Durner and Lipsius, HSA077b: Determining soil hydraulic propertiesEncyclopaedia of Hydrological Sciences
Title
HSA077b Determining Soil Hydraulic Properties
Contributor names
Wolfgang Durner1 and Kai Lipsius2
Affiliation
1 Institute of Geoecology, Braunschweig Technical University, Langer Kamp 19c, 38106 Braunschweig, Germany,
2 Institute of Geoecology, Braunschweig Technical University, Langer Kamp 19c, 38106 Braunschweig, Germany,
Keywords
Review; soil hydrology; unsaturated zone; vadose zone; hydraulic properties; measurement; estimation; hydraulic functions; hydraulic conductivity; water retention characteristic; one-step outflow; multistep outflow, evaporation method, parameter estimation, laboratory experiments, inverse modeling.
Abstract
1Introduction
1.1The purpose of hydraulic measurements
1.2The challenge of determining soil hydraulic properties
1.3Classification of methods
2Methods to determine soil hydraulic functions
2.1Indirect estimation of hydraulic functions
2.1.1Estimation by pedotransfer functions
2.1.2Conductivity estimation by statistical estimation methods
2.1.3Estimation of constitutive relationships by pore-network models
2.2Laboratory methods
2.2.1Hydrostatic equilibrium methods
2.2.2Steady-state laboratory methods
2.2.3Transient laboratory methods
2.3Field methods
2.3.1Internal Drainage Method
2.3.2Steady flow infiltration methods
2.3.3Pressure Ring Infiltration
2.3.4Tension Disc Infiltration
2.3.5Infiltration from wells and bore holes
2.3.6Use of tensiometer response to measure soil hydraulic conductivity
2.4Integrated determination by inverse modelling
3Concluding remarks
4References
Abstract
Hydraulic properties are required for modelling water and solute transport in unsaturated soils. The bottleneck for the successful application of numerical simulation models lays usually in their parameter estimation requirements. Methods to determine hydraulic properties can be classified into indirect and direct approaches. Indirect methods encompass the estimation of hydraulic properties by pedotransfer functions from more easily measured soil properties, and the prediction of the unsaturated hydraulic conductivity function from the water retention curve. In direct methods, observations of flow attributes from laboratory or field experiments are evaluated. This article reviews common methods to estimate the hydraulic conductivity function from the water retention characteristic and various direct measurement techniques in the laboratory and the field. We conclude with an outlook on contemporary developments in measurement techniques, stressing the key role of inverse modelling of experiments to derive optimum hydraulic properties and the importance of a future combination of non-invasive measurement techniques with inverse modelling by stochastic data fusion.
1Introduction
A proper characterization of water flow processes is needed in nearly all basic and applied aspects of soil, water, nutrient, and salinity management research (van Genuchten et al., 1999b). Water flow in soils is typically described with the Richards equation (Richards, 1931)
(1)
where t is time [T], z is a spatial coordinate [L], positive downwards, h is the matric potential, expressed as pressure head [L], C(h) is the specific water capacity [L–1], defined by the change of the volumetric water content [L3 L–3], with pressure head, , K(h) is the unsaturated hydraulic conductivity [L T–1], and s is a source/sink term [T–1]. The model is completed by appropriate initial and boundary conditions. Since C and K are non-linearly dependent on h, its solution requires generally numerical methods. Equation (1) is derived from the combination of the Darcy-Buckingham equation and continuity considerations in the framework of the continuum theory, and is valid for the measurement scale (Durner and Flühler, HS077). It is also frequently used as process model for water transport at much larger scales. The coefficients C and K are then used as effective properties, which have the same names as for the local scale, but their values are no longer necessarily consistent with the local definition. Determining hydraulic properties is the process of deriving the constitutive relationships (h) and K(), or K(h), as used in equation (1). The relationship (h) is called water retention curve, WRC. The dependence of the hydraulic conductivity, K(h), on water content or pressure head is called “hydraulic conductivity curve”.
