GEOLOGY, July 2010 639

INTRODUCTIONThe past two decades have seen an explo-

sion of research on the interaction of tectonics, climate, and denudation in evolving landscapes (Whipple, 2009). A great deal of this work has focused on rates and patterns of bedrock river incision, rock uplift, and precipitation (e.g., Bur-bank et al., 2003; Dadson et al., 2003; Reiners et al., 2003). The complexity of these processes has often led to simplifi ed conceptual or numeri-cal models in which landscapes are assumed to rise slowly and steadily above a base level, with bedrock rivers incising just as gradually into the landscape. However, active tectonic uplift is a sporadic process when driven by earthquakes, and the geomorphic consequences of such events are poorly understood and documented. We focus on the question of how bedrock rivers react to seismic events: does an event increase the rate of river erosion by locally generating relief along the surface rupture, or does it impede erosion by generating landslides, which swamp the rivers with sediment and thus inhibit ero-sion? We show that, in fact, both of these effects are currently occurring along rivers in central Taiwan in response to the 1999 Chi-Chi earth-quake, and that this one seismic event will cause spatially variable aggradation and incision in the river basin over periods of years to centuries. We begin with a description of the knickpoints generated by the fault scarp. We then explain our calculations of the mass of landslide mate-rial attributed to this earthquake and the time required for evacuating it. Finally, we discuss the results and implications for strath terrace devel-opment associated with the spatio-temporal vari-ation in river incision and fl oodplain widening processes in seismically active regions.

COSEISMIC KNICKPOINTSVertical offsets of surface ruptures from the

Mw = 7.6 Chi-Chi earthquake ranged from 0.5 to 8 m (Fig. 1) and were higher to the north (Chen et al., 2001), although local structural features introduce some variability compared to the mapped offsets (e.g., a backthrust across the Wu River and a pop-up structure across the Da-an River). A common feature among the major riv-ers that cross the fault is the presence of fl uvial knickpoints that have migrated upstream by

Geology, July 2010; v. 38; no. 7; p. 639–642; doi: 10.1130/G30883.1; 3 fi gures; Data Repository item 2010171.© 2010 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or editing@geosociety.org.

*Current address: Department of Geological Sci-ences, University of Michigan, Ann Arbor, Michi-gan, 48109, USA; E-mail: yanites@umich.edu.

How rivers react to large earthquakes: Evidence from central TaiwanBrian J. Yanites1*, Gregory E. Tucker1,2, Karl J. Mueller1, and Yue-Gau Chen3

1Department of Geological Sciences, University of Colorado, Boulder, Colorado 80309, USA2Cooperative Institute for Research in Environmental Science (CIRES), University of Colorado, Boulder, Colorado 80309, USA3Department of Geosciences, National Taiwan University, Taipei, Taiwan, Republic of China

ABSTRACTEarthquakes and bedrock river incision are fundamental processes in the evolution of tec-

tonically active landscapes, yet little work has focused on understanding how a bedrock river responds to a single large earthquake. Data from the 1999 Mw = 7.6 Chi-Chi earthquake in central Taiwan show dramatic differences in river response that depend on the proximity to the rupture zone. Near the fault scarp, vertical ground deformation intensifi es river inci-sion on a time scale of years to decades, while distal to the fault, landslides induced by the earthquake mantle bedrock on the river bed with sediment, impeding incision for decades to centuries as the material is evacuated. This surprising spatial and temporal variability in bedrock incision caused by earthquakes has implications for the paleoseismic interpretation of bedrock terraces and for the evolution of tectonically active landscapes.

