JOURNAL OF ENVIRONMENTAL GEOGRAPHY Journal of Environmental Geography 7 (1–2), 11–22. DOI: 10.2478/jengeo-2014-0002 ISSN: 2060-467X DATASET FOR CREATING PEDOTRANSFER FUNCTIONS TO ESTIMATE ORGANIC LIQUID RETENTION OF SOILS Hilda Hernádi * , András Makó Department of Crop Production and Soil Science, Georgikon Faculty, University of Pannonia, Deák F. u. 16, H-8360 Keszthely, Hungary *Corresponding author, e-mail: hhilduci@gmail.com Research article, received 27 January 2014, accepted 14 March 2014 Abstract Soil properties characterising pressure-saturation relationships (P-S), such as the fluid retention values or the fitting paramete r of retention curves are basic input parameters for simulating the behaviour and transport of nonaqueous phase liquids (NAPLs) in subsurface. Recent investigations have shown the limited applicability of the commonly used estimation methods for predicting NAPL retention values in environmental practice. Alternatively, building pedotransfer functions (PTFs) based on the easily measurable properties of soils might give more accurate and reliable results for estimating hydraulic propertie s of soils and enable the utilisation of the wide range of data incorporated in Hungarian and international datasets. In spite of the availability of several well-established PTFs to predict the water retention of soils only a limited amount of research has been done concerning the NAPL retention of soils. Thus, in our study, data from our recent NAPL and water retention mea s- urements were collected into a dataset containing the basic soil properties as well. Relationships between basic soil propert ies and fluid retention of soils with water or an organic liquid (Dun asol 180/220) were investigated with principal component analysis. NAPL retention of soil samples were determined with PTFs, based on basic soil properties and their d erived values, and using a scaling method. Result of the statistical analysis (SPSS 13.1) revealed that using PTFs could be a promising alte r- native and could give more accurate results compared to the scaling method both for determining the NAPL saturation or the volumetric NAPL retention values of soils. Keywords: NAPL retention, pedotransfer function, hydraulic characteristics, Leverett equation INTRODUCTION From the 1980s increased attention has been received to improve understanding the relevant characteri stics and processes in flow and transport of s ubsurface nonaqueous phase liquids (NAPL) firstly in petrol industry and later to help design remedial strategies. Knowledge of pressure-saturation (P-S) relationships is essential for simulating the fate and transport of nonaqueous phase liquids (NAPLs) in subsurface with any type of models. Nowadays, as the measur e- ments of hydraulic parameters are costly and time consuming the development of accurate estimation method is preferred. In environmental practice the NAPL retention of soils is commonly determined by the table of average empirical pressure-saturation (P-S) values (dePas- trovitch et al., 1979), table of average fitting parame- ters of the van Genuchten equations proposed by Ca r- sel and Parrish (1988) (e.g. in RETZ model – van Genuchten et al., 1991; HSSM model – Weawer et al., 1994), the different modified versions of the Leverett function (Leverett, 1941) or their combination. In addition, Beckett and Joy (2003) created a dataset of calculated fitting parameters based on the modified scaling method suggested by Lenhard and Parker (1987). Nevertheless, tables of average NAPL rete n- tion or fitting parameter values of fluid retention curves do not accurately represent the variability of soils with different physical and chemical properties. All type of the modified Leverett function is valid for ideal porous systems. These estimation methods do not take into account the different interactions b e- tween the various fluids and the porous media, thus their application might be limited for natural soils (i.e. well aggregated, higher organic matter or clay co n- tent, etc.) soils. Furthermore, they have not been properly validated (only a few methods validated with column experiments, but these were carried out in most cases with glass beads and/or sands) (Rathfelder and Abriola, 1996; Makó and Hernádi, 2013). In the commonly used fluid retention measure- ment methods usually only the main drainage curves are measured. The more accurate determination of the main drainage curve of soil has key importance in determining the transport parameters (e.g. residual organic liquid content, penetration depth and time). In addition, main drainage curves could be the basis of determining the hysteretic and scanning curves in hysteretic models. 12 Hernádi and Makó (2014) Predicting of the physical, chemical and biological properties of soils with pedotransfer functions (PTFs) is a fast-developing field and several well-established PTFs are available to predict the water retention of soils both in Hungary (Rajkai, 2004; Rajkai et al., 1996; 2004; Nemes, 2003; Makó et al., 2005 and Tóth, 2011) and abroad (Minasny et al., 1999; Wösten et al., 1999; Rawls et al., 2001, etc.). Most of these estimation meth- ods have already been incorporated into numerical algo- rithms such as SOILPAR 2.0 (Acutis and Donatelli, 2003), Neuro Multistep (Minasny et al., 2004), TALA- JTANonc 1.0 (Fodor and Rajkai, 2011) or the k-Nearest (Nemes et al., 2008), etc. However, only a few studies had begun focusing on creating PTFs concerning the soils organic liquid retention capacity. Experiences in creating PTFs for water retentions may be essential tool for obtain the best possible esti- mation method for predicting the NAPL retention. PTFs for water retention (as response variable) may be created for the measured point values of the pressure saturation curves (point estimation) or for the fitting parameters of the hydraulic functions (parameter esti- mation) (Brooks and Corey, 1964; Brutsaert, 1966; van Genuchten, 1980) based on the easily measurable basic soil properties or their derived values (predictor varia- bles) (Wösten, 1995). The parameter estimation methods are widely used in environmental and soil hydrological practices because the simulation models mainly use these fitting parameters as input variable. Moreover, these parame- ters are equal to those of the equations