HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (9) Citing Articles via Google Scholar Google Scholar Articles by Mecke, M. Articles by Ilvesniemi, H. Search for Related Content PubMed Articles by Mecke, M. Articles by Ilvesniemi, H. GeoRef GeoRef Citation Agricola Articles by Mecke, M. Articles by Ilvesniemi, H. Related Collections Structure and Properties Soil Physics

DIVISION S-1 - SOIL PHYSICS

Water Retention Capacity in Coarse Podzol Profiles Predicted from Measured Soil Properties

Dep. of Forest Ecology, Univ. of Helsinki, P.O. Box 27, FIN-00014, Finland

* Corresponding author (marja.mecke{at}helsinki.fi)

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
Fennoscandian podzols consist of coarse parent materials with a strongly developed secondary structure in the Spodic B-horizons. For such materials, predictive standard water retention models developed for soils with a wide range of texture, including fine clay-rich soils, do not give estimates of water retention capacity with high enough precision. In this study we describe an approach to predict water content of coarse podzolic soils separately at seven potentials using multiple linear regression approach and stepwise regression analysis (Rawls et al., 1982, 1983). One model is constructed using only the easily measurable soil texture and dry bulk density data. To include the effects of secondary structure, two other models are constructed by using several measured soil physical and chemical properties for all horizons, or for all horizons excluding the eluvial horizons. These models explain between 90 and 95% of variation in water content (i.e., R² = 0.90–0.95) at potentials -3.2 to -100 kPa and at saturation, but not at the -1585 kPa potential. The validity of the models is tested on soils from four different sites covering a large range of soil properties of Fennoscandian podzols. Two of the models give accurate estimates of water content with errors 0.02 ± 0.03 m³ m^-3 at potentials of -10 kPa, -100 kPa, and at saturation. Results show that podzolization processes, including formation of secondary structure by enriched Al and Fe, affect soil water retention, but estimates based on oxalate extractable Al and Fe content cannot be generalized to other soils using the same model calibration.

Abbreviations: (h), volumetric soil water content • _s, water content value at saturation • Csand, coarse sand • D_b, oven dry bulk density • Fsand, fine sand • h, water potential • P_tot, total porosity • P_max, the highest P-value of the single variable • SiCla, sum of silt and clay

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
THE HYDRAULIC PROPERTIES OF A SOIL MATRIX depend on soil structure, which can be described in terms of the spatial distribution of pore spaces (FitzPatrick, 1983). Water retention capacity depends primarily on total porosity and pore-size distribution, which are related to texture, bulk density, and secondary structure.

In podzolic forest soils, several soil-forming processes affect the water retention properties of soil surface horizons over time. In the upper part of illuvial horizon (B_hs, B_s) the enrichment by Al, Fe, and organic compounds leads to the formation of secondary structure that is completely different from single-grained parent materials (Bohn et al., 1985; Browne, 1995; Petersen, 1976; Liski and Westman, 1995). Enrichment by the Al and Fe around and between sand grains forms an isotropic material occuring as single or clustered fine granules. This results in an irregular pore system including a considerable amount of small pores. Additionally, in horizons, where the proportion of textural fraction <0.02 mm (including silt and clay) is greater than 20%, phyllosilicate mineralogy impacts structure development by producing colloidal material (Bohn et al., 1985).

Freezing and thawing, soil biological activity, movement of tree roots when the stems are bent by the wind (Hintikka, 1972), and downward transport of organic matter from the surface humus layer, affect soil structure primarily by decrease in density and increase in porosity in eluvial and upper illuvial horizons. An increase in the total volume of pores in the upper part of illuvial horizon is due to an increase in the amount of small and medium-sized pores accompanied by a decrease in the large-sized pores (FitzPatrick, 1983; Scheffer and Schachtschabel, 1989).

These young podzols differ from older podzols found elsewhere, primarily because of relative age. Only about 10 000 yr have passed since the last glaciation in Fennoscandia, and their soil minerals are only slightly weathered. Furthermore, during deglaciation the soil particles were sorted by the melt water and wave action washed away fine particles in surface layers near the shorelines, which changed continuously due to the land uplift. As a result, these soils usually consist of coarse parent materials with little clay (Tamminen and Starr, 1994).