This article reviews indirect and direct methods for the determination of soil hydraulic properties. Direct methods are based on flow experiments in the field or with soil samples in the laboratory. They rely on observations of flow attributes, such as water potential, water content or water flux density. These measurements are far from simple. Water content measurement is treated in this encyclopaedia in chapter HS076 (Topp and Ferré, 2005), soil water potential measurement is treated in chapter HS077a (Durner and Or, 2005). Water flux density can be determined at system boundaries, using scales or burettes in the laboratory. For measuring fluxes in situ, there are no reliable and accurate methods. Accordingly, in situ flux measurements are up to date not used for the determination of hydraulic properties (Dirksen, 1999b).
Indirect methods are used to estimate hydraulic properties from more easily measured data, using regression or neural network algorithms. In particular, the unsaturated conductivity function is seldom measured, but commonly estimated from the water retention curve and matched to a single measured conductivity value. Because of the shortcomings of direct measurement procedures, indirect estimation methods are gaining popularity. All estimation procedures, however, need the results of direct measurements as benchmarks for validation. Furthermore, reliable and efficient experimental procedures are critical to improve the understanding of flow and transport processes in variably saturated media, regardless of the advances in the formulation of indirect methods.
A review on measurement methods needs to address aspects of sensor technology, instrumentation, experimental design, techniques to evaluate observed data, scale issues, parameterisation of hydraulic functions, parameter estimation techniques, and uncertainty estimation and propagation. Covering all these issues in the required depth would be far beyond this contribution. So the focus is on principles of indirect estimation procedures (section 2.1), on direct laboratory measurement methods (section 2.2), and on field methods (section 2.3). Topics that are closely related to soil hydraulic measurements are treated in the companying contributions in this encyclopaedia. These are, in particular, soil water potential measurement (Durner and Or, HS077a), water content measurement (Topp and Ferrè, HS076), estimation of hydraulic properties by pedotransfer functions (Schaap, HS078), scale issues (Hopmans and Shoups, HS070), recent developments in parameter estimation techniques (Vrugt and Dane, HS079), and uncertainty propagation in hydrologic models (Brown and Heuvelink, HS081). The definition of hydraulic functions in the framework of the continuum theory, the parameterization of the hydraulic functions including hysteresis, and the issue of scale dependence of hydraulic properties are treated by Durner and Flühler (HS077).
Methods and techniques to measure soil hydraulic properties are described in a series of monographs. The reference with the widest distribution is the classic monograph “Methods in Soil Physics – Physical Methods” of the American Society of Agronomy. The previous edition (Klute, 1986) has recently been revised (Dane and Topp, 2002), and covers almost all practical measurement techniques including soil sampling, uncertainty, and sensor technology. Hydraulic measurement methods are further discussed in some textbooks, e.g., Kutilek and Nielsen (1994), Dirksen (1999a), and Flühler and Roth (2004). The scientific state of knowledge on direct, inverse and indirect measurement methods at the time of the millennium is documented on 1600 pages in the “Proceedings of the International Workshop on Characterization and Measurement of the Hydraulic Properties of Unsaturated Porous Media” (van Genuchten et al., 1999a). Direct measurement methods cover about a fourth of the volumes.
1.1The purpose of hydraulic measurements
Methods to determine hydraulic properties differ with respect to their accuracy, measurement range, difficulty of implementation, and demand for time and capital. Before selecting a specific measurement method, the purpose of the measurements must be manifest. Purposes may be classified into three groups.
(i)Soil hydraulic properties are often needed for a hydraulic classification of soils, in a similar manner as the particle size distribution is used for a textural classification. Knowledge of basic hydraulic properties, such as field capacity or plant available water content, is useful for a variety of purposes, where soil moisture storage, soil wetness (affecting oxygen supply for plant roots and redox state of soil), surface runoff, susceptibility to soil erosion, and other large-scale properties of soils are of interest. Indirect methods are often appropriate for this group (Schaap, HS078).
(ii)Today’s prevalent demand for hydraulic properties is their use in numerical simulation of water transport by Richards‘ equation. Estimation of water recharge through the vadose zone for water balance calculations are a classic application, but more important is now their use for agricultural, ecological and environmental purposes, such as irrigation control, fertilizer management, and contaminant fate modelling. The focus of subsurface models of water and solute transport has increasingly been shifted towards environmental research, with the primary concern on the subsurface fate and transport of various substances, such as nutrients, pesticides, pathogens, pharmaceuticals, viruses, bacteria, colloids and toxic trace elements (Simunek, HS080). The crucial bottleneck for the successful application of these models is their parameter estimation requirements.
