You are on page 1of 28

Geoderma, 55 (1992) 183-210 183

Elsevier Science Publishers B.V., Amsterdam

Towards the quantitative modeling of


pedogenesis - a review

Marcel R. Hoosbeek and Ray B. Bryant


Department of Soil, Crop and Atmospheric Sciences, Cornell University, Ithaca, NY, USA
(Received March 3, 1992; accepted after revision June 6, 1992 )

ABSTRACT

Hoosbeek, M.R. and Bryant, R.B., 1992. Towards the quantitative modeling of pedogenesis - a
review. Geoderma, 55: 183-210.

The historical development of our pedogenetic model is reviewed, and trends or directions toward
the future are discussed. Pedogenetic models are characterized with respect to relative degree of com-
putation, complexity, and level of organization. These three characteristics are used as a framework
for classification. Early qualitative models have well served the purpose of soil survey to describe the
distribution of soils in landscapes. Further development of qualitative models will contribute to the
understanding of soil genesis in areas where soil surveys are not completed. In the developed coun-
tries, however, the scientific questions have changed dramatically. Greater interest lies in understand-
ing the physical and chemical processes of soil formation acting within relatively short time frames
and assessing the interactions between natural processes and environmental change or anthropogenic
impacts. Quantification of pedogenetic processes is a life-line to other environmental disciplines, and
a mechanistic understanding would, in an ideal situation, describe and predict the behavior of a sys-
tem for a limited number of years under changing environmental conditions.
A proposed approach to mathematical simulation of the dynamic pedogenetic processes is to inte-
grate soil physical, chemical, and other ecological system models into a quantitative pedogenetic model.
Models may be developed as research tools in data-intensive studies at the pedon or horizon level, or
for the purpose of simulating a soil system at the catena or soil region level. Specific kinds of pedoge-
netic models have specific implications with respect to the availability and spatial variability of the
data needed for development and testing.

INTRODUCTION

Scientific knowledge is based on the interaction between observations and


the hypotheses that correspond with those observations (Popper, 1959; France
and Thornley, 1984). The continual interaction between hypotheses (how we
think or expect things to work) and observations (attempts to measure the
real world) leads to progress in our scientific knowledge. This scientific

Correspondence to: M.R. Hoosbeek, Cornell University, 701 Bradfield Hall, Ithaca, N Y 14853,
USA.

0016-7061 / 92/$05.00 1992 Elsevier Science Publishers B.V. All rights reserved.
184 M.R. HOOSBEEK AND R.B. BRYANT

method, applied to an empirical science like pedology, was described by Dijk-


erman (1974) in seven stages; (1) the selection of a system (e.g. horizon,
pedon, soil landscape); (2) the measurement of properties; (3) the ordering
and condensing of data (e.g. by classification); (4) the explanation of the
data by hypotheses; ( 5 ) the testing of these hypotheses against new data; (6)
the structuring of confirmed hypotheses into scientific laws which together
form a body of well-established formal theory, and (7) the use of scientific
laws in predicting new unknown phenomena. Soil survey usually involves
stages ( 1 ), (2), and (3), while soil genesis research emphasizes stages (4)
and ( 5 ). To use the several stages of scientific methodology it is necessary to
reduce the complex natural soil system to a level of abstraction that can be
dealt with by the pedologist. This abstract reduction or simplification results
in a conceptual model. Conceptual models, as opposed to concrete models
(real physical objects, e.g. a soil column in the laboratory ), will be referred to
as "models".
Cline ( 1961 ) emphasized that a pedogenetic model embodies the summa-
tion of our knowledge of soils at any given time in the development of soil
science. Periodic assessment of the historical development of our pedogenetic
model, its current state, and trends or directions toward future developments
is a valuable exercise that serves to coordinate scientific efforts and encourage
advances in the field. Smeck et al. ( 1983 ) warned that "models can be limit-
ing to progress if the model is accepted as fact rather than as a source of hy-
potheses. All models should ultimately destroy themselves in whole or in part
as models simply represent a series of approximations towards the truth."
This paper reviews and classifies a number of pedogenetic models. Changes
in the kinds of models that have been developed, their uses and limitations,
and future needs are discussed.

A FRAMEWORK FOR CLASSIFICATION

All pedological models may be characterized with respect to relative degree


of computation, complexity, and level of organization. These three character-
istics are used as a framework for classifying kinds of models that have been
developed. The first characteristic, "degree of computation" applied in the
model, distinguishes between qualitative and quantitative models (Fig. 1 ).

Oualitative Ouantitative
Mental/Verbal Mathematical
Descriptive Deterministic/Stochastic

I .................................. I
Degree of computation --~

Fig. 1. Classification based on degree of computation.


TOWARDS THE QUANTITATIVEMODELING OF PEDOGENESIS- A REVIEW 185

Mental models are concepts and ideas that exist only in the h u m a n m i n d to
enable a soil scientist to work with the complex natural soil system (Dijker-
man, 1974). Verbal models are mental models expressed in spoken or written
language. Mental, verbal and descriptive models are qualitative and are placed
at one extreme of the scale. Mathematical models express abstractions in the
form of algorithms using either a deterministic or a stochastic approach. De-
terministic models presume that a simulated system will return one uniquely
defined outcome as the result of an input data set. The outcome of a stochas-
tic model is not one determined solution, but rather a distribution around an
average. The stochastic model presupposes the outcome is uncertain and
therefore uses a simulation structure to take this uncertainty into account
(Addiscott and Wagenet, 1985 ). Quantitative models are placed at the op-
posite extreme of the scale. In actuality, many models that have been devel-
oped fall somewhere between these two extremes.
The second characteristic, complexity of the structure used in the model,
distinguishes between functional and mechanistic models (Fig. 2). Func-
tional models use a simplified or empirical relation between observed and
simulated data without making a claim to fundamentality with regard to the
processes involved. In these models, usually having a low degree of complex-
ity, cause and effect are related through a black box. Mechanistic models in-
corporate fundamental mechanisms of the processes involved to a degree of
complexity corresponding to the present state of scientific knowledge (Addis-
cott and Wagenet, 1985 ). Capacity and rate models are differentiated in terms
of whether the main parameter of the model was a capacity or rate parameter.
Within the framework, capacity or rate models are differentiated by complex-
ity of the structure. Capacity models simulate changes of quantities without
time as a direct variable, as opposed to rate models in which the change of a
variable is defined as a function of time. The latter are more complex because
they require the use of differential equations and iterative procedures to solve
for the individual variables. A parallel in degree of complexity can also be
made between management and research oriented models. Management
models help make decisions regarding natural or agricultural resources with-
out the aim of simulating a natural system in detail. Rather, they aim to pro-
vide a reasonable answer, which in many cases can be obtained by using an

Functional Mechanistic

Capacity Rate

Management Research

I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
Complexity --~

Fig. 2. C l a s s i f i c a t i o n b a s e d o n c o m p l e x i t y .
186 M.R. HOOSBEEK AND R.B. BRYANT

empirical relation with a low degree of complexity. Research models are used
to test hypotheses or as tools to explore less well understood areas of knowl-
edge, which makes them inherently more complex.
The third distinction is based on the organizational hierarchy which de-
scribes at which level a model aims to simulate a natural system. Each level
can be regarded as a system by itself, with its own terminology, and can be
seen as a combination of subsystems at lower levels or as a subsystem of higher
level systems. Each level integrates the knowledge of subsystems at lower lev-
els, which means that investigations at a subsystem level, e.g. i - l, provide a
mechanistic understanding of a model at the/-level. The pedon is placed at
the central/-level in the hierarchy presented in this paper. Soil Taxonomy
(Soil Survey Staff, 1975 ) defines a pedon as a three dimensional natural body
large enough to represent the nature and arrangements of its horizons and
variability. The other levels (modified after Dijkerman, 1974 ) were defined
with the same idea in mind, i.e., three-dimensional bodies large enough to
represent the nature and variability at a certain level (Table 1 ). The spatial
variability of a soil system needs to be investigated at a level of resolution
appropriate for the level at which a model aims to simulate the system. Spa-
tial variability in soils forms a continuum from megascopic to microscopic
levels of resolution (Wilding, 1985 ). The level of resolution being used de-
pends on the/-level at which a model aims to simulate, the soil properties
used in the model, and the sampling methods.
The three characteristics may be combined and depicted graphically as axes
in a three-dimensional framework for classification (Fig. 3 ). Both degree of
TABLE1

Organizational hierarchy of soil systems (modified after D i j k e r m a n , 1974)

Level System Examples of types of modeling

i+ 3 Soil region Global phenomena (CO2 studies )


i+ 2 Catena Soil landscape modeling. Catchment area
budget studies
i+ 1 Polypedon Pedological modeling as part of a dynamic
ecosystem
i Pedon Dynamics of genetic processes (mass flow
models; eluviation --, illuviation)
i- 1 Horizon Dynamics of horizonation (mineral stabil-
ity and weathering; OM accumulation )
i- 2 Peds, Micromorphological studies (formation or
aggregates degradation of cutans; aggregate stability)
i- 3 Molecular Ion exchange phenomena; complexation o f
interaction metal ions by organic matter
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW 187

Organizational
Hierarchy
~.i+3

.i+2

Functional
i+1 o~'
Qualitative ."