Water balance and solute transport models are widely used to study changes in forested ecosystems (Tiktak and van Grinsven, 1995). In many cases, lack of adequate information about soil water retention capacity restricts the applicability of these models despite the fact that data on other soil chemical and physical properties are available (Kareinen et al., 1998). Using a pressure plate extractor to determine soil water retention capacity is laborious, demanding special equipment available only in certain laboratories. Since hydraulic properties of these young podzol forest soils can vary considerably both horizontally and with depth, many measurements are needed to fully describe them. If the hydraulic properties of soils could be estimated from texture and bulk density, the utilization of the already available soil data and the accuracy of the model predictions could be markedly improved (Addiscott and Wagenet, 1985; Jarvis et al., 1991). Estimates of water retention capacity calculated from simple soil properties could then be used to improve forest management and site classification.

Several types of models and classification schemes are available in literature to estimate soil water retention characteristics from routinely measured textural and structural soil properties (Arya and Paris, 1981; Cosby et al., 1984; Saxton et al., 1986). The multiple linear regression approach of calculating water content at a number of fixed potentials has been applied by Gupta and Larson (1979), Rawls et al. (1982)(1983), and Puckett et al. (1985) to soils with generally high clay content. When comparing effectiveness of textural parameters for these models at -100 kPa potential, clay had the highest value of the regression coefficient. However, at -10 kPa sand (2–0.05 mm) and silt contents (USDA classification) were most influential with clay not far behind. Since in Fennoscandian podzols the content of coarse sand (Csand, 2–0.2 mm) is high, and the amount of clay is low (ICS classification), better fit at -10 kPa potential than at lower values would be expected.

In the Rawls et al. (1982)(1983) models, 2543 soil samples were included, and the correlation coefficients varied between 0.80–0.88 at potentials between -10 kPa and -1500 kPa. The potential values higher than -10 kPa were not modelled. Ahuja et al. (1985) tested Rawls' models with 189 soil samples, and Williams et al. (1992) tested Rawls' models and models of Gupta and Larson (1979) with data consisting of 366 samples of fine sandy loam from a research plot. In both studies, mean errors for several potentials and the standard deviations of the differences were high for all models. In a review article on water retention models, Rawls et al. (1991) concluded that the effects of Csand, organic matter, and chemical soil properties on water retention need to be more thoroughly evaluated.

The models of Gupta and Larson (1979) provide a good example of a limited application of multiple linear regression approach with high precision. The models were constructed on the data collected from 43 packed samples, and the correlation coefficients obtained were over 0.94 at potential ranges from -4.0 to -1500 kPa. However, when tested on 61 samples of Missouri soil, the prediction accuracy was not very high. Puckett et al. (1985) obtained good prediction accuracy (R² > 0.92) when applying Rawls' method (Rawls et al., 1982, 1983) for spatially limited soils with highly variable clay content (1–42%) but very low organic matter. Because the relationships between model parameters differed from those previously reported, Puckett et al. (1985) concluded that separate models should be developed for soils having similar mineralogy and genesis. This conclusion is supported also by the Ahuja et al. (1985) and Williams et al. (1992) studies.

The podzols in Fennoscandia form a distinct subgroup of soils clearly deviating in properties from other soils due to similar parent materials, preconditions, and climatic conditions during and after the last glaciation. These soils tend to be coarse-textured, and have a highly variable total porosity, organic C content, and structural attributes, which impact their water retention. Thus, models developed for finer textured-soils have only a limited use for these podzols.

The purpose of this study was to construct a multiple linear regression model and to investigate the functional relations between the model variables for coarse podzols, where water content at each measured potential is estimated from other soil properties.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
Site Description and Sampling
The study sites were located at Juupajoki in southern Finland, 190 km to the north of Helsinki (61°48' N, 24°19' E). Three sites of Haplic podzols (Food and Agriculture Organization of the United Nations-Unesco, 1990) under mixed Scots Pine and Norway Spruce stands were selected to develop and test the multiple linear regression models (S1-S3), and four sites were selected to validate the models (S4-S7). Table 1 shows the profile description for test and validation sites, while Tables 2 and 3 list the textural, chemical, and physical properties of each.

Intact 150 x 10^-6 m³ soil cores were extracted with a steel cylinder (diameter 0.057 m, height 0.059 m). The total number of test soil cylinder samples was 86. At Sites 1 and 2, seven randomly selected sampling points were taken from a 4 by 4 grid with 3-m intervals. At Site 3, eight sampling points, on the average of 7 m from each other, were selected to cover the variation in the soil depth to bedrock. The sampling depths were 0, 0.07, 0.23, and 0.50 m from the surface of the mineral soil. Same sampling depths were selected since the thickness of each morphological horizons was almost similar at all sites.

