Using the HYDRUS-1D and HYDRUS-2D Codes

for Estimating Unsaturated Soil Hydraulic

and Solute Transport Parameters
Jirka Òimçnek and Martinus Th. van Genuchten

U.S. Salinity Laboratory, USDA, ARS, Riverside, CA

Abstract. In this paper we describe a parameter estimation procedure which combines the
Levenberg-Marquardt nonlinear parameter optimization method involving weighted least squares
with either a one-dimensional numerical model (HYDRUS-1D) or a two- or quasi three-
dimensional model (HYDRUS-2D), which solve the governing equations for water flow and
solute transport in variably-saturated porous media. The procedure permits several unknown
parameters in the unsaturated soil-hydraulic functions to be estimated from observed water
contents, pressure heads, and/or instantaneous or cumulative boundary fluxes (e.g., infiltration or
outflow data) during transient water flow by numerical inversion of the Richards equation.
Additional retention or hydraulic conductivity data, as well as a penalty function for constraining
the optimized parameters to remain in some feasible region (Bayesian estimation) can be
optionally included in the parameter estimation procedure. Similarly, the procedure permits
solute transport and/or reaction parameters to be estimated from observed concentrations and/or
instantaneous or cumulative boundary solute fluxes during transient solute transport by numerical
inversion of the convection-dispersion equation. The unsaturated soil hydraulic and solute
transport and reaction parameters can be estimated either sequentially or simultaneously.
Depending upon the quality of observed data, soil hydraulic or solute transport parameters for
several soil layers can be estimated simultaneously. The parameter estimation procedure is
demonstrated for several laboratory and field experiments.

INTRODUCTION

As increasingly more complicated computer models are being developed for simulating
subsurface flow and transport processes, the accuracy of numerical simulations largely depends
upon the accuracy with which various model parameters can be estimated. Flow and transport
models for the unsaturated zone are often based on numerical solutions of the Richards equation
which requires knowledge of the unsaturated soil hydraulic functions, i.e., the soil water retention
curve, θ(h), describing the relationship between the water content θ and the pressure head h, and
the unsaturated hydraulic conductivity function, K(h), defining the hydraulic conductivity K as a
function of h. Accurate measurement of the hydraulic properties is difficult because of the highly
nonlinear nature of these properties, especially K(h), instrumental limitations, and the extreme
heterogeneity of the subsurface environment. Hence, methods for making relatively fast and
reliable measurements of the unsaturated soil-hydraulic properties remain sorely needed [van
Genuchten and Leij, 1992].
A variety of field methods are currently available for direct measurement of the hydraulic
conductivity, K, or the soil water diffusivity, D, as a function of h and/or θ [Klute and Dirksen,
1986; Green et al., 1986]. Popular field methods include the instantaneous profile method,
various unit-gradient type approaches, sorptivity methods following ponded infiltration, and the
crust method based on steady water flow. While relatively simple in concept, these direct
measurement methods have a number of limitations that restrict their use in practice. For

1523
1524 Òimçnek and van Genuchten

example, most methods are very time-consuming to execute because of the need to adhere to
relatively restrictive initial and boundary conditions. This is especially true for field gravity-
drainage experiments involving medium- and fine-textured soils. Methods requiring repeated
steady-state flow situations, or other equilibrium conditions are also tedious, while linearizations
and other approximations or interpolations to allow analytic or semi-analytic inversions of the
flow equation may introduce additional errors. Finally, information about uncertainty in the
estimated hydraulic parameters is not readily obtained using direct inversion methods.
A much more flexible approach for solving the inverse problem is the use of parameter
optimization methods [Hopmans and Šimçnek, 1999]. Optimization procedures make it possible
to simultaneously estimate the retention and hydraulic conductivity functions from transient flow
data [Kool et al., 1987]. While many possible scenarios exist for the application of parameter
optimization methods, numerical inversion of the Richards equation has thus far been limited
only, or nearly exclusively, to one-dimensional experiments [Kool et al., 1985; Russo et al.,
1991], mostly in conjunction with one-step or multi-step outflow experiments [Kool and Parker,
1988; van Dam et al., 1992, 1994; Eching et al., 1993]. Nevertheless, other types of
experiments, such as upward infiltration [Hudson et al., 1996] or evaporation methods [Ciollaro
and Romano, 1995; Santini et al., 1995; Šimçnek et al., 1998d, 1999b], were also reported.
Possible multi-dimensional applications involve the use of disc tension permeameters [Perroux
and White, 1988; Ankeny et al., 1991, Šimçnek and van Genuchten, 1996, 1997; Šimçnek et al.,
1998a,c], a modified cone penetrometer [Gribb et al., 1998; Kodešová et al., 1998, 1999;
Šimçnek et al., 1999a], a multistep soil-water extraction method [Inoue et al., 1998, 1999],
infiltration from a furrow, and surface or subsurface drip irrigation experiments.
In this paper we describe a parameter estimation procedure which combines the Levenberg-
Marquardt nonlinear parameter optimization method involving weighted least squares with either
a one-dimensional numerical model, HYDRUS-1D [Šimçnek et al., 1998b], or a two- or quasi
three-dimensional numerical model, HYDRUS-2D [Šimçnek et al., 1996], which solve the
variably-saturated water flow and solute transport equations. We demonstrate the proposed
parameter estimation procedure on laboratory and field data, and briefly summarize many other
applications of the HYDRUS models.

