Landscape Evolution

Embed Size (px)

Citation preview

  • 8/13/2019 Landscape Evolution

    1/28

    Landscape evolution models

    Frank J. Pazzaglia

    Department of Earth and Environmental Sciences, Lehigh University, 31 Williams,

    Bethlehem, PA 18015, USA; [email protected]

    Introduction

    Geomorphology is thestudy of Earths landforms andthe pro-

    cesses that shape them. From its beginnings, geomorphology

    has embraced a historical approach to understanding land-

    forms; the concept of evolving landforms is firmly ingrained

    in geomorphic thought. Modern process geomorphologists

    use the term landscape evolution to describe the interactions

    between form and process that are played out as measurable

    changes in landscapes over geologic as well as human

    time scales. More traditionally defined landscape evolution

    describes exclusively time-dependent changes from rugged

    youthful topography, through the rounded hillslopes of ma-turity, to death as a flat plain. Bridging the considerable gulf

    in these different views of landscape evolution is the task of a

    different paper altogether. Rather, this chapter provides some

    historical perspectives on landscape evolution, identifies the

    key qualitative studies that have moved the science of large-

    scale geomorphology forward, explores some of the new nu-

    meric models that simulate real landscapesand real processes,

    and provides a glimpse of future landscape evolution studies.

    In 1965, the year of the last INQUA meeting held in the

    United States, thoughts on landscape evolution were still

    dominated by the classic, philosophy-based arguments of

    William Morris Davis, Lester King, and John Hack. Geomor-

    phology, on the other hand, had become more quantitative

    and the number of process studies was growing rapidly(Leopold et al., 1964). Systems theory was on still on the

    horizon, but a landmark paper bySchumm & Licthy (1965)

    had laid the framework for scaling the coming generation

    of process and physical modeling studies into the landscape

    context. In 1965, landscape evolution models were becoming

    stale; process studies were viewed as more scientifically

    rigorous and more directly applicable to human dimension

    problems. At the same time, plate tectonics was emerging and

    thought at the orogen scale enjoyed growth and acceptance

    in the structure-tectonics community. Decades later that

    community would begin questioning basic characteristics of

    active orogens such as: what limits mean elevation or mean

    relief of a landscape or what controls the rate that an orogen

    erodes? Orogen-scale geomorphology became relevantagain to geologic and tectonic questions about the uplift and

    erosion of mountains. Particularly in the past decade, interests

    shared by the structure-tectonics community and the geomor-

    phic community have inspired new thinking at the orogen

    scale and a new generation of landscape evolution models.

    Landscape evolution models come in two basic flavors,

    qualitative and quantitative, that can be applied across a

    wide range of spatial and temporal scales. This chapter will

    primarily consider models that address large-scale landforms

    andprocesses over the graded andcyclic scales ofSchumm &

    Licthy (1965)(spatial dimensions equivalent to an orogen or

    physiographic province, temporal dimensions equivalent to

    104 to106 years). This scale of observation is usefulbecause it

    encompasses investigations common to physical geography,

    process geomorphology, paleoclimatology, and geodynam-

    ics. Qualitative models are well known to most students of

    geomorphology and they form the basis for the more quanti-

    tative approaches.Louis Agassiz (1840)is best figured as the

    grandfather of all landscape models. Agassizs approach to

    understanding the impact of glaciation on landscapes forcedhim to think in terms of form and process as well as irre-

    versible changes in the overall configuration of landforms as

    a function of time. Ironically, Agassizs integrated approach

    diverged with the next generation of geomorphologists and

    landscape evolution models. By the late19th century William

    Morris Davis had published his two seminal papers on the ge-

    ographic cycle (Davis, 1889, 1899a) and Grove Karl Gilbert

    laid the foundation for process-oriented approaches with his

    influential chapter on land sculpture in his monograph on the

    Henry Mountains (Gilbert, 1877). The middle part of the 20th

    century saw the blossoming of physical analogue models

    and more aggressive pursuit of process-oriented studies,

    particularly with respect to hillslope hydrology (e.g.Horton,

    1945) and fluvial geomorphology (e.g.Leopoldet al., 1964).Growth of interdisciplinary studies in the latter part of

    the 20th century allowed the qualitative and quantitative

    approaches to begin to find common ground, spurring a

    proliferation of numeric approaches (e.g. Willgoose et al.,

    1991) and ultimately, the coupled geodynamic-surface

    process model (e.g.Beaumontet al., 1992; Koons, 1989).

    This chapter begins by defining terms and suggesting a

    taxonomy for types of landscape evolution models. It then

    discusses the classic qualitative paradigms of landscape

    evolution as a basis for explaining where the science is

    today. It follows with an exploration of physical models

    where the physical bases for geomorphic processes were

    first explored. Finally, it summarizes four different types of

    numeric landscape evolution models.

    A Taxonomy of Landscape Evolution Models

    Any organization of the various types of landscape evolution

    models necessarily begins with a definition of some terms. A

    landform is a feature of topography that exerts an influence

    DEVELOPMENT IN QUATERNARY SCIENCEVOLUME 1 ISSN 1571-0866DOI:10.1016/S1571-0866(03)01012-1

    247

    2003 ELSEVIER B.V.ALL RIGHTS RESERVED

  • 8/13/2019 Landscape Evolution

    2/28

    248 Frank J. Pazzaglia

    on, and is in turn shaped by surficial processes. A hillslope, a

    river valley, a sand dune, anda colluvialhollow arelandforms.

    A landscape is an aggregate of landforms for a region. The

    evolution of a landscape here is not restricted to long-term

    changes in topographic metrics, but rather is more generally

    applied to describe any change of form-process interactions of

    constituent landforms. This chapter organizes landscape evo-

    lution models into three broad categories: qualitative, physi-

    cal, and surficial process models.

    Qualitative landscape models tend to describe the long-

    term changes in the size, shape, and relief of landforms over

    continental or subcontinental regions. They are not rooted

    in the principles of physics and they are strongly colored

    by the geography of their origin. Nevertheless, the quali-

    tative models that endure are based on good, reproducible

    field observations that must represent a common suite of

    geodynamic and surficial processes that shape topography.

    Close inspection of the several main qualitative models

    reveals that they have much more in common then their often

    incorrectly celebrated differences. Qualitative models tend

    to be heavily skewed towards the influence of numerical

    age on the resulting landforms. For cyclic time and space

    scales in a decaying orogen setting, the age dependenceis warranted. But for landscape evolution at the scale of

    individual watersheds, the resulting landforms many depend

    less on numerical age than on the time scale of response and

    adjustment to driving and resisting forces (e.g. Bull, 1991).

    Physical models are scale representations of landforms. A

    flume is a good example of a physical model used extensively

    to understand channel form and process and there have been

    remarkable accomplishments in the understanding of natural

    channels based on flume studies (Schumm et al., 1987).

    Well-known criticisms of physical models, based on the fact

    that it is difficult to correctly scale for material properties

    (Paola et al., 2001; Schumm et al., 1987), should not be

    viewed as dismissing the potential insights from physical

    experiments. For example, it would be easy to build a modelriver channel in a flume with scaled down grain size and den-

    sity (coal dust) and a fluid other than water with a similarly

    scaled down viscosity (acetone). The problem lies in the fact

    that the resulting acetone-like fluid does not erode the coal

    dust substrate by any process resembling what happens in

    a real channel, not to mention that acetone volatilizes and

    coal dust has high electrostatic charges. By retaining sand

    and water for the flume, what is lost in the scaling is more

    than made up for by retaining similar processes of grain

    entrainment and erosion observed in natural channels.

    Surface process models are rooted in the principles of

    physics and chemistry. Good surface process models are

    both inspired by and verified by field observations. Surficial

    process models can focus on a specific landform, such asa hillslope or river channel, they can explore the linkages

    between landforms, and/or they can consider the feedbacks

    between surface and tectonic processes. A example that

    focuses on a single landform is surface process modeling

    of the evolution of hillslopes in a humid environment. Field

    observations reveal these hillslopes to be concavo-convex in

    cross-section. Their regoliths originate near the slope crest

    and creep downslope. Evolution of the hillslope profile, if

    not the regolith itself, is elegantly modeled by diffusion

    where the diffusivity is a function of the regolith material

    properties. As outlined in detail below, the rate of diffusion

    is proportional to the slope, so form and process mutually

    adjust. Numeric surface process modeling opens the door for

    exploring the linkages between landforms and/or surficial

    and tectonic processes. Numeric models are constructed by

    determining the appropriate mathematical proxy for all, or at

    least the major processes acting on a landscape. Typically, the

    processes are landform specific such that one mathematical

    equation describes the dominant hillslope process, another

    describes regolith production, and another describes fluvial

    transport. The surface process model reconstructs long-term

    changes in a landscape by integrating simultaneously across

    these different process using a large number of time steps in

    a computer program. The simplest kinds of numeric models

    deal with mass balances and cannot truly reconstruct the

    complexity of a real landscape but rather attempt to track

    some metric of that landscape, such as mean elevation. More

    complex models are typically 1-, 2-, or 3-D finite difference

    or finite element code that actually attempt to build and

    shape a synthetic landscape. Numeric surface process modelshave been successfully linked with geodynamic models.