The texture of the four validation sites represented rather well a range of texture common to Fennoscandian podzols, considering the mean and the range of particle-size deviations and also by having three various basic formation mechanisms of the parent material (glaciofluvial sand, delta sand, and till). Variation of the soil properties was increased due to the different fertility of the forest sites. In addition, in a single soil profile, the soil properties varied considerably between horizons. As a result, the validation samples represented a range of values including the extremes of various soil properties.

Two of the validation sites had sorted parent material. One was a delta sand with a narrow particle-size distribution concentrated within the medium fraction, and the other a glaciofluvial sand with a wide particle-size distribution. The two other sites were tills, including a fine sand (Fsand) till having considerable proportion of particles in all texture classes, and a Csand till with high proportions of gravel. At each site the cylinder samples were taken in one measurement point at four successive morphological soil horizons. The deepest sample in Csand till was omitted, since 69% of its mass consisted of gravel and its water retention values would have been very low. Thus the number of samples used for model validation was 15. One erroneous water content value at potential (h) = -100 kPa was discarded since it was higher than the value at h = -10 kPa for the same sample.

At all seven sites, the eluvial horizon (A, E) was light grey or dark grey brown with a rather loose composition of soil or a medium (2–5 mm) granular large scale structure. The thickness of the eluvial horizons was 0.06 ± 0.03 m (Table 1). The upper illuvial horizon (B_hs, B_s) varied in color from yellowish red to dark redish brown and in composition from loose to very firm and sticky. The large scale structure was coarse granular, but also some angular blocks of soil of medium size (10–20 mm) could be found. At the depth of about 0.30 m, the illuvial horizon had gradually changed into single-grained parent material. When the clay content of parent material at Site 3 was over 8%, the soil had partly a fine (<2 mm) platy structure.

The soil core sampling techniques used with the test and validation soil sampling differed slightly with respect to soil compaction. This was because of minor differences in inserting and adjusting the upper edge of the soil core.

Determination of Soil Physical and Chemical Properties in Laboratory
Desorption curves were measured by a pressure plate extractor (Soilmoisture Equipment Corp., Goleta, CA) at potentials of h = -1.0, -3.2, -6.3, -10, -100 and -1585 kPa. For the lowest potential, a homogenized subsample with one-third of the volume of the original sample was used. The volumetric water content, (h), at each potential (Tables 4, 5) and oven dry bulk density, D_b, (Tables 2, 3) were determined gravimetrically. Total porosity, P_tot, was assumed to be equal to the water content value at saturation, _s. From a practical point of view, the measured water content also described the water retention capacity and available water in the soils that have high organic matter content. Particle-size distribution and oxalate extractable Al and Fe and total C contents were determined from soil samples collected adjacent to cores used for desorption.

View this table: [in this window] [in a new window]   Table 4. Mean values and coefficients of variation (in parenthesis) of soil water content (h) at potential h measured in kPa, for four soil horizons at three test sites (S1–S3), n = 7 at S1 and S2, n = 8 at S3.

View this table: [in this window] [in a new window]   Table 5. Soil water content (h) at potential h measured in kPa for four soil horizons at four validation sites (S4–S7).

Statistical Analyses
Soil properties of the 101 samples (n = 86 + 15) were statistically analyzed for (h) at seven measured potentials, D_b, C content, oxalate extractable Al and Fe content, and their sum AlFe. In addition, Csand (2–0.2 mm), Fsand (0.2–0.02 mm), sum of silt and clay (SiCla, <0.02 mm), Clay (<0.002 mm), and Gravel (>2 mm) fractions, were included in the analysis.

Pearson's correlation analysis, partial correlation analysis, and stepwise regression analysis were applied to data, and three multiple linear regression models were constructed (Freund and Minton, 1979; Neter et al., 1990). The applied forward stepwise regression analysis method was a combination of adding variables and controlling every new set of them to exclude any variable on basis of limits set on statistical model parameters. The homoscedasticity and linearity, normality and independence of residuals, and multicollinearity of parameters in the models were controlled. In the correlation analyses, both the correlation coefficient and its significance were considered. The analyses were performed using statistical program Statistix (NH Analytical Software, Roseville, MN).

