29
Braunschweig Technical University Institute of GeoEcology Measuring soil hydraulic properties of FIMOTUM porous media: Results for medium sand - Report- Andre Peters and Wolfgang Durner 16.10.2006

Braunschweig Technical University Institute of GeoEcology · 2006. 10. 16. · Braunschweig Technical University Institute of GeoEcology Measuring soil hydraulic properties of FIMOTUM

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

  • Braunschweig Technical UniversityInstitute of GeoEcology

    Measuring soil hydraulic properties of FIMOTUM porous media:Results for medium sand

    - Report-

    Andre Peters and Wolfgang Durner16.10.2006

  • Contents

    1 Materials and Methods 6

    1.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

    1.2 Analysis of Data and Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . 9

    1.2.1 Inversion of Richard’s Equation . . . . . . . . . . . . . . . . . . . . . . . .9

    1.2.2 Fit of Equilibrium Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

    2 Results 11

    2.1 Saturated Water Content, Porosity, Bulk Density and Saturated HydraulicConductivity 11

    2.2 Inversion of Richard’s Equation . . . . . . . . . . . . . . . . . . . . . . . . .. . . 11

    2.2.1 First Outflow Branch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

    2.2.2 First Imbibition Branch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

    2.2.3 Second Outflow Branch . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

    2.2.4 Hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

    2.3 Fit of Equilibrium Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

    2

  • List of Tables

    1.1 Matric head at lower boundary of the ceramic plate,hLB, for the multistep out-

    flow/inflow experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

    2.1 Estimated parameters for first drainage branch - column 1. . . . . . . . . .. . . . . 16

    2.2 Estimated parameters for first drainage branch - column 2. . . . . . . . . .. . . . . 17

    2.3 Estimated parameters for first imbibition branch - column 1. . . . . . . . . . . . .. 17

    2.4 Estimated parameters for first imbibition branch - column 2. . . . . . . . . . . . .. 21

    2.5 Estimated parameters for second drainage branch - column 1. . . . . . . .. . . . . . 25

    2.6 Estimated parameters for second drainage branch - column 2. . . . . . . .. . . . . . 25

    2.7 Data for equilibrium fits. First outflow branch. . . . . . . . . . . . . . . . . .. . . . 27

    2.8 Data for equilibrium fits. Second outflow branch. . . . . . . . . . . . . . . .. . . . 27

    2.9 Estimated parameters of equilibrium data - first drainage branch. . . . . .. . . . . . 28

    2.10 Estimated parameters of equilibrium data - second drainage branch. . .. . . . . . . 28

    3

  • List of Figures

    1.1 Experimetal setup of the multistep outflow/inflow experiments . . . . . . . . . . . .7

    1.2 Matric head at lower boundary of the ceramic plate,hLB, for the multistep out-

    flow/inflow experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

    2.1 First drainage branch, column 1: Measured data and fitted with 4 different hydraulic

    properties models. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumu-

    lative outflow forhLB ≥ −25 to show the better performance of the more flexible

    models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

    2.2 First drainage branch, column 2: Measured data and fitted with 4 different hydraulic

    properties models. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumu-

    lative outflow forhLB ≥ −25 to show the better performance of the more flexible

    models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

    2.3 First drainage branch, column 1: Hydraulic properties. . . . . . . . . .. . . . . . . 15

    2.4 First drainage branch, column 2: Hydraulic properties. . . . . . . . . .. . . . . . . 15

    2.5 First outflow branch, both columns: Hydraulic properties. solid lines: column 2;

    dashed lines: column 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

    2.6 First imbibition branch, column 1: Measured data and fitted with 4 differenthydraulic

    properties models. Top: Matric head; Bottom: Cumulative outflow. . . . . . . . .. . 18

    2.7 First imbibition branch, column 2: Measured data and fitted with 4 differenthydraulic

    properties models. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumu-

    lative outflow forhLB ≥ −25 to show the better performance of the more flexible

    models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

    2.8 First imbibition branch, column 1: Hydraulic properties. . . . . . . . . . . . .. . . 20

    2.9 First imbibition branch, column 2: Hydraulic properties. . . . . . . . . . . . .. . . 20

    2.10 Second outflow branch, column 1: Measured data and fitted with 4 different hydraulic

    properties models. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumu-

    lative outflow forhLB ≥ −25 to show the better performance of the more flexible

    models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

    4

  • LIST OF FIGURES 5

    2.11 Second outflow branch, column 2: Measured data and fitted with 4 different hydraulic

    properties models. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumu-

    lative outflow forhLB ≥ −25 to show the better performance of the more flexible

    models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

    2.12 Second outflow branch, column 1: Hydraulic properties. . . . . . . . .. . . . . . . 24

    2.13 Second outflow branch, column 2: Hydraulic properties. . . . . . . . .. . . . . . . 24

    2.14 Second outflow branch, both columns: Hydraulic properties. solid lines: column 1;

    dashed lines: column 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

    2.15 Hysteresis of Hydraulic Properties, Column 1. Solid lines: first outflow; dots: first

    imbibition; dashed lines: second outflow. . . . . . . . . . . . . . . . . . . . . . . . 26