    Such linked models are used to understand the feedbacks

    between surficial and tectonic processes in active orogens.

    Qualitative Models: Classic Paradigms of Long Term

    Landscape Evolution

    It has been said that the study of landscape evolution is first

    and foremost a study of hillslope form and process (Carson

    & Kirkby, 1972; Hooke, 2000). This is not to say that rivers

    do not play an important role in the shaping of landscapes or

    in limiting the overall rate of landscape evolution (Howard

    et al., 1994; Whipple & Tucker, 1999). But across widelyvariable climates, rock types, and rates of rock uplift, rivers

    tend to respond relatively quickly to driving forces and their

    forms lack a memory of the changes in driving forces. Hill-

    slopes respond more slowly on average and as a result, retain

    a richer memory of the changes in process and driving forces

    expressed in their forms, especially in landscapes underlain

    by rock types of different erodibility.

    Observations from a large number of field studies

    conclude that most hillslopes generally have rounded convex

    summitsand areseparated from streamvalleys at their base by

    shallow concave reaches. The degree of convexity or concav-

    ity, the linear separation between summits and the base, and

    average slope angle constitute the differences in overall form.

    Both the upslope convexity and downslope concavity are theresult of transport-limited processes. The former results from

    weathering, creep, and rainsplash (Gilbert, 1909; Schumm,

    1956) and the latter from hillslope retreat, not necessarily

    at the same angle, by surface wash and solution (Schumm,

    1956; Schumm & Chorley, 1964), or from deposition. The

    remaining slope profile between the upper convexity and

    lower concavity is called the main slope and its form is

  • 8/13/2019 Landscape Evolution

    3/28

    Landscape Evolution Models 249

    primarily a result of weathering limited processes such as

    mass movement and surface wash. The main slope can either

    retreat parallel to itself, decline by laying back about a fixed

    hinge point at its base, or shorten by having the upper convex-

    ity extend downward to join the lower concavity. These three

    main behaviors vary with rock type, layering of rock types of

    variable resistance, and climate. The behaviors are common

    to the important qualitative models of long-term landscape

    evolution.

    Base Level, Erosion, and Landscape Genesis Powell

    (1875)

    John Wesley Powell, a Civil War hero of the Battle of Shiloh

    in which he lost an arm, is the true father of the genetic prin-

    ciples of landscape evolution, many of which are incorrectly

    attributed to William Morris Davis. Powells views on land-

    scape evolution are vividly described in the accounts of his

    expeditions through the river gorges of the Rocky Mountains

    and Colorado Plateau (Powell, 1875). The concepts of base

    level and widespread erosion of great mountain ranges to low

    elevation and relief are the cornerstones of Powells work.These ideas were a natural consequence of the features that

    Powell sawduring his river trips: great gorgescarved by rivers

    attempting to lower their gradients, great torrents of sediment-

    laden water resulting from material washing off steep hill-

    slopes, and perhaps most influential, the Great Unconformity

    of the Grand Canyon. More than any other feature, the Great

    Unconformity forced Powell to appreciate the enormity of

    erosion that must have occurred across a once lofty mountain

    range to produce the strikingly horizontal boundary between

    deformed Precambrian and flat-lying Paleozoic rocks. That

    unconformity marked the beveling of a former world and it

    is this point that most influenced the peneplain concept of

    William Morris Davis. River gorges carved through variable

    rock types led Powell to consider a genetic classification ofstreams, hinging primarily on their broad tectonic setting. The

    concepts of antecedent, consequent, subsequent, and super-

    imposed drainages have their origin in Powells writings, but

    weregreatly popularized andfurther defined by Gilbert (1877)

    andDavis (1889).

    Form and Process Gilbert (1877, 1909)

    The work of Grove Karl Gilbert has had perhaps the most

    lasting impact on modern geomorphology (Yochelson, 1980).

    The core of Gilberts work on landscape evolution was ele-

    gantly laid out in his monograph on the geology of the Henry

    Mountains, Utah (Gilbert, 1877). Chapter V in that mono-graph is entitled Land Sculpture and it is here that Gilbert

    begins the science of modern process geomorphology. The

    overall message in Chapter V is that there is an interaction

    between form and process, or in Gilberts words, an equality

    of action. In this view, the landforms of the Henry Mountains

    reflect their underlying rock type, the tectonic processes that

    uplifted, emplaced (in the case of laccoliths) and deformed

    Fig. 1. Sketch, modified from Gilbert (1877), of monoclinal

    shifting. Gilbert noted that when channels traverse dipping

    sedimentary rocks of variable resistance (a), their courses

    adjust to maximize interaction with soft rocks and minimize

    interaction with hard rocks (b).

    them, and the predominant surficial processes of weathering

    and transport. This represents a wholly modern view of pro-

    cess geomorphology andit is ironic that it lay largely dormant

    forseveral decades following Gilberts monograph, only to be

    rediscovered a half a century ago (Hack, 1960).

    Gilbert built his ideas on the foundation laid by John

    Wesley Powell (Powell, 1875) and on pioneering work of

    European engineers, particularly Du Buat (1786), Dausse

    (1857), Beardmore (1851) and Taylor (1851). Gilberts

    genius was applying the results of these earlier studies to

    the long periods of time represented by landscape evolu-

    tion. Particularly noteworthy are Gilberts ideas about theprocesses and rates of erosion, land sculpture (landform

    evolution), and drainage adjustment (Fig. 1). For Gilbert,

    erosion encompasses mechanical and chemical weathering

    in soils as well as during transport. Erosion rates are most

    rapid where slopes are steep a concept not fully popularized

    untilAhnert (1970). Because such erosion would produce a

    landscape with a single slope angle, it is clear that rock type

    must play a primary role also in determining erosion rates.

    Gilbert cast the role of climate in controlling erosion rates in

    the context of competing interests of effective precipitation

    and vegetation with semi-arid landscapes having the highest

    rates of erosion an idea not fully appreciated untilLangbein

    & Schumm (1958). Fluvial channels erode in proportion

    to their slope-discharge product, and sediment transportin a channel depends on the shear stress on the bed. Most

    impressively, Gilbert was the first to describe the concave-up

    profile of rivers as graded, improving upon the original

    definition by European engineers by attributing the graded

    condition to a balance between available discharge and

    capacity for adjacent reaches. Gilbert described drainage

    divides in the context of opposing and interacting graded

  • 8/13/2019 Landscape Evolution

    4/28

    250 Frank J. Pazzaglia

    Fig. 2. Sketch of a hillslope profile fromGilbert (1909)show-

    ing the zone of creep (light shading) andtwo hillslope profiles

    (solid lines) separated by a period of erosion. Gilbert rea-

    soned that the prism of material between 1 and 2 had to be

    carried past 2 during the period of erosion in the same way

    that the prism between 2 and 3 was carried past 3 over the

    same period of time. The quantity passing 2 and 3 is propor-

    tional to the distance to the summit (point 1) if the creeping

    layer is of uniform thickness and velocity of the creeping ma-

    terial is proportional to slope.

    profiles. Divides are strictly fluvial features in this description

    and thus should migrate in the direction of their more gentle

    side. In this respect, Gilbert was perplexed as to why the law

    of divides seem to be violated in certain badlands where

    opposing slopes were not concave up, but rather distinctly

    convex. He surmised that it has something to do with a

    transition from hillslope processes at or very near the divideto fluvial processes further downslope. This hunch proved to

    be the correct answer which is ultimately laid out in a later

    paper (Gilbert, 1909) that describes modern understanding

    of transport and weathering limited hillslopes, creep as a

    hillslope process, and the basis for hillslope transport pro-

    portional to slope (Fig. 2). These contributions later form the

    basis for Hortons zone of no fluvial erosion (Horton, 1945).

    Gilberts findings were only slowly appreciated in terms

    of long-term changes in the landscape. Part of the problem is

    that Gilbert did not philosophically cast his observations into

    a temporal framework. He focused on current interactions

    between form and process with little regard for how those in-

    teractions would play out over long periods of time to change

    a landscape from its current form, to something different.Ironically, understanding the interactions between form and

    process and casting them into mathematical equations is the

    basis for numeric surface process modeling that has proven

    to be a powerful tool for predicting long-term landscape

    evolution.

    Fig. 3. The geographic cycle of Davis

    showing (a)a simplifiedversionof theorig-

    inal figure from the 1899 paper and (b)

    a modified interpretation of the original

    figure (from Summerfield, 1991). Stages

    1 through 4 in (a) refer to uplift, valley

    bottom deepening, valley bottom widening,and finally interfluve lowering. Upper line

    denotes mean elevation of interfluves and

    lower line, mean elevation of valley bot-

    toms.

    The Geographic Cycle Davis (1889, 1899a, 1932)

    The most influential paradigm on long-term landscape evolu-

    tion comes from thework of theAmerica physical geographer

    William Morris Davis. Much has been written regarding the

    obvious differences between the Davisian approach with re-

    spect to Gilberts process approach. There is a place for both

    approaches in modern views of longterm landscapeevolution.