Three multiple linear regression models were constructed, which included different amount of soil information or different number of soil horizons (Table 6). Model 1 was constructed for the cases, where only information on variables Csand, SiCla, Gravel, and D_b are available, and all samples were included (n = 86). Variable _s was considered as a dependent variable. The equation to predict separately (h) at different potentials, (h) (m³ m^-3), was

[1]

where a, b, c, d, and e are regression coefficients and texture is defined by the ICS system. For Model 2, all available information was used, and thus eleven variables were included (n = 86). The resulting equation for potentials from h = -1.0 kPa to -100 kPa was of the form

[2]

where f is a regression coefficient, and for potential of h = -1585 kPa of the form

[3]

[4]

and for potential of h = -1585 kPa of the form

[5]

Opposite to Models 1 and 2, in Model 3, proportions of different particle-size fractions were calculated from <2 mm fraction, excluding the amount of gravel. In (h) calculations, the volume of gravel particles was subtracted from the volume of the cylinder before calculating the (h). This was done in order to focus the analysis to the soil fabric and morphological characteristics.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
Table 2 shows the mean values and the coefficients of variation of test samples, and Table 3 shows the values of validation samples for textural, physical, and chemical soil properties. Tables 4 and 5 list the mean values and the coefficients of variation or values of (h) at measured potentials, also for the same test and validation sites, while Table 6 lists the regression coefficients of the three models for predicting (h) at measured potentials. In Table 6, the P-value of the t-statistics gives the risk level, according to which the coefficient of a single variable is zero when all other variables of the model are included. P_max is the highest P-value of the single variable in the submodel of each potential.

Coarse sand content was high and varied considerably between samples and sites. The mean values and coefficients of variation by site and horizon indicate that Csand samples were rather equally distributed over the measured range. Gravel content was mostly low and varied in an irregular way between samples and sites, while Clay content was low for all samples. Both the amount of C in the surface horizons and the amount of oxalate extractable Al and Fe in the upper illuvial horizons were high.

Multiple Linear Regression Models
The regression coefficients of the variables, the degrees of explanation, standard deviations, and P_max–values of the three multiple linear regression models (Eq. [1]–[5]) are listed in Table 6. For Model 1, the calculated vs. measured water content values at h = -10 kPa potential and at saturation are plotted in Fig. 1a and 1b , respectively, and for Model 2, the calculated vs. measured water content values at h = -10 kPa and -100 kPa potential are plotted in Fig. 1c and 1d, respectively.

View larger version (38K): [in this window] [in a new window]   Fig. 1. Comparison of soil water contents at a given potential calculated by a model with measured values at test sites and at validation sites; (a) Model 1 at h = -10 kPa potential, (b) Model 1 at h = 0 (saturation), (c) Model 2 at h = -10 kPa potential, and (d) Model 2 at h = -100 kPa potential.

For water content of the potentials between -3.2 kPa and -100 kPa, a textural variable (Csand for -3.2, -6.3, and -10 kPa, and SiCla for -100 kPa) was always chosen by the stepwise regression analysis as the first explaining variable for all three models, with the average degree of explanation of this single variable r² = 0.77 (additional analysis of data, not listed here). This is referred to as the textural pore region potential interval, according to Jarvis and Messing (1995). At these potentials, a density-related variable (P_tot, D_b, or AlFe) increased the degree of explanation of the models on average by 0.14. The water content at potential level -1585 kPa could not be explained without the variable Al. Model 3 explained the water content at -1585 kPa better than Model 2, even if R² for this water content in Model 3 was not more than 0.59 (Table 6).

In Model 1, with variables Csand, D_b, and SiCla or Gravel, the average degree of explanation of the water content of textural pore region potentials (between -3.2 kPa and -100 kPa) was R² = 0.91 and the standard deviation was 0.031 m³ m^-3 (Table 6). By using all available information in the stepwise regression analysis, the prediction accuracy was increased to R² = 0.93 and standard deviation decreased to 0.028 m³ m^-3 (Model 2).

When the eluvial horizons and the effect of gravel were excluded, high accuracy was achieved with only one textural variable and variable AlFe (Model 3). At textural pore region potentials, between -3.2 kPa and -100 kPa, average R² was 0.94, and at saturation, R² was 0.91.