    2.16 Hysteresis of Hydraulic Properties, Column 2. Solid lines: first outflow; dots: first

    imbibition; dashed lines: second outflow. . . . . . . . . . . . . . . . . . . . . . . . 26

    2.17 Retention functions resulting from equilibrium fit. Left: column 1, right: column 2.

    Solid lines: first outflow brach; dashed lines: second outflow branch. .. . . . . . . 27

  • Chapter 1

    Materials and Methods

    1.1 Experimental Setup

    The air dry sand was packed in a 7.3 cm high column with a diameter of 9.4 cm. In order to obtain a

    horizontally homogeneous packed soil column, the sand rinsed through a device with 5 consecutive

    sieves before it reached the column. Whereas the packing procedure the column was slowly lowered

    underneath the packing device so that the distance between device outlet and soil surface was alway

    ≈ 5 cm. This allowed also for a vertically homogeneous packed soil column. The column was

    placed on a porous ceramic plate with a height of 0.7 cm and a saturated hydraulic conductivity

    of 6.0 cm h−1. The ceramic plate in turn was in contact with a burette (Fig. 1.1). Details of the

    experimental setup are given by Zurmühl (1998).

    One pressure transducer tensiometer was placed in the center of the column, i. e. at a height of 3.65

    cm above the plate. To prevent the building of macropores during the experiment due to “escaping

    air bubbles” the soil surface was covered with a metal grid and a weight ofapproximately 1.5 kg.

    The width of the grid was> 3 mm, so that no “extra pores” due to the grid were created. During the

    experiment no setting or building of macropores was observed.

    Before the Multistep-Outflow/Inflow (MSO/I) experiment was started the whole system was slowly

    saturated and then flushed with approximately 5 liters (approximately 25 pore volumes) of degassed

    water from the bottom. Simultaneously the saturated hydraulic conductivity,Ks, was measured. To

    avoid evaporation at the upper boundary the column was covered with a lid.Contact to the atmosphere

    was allowed by a thin needle. For the MSO/I experiment the pressure at the lower boundary was

    controlled as shown in Fig. 1.2 and documented in Tab. 1.1. At each step the pressure change was

    linearly within 30 seconds. In the outflow branches the hydraulic equilibriumwas achieved until the

    pressure at the lower boundary of the ceramic plate,hLB, was−40 cm. This was done to be able to

    predict point values of the pressure heads and water contents without measuring the water contents at

    certain depths.

    The cumulative outflow/inflow was determined by measuring the water volume in theburette. Cumu-

    6

  • 1.1. EXPERIMENTAL SETUP 7

    tensiometer(3.65 cm)

    sand

    pressurecontrol

    valve

    vacuumreservoir

    9.4 cm7

    .3

    pressuretransducer

    positivepressurereservoir

    pressurereservoir

    pressuretransducer

    burette

    porousplate

    air inlet

    top cover

    referencepressure

    line

    metal weightmetal grid

    Figure 1.1: Experimetal setup of the multistep outflow/inflow experiments

    0 50 100 150 200 250 300

    −50

    −40

    −30

    −20

    −10

    0

    10

    time [h]

    h LB [c

    m]

    Figure 1.2: Matric head at lower boundary of the ceramic plate,hLB , for the multistep outflow/inflow experi-ments.

  • 8 CHAPTER 1. MATERIALS AND METHODS

    Table 1.1: Matric head at lower boundary of the ceramic plate,hLB , for the multistep outflow/inflow experi-ments.

