
Since the particle trajectory is given by
X
i
t ; a; t
0
ð Þ¼ a
i
þ
Z
t
t0
v
i
X t
0
; a; t
0
ð Þ ð Þdt
0
;
from (11) and (4) it follows that
Y
ii
¼ hhX
2
i
i
a
ihR
2
i
i ha
2
i
i
a
ha
i
i
2
a
h i
M
ii
¼hS
ii
i S
ii
0 ð Þ M
ii
: ð12Þ
[12] The agreement of the first-order estimations with the
Monte Carlo simulations suggests that the quantity pre-
sented in Figures 10–12 of Darvini and Salandin [2006]
might be an estimation of Y
ii
. With the first-order approx-
imation (1)–(2), this estimation is obtained from (11) by
replacing the argument X of v
i
by X
0
. However, as follows
from the definition of the uncertainty of the second moment
with respect to the one-particle displacements covariance
[Darvini and Salandin, 2006, p. 12], one can also suppose
that the authors estimated hX
ii
i
a
R
ii
. This can be achieved
by replacing in (11) v
i
by v
i
0
and X by X
0
, as in relations
(22) and (23) of Darvini and Salandin [2006]. From
(12) and (6) we have Y
ii
= hX
ii
i
a
R
ii
+ Q
ii
. Thus the
estimation of the second moment by hX
ii
i
a
R
ii
disregards
not only the memory term M
ii
(as the numerical approach
based on (12) does) but also the correlation of the mean
velocity Q
ii
(accounted for by the numerical simulations).
This choice is justified if a highly accurate numerical
solution for the zeroth-order potential head [Darvini and
Salandin, 2006, equation (6)] was obtained or if no numer-
ical solution was computed at all and a constant head
gradient, as resulting from boundary conditions, was im-
posed. This leads to a constant zeroth-order velocity and to
the cancellation of both M
ii
and Q
ii
in (9) – (10). In this case,
the comparison with the simulation results [Darvini and
Salandin, 2006, Figure 11] indicates that the variance of
the mean Lagrangian velocity near the left boundary,
corresponding to the finite element solution of the exact flow
equations in the Monte Carlo simulations, does not produce
significant terms Q
ii
.
[13] If, instead of following exactly the same approach as
Salandin and Fiorotto [1998], the expected second moment
was computed from the mean square displacement of the
simulated particles trajectories, then the Monte Carlo results
correspond to the exact relation (6). Hence, even if Q
22
can
be negligible, significant memory terms M
22
, as in our
example presented in Figure 1, can occur for the transverse
line source placed near the left boundary. Figure 11 of
Darvini and Salandin [2006] shows that this was not the
case. To help the interested reader to understand their
results, the authors should explain how the expected second
moment was computed in both the first-order stochastic
finite element approach and Monte Carlo simulations. In
addition, we suggest that in a reply to this comment the
authors of the paper commented here should present some
results on the path line analysis of the mean Lagrangian
velocity field. This can be achieved by evaluations of the
correlation terms (9) – (10), which are readily obtainable by
their approach, or, as in our example from Figure 2, by
presenting the dependence of the numerically derived mean
Lagrangian velocity on initial positions of the particles. If
one proves that such dependencies can be neglected, the
effect of spatial inhomogeneity of the random velocity field
caused by the finite size of the domain, considered by
Darvini and Salandin [2006], can be completely described
in terms of one- and two-particle velocity covariances.
[14] For physically relevant statistical inhomogeneity of
the velocity field, as that caused by a trend in the mean
hydraulic conductivity, the bias of the expected second
moment hS
ii
i S
ii
(0) with respect to the space average of
the one-particle variance hX
ii
i
a
is given, according to (6), by
M
ii
+ Q
ii
R
ii
. Since in the first-order approximation M
ii
(9)
and Q
ii
(10) quantify the effect of the space variation of the
mean Eulerian velocity, these terms are particularly relevant
for the inhomogeneous case. To complete the assessment of
ergodicity with respect to hX
ii
i
a
, in addition to the bias of
the expectation, the standard deviation of S
ii
must be
computed as well [Suciu et al., 2006a]. Both tasks can be
achieved with the first-order stochastic finite element method
of Darvini and Salandin [2006]. By solving a deterministic
problem for given mean and correlation functions of the
hydraulic conductivity, such an approach avoids cumber-
some repeated Monte Carlo computations.
[15] Acknowledgments. This work was supported by Deutsche
Forschungsgemeinschaft (grant SU 415/1-2) and Romanian Ministry of
Education and Research (grant 2-CEx06-11-96). We thank F. Radu for
helpful discussions.
References
Bellin, A., P. Salandin, and A. Rinaldo (1992), Simulation of dispersion in
heterogeneous porous formations: Statistics, first-order theories, conver-
gence of computations, Water Resour. Res., 28(9), 2211 – 2227.
Dagan, G. (1990), Transport in heterogeneous porous formations: Spatial
moments, ergodicity and effective dispersion, Water Resour. Res., 26,
1281 – 1290.
Darvini, G., and P. Salandin (2006), Nonstationary flow and nonergodic
transport in random porous media, Water Resour. Res., 42, W12409,
doi:10.1029/2005WR004846.
Eberhard, J., N. Suciu, and C. Vamos¸ (2007), On the self-averaging of
dispersion for transport in quasi-periodic random media, J. Phys. A Math.
Theor., 40, 597 – 610, doi:10.1088/1751-8113/40/4/002.
Fiori, A., and I. Jancovic´ (2005), Can we determine the transverse macro-
dispersivity by using the method of moments?, Adv. Water Resour., 28,
589 – 599, doi:10.1016/j.advwaters.2004.09.909.
Salandin, P., and V. Fiorotto (1998), Solute transport in highly heteroge-
neous aquifers, Water Resour. Res., 34(5), 949 – 961.
Sposito, G., and G. Dagan (1994), Predicting solute plume evolution in
heterogeneous porous formations, Water Resour. Res., 30(2), 585 – 589.
Suciu, N., C. Vamos¸, J. Vanderborght, H. Hardelauf, and H. Vereecken
(2006a), Numerical investigations on ergodicity of solute transport in
heterogeneous aquifers, Water Resour. Res., 42, W04409, doi:10.1029/
2005WR004546.
Suciu, N., C. Vamos¸, and J. Eberhard (2006b), Evaluation of the first-order
approximations for transport in heterogeneous media, Water Resour. Res.,
42, W11504, doi:10.1029/2005WR004714.
Vamos¸, C., N. Suciu, and H. Vereecken (2003), Generalized random walk
algorithm for the numerical modeling of complex diffusion processes,
J. Comput. Phys., 186(2), 527 – 544, doi:10.1016/S0021-9991(03)00073-1.
N. Suciu, Institute of Applied Mathematics, Friedrich-Alexander
University of Erlangen-Nuremberg, Martensstrasse 3, D-91058 Erlangen,
Germany. (suciu@am.uni-erlangen.de)
C. Vamos¸, ‘‘T. Popoviciu’’ Institute of Numerical Analysis, Romanian
Academy, P. O. Box 68-1, 400320 Cluj-Napoca, Romania. (cvamos@
ictp.acad.ro)
4 of 4
W12601 SUCIU AND VAMOS¸: COMMENTARY W12601
19447973, 2007, 12, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2007WR005946 by Cochrane Romania, Wiley Online Library on [28/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License