Correlation Analysis
When Pearson's correlation matrix for the eleven independent variables (n = 86) was constructed (Table 7), two subgroups, the texture subgroup consisting of Csand and SiCla (Csand vs. SiCla |r| = 0.81) and density-related subgroup D_b, C, and P_tot (D_b vs. C |r| = 0.88, P_tot vs. C and D_b |r| = 0.91, 0.95) could be discerned with large within-group pairwise correlations for both. However, between the variables of the two subgroups, the correlations were considerably lower. Thus, the correlation coefficients of density related variables with Csand and SiCla were for P_tot |r| = 0.42, 0.42, for D_b |r| = 0.32, 0.32, and for C |r| = 0.47, 0.55, respectively. Variables Al, Fe, and AlFe did not appear to be strongly correlated with variables of either group.

View this table: [in this window] [in a new window]   Table 9. Partial correlations of the oxalate extractable Al and Fe and their sum AlFe with volumetric water contents (h) at the measured potentials h in kPa (n = 64 eluvial horizons excluded). The variables for which the correlations are adjusted in the partial correlation analysis are given in parenthesis.

The absolute and relative differences between the predicted and measured (h) of the four independent validation sites are given in Table 10. For Models 1 and 2 at potentials -10 and -100 kPa and for Model 1 at saturation, the water retention at validation sites was predicted relatively accurately (Figs. 1a–d). In contrast for two earlier studies (Ahuja et al., 1985; Williams et al., 1992) where good validation could be obtained only with a model in which one, or two water content values measured at fixed potentials were included.

View this table: [in this window] [in a new window]   Table 10. The differences between the predicted and measured volumetric water contents (h) at four validation sites (n = 15).

Model 3, which included oxalate extractable Al and Fe, gave the highest deviation between the predicted and measured values. The prediction errors were also high when soil properties included gravel. Thus the effect of the secondary structure parameter, AlFe, was so site-specific that the model could not be generally applied, without at least some calibration.

Testing of Rawls' Models
When Rawls' models (Rawls et al., 1982, 1983) were tested with the soil materials from this study at h = -10 kPa, -100 kPa, and -1500 kPa they showed large differences between predicted and measured values. The mean errors and standard deviations of errors were found to be 0.08±0.06 m³ m^-3; 0.03±0.05 m³ m^-3; 0.02±0.04 m³ m^-3, respectively. Similar results were obtained by Ahuja et al. (1985) and Williams et al. (1992). If the 16 samples at 0 m and 0.07 m of Site 3, which had very high organic C content, were excluded, Rawls' models predicted the water content of the two lower potentials (h = -100 kPa, and -1500 kPa) well with the mean errors and standard deviations of errors of 0.01±0.02 m³ m^-3 and 0.004±0.022 m³ m^-3, respectively. The Rawls' models for h = -10 kPa were tested with a limited data set, where 14 samples with high content of Csand at 0.23 m and 0.50 m depth from Site 1 were excluded in addition to samples from Site 3 with very high organic C content. Using the values without gravel fraction for all variables the test result was 0.03±0.03 m³ m^-3. This implied that the characteristic features of our test soils (highly varying organic C content, dominant Csand particle fraction, and gravel content at higher potentials) explained why the estimates produced by Rawls' models were not precise.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
Due to the range of variation of Csand, 19–96%, and organic C, 0–5.7% (original data, not listed; average data in Table 2), the water retention capacity (h) in test samples varied considerably (Table 4). In these soils (Table 2) the average content of Csand (51%) was five times higher than that for SiCla (10%) and the range of variation of Csand (77%) was nearly three times more than that of SiCla (28%). According to additional analysis of data (not listed) at potentials between h = -3.2 kPa and h = -10 kPa in all three models the variation of Csand explained a high proportion of the variation in the water content (average r² = 0.78). Thus the volume of all pores with capillary equation equivalent diameter less than the value corresponding to potential h strongly depended on amount of Csand particles. As the average explanation capacity of SiCla of the variation of water content at h = -100 kPa for the three models was high (r² = 0.77) the proportionally very small textural fraction SiCla explained to a high degree the volume of all small pores at lower potentials. The single density related variables explained the variation of water content at potentials between h = -3.2 kPa and h = -100 kPa by r² = 0.36 - 0.60.

The large proportion of Csand particles in test soils was the main reason to high prediction accuracy of the models. In contrast in soils having high proportions of silt and clay (SiCla) particles, where several properties affect the same pore size region and thus more variables were needed to achieve the same prediction accuracy (Gupta and Larson, 1979). The generally used textural fractions of sand, silt and clay could be simplified in the models for coarse podzols to Csand and to SiCla.

