Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Advances in Water Resources 131 (2019) 103384

Contents lists available at ScienceDirect

Advances in Water Resources


journal homepage: www.elsevier.com/locate/advwatres

Dynamic evolution of the soil pore size distribution and its connection to
soil management and biogeochemical processes
Norman Pelak a,b,c,∗, Amilcare Porporato b
a
Department of Civil and Environmental Engineering, Duke University, Durham, NC, USA
b
Department of Civil and Environmental Engineering, Princeton University, Princeton, NJ, USA
c
Sierra Nevada Research Institute, University of California, Merced, CA, USA

a r t i c l e i n f o a b s t r a c t

Keywords: Soil properties are determined by a complex arrangement of pores, particles, and aggregates, which may change in
Soil hydraulic properties time as a result of both ecohydrological dynamics and land management processes. The soil pore size distribution
Pore size distribution (PSD), which typically is treated as a static component, is a key determinant of soil properties, and its accurate
Tillage
representation has the potential to improve hydrological and crop models. Following previous work by Or et al.
Soil biogeochemistry
(2000), a modeling framework is proposed for the time evolution of the PSD which takes into account processes
such as tillage, consolidation, and changes in organic matter. A time-varying power law PSD is obtained as
the solution of a special form of transport equation for the PSD, parameterized using data from the literature to
capture, in a parsimonious and efficient manner, the changes in the PSD as a result of the soil processes considered.
Alterations in soil properties brought about by tillage, consolidation, and organic matter are then discussed. The
potential benefit of this method for determining soil properties over the more widely used pedotransfer functions
(PTF) is that it allows for the history of the soil, rather than only its present state, to be taken into account when
estimating soil properties, and it does so in a physically consistent manner, leading to the widely used power law
expression for soil properties with few parameters.

1. Introduction types which were used to create the functions is unknown and uncertain.
Also, they provide little information about soil structure, such as the
The properties of a soil are determined by a complex arrangement shape of the pore size distribution (PSD), and although more recently
of pores, particles, and aggregates, but the temporal evolution of these Vereecken et al. (2010) suggested the development of temporal PTFs
properties and their links to biogeochemical cycles are typically not to better represent the time-evolution of soil properties, they still im-
taken into account in current biogeochemical and ecohydrological mod- plicitly assume static conditions and do not account for time-dependent
els. An accurate representation of soil properties such as the water re- processes.
tention curve (WRC) and hydraulic conductivity curve (HCC) is needed Although they are often taken to be constant in modeling applica-
for such models in both natural and agricultural environments, as they tions, soil properties do change even over relatively short timescales, es-
largely control the soil moisture balance, which plays a key role in the pecially under the influence of human activities (Vereecken et al., 2010).
hydrologic cycle (e.g. Manzoni and Porporato, 2009; Porporato et al., Combined with the well-recognized importance of the PSD in determin-
2015). However, direct measurement of these properties for a particu- ing soil properties, this has led to efforts to model the time-evolution of
lar soil is often infeasible, leading to the need to estimate them based the PSD and thereby the dynamics of the WRC and HCC (Or et al., 2000;
on more easily obtained information, such as the silt/sand/clay frac- Leij et al., 2002a; 2002b). Maggi and Porporato (2007) also coupled the
tions, bulk density, and organic matter. Pedotransfer functions (PTFs), PSD with biological processes to understand the effect of bioclogging
which use large soil databases to statistically relate such basic data to on soil hydraulic properties. We follow this work by developing a PSD
the soil properties of interest, have been widely used to estimate use- model which evolves in time as a result of time-dependent soil processes
ful soil properties from more readily available data (e.g. Wösten et al., and biogeochemical cycling.
2001; Vereecken et al., 2010), but they are largely empirical and limited Our model differs from previous efforts in that we make use of a
in their physical information content. Among their main limitations is simple power law description of the PSD, which has the advantage of
that their ability to predict soil properties outside of the regions or soil a clear connection to the well known power law WRC (e.g. Brooks and


Corresponding author: Now at Sierra Nevada Research Institute, University of California, Merced, CA, USA.
E-mail address: npelak@ucmerced.edu (N. Pelak).

https://doi.org/10.1016/j.advwatres.2019.103384
Received 12 April 2019; Received in revised form 19 July 2019; Accepted 29 July 2019
Available online 30 July 2019
0309-1708/© 2019 Elsevier Ltd. All rights reserved.
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