    The Davisian approach provides the temporal framework and

    the recognition that all landscapes are palimpsests, overwrit-

    ten by numerous tectonic and climatic processes (Bloom,

    2002). A Pennsylvanian by birth, Davis summered with his

    family in the Pennsylvania Ridge and Valley, a landscape that

    would forever color his geomorphology views. Davis fully

    appreciated the best geologic interpretations of his day and

    used these to constrain his models of landscape evolution.

    He was also careful to point out that his model was idealized.

    He never intended concepts like an instantaneous, impulsive

    uplift, or completion of the geographic cycle to be attained in

    every case or to be taken and applied literally (Davis, 1899b).

    Unfortunately, these andother idealizations of his model have

    been taken literally by many outside of the field of geomor-

    phology, leading to muchconfusionand stifling reconciliationof his ideas with modern process-oriented thought.

    Among Daviss numerous publications, three define

    and apply the concept of the geographic cycle to a range

    of landscapes. The first paper is The rivers and valleys of

    Pennsylvania (Davis, 1889) which focuses on the develop-

    ment and evolution of the current drainage of Pennsylvania.

    Basic observations of accordant summits, wind gaps, and

    water gaps routing the master streams transverse to structure

    in the Ridge and Valley anchor the justifications for repeated

    landscape uplift and beveling. The second paper is The

    geographic cycle (Davis, 1899a) where the Davisian model

    for long term landscape evolution is described. The third

    paper is Piedmont benchlands and Primarrumpfe (Davis,

    1932) where Davis confronts a major alternative paradigmto the geographic cycle and the formation of peneplains

    presented by the ideas ofWalter Penck (1924).

    The geographic cycle (Fig. 3) is a simple, but compelling

    treatment of how mean elevation and mean relief change as a

    landscape erodes. It also places a large emphasis on transport

  • 8/13/2019 Landscape Evolution

    5/28

    Landscape Evolution Models 251

    limited processes for the overall erosion and lowering of hill-

    slopes. The geographic cycle has four major stages. In stage

    1, a landscape is born by rapid, if not impulsive uplift of both

    rocks and the land surface above sea level. The uplifted land-

    scape has a high mean elevation, but low mean relief. Erosion

    is initially concentrated in the river which leads to the carving

    of deep, narrow valleys. The landscape passes into stage 2

    where it is youthful in appearance, it has both a relatively

    high mean elevation and high relief. Valleys rapidly lower to

    base level, at which point they begin to widen laterally. At the

    same time, hillslope erosion becomes important as interfluves

    are rounded. Hillslopes take on a concavo-convex form and

    weathered material is moved across this form to the river

    valley predominantly by creep. These conditions describe

    stage 3, the mature landscape. Both mean elevation and mean

    relief steadily decrease, and valleys continue to widen. Stage

    4 is reached when valleys can no longer deepen or widen and

    all landscape lowering is accomplished by the progressive

    rounding and lowering of interfluves by hillslope creep. The

    completely beveled landscape resulting from the completion

    of the geographic cycle, the penultimate plain, is called a

    peneplain. Davis envisioned the peneplain being formed at

    different rates depending on the size of the landscape beingbeveled and the relative erodibility of the rocks. Resistant

    islands of high standing topography are called monadnocks

    (for Mt. Monadnock in southern New Hampshire), and

    evidence of partial peneplains is common on softer rocks.

    Davis fully envisioned the difficulty in running the cycle

    to completion. He inferred that renewed tectonic uplift of a

    landscape occurs often enough to interrupt the cycle. This

    realization led to the preoccupation with finding peneplains

    and partial peneplains in the landscape, essentially an attempt

    to define an landscapes erosional stratigraphy. Time was the

    critical factor in determining a landscapes mean elevation

    and relief; no interactions between form and process or

    driving and resisting forces are considered. The concept of

    the peneplain proved the most controversial part of the geo-graphic cycle (Davis, 1899b). Ironically, Davis never claimed

    ownership of the term. In his view, he was just giving a name

    to a process and a feature described by geologists before him,

    most notable, Powells descriptionof the Great Unconformity

    in hisExploration of the Colorado River(Powell, 1875).

    The Geographic Cycle makes little reference to hillslope

    processes and what references do exist are considered

    together under the concept of agencies of removal. Here

    Davis suggests that the hillslopes are shaped by a myriad

    of processes that include surface wash, ground water,

    temperature change, freeze-thaw, chemical weathering, root

    and animal bioturbation. These collectively drive creeping

    regolith downslope. Relative amounts of weathered products

    and exposed bedrock are considered in the context of whetherthe slope is steep and youthful, or old and rounded.

    An important idea that has survived from the geographic

    cycle is that slopes in tectonically active landscapes rapidly

    attain steep, almost straight profiles, then more slowly decay

    into concavo-convex, sigmoid-shaped profiles as tectonism

    wanes, valleys reach their base level of erosion, and divides

    are lowered. Davis also suggested several modifications to

    the geographic cycle model. Most notable are the ideas about

    cyclic erosion in arid landscapes (Davis, 1930). In this paper,

    Davis introduces elements of slope retreat into his model.

    From this model Bryan (1940) and later King (1953) describe

    and expand upon the role that retreating escarpments play in

    long-term landscape evolution. Unfortunately, Davis appears

    to abandon his view of slope form and process being a

    consequence of climate in his 1932 paper where he returns

    only to the ideas originally espoused in the Geographic Cycle

    as a defense against the ideas of Walter Penck (see below).

    The triumph of the geographic cycle is that it was

    almost a century ahead of its time in showing how mean

    elevation and mean relief might evolve in a landscape. Only

    recently, with the realization that surficial and geodynamic

    processes arelinked in orogenesis (Molnar& England, 1990),

    has the broader geologic community come to appreciate the

    lag times and feedbacks between rock uplift and erosion. The

    four stagesof thegeographic cycleresemble modernconcepts

    about growing, steady-state, and decaying orogens (Hovius,

    2000). Furthermore, the erosional response to impulsive up-

    lift described by the geographic cycle is completelyconsistent

    with the observed flux of material from disturbed Earth sys-

    tems across a wide range of space and time scales (Schumm& Rea, 1995). This flux follows a distinct exponential decay

    after an impulsive increase. Over cyclic time scales, uplift

    and the initial erosional response are in fact impulsive, and

    the landscape response in terms of lowering mean elevation

    and mean relief is most simply described by an exponential

    decay function. The tragedy of the geographic cycle is that

    individual parts of it have been accepted too literally. A prime

    example is equating meandering streams with Daviss wide,

    low gradient valleys of mature landscapes. Meandering chan-

    nels do not necessairly indicate landscape maturity. Rather,

    meandering reflects the interaction of driving and resisting

    forces such as watershed hydrology, prevailing grain size,

    bank stability, and rock type erodibility in a fluvial system.

    Slope Replacement Penck (1924, 1953)

    The most influential European geomorphologist in the

    early part of the 20th century and a direct challenger to the

    Davisian model was Walter Penck. Penck was influenced by

    field observations made in Germany, particularly the Rhine

    graben area, in northern Argentina, and to a lesser degree,

    in Africa. Like Davis, Penck envisioned long term landscape

    evolution occurring in stages from youth to old age, when

    all relief in a landscape was beveled by erosion. And despite

    the assertions that Penck was the original proponent of slope

    retreat (King, 1953), he like Davis, believed in the flattening

    of slopes through time. Misunderstandings arose from hisobscure writing style (Simons, 1962), and from Daviss

    incorrect translation of some of his work (Davis, 1932). But

    unlike Davis, Penck was more process oriented, actually

    making measurements in the field, and he focused his process

    studies on hillslopes.

    Penck recognized concavo-convex hillslopes similar

    to what Davis was observing in eastern North America.

  • 8/13/2019 Landscape Evolution

    6/28

    252 Frank J. Pazzaglia

    Fig. 4. The Penck model. (a) Diagram showing slopereplacement(upper profile)caused by successiveretreat of an infinite number

    of inclined valley bottoms leading to a smooth concave profile (lower concave profile). Penck envisioned an initially steep linear

    cliff (t0), being replaced by progressively more gentle slope segments (t1 t3). (b) Pencks original figure, misinterpreted by

    Davis (1932), illustrating Pencks cycle of erosion. Much like Daviss model, this diagram shows that Penck considered erosion

    to be concentrated in river valleys first (t0), then extended to hillslopes (t1 t2). If adjacent, ever-widening concave valley

    profiles meet at an interfluve (t3), that interfluve becomes rounded and lowered (modified from Carson & Kirkby, 1972).