    ∆t[h] t[h] hUR[cm]

    0.25 0.25 7.50

    0.25 0.50 3.75

    0.50 1.00 0.00

    0.75 1.75 -5.00

    1.00 2.75 -10.00

    1.50 4.25 -15.00

    2.75 7.00 -20.00

    4.00 11.00 -25.00

    6.00 17.00 -30.00

    10.00 27.00 -35.00

    18.00 45.00 -40.00

    24.00 69.00 -45.00

    24.00 93.00 -50.00

    24.00 117.00 -45.00

    24.00 141.00 -40.00

    24.00 165.00 -35.00

    18.00 183.00 -30.00

    10.00 193.00 -25.00

    6.00 199.00 -20.00

    4.00 203.00 -15.00

    2.75 205.75 -10.00

    1.50 207.25 -5.00

    1.00 208.25 0.00

    0.75 209.00 3.75

    0.50 209.50 7.50

    0.50 210.00 7.50

  • 1.2. ANALYSIS OF DATA AND PARAMETER ESTIMATION 9

    lative outflow and the suction in the center of the column were measured with a frequency of 10 Hertz

    and logged with a frequency of 1 Hertz. The measurements were conducted in two replications.

    1.2 Analysis of Data and Parameter Estimation

    The measured data were split into three branches:

    1. first outflow

    2. first inflow

    3. second outflow

    1.2.1 Inversion of Richard’s Equation

    The data were thinned out, so that approximately 600 data points were left for evaluation. To obtain

    a maximum of information in the remaining data, the data density was highest close to the time when

    the boundary changes occurred.

    The governing equation for the flow simulation was the one dimensional Richards equation:

    dh

    ∂h

    ∂t−

    ∂z

    [

    K(h)

    (

    ∂h

    ∂z− 1

    )]

    = 0 , (1.1)

    whereh [cm] is the matric head,θ [−] is the volumetric water content,z [cm] is the vertical length

    unit, t [h] is time andK(h) [cm h−1] is the hydraulic conductivity function.

    The constitutive relationships, i.e. the soil water retention function,θ(h), and the hydraulic conductiv-

    ity function,K(h), were described by parametric models of different flexibility.θ(h) was described

    by:

    1. the constrained van Genuchten model (Genuchten, 1980) (vG cons):

    Se(h) = (1 + (α|h|)n)−m (1.2)

    with m = 1 − 1n whereSe is the water saturation, defined by

    Se =θ − θrθs − θr

    (1.3)

    andα[cm−1], n [−] andm [−] are curve shape parameters.

    2. the unconstrained van Genuchten model withm as an independent fitting parameter( vG un-

    cons)

  • 10 CHAPTER 1. MATERIALS AND METHODS

    3. the bimodal van Genuchten model (Durner, 1994) (vG bimod)

    Se(h) =2

    i=1

    wi

    [

    1

    1 + (αi|h|, )ni

    ]1− 1ni

    (1.4)

    wherewi [−] are the weighting factors for the subfunctions, subject to0 < wi < 1 and∑

    wi =

    1.

    4. and free form hermite spline functions (Bitterlichet al. , 2004) (ff).

    The hydraulic conductivity function was for the first 3 cases describedby the Mualem model

    (Mualem, 1976):

    Kr(Se(h)) = Sτe

    Se∫

    0

    h−1dS∗

    e (h)

    1∫

    0

    h−1dS∗

    e (h)

    2

    , (1.5)

    that has analytical solutions for cases 1 and 3 (Priesack & Durner, 2006):

    K(h) =

    [

    k∑

    i=1

    wi [1 + (αi|h|)ni ]−mi

    k∑

    i=1wiαi

    [

    1 − (αi|h|)ni [1 + (αi|h|)

    ni ]−mi]

    k∑

    i=1wiαi

    2

    , (1.6)

    wherek is the modality of the function, i.e.k = 1 for case one and 2 for case 3. For the second case

    (independentm), the mualem model was solved numerically.

    In the free form caseθ(h) andK(h) are treated as completely independent spline functions. The

    complexity of the models increases from case one to case 4.

    Equation 1.1 was fitted with the four different models for the constitutive relationships to the three

    data branches of both columns, so that 24 parameter sets forθ(h) andK(h) were obtained. For each

    model and soil column we got a primary drainage curve (first outflow), a “primary” imbibition curve

    (first inflow) and a secondary drainage curve (second outflow).

    The fitting procedure was carried out with the shuffeld complex evolution algorithm (Duanet al. ,

    1992) which is a global optimizer. The software tool was developed in our department and is called

    RIEHP (Robust Inverse Estimation of Hydraulic Properties).

    1.2.2 Fit of Equilibrium Data

    The hydraulic equilibrium data (values at the end of each pressure step until hLB = −40 cm) for the

    4 outflow branches we also fitted with the three parametric models as describedby (Peters & Durner,

    2006) with the software tool SHYPFIT2.0 (Soil Hydraulic Properties Fitting).