Corey, 1964; Clapp and Hornberger, 1978) and has a basis in the fractal Or et al. (2000) applied a related evolution equation (the Fokker-
fragmentation of soils (Tyler and Wheatcraft, 1992; Perrier et al., 1996). Planck equation) to the time evolution of the PSD and obtained a log-
While the power law model is not applicable in all cases and its form normal PSD, a distribution which was first used for this purpose by
imposes certain limitations on the shape of the distribution, due to its Brutsaert (1966) and later by Kosugi (1994, 1996). Unlike in that study,
widespread use, physical basis, and the ease with which it can be linked we do not include a diffusion term (i.e., a term depending on the sec-
to bulk soil properties, it constitutes a desirable, parsimonious model of ond derivative of f with respect to r, 𝜕 2 f/𝜕r2 ). The presence of such a
the PSD. We also consider biogeochemical processes by connecting the term requires careful justification, because it is linked to a random walk
parameters to the soil organic matter (SOM) content. We first develop of the individual pore sizes r(t) (e.g. a random walk in the top panel of
an evolution equation for the dynamics of the pore size distribution and Fig. 1, in addition to the drift and mortality terms). However, reasonable
derive the power law distribution as one of its solutions. We then derive forms of the PSD may be obtained in this way, and might be justified by
expressions for key soil properties from the power law PSD, and with the consideration of processes such as wet/dry and freeze/thaw cycles,
a careful consideration of experimental data from the literature, link which are not considered here. One could also consider additional in-
its parameters to key soil processes and management activities such as tegral terms describing sudden changes in pore structure (such as those
tillage, soil consolidation, and changes in SOM. Finally, we show how brought about by tillage). Rather than doing this, we assume that tillage
these processes alter the PSD and soil properties. resets the soil pore system with new initial conditions, so that Eq. (1) de-
scribes the continuous evolution of the system between tillage events.
2. Dynamic pore size distribution
2.2. Power law pore size distribution
2.1. Evolution equation
We assume that in its evolution the PSD retains its power law form
We model the time-evolution of the pore size distribution 𝑓 = 𝑓 (𝑟, 𝑡) with time-varying parameters
as a function of the pore radius r and the time t via a generic transport 𝑓 (𝑟, 𝑡) = 𝑎(𝑡)𝑟−𝑏(𝑡) 0 ≤ 𝑟 ≤ 𝑅𝑚 (𝑡), (2)
equation of the form
in which a(t) is a scaling parameter, b(t) is the power law exponent, and
𝜕𝑓 𝜕
= [𝑣𝑓 ] − 𝑚𝑓 , (1) Rm (t) is the maximum effective pore radius. The power law PSD can be
𝜕𝑡 𝜕𝑟
obtained as a solution to Eq. (1), which is discussed further in the next
where 𝑣 = 𝑣(𝑟, 𝑡) is a drift term, and 𝑚 = 𝑚(𝑟, 𝑡) is a source/sink term. section and in the Appendix. Rm is related to the air-entry or bubbling
This equation represents the evolution of the ensemble frequency distri- pressure, which can be defined as the matric pressure at which air enters
bution of pore radii. The term v represents the shrinking of a pore radii the soil pores (Brooks and Corey, 1964).
as a function of pore radius and time, while m represents the instanta- The integral of the PSD over r is equal to the porosity, 𝜙(t), which
neous rate of gain or loss of pores of radius r. This type of equation is is equivalent to the soil water content at saturation. We can therefore
typically used in nonequilibrium statistical mechanics and kinetic theo- solve for 𝜙(t) as
ries to describe the evolution of statistical distributions of multiparticle 𝑅𝑚 (𝑡) 𝑎(𝑡)𝑅𝑚 (𝑡)1−𝑏(𝑡)
systems (Balescu, 1997). A sample application to the PSD can be seen 𝜙(𝑡) = 𝑎(𝑡) 𝑟−𝑏(𝑡) 𝑑𝑟 = , (3)
in Fig. 1. In the top panel of Fig. 1, trajectories of individual pores of ∫0 1 − 𝑏(𝑡)
radius r gradually decrease according to v(r, t), and are randomly lost from which the scaling parameter a(t) is
according to m(r, t). The bottom left and right panels show the PSD, f(r,
𝜙(𝑡)(1 − 𝑏(𝑡))
t), at 𝑡 = 0 and 𝑡 = 18 years, respectively. The expressions for f(r, t), v(r, 𝑎(𝑡) = , (4)
𝑅𝑚 (𝑡)1−𝑏(𝑡)
t), and m(r, t) are discussed in Section 2.3.
yielding the full expression for the PSD
( )
𝜙(𝑡)(1 − 𝑏(𝑡)) −𝑏(𝑡)
𝑓 (𝑟, 𝑡) = 𝑟 . (5)
𝑅𝑚 (𝑡)1−𝑏(𝑡)

In order for the integral to converge, it is necessary that b(t) < 1, which is
a requirement for the model to be physically realistic. In the next section
we develop expressions for v and m which allow us to obtain a power
law form of the PSD.

2.3. Terms of the evolution equation

We now return to the as yet undefined terms of the evolution equa-


tion, Eq. (1), the drift v(t) and the source/sink m(t). In order to obtain a
power law PSD which evolves in time, these terms must be
𝑟 ( )
𝑣(𝑟, 𝑡) = 𝑎(𝑡)𝑏′ (𝑡) ln(𝑟) − 𝑎′ (𝑡) , (6)
𝑎(𝑡)𝑏(𝑡)
and
𝑏′ (𝑡) 𝑎′ (𝑡)
𝑚(𝑟, 𝑡) = (1 + ln(𝑟)) − , (7)
𝑏(𝑡) 𝑎(𝑡)𝑏(𝑡)
where the time-derivative of the parameters is indicated with a prime
Fig. 1. The top panel shows sample trajectories of pores of radius r as they drift (′) symbol. With these terms, we obtain a form of Eq. (1) which can be
and close according to v(r, t) and m(r, t), respectively. Note that only pore closure solved analytically using the method of characteristics (Logan, 2013)
is visible in this figure because the source/sink term leads to pore opening only leading to a power law solution (see the Appendix for details). The initial
for very small values of r. A tillage event at 𝑡 = 18 years (grey line) resets the condition for this problem is
PSD to the initial condition. The initial PSD at 𝑡 = 0 years (bottom left) and the
final PSD at 𝑡 = 18 years (bottom right) are also shown. 𝑓 (𝑟, 0) = 𝑓0 (𝑟) = 𝑎(0)𝑟−𝑏(0) , 0 < 𝑟 < 𝑅𝑚 (0), 𝑡 = 0. (8)
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

