J. Wind Eng. Ind. Aerodyn.

121 (2013) 116–130

Contents lists available at ScienceDirect

Journal of Wind Engineering

and Industrial Aerodynamics
journal homepage: www.elsevier.com/locate/jweia

Dynamic characteristics and wind-induced responses of a super-tall

building during typhoons
J. Yi, J.W. Zhang, Q.S. Li n
Department of Civil and Architectural Engineering, City University of Hong Kong, Hong Kong

art ic l e i nf o a b s t r a c t

Article history: This paper presents the field measurement results of dynamic characteristics and wind-induced
Received 22 September 2012 responses of a 420 m high tall building in Hong Kong during the passage of typhoons. The field data
Received in revised form such as wind speed, wind direction, acceleration and displacement responses were simultaneously and
28 July 2013
continuously recorded from a monitoring system installed in the super-tall building. The outputs of
Accepted 10 August 2013
Available online 14 September 2013
accelerometers and GPS mounted in the building are compared to examine the accuracy of the GPS. The
multipath effects are extracted by a combination of Empirical Mode Decomposition (EMD) and wavelet
Keywords: method, and then are removed by a high pass finite impulsive response (FIR) digital filter to improve the
Full-scale measurement performance of the GPS. After obtaining the accurate GPS′s data, the resonant and background responses
Super-tall building
of the super-tall building as well as their contributions to the total displacements are presented and
Typhoon
discussed. Finally, damping ratios of the building are evaluated by the random decrement technique,
Dynamic characteristics
Wind-induced response which demonstrate amplitude-dependent characteristics. The outcomes of this study are expected to be
Wind effect of interest and practical use to engineers and researchers who are involved in the wind-resistant design
GPS or structural health monitoring of super-tall buildings.
Empirical Mode Decomposition method & 2013 Elsevier Ltd. All rights reserved.
Wavelet method
Structural health monitoring

1. Introduction
The development of new materials and advanced construction
techniques in recent years has resulted in the emergence of many
super-tall buildings which are generally wind-sensitive structures.
A statistical report from the Council on Tall Building and Urban
Habitat (2011) shows that 37 tall buildings over 300 m have been
built throughout the world since 2000. Due to fast urban devel-
opments in some countries, more super-tall structures will be
built. For these high-rise structures, wind loads usually control
their structural design, resulting in a greater emphasis on under-
standing the structural behavior of super-tall buildings under
strong wind actions, in particular in typhoon or hurricane-prone
regions.
Although significant advances have been made for evaluation
of the wind effects on tall buildings by wind tunnel testing and
numerical simulation, many critical phenomena can only be
investigated by full-scale experiments on prototype structures.
Field measurement is still regarded as the most reliable way for
investigation of the wind effects on buildings and structures.
Numerous studies on wind effects on tall buildings have been
n
Corresponding author. Tel.: þ 852 34424677; fax: þ 852 27887612.
E-mail address: bcqsli@cityu.edu.hk (Q.S. Li).
carried out by field measurements over the past three decades
(Jeary, 1986; Ohkuma et al., 1991; Littler and Ellis, 1992; Tamura
and Suganuma, 1996; Campbell et al., 2007; Pirnia et al., 2007). In
particular, several full-scale measurement studies on the wind
effects on super-tall buildings are being conducted, including the
programs on four Chicago super-tall buildings by Notre Dame
University and the University of Western Ontario (Kijewski-Correa,
2003) and on ten super-tall buildings in Hong Kong, Taiwan and
mainland China by City University of Hong Kong (Li et al., 1998,
2003, 2004a, 2004b, 2005, 2006, 2011). However, opportunities to
conduct full-scale measurements are still quite rare, and thus the
data obtained are of significant value. Furthermore, it has been
recognized the importance of full-scale testing as a benchmark for
both wind tunnel and CFD modeling. Literature review reveals that
comprehensive full-scale measurements of wind effects on super-
tall buildings (buildings with a height 4400 m) have rarely been
conducted under typhoon or hurricane conditions. Thus, such a
database needs to be created.
A super-tall building with 420 m high and 88 floors, is located
in Central, Hong Kong Island. The sitting of the building is very
close to the seashore. As typhoons occur frequently in Hong Kong,
this super-tall building may be subjected to extreme horizontal
wind forces under typhoon conditions. Hence, a wind and move-
ment monitoring system has been installed in the building since
2007 to monitor the wind effects on the high-rise structure.

0167-6105/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jweia.2013.08.006
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130
Monitoring the performance of the super-tall building under harsh
typhoon conditions and the subsequent analysis of the data can
provide reliable information on the wind loading and the
responses of the super-tall building, which will be useful in
gaining a better understanding of the wind-resistant design and
habitability of modern tall buildings. Meanwhile, the measure-
ment results can also be used to improve wind tunnel test
techniques and numerical simulation methods.
This paper focuses on the investigation of the dynamic char-
acteristics (natural frequencies and damping ratios) and wind-
induced responses (acceleration and displacement) of the super-
tall building based on the field measurements from the wind and
movement monitoring system during typhoons. Accelerometers
have been widely used for measurements of structural vibration,
which are regarded as an effective tool to measure the resonant
responses of structures. However, the low frequency displacement
responses are difficult to be reconstructed by using in-situ mea-
sured accelerations (Tamura et al., 2002; Kijewski-Correa and
Kochly, 2007). For this reason, Global Positioning System (GPS)-
based techniques have been used to measure the mean and low
frequency deflections of civil structures in the past two decades
(Lovse et al., 1995; Ashkenazi and Roberts, 1997; Brown et al.,
1999; Fujino et al., 2000; Nakamura 2000; Wong et al., 2001;
Table 1
Three groups of sensors installed in three levels at the top part of the building.
Sensors
Anemometers: including one Gill Wind MasterPro ultrasonic
anemometer and two Young 05103 propellers
One GPS antenna: AX 1202 smart track antenna
Four accelerometers (A1, A2, A3, A4)
Fig. 1. Sensors installed atop the tall building: (a) propeller anemometers; (b) ultrasonic anemometer; (c) antenna of GPS; (d) locations of the accelerometers; and
(e) control system.
117
Tamura et al., 2002; Kijewski-Correa and Kochly, 2007; Li and Wu,
2007). It has been widely accepted that the combination of these
two types of measurements can provide comprehensive informa-
tion on these response components of structures. Hence, both
accelerometers and GPS were adopted in the monitoring system
installed in the super-tall building.
The aim of this study is to further the understanding of the
wind effects on the super-tall building during typhoons in order to
apply that knowledge to design similar high-rise structures in the
future. Meanwhile, it provides useful information for structural
health monitoring of super-tall buildings.
2. Introductions of the field measurement program
The monitored super-tall building is located in central, on the
north shore of Hong Kong Island. There are a great amount of
buildings including some tall buildings with heights of 70–150 m
on the north of the building in Kowloon where it is 1.5 km away
from the building location across the Harbor. In a sector spanning
from east clockwise to west, the building is surrounded by a large
number of tall buildings with heights of more than 150 m.
The northwest of the building corresponds to open-sea fetch.
Height level
Installed on a mast at height of 420.55 mPD, the highest level of the building. The diameter of the
mast at the location of anemometer is 0.2 m and the anemometer are installed away from the master
by about 1.5 m through a steel cantilever (Gill on one side and two propellers on the other side), at
about 14 m height above the roof of the building.
Installed on the roof with the installation height of 414.55 mPD
Fixed at the 88th floor (the top floor) at height of 400.10 mPD
118 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130

