Bulletin of the Seismological Society of America, Vol. 103, No. 2B, pp. 1336–1347, May 2013, doi: 10.

1785/0120110264

The 2011 M w 9.0 Tohoku Earthquake: Comparison

of GPS and Strong-Motion Data
by Rongjiang Wang, Stefano Parolai, Maorong Ge, Mingpei Jin,
Thomas R. Walter, and Jochen Zschau

Abstract Near-field ground-motion data are available in semi-real time either from
modern strong-motion or continuous Global Positioning System (GPS) networks,
allowing robust solutions for earthquake source parameters, which are useful for rapid
disaster assessment and early warning. These wide applications require the ground-
motion data to cover a very broad frequency band that, however, is usually not
available. This paper presents a case study on the 2011 M w 9.0 Tohoku earthquake,
showing how the ground-motion information from geodetic and seismic instrumen-
tations is complementary, and suggesting the joint use of both types of data, particu-
larly when the network coverage is sparse. First the strong-motion records from the
two Japanese networks, K-NET and KiK-Net, are analyzed using an automatic
empirical baseline correction tool. The static coseismic displacement data are obtained
by double integration and then used to derive the permanent slip distribution on the
earthquake fault. Comparisons with the corresponding GPS-based solutions yield a
quantitative estimation of uncertainties of the empirical baseline correction. Further-
more, a dozen nearby GPS and strong-motion station pairs are selected to demonstrate
that the information in their time series agrees with each other. Finally, methods for
combining both types of ground-motion observation systems are discussed, and the
wide applicability of this approach is highlighted.

Introduction
Transient and permanent ground movements caused by Therefore, the joint use of the GPS and strong-motion instru-
earthquakes are crucial for locating their source and for inves- ments can improve data coverage and can advance the devel-
tigating their rupture processes. The transient and dynamic opment of new strategies for data processing due to their
movements are usually measured by seismic sensors deployed complementary advantages (Emore et al., 2007; Bock et al.,
worldwide, resulting in velocity or acceleration seismograms, 2012). However, a systematic test and comparison of such
while permanent displacements are traditionally measured data streams as well as their joint influence on kinematic
through geodetic methods. source models has not yet been made.
On the one hand, the continuous and rapid development Japan has one of the densest GPS and seismometer
of Global Positioning System (GPS) hardware and the im- networks in the world, and presently includes the GeoNet,
provement in GPS data processing strategies have been such with 1200 permanent GPS stations, the F-Net, with 84 broad-
that estimating seismic displacements from high-rate GPS has band seismograph stations, the K-NET, with 1000 strong-
become a research topic referred to as GPS seismology (e.g., motion seismograph stations, the Hi-Net, with 777 high
Larson et al., 2003; Bock and Prawirodirdjo, 2004; Chen sensitivity seismograph stations (borehole installation), and
et al., 2004; Langbein and Bock, 2004; Miyazaki et al., 2004; the KiK-Net, co-located with the Hi-Net. Each KiK-Net sta-
Elósegui et al., 2006; Larson, 2009; Irwan et al., 2011). tion is equipped with two strong-motion seismographs at the
Currently, GPS seismology focuses on the acquisition of surface and borehole, respectively (see Data and Resources).
high-rate GPS data and the refinement of data processing ap- Additional data, such as InSAR, were also collected, though
proaches (Smalley, 2009). On the other hand, recent studies not further considered in this study. Nevertheless, after the
have shown that large near-source static displacements can 2011 Mw 9.0 Tohoku earthquake, an unprecedented amount
also be derived from high-quality accelerometer records after of ground-motion data were available from these observation
an empirical baseline correction (e.g., Graizer, 1979, 1983; networks, providing an ideal opportunity for evaluating the
Iwan et al., 1985; Boore, 2001; Wang, Boore, et al., 2003; consistency of the ground motion obtained by the different
Wu and Wu, 2007; Chao et al., 2009; Wang et al., 2011). observation instruments.

1336
Comparison of GPS and Strong-Motion Data of 2011 M w 9.0 Tohoku Earthquake 1337

In this study, we first focus on assessing the uncertain-

ties of the estimated displacement values, starting from
strong-motion data, and investigate their effect on the earth-
quake source inversion results. In particular, we present a
new approach to detect outliers in the derived dataset. This
allows us to evaluate whether a joint inversion of the two
datasets represents a valid option when networks are sparser
than in Japan. We also examine the validity of our results for
data coverage much sparser than in the present case. Finally,
we compare the displacement and velocity seismograms
derived from the high-rate GPS and the strong-motion re-
cords and show how they are complementary over a broad
frequency band.

