Professional Documents
Culture Documents
161–179
www.elsevier.comrlocatergeoderma
Abstract
The necessity for a quantitative analysis of pedogenesis has become more compelling in the
last few decades. In part one of this paper we proposed a rudimentary mechanistic model that
considers soil formation spatially at the catena scale. This paper extends the model for soil
formation in a landscape conditioned by a digital elevation model and further illustrates the
application of the model in quantifying pedogenesis. The current model states that the change in
soil thickness over time depends on the physical weathering rate of rock, the loss due to chemical
weathering and the transport of soil through erosion. The rate of physical weathering or lowering
of the bedrock surface is represented as an exponential decline with soil thickness. The chemical
weathering rate is modelled as a negative exponential function of both soil thickness and time.
Assuming uni-dimensional weathering, the rate of chemical weathering can be expressed as a
reduction in soil thickness. The movement of materials in the landscape is characterized by
diffusive transport. The model is solved numerically using the finite-difference approach and
applied to a digital elevation model. The results for simulation of soil formation in a landscape
after 10,000 years shows that the soil accumulates Žthickens. in the gullies and erodes Žthins. in
the ridges. The soil from the upper slope is transported down-slope by the erosion processes and
fills the gullies. Soil thickness is highly correlated with the profile curvature. The effect of
climate, rock type and land management is illustrated by different combinations of weathering rate
and erosive diffusivity. To illustrate the effect of irregularity and randomness on the stability of
)
Corresponding author. Tel.: q61-2-9351-5813; fax: q61-2-9351-3706.
E-mail address: budiman@acss.usyd.edu.au ŽB. Minasny..
0016-7061r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 0 1 6 - 7 0 6 1 Ž 0 1 . 0 0 0 7 5 - 1
162 B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
the solution and the soil development, spatially correlated normally distributed random numbers
were added to the initial soil elevation. The small randomness appears to cause instability Žchaos.
in the system. Dimensionality analysis of the soil thickness as a function of time confirms the
non-linear chaotic behaviour of the model rather than merely random noise. A difference plot of
the soil thickness time series unveiled the presence of a strange attractor. Alternatively, these
results may be a result of numerical instability. q 2001 Elsevier Science B.V. All rights reserved.
1. Introduction
2. Theory
Our model considers a landscape with surface elevation z, soil with thickness
h and soil–bedrock interface e along a horizontal x- and y-axes Ž Fig. 1.. The
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 163
change in soil thickness over time depends on the processes of the formation of
soil from weathering of bedrock, the loss of materials by chemical weathering
and the transport of soil by erosion.
Soil formation depends on the rate of breakdown or weathering of the
underlying parent materials under physical, chemical and biological processes.
Soil erosion is mainly governed by elevation z, which acts as a potential driving
the process. The process of erosion or sediment flow may occur by soil creep or
mass wasting.
The continuity equation for the soil thickness over time can be formulated as
the change in soil elevation Ž soil thicknessq bedrock weatheringy soil weather-
ing. over time is equal to the transport of soil:
Eh Ee El
rs q rr q rs s y= Ž rs qs . Ž1.
Et Et Et
where rs is the density of soil and rr is the density of rock, l is the reduction in
soil thickness as the result of chemical weathering or solutional denudation, and
qs is the flux of materials.
164 B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
where DVsoil and DVrock are the changes in volume of soil and rock due to
weathering. Positive values indicate volumes that increase Ž expansion. , while
negative values indicate volume reduction Ž collapse. . If only physical weather-
ing occurs, the volume change will only increase proportional to the density
ratio between the rock and soil. It is evident that the volume change can also
decrease as a result of mineral dissolution and leaching of soluble constituents
ŽMoran et al., 1988; Brimhall et al., 1991. . Therefore, a chemical weathering
factor is introduced in the current model.
We assumed uni-dimensional weathering and the rate of weathering is
defined as a reduction in the soil thickness ElrEt. For chemical weathering, the
equation that is analogous to the first-order reaction is given as:
El
s W0 exp Ž yk 0 tk 1 h y k 2 h . Ž4.