predicting the soil hydraulic conductivity and relative permeability (van Genuchten, 1980; Lenhard and Parker, 1987; Chen et al., 1999). If the results of two point PTFs were to be compared or if no measured water rete n- tion points are available for a particular PTF, (or vice versa) the comparison of the retention curves with using the fitted hydraulic parameters (Minasny et al., 1999; Rawls et al., 2001) or calculating the estimated water contents at the desired pressure heads by l inear interpolation were suggested (Tietje and Tapkenhi n- richs 1993; Schaap and Leij, 2001). For parametric methods it is common to predict logarithmic transformed values of α (ln α) and n (ln n - 1) to convert the distribution of the parameters into a more statistically normal distribution (e.g., Rawls and Brakensiek, 1985; Wösten et al, 1999). Both in case of water and NAPL retention estimation, application of similar soil properties (bulk density or texture class information, organic matter and carbonate co ntent, etc.) or their inherited values (e.g. the averaged values of particle size data) were suggested as independent variables (Makó, 2004; Makó and Elek, 2006). In case of predicting the water retention the texture, morphology etc. are commonly used as a grouping factor in developing PTFs (Wösten et al. 1995; Schaap et al. 1999, 2001; Pachepsky and Rawls 2004). Statistical attributes of comparing PTFs for predicting water retention can potentially be adapted for the investigation of NAPL retention estimation methods. Recently, the determination of accuracy (with the working dataset), uncertainty and reliability (with test data) of the predictions, R, R2, mean error (ME), mean square error (MSE), root mean square error (RMSE), the unbiased root mean square errors (URMSE), etc. and their complete calculation are recommended, to acchieve a comprehensive verific a- tion (Pachepsky and Rawls, 2004). For the compari- son of the accuracy of the fitted fluid retention curves the calculation of ZAPF values were suggested by Rajkai (2004). The AIC (Akaike Information Criter i- on) value offers the possibility to compare the effi- ciency of different estimation or fitting methods with various numbers of parameters and the models with lower error can be selected with Fisher’s test (Rajkai, 2004). The uncertainty in input data can be evaluated using Monte Carlo analysis (Minasny et al. 1999), procedures based on fuzzy rules (McBratney et al., 2002) or with the Bootstrap Method (Carsel and Par- ish, 1988). Nowadays, the development of inference sys- tems (e.g. SINFER) to select the proper PTFs with minimum variance based on logical rules (McBra t- ney et al., 2002)., or the application of data driven methods, e.g. support vector machines (Lamorsky et al., 2008) might be a challenging t opic. The stability of the estimated coefficients can be i nvestigated using the double cross-validation techniques (K-fold, Leave-one-out, Jackknife or Delete-d methods) with randomly split the data (Pachepsky and Rawls, 2004). According to Tóth et al., (2013) it may be sufficient to random spilt the data in proportion of 90:10 (working and test data). Besides, many authors suggested the detection of outliers for calibrating PTFs. To our current knowledge, there is only one research had begun for creating PTFs concerning the soils organic liquid retention in Europe (Makó, 1995; Makó, 2002; 2004). In spite of the large number of measurements to determine the NAPL rete ntion of soils, databases from measured NAPL rete ntion data have not been created until now. Beckett and Joy (2003) created a dataset of calculated fitting param e- ters based on the modified scaling met hod suggested by Lenhard and Parker (1987) and Parker et al. (1987) but this contains the scaled NAPL retention values calculated from the fitted water retention values of HYPRES database. Creating PTFs for NAPL-retention might be promising because a large amount of measured basic data is available in na- tional and Hungarian databases (HYPRES, UNS O- DA v2.0, HUNSODA, EU HYDI, MARTHA etc.), which have already been used effectively in develop- ing PTFs for estimating water retention (Wösten et al., 1995; Nemes et al., 2003, 2008; Makó et al., 2005; Tóth et al., 2006, 2013; Lilly, 2010). More o- ver, the up-to-date hydrodynamic and transport mo d- els enable the adaptation of GIS (Geographic Infor- mation Systems) datasets (e.g. GMS - Groundwater Modelling System and Argus Open Numerical Env i- ronments - ARGUS ONE), which allow for carto- graphic representation with different commonly used software applications in the environmental engineer- Dataset for creating pedotransfer functions to estimate organic liquid retention of soils 13 ing practice (such us SURFER-GRAPHER or AU- TOCAD). In the 1990s a series of investigations for measuring NAPL retention of soils and mineral mi x- tures with the pressure plate method to create PTFs for organic liquid retention, in Hu ngary were started (Makó, 1995; Makó, 2002). In this study a dataset from these recent mea s- urements was created and analysed by statistical methods (SPSS 13.1). After the preliminary analysis (descriptive statistics and outlier detections), PTFs were built for predicting the fitting parameters of NAPL retention curves. Then, the NAPL retention of soils was predicted with classPTFs (for selected four texture groups) and using the scaling method of Le n- hard and Parker (1987). Afterwards, the accuracy and reliability of predicting NAPL retention with different estimation methods were compared. DESCRIPTION OF THE DATASET The dataset contains the physical and chemical pro p- erties and the fluid retention data of five measurement series collected from 1991 to 2011 (Table 1-2). Da- taset contains 369 disturbed and undisturbed samples, 40 soil profiles with 107 genetic layers and various types of soils (Fig. 1) for 10 texture classes are repre- sented (Fig. 2). Table 1 The subsets of dataset Subset N % Origin 1 111 30.1 undisturbed samples of 9 soil profiles of the research program of the Hungarian National Long-term Fertilization (Hernádi and Makó, 2011a) 2 123 33.3 undisturbed samples of 12 soil profiles of an investigation for the Hungarian Gas & Oil Company Plc (MOL Rt.) (Makó, 2002; 2005) 3 45 12.2 disturbed samples of mineral mixture series (Hernádi et al., 2011b) (Makó and Marczali, 1999) 4 60 16.3 disturbed samples of aggregate series separated from the upper „A” layer of selected soils (2.0 mm>, 1.0 mm>, 0,5 mm>, 0,25 mm> and 0,056 mm>) (Makó and Elek, 2006) 5* 30 8.1 disturbed samples investigated in the course of the TÁMOP-4.2.1/B- 09/1/KONV-2010-0003 Mobility and Environment Project (Hernádi et al., 2011) Sum 369 100 * before fluid retention measurements, samples were held 24hr in water for desaggregation and dried on 40°C/24hr Table 2 Soil properties in the dataset N N% water retention NAPL retention vol.