Comparison between Static Coseismic Displacements

Derived from the Strong-Motion and GPS
Recordings
In order to obtain reliable estimates of permanent
displacement, we first analyze the complete strong-motion
dataset collected by both K-NET and KiK-net using the
automatic empirical baseline correction scheme proposed
by Wang et al. (2011). The empirical baseline correction
method is based on the same concept first proposed by Iwan
et al. (1985), in which a bilinear function is used to describe
the velocity trend induced by baseline errors in strong- Figure 1. GPS and strong-motion (SM) stations of the GeoNet,
motion records. Currently, several manual or automatic ver- the Ocean-Bottom GPS/Acoustic Network, the K-NET, and the
KiK-Net, respectively, from which the data for the 2011 Tohoku
sions of the method have been described (see Wang et al., earthquake used in this study originate. The star is the epicenter
2011 and references therein). The only difference between of the 2011 Tohoku earthquake (38.322° N, 142.369° E) as deter-
them is the choice of the two timing parameters of the mined by the USGS. The color version of this figure is available only
bilinear correction. In the method by Wang et al. (2011), the in the electronic edition.
timing parameters are determined in a way so that the cor-
rected displacement history best fits a step function. After the mainly for rapid response purposes, most stations are located
baseline correction, velocity and displacement seismograms in public offices, schools, parks and on free-surface of softer
are obtained through integration over time. The permanent
sediments than those of KiK-Net (Tsuda et al., 2006; Aoi
displacement, also commonly called coseismic displace-
et al., 2011). Possibly, the amplified response of the soft soil
ment, is then estimated from the postevent plateau of the cor-
below these stations led to large transient tilting of the accel-
rected displacement history. Although the latter might be
erometers (Graizer 2005, 2010). For this reason, coseismic
underrepresented in this routine, the strong-motion data
displacements derived from the K-NET data will not be further
may provide important information on the initial deformation
as described below. discussed. Some isolated anomalous displacement values that
Figure 1 shows a site map of the GeoNet, K-NET, and are not coherent with those obtained by nearby stations are
KiK-Net stations in the area of investigation. Coseismic dis- also observed for a few KiK-Net surface sensors. An example
placements are derived from K-NET, KiK-Net-SF (surface of such an anomaly is shown by the surface sensor at the KiK-
sensors), and KiK-Net-BH (borehole sensors), respectively, Net station FKSH18 (37.4894° N, 140.5380° E), which even
and are shown in Figure 2, where the high-precision GPS sol- shows an east–west displacement not consistent with the
utions are depicted as reference. In general, the KiK-Net-BH retrieved general trend (Fig. 3). We conjecture this kind of
dataset leads to the most stable displacement values (with anomaly to be mainly due to the nonlinear baseline shift
coherent spatial variability) that are also the closest to the affecting the strong-motion recordings, which cannot be
GPS observations. The displacement vectors show a central- properly accounted for using the bilinear correction scheme
ized direction to the earthquake source centroid as described (Fig. 3). Theoretically, the ground tilt does not affect the ver-
below. The maximum displacement in the order of about 5 m tical component so strongly. This is verified in the present
was observed in the coast near Sendai by both GPS and case. No extreme anomalous values appear in the vertical dis-
KiK-Net stations. The K-NET dataset shows a large spatial placement derived from both K-NET and KiK-Net datasets.
variability in the displacements, also between stations that Before comparing the source models in the following
are close to one another. In fact, as the K-NET was designed section, we compare the coseismic displacements derived
1338 R. Wang, S. Parolai, M. Ge, M. Jin, T. R. Walter, and J. Zschau

Figure 2. Coseismic displacement data derived from different GPS and strong-motion networks for the 2011 Tohoku earthquake, com-
pared with the values predicted by Models 1–3 (Fig. 4) inverted from the respective dataset. No model is derived from the K-NET dataset
because of its large uncertainty. The model for both the onshore and offshore GPS datasets is obtained by joint inversion. The color version of
this figure is available only in the electronic edition.

from the KiK-Net-BH data with those interpolated from the

GPS. The root mean square (rms) of the deviatoric displace-
ment vectors between the two datasets is 0.47 m, or 37% of
the rms of the absolute KiK-Net-BH displacement vectors,
which is 1.25 m. Therefore, given the large coseismic dis-
placement associated with the Mw 9.0 Tohoku earthquake,
those two datasets compare well and provide independent
observation points, which may have an effect on the source
models as shown below.

Comparison between Source Models Derived from

the GPS and Strong-Motion Derived Coseismic
Displacement Data
In order to assess the effect of the largest uncertainties in
the coseismic displacement data derived by strong-motion
recordings on the earthquake source modeling, we invert the
static slip distribution from the two KiK-Net datasets sepa-
rately, and we compare the obtained results with those
obtained by the high-precision GPS data.
The inversions are carried out using a self-developed
software (FORTRAN code “SDM” by R. Wang) based on
Figure 3. Ground velocity derived from the records of both sur- the constrained least-squares method, which has been used
face and borehole accelerometers at KiK-Net station FKSH18
(37.4894° N, 140.5380° E). The baseline corrections are estimated in a number of recent studies on geodetic slip models (Mo-
using the automatic scheme by Wang et al. (2011). The color tagh et al., 2008, 2010; Wang et al., 2009, 2011; Diao et al.,
version of this figure is available only in the electronic edition. 2010, 2011; Xu et al., 2010). For simplicity of numerical
Comparison of GPS and Strong-Motion Data of 2011 M w 9.0 Tohoku Earthquake 1339

