You are on page 1of 19

Bulletin of the Seismological Society of America, Vol. 94, No. 6B, pp.

S156–S174, December 2004

Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali
Fault Earthquake Sequence
by Natalia A. Ratchkovski, Stefan Wiemer, and Roger A. Hansen

Abstract In this article we analyze the spatial and temporal variations in the
seismicity and stress state within the central Denali fault system, Alaska, before and
during the 2002 Denali fault earthquake sequence. Seismicity for 30 years prior to
the 2002 earthquake sequence along the Denali fault was very light, with an average
of four events with magnitude ML ⱖ 3 per year. We observe a significant increase
in the seismicity rate prior to the MW 7.9 event of 3 November 2002 within its
epicentral region, starting about 8 months before its occurrence. The majority of the
aftershocks of the MW 7.9 event are located within the upper 11 km of the crust and
form several persistent clusters with a few aseismic patches along the ruptured fault.
The most active aftershock source is associated with the epicentral region of the
earthquake. The overall b-value of the aftershock sequence is 0.96 with the highest
b-values within the epicentral region. We estimate that it will take 14 years for the
seismicity rate to drop back to the background level. The stress regime across the
region varies in space and time. The inferred stress regime prior to the 2002 sequence
is predominately strike slip. Along the central part of the rupture zone, the orienta-
tions of the least- and intermediate-stress axes are reversed after the 2002 earthquake
sequence. The maximum compressive stresses along the Denali fault rotate clockwise
by up to 35⬚; the greatest rotations occur in the area of the rupture step-over from
the Denali to the Totschunda fault. The inferred stress regime after the 2002 sequence
reflects an interchanging thrusting and strike-slip faulting along the ruptured fault.
The thrust faulting is concentrated in the epicentral region of the MW 7.9 event and
along the rupture segments showing the largest surface offsets.

Introduction
The 2002 earthquake sequence, which ruptured a sig- (DF) with up to 0.9 m slip occurring at depth and very little
nificant segment of the central Denali fault system (DFS) in slip in the upper 4 km of the crust. The NME, however, was
Alaska, provides an exciting opportunity to study major con- just the beginning of a more significant sequence of events.
tinental strike-slip events. Because of the low population Ten days later on 3 November 2002, the magnitude MW
density and earthquake-resistant building style of central 7.9 Denali fault earthquake (DFE) ruptured nearly 340 km
Alaska, the event fortunately resulted in minimal economic of the DFS (Eberhart-Phillips et al., 2003). It initiated as a
damage and no loss of human life. At the same time, how- thrust event on the previously unrecognized Susitna Glacier
ever, the remoteness of the area makes it challenging to doc- Thrust (SGT) fault, a splay fault located south of the main
ument the coseismic earthquake effects and to install instru- DF strand. The rupture then transferred onto the main DF
mentation to monitor the postseismic phenomena. strand and continued as a predominately right-lateral strike-
The 2002 Denali fault earthquake sequence began with slip event for ⬃220 km until it reached the Totschunda fault
the magnitude MW 6.7 Nenana Mountain event (NME) on (TF) near 143⬚ W longitude. At that point, it right-stepped
23 October 2002. The source mechanisms from teleseismic onto the more southeasterly trending TF strand and stopped
(Ozacar et al., 2003) and regional data (Ratchkovski et al., after rupturing nearly 70 km. A team of geologists surveyed
2003) indicate right-lateral strike-slip faulting on a vertical the total length of the ruptured faults and reported maximum
fault plane. While there was no surface rupture observed vertical and horizontal offsets of 2.8 m and 8.8 m, respec-
(P. Haeussler, personal comm.), the aftershock distribution tively, west of the DF and TF junction (Eberhart-Phillips et
(Ratchkovski et al., 2003) and the interferometric synthetic al., 2003). Inversions of the seismic (Ozacar et al., 2003),
aperture radar (InSAR) inversion (Wright et al., 2003) sug- geodetic (Hreinsdóttir et al., 2003), InSAR data (Lu et al.,
gest that this event ruptured some 35 km of the Denali fault 2003), and combined datasets (Dreger et al., 2004; Oglesby

S156
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S157

et al., 2004) indicate variable subsurface slip, mainly re- Wrangellia and Peninsula terranes and the Alaska Range
stricted to the shallow depths with three to four patches of complex south of the DFS. The Wrangellia and Peninsula
high slip along the ruptured fault. terranes include metamorphosed and deformed, upper Paleo-
Both the NME and the DFE generated vigorous after- zoic, volcanic, and sedimentary rocks that were originally
shock sequences. This article presents an analysis of the re- part of an island arc. The Yukon-Tanana terrane includes
gional crustal seismicity prior to the 2002 earthquake se- intensely metamorphosed and ductilely deformed, Devonian
quence and of the aftershock sequence. Specifically, we have and older, sedimentary and subordinate volcanic rocks as
performed a double-difference relocation of the events, a well as Devonian metaplutonic rocks (Csejtey et al., 1982;
statistical analysis of pre- and post-Denali earthquake se- Redfield and Fitzgerald, 1993).
quence (DES) seismicity parameters, and a stress-tensor in-
version of pre- and post-DES focal mechanisms. Our goal is Regional Seismicity Prior to the 2002 Denali Fault
to combine these diverse analysis results into an accurate Earthquake Sequence
description and interpretation of the seismotectonic frame-
work. Seismic Network
A regional seismic network in Alaska was established
Tectonic Setting in the late 1960s and early 1970s and has undergone many
changes since then. One of the most significant transitions
Interaction of the Pacific and North American tectonic occurred in the late 1980s when the waveform data began
plates shapes the major tectonic characteristics of Alaska. to be recorded digitally. Presently, the Alaska seismic net-
In southern Alaska and along the Aleutian arc, the Pacific work consists of approximately 400 sites that are operated
plate subducts underneath the North American plate at a con- by several different agencies. The majority of the sites are
vergence rate increasing from 5.5 cm/yr in the east to 7.5 maintained by the Alaska Earthquake Information Center
cm/yr in the west (DeMets et al., 1990). In southeastern (AEIC), the Alaska Volcano Observatory, and the West
Alaska, the plate boundary lies along the Queen Charlotte- Coast and Alaska Tsunami Warning Center. Although most
Fairweather strike-slip fault system. Although most of the of the sites are equipped with the analog, short-period ver-
seismic energy is released along the plate boundary in major tical seismometers, in the past 5 years AEIC has upgraded
earthquakes, large earthquakes also occur within central some of the existing sites with digital broadband sensors and
Alaska with the 2002 NME and DFE being the most recent installed new broadband seismic stations, bringing the total
examples of this phenomenon. The buildup of tectonic number of the broadband sites in Alaska to nearly 40.
stresses in central Alaska is the consequence of the collision The AEIC, located in Fairbanks, receives, processes, and
of the Yakutat block into the southern Alaska margin (Page archives data recorded by the Alaska regional network and
et al., 1995; Fletcher and Freymueller, 1999). produces the Alaska earthquake catalog (www.aeic.alaska
The DFS extends for ⬃2000 km from southeast Alaska .edu/Seis/html_docs/db2catalog.html). Prior to July 1998 the
through Canada and central Alaska to the Bering Sea (Fig. earthquakes were routinely located with the use of HYPO-
1). Although the DFS has been generally regarded as a right- ELLIPSE location program (Lahr, 1989). Since then AEIC
lateral strike-slip fault system, its tectonic history remains has begun using the dbloc2 module of the Antelope software
poorly understood, particularly at its western end. Along dif- package from BRTT, Inc. (www.brtt.com) to process re-
ferent segments, the documented amount and timing of off- corded events. Both location programs use the same suite of
set varies by an order of magnitude, from estimates of tens the regional plane-layer velocity models.
of kilometers for the western end (Grantz, 1966; Reed and
Elliot, 1968) to hundreds of kilometers for the eastern end
(Nokleberg et al., 1985). Savage et al. (1981) and Plafker et Characteristics of the Regional Seismicity
al. (1989) suggested contemporary slip rates along the cen- Crustal seismicity in central Alaska is associated with
tral DF to be of the order of 1–2 cm/yr. Recent Global multiple structures (Fig. 1). The major tectonic structures are
Positioning System (GPS) studies have indicated a rate of the two large-scale right-lateral strike-slip fault systems:
⬃0.8 cm/yr for deep slip along the central part of the DFS Tintina-Kaltag in the north and Denali in the south. They
(Fletcher, 2002). bound the north-northeast-trending seismic zones that pro-
The DFS has been suggested to have played a significant duced magnitude 7Ⳮ events in the 1900s (Fletcher and
role in the tectonic development of southern Alaska (Stout Christensen, 1996). These zones of seismicity become less
and Chase, 1980; Redfield and Fitzgerald, 1993). As Csejtey active from west to east and fade out at about 147⬚ W lon-
et al. (1982) suggested and later geophysical studies con- gitude. Page et al. (1995) suggested a block-rotation model
firmed (Labson et al., 1988; Stanley et al., 1990; Beaudoin to characterize these seismic zones. In this model, the crustal
et al., 1992), the DFS has developed on the already accreted blocks are rotating clockwise in a dextral shear zone between
continental margin, partly along an older, Cretaceous suture. the Denali and Tintina fault systems accommodating short-
North of the DFS lie the Yukon-Tanana and Nixon Fork ening across the region in response to north-northwest com-
terranes, which were in place prior to the accretion of the pression resulting from plate convergence. In addition, cen-
S158 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Figure 1. Seismicity of central Alaska (1971 to September 2002, AEIC earthquake


catalog). Inset map illustrates general tectonic setting of Alaska: NAP, North American
plate; PP, Pacific plate; DFS, Denali fault system; YB, Yakutat block. Solid lines are
major faults.

