1 Damping Scaling Factors for Elastic

2 Response Spectra for Shallow Crustal

3 Earthquakes in Active Tectonic Regions:
4 “Average” Horizontal Component
5 Sanaz Rezaeian,a) Yousef Bozorgnia,b) M.EERI, I. M. Idriss,c) Hon.M.EERI,
6 Norman Abrahamson,d) M.EERI, Kenneth Campbell,e) M.EERI, and
7 Walter Silva,f) M.EERI

8 Ground motion prediction equations (GMPEs) for elastic response spectra are
9 typically developed at a 5% viscous damping ratio. In reality, however, structural
10 and nonstructural systems can have other damping ratios. This paper develops a
11 new model for a damping scaling factor (DSF) that can be used to adjust the 5%
12 damped spectral ordinates predicted by a GMPE for damping ratios between 0.5%
13 to 30%. The model is developed based on empirical data from worldwide shallow
14 crustal earthquakes in active tectonic regions. Dependencies of the DSF on poten-
15 tial predictor variables, such as the damping ratio, spectral period, ground motion
16 duration, moment magnitude, source-to-site distance, and site conditions, are
17 examined. The strong influence of duration is captured by the inclusion of
18 both magnitude and distance in the DSF model. Site conditions show weak influ-
19 ence on the DSF. The proposed damping scaling model provides functional forms
20 for the median and logarithmic standard deviation of DSF, and is developed for
21 both RotD50 and GMRotI50 horizontal components. A follow-up paper develops
22 a DSF model for vertical ground motion. [DOI: 10.1193/100512EQS298M]

23 INTRODUCTION
24 In seismic design, analysis, and hazard calculations for engineered facilities, ground-
25 motion prediction equations (GMPEs) are used to predict the intensity of ground shaking.
26 Traditionally, GMPEs for elastic response spectra are developed at a 5% reference damping
27 ratio (e.g., for NGA models, see Power et al. 2008). The damping ratio represents the level of
28 energy dissipation in structural, geotechnical, and nonstructural systems. Within the struc-
29 tural dynamics framework, two types of damping are usually considered: viscous and hys-
30 teretic. Our focus in this study is on the equivalent viscous damping ratio, which is an
31 idealized concept used to approximate the overall damping due to the many damping
32 mechanisms that are present in an actual structure (Chopra 2012). This measure is sometimes

a)
U.S. Geological Survey, Golden, CO, sanazr128@yahoo.com, srezaeian@usgs.gov
b)
Pacific Earthquake Engineering Research Center, University of California, Berkeley, CA
c)
University of California, Davis, CA
d)
Pacific Gas & Electric Company, CA
e)
EQECAT Inc., Beaverton, OR
f)
Pacific Engineering & Analysis, El Cerrito, CA

939
Earthquake Spectra, Volume 30, No. 2, pages 939–963, May 2014; © 2014, Earthquake Engineering Research Institute
 940 REZAEIAN ET AL.

33 used to approximately account for systems with hysteretic damping as well (e.g., Iwan and
34 Gates 1979).

35 Even though GMPEs are typically developed for a reference 5% damping ratio, in reality,
36 structures can have damping ratios other than 5%. The value of the damping ratio depends on
37 the structural type, construction material, and level of ground shaking, among other factors.
38 For example, base-isolated structures and structures with added energy-dissipation devices
39 can have damping ratios greater than 5%, while some nonstructural components can have
40 damping ratios less than 5%. As another example, the recent guidelines for performance-
41 based seismic design of tall buildings (PEER 2010) specify a damping ratio of 2.5% for
42 tall buildings at the serviceability hazard level. Generally, a lower damping ratio is expected
43 if the structure remains elastic; on the other hand, if the ground shaking is severe enough to
44 cause yielding or damage to the structural and nonstructural components, the damping ratio
45 in an equivalent linear analysis could increase significantly. Damping ratios for different
46 types of structures and ground motion levels are a subject of debate, but recommended values
47 and estimation techniques are available in the literature (e.g., Newmark and Hall 1982,
48 Regulatory Guide 1.61 2007, PEER/ATC72-1 2010).
49 If a system has a damping ratio other than 5%, for certain applications, such as analysis of
50 tall buildings, nuclear power plants, or base-isolated systems, the predicted 5% damped
51 ground motion intensity should be adjusted accordingly. Furthermore, there has been an
52 increased interest in such adjustment techniques in recent years due to the emergence of
53 direct displacement-based design approaches that require spectral ordinates with damping
54 values much larger than 5% (Priestley 2003). Such equivalent linearization approaches
55 have also been envisaged in recent building codes (e.g., Eurocode 8 (CEN 2004), FEMA
56 440 2005), increasing the need for models to adjust the 5% damped ground motion intensity.
57 An example of such adjustment techniques is the classic work of Newmark and Hall (1982),
58 or its variations, which have been used worldwide to scale design spectra for different damp-
59 ing ratios (Bozorgnia and Campbell 2004). The pioneering work of Newmark and Hall
60 (1982) was based on only 28 records from nine earthquakes that had occurred prior to
61 1973. In this paper, we use an extensive database of over 2,000 recordings from shallow
62 crustal earthquakes in active tectonic regions to develop a new model that can be used
63 to scale the 5% damped spectral ordinates predicted by a GMPE to spectral ordinates for
64 other damping ratios.
65 In the past two decades, a rather large number of studies have been conducted to obtain
66 response spectral ordinates for damping ratios other than 5%. A comprehensive summary of
67 the existing literature is provided in the report by Rezaeian et al. (2012). Few studies have
68 developed GMPEs that directly estimate the spectral ordinate at various levels of damping. In
69 these models, GMPE coefficients are different for each damping ratio. This is the approach
70 taken, for example, by Faccioli et al. (2004) and Akkar and Bommer (2007). A review of
71 similar methods (e.g., Trifunac and Lee 1989, Boore et al. 1993, Bommer et al. 1998, and
72 Berge-Thierry et al. 2003) is provided in Bommer and Mendis (2005). The GMPEs for multi-
73 ple damping levels can be used to adjust other existing or updated 5% damped GMPEs by
74 estimating scaling factors for any combination of their predictor variables. However, without
75 interpolation, this approach is limited to the discrete damping ratios covered by the original
76 GMPE and is dependent on its predictor variables. An alternative approach that has been
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 941