Mechanistic
;.I-3
,t

Fig. 3. Modelclassificationbasedon threecharacteristics.

computation and complexity are continua with models positioned some-


where along the axes. The hierarchial levels are divided into "positive/-lev-
els" (i, i+ 1, i+2, i + 3 ) and "negative/-levels" ( i - 1 , i - 2 , i - 3) along the
vertical axis. Groupings of models for further discussion are rather artificial,
especially for models with multiple characteristics (e.g. a model that is partly
descriptive and partly mathematical).

QUALITATIVE MODELS

Early qualitative models, particularly at the positive /-levels, have well


served the purpose of soil survey, to describe the distribution and origin of
soil properties in landscapes.

Qualitative-functional-positive i-level models


Independently, Dokuchaev in 1892 (Dokuchaev, 1948 ) and Hilgard in 1892
(Fanning and Fanning, 1989 ) were the first to recognize the existence of the
five soil forming factors. The development of these concepts, which have
dominated soil genesis the last century, merits review and discussion.
Dokuchaev presented his qualitative ideas in equation form (Volobuyev,
1984):
II=f(K, O, F)B
Or in western notation:
S=f(cl, o, p ) t
188 M.R. HOOSBEEK AND R.B. BRYANT

where S = soil; cl= climate of a given region; o = plants and organisms (orga-
nisms); p = geologic substratum (parent material); to = relative age. Relief was
mentioned in a footnote as being important for some azonal soils. It is under-
standable that relief was not a part of the equation given the topography of
the Chernozem-steppe area in European Russia.
Jenny's ( 1941 ) equation has the form:
S = f ( c l , o, r, p, t .... )
where cl= environmental climate; o = organisms and their frequencies refer-
ring to species germules rather than actual growth; r=topography, also in-
cluding certain hydrologic features; p = parent material, defined as state of
soil at soil formation time zero; ... =additional, unspecified factors (e.g. eolian
deposition ).
Jenny sees the factors not as formers, creators, or forces, but as state factors
that define the state of the soil system. In a subsequent paper, Jenny ( 1961 )
derives the "general state factor equation" in which the fluxes affecting an
ecosystem are driven by potentials:
l, s, v, a = f ( L o , Px, t)
w h e r e / = e c o s y s t e m properties, s = soil properties, v= vegetation properties
and a = a n i m a l properties, are a function of the three "state factors":
Lo=initial state of the system, Px= external flux potentials, t=age of the
system.
This ( 1961 ) derivation led to the following statement: "... each ecosystem
property and each soil property is related individually to the state factors."
Jenny suggests that the relation between a certain soil property and a state
factor can be investigated in the field in a region in which one state factor is
dominant as compared to the combined contributions of the other state fac-
tors. Where the influence of the state factor cl is by far dominant as compared
to o, r, p, and t in describing a soil property, a climofunction can be defined
as:
l, s, v, a = f ( c l , o, r,p, t, ...)
And respectively with the other state factors as dominant:

l,s, v , a = f (0, cL r, p, l, ... ) biofunction


l,s, v,a=f ( r, cl, o, p, t .... ) topofunction
l,s,v,a=f (p, cl, o, r, t .... ) lithofunction
l,s,v,a=f ( t, cZ o, r, t, ... ) chronofunction
l,s, v , a = f ( .... cZ o, r, p, l) unspecified ex-function

These functions are explored in greater detail in part B of Jenny's (1980)


book. The time scales involved in these qualitative models are typically in the
order of hundreds to thousand of years and the levels of organizational hier-
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW 189

archy are at least at the polypedon ( i + 1) level, primarily at the catena ( i + 2 )


level, and in some studies at soil region (i + 3 ) level. Several specific exam-
ples are described and in some cases a linear regression is applied to the rela-
tion between a certain soil property and a dominant state factor.
Birkeland's (1974) book applied Jenny's approach to the study of soil
forming factors and introduced the concepts to soil geomorphic research.
Yaalon (1975) reviewed the status of the state factor functions: litho- and
topo-functions have been approached in a semi-quantitative way by assigning
a code to components; climofunctions have been criticized for their assumed
constancy of climate; a few biofunctions have been described in dry climates,
but generally vegetation is considered as a dependent attribute of the soil eco-
system; and chronofunction studies resulted mostly in graphical solutions.
Several authors questioned the possibility of solving Jenny's model mainly
because of the interdependency of the factors (Stephens, 1947; Runge, 1973;
Chesworth, 1973 ). Chesworth pointed out that the differential form of Jen-
ny's equation is not valid because of the diminishing effect of parent material
on soil composition with time, which implies that at least two of the factors
are not independent of each other. Kline (1973) wrote: "It is reasonably safe
to say that Jenny's ( 1941 ) equation, if ever fully written in analytical form as
an isomorphic model, would be of unprecedented complexity, and probably
completely intractable." Still, Richardson and Edmonds (1987) used simple
linear regression to describe segments of a curve for soil properties plotted
against a state factor. The relative effectiveness as indicated by the slope was
examined for specific examples.
In an effort to quantify soil development over the long time periods used in
the factors approach to studies of soil genesis, Harden and Taylor (1983)
investigated four chronosequences at the i + 2 catena level in different cli-
matic zones with the use of the "soil development index". Points were as-
signed to properties of a horizon as compared to the reference state of the
parent material (e.g. parent material has hue 7.5YR, then a horizon with hue
5YR scored 10 points because of rubification). Quantified properties were
presented in time plots to determine whether they developed at similar rates.
The soil development index was found to be useful and was presented as a
quantitative method, but the scale is non-parametric, which still leaves the
point system as a qualitative description technique.
Following the modern Russian school of thought, Volobuyev (1984) ap-
plied "soil formation energetics"to Dokuchayev's original equation: / / =
f ( K , O, F)B. The relation between the state of inputs and outputs was ex-
pressed as the transformation of vector x into vector y:

O--
Fro , (x) T=y , soils

Mm
190 M.R. H O O S B E E K A N D R.B. BRYANT

Volobuyev (1984) did not have a mathematical model for the transforma-
tion operator T to describe the soil-formation processes, but instead used the
energetics approach. We quote: "It has been established that the expenditure
of radiant solar energy R on soil-formation processes depends on relative wet-
ness P and the activity of the biogeocenosis m, the relation of which is de-
scribed by:
Q=R e-,/mC
where e is the base of the natural logarithm." Other parameters are:
Q=energetics of soil-formation processes; w=rate of mineral transforma-
tions; Pc = precipitation fixed in substances of the biogeocenosis; r = radiation
balance; p=atmospheric precipitation; C was not defined in Volobuyev's
paper.

K(r,p) }
O (m) factors ecw R 0"67
F(Pc, w) ' , Q=R( +__r)exp mp( +_p) ~soils
M(p, r)
Volobuyev claimed to have used this scheme for actual calculations of energy
expenditures on soil formation for different soils. Unfortunately, no exam-
ples or references regarding this calculation were provided.
There are many papers in which an existing qualitative soil genesis model
was noted to fall short in representing a natural soil system. Preliminary in-
vestigations result in new hypotheses, followed by field and laboratory tech-
niques to test the hypotheses. Confirmation leads to the establishing of a new
qualitative model (e.g., Richardson and Bigler, 1984; McDaniel and Munn,
1985; Amba et al., 1990; Puckett et al., 1990).

Qualitative-functional-negative i-level models


Simonson's (1959) Generalized Theory of Soil Genesis is a qualitative
functional model that is primarily concerned with the differentiation of ho-
rizons in a profile at the i - 1 level. Four kinds of changes were described:
additions, removals, transfers, and transformations. Differences between soils
and their horizons are explained by differences in relative intensity of the four
changes. The four changes are not defined in terms of actual processes (Hug-
gett, 1975 ), which limits the model to a functional and descriptive approach.
Arnold ( 1965 ) introduced the concept of multiple working hypotheses to
soil genesis: "The application of multiple working hypotheses, as discussed
by Chamberlain ( 1897 ), consists of considering each component hypothesis
proposed to explain an observed relationship, and then confirming, invali-
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW 191

dating, or setting it aside for further study. The result is a gradual increase of
the reliable evidence. The use of multiple working hypotheses allows one to
consider and analyze simultaneously several rational explanations of a phe-
nomenon. The best explanation is often a synthesis of the more acceptable
ideas contained in the respective hypotheses." The concept was applied to the
genesis of light-colored, grainy coatings on B-horizon structural aggregates in
Udolls. Four working hypotheses (qualitative functional models), based on
Simonson's theory, were described. The four working hypotheses represented
various scenarios of changing environmental state factors with coincidental
changes in landscape stability, parent material, and vegetation. Morphology
changes and process rates resulted in similar sets of observable soil character-
istics. Although the genesis of the grainy gray ped coatings could not be fully
understood for this particular application, Arnold ( 1965 ) concluded that the
multiple working hypotheses are useful in organizing data collections and as
a basis for further research.
In 1987, Johnson and Watson-Stegner introduced an evolutional model of
pedogenesis and claimed that only this evolutional model meets the following
conditions: "Such a model should be consistent with the observation that soils
are complex open systems, continuously adjusting by variable degrees, scales,
and rates to variable energy and mass fluxes, thermodynamic gradients, and
other changing exogenous environmental conditions." The model is written
as:

s=f(e, R)
where S=soil; P = progressive pedogenesis, or soil progression, which in-
cludes those processes, factors, and conditions that promote differentiated
profiles leading to physicochemical stability, namely horizonation/leaching
processes, developmental (assimilative) upbuilding, and/or deepening;
R = regressive pedogenesis, or soil regression, which includes those processes,
factors, and conditions that promote simplified profiles leading to physico-
chemical instability, namely haploidization/rejuvenation processes, surface
removals, and/or retardant (non-assimilative) upbuilding. In fact, this model
is a rephrasing of earlier models developed by Jenny (1941 ), Simonson
( 1959 ), Runge ( 1973 ), and others, with added emphasis on certain aspects.
Three years later, Johnson et al. (1990) introduced a "dynamic-rate model
of soil evolution", which depends heavily on philosophical redefinitions of
existing concepts. The model is still descriptive despite the use of dynamic
and passive pedogenetic vectors and their derivatives:
S=f(D, P, dD/dt, dP/dt)
where S=state of the soil or profile evolution; D=set of dynamic vectors;
P=set of passive vectors (both sets of vectors vary spatially and fluctuate
through time); dD/dt and dP/dt = vector changes through time.
192 M.R. HOOSBEEK AND R.B. BRYANT

Qualitative-mechanistic-positive i-level models

In the line of Jenny and Simonson, Runge (1973) developed an energy


model with a more mechanistic approach in which gravity is the energy source.
"Soil development at a landscape segment is dependent on the relative amount
of water running off (no net soil development) versus the amount of water
infiltrating into the soil (producing soil development)". Runge based his ideas
on the first two laws of thermodynamics. A closed system will approach a
state of m i n i m u m energy and m a x i m u m entropy at equilibrium. The Gibb's
free energy, G, is defined as the difference between the change in enthalpy, H,
and entropy, S, at constant temperature and pressure:
A G = A H - TAS
A decrease in enthalpy and an increase in entropy will only occur if there is a
positive energy input, AG, into the closed system. Runge applied this concept
to the open soil system. A loess soil parent material has initially a high en-
tropy ( m a x i m u m disorder), profile development decreases the entropy (or-
dering) which will only occur at the expense of an external energy source. The
leaching of water provides this energy. The model was expressed as:

s=f(o, w, t)
where s = soil development; o = organic matter production, renewing vector;
w= amount of water available for leaching, development vector; both vectors
are intensity factors. Runge's ( 1973 ) energy model was utilized to study the
genesis of a fine clay horizon (fl horizon) in a stable loess landscape. Other
examples were discussed as well. Smeck et al. (1983) extended Runge's en-
ergy model and gave a graphical representation of entropy changes on a rela-
tive scale as a function of the contribution of selected soil-forming processes
to the development of the ten taxonomic soil orders. Both Runge and Smeck
used laws of thermodynamics applicable to closed systems. Although the en-
ergy model is useful in a qualitative way, soils are open systems, therefore
actual quantitative thermodynamical calculations are not possible. An ap-
proach based on non-equilibrium thermodynamics would be more desirable.
Another point is that gravity is not the only source of energy. Solar energy
dries soils and also provides energy for the plants and organisms that modify
soils.
Hugger ( 1975 ) used the soil landscape system as an approach to model
the soil system. The valley basin is defined as the basic organizational unit of
the soil system and exists of a three dimensional open system whose bounda-
ries are defined as drainage divides, the surface of the land, and the weather-
ing front at the base of the soil profile. Two subsystems are defined within the
organizational unit, the soil skeleton and the soil solution. The soil skeleton
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW 193

is comprised of relative large and stable elastic sediments (including loess)


and is part of a geomorphic sediment-transporting system. The very mobile
and unstable soil solution is part of the hydrological system. The model is
based on the examination of the material fluxes of the subsystems through the
valley basin.

Qualitative-mechanistic-negative i-level models

There are many fine examples in the literature of papers in which an exist-
ing qualitative pedogenetic model was found to fall short as an abstraction of
the natural system. Based on pedogenetic research, new hypothetical models
were validated, for example, Franzmeier et al.'s (1989) model of fragipan
formation in landscapes of southern Indiana describes a mechanistic way in
which amorphous silica is accumulated as a bonding agent in fragipans. Sim-
ilar studies by Munn and Boehm, 1983; Karathanasis et al., 1983; Pickering
and Veneman, 1984; Rabenhorst and Wilding, 1986; Chadwick et al., 1987,
and many others have made major contributions to the mechanistic under-
standing of pedogenetic processes.

QUANTATIVE MODELS

The need for quantification of soil forming processes has become increas-
ingly urgent over the last few decades. From the outside, quantification is
increasingly desired by neighboring environmental disciplines.

Quantitative-functional-positive i-level models

Several individual processes important to pedogenesis have been quanti-


fied in a functional manner. The Agricultural Research Service (ARS), in
cooperation with other agencies and universities, developed models for use
in agricultural management (Flach, 1985 ) that contain sub-routines of inter-
est to pedogenesis. CREAMS 1&2 (Chemical, Runoff, and Erosion of Agricul-
tural Management Systems) are field scale models based on functional rela-
tions between soil information (hydrologic classification, K value for the
universal soil loss equation) and non-point source pollution and are used to
assess the effects of alternative management practices. EPIC (Erosion Produc-
tivity Impact Calculator) uses pedon data, slope length, and slope percentage
to calculate the effect of erosion on potential soil productivity. The Produc-
tivity Index Model statistically relates soil properties and productivity for a
number of Missouri soils. The functional approach inherently results in the
presence of parameters that need to be calibrated. The unknown factors are
usually calibrated for a number of soils in one particular area. The accuracy
194 M.R. HOOSBEEK AND R.B. BRYANT

of such a model when applied in another environment is unknown, and may


result in the prediction of nonsense.
Parton et al. (1987) developed the CENTLrR model which simulates the
formation of soil organic matter (SOM). The model considers three SOM
fractions: ( 1 ) active SOM with a short turnover time ( 1-5 yr); (2) slow SOM
with an intermediate turnover time (20-40 yr), and (3) passive SOM with
the longest turnover time (200-1500 yr). The system is defined by a set of
four driving variables: (1) annual precipitation affects the decomposition and
production of SOM; (2) temperature controls decomposition; (3) soil tex-
ture influences turnover rates, and (4) plant lignin content controls decom-
position rates of above- and below-ground material as a function of climate.
The parameters for these relations were estimated from published data using
a nonlinear data fitting procedure. Most of these relations are of a functional
type, others are of a more mechanistic type, e.g. based on laboratory incuba-
tion experiments and data. The model was validated by comparing soil C and
N to mapped values at 24 sites on the Great Plains (USA). Regional trends
in SOM on the Great Plains were predicted with an overall error of 150/0. The
model provides insight into the factors controlling SOM levels which is valu-
able to taxonomic and pedogenetic studies.
Several quantitative models have been developed to serve soil survey. Shovic
and Montagne (1985 ) developed a statistical model of soil-landscape rela-
tions to serve as an aid in making maximum use of limited data. Thematic
mapper data and topographic information from a digital elevation model were
combined by Lee et al. (1988) to determine soil characteristics of hilly ter-
rain in southwestern Wisconsin. Havens (1988) developed a GIS-based sta-
tistical model which predicts the percentage of the soil surface covered with
rock fragments. Although the model failed to predict for fragments smaller
than 25 cm in diameter, the probability of encountering a particular stoniness
class was simulated correctly in 77% of the cases during validation in the field.
A similar approach of soil-landscape modeling was used by Bell (1990) to
produce soil drainage class maps. The model had a 74% overall agreement
rate with field observations in the area of development (calibration), but failed
when applied to another site within the same physiographic province. The
approach is innovative in the use of new technology and appears to be useful
when properly calibrated for a specific area, however functional models are
inherently restricted to areas for which they are calibrated.
Kirkby (1985) developed a mathematical model for soil profile develop-
ment with a strong link to geomorphology. A weathering profile, organic pro-
file, and inorganic profile are simulated based on the following processes: per-
colation, equilibrium solution, leaching, ionic diffusion, organic mixing, leaf
fall, organic decomposition, and mechanical denudation. The weathering
profile describes the "proportion of substance remaining", p, at any depth.
The accumulated deficit of weathered material is:
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW | 95

7
w(z)= ] (l-p)dz
z=0

where z is depth below the soil surface. A simple linear flow law for a convex
slope ( d g / d x constant) with uniform permeability, K, is written as:
q(z) = K g w ( z )
where q(z)=total flow below depth z in the soil; g = t h e surface gradient;
K = t h e flow velocity on unit hydraulic gradient. Maximum percolation at
depth z is:
F(z) = F o + d [ (Kg) w(z) ]/dx=Fo+2W(Z)
where Fo = the limiting percolation rate at large depths; ;t = a constant for any
given site. The process of weathering is expressed as:
-Sp/St=c(p)q'
where c (p) = the total solute concentration at p; q' = the rate of flow through
the soil in 1 rag- ~y r - 1. These equations and equations for the other processes
are combined to second-order linear partial differential equations for each of
the three profiles. Kirkby mentioned the neglect of processes such as physical
translocation of clay, complexing, and cheluviation as simplifications. Other
limitations are the neglect of ion exchange and adsorption phenomena. The
solution chemistry is based on the assumption of equilibrium with the min-
eral phase, which, as will be discussed later in this review paper, is usually not
the case and can be described more successfully in terms of kinetics. How-
ever, Kirkby's model makes a valuable connection between soil profile and
hill slope and, despite the simplifications, it provides a basis for long term
quantitative soil profile modeling on a global scale.

