Macropores and water flow in soils revisited
Keith Beven
1,2,3
and Peter Germann
4
Received 8 February 2012; revised 16 February 2013; accepted 18 February 2013; published 4 June 2013.
[1] The original review of macropores and water flow in soils by Beven and Germann is
now 30 years old and has become one of the most highly cited papers in hydrology. This
paper attempts to review the progress in observations and theoretical reasoning about
preferential soil water flows over the intervening period. It is suggested that the topic has
still not received the attention that its importance deserves, in part because of the ready
availability of software packages rooted firmly in the Richards domain, albeit that there is
convincing evidence that this may be predicated on the wrong experimental method for
natural conditions. There is still not an adequate physical theory linking all types of flow,
and there are still not adequate observational techniques to support the scale dependent
parameterizations that will be required at practical field and hillslope scales of application.
Some thoughts on future needs to develop a more comprehensive representation of soil
water flows are offered.
Citation: Beven, K., and P. Germann (2013), Macropores and water flow in soils revisited, Water Resour. Res., 49, 3071–3092,
doi:10.1002/wrcr.20156.
1. Introduction
[2] The general topic of macropore flow in soils and sim-
ilar permeable media, and related topics like preferential
flow, nonequilibrium flow, and dual-porosity flow contin-
ues to increase in popularity among researchers in various
fields. Gerke et al. [2010], for instance, reported a decadal
increase of more than 50% of annually published relevant
papers on these topics. The review of macropores and water
flow in soils of Beven and Germann [1982] (BG82 from
hereon) continues to be cited and is now one of the most
frequently referenced papers in hydrology journals [Kout-
soyiannis and Kundzewicz, 2007]. The review paper
stemmed from work that was started at the Institute of
Hydrology, Wallingford, UK, during a study visit by PG
supported by the Swiss National Science Foundation
[Figure 1; Germann and Beven, 1981a, 1981b ; Beven and
Germann, 1981] and continued when we were both based
at the University of Virginia in Charlottesville. We shared
the experience from the UK and Switzerland of faster infil-
tration and downslope macropore flows than the Richards
[1931] equation would normally predict.
[
3] At that time, there had already been other reviews
about the influence of macropores on water flows in the
soil profile [Thomas and Phillips, 1979; Bouma, 1981a],
discussions of microporosity, mesoporosity and macro-
porosity [Luxmoore, 1981; Bouma, 1981b; Skopp, 1981;
Beven1981a], and recognition of the much earlier work of
Schumacher [1864] and Lawes et al. [1882]. We also now
know that Robert Horton was aware of the work of Lawes
et al. and, in an unpublished monograph on infiltration, rec-
ognized the importance of macropores in both water and
air flows in the upper layers of the soil [Beven, 2004a]. He
also rejected the idea of profile (capillary gradient) controls
on infiltration rates in favor of surface controls. He saw his
infiltration equation as an ‘‘extinction’’ equation, with sur-
face processes restricting inputs of water into larger flow
pathways [Beven, 2004a].
[
4] In 1982, however, the vast majority of soil hydrolo-
gists and soil physici sts still adhered strongly to the Rich-
ards approach to water movement and our experience was
that papers suggesting a need for alternative concepts
tended to be refereed quite harshly. Only 10 years later,
400 people turned up at the ASAE special meeting on Pref-
erential Flow in Soils in Chicago [Gish and Shirmoham-
madi, 1991]. A major reason for this change was the need
to understand why pesticides and other pollutants that
should sorb strongly onto soil particles in the near surface
were being found widely in routine water quality observa-
tions of groundwaters and field drains. Even if the pollu-
tants had sorbed onto fine colloidal materials, it was
difficult to understand how there could be continuous flow
pathways that would allow even fine particles to reach sig-
nificant depths without being filtered out by the soil matrix.
It seemed then that there would be a real impetus towards a
more realistic approach to representing preferential water
movement in field soils and exploring the implications at
larger hillslope and catchment scales. Thus, 30 years after
the BG82 review, it seems worthwhile to as sess the innova-
tions in the stud y of soil water processes at different scales
since that time. So what progress has been made?