77 taken in the majority of the existing literature and building codes is to develop models for
78 multiplicative factors that scale the 5% damped spectral ordinates predicted by any GMPE to
79 ordinates for other damping ratios. We follow the second approach and define the damping
80 scaling factor (DSF) as
PSAβ%
DSF ¼ (1)
PSA5%
EQ-TARGET;temp:intralink-;e1;62;591

81 where β denotes the damping ratio of interest (in percentage of critical damping), and
82 PSA represents the elastic pseudo-spectral acceleration at a damping ratio indicated in
83 the subscript.

84 We divide various methods in the literature for the DSF modeling into three categories.
85 The first category is methods based on random vibration theory (e.g., McGuire et al. 2001).
86 The procedure recommended by McGuire et al. (2001) uses different formulas for different
87 ranges of spectral period. For periods between 0.2 s to 1 s, the DSF is based on the procedure
88 developed by Rosenblueth (1980) and depends on the damping ratio, spectral period, and
89 duration of motion. For periods shorter than 0.2 s, the DSF is based on the procedure devel-
90 oped by Vanmarcke (1976) and depends additionally on peak ground acceleration (PGA).
91 A shortcoming of this method is that it is only applicable to periods less than 1 s. The second
92 set of methods is based on analytical studies that examine the dependence of the DSF on
93 various parameters. For example, Cameron and Green (2007) examined the analytical
94 response of a single-degree-of-freedom elastic oscillator to finite-duration, sinusoidal excita-
95 tions in order to show the dependence of the DSF on frequency content and duration of
96 motion. The third set of methods is based on empirical evaluations. Building code guidelines
97 are usually based on empirical models. Reviews of empirical models and building code
98 guidelines are presented in Naeim and Kircher (2001), Bozorgnia and Campbell (2004),
99 Lin et al. (2005), Bommer and Mendis (2005), Mendis and Bommer (2006), Cameron
100 and Green (2007), and Stafford et al. (2008).
101 We empirically develop a predictive equation of the following generic form
EQ-TARGET;temp:intralink-;e2;62;285 lnðDSFÞ ¼ μðβ; T; earthquake; site; bÞ þ ε (2)
102 where μ represents the mean of lnðDSFÞ, which is a function of the damping ratio β, the
103 spectral period T, and various earthquake and site characteristics such as earthquake mag-
104 nitude, source-to-site distance, and site conditions; b is the vector of regression coefficients;
105 and ε represents the error term that is assumed to be normally distributed with zero mean.
106 To identify the predictor variables, we first select those variables that have obvious phy-
107 sical influence on the DSF (i.e., damping ratio, period, duration, etc.) based on analytical
108 studies (e.g., random vibration theory, approximate analytical studies similar to that of
109 Mendis and Bommer 2006 or Cameron and Green 2007). Then, we extract patterns and
110 examine the trends between the DSF and various variables in our database. We start
111 with the variables already identified in the literature to have influence on the DSF. β is a
112 common predictor variable in all existing empirical models, and in fact, it is the only pre-
113 dictor variable in Ashour (1987), Tolis and Faccioli (1999), and Priestley (2003). Another
114 important predictor variable considered in the majority of existing models is period T (e.g.,
115 Newmark and Hall 1982, Wu and Hanson 1989, Idriss 1993, and Lin and Chang 2003).
 942 REZAEIAN ET AL.

116 Abrahamson and Silva (1996) additionally included the earthquake magnitude as one of their
117 predictor variables. Lin and Chang (2004) and Cameron and Green (2007) considered the
118 influence of site effects on the DSF. In more recent studies, the effects of duration, magni-
119 tude, and distance on the DSF have been the subject of interest. However, the majority of
120 these recent studies (e.g., Naeim and Kircher 2001, Atkinson and Pierre 2004, Bommer and
121 Mendis 2005, Cameron and Green 2007) limited their results to tabulating or plotting the
122 DSF for various magnitude–distance bins, different soil conditions, or different tectonic set-
123 tings, and they did not provide a single unified predictive equation for the DSF. Mendis and
124 Bommer (2006) and Stafford et al. (2008) directly included a measure of duration in their
125 predictive equation, but T was not a predictor variable in their models. (These models are
126 only applicable for intermediate periods.)
127 According to the literature reviews (see Rezaeian et al. 2012 for examples), there are
128 significant disagreements among the existing empirical models. One should keep in mind
129 that different models have used different databases and considered different ranges of the
130 damping ratio and spectral period, which may contribute to the discrepancies. Despite
131 these discrepancies, the majority of the models qualitatively agree on the general trends
132 of the DSF with potential predictor variables.
133 This paper starts by describing the database of strong ground motion records that are used
134 in this study for empirical modeling. This is followed by a summary of the observed depen-
135 dencies between the DSF and potential predictor variables. Next, models for the median DSF
136 and its logarithmic standard deviation for the horizontal component of ground motion are
137 proposed. Finally, the proposed model is compared to data and several existing models
138 in the literature.