tral Alaska is an active source of subduction zone seismicity and Hansen, 2002a). See Table 1 for locations of the above-
(Fig. 1). Earthquakes as deep as 150 km extending as far mentioned earthquakes.
north as 64⬚ N latitude are located routinely by the AEIC.
This intermediate-depth seismicity underlies a section of the
DFS between 148 and 153⬚ W. Earthquake Relocations
Recorded seismicity along the central DF during the 3 The AEIC earthquake catalog contains over 8000 crustal
decades prior to the 2002 DES shows a very low level of earthquakes (0–30 km depth) within the greater central DFS
activity (Fig. 2a). In general, the seismic behavior of the DF region with available phase data (January 1971 to September
is characterized by infrequent large earthquakes. The largest 2002). The station coverage in central Alaska is rather sparse
recorded earthquake prior to the 2002 DFE that can be at- and variable. Station spacing ranges from less than 50 km
tributed to the faulting along the DF was the magnitude 7.2 to more than 100 km (Fig. 2a). Mean horizontal and vertical
earthquake on 7 July 1912 (Doser, 2004; Plafker et al., location uncertainties of the selected events are 2.3 and 3.4
2004). A magnitude 5.9 shock occurred on 31 August 1958 km, respectively (66.7% confidence ellipsoid). The routine
within the eastern part of the DFE rupture zone, but its lo- earthquake locations can be improved with the use of various
cation is north of the DF trace. A magnitude 6.0 event on relocation techniques (e.g., in Alaska, Ratchkovski and Han-
8 August 1959 occurred to the west of the NME aftershock sen, 2002a,b; Ratchkovski et al., 2003). We use a double-
zone at an estimated depth of 44 km. Its depth and focal difference method (Waldhauser and Ellsworth, 2000) to re-
mechanism obtained from waveform modeling (Doser, locate selected crustal events within the central DFS region.
2004) is not consistent with faulting along the DF. More The double-difference algorithm links pairs of event neigh-
recent notable seismicity along the western part of the DFE bors through both P- and S-wave travel-time differences re-
rupture was three moderate earthquakes (ML 5.0–5.3) that corded at the same stations and uses differentials of these
occurred in 1967–1970. The most recent event of note was travel-time differences in its event location algorithm.
the magnitude MW 5.7 event on 22 October 1996 located We used only phase readings within a 300-km epicen-
east of the Richardson Highway. As evidenced by its after- tral distance for the relocation. There were 1,136,764 P-
shock distribution and focal mechanism, this was a thrust phase and 392,650 S-phase pairs available for the selected
event located just north of the main DF trace (Ratchkovski 8178 events, of which 7715 events were clustered and 463
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S159

Figure 2. (a) Maps of the double-difference relocated crustal seismicity (0–30 km


depth) for the interval between 1971 and 22 October 2002. Small and large open stars
denote epicenters of the 2002 MW 6.7 and MW 7.9 events, respectively. Filled triangles
denote seismic stations. Solid lines are Paleogene faults from Plafker et al. (1994).
Dashed rectangle deleniates events included into the cross section shown in b. (b) Cross
section of the relocated hypocenters outlined in a.

were isolated events. The average offset between linked The 2002 Denali Fault Earthquake Sequence
events was 10.9 km with the maximum offset of 20 km. S-
wave arrivals were assigned half the weight of the P-wave Aftershock Processing
arrivals. A total of 7139 events were relocated, that is, 87% Following the MW 6.7 NME, AEIC staff installed a net-
(Fig. 2a). The rest of the events did not relocate because of work of seven instruments along the Parks and Denali High-
large-event separation and/or travel-time residuals. ways, west and south of its aftershock zone, respecively (Fig.
Although the relocated events show significant im- 3a). This initial network consisted of three strong-motion
provement in delineation of the north-northeast-trending and four broadband stations. From this portable network two
seismic zones in the northwestern part of the study area, the strong-motion and three broadband instruments recorded the
seismicity along the DF is still rather diffuse. Most of the MW 7.9 mainshock. The nearest site was 34 km from its
relocated events along the DF (95%) lie within the upper 15 epicenter. Within a few days of the DFE an additional 19
km of the crust (Fig. 2b). The earthquakes along the DF trace temporary sites were installed for monitoring the central and
are concentrated within the epicentral region of the NME and eastern segments of the DFE rupture zone (Ratchkovski et
DFE events and east of the junction of the Hines Creek and al., 2003). The temporary installations greatly improved sta-
Talkeetna faults with the main Denali strand. A scattered tion coverage around the DF and provided valuable data. The
band of earthquakes is associated with the TF branch. There temporary network consisted of a mixture of strong-motion
is no recorded seismicity along the stretch of the fault be- and broadband instruments. The sites were serviced every 3
tween 146 and 147⬚ W. to 4 weeks, when the data were retrieved from the instru-
S160 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Table 1 with ML ⱖ 1.5, which is close to the overall magnitude of