% N N% vol.% N N% Surface area (BET) (cm 2 g -1 ) (Brunaer et al., 1938) 144 39.0 pF 0.0 308 83.5 0 mbar 245 66.4 Plasticity according to Arany (%) (Buzás, 1993) 339 91.9 pF 0.2 170 46.1 2 mbar 361 97.9 Particle size distribution (%)* 312 84.6 pF 1.0 111 30.1 20 mbar 242 65.6 Particle size distribution (%) ** 369 100.0 pF 1.3 59 16.0 50 mbar 245 66.4 Organic matter content (%) *** 369 100.0 pF 1.5 279 75.6 100 mbar 116 31.4 CaCO3 (%)*** 369 100.0 pF 1.7 59 16.0 150 mbar 245 66.4 Acidity (Na4OAc) (cmol (+) kg -1 )*** 66 17.9 pF 2.0 156 42.3 200 mbar 116 31.4 Acidity (KCl) (cmol (+) kg -1 )*** 66 17.9 pF 2.2 59 16.0 400 mbar 245 66.4 Salt content (mass %)*** 93 25.2 pF 2.3 111 30.1 500 mbar 116 31.4 Exchangeable Na (cmol(+)/kg)*** 87 23.6 pF 2.5 168 45.5 1000 mbar 361 97.8 Base saturation (w %)*** 24 6.5 pF 2.6 59 16.0 1500 mbar 30 8.1 Aggregate size distribution (%)* 105 28.5 pF 2.7 111 30.1 pH (distilled water 1 : 2,5)*** 273 74.0 pF 3.0 59 16.0 pH (KCl suspension 1 : 2,5)*** 189 51.2 pF 3.2 30 8.1 Bulk density (water) (g cm -3 )* 348 94.3 pF 3.4 111 30.1 Bulk density (NAPL) (g cm -3 )* 369 100.0 pF 4.2 279 75.6 pF 6.2 369 100.0 * Physical properties of soils were determined according to MSZ 08 0205:1978 standard (PSD – pipette method) **PSD of soils is determined according to ISO 11277:1995 with pipette method, after the total desaggregation of soil samples (eliminating the soil organic matter and CaCO3 content and the iron oxides). ***Chemical properties of soils were determined according to MSZ 08 0206/2:1978 standard 14 Hernádi and Makó (2014) METHODS The basic soil properties, used in this investigation were measured according to the Hungarian standards (MSZ08 0205:1978 and MSZ08 0206/2:1978). The texture of soil samples was determined according to the ISSS (International Society of Soil Science) texture triangle (Table 3). The fluid retention measurements were performed with distilled water and a special nonaromatic organic liquid, DUNASOL 180/220 (Hungarian Gas & Oil Company Plc. - MOL Rt.). Water retention measurements were carried out with porous pressure plate extractors (Soilmoisture Corp. LAB 023). For the purpose of determining the NAPL retention of the samples, a modified version (as de- scribed by Makó, 1995) of these pressure plate extrac- tors were used. Samples showing extremely high or low NAPL re- tention values were selected with outlier detection. Only the samples, were the casewise diagnostic of the preliminary linear regression between soil parameters and fluid retention values showed significant differ- ences with more than two times the standard deviation at a given pressure level, were eliminated. Then only the records where both the water and NAPL retention were measured were retained. Therefore, the final dataset contained 316 samples. The fluid retention measurements were performed with distilled water and a special nonaromatic organic liquid, DUNASOL 180/220 (Hungarian Gas & Oil Company Plc. - MOL Rt.). Water retention measurements were carried out with porous pressure plate extractors (Soilmoisture Corp. LAB 023). For the purpose of determining the NAPL retention of the samples, a modified version (as de- scribed by Makó, 1995) of these pressure plate extractors were used. Fig. 1 WRB (World Reference Base) soil types, clay minerals and mineral mixtures in the dataset (1: Benite, 2: Kaolin, 3: Illite, 4: Loess, 5: Pannon sand, 6: Mineral mixtures, 10: Eutric Cambisol, 11: Hortic Terric Cambisol, 12: Dystric Cambisol, 20: Haplic Arenosol, 30: Gleyic Luvisol, 40: Calcic Phaeosem, 41: Luvic Phaeozem, 42: Calcaric Phaeozem, 43: Haplic Phaeozem, 50: Vermic Calcic Chernozem, 60: Vertic Stagnic Solonetz, 61: Orthic Solonetz, 70: Gleyic Vertisol, 80: Cal- caric Gleysol, 90: Eutric Regosol, 100: Calcaric Fluvisol) Fig.2 Texture and sample types of soil, clay minerals and mineral mixture samples (1: Clay, 2: Silty clay, 4:Clay loam, 5: Silty clay loam, 6: Sandy clay loam, 7: Loam, 8: Silt loam, 9: Sandy loam, 11:Loamy sand, 12: Sand; A: Disturbed samples, B: Undisturbed samples, C: Mineral mixtures, D: Aggregate series of soil samples) Dataset for creating pedotransfer functions to estimate organic liquid retention of soils 15 Samples showing extremely high or low NAPL r e- tention values were selected with outlier detection. Only the samples, were the casewise diagnostic of the preliminary linear regression between soil parameters and fluid retention values showed significant diffe r- ences with more than two times the standard deviation at a given pressure level, were eliminated. Then only the records where both the water and NAPL retentio n were measured were retained. Therefore, the final dataset contained 316 samples. For fitting the NAPL retention curves the van Genuchten equation with three parameters (Eq. 1) (van Genuchten et al., 1980) were used as initial values in nonlinear regression (SPSS 13.1, Nonlinear regres- sion/quadratic sequential programming) with the average fitting parameters proposed by Carsel and Parish (1988) (used in e.g. RETZ and HSSM models).    