The super-tall building is an 88-storey office tower with five were continuously and simultaneously acquired and digitized at
additional basement levels going down to  32mPD. The aspect 20 Hz by a high-resolution digital data logger (NI-6281). The data
ratio of the building′s height to transverse width is about 8. The analyzed and presented in this paper were measured from 2008 to
general footprint of the building is about 57 m  57 m, while at the 2011, including the passages of five typhoons (see Table 2).
roof level this area is reduced to 39 m  39 m. The structural The GPS used in this study contained two stations: one at the
system consists of a central reinforced concrete core wall linked measurement point (measurement station) and another for refer-
by steel beams and outriggers to eight exterior composite mega- ence purpose (reference station). The measurement station with a
columns. Two secondary columns are located at each corner of the high-precision Leica GMX 902 GG GPS receiver and AX 1202 Smart
building to support the gravity load. Track antenna that can reduce the multipath effects was installed
The wind and movement monitoring system contains a data on the roof of the building. Meanwhile, the monitoring system
acquisition unit (DAU) controlling three groups of sensors as listed utilizes a continuously operating reference station established by
in Table 1. Photos of the sensors of the monitoring system are Survey and Mapping Office (SMO) of Hong Kong Lands Depart-
shown in Fig. 1. ment for the RTK-GPS measurements. This arrangement is differ-
Wind speed, wind direction atop the building, and the dis- ent from previous studies on the applications of GPS to monitor
placement and acceleration responses of the super-tall building the displacements of tall buildings, which has advantages of
obtaining higher quality outputs of GPS, but less cost. The
reference station selected in this study is located at Stonecutters
Table 2 Island, a former island in Victoria Harbor. An high-precision Leica
Typhoon information and the peak gust speeds measured during the typhoons. System GRX1200 Pro receiver with Leica AT504 Choke-ring
antenna were installed on a low-rise building which is about
Typhoons Date Peak gust (m/s) 4 km away from the measurement point (the building site), as
Propeller Ultrasonic
shown in Fig. 2. It is one of 12 continuously operating reference
stations of Hong Kong Satellite Positioning Reference Station
“Neoguri” April 18, 2008 30.1 30.2 Network (SatRef). The SatRef system (see Fig. 2a) has been
“Fengshen” June 24, 2008 40.4 42.9 developed by the SMO of Hong Kong Lands Department in recent
“Hagupit” September 23, 2008 39.2 42.9
years, which can produce the RTK correctional data stream round
“Koppu” September 14, 2009 40.0 42.2
“Nesat” September 28, 2011 32.4 33.1 the clock to the monitoring system installed on the super-tall
building through the Internet.

Fig. 2. (a) Hong Kong satellite positioning reference station network: Rp for reference point; Mp for measurement point and (b) reference station at Stonecutters Island.
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130 119

30
250

Mean Wind Direction (degree)

Mean Wind Speed (m/s)
25
200
20
150
15
100
10

5 50

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (hour) Time (hour)
Fig. 3. 10 min averaged wind speed and wind direction during Typhoon Koppu (on September 14, 2009).

8 8
Direction X Direction Y
6 6

Acceleration (gal)
Acceleration (gal)

4 4
2 2
0 0
-2 -2
-4 -4
-6 -6
-8 -8
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (hour) Time (hour)

6 6
S.D. Acceleration
Acceleration in Direction Y (gal)

S.D. Acceleration
Acceleration in Direction X (gal)

Peak Acceleration Peak Acceleration

5 5

4 4

3 3

2 2

1 1

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (hour) Time (hour)

Fig. 4. (a) Time series of acceleration responses during Typhoon Koppu (on September 14, 2009). (b) Variations of the peak and standard deviation values of acceleration
responses during Typhoon Koppu (on September 14, 2009).

6 6
S.D. Acceleration S.D. Acceleration
fitted curve fitted curve
Acceleration in Direction X (gal)

Acceleration in Direction Y (gal)

5 Peak Acceleration 5 Peak Acceleration

fitted curve fitted curve

4 4

3 3

2 2

1 1

0 0
10 15 20 25 10 15 20 25
Mean Wind Speed (m/s) Mean Wind Speed (m/s)
Fig. 5. Relationship between acceleration responses (the peak and standard deviation values) and the 10-min mean wind speed.
120 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130