FORWARD PROBLEM

Variably-Saturated Water Flow

The governing equation for two-dimensional isothermal Darcian water flow in a variably-
saturated rigid isotropic porous medium is given by the following modified form of the Richards
equation:
∂θ ∂ ∂h
= [ K ( K ijA + K iA2 )] - S (1)
∂t ∂ xi ∂ xj

where xi (i=1,2; with x2=z being the vertical coordinate positive upwards) are the spatial
coordinates, t is time, S is a sink term, KijA are components of a dimensionless anisotropy tensor
A
K , and K is the unsaturated hydraulic conductivity function given as the product of the relative
hydraulic conductivity, Kr, and the saturated hydraulic conductivity, Ks.
Equation (1) can be solved numerically for a given set of initial and boundary equations.
The HYDRUS-1D and -2D codes implement three different types of boundary conditions:
specified pressure head (Dirichlet type) conditions of the form
h( x , t ) = ψ ( x , t ) (2)
specified flux (Neumann type) conditions given by
Estimating Flow and Transport Parameters with HYDRUS Codes 1525

∂h
- [ K ( K ijA + K iA2 )] ni = σ ( x, t ) (3)
∂x
j

and specified gradient conditions (e.g., free drainage associated with a unit hydraulic gradient) as
follows
∂h
( K ijA + K iA2 ) ni = ζ ( x , t ) (4)
∂x
j

where ψ, σ, and ζ are the prescribed Dirichlet, Neumann, and gradient type boundary conditions,
respectively, as functions of x and t, x is the spatial coordinate of a boundary, and ni are the
components of the outward unit vector normal to boundary.
The above boundary conditions can be implemented in HYDRUS-1D and -2D in several
ways: as (a) constant boundary conditions (either flux or head), (b) variable boundary conditions
(again either flux or head), (c) seepage faces, (d) atmospheric boundaries, and (e) free or deep
drainage boundaries. Boundary classes (a) and (b) represent system-independent boundary
conditions, while (c), (d), and (e) are system dependent, i.e., they depend on the prevailing
transient soil moisture or flux conditions. As explained later, instantaneous or cumulative
boundary fluxes across any of the boundaries, and water contents and pressure heads measured
anywhere in the transport domain, can be included in the objective function for purposes of
parameter identification.
While different functions for the unsaturated soil-hydraulic properties may be used in the
inverse problem, the expressions adopted in HYDRUS codes are those of van Genuchten [1980]:
θ (h) - θr n −m
θe(h) = =(1 + α h ) (5)
θs - θr
2
K ( θ ) = K s θ el [ 1 - ( 1 - θ e1 / m ) m ] (6)

and Brooks and Corey [1966]:

α h − n h < − 1/α
θ (h) - θr 
θe(h) = =  (7)
θs - θr  1
 h ≥ − 1/α
K ( θ ) = K s θ le+ 2+ 2 / n (8)
where θe is the effective water content, θr and θs denote the residual and saturated water contents,
respectively, and α, n, m (= 1 - 1/n), and l are empirical parameters. The hydraulic characteristics
defined by (5) through (8) contain 6 unknown parameters: θr , θs , α, n, l, and Ks. Of these, θr, θs,
and Ks have a clear physical meaning, whereas α, n and l are essentially empirical parameters
determining the shape of the retention and hydraulic conductivity functions [van Genuchten,
1980].
For the hysteretic case the HYDRUS codes use the formulation of Kool and Parker [1987]
who coupled the van Genuchten-Mualem model with a simplified scaling approach proposed by
Scott et al. [1983] to describe the scanning curves. Scott et al. [1983] assumed that the shape
parameters α and n for all drying scanning curves are the same as those for the main drying curve
and, similarly, the shape parameters for all wetting scanning curves are the same as those for the
main wetting curve. Scanning curves are then calculated by varying the residual and saturated
water contents for the wetting and drying scanning curves, respectively. Kool and Parker [1987]
further assumed that the shape parameter n is the same for both wetting and drying, thus
decreasing the number of required parameters. Using the additional restrictions that θr and θs are
1526 Òimçnek and van Genuchten

the same for both drying and wetting, the only additional parameter describing hysteresis is a
third shape parameter αd for the drying retention curve (we use αw for wetting).
Equation (1) subject to initial and boundary conditions (2), (3) and (4) was solved
numerically by means of the finite element method. The solution scheme was based on the mass-
conservative numerical iterative scheme used by Celia et al. [1990]. A simple modification of
this numerical scheme also permitts similar mass-conservative solutions of quasi-three-
dimensional axisymmetrical flow problems [Šimçnek et al., 1996].