nn s h h 1 1 )( 1       (1) where: θ(h) is the volumetric fluid content at potential h (kPa); θs,is the saturated fluid content; α, and n are fitting parameters. Afterwards, NAPL retention values for 10 pressure levels (0.01, 0.3, 10, 33, 100, 200, 330, 1000, 15849 and 1584893 mbar) were calculated. The accuracy of fitting was verified by the calcula- tion of Pearson R 2 , RMSE (Eq. 2) and ZAPF values (Eq. 3) (Rajkai, 2004). N yy RMSE N ii   1 2 )ˆ( (2) where: yi is the fitted NAPL retention, ŷi the estimated NAPL retention and N is the number of samples. In case of ZAPF values (Eq. 3.), the lower |ZAPF| than measurement error (ME) of NAPL retention was regard- ed as sufficiently fitted (as suggested by Rajkai, 2004). n ZAPF n i me    1  (3) where: θm is the measured fluid retention value θe estimated fluid retention values; n is the number of fitting point of fluid retention curves. Before creating PTFs the dataset was randomly split with the ratio of 90:10 (calibration and validation data set), thus we had opportunity to investigate the accuracy and the also the reliability of estimations (Table 4). In further investigations the fitted fluid retention data was used, which enabled us to compare the retention data determined with both fluid types measured at different pressure levels. Relationships between basic soil properties and fluid retentions, as well as between soil properties and fitting parameters were preliminarily investigated with principal comp o- nent analysis (PCA) (SPSS 13.1/Varimax rotation with Kaiser normalization). Subsequently, parametric PTFs were built as class PTFs for selected four texture groups (silty clay, silty clay loam, silt loam and sandy loam) with multiple linear regression (SPSS 13.1 Linear regres- sion/Stepwise method). In linear regression, bulk density, clay, silt, CaCO3 and organic matter content and their derived values, such as the logarithmic, reciprocal, squared values and their interactions were used as independent variables (as suggested by Wösten, 1995). The NAPL retention of the soil samples were pr e- dicted with a modified fitting method proposed by Lenhard and Parker (1987) (Eq. 4) which was a co m- bination of the Leverett equation (1941) and a four - parametric fitting procedure suggested by van Genuchten et al. (1980).   m n P S         1 1 (4) where: S is the fluid saturation; P is the pressure head; β is a scaling factor calculated from the interfacial tension of the phase pair and α, n and m are fitting parameters. In scaling method the m = 1-1/n constraints was sup- posed as suggested by van Genuchten (1991). In order to compare the accuracy and reliability of the different predictions, the RMSE and R 2 values were determined. The NAPL saturation values of soil samples were also calculated on the validation test in order to make a preliminary investigation to compare the reliability of PTFs with scaling method in predict- ing NAPL saturation values, versus the pressure. Table 3 Descriptive statistics of the selected basic soil properties used in this study Method Unit Mean SD SE Med. Min. Max. N Clay MSZ 08 0205:1978 w % 27.58 0.95 16.75 24.08 1.25 67.67 369 Silt 44.22 0.91 16.11 44.39 0.05 78.50 369 Sand 28.21 1.43 25.20 21.43 0.63 98.62 369 OM MSZ 08 0206/2:1978 w % 1.43 0.06 1.24 1.21 0.00 5.66 369 CaCO3 MSZ 08 0206/2:1978 w % 5.57 0.43 8.26 0.16 0.00 30.71 369 Bulk density (water)* MSZ 08 0205:1978 g cm -3 1.43 0.01 0.19 1.44 0.89 2.20 369 Bulk density* (NAPL) MSZ 08 0205:1978 g cm -3 1.42 0.01 0.26 1.45 0.80 1.99 369 *Samples for water and NAPL retention measurements had different bulk density. 16 Hernádi and Makó (2014) RESULTS AND DISCUSSION Fitting van Genuchten equation As the determination coefficient showed NAPL rete n- tion curves can be accurately fitted with the hydraulic function proposed by van Genuchten (1980). Moreo- ver, according to the ZAPF values, the differences between the fitted and measured NAPL retention va l- ues were lower than the measurement error in more than 98.4% of the samples (Table 5). Table 5 Statistics of the fitted van Genucthen equation to the measured NAPL retention values Equation ZAPF (<0,75 – ME*) RMSE R R 2 N Frequency (%) van Genuchten (3 parameter) 2286.0 98.4 1.2 0.99 0.99 * ME is the average error of NAPL retention measurement Principal component analysis Principal components were extracted from 17 variables which showed strong correlations between the selected soil properties. The maximum possible variance of 90.92% can be explained by the three components. Scores of component I (CI) give information about 69.56% scores of component II (CII) about 15.48% and scores of component III (CIII) about 5.87% of the vari- ance of variables. After Kaiser normalization, all components with eigenvalues under 1.0 were eliminated, and only three components were taken into account. In Table 6 and 7, component scores less than the variance explained criteria (0.4) are shown in gray. Communality values show, that all the variances of each correlated vari a- bles might be accounted for by the co mponents. The finer textured fractions belonged to the same comp o- nent with the NAPL retention at higher pressure lev- els, where the organic liquid can be retained against gravity mainly by capillary and adsorptive forces. The joint variation of these soil properties were a f- fected by CI as the unobserved latent variable. Mor e- over, the component scores of finer particle size frac- tions exceeded the 0.4 criteria only in CI. Differences in the role of components on the NAPL retention of soils were observed. Increasing value of component scores with increasing pressure can refer to the strong correlations between particle size distribution and fluid retention, especially in the higher pressure range. Higher component scores of CII were observed than CI on the 0-50mbar pressure range. In CII only the fluid retention less than 150 mbar were represented while the component scores of CI in case of NAPL retention at Table 4 Statistics of the fitted fluid retention values in calibration and validation datasets Pressure (mbar) Calibration set Validation set N = 264 N = 26 Mean SE SD Med. Min. Max. Mean SE SD Med. Min. Max. NAPL retention 0.0 41.0 1.0 10.0 41.0 22.0 68.0 38.0 2.0 10.0 37.0 25.0 57.0 0.33 41.0 1.0 10.0 41.0 22.0 68.0 38.0 2.0 10.0 36.0 25.0 57.0 10 36.0 1.0 10.0 35.0 17.0 64.0 33.0 2.0 8.0 32.0 21.0 49.0 33 31.0 1.0 8.0 31.0 9.0 62.0 28.0 1.0 5.0 29.0 20.0 38.0 100 25.0 0.0 6.0 25.0 5.0 60.0 24.0 1.0 5.0 24.0 16.0 33.0 200 22.0 0.0 6.0 21.0 3.0 55.0 22.0 1.0 6.0 22.0 10.0 32.0 330 20.0 0.0 6.0 19.0 2.0 49.0 21.0 1.0 6.0 22.0 7.0 32.0 1000 17.0 0.0 6.0 16.0 1.0 34.0 18.0 1.0 7.0 18.0 3.0 30.0 15849 11.0 0.0 7.0 11.0 0.0 29.0 14.0 1.0 7.0 14.0 0.0 26.0 1584893 7.0 0.0 6.0 6.0 0.0 22.0 10.0 1.0 6.0 9.0 0.0 21.0 Water retention 0.0 50.0 1.0 11.0 47.0 27.0 81.0 49.0 2.0 9.0 47.0 37.0 70.0 0.33 50.0 1.0 11.0 47.0 26.0 81.0 49.0 2.0 9.0 47.0 37.0 70.0 10 48.0 1.0 11.0 46.0 16.0 78.0 47.0 2.0 9.0 46.0 36.0 69.0 33 45.0 1.0 11.0 43.0 12.0 73.0 45.0 2.0 10.0 42.0 31.0 