(iii)Finally, accurate measurement of soil hydraulic properties is required for a further improvement in the basic understanding of soil hydraulic processes, i.e., in order to test and improve the process knowledge we have. Examples are the understanding and assessment of non-equilibrium phenomena in water flow, further progress in describing hysteresis in soil water flow, and the question up to which scale the Richards equation process model provides an appropriate effective process representation for unsaturated water transport (Durner and Flühler, HS077).
Whereas the demand on accuracy, resolution, precision, and reliability of the measurements is moderate for the first group, it is higher for the second and third. For the second group, we are particularly faced with scale considerations. For the third group, the precision, reliability and validity of measurements are of utmost importance, in order to avoid misconceptions.
1.2The challenge of determining soil hydraulic properties
Determining soil hydraulic properties is demanding for a variety of reasons. Soils are porous media with a three-dimensional arrangement of interconnected voids that form a highly complex pore system. The topology of this system shows, in general, a hierarchical arrangement, with spatial and temporal variability on a multitude of scales. The microscopic properties of the pore system determine the macroscopic hydraulic behaviour. A complete understanding of water flow in soils requires a thorough understanding of processes on scales much smaller than the usual measurement scale and the ability to express effective hydraulic properties at scales much larger than the measurement scale (Durner and Flühler, HSA077; Hopmans and Shoups, HS070).
A specific problem in the determination of soil hydraulic properties lays in the fact that quality control and validation of measurement results is extremely difficult. Contrary to soil chemical analysis, there is virtually no possibility for reliable inter-laboratory comparisons (Dirksen, 1999b). The reasons for this are: (i) as opposed to consolidated porous media, the soil pore system is not a stable structure. Just those parts of the pore system, which control the water transmission near saturation, are most fragile, and there is always a danger that the measurement process itself changes the system (Ghezzehei and Or, 2003). Therefore, repetitive measurements on the same soil sample by different laboratories are impractical. The sampling process itself often causes the most severe disturbance, when an “undisturbed” soil sample is isolated from the natural embedding. (ii) Soils exhibit considerable temporal variability (Mapa et al., 1986; Ahuja et al., 1998; Leij et al., 2002). Thus, measurements at the same site may yield different results if applied at different times (van Es et al., 1999). (iii) Soil is a living organism and the pore system is affected by a variety of interacting biological, chemical and physical processes. Matrix surface properties are variegated and may change dependent on physical, chemical and biological factors, thereby changing the macroscopic hydraulic behaviour. (iv) Spatial variability of hydraulic properties, finally, is probably the biggest problem (Nielsen et al., 1973, 1986). Different measurement methods use different sample volumes and sample numbers, dictated by standard procedures and the apparatus available for the various methods. This implies that, in a comparison of measurement methods, considerable uncertainty about the result of the comparison will always be induced by natural variability (Stolte et al., 1994, Munoz-Carpena et al., 2002).For example, the determination of field-saturated conductivity, Kf, may vary two or more orders of magnitude among different field and laboratory methods.
Testing measurement methods on synthetic soil-like porous media is not a solution of this dilemma. The pore system properties of repacked soil samples are often very different from the properties of an undisturbed soil (Torquato, 2001). It is notable that especially uniform fine sand, being the favourite material used for research purposes in soil physics, has properties that are quite untypical for a structured natural soil. Since the early observations of Kozeny (1927) on particle segregation during packing, the problem of constructing a synthetic porous medium in a fully reproducible manner, with pore system properties comparable to a natural soil, has remained unresolved (Lebron and Robinson, 2003).
An overview on various field and laboratory methods for determining unsaturated hydraulic properties shows that many techniques have been proposed, but most of them are limited to relatively narrow ranges of water potential (h) or water content (. Fig. 1 illustrates this, showing results of five different measurement methods to determine the hydraulic conductivity. The existing experimental procedures all have their own unique advantages and limitations (Gee and Ward, 1999), and selecting the most appropriate measurement for a specific task is usually not trivial.