    However, the Penckian model for their origin was altogether

    different. Penck is the father of slope replacement as a

    mechanism for hillslopes evolution (Fig. 4). Whereas Daviss

    hillslopes are transport limited and always covered with a

    creeping regolith mantle, Pencks hillslopes are predomi-

    nantly weathering limited with little to no regolith cover,

    except, implicitly at their base where regolith transported

    downslope is allowed to accumulate. Originally steep and

    straight slope profiles weather parallel to themselves except

    for a small, step-like flattening at the base of the slope, the

    haldenhang, which is presumably controlled by the angle

    of repose of the hillslope debris. Retreat of the haldenhang

    results in an even lower-gradient basal slope called theabflachungshang. The process continues with successive

    abflachungshang retreating, each leaving a basal slope of

    lower angle than itself. The integrated result is a concave-up

    slope that has replaced the original steep, straight slope.

    Penck went on to propose mechanisms of how the upper

    part of the concave profile would flatten to produce an upper

    slope convexity using arguments similar to those used by

    Gilbert (1909).

    The concept of waxing and waning slopes is not as much

    a part of Pencks original model as it is Daviss interpretation

    of it. But Pencks views on the effects of base level change

    on slope form are notable. Penck argued that given a stable

    base level, a hillslope will attain a characteristic profile deter-

    mined by the erodibility of underlying rock. Steep slopes areunderlain by hard rocks whereas gentle slopes are underlain

    by soft rocks. Furthermore, Penck was influenced by his

    studies in the Alpine foreland, where it appeared certain

    that orogeny was protracted and occurred at variable rates,

    rather than being impulsive as in the Davisian model. These

    observations led Penck to believe that base level fall, if linked

    to orogeny, starts slow, accelerates to a maximum (waxing),

    and decelerates back to stability (waning). Hillslopes adjust

    to that base level fall throughout the waxing and waning

    cycle. A fall in base level causes that segment of the hillslope

    immediately adjacent to the base level fall to steepen, and that

    steepened hillslope segment then causes the next segment

    upslope to steepen and so on. The result is the replacement of

    a gentle slope with a more steep one as the base level fall sig-

    nal is translated upslope. If base level fall is a one-time event,

    the steep hillslope segment propagates upslope, and in turn is

    replaced by a less gentle segment downstream. But if the base

    level fall continues or accelerates, there is a downslope trend

    of progressively steeper slope segments, such that the hills-

    lope becomes convexin profile. Alternatively under base levelrise or long-term stability, there is a downslope shallowing of

    slope segment gradientsand thehillslope is concave in profile.

    The Penckian model predicts convex hillslopes when base

    level is actively falling (waxing slopes) as part of a youthful

    landscape; the Davisian model predicts concavo-convex

    hillslopes where base level is stable and a mature landscape is

    slowly being beveled. In summary, Penck viewed base level

    changes (uplift) as long-lived, rather than impulsive, and the

    overall landscape response to those changes as having short

    lag times.

    Even though Pencks ideas on landscape evolution dif-

    fered significantly from the Geographic Cycle, his landforms

    are still time-dependent features (Fig. 4). The Penck model

    holds that landscapes are born from a base level fall thataffects a low relief landscape called a primarrumpf, resulting

    in stream incision and convex hillslopes. Acceleration in

    the rate of base level fall increases hillslope convexity and

    results in the formation of benches (piedmottreppen) near

    drainage headwaters that have not been affect by the slope

    replacement process. Eventually, base level fall decelerates

    and stops completely. Slopes are replaced to lower and lower

  • 8/13/2019 Landscape Evolution

    7/28

    Landscape Evolution Models 253

    Fig. 5. The pediplanation model ofKing (1953)(modified fromSummerfield, 1991). Pencks model stresses the development of

    concave-up slopes that retreat faster than interfluves lower, resulting in widespread pediments that coalesce into a pediplain.

    Note that mean elevation decreases in this model, but relief persists along escarpments or inselbergs.

    declivities until the relatively straight profiles of base level

    stability retreat headward leaving a beveled plain called

    an endrumpf with large, residual inselbergs at the drainage

    divides.

    Pediplanation L.C. King (1953, 1967)

    Drawing upon work conductedprimarilyin Africa, but versed

    on landscapes worldwide, Lester King championed a modelfor long term landscape evolution very similar to the geo-

    graphic cycle, but differing in the dominant mode of hillslope

    evolution (King, 1953, 1967). Like Davis, King envisioned

    impulsive uplift and long response times of landscape ad-

    justment. King never accepted the Davisian concavo-convex

    slope; instead, he favored Pencks view of concave hillslopes

    and slope replacement. In fact, King took the Penckian model

    of slope replacement literally, to conclude that the landscape

    assumes the form of a series of nested, retreating escarpments

    (Fig. 5). King called the low-gradient footslope extensions of

    the steep escarpments pediments and the flat beveled surface

    they leave in their retreating wake a pediplane. Such a re-

    stricted interpretation of the Penckian model is unwarranted

    because Penck made it clear that the rates of slope modifi-cation are proportional to gradient, a relationship that would

    continually produce slope replacement, not parallel retreat. In

    fact, Kings model is based more on the earlier work ofKirk

    Bryan (1922, 1940)than it is on Penck.

    The notionof a landscapedominated by pediments should

    not come as a surprise to a geomorphologist influenced by

    South African landscapes. But King was wholly convinced

    that these landforms are not restricted to just the arid and

    semi-arid climates of his homeland. Rather, he argued, Davis

    himself had described precisely the same features in New

    England where Mt. Monadnock, the supposed type locality of

    a remnant, rounded, concavo-convex hillslope has a concave-

    up pediment profile. To Kings credit he built upon the earlier

    work of Penck in making measurements in the field andincorporating the process work on pediment formation from

    other noted geomorphologists includingKirk Bryan (1922).

    King proposed that once pediments form, that they persist

    indefinitely until consumed by younger retreating escarp-

    ments following renewed base level fall. The base level fall

    in Kings model is inherently episodic because it occurs on

    passive margins, which at the time, were thought to respond

    isostatically to episodes of erosional unloading and offshore

    deposition (Schumm, 1963) a concept viewed today as

    untenable.

    King summarizes his pediplane model as well as his over-

    all views on landscape evolution in fifty canons, published in

    the1953 paper. Of these canons, thefollowing areparticularly

    insightful, and research continues on many of them: (No. 1)

    Landscape is a function of process, stage, and structure;

    the relative importance of these is indicated by their order.

    (No. 3) There is a general homology between all (fluviallysculpted) landscapes. The differences between landforms of

    humid-temperate, semiarid, and arid environments are dif-

    ferences only of degree. Thus, for instance, monadnocks and

    inselbergs are homologous. (No. 6) The most active elements

    of hillslope evolution are the free face and the debris slope. If

    these are actively eroded, the hillslope will retreat parallel to

    itself. (No. 9) When the free face anddebris slope are inactive,

    the waxing slopebecomes strongly developed and may extend

    down to met the waning slope. Such concavo-convex slopes

    are degenerate. (No. 42) Major cyclic erosion scarps retreat

    almost as fast as the knickpoints which travel up the rivers

    transverse to the scarp. Such scarps therefore remain essen-

    tially linear and lack pronounced re-entrants where they cross

    the rivers.

    Dynamic Equilibrium Hack (1960)

    The simple, embodying concept of driving and resisting

    forces the interaction between form and process proposed

    by Gilbert (1877) was rediscovered and presented as an

    alternative to the time-dependent paradigms of landscape

    evolution by J.T. Hack of the U.S. Geological Survey ( Hack,

    1960). By his own admission, Hack had been carefully study-

    ing the same Appalachian landscape of Davis and making

    a conscious effort to seek alternatives to cyclic theories of

    landscape evolution. The alternative favored by Hack is that

    landscapes arein a state of dynamic equilibrium (Fig.6). Theyare in equilibrium in the sense that given the same driving

    and resisting forces over a long period of time, a time-

    independent characteristic landscape will emerge. This is a

    landscape where the rivers and hillslopes are all graded and

    the processes acting on the interfluves and channel bottoms,

    although different, are lowering their respective parts of the

    landscape at the same rate. The landscape is dynamic in

  • 8/13/2019 Landscape Evolution

    8/28

    254 Frank J. Pazzaglia

    Fig. 6. Attainment of a near time-invariant relief and mean

    elevation of a dynamic equilibrium landscape (a) attained

    over graded time during a protracted period of decay (cyclic

    time) and (b) attained as a flux steady-state between the input

    of rocks by tectonic processes and output by erosion.

    the sense that climate, tectonics, and rock type change as

    subaerial erosion progresses so there is a constant adjustment

    between the principal driving and resisting forces such that

    true equilibrium is probably only asymptotically reached.

    Strictly speaking, a truely characteristic or steady-state

    landscape cannot exist, except in a tectonically active setting

    where the mass flux in and out of the orogen is conserved

    (Pazzaglia & Knuepfer, 2001; Willett & Brandon, 2002). In

    a decaying orogen, mean elevation and mean relief must be

    reduced, as Davis surmised, because isostatic rebound of

    a low density crustal root only recovers 80% of what is

    removed via erosion.