67.0 100 41.0 1.0 12.0 39.0 5.0 73.0 42.0 2.0 10.0 40.0 26.0 63.0 200 37.0 1.0 12.0 36.0 1.0 72.0 39.0 2.0 10.0 39.0 23.0 62.0 330 35.0 1.0 12.0 33.0 0.0 70.0 37.0 2.0 10.0 38.0 21.0 61.0 1000 29.0 1.0 11.0 27.0 0.0 65.0 32.0 2.0 10.0 32.0 17.0 56.0 15849 17.0 0.0 8.0 16.0 0.0 41.0 19.0 1.0 7.0 19.0 7.0 37.0 1584893 7.0 0.0 4.0 6.0 0.0 19.0 8.0 1.0 3.0 8.0 1.0 15.0 Dataset for creating pedotransfer functions to estimate organic liquid retention of soils 17 higher than 400 mbar exceed the 0.9 value (Table 6.) This difference might be resulted from reaching a threshold pressure range, which can be in the range of 20-50 mbar. At this pressure range the larger pores might be drained by gravity during NAPL retention measurements with Duna- sol 180/220, as similar results were observed by Makó et al. (2011a). This is lower than the pressure of the field capacity of soils, which is referred approximately 400 mbar (pF 2.5) in the case of water retention. This is might not be contradicted to the results of other researchers for water retention (Vere ecken et al., 1989; Saxton and Rawls, 2006). Vereecken et al., (1989) resulted that at pressure higher than 1500 kPa water retention is determined mainly by texture then in lower pressure range the fluid retention strongly influenced by the aggregation of soil particles and organic matter content as well. However, in case of NAPL retention the variance of organic matter content affects the NAPL retention but in decreasing rate with increasing pressure in CI and CII, and calcium carbonate content of samples appeared only in CIII. Variance in bulk densities can be explained by all three components, thus it may influence the NAPL retention at every pressure level. In all cases bulk density had negative correlation with other variables which refers inverse relationships between bulk dens i- ty and fluid content of soils samples. More complex and less correlation between pa- rameters of water and NAPL retention curves related to basic soil properties were observed. In case of wa- ter retention only the 55.59% of the variance could be explained by the components as compared to NAPL retention (76.55% (Table 7). Parameters α, n and θs have substantial roles in CI of NAPL retention in contrast to the parameters of water retention curves which belong to separated principal components. Descriptive statistics show (Table 4), that the mean saturated water content is significantly higher (50 vol %) than the possible average maximum NAPL retention (41 vol %), which might influence the variabil- ity of the other two fitting parameters. Parameters α and n are shape factors, strongly in- fluenced by the pore size distribution of soils and structure (van Genuchten et al., 1991; Vereecken et al., 1989). However, there is a difference in effective pore size distribution of soils saturated by fluids with various chemical and physical properties. In addition, interactions between water and the solid phase during drainage and imbibition processes, such as swelling and shrinking, have not occurred in case of the soil pores filled with NAPLs. Thus, the soil properties which might influence the shape of fluid retention curves might be difference. Strong correlation between the fitting parameters was experienced when the van Genuchten equation was fitted to NAPL retention data. The correlation between α and θs of the selected four texture group varied between 0.553-0.887. This might refers the inability of using the same initial parameters or boundary conditions for fitting NAPL and water retention curves. Lower communality values were experienced in the n and θs values of water retention curves than those of NAPL retention curves. This could be result- ed from the variability in NAPL retention of soils might be explained better with the selected basic soil properties (communalities of n was 0.821, and θs was 0.739) than in case of water retention (where the c u- mulated variance was only 0.282 and 0.298). Table 6 Components of variance and scores of the principal component analysis Variables CI CII CIII Communalities Basic soil properties 0.0002 mm > 0.93 0.22 -0.19 0.94 0.005-0.002 mm 0.87 0.36 0.11 0.89 0.01-0.005 mm 0.91 0.30 0.18 0.94 0.02-0.01 mm 0.75 0.35 0.50 0.93 0.05-0.02 mm 0.02 0.41 0.88 0.95 0.05-0.25 mm -0.94 0.05 -0.33 0.99 Organic matter (w %) 0.56 0.46 0.02 0.53 CaCO3 (w %) -0.07 0.05 0.95 0.91 Bulk density (g cm -3 ) -0.51 -0.53 -0.53 0.82 NAPL retention 0mbar (w %) 0.48 0.66 0.46 0.88 2mbar (w%) 0.49 0.65 0.45 0.87 20mbar (w %) 0.46 0.76 0.39 0.94 50mbar (w %) 0.56 0.59 0.54 0.96 150mbar (w %) 0.78 0.46 0.40 0.98 400mbar (w %) 0.91 0.34 0.14 0.96 1000mbar (w %) 0.93 0.34 0.04 0.98 1500mbar (w %) 0.90 0.41 -0.05 0.98 Statistics Variance % 69.56 15.48 5.87 18 Hernádi and Makó (2014) In both cases bulk density is presented in CI with negative effect on saturated fluid content, which can be a consequence the bulk density and saturated fluid content are closely related. In case of water retention the role of organic matter content is apparent only in CIII separately from clay content. This can be a consequence of the effect of high clay content might mask the effects of increasing organic matter for water retention, as Saxton and Rawls (2006) and Rawls et al. (2003) experienced as well. Similar relationships observed between organic matter content and the fitting parameters of water retention curves than Petersen et al. (1968) and Ve- reeckeen et al. (1989) that is the organic matter co n- tent of soils affects primary the α parameter of water retention curves, which refers the fluid rete ntion values at inflection point. In contrary with texture, organic matter content of soils influence the position of the retention curves, rather than its shape. Communality values showed the same important role of calcium carbonate (0.722, 0.709) and organic matter (0.579, 0.454) on water and NAPL retention. However, a separated component (CIII) was com- posed by organic matter, CaCO 3 and clay content in case of NAPL retention data. In addition the role of clay content on water retention seemed to be higher than its role on NAPL retention. All of these o b- served relationships refer to the different rate of interactions between various fluids and the solid phase considering these express mainly the linear relationships between these