Whipple y Tucker, 1999
Whipple y Tucker, 1999
Whipple y Tucker, 1999
net/publication/259599257
CITATIONS READS
1,738 1,667
2 authors, including:
Gregory E Tucker
University of Colorado Boulder
264 PUBLICATIONS 14,113 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Gregory E Tucker on 08 June 2016.
Gregory E. Tucker
Department
of Civil andEnvironmental
Engineering,
Massachusetts
Instituteof Technology,
Cambridge
All uncertainties
indicate1-sigmaerrorbars.
aAcdeftnedby breakin slope-area
scalingin longitudinal
profiledata(only).
bThe
condition
0 = m/nholds
if andonlyif channels
areinequilibrium
andbothUandK areconstants.
Reported
values
werefit to longprofiledatabetweenAcandthebedrock-alluvial
transitiononly.
lar, carries information about the physics of the erosion [Hack,1973],a valueconsistentwith approximately
loga-
process. rithmicchannelprofilesobservedin nature[Hack,1957].
Several researchershave recently argued that erosion rate However,besidesthe pioneering
workof Suzuki[1982],no
is a function of the ratio of sedimentflux qsto sedimenttrans- comprehensivestudy of the controls on bedrock channel
port capacity qc [Sklar and Dietrich, 1997; Slingerland et width has been done.
al., 1997; Sklar and Dietrich, 1998], which likely varies Combining(2) -(7), the bedrockerosionrate for shear-
with uplift rate, climate,and position in a catchment.A sim- stressdependenterosioncan be written as
ple way to denote this dependencewithin the frameworkof
the stream-powererosionlaw is to write
œ
=rA2aC(1-b)/3S2a/3
(8a)
kb=kef(qs,qc) (3) K= k•k•
2432•(1-o)/3f(q,)C?p•g2•/3
kq (8b)
where ke depends on rock massquality and erosion process
Comparing(8) with (1), it canbe seenthatexponents
rn and n
andj(qs,qc) is an unspecifiedfunction. As argued by Sklar
arerelatedto erosionprocess,hydraulicgeometry,and basin
and Dietrich [1998], the role of sedimentflux (here denoted
hydrologyaccordingto
asj(qs,qc))encapsulates
at leasttwo competing
effects:
(1) ac-
celerated erosion due to an increased number of tools in the
flowand(2) reduced
erosion
asa resultof partialshieldingof m=2ac(1-
b)/3 (9a)
thebedfrom particleimpactandotherprocesses. For the sake n= 2a/3 (9b)
of simplicity and in keepingwith the standardformulationof
the stream-power law we restrictour analysisto the condi- •C --
(9c)
tion of constantkb(for a given lithology), an assumption
incorporatedinto mostlandscapeevolutionmodels. Similarresultsarereadilyfoundfor the unit stream-power
Coupling either (2a) or (2b) and (3) with relations case
describingflow hydraulics,channelgeometry,and basin
hydrologyresultsin a simpleexpressionfor channelerosion K- kek•ak•(1-t•)f(qs)paga (lea)
ratein termsof streamgradientand drainagearea,in the form
of(1). In this analysis,hydrologicand hydraulic variables m=ac(1-b) (10b)
(dischargeQ, flow depthD, flow width W, flow velocity V, (10c)
shearstress'•b,and erosionrateœ)aretakenastime-averaged
quantities,suchthat dischargecanbe takenas a simplefunc- (led)
tion of drainageareaA. Thus it is implicitly assumedthat an
effectivedischargecan be definedthat adequatelyrepresents Thusthe shear-stress andunit stream-power versionsof the
the integratedeffectsof the full-time history of flood dis- erosionlaw differin detail butare notfundamentally
different.
charges [Wolman and Miller, 1960; Willgoose, 1989; Moreover,giventhat the exponenta in (2) is unknown,it
Tuckerand Bras, 1997]. Given this assumption, the internal wouldbe difficult at presentto discriminatebetweenthe unit
relationsare conservationof mass(water) stream-power and shear-stressmodels on the basis of field
data.
dt Uok•-mLn-hmH-n (19)
N/r=-•-
where z is the elevation of the river bed, x is the distance
downstream,and U is the rock uplift rate definedrelative to Note thatby definition,if the rockuplift rate U is steadyand
the erosionalbaselevel. Combining(11) with (1) and em- uniform,the dimensionless uplift rate U, is unity (equations
ployingHack's law [Hack, 1957], (15) and (16)).