Fig. 2. Example of using the Ψ − 𝜃 data to obtain the model parameters. Here,
the data is from Teiwes (1988). The ‘NoTill’ data are from the long term plot
without tillage, and the ‘Mid’ data are from samples which were taken at ap-
Fig. 3. The PSD for the tilled and untilled soils of Teiwes (1988).
proximately the midpoint between annual tillage, roughly 180 days after tillage
occurred.

As previously stated, we assume that the system resets with new initial parallel tubes are cut normal to the flow direction and randomly re-
conditions at each tillage event. In the following sections, we will de- arranged (Brutsaert, 2005), leading to the following expression for the
velop expressions for key soil properties, as well as discuss the model hydraulic conductivity
parameterization and the dependence of the parameters on both time 𝛾𝑤 𝐺𝑒 𝜙(𝑡)2 𝑅𝑚 (𝑡)2 (1 − 𝑏(𝑡))2 4−2𝑏(𝑡)
𝐾(𝑠, 𝑡) = 𝑠 1−𝑏(𝑡) , (11)
and soil biogeochemical cycling. 𝜇(3 − 𝑏(𝑡))(2 − 𝑏(𝑡))
where 𝛾 w is the specific weight and 𝜇 is the dynamic viscosity of water.
3. Soil hydraulic properties If the flow in the pores follows the Hagen-Poiseuille equation, 𝐺𝑒 = 1∕8
(Brutsaert, 2005). Another soil property of interest, the field capacity,
In order to utilize the results of this study in an ecohydrological may be defined as the value of s at which the hydraulic conductivity
model, we obtain expressions for the evolution of key soil properties, becomes negligible compared to the maximum daily evapotranspiration
in particular the WRC and the HCC. Note that the PSD is intended as an rate (Rodríguez-Iturbe and Porporato, 2004).
average over the soil depth, as we do not consider different soil horizons
or a vertically explicit representation of soil properties, and so these ex- 3.1. Model parameterization
pressions should likewise be taken as average values. Using the methods
outlined by Mualem and Dagan (1978) and Brutsaert (2005), we model As the WRC curve is generally easier to measure than pore size data,
the flow of water through the soil by assuming that the soil pores act it is useful to estimate the latter from the former. We can relate the soil
as capillary tubes, with a distribution of pore sizes determined by the moisture s to the water content 𝜃(t) and the porosity 𝜙(t) as
normalized PSD (i.e., the probability density function of r), and derive
𝜃(𝑡)
the following expression for the matric potential Ψs 𝑠(𝜃(𝑡), 𝑡) = , (12)
( ) 𝜙(𝑡)
𝐶𝑠 which allows us to rewrite Eq. (9) as
Ψ𝑠 (𝑠, 𝑡) = − 𝑠−1∕(1−𝑏(𝑡)) , (9) ( )( )
𝑅𝑚 (𝑡)
𝐶𝑠 𝜃(𝑡) −1∕(1−𝑏(𝑡))
Ψ𝑠 (𝜃(𝑡), 𝑡) = − . (13)
where s is the relative soil moisture and Cs is a constant related to the 𝑅𝑚 (𝑡) 𝜙(𝑡)
surface tension of water (a derivation can be found in the Appendix).
We estimated the parameter values by using the Matlab constrained min-
We can invert this expression to solve for the relative soil moisture at a
imization function ‘fmincon’ to minimize the root mean square error
particular matric potential. For example, we can obtain the soil moisture
(RMSE) between Eq. (13) and the data. While developing this model,
at the wilting point sw by solving for s with the wilting matric potential,
we explored other parameterizations of the WRC (for example, includ-
Ψ𝑠𝑤 , as
ing a residual porosity or a minimum pore radius) but found that these
( )
−Ψ𝑠𝑤 𝑅𝑚 (𝑡) additional parameters had little effect on the model results, and so in the
𝑏(𝑡)−1
𝑠𝑤 (𝑡) = . (10) interest of parsimony we have neglected them. Fig. 2 shows an example
𝐶𝑠
of this fitting to data of Teiwes (1988) (obtained from Leij et al., 2002b),
The same procedure can be followed to obtain other values of interest which compares data from two experimental sites, one which had been
for soil moisture, such as the hygroscopic point and the point of incipient tilled annually and one which was managed using no-till methods.
water stress. Values for the wilting point vary with the plant type, but
in this study we use Ψ𝑠𝑤 = −1.5 MPa (this is a typical value for a crop 4. Connection of parameters to temporal and biogeochemical
such as maize, though a value of −3 MPa is often used for stress tolerant processes
plants (Rodríguez-Iturbe and Porporato, 2004)).
An expression for the HCC can also be obtained by treating the soil We separate the important factors which affect the PSD parameters
pores as a series of parallel tubes with a distribution of radii determined into those which are explicitly time-dependent (including external forc-
by the PSD (Mualem and Dagan, 1978; Brutsaert, 2005). As this tends ing such as soil tillage and internal processes such as consolidation) and
to lead to an unrealistically high value for the saturated hydraulic con- those which depend on the soil biogeochemistry. The effect of soil bio-
ductivity, a so-called series-parallel model can be used, in which the geochemistry on soil structure and properties are complex. To account
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