Et
where W0 wL Ty1 x is the potential chemical weathering rate and k 0 wTy1 x is the
rate constant for time, k 1 wLy1 x and k 2 wLy1 x are the rate constants for soil
thickness. The above equation expresses the loss of material through chemical
weathering as a combination of both soil thickness and time Ž Fig. 2. . Some
modellers have suggested an alternative ‘bell-shaped’ curve for chemical weath-
ering ŽAhnert, 1977..
The movement of materials in the landscape may be described as diffusive
transport Ž Scheidegger, 1991.:
qs s yD=z Ž5.
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 165
Fig. 2. Rate of chemical weathering as a function of soil thickness and time Žnumbers near the
curves represent time in years..
in which q s is the volume of material that flows across a slope profile width per
unit time wL3 Ly1 Ty1 x or sediment flux wL2 Ty1 x, and D is the erosive
diffusivity of the material wL2 Ty1 x, = s partial derivative vector ŽŽ ErE x . q
Ž ErE y .., =z is the first derivative of z with respect to space, i.e. the slope or
gradient. The second derivative of z with respect to space 2 Ž = 2 z s Ž E 2 zrE x 2 .
q Ž E 2 zrE y 2 .. is the curvature Žor profile curvature. , defined as the rate of
change in gradient. A positive curvature reflects a concave slope Ž valley. , and
negative one is a convex slope Ž hill..
If we combine Eqs. Ž1. and Ž5. and assume constant density and diffusivity,
then we can define the general equation for soil formation in a landscape:
Eh rr Ee El
sy y q D= 2 z Ž6.
El rs Et Et
The above equation states that the change in soil thickness over time depends
on the weathering of bedrock, the loss of materials due to chemical weathering
and the transport of soil through erosion.
Given appropriate initial and boundary conditions, Eq. Ž 6. can be solved
numerically using the finite-difference approach. The formulation using the
166 B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
h tq1 t
x, y y h x, y rr d e dl
sy q
Dt rs dt dt
t t t
z xq1, y y 2 z x , y q z xy1, y z xt , yq1 y 2 z xt , y q z xt , yy1
qD 2 q 2
ŽD x. ŽD y.
Ž7.
h s hi , z s zi t s 0, x , y G 0 hG0 tG0
de
s yP0 exp Ž ybh . t ) 0, 0 - x , y - `
dt
dl
s W0 exp Ž yk 1 h y k 2 t . t ) 0, 0 - x , y - `
dt
Eh Eh
s 0, s0 t ) 0, x y s 0, and x , y ™ `
Ex Ey
The initial condition describes the soil thickness and elevation at h i and z i . The
boundary conditions express that the soil thickness should be greater or equal to
zero at all times. The physical weathering rate is assumed to be an exponential
decay function of the soil thickness, and chemical weathering reduces the soil
Table 1
Parameters used in the simulation
Parameters Values
D 0.008 m2 yeary1
P0 1.0=10y3 m yeary1
b 1.5 my1
W0 1.6=10y3 m yeary1
k0 0.001 yeary1
k1 0.8 my1
k2 1.5 my1
rr 2.6 Mg my3
rs 1.6 Mg my3
Dt 100 year
D x, D y 2.0 m
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 167
Table 2
Parameters illustrating the combination low and high values of weathering and erosive diffusivity
Low High
2 y1 .
D Žm year 0.008 0.080
P0 Žm yeary1 . 0.1=10y3 1.0=10y3
W 0 Žm yeary1 . 0.1=10y3 3.0=10y3
thickness exponentially. The last condition defines the lower and upper bound-
ary for the finite-difference space grid.
3. Methods
Fig. 3. Semivariogram of the random noise added to the initial soil elevation.
168 B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
Fig. 4. Soil formation in a landscape after 10,000 years, shading represents soil thickness Žin m..
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 169
curvature. The soil needs to fill in the gullies before it can be transported
further. In reality, D must vary across the landscape depending on the slope.