analysis, the fault plane is represented by a number of small scatters between 30 and 60 m (e.g., Ammon et al., 2011;
rectangular dislocation patches with uniform slip. The Lay et al., 2011). The moment magnitude of the earthquake
observed displacement data are related to the discrete fault is estimated to be 8.90, in agreement with that given by the
slips through Green’s functions of the earth model, which global seismic network GEOFON. It should be noted that
are calculated using linear elastic dislocation theory. For the when calculating the seismic moment, we adopted the com-
discrete slips to be an adequate representation of the true con- monly used constant shear modulus μo  30 GPa instead of
tinuous slip distribution, the patch size must be reasonably using the depth dependent shear modulus given by the earth
small (Segall and Harris, 1987). In fact, if the available data model. If the latter is used, the geodetic moment magnitude
do not include enough information for determining the slip would reach 9.01. The reason for using the constant shear
distribution with the desired resolution, the inversion system modulus is that here we want to compare the different slip
becomes underdetermined. To overcome the problem of non- models in terms of slip magnitude instead of the associated
uniqueness and instability inherent in such an underdeter- seismic moment. Hence, the shear modulus used here only
mined system, a priori conditions (fixed fault geometry plays the role of a scaling factor.
and restricted variation range for the rake angle) and physical Using the best geodetic slip model as a reference, Model
constraints (smooth spatial distribution of slip or stress drop) 3 (see Fig. 4) derived from the KiK-Net-BH dataset is very
are considered (Du et al., 1998). satisfactory and considerably better than that from the
In the present case, we employ a slightly curved fault KiK-Net-SF dataset (Model 2, Fig. 4), which strongly under-
plane, parallel to the assumed subduction slab. The dip angle estimates the magnitude of the earthquake. It is interesting that
increases linearly from 10° on the top (ocean bottom) to 20° peak slip values as large as 50 m can only be obtained when
at about 80 km depth, which is consistent with that suggested seafloor GPS data are used. Without this data, the GPS-derived
by Pollitz et al. (2011). To avoid any artificial bounding slip model (Model 4) is very close to that from the KiK-Net-BH
effect, a large enough potential rupture area (approximately dataset (Model 3), with both showing similar patterns where the
650 km × 300 km) is used. The upper edge of the fault is peak slip is less than 25 m. The datafit of Model 3 to the onshore
located along the trench east of Japan, on the boundary be- GPS data is equally as good as the datafit of Model 1.
tween the Pacific plate and the Eurasian plate. The patch size
is about 10 km × 10 km. The rake angle determining the slip A Model-Based Approach for Detecting Outliers in
direction at each fault patch is allowed to vary between the Strong-Motion Derived Coseismic Displacements
90°
20°, corresponding to the upper bound of the uncer-
tainties of the teleseismic focal solutions. Green’s functions In the previous two sections we have shown, consistent
are calculated based on a multilayered elastic half-space with previous studies elsewhere, that the large coseismic dis-
model (Wang, Lorenzo-Martin, Roth, 2003; Wang et al., placements during the Japan 2011 Tohoku earthquake can
2006). The model parameters are chosen according to the also be retrieved reliably from strong-motion records using
database CRUST2.0 (Bassin et al., 2000). The weighting fac- an empirical baseline correction tool. Although the displace-
tor for the smoothing term, also called the smoothing factor, ment values obtained at a few individual stations may include
is optimized first for the inversion of the complete GPS data large uncertainties, a robust source solution can still be
including the offshore observations adopted from Sato et al. derived when the whole dense network is considered, pro-
(2011), and then is fixed for all other inversions. vided that the sensors are installed in boreholes or at least on
The inversion results are shown in Figure 4. Because of hard ground. In practice, such ideal installation conditions are
the high station density of all considered networks, we not always fulfilled, not only due to financial or logistic lim-
assume that differences in the results are related to the differ- itations, but also because many strong-motion stations were
ent quality of the input data for the inversion. The best installed intentionally on soft sediments and urban areas for
resolved slip model is assumed to be that derived from the engineering purposes. Therefore, large baseline errors may be
high-precision GPS data, including both onshore and off- induced by the amplified and transient tilt effect, and in
shore observations. In fact, this model indicates that the peak general they cannot fully be corrected using the empirical bi-
coseismic slip of the earthquake reached about 50 m which is linear correction method. Consequently, obvious outliers ap-
in agreement with previous results obtained by the joint pear in the strong-motion derived displacement data (see
inversion of onshore GPS and deep ocean tsunami data by Fig. 2). Because these outliers may affect the results of the slip
Simons et al. (2011), but larger than the peak slip of 33 m inversion, we propose a model-based approach to detect them.
obtained by Pollitz et al. (2011) who also used both the on- First we invert, for example, the total KiK-Net-SF data-
shore and offshore GPS data but a slightly different fault set to obtain a preliminary slip model. A certain observed
geometry (three connected rectangles) and approximated surface displacement in the input dataset is defined as an
the slip distribution by analytical Hermite-Gauss functions. outlier if its misfit to the predicted one oversteps a certain
Other recently published slip models that are only based on threshold. As an example, Figure 5 shows the two KiK-Net
the onshore GPS data show peak slips systematically below datasets after having removed the outlier vectors, which
35 m (e.g., Iinuma et al., 2011; Ozawa et al., 2011), whereas deviate from the model by more than 15° in direction. Here
those only based on teleseismic waveform data show large the 15° threshold in direction is chosen so that the filtered
1340 R. Wang, S. Parolai, M. Ge, M. Jin, T. R. Walter, and J. Zschau

Figure 4. Slip models inverted from different coseismic displacement datasets (OB, ocean bottom; SM, strong motion derived; SF, surface
sensor; and BH, borehole sensor). The data are shown in Figure 2. The selected KiK-Net surface and borehole data are shown in Figure 5.
Both horizontal and vertical components are used for the inversion. The contour step is 5 m. The color version of this figure is available only
in the electronic edition.