The uplift-erosionnumbercanbe immediatelyidentifiedas
A= k•x• (12) the critical dimensionlessgroup governingthe dynamicsof
the bedrockchannelprofile evolution equation(18). More-
wherek, is a dimensionalconstantand h is the reciprocalof over, as with the familiar Reynolds and Froude numbersin
the Hack exponent,showsthatriver profilesare governedby fluid mechanics,dynamicresponses associated
with perturba-
a nonlinearkinematicwave equation[Whitham, 1974]: tions of the suite of variables Uo, K, ka,L, m, and H can be
fully capturedby simply consideringresponsesto perturba-
dt -(x,O
-amxhmn-'l-
I Xc
-<
X-< r tions in the uplift-erosionnumberNE.For instance,changes
in the rock uplift relative to baselevel U are dynamically
equivalentto changesin the coefficientof erosionK. In addi-
withwavespeed-Kk,mXhmS
•4, whereS = I(az/ax)l,
L is the tion, covarianceof empirically determinedK values and the
bedrockchannelstreamlength measuredfromthe divide, and
exponentm (the dimensionsof K depend on m) [Sklar and
the area-length exponent h is seen to vary over a narrow
Dietrich, 1998; Stock and Montgomery, 1999] does not
rangefi'om1.67 to 1.92 [Hack, 1957; Marltan et al., 1996;
complicatethe dynamic behavior of the profile evolution
Rigon et al., 1996].
equation as this effectis encapsulatedwithin the uplift-
Nondimensionalizationof the bedrock river profile evolu-
erosion number. Consideration of steady state conditions
tion equation first requiresthat we write (13) in general
terms:
will reveal the roles of the exponentsh, m, and n in the form
and dynamicsof modeledriver profiles.
z F(U,K,k
=
a,x,dx
dz' ) • t (14)
4. Steady State River Profiles
The right-handsideof (14) hassix independent variablesin In this section we explore the behavior of bedrock chan-
two dimensions(length and time), which thereforecan be nels as predictedby the shear-stress/stream-power model in
writtenasfour independentnondimensional groups. Note orderto draw out the significanceof the issuesoutlined ear-
thattheexponents h, m, and n do not appearas variableson lier in regard to channel profile form,the relationship be-
the right-handsideasthesearepartof the unspecified func- tween equilibrium channel gradient and environmental
tion F. Similarly,the variableXcintroduced earlier(seeFig- controls(climate,lithology,and uplift rate),and the equilib-
ure 1) doesnotappearasthisentersonly as a boundarycon- rium height of mountainranges. In the analysisthese envi-
ditionto theunspecified functionF. In orderto proceedwith ronmentalcontrolsare all representedby the uplift-erosion
thenondimensionalization we introducethreerepresentative numberNE introducedin section3, which can be quantita-
scales(H, L, and Uo)to definethe following dimensionless tively interpreted as either reflecting tectonic forcing
variables:
(throughUo)or climaticandlithologicforcing(throughK).
17,666 WHIPPLE AND TUCKER: DYNAMICS OF THE STREAM-POWER LAW
--½ 1 (21a)
E 0.1 o.1
::] / / ill:
ß-
ß
__-- /
/
z,(x,)-z,(1)-NE1/nu,1/nln(x,)
hm-1(2lb)
'-•
O'
0 ' 01 ' ' ' •'"1 ........ ', ........ ', ....... 0.01
where L is total bedrockstreamlength and z,(1) is the dimen-
0.01 0.1 1 10 100
sionless elevation at the basin outlet (or at the bedrock-
Uplift/Erosion Ratio (N E/[N E] r ) alluvial transition). Equations (21a) and (2lb) arevalid for
X,c _<x, _<1 only, where X,c is the dimensionlessdistance
Figure 2. Sensitivityof dimensionless equilibriumchannel downstream fi'om the divide at which fluvial processes
gradientand dimensionless equilibriumfluvial relief to the become dominant [Montgomery and Foufoula-Georgiou,
uplift-erosion
numberNE as a functionof the slopeexponent
1993] (See Figure 1).