A cross-section of the landscape shows the accumulation of the soil with time
ŽFig. 5.. The soil from the upper slope is transported down-slope by erosion and
filled the gullies. The soil thickness is highly correlated with the surface
curvature.
Fig. 6a shows the formation of the soil with time. The rate of formation is
initially rapid then gradually increases. The plot of the rate of soil thickness
changes over time is shown in Fig. 4b. In Minasny and McBratney Ž 1999. , we
presented a model with only physical weathering and the change in soil
thickness over time decreased exponentially. With the extension to chemical
weathering we can see that initially the rate is slow and increases due to the high
physical weathering of the bedrock and also the loss of soil caused by chemical
dissolution. After reaching a maximum weathering rate, the rate decreases
exponentially. Chemical weathering depends on the presence of water. At the
initial stage of weathering, the soil is thin and there is little water held by the
soil, as the soil grows thicker the chemical processes become more effective
until reaching a maximum rate at a certain thickness before it slowly decreases
following first-order kinetics.
The volume change due to weathering is usually measured as a strain
ŽBrimhall et al., 1991. . In our model since the weathering is assumed to be
Fig. 5. Cross-section across the landscape illustrating the increase in soil thickness and its relation
to surface elevation Žthick line.. Each line Žfrom the bottom. represents an increment of 200 years.
170 B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
Fig. 6. The change in soil thickness Ža., and its rate Žb., with time.
uni-dimensional, the strain can be calculated as the change in soil thickness over
the change in the soil–bedrock interface:
Dh
´s y1
De
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 171
The strain of the landscape after 10,000 years is illustrated in Fig. 7. For the
model without erosion, the strain will be uniform across the landscape. Because
of the transport of materials across the landscape the volume change is variable
depending on the position on the landscape. This has implications for detailed
weathering studies.
The effect of climate, rock types and land management is illustrated by
different combinations of weathering rates and erosive diffusivity. Results in
Fig. 8 show such combinations. At slow weathering and slow erosion rates soil
formation is small and distributed evenly in the landscape. When weathering is
intense the landscape progressively produces a thick soil mantle until it reaches
equilibrium. In extreme situations where the weathering is slow and erosion is
great, the soil in the ridges or hills are eroded leaving a bare rock exposure in
the surface where most of the soil is accumulated in the gullies or valleys. In an
intense weathering and erosion environment as the soil thickens it is also
distributed down through the landscape.
Addition of a small amount of randomness on the initial surface elevation or
weathering parameter produces a chaotic behavior in the results for the model.
The result for the soil thickness over time ŽFig. 9. shows bifurcation Ž Gleick,
1987.. The overall shape of the final elevation does not change dramatically, but
small variations are present locally, which are due to the variation of soil
thickness ŽFigs. 9 and 10. .
Semivariograms of elevation and soil thickness after 10,000 years of simula-
tion in the stable and apparently chaotic system are shown in Fig. 11. The soil
Fig. 7. Volume change Žstrain. in the landscape after 10,000 years. Continuous lines represent
areas with zero strain.
172
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
Fig. 8. Combinations of different weathering rates and erosive diffusivity on the shape of the landscape.
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 173
thickness ŽFig. 11b. of a chaotic system has a similar shape to the stable system
but initially has a much higher nugget variance Ž f 0.2 m2 .. Consequently, the
surface elevation of the apparently chaotic system exhibits higher variance
compared with the stable system at short distances. The variance difference
between the apparently chaotic and stable systems at the shortest distance Ž 2 m.
is approximately equal to the nugget variance of the soil thickness Ž f 0.2 m2 ..
In order to determine whether this result is merely random noise or chaotic
behaviour, dimensionality analysis is required Ž Fowler and Roach, 1993. . For a
nonlinear system, a time series can be described by multidimensional vectors of
time, the so-called phase space. A plot of the phase space of the hŽ t . may reveal
the chaotic behaviour of the system, which is usually characterized by a fractal
object known as a strange attractor.