Fig. 5. Timeseries for the varying SOM levels representing low, middle, and
high inputs (of fertilizer, manure, or other amendments which tend to increase
SOM) and which were used to drive the results of Figs. 6 and 7.

that they can be accounted for with two multiplicative terms. In real-
ity, of course, they are not likely to be completely independent, but this
is a useful first step to separate these two important influences. In this
way, it represents an improvement on a typical pedotransfer function,
for which, all else being equal, a given value of SOM will always result
in the same soil properties, because time-dependent information such as
the time when the soil was last tilled are not taken into account.
Below, we present only the results for the power law exponent b, but
the same procedure was followed for all parameters. We separate the
factors which control the value of b into two parts: first, a function 𝛾 b (t)
which is used to account for explicitly time-dependent changes in the
PSD, and second, a component bC (C(t)), which relates the value of b to
the SOM content. These terms are then multiplied together to obtain an
expression for b(t, C(t))

𝑏(𝑡, 𝐶(𝑡)) = 𝛾𝑏 (𝑡) ⋅ 𝑏𝐶 (𝐶(𝑡)). (14)

The next subsections discuss 𝛾 b (t) and bC (C(t)), respectively.

4.1. Tillage and consolidation term

Differences in the soil properties of tilled and untilled soils are evi-
dent in many studies (Teiwes, 1988; Arshad et al., 1999; Or et al., 2000;
Leij et al., 2002b) and can be seen for example in Fig. 2. While tillage
induces an immediate change in the PSD, there is also a consolidation
process which occurs over a larger timescale as larger pores close and
the distribution of pore sizes changes (Hillel, 1980; Jiang et al., 2018).
In order to incorporate the effect of tillage and consolidation into
our model, we utilized data (Teiwes, 1988) which presents contrasting
WRC curves from plots which were tilled annually and those which were
left untilled for an extended period (in this case, 18 years). Fig. 3 shows
Fig. 4. The three model parameters as a function of soil organic matter (C). The the fitted PSD model before and after tillage. The samples were taken
data points are from Naveed et al. (2014) and the lines are fit as desctibed in approximately 6 months ( ≈ 180 d) after the soil was tilled. For each pa-
Section 4.2. rameter, we obtained the ratio between the no-till and the tilled value.
Based on the empirical differences between the PSD of tilled and no-till
soils (Fig. 2) we assume that there is a consistent ratio of the value of
for these effects in our model, we adopt the soil organic matter con- each parameter at two given points in time (e.g., the ratio of a parameter
tent, C(t), as a means to encapsulate the effect of soil biogeochemical immediately after tillage and after maximum consolidation is constant),
changes on the soil. While this is certainly a simplification, it is a nec- and that this ratio is independent of the current SOM content. Given the
essary one to maintain the parsimonious description of the soil PSD. In time scales of interest in this work, ranging from days to multiple grow-
order to incorporate the effects of time dependent processes and soil or- ing seasons, tillage is assumed to occur instantaneously, and is followed
ganic matter, we posit that these changes are sufficiently independent by consolidation, which is modeled by an exponential decline towards
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

Fig. 6. Timeseries for the wilting point sw (top left) and the saturated hydraulic conductivity Ksat (bottom left) for seven years of tillage followed by a no till period
and varying SOM levels. Also shown are the WRC (top right) and the HCC (bottom right) at time points A, B, and C.

the untilled state. Therefore, we write the settling term 𝛾 b as 4.2. SOM relationship

𝛾𝑏 (𝑡) = 𝑟𝑏 + (1 − 𝑟𝑏 ) exp{−𝑘𝑏 (𝑡 − 180)}, (15) Organic matter is an important indicator of soil structure and func-
tion (e.g. Dexter, 2004; Dexter et al., 2008), which is demonstrated by
the inclusion of SOM as a predictor variable in many PTFs (e.g. Wösten
where rb is the ratio of the no-till parameter value to the base value,
et al., 2001; Vereecken et al., 2010). Soils which have different levels
kb is the time-rate of settling, and t is the time since the soil was tilled
of organic matter clearly have different properties. In reality, there is
(at which point the PSD is reset with a new initial condition). In this
likely some overlap between the effect on the PSD of management prac-
way, when 𝑡 = 180 days (the soil has been tilled 180 days prior), the
tices which tend to increase or decrease SOM content and the physical
consolidation term 𝛾𝑏 = 1 and the modeled PSD matches the data. As
properties of SOM itself, but in this study we take the amount of SOM
t → ∞ (after a long period without tillage) the consolidation term 𝛾 b → rb
as a proxy for these joint effects. The expression for bC is
and the PSD approaches the untilled state. By assuming a constant value
for kb , the parameter values at tillage can be obtained by setting 𝑡 = 𝑏𝐶 (𝐶(𝑡)) = 𝑏0 + 𝜎𝑏 𝐶(𝑡), (16)
0 days. The effect of the 𝛾 term on the parameter values can be seen
in the black lines in Fig. 7, which show the parameter evolution with where b0 is the parameter value at 𝐶 = 0 and 𝜎 b is the slope of the 𝑏 − 𝐶
constant C. relationship. Here we have used a linear fit, though other relationships
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