1.3Classification of methods
Determining hydraulic properties of soil encompasses direct measurements and indirect estimation methods. Because of the shortcomings of direct measurement procedures, indirect estimation methods are gaining popularity. Computers offer the possibility to generate indirect estimates using regression or neural network algorithms. In practice, unsaturated conductivity is seldom measured, but estimated from the saturated conductivity (or preferably an other matching point) and the water retention curve. The principles of indirect estimation procedures are outlined in section 2.1.
Fig. 1: Hydraulic properties of a fluviatile silt loam. Left: Water retention curve measurements. Right: Unsaturated conductivity measurements, determined by five different methods. [Reproduced by permission of the American Soil Science Society from J. .Stolte, J.I. Freijer, W. Bouten, C. Dirksen, J.M. Halbertsma, J.C. Van Dam, J.A. Van den Berg, G.J. Veerman, and J.H.M. Wösten, Soil Sci. Soc. Am. J., 58, 1596-1603 (1994).].
All estimation procedures need the results of direct measurements as benchmarks for validation of their results. Reliable and efficient experimental procedures are critical to improve the understanding of flow and transport processes in variably saturated media, regardless of the advances in the formulation of indirect methods. Experiments for measuring hydraulic properties are based on hydrostatic, steady state, or transient flow conditions. The methods can be grouped in those that aim at measuring (i) water retentivity, (ii) saturated conductivity, (iii) unsaturated conductivity or water diffusivity, and (iv) simultaneously retentivity and conductivity. We may further distinguish between laboratory methods and field methods. The reason for treating laboratory (section 2.2) and field experiments (section 2.3) in this review in separate sections is less motivated by the different scale, but by the fact that in the laboratory a much better control of boundary conditions, fluxes across boundaries, and integral water content measurements can be achieved.
Finally, methods differ in how the observations of flow attributes, such as pressure heads, water contents or fluxes are evaluated. In the early stages of soil physics, methods have been developed for determining the water retention curve or the saturated conductivity (Gardner, 1986). Determination of the water retention characteristic can be done directly, by pairing water content and water potential measurements in the laboratory or in the field. The determination of saturated conductivity is achieved by a closed-form inversion of the flow equation. Analytical solutions for the unsaturated conductivity can be obtained in the laboratory by a series of consecutive unit-gradient experiments, where a constant flux or pressure head is applied at the top of a sample, and a corresponding suction at the bottom (Dirksen, 1991, 1999a, 1999b). These methods are conceptually straightforward and easy to implement. Their main disadvantage is that they take a long time, and are therefore tedious and expensive. In order to achieve hydraulic equilibrium, the sample sizes must be kept small and sizes are often below the representative elementary volume, REV (Durner and Flühler, HS077). Accordingly the resulting properties are generally highly variable.
Measurement of unsaturated hydraulic conductivity and water diffusivity poses greater obstacles. The methods are based on solving the inverse problem, where a model of the flow process is optimized to match observations (Russo et al., 1991; Hopmans et al., 2002; Vrugt and Dane, HS079). For some simplified cases, solving the inverse problem is accomplished by closed-form solutions, such as the determination of diffusivity from a quasi-analytical solution of the Richards equation. In most other cases, solution of the inverse problem can only be achieved by fitting numerically simulated data to observations. This requires the use of non-linear parameter estimation techniques. In doing this, the logical next step is to simultaneously estimate retentivity and conductivity parameters by inverse modelling.
2Methods to determine soil hydraulic functions
2.1Indirect estimation of hydraulic functions
2.1.1Estimation by pedotransfer functions
In pedotransfer functions, the water retention and conductivity functions are derived from more easily measured soil properties, such as soil texture, bulk density, and organic matter content. Methods to derive soil hydraulic properties indirectly include multiple regression, classification, and neural network predictions (Genuchten et al., 1999a). Comparisons of indirectly determined hydraulic properties to directly measured properties are manifold and remain a topic of ongoing research (Schaap, HS078). In general, it is found that the accuracy of pedotransfer estimates of hydraulic functions for characterizing field average is comparable to simple direct measurements, if spatial variability is considered. Depending on the desired use of these properties, indirect methods have evolved to a point where they provide reliable answers for many problems. However, further improvement of indirect methods hinges on experimental data, which must be obtained with direct procedures.