Quantitative-functional-negative i-levelmodels

Levine and Ciolkosz (1986) developed "A Computer Simulation Model


for Soil Genesis Applications." The model uses Kline's (1973) compartment
design to represent the zone above the soil surface, and an upper and lower
horizon. Each compartment has an input and output flow and a parameter
that describes its content. Input consists of annual precipitation volume and
chemistry (the major ions) and standard soil characterization information.
The simulation of the major soil processes is based on functional relations,
e.g., the pH values are calculated from base saturation (BS) using linear
regression equations derived from The Pennsylvania State University soil data
base:
196 M.R. HOOSBEEK AND R.B. BRYANT

A horizons: pH = 0.04 BS+ 3.48 (r 2= 0.92 )

B horizons: p H = 0.04XBS+3.34 (r2=0.82)


The model was tested with soil analysis data from 21-week and 5-year sam-
plings, t-Test analysis indicated an effective simulation for 1- and 5-year pe-
riods. However, 1- and 5-year test periods are very short compared to the
1-year time increments of the model. Despite the non-mechanistic approach,
a major advantage of the model is its ability to simulate based on standard
soil data. In 1988, Levine and Ciolkosz used the same model to determine the
sensitivity of Pennsylvania soils to acid deposition.
The Newhall Simulation model is a simple functional model useful in de-
termining the soil moisture regimes used in Soil Taxonomy (Van Wambeke
et al., 1986). It approximates daily moisture conditions in a soil based on
limited monthly climatic data. Eight layers are divided by eight compart-
ments (8 X 8 matrix) in which water is held at a tension between field capac-
ity and permanent wilting point. The water holding capacity is fixed, accre-
tion takes place after precipitation, and depletion depends on the potential
evapotranspiration. The model was tested in the U.S.A. and proved to be
useful in parts of the world where data are sparse.
CALSOIL (Mayer, 1985 ) is a functional model that simulates the develop-
ment of calcic horizons. The simulation of water movement is similar to the
Newhall model. Precipitation enters the first compartment and fills it to field
capacity, excess water "flows over" to the next lower compartment. Carbon-
ate can cross compartment boundaries only in solution and is modeled as a
function of calcite equilibria, which depend on temperature and soil Pco2.
Another model of interest for pedogenesis is the Trace Element Transport
Model (TETrans)(Corwin and Waggoner, 1990). TETrans is a one-dimen-
sional capacity model which defines changes in amounts of solute driven by
the amounts of rainfall, irrigation or evapotranspiration with time as an in-
direct variable. The model utilizes a mass-balance approach to determine so-
lution and adsorbed (or exchangeable) solute concentration distributions.
Model options include: plant water uptake, hydraulic bypass, exchange, and
adsorption. Hydraulic bypass occurs when water moves through pores where
stagnant areas of immobile water exist (immobile water films, dead-end po-
res) and when water moves through large pores thereby bypassing water in
smaller pores. Several simplifying assumptions are made in TETrans; how-
ever, this functional approach has the advantage that the model is applicable
to situations where available data are limited.
Chadwick et al. (1990) presented a methodology, referred to as "a mass
balance interpretation of pedogenesis", to estimate the rates of weathering
and pedogenesis of the Cenozoic period. A set of analytical mass-balance
functions consisting of basic conservation equations, strain equations (to ac-
TOWARDS THE QUANTITATIVE M O D E L I N G OF PEDOGENESIS - A REVIEW 197

count for volumetric changes), and flux equations were combined with tra-
ditional selective extraction and particle-size separation procedures to inves-
tigate the overall long term pedogenetic processes that have turned beach sand
into an Alfisol.

Quantitative-mechanistic-positive i-level models

The soil compartment in an ecological study is often the least understood


and least quantified part of the ecosystem. The soil is frequently represented
as a black-box between the more quantitative mechanistic models of clima-
tologists, biologists, and hydrologists. Quantitative mechanistic pedogenetic
models are needed to interface with ecological models. This approach should
lead to a better understanding of how soils function in a changing environment.
Marion et al. (1985 ) developed a regional soil genesis model for CaCO3
deposition in desert soils of the southwestern United States. The CALDEP
model consists of five major components; (1) a stochastic precipitation model
based on monthly data resulting in three precipitation seasons; (2) an eva-
potranspiration model (Thornthwaite's equation ~ pan evaporation --* ac-
tual evapotranspiration); (3) chemical thermodynamic relationships of the
carbonate system; (4) soil parameterization (Pco2 per horizon, water-hold-
ing capacity), and (5) soil water and CaCO3 fluxes (only saturated flow was
considered, influx of calcium is through weathering and from dust and pre-
cipitation). The model was run using the present climate and three Pleisto-
cene climate scenarios and was highly sensitive to the frequency of storm
events, water holding capacity, and biotic control of Pcov Predicted CaCO3
deposition rates agreed with the rates for most field studies (1 to 5 g m -2
y r - l ) . The CALDEP model is one of the first soil models that used several
component models to simulate a soil forming process. Each component model
borrowed knowledge from other disciplines (statistics, climatology, soil phys-
ics, thermodynamics, soil chemistry, and soil characterization).
The Integrated Lake-Watershed Acidification Study (mWAS, Chen et al.,
1982 ) yielded an early acid rain model to simulate lake-watershed acidifica-
tion processes. The model is comprised of several modules including wa-
tershed hydrology, stream and lake hydraulics, canopy, snow, soil, stream and
lake chemistry. Incoming precipitation is routed through a layered soil profile
to the streams and ultimately to the lake. Field studies indicated the impor-
tance of the several possible flow-paths (surface-, shallow subsurface-, and
ground water-flow) in determining lake water acidity. ILWAScan be useful to
pedogenetic modeling in providing a shell consisting of non-soil compart-
ments, leaving the soil compartment for further improvements.
198 M.R. HOOSBEEK AND R.B. BRYANT

Quantitative-mechanistic-negative i-level models

Kline ( 1973 ) discussed a linear, constant-coefficient, compartment struc-


ture for quantitative mechanistic models which can be used for soil system
simulation. A series of interconnected compartments represents the actual
system. The behavior of a compartment is described by; ( 1 ) the flow rate
through the compartment (F), and (2) its content (C). From these two
measurable parameters, the rate constant 2, or transfer coefficient, is defined
as: 2=F/C. Many other, more complicated, methods describing compart-
mental transfers are available. De Wit's and van Keulen's ( 1972 ) book "Sim-
ulation of transport processes in soils" was an early treatise on the transport
of heat, salts, ions and water in the unsaturated phase. Algorithms and pro-
gram examples written in the simulation language CSMP are provided. Camp-
bell ( 1985 ) described the transport of water, gases, heat, and nutrients in soil
and plant systems with the use of example programs written in aASIC. Richter
( 1987 ) discussed several modeling approaches and transport processes in his
book "The Soil as a Reactor".
Bryant and Olson (1987) called for orientation of soil genesis modeling
efforts towards finding solutions to anthropogenically created environmental
problems. They proposed using a systems approach to combine subsystems
and compartments of various levels of resolution into a pedogenetic model.
The compartments dealing with soil processes would have a greater concen-
tration of interior detail than, for example, a compartment dealing with the
plant canopy. But, from a mathematical point of view, one has to keep in
aaind that "the chain is as strong as the weakest link". The accuracy of input
:o a soil compartment (e.g. the O horizon) heavily depends on the accuracy
~f output from a neighboring compartment (e.g. the canopy).
Pedogenetic modelers should not try to reinvent the wheel, but rather in-
corporate established models in biology, soil physics and chemistry as subsys-
tems into larger systems. Water and solute transport models and solution
chemistry models are two key categories of submodels for pedogenetic models
and will be discussed in greater detail.