3. Acceleration responses relationship, while the estimated results of c1 , c2 , c3 and c4 by

3.1. Structural dynamic responses during typhoons
Fig. 3 presents 10 min averaged wind speed and wind direction
atop the building during Typhoon Koppu, while Fig. 4 shows time-
histories of the acceleration responses at the top floor of the
building along its two main axes during the passage of the
typhoon. As expected, the responses increased with increasing of
the wind speed. The peak accelerations were 4.1 gal and 5.1 gal in
direction X and Y, respectively, which were recorded during the
strongest windstorm period of the typhoon.
Fig. 5 shows the relationships between the peak and standard
deviation values of 10-min acceleration responses and 10-min
mean wind speed (U) for southeast winds during Typhoon Koppu.
For the data presented in Fig. 5, the regression curves of the
acceleration response are expressed as follows:
sa ¼ c 1 U c 2
pa ¼ c 3 U c 4
where sa is the standard deviation of the response, pa is the peak
response, U is the mean horizontal wind speed, c1 , c2 ,c3 and c4 are
coefficients which (typically c2 and c4 ) reflect the sensibility of the
Table 3
Curve fitted parameters in Eq. (1).
Direction S.D. acceleration
c1 c2
X 7.589E  5 3.107
Y 6.916E  5 3.105
R2 is the square of the correlation between the response values and the predicted
response values.
Power Spectral Density
curve fitting method are listed in Table 3.
It can be seen from Fig. 5 that the increases in the dynamic
acceleration responses were associated with the increase in the
approaching mean wind speed.
3.2. Natural frequencies
To extract the modal parameters of the high-rise structure from
the dynamic response measurements, the Peak-Picking (PP)
method was used based on spectral analysis results of the dynamic
responses. Welch method was adopted for spectral estimation. In
the spectral analysis, the field data measured under steady wind
speed and wind direction conditions were adopted to determine
the dynamic characteristics of the super-tall building. Fig. 6 and
Table 4 show the natural frequencies of several sway modes of the
building. Meanwhile, the natural frequency of the fundamental
rotational mode was also identified based on the acceleration
ð1aÞ
response data. The spectral analysis results show that the accel-
eration responses of the super-tall building were primarily in the
ð1bÞ
two fundamental sway modes of vibration, while higher mode
components also made contributions.
4. Comparsion of the accelerations derived from the GPS data
and measured by the accelerometers
As introduced previously, the monitoring system installed in
Peak acceleration the super-tall building included a GPS. However, its deployment in
R2 c3 c4 R2
such a complex urban area required careful considerations of
several issues to ensure obtaining accurate outputs. Hence, it
0.833 7.766E  4 2.730 0.816 was necessary to validate the GPS data. In this study, the resonant
0.851 1.536E  3 2.455 0.818 responses of the GPS′s records are converted to accelerations via
the double differencing method, which are then compared with
the outputs of the accelerometers to examine the accuracy of the
GPS′s measurements.
Displacement data were band-pass filtered for the first mode
by the Kaiser FIR digital filter. Then the results were differentiated

102
Acceleration 1
in frequency domain based on the FFT and the inverse FFT (IFFT) to
Acceleration 3
0 generate the acceleration derived from the GPS data. On the other
10
hand, the acceleration data recorded by the accelerometers were
also band-pass filtered for the first mode. Comparisons of the
10-2 standard deviation values of the acceleration responses derived
from the GPS data with those recorded by the accelerometers are
10-4 plotted in Fig. 7. It can be seen from the figure that the acceleration
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
responses in the two orthogonal directions determined by the two
Frequence (HZ) kinds of devices are in good agreements, suggesting that the
102
Power Spectral Density

Acceleration 2 4
Acceleration 4 Direction X Acceleration
3 GPS
(gal)

100
2
acc

1
10-2
0
0 2000 4000 6000 8000 10000

10 -4 Time (second)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
4
Frequence (HZ) Direction Y Acceleration
3 GPS
(gal)

Fig. 6. The power spectral densities of the measured acceleration responses.

2
acc

1
Table 4 0
Natural frequencies (Hz) of the fundamental modes. 0 2000 4000 6000 8000 10000
Time (second)
Frequency (direction X) Frequency (direction Y) Frequency (rotation)
Fig. 7. Comparison of the standard deviations of accelerations measured by
0.140 0.144 0.368 the accelerometer and derived from the GPS data during Typhoon Hagupit (on
September 23, 2008).
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130 121

104 104
GPS GPS
Accelerometer Accelerometer

102 102
Power spectral density

Power spectral density

100 100

10-2 10-2

10-4 10-4

0.1 0.2 0.3 0.1 0.2 0.3

Frequence(Hz) Frequence(Hz)

