FP-Experiment Water and solute transport in porous media Klaus Schneider, Ute Wollschläger, and Kurt Roth Institut für Umweltphysik, Universität Heidelberg Version: April 2008 Location of the experiment: INF 229, room 204 Contents 1 Overview 2 2 Part I: Estimation of hydraulic properties by inverse modelling 3 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.3 Dynamics of water movement . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.1 Hydraulic properties . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.2 Experimental determination of hydraulic properties by inverse mod- elling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.4 Multi-Step Outflow Experiment . . . . . . . . . . . . . . . . . . . . . . . . 11 2.4.1 Setup of the experiment . . . . . . . . . . . . . . . . . . . . . . . . 11 2.4.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.4.3 Schedule of the practical course (part I) . . . . . . . . . . . . . . . 13 2.5 What to learn in this experiment . . . . . . . . . . . . . . . . . . . . . . . 15 2.6 Questions you should be able to answer before starting . . . . . . . . . . . 16 3 Part II: Measurement and parameterisation of solute transport 17 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.2.1 Convection-Dispersion model (CD) . . . . . . . . . . . . . . . . . . 17 3.2.2 Extension to multi-domain models (MIM) . . . . . . . . . . . . . . 20 3.3 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.3.1 Setup of the experiment . . . . . . . . . . . . . . . . . . . . . . . . 20 3.3.2 Schedule of the practical course (part II) . . . . . . . . . . . . . . . 21 3.4 What to learn in this experiment . . . . . . . . . . . . . . . . . . . . . . . 23 3.5 Questions you should be able to answer before starting . . . . . . . . . . . 23 1 Overview Background: Porous media are part of our daily life and include technical devices like filter materials and fuel cells as well as natural systems like geologic formations and soils. An important characteristic of porous media are capillary forces that play a dominant role for the dynamics of such materials. It is a challenge for physicists to model fluid flow and transport of dissolved chemicals in porous media e.g. in order to predict the behaviour of a system as response to various boundary conditions or to construct porous media with the properties desired. The focus here is on soil, which is typically characterised by a complex structure. It consists of minerals, organic material and the pore space in-between. Moreover, this structure is typically not static but continuously changed by plants, animals and various physical boundary conditions due to temperature and precipitation. Crucial questions that require a quantitative understanding of the processes in porous media are forexample: Howtostoreinfiltratingwater insoiltowarrant thegrowthofplants? Howto prevent pollution in the environment of waste disposals? How to treat soils contaminated by anthropogenic wastes? What is the role of soils for global mass and energy fluxes? Subject of the course: In principle, the microscopic structure of the pore space deter- mines the behaviour of porous media. However we are typically not in the position to describe this 3-dimensional structure in all detail. We aim at modelling and predicting fluid flow and solute transport in porous media at a scale of some 10−1m within a soil up to some 103m within a catchment. This range of scales is significantly larger than the typical size of pores in soil that are in the range of 10−6m to 10−3m. (Just think about the amount of computer memory you would need to store the structure of 1dm3 soil at a resolution of 1µm). This leads to the introduction of “effective” models based on “effective” material properties that translate the effects of the microscopic structure into macroscopic properties. In the first part of this course we only consider water flow in porous media and we intro- duce two effective material properties: 1) The relation between the energy density of water (capillary pressure) and the water saturation which is referred to as water characteristic and 2) the relation between the water saturation of the media and its hydraulic conduc- tivity which is termed the hydraulic conductivity function. An experiment is performed to determine these basic material properties which are essential for modelling the dynamics of fluids in porous media at larger scales. In the second part we consider solute transport through porous media which also may be related to macroscopic material properties as for instance a dispersion tensor. Another experiment is performed to identify a convenient transport model and to determine the related parameters. Preparation: To prepare this course, this short script should be sufficient. Here, the basic theory of water dynamics in porous media is described and some practical de- tails on the set-up of the experiment are provided. Finally the evaluation of the mea- sured data through inverse modelling is described. For a deeper study of the subject you may consider Roth (2006) at www.iup.uni-heidelberg.de/institut/forschung/ groups/ts/soil_physics/students/lecture_notes05. 