genuchten et al., 1991).

us 2 u r
uðhÞ ¼ ur þ ð1Þ
ð1 þ lahln Þ121=n

ðð1 þ lahln Þ121=n 2 lahln21 Þ2

KðhÞ ¼ Ks ð2Þ
ð1 þ lahln Þð121=nÞðLþ2Þ

As a result the measured hydraulic characteristics became available via optimised

model parameters (i.e., ur, us, Ks, a, L and n).
DEVELOPMENTS IN SOIL SCIENCE q 2004 Elsevier B.V.
VOLUME 30 ISSN 0166–2481/DOI 10.1016/S0166-2481(04)30022-X All rights reserved.
432

PTFs for the different texture classes were derived by firstly using the optimized
parameters to determine water contents and hydraulic conductivities at 13 pressure heads:
0, 1, 2, 3, 5, 10, 25, 50, 100, 250, 500, 1000, 1600 kPa. Because the water retention ðuðhÞÞ
and hydraulic conductivity ðKðhÞÞ values are log-normally distributed in nature, the
geometric mean water contents and hydraulic conductivities at the 13 pressure heads were
calculated. In addition to the geometric mean values, also the u and K values plus and
minus one standard deviation were calculated. These standard deviations gave an
indication of the degree of variation of the individual curves around the geometric mean
curve. Next Equations (1) and (2) were fitted to the geometric mean values at the 13
pressure heads to arrive at optimized model parameters for the mean characteristics. Since
these parameters represent the mean hydraulic characteristics for a soil texture class, they
are called class PTFs. In total, class PTFs for 11 texture classes have been established. The
11 texture classes consist of five FAO texture classes, each subdivided into topsoil and
subsoil subclasses, plus the FAO class of organic soils used to prepare the European soil
map on a scale 1:1,000,000. Figure 1 shows the calculated geometric mean water retention
and hydraulic conductivity characteristic and the standard deviations for the texture class
“Medium Fine Topsoils.” The figure shows the characteristic shape of the hydraulic
characteristics to be expected for this texture class. At the same time, the standard
deviations demonstrate that there is considerable variability in hydraulic characteristics
within the texture class. Table 1 gives the Mualem – van Genuchten parameters of the
geometric-averaged curves for each texture class of the HYPRES database.

Figure 1. Geometric mean water retention (left graph) and hydraulic conductivity (right
graph) characteristic (solid lines), standard deviations (bars) and Mualem – van Genuchten
fits (dotted lines) for the texture class “Medium Fine Topsoil” (after Wösten et al., 1999).

In addition to the development of class PTFs, linear regression was used to investigate
the dependency of each model parameter on more easily measured, basic soil properties.
To comply with a number of physical boundary conditions, transformed parameters rather
 433

Table 1
Mualem – van Genuchten parameters for geometric averaged water retention and hydraulic
conductivity curves of the HYPRES database
ur us a n m L Ks
Topsoils Coarse 0.025 0.403 0.0383 1.3774 0.2740 1.2500 60.000
Medium 0.010 0.439 0.0314 1.1804 0.1528 2 2.3421 12.061
Medium fine 0.010 0.430 0.0083 1.2539 0.2025 2 0.5884 2.272
Fine 0.010 0.520 0.0367 1.1012 0.0919 2 1.9772 24.800
Very fine 0.010 0.614 0.0265 1.1033 0.0936 2.5000 15.000
Subsoils Coarse 0.025 0.366 0.0430 1.5206 0.3424 1.2500 70.000
Medium 0.010 0.392 0.0249 1.1689 0.1445 2 0.7437 10.755
Medium fine 0.010 0.412 0.0082 1.2179 0.1789 0.5000 4.000
Fine 0.010 0.481 0.0198 1.0861 0.0793 2 3.7124 8.500
Very fine 0.010 0.538 0.0168 1.0730 0.0680 0.0001 8.235
Organica 0.010 0.766 0.0130 1.2039 0.1694 0.4000 8.000
a
Within the organic soils no distinction is made in topsoils and subsoils.

