the inﬂuence of geomorphic processes on plant distribution...

Ecological Monographs, 82(4), 2012, pp. 429–447� 2012 by the Ecological Society of America

The influence of geomorphic processes on plant distributionand abundance as reflected in plant tolerance curves

M. N. CHASE,1 E. A. JOHNSON,1,2,4 AND Y. E. MARTIN2,3

1Department of Biological Sciences, University of Calgary, Calgary, Alberta T2N 1N4 Canada2Biogeoscience Institute, University of Calgary, Calgary, Alberta T2N 1N4 Canada

3Departments of Geography and Geoscience, University of Calgary, Calgary, Alberta T2N 1N4 Canada

Abstract. Ecologists describe plant distribution using direct gradient analysis, by which atolerance curve of species abundance is described along an environmental gradient (anyenvironmental variable that affects plant distribution). Soil moisture is generally the gradientin low-relief areas that explains the most variation. Traditional direct gradient analyses haveused terrain structure (i.e., transects up or down hillslopes) as a correlate to soil moisture.Here we use a numerical tectonic and geomorphic process-based landscape developmentmodel to create two landscapes with different geomorphic characteristics: (1) to demonstratethe influence of geomorphic processes on soil moisture patterns and plant distribution and (2)to evaluate the effectiveness of transects in describing moisture gradients and tolerance curveson landscapes dominated by creep or overland flow. We use a topographic index toapproximate the distribution of soil moisture as it is determined by the shape of these differentlandscapes. Transects are placed on hillslopes in each model landscape and used to constructtolerance curves.

Results show that transect methods that use the distance from the channel to the ridgelineas an approximation of soil moisture create variable tolerance curves for the same plant bothwithin a single landscape and between different landscapes. The reason for these differences isthat transects do not take into account the three-dimensional landscape form that explainswater movement. Landscapes have regions of convexity and flow path divergence and regionsof concavity and flow path convergence that, along with hillslope length, determinecontributing area. In addition, hillslope curvature results in varying capacities to retainwater. However, when the topographic index is used instead of hillslope transect position,tolerance curves from the same and different landscapes reflect the differences the topographicstructure has on soil moisture. We thus show that traditional methods of direct gradientanalysis are not always adequate as they do not tend to consider that soil moisture depends onhillslope length, curvature, and slope. Furthermore, we show that within and betweenlandscapes there are differences in spatial distributions of soil moisture that are reflections ofthe geomorphic processes that created them.

Key words: direct gradient analysis; geomorphic process; geomorphology; gradient analysis; landscapedevelopment models; plant distribution; plant ecology; tolerance curves; topographic index; transect.

INTRODUCTION

One of the basic goals of plant ecology is to describe

and understand the distribution of plants on the

landscape. Tolerance curves (Shelford 1913) are one

tool used to describe the local distribution of plants.

‘‘Local’’ here will refer to the scale of hillslopes and low-

order watersheds (this will be defined more carefully

later). Tolerance curves are the abundance of a species

distributed along one or more orthogonal environmental

gradient(s). They are generally symmetrical and unim-

odal; however, several studies have also found skewed

and bimodal distributions, and the shape of these curves

has been a subject of some debate (e.g., Austin et al.

1984, Austin 1987, Minchin 1989, Oksanen and Minchin

2002). Regardless of the discrepancy in the shape of the

curve from one study to the next, gradient methods used

to derive the curves often come to similar conclusions,

that soil moisture, often along with nutrients, is one of

the principal axes of variation in abundance (e.g., Curtis

and McIntosh 1951, Whittaker 1956, Waring and Major

1964, Bridge and Johnson 2000, Zinko et al. 2005, and

many others). Temperature/elevation is another primary

axis and in mountainous landscapes is often the first axis

(Whittaker 1967, Urban et al. 2000).

However, the reasons why landscape organization

influences plant distribution and abundance via soil

moisture are rarely explored in any detail in ecology (for

an exception, see Hack and Goodlett [1960]). Over the

last several decades, geomorphological research has

resulted in a better understanding of how geomorphic

processes act on landscapes to give them their charac-

Manuscript received 30 November 2011; revised 18 April2012; accepted 20 April 2012; final version received 22 May2012. Corresponding Editor (ad hoc): M. W. Davis.

4 Corresponding author. E-mail: [email protected]

429

teristic forms (e.g., Dietrich and Montgomery 1998),

while hydrologists have demonstrated the connections

between topography and soil moisture (e.g., Beven and

Freer 2001). An understanding of these earth-surface

processes can help explain what makes moisture (and

nutrient) gradients so universally important in deter-

mining local plant distribution.

Traditional direct gradient analysis (Gauch 1982),

uses landscape structure as a correlate to gradients. For

example, transects up or down hillslopes may be used to

approximate soil moisture gradients in low to moderate

relief landscapes, and elevation as a correlate of

temperature in mountainous landscapes (e.g., Whittaker

1967). However, knowledge of the processes that shape

landscapes is necessary to understand and explain

gradients of plant distribution and abundance on a

landscape.

This study uses a numerical model of landscape

development that combines tectonics and geomorphol-

ogy to create landscapes with known processes. The two

objectives of this study are (1) to demonstrate the effect

different geomorphic processes have on soil moisture

patterns and the subsequent implications for plant

distribution and (2) to analyze patterns of soil moisture

by evaluating the effectiveness of three different direct

gradient analysis transect methods (channel to ridgeline,

path of steepest ascent from channel to ridgeline, and

path of steepest descent from ridgeline to channel) in

describing consistent moisture gradients, moisture dis-

tribution, and tolerance curves on two landscapes

created by different combinations of geomorphic pro-

cesses.

Numerical models of combined tectonic and geomor-

phic processes have been used to explore the develop-

ment of landscapes over a range of spatial and temporal

scales (e.g., Harmon and Doe 2001, Pelletier 2008).

Numerical modeling has been an effective tool for such

studies because of the complexity of the process

operation and the often very large spatial and temporal

scales of operation (for examples see Martin and Church

2004). We will use a landscape development model

(Tucker et al. 2001) to create two landscapes (dominated

respectively by soil creep and erosion by overland flow)

in which the geomorphic processes acting upon each

landscape are defined and the relative strengths of these

processes can be controlled. Although the landscapes

were created under different precipitation regimes, we

will use them only as a first approximation of the

representative topographic forms (e.g., leaving out

precipitation and heat budgets). Real-world landscapes

are subject to a variety of historical contingencies

(Simpson 1963, Martin and Church 2004) that lead to

the unique properties of particular landscapes, their

morphology and other associated properties (bedrock

types, soil properties, morphologic irregularity). These

historical contingencies provide real-world variation in

landscapes that make each landscape unique despite

some basic similarities. To eliminate the possibility of

historical contingencies affecting the landscapes being

analyzed, we use model prototypes created using anumerical model of landscape development in our

analysis.A topographic index is used as an approximation of

the relative spatial distribution of soil moisture ascontrolled by a landscape (Kirkby 1975, Grayson et al.

1997); this index is based on water movement overhillslopes and its contribution to soil moisture at aparticular location. Because of the manner in which the

index is constructed, locations with similar indexes willhave similar topographically determined soil moisture if

the index’s assumptions are met (Beven 2001). Thisallows comparison of similar dry, mesic, and wet

locations on landscapes produced by different geomor-phic processes. The gradient of the topographic index

can then be related to species abundance (i.e., tolerancecurves).

CONCEPTUALIZING BACKGROUND

Gradients and tolerance curves

Generally two approaches are used to construct soilmoisture gradients and plant tolerance curves. The firstapproach determines gradients of soil moisture indirect-

ly by using the similarity in stand/plot species compo-sition and abundance. Multivariate statistical techniques

are used to define the axis of principal variation (Gauch1982). The gradients, and thus the tolerance curves, are

constructed by placing stands/plots with similar speciescomposition close together on the axes, there is not

necessarily any relationship of these gradients to specificlandscape positions (but see Loucks 1962). The second

approach defines moisture gradients directly based onthe assumption that a one-dimensional transect up a hill

will, as a first approximation, give a gradient ofdecreasing moisture, or that the landscape can be

organized into regions of high and low soil moistureusing topographic features such as valley bottoms and

crests of ridges (e.g., Whittaker 1956, 1967, Racine1971). This assumption probably followed from the one-dimensional idea of tolerance curves (Shelford 1913).