variables. In spite of the cumulated variance was only 55.59 % in case of water retention, the Kaiser -Meyer-Olkin test showed the applicability of PCA, sampling ad e- quacy (as it refers that the correlations between pairs of variables can be explained with other variables) was higher than 0.5, and the Barthlett’s test indicated significant correlation between variables (p<0.05). Pedotransfer functions Accurate prediction was given by classPTFs for the selected four texture groups for estimating NAPL reten- tion of soils (Table 8). Table 8 Accuracy of PTFs to estimate fitting parameters of NAPL retention curves Parameter R 2 RMSE α 0.978 13.59 n 0.900 17.75 θs 0.988 7.50 According to the RMSE values classPTFs give more accurate estimations to saturated fluid co ntent than the shape parameters α or n. The lower efficien- cy of estimating parameter α might be caused by the strong correlations experienced b etween parameters α and θs and might also resulted from the weaker correlation with independent variables. Table 7 Components of variance and scores of the principal component analysis with parameters of fluid retention curves Variables Water retention NAPL retention 4 CI CII CIII Communalities CI CII CIII Communalities a 0.025 -0.072 0.753 0.665 -0.738 0.451 -0.041 0.749 n -0.155 0.581 -0.101 0.282 0.468 -0.775 0.021 0.821 θs -0.602 0.171 -0.228 0.298 0.841 0.155 0.085 0.739 Clay content (w %) 0.847 -0.163 -0.237 0.933 0.409 0.657 -0.455 0.806 Silt content (w %) 0.469 0.478 0.077 0.932 0.097 0.815 0.337 0.788 Organic matter (w %) 0.014 0.123 0.784 0.579 -0.187 0.074 0.643 0.454 CaCO3 (w %) 0.084 0.801 0.137 0.722 0.426 0.03 0.726 0.709 Bulk density (g cm -3 ) -0.638 -0.315 -0.034 0.686 -0.889 0.016 0.136 0.808 Variance % 22.768 17.792 22.768 33.191 29.179 14.184 Cumulated variance % 22.768 40.560 55.599 33.191 62.370 76.554 Bartlett's Test of Sphericity 0.000 0.000 Kaiser-Meyer-Olkin Measure of Sampling Adequacy 0.597 0.634 Dataset for creating pedotransfer functions to estimate organic liquid retention of soils 19 The applicability of PTFs to predict the NAPL reten- tion values was proven by the statistical attributes (R 2 and RMSE). In comparison the accuracy of parametric PTFs with the scaling method, NAPL retention might be better predicted with PTFs at a given pressure level (Fig. 3). Applicability of these estimation methods for pre- dicting NAPL saturation values of soils was compared with the determination of R 2 and RMSE values on vali- dation data (10 % of the dataset, 26 samples). Lower reliability of PTFs was found to estimate the NAPL saturation with PTFs than the scaling method at higher than 100 mbar pressure level. However, PTFs could provide estimates sufficiently reliable for NAPL reten- tion at all pressure level, especially near saturation (Fig. 4). On Fig. 4 the statistics of saturated fluid con- tent were not presented because their values are 1 for both fluids. Supposedly, scaling method gave increasing better estimation for NAPL retention as a consequence of decreasing difference between fluid retentions with NAPL and water. As it shown formerly from descrip- tive statistics (Table 4), the difference between the average water and NAPL retention of the investigated soil samples was 26 vol. % at 100 mbar but practically 0 vol. % at 1584893 mbar. At the same time RMSE values indicated lower difference in reliability of estimation methods for predicting the NAPL saturation. Only moderate reliability of PTFs was found when predicting the volumetric NAPL retention values (Fig. 5). In spite of the fact that higher RMSE values were ob- served for PTFs in lower pressure ranges, the R 2 values show that the reliability of PTFs exceed that of the scaling method at lower pressure levels (< 330 mbar) and is al- Fig. 3 Accuracy of estimating volumetric NAPL retention values with different estimation methods PAR: the parametric method based on basic soil properties and their derived values; SC: Scaling method (Lenhard and Parker, 1987) Fig. 4 Reliability of estimating NAPL saturation values with PTF and scaling methods PAR: the parametric method based on basic soil properties and their derived values; SC: Scaling method (Lenhard and Parker, 1987) 20 Hernádi and Makó (2014) most equal to it at pressure higher than 1000 mbar (Fig. 5). Moreover both the decreasing R 2 values of the scaling method and the increasing RMSE values in the mid- pressure range (near the inflection point of the NAPL retention curves) suggest the inability of scaling methods for predicting NAPL retention of soil based on water retention. CONCLUSION In the last fifteen years several investigations were per- formed in order to create estimation methods to predict the NAPL retention of soils. However, the results of these measurements have not been collected into com- prehensive and well established databases, yet. Building PTFs using the data of easily measured soil property data might be a promising alternative for pre- dicting the NAPL retention of soils. By the creation of an organic liquid retention database representative for most soil types we can generate well established point and parametric PTFs based on basic soil properties. After the validation of PTFs with the results of column experiments it might be inserted to transport models. Moreover, databases of soil hydraulic and basic proper- ties which have already been used in developing PTFs for estimating water retention might be applicable to create NAPL pollution sensitivity maps. In our study an organic liquid retention dataset from re- cently measured NAPL and water retention data of soil and mineral mixture samples was created. The fluid retention measurements were carried out with the pressure plate extractor method. Furthermore, the basic soil properties, such as particle size distribution, organic matter and CaCO3 contents, and bulk density, etc. were also measured. According to our results, the van Genuchten equa- tion, having three parameters, could be perfectly fitted to the measured NAPL retention data as well as to water retention curves. Preliminary analysis of the filtered dataset of 316 soil samples with PCA showed strong correlations between the selected basic soil properties as well as fluid reten- tions of water and NAPL. The important role of organic matter and CaCO3 content in fluid retention