[
5] Well, there has certainly been lots of activity in the
field but the dominant concept of soil physics in recent
hydrological textbooks remains the Darcy-Richards equa-
tion [e.g., Brutsaert, 2005 ; Shaw et al., 2010]. The domi-
nant concept underlying ‘‘physically based’’ hydrological
1
Lancaster Environment Centre, Lancaster University, Lancaster, UK.
2
Department of Earth Sciences, GeoCentrum, Uppsala University,
Sweden.
3
ENAC Laboratoire d’Ecohydrologie, EPFL Lausanne, Switzerland.
4
Geographisches Institut, Universit
€
at Bern, Bern, Switzerland.
Corresponding author: K. Beven, Lancaster Environment Centre, Lan-
caster LA1 4YQ, UK. (K.Beven@lancaster.ac.uk)
©2013. American Geophysical Union. All Rights Reserved.
0043-1397/13/10.1002/wrcr.20156
3071
WATER RESOURCES RESEARCH, VOL. 49, 3071–3092, doi :10.1002/wrcr.20156, 2013
models remains the Darcy-Richards equation [e.g., Loague
et al., 2006 ; Qu and Duffy, 2007; Ivanov et al., 2008]. The
widespread use of pedotransfer functions (mostly derived
from experiments on small soil samples) presumes that the
Darcy-Richards equation holds at larger scales of applica-
tion, and that the parameter values are constant in time and
space for a given soil horizon [see, for example, Wösten,
1999; Acutis and Donatelli, 2003; Schaap et al. 2001;
McBratney et al., 2002]. The dominant concept in solute
transport remains the advection-dispersion equation, nor-
mally implement ed as a symmetric Gaussian velocity dis-
tribution around a local mean pore water velocity derived
from Darcy or Darcy-Richards theory. If preferential flows
are important, this will not properly represent their effect
on transport, even if the mean velocities are of the correct
magnitude. A reason for continuing to use such models has
been expressed in terms of understanding why they fail in
comparison with field observations, so that they can be
improved in future [Ebel and Loague, 2006; Loague et al.,
2006; James et al., 2010] even though there should already
long have been an expectation of failure based on the field
tracer evidence and from a consideration of the physics of
Darcy-Richards flow in unsaturated heterogeneous soils
[Beven, 1989, 2001].
[
6] So why has the Darcy-Richards representation of
flow in partially saturated soils remained so popular, de-
spite the evidence? One reason surely is that computing
power has increased to such an extent in the last thirty
years that the use of software packages to solve the Rich-
ards equation has become comfortable for a wide range of
users. In addition, modifications of the Richards approach
evolved to take some account of preferential flows. Gener-
ally accessible codes, like HYDRUS now provide the capa-
bility for applying different dual porosity and dual
permeability representations [
Simu
°
nek and van Genuchten,
2008; Radcliffe and
Simu
°
nek, 2010]. Other codes have
been developed to represent preferential flows as an addi-
tional nonequilibrium flow component but often without a
secure physical foundation (see later under Theoretical
Studies).
[
7] Arguably, the availability of such tools has diverted
attention from more fundamental research on macropore and
preferential flow, but there has been a wide range of experi-
mental and modeling studies published on these topics since
BG82. These have predominantly been at the profile to ly-
simeter or small plot scale, where vertical flows dominate.
In moving from the profile scale to hillslope scale subsurface
stormflows it is necessary to consider the integration of verti-
cal and lateral or downslope flows, including the effects of
the capillary fringe on celerities and displacement of stored
water mentioned in BG82. It is important to differentiate
between pore water velocities that control transport proc-
esses, and celerities that control the propagation of perturba-
tions through the system and therefore control hydrograph
responses [e.g., Beven, 2012, chap. 5]. At local scales, a plot
or field is often treated as a collection of vertical columns
for both flow and transport calculations [e.g., Beven et al.,
2006], but in shallow soils, or where the water table is close
to the surface, spatial interactions become important. In par-
ticular, the importance of macropores and preferential flows
in facilitating rapid subsurface stormflow responses has been
incorporated into developing perceptual models of hillslope
hydrology (See section 5 below). There is now a better
appreciation of the different roles of wave celerities and
water velocities in these interactions (though note that the
discussion of ‘‘translatory waves’’ goes back at least to the
1930s, see Beven [2004b]).