than the original model parameters were used in the regression analysis. In this case, the
imposed boundary conditions were: Ks . 0; a . 0; n . 1 and 210 , L , þ10 (Wösten
et al., 1999). The following basic soil properties were used as input variables: percentage
clay, percentage silt, percentage organic matter; bulk density and also the qualitative
variable topsoil or subsoil. The latter variable was entered in the regression equation by
assigning the value 0 or 1. Linear, reciprocal, and exponential relationships of these basic
soil properties were used in the regression analysis, and possible interactions were also
investigated. As a consequence, the resulting regression model consists of various basic
soil properties and the product of different basic soil properties, all of which contribute
significantly to the description of the transformed model parameters. The models were
selected using a subset selection method (Furnival and Wilson, 1974). Since these PTFs
require basic soil properties for a specific point in the field instead of class average
texture data, they are called continuous PTFs (Vereecken et al., 1992; Tietje and
Tapkenhinrichs, 1993). The continuous PTFs developed using the HYPRES database, are
presented in Table 2.
While class PTFs predict the hydraulic characteristics for rather broadly defined soil
texture classes, and therefore cannot provide site-specific information, continuous PTFs
can be applied in case of more site-specific applications. However, the R2 values obtained
indicate that the predictions of the hydraulic characteristics when using continuous PTFs
are fairly inaccurate (Table 2). Subdividing the complete dataset into subsets of similar
soil texture might improve these predictions.
The class PTFs for the 11 FAO texture classes are used to translate the representative
profiles for the mapping units of the European soil map at a scale of 1:1,000,000 into soil
hydraulic profiles. The result is a transformed European soil map that gives a spatial
picture of the soil hydraulic properties of the unsaturated zone of European soils. Based on
this map it was possible to prepare a map showing total available water on a European
scale (Wösten et al., 1999). This map is just one example of the type of new spatial
information that can be generated when PTFs are used in combination with other existing
434

Table 2
Continuous pedotransfer functions developed from the HYPRES database
us ¼ 0:7919 þ 0:001691C 2 0:29619D 2 0:000001491S2 þ 0:0000821OM2
þ 0:02427C 21 þ 0:01113S21 þ 0:01472 lnðSÞ 2 0:0000733 £ OM £ C
2 0:000619DC 2 0:001183 £ D £ OM 2 0:0001664 £ topsoil £ S ðR2 ¼ 76%Þ
p
a ¼ 214:96 þ 0:03135C þ 0:0351S þ 0:646OM þ 15:29D 2 0:192topsoil
2 4:671D2 2 0:000781C 2 2 0:00687OM2 þ 0:0449OM21 þ 0:0663 lnðSÞ
þ 0:1482 lnðOMÞ 2 0:04546DS 2 0:4852 £ D £ OM þ 0:00673 £ topsoil £ C
ðR2 ¼ 20%Þ
np ¼ 225:23 2 0:02195C þ 0:0074S 2 0:1940OM þ 45:5D 2 7:24D2 þ 0:0003658C 2
þ 0:002885OM2 2 12:81D21 2 0:1524S21 2 0:01958OM21 2 0:2876 lnðSÞ
2 0:0709 lnðOMÞ 2 44:6 lnðDÞ 2 0:02264DC þ 0:0896 £ D £ OM þ 0:00718
£ topsoil £ C ðR2 ¼ 54%Þ
Lp ¼ 0:0202 þ 0:0006193C 2 2 0:001136OM2 2 0:2316 lnðOMÞ 2 0:03544DC
þ 0:00283DS þ 0:0488 £ D £ OM ðR2 ¼ 12%Þ
p
Ks ¼ 7:755 þ 0:0352S þ 0:93topsoil 2 0:967D2 2 0:000484C2 2 0:000322S2
þ 0:001S21 2 0:0748OM21 2 0:643 lnðSÞ 2 0:01398DC 2 0:1673 £ D £ OM
þ 0:02986 £ topsoil £ C 2 0:03305 £ topsoil £ S ðR2 ¼ 19%Þ
us is a model parameter, ap, np, L p and Kps are transformed model parameters in the
Mualem – van Genuchten equations; C, percent clay (i.e., percent , 2 mm); S, percent silt
(i.e., percent between 2 and 50 mm); OM, percent organic matter; D, bulk density; topsoil
and subsoil are qualitative variables having the value of 1 or 0; ln, natural logarithm.