of soils beyond the particle size distribution and bulk density was proved. The estimation methods for predicting the volu- metric NAPL retention values of soils PTFs gave better results than the scaling method. As the parame- ters of NAPL retention curves were correlated (e. g. fitting with nonlinear regression might offer more than one possible solution), further research is needed to investigate the role of initial parameter values in fitting hydraulic functions to NAPL retention data. Reliability of PTFs and the scaling method to pre- dict the saturation values versus the pressure were compared. Similar reliable estimations found when predicting the NAPL saturation values with scaling method and with PTFs. However, PTFs has been suc- cessfully applied to estimate the volumetric NAPL retention, more reliable than the scaling method. Acknowledgements The authors gratefully acknowledge Ágota Horel and Gábor Barton, for their assistance in preparing the manuscript. We would also like to thank Jenőné Borbély, for the support in laboratory experiments at Pannon University. This research was supported by the European Union and the State of Hungary, co-financed by the European So- cial Fund in the framework of TÁMOP-4.2.4.A/ 2-11/1- 2012-0001 ‘National Excellence Program’. References Acutis, M., Donatelli, M. 2003. SOILPAR 2.00: Software to estimate soil hydrological parameters and functions. European Journal of Agronomy 8 (3–4), 373–377. DOI: 10.1016/s1161- 0301(02)00128-4 Fig. 5 Reliability of estimating volumetric NAPL retention values with PTF and scaling methods PAR: the parametric method based on basic soil properties and their derived values; SC: Scaling method (Lenhard and Parker, 1987) Dataset for creating pedotransfer functions to estimate organic liquid retention of soils 21 Beckett, G.D., Joy, S. 2003. Light Non-Aqueous Phase Liquid (LNAPL) Parameters Database. Version 2.0. Users Guide. American Petroleum Institute. Publ. 4731. Washington, DC. Brooks, R. H., Corey, A. T., 1964. Hydraulic properties of porous media. Hydrology Paper No. 3. Colorado State University. Fort Collins, Colorado. Brunauer, S., Emmett, P. H., Teller, E. 1938. Adsorption of gases in multimolecular layers. Journal of the American Chemical Socie- ty 60 (2) 309–319. DOI: 10.1021/ja01269a023 Brutsaert, W. 1966. Probably laws for pore size distributions. Soil Science 101 85–92. DOI: 10.1097/00010694-196602000-00002 Buzás, I. 1993. Manual for soil and agrochemical analysis. INDA Kiadó, Budapest. (in Hungarian) Carsel, R.F., Parrish, R.S. 1988. Developing joint probability distribu- tion of soil water retention characteristics. Water Resources Re- search 24 755–769. DOI: 10.1029/wr024i005p00755 Chen, J., Hopmanns, J. W., Grismer, M. E. 1999. Parameter estimation of two-fluid capillary pressure-saturation and permeability func- tions. Advances in Water Resources 22, 479–493. DOI: 10.1016/s0309-1708(98)00025-6 DePastrovitch, T. L., Baradat, Y., Barthel, Y., Chiarelli, A., Fussel, D. R., 1979. Protection of groundwater from oil pollution. CONCAWE Rep. No 3/79. CONCAWE (Conservation of Clean Air and Water – Europe). The Hague. The Netherlands. Fodor, N., Rajkai, K. 2011. Computer program (SOILarium 1.0) for estimating the physical and hydrophysical properties of soils from other soil characteristics. Agrokémia és Talajtan 60, 27–40. Hernádi, H., Makó, A. 2010. Predicting oil retention of soils polluted with hydrocarbon derivates with pedotransfer functions. Mé- rnökgeológia-kőzetmechanika. Műegyetem Kiadó. Budapest. (in Hungarian) Hernádi, H., Makó, A. 2011a. Predicting the oil retention of soils with different methods. In Farsang, A., Ladányi, Zs. (eds.) Tala- jvédelem supplementum (3-4), 363–371 (In Hungarian) Hernádi, H., Makó, A., Kovács, J., Csatári, T. 2011b. Nonaqueous- phase liquid retention of mineral mixture series containing dif- ferent clay minerals. Communications in Soil Science and Plant Analysis 44, 390–396. DOI: 10.1080/00103624.2013.742314 Lamorsky, K., Pachepsky, Y., Słavińsky, C., Walczak, R.T. 2008. Using support vector machines to develop pedotransfer func- tions for water retention of soils in Poland. Soil Science Society America Journal 72, 1243–1247. DOI: 10.2136/sssaj2007.0280n Lenhard, R .J., Parker, J. C., 1987. Measurement and prediction of saturation-pressure relations in three phase porous media sys- tems. Journal of Contaminant Hydrology 1. 407–424. DOI: 10.1016/0169-7722(87)90017-9 Leverett, M. C. 1941. Capillary behavior in porous solids. Transactions of the Society of Petroleum Engineers. American Institute of Mechanical Engineers 142, 152–169. Lilly, A. 2010. A hydrological classification of UK soils based on soil morphological data. 19th World Congress of Soil Science, Soil Solutions for a Changing World 1 – 6 August 2010, Brisbane, Australia. Published on DVD. Makó, A. 1995. Interactions between the porous phase of soils and organic liquids. PhD dissertation. Keszthely, Hungary. (in Hun- garian) Makó, A. 2002. Measuring and estimating the pressure-saturation curves on undisturbed soil samples using water and NAPL. Agrokémia és Talajtan 51, 27–36. Makó, A. 2005. Measuring the two-phase capillary pressure-saturation curves of soil samples saturated with nonpolar liquids. Commu- nication in Soil Science and Plant Analysis 36, 439–453. DOI: 10.1081/css-200043170 Makó A., Elek, B. 2006. Comparison of soil extraction isotherms of soil samples saturated with nonpolar liquids. Water, Air and Soil Pollution 6, 331–342. DOI: 10.1007/s11267-005-9026-x Makó, A., Hernádi H. 2013. Hydrocarbon derivates in soils: Soil physical researches. Monography. (in Hungarian) Makó, A., Marczali, Zs. 1999b. Laboratory measurement of the soils fluid retention concerning the organic liquid retention. XIII. Or- szágos Környezetvédelmi Konferencia és Szakkiállítás. Siófok. 14-16. szept. 1999. 147-153. (In Hungarian) Makó, A., Máté, F., Németh, T., Hernádi H. 2004. The temperature- dependence of the NAPL retention of different soils. 12th Inter- national Poster Day and Institute of Hydrology Open Day: Transport of water, chemicals and energy in the soil-plant- atmosphere system. 