Soil-forming processes markedly influenced D_b, as well as soil C and oxalate extractable Al and Fe contents and, consequently, also P_tot. The functional relationship and high correlation (Tables 7, 8) between these variables (excluding the oxalate extractable Al and Fe contents of the eluvial horizon) was another reason for the good explanation capacity of the water retention equations presented. However, it appeared that a simultaneous increase of P_tot and secondary structure was essential for this to occur. This is indicated by the correlation between P_tot and AlFe (|r| = 0.93, n = 64), Table 8, and density related variables (|r| = 0.88–0.95, n = 86), Table 7. In podzolic forest soils, the close connection between D_b, P_tot, and C content was already well-known from earlier studies (Rawls, 1983; Tamminen and Starr, 1994; Westman, 1990).

The combined effect of the soil-forming processes and texture on soil structure was very regular both over the area and with depth for model soils, and was most obvious in the very coarse soil of Site 1 (Table 2 and 4). In coarse podzolic soils of this study, at textural pore region potentials, water content decreased with increasing proportion of Csand fraction, and increased in accordance with the degree of formation of the secondary structure. The variables associated with these two soil properties (Csand; D_b, P_tot, and AlFe) behaved in the models in a linear and additive way. With respect to AlFe, however, the relations between model parameters were not universal, since Model 3 gave less accurate estimates for the validation soils than the two other models (Table 10).

At lower potentials in Model 3 (Table 6) the regression coefficient of AlFe is about 10 times higher than that of SiCla, and thus by the same mass of each, AlFe contributes considerably more to (h) than SiCla. Enrichment of Al and Fe compounds changes soil structure efficiently, as it adds thin coatings on the surface of the already existing particles and between them (FitzPatrick, 1983; Scheffer and Schachtschabel, 1989). The roles of extractable Al and Fe in the Pearson's correlation matrix (Tables 7, 8) differed somewhat from each other (r = 0.72). Variable Al explained the water retention better than Fe (Table 9), although the range of Al and Fe content in test soil samples was approximately the same (Table 2).

In the study of Rawls et al. (1982) of fine textured soils where the amount of clay was high, the water retention at -1500 kPa was explained by clay content and the amount of organic material (R = 0.80), while the amount of silt had no effect. In the coarser soils of this study, the water content at -1585 kPa varied from 0.01 to 0.14 m³ m^-3 and the clay content varied from 0.1 to 11.5% (original data not listed; average data in Tables 2, 4). In spite of its high variation in the soil samples, the variable Clay consisted of a very small proportion of the texture, and the influence of Clay on water content was covered by other variation.

In the validation materials, predicted water content values were systematically lower than the measured ones. As mentioned earlier in the Methods section, techniques used with the test and validation samples differed with respect to soil compaction. In coarse-textured soils, the volume of large pores between sand grains can easily be changed even by slight compaction. This assumption was supported by the result that at saturation, Model 1 gave accurate test results, while at lower potentials there was a systematic difference between predicted and measured water contents. At saturation, Model 1 (Table 6) was based mainly on D_b (R² = 0.91), and thus both the water content and the explaining variable included the effect of possible compaction. At potentials between h = -3.2 kPa and h = -100 kPa, the water content was explained to a high degree by the mass proportions of textural fractions, which are independent of compaction (see results section, Multiple Linear Regression Models).

The water retention curve can be interpreted to represent pore size distribution since, according to capillary equation, each value of potential corresponds to an equivalent pore diameter. The comparison of the average pore-size distribution (Fig. 2 , Table 4) of the upper illuvial horizon (0.07 m) with the parent material (0.50 m) soil at Site 1 and Site 3 revealed the different effects of the soil-forming processes on soils of opposite texture. According to the definition of Luxmoore (1981), the limit between micropores and larger (macro, meso) pores is 0.01 mm of pore diameter, which in our data is closest to the 0.03 mm equivalent diameter of -10 kPa potential. At Site 1, the amount of large pores had not changed much from parent soil at the depth of 0.50 m (0.37 m³ m^-3) to upper illuvial horizon at 0.07 m (0.35 m³ m^-3). The decisive change was the increase of volume of micropores from 0.04 m³ m^-3 at the depth of 0.50 m to 0.14 m³ m^-3 at the depth of 0.07 m. This implies that after the channel flow and the flow due to gravitational force have ceased in the soil profile, the water storage (in volume units) in the upper illuvial horizon is over three fold compared with parent soil.