139 GROUND MOTION DATABASE

140 A comprehensive new database of over 8,000 three-component recordings from over 300
141 worldwide earthquakes with moderate to large magnitudes has been developed for the NGA-
142 West2 project (see Ancheta et al. 2012a and 2012b, and Bozorgnia et al. 2012). NGA-West2
143 is a research program supported by the Pacific Earthquake Engineering Research Center
144 (PEER) to update the 2008 Next Generation Attenuation (NGA) GMPEs for shallow crustal
145 earthquakes in active tectonic regions (Power et al. 2008, Bozorgnia et al. 2012). In the NGA-
146 West2 database, ground motions recorded in numerous events since 2003 have been added;
147 thus, the new database is larger than that in NGA by a factor of almost 2.2. In this database,
148 elastic response spectra for the horizontal and vertical components have been calculated for
149 11 damping ratios: β ¼ 0.5%, 1%, 2%, 3%, 5%, 7%, 10%, 15%, 20%, 25%, and 30%. Our
150 focus in this paper is on horizontal components. (For details on the record processing and
151 computation of spectral values, refer to Ancheta et al. 2012b).
152 The elastic response spectra for horizontal components in the NGA-West2 database
153 include the two “as-recorded” horizontal components, as well as the GMRotI50 (Boore
154 et al. 2006) and RotD50 (Boore 2010) horizontal components. The last two are representa-
155 tives of the “average” horizontal ground motion and are independent of the in-situ orientation
156 of a seismometer. GMRotI50 is based on the “geometric mean” of the two individual
157 horizontal components, and a single rotation angle is used for all oscillator periods (i.e.,
158 period-independent). RotD50 is obtained without computing geometric means and uses a
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 943

159 period-dependent rotation angle. While GMRotI50 was used in the development of the
160 original NGA models (Power et al. 2008), RotD50 will be used in the development of
161 the NGA-West2 models (Bozorgnia et al. 2012). In this paper, we develop DSF models
162 for GMRotI50 and RotD50 horizontal components. A model for the DSF of the vertical
163 component is developed in a follow-up paper, a summary of which is presented in the report
164 by Rezaeian et al. (2012).
165 We use the 2011 version of the NGA-West2 database, which is the latest (at the time of
166 our work) version of the database for multiple damping ratios other than 5%. To ensure a
167 proper damping scaling model for the near-source data, 2,250 records from 218 earthquakes
168 with closest distance to rupture, Rrup , of less than 50 km are selected to develop the model.
169 The validity of the model for distances beyond 50 km is later verified by examining residuals
170 of the remaining records. The moment magnitude, M, of the selected database ranges
171 between 4.2 to 7.9. The magnitude–distance distributions of the NGA-West2 database
172 and the selected records are shown in Figure 1.
173 To measure ground motion duration, we use D5–75 , the significant duration of motion
174 between cumulative values of 5% and 75% of Arias intensity. Another duration measure,
175 D5–95 , is also considered and mentioned later in this paper. These measures are calculated
176 for each record in the database to examine the expected dependence of the DSF on the dura-
177 tion of motion. We estimate durations of GMRotI50 and RotD50 components by taking the
178 arithmetic average of D5–75 for the two “as-recorded” horizontal components. This measure
179 of duration for the selected records in the database ranges between 0.25 s to 59.32 s with a
180 mean of about 7.5 s.
181 The DSF is calculated according to Equation 1 for each record in the database at all
182 11 damping ratios and at the 21 spectral periods considered in the NGA project:
183 T ¼ 0.01 s, 0.02 s, 0.03 s, 0.05 s, 0.075 s, 0.1 s, 0.15 s, 0.2 s, 0.25 s, 0.3 s, 0.4 s,

Figure 1. Magnitude–distance distribution of the (a) NGA-West2 database, and (b) selected
database.
 944 REZAEIAN ET AL.

184 0.5 s, 0.75 s, 1 s, 1.5 s, 2 s, 3 s, 4 s, 5 s, 7.5 s, and 10 s. Validity of the calculated PSA at long
185 periods depends on the filtering process for each record. If the filter corner period is not
186 adequate for the period of interest—that is, the maximum usable period is less than the period
187 of interest (see Akkar and Bommer 2006 or Ancheta et al. 2012b)—the record is eliminated
188 for the long period calculations of the DSF.

189 PREDICTOR VARIABLES

190 As previously mentioned, the predictor variables in Equation 2 are identified by exam-
191 ining the dependence of the DSF on various variables in our database. We consider the fol-
192 lowing variables: the damping ratio β, which is a common predictor variable in all existing
193 empirical models; the spectral period T, which is considered in the majority of existing mod-
194 els; duration, magnitude, and distance, which have been the subject of interest in more recent
195 studies; and site conditions, which have been considered in very few studies.
196 The most fundamental predictor variables for the DSF are β and T. Dependence of the
197 DSF on these two variables is apparent from the definition of the DSF and the dependence of
198 PSA on T. However, different degrees of dependence have been reported in the literature. For
199 example, mild, weak, and very weak dependence on T has been reported by Stafford et al.
200 (2008), Bommer and Mendis (2005), and Naeim and Kircher (2001), respectively. It is
201 important to note that these studies focus on intermediate period ranges. Statistical analysis
202 of our database reveals systematic patterns between the DSF, and these two variables as seen
203 in Figure 2. In this figure, the median of the DSF for all recordings in our database is cal-
204 culated and plotted for various β and T. (Similar trends are observed if the data are divided
205 into smaller magnitude and distance bins.) There is almost no dependence on T between
206 0.2–2 s for β ≥ 2%; however, there is a strong dependence outside this period range
207 until the DSF approaches unity for very short and very long T (Figure 2a). This is expected
208 because the forces in a very stiff or a very flexible system are relatively independent of the
209 damping ratio. The dependence on T is much stronger for β ≤ 1%.