[
8] As noted earlier one of the primary drivers for inter-
est in preferential flows and macropores in soils was the
problem of explaining how pesticides and other sorbing
pollutants were being transported to field drains, ground-
waters and rivers [Flury, 1996]. This interest has not
abated, and there are many recent papers that address this
problem [see, for example, Kladivko et al., 2001 ; Zehe and
Fl
€
uhler, 2001a; Reichenberger et al., 2002; Ciglasch et
al., 2005 ; Beven et al., 2006], including the transport of
viruses and bacteria [Darnault et al., 2003; Germann et al .
1987]; the role of colloids in facilitating transport of sorb-
ing compounds through large soil pores [Villholth et al.,
2000; Germann et al., 2002a]; and factors such as water
repellency in inducing preferential flow in certain locations
[e.g., Bauters et al., 1998 ; Blackwell, 2000 ; Cerd
aetal.
1998]. The importance of macropore connectivity has also
been demonstrated [Andreini and Steenhuis, 1990 ; Allaire-
Leung et al., 2000 ; Rosenbom et al., 2008].
[
9] This has also become a regulatory issue. The licens-
ing of potentially harmful products depends on som e
assessment of how easily they are transported in the envi-
ronment. That assessment, however, must be done across a
wide range of soil characteristics, land management, and
soil water conditions. Classification or indexing [e.g., Rao
et al., 1985; Quisenberry et al., 1993; Schlather and
Huwe, 2005; Sinkevich et al., 2005 ; Stenemo et al., 2007;
McGrath et al., 2009] and modeling [Steenhuis et al.,
1994; Jarvis et al., 1994, 1997 ; Stewart and Loague, 1999]
approaches have been used to assess the relative risk of
transport of different types of products that might be harm-
ful to the environment in ways that reflect the possibility of
preferential flow.
Figure 1. Red filtered photograph of Rhodamine WT
tracing (darkest black tones) from the soil surface to a mole
drain in a heavy clay subsoil (Denchworth series) at Gren-
don Underwood, Oxford, UK; the first field tracing experi-
ment on preferential flow by Beven and Germann in March
1979. In this soil, there is some fingering of tracer in the A
horizon, but below this the tracer takes pathways around
the soil peds without penetrating (as do the grass roots).
BEVEN AND GERMANN : REVIEW
3072
[10] The questions raised in BG82 about ‘‘When does
water flow through macropores in the soil? How does water
flow through macropores in the soil? How does water in a
macropore interact with water in the surrounding soil?
How important are macropores in terms of volumes of flow
at the hill slope or catchment scale ? What are the implica-
tions of macropores for movement of solutes and chemical
interactions in the soil?’’ are still relevant today [see also
Jury, 1999] but relatively few testable concepts have
evolved since that time. Although all of these questions
have been worked on during the last 30 years, revisiting
BG82 may offer an opportunity to reassess the achieve-
ments and the needs in this area of soil physics.
2. The Aims of This Paper
[11] Preferential flows of different types have been the
subject of significant interest over this period and there
have been a number of review articles and special issues of
journals published in the last decade or so [see Bryan and
Jones, 1997 ; Blackwell, 2000 ; Sidle et al., 2000; Uchida et
al., 2001 ;
Simu
°
nek et al., 2003; Gerke, 2006; Jarvis,
2007; Coppola et al., 2009; Allaire et al., 2009; Köhne
et al., 2009a, 2009b; Clothier et al ., 2008 ; Morales et al.,
2010; Beven, 2010; Chappell, 2010; Jones, 2010 ;
Hencher, 2010 ; Bachmair and Weiler, 2011]. Our intention
here is not to repeat the information presented in these
papers, but rather to focus on the issues that remain and
make some suggestions about ways in which they might be
addressed. Throughout, given the large body of relevant lit-
erature, the choice of citations is intended to be illustrative
of the issues rather than complete. We will suggest more
radical alternatives to the ‘‘safe’’ Darcy-Richards based
concepts incorporated into dual-continuum and dual-per-
meability flow models. We will follow the broad structure
of the original paper in moving from local scale to catch-
ment scale issues, and differentiating experimental evi-
dence from theoretical studies, with the aim of encouraging
a more realistic physical basis for future studies on water
flows in soils.