(((((((((((((((((((

(((((((

((

((

((((((((((((((((((((((((((((((((((((

((((((((((((((((((((((((((((((((((

((((((((((((((((((((((((((((((((((((

((((((((((((((((((((((((((

((((((((

(((

((((((((((

(((

((

(((((((((((((((((((((((((((((

((

((

((((

((((

(

((((

((((((((

((((((((((

((((

((((((((

((((

((((((

((((((((((((((((((

(((((

(

((((((((((((((((((((

(((((((((((((((((( ((

((((((

((

((

B

(((

220000

220000

240000

240000

2640000 00

0000

2660000 00

0000

2680000 00

0000

10km

Knickpoint height (m)# 1–2# 2–4# 4–6# > 6

Vertical componentcoseismic (m) –1.1–0.0

0.0–0.10.1–0.50.5–11–22–33–44–55–66–8.5

C

D

PrePreP -- rththquaqqq ke k chahchh bbbb ddd

Knickpoint

B

C

D

~2 ~2~2 ~~2 ~2 ~2

FloFloFloFFloFFFF w dw dwww irei ctictic oo

FloFloFloFFFloFF w dw dww w dw ireii cticctictiicctiooo

Pre-earthquake channel bed

~2 m

Flow direction

Flow direction

Ch

elu

ng

pu

fau

lt

((((

(((( (((( (((((((( (

1999 Chi-chi Epicenter

Study reachPeikang

((((Gauging Station

A

A’

(((((((((

Shu

ilike

ng

fau

lt

Da’an

Tachia

Wu

Cho Shui

##

#

#

1999 Surfacerupture

A

Figure 1. Overview of fi eld area. A: Shaded relief image showing location of Chelungpu fault (red) and other mapped faults (Chen et al., 2001; Yanites et al., 2010). Color circles show vertical coseismic displacement. Triangles are locations of knickpoints in January 2008 and are colored according to their height as of 2008: Choshui: ~2–4 m; Wu River: 1–2.5 m; Tachia River: 4–6 m; Da-an River: 8–18 m. Red square denotes location of gauging station that doc-uments signifi cant aggradation following earthquake (Fig. DR2; see footnote 1). Stars are lo-cations of photographs found in B–D. Map coordinates are in Taiwan Datum 1967. Line A-A′ is location of cross section in Figure DR1. B: Bridge in low-capacity reach that has almost been buried in fl uvial sediment following earthquake. C: Looking downstream of knickpoint on Choshui. Former river bottom is now vegetated terrace. Inner banks become more diffuse downstream. D: Knickpoint on Choshui River located 4 km from surface rupture (as of 2008). Inner channel depth is ~2 m.

640 GEOLOGY, July 2010

~0.5 km to ~4 km since 1999. The depth of mea-sured knickpoint erosion ranged from 1 to 18 m in January 2008, and mirrors the general magni-tude of the coseismic vertical offset (i.e., higher to the north). The vertical coseismic displace-ment rapidly decays upstream and is minimal ~15 km east of the fault (Fig. 1). We expect the vertical incision from the migrating knickpoints to decay at a comparable distance.

Paleoseismic work along this fault suggests an earthquake recurrence interval of ~500 yr (W.S. Chen et al., 2004). Thus, the vertical inci-sion accomplished within the fi rst several years of the 1999 event corresponds to a 2–36 mm/yr incision rate when averaged over the entire seismic cycle. This is comparable to or greater than the ~3–5 mm/yr vertical component of slip over the Holocene for the Chelungpu fault (W.S. Chen et al., 2004; Y.G. Chen et al., 2004) (Fig. DR1 in the GSA Data Repository1) and may explain the lack of a persistent knickpoint at the fault. The inner channel on the Choshui transitions from a sharp rectangular geometry immediately downstream of the knickpoint to a diffuse and almost nonexistent inner chan-nel within 2 km downstream, suggesting rapid destruction of the inner channel after the knick-point passes (Figs. 1C and 1D).