KiK-Net-SF data can still show a reasonable coherence of have shown that the strong-motion data derived displacement
their spatial variability. We do not filter the outlier displace- has an acceptable quality when compared with that directly
ments on the basis of the vector amplitude because it has less measured by GPS, the combination of the individual observa-
influence on constraining the source location. After filtering tion systems could improve the data coverage in those regions
the outliers, the statistical deviation of the KiK-Net-BH and thus reduce the uncertainties of earthquake source inver-
dataset from the GPS is reduced from 37% to 20% in the sion when only based on GPS data. To simulate such a situa-
present case. tion, we make a test in which 25 displacement vectors are
In areas with a high station density, the outliers can also randomly selected from the GPS and KiK-Net-SF datasets
be easily detected by comparison between neighboring sta- (without the outliers), respectively. As shown in Figure 6,
tions, but the model-based detection method also works for although only about 10% of the original data are used, the
earthquake source can be resolved satisfactorily. We attribute
inhomogeneous networks. It is worth noting that, when using
the result to the remaining azimuth coverage of the data and
the filtered data, we can improve the slip model considerably,
the improved data quality after filtering the outliers.
particularly with respect to the moment magnitude in the
case of the KiK-Net-SF dataset (Model 5 in Fig. 4). Because
Comparison between Time Series of GPS and
of their relatively small amplitudes, most of the vertical com-
Strong-Motion Records
ponents are within the uncertainty range of the method.
In most other regions of the world, both geodetic and seis- Another important utility of high-rate GPS and strong-
mic networks are generally sparser than in Japan. Because we motion networks is to provide near-field seismic waveform
Comparison of GPS and Strong-Motion Data of 2011 M w 9.0 Tohoku Earthquake 1341

data, which are useful for constraining the kinematic earth- source time function. Thus, we need to derive velocity or
quake source process. In practice, the low-frequency band of displacement seismograms from strong-motion records, for
waveform information is particularly useful for inversion of which the empirical baseline correction is also necessary.
Without the baseline correction, certain high-pass or band-
pass filters have to be applied to the strong-motion records,
leading to loss of the information about low-frequency as
well as static ground deformation that is important for esti-
mating the moment magnitude.
The uncertainties of the strong-motion derived velocity
or displacement seismograms can be best evaluated using co-
installed high-rate GPS as reference. Although Japan has the
densest GPS and strong-motion networks in the world, there
are rare stations where both types of instruments are co-in-
stalled. Nevertheless, we found more than a dozen collocated
(or nearly collocated) GeoNet and KiK-Net station pairs. Us-
ing the automatic baseline correction scheme by Wang et al.
(2011), we derive the velocity and displacement seismo-
grams from the strong-motion recordings and compare them
with the GPS data for all 14 station pairs with distance
≤ 4 km (Fig. 7). In only a few cases the strong-motion de-
rived displacement seismograms show significant deviations
from the high-rate GPS data, mostly in the period after the
permanent displacement has arrived. The worst results are
obtained for the borehole sensor of KiK-Net station FKSH07
(No. 14 in Fig. 7), where the derived east–west displacement
converges to an opposite final value with respect to the GPS.
This is partly due to the fact that in this case, the strong-mo-
tion records do not allow us to select enough pre-event and
post-event times, which are necessary for a successful em-
pirical baseline correction using the automatic scheme. In
all other cases, however, the velocity seismograms derived
from the high-rate GPS and strong-motion records agree sat-
Figure 5. Coseismic displacement data derived from strong- isfactorily with each other, supporting previous observations
motion records at selected KiK-Net stations, where the observed
3D displacement vector deviates from that predicted by Models 2 that the uncertainties of the empirical baseline correction can
and 3 in Figure 4, respectively, by less than 15° in direction. The color affect the derived velocity seismograms, but not as dramati-
version of this figure is available only in the electronic edition. cally as for the displacement seismograms (Boore, 2001).

Figure 6. Coseismic displacement data randomly selected from the GPS and KiK-Net-SF datasets (Fig. 2) and the associated slip model
by joint inversion. The color version of this figure is available only in the electronic edition.
1342 R. Wang, S. Parolai, M. Ge, M. Jin, T. R. Walter, and J. Zschau

Figure 7. Displacement and velocity seismograms of the 2011 Tohoku earthquake derived from the KiK-Net borehole strong-motion
records using the automatic baseline correction tool by Wang et al. (2011), compared with nearby GPS observations. The number for each
station pair corresponds to the number in the station map. The time window is uniformly 250 s. The color version of this figure is available
only in the electronic edition.

Joint Use of High-Rate GPS and Strong-Motion empirical baseline correction can be improved upon using
Records the known coseismic displacement as reference. We suggest
a modification of the criterion for choosing the timing param-
In previous studies dealing with other major earth- eters so that the final displacement is closest to the GPS refer-
quakes, such as the 1999 Chi-Chi (Taiwan), 2007 Tocopilla ence. Our results show that the use of the GPS reference can
(Chile), and 2008 Wenchuan (China) earthquakes (Wu and obviously improve the recovery of the displacement history
Wu, 2007; Wang et al., 2011), the high-rate GPS data from from the strong-motion records (Fig. 8), providing more
permanent stations were not available, but the static coseis- accurate low-frequency waveform data that are useful for
mic displacement of the strong-motion stations was well investigating the rupture process of the earthquake.
estimated from temporary GPS networks or InSAR observa- For nearly co-installed GPS and strong-motion stations,
tions. Therefore, it is interesting to investigate whether the the true ground-motion information at a broad frequency
Comparison of GPS and Strong-Motion Data of 2011 M w 9.0 Tohoku Earthquake 1343

Figure 8. Same as Figure 7, but the permanent GPS displacement is used as a reference in the baseline correction of strong-motion
records. The color version of this figure is available only in the electronic edition.