n. In orderto emphasizethe sensitivity to changesin the
uplift-erosionnumberboth dimensionless channelgradient Although calculationsusing the restrictive assumptions
and dimensionless fluvial relief are shown relative to a refer- incorporatedinto (2 la) and (2 lb) are illustrative (Figure3),
encevalue notedwith the subscriptr. Referencevalues in all we stress that nonuniform uplift rates [i.e., Adams, 1985;
figuresare computedwith the parameters
listed in Table 2. Koons, 1989], orographic precipitation [Beaumont et al.,
Note the log-log scale. 1992; Masek et al., 1994b], and systematic downstream
WHIPPLE AND TUCKER: DYNAMICS OF THE STREAM-POWER LAW 17,667
1
variations in sediment loading in streams [Sklar and
Dietrich, 1997; Slingerland et al., 1997; Sklar and Dietrich,
1998] all play potentially important roles in even the sim-
plest realistic scenario. For steady,uniform uplift, constant
• '"•0.6 =
coefficientof erosion,and constantareaand slope exponents
(m and n) the ratio hm/n dictatesthe equilibrium formof river
o., profiles(Figure 3a). For typical valuesof h (1.67 < h < 1.92)
and the rn/n ratio (0.35 < rn/n < 0.6) the ratio hrn/n ranges
fi'om0.58 to 1,15, and predicted river profiles are approxi-
• 0.2
Z mately logarithmic,as documentedby Hack [1957] and illus-
trated in Figure 3b.
0
Althoughthe steadystateform (concavity)of river profiles
0 0.2 0.4 0.6 0.8 1
subjectedto the constraintsoutlined above is predictedto be
relatively insensitiveto the multivariate controls on bedrock
DimensionlessDistance (x I L )
erosion processes,any deviations fi'omsteady state, or any
systematicdownstreamchangesin uplift rate (e.g., tilting) or
3 1\.. erodibility (e.g., change in dominant process, sediment
supply, and cover)will complicatethe interpretation of river
•"•..... ß...... schematic
ridge profile data in terms of the m/n ratio. For instance, where
•' I!\. --- '-.. / profile uplift rate increasesmonotonically downstream(dU/dx > 0;
• 2 ul-:"•
• ..... \ '-,•-'---_1•
I', %
.... x •,R hc --",. back-tilting), profile concavity will be diminished and vice
versawhere uplift rate decreasesdownstream(dU/dx < 0). In
addition, spatially variable controls on erodibility (K)may
play an important role in channel profile form [Sklar and
o
Dietrich, 1998]. Because of such difficulties, Seidl and
0x c 5 10 15 20 25 Dietrich [1992] proposeda methodfor extractingrn/n ratios
from differencesof channelandtributarygradientsat tributary
Distance from Divide (km) junctions. Although their analysis did not accountfor pos-
sible differences in alluvial cover or channel width between
tributary channels, their method requires no assumptions
4
regarding steady state or equilibrium conditions. However,
their finding of m/n = 1 for streams in the Oregon Coast
• n =1
Range is at odds with other data (i.e., reasonable values for
mm' 3
exponentsin (6) and (7) and logarithmic channel profiles)
and hasnot yet been explained.
4.2.2. Equilibrium fluvial relief and the height of
mountain ranges. Over long timescalesthe height of moun-
• 'g 1 tain rangesis limited by either crustal strength[e.g.,Molnar
and Lyon-Caen, 1988; Bird, 1991; Masek et al., 1994a] or by
a balancebetweenrock uplift and erosion[e.g.,Adams, 1985;
0
Koons, 1989], whichever is more restrictive. In the case
0 25 50 75 100
where crustalstrengthis not limiting, the equilibrium height
c Range Half Width (L) (km) of a fluvially sculptedmountainrangeis dictatedby four fun-
dmental geomorphiccontrols: (1) rangewidth, (2) longitu-
Figure3. Equilibriumchannelprofilesand fluvial relief dinal profiles of transversebedrock streams,(3) the length
(assuming spatiallyconstantU, K, m, andn). (a) Longitudi- and gradientof colluvial channelsabovethe fluvial network,
nalprofileconcavity is controlledbythehm/nratio(concav- and (4)the length (= drainagedensity) and gradient of hill-
ity index). Exponenth is held constantat the observedvalue slopes(see Figure 1).