Solute Transport
We assume that solutes can exist in all three phases (liquid, solid, and gaseous) and that
production and decay processes can be different in each phase. We further assume that solutes are
transported by convection and dispersion in the liquid phase, as well as by diffusion in the gas
phase. The partial differential equations governing the nonequilibrium chemical transport of
solutes involved in a sequential first-order decay chain during transient water flow in a variably
saturated rigid porous medium are taken as [Šimçnek et al., 1998b]:
∂θ c k ∂ρ s k ∂ a v g k ∂ ∂c ∂ ∂ g k ∂q c k
+ + = (θ D kw k ) + (a v D kg )- -
∂t ∂t ∂t ∂x ∂x ∂x ∂x ∂x
- ( µ w,k + µ 'w,k )θ c k - ( µ s,k + µ 's,k ) ρ s k - ( µ g,k + µ 'g,k ) a v g k + µ 'w,k -1θ c k -1 (9)
+ µ 's,k -1 ρ s k -1 + µ 'g , k -1 a v g k-1 + γ w,k θ + γ s,k ρ + γ g,k a v - S c r,k k ∈ ( 2, n s )

where for simplicity the one-dimensional formulation is used. In (9) c, s, and g are solute
concentrations in the liquid, solid and gas phases, respectively; q is the volumetric flux density,
µw, µs , and µg are first-order rate constants for solutes in the liquid, solid and gas phases,
respectively; µwN, µsN, and µgN are similar first-order rate constants providing connections between
individual chain species, γw , γs, and γg are zero-order rate constants for the liquid, solid and gas
phases, respectively; ρ is the soil bulk density, av is the air content, S is the sink term in the flow
equation (1), cr is the concentration of the sink term, Dw is the dispersion coefficient for the liquid
phase, and Dg is the diffusion coefficient for the gas phase. The subscripts w, s, and g correspond
with the liquid, solid and gas phases, respectively; subscript k represents the kth chain number,
and ns is the number of solutes involved in the chain reaction. The nine zero- and first-order rate
constants in (9) may be used to represent a variety of reactions or transformations including
biodegradation, volatilization, and precipitation.
The HYDRUS codes may be used to simulate nonequilibrium interactions between the
solution (c) and adsorbed (s) concentrations, and equilibrium interaction between the solution (c)
and gas (g) concentrations of the solute in the soil system. The equilibrium adsorption isotherm
relating s and c is described by a generalized nonlinear equation of the form
k s cβ
s = (10)
1+ η c β

where ks, β and η are empirical coefficients. The Freundlich, Langmuir, and linear adsorption
equations are special cases of (10). The concentrations g and c are related by a linear expression
of the form
g = kg c (11)

where kg is an empirical constant, often referred to as Henry’s constant.

The concept of two-site sorption [Selim et al., 1977; van Genuchten and Wagenet, 1989] is
implemented in the HYDRUS codes to permit consideration of nonequilibrium adsorption-
desorption reactions. The two-site sorption concept assumes that the sorption sites can be divided
Estimating Flow and Transport Parameters with HYDRUS Codes 1527

into two fractions. Sorption, se, on one fraction of the sites (the type-1 sites) is assumed to be
instantaneous, while sorption, sk, on the remaining (type-2) sites is considered to be time-
dependent. The mass balance equation for the type-2 sites in the presence of production and
degradation is given by
∂ sk  k s cβ 
= ω ( 1 - f ) - s k  - ( µ s + µ 's ) s k + ( 1 - f ) γ s (12)
∂t  1+ η c β

where ω is the first-order rate constant and f is the fraction of exchange sites assumed to be in
equilibrium with the solution phase.
The HYDRUS models also implement the concept of two-region, dual-porosity type solute
transport [van Genuchten and Wierenga, 1976] to permit consideration of physical
nonequilibrium transport. The two-region concept assumes that the liquid phase can be
partitioned into mobile (flowing), θm, and immobile (stagnant), θim, regions and that solute
exchange between the two liquid regions can be modeled as a first-order process, i.e.,
 k s β cim  ∂ cim
β -1
 β -1
k s cim 
θ im + ρ ( 1 - f )  = - θ im ( µ + µ ' ) + ρ ( µ + µ ' ) ( 1 - f ) β  cim
+
 (1 + η cim
β 2
)  ∂t 
w w s s
1 + η cim  (13)
ω (c - cim ) + γ w θ im + ( 1 - f ) ρ γ s

where cim is the concentration in the immobile region and ω is the mass transfer coefficient.
By selecting certain values of the γw, γs, γg, µw, µs, µg, µw', µs', µg', η, ks, kg, f, θim, β and ω
in (9) through (13), the entire system can be simplified significantly.