One problem with this approach is the lack of astandardized model. Often the hills used, the arrange-

ment of transects, and the position of transects orcategories in the landscape are not well specified with

respect to the landscape’s geomorphology and hydrol-ogy. This approach, as we will show, is often at variance

with the three-dimensional manner in which water isknown to move on hillslopes.

Landscape organization

Landscapes have characteristics that are universallyshared; watersheds consist of peaks (local elevation

maxima), streams (local minima), and hillslopes thatconnect these relative high and low points (Warntz 1975,Werner 1988, 1991, Dawes and Short 1994). All

landscapes can be delineated using contour lines(locations of equal elevation) and flow lines (lines

M. N. CHASE ET AL.430 Ecological MonographsVol. 82, No. 4

orthogonal to contours). Flow lines indicate the

direction of steepest descent and usually connect

ridgelines and stream courses.

While these basic elements are similar for all

landscapes, the degree of dissection will vary. A highly

dissected landscape will have more channelization per

unit area than a less dissected landscape. This charac-

teristic is referred to as the drainage density, Dd (Horton

1945):

Dd ¼X

L

Að1Þ

where L is the length of each channel link in the entire

channel network and A is the total area of the basin. For

a list of symbols found in equations, see Table 1. The

average length of a hillslope (or length of overland flow),

lo (Horton 1945), is

lo ¼1

2Dd

: ð2Þ

According to this relationship, a landscape that is more

dissected will have shorter hillslopes on average than one

that is less dissected.

Landscapes develop through possible tectonic activity

that causes uplift and geomorphic processes that erode

and redistribute sediment across the landscape (e.g.,

Anderson and Anderson 2010). When erosive forces

overcome the resistive forces imparted by substrate and

vegetation (e.g., Istanbulluoglu and Bras 2005), material

is transported downslope. Two hillslope processes that

transfer sediment are soil creep and overland flow.

Soil creep is the gradual downhill movement of soil

due to gravity (Culling 1963), aided by events such as

freeze–thaw (Davison 1889) and wetting and drying

cycles (Fleming and Johnson 1975), bioturbation (Black

and Montgomery 1991, Roering et al. 2002, Yoo et al.

2005), and rain splash (Moeyersons 1975). Vegetation

may provide resistance to erosive processes such as creep

because roots help to anchor soil and hinder its

movement (Thornes 1990). Alternatively, vegetation

may contribute to erosion through tree throw events

(e.g., Thornes 1990). Soil creep rates increase linearly as

slope increases for low to moderate slopes (Gilbert 1909,

Culling 1960, McKean et al. 1993, Small et al. 1999);

however, in regions characterized by steep slopes, soil

creep may become an increasingly nonlinear process as

slope increases (Howard 1994b, Martin and Church

1997, Roering et al. 1999). Soil creep is most likely to be

a dominant process in lower-gradient landscapes and

produces hillslopes that are convex (Davis 1892, Gilbert

1909, Anderson 1994), due to the nature of the sediment

transport processes (which are reflected in the diffusion

equation that is often used to represent soil creep in

numerical models).

Two types of overland flow that contribute to erosion

are infiltration excess overland flow (Hortonian over-

land flow) (Horton 1933) and saturation excess overland

flow (Kirkby and Chorley 1967). Under conditions of

poor infiltration, infiltration excess overland flow may

occur uniformly over the hillslope and primarily occurs

in areas of arid vegetation and low drainage density

(Horton 1933). Saturation excess overland flow is the

more common form of overland flow in humid totemperate watersheds with a higher density of vegetation

(Kirkby and Chorley 1967). In these latter watersheds,

water moves primarily as subsurface flow, travelling

through the porous spaces in soil and rock along paths

of steepest descent that are orthogonal to contour lines.The flow of water downslope causes the water table to

rise. If the rise of the water table is great enough, it may

result in groundwater returning to the surface. Any

subsequent precipitation is unable to infiltrate the soil

and further contributes to saturation overland flow.Since the water table is closest to the surface at the bases

of hillslopes and water moves downhill, saturation

overland flow will first occur at the bases of slopes and

then progress uphill with increased precipitation. The

volume of water flowing over a location per unit time

(i.e., the water discharge) may be approximated bymultiplying runoff rate and contributing area (Tucker

and Bras 1998). If the water discharge is great enough, it

may become an erosive force by overcoming the resisting

stresses and result in the formation of miniature

channels or rills. Overland flow tends to produce

TABLE 1. Definition of symbols used in equations.

Symbol Definition

Dd Drainage density, length of channels per area(m/m2)

L Channel length (m)A Contributing area (m2)lo Hillslope length or length of overland flow (m)S Slope gradient, rise over run, dimensionlessz Elevation (m)t Time (yr)U Rate of tectonic uplift (m/yr)Q Surface discharge (m3/yr)R Runoff rate (m/yr)Qtotal Total discharge (m3/yr)Qsubsurface Subsurface flow (m3/yr)b Contour width (m)T Transmissivity (m2/yr)qs Rate of sediment transport (m3/yr)Kf Sediment transport efficiency factor

(m3�yr�1�Pa�1)W Channel (or rill) width (m)s Shear stress (Pa)sc Critical shear stress to be overcome for erosion

to occur (Pa)P Shear stress exponent, dimensionlessKt Coefficient for shear stress

(Pa�yr�1�m�2)M Detachment discharge exponent, dimensionlessN Detachment slope exponent, dimensionlessDc Erosion rate, or detachment capacity (m/yr)KB Bedrock erodibility (m�yr�1�Pa�1)Pb Bedrock shear stress parameter, dimensionlessqs(diffusion) Diffusive sediment flux (m3/yr)KD Diffusion coefficient (m2/yr)rz Tangent slopeSc Critical slope gradient, dimensionless

November 2012 431GEOMORPHIC PROCESSES INFLUENCE PLANTS

hillslopes that are concave (Howard 1994a, Tucker and

Bras 1998, Dietrich and Perron 2006).

Linking landscape organization to soil moisture gradients

and plant tolerance curves

The topographic index (Kirkby 1975) provides a

reasonable approximation of the relative distribution of

soil moisture at a location in the basin:

ln ðA=SÞ ð3Þ

where A is contributing area at a given contour width

and S is the local slope. See Kirkby (1975) for a

derivation of the index from near first principles. This is

a very simplified approach as many factors influence soil

moisture (e.g., temperature, precipitation); however, in

the present study we use the simple topographic index to

examine the role of topography as it influences plant

distribution and assume that, once the landscapes have

formed, all other factors are equal. The advantage of the

topographic index is that, as a similarity index, it allows

for direct comparison of sites with the same index while

considering the role of the three-dimensional shape of

the landscape in determining water flow and soil

moisture distributions. Other similarity indices approx-

imating saturated soil moisture have also been devel-

oped (O’Loughlin 1981, 1986, Hjerdt et al. 2004), most

of which are expanded versions of Kirkby’s original

derivations (Beven 2001). Due to a much larger area of

the landscape being composed of hillslopes than

channels, the topographic index distributions for water-

sheds are generally right skewed (Beven 2001).

Hillslopes can be subdivided by curvature (i.e.,

convergence and divergence), creating nine categories of

hillslopes (Fig. 1) that should each show hydrological

similarity. Two measures of curvature are considered for

this categorization (Suzuki 1977 presented in Tsukamoto

and Ohta 1988, Aryal et al. 2002). The first, planform

curvature, is determined by the convergence ratio, which

is the ratio of the length of the ridgeline to the length of

outlet for a hillslope (according to the definition by

Suzuki 1977 presented in Tsukamoto and Ohta 1988,

Aryal et al. 2002). Headslopes should have large

convergence ratios because the ridgeline of the hillslope

is much longer than the headwater channel toward which

the subsurface flowpaths of the hillslope drain. Sideslopes

can be separated into parallel hillslopes and divergent

hillslopes. Parallel hillslopes have convergence ratios that

are close to one because the top and the base of the

hillslope have nearly equal lengths and flow paths are

parallel. Divergent hillslopes have a ratio that is less than

one because flow paths diverge and water drains from a

smaller area at the ridgeline to a larger area at the channel

link. The second measure of curvature is profile curva-

ture, which is the change in slope along a profile. Profile

curvature separates hillslopes into concave, planar, or

convex types. When profile curvature is combined with

the three planform types, it allows for a total of nine

distinct hillslope types to be identified (Fig. 1).