3. The Occurrence of Preferential Flow in Soils
[12] BG82 cautiously staked out some common ground
by stating that : ‘‘There has long been speculation that large
continuous openings in field soils (which we will call mac-
ropores) may be very important in the movement of
water—at least under certain conditions. Such voids are
readily visible, and it is known that they may be continuous
for distances of at least several meters in both vertical and
lateral directions.’’ The implication then was that water
flow in such readily visible pores would be subject to rather
different process controls than water flow in the soil matrix.
There is basic agreement that soil macropores originate
from the processes of desiccation, growth and decay of
roots and mycelia, and burrowing animals [BG82, and
more recently Coppola et al., 2009 and Bachmair and
Weiler, 2011]. Extension to consideration of sediments and
rocks may include fissures and karst formations that have
been the subject of interest in hydrogeology, radioactive
waste disp osal and petroleum engineering. Soil features,
including cracks, root channels, worm holes and other bio-
logically induced macropores, are typically restricted to the
depth of the soil profile, but may link to openings in the
regolith and bedrock which may extend over tens or hun-
dreds of meters in fractured rock s ystems. Dubois [1991],
for instance, injected dye into the granite formation approx-
imately 2000 vertical meters above the Mt. Blanc tunnel
between France and Italy. About a hundred days later he
found the tracer in water samples taken from the tunnel’s
drainage system. Hillslope tracer measurements also sug-
gest connected flow pathways extending over at least tens
of meters [e.g., Nyberg et al., 1999; Weiler and McDon-
nell, 2007; McGuire et al., 2007 ; Graham and McDonnell,
2010].
[
13] There is an important point to be made here about
the Darcy and Richards equations. Both are based on exper-
imentation under particular conditions. Both are consistent
with those conditions by back-calculation of a coefficient
or function of proportionality that we now call the hydrau-
lic conductivity. In the case of Darcy [1856], who looked
at saturated conditions in steel cylinders filled with sorted
sand, the linearity between hydraulic gradient and flux rate
should hold in the field, provided that the flow remains in
the laminar and stable flow regime (although Darcy’s data
were only very nearly linear, Davis et al. [1992]). In the
case of Richards [1931], who looked at unsaturated condi-
tions imposed by sequentially decreasing capillary pressure
in a confined sample using a hanging water column, the ex-
perimental conditions preclude preferential flow in larger
soil pores, which are a priori drained at each step of the
decrease in capillary pressure. By then imposing a pressure
gradient, Richards experimentally created a steady flux at
different points on the retention curve.
[
14] Richards [1931, p. 322] explained: ‘‘When the con-
ditions for equilibrium under gravity ... are fulfilled, the
velocity and acceleration of the capillary liquid are every-
where zero .... which means that the force arising from the
pressure gradient just balances gravity.’’. He used pressure
interchangeably with capillary potential in unsaturated po-
rous media. Because neither velocity nor acceleration
occurs, hydrostatic conditions are assumed according to the
material-specific retention curve. Further ‘‘If this condit ion
does not obtain there will be a resultant water moving force
and in general there will be capillary flow.’’ (p. 322).
Thus, in his view (and resulting equations), only gravity
and capillarity drive capillary flow, while the flow-restrain-
ing forces are summarized in the bulk ‘‘ ... factor of pro-
portionality ...’’ (p. 323) as Richards called the hydraulic
conductivity.