FORMULATION OF INVERSE PROBLEM

The inverse problem may be carried out using several direct and indirect methods [Neuman,
1973]. Direct methods treat the model parameters as dependent variables in a formal inverse
boundary value problem [Yeh, 1986]. Indirect approaches, such as the one used in this paper,
attempt to minimize a suitable objective function which expresses the discrepancy between
observed and predicted system response. Initial estimates of the assumed unknown hydraulic and
transport parameters are then iteratively adjusted and improved upon during the minimization
procedure until a desired precision is obtained.
When measurement errors follow a multivariate normal distribution with zero mean and
covariance matrix V, the likelihood function can be written as [Bard, 1974]
1
L(b) = (2 π )- n / 2 det - 1 / 2 V exp { - [ q* - q (b) ]T V - 1 [ q* - q(b) ] } (14)
2
where L(b) is the likelihood function, b = {b1, b2,..., bm} is the vector of unknown parameters (θr,
*
θs, α, n, l, Ks, λ, µw, µs, β, η, and/or others), m is the number of parameters to be estimated, q =
* * *
{q1 , q2 ,..., qn } is a vector containing the observations (e.g., observed pressure heads, water
contents, concentrations, and/or cumulative and actual water or solute infiltration or outflow
rates), q(b) = {q1, q2,..., qn} is a vector of corresponding model predictions obtained with the
unknown parameters, and n is the number of observations. The maximum likelihood estimate is
that value of the unknown parameter vector b that maximizes the value of the likelihood function.
Assuming that the covariance matrix V is diagonal, i.e., the measurement errors are uncorrelated,
the problem of maximizing the likelihood function simplifies in a weighted least-squares
minimization problem
1528 Òimçnek and van Genuchten

n
Φ (b) = ∑ wi [ qi* - qi (b) ]2 (15)
i =1

where wi is the weight assigned to a particular measured value.

Estimating Flow and Transport Parameters with HYDRUS Codes 1529

If something about the distribution of the fitted parameters is known before the inversion,
that information can be included into the parameter identification procedure by multiplying the
likelihood function with the prior probability density function (pdf), p0(b), which summarizes the
prior information. Estimates which make use of prior information are known as Bayesian
estimates, and lead to the maximization of a posterior pdf, p*(b), given by
p* (b) = c L(b) p0 (b) (16)

in which c is a constant. The posterior density function is proportional to the likelihood function
when the prior distribution is uniform
The objective function Φ to be minimized during the parameter estimation process in both
HYDRUS models is defined as [Šimçnek et al., 1998b,c]:
mq nq j
Φ (b , q , p ) = ∑ v j ∑ wi, j [ q*j ( x, 2
t i ) - q j ( x , t i , b) ] +
j =1 i =1
mp npj
(17)
nb
∑ v j ∑ wi, j [ *
pj( 2
θ i ) - p j (θ i , b ) ] + ∑ vˆ j [ b*j - b j ]
2

j =1 i =1 j =1

where the first term on the right-hand side represents deviations between the measured and
calculated space-time variables (e.g., observed pressure heads, water contents, and/or
concentrations at different locations and/or time in the flow domain, or actual or cumulative
fluxes versus time across a boundary of specified type). In this term, mq is the number of
different sets of measurements, nqj is the number of measurements within a particular
measurement set, qj*(x,ti) represents specific measurements at time ti for the jth measurement set
at location x, qj(x,ti,b) represents the corresponding model predictions for the vector of optimized
parameters b (e.g., θr, θs, α, n, l, Ks, Dl, kg, ...), and vj and wi,j are weights associated with a
particular measurement set or point, respectively. The second term on the right-hand side of (17)
represents differences between independently measured and predicted soil hydraulic properties
(e.g., retention, θ(h), and/or hydraulic conductivity, K(θ) or K(h), data), while the terms mp, npj,
pj*(θi), pj(θi, b), vj and wi,j have similar meanings as for the first term but now for the soil
hydraulic properties. The last term of (17) represents a penalty function for deviations between
prior knowledge of the soil hydraulic parameters, bj*, and their final estimates, bj, with nb being
the number of parameters with prior knowledge and vj representing pre-assigned weights. We
note that the covariance (weighting) matrices which provide information about the measurement
accuracy, as well as any possible correlation between measurement errors and/or parameters, are
assumed to be diagonal in both models. The weighting coefficients vj may be used to minimize
differences in weighting between different data types because of different absolute values and
numbers of data involved, and are given either by [Clausnitzer and Hopmans, 1995]:
2
v j =1/ n j σ j (18)

which causes the objective function to become the average weighted squared deviation
normalized by the measurement variances σj2, or by
 nj

1


∑ q i ni 

vj = min q j ; i =n1  (19)
qj 
j

 ∑ ni 
 i =1 
where qj is the mean of a particular measurement set, or can be specified independently as input.
1530 Òimçnek and van Genuchten

SOLUTION OF THE INVERSE PROBLEM

Many techniques are available for solving the nonlinear minimization/maximization problem
[Bard, 1974; Yeh, 1986; Kool et al., 1987]. Most methods are iterative by starting first with a
given initial estimate bi of the unknown parameters to be estimated, followed by a study of how
the objective function Φ(b) behaves in the vicinity of the initial estimate. Based upon this
behavior one selects a direction vector vi such that the new value of the unknown parameter
vector, i.e.,
bi +1 = bi + ρ i v i (20)

decreases the value of the objective function:

Φ i +1 < Φ i (21)
where Φi and Φi+1 are the objective functions at the previous and current iteration level, and ρi is
a scalar which insures that the iteration step is acceptable. Methods based on (20) are called
gradient methods. Differences among the various gradient methods presented in the literature
(e.g., steepest descent, Newton's method, directional discrimination, Marquardt's method, Gauss
type methods, variable metric methods, and interpolation-extrapolation methods) are a result of
differences in choosing the step direction vi and/or the step size ρi [Bard, 1974]. In the HYDRUS
models we use Marquardt's [1963] method which has proven to be very effective in many
applications involving nonlinear least-square fitting. The method represents a compromise
between the inverse-Hessian and steepest descend methods by using the steepest-descent method
when the objective function is far from its minimum, and switching to the inverse-Hessian
method close to the minimum.