Hypocentral Parameters for Selected Earthquakes completeness of the fully processed aftershock catalog. This
Date Latitude Longitude Depth
relocation effort includes more aftershock data and has a
(mm/dd/yyyy) N W (km) MS ML MW lower magnitude cutoff than the previously published report
by Ratchkovski et al. (2003). We again limited arrivals to
07/07/1912 63.07 ⳮ146.14 25.0 7.2
10/16/1947 64.28 ⳮ148.23 0.0 7.2 only those within the 300-km epicentral distance and as-
08/31/1958 63.22 ⳮ144.32 17.0 5.9 signed S-arrivals half the weight of the P-arrivals. There
08/28/1959 63.42 ⳮ148.85 44.0 6.00 were 927,463 P-phase and 304,387 S-phase pairs. The av-
11/22/1967 63.60 ⳮ147.00 10.0 5.3 erage offset between linked events was 5.72 km with the
05/04/1969 63.50 ⳮ148.30 40.0 5.0
maximum allowed offset of 20 km. Of 9926 events selected
05/01/1970 63.50 ⳮ148.90 0.0 5.0
10/22/1996 63.3470 ⳮ145.3588 3.9 5.4 5.5 for the relocation, there were 9888 clustered and 38 isolated
10/23/2003 63.5144 ⳮ147.9116 4.2 6.7 6.7 events. A total of 8895 events were relocated, of which 396
11/03/2003 63.5285 ⳮ147.4455 0.2 4.4 were the NME aftershocks that occurred prior to the 3 No-
18:47:15.2 vember event. Therefore, 10% of the events did not relocate
11/03/2003 63.5141 ⳮ147.4529 4.2 8.5 7.9
22:12:41.5
because of large-event separation and/or travel-time residu-
11/03/2003 63.3001 ⳮ145.6795 0.9 5.8 als. Maps and cross sections of the relocated events are
22:32:17.7 shown in Figure 3 c–e.
Overall, the double-difference locations are more tightly
Data sources for pre- and post-1973 data are the USGS/NEIC Catalog
of Significant U.S. Earthquakes (http://earthquake.usgs.gov/) and AEIC clustered and show more detail than the original catalog lo-
Earthquake Catalog (http://www.aeic.alaska.edu/), respectively. cations. The NME aftershocks are located mostly to the west
of the Mw 6.7 epicenter along a 35-km-long stretch of the
DF with d5 ⳱ 2 km and d95 ⳱ 13 km (Fig. 3b), where d5
ments and brought back to the AEIC to be merged with the and d95 are the depths above which 5% and 95% of the
permanent network data. Data recovery continued through- events are located, respectively. The DFE epicenter is located
out the difficult Alaska winter leading to a recovery rate of 22 km east of the NME. The ML 4.4 foreshock (smallest star,
less than 75%. The temporary sites were dismantled in June Fig. 3b) is located within 1 km of the mainshock but is
2003. slightly shallower. The majority of the DFE aftershocks are
The AEIC located over 1000 aftershocks of the NME located between d5 ⳱ 1.5 km and d95 ⳱ 10.7 km depth.
before the M 7.9 event struck and over 16,000 total DES Aftershock locations within the epicentral region of the
events through the end of December 2002. Because of the DFE indicate complex interactions between the Susitna Gla-
great increase in the processing load, AEIC has processed so cier and Denali faults. In the south, the aftershock epicenters
far only events of ML ⱖ 3 for the interval January to July are bounded by the mapped rupture along the SGT fault, and
2003. The largest aftershocks of the NME were two ML 3.8 most hypocenters are confined within the hanging wall. In
earthquakes. The DFE was preceded by a magnitude ML 4.4 the north, the aftershocks extend beyond the DF trace and
foreshock that occurred 3.5 hr before the mainshock. The occupy a somewhat large volume. Thrust faults have been
foreshock had a strike-slip focal mechanism, which is con- mapped within close proximity of the DF north of its main
sistent with the right-lateral faulting along the DF. The larg- trace (Ridgway et al., 2002). We interpret the cloud of seis-
est aftershock of the DFE was an ML 5.8 event that occurred micity north of the DF trace within the epicentral area to be
20 min after the mainshock and was located 95 km east of associated with one or multiple thrusts similar to those
it. The map in Figure 3a shows 9937 earthquakes from the shown in Figure 2a of Ridgway et al. (2002).
AEIC catalog having depths less than 30 km and ML ⱖ 1.5, East of the SGT fault, the DFE aftershocks cluster tightly
that occurred between 23 October 2002 and 30 November along the mapped DF trace for a distance of 35 km (Fig. 3b).
2003. The recorded aftershocks present a complex pattern; Farther east, the aftershock pattern becomes once again more
they cluster in pockets of activity separated by less active complex (Fig. 3c). The second and second largest subevent
patches. of the DFE was located within this area (Eberhart-Phillips et
al., 2003). The aftershock patterns (cross sections D and C)
are consistent with the “flower” structures (thrust faults dip-
Aftershock Relocations ping toward the main fault trace) imaged on both sides of
Although the temporary network stations improved cov- the ruptured fault (Brocher et al., 2004; Fisher et al., 2004).
erage around the ruptured fault and provided valuable data Considerably fewer events are located along a short
for the aftershock detections and locations, the routine earth- stretch of the fault coinciding with the Richardson Highway
quake locations can be improved with the use of relocation crossing of the DF and the regional topographic low (Fig.
techniques. To better understand complexities of the after- 3c). This is where the Hines Creek fault and the Talkeetna
shock distribution, we again apply the double-difference Thrust are merging with the main DF trace from the north
method (Waldhauser and Ellsworth, 2000) to relocate the and south, respectively. A wide, low-velocity zone has been
DES events. We selected for relocation only earthquakes imaged within this aftershock gap (Brocher et al., 2004). The
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S161

Figure 3. (a) Double-difference relocated 2002 Denali fault sequence events ML ⱖ 1.5. Red
lines are the Peleogene faults from Plafker et al. (1994); yellow line, mapped surface rupture of
the M 7.9 earthquake (Eberhart-Phillips et al., 2003); black dashed lines, roads (DH, Denali High-
way; PH, Parks Highway; RH, Richardson Highway). Filled triangles denote seismic stations. (b–
e). Maps and cross sections of the relocated 2002 Denali earthquake sequence events. Filled tri-
angles and inverted open triangles denote permanent and temporary seismic stations, respectively.
Dashed lines, roads; thin black lines, Paleogene faults from Plafker et al. (1994) (DF, Denali fault;
TF, Totschunda fault; TT, Talkeetna thrust; SGT, Susitna Glacier Thrust; HC, Hines Creek fault;
MGF, McGinnis Glacier fault); thick solid line, mapped surface rupture (Eberhart-Phillips et al.,
2003); solid arrows, cross-section trends. Dashed lines in the cross sections are our interpretation
of the dipping (“flower”) structures based on the earthquake hypocenters. (continued)
S162 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Figure 3. (Continued)

crust may be so highly fractured because of the multiple fault determined by Lu et al. (1997), Ratchkovski and Hansen
interactions that it can not support sufficient strain accu- (2002a), and Ratchkovski (2003) for their studies. All the
mulations to produce earthquakes. The largest aftershock in mechanisms were obtained by using the computer program
the sequence (ML 5.8) occurred just east of this aftershock FPFIT (Reasenberg and Oppenheimer, 1985).
gap (Fig. 3c, open star). There is a wide variety of the faulting types. The events
Near the rupture step-over from the DF onto the TF, the within the north-northeast-trending seismic zones north of
aftershocks abandon the main DF trace and follow the the DFS indicate left-lateral strike-slip motion, consistent
mapped rupture onto the TF branch (Fig. 3d,e). The after- with the block-rotation model for this region (Page et al.,
shocks delineate a simple vertical plane along the TF (cross 1995). South of the north-northeast-trending seismic zones
section G), but indicate more branching complexity near the and north of the Hines Creek strand, there is an east–west
rupture step-over from one fault onto the other (cross section trending band of thrust-type events that follows the northern
F) and near the rupture arrest (cross section H). There is a foothills of the Alaska Range. The 1947 MS 7.3 event that
small cluster of events near the end of the mapped rupture had a thrusting source mechanism was located within this
extending eastward to the main strand of the DF (Fig. 3a). zone (Fletcher and Christensen, 1996). Many of the earth-
quakes along the DF are consistent with right-lateral strike-
Stress Tensor Inversion slip faulting; there are, however, some outliers.
We determined P-wave first-motion focal mechanisms
Earthquake Focal Mechanisms for the aftershocks with ML ⱖ 3.8 through the end of No-
Because of the sparse station coverage in the central vember 2002 and ML ⱖ 3.5 later on through the end of
Alaska, well-constrained focal mechanisms generally cannot November 2003. The focal mechanisms are predominantly
be determined for events with ML ⬍ 3.5. The map in Figure right-lateral strike-slip and reverse types (Fig. 5). Only a
4 shows 123 P-wave first-motion focal mechanisms avail- very few mechanisms indicate normal faulting; they are con-
able for the regional events shown in Figure 2a, starting in centrated in the epicentral region of the DFE and in the area
1987. The majority of the mechanisms were routinely de- of the rupture step-over onto the TF strand. The mechanisms
termined by the AEIC. We also included focal mechanisms in the epicentral area are a mixture of right-lateral strike-slip
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S163

Figure 4. P-wave first-motion focal mechanisms for crustal events (depth less than
30 km) prior to the 2002 Denali earthquake sequence (lower hemisphere projection).
Solid lines, faults from Plafker et al. (1994). Earthquake locations are from the AEIC
earthquake catalog.

and reverse mechanisms. East of the junction of the SGT lowing the classification of Zoback (1992). In addition, the
with the DF and west of the Richardson Highway, the mech- inversion computes maps of the stress-field homogeneity
anisms are predominantly right-lateral strike-slip, including based on the variance of the resulting stress tensor.
events located along the McGinnis Glacier fault. East of A low variance is indicative of homogeneous stress con-
the Richardson Highway between 145.8⬚ W and 144.8⬚ W, ditions. Large variance values indicate poor fit to the result-
events are again a mixture of reverse and strike-slip mech- ing stress tensor, suggesting that the stress field is hetero-
anisms. Further east, between 144.8⬚ W and 144⬚ W, within geneous (Wiemer et al., 2002). Previous studies of the stress
the area of the largest surface displacements, there is an over- orientations in the crust within the central DFS based on the
whelming predominance of the reverse mechanisms. Along earthquake focal mechanisms have shown that the stress
the TF, the mechanisms are mostly strike slip. field is heterogeneous across the region, and the stress ori-
entations changed after the DFE (Ratchkovski and Hansen,
2002a; Ratchkovski, 2003).
Mapping the Stress Orientations in the Crust There are 123 focal mechanisms in the pre-DES dataset
Based on the earthquake focal mechanisms, we map (Fig. 4) and 178 mechanisms for the DES events (Fig. 5).
stress orientations in the crust using the stress-tensor inver- The average number of first motions for the pre-DES earth-
sion method of Michael (1984, 1987) implemented as part quakes is about 30, and almost 50 for the aftershocks, with
of the ZMAP seismological software package (Wiemer, a minimum of 15 polarities for any given event. The pre-
2001). The stress inversion resolves orientations of the three DES dataset has fewer events per unit area than the DES set;
principal stresses (r1, r2, and r3), their relative amplitude therefore, we used different grids to map the stress direc-
and homogeneity of the stress field. A sufficiently hetero- tions: 20 ⳯ 20 km for the former and 10 ⳯ 10 km for the
geneous set of focal mechanisms, as required for reliable latter dataset, with 10 nearest mechanisms for each grid node
inversion results, is available for the inversion (Figs. 4 and taken for the inversion. To verify the inversion results and
5). Because the fault and auxiliary planes can not be dis- the sampling method, we computed maps with varying sam-
tinguished, they are assumed to be equally likely in the in- pling radii, grid size, and number of mechanisms per sample
version. The inversion also distinguishes between various and found no significant difference among the inversion
tectonic regimes (normal, strike-slip, and thrust faulting) fol- results.
S164 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Figure 5. P-wave first-motion focal mechanisms for the 2002 Denali earthquake
sequence events (lower hemisphere projection). Solid lines, faults from Plafker et al.
(1994); dashed lines, roads (RH, Richardson Highway). Crosses are the double-
difference event locations. Upper panel shows measured horizontal surface offsets
(Haeussler et al., 2004) for comparison.