[
15] While Richards’ equation may be valid within that
experimental framework, it should not be a surprise that the
concept may not carry over to field conditions when the air
pressure within well-aerated soils is atmospheric, the soil is
heterogeneous in its characteristics and the fluxes are sub-
ject to dynamic effects (including preferential flows). Al-
ready in the capillary range of pore sizes, where it is
commonly assumed that a Richards type of description is
appropriate, there is a problem in applying the theory at
scales of interest in heterogeneous, unsaturated soils. The
problem is that capillary potentials are then rarely in equi-
librium such that in there is no consistent hydraulic gradi-
ent, even in the absence of preferential flows. Since the
constant of proportionality is a nonlinear function of
BEVEN AND GERMANN : REVIEW
3073
capillary potential or water content, it will also vary with
the heterogeneity. Thus, at the scale of a useful landscape
element, the heterogeneity means that the Richards equa-
tion will not hold the physics tells us that some other form
of equation should be used even if the Richards equation
holds at the small scale. Some attempts have been made to
define such an equation given assumptions about the heter-
ogeneity [e.g., Yeh et al., 1985a, 1985b; Russo, 1995,
2010] or to simulate the effect of randomly generated fields
[Mantoglou and Gelhar, 1987; Binley et al., 1989; Binley
and Beven, 1992; Russo et al., 2001; Fiori and Russo,
2007, 2008]. Basically, however, knowledge about the het-
erogeneity will never be available so that the impact of het-
erogeneity has been consistently ignored. Richards
equation is applied as if the soil were homogeneous at the
scale of the calculation element. Similar considerations
apply when equilibrium, dual porosity or dual permeability
representations of preferential flow are applied within a
Richards domain (see below). This is not a new insight, it
has been known for decades [e.g., Dagan and Bresler,
1983], but with very little effect on modeling practice [see
discussion in Beven, 1989, 2001, 2006, 2012]. Given these
limitations of the Richards approach when used in practice
(that relationships derived under steady equilibrium
assumptions should not be expected to apply in dynamic
cases and that local equilibration will not apply in heteroge-
neous unsaturated soils) we hope to encourage a radical
rethink of how preferential flows are represented at the
plot, field, hillslope and catchment scales of interest. We
suggest that the Richards approach to representing fluxes in
field soils should not be considered to be physically based,
but a convenient conceptual approximation.
[
16] Flow in the structural voids of the soil will result in
nonequilibrium conditions when water cannot move fast
enough into the smaller pores of the surrounding matrix to
spontaneously and continuously achieve equilibrium
according to the retention curve. Germann et al. [1984], for
instance, reported bromide concentrations to decrease with
increasing horizontal distance from stained macropores.
Jarvis [2007] describes nonequilibrium conditions as fol-
lows: ‘‘As water starts to flow into large structural macro-
pores, the sharp contrast in pore size and tortuosity with the
surrounding textural pores leads to an abrupt increase in
water flow rate for only a small increase in soil water pres-
sure. The resulting non-uniform flow (physical non-equilib-
rium) can be illustrated by imagining a soil block that
contains macropores wetting up towards saturation during
infiltration.’’ This implies that macropores carry water quite
independently from antecedent soil moisture and capillary
flow such that (at least) a dual-pore structure containing the
matrix and the preferential flow supporting macropores is
needed.
4. Experimental Evidence
4.1. Profile and Plot Scale
[
17] Some of the most convincing evidence of preferen-
tial flow in soils has come from the use of tracers and dye
staining. Most of these experiments have been carried out
at the profile and plot scales [e.g., Figure 1 ; Flury et al.,
1994; Abdulkabir et al., 1996; Henderson et al., 1996;
Villholth et al., 1998; McIntosh et al., 1999; Kung et al.,
2000, 2001b; Zehe and Fl
€
uhler, 2001a, 2001b; Stamm et
al., 2002 ; Weiler and Naef, 2003b; Weiler and Fl
€
uhler,
2004; Bachmair et al., 2009; Blume et al., 2009; van
Schaik, 2009; and many others]. Preferential flow has to be
inferred from tracer data ; Kung et al. [2000] for example
conclude from the similar early breakthroughs of both sorb-
ing and conservative tracers that flow must be through pref-
erential pathways. Dye staining may not reveal all
preferential pathways; only those that have been connected
to a source of dyed input water. It is in fact quite common
in such studies to see readily visible macropores that are
not stained, or have been only partly stained. This may be
because of a lack of connectivity to the source, or because
they were not flowing at capacity, or that they contained
water displaced from the matrix by the input water.