Figure 2. Influence of spectral period (T) and damping ratio (β) on DSF. Median DSF is plotted
against (a) period for various damping ratios and (b) damping ratio for various periods.
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 945

210 Duration of the ground motion can be an important factor controlling the DSF, as the
211 number of energy dissipating cycles can be influential. The influence of duration on the DSF
212 is demonstrated both empirically and through fundamental dynamic analysis by Bommer and
213 Mendis (2005), Cameron and Green (2007), and Stafford et al. (2008). Our data show an
214 increase in the DSF with D5–75 for β < 5% and a decrease for β > 5%. Figure 3 shows the
215 data at T ¼ 1 s and β ¼ 2% and 20%, along with a fitted line that shows the linear correlation
216 between DSF and logðD5–75 Þ for visual purposes. In general, the dependence of the DSF on
217 duration becomes stronger as T increases and as β deviates from 5%.
218 Explicit inclusion of duration in the model is not ideal in practice because duration is
219 usually not specified as part of a seismic design scenario. In general, there is a strong positive
220 correlation between duration and earthquake magnitude, and a moderate positive correlation
221 between duration and distance (e.g., Kempton and Stewart 2006, Bommer et al. 2009).
222 Therefore, we consider whether the influence of duration on the DSF can be captured by
223 including magnitude and distance in our DSF model. Our data show a strong dependence
224 between DSF and M. An example is shown in Figure 4, at T ¼ 1 s and β ¼ 2% and 20%.

Figure 3. Influence of duration on DSF at T ¼ 1 s.

Figure 4. Influence of magnitude on DSF at T ¼ 1 s.

 946 REZAEIAN ET AL.

Figure 5. Influence of distance on DSF at T ¼ 1 s.

225 A fitted line shows the linear trend in each plot. Similar to D5–75 , the dependence on M is
226 more pronounced at longer T and as β deviates from 5%. Similar patterns, but far less sig-
227 nificant, are seen between DSF and Rrup (see Figure 5). By performing regression analysis
228 and scrutinizing the residuals, we find that most of the influence of duration on the DSF can
229 be captured through inclusion of M in the model. Furthermore, we find that despite the weak
230 influence of distance, some of the residual effects of duration left after including M can be
231 captured by including Rrup in the model.
232 Influences of site conditions and tectonic setting have been considered in few studies. Our
233 focus is on shallow crustal events in active tectonic regions; thus, we do not consider the
234 tectonic setting as a predictor variable. To consider the effect of shallow site conditions, we
235 examine the influence of V S30 (time-averaged shear-wave velocity in the top 30 m of the site)
236 on the DSF. A negligible dependence of the DSF on V S30 is observed based on inspection of
237 data and examining plots similar to Figures 3–5 (example plots are presented in Rezaeian
238 et al. 2012). The effect of sediment depth is also examined by scrutinizing the residual plots
239 of the model developed in the next section versus available data for Z 1.0 and Z 2.5 , which
240 respectively represent the depth to the 1.0 km∕s and 2.5 km∕s shear-wave velocity horizons.
241 Except at very long periods (e.g., 7.5 s) and very large depths (e.g., Z 1.0 greater than about
242 1.0 km and Z 2.5 greater than about 3 km), where there are minor trends at low and high
243 damping ratios, the residual plots do not show bias. Therefore, we do not include site con-
244 ditions as predictor variables in our model. This is generally consistent with the literature. For
245 example, Bommer and Mendis (2005) reported that soft soil influences the DSF, but to a
246 much lesser degree than magnitude and distance. Lin and Chang (2004) included site
247 class in their model, but they reported that this factor can be neglected when the DSF is
248 calculated for pseudo-spectral acceleration (as in Equation 1).

249 MODEL DEVELOPMENT

250 The predictor variables of the model are selected as described in the previous section.
251 After scrutinizing the data and performing statistical analyses, a lognormal distribution is
252 assigned to the random variable DSF. The functional form for the mean of lnðDSFÞ is
253 selected by examining the observed trends between the DSF and the predictor variables
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 947

254 in our database, as well as by reviewing various functions used in the literature. A summary
255 of the relevant functions used in the literature is given in Rezaeian et al. (2012). Regression
256 analysis is then performed to estimate the model coefficients and the variance using the
257 selected database of recorded ground motions with Rrup < 50 km.
258 To arrive at the final form of the model for median DSF, we follow a step-by-step model-
259 building process, details of which are given in Rezaeian et al. (2012). At each combination of
260 the 21 specified periods and the 11 selected damping ratios, we regress lnðDSFÞ on various
261 functions of the predictor variables M and Rrup . Each term is added to the model one at a time,
262 and the residuals (i.e., differences between the observed values of DSF and the values pre-
263 dicted by the model) versus M, Rrup , and D5–75 are examined. In our view, a linear magnitude
264 term is necessary and sufficient to capture the dependence of data on M and most of the
265 dependence on D5–75 . The addition of a logarithmic function of Rrup further reduces (although
266 not as much as the magnitude term) the dependence on D5–75 . It is convenient to directly
267 include β as a predictor variable in the model. To achieve this goal, dependences of the con-
268 stant term, the coefficient of the magnitude term, and the coefficient of the distance term on β
269 are examined. This dependence is captured best by a quadratic function of lnðβÞ. The result-
270 ing model is validated by examining the scatter plots of the residuals versus the predictor
271 variables β, M, and Rrup , and versus other variables, such as D5–75 , D5–95 , V S30 , Z 1.0 , and Z 2.5 .
272 The results show that the residuals are symmetrically scattered above and below the zero
273 level with no obvious systematic trends.