The inversion results for the pre-DES dataset (Fig. 6) 2002) and the 1947 MS 7.3 thrust event occurred (Fletcher
indicate predominance of the strike-slip stress regime across and Christensen, 1996). The second subregion with pre-
the region. The area north of the DF is characterized by dominantly thrust faulting is located east of the merger of
north–south compression and east–west extension (R1). In- the Hines Creek and Talkeetna faults with the main DF
version results for the area south of the DF (R2) indicate a strand. This area exhibits the fault-normal r1 orientations.
high degree of stress heterogeneity (variance of 0.22) and a From GPS measurements of crustal movements in cen-
DFS-parallel orientation of the maximum compressive stress tral Alaska, Fletcher (2002) found principal strain directions
r1. Along the DF, the r1 directions change from a northwest- north of the DFS, within the region of the north-northeast-
trend west of the fault apex (R3) to fault-normal along the trending seismic zones, to be N57⬚W for compression and
fault segment east of the merger of the Hines Creek and N33⬚E for extension. These results are consistent with our
Talkeetna faults with the main DF strand (R4). Along the findings. However, the sparsity of the GPS data resulted in
TF, r1 is at a ⬃40⬚ angle to the fault trace (R5). These results a high uncertainty of the strain direction estmates (up to 40⬚)
are consistent with the stress directions obtained from the and an inability to detect spatial variations across the region.
earthquake focal mechanisms by Ratchkovski and Hansen Another study of geodetic data from two small triangulation
(2002a) and Ratchkovski (2003). networks across the DF near the Parks and Richardson High-
The most prominent exception from the strike-slip fault- way crossings by Savage and Lisowski (1991) identified a
ing regime is the area that follows the northern foothills of minor northeastern uniaxial extension. Our results agree
the Alaska Range north of the DF apex. This is the area with these findings near the Parks Highway crossing (at
where multiple thrust features were mapped (Ridgway et al., 149⬚ W), whereas near the Richardson Highway (145⬚ W)
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S165

Figure 6. Map of the stress-tensor inversion results for the dataset prior to the 2002
Denali fault earthquake sequence. Directions of the maximum compressive stresses r1
are color-coded according to the inferred faulting regime (blue, strike slip; red, thrust-
ing). Results of the stress inversions are shown for the selected regions (outlined by
dashed polygons) as stereographic projections of the best-fitting stress tensor with 95%
confidence regions.

we find an east–west orientation of the minimum principal reflects a higher degree of stress heterogeneity within an
stress axis. environment of complex fault interactions (Hines Creek and
The inversion results for the DES show varying stress Talkeetna versus DF). The two areas with predominant thrust
orientations and degrees of stress homogeneity along the faulting (A3 and A2) correspond to the fault segments with
ruptured faults (Fig. 7). The highest variance values corre- the largest and second largest surface slip (Eberhart-Phillips
spond to the area of the second largest slip event (A2), which et al., 2003; Haeussler et al., 2004). The third (smaller) re-
S166 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Figure 7. Map of the stress-tensor inversion results for the 2002 Denali fault earth-
quake sequence dataset. Directions of the maximum compressive stresses r1 are color-
coded according to the inferred faulting regime (blue, strike-slip; red, thrusting). Yellow
line is the mapped rupture. Results of the stress inversions are shown for the selected
regions (outlined by dashed polygons) as a stereographic projections of the best-fitting
stress tensor with 95% confidence regions. Measured horizontal surface offsets (Haeus-
sler et al., 2004) are shown for comparison below the map.

gion of thrust faulting (A1) is in the epicentral area of the ture, r1 rotated clockwise by up to 35⬚, with the maximum
DFE. Within this area, the orientations of the least and in- rotations just west of the TF branching point (A4). Along
termediate-stress axes are reversed after the DFE and the the TF, the r1 direction changed to fault-normal (A5). These
maximum stresses r1 generally rotated ⬃15⬚ clockwise from stress directions agree with the results obtained by Ratch-
the pre-DES orientations. Along the central part of the rup- kovski (2003).
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S167

The implications of the r1 rotations along the central when performing a precursory activity study is the question
part of the rupture zone are not clear. Taking into account of catalog homogeneity. Changes in the reporting capability
that r1 was fault-normal prior to the DFE and that a major of the network, magnitude shifts, and quarry blasts can either
earthquake is likely to change the stress state at least tem- mimic or mask natural signals and make the analysis of
porarily, the clockwise r1 rotations are expected after a right- microseismicity rates a challenging endeavor (Habermann,
lateral event. However, the new stress field, at least within 1987; Zuniga and Wiemer, 1999). We start our analysis in
the areas of the strike-slip faulting, is consistent with left- 1989, because catalog quality dramatically improved after
lateral motions. this date. We limit the analysis to crustal seismicity (depth
Our findings of significant stress state changes follow- ⱕ 30 km) within the central DFS, a total of 7250 events of
ing the DFE agree with previous results from the studies of magnitude ML ⱖ ⳮ0.5 excluding events without magni-
other major earthquakes. A high diversity of focal mecha- tudes (Fig. 2a).
nisms has been observed in the aftershock sequences of such First, earthquake clusters are removed from the catalog
earthquakes as the 1989 MS 7.1 Loma Prieta (Michael et al., by using the Reasenberg algorithm with its original param-
1990; Zoback and Beroza, 1993), 1992 MW 7.3 Landers eter setting (Reasenberg, 1985). This process removed 770
(Hauksson, 1994; Wiemer et al., 2002), 1994 MW 6.7 North- events in 192 clusters. The cumulative number with time
ridge (Thio and Kanamori, 1996; Hardebeck et al., 1998), plot (Fig. 8) reveals several apparent changes in seismicity
1999 MW 7.1 Hector Mine (Wiemer et al., 2002), and 1999 rate. The first one, a drop in rate by about 50% around mid-
MW 7.6 Izmut (Polat et al., 2002) events. For the Loma Prieta 1992, was discussed in detail in Zuniga and Wiemer (1999),
earthquake, Michael et al. (1990) found a stress field favor- who showed that it was caused by a change in the number
ing the right-lateral strike-slip faulting on planes subparallel of required number of station triggers for event detections
to the San Andreas fault before the mainshock, whereas the from three to four in the wake of the Mount Spurr volcanic
aftershocks indicated an extremely heterogeneous stress eruption. Overall magnitude of completeness Mc increased
field. In the Landers sequence, Hauksson (1994) docu- at this time from ⬃1.4 to ⬃1.6. A second drop in seismicity
mented a significant change in the stress state following the rate occurs in the period January 1999 through mid-2001.
earthquake sequence with the main changes in variations in Its characteristics are less clear and not readily explained as
the trend of the maximum and minimum principal stress a change in completeness, but it is likely related to the on-
axes. A detailed study of the stress field before and after the going changes in the AEIC data-recording and -processing
Izmut earthquake (Polat et al., 2002) found that the stress
regime before the mainshock was a strike slip, whereas the
stress regime after the event was between the strike slip and
extension with the direction of the minimum stress axis r3
changing only slightly.