band can be retrieved by the joint use of the two instruments. ferentiations. In the following we propose a simple approach
As an example, we select the nearest station pair between the for the utility of joint GPS and strong-motion records.
strong-motion and GPS networks, that is, K-NET station We first remove the pre-event offset from the GPS data.
AKT006 (40.2152° N, 140.7873° E) and the GeoNet station Then we assume that the baseline error of each component of
0183 (40.2154° N, 140.7873° E), being separated by 20 m. the accelerometer is time dependent and can be expressed
Figure 9 shows clearly the problems of both observation generally by a linear function combined with a sinusoidal
techniques. The ground tilting leads to large erroneous trends series,
in the two horizontal displacement components obtained by

double integration of the strong-motion records, whereas the X
N
t
Bt  a  bt  cn sin 2nπ ; t∈0; T;
ground acceleration is strongly underestimated by the double n1
T
derivatives of the GPS displacement. Because of the low sam-
pling rate (1 Hz) of GPS, high-frequency content is missing where 0; T is the time window of the data, and a, b, and
and possible aliasing effects might appear after double dif- cn n  1; 2; …; N are free parameters to be determined.
1344 R. Wang, S. Parolai, M. Ge, M. Jin, T. R. Walter, and J. Zschau

Figure 9. (a) The 100 Hz strong-motion records (top) at K-NET Station AKT006, the velocity (middle) and displacement seismograms
(bottom) integrated from the strong-motion records, respectively. (b) The acceleration (top) and velocity seismograms (middle) differentiated
from the 1 Hz GPS time series (bottom) measured at GeoNet Station 0183, respectively. (c) The acceleration (top), velocity (middle) and
displacement seismograms (bottom) obtained using joint use of the strong-motion and GPS data. The thick gray curves in the right-top and
right-bottom panels are the estimated strong-motion baseline errors and GPS positioning errors, respectively. The time window is uniformly
225 s. The color version of this figure is available only in the electronic edition.

The linear term represents the permanent as well as the slow In principle, the larger the degree N, the better the
trend of the baseline, whereas the sinusoidal series, which strong-motion derived displacement fits the GPS data. On
starts and ends with zero, covers the transient baseline errors the other hand, N should be as small as possible to avoid
during the strong shaking period. Because the two stations any overweighting of the GPS in the joint processing because
are nearly co-located, their seismic movements should be al- the GPS is known to have larger high-frequency noise. Gen-
most the same for the great Tohoku earthquake that nucleated erally we start with N  0 and then, if necessary, increase it
at about 250 km distance. Thus, we propose to determine the until the required fitting is satisfied or a certain upper limit of
free parameters in a way where the displacement obtained by N is arrived. The results shown in Figure 9 are obtained by
double integration of the strong-motion recordings, after re- requiring the fitting error not larger than
1 cm for the two
moving the baseline errors, best fits the GPS data in a least- horizontal components and
2 cm for the vertical compo-
squares sense. The number of the free parameters for the nent, corresponding roughly to the measurement errors of
baseline correction depends on the required fitting error of the GPS. It turns out that N  5, 8, and 0 for the east, north,
the strong-motion derived displacement to the GPS data. and vertical components, respectively. Though the functions
Comparison of GPS and Strong-Motion Data of 2011 M w 9.0 Tohoku Earthquake 1345