    Hacks concept of dynamic equilibrium can be usedto describe the origin of the same landscape that Davis

    worked in, but without cycles of erosion and peneplains. The

    Hack model predicts accordant summits among resistant

    rock types because the similar processes have established

    themselves on slopes of similar declivity, and on rocks

    sharing similar structure over a long period of time. Hack did

    not propose an erosional stratigraphy in landscapes. Rather,

    he proposed that all landscapes are essentially modern and a

    reflection of modern processes. Truly flat landscapes are not

    erosional, but rather aggradational. At the time, Hacks paper

    was only the most recent of a long list of cycle of erosion

    critics to point out the conspicuous global lack of peneplains

    at or near sea level.

    Hacks contribution very much refocused the geomor-phologic community back towards process and as such,

    helped to lay the foundation upon which modern analogue

    and surface process models are built. This is not to say that

    process geomorphology was absent before Hack. But his

    concept of dynamic equilibrium moved geomorphology back

    into the wider geologic arena of basic principle chemical,

    and biologic principles that govern landscape evolution.

    Process Linkage Bull (1991)

    Several decades of process studies (e.g. Leopold et al.,

    1964), watershed hydrology (Dunne, 1978; Horton, 1945),

    and the Hack tradition of dynamic equilibrium (Hack, 1960)

    collectively form the basis for understanding landscape

    evolution from the perspective of linked processes. This

    view of landscape evolution, colored by the dramatic

    effects of Quaternary climate change, is best representedin the textbook Climatic Geomorphology (Bull, 1991).

    Many geomorphologists in North America and elsewhere

    contributed to a modern understanding of process linkage.

    But it was Bill Bull, working in the American southwest

    and with colleagues abroad, particularly in Israel, who led

    this effort.

    Process linkage refers to the direct and indirect ways

    in which individual components of a watershed mutually

    respond to an external driving force, the two most obvious

    being tectonics and climate. Implicit in the linkage among

    individual components is the realization that there may not

    be a unique response for a given external stimuli. Such

    non-unique responses arise because thresholds (Patton

    & Schumm, 1975) are embedded in most geomorphicprocesses, and because a large external stimulus is required

    to elicit the same response in multiple watersheds.

    Process linkage has a particular focus on landscape

    change as a function of the creation and routing of sediment

    thorough a watershed. In this respect, it differs from other

    models of landscape evolution that focus on the evolution

    of hillslope forms. Climate change, an important control on

    sediment creation and routing affects watersheds at all spatial

    scales and has been particularly acute during Pleistocene

    glacial-interglacial cycles (Pederson etal., 2000, 2001). Qua-

    ternary stratigraphy, like that of an alluvial fan, commonly

    reveals relatively brief pulses of deposition interspersed

    with longer periods of landscape stability and pedogenesis.

    The Bull model considers weathering, the production rateof regolith, the liberation of that regolith off hillslopes, the

    response time, thresholds, and equilibrium all interacting

    in the watershed as the primary controls on the resulting

    alluvial fan stratigraphy (Fig. 7). Although many factors in-

    cluding rock type, climate, seasonality, relief, and vegetation

    influence the hydrology and the unique response of a given

    watershed, a simple, representative scenario best illustrates

    Bulls model.

    Particularly in arid and semi-arid climates, regolith

    production is thought to be maximized under relatively cool,

    moist climatic conditions when vegetative cover on the hills-

    lope accelerates chemical and physical weathering. Stripping

    of that regolith occurs when a switch to drier, warmer con-

    ditions leads to the loss of the vegetative cover, leaving theregolith prone to gulling and removal by overland flow. Even

    without much loss of vegetation, changes such as an increase

    in precipitation seasonality or the lowering of infiltration

    rates in well-developed, mature soils can eventually provoke

    loss of a hillslopes regolith to increased overland flow. The

    time between the climatic perturbation and the beginning of

    hillslope stripping involves crossing a geomorphic threshold

  • 8/13/2019 Landscape Evolution

    9/28

    Landscape Evolution Models 255

    Fig. 7. Reaction time, relaxation time, and response times of

    theBull (1991)process-response model (modified fromBull,

    1991).

    and is called the reaction time (Fig. 7). The time needed to

    achieve a new equilibrium condition between the new climate

    and hillslope process is called the relaxation time. The sumof the reaction time and relaxation time is the response time

    (Brunsden & Thornes, 1979). Whipple & Tucker (1999)

    describeda similar definitionof theresponsetime as it applies

    to channel headwaters adjusting to changes in the rate of

    rock uplift.

    The transition from the reaction to relaxation times of

    when the hillslopes begin to liberate regolith is affected

    by the overall climatic regime (arid, semi-arid, semi-arid

    and glaciated) of the watershed (Ritter et al., 2000). That

    regolith is delivered to the fluvial system which moves it

    out of the watershed. The hydraulic geometry may change

    as sediment is routed through them, and the streams may

    aggrade or incise depending on the sediment flux from the

    hillslopes. Even so, the residence time of sediment in thefluvial system is typically short compared with the residence

    time of regolith on the hillslopes. The sediment is delivered

    to a basin where its depositional architecture retains clues

    about climatically influenced transport processes as well as

    the weathering processes in the source (Smith, 1994). Some

    watersheds have very short response times between the

    hillslopes, channels, and depositional basin where a direct

    temporal link can be established between the climatic event

    that affected the hillslopes and the depositional response

    far downstream (Fig. 7). In other cases, long response times

    between the hillslopes and depositional basin make it far

    more difficult to link a climate change to the depositional

    response.

    Among the numerous studies that view process linkageas an important agent of landscape evolution, three are

    noteworthy in illustrating the relative influences of rock type

    (Bull & Schick, 1979), pedogenesis (Wells et al., 1987),

    and climate-watershed hydrology (Meyer et al., 1995). The

    Bull & Schick (1979) study is one of the first to show the

    response of two adjacent watersheds in an arid environment

    to the same climate change. The watersheds are underlain

    by different rock types, leading to different regoliths, soil

    infiltration rates, and response times to late Pleistocene

    and Holocene climate change. Some watersheds continue

    to deliver sediment from hillslopes as a response to late

    Pleistocene to Holocene climate change, so that alluvial fans

    at the mouths of these watersheds have a primarily aggrada-

    tional history. In contrast, other watersheds cease liberating

    sediment from their hillslopes and the corresponding alluvial

    fans pass into a period of incision as the sediment supply

    wanes.

    Wells et al. (1987)showed how the landscape evolution

    of an alluvial fan surface, including the location of fan

    deposition, is strongly controlled by the linked factors of

    soil genesis and how those soils control runoff. Alluvial

    fan deposition is initially affected by the late Pleistocene-

    Holocene climate change with the watershed hillslopes

    being the dominant source of sediment delivered to the

    fan. Subsequent Holocene fan deposition reflects a change

    in the source from the hillslopes, where runoff has been

    reduced by renewed colluvial mantle development, to the

    piedmont where runoff has been increased by development of

    progressively impermeable clay and calcic horizons in eolian

    soils.Meyer et al. (1995) described the landscape evolution

    of rugged montane valleys in the Rocky Mountains, where

    liberation of sediment from hillslopes is influenced by large

    slope-clearing fires as well as climate. The fires recur at

    intervals that average approximately 200 years. Relatively dry

    climate over a period of years increases the chances of large,

    destructive, slope clearing fires. Drier climatic conditions

    also correlate with relatively small winter snowpacks and

    intense summer convective storms. The hydrology associated

    with these relatively dry climatic conditions favors debris

    flow activity in burned regions during the summer. Sediment

    accumulates in debris fans along the valley margins and the

    fluvial system has flashy discharges in a braided, incised

    channel with a low base flow. Wet years, in contrast, arecharacterized by a climate with heavy winter snowpacks,

    depressed summer convective storm activity, and decreased

    numbers of fires and debris flows. The result is a meandering

    stream fed by the high, stable base flows and ample sediment

    supply recruited from the debris fans as the channel mean-

    ders across a widening valley floor. The transition between

    relative dry and relatively wet climatic conditions occurs at

    the millennial scale. In their long-term evolution, hillslopes

    and rivers in this setting are clearly linked, but there is a

    millennial-scale response time of the river system following

    the initial reaction time of the hillslope to the climatic

    perturbation.

    Geodynamic and Surficial Process Feedbacks Molnar

    & England (1990)

    A new way of thinking about orogen-scale landscape evolu-

    tion, in terms of surface processes that limit rates of tectonic

    deformation and the uplift of rocks originated withMolnar

    & England (1990)who linked the building of high-standing

  • 8/13/2019 Landscape Evolution

    10/28

    256 Frank J. Pazzaglia

    Fig. 8. Changes in mean elevation and mean

    localreliefin response to erosionthat deepens

    and widens valleys. Rocks are uplifted as an

    isostatic response to this erosion because the

    topography is compensated by a low-density

    crustal root protruding into the mantle.

    Despite rock uplift and surface uplift of the

    summits, the mean elevation of the landscape

    decreases as the crustal root is consumed

    (modified fromBurbank & Anderson, 2001).

    mountains to cool climates. The idea of coupled surficial

    and geodynamic processes is not new, extending back as

    least as far as the original considerations of isostasy and

    erosion (Ransome, 1896). More recently,King (1953)argued

    that nested, inset pediments were the result of epeirogenic

    upheavals caused by the unloading effects of retreating es-

    carpments. Schumm (1963) forwarded a similar argument foruplift in the Appalachians caused by erosion crossing an iso-

    static threshold after which the land surface would rapidly re-

    bound. Thesearguments of episodic uplift(or subsidence) due

    to erosion have largely been shown to be incorrect ( Gilchrist

    & Summerfield, 1991), but the general idea that uplift and

    erosion are linked persists.