Statistical Analysis
Analysis of Pre-Denali Earthquake
Sequence Seismicity
In this section we address the question of possible
changes in the seismicity rate prior to the DFE that might, in
hindsight, be interpreted as precursors to the subsequent
events. In particular, we investigate seismicity rates for signs
of anomalous increase or decrease in activity.
Numerous publications have investigated a precursory
quiescence based on observational data (Reasenberg and
Matthews, 1988; Kisslinger and Kindel, 1994; Wiemer and
Wyss, 1994; Dieterich and Okubo, 1996; Yoshida et al.,
1996; Wiemer and Katsumata, 1999; Zoeller et al., 2002) or
laboratory-based studies (Lockner and Byrlee, 1991; Kato
and Hirasawa, 1999). Precursory seismic quiescence is best
defined as a significant decrease in the microseismicity rate
in the immediate vicinity of an upcoming mainshock within
the preceding years to months. However, it is fair to say that Figure 8. Cumulative number of earthquakes as a
function of time for crustal seismicity in central
no consensus exists whether precursory quiescence is indeed Alaska (Fig. 2a). The figure is based on the declus-
a significant precursory phenomenon. tered catalog and includes all magnitudes. Note the
The first and probably most important question to ask significant changes in slope, marked by arrows.
S168 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

systems in 1998–1999 (Lindquist and Hansen, 1998). Seis-


micity rates increased again to pre-1992 levels after mid-
2001, when the AEIC analysts adjusted their processing rou-
tines to better identify events missed by the automatic
detection system. In addition, new broadband stations were
introduced in central Alaska starting in 1998. The complete-
ness also varies dramatically as a function of space, which
we investigated by mapping Mc according to the approach
by Wiemer and Wyss (2000). Mc is found to be generally
lower (⬃1.0) in the northern part of the study area and higher
(⬃2.0) near the DF.
We illustrate the characteristics of these changes in re-
porting by plotting the annual frequency of events for each
magnitude bin (Fig. 9), focusing on the northwestern quad-
rant of the study region only. If recording were homoge- Figure 9. Rate of events as a function of magni-
tude for four different periods. The rate of each mag-
neous, all curves would fall approximately on top of one
nitude bin is normalized to number of events per year,
another. However, the differences in completeness between and smoothed by using a moving average filter. This
the periods, as approximated by the bin with the highest figure is based on the northwestern quadrant of the
frequency, is very clear. The period from 1998 to mid-2001 study area only.
shows a different shape of the frequency-magnitude distri-
bution, broader and overall much lower. This raises the sus-
picion that the gradual rate decrease starting around 1998 prior to the DES, as compared with a background rate of
(Fig. 8) is an artifact, possibly related to poor performance ⬃10 events annually.
of the new data-processing system in the early years of its
operation. From this analysis we conclude that the catalog
is highly heterogeneous in its properties, considerably com- Analysis of the 2002 Denali Fault
plicating seismicity rate studies. Earthquake Sequence
To spatially map seismicity rate changes, we image the In this section, we map selected seismicity parameters
b-value that describes the significance of seismicity rate of the DFE aftershock sequence both spatially and tempo-
changes (Matthews and Reasenberg, 1988), comparing the rally, with the objectives of obtaining a detailed statistical
background period of 1989–2000 with the pre-DFE period description of the sequence and of investigating physical
of 2000 to 22 October 2002 (Fig. 10). We sample the nearest mechanisms of aftershocks in general. When assessing after-
100 events to each grid node spaced 0.1⬚ ⳯ 0.1⬚, using seis- shock rates or their size distribution, it is vital to have a
micity with ML ⱖ 1.6. The sampling volumes thus vary in detailed knowledge of the spatial and temporal variability in
radii, with typical radii between 20 and 100 km (Wiemer magnitude of completeness Mc (Wiemer and Katsumata,
and Wyss, 1994). We also analyzed other periods; however, 1999; Wiemer et al., 2002; Wiemer and Wyss, 2002). We
for space considerations we only show this selection, which use the entire magnitude range (EMR) approach defined by
displays two interesting features. Values of b ⬎ 1.96 indicate Woessner and Wiemer (2004) to assess Mc. We first inves-
a significance level of 5% or two standard deviations. Blue tigate it as a function of time, because in past sequences
colors, found near the hypocenter of the subsequent DFE, Mc(t) variability has been shown to be dominating (Wiemer
signify a relative increase in the seismicity rate in the 2.8 and Wyss, 2002). We analyze the 15,424 aftershocks until
years preceding the mainshock. Red colors, found to the 21 December 2002. The largest aftershock in this dataset is
north of the DFE, indicate a relative decrease. For two se- ML 5.8, the smallest is ML 0.1. Using a moving window of
lected nodes (A and B) we plot the cumulative number of 150 events with a 25-event increment, we compute Mc(t)
events (Fig. 10, inset). We find that the rate drop to the north (Fig. 12). The resulting curve shows a strong and typical
of the hypocenter is quite sudden in its onset, starting at decrease of Mc as the sequence evolves: Mc is greater than
about mid-1999. 4.0 in the first hour after the mainshock, and about 1.5 later
We confirm the rate increase near the DFE epicenter in in the sequence. The same decrease of Mc with time has been
the original, unclustered catalog and for various magnitude documented for other sequences such as the Landers or Hec-
cutoffs. We find that the rate increase near the epicenter is tor Mine (Wiemer et al., 2002; Woessner et al., 2004); it is
persistent in all representations of the analysis. However, the a consequence of the fact that in the early hours and days,
rate decrease to the north, while still visible, appears less small aftershocks are hidden in the codas of larger ones, and
sharp. The time series in Figure 11 reveals that indeed the that there is a tendency in network processing to neglect the
seismicity rate increased, due to small, swarm-like foreshock smallest aftershocks when faced with hundreds of events
activity that began around January 2002. The largest of these each day.
events had an ML 4.7; ⬃80 events occurred in the months To establish the overall p- and b-value of the entire se-
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S169

Figure 10. Map of the study region, colored is b-value that measures the signifi-
cance of rate changes between the two periods 1989–2000 and 2000–22 October 2002.
The map is computed using the declustered earthquake catalog with magnitude ML ⱖ
1.6. At each node, the nearest 100 events are sampled. A significant increase in seis-
micity rate in the later period, shown in blue and marked A, is found at the subsequent
hypocenter location. The cumulative number at this node, plotted in the inset in blue,
illustrates that the seismicity rate increased in 2002. Nodes north of the DF show a
significant decrease in seismicity (red); an example can be seen in the inset (B).

quence we analyze the subset of 911 aftershocks with ML ⱖ first day; this dataset is complete at Mc 2.3 (total of 2129
3 from 0.3 days after the mainshock time to 30 October events). In addition, we analyze the parameter pair that max-
2003. We fit a modified Omori law (Utsu et al., 1995; Ogata, imizes the number of available data (Mc 1.4, Tcut ⳱ 4.2 days,
1999) to the sequence (Fig. 13). The daily rate of aftershock N ⳱ 6300 events). These results, however, are quite com-
activity as a function of time can be well explained by an parable and are thus not shown due to space considerations.
Omori law with typical values of the parameters (p ⳱ 1.13, While completeness also varies as a function of space, which
c ⳱ 0.07, k ⳱ 150; Fig. 13, left panel). The goodness of fit we investigated by computing Mc maps for different periods,
of the aftershock sequence to a modified Omori law is judged this spatial variability is found to be small when compared
to be acceptable at 98% confidence by using a Kolmogoroff- to the temporal variability.
Smirnoff test (Woessner et al., 2004). We compute the ap- We map the b-value on a 1 ⳯ 1 km grid using overlap-
proximate duration of the aftershock sequence by comparing ping volumes containing 150 events each. The resulting b-
the daily rate with the background rate. From the 30 years value map (Fig. 14, top) shows spatial variations in b of up
of seismicity prior to the DFE we extracted a daily rate of to a factor of 2. The highest b-values (1.3) are found near
four ML ⱖ 3 events per year. Comparing this with the af- the hypocenter, whereas the eastern part of the rupture is
tershock rate we find that after about 14 years, the aftershock characterized by much lower b-values (0.7). In Figure 15 for
activity will have decreased to the previous background the two selected nodes, marked as A and B in Figure 14, we
level. This value is quite comparable with aftershock se- plot the frequency-magnitude distribution of the events
quences of large events in California (Dieterich, 1994). An within 10 km radii of these nodes. The two curves are quite
overall b-value for the sequence is 0.96 (Fig. 13, right panel). different to the eye and the difference between them can be
The fit to a power law is good up to a magnitude ⬃4.5; established to be statistically significant at 10ⳮ5 when ap-
above, fewer aftershocks than expected occur. plying Utsu’s test (Utsu, 1999).
To obtain a complete dataset for mapping purposes, we We also investigated the b-value as a function of dis-
use the more detailed 2002 data set. To ensure a complete tance from the epicenter and tried to correlate these functions
set, we again need to cut the time series in time and mag- with the surface-slip distribution of the mainshock. No sig-
nitude space. The choice of the [Mc, Tcut] pair is somewhat nificant correlation could be obtained by this analysis. Thus,
arbitrary. We chose to analyze the seismicity excluding the we do not find evidence in support of the hypothesis offered
S170 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Figure 12. Magnitude of completeness, Mc, as a


function of time, measured in days from the 3 No-
vember 2002 mainshock. Mc is computed by using
overlapping windows containing 150 events each, fol-
lowing the approach by Wiemer and Wyss (2000).