spatial soil data. Other possible new products could be a travel time map for solutes or an
infiltration rate map for erosion studies.
The HYPRES database served as input in a study by Nemes et al. (2003) in which the
authors used national, continental and intercontinental scale data collections to derive
PTFs for the estimation of soil water retention, using artificial neural networks. The goal of
the study was to identify the relevance of PTF from international data collections for
individual countries that contributed to the database or for countries located in areas with
comparable climatic conditions. The same methodology and the same sets of predictors
were used while the source database was changed. In order to study the influence of
different combinations of predictors, 11 different PTFs were developed from each of the
data sets. Larger residuals were found for soils that were underrepresented in the data sets.
Water retention estimates were functionally tested. Soil moisture time series of seven soils
were simulated, using water retention characteristics estimated by the different PTFs. Sum
of square residuals were evaluated for simulated water contents and water contents
observed in the field. Small differences were found for PTFs derived from different scale
data collections. Moreover, PTF estimates resulted in simulations that were only
marginally worse than simulations with estimates using laboratory measured water
retention. In this case, differences in scales of PTF estimations contributed little to
differences in simulated and observed water contents.
 435

Application of HYPRES showed that it is possible to assign soil hydraulic

characteristics to soils with a textural composition comparable to the soils for which the
PTFs have been derived. However, care should be taken to use the functions for prediction
of hydraulic characteristics of soils outside the range of the original database. Class PTFs
give the mean hydraulic characteristics for rather broadly defined soil texture classes. As a
consequence, these functions are generally better applicable, however, they can give
limited site-specific information. In contrast, continuous PTFs can be more site-specific,
but their general applicability is limited as they require more specific input data.

REFERENCES

FAO (Food and Agriculture Organisation), 1990. Guidelines for Soil Description, Third
edition. FAO/ISRIC, Rome.
Furnival, G.M., Wilson, R.W., 1974. Regression by leaps and bounds. Technometrics 16,
499-511.
Nemes, A., Wösten, J.H.M., Lilly, A., Oude Voshaar, J.H., 1999. Evaluation of different
procedures to interpolate particle-size distributions to achieve compatibility within soil
databases. Geoderma 90, 187-202.
Nemes, A., Schaap, M.G., Wösten, J.H.M., 2003. Functional evaluation of pedo-
transfer functions derived from different scales of data collection. Soil Sci. Soc. Am. J.
67, 1093-1102.
Tietje, O., Tapkenhinrichs, M., 1993. Evaluation of pedo-transfer functions. Soil Sci. Soc.
Am. J. 57 (4), 1088-1095.
USDA (United States Department of Agriculture), 1951. Soil Survey Manual, U.S. Dept.
Agriculture Handbook No. 18. USDA, Washington, DC.
van Genuchten, M.Th., Leij, F.J., Yates, S.R., 1991. The RETC code for quantifying the
hydraulic functions of unsaturated soils. USDA, US Salinity Laboratory, Riverside, CA.
United States Environmental Protection Agency, document EPA/600/2-91/065.
Vereecken, H., Diels, J., van Orshoven, J., Feyen, J., Bouma, J., 1992. Functional
evaluation of pedotransfer functions for the estimation of soil hydraulic properties. Soil
Sci. Soc. Am. J. 56, 1371-1378.
Wösten, J.H.M., Lilly, A., Nemes, A., Le Bas, C., 1999. Development and use of a
database of hydraulic properties of European soils. Geoderma 90, 169-185.