  • Chapter 2

    Results

    2.1 Saturated Water Content, Porosity, Bulk Density and Saturated

    Hydraulic Conductivity

    The saturated hydraulic conductivity,Ks, was 144 and 68.3cm h−1, and the saturated water content,

    θs, was 0.396 and 0.385, for column 1 and 2, respectively. For the imbibition branch and the second

    outflow branch theθs was assumed to be equal to the water content at the end of the measurement

    (hLB = +7.5 cm). Hereθs was 0.351 and 0.343. The bulk density,ρb was very similar with 1.567

    and 1.563g cm−3 for the two columns. The porosity,φ, was calculated according toφ = ρb/ρswhereρs is the particle density. The value ofρs was assumed to be 2.65g cm−3. The results are

    summarized in Tables 2.1 to 2.6.

    2.2 Inversion of Richard’s Equation

    2.2.1 First Outflow Branch

    The measured data and model fits are shown in Figures 2.1 and 2.2. For the last two steps

    (hLB ≤ −45 cm) hydraulic equilibrium was not achieved. For all other steps the point water con-

    tents and pressures may be predicted at any height. Although we assume a homogeneous soil without

    any macropores we observed that the soil column drained already slightly at very high values for

    hLB. That behavior can only be described by very flexible hydraulic models such as the bimodal van

    Genuchten model or free form approaches.

    The two outflow plateaus athLB = −25 cm in column 2 (Fig. 2.1 bottom, time 10 to 15 hours)

    may be explained by a break of an instability after some fluctuations at the lowerboundary occurred.

    These fluctuations have no effect on the matric head in the middle of the column.

    The resulting hydraulic properties functions are plotted in Figures 2.3 to 2.5.The shapes of the

    different models look for both columns quite similar (Fig. 2.5), so that we regard the experiment as

    well reproducible. The corresponding parameter values and goodness of fits are listed in Tables 2.1

    11

  • 12 CHAPTER 2. RESULTS

    and 2.2. Since the parameter correlation for the bimodal van Genuchten model is very high, the single

    parameter values should not be interpreted. The more valuable information isthe mere shape of the

    curves curve.

  • 2.2. INVERSION OF RICHARD’S EQUATION 13

    0 20 40 60 80 100

    −50

    −40

    −30

    −20

    −10

    0

    time [h]

    mat

    ric h

    ead

    [cm

    ]

    measuredvG consvG unconsvG bimodfree form

    0 20 40 60 80 100

    −2

    −1.5

    −1

    −0.5

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    0 3 6 9 12 15

    −1

    −0.8

    −0.6

    −0.4

    −0.2

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    Figure 2.1: First drainage branch, column 1: Measured data and fitted with 4 different hydraulic propertiesmodels. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumulative outflow forhLB ≥ −25 to show

    the better performance of the more flexible models.

  • 14 CHAPTER 2. RESULTS

    0 20 40 60 80 100

    −50

    −40

    −30

    −20

    −10

    0

    time [h]

    mat

    ric h

    ead

    [cm

    ]measuredvG consvG unconsvG bimodfree form

    0 20 40 60 80 100−2.5

    −2

    −1.5

    −1

    −0.5

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    0 5 10 15−1

    −0.8

    −0.6

    −0.4

    −0.2

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    Figure 2.2: First drainage branch, column 2: Measured data and fitted with 4 different hydraulic propertiesmodels. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumulative outflow forhLB ≥ −25 to show

    the better performance of the more flexible models.

  • 2.2. INVERSION OF RICHARD’S EQUATION 15

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ]

    vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.3: First drainage branch, column 1: Hydraulic properties.

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ]

    vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.4: First drainage branch, column 2: Hydraulic properties.

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ]

    vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.5: First outflow branch, both columns: Hydraulic properties. solid lines: column 2; dashed lines:column 2.

  • 16 CHAPTER 2. RESULTS

    Table 2.1: Estimated parameters for first drainage branch - column 1.

    Parameter unit measured

    ρb g cm−3 1.567

    φ 1 - 0.409

    Ks cm h−1 144

    θs - 0.396

    estimated

    vG cons vG uncons vG bimod ff r=10

    α1 cm−1 0.0340 0.0357 0.0330 −

    n1 − 11.42 15.13 12.77 −

    m1 − = 1 − 1/n1 0.479 = 1 − 1/n1 −

    θr − 0.076 0.070 0.072 −

    α2 cm−1

    − − 0.076 −

    n2 − − − 1.87 −

    m2 − − − = 1 − 1/n2 −

    w2 − − − 0.061 −

    τ − 0.625 1.039 0.902 −

    RMSEh2 cm 5.987E-01 5.763E-01 5.849E-01 5.780E-01

    RMSEQ3 cm 4.178E-02 3.948E-02 3.419E-02 2.040E-02

    1 calculated with assumption thatρs = 2.65 g cm−3: φ = 1 − ρb/ρs2 root mean square error of the measurement and model prediction ofmatric head,h3 root mean square error of the measurement and model prediction ofcumulative outflow,Q

    2.2.2 First Imbibition Branch

    In Figures 2.6 and 2.7 the inflow data and corresponding fits are shown. In column 1 hydraulic

    equilibrium is only reached afterhLB ≥ −10 cm. This late response on the change of the pressure at

    the lower boundary could possibly be explained by the existence of some small air bubbles underneath

    the ceramic plate which cause contact problems of the water phase. These air bubbles were not visible.