Fig. 8. (a) Time histories of acceleration responses measured by the accelerometers and derived from the GPS data during Typhoon Hagupit (on September 23, 2008),
(b) correlation between the acceleration responses measured by the accelerometers and derived from the GPS data during Typhoon Hagupit (on September 23, 2008) and (c)
power spectral densities of the acceleration responses measured by the accelerometers and derived from the GPS data.
122 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130
precision of the GPS for the resonant response measurements is
satisfactory. Fig. 8(a) shows time-histories of acceleration
responses derived by the GPS data and measured by the accel-
erometers during Typhoon Hagupit, which illustrates that the two
sets of data match well. The extent of correlation between the
acceleration responses measured by accelerometer and derived
from the GPS data is demonstrated in Fig. 8(b), and clearly a
distinct complete correlation is evident. Fig. 8(c) shows the spectra
of the acceleration responses for the two fundamental sway
modes of vibration. Both the spectra of acceleration responses
measured by the accelerometer and derived from the GPS data in
Direction X have a peak at 0.140 Hz and have a peak at 0.144 Hz for
Direction Y, corresponding to the two fundamental natural fre-
quencies of the building. The cross-comparison of the results from
the two measurement devices verified the accuracy and reliability
of the GPS and the accelerometers.
5. Displacement responses by the GPS
Multipath effects are caused by a receiver received satellite
signals from several paths because of reflection or diffraction of
signals by surrounding objects. The reflected signals arrive with a
slight time delay, compared with the directly arrived signals,
resulting in a long period distortion in the displacement measure-
ments from GPS. As the differential GPS techniques can effica-
ciously remove or mitigate most of the error associated with
autonomous GPS surveying such as atmospheric and clock errors,
multipath error is the main issue for users of GPS to be eliminated,
especially when a GPS system is established in a complex urban
area. It may introduce errors of measurements up to several
meters (Lovse et al., 1995). For the monitoring system installed
in the super-tall building located in such a complex urban area, the
multipath effects on the accuracy of the displacement measure-
ments by the GPS may be significant. The following part will focus
on identifying and removing the multipath effects from the data
recorded from the GPS.
Providing a GPS antenna with a clear view of the sky and away
from the reflect surfaces should be efficient to ensure the best
possible accuracy. In addition, newly developed commercial hard-
ware and software are helpful for removing the multipath effects.
Moreover, there are several data post processing methods which
have been demonstrated to be useful to mitigate the errors caused
by the multipath effects, such as signal-to-noise (SNR) analysis
(Comp and Axelrad, 1998), FIR filters (Han and Rizos, 1997),
wavelet filters (Satirapod and Rizos, 2005), adaptive filters (Ge
et al., 2000), wavelet transform based on cross-validation method
(Zhong et al., 2008), EMD-Wavelet trend extraction model for
baseline solution (Wang et al., 2007).
5.1. Multipath trend extraction
Multipath has some special features which are helpful for
identifying and removing the error source (Kijewski-Correa and
Kochly, 2007). The first feature is that multipath brings in long
period or low frequency distortions in the records from GPS. In this
study, these low frequency trends are well below the natural
frequencies of the monitored structure. The second feature is the
identical multipath frequency components in the responses of the
super-tall building in both X and Y directions. Another feature is
the fact that the satellite configuration repeats itself every sidereal
day (every 23 h 56 min). Since the reflective surfaces causing
multipath remain unchanged, the multipath distortions should
also reoccur in the displacement measurement every sidereal day.
Based on the understanding of the multipath effects, a combi-
nation of Empirical Mode Decomposition (EMD) and Wavelet
method is adopted in this study to extract the trend of multipath
error. Previous research efforts (Wang et al., 2007) have demon-
strated the effectiveness of this strategy.
Empirical Mode Decomposition (EMD) is an adaptive time–
frequency analysis method proposed by Huang et al. (1998, 1999),
which has showed a great advantage for signal trend extraction
(Wang et al., 2007). It decomposes the given signal x(t) as the
following terms for signal trend extraction (Huang et al., 1998, 1999):
n
xðtÞ ¼ ∑ imf i ðtÞ þ r n ðtÞ; ð2Þ
i¼1
where imf i ðtÞ denotes the ith Intrinsic Mode Function (IMF), which
can be defined as: (1) for every single IMF, the number of extrema
and the number of zero-crossings must either be equal or differ at
most by one; (2) the envelope defined by the local maxima and the
envelope defined by the local minima are symmetric.r n ðtÞis the
residue of signal x(t), after n number of IMFs are extracted.
In practice, the EMD is carried out through a sifting process that
uses only the local extrema, and the procedure is as follows:
(1) Identify all extrema of x(t).
(2) Connect all these local maxima and minima with a cubic spline
as the upper and lower envelopes, emax ðtÞ and emin ðtÞ.
(3) Calculate mean value with mðtÞ ¼ ðemin ðtÞ þ emax ðtÞ=2Þ.
(4) Take the difference between x(t) and mean value: d(t) ¼x(t) 
m(t),
(5) Iterate steps (1) to (4) on d(t) until the envelopes are
symmetric with respect to zero mean certain criteria. The
obtained d(t) is referred to as an IMF.
(6) Calculate mðtÞ ¼ xðtÞimf i ðtÞ.
(7) Repeat steps (1) to (6) until no more IMFs are available.
Eventually, the original signal x(t) is first decomposed through
the main loop as
XðtÞ ¼ imf 1 ðtÞ þ m1 ðtÞ
and the first residual m1(t) is itself decomposed as
m1 ðtÞ ¼ imf 2 ðtÞ þ m2 ðtÞ
So that
XðtÞ ¼ imf 1 ðtÞ þ m1 ðtÞ ¼ imf 1 ðtÞ þ imf 2 ðtÞ þ m2 ðtÞ…
n
¼ ∑ imf i ðtÞ þmn ðtÞ ð3Þ
i¼1
Wavelet transform has been extensively used for denoising due
to its good localized time–frequency features. Wavelet shrinkage
noise reduction involves the following major steps (Taswell,
2000): forward transformation of the signal to the wavelet
domain, the obtained wavelet decomposition coefficients wj,k (j,k
represent the wavelet decomposition level and scale) are shrunk
by considering threshold rules based on the idea that only very
few coefficients contribute to the signal, and transformation of the
wavelet coefficients back to the original signal domain.
The hard-threshold and the soft-threshold for obtaining new
wavelet coefficients w^j;k , are presented as
Hard-threshold
(


wj;k ;
wj;k
Z λ
w^ j;k ¼

ð4Þ
0;
wj;k
o λ
Soft-threshold
(




signðwj;k Þð
wj;k
λÞ;
wj;k
Z λ
^ j;k ¼
w

ð5Þ
0;
wj;k
o λ
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130 123

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where λ ¼ s 2logðnÞ=n is a predefined threshold value, n is the
length of the discrete signal. s can be estimated by
s^ ¼ MAD=0:6754; MAD is the median of the absolute deviation
value. Thereafter, wavelet reconstruction algorithm is applied to
obtain the de-noised signal. Soft-threshold noise reduction is
adopted in this study.
The data sets recorded from the GPS over a period of 2 con-
secutive days were selected for the analysis so that every potential
satellite configuration can be observed twice and the repeatability
of the satellite constellations can be exploited to diagnose the
multipath effects. At the first stage, the records by the monitoring
system from 4 June to 5 June in 2008, while the wind speeds were
very low during the period, were used herein, as such data sets are
usually regarded to be idea for the multipath trend extraction. As
mentioned previously, multipath brings in long period or low
frequency distortions in the records, so at this stage we focused on
how to extract the slowly varying trend from the signals which
may be superimposed with random noises. The data were firstly
decomposed by the EMD method. The obtained IMFs were
distinguished as high frequency components and low frequency
components. The means of the accumulated standardized modes,
defined below, can be considered for the discrimination of high
frequency and low frequency components of IMFs (Huang et al.,
1998).
m1
hm ¼ meanf ∑ ½imf i ðtÞmeanðimf i ðtÞÞ=stdðimf i ðtÞÞg;
i¼1
where imf i ðtÞ was defined in Eq. (1), m makes hm depart sig-
nificantly from zero, then the scale m is fixed as to distinguish the
high frequency mode from the low one. The high frequency IMFS
are deal with by Wavelet Shrinkage noise reduction model in this
study. Then, the extracted multipath trend can be reconstructed as
m1 n
x^ ðtÞ ¼ ∑ wave shrinkðimf i ðtÞÞ ∑ imf i ðtÞ þr n ðtÞ
i¼1 i¼m
where waveshrink (.) stands for Wavelet Shrinkage noise reduction.
Since the full 24-h sidereal day’s data plot is too long to be
illustrated here, the results for a period of 2 h over the 2 consecutive
Direction X (cm)
days are shown in Fig. 9 for illustration purposes. Fig. 10 shows the
EMD decomposed results for the data recorded on 4 June, 2008
along direction X and 14 IMFs were obtained by the analysis
procedure introduced above.
The relationship between the means of the accumulated
standardized modes (MSAM) and the scale is given in Fig. 11.
Obviously, the value of MSAM departs from zero, when the IMF
index is greater than 8. So, the first 8 IMFs can be selected for
denoising with the wavelet shrinkage reduction model. Then, the
EMD reconstruction gives the extracted systematic error (Fig. 12).
The time-history of the extracted residual series shows long-
period oscillations which represent the typical phenomenon of
the multipath effects that occurs in both response components
and repeat over the consecutive sidereal days.
5.2. Time–frequency spectra
Traditional spectral analysis methods, such as FFT, are only
appropriate for stationary signals whose statistical properties
make no variation with time. But, non-stationary signals more
commonly exist in engineering practice. A typical feature of such
signals is the variation of characteristics with respect to time.
Hence, spectra of non-stationary signals should be depicted in
both time and frequency domains. Wavelet transform (WT) is a
popularly used technique to obtain time–frequency spectra
mrn ð6Þ (Kijewski and Kareem, 2003; Li and Wu, 2007). It has been
recognized that wavelet transform is very effective in analyzing
data with gradual frequency change (Huang et al., 1996) and has
the merit of a continuous expression of spectra in both time and
frequency domains. WT can be expressed as
Z þ1  
  1 tb
W ψ ða; b; X; ψÞ ¼ xðtÞ; ψ a;b ðtÞ ¼ pffiffiffiffiffiffi xðtÞψ n dt: ð8Þ
jaj 1 a
ð7Þ where Ψ is the basic wavelet function that satisfies certain
conditions, a is the dilation factor and b is the translation of the
origin. The frequency information in the WT is involved in the
dilation factor, which can be converted to traditional Fourier
spectrum through:
f a ðnÞ ¼ f c ðnÞf s ðnÞ=a ð9Þ
in which f a (.) is the converted traditional spectrum, f c (.) is the