FIG. 1. Nine different hillslope types that may be identified based on planform and profile curvature (from R. Suzuki, aspresented in Tsukamoto and Ohta [1988] and Aryal et al. [2002]). Reprinted from Tsukamoto and Ohta (1988) with permissionfrom Elsevier; Aryal et al. (2002) reproduced/modified by permission of the American Geophysical Union.


THE LANDSCAPE DEVELOPMENT MODEL

A summary outline of the landscape development

model is now described. Herein, the Channel Hillslope

Integrated Landscape Development model (CHILD)

(Tucker et al. 2001) was utilized to create representative

digital elevation models (DEMs) for our study. The

following is a summary of major model components for

sediment transport contained in CHILD version R18.12.

A model of this magnitude entails lengthy and detailed

source code; for a more full discussion of model details,

see Tucker et al. (2001), and for technical material see

Tucker (2010).

CHILD simulates uplift and erosion/deposition ac-

cording to the following general equation:

]z

]t¼ U þ ]z

]t

��water

þ ]z

]t

��creep

ð4Þ

where z is elevation, t is time, and U is uplift rate. The

primary sediment transport equations for each of these

model components are outlined in the following

subsections.

Water-based erosion

The equations for water-based transport are applied

to all grid cells across the landscape; particular values of

discharge may result in (1) no erosion (if discharge

values are very low), (2) erosion due to overland flow, or

(3) fluvial transport (for high values of concentrated

flow). The change in elevation with time for each cell is

determined by the difference between the sediment

transport due to water-based erosion entering and

exiting each cell.

Input values of water discharge are required for the

sediment transport equations. Water-based erosion

occurs during storms, and storm intensity, duration,

and time between storms may either be modeled as

constant values or as exponential distributions centered

around a mean (Eagleson 1978, Tucker and Bras 2000).

In our study, constant values are used. Overland flow

may be modeled as infiltration excess overland flow

(Hortonian) or as saturation excess flow. In the case of

infiltration excess overland flow, surface discharge, Q, is

Q ¼ R 3 A ð5Þ

where A is contributing area and runoff, R, is the

difference between rainfall intensity and infiltration

capacity of the soil. For saturation overland flow,

surface discharge is calculated using equations based

upon O’Loughlin’s (1986) topographic index:

Q ¼ Qtotal � Qsubsurface ð6Þ

where total discharge, Qtotal, is area multiplied by the

precipitation rate. Qsubsurface (subsurface flow) is given

by

Qsubsurface ¼ S 3 b 3 T ð7Þ

where S is local slope, b is contour width, and T is soil

transmissivity.

Channel geometry for rills and channels is also

necessary input information for model runs. The

approach followed in CHILD follows Leopold and

Maddock’s (1953) empirical relationships.

CHILD includes the possibility for two water-erosion

situations, that of transport-limited erosion (i.e., sedi-

ment supply is not a limiting factor) and a second case in

which material is solid (i.e., rock) or has notable

cohesion or vegetation anchoring. The sediment trans-

port rate (qs) is modeled as a function of excess shear

stress above some critical threshold:

qs ¼ KfWðs� scÞP ð8Þ

where Kf is a sediment transport efficiency parameter, W

is width, s is average bed shear stress, sc is critical shearstress for sediment entrainment, and P is an exponent

with values typically in the range of 1.5 to 2.5 depending

on the type of load being transported. Bed shear stress,

s, is given by

s ¼ Kt

Q

W

� �M

SN ð9Þ

where Q is discharge, andW is width. Parameters Kt, M,

N are described in Table 1.

The rate of lowering due to detachment of material

(Dc) can be modeled as

Dc ¼ KBðs� scÞPb ð10Þ

where KB is bedrock erodibility, s is shear stress, sc is thecritical shear stress to be overcome for erosion to occur,

and Pb is the bedrock shear stress parameter. Note that

shear stress, s, is calculated similarly to Eq. 9.

Vegetation has been incorporated into the model as a

resistance to erosion. The reader is referred to Collins et

al. (2004) for details regarding these model extensions.

Soil creep

Soil creep has been modeled using the Roering et al.

(1999) adaptation of the Howard (1994b) equation that

models creep as a nonlinear process:

qsðdiffusionÞ ¼KDrz

1� ðjrzj=ScÞ2ð11Þ

where qs represents the diffusive sediment flux, KD is a

diffusion coefficient, rz represents the tangent slope of

the hillslope, and Sc is a critical hillslope gradient.

METHODS

Landscape simulations

The Channel Hillslope Integrated Landscape Devel-

opment model (CHILD) version RI8.12 (Tucker et al.

2001) was used to generate two digital elevation models

(DEMs) of landscapes having spatial dimensions of 3 3

2 km with elevation contours of 15 m. Both landscapes


had a single outlet point through which water and

sediment transported within them exited.

Table 2 provides the parameters used in the creation

of each of the landscapes. The landscapes were not

designed to represent any specific real landscapes, but

rather to represent a general prototype for landscapes

formed by certain dominant processes. Parameter values

for numerical model runs were chosen so that they

created landscapes characteristic of different geomor-

phic processes (e.g., Tucker and Bras 1998). For each of

the landscapes, a DEM and slope (contributing)–area

plot were generated. Slope–area plots are simple two-

dimensional portrayals of the distribution of these two

topographic variables and can be used as an approxi-

mate means of showing the different geomorphic

processes in the landscape (for explanation of slope–

area plots see, e.g., Willgoose et al. [1991], Montgomery

and Foufoula-Georgiou [1993], Tucker and Bras [1998]).

Calculations of landscape characteristics

After the DEMs were created, they were converted

from triangular networks into raster grids, and TAR-

DEM (Tarboton 1989, 1997) was used to calculate slope

gradients, flow directions, and contributing areas. The

flow directions were then used to define channels and

hillslopes using the hsB toolkit (Troch et al. 2003,

Bogaart and Guardiola 2007).

Each basin was divided into hillslopes of the nine

planform and profile curvature types (cf. Tsukamoto

and Ohta 1988) using the hsB toolkit (Troch et al. 2003,

Bogaart and Guardiola 2007). Channels were delineated

as cells with a contributing area greater than 50 000 m2

because this provided reasonable looking drainage

basins on both landscapes. This value is lower than

typical real world values; however, since it does not

differ between the landscapes it does not affect the

comparison between them. The spatial distribution of

soil moisture for each landscape was then approximated

by calculating soil moisture for each cell on the

landscape in question using Kirkby’s (1975) topographic

index.

Transect sampling and tolerance curve construction

The assumption was made that the distance from the

channel to the ridgeline indicated a gradient of decreased

topographic index (i.e., relative soil moisture), and the

abundance of a species as a function of distance along

the transect yielded a tolerance curve. Transects were

then established on each landscape within each of the

nine hillslope categories using the three different

methods illustrated in Fig. 2. The first method, the

channel-to-ridgeline transect, cut straight up a hillslope

from the middle of the base of the hillslope to the middle

of the ridgeline (Fig. 2). Contributing area, slope, and

topographic index were calculated and plotted every 25

m along each channel-to-ridgeline transect. The second

method, the steepest ascent transect, traced a path

through the grid cells of the landscape from the middle

of the base of the hillslope to the ridgeline. The third

method, the steepest descent transect, traced a path

through the grid cells from the middle of the ridgeline to

the base of the hillslope. Contributing area, slope, and

topographic index were calculated and plotted for every

grid cell along each steepest ascent and steepest descent

transect (Fig. 2).

Next, a ‘‘mesic’’ Gaussian plant tolerance curve was

created (Fig. 3). This theoretical plant was designed with

a distribution along the topographic index so that it

would be present on both landscapes, and sampled by

each transect. The mesic plant’s abundance was then

mapped on each landscape using the topographic index

at each location. Tolerance curves were constructed by

defining the soil moisture gradient either (1) as the value

of the topographic index or (2) by using the distance

from the channel to the ridgeline as an approximation of

decreasing soil moisture for each of the three transect

methods.

RESULTS

The results are divided into four parts. The first part

examines how different combinations and magnitudes of

tectonic and geomorphic processes produce landscapes

with characteristic topographic structures. The second

part then analyzes how the resulting topography might

affect soil moisture patterns and the resultant plant

distribution. The third part investigates whether the

traditional one-dimensional transects used in direct

gradient analysis produce consistent gradients of topo-

graphic indexes of relative soil moisture. The fourth and

TABLE 2. Parameter values used in the landscape evolutionmodel simulations to generate the landscapes.