    Due to this late response almost the complete inflow happened at the last coupleof pressure steps,

    that can only be described by a very steep retention function close to saturation. The fit with the both

    unimodal models failed completely. The data for column 2 look more reasonable.But also here the

    system behavior can only adequately be described by the more flexible models, i.e. the bimodal or

    the free form approaches.

    The hydraulic properties functions and the corresponding parameter values are shown in Figures 2.8

    and 2.9 and Tables 2.3 and 2.4. Due to the assumption that the imbibition measurement for column 1

    was influenced by some measurement artifacts we suggest that the hydraulic properties functions for

    the imbibition branch are interpreted only from column 2.

    2.2.3 Second Outflow Branch

    The measured data and resulting fits are shown in Figures 2.10 and 2.11. Again, the behavior close to

    saturation can only adequately be described by the more flexible functions.

  • 2.2. INVERSION OF RICHARD’S EQUATION 17

    Table 2.2: Estimated parameters for first drainage branch - column 2.

    Parameter unit measured

    ρb g cm−3 1.563

    φ 1 - 0.410

    Ks cm h−1 68.3

    θs - 0.385

    estimated

    vG cons vG uncons vG bimod ff r=9

    α1 cm−1 0.0327 0.0329 0.0324 −

    n1 − 10.91 11.05 12.04 −

    m1 − = 1 − 1/n1 0.870 = 1 − 1/n1 −

    θr − 0.095 0.0946 0.0976 −

    α2 cm−1

    − − 0.107 −

    n2 − − − 4.929 −

    m2 − − − = 1 − 1/n2 −

    w2 − − − 0.032 −

    τ − 0.370 0.408 0.216 −

    RMSEh2 cm 2.738E-01 2.730E-01 2.895E-01 3.273E-01

    RMSEQ3 cm 3.528E-02 3.527E-02 2.207E-02 7.565E-03

    1 calculated with assumption thatρs = 2.65 g cm−3: φ = 1 − ρb/ρs2 root mean square error of the measurement and model prediction ofmatric head,h3 root mean square error of the measurement and model prediction ofcumulative outflow,Q

    Table 2.3: Estimated parameters for first imbibition branch - column 1.

    Parameter unit measured

    ρb g cm−3 1.567

    φ 1 - 0.409

    Ks cm h−1 144

    θs - 0.351

    estimated

    vG cons vG uncons vG bimod ff r=9

    α1 cm−1 0.0643 0.0628 0.062 −

    n1 − 14.45 13.22 15.02 −

    m1 − = 1 − 1/n1 1.031 = 1 − 1/n1 −

    θr − 0.120 0.119 0.02 −

    α2 cm−1

    − − 0.450 −

    n2 − − − 1.104 −

    m2 − − − = 1 − 1/n2 −

    w2 − − − 0.366 −

    τ − 1.606 1.031 2.02 −

    RMSEh3 cm 5.346E+00 5.512E+00 2.081E+00 7.686E-01

    RMSEQ4 cm 9.324E-02 9.093E-02 7.856E-02 5.023E-02

    1 calculated with assumption thatρs = 2.65 g cm−3: φ = 1 − ρb/ρs2 estimated parameter reached boundary of parameter space3 root mean square error of the measurement and model prediction ofmatric head,h4 root mean square error of the measurement and model prediction ofcumulative outflow,Q

  • 18 CHAPTER 2. RESULTS

    0 20 40 60 80 100

    −50

    −40

    −30

    −20

    −10

    0

    time [h]

    mat

    ric h

    ead

    [cm

    ]

    measuredvG consvG unconsvG bimodfree form

    0 20 40 60 80 100

    0

    0.5

    1

    1.5

    2

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    Figure 2.6: First imbibition branch, column 1: Measured data and fitted with 4 different hydraulic propertiesmodels. Top: Matric head; Bottom: Cumulative outflow.