Modeling water and solute transport


The movement of water and its dissolved constituents in the soil is funda-
mental to any quantitative mechanistic pedogenetic model since water plays
a major role in soil forming transfer and transformation processes. Realistic
representation of saturated and unsaturated flow in a typically non-isotropic
medium like soil is a complex problem. Both deterministic and stochastic
transport models have been developed by soil physicists. The deterministic
approach is generally limited to one dimension, while the stochastic approach
aims to incorporated the spatial variability as it occurs in the natural environ-
ment. Examples of both approaches will be discussed.
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW 199

SWATR (Feddes et al., 1978), or the later version SWATRE (Belmans et al.,
1983 ), are deterministic models that simulate the flow of water in the unsat-
urated zone based on Darcy's law and the continuity principle:

6H l) 60 6q ( d a y - 1)
q= -K~-~z (cm d a y - and ~-~= - ~ z

Combination yields a partial differential equation in terms of hydraulic head,


called the Richard's equation:
60 6{ 8H'X

Using the pressure head form of the flow equation and definition of the dif-
ferential moisture capacity C, C = d 0 / d ~ yields the flow equation for predict-
ing water movement in layered soils:

5t - C(g/) 6z ~-
The model deals with transient flow in a heterogeneous soil-root system which
either is or is not under groundwater influence. Water uptake by roots is sim-
ulated by a sink term depending on pressure head, rooting depth, and poten-
tial transpiration.
The water flow subroutine of LEACHM (Leaching Estimation And Chemis-
try Model; Wagenet and Hutson, 1987 ) uses a numerical solution to the Rich-
ard's equation. Soil hydrological characteristics (K-O-h relationships),
boundary conditions, and source and sink terms need to be defined or can be
calculated (e.g. the subroutine RETPRED calculates retentivity and conductiv-
ity parameters from particle size distribution data; POTET calculates daily po-
tential evaporation and transpiration from weekly pan evaporation totals).
Once the water fluxes are calculated, the corresponding chemical fluxes can
be estimated using a numerical solution to the diffusion-convection equa-
tion, taking into account concurrent sources and sinks of solute and sorption
on the solid phase. The derivation of the diffusion-convection equation, the
numerical solutions to it, and the theory of solute movement in general are
discussed in several papers (e.g. Wagenet, 1983; Wagenet, 1986; Wagenet and
Hutson, 1987).
Another model that might be of interest for pedogenesis is the "General
Purpose Simulation Model of Water Flow in the Soil-Plant-Atmosphere
Continuum" (GAPS), developed by Buttler and Riha (1989). Procedures
within the model can be selected based on desired components of the contin-
uum, modeling objective, and data availability. For instance, the water bal-
ance can be calculated either with the 'tipping-bucket-flow' procedure, a sim-
ple functional model based on a series of buckets that overflow into a lower
200 M.R. HOOSBEEK AND R.B. BRYANT

bucket when water-holding capacity is exceeded, or with the mechanistic


'Richard's equation' procedure.
Jury and Sposito (1985) mentioned the limitations of one-dimensional
deterministic models based on the convection-dispersion equation (e.g.
LEACHM ) when applied to solute transport in a three-dimensional experi-
ment: "...if the area scale is large and pronounced local variations exist in the
pore-water velocity, the convection-dispersion equation may not describe well
the situation wherein area-averaged soil solute concentrations develop from
a distributed surface input-source." The stochastic transfer-function model
(TFM), developed by Jury ( 1982 ), takes the spatial variability of water flux
into account and can be applied to the upper part of the unsaturated zone.
Solute movement is described by a travel-time probability density function
fL(t), such that fL(t)dt yields the probability of a solute reaching a certain
depth, z=L, in time interval, t to t+dt, after it was initially at the surface,
z = 0, at time zero. The model parameters can be determined by applying a
narrow pulse of a non-reactive solute at the soil surface, z = 0, with concentra-
tion Cin. The flux concentration at z=L is:

C(L, t) =~( Cin(t-t' )fL(t') dt'


0

With the travel-time density function represented in parametric form, the re-
sult of integration can be fitted to the experimental concentration data. Area-
averaged field-scale solute concentration data were used to determine the
shape of the two-parameter lognormal density function:
fL(t) = [ (2n)l/Eat]-1 exp[ (In t--/.t)2/2o "z]
where/z and a are mean and variance parameters.
Jury and Sposito ( 1985 ) examined the parameter estimation problem for
the lognormal transfer function model (TFM) and the convection dispersion
equation (CDE). The two models represent two different hypotheses for
mechanisms of solute movement at field scale. Chloride and bromide concen-
tration data were obtained from soil solution samplers and from soil coring.
Three calibration procedures, sum of squares optimization, method of mo-
ments, and maximum likelihood were used to estimate the parameters of each
model. "Calculated uncertainty in the parameter estimates was high enough
that no conclusion could be drawn about which model better described the
data from the solution sampler experiment, but a'ZM provided a better de-
scription of the coring experimental data after calibration than the CDE."
In 1986, Jury et al. generalized the transfer function model to describe sol-
ute movement that may undergo physical, chemical, and biological transfor-
mations in a natural soil. A soil unit with a transport volume, Vst and its
TOWARDS THE QUANTITATIVE MODELING OF PEDOGENESIS - A REVIEW 201

boundary surface, Sst, was defined in which the flow of water may vary in
space and time. The general transfer function equation is written as:
O0

Qout(t)=J g ( t - t ' It') Qin(t') dt'


o

where Qout(t) is the rate of solute loss from the soil unit; Qin(t' ) is the mar-
ginal probability density function for the solute input time and can be inter-
preted as the rate at which solute mass enters Vst for the first time divided by
the total mass of solute input; g(t- t' [t') is the lifetime density function which
represents the net effect of soil processes, e.g. convection, dispersion, sorp-
tion, transformations, etc. In order to solve the lifetime density function for
a particular solute transport, the transfer function equation needs to be solved
with the use of experimental field data that relate solute mass input and loss
rates as functions of time. White et al. (1986) applied the transfer function
model to solute transport through undisturbed soils with Br and C1 as tracers.
Travel times varied with soil type, initial water content, and rate of water
input. The lognormal distribution of travel times, as initially assumed by Jury
( 1982 ), described reasonably well the probability density function of some
soils, but was found inadequate when applied to well-structured soils. The
existence of a component of fast solute transport through large, interped voids
and a component of slow transport through intraped voids, resulted in prob-
ability density functions with different shapes. This phenomenon should be
incorporated in the transfer function model of a well-structured soil. In an
other field study comprising chemical outflow in tile drains, the transfer func-
tion model was found to be a successful tool in interpreting and predicting
pesticide travel time characteristics (Utermann et al., 1990).
Both the deterministic and stochastic approach to modeling solute move-
ment are of interest for pedogenetic modeling. The deterministic approach,
based on the Richard's equation for example, might be more accessible in
case standard soil data are available. The stochastic approach has the theoret-
ical advantage of incorporating spatial variability, but the intensity of specific
experimental field data which is needed to solve the transfer function equa-
tions for each soil, might be a disadvantage.

Modeling solution chemistry


Sposito ( 1989 ) demonstrated that the kinetics of complexation in the soil
solution is fast enough to assume instant equilibrium. There are many chem-
ical equilibrium models that calculate ion activities and speciations based on
thermodynamic relations between solutions and solid phases. Chemical spe-
ciation can be based on the minimization of the Gibbs' free energy or on the
equilibrium constant approach (Baham, 1984). The latter is most widely em-
ployed in geochemical models, e.g. MINEQL ( W e s t a l l et al., 1976) and
202 M.R. HOOSBEEK AND R.B. BRYANT

GEOCHEM (Sposito and Mattigod, 1979). These models calculate the distri-
bution of chemical species in an aqueous system, and model the precipitation
of solid phases.
The kinetics of dissolution reactions is generally too slow for the soil solu-
tion to reach equilibrium with the solid phase under field conditions (Manley
et al., 1987; Sposito, 1989). Kinetics may play a more important role than
thermodynamics in controlling Al solubility in soils (McBride and Erich,
1985). The rates of chemical weathering reactions depend on mineralogy,
temperature, flow rate, surface area, pH, and ligand (DOC) and CO2 concen-
tration in soil solution (Sparks, 1989). Controlled laboratory experiments
led to the distinction of three major rate-limiting mechanisms for mineral
dissolution (at undersaturation); (1) transport of solute away from the dis-
solved mineral; (2) surface reaction-controlled kinetics where ions are de-
tached from the mineral surface, and (3) a combination of (1) and (2)
(Sparks, 1989). Models that simulate these processes are very sensitive to the
type of experimental method being used. Van Grinsven (1988) compared
batch and column techniques in obtaining dissolution rates of C horizon ma-
terial from a Dystrochrept. Intensive stirring increased rates by more than an
order of magnitude compared with column experiments and unstirred batch
experiments. All rates were found to be 5 to 10 times higher than estimated
rates from field studies. Hydrodynamic factors and the condition of the min-
eral surface (freshness, roughness, presence of coatings) are major factors in
the discrepancies between laboratory experiments of different researchers and
between laboratory and field conditions. Both Stumm et al. ( 1985 ) and Spos-
ito (1989) described the dissolution of clay minerals and metal hydrous ox-
ides as a surface controlled zero-order reaction in which the rate is indepen-
dent of the aqueous-phase concentration of an ionic constituent of the mineral.
Bohn et al. ( 1985 ) argued that weathering in a soil is better characterized by
a first-order reaction in which the mineral concentration is incorporated in
the equation.
Warfvinge and Sverdrup (1989) coupled four reactions, dissolution of cal-
cite in a stagnant aqueous system, cation exchange, leaching and accumula-
tion of dissolved components, and the carbonate equilibrium system, into
one mathematical model for dissolution of limestone in acid soils. The differ-
ential conservation equations constituting the framework for the model are
derived and simplified for these specific conditions as:
d[ANC]/dt= 1/z ( [ANC]0-- [ANC] )+Rd+Rx
d [Ca 2+ ] / d t = 1/z ( [Ca 2+ ] o - [Ca 2 ] ) + (Rd+Rx)/2
where the acid neutralizing capacity is defined as:
ANC= IOH- ] + [HCO~- ] +2 [CO2- ] - [ n ]
TOWARDS THE QUANTITATIVEMODELING OF PEDOGENESIS- A REVIEW 203