Fig. 7. In the top row and bottom left are shown timeseries for the model parameters b, Rm , and 𝜙 with exponential decay from the tilled to the untilled state and
with tillage causing instantaneous jumps back to the tilled state, for seven years of tillage followed by a no till period. In the bottom right is the PSD at time points
A, B, and C.

may be supported by additional data. Fig. 4 shows fitted relationships 5. Evolution of soil hydraulic properties and parameters
for each of the three model parameters. We note that the data to which
this relationship was fit extends over a limited range, and it therefore In the preceding sections we have developed a model which accounts
may not hold outside of it (for example, it may not be valid as low as for the evolution of the PSD as a result of both temporal (tillage and con-
𝐶 = 0, despite the fact that b0 appears in the equation). The magnitude solidation) and biogeochemical (SOM content) changes in the soil, and
of parameter changes over time due to changing C can be seen in Fig. 7. which connects the resulting dynamic PSD to important soil hydraulic
The three lines in each panel correspond to increasing, constant, and properties. We will now see the joint effects of these changes on the
decreasing C. model. The effects of biogeochemical processes are shown by varying
All the data used in this study is from silt-loam soils, and it is the SOM levels (see Fig. 5). We have imposed three theoretical regimes
possible (or even likely) that other types of soils would produce sub- of SOM dynamics, in which SOM declines, remains constant, and in-
stantially different results. Additionally, the tillage method used in creases, respectively. These temporal variations in SOM are consistent
Teiwes (1988) was moldboard plowing, while multiple tillage types with a minimal SOM model with constant inputs and first order decom-
were used in Naveed et al. (2014), including an offset disc plow and position, and the three regimes are intended to represent low, medium
rotary harrow. Here, we do not consider different effects for different and high input levels of fertilizer, manure, or other amendments which
types of tillage, though there are differences (Heard et al., 1988), which tend to promote SOM accumulation. The upper and lower limits of SOM
could possibly be taken into account by adjusting the ratios between till are chosen so that the results are shown over the range of SOM values
and no till parameters. contained in the data from Naveed et al. (2014) which were used to
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