[
18] Watson and Luxmoore [1986] were probably the
first to apply tension infiltrometers to relate flow rates to
pore diameters in situ (according to Clothier and White
[1981]). They applied water to the soil surface at preset
capillary potentials of 0, 3, 6, and 15 hPa, thus imple-
menting a Dirichlet-boundary condition. An analysis of
their results suggested that 95% of flow occurred in pores
that were wider than 250 m, which occupied just 0.32%
of the pore volume. These field data support the model
results of Beven and Germann [1981] who demonstrated
the paramount contribution to flow by only a very small
portion of a soil volume containing wide continuous voids.
The resulting parameters are frequently related with the
corresponding water contents in order to construct an appa-
rently physical relationship. These sequential steady state
experiments are consistent with the equilibrium conditions
of Richards. However, the imposed Dirichlet boundary
condition is far from realistic. Water typically arrives at the
ground surface either in discrete drops, as in rain, in
streams as in concentrated stem flow, or ponded in surface
hollows, such that pressure in the arriving water remains
positive or at least atmospheric. Thus, preference should be
given to the more realistic flux-controlled infiltration
observing a Neumann boundary condition, albeit that the
spatial pattern of fluxes might be diffic ult to assess under
other than controlled experimental conditions as a result of
throughfall, stemflow, surface irregularities, collection area
to a surface connected macropore, other surface controls,
etc. [e.g., Weiler and Naef, 2003a]. This raises the question
when and where in a permeable medium significant capil-
lary forces start to act on the infiltrating water. Wide
enough macropores will preclude any significant effect of
capillarity. Turbu lent pipe flows [see, for instance, Jones,
2010] are the most impressive representation of such flows.
Capillary sorption of water has still to be considered
regardless of the width of the preferred flow path but may
be restricted by the local microstructure of the soil (such as
the cutans of translocated clay particles, earthworm mucus
at the edges of larger pores, or hydrop hobic excretions by
plants, e.g., Cerd
aetal. [1998]).
[
19] Other configurations may also lead to flows that
occur close to atmospheric pressure in nonsaturated perme-
able media. Germann and al Hagrey [2008], for example,
reported from studies of data collected in the 2 m deep
sand tank in Kiel that the capillary potential collapsed to
close to atmospheric pressure behind the progressing wet-
ting front. The tank was uniformly filled with sand in the
BEVEN AND GERMANN : REVIEW
3074
texture range of 63–630 m, and the buildup of any macro-
pore-like structures was purposefully avoided during the
filling process [al Hagrey et al., 1999]. Tensiometers and
time domain reflectrometry (TDR) [see, for instance, Topp
et al., 1980] were used to record capillary potentials and
volumetric water contents at nine depths. Depth averages
of the initial and maximum water contents were 0.08 and
0.26 m
3
m
3
, respectively, while the maximum degree of
saturation was 0.54 of the porosity of 0.47 m
3
m
3
, demon-
strating that water contents were always far from satura-
tion. But maximum capillary potentials during infiltration
were between 17 and 25 hPa, indicating that infiltration
behind the wetting front occurred close to atmospheric
pressure. The data from tensiometers and TDR waveguides
showed that the wetting front moved with a constant aver-
age velocity of 3.2 10
5
ms
1
from top to bottom during
the 16 h of sprinkler infiltration. The average wetting front
velocity followed from the slope of the linear relationship
between the depth of the wetting front and its respective ar-
rival time with a coefficient of determination of 0.994 (Fig-
ure 2). Richards equation will predict a linear propagation
of the wetting front only if the effect of the capillary gradi-
ent is negligible relative to gravity but this should not be
the case, of course, for wetting of a dry soil.