of 1.67 [Hack, 1957]. Natural channelsare approximately Equilibrium fluvial bedrockchannelreliefRf is given by
logarithmic in form: consistentwith h = 1.67- 1.92 and m/n = the difference between the elevation at the headwater of the
1/2. (b) At steadystatethe topographic envelopeof moun-
tain rangesis setby the longitudinalprofilesof the stream fluvial channel(i.e., at x = Xc)and the elevation of the basin
outlet or the bedrock-alluvial channel transition (i.e., at x =
networkplus the relief on hillslopes and colluvial channels
(Rhc).A theoreticalprofile,in dimensional form(trunk chan- L). In termsof dimensionless
variables,from(21), fluvial bed-
nel' solid black line; tributary channels(projectedinto rock channelrelief is given by
plane)'dashed blacklines),computed with U = 5xl 0-3ma-], K
= 1.2x10
-s,andn= 1(other
divide-to-outlet
parameters
longitudinal
aslistedin Table2) is
shown(forXc_(x _(L) for directcomparison againstdatafor a
profile(trunkchannelonly) for
g,f
:NE1/ns,1/nll-•!-l(1
-X,c
1-hm/n) hrn
an intermediate-sized basinin northernTaiwan (Table 1). ½1
(22a)
Stairstepsin theprofilefromTaiwanprobablyreflectnoisein n
Unsurprisingly, equilibrium fluvial relief varies in exactly in the responseof hillslopes to channel incision [e.g.,
the sameway with the uplift-erosion numberNE as does the Fernandes and Dietrich, 1997], nor any feedbacksdue to
equilibrium channel gradient (see Figure 2). Thus the slope eitherinteractionbetweenchannelsand hillslopes [Schumm
exponentn in (1) emergesonce again as a critical unknown. and Parker, 1973; Schumm,1979] or coupling with down-
Recall that in the caseof uniform rock uplift the dimension- streamdepositional systems [e.g., Humphrey and Heller,
less uplift rate U, is unity, and the uplift-erosion number 1995].
fully capturesthe dependenceof fluvial relief on environ- Before deriving results for a dimensionlessresponse
mental controls(lithology, climate, tectonics,and basin size). timescaleT, it is necessaryto return to the dimensional
Thus fluvial relief depends on rock uplift rate to the 1/n analysispresentedearlier. We arguedthat the representative
power,the coefficientof erosionto the -1/n power, and stream rock uplift rate Uo was the appropriatescaling term for the
length to a small power (-0.15 - 0.42 for typical values of rate of changeof channelbed elevation(dz/dt). However, we
hm/n), as illustratedin Figures 2 and 3c. In plotting Figure could equally well have introduceda characteristictimescale
3c it is assumedthat the largest transversedrainages(length T such that
L) scalewith rangehalf width.
Equation (22) makesthe direct prediction that all else
being equal, greaterrelief is expectedfor lower values of the
'•7=-•(dt,• =Uødt, (23a)
coefficientof erosionK. That is, greaterrelief is expectedfor where
U.I ,':•
!
fo(U,K,x,n)
sudden
climatic setting?, and (2) over what timescaleswill a river 200/"
|',
.... ""••..b.a.s.e
knickpoint ......
level
drop
systemreturn to equilibrium following a changein tectonic 0/; I I I
or climatic conditions? We focus our discussion on tectonic
0x c 1000 2000 3000 L 4000
perturbations, but the analysis is not significantly different
for suddenchangesin climate. Distance (m)
We considertwo typesof tectonic perturbation away from
Figure4. Schematic
illustrationof knickpointmigrationin
a base steadystate:(1)a single, suddenfall in base level and
responseto a sudden base level fall (coseismic,eustatic,
(2) a step function increase in uplift rate. In all cases we streamcapture,etc.)on a channelotherwisein a steadystate
assumethat K, U, m, and n do not vary along the river profile. condition.KnickpointmigrationspeedC, is givenby (26).