HYDRUS-1D EXAMPLES

Because of a very general formulation of the inverse problem and the possibility to use
different combinations of boundary conditions, the HYDRUS models can be used for a wide
variety of parameter optimization problems. Typical applications include onestep [Kool et
al., 1985] and multistep [van Dam et al., 1992, 1994; Eching et al., 1993] outflow
experiments, upward infiltration [Hudson et al., 1996], and evaporation experiments
[Ciollaro and Romano, 1995; Santini et al., 1995; Šimçnek et al., 1998d, 1999b]. Below we
demonstrate the use of HYDRUS-1D for estimating the soil hydraulic parameters from
multistep outflow data, and a horizontal infiltration experiment followed by redistribution.
The latter example demonstrates the use of HYDRUS-1D for evaluating water flow involving
hysteresis. We will also use HYDRUS-1D here to estimate nonlinear parameters for solute
transport involving Freundlich adsorption by analyzing a measured breakthrough curve.

Inverse Analysis of a Multistep Outflow Experiment

In this test example we analyze a multistep outflow experiment with simultaneous
measurement of the pressure head inside the soil sample [Hopmans, personal communication].
The experimental setup consisted of a 6-cm long soil column in a Tempe pressure cell modified
to accommodate a microtensiometer-transducer system. A tensiometer was installed, with the
cup centered 3 cm below the soil surface. The soil sample was saturated from the bottom and
subsequently equilibrated to an initial soil water pressure head of -25 cm at the soil surface.
Pressures of 100, 200, 400, and 700 cm were subsequently applied in consecutive steps at 0,
12.41, 48.12, and 105.92 hours, respectively. Figure 1 compares the measured and optimized
cumulative outflow curves for the soil sample, while Figure 2 compares measured and optimized
Estimating Flow and Transport Parameters with HYDRUS Codes 1531

pressure heads. Excellent agreement was obtained for both variables. The final fit for the
optimized soil hydraulic parameters (θr=0.197, θs=0.438, α=0.0101 cm-1, n=1.43, l=3.80, and
Ks=0.521 cm h-1) had an r2 of 0.9995.

Fig. 1. Measured and optimized cumulative bottom flux during a multistep outflow experiment.

Fig. 2. Measured and optimized pressure heads in soil sample during a multistep outflow experiment.

Horizontal Infiltration Followed by Redistribution

This example demonstrates the use of HYDRUS-1D for analyzing transient hysteretic
flow. Data used in this example were published by Vachaud [1968]. A horizontal soil
column of 60 cm length and having an internal diameter of 9 cm was used. The initially air
dry silty soil was subjected to a zero pressure at one end of the column for 620 minutes, after
which water was allowed to redistribute. Although water contents were measured for about
25 days with a γ-ray attenuation technique at about 20 points in the column, we used data
from only 10 points for the inversion. The soil hydraulic parameters in the hysteresis model
of Kool and Parker’s [1987] assuming different α values for the wetting and drying curves (αw,
αd), were optimized. Figure 3 shows measured and fitted water contents during the entire
1532 Òimçnek and van Genuchten

experiment. An excellent fit could be obtained only when hysteresis was considered. The
following soil hydraulic parameters were obtained: θr=0.009, θs=0.423, αd=0.0637 cm-1,
-1 -1
αw=0.0910 cm , n=3.86, l=1.47, and Ks=0.0202 cm min .

Fig. 3. Measured and optimized water contents at 10 locations in a soil column

during horizontal infiltration followed by redistribution.

Nonlinear Solute Transport

This example demonstrates the use of HYDRUS-1D to estimate nonlinear solute
transport parameters from breakthrough curves. A 10.75-cm long soil column was first
saturated with a 10 mmolcL-1 CaCl2 solution. The experiment consisted of applying a 14.26
pore volume pulse (t=358.05 h) of a 10 mmolcL-1 MgCl2 solution, followed by the original
CaCl2 solution. The adsorption isotherm was determined independently with the help of
batch experiments [Selim et al., 1987], and fitted with the Freundlich equation to yield
ks=1.687 cm3g-1 and β=1.615.
Only the coefficients of the Freundlich isotherm (i.e., ks and β) were optimized. Since
the governing solute transport equation is nonlinear, one can not use an analytical solution in
this case but must resort to a numerical model. The observed Mg breakthrough curve is
shown in Figure 4, together with the fitted breakthrough curve obtained with HYDRUS-1D.
The results indicate a reasonable prediction of the measured breakthrough curve for the final
estimates of the optimized solute transport parameters (ks=0.943, and β=1.774).
Estimating Flow and Transport Parameters with HYDRUS Codes 1533