by Wiemer and Ktasumata (1999) that areas of high (low)


b-values correlate with areas of high (low) slip. In the DFE
case, the highest b-values are observed near the epicenter,
where comparatively little slip occurred, whereas the areas
of highest slip in the central part of the rupture show a low
b-value.
Figure 11. Time series of the cumulative number
of events for different magnitude thresholds (top We also compute a map of activity rate as the logarithm
frame) and magnitude stem plot (bottom frame). The of the number of events of magnitude M 2.3 (the a-value at
figure is based on the original, clustered catalog for M 2.3) for volumes of radii 5 km, and based on the data
the area near the 2002 Denali hypocenter (white rec- starting 4 days after the mainshock (Fig. 14, bottom). The
tangle in Fig. 10). aftershock activity varies by almost two orders of magni-
tude; the highest activity is near the hypocenter and the low-
est near the eastern end of the rupture. The elevated levels
of activity also correlate with the areas of the largest surface
offsets. The Richardson Highway aftershock gap is clearly
visible.

Figure 13. (Left) Rate of ML ⱖ 3 earthquakes as a function of time after the Denali
fault event (Tcut ⳱ 0.3 days). The observed data (circles) are fitted with a modified
Omori law (solid black line). The dashed line is the extrapolation into the future: After
about 14 years the aftershock rate will have reached the previous background rate of
four ML ⱖ 3 events per year. (Right) Frequency-magnitude distribution of aftershocks
with magnitude ML ⱖ 3. The resulting b-value is 0.96 (solid line).
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S171

Figure 15. Cumulative frequency of events as a


function of magnitude for two selected nodes, marked
A and B in Figure 14.

are foreshocks in a broad sense to the subsequent mainshock.


Such foreshock activity prior to large events is quite com-
mon (Reasenberg, 1999), although foreshock sequences dif-
fer in extent, magnitude distribution, and duration. So far,
the only way seismologists can explore their occurrence is
in a probabilistic sense: Time-dependent hazard maps that
include clustering, and thus foreshock activity, do show a
Figure 14. (Top) Map of the b-value of the after-
shocks (Mc 2.3, Tcut ⳱ 1.0 days, N ⳱ 2129 events) small increase in hazard through such sequences (Wiemer,
computed on a 1 ⳯ 1 km grid with 150 nearest events 2000).
per each grid node. The highest b-values, shown in The observed decrease in seismicity preceding the DFE
red, are found near the epicenter of the Denali fault in the north is in our opinion less likely to be related to the
event. For two nodes, marked as A and B, we plot the
upcoming events. Although it is physically plausible to ex-
frequency-magnitude distribution in Figure 15. (Bot-
tom) Map of logarithm of the number of events at pect quiescence due to a stress shadow at some distance from
magnitude M 2.3 (the a-value at M 2.3). This map is a subsequent mainshock, the distance between the anomaly
based on sampling volumes with a constant radii of and epicenter do not support a direct correlation. In addition,
5 km, using the same dataset with Mc 2.3, Tcut ⳱ 1 seismicity rate decreases are notoriously more difficult to
day. High values, shown in purple, indicate regions
establish and more prone to be introduced by monitoring
with the highest density of earthquakes.
changes. Given the overall low seismicity rate in the area of
the decrease, and given the heterogeneity in reporting, we
feel that we cannot establish with certainty that the observed
decreased seismicity is casually linked to the upcoming DFE.
Conclusions We applied the double-difference relocation technique
to relocate the aftershock catalog. The majority (95%) of the
We analyzed spatial and temporal variations in the relocated aftershocks of the DFE are located within the upper
crustal seismicity and stress state within the central Denali 11 km of the crust. The NME aftershocks are slightly deeper
fault system before and during the 2002 Denali fault earth- (2–13 km depth). The aftershocks form several persistent
quake sequence. Seismicity along the DF for 3 decades prior clusters and define a few aseismic patches along the ruptured
to the 2002 DES was very light, with an average of four fault. The most active aftershock source is associated with
events of ML ⱖ 3 per year. However, a significant increase the epicentral region of the DFE. We interpret the aftershock
of seismicity of up to 80 small events was observed 8 months “cloud” within the epicentral area of the DFE as associated
prior to the sequence onset within the epicentral region of with seismicity largely confined within the hanging walls of
the MW 7.9 DFE. Although the AEIC catalog is heteroge- thrust faults south (SGT fault) and north of the main DF trace
neous in reporting quality, much of this increase in seismic- (Ridgway et al., 2002). Aftershock clusters along the central
ity is in our opinion real, and not simply a result of improved part of the rupture similarly resemble seismicity concen-
monitoring capability. Although such swarm-like activation trated within the hanging walls of the thrust structures dip-
is not unique in the data, its spatial and temporal correlation ping toward the main fault trace, so called “flower” struc-
with the subsequent DFE strongly suggests that these events tures (Brocher et al., 2004; Fisher et al., 2004).
S172 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