SEISMICALLY INDUCED LANDSLIDESThe earthquake triggered ~20,000 landslides

within the 0.2 g coseismic vertical ground accel-eration contour (Dadson et al., 2004). Two years later, Typhoon Toraji initiated another 30,000 landslides in this area, an unusually large num-ber for a typhoon, which likely represents a reduction in cohesion and friction by the shaking during the earthquake (Dadson et al., 2004). The volume of material released by these landslides is calculated with a bounded power-law distribu-tion of landslide area (Hovius et al., 1997):

( )= =p area A

A AAl l

max min

−−− −

−11 1

αα α

α

n m≤ ≤A A Almi ax, (1)

where Al is the landslide area. The minimum land-slide area, Amin, is 3600 m2, the detection limit of Dadson et al. (2004), and 1 km2 is the maximum area, Amax. We are not aware of any landslides larger than this size in our basin nor do we see any evidence that such a slide occurred. Incor-porating very large landslides would increase the sediment mass needing to be removed and thereby increase calculated evacuation times. The exponent, α, was calculated by matching

the observed total landslide area (Dadson et al., 2004; Meunier et al., 2007) with that predicted by Equation 1 for both the earthquake, α = 2.1, and the typhoon, α = 2.7. Next, we assumed a landslide depth, d, using the scaling relationship proposed by Hovius et al. (1997):

d l A≅ ≅ε ε l , (2)

where ε is a scaling factor, and l is the landslide width (assumed to be Al

0.5). This allows a cal-culation of landslide volume and mass for the 50,000 landslides caused by the Chi-Chi earth-quake and subsequent typhoon Toraji. The range in α found above leads to a range in landslide volume and therefore a range in the evacuation time calculated below and shown in Figure 2.

Our estimated landslide parameters suggest that the earthquake and subsequent typhoon produced ~0.5–1.3 × 104 Mt of debris within the 0.2 g ground acceleration zone, equivalent to stripping ~0.6–1.7 m of material off the land-scape. This is a minimum value for the seis-mic and postseismic eroded material, as other typhoons caused enhanced landslide activity (Chen and Petley, 2005; Lin et al., 2008) that we do not account for here. This hillslope erosion accounts for a denudation rate of at least 1.2–3.4 mm/yr for an ~500 yr earthquake recurrence interval, comparable to island-wide average denudation rates of ~5.2 mm/yr from sediment load data and 1.5–6 mm/yr from thermochro-nology (Dadson et al., 2003).

SEDIMENT EVACUATIONThe rivers evacuating this pulse of sediment

and not affected by the coseismic knickpoint are overwhelmed with sediment, effectively halt-ing bedrock incision until most of the sediment pulse has washed downstream (Fig. 1B). To explore the infl uence of this event on the long-term dynamics of these river systems, we calcu-late the evacuation time of this sediment pulse to compare to the frequency of a Chi-Chi type event. Landslide mass and sediment transport capacity are estimated for the Peikang River. The river is located 15–25 km east of the fault scarp (Fig. 1) but within the 0.2 g ground accel-eration zone (Dadson et al., 2004). Bed-eleva-tion records from a gauging station in this reach reveal 3.4 m of aggradation in the 6 yr following the earthquake, and fi eld observations between 2006 and 2008 revealed that this aggradation persists (Fig. DR2). Further, electrical resistiv-ity surveys reveal signifi cant sediment cover throughout the study reach (Hsu et al., 2010).

The study area of the Peikang River basin covers ~10% (at the upstream end) to 18% (at the downstream end) of the area affected by intense landslide activity as mapped by Dadson et al. (2004), requiring ~1–2 × 103 Mt of mate-rial to be removed. We assume that 10% of this

sediment is removed by bedload. Our value of 10% is a conservative estimate as landslides tend to generate coarse material (e.g., Attal and Lavé, 2006); however, the study reaches are downstream from the low-order basins where most of the material is produced and thus fl u-vial processes (abrasion, sorting) lower the fraction of bedload.

0.20.40.60.81

1.21.41.61.8

x 109

Ann

ual Q

T (

kg/y

r)

3000

4000

5000

6000

7000

8000

9000

QT fo

r 10

yr

flood

(kg

/s)

3/2

3/2*8

( )

( )b s

T s cs

Q gWgD

Dτ ρ ρρ τ

ρ ρ⎡ ⎤

= ⎢ ⎥−

⎣ ⎦−

−

B

4550556065-100

0

100

200

300

400

500

Distance from headwaters (km)

Tim

e sc

ale

for

rem

oval

of

bed

load

mat

eria

l (yr

)

DFlow direction

A

0

4

8

12

16

Inci

sion

rat

e (m

m/y

r)

1

2

3

4

x 1011

Ups

trea

m la

ndsl

ide

mas

s (k

g) C

73 km

ρ

~9–20 ~11–28

1GSA Data Repository item 2010171, Figures DR1–DR3, is available online at www.geosociety.org/pubs/ft2010.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.