Parameter Value

Landscape 1: creep dominated

Spacing between nodes 15 mUplift rate, U 0.00003 m/yrDiffusion coefficient, KD 0.005 m2/yrBedrock erodibility, KB 0.000003 m�yr�1�Pa�1Transmissivity, T 6209 m2/yrCritical shear stress of soil, sc;s 5 PaShear stress imparted by

vegetation, sc;v

80 Pa

Erodibility of vegetation, Kv 1 yr/PaTime for vegetation to grow to

point where limits erosion, Tv

5 yr

Storm intensity 5.22 m/yrStorm duration 0.885 yrBetween-storm duration 29.5 yr

Landscape 2: overland flow dominated

Uplift rate, U 0.00007 m/yrBedrock erodibility, KB 0.000004 m�yr�1�Pa�1Infiltration rate, I 1.74 m/yrShear stress imparted by

vegetation, sc;v

40 Pa

Note: Values for the creep- and overland-flow-dominatedlandscapes are the same unless otherwise indicated.


final piece of the results deals with how an understand-

ing of landscape geomorphic processes and hydrology

may provide a starting point for estimating improved

moisture gradients for plant ecologists.

Landscapes dominated by different processes

The DEM of the creep-dominated landscape (Fig. 4a)

displays the expected characteristic rounded (convex)

hilltops. The slope–area plot (Fig. 4b) shows a strong

initial increase in slope with contributing area, which

indicates slope convexity and the dominance of soil

creep (Willgoose et al. 1991, Tucker and Bras 1998).

Real-world examples of such a landscape are the rolling

hills found in coastal Marin County, California, USA

(Black and Montgomery 1991, Dietrich et al. 1993).

Of the two landscapes, the creep-dominated landscape

has the lowest relief (the difference in elevation between

the highest and lowest points of a landscape) as well as

the lowest slopes. The average hillslope length calculated

using Eq. 2 is about 162 m.

The DEM of the overland-flow-dominated landscape

(Fig. 5a) has sharper peaks than the creep-dominated

landscape. A decrease in slope with increasing area is the

predominant trend in the slope-area plot (Fig. 5b),

indicating dominance of overland-flow erosion (Mont-

gomery and Dietrich 1988, 1989, 1992, 1994, Willgoose

et al. 1991, Tucker and Bras 1998). This landscape

experienced infiltration excess overland flow rather than

the saturation excess overland flow present in the creep-

dominated landscape. The average hillslope length is 167

m. A similar real-world landscape would be the desert of

Walnut Gulch, Arizona (Parsons et al. 1991, Grayson et

al. 1992, Wainwright et al. 2000).

Each of the modeled landscapes is dominated by a

different geomorphic process as indicated in the

distinctive signatures found in their slope–area plots

and their topographic structure. Next, we evaluate

how these differences in landscape structure are reflected

in relative soil moisture distributions using the topo-

graphic index.

Soil moisture pattern on landscapes dominated by

different geomorphic processes

The frequency distributions of the topographic index

(Fig. 6a) for all pixels on both landscapes are right

FIG. 3. The plant species created where abundance is a function of the topographic index.

FIG. 2. The three different types of transects on a hillslopeof the creep-dominated landscape: the straight line representsthe channel to ridgeline path, solid circles represent the path ofsteepest ascent, and eight-pointed stars represent the path ofsteepest descent. The different shaded regions indicate differenthillslopes. The white grid cells represent the channel network.Although the topographic index was calculated using contrib-uting areas based on a dinf flow routing scheme, the d8 (path ofsteepest descent, and direction most of the flow will follow) flowlines are depicted as thin blue lines instead for the sake ofsimplicity.


skewed with different means and ranges. The creep-

dominated landscape is the wetter landscape of the two

with greater mean and median topographic index values

than the overland-flow-dominated landscape. This result

is due to greater median contributing areas and lower

mean and median slopes on the creep-dominated

landscape than the overland-flow-dominated landscape

(Fig. 6b and c). However, the creep-dominated land-

scape has minimum and maximum topographic index

values (7.0 to 22.0) that are drier than the overland-flow-

dominated landscape (7.7 to 22.6; Fig. 6a).

The differences in topographic index values are shown

on the landscape DEMs and illustrate how dry, mesic,

and wet species would be distributed on the landscape

(see Figs. 4 and 5). On the creep-dominated landscape

(Fig. 4a), wet areas radiate outward from headwaters or

concave valleys, where contributing areas are large, and

the driest areas are restricted to ridgelines. The overland-

flow-dominated landscape (Fig. 5a) is drier than the

creep-dominated landscape, and wet regions tend to

occur lower on the hillslopes.

The frequency distribution of hillslopes in the nine

groups of Aryal et al. (2002) (Fig. 7) is similar for both

landscapes. Convergent-concave and divergent-concave

hillslope types are the most common in both landscapes.

Convergent-concave hillslopes are the dominant hill-

slope type at first-order channels, the most common

channel order. Divergent-concave hillslopes are the

dominant hillslope type at second-order channels.

The abundance distribution for a mesic plant is shown

in Fig. 8 for the landscapes. On the creep-dominated

landscape, the mesic plant is abundant on and near

ridges (Fig. 8a) but abundance rapidly declines down-

slope. On the overland-flow-dominated landscape, the

mesic plant is widely distributed (Fig. 8b) with

intermediate abundance on the ridges, high abundance

below the ridges, and intermediate to low abundance

further downslope. When the plant’s abundance is

plotted as a function of the topographic index on

samples of each type of transect (Fig. 9), the overland-

flow-dominated landscape captures a larger portion of

the species range than the creep-dominated landscape.

It has thus been demonstrated that, although all

landscapes share the same basic structure, patterns of

soil moisture, and consequently plant abundance, differ

on topography shaped by different geomorphic process-

es. Next, nine hillslope types are analyzed to illustrate

how the flow of water moves and the effectiveness of

simple transects to create consistent species tolerance

curves.

Defining the moisture gradient using transects

In this section, we evaluate how well traditional

transect methods define moisture gradients in each of

the nine hillslope categories by examining changes in the

topographic indexes along individual and averaged

transects. The moisture gradient is expected to decrease

from the channel toward the ridgeline. The nine hillslope

categories do not yield any patterns in the topographic

FIG. 4. Creep-dominated landscape: (a) digital elevation model (DEM) colored by topographic index and (b) slope–area plot.


index with distance from the channel for either of the

landscapes because the shapes of the curves for each

transect differ both within and among hillslope types

(see Appendix A). To demonstrate typical patterns, a

sample of nine hillslope transects (one from each

hillslope type) are shown in Figs. 10–12 for the

channel-to-ridgeline (Fig. 10), steepest ascent (Fig. 11),

and steepest descent (Fig. 12) transects. We have chosen

to present individual transects rather than means of

transects, because the individual transects are more

descriptive even though at times there are multiple

overlapping lines. Appendix B shows the nine hillslope

transects of Figs. 10–12 combined by taking the means

and standard deviations. The means shown in Appendix

B do not indicate any discernible trend in the data.

However, as we will show, this is because individual

transects sample points with variable slope and contrib-

uting areas, which results in a large amount of variation

in not only topographic index values at similar

distances, but also variable rates of change of the index

from one hillslope transect to the next.

The topographic index is variable along each channel-

to-ridgeline transect but tends to decrease as distance

from the channel increases for both landscapes (Fig. 10).

FIG. 5. Overland-flow-dominated landscape: (a) digital elevation model colored by topographic index and (b) slope–area plot.

FIG. 6. Probability density function of the (a) topographic index, (b) contributing areas, and (c) slopes for the landscapes. Thesolid black line represents the creep-dominated landscape, and the gray line represents the overland-flow-dominated landscape.Note that the contributing area plot has been cut off at 3000 m2 because both distributions are very right-skewed.


However, on several transects the topographic index

values increase and decrease repeatedly along the length

of the transect (i.e., do not consistently decrease).

One reason that the transects have variable topo-

graphic index values with distance from the channel is

because they do not always follow the flow paths of

water down slopes (Fig. 2). For example, a transect may

intercept a region where the flow of water is largely

parallel or divergent, resulting in a low contributing

area, and then further upslope intercept a region where

several flow paths converge, resulting in a high

contributing area (Appendix C: Fig. C1). This could

result in a lower topographic index closer to the channel

than at a point closer to the ridgeline. Variable slope

gradients (Fig. C1) can also contribute to fluctuations in

the index by a similar argument.