We do not consider any time lags associated with an Total knickpoint propagationtime fromL to x½definesthe
isostatic responseto denudationalunloading, any time lags systemresponsetime.
WHIPPLE AND TUCKER: DYNAMICS OF THE STREAM-POWER LAW 17,669
100.00
Ce,
=_NE-lx,
hm(
dg*)ln-1
•,•,-, (26)
Thus(18) dictatesthat to firstordera suddendrop in base
level will producea knickpointthat propagatesupstreamat
• I...lO.OO n=i/3
n --
o E
aneverdecreasing ratefromthepointof disturbance,without '• i: •.oo
attenuation (Figure4). However,thedependence of the wave
speed onchannel gradientforn ½1 maycausechanges to the E =
shapeof the kinematicwave, possiblyproducingshock i5o o.o
wavesthat alter the rate of translation [Whitham, 1974]. For
this reasonthe derivationpresentedhereshouldbe consid-
eredpreliminaryin natureand only valid for infinitesimal, 0.01
(Ce*
)s
s=-NE
-1/nU*l-1/nx*hm/n
(27)
To a first approximation,
dimensionless response
time T,b 0.10
that have not yet respondedto rapid adjustmentsof the chan- Since the upper reach of the fluvial channel does not
nel system. Further, a field-testableprediction of this model respondto the changein uplift rate until the wave of erosion
is the location of knickpoints correlatedwith known (co- reachesthispoint,from(11) thetimerateof changeof the ele-
seismicor eustatic) base level drops [e.g., see Rosenbloom vationof the fluvial channeltips Z(Xc)is given by the differ-
and Anderson, 1994]. encebetweenthenewlyimposed
uplift rate Uf andthe previ-
ouslyestablishederosionrate (ci(Xc)):
5.2. Sudden Increase in Uplift Rate
Responsetime for a step function increasein uplift rate Tt•
is derivedin a differentmanner. Application of the kinematic Rearrangingand noting that for perturbationaway froman
wave solution is not straightforwardin this casebecausethis initialsteadystatetheerosionrateat the fluvialchanneltips
problementails a significantdeviation fromthe base steady is equalto the initial uplift rate(c•(Xc)=UO, we obtain the
state and becausethe channel is subjectedto a sustainedrate simple result:
of base level fall, rather than a discrete perturbation. In fact,
the derivation presentedhere is more robust than the kine-
matic wave solution above as it is not subject to any uncer-
tainties related to knickpoint shape evolution during
Tu= Uf - Ui (32)
upstreammigration. Using(32), the systemresponse
timescale
canbe directly
An increasein uplift rate initiates a wave of erosion that estimatedfrom field observations, providedcertainrestrictive
propagatesupstream(Figure 6). Equilibrium within the flu-
conditionsare met: (1) adjacentterranesof similarclimateand
vial system is reached when the wave of erosion reachesthe lithologyare experiencing different,
knownrockuplift rates,
fluvial channeltips (x = Xc),as argued above. We reasonthat and(2) channelprofilesappearto haveadjustedto imposed
arrival of this migratingknickzoneat the fluvial channeltips uplift rates;profileshavesmoothlogarithmicformswith no
coincides with the time that they attain their final elevation indicationof an active,propagating knickzone[Snyderand
(ZAXc)).In other words, the migratingknickzonedefinesthe Whipple,1998; Snyderet al., submittedmanuscript, 1999].
Theseconditionswill rarelybe met, however,and it is useful
boundary between a downstreamsegmentof the profile that
has achieved its final equilibrium gradient and an upstream to write (32) in termsof nondimensionalrock uplift rate,
segmentthat has not yet steepenedin responseto the change systemlength,and rock erosionparameters,
using(24) for
dimensionless timescale:
in rock uplift rate. Barring the effectsof numericaldiffusion,
this is precisely the behavior observed in numerical solu-
tions of the profile evolution equation (18). Thus response
time is set by the time required to elevate the fluvial channel U,f(]_$) (33a)
tips from an initial equilibrium position (zi(xc)) to a final
equilibrium position (Figure 6): Si
•f=f (33b)
(xc)- + (30)whereUo/ischosenasthereferencerockupliftrate.