Figure 2. Incision and morphological data from Peikang River. A: Bedrock incision rate calculated using optically stimulated luminescence dating of fl uvial sediments on bedrock strath terraces (Yanites et al., 2010). Shading denotes different age models used in dating procedure (central age model in gray and minimum age model in black). B: Black bars are total annual transport capac-ity (Q) averaged for each reach along river (Equation 3). Transport capacity for 10 yr fl ood is also shown in gray for comparison. C: Cumulative mass of landslide material produced using exponents of 2.1 (black) and 2.7 (gray) in Equation 1. See text for expla-nation. D: Evacuation time of Chi-Chi earth-quake–induced landslide pulse. Black rep-resents t1, evacuation time of all upstream landslide material if a reach is able to trans-port at capacity. Gray represents transport time of local material, t2, assuming mate-rial is evacuated at rate of excess transport capacity compared to upstream reaches. Negative values arise when upstream reach transport capacity is greater than local.

GEOLOGY, July 2010 641

We divide the river in the study area into fi ve reaches based on similar morphology and Holo-cene incision rates (Yanites et al., 2010) and cal-culate two time scales for each reach. The fi rst time scale, t1, assumes that the reach can evacu-ate all of the material upstream of and local to the reach at its local transport capacity, and is calculated t1 = Mh/QT, where QT is the local transport capacity integrated over 32 yr of daily fl ows (linearly scaled upstream and downstream with drainage area), and Mh is the total volume of landslide material that must be evacuated. QT is calculated with a Meyer-Peter and Müller (1948) formulation of bedload transport capac-ity (Fig. 2B):

Q WDT s

b

sc*g

=−( ) −

⎡

⎣⎢

⎤

⎦⎥8

3 2

ρ τρ ρ

τ

D s g−( )3 2 ρ ρρ

, (3)

where ρs is the density of a sediment grain (assumed to be 2.65 g/cm3), ρ is the density of water, W is channel width, D is the median grain size (assumed to be uniform at 10 cm based on >2000 measured clasts), τc* is the critical Shields stress (0.03), and τb is the local shear stress calculated with fi eld and remote measure-ments of channel morphology.

The second time scale, t2, is a lower bound of the evacuation time and accounts for situations in which the upstream reach has a lower trans-port. In these cases, we assume that the excess transport capacity (compared to the upstream reach) is used to transport locally derived land-slide material. This is the material that enters between the upstream and downstream ends of the local reach and is calculated by using the dif-ference in contributing drainage area between the ends. This time scale is calculated t2 = ML/(QLT − QUT ), where QLT and QUT are the local and upstream transport capacities, and ML is the local landslide mass.

DISCUSSION AND CONCLUSIONSAlong the reach located ~46–50 km from the

headwaters, Holocene bedrock incision rates are ~1–2 mm/yr (Yanites et al., 2010). Field evi-dence and electrical resistivity data (Hsu et al., 2010) indicate signifi cant modern aggradation, including buried bridge piers (Fig. 1B). Other rivers infl uenced by landslide debris aggraded 18 m in places (Chen and Petley, 2005). Even during the large fl oods associated with typhoons, this reach currently undergoes no vertical bed-rock incision as the upstream landslide material is evacuated from the hillslopes and low-order basins. The reach will require ~250–600 yr to remove the upstream sediment pulse (Fig. 2D) if

the calculated transport capacity is maintained. Following evacuation, enhanced erosion rates (>1–2 mm/yr) must occur if the reach continues to incise at the measured Holocene rates.