274 THE PROPOSED MODEL FOR MEDIAN DSF

275 In the model building process, various functions were considered, examined, and kept or
276 discarded. The final model has the following form

EQ-TARGET;temp:intralink-;e3;62;335 lnðDSFÞ ¼ b0 þ b1 lnðβÞ þ b2 ðlnðβÞÞ2 þ ½b3 þ b4 lnðβÞ þ b5 ðlnðβÞÞ2 M

(3)
þ ½b6 þ b7 lnðβÞ þ b8 ðlnðβÞÞ2  lnðRrup þ 1Þ þ ε

277 where β is the damping ratio in percentage (e.g., β ¼ 2 for 2% damping); Rrup is in km; bi ,
278 i ¼ 0; : : : 8, are the period-dependent regression coefficients listed in Tables 1 and 2; and ε is
279 a zero-mean normally distributed random variable with standard deviation σ. A model for σ is
280 presented in the next section.
281 The dependence on the damping ratio is captured best by a quadratic function of lnðβÞ
282 (see Equation 3), as is also seen in the models by Stafford et al. (2008) and Hatzigeorgiou
283 (2010). A linear function of lnðβÞ, which has been used in many other studies, works well
284 only at certain periods. There are minor differences between the model coefficients for the
285 RotD50 and GMRotI50 horizontal components. Therefore, the differences between the
286 resulting DSF models for these two components are very small (see Rezaeian et al. 2012
287 for comparison). Figure 6 shows the predicted DSF values according to Equation 3 for
288 the RotD50 component for M ¼ 5, 6, 7, 8 and Rrup ¼ 10 km. As an example, the damping
289 scaling factor is applied to the geometric mean of the NGA GMPEs at M ¼ 6.7,
290 Rrup ¼ 10 km, V S30 ¼ 255 m∕s and is plotted versus period in Figure 7. A simpler version
291 of the model in Equation 3 can be obtained by excluding the distance term. This version,
292 which is presented in Rezaeian et al. (2012), is simpler as it has fewer coefficients
Table 1. Regression coefficients for the horizontal component RotD50
948
T (s) b0 b1 b2 b3 b4 b5 b6 b7 b8 a0 a1
0.01 1.73E-03 2.07E-04 6.29E-04 1.08E-06 8.24E-05 7.36E-05 1.07E-03 9.08E-04 2.02E-04 3.70E-03 2.30E-04
0.02 5.53E-02 3.77E-02 2.15E-03 4.30E-03 3.21E-03 3.32E-04 4.75E-03 2.52E-03 2.29E-04 2.19E-02 2.11E-03
0.03 1.22E-01 7.02E-02 2.28E-03 3.21E-03 6.91E-05 9.82E-04 1.30E-02 7.82E-03 2.27E-04 5.21E-02 4.60E-03
0.05 2.39E-01 1.06E-01 2.63E-02 8.57E-04 7.43E-03 4.87E-03 1.69E-02 8.08E-03 1.71E-03 9.57E-02 1.31E-03
0.075 3.05E-01 7.32E-02 7.29E-02 2.02E-04 1.64E-02 1.03E-02 9.26E-04 6.40E-03 4.42E-03 1.21E-01 5.79E-03
0.1 2.69E-01 4.18E-03 1.07E-01 5.80E-03 2.49E-02 1.34E-02 2.35E-02 2.37E-02 5.84E-03 1.24E-01 1.08E-02
0.15 1.41E-01 1.00E-01 1.18E-01 3.01E-02 4.09E-02 1.41E-02 3.16E-02 2.47E-02 3.15E-03 1.15E-01 1.14E-02
0.2 5.01E-02 1.45E-01 1.11E-01 4.69E-02 4.77E-02 1.18E-02 3.10E-02 2.29E-02 2.41E-03 1.08E-01 8.85E-03
0.25 2.28E-02 1.43E-01 9.73E-02 5.20E-02 4.70E-02 9.47E-03 2.71E-02 2.02E-02 1.31E-03 1.04E-01 7.35E-03
0.3 1.58E-02 1.48E-01 8.83E-02 5.21E-02 4.36E-02 7.33E-03 3.87E-02 2.66E-02 1.76E-03 1.01E-01 6.90E-03
0.4 2.24E-02 1.03E-01 7.41E-02 4.63E-02 3.58E-02 4.65E-03 3.63E-02 2.45E-02 1.18E-03 1.02E-01 6.71E-03
0.5 3.19E-02 7.04E-02 5.57E-02 4.25E-02 2.94E-02 1.88E-03 3.87E-02 2.47E-02 3.13E-04 1.01E-01 6.22E-03
0.75 1.04E-02 5.33E-02 3.72E-02 4.47E-02 2.40E-02 2.40E-03 3.47E-02 2.59E-02 2.90E-03 1.01E-01 5.86E-03
1 8.84E-02 8.92E-02 2.14E-02 4.98E-02 2.36E-02 4.70E-03 5.02E-02 3.43E-02 2.32E-03 1.02E-01 7.31E-03
1.5 1.57E-01 9.33E-02 3.28E-03 5.85E-02 2.36E-02 8.02E-03 4.81E-02 3.30E-02 2.10E-03 1.02E-01 8.75E-03
2 2.96E-01 1.50E-01 2.09E-02 7.30E-02 2.96E-02 9.95E-03 5.24E-02 3.32E-02 6.86E-04 1.03E-01 9.22E-03
3 4.07E-01 1.97E-01 3.28E-02 8.35E-02 3.54E-02 1.01E-02 5.57E-02 2.91E-02 3.17E-03 9.63E-02 1.07E-02
4 4.49E-01 2.07E-01 4.42E-02 8.75E-02 3.59E-02 1.14E-02 5.07E-02 2.43E-02 4.67E-03 9.83E-02 1.37E-02
5 4.98E-01 2.17E-01 5.36E-02 9.03E-02 3.48E-02 1.29E-02 5.19E-02 2.30E-02 5.68E-03 9.42E-02 1.53E-02
7.5 5.25E-01 2.06E-01 7.79E-02 9.88E-02 3.76E-02 1.51E-02 2.91E-02 4.93E-03 9.02E-03 8.95E-02 1.63E-02
10 3.89E-01 1.43E-01 6.12E-02 7.14E-02 2.36E-02 1.30E-02 2.33E-02 5.46E-03 5.92E-03 6.89E-02 1.43E-02
REZAEIAN ET AL.
Table 2. Regression coefficients for the horizontal component GMRotI50