  • 2.2. INVERSION OF RICHARD’S EQUATION 19

    0 20 40 60 80 100

    −50

    −40

    −30

    −20

    −10

    0

    time [h]

    mat

    ric h

    ead

    [cm

    ]

    measuredvG consvG unconsvG bimod

    0 20 40 60 80 1000

    0.5

    1

    1.5

    2

    2.5

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m] measured

    vG consvG unconsvG bimod

    75 80 85 90 950

    0.5

    1

    1.5

    2

    2.5

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m] measured

    vG consvG unconsvG bimod

    Figure 2.7: First imbibition branch, column 2: Measured data and fitted with 4 different hydraulic propertiesmodels. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumulative outflow forhLB ≥ −25 to show

    the better performance of the more flexible models.

  • 20 CHAPTER 2. RESULTS

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ] vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.8: First imbibition branch, column 1: Hydraulic properties.

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ] vG consvG unconsvG bimod

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimod

    Figure 2.9: First imbibition branch, column 2: Hydraulic properties.

  • 2.2. INVERSION OF RICHARD’S EQUATION 21

    Table 2.4: Estimated parameters for first imbibition branch - column 2.

    Parameter unit measured

    ρb g cm−3 1.563

    φ 1 - 0.410

    Ks cm h−1 68.3

    θs - 0.343

    estimated

    vG cons vG uncons vG bimod ff r=9

    α1 cm−1 0.0556 0.0686 0.049 −

    n1 − 4.93 8.338 10.28 −

    m1 − = 1 − 1/n1 0.278 = 1 − 1/n1 −

    θr − 0.130 0.122 0.094 −

    α2 cm−1

    − − 0.088 −

    n2 − − − 1.972 −

    m2 − − − = 1 − 1/n2 −

    w2 − − − 0.4567 −

    τ − 1.084 2.02 2.02 −

    RMSEh3 cm 1.906E+00 1.534E+00 6.523E-01 xxx

    RMSEQ4 cm 4.751E-02 5.311E-02 2.180E-02 xxx

    1 calculated with assumption thatρs = 2.65 g cm−3: φ = 1 − ρb/ρs2 estimated parameter reached boundary of parameter space3 root mean square error of the measurement and model prediction ofmatric head,h4 root mean square error of the measurement and model prediction ofcumulative outflow,Q

    The resulting hydraulic properties functions are plotted in Figures 2.12 to 2.14. The shapes of the

    different models look for both columns quite similar (Fig. 2.14), so that we regard the experiment as

    well reproducible. The corresponding parameter values and goodness of fits are listed in Tables 2.1

    and 2.2.

  • 22 CHAPTER 2. RESULTS

    0 20 40 60 80 100

    −50

    −40

    −30

    −20

    −10

    0

    time [h]

    mat

    ric h

    ead

    [cm

    ]

    measuredvG consvG unconsvG bimodfree form

    0 20 40 60 80 100

    −2

    −1.5

    −1

    −0.5

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    0 5 10 15−1

    −0.8

    −0.6

    −0.4

    −0.2

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    Figure 2.10: Second outflow branch, column 1: Measured data and fitted with4 different hydraulic propertiesmodels. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumulative outflow forhLB ≥ −25 to show

    the better performance of the more flexible models.

  • 2.2. INVERSION OF RICHARD’S EQUATION 23

    0 20 40 60 80 100 120

    −50

    −40

    −30

    −20

    −10

    0

    time [h]

    mat

    ric h

    ead

    [cm

    ]measuredvG consvG unconsvG bimodfree form

    0 20 40 60 80 100 120−2.5

    −2

    −1.5

    −1

    −0.5

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    0 5 10 15−1

    −0.8

    −0.6

    −0.4

    −0.2

    0

    time [h]

    cum

    ulat

    ive

    outfl

    ow [c

    m]

    measuredvG consvG unconsvG bimodfree form

    Figure 2.11: Second outflow branch, column 2: Measured data and fitted with4 different hydraulic propertiesmodels. Top: Matric head; Middle: Cumulative outflow; Bottom: Cumulative outflow forhLB ≥ −25 to show

    the better performance of the more flexible models.

  • 24 CHAPTER 2. RESULTS

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ] vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.12: Second outflow branch, column 1: Hydraulic properties.

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ] vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.13: Second outflow branch, column 2: Hydraulic properties.

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ]

    vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.14: Second outflow branch, both columns: Hydraulic properties.solid lines: column 1; dashed lines:column 2.

  • 2.2. INVERSION OF RICHARD’S EQUATION 25

    Table 2.5: Estimated parameters for second drainage branch - column 1.