    The central idea in coupled geodynamic-surficial pro-

    cesses is best illustrated by the simple Airy isostatic case of

    a mountain range where the topography is supported by a

    low-density crustal root protruding into the mantle (Fig. 8).

    Uniform erosion across the range results in isostatic rebound

    of the root proportional to the density difference between

    the mantle and the crust, but the land surface and mean

    elevation are lowered with respect to sea level. Airy isostaticrebound recovers approximately 80% of the mass removed

    by erosion; a column of rock 5 km thick must be consumed

    to reduce mean elevation by a kilometer. The shape of a

    mountain range responds dramatically to the rebound if the

    erosion rate is not uniform across the range. When erosion

    is concentrated in valleys and interfluves are lowered much

    more slowly than valley bottoms, local relief is increased

    and the resulting isostatic rebound pushes the interfluves

    (mountain peaks) higher even though mean elevation has

    been lowered (Fig. 8). The effect is muted below the flexural

    wavelength of the lithosphere (Small & Anderson, 1998)

    but it still may be an important source of recent surface

    uplift.

    Feedbacks between surficial and geodynamic processesgo beyond simple consideration of surface uplift by isostasy.

    The concept encompasses climatically controlled limits on

    mean elevation (Brozovicet al., 1997), the width of orogens

    (Beaumont et al., 1992), the metamorphic grade of rocks

    exposed in ancient orogens (Hoffman & Grotzinger, 1993),

    the dominant river long profiles and hillslope erosion pro-

    cesses (Hovius, 2000), the structure of convergent mountain

    belts (Willett, 1999), and a geomorphic throttle on orogenic

    plateau evolution (Zeitler et al., 2001). Consideration of

    surficial-geodynamic feedbacks releases geomorphic think-

    ing from viewing landscape evolution at the large scale as

    only reacting, rather than interacting to impulsive climatic

    or tectonic changes. At the scale of the orogen, landscapes

    evolve as a consequence of both tectonics and climate and theevolutionary path itself plays a role in defining the feedbacks

    between the two.

    Physical Models

    Physical models of landscape evolution existed in parallel

    with, but originally did not enjoy the spotlight focused on

    the philosophical approaches. A history of the rise and

    development of physical models is provided by Schummetal. (1987). Daubree (1879) wasamongthe first to recognize

    that experiments can reveal basic relationships and general

    hypotheses to guide field studies. Early experiments focused

    on formation of drainage networks (Hubbard, 1907), theerosional evolution of an exhumed laccolith (Howe, 1901),

    the downslope movement of scree (Davisson, 1888a, b),

    fluvial transport of sediment (Gilbert, 1914, 1917), and

    behavior of channel meanders (Jaggar, 1908). Evolution

    of drainage networks within a given watershed were first

    simulated byGlock (1931)andWurm (1935, 1936).

    There are three broad classes of physical models: (1)

    segments of unscaled reality, (2) scale models, and (3) analog

    models (Chorley, 1967). A segment of unscaled reality does

    not imply that the model is constructed at some scale very

    different than what is found in nature. Rather, it refers to

    a model that can actually be of a natural system, like the

    segment of an alluvial river channel that is considered repre-

    sentative of a broader range of river channels in form, process,and geography. Simple sand bed flumes or stream tables

    also illustrate unscaled reality. There are particularly useful

    applied geomorphology practices where data from segments

    of unscaled reality have led to important policies in land man-

    agement. An example widely used by agricultural engineers

    is the Universal Soil Loss Equation (Wischmeier & Smith,

    1965), which wasderived from both field studies andphysical

  • 8/13/2019 Landscape Evolution

    11/28

    Landscape Evolution Models 257

    models of soil erosion. Commonly, space and time are substi-

    tuted in segments of unscaled realitymodels so that processes

    that unfold over millions of years in large watersheds can

    be investigated over a period of years. For example, Ritter

    & Gardner (1993)linked dependence of infiltration recovery

    with channel development on mined landscapes in a physical

    model that plays out over a period of years to mimic the

    evolution of larger watersheds. In the process, the model

    provides insight as to how climate change affects basin

    hydrology.

    Scale models make an attempt to reproduce natural

    landforms in such a way that ratios of significant dimensions

    and forces are equal to those found in nature. True scale

    models are difficult to construct because of the limited range

    of fluid and substrate material properties that are practically

    available to maintain ratios of dimension and force. Jurassic

    Tank at the University of Minnesota St. Anthonys Falls

    laboratory illustrates a scale model dedicated to the study

    of depositional landforms and their resulting underlying

    stratigraphy (Paola et al., 2001).

    Analogue models are designed to mimic a given natural

    phenomena without having to reproduce the same driving

    forces, processes, or materials. Examples include modelingthe flow of glaciers using oatmeal (Romey, 1982) or mod-

    eling landslides using a flume filled with beans ( Densmoreet al., 1997).

    Physical models typically isolate one part of the geomor-

    phic system, or even one part of a watershed and, as a result,

    rarely show the behavior of an alluvial channel, for example,

    in the context of broader landscape change. Fortunately, vir-

    tually all geomorphic processes produce erosion which phys-

    ical models can reproduce as a general proxy for landscape

    change. Erosion, channel incision, and sediment transport,

    common to many physical models, are considered below.

    Drainage Networks

    Drainage networks have attracted a great deal of attention in

    the overall evolution of a watershed. Given the growing ap-

    preciation for the rate of fluvial incision and characteristic

    drainage density in limiting rates of rock uplift and erosion

    in tectonically active landscapes, the initiation and evolution

    of drainage networks remains a central pursuit in landscape

    evolution models. There are three main hypotheses regard-

    ing drainage network evolution: (1) network growth by over-

    land flow (Horton, 1945); (2) headward growth of established

    channel heads (Howard, 1971; Smart & Moruzzi, 1971); and

    (3) dominance of rapidly elongated channels (Glock, 1931).

    The Glock model incorporates components of both overland

    flow andheadwardretreat. In this model, a trunk channel initi-ates, rapidly elongates, elaborates, and then abstracts, settling

    into a characteristic drainage density consistent with the pre-

    vailing infiltration characteristics of contributing hillslope. In

    fact, it is the evolution of infiltration characteristics on the

    hillslope that determines the speed at which the network pro-

    gresses through the various stages. Predictions of the Glock

    model have been directly observed in field studies (Morisawa,

    1964; Ritter & Gardner, 1993) and have been reproduced and

    refined with other physical models (Parker, 1977).

    Hillslopes

    Physical models are common in studies of hillslope form and

    process. Geological engineering and watershed hydrology

    literature is rich in examples of physical models designed

    to understand landslides, earthflows, and debris flows. The

    U.S. Geological Survey debris flow experimental flume (e.g.

    Denlinger & Iverson, 2001; Iverson & Vallance, 2001) is

    an example of recent approaches that have greatly advanced

    the understanding of Coulomb and granular dispersion

    processes in hyperconcentrated and debris flows. However, it

    is typically more difficult to reproduce scale representatives

    of hillslopes in the laboratory, so many physical models

    are actually carefully controlled field monitoring studies

    designed to isolate one process acting over a limited area.

    The fall of Threatening Rock, a well-documented example

    of slab-failure processes, is one such field study that passes

    for a physical model (Schumm & Chorley, 1964). This study

    revealed that movement of the slab away from the cliffface accelerated exponentially approaching the moment of

    failure. The rates were fastest in the winter months, probably

    because of frost action. Studies of hillslope creep are another

    class of physical models, typically conducted in the field.

    Creep experiments consider a column or known shape of

    regolith or soil periodically measured with respect to fixed

    reference points. The measurements may include markers or

    lines on the surface of the regolith, or horizontal pins inserted

    in the free face of a Young Pit (Young, 1960).

    Weathering and rock disaggradation in the hillslope envi-

    ronment are easily simulated by repeated mass and volume

    measurements of representative rock samples under different

    weathering environments. Rates of disintegration and mass

    loss depend on whether the rock sample remains exposed atthe surface, or if it remains buried in a soil profile, exhumed

    only for the purpose of measurement. Such experiments may

    influence the resulting measured rates, but they do provide

    insights into processes. The classic experiment of this type

    was conducted on sandstone collected from cliff faces in the

    Colorado Plateau (Schumm & Chorley, 1966). The results

    from measurements over a period of two years, showed

    that granular disintegration, rather than fracturing into small

    pieces, was the dominant weathering process. Steep cliff

    faces are probably maintained in the Colorado Plateau as

    the rock disintegrates and is blown away; little accumulates

    against the cliff base.