The reach located ~54–60 km downstream of the headwaters has the highest transport capac-ity (Fig. 2). However, sediment cover is still signifi cant and clearly on the order of meters above the thalweg (Fig. DR2; Hsu et al., 2010). Since this reach should be able to transport out the material delivered by the lower transport capacity reach upstream, the persistence of the alluvial cover in this reach can most simply be explained by the addition of locally derived material. Assuming that the reach immediately upstream will deliver bedload at its capacity, these reaches will take ~10–30 yr to evacuate the excess local material delivered by tributar-ies and adjacent hillslopes (Fig. 2D). Although bedrock incision may occur during fl ood events that locally scour to bedrock, it likely occurs at a much reduced rate due to the thickness of allu-vial cover (Hsu et al., 2010). After this excess material is gone from the lower order basins, the river will have access to bedrock along the bed and vertical bedrock incision will resume until the next earthquake.

Lateral widening is not reduced by bed aggradation in either the high- or low-capac-ity reaches. In fact, repeat fi eld observations between 2006 and 2008 revealed erosion of meter-scale blocks along the river banks during typhoon seasons. Periods of lateral widening and no vertical incision produce planar strath terrace surfaces that are abandoned when vertical inci-sion is recommenced (or, for example, when a large earthquake on the nearby Shuilikeng fault generates a local knickpoint). These planar sur-faces, far from the coseismic deformation fi eld, can become especially wide during periods of enhanced earthquake frequency on more distant faults such as the Chelungpu.

The short evacuation time in the high-capac-ity reach could explain the difference in incision rates between the low-capacity and high-capac-ity reaches (Fig. 2), as the latter is exposed to active bedrock incision more frequently than the former. The greater incision in the high-capacity reach corresponds to a zone of higher rock uplift due to the Shuilikeng fault (Yanites et al., 2010). This suggests that in central Taiwan, along-stream variations in stream power and shear stress correlate with incision rate and rock-uplift rate because they infl uence the response time of each river reach to the seismically induced increase in sediment delivery. For a channel to attain steady state over millennial time scales (i.e., incision rate equals rock-uplift rate), it must have a morphology (width and slope) that is able to both transport earthquake-generated sediment pulses out of the basin and also make up for the reduction in erosion rate while this

major evacuation period was ongoing. Hence, channel morphology along the Peikang River is controlled both by local base-level effects and by seismically triggered landsliding.

We conclude that following a large earth-quake, incision rates are signifi cantly lowered and in some cases arrested in reaches distal to the fault. The duration of this reduction is a funda-mental control on long-term bedrock river inci-sion and will depend on local transport capacity, upstream transport capacity, and the magnitude of earthquake-generated sediment supply. For reaches with relatively low transport capacity and far from the fault, the majority of vertical bedrock incision occurs late in the seismic cycle (Fig. DR3), while lateral erosion processes occur throughout. A second important time scale is the time required for the earthquake-generated knickpoint to arrive at the local river reach. These time scales, along with the height of the knick-point when it arrives (relative to the total incision depth between earthquakes), will dictate whether vertical bedrock incision is accomplished primar-ily via knickpoint migration or fl at-bed processes of erosion. In any case, it is clear that the bed state of a bedrock stream in a seismically active region is variable through time.

Our results lead to a conceptual model than can be applied to many active tectonic settings (Fig. 3) and that has signifi cant implications for the development and preservation of bedrock strath terraces in seismically active areas. In channel reaches distal to the fault and further into the thrust belt (but still subject to ground acceleration), periods of frequent earthquake activity result in the reduction or cessation of vertical incision processes, yet continued lateral widening will generate planar bedrock surfaces mantled by alluvium (Fig. 3; Fig. DR3). Long after a major earthquake, the river may have had enough time to evacuate the sediment pulse and reinitiate downcutting. If the vertical inci-sion is signifi cant enough, the river will aban-don the strath levels that were initially gener-ated during previous earthquake cycles. Thus terrace surface ages (i.e., the age at which the surface is no longer fl ooded) will correlate to periods of seismic quiescence. Further, as our schematic drawing suggests (Fig. 3), reaches with relatively low transport capacity (and dis-tal to the fault) will have relatively wide planar strath surfaces, as observed along our study river (Yanites et al., 2010).