Fig. 4. Measured and optimized breakthrough curve for a nonlinear solute transport problem.
HYDRUS-2D EXAMPLES

Similarly as HYDRUS-1D, HYDRUS-2D [Šimçnek et al., 1996] can be used for a broad
range of inverse problems. Recent applications with HYDRUS-2D include estimating soil
hydraulic parameters from data collected with a tension disc infiltrometer [Šimçnek and van
Genuchten, 1996, 1997; Šimçnek et al., 1998a,c], a modified cone penetrometer [Gribb et al.,
1998; Kodešová et al., 1998, 1999; Šimçnek et al., 1999a], and using multistep soil water
extraction device [Inoue et al., 1998, 1999]. Below is an example of the use of HYDRUS-2D
for analyzing tension disc infiltrometer data.

Tension Disc Infiltrometer Example

Tension infiltrations are increasingly being used for evaluating saturated and unsaturated
hydraulic conductivities, and for quantifying the effects of macropores and preferential flow
paths on infiltration. A relatively standard way for estimating unsaturated hydraulic
conductivities from tension infiltrometer data has been to invoke Wooding's [1968] analytical
solution. This approach requires steady-state infiltration rates for two different supply pressure
heads, and assumes applicability of an exponential function for K(h). Šimçnek and van
Genuchten [1996] suggested the combined use of transient infiltration data obtained during a
single tension infiltration experiment, and tensiometer or TDR data measured in the soil below
the disc, to estimate the unknown soil hydraulic parameters via parameter estimation. We later
revised this method by using multiple tension infiltration experiments in combination with
knowledge of the initial and final water contents [Šimçnek and van Genuchten, 1997]. This
modification avoided the cumbersome use of tensiometers and TDRs. An evaluation of the
numerical stability and parameter uniqueness using numerically generated data with
superimposed stochastic and deterministic errors showed that a combination of multiple
cumulative tension infiltration data, a measured final water content, and an initial condition
expressed in terms of the water content, provided the most promising parameter estimation
approach for practical applications [Šimçnek and van Genuchten, 1997].
The numerical inversion method was later used to estimate the soil hydraulic characteristics
of a two-layered crusted soil system in the Sahel region of Africa [Šimçnek et al., 1998c]. Here
we will report only results for the sandy subsoil obtained with a tension disc diameter of 25 cm
and with supply tensions of 11.5, 9, 6, 3, 1, and 0.1 cm. Figure 5 shows measured and optimized
cumulative infiltration curves. The small breaks in the cumulative infiltration curve were caused
by brief removal of the infiltrometer from the soil surface to resupply the instrument with water
and adjust the tension for a new time interval. Very close agreement between the measured and
optimized cumulative infiltration curves was obtained; the largest deviations were generally less
than 60 ml, which was only about 0.5% of the total infiltration volume. Figure 6 shows a
comparison of the parameter estimation results against results obtained with Wooding's analysis.
Both methods give almost identical unsaturated hydraulic conductivities for pressure heads in the
interval between -2 and -10.25 cm. However, the hydraulic conductivity in the highest pressure
head interval was overestimated by a factor of two using Wooding's analysis. Šimçnek et al.
[1998c] further compared the numerical inversion results with hydraulic properties estimated
from available soil textural information using a neural-network-based pedotransfer function
approach. They reported relatively good agreement between the inverse and neural network
predictions.
1534 Òimçnek and van Genuchten

Fig. 5. Measured and optimized cumulative infiltration curves for a tension disc infiltrometer
experiment carried out in the Sahel region.

Fig. 6. Unsaturated hydraulic conductivities calculated using Wooding's analytical solution for
particular pressure heads, and the complete function obtained with numerical inversion.

CONCLUSIONS

Two numerical codes (HYDRUS-1D and HYDRUS-2D) were developed for identifying
soil-hydraulic and solute transport parameters from unsaturated flow and transport data in a one-,
two-, and quasi-three-dimensional porous media. The utility of the two codes was demonstrated
using data typically obtained during multistep outflow experiment, horizontal infiltration
followed by redistribution, a column miscible displacement (breakthrough) study, and a three-
dimensional disc permeameter infiltration experiment. Because of their generality (in terms of
the definition of the objective function, the possible combination of different boundary and initial
conditions, and options for considering multi-layered systems), both models are extremely useful
tools for analyzing a broad range of steady-state and transient laboratory and field flow and
Estimating Flow and Transport Parameters with HYDRUS Codes 1535