We next test how the topographic index (soil

moisture) changes depending whether the steepest path

transect goes from channel to ridge or vice versa. The

steepest ascent transects give topographic indexes (Fig.

11) that are mid to low values for the first couple

hundred meters from the channel and lower values after

that point. Similar to the channel-to-ridgeline method,

the topographic index of most of the steepest ascent

transects increase and decrease over the first couple

hundred meters, but then tend to level out around index

values of 8–9.

The steepest ascent transects tend to have low

topographic index values along most of their distance

because the path of steepest ascent is not likely to be the

same as the path of steepest descent that water tends to

follow (Fig. 11; see flow lines in Fig. 2). Flow paths tend

to follow directions of steepest descent and, in areas of

convergence, will accumulate in regions of low elevation.

Taking a path of steepest ascent tends to avoid these

regions because, while descent seeks out low elevation

values, ascent seeks out high elevations. Therefore the

contributing areas at points along a path of steepest

ascent tend to be low (Appendix C: Fig. C2). Slope

gradients are also variable (Fig. C2) and contribute to

fluctuations in the index.

The steepest descent transects tend to show an

increase in the topographic index in a convex pattern

as the distance to the channel decreases along transects

on both landscapes (Fig. 12). Some of the transects do

show increases and decreases in the topographic index,

but these oscillations are not as pronounced as in the

other two transect methods.

The steepest descent transects tend to increase in a

convex form from the ridgeline to the channel because

these transects consider the flow paths of water (Fig. 2).

However, these transects also have some increases and

decreases in the topographic index. This is due in part to

variable slope gradients and also to divergent flow

(Appendix C: Fig. C3). The steepest descent transect

FIG. 7. Proportion ([no. hillslopes of each type]/[total no. hillslopes in DEM]) of each hillslope type in the landscapes.


method produces the best gradient of topographic

indices because the method tends to follow flow lines.

Constructing plant tolerance curves using transects

In this section, we evaluate how the different transect

methods affect the shape of the resulting ‘‘mesic’’ plant

tolerance curves. Moisture is assumed to be a decreasing

function of distance from channel to ridge. We know

from the previous section that this is not necessarily the

case, at least with respect to using the topographic index

as a measure of soil moisture. As is the case for the

transect data, individual tolerance curves are presented

rather than only means because, as discussed in the

previous section, it does not make sense to produce

average tolerance curves for transects with such variable

topographic index values at similar distances. However,

as this has often been done in direct gradient analysis

(Gauch 1982), we present mean values with standard

deviations in Appendix D. Figs. 13–15 show the

individual tolerance curves corresponding to the same

transects taken on the same nine hillslopes (one

representative from each hillslope type) as presented in

Figs. 10–12. Appendix E shows the individual tolerance

curves categorized by hillslope type. As discussed in the

previous section on moisture gradients defined by

transects, there are no patterns apparent among the

nine hillslope groups transects separated into the nine

groups, and consequently no patterns are revealed in the

individual tolerance curves among the nine groups.

Mesic plant tolerance curves are variable for each

channel-to-ridgeline transect (Fig. 13a and d). Normal-

izing transect distances by total transect distance, one

technique used in gradient analysis (Racine 1971), does

little to resolve this variation (Fig. 13b and e). When

transects are combined by binning data points (accord-

ing to the square root of sample size) and taking means

for each of the 9 hillslope transects within each transect

category and landscape, modes and degree of modality

of the average tolerance curve also vary (Fig. 13c and f ).

Binning is necessary due to the differences in normalized

distances for each individual tolerance curve that results

from differences in total transect length.

Tolerance curves obtained from the steepest ascent

transects (Fig. 14a and d) tend to have values that

oscillate around mid- to high abundance for the mesic

plant species over several hundred meters for the

overland-flow- and creep-dominated landscapes.

Normalization of distance (Fig. 14b and e) fails to

resolve much of the variation in the curves for both

landscapes. Tolerance curves obtained by binning for

each of the nine hillslope types tends to indicate

unimodal curves for the creep-dominated landscape

(Fig. 14c). The curves of the overland-flow-dominated

landscape remain at mid- to high values across the entire

gradient (Fig. 14f ).

The tolerance curves obtained using the path of

steepest descent method (Fig. 15a and d) suggest that

only part of the species range is sampled (refer back to

the Gaussian curve of Fig. 3) on transects from each of

the nine hillslope types for both the landscapes.

Tolerance curves obtained from the creep-dominated

landscape appear to sample approximately half of the

species range (Fig. 15a). However, tolerance curves from

the drier overland-flow-dominated landscape appear to

sample more than half the species range such that the

drier end of the species distribution is more fully

FIG. 8. Distributions of a mesic-tolerant plant (Fig. 3) onthe (a) creep-dominated and (b) overland-flow-dominatedlandscape.


FIG. 9. Mesic plant tolerance curves (Fig. 3) for one of each of the nine hillslope types when the topographic index is plottedagainst plant abundance (no. plants/grid cell) for the (a, d) channel to ridgeline, (b, e) ascent, and (c, f ) descent methods for the (a–c) creep-dominated landscape and (d–f ) overland-flow-dominated landscape. Plus signs (þ) represent divergent-convex hillslopetransects, circles represent divergent-planar, eight-pointed stars represent divergent-concave, x’s represent parallel-convex, squaresrepresent parallel-planar, diamonds represent parallel-concave, upward triangles represent convergent-convex, downward trianglesrepresent convergent-planar, and right-pointing triangles represent convergent-concave hillslope transects.

FIG. 10. Channel-to-ridgeline transects topographic index as a function of distance from the channel for nine hillslope types(one of each hillslope type) for the (a) creep-dominated and (b) overland-flow-dominated landscape. Symbols are as in Fig. 9.


FIG. 11. Ascent transects topographic index as a function of distance from the channel for nine hillslope types (one of eachhillslope type) for the (a) creep-dominated and (b) overland-flow-dominated landscape. Symbols are as in Fig. 9.

FIG. 12. Descent transects topographic index as a function of distance from the channel for nine hillslope types (one of eachhillslope type) for the (a) creep-dominated and (b) overland-flow-dominated landscape. Symbols are as in Fig. 9.


FIG. 13. Channel to ridgeline transects mesic plant tolerance curves (Fig. 3) for one of each of the nine hillslope types where soilmoisture is approximated by (a, d) distance from the channel (wet) to the ridgeline (dry), (b, e) normalized distance (distance fromchannel divided by total transect distance), and (c, f ) by binning distance for the (a–c) creep-dominated and (d–f ) overland-flow-dominated landscape. Symbols are as in Fig. 9. The solid line in panels (c) and (f ) indicates the mean obtained from binning usingall data points (shown as dots) for the nine transects. Plant abundance units are number of plants per grid cell.

FIG. 14. Ascent transects for mesic plant tolerance curves (Fig. 3) for one of each of the nine hillslope types where soil moistureis approximated by (a, d) distance from the channel (wet) to the ridgeline (dry), (b, e) normalized distance (distance from channeldivided by total transect distance), and by binning distance for the (a–c) creep-dominated and (d–f ) overland-flow-dominatedlandscape. Symbols are as in Fig. 9. The solid line in panels (c) and (f ) indicates the mean obtained from binning using all datapoints (shown as dots) for the nine transects.


represented (Fig. 15d). Ranges vary from transect to

transect, as does the position of the mode, which is

located near the ridge (dry end of the gradient) for all

tolerance curves (Fig. 15a and d). Distance normaliza-

tion does not resolve much of the variation in transects

for the landscapes (Fig. 15b and e). Binning results in

variable modes, modality, and skew for both landscapes

(Fig. 15c and f; see also Appendix D, E).

DISCUSSION AND CONCLUSIONS

The purpose of this study was twofold. First, to

demonstrate the role geomorphic processes have on

landscape structure and the subsequent implications for

soil moisture and the shaping of plant abundance and

distribution; and second, to evaluate the ability of three

direct gradient analysis techniques to produce consistent

moisture gradients on two different landscapes created

by different combinations of geomorphic processes.