25th November 2004. Bratislava. Slovak Republik. Makó, A., Rajkai, K., Tóth, G., Hermann, T. 2005. Estimating soil water retention characteristics from the soil taxonomic classifi- cation and mapping informations: consideration of humus cate- gories. Cereal Research Communications 34, 199–201. DOI: 10.1556/crc.33.2005.1.27 McBratney, A. B., Minasny B., Cattle S. R., Vervoort R. W. 2002. From pedotransfer functions to soil inference systems. Ge- oderma 109, 41–73. DOI: 10.1016/s0016-7061(02)00139-8 Minasny, B., McBratney, A. B., Bristow, K. L. 1999. Comparison of different approaches to the development of pedotransfer func- tions for water-retention curves. Geoderma 93, 225–253. DOI: 10.1016/s0016-7061(99)00061-0 Minasny, B., J., Hopmans, J. W., Harter, T., Eching, S. O., Tuli, A., Denton, M. A. 2004. Neural networks prediction of soil hydrau- lic functions for alluvial soils using multistep outflow data. Soil Science Society of America Journal 68, 417–429. DOI: 10.2136/sssaj2004.4170 Nemes, A., Schaap, M. G., Wösten, J. H. M. 2003. Functional evalua- tion of pedotransfer functions derived from different scales of data collection. Soil Science Society of America Journal 67, 1093–1102. DOI: 10.2136/sssaj2003.1093 Nemes, A., Roberts, R. T., Rawls, W. J., Pachepsky, Y. A., van Genuchten, M. T. 2008. Software to estimate -33 and -1500 kPa soil water retention using the non-parametric k-nearest neighbor technique. Environmental Modelling & Software 23 (2), 254– 255. DOI: 10.1016/j.envsoft.2007.05.018 Pachepsky, Y. A., Rawls, W. J. (Eds.) 2004. Development of Pedo- transfer Functions in Soil Hydrology. Developments in Soil Sci- ence. Elsevier. Amsterdam. Parker, J.C., Lenhard, R.J., Kuppusamy,T. 1987. A parametric model for constitutive properties governing multi-phase flow in porous media. Water Resources Research 23, 618–624. DOI: 10.1029/wr023i004p00618 Petersen, G. W., Cunnungham, R. L., Matelski, R.P. 1968. Moisture characteristics of Pennsylvania soils II. Soil factors affecting moisture retention within a textural class - silt loam. Soil Science Society America Proceedings 32, 866–870. DOI: 10.2136/sssaj1968.03615995003200060042x Rajkai, K. 1988. Relationships between water retention and different soil properties. Agrokémia és Talajtan 36–37, 15–30. (In Hun- garian) Rajkai K. 2004. The quantity, distribution and transport of water in soil. RISSAC, Research Institute for Soil Sciences and Agricul- tural Chemistry. Budapest. (In Hungarian) Rajkai, K., Kabos, S., van Genuchten, M. T., Jansson, P. E. 1996. Estimation of water-retention characteristics from the bulk den- sity and particle-size distribution of Swedish soils. Soil Science 161, 832–845. DOI: 10.1097/00010694-199612000-00003 Rajkai, K., Kabos, S., van Genuchten, M. T., 2004. Estimating the water retention curve from soil properties: Comparison of linear, nonlinear and concomitant variable methods. Soil and Tillage Research 79. (2) 145–152. DOI: 10.1016/j.still.2004.07.003 Rathfelder, K., Abriola, L. M. 1996. The influence of capillarity in numerical modelling of organic liquid redistribution in two- phase systems. Advances in Water Resources 21 (2), 159–170. DOI: 10.1016/s0309-1708(96)00039-5 Rawls, W. J., Brakensiek, D. L., 1985. Prediction of soil water proper- ties for hydrologic modeling. In: Watershed Management in the 1980s. Proceeding of Symposium of Irrig. Drainage Div., Den- ver, CO., April 30–May 1, 1985. American Society pf Civil En- gineers. New York. 293-299. Rawls, W. J., Pachepsky, Y. A., Shen, M. H., 2001. Testing soil water retention estimation with the MUUF pedotransfer model using data from the southern United States. Journal of Hydrology 251. 177–185. DOI: 10.1016/s0022-1694(01)00467-x Rawls, W.J., Pachepsky, Y. A., Ritchie, J. C., Sobecki, T. M., Blood- worth, H. 2003. Effect of Soil Organic Carbon on Soil Water Retention. Geoderma 1 (16), 61–71. DOI: 10.1016/s0016- 7061(03)00094-6 Saxton, K. E., Rawls, W. J. 2006. Soil Water Characteristic Estimates by Texture and Organic Matter for Hydrologic Solutions. Soil Science Society America Journal 70, 1569–1578. DOI: 10.1016/s0016-7061(03)00094-6 Schaap, M. G., Leij, F. J., van Genuchten, M. T. 1999. A bootstrap- neural network approach to predict soil hydraulic parameters. In: 22 Hernádi and Makó (2014) Proceedings of International Workshop Characterization and Measurements of the Hydraulic Properties of Unsaturated Po- rous Media 1237–1250. Schaap, M. G., Leij, F. J., van Genuchten, M. T. 2001. Rosetta: A computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. Journal of Hydrology 251, 163–176. DOI: 10.1016/s0022-1694(01)00466-8 Tietje, O., Tapkenhinrichs, M. 1993. Evaluation of pedo-transfer functions. Soil Science Society of America Journal 57, 1088– 1095. DOI: 10.2136/sssaj1993.03615995005700040035x Tóth, B. 2011. Calculation and characterization of water retention of major Hungarian soil types using soil survey information. PhD dissertation. Keszthely. (in Hungarian) Tóth, B., Makó, A., Rajkai, K., Kele, Sz. G., Hermann, T., Marth, P. 2006. Use of soil water retention capacity and hydraulic conduc- tivity estimation in the preparation of soil water management maps. Agrokémia és Talajtan 55, 49–58. DOI: 10.1556/agrokem.55.2006.1.6 Tóth, B., Makó, A., Tóth, G., Farkas, C., Rajkai, K. 2013. Comparison of pedotransfer functions to estimate the van Genuchten parame- ters from soil survey information. Agrokémia és Talajtan 62 (1), 5-22. (in Hungarian) van Genuchten M.T. 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society American Journal 44, 892–989. DOI: 10.2136/sssaj1980.03615995004400050002x Vereecken, H., Maes, J., Feyen, J., Darius, P. 1989. Estimaing the soil moisture retention characteristic from texture, bulk density and carbon content. Soil Science 148, 389–403. DOI: 10.1097/00010694-198912000-00001 Wösten, J. H. M., Schuren, C. H. J. E. Bouma, J., Stein, A. 1990. Functional sensivity analysis of four methods to generate soil hydraulic functions. Soil Science Society of America Journal 54, 832–836. DOI: 10.2136/sssaj1990.03615995005400030036x Wösten, J.H.M., Finke, P.A., Janses, M.J.W. 1995. Comparison of class and continuous pedotransfer functions to generate soil hy- draulic characteristics. Geoderma 66, 227-237. DOI: 10.1016/0016-7061(94)00079-p 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. DOI: 10.1016/s0016-7061(98)00132-3 Weawer, J. L., Charbeneau, R. J., Tauxe, J. D., Lien, B. K., Provost, J. B. 1994. The hydrocarbon spill screening model (HSSM). 1. US EPA. EPA/600/R-94/039a.