6 central frequency of the adopted wavelet, and f s (.) is the sampling

04-06-2008
4
frequency, n denotes frequency in Hz.
2
Fig. 13 shows the WT results of the extracted residual series
0
during the two day’s observation. The energy distribution in both
-2
0 0.5 1 1.5 2 time and frequency domains can be clearly seen in the figure. The
majority of the energy is distributed below 0.02 Hz. There are
Direction X (cm)

6
4
05-06-2008 some frequency components (such as at 0.001 Hz and 0.003 Hz)
2 presented in both response directions and repeated in the second
0 sidereal day’s record. That was theorized to be a result of the
-2 multipath.
0 0.5 1 1.5 2
Time (hour)
5.3. Multipath removal by FIR filter
Direction Y (cm)

4
2 04-06-2008
0 Based on the understanding of the multipath effects as discussed
-2 above, the technique of filter is used herein to remove them.
-4
Filtering process can be conducted in both frequency and time
-6
0 0.5 1 1.5 2 domains. Frequency domain filters are based on FFT and the inverse
FFT (IFFT). They are regarded as efficient and are considerably easy
Direction Y (cm)

4
2 05-06-2008 to handle. But, due to spectrum leakage and abrupt truncation of
0 undesirable frequency contributions, the phase is to be distorted
-2
-4 inevitably. Therefore, the filtered signals could be hardly adjusted to
-6 keep a synchronous pace with the original signals.
0 0.5 1 1.5 2
A time domain filter, mathematically expressed as a differential
Time (hour)
equation, deals with time series of signal directly. This kind of
Fig. 9. Two-hour time histories of GPS records. filters can be divided into two categories: the infinite impulsive
124 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130

IMF1
0

-2
0 1000 2000 3000 4000 5000 6000 7000
2

IMF2
0

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF3

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF4

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF5

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF6

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF7

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF8

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF9

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF10

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF11

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF12

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF13

-2
0 1000 2000 3000 4000 5000 6000 7000
2
IMF14

-2
0 1000 2000 3000 4000 5000 6000 7000

Fig. 10. EMD decomposed results: IMF components of the GPS data in direction X recorded on 04-06-2008.

response (IIR) filters and the finite impulsive response (FIR) filters. But, there is no guarantee of yielding an accurate linear phase in
IIR filters adopt a recursive model. They generally call for fewer the frequency response function (FRF), resulting in that the
orders than the FIR filters. Thus, IIR filters perform more efficiently. problem of phase distortion still exists. By contrast, a FIR filter is
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130 125

0.05 reasonable than those presented previously. The large amount of

energy concentrated in the low-frequency range of the PSDs of the
0 measured displacement responses is cut down, while the back-
ground and resonant response components are left. The spectra
also show that the wind-induced responses of the building were
hm

-0.05
primarily in the two fundamental sway modes of vibration. The
fundamental natural frequencies in the X and Y directions can be
-0.1
identified from the spectra of the displacement responses and they
are 0.139 Hz and 0.144 Hz, respectively. They are in good agree-
-0.15 ment with 0.140 Hz and 0.145 Hz determined by the spectra of the
2 4 6 8 10 12
IMF index
acceleration responses, thus again verifying the accuracy of the
GPS and the accelerometers installed atop the super-tall building.
Fig. 11. Standardised Empirical Mean (SEM) of fine-to-coarse EMD reconstruction.

5.4. Resonant and background responses

Direction X (cm)

10
04-06-2008 Before GPS was applied in structural health monitoring, most
0 full-scale measurements of wind-induced responses of structures
focused on resonant response components, as accelerometers
-10 essentially only measure the resonant responses. Such information
0 0.5 1 1.5 2
is very useful for human comfort, which is of primary concern in the
Direction X (cm)

10 design of tall buildings. However, for the structural design of stress

05-06-2008
and strength, there are two other components that are also
0
important: the mean and the background quasi-static components
of the response, since the major contribution to wind-induced
-10
0 0.5 1 1.5 2 loading may be from these components. The GPS technique is
Time (hour) regarded as a viable means of obtaining the mean, background, and
resonant responses of structures. In order to further investigating
Direction Y (cm)

10
04-06-2008 the performance of the super-tall building under typhoon condi-
tions and how the background and resonant response components
0
contribute to the total displacement response, the data presented in
-10 Fig. 15 were divided into the resonant and background components
0 0.5 1 1.5 2 by a band-pass filter and low-pass filter, respectively. These two
Direction Y (cm)