The b-value mapping of the aftershocks shows spatial improvement of the manuscript. This research was funded in part by USGS
Contract 01HQAG0138 and NSF contract EAR-0328043 (N.A.R. and
variations of up to a factor of 2 along the ruptured faults. R.A.H.). This is a contribution of the Geophysical Institute, ETH Zurich
The highest b-values are associated with the epicentral re- (S.W.).
gion of the DFE. The lowest b-values are mapped along the
TF branch. No significant correlations between the surface
References
slip and the mapped b-values could be observed, leaving
open the question why aftershock seismicity parameters vary Beaudoin, B. C., G. S. Fuis, W. D. Mooney, W. J. Nokleberg, and N. I.
Christensen (1992). Thin, low-velocity crust beneath the southern
so dramatically among different sequences. The sequence Yukon-Tanana terrane, east central Alaska: Results from Trans-
can be well fit with an Omori law with typical parameters. Alaska Crustal Transect refraction/wide-angle reflection data, J. Geo-
We estimate that it will take 14 years for the seismicity rate phys. Res. 97, 1921–1942.
Brocher, T. M., G. S. Fuis, W. J. Lutter, N. I. Christensen, and N. A.
to drop back to the background level. Ratchkovski (2004). Seismic velocity models for the Denali fault zone
The stress regime across the region varies in space and along the Richardson Highway, Alaska, Bull. Seism. Soc. Am. 94,
time. The inferred stress regime prior to the 2002 sequence no. 6B, S85–S106.
is predominately strike-slip. North of the fault, the maximum Csejtey, B., D. P. Cox, and R. C. Evarts (1982). The Cenozoic Denali fault
system and the Cretaceous accretionary development of southern
compressive stress r1 is oriented north-south while south of Alaska, J. Geophys. Res. 87, 3741–3754.
the fault most of the focal mechanisms are consistent with DeMets, G., R. G. Gordon, D. F. Argus, and S. Stein (1990). Current plate
the fault-parallel compression. Two smaller areas with pre- motions, Geophys. J. Int. 101, 425–478.
Dieterich, J. (1994). A constitutive law for rate of earthquake production
dominantly thrust faulting were identified within the pre- and its application to earthquake clustering, J. Geophys. Res. 99,
DES dataset. One is located north of the DF apex and coin- 2601–2618.
cides with the location of the 1947 MS 7.3 event. Another is Dieterich, J. H., and P. G. Okubo (1996). An unusual pattern of seismic
located east of the area where the Hines Creek and Talkeetna quiescence at Kalapana, Hawaii, Geophys. Res. Lett. 23, 447–450.
Doser, D. I. (2004). Seismicity of the Denali fault zone (1912–1988), Bull.
Thrust faults join with the main DF strand (central part of Seism. Soc. Am. 94, no. 6B, S132–S144.
the DFE rupture zone). The pre-DES r1 stress directions Dreger, D., D. D. Oglesby, R. A. Harris, N. Ratchkovski, and R. Hansen
along the DF vary from northwest orientation west of the (2004). Kinematic and dynamic rupture models of the November 3,
2002 Mw 7.9 Denali, Alaska, earthquake, Geophys. Res. Lett. 31,
fault apex to fault-normal along the fault segment east of L04605, doi 10.1029/2003GL018333.
the merger of the Hines Creek and Talkeetna faults with the Eberhart-Phillips, D., P. J. Haeussler, J. T. Freymueller, A. D. Frankel,
main Denali strand. Along the TF, r1 is at a ⬃40⬚ angle to C. M. Rubin, P. Craw, N. A. Ratchkovski, G. Anderson, G. A. Carver,
A. J. Crone, T. E. Dawson, H. Fletcher, R. Hansen, E. L. Harp, R. A.
the fault trace. Harris, D. P. Hill, S. Hreinsdóttir, R. W. Jibson, L. N. Jones, R. Kayen,
The stress directions along the DF changed after the D. K. Keefer, C. F. Larsen, S. C. Moran, S. F. Personius, G. Plafker,
2002 earthquake sequence. The maximum compressive B. Sherrod, K. Sieh, N. Sitar, and W. Wallace (2003). The 2002 Den-
stresses rotated clockwise by up to 35⬚. The greatest rota- ali Fault Earthquake, Alaska: A large magnitude, slip-partitioned
event, Science 300, 1113–1118.
tions correspond to the area of the rupture step-over from Fisher, M. A., N. A. Ratchkovski, W. J. Nokleberg, L. Pellerin, and J. M.
the DF onto the TF. The inferred stress regime after the 2002 Glen (2004). Geophysical data reveal the crustal structure of the
sequence indicates an interchanging thrusting and strike-slip Alaska Range orogen within the aftershock zone of the M 7.9 Denali
Fault earthquake, Bull. Seism. Soc. Am. 94, no. 6B, S107–S131.
faulting along the ruptured fault. The thrust faulting is con- Fletcher, H. J. (2002). Crustal deformation in Alaska measured using the
centrated in the epicentral region of the DFE and along the Global Positioning System, Ph.D. Thesis, University of Alaska, Fair-
rupture segments with the largest measured surface offsets. banks, 257 pp.
Fletcher, H. J., and D. H. Christensen (1996). A determination of source
The implications of the r1 rotations along the central part of properties of large intraplate earthquakes in Alaska, Pageoph 146,
the rupture zone are not clear. Because r1 was fault-normal 21–41.
prior to the DFE and a major earthquake is likely to change Fletcher, H. J., and J. T. Freymueller (1999). New GPS constraints on the
motion of the Yakutat block, Geophys. Res. Lett. 26, 3029–3032.
the stress state at least temporarily, the clockwise r1 rota- Grantz, A. (1966). Strike slip faults in Alaska, U.S. Geol. Surv. Open File
tions are expected after a right-lateral event. However, the Rept. 267, 82 pp.
new stress field, at least within the areas of the strike-slip Habermann, R. E. (1987). Man-made changes of seismicity rates, Bull.
Seism. Soc. Am. 77, 141–159.
faulting, is consistent with left-lateral motions.
Haeussler, P. J., D. P. Schwartz, T. E. Dawson, H. D. Stenner, J. J. Lien-
kaemper, B. Sherrod, F. R. Cinti, P. Montone, P. A. Craw, A. J. Crone,
and S. F. Personius (2004). Surface rupture and slip distribution of
the Denali and Totschunda faults in the 3 November 2002 M 7.9
Acknowledgments earthquake, Alaska, Bull. Seism. Soc. Am. 94, no. 6B, S23–S52.
Hardebeck, J. L., J. J. Nazareth, and E. Hauksson (1998). The static stress
We thank Incorporated Research Institutions for Seismology, Na- change triggering model: constraints from two southern California
tional Science Foundation, U.S. Geological Survey, Geophysical Institute, aftershock sequences, J. Geophys. Res. 103, 24,427–24,437.
University of Alaska Fairbanks, and the University of Wisconsin for pro- Hauksson, E. (1994). State of stress from focal mechanisms before and
after the 1992 Landers earthquake sequence, Bull. Seism. Soc. Am.
viding technical and financial support in responding to the 2002 Denali
84, 917–934.
fault earthquake crisis. We also thank AEIC field crew Martin LaFevers, Hreinsdóttir, S., J. T. Freymueller, H. J. Fletcher, C. F. Larsen, and R.
Bart MacCormack, and Josh Stachnik and analysts Ed Clark, Trilby Cox, Burgmann (2003). Coseismic slip distribution of the 2002 Mw 7.9
Otina Fox, Lalitha Rao, and Jamie Roush for their hard work in collecting Denali Fault Earthquake, Alaska, determined from GPS measure-
and processing the aftershock data. Thoughtful reviews by Karl Kisslinger, ments, Geophys. Res. Lett. 30, 1670, doi 10.1029/2003G1017447.
Robert Page, and the associate editor Charlotte Rowe contributed to the Kato, N., and T. Hirasawa (1999). A model for possible crustal deformation
Seismotectonics of the Central Denali Fault, Alaska, and the 2002 Denali Fault Earthquake Sequence S173