used for the baseline correction (thick gray curves in the to 20% in the static surface displacement (Pollitz, 1996).
right-top panel, Fig. 9c) are very smooth, the misfits between Moreover, recent studies have shown that when the inversion
the strong-motion based displacements and the GPS data is made based on models incorporating depth-dependent
(thick gray curves in the right-bottom panel, Fig. 9c) are shear moduli, the derived seismic moment can be up to 40%
likely uncorrelated with the high-frequency seismic signals. greater because of deeper source location than that based on
Therefore, these misfits might be attributed to stochastic the uniform half-space model (Hearn and Bürgmann, 2005).
observation noises in both strong-motion data and GPS data, In order to determine the moment magnitude independently,
but also to some effects caused by different installation con- the effect of the layered crustal structure should be accounted
ditions between the two instruments. Figure 9 also shows the for when inverting the geodetic data. We note that a more
ground velocity (right-middle panel) recovered by the joint complex fault geometry may have an effect on the interpre-
use of the GPS and strong-motion data, which provides a ba- tation of the data.
sis for the broadband ground-motion data. On this basis, one After an empirical correction for the baseline errors, the
can obtain the ground displacement (right-bottom panel) by ground velocity and displacement histories are derived from
integration and the ground acceleration (right-top panel) by the strong-motion records through the integration over time.
differentiation, both with a resolution that is much higher Generally, the slip model results obtained from inverting the
than is derived from the geodetic and seismic observations KiK-Net data agree considerably better with those derived
separately. from the high-rate GPS observations than those from the
K-NET data. The latter appear to be strongly affected by site
Discussion and Conclusions effects (strong nonlinear trends in the recordings that could
be due to large and transient tilting of the sensor installed on
Japan has the densest geodetic and seismic networks in soft ground) and thus cannot be used until an efficient pro-
the world. In the case of major earthquakes, information cedure for correcting them is found. The velocity time his-
about ground motion is available near real time through these tories derived from the KiK-Net borehole records are barely
networks. Using an automatic empirical baseline correction influenced by the residual uncertainties still existing after the
scheme, we analyzed the accelerograms recorded by the empirical baseline correction and are more informative, be-
K-NET and KiK-Net strong-motion networks for the 2011 cause of the higher sampling rate (100 Hz), than the 1-Hz
Tohoku earthquake. The coseismic displacements derived GPS data for the ground-motion characterization in terms
from the KiK-Net borehole dataset are the closest to the high- of velocity and therefore energy.
precision GPS solutions. In comparison, the KiK-Net surface It is desirable that the ground-motion data cover a
dataset shows larger uncertainties, but still appears to be ro- frequency band as broad as possible so that they can be used
bust when the whole network is considered. Less confident for different geophysical and engineering investigations. In
results for the derived coseismic displacements are those order to verify if this is the case when strong-motion and GPS
from the K-NET dataset showing large spatial variability. data are considered, we made a comparison of the datasets in
Probably the strongly amplified soil response including sur- the acceleration, velocity, and displacement domains, which
face tilting (Bonilla et al., 2011) caused complicated and dominate different frequency bands of the surface deforma-
multiple baseline shifts for most of the K-NET sensors, as tion information. On the one hand, we integrated the KiK-
well as for some of the KiK-Net surface sensors, leading Net-BH strong-motion records after the empirical baseline
to the failure of the simple bilinear correction scheme. correction to obtain velocity and displacement seismograms.
We inverted the static slip distribution from the two On the other, we differentiated the high-rate GPS time series
KiK-Net datasets, respectively, and compared it with that to receive the velocity seismograms, but did not derive the
obtained from the high-precision GPS data. When the com- acceleration seismograms, since the low 1-Hz sampling rate
monly used constant shear modulus μo  30 GPa is adopted of the GPS data would only allow a comparison with a low-
to convert the slip to the seismic moment, the moment mag- pass filtered version of the accelerogram. Although theoreti-
nitude of the earthquake derived from either the strong- cally the highest sampling rate of GPS receivers nowadays
motion based or GPS coseismic displacement field is about may reach 50 Hz, it is expected that the derived acceleration
8.8 to 8.9, slightly less than the final CMT solution of 9.0 by would be very noisy and could not, with any real confidence,
the U.S. Geological Survey (USGS), but consistent with the be used as an alternative to the strong-motion recordings.
GEOFON value. Currently, most slip models derived by the We also attempted to evaluate, and verify, whether the
inversion of near-field geodetic data are based on a uniform baseline correction of the strong-motion records proposed by
elastic half-space mainly because it allows us to calculate Wang et al. (2011) can be improved upon when using the
Green’s functions in a closed analytical form. It is known GPS coseismic displacement as reference. The results we
that the static displacement due to a dislocation source in obtained are successful in terms of comparability of the
a uniform elastic half-space is dependent upon the Poisson velocity and displacement histories estimated from the GPS
ratio, but independent of the absolute value of the shear and strong-motion data, respectively. Furthermore, we dem-
modulus of the medium. This might not be valid for the lay- onstrated an example in which both high-rate GPS and
ered structure; neglecting this issue may lead to errors of up strong-motion data from a co-located site are combined to
1346 R. Wang, S. Parolai, M. Ge, M. Jin, T. R. Walter, and J. Zschau