Proximal to the fault, vertical coseismic offset causes rapid downcutting, which will abandon previous fl oodplain levels where water depths are less than the magnitude of ground deforma-tion. After the passage of the coseismic knick-point, lateral widening and the creation of wide planar surfaces will continue until the surface is abandoned by another large seismic event. Thus near the fault, abandoned strath surfaces should

642 GEOLOGY, July 2010

correlate with seismic events. Another predic-tion our model suggests is that strath terraces proximal to the seismic event will be truncated at the fault, which is observed (Fig. 1).

The results presented show that surface-rupturing earthquakes infl uence river-incision processes in a manner that depends on the prox-imity of the river to the fault. Near the Chi-Chi fault scarp, the majority of vertical bedrock inci-sion is accomplished within years to decades of the earthquake. Conversely, the Peikang River, located 15–25 km east of the fault scarp, under-goes strongly reduced or even halted incision due to the pulse of sediment associated with earthquake-induced landslides. Nonsteady ero-sion at time scales of decades to centuries, along with the spatial variability in erosion processes, has implications for measuring bedrock channel incision rates and morphology, the formation of strath terraces, and the controls of landscape evolution within seismically active regions.

ACKNOWLEDGMENTSFunding was provided by National Science Foun-

dation grant EAR-0510971 and a National Defense Science and Engineering Graduate Fellowship to Yanites. The manuscript benefi ted from thorough reviews from Mikael Attal, Blanche Poisson, and an anonymous reviewer.

REFERENCES CITEDAttal, M., and Lavé, J., 2006, Changes of bedload

characteristics along the Marsyandi River (cen-tral Nepal): Implications for understanding hill-slope sediment supply, sediment load evolution along fl uvial networks, and denudation in ac-tive orogenic belts, in Willett, S.D., et al., eds., Tectonics, climate, and landscape evolution: Geological Society of America Special Paper 398, p. 143–171, doi:10.1130/2006.2398(09).

Burbank, D.W., Blythe, A.E., Putkonen, J., Pratt-Si-taula, B., Gabet, E., Oskin, M., Barros, A., and Ojha, T.P., 2003, Decoupling of erosion and precipitation in the Himalayas: Nature, v. 426, p. 652–655, doi: 10.1038/nature02187.

Chen, H., and Petley, D.N., 2005, The impact of land-slides and debris fl ows triggered by Typhoon Mindulle in Taiwan: Quarterly Journal of En-gineering Geology and Hydrogeology, v. 38, p. 301–304, doi: 10.1144/1470-9236/04-077.

Chen, W.S., Lee, K.J., Lee, L.S., Ponti, D.J., Pren-tice, C., Chen, Y.G., Chang, H.C., and Lee, Y.H., 2004, Paleoseismology of the Chelungpu fault during the past 1900 years: Quaternary International, v. 115, p. 167–176, doi: 10.1016/S1040-6182(03)00105-8.

Chen, Y.G., Chen, W.S., Lee, J.C., Lee, Y.H., Lee, C.T., Chang, H.C., and Lo, C.H., 2001, Sur-face rupture of 1999 Chi-Chi earthquake yields insights on active tectonics of central Taiwan: Seismological Society of America Bulletin, v. 91, p. 977–985, doi: 10.1785/0120000721.

Chen, Y.G., Shyu, J.B.H., Ota, Y., Chen, W.S., Hu, J.C., Tsai, B.W., and Wang, Y., 2004, Active

structures as deduced from geomorphic fea-tures: A case in Hsinchu area, northwestern Tai-wan: Quaternary International, v. 115, p. 189–199, doi: 10.1016/S1040-6182(03)00107-1.