    Parameter unit measured

    ρb g cm−3 1.567

    φ 1 - 0.409

    Ks cm h−1 144

    θs - 0.351

    estimated

    vG cons vG uncons vG bimod ff r=9

    α1 cm−1 0.0337 0.0311 0.0332 −

    n1 − 8.016 7.148 9.467 −

    m1 − = 1 − 1/n1 1.372 = 1 − 1/n1 −

    θr − 0.062 0.068 0.0625 −

    α2 cm−1

    − − 0.0672 −

    n2 − − − 2.276 −

    m2 − − − = 1 − 1/n2 −

    w2 − − − 0.097 −

    τ − 1.380 1.018 1.315 −

    RMSEh2 cm 5.758E-01 5.922E-01 5.689E-01 5.727E-01

    RMSEQ3 cm 3.042E-02 2.857E-02 1.630E-02 1.363E-02

    1 calculated with assumption thatρs = 2.65 g cm−3: φ = 1 − ρb/ρs2 root mean square error of the measurement and model prediction ofmatric head,h3 root mean square error of the measurement and model prediction ofcumulative outflow,Q

    Table 2.6: Estimated parameters for second drainage branch - column 2.

    Parameter unit measured

    ρb g cm−3 1.563

    φ 1 - 0.410

    Ks cm h−1 68.3

    θs - 0.343

    estimated

    vG cons vG uncons vG bimod ff r=10

    α1 cm−1 0.0336 0.0333 0.0332 −

    n1 − 8.77 8.661 9.900 −

    m1 − = 1 − 1/n1 0.931 = 1 − 1/n1 −

    θr − 0.061 0.062 0.064 −

    α2 cm−1

    − − 0.150 −

    n2 − − − 2.350 −

    m2 − − − = 1 − 1/n2 −

    w2 − − − 0.0464 −

    τ − 0.783 0.744 0.561 −

    RMSEh2 cm 3.447E-01 3.462E-01 3.456E-01 3.786E-01

    RMSEQ3 cm 3.344E-02 3.340E-02 2.043E-02 1.110E-02

    1 calculated with assumption thatρs = 2.65 g cm−3: φ = 1 − ρb/ρs2 root mean square error of the measurement and model prediction ofmatric head,h3 root mean square error of the measurement and model prediction ofcumulative outflow,Q

  • 26 CHAPTER 2. RESULTS

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ]

    vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.15: Hysteresis of Hydraulic Properties, Column 1. Solid lines:first outflow; dots: first imbibition;dashed lines: second outflow.

    0 10 20 30 40 50 60

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    vol.

    wat

    er c

    onte

    nt [−

    ]

    vG consvG unconsvG bimodfree form

    0 10 20 30 40 50 60−5

    −4

    −3

    −2

    −1

    0

    1

    2

    3

    h [cm]

    log 1

    0 (K

    in c

    m h

    −1 )

    vG consvG unconsvG bimodfree form

    Figure 2.16: Hysteresis of Hydraulic Properties, Column 2. Solid lines:first outflow; dots: first imbibition;dashed lines: second outflow.

    2.2.4 Hysteresis

    In Figures 2.15 and 2.16 the hydraulic properties functions are combined for both columns. As

    expected the imbibition curves show a air entry closer to saturation. The retention functions of the

    imbibition branch for column 2 are also less steep.

    2.3 Fit of Equilibrium Data

    The model fits of the equilibrium data gave very similar results as the inversion of Richards Equation.

    This is remarkable since only the data up tohLB ≥ −40 cm could be used. With that restriction

    10 data points were available for evaluation (Tables 2.7 and 2.8). The resulting parameter values are

    given in Tables 2.9 and 2.10. The retention functions are plotted in Figure 2.17.

  • 2.3. FIT OF EQUILIBRIUM DATA 27

    Table 2.7: Data for equilibrium fits. First outflow branch.column 1 column 2

    hm θ̄ hLB θ̄

    cm [−] cm [−]

    -0.810 0.396 -0.680 0.385

    -4.710 0.395 -4.430 0.384

    -9.460 0.393 -9.590 0.381

    -14.550 0.388 -14.620 0.376

    -19.450 0.381 -19.340 0.370

    -24.670 0.363 -24.370 0.356

    -29.430 0.254 -29.440 0.268

    -34.420 0.147 -34.420 0.152

    -39.400 0.105 -39.480 0.102

    -43.720 0.089 -44.410 0.084

    1 rmean of matric head in soil column at each pressure step.hm = hLB − L/2, whereL [cm] is the height of the soil column with

    cermaic plate.