This study demonstrated that plant abundance and

distribution depend on the template created by geomor-

phic processes. While Hack and Goodlett (1960) first

considered the role of topography in determining soil

moisture via contributing areas and related this to

vegetation patterns, we have taken this a step further to

show how the geomorphic processes at work on

landscapes contribute to the plant communities found

on them. We have shown how soil moisture distributions

differed between two landscapes that were shaped by

different geomorphic processes and how subsequently

plant distribution and abundance could also differ.

The most notable effect the two different landscapes

had on the tolerance curves was through the range of the

topographic index represented on each landscape. The

overland-flow-dominated landscape lacked the driest

values of the mesic species range. The creep-dominated

landscape had only the wetter half of the mesic species

range. This resulted in nearly complete curves on the

overland-flow-dominated landscape (Fig. 9d–f ), and

partial curves for the creep-dominated landscape (Fig.

9a–c) when the topographic index was used to construct

the moisture gradient. Thus landscapes shaped by

different geomorphic processes exhibited different dis-

tribution and abundances for the same plant.

Furthermore, results in this study found that transects

from channel to ridgeline do not consider landscape

curvature or hillslope length and therefore do not give

consistent moisture gradients. These transects often

intersect flow paths that have converged and/or diverged

with other paths, each of varying lengths, and, as a

result, soil moisture along them is likely to fluctuate in a

nonlinear way (Figs. 2 and 10). A path of steepest ascent

was also shown to inconsistently define the moisture

gradient because it preferentially seeks out high eleva-

tions whereas flow paths seek out low elevations (Figs. 2

and 11).

Transects taken using a path of steepest descent do

follow flow paths, and therefore contributing area

FIG. 15. Descent transects for mesic plant tolerance curves (Fig. 3) for one of each of the nine hillslope types where soilmoisture is approximated by (a, d) distance from the channel (wet) to the ridgeline (dry), (b, e) normalized distance (distance fromchannel divided by total transect distance), and (c, f ) binning distance for the (a–c) creep-dominated and (d–f ) overland-flow-dominated landscape. Symbols are as in Fig. 9. The solid line in panels (c) and (f ) indicates the mean obtained from binning usingall data points (shown as dots) for the nine transects.


tended to increase toward the channel and gave more

consistent moisture gradients than the other two

transect methods (Figs. 2 and 12). As a result, moisture

gradients obtained by steepest descent transects were not

as variable. However, flow path divergence means that

contributing area can decrease as well as increase along

one flow path toward the channel and potentially cause

decreases in soil moisture. Furthermore, slope gradient

is also variable, adding to fluctuations in soil moisture

(Figs. 4b, 5b, and Appendix A).

Direct gradient analysis studies sometimes average

transects or categories to construct tolerance curves

(Whittaker 1967). For transects, this assumes that soil

moisture values are comparable at similar distances

upslope. However, variable hillslope lengths and slope

gradients make this unlikely. For example, consider

transects taken on two hillslopes with parallel flow and

equal slope gradients but different lengths. Topographic

index values will differ at the same distance from the

channel because one hillslope has a greater contributing

area to that point than the other. Alternatively two

hillslopes with parallel flow and same lengths but

different slope gradients will differ in their ability to

hold water and therefore will also differ in their

topographic indexes. Furthermore, rates of change of

the topographic index along transects in the above

examples will differ. Convergence, divergence, and

variable slope gradients on hillslopes may introduce

further variation. As a result, it is very unlikely that

conditions would be such that transects from two

different hillslopes could be considered to have equal

representations of the environmental gradient. Normal-

izing transects does little to resolve this variation (Figs.

13, 14, and 15).

As a consequence of different topographic index

values and rates of change of the values along transects,

plant tolerance curves differed from transect to transect

for all three transect methods (Figs. 13, 14, and 15).

However, the steepest descent method did tend to

produce tolerance curves most similar to the original

curve (Fig. 3) distributed on the landscapes. Thus, when

the topographic index fluctuated, the tolerance curves

were different. Different index values on different

transects but at the same distance from the channel

resulted in different modes (Figs. 10–15). Furthermore,

different rates of change of the topographic index

influenced skew in the tolerance curves (Figs. 10–15).

Plant abundance is primarily determined by soil

moisture (e.g., Curtis and McIntosh 1951, Whittaker

1956, Waring and Major 1964, Bridge and Johnson

2000, Zinko et al. 2005, and many others) and therefore

as demonstrated in this study, the location of any plant

will be dependent in part on the structure of the

landscape as determined by the geomorphic processes.

Using transects or categories, as in traditional direct

gradient analysis, will result in tolerance curves that

have variable modes, ranges, and skew from one

transect to the next because the one-dimensional

definition of the soil moisture gradient is inconsistent

with the processes of water movement. Since soil

moisture is determined by processes that act in three

dimensions, a similarity index such as the topographic

index should result in better agreement of tolerance

curves than traditional one-dimensional transects or

categories when extended to actual landscapes.

We should seek to understand the processes creating

the gradient. The moisture gradient is not simply wet to

dry from channel to ridge because water moves

downhill. While techniques in direct gradient analysis

have advanced from the first approximations to

represent a decrease in soil moisture (e.g., ter Braak

and Prentice 1988, De’ath 1999, Lookingbill and Urban

2005), landscapes are complex structures and hillslope

curvature and lengths dependent on geomorphic pro-

cesses dictate where water will collect and how much.

Slope gradient will dictate the capacity for water to

remain at any one point. Defining the gradient by using

a topographic index that considers these variations

makes sites directly comparable, which is not the case

using transect or categorical methods.

The implications for plant ecology are interesting.

Often as ecologists we search for patterns to aid in

interpretation, hence the attractiveness of dividing

landscapes into valley bottoms and ridges (Whittaker

1956, 1967), or hillslopes sharing similar profile and

contour curvatures (Aryal et al. 2002). If the search for

patterns is extended one step further to processes, we are

able to achieve a better understanding and intuition

regarding plant distributions. As demonstrated in this

study, a process-based understanding allows one to

recognize that landscapes dominated by soil creep with

saturation excess overland flow (characterized by convex

hilltops) may support plants that are more adapted to a

wetter environment than landscapes dominated by

infiltration excess overland flow (characterized by

peaked hilltops in low slope landscapes). Part of the

reason the creep-dominated landscape was wetter than

the overland-flow-dominated landscape may potentially

be the difference between the two types of overland flow.

The creep-dominated landscape had shorter hillslopes

than the overland-flow-dominated landscape; however,

it had a greater median contributing area, which may

indicate that saturation overland flow results in

increased topographic convergence in comparison to

infiltration excess overland flow. The process approach

helps to focus these kinds of questions.

The links between topography and vegetation are very

direct and there has been an increasing interest in the

geomorphological literature concerning the connections

between vegetation and geomorphic processes (e.g.,

Collins et al. 2004, Istanbulluoglu and Bras 2005,

Martin 2007, Collins and Bras 2010). For example, by

adding resistance to counteract shear stress acting on the

landscape, vegetation may contribute to landscapes of

greater relief that are steeper and have lower drainage

densities (i.e., longer hillslopes) (Collins et al. 2004). By


adopting a more interdisciplinary approach and consid-

ering the mechanisms of landscape development, an

improved understanding can be obtained about the

controlling factors of plant communities, and, in turn,

how these communities influence the shape of land-

scapes.

However, there are issues that warrant further

consideration and study when using landscape develop-

ment models to investigate linkages between ecology and

geomorphology. For example, the landscapes developed

in our study do not incorporate the effects of glaciation,

which can have notable effects on topographic structure.

Also, the effect of vegetation is similar over the

landscape, i.e., differences in types of vegetation and

the varying resistances to erosion they contribute have

not been incorporated. Despite recent rapid develop-

ments in the numerical modeling of landscape change,

there remain many unanswered questions because of the

great complexity in the development of landscapes over

large spatial and temporal scales. Finally, ecological

disturbance is not considered in most models of

landscape development (but see Istanbulluoglu and Bras

2005, Martin 2007). Herein, the assumption is made (as

is often made in gradient analysis) that we are interested

in the vegetation at some notable amount of time

following the last disturbance, such that it is not affected

by rapid compositional change (but see Johnson 1981).

Given the multitude of remaining questions about

landscape development and its interactions with ecolo-

gy, this study should be viewed as a modest yet

important beginning to contribute to improved under-

standing.