parameterize the model. The effects of temporal processes are demon- possibilities for optimizing management strategies in order to achieve
strated by imposing a ‘Till’ period of several years of annual tillage fol- desired soil property outcomes. However, we emphasize that the results
lowed by a longer ‘No Till’ period. These periods are separated by the shown in Figs. 5–7 do not represent a full coupling of the dynamic PSD
gray lines in Figs. 6 and 7. Additionally, we have denoted 3 points on model to an ecohydrological model. This would include feedbacks be-
Fig. 5 as A, B, and C, which we will use to indicate the differences be- tween, for example, the soil hydraulic properties and the SOM dynamics,
tween the soil PSD and hydraulic properties across varying times and and would also require estimations of other soil properties such as the
SOM levels. soil depth from the changes in the PSD. Such a coupling will be explored
As previously discussed, the WRC and the HCC are two key soil hy- in a follow-up publication.
draulic properties which are largely controlled by the PSD. In Fig. 6, the
effects of tillage on these two properties are shown with a timeseries of 6. Conclusion
the soil wilting point sw (top left) and the saturated hydraulic conduc-
tivity Ksat (bottom left) which were estimated using Eqs. (9) and (11). The model described in this study represents realistic trends of key
Additionally, the right side of Fig. 6 shows the WRC and the HCC at soil properties as they evolve in time and change with soil biogeochem-
points A, B, and C, to demonstrate how these changing soil properties istry in a physically consistent manner. Changes in the key soil proper-
can be integrated into an ecohydrological model. ties (indicated here by the WRC and the HCC as well as sw and Ksat ) and
In the top panel of Fig. 6, the wilting point decreases with tillage and the parameters which control them (b, Rm , 𝜙) are shown in Figs. 6 and
for higher levels of SOM. This reflects the general purpose of tillage, 7. Each panel shows a timeseries of the parameter or variable, with the
which is to improve soil structure for crop growth (Hillel, 1980), as soil undergoing several years of annual tillage followed by a longer pe-
lowered wilting points allow crops to continue transpiring at lower soil riod of consolidation (corresponding to a no till period), for increasing,
moisture levels. The negative relationship between sw and SOM is also decreasing, and constant SOM. The general trends shown by the soil
consistent with other studies which considered the impact of SOM on properties in Figs. 6 and 7 have support in the literature.
soil properties (for example, Dexter, 2004 found a positive correlation Via their dependence on SOM, these dynamic variables can be read-
between SOM and a measure of soil physical quality). ily coupled to an ecohydrological or soil biogeochemical model. We
In the bottom panel, we observe that the hydraulic conductivity Ksat note that the model presented here does not account for very short
is increased with tillage and for higher values of SOM. Some studies term changes in soil properties, such as those due to wetting and dry-
show an increase of soil hydraulic conductivity with tillage (Mielke ing cycles. However, the long-term effects of such hydrologic forcing
et al., 1986; Heard et al., 1988; Pikul Jr et al., 1990), which may rep- on soil properties could be indirectly accounted for via the changes in
resent an increase in the number of macropores (although macropores SOM which would result from different hydrologic regimes or irrigation
are more continuous in no-till soils, Heard et al. (1988) found them strategies. To create Figs. 6 and 7, we imposed three contrasting SOM
to be more numerous in tilled soils). Macroporosity is not separated regimes, which were intended to represent low, middle, and high inputs
from total porosity in our model, but is related to the Rm parameter, of amendments which tend to promote the accumulation of SOM. The
which also increases with tillage (see the middle panel of Fig. 7). How- power law model may in general perform less well on fine textured soils
ever, it is also possible that undisturbed soils could have a higher hy- (Tyler and Wheatcraft, 1990), and because this model is vertically av-
draulic conductivity because their macropores (formed by earthworms, eraged over the soil depth, it may also be less applicable to soils with
insects, or roots) are more continuous and are not disrupted during strong vertical variation in soil type or properties.
tillage (Ehlers and Van Der Ploeg, 1976; Klute, 1982; Wu et al., 1992). The results in Figs. 6 and 7 point toward a means to account for the
The positive relationship between hydraulic conductivity and SOM can considerable changes which can be brought about in soil properties as
be attributed to an associated increase in macro- and mesoporosity a result of contrasting management strategies–specifically, those which
(Mahmood-ul Hassan et al., 2013). The study from which we obtained increase or decrease SOM, and the effect of tillage. This can be done by
SOM data (Naveed et al., 2014) did not present hydraulic conductivity providing a real coupling of this model to ordinary differential equations
data, though it did show a general trend of increasing air permeability describing soil biogeochemical processes and plant growth. Future con-
with increasing SOM, and it is to be expected that hydraulic conductivity tributions will explore the impact of contrasting management strategies
would trend in the same direction as air permeability. (e.g. irrigation, fertilization, organic matter inputs) on the soil structure
Fig. 7 shows timeseries of the model parameters for a tilled followed and properties, as well as a coupling with a soil biogeochemistry and
by a no-till period and for contrasting SOM levels, as in Fig. 6. Although plant growth model.
the power law PSD has certain limitations in the variety of soil types
that it can adequately represent, a decrease in the power law exponent Acknowledgments
b generally means that for a particular matric pressure, the soil is capa-
ble of maintaining a higher water content, which is favorable for crops. The authors would like to thank Stefano Manzoni for early dis-
The top left panel of Fig. 7 shows that b decreases with tillage and with cussions on this topic. This work was funded by the US National Sci-
higher levels of SOM, which is consistent with the positive correlation ence Foundation under grants EAR-1331846, EAR-1316258, and FESD-
of SOM with a soil quality index found elsewhere (Dexter, 2004). The 1338694, and the authors also acknowledge support from BP through
positive relationship of Rm with both tillage and SOM (top right panel) the Carbon Mitigation Initiative (CMI) at Princeton University.
can be connected to the associations of both with increased macroporos-
ity, as discussed previously, and is shown in the middle of Fig. 7. The Appendix A
positive relationships between porosity and SOM and the decrease in 𝜙
in the absence of tillage (bottom left panel) can be seen in the bottom A1. Solution of the PSD evolution equation
panel of Fig. 7 and have support in the literature (Teiwes, 1988; Dexter
et al., 2008; Naveed et al., 2014). The bottom right panel shows the com- The method of characteristics (e.g., Logan, 2013) can be used to an-
bined effects of different parameter combinations on the PSD at points alytically solve the evolution equation for the soil PSD,
A, B, and C. One potential implication of the results shown in Figs. 6 and
𝜕𝑓 (𝑟, 𝑡) 𝜕
7 is that there may be tradeoffs between management strategies (here = [𝑣(𝑟, 𝑡)𝑓 (𝑟, 𝑡)] − 𝑚(𝑟, 𝑡)𝑓 (𝑟, 𝑡). (A.1)
𝜕𝑡 𝜕𝑟
encapsulated by tillage and changes in SOM) and soil properties. For
example, in the bottom left panel of Fig. 7, despite 18 years of no-till Assuming a drift term of the form
management, at the end of the 25 years modeling period the porosity for 𝑟 ( )
𝑣(𝑟, 𝑡) = 𝑎(𝑡)𝑏′ (𝑡) ln(𝑟) − 𝑎′ (𝑡) , (A.2)
the high SOM treatment was over 90% of its initial value. This suggests 𝑎(𝑡)𝑏(𝑡)
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

and a source/sink term of the form the matric pressure, Δ𝑝 = −Ψ𝑠 . We then obtain the following expression
𝑏′ (𝑡) 𝑎′ (𝑡) for the matric pressure as a function of r
𝑚(𝑟, 𝑡) = (1 + ln(𝑟)) − , (A.3)
𝑏(𝑡) 𝑎(𝑡)𝑏(𝑡) 𝐶𝑠
Ψ𝑠 ( 𝑟 ) = − , (A.15)
Eq. (A.1) becomes 𝑟
where 𝐶𝑠 = 2𝜎. By substituting the previous expression for r(s), we can
𝜕𝑓 (𝑟, 𝑡) 𝑟 ( ) 𝜕𝑓 (𝑟, 𝑡)
= 𝑎(𝑡)𝑏′ (𝑡) ln(𝑟) − 𝑎′ (𝑡) , (A.4) obtain the matric pressure as a function of s (i.e. the soil water retention
𝜕𝑡 𝑎(𝑡)𝑏(𝑡) 𝜕𝑟
curve)
with the initial condition 𝐶𝑠
Ψ𝑠 ( 𝑠 ) = − , (A.16)
𝑓 (𝑟, 0) = 𝑓0 (𝑟) = 𝑎(0)𝑟−𝑏(0) , 0 < 𝑟 < 𝑅𝑚 (0), 𝑡 = 0. (A.5) 𝑃 −1 (𝑠)