2 1cm (a) (b) (c) Figure 1: 2D and 3D structure of soil at the pore scale with a resolution of 0.04mm/pixel obtained from serial sections (a and b). Structure of porosity including few large pores (black spots) at a larger scale with a resolution of 0.4mm/pixel obtained by X-ray tomography (c). 2 Part I: Estimation of hydraulic properties by inverse modelling 2.1 Introduction To model the dynamics of water in porous media at a scale much larger thanthe individual pores, we introduce so called “effective material properties” which are based on variables that are visible at the larger scale. These properties and the related variables are obtained by averaging over the complicated porous structure at the microscopic scale. Thereby, the details of the porous structure are lost, but the aim is to preserve those properties that are important to quantitatively understand the processes at the larger scale – this is why we call them “effective”. At the pore scale (Figure 1b), flow and transport processes may be described using clas- sical concepts of fluid dynamics where the acceleration of a fluid element is related to the forces acting on it, most importantly to gravity, pressure gradients and friction. This leads to the well-known Navier-Stokes equation. This equation is not invoked here, however, be- cause (i) the inertia term can be neglected for the low flow velocities typically encountered, 3 (ii) implementing the boundary conditions for the pore space is next to impossible and the computational effort is excessive, and (iii) for most phenomena a detailed pore-space description is not required. Hence, we go for a continuum description that is obtained from averaging. In formal analogy to the averaging used in thermodynamics, the dynamics of water flow is described over a representative region of the pore space. The remaining task then is to estimate pertinent effective material properties. 2.2 Theory The complicated structure of the soil cube in Figure 1b is replaced by “effective” measures and state variables with respect to the fluids. This can be interpreted as averaging over microscopic properties or upscaling from microscopic to macroscopic properties. Consider a porous medium which consists of the solid and undeformable “soil matrix”, i.e. the material the soil is made of, and the intermediate pore space. Pores are filled with a fluid, e.g. water or air. At every point of the pore space, there is precisely one phase: soil matrix, water, or air. State Variables We define the state of a fluid element of a fluid i in a porous medium by the following state variables: its height z , its pressure p , its temperature T and the concentrations C i i i ij of solute chemicals j. In the following, only pure fluids are considered: C = 0. ij Fluid Content The distribution of a phase i (soil matrix, air or water) is averaged over a volume ∆V around the spatial location r. One obtains the volume fraction θ := ∆V /∆V . (1) i i It specifies the fraction of the volume ∆V around r which is taken up by the phase i. θ (r) is a continuous function. Note that at the pore scale, every point belongs exactly to i one phase, while in contrary, in the macroscopic description, one point belongs to several phases. The volume fraction of the pore space is called porosity φ = θ + θ = 1 − θ . This a w m corresponds to the water content at complete water saturation (θ = 0, θ = φ). If we a w know the density of the solid material ρ (for quartz it is 2.65g/cm3) we can calculate the s porosity from bulk density: φ = 1−(ρ /ρ ). b s Representative Averaging Volume The volume ∆V should be chosen such that θ does not depend on the actual size or shape i of ∆V, i.e. that it is large enough to contain all the microscopic heterogeneities. If ∆V is about the size of a pore, θ will hardly differ from the microscopic description. When ∆V i is enlarged, the average value changes abruptly, depending crucially on the choice of ∆V, and eventually becomes stable if ∆V is above some critical value ∆V . This is the point c where the averaging begins to make sense. Such a volume which leads to a representative value of θ is called Representative Ele- i mentary Volume (REV). If no REV can be found, a reasonable averaging is not possible. 