ACKNOWLEDGMENTS

This research was supported by an NSERC-Discovery Grantand NSERC’s GEOIDE Network of Centres of Excellence. Wethank Greg Tucker and Erkan Istanbulluoglu for theirguidance in using CHILD and for valuable discussions withthem about the connections between topographic structure andecology. We also thank Petro Babak and Olga Chaikin for helpwith Matlab scripts and programming. Finally, comments bytwo anonymous reviewers helped significantly improve the finalpaper.

LITERATURE CITED

Anderson, R. S. 1994. Evolution of the Santa Cruz Mountains,California, through tectonic growth and geomorphic decay.Journal of Geophysical Research 99:161–120.

Anderson, R. S., and S. P. Anderson. 2010. Geomorphology:the mechanics and chemistry of landscapes. CambridgeUniversity Press, Cambridge, UK.

Aryal, S. K., E. M. O’Loughlin, and R. G. Mein. 2002. Asimilarity approach to predict landscape saturation incatchments. Water Resources Research 38:1208.

Austin, M. P. 1987. Models for the analysis of species’ responseto environmental gradients. Vegetatio 69:35–45.

Austin, M. P., R. B. Cunningham, and P. M. Fleming. 1984.New approaches to direct gradient analysis using environ-mental scalars and statistical curve-fitting procedures. Veg-etatio 55:11–27.

Beven, K. J. 2001. Rainfall–runoff modelling—the primer. JohnWiley, West Sussex, UK.

Beven, K., and J. Freer. 2001. A dynamic TOPMODEL.Hydrological Processes 15:1993–2011.

Black, T. A., and D. R. Montgomery. 1991. Sediment transportby burrowing mammals, Marin County, California. EarthSurface Processes and Landforms 16:163–172.

Bogaart, P. W., and M. Guardiola. 2007. The hsB toolkit inMATLAB. Wageningen University, Wageningen, The Neth-erlands.

Bridge, S. R. J., and E. A. Johnson. 2000. Geomorphicprinciples of terrain organization and vegetation gradients.Journal of Vegetation Science 11:57–70.

Collins, D. B. G., and R. L. Bras. 2010. Climatic and ecologicalcontrols of equilibrium drainage density, relief, and channelconcavity in dry lands. Water Resources Research46:W04508.

Collins, D. B. G., R. L. Bras, and G. E. Tucker. 2004. Modelingthe effects of vegetation-erosion coupling on landscapeevolution. Journal of Geophysical Research 109:F03004.

Culling, W. E. H. 1960. Analytical theory of erosion. Journal ofGeology 68:336–344.

Culling, W. E. H. 1963. Soil creep and the development ofhillside slopes. Journal of Geology 71:127–161.

Curtis, J. T., and R. P. McIntosh. 1951. An upland forestcontinuum in the prairie–forest border region of Wisconsin.Ecology 32:476–496.

Davis, W. M. 1892. The convex profile of bad-land divides.Science 20:245.

Davison, C. 1889. On the creeping of the soil-cap through theaction of frost. Geological Magazine 6:255.

Dawes, W. R., and D. Short. 1994. The significance of topologyfor modeling the surface hydrology of fluvial landscapes.Water Resources Research 30:1045–1055.

De’ath, G. 1999. Principal curves: a new technique for indirectand direct gradient analysis. Ecology 80:2237–2253.

Dietrich, W. E., and D. R. Montgomery. 1998. Hillslopes,channels, and landscape scale. Pages 30–60 in G. Sposito,editor. Scale dependence and scale invariance in hydrology.Cambridge University Press, New York, New York, USA.

Dietrich, W. E., and J. T. Perron. 2006. The search for atopographic signature of life. Nature 439:411–418.

Dietrich, W. E., C. J. Wilson, D. R. Montgomery, and J.McKean. 1993. Analysis of erosion thresholds, channelnetworks, and landscape morphology using a digital terrainmodel. Journal of Geology 101:259–278.

Eagleson, P. S. 1978. Climate, soil, and vegetation. 2. Distri-bution of annual precipitation derived from observed stormsequences. Water Resources Research 14:713–721.

Fleming, R. W., and A. M. Johnson. 1975. Rates of seasonalcreep of silty clay soil. Quarterly Journal of EngineeringGeology 8:1–29.

Gauch, H. G. J. 1982. Multivariate analysis in communityecology. Cambridge University Press, Cambridge, UK.

Gilbert, G. K. 1909. The convexity of hilltops. Journal ofGeology 17:344–350.

Grayson, R. B., I. D. Moore, and T. A. McMahon. 1992.Physically based hydrologic modeling. 1. A terrain-basedmodel for investigative purposes. Water Resources Research28:2639–2658.

Grayson, R. B., A. W. Western, and F. H. S. Chiew. 1997.Preferred states in spatial soil moisture patterns: local andnonlocal controls. Water Resources Research 33:2897–2908.

Hack, J. T., and J. C. Goodlett. 1960. Geomorphology andforest ecology of a mountain region in the central Appala-chians. U.S. Geological Survey Professional Paper. 347. U.S.Geological Survey, Reston, Virginia, USA.

Harmon, R. S., and W. W. I. Doe, editors. 2001. Landscapeerosion and evolution modeling. Kluwer Academic/PlenumPublishers, New York, New York, USA.

Hjerdt, K. N., J. J. McDonnell, J. Seibert, and A. Rodhe. 2004.A new topographic index to quantify downslope controls onlocal drainage. Water Resources Research 40:W05602.


Horton, R. E. 1933. The role of infiltration in the hydrologiccycle. Transactions of the American Geophysical Union14:446–460.

Horton, R. E. 1945. Erosional development of streams andtheir drainage basins—hydrophysical approach to quantita-tive morphology. Geological Society of America Bulletin56:275–370.

Howard, A. D. 1994a. Badlands. Pages 213–242 in A. D.Abrahams and A. J. Parsons, editors. Geomorphology ofdesert environments. Chapman and Hall, London, UK.

Howard, A. D. 1994b. A detachment-limited model of drainagebasin evolution. Water Resources Research 30:2261–2285.

Istanbulluoglu, E., and R. L. Bras. 2005. Vegetation-modulatedlandscape evolution: Effects of vegetation on landscapeprocesses, drainage density, and topography. Journal ofGeophysical Research 110:F02012.

Johnson, E. A. 1981. Vegetation organization and dynamics oflichen woodland communities in the Northwest Territories,Canada. Ecology 62:200–215.

Kirkby, M. J. 1975. Hydrograph modelling strategies. Pages69–90 in R. Peel, M. Chisholm, and P. Haggett, editors.Processes in physical and human geography. HeinemannEducational Books, London, UK.

Kirkby, M. J., and R. J. Chorley. 1967. Throughflow, overlandflow and erosion. Hydrological Sciences Journal 12:5–21.

Leopold, L. B., and T. J. Maddock. 1953. The hydraulicgeometry of stream channels and some physiographicimplications. U.S. Geological Survey Professional Paper252. U.S. Geological Survey, Reston, Virginia, USA.

Lookingbill, T. R., and D. L. Urban. 2005. Gradient analysis,the next generation: towards more plant-relevant explanatoryvariables. Canadian Journal of Forest Research 35:1744–1753.

Loucks, O. L. 1962. Ordinating forest communities by means ofenvironmental scalars and phytosociological indices. Ecolog-ical Monographs 32:137–166.

Martin, Y. E. 2007. Wildfire disturbance and shallow landslid-ing in coastal British Columbia over millennial time scales: anumerical modelling study. Catena 69:206–219.

Martin, Y., and M. Church. 1997. Diffusion in landscapedevelopment models: on the nature of basic transportrelations. Earth Surface Processes and Landforms 22:273–279.

Martin, Y., and M. Church. 2004. Numerical modelling oflandscape evolution: geomorphological perspectives. Pro-gress in Physical Geography 28:317–339.

McKean, J. A., W. E. Dietrich, R. C. Finkel, J. R. Southon,and M. W. Caffee. 1993. Quantification of soil productionand downslope creep rates from cosmogenic Be-10 accumu-lations on a hillslope profile. Geology 21:343–346.

Minchin, P. R. 1989. Montane vegetation of the Mt. Fieldmassif, Tasmania: a test of some hypotheses about propertiesof community patterns. Vegetatio 83:97–110.

Moeyersons, J. 1975. An experimental study of pluvialprocesses on granite gruss. Catena 2:289–308.