recover the ground-motion information over a broad fre- Boore, D. M. (2001). Effect of baseline corrections on displacement and
quency band (0–100 Hz). It is expected that there will be response spectra for several recordings of the 1999 Chi-Chi, Taiwan,
earthquake, Bull. Seismol. Soc. Am. 91, 1199–1211.
increasingly co-installed high-rate GPS and strong-motion Chao, W., Y. Wu, and L. Zhao (2009). An automatic scheme for baseline
networks and the joint GPS and strong-motion data process- correction of strong-motion records in coseismic deformation determi-
ing will become useful for real-time inversion of the kin- nation, J. Seismol. 14, 495–504.
ematic source process and therefore of great impact for Chen, J., K. M. Larson, Y. Tan, K. W. Hudnut, and K. Choi (2004). Slip
history of the 2003 San Simeon earthquake constrained by combining
earthquake and tsunami early warning.
1-Hz GPS, strong motion, and teleseismic data, Geophys. Res. Lett. 31,
L17608, doi: 10.1029/2004GL020448.
Diao, F., X. Xiong, and R. Wang (2011). Mechanism of transient post-
Data and Resources seismic deformation following the 2001 M w 7.8 Kunlun (China)
earthquake, Pure Appl. Geophys. 168, 767–779.
The K-NET and KiK-Net strong-motion data for the Diao, F., X. Xiong, R. Wang, Y. Zheng, and H. Hsu (2010). Slip model of the
2011 Tohoku earthquake were provided by the National Re- 2008 Mw 7.9 Wenchuan (China) earthquake derived from the co-
search Institute for Earth Science and Disaster Prevention seismic GPS data, Earth Planets Space 62, 869–874.
(NIED) of Japan and are available from http://www.k‑net Du, Y., A. Aydin, and P. Segall (1998). Comparison of various inversion
techniques as applied to the determination of a geophysical deforma-
.bosai.go.jp/ and http://www.kik.bosai.go.jp/ (last accessed
tion model for the 1983 Borah Creek earthquake, Bull. Seismol. Soc.
June 2011), respectively. All original GeoNet GPS data were Am. 82, 1840–1866.
provided by the Geospatial Information Authority (GSI). The Elósegui, P., J. L. Davis, D. Oberlander, R. Baenaand, and G. Ekström
GPS coseismic displacement data produced by the ARIA (2006). Accuracy of high-rate GPS for seismology, Geophys. Res. Lett.
team at JPL and Caltech were downloaded from ftp:// 33, L11308, doi: 10.1029/2006GL026065.
Emore, G. L., J. S. Haase, K. Choi, K. M. Larson, and A. Yamagiwa (2007).
sideshow.jpl.nasa.gov/pub/usrs/ARIA/ (last accessed March
Recovering seismic displacements through combined use of 1 Hz
2011). The 1-Hz GPS data (of the GSI via Nippon GPS Data GPS and strong-motion accelerometers, Bull. Seismol. Soc. Am. 97,
Services Company [NGDS]) were downloaded from http:// 357–378.
rtgps.com/rtnet_dl_eq.php (last accessed April 2011). The Graizer, V. M. (1979). Determination of the true ground displacement by
teleseismic focal solution of the 2011 Tohoku earthquake using strong motion records, Izvestiya Acad. Sci. USSR, Phys. Solid
Earth (English edition) published by the American Geophysical Union
was adopted from http://geofon.gfz-potsdam.de/data/alerts/
and Geological Society of America, 15, no. 12, 875–885.
2011/gfz2011ewla/mt.txt (last accessed March 2011). The Graizer, V. M. (1983). The movement near the fault of the Gazli earthquake,
CRUST2.0 model was downloaded from http://igppweb. Izvestiya Acad. Sci. USSR, Phys. Solid Earth (English edition) pub-
ucsd.edu/~gabi/crust2.html (last accessed March 2011). lished by the American Geophysical Union and Geological Society
of America, 19, no. 3, 172–178.
Graizer, V. M. (2005). Effect of tilt on strong motion data processing, Soil
Acknowledgments Dynam. Earthq. Eng. 25, 197–204.
Graizer, V. M. (2010). Strong motion recordings and residual displacements:
We thank Walter Zürn and Adrien Oth for valuable discussion, and What are we actually recording in strong motion seismology? Seismol.
Kevin Fleming for proofreading of the manuscript. This work is partly sup- Res. Lett. 81, 635–639.
ported by the Towards Real-Time Earthquake Risk Reduction (REAKT) project Hearn, E. H., and R. Bürgmann (2005). The effect of elastic layering on
of the European Seventh Framework Programme (Grant Agreement Number inversions of GPS data for coseismic slip and resulting stress changes:
282862). Mingpei Jin is supported by the China Scholarship Council and Strike-slip earthquakes, Bull. Seismol. Soc. Am. 95, 1637–1653.
China Earthquake Administration. The comments of two anonymous reviewers Iinuma, T., M. Ohzono, Y. Ohta, and S. Miura (2011). Coseismic slip dis-
were very helpful for improving the quality and readability of the paper. tribution of the 2011 off the Pacific coast of Tohoku earthquake
(M 9.0) estimated based on GPS data—Was the asperity in Miyagi-
oki ruptured? Earth Planets Space 63, 643–648.
References Irwan, M., F. Kimata, K. Hirahara, T. Sagiya, and A. Yamagiwa (2011).
Measuring ground deformations with 1 Hz GPS data: The 2003
Ammon, C. J., T. Lay, H. Kanamori, and M. Cleveland (2011). A rupture Tokachi-oki earthquake (preliminary report), Earth Planets Space
model of the 2011 off the Pacific coast of Tohoku Earthquake, Earth 56, 389–393.
Planets Space 63, 693–696. Iwan, W., M. Moser, and C. Peng (1985). Some observations on strong-
Aoi, S., T. Kunugi, H. Nakamura, and H. Fujiwara (2011). Deployment of motion earthquake measurement using a digital acceleration, Bull.
new strong motion seismographs of K-NET and KiK-net, Geotech. Seismol. Soc. Am. 75, 1225–1246.
Geol. Earthq. Eng. 14, 167–186, doi: 10.1007/978-94-007-0152-6_12. Langbein, J., and Y. Bock (2004). High-rate real-time GPS network at
Bassin, C., G. Laske, and G. Masters (2000). The current limits of resolution Parkfield: Utility for detecting fault slip and seismic displacements,
for surface wave tomography in North America, EOS Trans. AGU 81, Geophys. Res. Lett. 31, L15S20, doi: 10.1029/2003GL019408.
F897. Larson, K. M. (2009). GPS seismology, J. Geod. 83, 227–233, doi: 10.1007/
Bock, Y., and L. Prawirodirdjo (2004). Detection of arbitrarily large dynamic s00190-008-0233-x.
ground motions with a dense high-rate GPS network, Geophys. Res. Larson, K. M., P. Bodin, and J. Gomberg (2003). Using 1 Hz GPS data to
Lett. 31, L06604, doi: 10.1029/2003GL019150. measure deformations caused by the Denali fault earthquake, Science
Bock, Y., D. Melgar, and B. W. Crowell (2012). Real-time strong-motion 300, 1421–1424.
broadband displacements from collocated GPS and accelerometers, Lay, T., C. J. Ammon, H. Kanamori, L. Xue, and M. J. Kim (2011). Possible
Bull. Seismol. Soc. Am. 101, 2904–2925, doi: 10.1785/0120110007. large near-trench slip during the great 2011 Tohoku (Mw 9.0) earth-
Bonilla, L. F., K. Tsuda, N. Pulido, J. Régnier, and A. Laurendeau (2011). quake, Earth Planets Space 63, 687–692.
Nonlinear site response evidence of K-NET and KiK-net records from Miyazaki, S., K. M. Larson, K. Choi, K. Hikima, K. Koketsu, P. Bodin, J.
the 2011 Tohoku earthquake, Earth Planets Space 63, 1–7. Haase, G. Emore, and A. Yamagiwa (2004). Modeling the rupture
Comparison of GPS and Strong-Motion Data of 2011 M w 9.0 Tohoku Earthquake 1347