    Table 2.8: Data for equilibrium fits. Second outflow branch.column 1 column 2

    hm θ̄ hLB θ̄

    cm [−] cm [−]

    -0.620 0.351 -0.730 0.343

    -4.570 0.351 -4.430 0.341

    -9.450 0.347 -9.550 0.337

    -14.430 0.341 -14.490 0.333

    -19.400 0.332 -19.390 0.326

    -24.340 0.298 -24.460 0.298

    -29.400 0.224 -29.500 0.216

    -34.340 0.142 -34.440 0.136

    -39.410 0.102 -39.390 0.097

    -44.420 0.084 -44.500 0.078

    1 rmean of matric head in soil column at each pressure step.hm = hLB − L/2, whereL [cm] is the height of the soil column with

    cermaic plate.

    0 10 20 30 40 50 600.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    volu

    met

    ric w

    ater

    con

    tent

    [−]

    vG consvG unconsvG bimod

    0 10 20 30 40 50 600.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0.35

    0.4

    h [cm]

    volu

    met

    ric w

    ater

    con

    tent

    [−]

    vG consvG unconsvG bimod

    Figure 2.17: Retention functions resulting from equilibrium fit. Left: column 1, right: column 2. Solid lines:first outflow brach; dashed lines: second outflow branch.

  • 28 CHAPTER 2. RESULTS

    Table 2.9: Estimated parameters of equilibrium data - first drainage branch.

    Parameter unit column 1 column 2

    vG cons vG uncons vG bimod vG cons vG uncons vG bimod

    α1 cm−1 0.0337 0.0346 0.0335 0.0328 0.0320 0.0325

    n1 − 11.85 12.78 13.11 10.88 10.31 12.15

    m1 − = 1 − 1/n1 0.727 = 1 − 1/n1 = 1 − 1/n1 1.089 = 1 − 1/n1θr − 0.086 0.083 0.089 0.0741 0.0757 0.0771

    α2 cm−1

    − − 0.0940 − − 0.1038

    n2 − − − 11.43 − − 7.83

    m2 − − − = 1 − 1/n2 − − = 1 − 1/n2

    w2 − − − 0.0294 − − 0.0337

    RMSE 1 cm 4.877E-03 4.853E-03 3.048E-03 5.188E-03 5.166E-03 2.413E-03

    1 root mean square error of the measurement and model prediction ofmean water content,̄θ

    Table 2.10:Estimated parameters of equilibrium data - second drainagebranch.

    Parameter unit column 1 column 2

    vG cons vG uncons vG bimod vG cons vG uncons vG bimod

    α1 cm−1 0.0342 0.0217 0.0344 0.0340 0.0281 0.1258

    n1 − 7.57 5.87 2.5409 8.45 6.95 4.07

    m1 − = 1 − 1/n1 9.456 = 1 − 1/n1 = 1 − 1/n1 2.6609 = 1 − 1/n1θr − 0.063 0.084 0.0356 0.0643 0.0766 0.0718

    α2 cm−1

    − − 0.0336 − − 0.0336

    n2 − − − 10.52 − − 9.87

    m2 − − − = 1 − 1/n2 − − = 1 − 1/n2

    w2 − − − 0.6842 − − 0.9567

    RMSE 1 cm 3.992E-03 2.969E-03 8.035E-04 4.493E-03 4.155E-03 1.100E-03

    1 root mean square error of the measurement and model prediction ofmean water content,̄θ

  • Bibliography

    Bitterlich, S., Durner, W., Iden, S. C., & Knabner, P. 2004. Inverse estimation of the unsaturated soil

    hydraulic properties from column outflow experiments using free-form parameterizations.Vadose

    Zone J., 3, 971–981.

    Duan, Q., Sorooshian, S., & Gupta, V. 1992. Effective and efficient global optimization for conceptual

    rainfall-runoff models.Water Resources Research, 28, 1015–1031.

    Durner, W. 1994. Hydraulic conductivity estimation for soils with heterogeneous pore structure.

    Water Resour. Res., 30, 211–223.

    Genuchten, M. Th. van. 1980. A closed-form equation for predicting the hydraulic conductivity of

    unsaturated soils.Soil Sci. Soc. Am. J., 44, 892–898.

    Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous

    media.Water Resour. Res., 12(3), 513–521.

    Peters, A., & Durner, W. 2006. Improved estimation of soil water retention characteristics from

    hydrostatic column experiments.Water Resour. Res. in press.

    Priesack, E., & Durner, W. 2006. Closed fom expression for the multi-modal unsaturated conductivity

    function. Vadose Zone J., 5, 121–124.

    Zurmühl, T. 1998. Capability of convection dispersion transport models to predict transient water and

    solute movement in undisturbed soil columns.J. Contam. Hydrol., 30, 99–126.

    29