transport experiments.
1536 Òimçnek and van Genuchten

REFERENCES

Ankeny, M. D., M. Ahmed, T. C. Kaspar, and R. Horton. 1991. Simple field method for
determining unsaturated hydraulic conductivity. Soil Sci. Soc. Am. J. 55:467-470.
Bard, Y. 1974. Nonlinear Parameter Estimation, Academic Press, New York, N.Y., 341pp.
Brooks, R. H., and A. T. Corey. 1966. Properties of porous media affecting fluid flow. J. Irrig.
Drainage Div., ASCE Proc. 72(IR2):61-88.
Celia, M. A., and E. T. Bouloutas, R. L. Zarba. 1990. A general mass-conservative numerical
solution for the unsaturated flow equation. Water Resour. Res. 26(7):1483-1496.
Ciollaro, G., and N. Romano. 1995. Spatial variability of the soil hydraulic properties of a
volcanic soil. Geoderma 65:263-282.
Clausnitzer, V., and J. W. Hopmans. 1995. Non-linear parameter estimation: LM_OPT.
General-purpose optimization code based on the Levenberg-Marquardt algorithm. Land,
Air and Water Resources Paper No. 100032, University of California, Davis, CA.
Eching, S. O., J. W. Hopmans, and O. Wendroth. 1993. Optimization of hydraulic functions from
transient outflow and soil water pressure data. Soil Sci. Soc. Am. J. 57:1167-1175.
Green, R. E., L. R. Ahuja, and S. K. Chong. 1986. Hydraulic conductivity, diffusivity, and
sorptivity of unsaturated soils: field methods. In: A. Klute (ed.), Methods of Soil Analysis,
Part 1. 2nd ed., pp. 771-798, Agronomy Monogr. 9, ASA and SSSA, Madison, WI.
Gribb, M. M., J. Šimçnek, and M. F. Leonard. 1998. Development of a cone penetrometer
method to determine soil hydraulic properties. ASCE J. of Geotech. and Geoenviron. Eng.
124(9):820-829.
Hopmans, J. W., and J. Šimçnek. 1999. Review of inverse estimation of soil hydraulic properties,
In: M. Th van Genuchten, F. J. Leij, and L. Wu (eds.), Proc. Int. Workshop,
Characterization and Measurement of the Hydraulic Properties of Unsaturated Porous
Media, pp. 643-660, University of California, Riverside, CA.
Hudson, D. B., P. J. Wierenga, and R. G. Hills. 1996. Unsaturated hydraulic properties from
upward flow into soil cores. Soil Sci. Soc. Am. J. 60:388-396.
Inoue, M., J. Šimçnek, J. W. Hopmans, and V. Clausnitzer. 1998. In-situ estimation of soil
hydraulic functions using a multi-step soil water extraction technique. Water Resour. Rer.
34(5):1035-1050.
Inoue, M., J. Šimçnek, J. W. Hopmans, and V. Clausnitzer. 1999. Using a multi-step soil-water
extraction technique for in-situ estimation of soil hydraulic properties, In: M. Th van
Genuchten, F. J. Leij, and L. Wu (eds.), Proc. Int. Workshop, Characterization and
Measurement of the Hydraulic Properties of Unsaturated Porous Media, pp. 725-736, Univ.
of California, Riverside, CA.
Klute, A. and C. Dirksen. 1986. Hydraulic conductivity and diffusivity: Laboratory methods, In:
A. Klute (ed.), Methods of Soil Analysis, Part 1. 2nd ed., pp. 687-734, Agronomy Monogr. 9,
ASA and SSSA, Madison, WI.
Kodešová, R., M. M. Gribb, and J. Šimçnek. 1998. Estimating soil hydraulic properties from
transient cone permeameter data. Soil Sci. 163(6):436-453.
Kodešová, R., M. M. Gribb, and J. Šimçnek. 1999. Use of the cone permeameter method to
determine soil hydraulic properties, In: M. Th van Genuchten, F. J. Leij, and L. Wu (eds.),
Proc. Int. Workshop, Characterization and Measurement of the Hydraulic Properties of
Unsaturated Porous Media, pp. 527-540, University of California, Riverside, CA.
Kool, J. B., J. C. Parker, and M. Th. van Genuchten. 1985. Determining soil hydraulic properties
from one-step outflow experiments by parameter estimation: I. Theory and numerical
studies. Soil Sci. Soc. Am. J. 49:1348-1354.
Kool, J. B., and J. C. Parker. 1987. Development and evaluation of closed-form expressions for
hysteretic soil hydraulic properties. Water Resour. Res. 23(1):105-114.
Kool, J. B., J. C. Parker, and M. Th. van Genuchten. 1987. Parameter estimation for unsaturated
Estimating Flow and Transport Parameters with HYDRUS Codes 1537

flow and transport models - A review. J. Hydrol. 91:255-293.