process of the 2003 September 25 Tokachi-Oki (Hokkaido) earthquake Wang, L., R. Wang, F. Roth, B. Enescu, S. Hainzl, and S. Ergintav (2009).
using 1-Hz GPS data, Geophys. Res. Lett. 31, L21603, doi: 10.1029/ Afterslip and viscoelastic relaxation following the 1999 M 7.4 Izmit
2004GL021457. earthquake from GPS measurements, Geophys. J. Int. 178, 1220–1237.
Motagh, M., B. Schurr, J. Anderssohn, B. Cailleau, T. R. Walter, R. Wang, Wang, R., F. Lorenzo-Martín, and F. Roth (2003). Computation of defor-
and J.-P. Villotte (2010). Subduction earthquake deformation associ- mation induced by earthquakes in a multi-layered elastic crust—
ated with 14 November 2007, Mw 7.8 Tocopilla earthquake in Chile; FORTRAN programs EDGRN/EDCMP, Comput. Geosci. 29, 195–207.
Results from InSAR and aftershocks, Tectonophysics 490, 60–68. Wang, R., F. Lorenzo-Martín, and F. Roth (2006). PSGRN/PSCMP—A new
Motagh, M., R. Wang, T. R. Walter, R. Bürgmann, E. Fielding, code for calculating co- and post-seismic deformation, geoid and grav-
J. Anderssohn, and J. Zschau (2008). Coseismic slip model of the ity changes based on the viscoelastic-gravitational dislocation theory,
August 2007 Pisco earthquake (Peru) as constrained by Wide Swath Comput. Geosci. 32, 527–541.
radar observations, Geophys. J. Int. 174, 842–848. Wang, R., B. Schurr, C. Milkereit, Zh. Shao, and M. Jin (2011). An im-
Ozawa, S., T. Nishimura, H. Suito, T. Kobayashi, M. Tobita, and T. Imakiire proved automatic scheme for empirical baseline correction of digital
(2011). Coseismic and postseismic slip of the 2011 magnitude-9 strong-motion records, Bull. Seismol. Soc. Am. 101, 2029–2044.
Tohoku-Oki earthquake, Nature 475, 373–376, doi: 10.1038/ Wu, Y., and C. Wu (2007). Approximate recovery of coseismic deformation
nature10227. from Taiwan strong-motion records, J. Seismol. 11, 159–170.
Pollitz, F. F. (1996). Coseismic deformation from earthquake faulting on a Xu, C., Y. Liu, Y. Wen, and R. Wang (2010). Coseismic slip distribution of
layered spherical earth, Geophys. J. Int. 125, 1–14. the 2008 M w 7.9 Wenchuan earthquake from joint inversion of GPS
Pollitz, F. F., R. Bürgmann, and P. Banerjee (2011). Geodetic slip model of and InSAR data, Bull. Seismol. Soc. Am. 100, 2736–2749.
the 2011 M 9.0 Tohoku earthquake, Geophys. Res. Lett. 38, L00G08,
doi: 10.1029/2011GL048632.
Sato, M., T. Ishikawa, N. Ujihara, S. Yoshida, M. Fujita, M. Mochizuki, and
Helmholtz Centre Potsdam
A. Asada (2011). Displacement above the hypocenter of the 2011
GFZ German Research Centre for Geosciences
Tohoku-Oki earthquake, Science 332, 1395. Telegrafenberg
Segall, P., and R. Harris (1987). Earthquake deformation cycle on the 14473 Potsdam, Germany
San Andreas fault near Parkfield, California, J. Geophys. Res. 92, wang@gfz‑potsdam.de
10,511–10,525. parolai@gfz‑potsdam.de
Simons, M., S. E. Minson, A. Sladen, F. Ortega, J. Jiang, S. E. Owen, L. maor@gfz‑potsdam.de
Meng, J.-P. Ampuero, Sh. Wei, R. Chu, D. V. Helmberger, H. [email protected]
Kanamori, E. Hetland, A. W. Moore, and F. H. Webb (2011). The (R.W., S.P., M.G., T.R.W., J.Z.)
2011 Magnitude 9.0 Tohoku-Oki earthquake: Mosaicking the mega-
thrust from seconds to centuries, Science 331, 1421–1425.
Smalley, R. (2009). High-rate GPS: How high do we need to go? Bull. Western Yunnan Earthquake Prediction Test Site
Seismol. Soc. Am. 80, 1054–1061. China Earthquake Administration
Tsuda, K., J. Steidl, R. Archuleta, and D. Assimaki (2006). Site-response 1-1-301 Yuyuan Community
estimation for the 2003 Miyagi-Oki earthquake sequence considering Wenhua Road, Dali 671000
nonlinear site response, Bull. Seismol. Soc. Am. 96, 1474–1482. People’s Republic of China
Wang, G., D. M. Boore, H. Igel, and X. Zhou (2003). Some observations [email protected]
on colocated and closely spaced strong ground-motion records of (M.J.)
the 1999 Chi-Chi, Taiwan, earthquake, Bull. Seismol. Soc. Am. 93,
674–693. Manuscript received 15 September 2012