    There are at least two recent examples of small-scale

    physical models with applications to long-term hillslope evo-lution based on the study of avalanches. Field studies support

    creep processes to be dominant on low-gradient hillslopes

    or for the convex upper slope portion, whereas landslides

    dominate steep hillslopes or on the steep, straight main slope.

    A physical model using dried beans of various sizes and

    shapes (Densmoreet al., 1997) helps show the genesis and

    role of landslides on the steep main slope. The apparatus used

  • 8/13/2019 Landscape Evolution

    12/28

    258 Frank J. Pazzaglia

    in these experiments wasa narrow flume made of clear acrylic

    walls 2.5 cm apart. The front of the flume could be lowered to

    simulate base level fall. The flume was alternately filled with

    red, oval-shaped beans or white, more spherically-shaped

    beans. Base level was lowered andthe fluxof beans shed from

    the flume was measured for each base level lowering step.

    The beans are thought to represent strong, coherent blocks

    of rock with weak grain boundaries and as such, collectively

    approximate rock behavior where strength is a function of

    both zones of weakness andblock anisotropy. Results showed

    that base level lowering caused a step to develop at the toe

    of the slope. As the step grew, the slope above it destabilized

    and was swept clear of a layer of beans in a slope-clearing

    landslide. The landslides happened at irregular intervals and

    their frequency depended on bean anisotropy. The slope

    clearing events, analogous to landslides, accounted for 70%

    of the total mass removed from the model hillslope. In real

    landscapes, the steep inner gorges in mountainous topogra-

    phy may be analogous to the lower step seen in the model.

    The applicability of this model to real landscapes has been

    questioned because subsequent experiments demonstrate

    that the narrow width of the flume may have influenced the

    initiation of the inner gorges and slope-clearing events (Aaltoet al., 1997). Nevertheless, the study does underscore the im-

    portance of both creep and landslide processes on hillslopes

    have creep and landslide components of mass removal.

    Creep and landslide removal of mass from a hillslope

    has been treated as linear and non-linear diffusive processes

    respectively (Martin, 2000). Non-linear sediment transport

    from model hillslopes was investigated by a physical model

    (Roering et al., 1999, 2001) consisting of a plexiglass box,

    with open ends, filled with a hill of sand. The sand was

    subjected to acoustic vibrations that caused the outermost

    layer of grains to vibrate, dilate, and creep downslope,

    simulating the biologic and freeze-thaw turbation of regolith

    on real hillslopes. A key result of the experiment is that the

    flux of sediment from this model hillslope approximate alinear diffusive behavior when gradients were low (

  • 8/13/2019 Landscape Evolution

    13/28

    Landscape Evolution Models 259

    Many iterations of solving the equations lead to changes

    in cell height that mimic real time-dependent changes in

    landscape elevation and relief. Numeric models provide

    alternatives to cumbersome physical models where scaling

    relationships may be violated. They also facilitate the focused

    investigation of an individual process or suite of processes.

    There is a certain elegance and level of objectivity to numeric

    approaches because the outcome of a good model is the result

    of independent equations based on physical laws. Lastly, and

    perhaps most importantly, numeric models can be used to

    test paradigms of landscape evolution and to predict future

    changes in the landscape, using the modern topography as a

    starting point. However, erroneous interpretations result if the

    boundary conditions processes are not properly identified and

    defined.

    There are several different types of numeric models

    that have been developed over the past three decades. Their

    rapid expansion into studies of landscape evolution has

    been greatly enhanced by computers. There are at least five

    reviews or compilations of numeric landscape models from

    the past ten years: the Tectonics and Topography special

    volume published in the Journal of Geophysical Research

    (Merritts & Ellis, 1994), Koons (1995), Beaumont et al.(2000),Landscape erosion and evolution modeling (Harmon

    & Doe, 2001), and the Steady-State Orogen special volume

    published by the American Journal of Science (Pazzaglia

    & Knuepfer, 2001). The Beaumont et al. (2000) paper is

    particularly noteworthy for its attempt to categorize numeric

    models based on relative complexity. The organization and

    explanations of numeric models in this chapter borrows heav-

    ily upon these sources. The interested reader is directed to

    them for detailed descriptions of the models describedherein.

    The three major types of numeric landscape evolution

    models are surrogate models, multi-process models, and

    coupled geodynamic-surface process models. A surrogate

    model typically tracks a single metric of the landscape, such

    as mean elevation or mean relief, and does not discriminateamong the myriad of processes that redistribute mass in real

    landscapes. The multi-process model attempts to isolate and

    represent all processes that contribute to the redistribution of

    mass with individual mathematical equations. A recent trend

    is to link a robust multi-process model with a geodynamic

    model of rock deformation in the Earths crust, to yield a

    coupled geodynamic-surface process model. A true coupled

    model is one where the surface process model predicts

    lateral mass fluxes throughout the model landscape and the

    corresponding geodynamic model calculates the solid earth

    mass fluxes in response to the surface processes and as a

    consequence of tectonic forces or velocity fields (Beaumont

    et al., 2000). Such coupled models are particularly intriguing

    because their results provide insights into the feedbacksbetween geodynamic and surficial processes.

    All numeric models share the common goal of represent-

    ing erosion of the landscape, either explicitly or implicitly as

    a key component of landscape evolution. Nearly all models

    treat erosion as proportional to local relief, mean elevation

    (Fig. 9), or local gradient and hydrology. There are separate

    justifications for all three. Erosion proportional to mean

    relief is supported by studies that show a correlation between

    river suspended sediment yields and relief (Ahnert, 1970).

    Erosion proportional to mean elevation is supported by

    Ruxton & McDougall (1967) and Ohmori (2000) among

    other studies, and a correlation between mean elevation and

    mean relief has been established for landscapes with well

    integrated drainages (Ohmori, 2000; Summerfield & Hulton,

    1991). An important caveat to models that appeal to relief

    dependent erosion is that they can only be applied to decaying

    landscapes. Constructional or steady-state landscapes may

    exhibit no correlation between mean elevation, mean local

    relief, and rate of erosion. Erosion proportional to local

    gradient and hydrology is predicated on the field and lab

    studies that began with Gilbert (1877, 1909) and continues

    with experiments like those ofRoeringet al. (2001).

    Another feature commonly shared by numeric models is

    the concept of landscape response time. This numeric formal-

    ization of the response timeofBull (1991) is typically derived

    from linear system behavior. Response times are particularly

    important in coupled geodynamic-surface process models

    because the tectonic forcing in these models is commonly

    treated as a step function. Linear system behavior treats the

    system response to a step-like forcing as an exponential andthe response time (sometimes also called the characteristic

    time) is defined as (1 1/e) where e = 2.71828. In other

    words, the response time indicates the amount of time needed

    to accomplish 63% of the landscape change or adjustment

    following a perturbation. Response times are sensitive to

    climate, model rock type, and the spatial-temporal scale of a

    model tectonic forcing. In real orogens that are well-drained,

    the response time scales approximately with the size of the

    orogen. Small orogens like Taiwan have short response times

    on the order of millions of years. Large orogens with orogenic

    plateaux like the Himalaya mayhaveresponse times in excess

    of 100 million years. Linear system behavior is supported

    by real observations of how perturbed Earth systems liberate

    sediment as they seek new equilibria (Schumm & Rea, 1995).Flumes, watersheds, and orogens all show an exponential de-

    cay in sediment yield following the perturbation. In the case

    of a whole orogen, theresponsetimes are long enoughthat the

    simplest explanation for exponential reduction in sediment

    yield is a decrease inthe mean elevation and meanrelief ofthe

    source. At cyclic space and time scales, the Davisian cycle of

    erosion is probably an appropriatelandscape evolution model.

    Analytical solutions to the landscape response timehavebeen

    proposed byWhipple & Tucker (1999)andWhipple (2001).

    Surrogate Models

    The goal of the surrogate landscape model is to track a metricof the landscape, such as mean elevation above base level, or

    evolution of a single landscape component like a streamchan-

    nel (Howard et al., 1994; Whipple & Tucker, 1999), using a

    physically based equations or established functional relation-

    ships between a landscape metrics or processes. Erosion rate

    is commonly indexed to mean elevation (Fig. 9). Several early

    models use an elevation-erosion rate relationship to solve for

  • 8/13/2019 Landscape Evolution

    14/28

    260 Frank J. Pazzaglia

    Fig. 9. (a) Correlation between mean local relief (measured in a circular window 10 km in diameter) and mean denudation rate,

    calculated from suspended sediment yield data from large, mid-latitude, well-drainage watersheds (modified from Pazzaglia

    & Brandon, 1996). (b) Correlation between mean elevation and mean local relief for well drained landscapes (modified from

    Summerfield & Hulton, 1991).

    rockuplift, evolution of mean elevation, and flexural deforma-tion of thelithosphere(Moretti & Turcotte, 1985; Stephenson,

    1984; Stephenson & Lambeck, 1985). Such models have also

    simulated critical taper on an emergent accretionary wedge

    (Dahlen & Suppe, 1988), predicted the height limit of moun-

    tains (Ahnert, 1984; Slingerland & Furlong, 1989; Whipple

    & Tucker, 1999), and predicted the thermal structure of an

    eroding crustal column (Batt, 2001; Zhou & Stuwe, 1994).