10 components are plotted in Figs. 19 and 20, respectively. It was noted

05-06-2008
0 were approximately corresponding to the across-wind and along-
-10
0 0.5 1 1.5
Time (hour)
Fig. 12. Extraction of multipath trend for 2 h data during 2 consecutive days.
able to provide an accurate linear phase. As a result, the filtered
signal can be easily adjusted to behave synchronously with the
original signal. So, in this study, the Kaiser FIR digital filter was
adopted to deal with the multipath effects.
Fig. 14 shows the variation of the displacement responses over
12-h on June 24, 2008, during the passage of Tyoon Fengshen. For
the record, the wind was approaching from the southeast with
10-min mean values of 5–25.1 m/s (see Fig. 15). It is noted that the
displacement responses in X and Y directions were approximately
corresponding to the across-wind and along-wind responses,
respectively. Obviously, there were long-period distortions in the
time histories of the displacement responses, which were not
correlated with any similar fluctuations in wind speed or wind
direction. The power spectral densities of the displacement
responses of the building along the X and Y directions (Fig. 16)
demonstrate significant energy concentrations in the low-
frequency range, which are larger than the resonant peaks. This
phenomenon caused by the multipath effects should be removed
from the recorded signals. Hence, the measured data were high-
pass filtered with a cut-off frequency of 0.02 Hz, in order to
remove the low-frequency contents. After the filtering operation,
the time-histories of the displacement responses are plotted in
Fig. 17, along with their PSDs presented in Fig. 18. It can be
observed from the figures that the responses became more
that the responses in direction X and direction Y for this wind event
wind responses, respectively. It can be seen from Fig. 19 that the
resonant response in direction X (across-wind response) was larger
2
than that in direction Y (along-wind response), while this tendency
also exited for the background displacement responses (see Fig. 20).
This suggests that the across-wind responses of the super-tall
building including the resonant and background components may
be larger than those in along-wind direction. Furthermore, it can
also be observed from the figures that the resonant displacement
responses in the two directions were greater than those of the
background displacement responses.
To further analyze how much of each component contributes to
the total fluctuating displacement response, the contributions of
the background and resonant RMS displacements recorded by the
GPS during several typhoons in 2009 are plotted in Fig. 21. The
10-min mean wind speeds of these records were between 15 m/s
and 24.4 m/s, and the wind directions were between 1101 and
1301. As can be seen from the figure, the background portion of the
displacement responses was significant at lower wind speeds,
where it was about half of the total response. Even at higher wind
speeds, the background component was still noteworthy.
6. Damping estimation
In general, the dynamic response of a structural system is greatly
affected by the amount of damping exhibited by each mode of
vibration. The importance of damping is becoming increasingly
significant as modern tall buildings are becoming taller and more
flexible. However, as yet there is no widely accepted method
available for estimating damping ratios of a structure prior to
construction. Reliable evaluations of structural damping can be
126 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130

Fig. 13. Time–frequency spectra of extracted residual series in the two directions, obtained by wavelet transform, based on 2 h data recorded during 2 consecutive days.
Colors indicate the energy density of the signals (in db) (a) direction X (04-06-2008), (b) direction Y (05-06-2008), (c) direction Y (04-06-2008) and (d) direction Y (05-06-2008).
X Displacement• cm•

20

10

-10
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Time• hour•
Y Displacement• cm•

20

10

-10
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Time• hour•

Fig. 14. Measured displacement responses during Typhoon Fengshen (on June 24,
2008). Fig. 15. Wind direction for the wind data recorded on June 24, 2008.

obtained from experimental measurements of a prototype building
only. It is thus desirable to collect more damping data by full-scale
measurements, in particular from super-tall buildings, since there
are lacking such information in the literature.
The measured acceleration data were used to obtain the
dynamic characteristics of the building (damping, natural frequen-
cies, etc.). The random decrement technique was employed to
evaluate the amplitude-dependent damping ratios of this building.
The random decrement technique represents a quick and practical
method for establishing the non-linear damping characteristics.
The determination of the amplitude-dependent damping ratios for
the building included the following steps:
(1) As discussed above, the wind-induced response of the building
is primarily in the two fundamental sway modes of vibration,
but higher modes are also present. In order to obtain the
damping ratio of each mode, the measured signals of the
acceleration responses were band-pass filtered before proces-
sing the random decrement technique to remove the compo-
nents not concerned.
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130 127

104 104
Direction X Direction Y

103 103

Power spectral density

Power spectral density

102 102

101 101

100 100

10-1 10-1

10-2 10-2
10-2 10-1 100 10-2 10-1 100
Freuence (Hz) Freuence (Hz)
Fig. 16. Power spectral densities (PSDs) of the measured displacement responses during Typhoon Fengshen (on June 24, 2008): (a) PSD in X direction and (b) PSD in Y
direction.
X Displacement• cm•

15 directions are shown in Figs. 22 and 23, respectively. The obtained

10 damping ratios actually comprise both structural damping and
5 aerodynamic damping. Information on the amplitude-dependent
0 damping ratio obtained from the super-tall building should be
-5
very valuable since similar measurements are still very limited.
-10
The damping ratio curves (damping ratio versus amplitude)
-15
0 1 2 3 4 5 6 7 8 9 10 11 12 13 presented in these figures demonstrate non-linear energy dissipa-
Time• hour• tion characteristics of the building. The damping curves shown in
these figures demonstrate that the damping ratios vary with the
Y Displacement• cm•

15
10 increase of vibration amplitude. The overall tendency is that the
5 damping ratios increase with increase in amplitude. From the
0 measurements of damping, it appears that damping values within
-5 1.0–2.0% of critical are reasonable for wind-resistant analysis of
-10 super-tall buildings with similar height and structural system for
-15 serviceability consideration.
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Time• hour•

Fig. 17. Displacement responses after filtering the measured data recorded during 7. Concluding remarks
Typhoon Fengshen (on June 24, 2008).