4 Potential The density of potential energy ψ (r) of a fluid i is defined by the energy that is necessary i to move a unit volume of fluid from the reference state z = z , p = p , T = T to the state 0 0 0 at position r. p is normally chosen to be the ambient air pressure. 0 By convention, z points downwards into the ground. In an isothermal situation with no dissolved chemicals, only pressure and gravitation contribute to the energy density. Considering an incompressive medium we obtain ψ (r) = p (r)−p −(z −z )ρ g . (2) i i 0 0 i =:ψim =:ψig | {z } | {z } ψ is called matric potential of the fluid i. In unsaturated porous media, this poten- im tial describes the energy densities associated with the boundary layers solid-liquid and liquid-gas, i.e. caused by capillary forces. In mineral soils, the solid soil matrix is typi- cally wettable so that the solid material is completely covered by a thin water film (i.e. the energy density of the boundary layer solid-liquid is smaller than that of liquid-gas). Hence, the matric potential of water ψ is only determined by the interface liquid-gas σ m wa (0.0725Jm−2 for pure water at 20◦C). In the following, only water is considered as fluid and the fluid index i is omitted. Therewith the total energy density below is denoted by ψ (w for water). w At equilibrium, when the free energy is minimal, the surface area of the liquid-gas interface is also minimal and one can show that this surface is described by the Young- Laplace Equation (Landau and Lifschitz, 1991) 1 1 ψ = p −p = σ + , (3) m w a wa r r (cid:20) 1 2(cid:21) where p and p are the pressures in water and air, respectively, and r , r are the principal w a 1 2 radii of curvature (the two radii of a local elliptic approximation of the curved interface area). We use the convention that the sign of the radius is negative if it is within the air phase. In the simplest case of a cylindrical capillary (3) reduces to ψ = 2σ /r. In the m wa convention used here, ψ is negative if the fluid is bound (p < p ) and positive if it is m i 0 free. The energy density ψ [Jm−3] equals the pressure jump across the interface p −p m w a [Nm−2]. In soil physics, often the energy density per unit weight is used: ψ m h := (4) m ρ g w h is called “hydraulic head” which is usually expressed in “cm water column” [cmWC]. The idea is that h can be interpreted as equivalent height of a fluid column; negative values correspond to a hanging column. Additional remarks 1. We restrict the description of the matric potential to the most simple case. In reality the situation is much more complicated. In particular, σ is not a constant wa but depends on temperature and on the composition of the water phase (dissolved chemicals). Moreover, thewettability ofsoilconstituents mayvaryintimeandspace. 5 Finally, σ changes with the thickness of the water layer when going to very dry wa conditions. 2. Besides gravity and capillarity, there may be additional forces: osmotic forces due to gradients in salt concentration, mechanical forces in case the solid material is deformable (this is the rule especially in clay rich soils), and the air pressure may be locally increased when air is entrapped by the water phase. 2.3 Dynamics of water movement Flux law In porous media, typical pore dimensions and typical velocities are so small compared to the viscosity of water and the time scale of typical external forcing that there is no turbulence. With this prerequisite, the volume flux of water j [ms−1] is linearly related w to the gradient of the potential energy density ψ of the water, w j = −K(θ)∇ψ , (5) w w where the tensorial quantity K(θ) is the hydraulic conductivity that depends on the vol- umetric water content and ψ is given by (2) applied for water. With decreasing water w content, less area is available for water transportation. As large pores are drained first and flow velocities are higher in larger pores, the conductivity rapidly decreases with water content. K(θ) is a material property. In a homogeneous medium, the tensor K reduces to a scalar K. This flux law is called the Buckingham-Darcy law. Notice the formal analogy with flux laws for heat (Fourier), dissolved substances (Fick), and electric charge (Ohm). Richards’ Equation We formulate the conservation of mass of water – actually of volume since water may be considered as incompressible – as ∂ θ+∇·j = 0 , (6) t w where θ is the volumetric water content. Inserting the flux law (5) into the conservation equation yields Richards’ equation ∂θ ∂ψ m −∇·[K(θ)∇ψ ] = 0 (7) w ∂ψ ∂t m for the dynamics of water movement. 