Sugihara and May Ž1990. predicted the short-term future values in a time
series based on a library of the existing patterns. A plot of the correlation
coefficient R between the predicted and actual values over the embedding
dimension Ž E . ŽFig. 12. provides insight into the dimensionality of a system.
For a nonlinear dynamic system, the correlation should be high and decrease
with increasing number of dimensions, while random noise will have a low
correlation that does not change with increasing number of dimensions. The
correlation plot in Fig. 13 shows a high correlation at E s 2. This means that
174 B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
Fig. 10. Cross-section of the landscape showing the elevation at initial condition Žsmooth line. and
after 10,000 years in the apparent chaotic system in Ža. upper slope and Žb. lower slope.
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 175
Fig. 11. Semivariogram of Ža. elevation and Žb. soil thickness of the apparent chaotic system.
when plotted in two dimensions the time series will unfold the chaotic character-
istics of the system.
176
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179
Fig. 12. Difference plot of soil thickness over time and its magnification revealing the strange attractor.
B. Minasny, A.B. McBratneyr Geoderma 103 (2001) 161–179 177
Fig. 13. Correlation coefficient with the embedding dimension of the chaotic soil thickness
time-series.
Fig. 14. Stability ratio of numerical solution against the nugget variance of soil thickness.
appears stable only when the ratio F F 0.25. Therefore, the apparently chaotic
behaviour of this model may well be due to numerical instability.
5. Conclusions
In spite of the assumptions and limitations, the results are promising in terms
of quantitative modelling of pedogenesis. Application of the model can include a
prediction of the effect of land management on soil development and suggest
management practices in a landscape. However, to utilise and validate the model
in the landscape much field and laboratory work is needed to collect data to
estimate the parameters of the model.
Acknowledgements
References
Ahnert, F., 1977. Some comments on the quantitative formulation of geomorphological process in
a theoretical model. Earth Surf. Processes Landforms 2, 191–201.
Brimhall, G.H., Lewis, C.J., Ford, C., Bratt, J., Taylor, G., Warin, C., 1991. Quantitative
geochemical approach to pedogenesis: importance of parent material reduction, volumetric
expansion, and eolian influx in laterization. Geoderma 51, 51–91.
Fowler, A.D., Roach, D.E., 1993. Dimensionality analysis of time series data: nonlinear methods.
Comput. Geosci. 19, 41–52.
Gleick, J., 1987. Chaos. Making A New Science. Sphere Books, London.
Heimsath, A.M., Dietrich, W.E., Nishiizumi, K., Finkel, R.C., 1997. The soil production function
and landscape equilibrium. Nature 388, 358–388.
Hoosbeek, M.R., Bryant, R.B., 1992. Towards the quantitative modeling of pedogenesis—a
review. Geoderma 55, 183–210.
Minasny, B., McBratney, A.B., 1999. A rudimentary mechanistic model for soil production and
landscape development. Geoderma 90, 3–21.
Moran, C.J., McBratney, A.B., Koppi, A.J., 1988. A micromorphometric method for estimating
change in the volume of soil induced weathering. J. Soil Sci. 39, 357–373.
Press, W.H., Flannery, B.P., Teukolsky, S.A., Vetterling, W.A., 1992. Numerical Recipes: The
Art of Scientific Computing. Cambridge Univ. Press, Cambridge.
Roering, J.J., Kirchner, J.W., Dietrich, W.E., 1999. Evidence for nonlinear, diffusive transport on
hillslopes and implications for landscape morphology. Water Resour. Res. 35, 853–870.
Scheidegger, A.E., 1991. Theoretical Geomorphology. 3rd edn. Springer-Verlag, Berlin.
Sugihara, G., May, R.M., 1990. Nonlinear forecasting as a way of distinguishing chaos from
measurement error in time series. Nature 344, 734–741.
Warke, P.A., Smith, B.J., 1998. Effects of direct and indirect heating on the validity of rock
weathering simulation studies and durability tests. Geomorphology 22, 347–357.