The dynamic characteristics and wind-induced responses of

a super-tall building in Hong Kong were investigated based on
the in-situ measured data from the wind and movement
(2) There are several ways to achieve the random decrement monitoring system including the GPS, accelerometers and
signatures. In this study, a set of thresholds were used to find anemometers installed in the building during the passage of
segments for obtaining a set of random decrement signatures five typhoons. Some conclusions from this study are summar-
with different initial amplitudes. To obtain one random decre- ized as follows:
ment signature, it was required at least 2000 segments for
taking average. The length of each segment (lag points) was (1) The natural frequencies measured from the accelerometers
more than ten times of the period of the corresponding and the GPS during the typhoons were almost identical. In
vibration mode. addition, the time-histories and RMS acceleration responses
(3) Regarding the random decrement signature as a free vibration which were obtained from the two kinds of measurement
signal, the damping ratio was determined using the least- devices were in good agreement. These observations illu-
squares approximation accordingly. strated that the precision of the GPS was satisfactory to
capture the dynamic responses of the tall building.
The information on the damping ratios against vibration (2) Multipath effects brought in long period or low frequency
amplitude for the first and second sway modes in the X and Y distortions in the displacement records from the GPS, which
128 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130

104 104
Direction X Direction Y

103 103

102 102
Power spectral density

Power spectral density

101 101

100 100

10-1 10-1

10-2 10-2
10-2 10-1 100 10-2 10-1 100
Freuence (Hz) Freuence (Hz)
Fig. 18. Power spectral densities of the filtered displacement responses (a) PSD in X direction and (b) PSD in Y direction.
displacement• cm•
displacement• cm•

15 4
X background

10
2
X resonant

5
0 0
-5 -2
-10
-15 -4
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13

Time • hour• Time• hour•

displacement• cm•

4
displacement• cm•

15
X background

10 2
X resonant

5
0
0
-5 -2
-10
-4
-15 0 1 2 3 4 5 6 7 8 9 10 11 12 13
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Time• hour•
Time• hour•
Fig. 20. The background displacement responses during Typhoon Fengshen
Fig. 19. The resonant displacement responses during Typhoon Fengshen (on June
(on June 24, 2008).
24, 2008).

resulted in identical frequency distortion components in the well below the natural frequencies of the monitored structure.
responses in both X and Y directions. Meanwhile, the repeated A high pass Kaiser FIR digital filter with a cut-off frequency of
distortions occurred in the displacement measurement every 0.02 Hz can eliminate the multipath effects.
sidereal day. Based on the identified features of the multipath (5) The GPS can capture both the resonant and background
effects, a combination of Empirical Mode Decomposition displacement responses of the monitored structure, while
(EMD) and wavelet method was found to be able to extract the accelerometers can only provide the resonant responses.
the trend of the multipath errors efficiently. Although the measured resonant components of the displace-
(3) Through the time–frequency spectral analysis results, it was ment responses in the two directions (X and Y) were larger
found that the energy distribution of the multipath effects than those of the background displacement responses, the
concentrated in the low-frequency region for both directions contribution of the background responses to the total displa-
(X and Y). There were some low frequency components cement response cannot be ignored.
presented in both response directions and repeated in the (6) The random decrement technique was used to estimate the
second sidereal day’s observation. damping ratios of the super-tall building during typhoons. The
(4) It was found that the multipath effects induced frequency variations of the damping ratios with vibration amplitude
components were generally lower than 0.01 Hz, which are were presented. The damping ratio curves (damping ratio
 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130 129

Fig. 21. Contributions of the background and resonant RMS responses to the total fluctuating displacements (excluding the mean component).

3 1.4
Measured total damping in X Direction Measured total damping in X Direction
Fitted total damping in X Direction Fitted total damping in X Direction
2.5
1.2
Damping ratio (%)
Damping ratio (%)

2
1
1.5
0.8
1

0.6
0.5

0 0.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

Amplitude (gal) Amplitude (gal)

1.4
3 Measured total damping in Y Direction
Measured total damping in Y Direction Fitted total damping inY Direction
Fitted total damping in Y Direction 1.2
2.5
Damping ratio (%)
Damping ratio (%)

2 1

1.5
0.8

1
0.6
0.5
0.4
0 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
0 0.2 0.4 0.6 0.8 1 1.2 1.4 Amplitude (gal)
Amplitude (gal)
Fig. 23. (a) Variation of damping ratio with vibration amplitude (for the second
Fig. 22. (a) Variation of damping ratio with vibration amplitude (for the first sway sway mode in X direction). (b) Variation of damping ratio with vibration amplitude
mode in X direction). (b) Variation of damping ratio with vibration amplitude (for (for the second sway mode in Y direction).
the first sway mode in Y direction).