Using (21) to determine
z,i(X,c)andz,•(X,c),substituting
into(33), andrearranging
givesan algebraicrelationshipfor
dimensionless
responsetimescaleT,t•:
4
1-hrn/
n
E
•
zt(x
c)'••final
steady
state
profiledz/dt =Uf '•i =Uf -Ui
E ta,f
o
n
• 1 (34a)
• of erosion
z,(xc)
•,.j .•..•..œ,• -"ßrE1/nrr
=
l/n-1
•'*f c(1_fl/n)--=
lnx, hm1 (34b)
T*u (l-f) n
0 5 10 15 20 25 Note that for the case of uniform rock uplift rate treated here
the dimensionlessuplift rate term U,/is unity. As in the
Distance from Divide (kin)
derivation of Tb,we have assumedblock uplift (dU/dx = O)
Figure6. Schematic illustrationof the transientresponse andspatiallyuniform coefficientof erosion and erosion proc-
(solidline)of anequilibrium channel(shaded line)to a sud- ess(dK/dx = 0; n and m are constants).For Ui = 0 the initial
den increasein rock uplift rate used in the derivation of
condition is a horizontal plane, and the erosion rate at the
response timescaleTs. Rockuplffi rates•e indicatedby fluvial channel headwater is zero until the wave of erosion
solid arrows;headwatererosionrates•e indicatedby
shadedarrows.An increasein uplffirateinstigates has swept throughthe entire fluvial system. In this casethe
a waveof
erosion that propagates upstream. Channel headwater responsetime is simply the quotientof the equilibrium eleva-
reaches•e uplffiedbut do not responduntil the wave of ero- tion above the base level of the channel at x = Xc and the
sionreaches them. Accordingly,z(xO increasesat a constant uplift rate Uo/. Interestingly, (34) reducesto the kinematic
rate (dz/dt = U•- UO until a new equilibriumis reached wave solution (equation(28)) in this case (Ui = 0; f= 0). In
(dashedshadedline). otherwords, it takesas long for a discretebaselevel fall to be
WHIPPLE AND TUCKER: DYNAMICS OF THE STREAM-POWER LAW 17,671
Fixed Final Uplift Rate (U f ) than with the coefficientof erosionK. For this reason,
responsetime is plotted in dimensionalformas a function of
10
E ....... n = 2/3 therockupliftraterelativeto a reference
condition
(Figure7).
Additionally, the appearanceof both initial and final rock
•n=l
uplift rates in both the numerator and the denominator
.... n=2
(throughfand NE) indicatesthat there is an additional level
of complexityin the dependence of responsetime on rock
uplift rates. Accordingly,Figure7a is plotted to illustrate
therelationship
between
response
timeandfinalupliftrateUf
andFigure7b illustratesthe relationship
betweenresponse
o time andinitial uplift rate Ui.
Somenonintuitiveeffects
arerevealed
in the relationship
0.1 i i i i i iiI I i i i i i i •1 betweenthe magnitudeof the changein uplift rate and
0.01 0.1 responsetime (Figure7). As seenin the caseof a suddenbase
levelfall,forn = 1, responsetimeis independent of both the
a Uplift Rate Ratio (f =U / IU f ) finalupliftrateUf andthemagnitude of thechange in uplift
rate(Figures 7a and7b). Interestingly,
forthecasen > 1, sys-
temresponse time decreasesfor smallerchanges in uplift rate
Fixed Initial Uplift Rate (U • ) (i.e.,asUdUfapproaches 1)whereUfisheldconstant (Figure
lO 7a) but actuallyincreases for smallerchanges in uplift rate
2/3 whereU• is held constant(Figure7b). Conversely, for the
•n=l
caseof n < 1, response timeincreases forsmallerchanges in
upliftratewhereUf is heldconstant(equation(34); Figure
7a) anddecreases for smallerchangesin uplift ratewhereU• is
held constant(equation(34); Figure 7b). This yields the
_
-
seeminglyodd result that for n < 1 it takes significantly
longerto adjustto a minor changein uplift rate than it does
to raisetheentirerangestartingfrom a horizontalplane. The
reasonfor thisis twofold:(1) the relationshipbetweenuplift
rate and equilibrium channel headwater elevation is non-
0.1 i i i , i i ii I i i i i i i i i linearin the slopeexponentn (equation(21)), and(2) the rate
0.01 0.1
at which the channelheadwateris elevateddependson the
initial slopeat x = Xc.