Kool, J. B., and J. C. Parker. 1988. Analysis of the inverse problem for transient unsaturated
flow. Water Resour. Res. 24(6):817-830.
Marquardt, D. W. 1963. An algorithm for least-squares estimation of nonlinear parameters, SIAM
J. Appl. Math. 11:431-441.
Neuman, S. P. 1973. Calibration of distributed parameter groundwater flow models viewed as a
multiple-objective decision process under uncertainty. Water Resour. Res. 9(4):1006-1021.
Perroux, K. M., and I. White. 1988. Design for disc permeameters. Soil Sci. Soc. Am. J. 52:1205-
1215.
Russo, D., E. Bresler, U. Shani, and J. C. Parker. 1991. Analysis of infiltration events in relation
to determining soil hydraulic properties by inverse problem methodology. Water Resour.
Res. 27(6):1361-1373.
Santini, A., N. Romano, G. Ciollaro, and V. Comegna. 1995. Evaluation of a laboratory inverse
method for determining unsaturated hydraulic properties of a soil under different tillage
practices. Soil Sci. 160:340-351.
Scott, P. S., G. J. Farquhar, and N. Kouwen. 1983. Hysteresis effects on net infiltration,
Advances in Infiltration, Publ. 11-83, pp. 163-170, Am. Soc. Agri. Eng., St. Joseph, Mich.
Selim, H. M., R. Schulin, H. Flühler. 1987. Transport and ion exchange of calcium and
magnesium in an aggregated soil. Soil Sci. Soc. Am. J. 51(4):876-884.
Šimçnek, J., M. Šejna, and M. Th. van Genuchten. 1996. The HYDRUS-2D software package
for simulating water flow and solute transport in two-dimensional variably saturated media.
Version 1.0, IGWMC-TPS - 56, International Ground Water Modeling Center, Colorado
School of Mines, Golden, CO.
Šimçnek, J., and M. Th. van Genuchten. 1996. Estimating unsaturated soil hydraulic properties
from tension disc infiltrometer data by numerical inversion. Water Resour. Res. 32(9):2683-
2696.
Šimçnek, J., and M. Th. van Genuchten. 1997. Estimating unsaturated soil hydraulic properties
from multiple tension disc infiltrometer data. Soil Sci. 162(6):383-398.
Šimçnek, J., R. Angulo-Jaramillo, M. Schaap, J.-P. Vandervaere, and M. Th. van Genuchten.
1998a. Using an inverse method to estimate the hydraulic properties of crusted soils from
tension disc infiltrometer data. Geoderma 86(1-2):61-81.
Šimçnek, J., M. Šejna, and M. Th. van Genuchten. 1998b. The HYDRUS-1D software package
for simulating water flow and solute transport in two-dimensional variably saturated media.
Version 2.0. IGWMC - TPS - 70. International Ground Water Modeling Center, Colorado
School of Mines, Golden, CO.
Šimçnek, J., M. Th. van Genuchten, M. M. Gribb, and J. W. Hopmans. 1998c. Parameter
estimation of unsaturated soil hydraulic properties from transient flow processes. Soil &
Tillage Research 47/1-2:27-36.
Šimçnek, J., O. Wendroth, and M. Th. van Genuchten. 1998d. A parameter estimation analysis of
the evaporation method for determining soil hydraulic properties. Soil Sci. Soc. Am. J.
62(4):894-905.
Šimçnek, J., R. Kodešová, M. M. Gribb, and M. Th. van Genuchten. 1999a. Estimating
hysteresis in the soil water retention function from modified cone penetrometer test. Water
Resour. Res. (in press).
Šimçnek, J., O. Wendroth, and M. Th. van Genuchten. 1999b. Soil hydraulic properties from
laboratory evaporation experiments by parameter estimation, In: M. Th van Genuchten, F. J.
Leij, and L. Wu (eds.), Proc. Int. Workshop, Characterization and Measurement of the
Hydraulic Properties of Unsaturated Porous Media, pp. 713-724, University of California,
Riverside, CA.
Vachaud, G. 1968. Contribution to the study of flow problems in unsaturated porous media,
Ph.D. Thesis, A.O. 2655, School of Sciences, University of Grenoble, France.
1538 Òimçnek and van Genuchten

van Dam, J. C., J. N. M. Stricker, and P. Droogers. 1992. Inverse method for determining soil
hydraulic functions from one-step outflow experiment. Soil Sci. Soc. Am. Proc. 56:1042-
1050.
van Dam, J. C., J. N. M. Stricker, and P. Droogers. 1994. Inverse method to determine soil
hydraulic functions from multistep outflow experiment. Soil Sci. Soc. Am. Proc. 58:647-652.
van Genuchten, M. Th. 1980. A closed-form equation for predicting the hydraulic conductivity of
unsaturated soils. Soil Sci. Soc. Am. J. 44:892-898.
van Genuchten, M. Th., and F. J. Leij. 1992. On estimating the hydraulic properties of
unsaturated soils, In: M. Th. van Genuchten, F. J. Leij and L. J. Lund (eds.), Indirect
Methods for Estimating the Hydraulic Properties of Unsaturated Soils, pp. 1-14, Univ. of
California, Riverside, CA.
van Genuchten, M. Th., and R. J. Wagenet. 1989. Two-site/two-region models for pesticide
transport and degradation: Theoretical development and analytical solutions. Soil Sci. Soc.
Am. J. 53:1303-1310.
van Genuchten, M. Th., and P. J. Wierenga. 1976. Mass transfer studies in sorbing porous media,
I. Analytical solutions. Soil Sci. Soc. Am. J. 40:473-481.
Wooding, R. A. 1968. Steady infiltration from large shallow circular pond. Water Resour. Res.
4:1259-1273.
Yeh, W. W-G. 1986. Review of parameter identification procedures in groundwater hydrology:
The inverse problem. Water Resour. Res. 22(2):95-108.