and the average turnover time of the soil solution in the limed layer is calcu-
lated as:
r=zO/Q
where z = height of limed layer; 0= volumetric water content; Q = flow inten-
sity through the layer.
Dissolution of calcite powder and contribution from the cation exchange re-
action are calculated respectively as:
Ra= -2/(MOz) Y~j(dmj/dt)
where M = molar mass of calcite; dmj/dt= rate in change of mass due to dis-
solution for particle size fractionj and:
Rx = - (fEE "p)/0 dXca/dt = (CEC-/9) / 0 dXAc/dt
where p = bulk density of dry soil; dXca and dXkc = change of the ratio of
moles of charge of exchangeable Ca and acidity to the total CEC. Despite the
many assumptions and the limited chemistry, the model is interesting as an
example for its integration of several process equations.
For the pedogenetic modeler, submodels that describe chemical phenom-
ena based on thermodynamics are readily available. Submodels dealing with
other chemical phenomena such as the kinetics of mineral dissolution and ion
exchange need further development, however the equations based on the pre-
sented theory may be incorporated directly into pedogenetic models.

AN APPROACH TO PEDOGENETIC SIMULATION MODELING

A proposed approach to mathematical simulation of the dynamic pedoge-


netic processes of the natural soil system is to integrate physical, chemical,
and other components of the ecological system into one quantitative pedoge-
netic model. An important question is for what purpose this genetic model is
to be developed. Will the model be used as a research tool in a data-intensive
study at the pedon or horizon level, or is the model supposed to simulate a
soil system at the catena or soil region level? Both extremes have specific im-
plications with respect to the availability and spatial variability of the data
needed for simulation and will be discussed separately.

Pedogenetic modeling at the negative i-level


The following discussion of pedogenetic modeling at the negative i-level is
a brief account of the authors' experience with developing a pedogenetic model
primarily for use as a research tool to provide insight into the mechanisms of
podzolization and as an aid in testing several hypotheses. A detailed study
site is installed in a boreal forest of the Adirondack Mountains in New York.
204 M.R. HOOSBEEK AND R.B. BRYANT

Well-expressed horizons of a Typic Haplorthod developed in sandy glacial


outwash deposits are being monitored using tensiometers, samplers with ce-
ramic porous cups, samplers with non-ceramic porous plates, and thermis-
ters. Soil solutions, precipitation, and canopy throughfall are regularly col-
lected and chemically analyzed for organic and inorganic components.
Undisturbed core samples are used in laboratory experiments to provide data
on ion exchange and metal- organic interactions.
The soil profile is divided into layers whose properties are assumed to be
concentrated at the nodes (a horizon may have several layers). The pedoge-
netic model is an integration of submodels dealing with the following pro-
cesses: (1) solute movement; (2) soil temperature profiles; (3) microbial de-
composition of soil organic carbon (SOC) and dissolved organic carbon
(DOC); (4) mineral dissolution; ( 5 ) ion exchange; ( 6 ) adsorption; ( 7 ) spe-
ciation and complexation, and (8) precipitation. Each submodel simulating
a process will be described briefly.
Solute movement and soil temperature profiles are simulated with the ex-
isting LEACHM model (Wagenet and Hutson, 1987) and are based on real-
time weather data.
The submodel dealing with microbial decomposition is in part based on the
CENTURY model (Parton et al., 1987). Microbial decomposition takes place
in six pools, structural and metabolic plant remains, active soil organic car-
bon (SOC) (decomposing plant residues, live microbes), slow SOC (micro-
bial metabolites), passive SOC (humified material), and DOC. The general
equation for decomposition within a soil environment is
dCx/dt = KMAXx X THETA X SLTEMP X Cx
where Cx=Carbon state variable of pool x; KMAXx=maximum decomposi-
tion rate for pool x; THETA= volumetric water content; SLa'EMP= soil temper-
ature. The products are CO2 + H20, SOC flowing to other pools, and DOC.
The ratio in which these products are produced depends on the volumetric
water content, which is hypothesized to be an important factor in the podzo-
lization process.
The rate equation used to simulate the kinetics of dissolution reactions is a
compilation of equations developed by Bohn et al. ( 1985 ) and Stumm et al.
(1985):
--d[MIN]/dt=KH [H + ] ' [MIN] + K L [LIGAND- ] X [MIN]
where MIN ----mineral concentration ( m o l / m 3); KH = rate constant related to
pH ( tool/day ); H + -- proton concentration (mol/1 ); n = fractional hydroly-
sis reaction order constant; KL = rate constant related to organic ligand con-
centration; LIGAND= dissolved organic ligand concentration (mol/1 ). The
dissolution rate is composed of two additive rates, the effect of surface pro-
tonation and the effect of organic ligands.
TOWARDS THE QUANTITATIVEMODELING OF PEDOGENESIS - A REVIEW 205

The cation exchange capacity (CEC) can be thought of as the result of iso-
morphic substitution, and ionization of functional groups. The CEC can be
measured per horizon or if not available estimated from empirical formulae.
The CEC due to isomorphic substitution is calculated from the structural for-
mula of a clay mineral. The contribution of SOM is estimated from (Mc-
Bride, 1991 ):
CEC (mmol/kg of SOM) = -- 600 + 500 X pH
But in acidic soils free Fe 3+ and A 1 3 + in solution may block potential CEC
sites on SOM by the formation of stable complexes. Due to the concentration-
charge effect, which applies to the study area (excessive leaching, small elec-
trolyte concentrations), Ca +, Mg 2+, and A13+ dominate the exchange sites.
Also, from a pedogenetic point of view, the base cations can be represented
by one variable in the following selectivity equation (McBride, 1991 ):
Ks= (NAI)2 " (Mca+Mg)3XMT
(MA02
( N c a + Mg) 3 , x

where N = fraction of exchange sites occupied; M = m o l e fractions, e.g.


MA = mA/mT; MT = total molarity of the solution.
Chemical speciation in soil solution, surface adsorption, and the precipi-
tation of solid phases, are calculated with subroutines of the chemical equilib-
rium program M I N E Q L + (Westall et al., 1976; Schecher and McAvoy, 1991 ).
The pedogenetic model as discussed is a one dimensional model. Spatial
variability is not represented in the calculations. By applying the model to
individual sites (three profiles are being monitored), statistical analysis can
be performed on the results of the simulations to obtain information on the
variability among the sites. In addition to its primary purpose, the pedoge-
netic model provides an impression of how simulated changes in environ-
mental factors influence the dynamics of the soil, under the limitation that
extrapolation in time is only realistic for a limited number of years.
Pedogenetic modeling at the positive i-level
The following discussion of pedogenetic modeling at the positive i-level
represents the authors' opinions and reflections on present concepts and fu-
ture directions for research, particularly with respect to the problem of spatial
variability in soils and availability of data. Spatial variability in a catena or
landscape presents a major challenge for pedogenetic modeling. One method
addressing spatial variability is to apply a deterministic (one-dimensional)
simulation model to individual points in a landscape and perform statistical
analyses on the results. In a pesticide movement study, Petach and Wagenet
(1989) grouped soils from a 7 by 10 km 2 site into six groups with similar
hydraulic properties. For each of the hydraulic groups, the LEACnM model
(Wagenet and Hutson, 1987 ) was executed 100 times with randomly selected
hydraulic parameters, based on Monte Carlo techniques. The generated fre-
quency distributions for each of the output parameters, mass, concentration,
206 M.R. H O O S B E E K A N D R.B. BRYANT

and surface matrix potential, were used to produce maps with the help of a
geographic information system.
Another way of incorporating spatial variability is to use stochastic models.
The TFM approach, as discussed previously, could be used in pedogenetic
studies to calculate travel times of mass movement resulting in horizonation
or other features.
Only research sites provide the luxury of sufficient data for developing and
testing simulation models. Bouma (1989) discussed "the challenge for soil
science to translate data we have ( soil survey ) to data we need." Pedotransfer
functions define relationships between different soil characteristics and prop-
erties (Bouma and van Lanen, 1987; Wosten et al., 1990). Bouma (1989)
provided examples of flow diagrams in which soil characteristics are related
by several levels of pedotransfer functions to a particular land quality, e.g.
soil water deficit. Soil structure descriptions from soil surveys can be benefi-
cial in explaining highly variable measurements and simulation results of hy-
draulic properties (Bouma, 1989 ). Bypass flow, defined as the vertical move-
ment of "free" water through an unsaturated soil matrix (Bouma, 1981 ),
might explain observed variability among hydraulic data simulated with the
Richards equation for example. Structure data also provide guidance in es-
tablishing representative elementary volumes (REV'S)of soil samples (Bouma,
1983).