Uplift Rate Ratio (f =U / I U f ) Simplifying (34) and writing it in dimensionalform, we
seethat the changein channelheadwaterelevationrequired
Figure 7. Timescaleof responseto a step function increasein with an increasein uplift rate scalesas
uplift rate (Tu; assumingspatially constant U, K, m, and n).
Referencevaluesof rock uplift rate Ur and the responsetime-
scaleTurusedto normalizethe plots are given by or derived
fromparametervalues reported in Table 2. Note that if the while the rate of changeof channelheadwaterelevation
initial uplift rate is zero (Ui = 0), the responsetime is given scales
withthedifference
between
finalandinitial rockuplift
by the base level fall solution (Figure 5). (a) Sensitivity of rates'
responsetime to the magnitudeof the increasein uplift rate
(UdUf) for the casewherethe final uplift rate Uf is held con-
stant. Responsetime increaseswith U• (increasingUdUf) for dt oc
Uf - Ui (36)
n < 1, is independentof Ui for n = 1, and decreaseswith U• for
n > 1. (b) Sensitivityof responsetime to the magnitudeof the This findinghas importantimplicationsfor the differing
increasein uplift rate (UdUf) for the casewhere the initial dynamic responseof landscapes (both real and simulated)
uplift rate U• is held ccnstant. Responsetime increaseswith etchedby differentsetsof erosional
processes (e.g.,abrasion
Uf (decreasingUdUf)for n < 1, is independent of Uf for n =1, by suspended load,abrasion by saltationload,andpluck-
anddecreases with Uf for n > 1. ing),throughtheircontrolof theslopeexponent n.
group, here termedthe uplift-erosionnumber,encapsulates Which processis dominant under what conditions? What
the dependenceof predicted erosion rates on tectonic, are the appropriateforms of (2) and (3)? Over what distance
lithologic,andclimaticvariables.Our sensitivityanalysisof is an appropriategradient measured? Similarly, the need to
the dimensionlessriver profile evolution equation (18) integrateover the full spectrumof flood dischargesto derive a
revealsthat boththe magnitudeand timescaleof the bedrock meaningfullong-term average rate is a problem [Willgoose,
channelresponseto an imposedtectonicor climaticforcing 1989; Tuckerand Bras, 1997]. One approachthat may help
arelargelygovernedby the uplift-erosionnumberraisedto a to bridge this gap is to pursue small-scaleprocess-oriented
power determinedby the slope exponentn in the stream- field studies in conjunction with reach-scalemodeling
powererosionlaw (equation(1)). Also, the muchdiscussed studies, pursuing top-down and bottom-up approachesin
m/n ratio is shown to influencesignificantly the shape(and concert. Field areas encompassing known differences in
therebyresponsetimescale)ofriver profiles. However,the either climate, lithology, or uplift histories but similar in
m/n ratio neither influencesthe sensitivity of channel gradi- other respectswill be critical to such studies [e.g.,Merritts
ent, relief, or responsetimescaleto changesin the uplift- and Vincent, 1989; Snyder and Whipple, 1998; Snyder et al.,
erosionnumbernorthe dependence of this sensitivity on the submitted manuscript,1999]. Moreover, field areas where a
slopeexponent n. Furthermore,
for thebroadsubsetof fluvial transientresponseto a recent climatic or tectonic perturba-
erosionprocesses adequatelydescribedby (2)- (7)the m/n tion from a known initial condition can be studied would be
ratio is shownto be restrictedto a narrow range(0.35 - 0.6). most advantageousbecauseo?the sensitivity of predicted
The slopeexponentn, whichemerges
fromtheanalysisas a transientresponsesto critical model parameters.