    Surrogate landscape models are not designed to capture

    the full range of changes in the landscape as it is eroded, but

    rather are useful in exploring the relative scaling relation-

    ships between landscape metrics and driving and resisting

    forces. For example,Pitman & Golovchenko (1991)justified

    an erosion proportional to mean elevation relationship to

    define the necessary and sufficient conditions for generatingpeneplains in the Appalachian landscape. Pazzaglia &

    Brandon (1996)followed a similar approach to reconstruct

    the evolution of mean elevation of the post-rift central

    Appalachians (Fig. 10a). The erosion rate was reconstructed

    from known quantities of sediment trapped in the Baltimore

    Canyon Trough, and a simple linear equation was solved

    for the general relationship between the flux of rocks into

    the mountain belt and erosional flux out of the belt. The

    results of a tectonic model, in which the flux of rocks into

    the belt was allowed to vary, but climate (rock erodibility)

    was held constant shows that three kilometers of rock would

    have to be fluxed through the belt in the past 20myr to

    account for the recent offshore sediment volumes. Mean

    elevation of the range would have had to increase from nearsea level before the Miocene to 1100 m, before diminishing

    to its present value of about 350 m. Tested by, for example,

    thermochronology, such predictions help place limits on the

    range of landscape responses that might be expected from a

    tectonic or climatic perturbation.

    A less common surrogate approach is to apply a stochastic

    technique where landscape change is determined by the

    Fig. 10. (a) Reconstruction of rock uplift (source, triangles),

    mean elevation (squares), and total erosion (inverted trian-

    gles) for the post-rift Appalachian mountains (modified fromPazzaglia & Brandon, 1996). This simple single-process

    landscape model provides insights to the scaling relation-

    ships between mean elevation and rock uplift to explain the

    known volume of sediment delivered to offshore basins. (b)

    Illustration of detrital mineral age spectra during exhumation

    of the landscape to the steady-state relief condition (modified

    fromStock & Montgomery, 1996).

  • 8/13/2019 Landscape Evolution

    15/28

    Landscape Evolution Models 261

    mean and variance of landscape slope elements (Craig,

    1982, 1989). This particular type of model is one of the

    few that carefully incorporates map-scale considerations

    of the influence of structure, rock type, and stratigraphy

    on landscape evolution. One-kilometer-long slope elements

    called draws are constructed between adjacent cells in a

    DEM where the rock type underlying the cells is noted. The

    mean and variance of draws for any particular rock type

    pair is calculated and recorded. The slope of the draws is

    assumed to be well adjusted to the rock type and structure.

    The model can only be applied in settings where driving and

    resisting forces approach a dynamic equilibrium. Erosion is

    introduced as a random perturbation by lowering one cell.

    All adjacent cells are adjusted randomly by choosing a new

    draw that falls within the variance limits of all possible draw

    slopes developed between specific rock type pairs. The result

    is a diffusive rippling effect of slopes adjusting away from

    the perturbation. Relief can increase or decrease depending

    on the rock types encountered or uncovered. Erosion stops

    when the landscape has been lowered to base level.

    Another type of surrogate model involves inversion

    techniques to estimate paleoelevation or paleorelief using

    thermochronology in the source and/or detrital grain ages(e.g. Bernet et al., 2001; Brandon & Vance, 1992; Garver

    et al., 1999; Stock & Montgomery, 1996; Fig. 10b). Con-

    ceptually these models assume a relief or elevation depen-

    dence on the exhumation of bedrock in the source. Mineral

    isochrons commonly increase in age with elevation especially

    where isotherms are a horizontally subdued reflection of to-

    pography. Measurable mineral age gradients are apparent in

    landscapes with moderate relief (3 km) and erosion rates

    around0.5 mm/yr. Thenumerical ages of theminerals depend

    on relief, the geotherm, and erosion rate. If erosion is uniform

    across the landscape and relief remains more or less steady as

    the landscape is unroofed, the vertical distribution of mineral

    ages in a depositional basin essentially reflects that unroofing

    process with a constant age gradient. If relief instead changesas the land erodes, the age gradients in the depositional basin

    also change. An inversiontechnique hasbeen proposed to con-

    vert that change in age gradient to changes in relief (Fig. 10b;

    Stock & Montgomery, 1996). The technique requires high

    precision single-grain ages, no change in mineral age during

    transport and deposition, mixing of the eroded sediment, and

    minimal sediment storage between source and sink.

    Multi-Process Landscape Models

    A multi-process landscape model considers two or more

    mutually interacting geomorphic processes that sculpt real

    landscapes. The model mathematically describes these pro-cesses, links the mathematical descriptions together in such a

    way that the continuity of mass is maintained, provides inputs

    of rock type, climate, and tectonics, and predicts the resulting

    landscape evolution (Slingerlandet al., 1994). Multi-process

    landscape models differ from the surrogate models in that the

    former actually strives to simulate the form and process of

    real landscapes, as opposed to simply exploring the scaling

    relationships among landscape metrics. Two generations of

    multi-process landscape models are recognized, the second

    borrowing heavily on the first. First generation multi-process

    models include attempts to mathematically capture bedrock

    weathering, hillslope creep, landsliding, fluvial transport

    of sediment, and channel initiation (Ahnert, 1976, 1987;

    Anderson & Humphrey, 1990; Armstrong, 1976; Chase,

    1992; Gregory & Chase, 1994; Kirkby, 1986; Musgrave etal.,

    1989; Willgoose et al., 1991). Second generation models

    specifically explore broader, more complex interactions

    among processes by introducing, for example, climatic

    perturbations (Tucker & Slingerland, 1997) or migration of

    drainage divides (Kooi & Beaumont, 1994).

    The basic physical principle underlying all multi-process

    models is the conservation of mass. Mass rates into and out

    of model cells are driven by the geomorphic processes acting

    on the landscape (Fig. 11). These include but are not limited

    to bedrock weathering, hillslope creep, hillslope landsliding,

    glacial erosion, bedrock channel erosion, alluvial channel

    erosion, and sediment transport. A simple multi-process

    model collapses all hillslope processes into one mathematical

    equation that treats them collectively as diffusion. Similarly, it

    treats all fluvial transport processes collectively as advection,with sediment transport proportional to unit streampower (the

    discharge-slope product;Howard, 1994). More sophisticated

    approaches consider non-linear processes such as hillslope

    landsliding. A fine example of a robust multi-process land-

    scape model, GOLEM, with examples of the various model

    inputs is discussed in detail inSlingerlandet al. (1994).

    Bedrock weathering is typically modeled as an alter-

    ation front that penetrates downward at a rate inversely

    proportional to the thickness of the regolith cover (Ahnert,

    1976; Anderson & Humphrey, 1990; Armstrong, 1976;

    Rosenbloom & Anderson, 1994). The justification for this re-

    lationship is that soil genesis is a self-limiting process. Thick,

    well developed soils (or regoliths) tend to influence infiltra-

    tion rates and generate runoff resulting in overland flow thatstrips the upper horizons. Regolith thickness approximates

    a steady state when the rate of stripping equals the rate of

    descent of theweathering front (Heimsath etal., 1999; Pavich

    et al., 1989).

    Hillslope creep is typically modeled as a linear diffusive

    process where sediment transport is proportional to slope.

    Diffusion has attracted the greatest attention of all modeled

    geomorphic processes both within and outside the modeling

    community because of its simplicity and its apparent veri-

    fication from slope degradation studies (e.g. Culling, 1960,

    1963; Hanks et al., 1984; Nash, 1980). On any uniformly

    eroding hillslope, it has been long known that sediment

    flux must increase systematically away from the hilltop

    (Gilbert, 1909). If the sediment flux is dependent on theslope, the combination of a slope-dependent transport law

    and the continuity equation leads to mathematical equation

    that looks like linear diffusion (Culling, 1960; Nash, 1980).

    However, this equation describes change in the hillslope

    profileand is not meant to describe the downslope transport

    of individual particles of sediment, a process that is probably

    non-diffusive. The diffusion constant or diffusivity in the

  • 8/13/2019 Landscape Evolution

    16/28

    262 Frank J. Pazzaglia

    Fig. 11. Schematics showing (a) the arrangement of cells in a traditional 2-D, multi-process model where flow is directed down

    the steepest paths between rectilinear cells and (b) the major processes that are represented mathematically in model landscapes

    (both modified fromTucker & Slingerland, 1994). Most models now employ a triangulated network, rather than rectilinear cells.

    diffusion equation is a scale dependent parameter that

    implicitly includes climate and substrate characteristics.

    The idea of simple hillslope diffusion is complicatedby field observations and physical models of slope-clearing

    landslides as important agents for removing mass from

    hillslopes, especially in steep landscapes undergoing tectonic

    uplift. Multi-process models that account for landsliding

    typically assign a threshold slope angle, below which

    sediment is transported by diffusion, and above which,

    sediment is transported by slope-clearing landslides until the

    slope diminishes to a threshold angle (Kirkby, 1984). More

    soph