2.3.1 Hydraulic properties Soil water characteristic The matric potential ψ = p − p is related to the curvature of the water-air interface m w a via the Young-Laplace Equation (3). This radius cannot be smaller than the radius of the pore the interface is located in. 6 6.6 mm = (cid:0)1 = (cid:0)38 hPa = (cid:0)16 hPa = (cid:0)11 hPa = (cid:0)8 hPa Figure 2: Distribution of different phases (water: grey, air: black) within a 2D section through a porous media at different matric potentials ψm. Note the different radii of curvature of the interfaces according to the Young-Laplace equation (3). If ψ is made more negative (e.g. by changing ∆p by applying an external pressure), m the radius r of the interface must become smaller according to (3). If r gets smaller than the radius r¯ of the pore i, r < r¯, there cannot be an interface any more, and the pore will i i be drained. This point is called the air entry point. This line of thoughts reveals that the water content θ is a function of the matric potential ψ : θ = θ (ψ ). This relation is w m w w m illustrated in Figure 2: water in wettable porous media tends to fill the small pores while air resides in the large pores. Figure 3 shows typical examples for a loamy and a sandy soil. The loamy soil has a higher water content at saturation, θ , because of a higher porosity. Going from complete s saturation to drier conditions by lowering the matric potential the water content of the sandy material decreases very rapidly below a potential of h ≈ −30cmWC. This is m 7 ) C℄ W ) m 4 m/h℄ 0 [ m [ h K g((cid:0) 3 og( -5 o l l y ntial 2 tivit -10 te 1 u o d p n -15 tri 0 Co a m 0 0.20 0.40 0 1 2 3 4 water ontent(cid:18) matri potentiallog((cid:0)hm [ mWC℄) (a) Water characteristic (b) Hydraulic conductivity function Figure 3: Typical hydraulic properties of a loamy soil (solid line) and a sandy soil (dashed line). The thin dashed line represents a sand with larger grains and thus, larger pores. because this sand is a rather homogeneous material where all pore sizes are within a narrow range. According to the Young-Laplace equation (3) and using (4) we find an equivalent pore radius for h = −30cmWC of r ≈ 0.05mm. When going from −30cmWC m to about −100cmWC the water within these pores is removed. In the loamy soil the water content decreases more continuously with decreasing water potential, meaning that the pore size distribution is much broader compared to the sand. Clearly, the shape of the soil water characteristic is an effective description of the underlying pore structure. However, it should be noted that not only the pore size distribution determines the shape of θ(ψ ) m but also the connectivity of the pores. Note that we assume that there is an instantaneous equilibrium between θ and ψ which m is only the case at high water contents and herewith small resistivity to water flow. In the dry range however, water in adsorbed films is barely mobile and equilibration has to be reached through vapour diffusion which is a very slow process. Thecomplicatedconnectivity ofporesinsoilsalsoleadstothephenomenon ofhysteresis. Tounderstandthisphenomenon,lookatasingleporeinporespace(Figure4a–c). Initially, the poreissaturated withwater. Now, pressure atthelower end oftheporeiscontinuously decreased. The water-air interface stays constant until the potential drops below ψ which 1 corresponds to the radius of the smaller capillary according to (3). Reducing the pressure any further empties the whole pore: the possible radii in the cavity are too large to sustain the interface as the pressure difference is too high. This leads to Figure 4 (b). At this point the process is reverted and the pressure is increased again. This does not lead back to (a); instead, at ψ , the configuration (c) will be found, which has the same 2 curvature radius as (a) but a different water content θ. The water characteristic is often described by the parameterisation of van Genuchten (van Genuchten, 1980) which can be fitted to a wide range of different materials: θ(ψ )−θ Θ(ψ ) = m r = [1+[αψ ]n]−1+1/n (8) m m θ −θ s r with the parameters Θ [-] water saturation [0,1] 8 drainage wetting (a) (b) (c) ψ m ψ 1 ψ 2 θ (d) Figure 4: Effect of pore connectivity on the hysteretic behaviour of the water characteristic shown for an idealised pore: at ψ1 > ψm > ψ2 we can find very different water