critical unknown, has been shown above to be directly
related to the degree of nonlinearity in the relationship Notation
betweenerosionrate and shearstress(or streampower). Thus
the dominant fluvial erosion processesdirectly and pro- Variables
foundly influencethe dynamicbehaviorof fluvial bedrock vertical
erosion
rate[LT-•].
streamchannels. Clearly, it is not satisfactorysimply to density
ofwater[ML-3].
assumethat erosion is linearly related to shear stress (or basalshear
stress
[ML'•T'2].
streampower); the relationship between channel gradient upstreamdrainage
area[L2].
and erosionrate and its potentialvariation betweenfield set- critical upstreamdrainagearea for fluvial erosion
tingsarefirst-orderproblemsin tectonicgeomorphology. processes
[L2].
As the slopeexponentn is directly relatedto the physics re kinematic
wavespeed
[LT'•].
of the activeerosionprocesses,directedsmall-scalefield and hydraulic friction factor.
modeling studies of these processesare greatly needed. averageflow depth[L].
Potential erosion processesinclude plucking, bashing by f(qs) erodibility scalingfactor for sedimentloading.
bedload,abrasion by suspendedload, cavitation, solution, f rock uplift rate ratio.
and weathering [Alexander, 1932; Maxson and Campbell, g gravitational
acceleration
[LT'2].
1935; Barnes, 1956; Foley, 1980; Wohl, 1993; Zen and H representative
verticallengthscale[L].
Prestegaard, 1994; Hancock et al., 1998; Sklar and L total bedrockstreamlength [L].
Dietrich, 1998; Whipple et al., submittedmanuscript,1999]. Q discharge
[L3T-•].
A numberof field and laboratory studies are underway [e.g., fluvial bedrockchannelrelief [L].
Slingerland et al., 1997; Sklar and Dietrich, 1998; Snyder R•c hillslopeand colluvial channelrelief [L].
and Whipple, 1998; Snyder et al., submittedmanuscript, s streamwisechannelbed gradient.
1999],but manyquestionsremainunanswered. Sc average gradient of hillslope/colluvial channel
Studyof the variationin the effectiveerosioncoefficientK profile.
accompanying adjustments to imposedboundaryconditions t time [7].
is also greatly needed. In most modeling studiesto date, responsetime for suddenbaselevelfall [7].
including our analysispresentedabove, the coefficientof responsetimefor stepincreasein upliftrate[T].
erosionhas been treated as a constant(dK/dt = 0; dK/dx = 0). rock uplift rate definedrelative to erosional base
In addition to the obvious assumptionthat lithology and level[LT4].
precipitationbe held constantin spaceandtime,holding K averagerockuplift rate definedrelative to erosional
constantcarriesthe implicit assumptionthat slope is the baselevel[LT-•].
only morphologicvariable that may adjust in responseto a v meanvelocity[LT'•].
changein boundaryconditions. Even in the simplestcases, w channelwidth [L].
this assumptionwill often be violated, with either the x streamwisedistancefrom divide [L].
amountof alluvial/colluvial cover [e.g., Howard, 1998; Sklar Xc critical distancefor transitionto fluvial erosion [L].
and Dietrich, 1998] or the channelwidth changingin con- elevationof streambed [L].
certwith the gradient. Spatialand temporalvariationsin the
coefficientof erosionmay importantlyinfluencethe dynamics
of river responseto tectonicandclimaticforcing. Exponents
The classicproblemof scaling observationsof local ero- a shear-stress
or stream-power
exponent.
sion rates and processesup to the reachscalerelevant in b exponentin channelwidth-dischargerelation.
landscape evolutionmodelsis, of course,a difficultyfacedby c exponentin discharge-arearelation.
small-scaleprocessstudies.A majorhurdlein this effortwill h exponent in area-lengthrelation (reciprocal
be finding an effectiveway to constrainthe reach-averaged Hack's exponent).
competitionand interactionof the variouserosionprocesses area exponent,erosionrule.
active at the bed. Answers to several questionsare needed. slope exponent,erosionrule.
WHIPPLEAND TUCKER:DYNAMICS OF THE STREAM-POWER
LAW 17,673