This condition and the form of the drift and source/sink term are chosen and with the power law PSD as our starting point, we obtain the follow-
to ensure that the obtained solution is a power law at all times, as will ing expression for the WRC
( )
appear from the following solution. In the method of characteristics, the 𝐶𝑠
coordinate system in (r, t) is changed to a new one in (r0 , s) in such a Ψ𝑠 ( 𝑠 ) = − 𝑠−1∕(1−𝑏) . (A.17)
𝑅𝑚
way that the PDE becomes an ODE along the curve, (r(s), t(s)), referred
𝑑𝑟 𝑑𝑡
to as the characteristic curve. 𝑑𝑠 and 𝑑𝑠 are given by At the extreme values of s, this function is no longer dependent on the
power law exponent b:
𝑑𝑟 𝑑𝑟 𝑟 ( )
= = 𝑎(𝑡)𝑏′ (𝑡) ln(𝑟) − 𝑎′ (𝑡) , (A.6) 𝐶𝑠
𝑑𝑠 𝑑𝑡 𝑎(𝑡)𝑏(𝑡) Ψ𝑠 (𝑠 = 1) = , (A.18)
𝑅𝑚
and
𝑑𝑡 which is known as the air-entry pressure, and at the other extreme,
= 1. (A.7)
𝑑𝑠 Ψ𝑠 (𝑠 = 0) → ∞. (A.19)
From Eq. (A.7), 𝑠 = 𝑡. From Eq. (A.6), using the initial condition 𝑟(0) = 𝑟0 ,
we obtain References
( 𝑏(0) )
𝑟0 𝑎(𝑡) 1 Arshad, M., Franzluebbers, A., Azooz, R., 1999. Components of surface soil structure under
𝑟(𝑡) = 𝑏(𝑡) , (A.8) conventional and no-tillage in northwestern canada. Soil Tillage Res. 53 (1), 41–47.
𝑎(0)
Balescu, R., 1997. Statistical Dynamics: Matter out of Equilibrium. Imperial Coll..
Brooks, R.H., Corey, A.T., 1964. Hydraulic properties of porous media. Hydrology Papers.
and solving for r0 , Colorado State University.
( ) 1 Brutsaert, W., 1966. Probability laws for pore-size distributions. Soil Sci. 101 (2), 85–92.
𝑎(0)𝑟𝑏(𝑡) 𝑏(0)
𝑟0 (𝑟, 𝑡) = . (A.9) Brutsaert, W., 2005. Hydrology: An Introduction. Cambridge University Press.
𝑎(𝑡) Clapp, R.B., Hornberger, G.M., 1978. Empirical equations for some soil hydraulic proper-
ties. Water Resour. Res. 14 (4), 601–604.
Inserting this into the initial condition yields the power law as the solu- Dexter, A., Richard, G., Arrouays, D., Czyż, E., Jolivet, C., Duval, O., 2008. Complexed
tion, organic matter controls soil physical properties. Geoderma 144 (3–4), 620–627.
Dexter, A.R., 2004. Soil physical quality: Part I. Theory, effects of soil texture, density,
𝑓 (𝑟, 𝑡) = 𝑓0 (𝑟0 ) = 𝑎(𝑡)𝑟−𝑏(𝑡) , 0 < 𝑟 < 𝑅𝑚 (0), 𝑡 = 0. (A.10) and organic matter, and effects on root growth. Geoderma 120 (3–4), 201–214.
Ehlers, W., Van Der Ploeg, R., 1976. Evaporation, drainage and unsaturated hydraulic
conductivity of tilled and untilled fallow soil. Z. Pflanzenernähr. Bodenkd. 139 (3),
A2. Water retention curve 373–386.
Mahmood-ul Hassan, M., Rafique, E., Rashid, A., 2013. Physical and hydraulic properties
of aridisols as affected by nutrient and crop-residue management in a cotton-wheat
The relative soil moisture s is defined as system. Acta Sci. 35 (1), 127–137.
𝜃 Heard, J.R., Kladivko, E.J., Mannering, J.V., 1988. Soil macroporosity, hydraulic con-
𝑠= , (A.11) ductivity and air permeability of silty soils under long-term conservation tillage in
𝜙
Indiana. Soil Tillage Res. 11 (1), 1–18.
where 𝜃 is the volumetric water content and 𝜙 is the porosity. Follow- Hillel, D., 1980. Introduction to Soil Physics. Academic Press.
Jiang, H., Han, X., Zou, W., Hao, X., Zhang, B., 2018. Seasonal and long-term changes in
ing Brutsaert (2005), we assume that soil moisture will occupy smaller soil physical properties and organic carbon fractions as affected by manure application
pores first, allowing us to set the relative soil moisture s(r) equal to P(r), rates in the mollisol region of northeast china. Agric. Ecosyst. Environ. 268, 133–143.
the cumulative distribution function (CDF) of the pore size distribution Klute, A., 1982. Tillage effects on the hydraulic properties of soil: a review. In: Predicting
Tillage Effects on Soil Physical Properties and Processes, pp. 29–43.
(where r is the effective radius of the largest pore which is filled at the Kosugi, K., 1994. Three-parameter lognormal distribution model for soil water retention.
specified soil moisture level), Water Resour. Res. 30 (4), 891–901.
Kosugi, K., 1996. Lognormal distribution model for unsaturated soil hydraulic properties.
𝑠(𝑟) = 𝑃 (𝑟). (A.12) Water Resour. Res. 32 (9), 2697–2703.
Leij, F.J., Ghezzehei, T.A., Or, D., 2002. Analytical models for soil pore-size distribution
We can invert this expression to solve for 𝑟 = 𝑃 −1 (𝑠).
We then apply the after tillage. Soil Sci. Soc. Am. J. 66 (4), 1104–1114.
Young-Laplace equation, which gives the pressure difference across the Leij, F.J., Ghezzehei, T.A., Or, D., 2002. Modeling the dynamics of the soil pore-size dis-
tribution. Soil Tillage Res. 64 (1–2), 61–78.
interface between two immiscible fluids (Brutsaert, 2005) Logan, J.D., 2013. Applied Mathematics. John Wiley & Sons.
( ) Maggi, F., Porporato, A., 2007. Coupled moisture and microbial dynamics in unsaturated
1 1
Δ𝑝 = 𝜎 ± , (A.13) soils. Water Resour. Res. 43 (7).
𝑅1 𝑅2 Manzoni, S., Porporato, A., 2009. Soil carbon and nitrogen mineralization: theory and
models across scales. Soil Biol. Biochem. 41 (7), 1355–1379.
in which R1 and R2 are the radii of curvature of the interface and 𝜎 Mielke, L., Doran, J., Richards, K., 1986. Physical environment near the surface of plowed
is the surface tension. We define the radii of curvature as 𝑅 = 𝑟∕ cos 𝛽, and no-tilled soils. Soil Tillage Res. 7 (4), 355–366.
where 𝛽 is the contact angle of the air-water interface, and assume that Mualem, Y., Dagan, G., 1978. Hydraulic conductivity of soils: unified approach to the
statistical models 1. Soil Sci. Soc. Am. J. 42 (3), 392–395.
the radii of curvature are equal, so that we can now write the previous Naveed, M., Moldrup, P., Vogel, H.-J., Lamandé, M., Wildenschild, D., Tuller, M., de
expression as Jonge, L.W., 2014. Impact of long-term fertilization practice on soil structure evo-
lution. Geoderma 217, 181–189.
2𝜎 cos 𝛽 Or, D., Leij, F.J., Snyder, V., Ghezzehei, T.A., 2000. Stochastic model for posttillage soil
Δ𝑝 = . (A.14)
𝑟 pore space evolution. Water Resour. Res. 36 (7), 1641–1652.
Perrier, E., Rieu, M., Sposito, G., de Marsily, G., 1996. Models of the water retention
The contact angle for water and typical soil materials is generally small, curve for soils with a fractal pore size distribution. Water Resour. Res. 32 (10),
so we let 𝛽 → 0, and define the pressure difference across the interface as 3025–3031.
N. Pelak and A. Porporato Advances in Water Resources 131 (2019) 103384