contents during drainage ((d), blue curve) and wetting (red curve) at the same matric potential (same interface radius). Subframes (a)–(c) show the water content of that pore at the different potentials: (a) shows the pore at ψm > ψ1, sub-frame (b) at ψm < ψ1 and sub-frame (c) at ψ1 > ψm > ψ2 after re-irrigation, with the same curvature radius as (a), but different water content. Without the effect of connectivity (in case of parallel capillaries) we would find the non-hysteretic dotted line. θ [-] water content at saturation s θ [-] residual water content r α [cm−1] empirical parameter n [-] empirical parameter The residual water content θ represents the small amount of water which is adsorbed r in thin films at low water potentials and which is barely mobile (see Fig. 3 for log(−ψ ) > m 3 log(cmWC) in the sand). The parameter α scales h and thus determines the position m of the curve relative to the axis of matric potential and the parameter n determines the shape of the curve. Thus, α is related to the mean pore size and n to the width of the pore size distribution. However both parameters are fitting parameters without immediate physical meaning. The parameters for the curves in Figure 3 are give in Tab. 1. Hydraulic conductivity function The conductivity K depends on the geometry of pores as well: large connections are better conductors than narrow necks. Pores which are filled with air are not available for liquid water transportation. 9 With decreasing water content, large pores are drained first and the remaining water is located in smaller pores. Because the conductivity K depends on the radius of the flow channel (j ∼ r4), it decreases rapidly with decreasing water content. As additional w effect, the fraction of pores which are used for water conduction decreases as well, and the way between two points A and B which only consists of water filled pores gets longer. Consequently, the conductivity of water is a function of the water content: K = K(θ ). w As θ is a function of ψ , K depends on ψ as well. w m m In Figure 3 the hydraulic conductivity functions K(ψ ) are shown for the sandy and the m loamy soil. Although the porosity of the sand is lower, the hydraulic conductivity at water saturation is higher in the sand. This is because the pores are larger (see θ(ψ)). However, for decreasing water potential the conductivity in the sand drops very rapidly. This is intuitively clear, because at the critical water potential below h ≈ −30 cmWC also the m water content drops significantly. Hence, the shape of θ(ψ ) tells us something about the m shape of K(ψ ) which is also used for the parameterisation. Using arguments derived m from statistical pore models Mualem (1976) proposed to derive the hydraulic conductivity curve through integration of the water characteristic Θh (θ)−1dθ 2 K(Θ) = K Θτ 0 m . (9) 0 "R01hm(θ)−1dθ# R Inserting the van Genuchten parameterisation (8) this leads to K(Θ) = K Θτ 1−[1−Θn/[n−1]]1−1/n 2 (10) 0 h i where now the conductivity is given as a function of Θ. Note that the shape of the curve is determined by the same shape parameter n as for the soil water characteristic. The hydraulic conductivity at water saturation, K , is introduced as a parameter to fix the 0 absolute height of the hydraulic conductivity. Moreover an additional parameter, τ, is used which is termed “tortuosity” and which is thought to account for the change in the topology of the water phase with decreasing water content. However, τ is considered to be a pure fitting parameter which frequently is fixed at a value of 0.5. At this point we arrive at the complete set of hydraulic parameters, describing the hy- draulic properties of porous media. These parameters are listed in Tab. 1 for the examples shown in Figure 3. The aim of the first part of this course is to experimentally determine such a parameter set for a given soil. 2.3.2 Experimental determination of hydraulic properties by inverse modelling Direct measurements of hydraulic properties are very time consuming and, in case of the hydraulic conductivity, also very difficult. In this practical course, we apply an indirect method, the approach of inverse modelling, which is based on the theory given in the previous chapter. The basic idea is that we can describe the dynamics of the system through Richards’ equation (7). Thus, for a given set of initial and boundary conditions, we can predict the behaviour of the system in case the hydraulic properties areknown — which is not the case here. But we can observe the behaviour of the system and choose the hydraulic properties 10