View larger version (29K): [in this window] [in a new window]   Fig. 2. Water retention for Site 1 (S1) high in coarse sand (Csand), and for Site 3 (S3) high in sand and silt, for upper illuvial horizons (0.07 m) and in parent material (0.50 m).

	CONCLUSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
In the studied Fennoscandian podzols, effects of enriched fine materials together with increased total porosity were superimposed on a coarse single-grained basic structure. Accordingly, the models of these soils were relatively simple. Similar to the models of Gupta and Larson (1979), (h) and related parameters varied over a large range of values, thus contributing to the high degree of explanation and making possible the estimation of water content also at potentials higher than -10 kPa. Even if the small number of validation soils covered a large range of soil properties in Fennoscandian podzols, the applicability of Models 1 and 2 should be tested with a wider set of podzolic soils in this area. It was shown in Model 3 that oxalate extractable Al and Fe, which were accumulated in the illuvial horizons, affected water retention but the validation of Model 3 did not succeed. Modelling using the sum of extractable Al and Fe should be investigated further.

The derived statistical relationships were explained in terms of the physical and chemical characteristics of these soils by considering the effects of various soil-forming processes on pore-size distribution. The results support the conclusion made on some other soils that only models constructed for a subgroup of soils with similar genesis, mineralogy, and morphology can provide a high prediction accuracy (e.g., Puckett et al., 1985; Williams et al., 1992).

	ACKNOWLEDGMENTS

 
The authors would like to thank Silja Aho, M.Sc., for laboratory assistance, and Jukka Pumpanen, M.Sc., for help in the field work. The work was supported by research funds of the University of Helsinki.

	NOTES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 
¹ Since at Sites 1 and 2 the thicknesses of soil layer above the bedrock has not been measured, we gave the depth of ground water table, which tells the minimum thickness of this layer.

Received for publication June 21, 1999.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS NOTES RESULTS DISCUSSION CONCLUSION REFERENCES

 