prior to a coming large interplate earthquake in the Tokai district, Turkey earthquake of 17 August, 1999: Previous seismicity, after-
central Japan, Bull. Seism. Soc. Am. 89, 1401–1417. shocks, and seismotectonics, Bull. Seism. Soc. Am. 92, 361–386.
Kisslinger, K., and B. Kindel (1994). A comparison of seismicity rates near Ratchkovski, N. A. (2003). Change in stress directions along the central
Adak island, Alaska, September 1988 through May 1990 with rates Denali fault, Alaska after the 2002 earthquake sequence, Geophys.
before the 1982 to 1986 apparent quiescence, Bull. Seism. Soc. Am. Res. Lett. 30, 2017, doi 10.1029/2003GL017905.
84, 1560–1570. Ratchkovski, N. A., and R. A. Hansen (2002a). New constraints on tecton-
Labson, V., M. Fisher, and W. Nockleberg (1988). An integrated study of ics of interior Alaska: Earthquake locations, source mechanisms and
the Denali fault from magnetotelluric sounding, seismic refraction, stress regime, Bull. Seism. Soc. Am. 92, 998–1014.
and geologic mapping (abstract), EOS Trans. AGU 69, 1457. Ratchkovski, N. A., and R. A. Hansen (2002b). New evidence for segmen-
Lahr, J. C. (1989). HYPOELLIPSE/version 2.00: a computer program for tation of the Alaska subduction zone, Bull. Seism. Soc. Am. 92, 1754–
determining local earthquakes hypocentral parameters, magnitude, 1765.
and first motion pattern, U.S. Geol. Surv. Open File Rept. 89-116, Ratchkovski, N. A., A. Hansen, J. C. Stachnik, T. Cox, O. Fox, L. Rao, E.
89 pp. Clark, M. Lafevers, S. Estes, J. B. MacCormack, and T. Williams
Lindquist, K. G., and R. A. Hansen (1998). Seismic network improvements (2003). Aftershock sequence of the Mw 7.9 Denali, Alaska, earthquake
in Alaska: Broadband stations and digital telemetry, EOS Trans. 79, of 3 November 2002 from regional seismic network data, Seism. Res.
F567. Lett. 74, 743–752.
Lockner, D. A., and J. D. Byrlee (1991). Precursory AE patterns leading Reasenberg, P. A. (1985). Second-order moment of Central California Seis-
to rock fracture, in Proceedings 5th Conference on Acoustic Emisson/ micity, J. Geophys. Res. 90, 5479–5495.
Microseismic Activity in Geologic Structures and Materials, H. R. Reasenberg, P. A. (1999). Foreshock occurrence rate before large earth-
Hardy (Editor), Pennsylvania State University. Clausthal-Zellerfeld, quakes worldwide, Pageoph 155, 355–379.
Germany, Trans-Tech. Publications, 1–14, June 1991. Reasenberg, P. A., and M. V. Matthews (1988). Precursory seismic qui-
Lu, Z., T. J. Wright, and C. Wicks (2003). Deformation of the 2002 Denali escence: a preliminary assessment of the hypothesis, Pageoph 126,
Fault earthquakes, Alaska, mapped by Radarsat-1 interferometry, 373–406.
EOS Trans. 84, 425–431. Reasenberg, P., and D. H. Oppenheimer (1985). FPFIT, FPPLOT, and
Lu, Z., M. Wyss, and H. Pulpan (1997). Details of stress directions in the FPPAGE: Fortran computer programs for calculating and displaying
Alaska subduction zone from fault plane solutions, J. Geophys. Res. earthquake fault-plane solutions, U.S. Geol. Surv. Open File Rept. 85-
192, 5385–5402. 739, 37 pp.
Matthews, M. V., and P. Reasenberg (1988). Statistical methods for inves- Redfield, T. F., and P. G. Fitzgerald (1993). Denali fault system of southern
tigating quiescence and other temporal seismicity patterns, Pageoph Alaska: An interior strike-slip structure responding to dextral and sin-
126, 357–372. istral shear coupling, Tectonics 12, 1195–1208.
Michael, A. J. (1984). Determination of stress from slip data: faults and Reed, B. L., and R. L. Elliot (1968). Geochemical anomalies and metallif-
folds, J. Geophys. Res. 89, 11,517–11,526. erous deposits between Windy Fork and Post River southern Alaska
Michael, A. J. (1987). Use of focal mechanisms to determine stress: a con- Range, Alaska, U.S. Geol. Surv. Circ. 569, 22 pp.
trol study, J. Geophys. Res. 92, 357–368. Ridgway, K. D., J. M. Trop, W. J. Nokleberg, C. M. Davidson, and K. R.
Michael, A. J., W. L. Ellsworth, and D. Oppenheimer (1990). Co-seismic Eastham (2002). Mesozoic and Cenozoic tectonics of the eastern and
stress changes induced by the 1989 Loma Prieta, California earth- central Alaska Range: progressive basin development and deforma-
quake, Geophys. Res. Lett. 17, 1441–1444. tion in a suture zone, Geol. Soc. Am. Bull. 114, 1480–1504.
Nokleberg, W. J., D. L. Jines, and N. J. Silberling (1985). Origin and tec- Savage, J. C., and M. Lisowski (1991). Strain accumulation along the Den-
tonic evolution of the Maclaren and Wrangellia terranes, eastern ali fault at the Nenana river and delta river crossings, Alaska, J. Geo-
Alaska range, Alaska, Geol. Soc. Am. Bull. 96, 1251–1270. phys. Res. 96, 14,481–14,492.
Ogata, Y. (1999). Seismicity analysis through point-process modeling: a Savage, J. C., M. Lisowski, and W. Prescott (1981). Strain accumulation
review, Pure Appl. Geophys. 155, 471–507. across the Denali fault system in the Delta River Canyon, Alaska,
Oglesby, D. D., D. S. Dreger, R. A. Harris, N. A. Ratchkovski, and R. J. Geophys. Res. 86, 1005–1014.
Hansen (2004). Inverse kinematic and forward dynamic models of the Stanley, W., V. Labson, B. Nockleberg, B. Csejtey, and M. Fisher (1990).
2002 Denali, Alaska earthquake, Bull. Seism. Soc. Am. 94, no. 6B, The Denali fault system and Alaska Range of Alaska: Evidence for
S214–S233. underplated Mesozoic fysch from magnetotelluric surveys, Geol. Soc.
Ozacar, A. A., S. L. Beck, and D. H. Christensen (2003). Source process Am. Bull. 102, 160–173.
of the 3 November 2002 Denali fault earthquake (central Alaska) from Stout, J. H., and C. G. Chase (1980). Plate kinematics of the Denali fault
teleseismic observations, Geophys. Res. Lett. 30, 1638, doi 10.1029/ system, Can. J. Earth Sci. 17, 1527–1537.
2003GL017272. Thio, H. K., and H. Kanamori (1996). Source complexity of the 1994 North-
Page, R. A., G. Plafker, and H. Pulpan (1995). Block rotation in east-central ridge earthquake and its relation to aftershock mechanisms, Bull.
Alaska: A framework for evaluating earthquake potential? Geology Seism. Soc. Am. 86, S84–S92.
23, 629–632. Utsu, T. (1999). Representation and analysis of the earthquake size distri-
Plafker, G., G. Carver, M. Metz, and L. Cluff (2004). Repeated historic bution: a historical review and some new approaches, Pageoph 155,
surface ruptures of the Denali fault at Delta River, Alaska, during 509–535.
large earthquakes in 1912 and 2002, EOS Trans. AGU 85, 47, (Fall Utsu, T., Y. Ogata, and R. S. Matsu’ura (1995). The centenary of the Omori
Meet. Suppl.), G11A-0780. formula for a decay law of aftershock activity, J. Phys. Earth 43,
Plafker, G., L. M. Gilpin, and J. C. Lahr (1994). Neotectonic map of Alaska, 1–33.
in Geology of Alaska, Geology of North America, in Decade of North Waldhauser, F., and W. L. Ellsworth (2000). A double-difference location
American Geology, G. Plafker and H. C. Berg (Editors), Geological algorithm: Method and application to the northern Hayward fault,
Society of America, Vol. G-1, plate 12, 1 sheet, 1:2,500,000 scale. California, Bull. Seism. Soc. Am. 90, 1353–1368.
Plafker, G., W. J. Nokleberg, and J. S. Lull (1989). Bedrock geology and Wiemer, S. (2000). Introducing probabilistic aftershock hazard mapping,
tectonic evolution of Wrangellia, Peninsular, and Chugach terranes Geophys. Res. Lett. 27, 3405–3408.
along the Trans-Alaska Crustal Transect in the Chugach Mountains Wiemer, S. (2001). A software package to analyze seismicity: ZMAP,
and Southern Cooper river basin, J. Geophys. Res. 94, 4255–4296. Seism. Res. Lett. 72, 373–382.
Polat, O., H. Haessler, A. Cisternas, H. Philip, H. Eyidogan, M. Aktar, Wiemer, S., and K. Katsumata (1999). Spatial variability of seismicity pa-
M. Frogneux, D. Comte, and C. Gurbuz (2002). The Izmut (Kocaeli), rameters in aftershock zones, J. Geophys. Res. 104, 13,135–13,151.
S174 N. A. Ratchkovski, S. Wiemer, and R. A. Hansen

Wiemer, S., and M. Wyss (1994). Seismic quiescence before the Landers Zoback, M. L. (1992). First and second order patterns of stress in the lith-
(M ⳱ 7.5) and Big Bear (M ⳱ 6.5) 1992 earthquakes, Bull. Seism. osphere: the world stress map project, J. Geophys. Res. 97, 11,703–
Soc. Am. 84, 900–916. 11,728.
Wiemer, S., and M. Wyss (2000). Minimum magnitude of complete re- Zoback, M. L., and G. C. Beroza (1993). Evidence for near-frictionless
porting in earthquake catalogs: examples from Alaska, the Western faulting in the 1989 (M 6.9) Loma Prieta, California, earthquake and
United States, and Japan, Bull. Seism. Soc. Am. 90, 859–869. its aftershocks, Geology 21, 181–185.
Wiemer, S., and M. Wyss (2002). Mapping spatial variability of the Zoeller, G., S. Hainzl, J. Kurths, and J. Zschau (2002). A systematic test
frequency-magnitude distribution of earthquakes, Adv. Geophys. 45, on precursory seismic quiescence in Armenia, Natural Hazards 26,
259–302. 245–263.
Wiemer, S., M. C. Gerstenberger, and E. Hauksson (2002). Properties of Zuniga, F. R., and S. Wiemer (1999). Seismicity patterns: are they always
the 1999 Mw 7.1, Hector Mine earthquake: implications for aftershock related to natural causes?, Pure Appl. Geophys. 155, 713–726.
hazard, Bull. Seism. Soc. Am. 92, 1227–1240.
Woessner, J., and S. Wiemer (2004). Assessing the quality of earthquake Alaska Earthquake Information Center
catalogs: Estimating the minimum magnitude of completeness and its Geophysical Institute, University of Alaska Fairbanks
uncertainty, Bull. Seism. Soc. Am. (submitted for publication). 903 Koyukuk Drive
Woessner, J., E. Hauksson, S. Wiemer, and S. Neukomm (2004). The 1997 Fairbanks, Alaska 99775
Kagoshima (Japan) Earthquake doublet: quantitative analysis of af- (N.A.R., R.A.H.)
tershock rate changes, Geophys. Res. Lett. 31, L03605, doi 10.1029/
2003GL018858. Swiss Seismological Service
Wright, T. J., Z. Lu, and C. Wicks (2003). Source model for the Mw 6.7, Geophysical Institute
23 October 2002, Nenana Mountain Earthquake (Alaska) from InSAR, ETH-Hönggerberg
CH-8093 Zurich, Switzerland
Geophys. Res. Lett. 30, 1974, doi 10.1029/2003GL018014.
(S.W.)
Yoshida, A., H. Ito, and K. Hosono (1996). Precursory seismic quiescence
appearing along tectonic zone just before the occurrence of intraplate
earthquake, J. Geogr. 105, 15–25. Manuscript received 19 February 2004.

You might also like