Montgomery, D. R., and W. E. Dietrich. 1988. Where dochannels begin? Nature 336:232–234.

Montgomery, D. R., and W. E. Dietrich. 1989. Source areas,drainage density, and channel initiation. Water ResourcesResearch 25:1907–1918.

Montgomery, D. R., and W. E. Dietrich. 1992. Channelinitiation and the problem of landscape scale. Science255:826–830.

Montgomery, D. R., and W. E. Dietrich. 1994. Landscapedissection and drainage area–slope thresholds. Pages 221–246in M. J. Kirkby, editor. Process models and theoreticalgeomorphology. John Wiley, New York, New York, USA.

Montgomery, D. R., and E. Foufoula-Georgiou. 1993. Channelnetwork source representation using digital elevation models.Water Resources Research 29:3925–3934.

Oksanen, J., and P. R. Minchin. 2002. Continuum theoryrevisited: what shape are species responses along ecologicalgradients? Ecological Modelling 157:119–129.

O’Loughlin, E. M. 1981. Saturation regions in catchments andtheir relations to soil and topographic properties. Journal ofHydrology 53:229–246.

O’Loughlin, E. M. 1986. Prediction of surface saturation zonesin natural catchments by topographic analysis. WaterResources Research 22:794–804.

Parsons, A. J., A. D. Abrahams, and S. Luk. 1991. Sizecharacteristics of sediment in interrill overland flow on asemiarid hillslope, Southern Arizona. Earth Surface Process-es and Landforms 16:143–152.

Pelletier, J. D. 2008. Quantitative modeling of earth surfaceprocesses. Cambridge University Press, New York, NewYork, USA.

Racine, C. H. 1971. Reproduction of three species of oak inrelation to vegetational and environmental gradients insouthern Blue Ridge. Bulletin of the Torrey Botanical Club98:297–310.

Roering, J. J., P. Almond, P. Tonkin, and J. McKean. 2002.Soil transport driven by biological processes over millennialtime scales. Geology 30:1115–1118.

Roering, J. J., J. W. Kirchner, and W. E. Dietrich. 1999.Evidence for nonlinear, diffusive sediment transport onhillslopes and implications for landscape morphology. WaterResources Research 35:853–870.

Shelford, V. E. 1913. Animal communities in temperateAmerica. University of Chicago Press, Chicago, Illinois,USA.

Simpson, G. G. 1963. Historical science. Pages 24–48 in C.Albritton, editor. The fabric of geology. Freeman, Stanford,California, USA.

Small, E. E., R. S. Anderson, and G. S. Hancock. 1999.Estimates of the rate of regolith production using 10Be and26Al from an alpine hillslope. Geomorphology 27:131–150.

Tarboton, D. G. 1989. The analysis of river basins and channelnetworks using digital terrain data. Dissertation. Massachu-setts Institute of Technology, Cambridge, Massachusetts,USA.

Tarboton, D. G. 1997. A new method for the determination offlow directions and upslope areas in grid digital elevationmodels. Water Resources Research 33:309–319.

ter Braak, C. J. F., and I. C. Prentice. 1988. A theory ofgradient analysis. Advances in Ecological Research 34:271–317.

Thornes, J. B., editor. 1990. Vegetation and erosion: processesand environments. John Wiley and Sons, Chichester, UK.

Troch, P. A., C. Paniconi, and E. Emiel van Loon. 2003.Hillslope-storage Boussinesq model for subsurface flow andvariable source areas along complex hillslopes: 1. Formula-tion and characteristic response. Water Resources Research39. http://dx.doi.org/10.1029/2002WR001728

Tsukamoto, Y., and T. Ohta. 1988. Runoff process on a steepforested slope. Journal of Hydrology 102:165–178.

Tucker, G. E. 2010. CHILD users guide for version R9.4.1.http://csdms.colorado.edu/mediawiki/images/Child_users_guide.pdf

Tucker, G. E., and R. L. Bras. 1998. Hillslope processes,drainage density, and landscape morphology. Water Re-sources Research 34:2751–2764.

Tucker, G. E., and R. L. Bras. 2000. A stochastic approach tomodeling the role of rainfall variability in drainage basinevolution. Water Resources Research 36:1953–1964.

Tucker, G. E., S. T. Lancaster, N. M. Gasparini, and R. L.Bras. 2001. The Channel-Hillslope Integrated LandscapeDevelopment (CHILD) model. Pages 349–388 in R. S.Harmon and W. W. Doe, editors. Landscape erosion andevolution modeling. Kluwer Academic/Plenum Publishers,New York, New York, USA.


Urban, D. L., C. Miller, P. N. Halpin, and N. L. Stephenson.

2000. Forest gradient response in Sierran landscapes: the

physical template. Landscape Ecology 15:603–620.

Wainwright, J., A. J. Parsons, and A. D. Abrahams. 2000. Plot-

scale studies of vegetation, overland flow and erosion

interactions: case studies from Arizona and New Mexico.

Hydrological Processes 14:2921–2943.

Waring, R. H., and J. Major. 1964. Some vegetation of the

California coastal region in relation to gradients of moisture,

nutrients, light, and temperature. Ecological Monographs

34:167–215.

Warntz, W. 1975. Stream ordering and contour mapping.

Journal of Hydrology 25:209–227.

Werner, C. 1988. Formal analysis of ridge and channel patterns

in maturely eroded terrain. Annals of the Association of

American Geographers 78:253–270.

Werner, C. 1991. Several duality theorems for interlockingridge and channel networks. Water Resources Research27:3237–3247.

Whittaker, R. H. 1956. Vegetation of the Great SmokyMountains. Ecological Monographs 26:1–80.

Whittaker, R. H. 1967. Gradient analysis of vegetation.Biological Reviews of the Cambridge Philosophical Society49:207–264.

Willgoose, G., R. L. Bras, and I. Rodrı́guez-Iturbe. 1991. Aphysical explanation of an observed link area–slope relation-ship. Water Resources Research 27:1697–1702.

Yoo, K., R. Amundson, A. M. Heimsath, and W. E. Dietrich.2005. Process-based model linking pocket gopher (Thomomysbottae) activity to sediment transport and soil thickness.Geology 33:917–920.

Zinko, U., J. Seibert, M. Dynesius, and C. Nilsson. 2005. Plantspecies numbers predicted by a topography-based ground-water flow index. Ecosystems 8:430–441.

SUPPLEMENTAL MATERIAL

Appendix A

Topographic index, contributing area, and slope as a function of distance from the channel for the channel-to-ridgeline, ascent,and descent transects within each of the nine different hillslope categories on the creep-dominated and overland-flow-dominatedlandscapes (Ecological Archives M082-015-A1).

Appendix B

Mean topographic index, contributing area, and slope values as a function of distance from the channel for the channel-to-ridgeline, ascent, and descent transects averaged for the nine sample hillslopes (one of each hillslope type) depicted in Figs. 9–11 forthe creep-dominated and overland-flow-dominated landscapes (Ecological Archives M082-015-A2).

Appendix C

Contributing area and slope as a function of distance from the channel for the channel-to-ridgeline, ascent, and descent transectsfor the nine sample hillslopes (one of each hillslope type) depicted in Figs. 8–10 for the creep-dominated and overland-flow-dominated landscapes (Ecological Archives M082-015-A3).

Appendix D

Channel-to-ridgeline, ascent, and descent transects mesic plant tolerance curves (Fig. 2) constructed using one of each of the ninehillslope types and taking the mean where soil moisture is approximated by distance from the channel (wet) to the ridgeline (dry)and normalized distance for the creep-dominated and overland-flow-dominated landscapes (Ecological Archives M082-015-A4).

Appendix E

Channel-to-ridgeline, ascent, and descent transects mesic plant tolerance curves (Fig. 2) for each of the nine different hillslopecategories where soil moisture is approximated by distance from the channel (wet) to the ridgeline (dry), normalized distance fromchannel, or topographic index for the creep-dominated and overland-flow-dominated landscapes (Ecological Archives M082-015-A5).

Supplement

CHILD input files used to create the creep- and flow-dominated landscapes and Matlab scripts used to take transects (EcologicalArchives M082-015-S1).

Data Availability

Data associated with this paper have been deposited in Dryad: http://dx.doi.org/10.5061/dryad.4pf4t


the inﬂuence of geomorphic processes on plant distribution...

Documents