T (s) b0 b1 b2 b3 b4 b5 b6 b7 b8 a0 a1
0.01 3.01E-03 2.78E-03 1.52E-03 3.17E-04 2.43E-04 1.55E-04 9.93E-04 5.77E-04 1.98E-04 2.43E-02 1.37E-03
0.02 6.10E-02 3.83E-02 1.85E-03 4.92E-03 3.20E-03 2.58E-04 5.43E-03 2.98E-03 5.81E-05 3.21E-02 1.64E-04
0.03 1.22E-01 6.98E-02 2.11E-03 3.36E-03 3.48E-04 8.94E-04 1.33E-02 7.75E-03 1.56E-04 5.48E-02 3.34E-03
0.05 2.34E-01 9.75E-02 2.74E-02 4.11E-04 8.08E-03 5.03E-03 1.79E-02 8.12E-03 1.48E-03 9.45E-02 9.22E-04
0.075 3.00E-01 6.77E-02 7.37E-02 5.64E-04 1.61E-02 1.03E-02 3.11E-05 6.23E-03 4.00E-03 1.20E-01 6.18E-03
0.1 2.54E-01 1.55E-02 1.08E-01 7.09E-03 2.54E-02 1.34E-02 2.24E-02 2.32E-02 5.59E-03 1.22E-01 1.04E-02
0.15 1.49E-01 9.41E-02 1.17E-01 2.77E-02 3.88E-02 1.37E-02 2.97E-02 2.39E-02 3.29E-03 1.14E-01 1.08E-02
0.2 2.57E-02 1.54E-01 1.11E-01 4.76E-02 4.73E-02 1.18E-02 3.20E-02 2.38E-02 2.49E-03 1.07E-01 8.14E-03
0.25 7.91E-03 1.50E-01 9.77E-02 5.14E-02 4.60E-02 9.24E-03 2.98E-02 2.22E-02 1.79E-03 1.03E-01 6.91E-03
0.3 1.32E-02 1.39E-01 8.50E-02 5.05E-02 4.13E-02 6.75E-03 3.69E-02 2.55E-02 1.59E-03 1.01E-01 6.37E-03
0.4 4.02E-02 8.04E-02 6.86E-02 4.39E-02 3.31E-02 4.32E-03 3.14E-02 1.99E-02 1.26E-04 1.01E-01 6.38E-03
0.5 4.76E-02 6.49E-02 5.60E-02 3.83E-02 2.72E-02 1.87E-03 3.89E-02 2.50E-02 3.41E-04 1.01E-01 6.61E-03
0.75 1.93E-02 4.86E-02 3.90E-02 4.16E-02 2.20E-02 2.40E-03 3.46E-02 2.58E-02 3.26E-03 1.02E-01 6.23E-03
1 6.40E-02 8.34E-02 2.47E-02 4.64E-02 2.24E-02 4.30E-03 4.63E-02 3.28E-02 2.49E-03 1.03E-01 6.82E-03
1.5 1.52E-01 8.58E-02 5.17E-03 5.63E-02 2.21E-02 8.06E-03 4.76E-02 3.17E-02 1.65E-03 1.02E-01 8.91E-03
2 2.61E-01 1.38E-01 1.85E-02 6.94E-02 2.85E-02 9.59E-03 4.61E-02 2.97E-02 6.49E-04 1.04E-01 8.98E-03
3 3.65E-01 1.71E-01 3.48E-02 7.88E-02 3.34E-02 9.86E-03 4.86E-02 2.33E-02 4.26E-03 9.84E-02 1.05E-02
4 4.38E-01 1.97E-01 4.19E-02 8.53E-02 3.43E-02 1.11E-02 4.91E-02 2.31E-02 4.36E-03 9.85E-02 1.22E-02
5 4.97E-01 2.21E-01 5.20E-02 8.98E-02 3.59E-02 1.23E-02 4.93E-02 2.07E-02 6.11E-03 9.60E-02 1.45E-02
DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES

7.5 5.05E-01 1.89E-01 7.36E-02 9.27E-02 3.33E-02 1.48E-02 3.22E-02 7.26E-03 7.98E-03 9.21E-02 1.53E-02
10 3.98E-01 1.41E-01 5.92E-02 7.22E-02 2.26E-02 1.31E-02 2.29E-02 7.35E-03 4.33E-03 7.23E-02 1.19E-02
949
 950 REZAEIAN ET AL.