EPILOGUE

Soil genesis models have been reviewed within a framework based on the
degree of computation, complexity, and hierarchical level. Early qualitative
models have well served the purpose of soil survey to describe the distribution
of soils in landscapes, and they will continue to do so. Further development
of qualitative models will contribute to the understanding of soil genesis in
areas where soil surveys are not completed. On a global scale, there is much
work left to be done.
In the developed countries, however, soil survey is nearer completion and
the scientific questions have changed dramatically over the last few decades.
How do soils function in today's environment and how will they function and
respond to a changing environment in the future? A first necessity is quanti-
fication because only numerically quantifiable hypotheses can be properly re-
lated with numerical observations, and quantification is a life-line to other
environmental disciplines. Secondly, a mechanistic understanding would, in
an ideal situation, describe and predict the behavior of a system for a limited
number of years under changing environmental conditions.
An approach for developing pedogenetic simulation models by integrating
submodels from related disciplines is proposed. Pedologist working in the dis-
cipline of soil genesis are challenged to adopt simulation modeling tools and
TOWARDSTHEQUANTITATIVEMODELINGOF PEDOGENESIS- A REVIEW 207

begin the arduous task of collecting data for building and testing quantitative
pedogenetic models.

ACKNOWLEDGEMENTS

The authors thank Dr. Johan Bouma of the Agricultural University Wag-
eningen, The Netherlands, for reviewing an earlier draft of this paper and
wish to acknowledge the financial support of the USDA Soil Conservation
Service.

REFERENCES

Addiscott, T.M. and Wagenet R.J. 1985. Concepts of solute leaching in soils: a review of mo-
delling approaches. J. Soil Sc., 36:411-424.
Amba, E.A., Smeck, N.E., Hall, G.F. and Bigham, J.M. 1990. Gcomorphic and Pedogcnic Pro-
cesses Operative in Soils of a Hillslope in the Unglaciated Region of Ohio. Ohio J. Sci., 90
(1):4-12.
Arnold, R.W., 1965. Multiple working hypotheses in soil genesis. Soil Sci. Soc. Am. Proc., 29 :
717-725.
Baham, J. 1984. Prediction of ion activities in soil solution: computer equilibrium modeling.
Soil Sci. Soc. Am. J., 48: 525-531.
Bell, J.C., 1990. A GIS-based soil-landscape modeling approach to predict soil drainage class.
Ph.D. Thesis, Pennsylvania State Univ.
Belmans, C., Wesseling, J.G. and Fcddes, R.A. 1983. Simulation model of the water balance of
a cropped soil: SWATRE.J. of Hydrology, 63:271-286.
Birkeland, P.W., 1974. Pedology, Weathering, and Gcomorphological Research. Oxford Uni-
versity Press, New York.
Bohn, H.L., McNeal, B.L. and O'Connor, G.A., 1985. Soil Chemistry.Wiley, New York.
Bouma, J., 1981. Soil Morphology and Preferential Flow along Macropores. Agric. Water Man-
age., 3: 235-250.
Bouma, J., 1983. Use of soil survey data to select measurement techniques for hydraulic con-
ductivity. Agric. Water Manag. 6 (2/3): 177-190.
Bouma, J., 1989. Using soil survey data for quantitative land evaluation, pp. 177-213. In: B.A.
Stewart (Editor) Advances in Soil Science, 9. Springer, New York.
Bouma, J. and van Lanen, H.A.J., 1987. Transfer functions and threshold values: From soil
characteristics to land qualities. In: K.J. Beck, P.A. Burrough and D.E. MeCormack (Edi-
tors), Proc. ISSS/SSSA Workshop on Quantified Land Evaluation Procedures. Int. Inst. for
Aerospace Sure. and Earth Sci. Publ. No. 6, Enschede, pp. 106-111.
Bryant, R.B. and Olson, C.G., 1987. Soil genesis: opportunities and new directions for research.
In: Future Developments in Soil Science Research. Soil Sci. Soc. of America. WI.
Buttler, I.W. and Riha, S.J., 1989. GAPS: A General Purpose Simulation Model of the Soil-
Plant-Atmosphere System. Cornell University Press, Ithaca, NY.
Campbell, G.S., 1985. Soil Physics with BASIC - Transport Models for Soil-Plant Systems.
Elsevier, Amsterdam.
Chadwick, O.A., Brimhall, G.H. and Hendricks, D.M., 1990. From a black to a gray box - a
mass balance Interpretation ofpedogcncsis. Geomorphology, 3: 369-390.
Chadwick, O.A., Hendricks, D.M. and Nettleton, W.D., 1987. Silica in duric Soils: I. A Depo-
sitional Model. Soil Sci. Soc. Am. J., 51: 975-982.
208 M.R.HOOSBEEKANDR.B.BRYANT

Chen, C.W., ASCE, M., Dean, J.D., Gherini, S.A. and Goldstein, R.A., 1982. Acid rain Model:
hydrologic module. J. Environ. Eng. Div., Proc. Am. Soc. Civil Eng., Vol. 108, No. EE3.
Chesworth, W., 1973. The parent rock effect in the genesis of soil. Geoderma, 10:215-225.
Cline, M.G., 1961. The changing model of soil. Soil Sci. Soc. Proc., 25: 442-446.
Corwin, D. L. and Waggoner, B., 1990. Trace Element Transport Model; Solute Transport
Modeling Software, IBM-Compatible Version 1.5 USDA-ARS, U.S. Salinity Laboratory,
Riverside, CA.
De Wit, C.T. and Van Keulen, H., 1972. Simulation of Transport Processes in Soils. Pudoc,
Wageningen.
Dijkerman, J.C. 1974. Pedology as a science: the role of data, models and theories in the study
of natural soil systems. Geoderma, 11: 73-93.
Dokuchaev, V.V. 1948. Russian Chernozem--Selected Works of V.V. Dokuchaev, Volume I.
Moskva, 1948. Translated from Russian by Israel Program for Scientific Translations, Jeru-
salem, 1967.
Fanning, D.S. and Fanning, M.C.B. 1989. SOIL, Genesis, and Classification. Wiley, New York.
Feddes, R.A., Kowalik, P.J. and Zaradny, H., 1978. Simulation of Field Water Use and Crop
Yield. Pudoc, Wageningen.
Flach, K.W., 1985. Modeling and Soil Survey. Soil Survey Horizons. Summer 1985, p. 15-20.
France, J. and Thoruley, J.H.M., 1984. Mathematical Models in Agriculture. Butterworths,
London.
Franzmeier, D.P., Norton, L.D. and Steinhardt, G.C., 1989. Fragipan formation in loess of the
Midwestern United States. In: N.E. Smeck, and E.J. Ciolkosz, (Editors) Fragipans: Their
Occurrence, Classification, and Genesis. SSSA Special Publ.Nr. 24. WI.
Harden, J.W. and Taylor, E.M., 1983. A qualitative comparison of soil development in four
climatic regimes. Quartemary Res., 20: 342-359.
Havens, M.W., 1988. A GIS-based soil-landscape, modeling approach to predict surface rock
fragment distributions. M.S. Thesis.The Pennsylvania State Univ.
Huggett, R.J., 1975. Soil landscape systems: a model of soil genesis. Geoderma, 13: 1-22.
Jenny, H., 1941. Factors of Soil Formation --A System of Quantitative Pedology. McGraw-
Hill, New York.
Jenny, H., 1961. Derivation of state factor equations of soils and ecosystems. Proc. Soil Sci.
Soc. Am., 25 : 385-388.
Jenny, H. 1980. The Soil Resource---Origin and Behavior. Springer, New York.
Johnson, D. L. and Watson-Stegner, D., 1987. Evolution Model of Pedogenesis. Soil Sci., 143
(5): 349-366.
Johnson, D.L., Keller, E.A. and Rockwell, T.K., 1990. Dynamic pedogenesis: new views on some
key soil concepts, and a model for interpreting quaternary soils. Quaternary Res. 33: 306-
319.
Jury, W.A., 1982. Simulation of solute transport using a transfer function model. Water Resour.
Res., 18 (2): 363-368.
Jury, W.A. and Sposito, G., 1985. Field calibration and validation of solute transport models
for the unsaturated zone. Soil Sci. Soc. Am. J., 49: 1331-1341.
Jury, W.A., Sposito, G., and White, R.E., 1986. A transfer function model of solute transport
through soil 1. Fundamental concepts. Water Resour. Res. 22 (2): 243-247.
Karathanasis, A.D., Adams, F. and Hajek, B.F., 1983. Stability relationships in kaolinite, gibb-
site, and Al-hydroxyinter-layered vermiculite soil systems. Soil Sci. Soc. Am. J. 47: 1247-
1251.
Kirkby, M.J., 1985. A basis for soil profile modeling in a geomorphic context. J. Soil Sci., 36:
97-121.
Kline, J.R., 1973. Mathematical simulation of soil-plant relationships and soil genesis. Soil Sci.,
115 (3): 240-249.
TOWARDSTHEQUANTITATIVEMODELINGOFPEDOGENESIS- AREVIEW 209