Pikul Jr, J., Zuzel, J., Ramig, R., 1990. Effect of tillage-induced soil macroporosity on Tyler, S.W., Wheatcraft, S.W., 1992. Fractal scaling of soil particle-size distributions: anal-
water infiltration. Soil Tillage Res. 17 (1–2), 153–165. ysis and limitations. Soil Sci. Soc. Am. J. 56 (2), 362–369.
Porporato, A., Feng, X., Manzoni, S., Mau, Y., Parolari, A.J., Vico, G., 2015. Ecohydrologi- Vereecken, H., Weynants, M., Javaux, M., Pachepsky, Y., Schaap, M., Genuchten, M.T.,
cal modeling in agroecosystems: examples and challenges. Water Resour. Res. 51 (7), et al., 2010. Using pedotransfer functions to estimate the van Genuchten–Mualem soil
5081–5099. hydraulic properties: A review. Vadose Zone J. 9 (4), 795–820.
Rodríguez-Iturbe, I., Porporato, A., 2004. Ecohydrology of Water-Controlled Ecosystems: Wösten, J., Pachepsky, Y.A., Rawls, W., 2001. Pedotransfer functions: bridging the gap
Soil Moisture and Plant Dynamics. Cambridge University Press. between available basic soil data and missing soil hydraulic characteristics. J. Hydrol.
Teiwes, K., 1988. Einfluß von Bodenbearbeitung und Fahrverkehr auf Physikalische Eigen- 251 (3–4), 123–150.
schaften Schluffreicher Ackerböden. PhD Thesis. Wu, L., Swan, J., Paulson, W., Randall, G., 1992. Tillage effects on measured soil hydraulic
Tyler, S.W., Wheatcraft, S.W., 1990. Fractal processes in soil water retention. Water Re- properties. Soil Tillage Res. 25 (1), 17–33.
sour. Res. 26 (5), 1047–1054.

You might also like