Dadson, S.J., Hovius, N., Chen, H.G., Dade, W.B., Hsieh, M.L., Willett, S.D., Hu, J.C., Horng, M.J., Chen, M.C., Stark, C.P., Lague, D., and Lin, J.C., 2003, Links between erosion, runoff variability and seismicity in the Taiwan oro-gen: Nature, v. 426, p. 648–651, doi: 10.1038/nature02150.

Dadson, S.J., Hovius, N., Chen, H., Dade, W.B., Lin, J.C., Hsu, M.L., Lin, C.W., Horng, M.J., Chen, T.C., Milliman, J., and Stark, C.P., 2004, Earth-quake-triggered increase in sediment delivery from an active mountain belt: Geology, v. 32, p. 733–736, doi: 10.1130/G20639.1.

Hovius, N., Stark, C.P., and Allen, P.A., 1997, Sedi-ment fl ux from a mountain belt derived by land-slide mapping: Geology, v. 25, p. 231–234, doi: 10.1130/0091-7613(1997)025<0231:SFFAMB>2.3.CO;2.

Hsu, H.L., Yanites, B.J., Chen, C.C., and Chen, Y.G., 2010, Bedrock detection using 2D electrical re-sistivity imaging along the Peikang River, cen-tral Taiwan: Geomorphology, v. 114, p. 406–414, doi: 0.1016/j.geomorph.2009.08.004.

Lin, G.W., Chen, H., Hovius, N., Horng, M.J., Dadson, S., Meunier, P., and Lines, M., 2008, Effects of earthquake and cyclone sequencing on landsliding and fl uvial sediment transfer in a mountain catchment: Earth Surface Pro-cesses and Landforms, v. 33, p. 1354–1373, doi: 10.1002/esp.1716.

Meunier, P., Hovius, N., and Haines, A.J., 2007, Regional patterns of earthquake-triggered landslides and their relation to ground motion: Geophysical Research Letters, v. 34, L20408, doi: 10.1029/2007GL031337.

Meyer-Peter, E., and Müller, R., 1948, Formulas for bed-load transport: Proceedings, International Association for Hydraulic Structures Research Meeting, 2nd, Stockholm, Sweden: Delft, Neth-erlands, International Association for Hydrau-lic Structures Research, p. 39–64.

Reiners, P.W., Ehlers, T.A., Mitchell, S.G., and Montgomery, D.R., 2003, Coupled spatial vari-ations in precipitation and long-term erosion rates across the Washington Cascades: Nature, v. 426, p. 645–647, doi: 10.1038/nature02111.

Whipple, K.X, 2009, The infl uence of climate on the tectonic evolution of mountain belts: Na-ture Geoscience, v. 2, p. 97–104, doi: 10.1038/ngeo413.

Yanites, B.J., Tucker, G.E., Mueller, K., Chen, Y.G., Wilcox, T., and Huang, S.Y., 2010, Incision and channel morphology across active structures in the Peikang River, central Taiwan: Implications for the importance of channel width: Geologi-cal Society of America Bulletin, doi: 10.1130/B30035.1 (in press).

Manuscript received 21 November 2009Revised manuscript received 10 February 2010Manuscript accepted 15 February 2010

Printed in USA

T1

T2

T3

T4

Figure 3. Schematic representation of how bedrock rivers can react to large earthquakes. Proximal and distal reaches are separated by black and white dashed line. At T1 (far left), earthquake generates surface uplift that decays upstream of fault. Landslides are gener-ated in both proximal and distal reaches. In years to decades following earthquake (T2), knickpoint propagates upstream in proximal reach, while in distal reaches, channels widen as bedrock incision is inhibited during evacuation of landslide material. After a few decades (T3), distal reaches with relatively high transport capacity have evacuated local landslide ma-terial and reinitiated vertical bedrock incision. Further into seismic cycle (T4), knickpoint has dissipated as it migrated through coseimically deformed reach. In these proximal reaches, bed is likely alluviated as river widens its fl oodplain. In distal reaches, landslide material has been evacuated and bedrock incision has resumed in some locations, but not necessarily all. Finally, centuries into seismic cycle (T5), all landslide material has been evacuated from distal reaches and bedrock incision occurs along entire distal river.