Lee, K-S., G.B. Lee, and E.J. Tyler. 1988. Thematic mapper and digital elevation modeling of
soil characteristics in hilly Terrain. Soil Sci. Soc. Am. J., 52:1104-1107.
Levine, E.R. and Ciolkosz, E.J., 1986. A computer simulation model for soil genesis applica-
tions. Soil Sci. Soc. Am. J., 50: 661-667.
Levine, E.R. and Ciolkosz, E.J., 1988. Computer simulation of soil sensitivity to acid rain. Soil
Sci. Soc. Am. J., 52: 209-215.
Manley, E. P., Chesworth, W., and Evans, L.J., 1987. The solution chemistry of podzolic soils
from the eastern Canadian shield: a thermodynamic interpretation of the mineral phases
controlling soluble Al3+ and H4SiO4. J. Soil Sci.,38: 39-51.
Marion, G.M., Schlesinger, W.H. and Fonteyn, P.J., 1985. CALDEP: A regional model for soil
CaCO3 (Caliche) deposition in southwestern deserts. Soil Sci., 139 (5): 468-481.
Mayer, L. 1985. The Distribution of Calcium Carbonate in Soils: A Computer Simulation using
Program CALSOIL. U.S. Geol.Surv. 975, Menlo Park, CA.
McBride, M.B. 1991. Ion Exchange. Dept. of Soil, Crop, and Atmospheric Sciences, Cornell
University. Class Notes (unpubl.).
McBride, M.B. and Erich, M.S., 1985. The Chemistry of Aluminum in Soils. Dept. of Agron-
omy, Cornell University, Ithaca, NY.
McDaniel, P.A. and Munn, L.C., 1985. Effect of temperature on organic carbon-texture rela-
tionships in Mollisols and Aridisols. Soil Sci. Soc. Am. J., 49:1486-1489.
Munn, L.C. and Boehm, M.M., 1983. Soil genesis in a Natrargid-Haplargid complex in North-
ern Montana. Soil Sci. Soc. Am. J., 47:1186-1192.
Parton, W.J., Schimel, D.S., Cole, C.V. and Ojima, D.S., 1987. Analysis of factors controlling
soil organic matter levels in Great Plain grasslands. Soil Sci. Soc. Am. J., 51:1173-1179.
Petach, M. and Wagenet, R.J., 1989. Integrating and analyzing spatial variable soil properties
for land evaluation. In: J. Bouma and A.K. Bregt, (Editors), Land Qualities in Space and
Time. Pudoc, Wageningen, pp. 145-154.
Pickering, E.W. and Veneman, P.L.M., 1984. Moisture regimes and morphological character-
istics in a hydrosequence in central Massachusetts. Soil Sci. Soc. Am. J., 48:113-118.
Popper, K.R. 1959. The Logic of Scientific Discovery. Hutchinson, London.
Puckett, W.E., Collins, M.E. and Schellentrager, G.W., 1990. Design of soil map units on a
karst-area in west central Florida. Soil Sci. Soc. Am. J., 54: 1068-1073.
Rabenhorst, M.C. and Wilding, L.P., 1986. Pedogenesis on the Edwards plateau, Texas: III.
New model for the formation of petrocalcic horizons. Soil Sci. Soc. Am. J., 50: 693-699.
Richardson, J.L. and Bigier, R.J., 1984. Principal component analysis of prairie pothole soils in
North Dakota. Soil Sci. soc. Am. J., 48: 1350-1355.
Richardson, J.L. and Edmonds, W.J., 1987. Linear regression estimations of Jenny's relative
effectiveness of state factors equation. Soil Sci., 144 (3): 203-208.
Richter, J. 1987. The Soil as a Reactor. Catena, Cremlingen.
Runge, E.C.A. 1973. Soil development sequences and energy models. Soil Sci., 115 (3): 183-
193.
Schecher, W.D. and McAvoy, D.C., 1991. MINEQL+: A chemical equilibrium program for
personal computers. User's manual, version 2.1. Environmental Research Software, Edge-
water, MD.
Shovic, H.F. and C. Montagne. 1985. Application of a statistical soil-landscape model to an
order III wildland soil survey.Soil Sci Soc. Am. J., 49: 961-968.
Simonson, R.W. 1959. Outline of a generalized theory of soil genesis. Soil Sci. Soc. Am. Prec.,
23: 152-156.
Smeck, N.E., Runge, E.C.A., and Mackintosh, E.E., 1983. Dynamics and genetic modelling of
soil systems. In: L.P. Wilding, N.E. Smeck, G.F. Hall, (Editors) Pedogenesis and Soil Tax-
onomy 1. Concepts and Interactions. Elsevier, Amsterdam, pp. 23-49.
210 M.R. HOOSBEEK AND R.B. BRYANT

Soil Survey Staff, 1975. Soil Taxonomy. Soil Conservation Service, U.S. Dept. of Agriculture,
Washington, DC.
Sparks, D.L. 1989. Kinetics of Soil Chemical Processes. Academic Press, San Diego.
Sposito, G. 1989. The Chemistry of Soils. Oxford University Press, New York.
Sposito, G. and S.V. Mattigod. 1979. GEOCHEM: a computer program for the calculation of
chemical equilibrium in soil solutions and other natural water systems. Kearney Foundation
of Soil Science, University of California, Riverside, CA.
Stephens, C.G. 1947. Functional synthesis in pedogenesis. Trans. R. Soc. Austr., 71: 168-181.
Stumm, W., Furrer, G., Wieland, E. and Zinder, B., 1985. The effects of complex-forming li-
gands on the dissolution of oxides and aluminosilicates. In: J.I. Drever (Editor), The Chem-
istry of Weathering. Reidel, Dordrecht, pp. 55-74.
Utermann, J., Kladivko, E.J. and Jury, W.A., 1990. Evaluating pesticide migration in tile-drained
soils with a transfer function model. J. Environ. Qual., 19: 707-714.
Van Grinsven, J.J.M. 1988. Impact of acid atmospheric deposition on soils: quantification of
chemical and physical processes. Ph.D. thesis, Agricultural University Wageningen.
Van Wambeke, A., Hastings, P., and Tolomeo, M., 1986. Newhall Simulation Model; a BASIC
program for the IBM PC. Cornell University Press, Ithaca, NY.
Volobuyev, V.R. 1984. Two key solutions of the energetics of soil formation. In: Genesis and
Geography of Soils. 1985. Scripta Publishing Co. Translated from: Pochvovedeniye, 1984,
No.7:5-11.
Wagenet, R.J. 1983. Principles of salt movement in soils. In: D.W. Nelson (Editor), Chemical
Mobility and Reactivity in Soil Systems. Soil Sci.Soc. Am. Special Pubt. No. 11, 123-140.
ASA, WI.
Wagenet, R.J. 1986. Water and solute flux. In: A. Klute (Editor), Methods of Soil Analysis,
Part I. Agronomy Monograph no. 9, ASA, WI.
Wagenet, R.J. and Hutson, J.L., 1987. Leaching Estimation and Chemistry Model. Dept. of
Agronomy, CorneU University, Ithaca, NY.
Warfvinge, P. and H. Sverdrup. 1989. Modeling limestone dissolution in Soils. Soil Sci. Soc.
Am. J., 53: 44-51.
Westall, J.C., Zachary, J.L., and Morel, F.M.M., 1976.MINEQL, A computer program for the
calculation of chemical equilibrium composition of aqueous systems. Mass. Inst. Technol.
Dept. Civil Eng., Tech. Note 18. Cambridge, MA.
White, R.E., Dyson, J.S., Haigh, R.A., Jury, W.A. and Sposito. G., 1986. A Transfer Function
Model of Solute Transport Through Soil 2. Illustative applications. Water Resources Res.,
22 (2): 248-254.
Wilding, L.P. 1985. Spatial variability: its documentation, accommodation and implication to
soil surveys. In: D.R. Nielsen and Bouma, J. (Editors), Soil Spatial Variability. Pudoc,
Wageningen.
Wtisten, J.H.M., Schuren, C.H.J.E., Bouma, J. and Stein, A., 1990.Functional sensitivity anal-
ysis of four methods to generate soil hydraulic functions. Soil Sci. Soc. Am. J., 54: 832-836.
Yaalon, D.H., 1975. Conceptual models in pedogenesis. Can soil-forming factors be solved?
Geoderma, 14:189-205.

You might also like