Acknowledgment

versus amplitude) demonstrated non-linear energy dissipation
characteristics of the building. The damping curves illustrated
that the damping ratios varied with vibration amplitude. The
overall tendency was observed to be that the damping ratios
increased with the increase of amplitude. From the measure-
ments of damping, it appeared that damping ratios within 1.0–
2.0% of critical are reasonable for the wind-resistant analysis of
super-tall buildings with similar height and structural system
for serviceability consideration.
The work described in this paper was supported by a grant
from the Research Grants Council of Hong Kong Special Adminis-
trative Region, China (Project no. CityU 117709) and a research
grant from National Natural Science Foundation of China (Project
no. 50778059). The authors are grateful to the owners and
management officials of the building for their supports to the
monitoring project. Thanks are due to Dr. L.H. Zhi, Prof. S.X. Zhang,
Dr. Z.M. Ma, Dr. Alex To, Dr. H.W. Li and Prof. Y.Q. Xiao for their
valuable contributions to this study.
130 J. Yi et al. / J. Wind Eng. Ind. Aerodyn. 121 (2013) 116–130
References
Ashkenazi, V., Roberts, G.W., 1997. Experimental monitoring of the Humber Bridge
using GPS. Proceedings of Institution of Civil Engineers, Civil Engineering 120,
177–182.
Brown, C, Karuna, R, Askenazi, V, Roberts, G, Evans, R., 1999. Monitoring of
structures using the Global Positioning System. Proceedings of Institution of
Civil Engineers, Structures and Buildings 134, 97–105.
Campbell, S., Kwok, K.C.S., Hitchcock, P.A., Tse, K.T., Leung, H.Y., 2007. Field
measurements of natural periods of vibration and structural damping of
wind-excited tall residential buildings. Wind and Structures 10 (5), 401–420.
Comp, C.J., Axelrad, P., 1998. Adaptive SNR-based carrier phase multipath mitigation
technique. Transactions of the Aerospace and Electronic Systems 34 (1),
264–276.
Council on Tall Building and Urban Habitat, Chicago, 2011. The tallest 20 in 2020:
entering the era of the megatall.
Fujino, Y., Murata, M., Okano, S., Takeguchi, M., 2000. Monitoring system of the
Akashi Kaikyo Bridge and displacement measurement using GPS. In: Proceed-
ings, Nondestructive Evaluation of Highways, Utilities, and Pipelines IV. SPIE,
Bellingham, Washington, pp. 229–236.
Ge, L., Han, S., Rizos, C., 2000. Multipath mitigation of continuous GPS measure-
ments using an adaptive filter. GPS Solutions 4 (2), 19–30.
Han, S., Rizos, C., 1997. Multipath effects on GPS in mine environments. In:
Proceedings of the International Congress of the International Society for Mine
Surveying, Fremantle, Australia, 2–6 November, pp. 447–457.
Huang, N.E., Shen, Z., Long, S.R., Wu, M.L., Shih, H.H., Zheng, Q., Yen, N.C., Tung, C.C.,
Liu, H.H., 1998. The Empirical Mode Decomposition and the Hilbert Spectrum
for nonlinear and non-stationary time series analysis. Proceedings of the Royal
Society of London A 454, 903–995.
Huang, N.E., Shen, Z., Long, R.S., 1999. A new view of nonlinear water waves – the
Hilbert spectrum. Annual Review of Fluid Mechanics 31, 417–457.
Huang, S.Y., Qi, G.Z., Yang, J.C.S., 1996. Orthonormal wavelets for system identifica-
tion. In: Proceedings of the 14th Modal Analysis Conference, vol. 1, pp. 696–701.
Jeary, A.P., 1986. Damping in tall buildings, a mechanism and a predictor. Earth-
quake Engineering and Structural Dynamics 14, 733–750.
Kijewski-Correa, T.L., 2003. Full-scale measurements and system identification: a
time–frequency perspective (Ph.D. thesis), The University of Notre Dame.
Kijewski, T., Kareem, A., 2003. Wavelet transforms for system identification:
considerations for civil engineering applications. Computer-Aided Civil and
Infrastructure Engineering 18, 341–347.
Kijewski-Correa, T., Kochly, M., 2007. Monitoring the wind-induced response of tall
buildings: GPS performance and the issue of multipath effects. Journal of Wind
Engineering & Industrial Aerodynamics 95, 1176–1198.
Li, Q.S., Fang, J.Q., Jeary, A.P., Wong, C.K., 1998. Full scale measurement of wind
effects on tall buildings. Journal of Wind Engineering and Industrial Aero-
dynamics 74–76, 741–750.
Li, Q.S., Xiao, Y.Q., Wong, C.K., Jeary, A.P., 2003. Field measurements of wind effects
on the tallest building in Hong Kong. The Structural Design of Tall and Special
Buildings 12, 67–82.
Li, Q.S., Xiao, Y.Q., Wong, C.K., Jeary, A.P., 2004a. Field measurements of typhoon
effects on a super tall building. Engineering Structures 26, 233–244.
Li, Q.S., Wu, J.R., Liang, S.G., Xiao, Y.Q., Wong, C.K., 2004b. Full-scale measurements
and numerical evaluation of wind-induced vibration of a 63-story reinforced
concrete super tall building. Engineering Structures 26, 1779–1794.
Li, Q.S., Xiao, Y.Q., Wong, C.K., 2005. Full-scale monitoring of typhoon effects on
super tall buildings. Journal of Fluids and Structures 20, 697–717.
Li, Q.S., Fu, J.Y., Xiao, Y.Q., Li, Z.N., Ni, Z.H., Xie, Z.N., Gu, M., 2006. Wind tunnel and
full-scale study of wind effects on China’s tallest building. Engineering
Structures 28, 1745–1758.
Li, Q.S., Wu, J.R., 2007. Time-frequency analysis of typhoon effects on a 79-storey
tall building. Journal of Wind Engineering and Industrial Aerodynamics 95,
1648–1666.
Li, Q.S., Zhi, L.H., Tuan, A.Y., Kao, C.S., Su, S.C., Wu, C.F., 2011. Dynamic Behavior of
Taipei 101 Tower: field measurement and numerical analysisJournal of Struc-
tural Engineering, ASCE 137143–155
Littler, J.D., Ellis, B.R., 1992. Full scale measurements to determine the response of
Hume Point to wind loading. Journal of Wind Engineering and Industrial
Aerodynamics 42, 1085–1096.
Lovse, J.W., Teskey, W.F., Lachapelle, G., Cannon, M.E., 1995. Dynamic deformation
monitoring of tall structure using GPS technology. Journal of Surveying
Engineering 121 (1), 35–40.
Nakamura, S., 2000. GPS measurement of wind-induced suspension bridge girder
displacements. Journal of Structural Engineering, ASCE 126 (12), 1413–1419.
Ohkuma, T., Marukawa, H., Niihori, Y., Kato, N., 1991. Full-scale measurement of
wind pressures and response accelerations of a high-rise building. Journal of
Wind Engineering and Industrial Aerodynamics 38, 185–186.
Pirnia, J.D., Kijewski-Correa, T., Abdelrazaq, A., Chung, J.Y., Kareem, A., 2007. Full-
scale validation of wind-induced response of tall buildings: investigation of
amplitude-dependent properties. In: Proceedings of 2007 Structures Congress,
ASCE, Long Beach, CA, pp. 1–10.
Satirapod, C., Rizos, C., 2005. Multipath mitigation by wavelet analysis for GPS base
station applications. Survey Review 38 (295), 2–10.
Tamura, Y., Suganuma, S., 1996. Evaluation of amplitude-dependent damping and
natural frequency of buildings during strong winds. Journal of Wind Engineer-
ing and Industrial Aerodynamics 59, 115–130.
Tamura, Y., Matsui, M., Pagnini, L.C., Yoshida, A., 2002. Measurement of wind
induced response of buildings using RTK-GPS. Journal of Wind Engineering and
Industrial Aerodynamics 90, 1783–1793.
Taswell, C., 2000. The what, how and why of wavelet shrinkage denoising.
Computing in Science and Engineering 2, 12–19.
Wang, J., Wang, J.L., Roberts, C., 2007. Reducing carrier phase errors with
EMD2Wavelet for precise GPS positioning. In: Proceedings of the Institute of
Navigation National Technical Meeting, pp. 919–928.
Wong, K.Y., Man, K.L., Chan, W.Y., 2001. Application of global positioning system to
structural health monitoring of cable-supported bridge. In: Proceedings, Health
Monitoring and Management of Civil Infrastructure Systems. SPIE, Bellingham,
Washington, pp. 390–401.
Zhong, P., Ding, X.L., Zheng, D.W., Chen, W, Huang, D.F., 2008. Adaptive wavelet
transform based on cross-validation method and its application to GPS multi-
path mitigation. GPS Solutions 12, 109–117.