[
20] A variety of other researchers have found constant
velocities of wetting fronts in sand boxes, including in the
absence of natural macropores (see Table 1). The boxes of
Selker et al. [1992] and Hincapi
e and Germann [2010]
were 0.9 m and 0.4 m long, respectively. At larger scales,
Rimon et al. [2007] and Ochoa et al. [2009] found in situ
constant wetting front velocities due to percolation in
undisturbed sediments to depths of 21 and 3 m. Constant
wetting front velocities that persist over prolonged time
periods and considerable depths are a strong indication of
non-Richards behavior during infiltration and percolation
demanding a fundamentally different approach to the repre-
sentation of the flow. Figure 3 shows the results of an anal-
ysis of 215 such sets of observations in the form of a
distribution of derived wetting front velocities.
4.2. Field Scale
[
21] Studies at the profile and plot scale are already diffi-
cult. At the field scale, additional complexity is introduced.
Transport to field drains, groundwaters, and rivers will
depend on preferential flows induced by the heterogeneity of
soil properties, including in some cases natural pipes or agri-
cultural drains (e.g., Figure 1). Such heterogeneities are not
necessarily obvious from the soil surface and may be the
result of the history of soil development and land use, ranging
from deep tertiary weathering and Holocene periglacial struc-
tures to recent land management and drainage histories. Some
larger scale staining experiments have been carried out [e.g.,
Kung, 1990; Noguchietal., 1999; Wienhöfer et al., 2009;
Anderson et al., 2009a, 2009b] but the evidence for the im-
portance of preferential flows at this and larger scales tends to
be indirect, inferred from the bulk responses of natural or arti-
ficial tracer concentrations at some measurement point [e.g.,
Hornberger et al., 1990; Nyberg et al., 1999; Rodhe et al.,
1996, Kienzler and Naef,2008;McGuire and McDonnell,
2010]. Kung et al. [2005, 2006], however, show how tracer
experiments to field drains under different imposed flux rates
can be interpreted, under laminar flow assumptions, in terms
of consistent distributions of effective pore radii.
[
22] One particular area of interest in respect of preferen-
tial flow at the field scale is for understanding and predict-
ing groundwater recharge. There have been many recharge
studies where observations of rapid water table responses
have been reported, or where tracers or pollutants have
been reported at depths much greater than would be
expected without inferring preferential flow. A recent
example is the study of Cuthbert et al. [2010], where rapid
groundwater resp onses in a sandstone aquifer overlain by a
till soil, even during summer periods, were interpreted as
recharge that occurred when a dry soil matrix was bypassed
by preferential flow throu gh fractures in the clay.
[
23] Macropores are considered to be the soil structures
most vulnerable to mechanical compaction due to heavy
machinery used in agriculture, forestry, and on construction
sites [e.g., Chappel l, 2010]. In addition to the earlier papers
cited in BG82, Alaoui and Helbling [2006] investigated
with sprinkler infiltration and dye-staining tests the effects
of cattle trampling, driving with a six-row sugar beet har-
vester, and soil reconstruction procedures on macropore
flow. They concluded that trafficking sealed the soil surface
almost completely. Cattle trampling, on the other hand,
negatively affected mostly fine pores, while increasing
macroporosity in some instances. Infiltration into recon-
structed soils was mainly through the soil matrix because
no macropores had formed since reconstruction.
4.3. Hillslope Scale
[
24] The transition from preferential vertical flow in soil
profiles to preferential lateral flow i.e., subsurface
Figure 2. Linear progression of a wetting front into the
Kiel sand tank under steady artificial rainfall at a rate of
15.6 mm h
1
applied for a period of 16 h 17 min. [from
Germann and al Hagrey, 2008].
BEVEN AND GERMANN : REVIEW
3075
![Figure 3. Distribution of 215 wetting front velocities determined in situ for 25 soils covering Udalf, Ochrept, Umbrept, Aquod, Ferrod, Humod, and Udult suborders from Germann and Hensel [2006]; and wetting front velocities from (1) al Hagrey et al. [1999], (2) Dubois [1991], and (3) Selker et al. [1992].](/figures/figure-3-distribution-of-215-wetting-front-velocities-3rkydej4.webp)
![Figure 4. Wetting front velocity, v, as function of the input rate, q, according to equation (A9). (Data from Hincapi e and Germann [2009b]; figure from Germann [2013], with permission.)](/figures/figure-4-wetting-front-velocity-v-as-function-of-the-input-n0nlilg1.webp)