• Addiscott, T.M., and R.J. Wagenet. 1985. Concepts of solute leaching in soils: a review of modelling approaches. J. Soil Sci. 36:411–424.
• Ahuja, L., J. Naney, and R. Williams. 1985. Estimating soil water characteristics from simpler properties or limited data. Soil Sci. Soc. Am. J. 49:1100–1105.[Abstract/Free Full Text]
• Arya, L.M., and J.F. Paris. 1981. A physicoempirical model to predict the soil moisture characteristics from particle-size distribution and bulk density data. Soil Sci. Soc. Am. J. 45:1023–1030.[Abstract/Free Full Text]
• Bohn, H.L., B.L. McNeal, and G.A. O'Connor. 1985. Soil chemistry. 2nd ed. John Wiley and Sons, New York.
• Browne, B.A. 1995. Toward a new theory of podzolization. p. 253–273. In W.W. McFee and J.M. Kelly (ed.) Carbon forms and functions in forest soils. SSSA, Madison, WI.
• Cosby, B.J., G.M. Hornberger, R.B. Clapp, and T.R. Ginn. 1984. A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res. 20:682–690.
• Elonen, P. 1971. Particle-size analysis of soil. Acta Agric. Fenn. 122:1–122.
• Food and Agriculture Organization of the United Nations-Unesco. 1981. Soil map of the world. 1:5,000,000 Volume V Europe. FAO, Paris.
• FitzPatrick, E. 1983. Soils. Their formation, classification and distribution. Longman, New York.
• Food and Agriculture Organization of the United Nations-Unesco. 1990. Soil map of the world. Revised legend. World Soil Resources Report no. 60. FAO, Rome.
• Freund, R.J., and P.D. Minton. 1979. Regression methods, a tool for data analysis. Marcel Dekker, New York.
• Gupta, S., and W. Larson. 1979. Estimating soil water retention characteristics from particle size distribution, organic matter percent, and bulk density. Water Resour. Res. 15:1633–1635.
• Hintikka, V. 1972. Wind induced root movements in forest trees. Commun. Inst. For. Fenn. 76:1–56.
• Jarvis, N.J., P.-E. Jansson, P.E. Dik, and I. Messing. 1991. Modelling water and solute transport in macroporous soil. I. Model description and sensitivity analysis. J. Soil Sci. 42:59–70.
• Jarvis, N.J., and I. Messing. 1995. Near-saturated hydraulic conductivity in soils of contrasting texture measured by tension infiltrometers. Soil Sci. Soc. Am. J. 59:27–34.[Abstract/Free Full Text]
• Kareinen, T., A. Nissinen, and H. Ilvesniemi. 1998. Analysis of forest soil chemistry and hydrology with a dynamic model ACIDIC. Acta For. Fenn. 262:1–42.
• Liski, J., and C.J. Westman. 1995. Density of organic carbon in soil at coniferous forest sites in southern Finland. Biogeochemistry 29:183–197.
• Luxmoore, R.J. 1981. Micro-, meso-, and macroporosity of soil. Soil Sci. Soc. Am. J. 45:671–672.[Free Full Text]
• Neter, J., W. Wasserman, and M.H. Kutner. 1990. Applied statistical regression models. 3rd ed. Richard D. Irwin, Homewood, IL.
• Petersen, L. 1976. Podzols and podzolization. DSR Forlag, Copenhagen.
• Puckett, W., J. Dane, and B. Hajek. 1985. Physical and mineralogical data to determine soil hydraulic properties. Soil Sci. Soc. Am. J. 49:831–836.[Abstract/Free Full Text]
• Rawls, W.J. 1983. Estimating soil bulk density from particle size analysis and organic matter content. Soil Sci. 135:123–126.
• Rawls, W.J., D.L. Brakensiek, and K.E. Saxton. 1982. Estimation of soil water properties. Trans. ASAE 25:1316–1320, 1328
• Rawls, W.J., D.L. Brakensiek, and B. Soni. 1983. Agricultural management effects on soil water processes, Part I. Soil water retention and Green and Ampt infiltration parameters. Trans. ASAE 26:1747–1752.
• Rawls, W.J., T.J. Gish, and D.L. Brakensiek. 1991. Estimating soil water retention from soil physical properties and characteristics. Adv. in Soil Sci. 16:213–234.
• Saxton, K., W. Rawls, J. Romberger, and R. Papendick. 1986. Estimating generalized soil-water characteristics from texture. Soil Sci. Soc. Am. J. 50:1031–1036.[Abstract/Free Full Text]
• Scheffer, F., and P. Schachtschabel. 1989. Lehrbuch der Bodenkunde. 12th ed. p. 159, 248, 336. Ferdinand Enke, Stuttgart, Germany.
• Tamminen, P., and M. Starr. 1994. Bulk density of forested mineral soils. Silva Fenn. 28:53–60.
• Tiktak, A., and H.J. van Grinsven. 1995. Review of sixteen forest-soil-atmosphere models. Ecol. Modell. 83:35–53.
• Wang, C. 1981. Extractable Al, Fe and Mn. p. 98–105. In J.A. McKeague (ed.) Manual on soil sampling and methods of analysis. Can. Soc. of Soil Sci., Ottawa, ON, Canada.
• Westman, C.J. 1990. Soil physical and physico-chemical properties of Finnish upland forest sites. Silva Fenn. 24:141–158.
• Westman, C.J., H. Fritze, H.-S. Helmisaari, H. Ilvesniemi, T. Lehto, J. Liski, M. Mecke, J. Pietikäinen, and A. Smolander. 1996. Carbon in boreal coniferous forest soil. p. 226–227. In J. Roos (ed.) The Finnish Research Programme on Climate Change. Final report, Publications of the Academy of Finland. Edita, Helsinki, Finland.
• Williams, R., L. Ahuja, and J. Naney. 1992. Comparison of methods to estimate soil water characteristics from soil texture, bulk density, and limited data. Soil Sci. 153:172–184.

This article has been cited by other articles:

				K. A. Winfield, J. R. Nimmo, J. A. Izbicki, and P. M. Martin Resolving Structural Influences on Water-Retention Properties of Alluvial Deposits Vadose Zone J., May 26, 2006; 5(2): 706 - 719. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (9) Citing Articles via Google Scholar Google Scholar Articles by Mecke, M. Articles by Ilvesniemi, H. Search for Related Content PubMed Articles by Mecke, M. Articles by Ilvesniemi, H. GeoRef GeoRef Citation Agricola Articles by Mecke, M. Articles by Ilvesniemi, H. Related Collections Structure and Properties Soil Physics