Figure 6. Predicted median DSF according to Equation 3 for the RotD50 component.

293 (i.e., b0 ; : : : b5 ), but it introduces an increase in the observed trends between the residuals and
294 the duration of motion. Therefore, we recommend the use of Equation 3.
295 To validate our model, we examine the residual diagnostic plots. Figure 8 shows exam-
296 ples of these plots at T ¼ 0.4 s. Only data with distances less than 50 km are plotted. In each
297 plot, a black line indicates the average values of residuals over equally spaced bins of data for
298 visual inspection of patterns. More examples are presented at select periods and damping
299 ratios and for additional variables (e.g., V S30 , Z 1.0 , Z 2.5 ) in Rezaeian et al. (2012). We recom-
300 mend the use of our model for moment magnitudes between 4.5 and 8.0. As previously men-
301 tioned, the regression is performed using data with Rrup < 50 km. The applicability of the
302 model for longer distances is investigated by studying the residual diagnostic plots for
303 records with Rrup ≥ 50 km. Figure 9 shows example plots for data with distances between
304 50 and 200 km (more examples are presented in Rezaeian et al. 2012). We conclude that our
305 proposed model can be used for distances of up to 200 km without any modification.
306 Figures 10 and 11 show the variation of DSF with magnitude and distance. This variation
307 is most sensitive to the damping ratio and spectral period. It can be negligible for damping
308 ratios close to 5%, or at periods shorter than about 0.1 s. But, it can be high at very low or very
309 high damping ratios, or at very long periods. Figure 10 shows the variation of the
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 951

Figure 7. The geometric mean of four NGA GMPEs (red) is scaled to adjust for various damping
ratios from 0.5% to 30%. The DSF model for the RotD50 component is used. Assumptions
in GMPE calculations: reverse fault, dip ¼ 45°, hanging wall, fault rupture width ¼ 15 km,
Rjb ¼ 0 km, Rx ¼ 7 km.

310 DSF with M at a distance of 10 km for eight different damping ratios. The variation with
311 magnitude is high for 0.5% and 30% damping levels; it is almost insignificant for 3% and 7%
312 damping levels. It is interesting to note that there is a “critical period,” where the direction of
313 the dependence between the DSF and M changes. For example, at 30% damping ratio, this
314 “critical period” is around 0.25 s; the DSF increases with M for periods less than 0.2 s, but
315 decreases with M for periods greater than 0.3 s. This critical period changes with the damping
316 ratio. Figure 11 shows the variation of the DSF with Rrup at a magnitude of 7.0. Again, the
317 variation is more extreme at very low and very high damping levels, and at long periods.
318 Observe that, similar to Figure 10, there is a damping-dependent “critical period,” where the
319 direction of the dependence between the DSF and Rrup changes.

320 THE PROPOSED MODEL FOR STANDARD DEVIATION

321 The standard deviation σ in Equation 3 is calculated for all combinations of T and β to
322 estimate the logarithmic standard deviation of the DSF. The data suggest dependence of the
323 variance on the damping ratio. As expected, the standard deviation is zero at 5% damping
324 (i.e., DSF ¼ 1 for β ¼ 5%), and it increases as the damping ratio deviates from 5%, reaching
325 a maximum of about 0.2 (Figure 12a shows an example at T ¼ 1 s). This dependence, at a
326 specified period, can be captured by the following equation:
 952 REZAEIAN ET AL.

Figure 8. Sample residual plots at T ¼ 0.4 s using data with Rrup < 50 km.

σ lnðDSFÞ ¼ ja0 lnðβ∕5Þ þ a1 ðlnðβ∕5ÞÞ2 j

EQ-TARGET;temp:intralink-;e4;41;316 (4)

326
327 where σ lnðDSFÞ represents the logarithmic standard deviation of the DSF. The model coeffi-
328 cients, a0 and a1 , are obtained by fitting (using least squares regression) Equation 4 to the
329 data from 11 damping ratios. Their values are given in Tables 1 and 2, respectively, for
330 the RotD50 and GMRotI50 components. The predicted standard deviation according to
331 Equation 4 is plotted in Figure 12b for the RotD50 component.
332 Few studies in the literature reported estimates of the standard deviation or coefficient of
333 variation (i.e., Atkinson and Pierre 2004, Lin and Chang 2004, Cameron and Green 2007,
334 and Stafford et al. 2008). The reported values ranged between 0 to about 0.2, depending on
335 the damping ratio and spectral period. Other parameters did not show significant effects on
336 the variance. This is consistent with our findings.
337 The standard deviation of the scaled response spectrum can be calculated using the defi-
338 nition of DSF. Taking the natural logarithm of both sides in Equation 1, rearranging the
339 equation, then taking the variance and the square root results in

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
σ lnðPSAβ% Þ ¼
EQ-TARGET;temp:intralink-;e5;41;123 σ 2lnðPSA5% Þ þ σ 2lnðDSFÞ þ 2σ lnðPSA5% Þ σ lnðDSFÞ ρ (5)
DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 953

Figure 9. Sample residual plots at T ¼ 0.4 s using data with 50 ≤ Rrup < 200 km.

Figure 10. Variation of the DSF with magnitude at 10 km distance.

 954 REZAEIAN ET AL.

Figure 11. Variation of the DSF with distance at magnitude of 7.0.

Figure 12. (a) Dependence of standard deviation on β and the fitted function according to
Equation 4. (b) Predicted logarithmic standard deviation according to Equation 4.

339
340 where ρ represents the correlation coefficient between lnðDSFÞ and lnðPSA5% Þ. Based on the
341 relatively small values of σ lnðDSFÞ and the estimated values of ρ found by Rezaeian et al.
342 (2012), we expect σ lnðPSA5% Þ to dominate the overall standard deviation of the scaled spec-
343 trum. More information on ρ for horizontal motion is presented in Rezaeian et al. (2012).
344 Based on sample correlation coefficients, ρ is very small at lower periods, is negative for
345 β > 5%, and reaches a maximum of about 0.5 at very long T and β < 5%. Note that we
346 are not making any recommendations on whether to use ρ ¼ 0, but merely stating that
347 the value of σ lnðPSAβ% Þ is driven by σ lnðPSA5% Þ .
DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 955

Figure 13. Data binned for M and Rrup ð6 ≤ M ≤ 7 and 0 ≤ Rrup < 50 kmÞ is superimposed on
the plots of the proposed model for M ¼ 6.5 at Rrup ¼ 20 km. Recall that 11 lines correspond to
11 damping ratios.

Figure 14. The proposed model is plotted for all 11 damping ratios from 0.5 to 30%. Idriss
(1993) is plotted for β ¼ 1%, 2%, 3%, 5%, 7%, 10%, 15%. It is applicable to T ¼ 0.03  5 s,
and is not a function of M or Rrup .
 956 REZAEIAN ET AL.

348 COMPARISON WITH DATA AND WITH EXISTING MODELS

349 In this section, we compare our final model with the computed DSF values from the
350 database of recorded ground motions and with selected existing models in the literature.
351 Since our proposed model is empirical and was designed to capture the observed trends
352 in the database, we expect close agreement between the model and the data. Plots similar
353 to Figure 13 are generated to validate the model. In Figure 13, the predicted median DSF is
354 plotted for a moment magnitude of 6.5 and Rrup ¼ 20 km. The median DSF calculated from
355 records in a magnitude–distance bin of 6 ≤ M ≤ 7 and 0 ≤ Rrup < 50 km is superimposed.
356 Overall, there is good agreement between the model and data. The most pronounced differ-
357 ence is seen around shorter periods. The discrepancies are likely due to the wide ranges of
358 magnitude and distance bins used for the observed data. Depending on the exact values of M
359 and Rrup , narrowing the magnitude and distance bins may give a better match to the corre-
360 sponding prediction if enough data points are available in the selected bins, or due to the
361 reduction in the sample size it may have the opposite effect.
362 We also compare our model to selected existing models in the literature. Recall that dif-
363 ferent models use different databases and are applicable to different ranges of β, T, M, and
364 Rrup . In Figures 14–18, the median DSF for our proposed model is plotted versus period at the
365 11 damping ratios for M ¼ 6.5 and 7.5 and Rrup ¼ 5 km and 10 km. In Figure 14, the model

Figure 15. The proposed model is plotted for all 11 damping ratios from 0.5 to 30%.
Abrahamson and Silva (1996) is plotted for β ¼ 0.5%, 1%, 2%, 3%, 7%, 10%, 15%, 20% at select
periods and interpolated in between. It is applicable to T ¼ 0.02  5 s, and is a function of M, but
not Rrup .
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 957

366 developed by Idriss (1993) is superimposed on each plot. This model is only given and
367 plotted for β ¼ 1%, 2%, 3%, 5%, 7%, 10%, 15%. It is applicable to T ¼ 0.03  5 s, and
368 is not a function of M or Rrup . It best agrees with our proposed model at higher magnitudes
369 and periods greater than 0.1s.
370 In Figure 15, the model developed by Abrahamson and Silva (1996) is superimposed on
371 each plot for β ¼ 0.5%, 1%, 2%, 3%, 5%, 7%, 10%, 15%, 20%. This model is applicable to
372 T ¼ 0.02  5 s. It is calculated and plotted at select periods and is linearly interpolated in
373 between. It is a function of M, but not Rrup . Except for very low damping, where the peak is at
374 a longer period than the peak DSF in our model, there is a good agreement between this
375 model and our proposed model.
376 In Figure 16, our model is compared to the model developed by Newmark and Hall
377 (1982), which is the basis for most U.S. building codes. The model of Newmark and
378 Hall (1982) is only applicable for β ≤ 20% and is plotted for β ¼ 0.5%, 1%, 2%, 3%,
379 5%, 7%, 10%, 15%, 20%. Furthermore, it is applicable for T ¼ 0.125  10 s and is not a
380 function of M or Rrup . Considering the limited number of records they used, their model
381 is in good agreement with our model, particularly for periods around 0.125 s to 1 s.

Figure 16. The proposed model is plotted for all 11 damping ratios from 0.5 to 30%. Newmark
and Hall (1982) is applicable for β ≤ 20% and T ¼ 0.125  10 s. It is plotted for β ¼ 0.5%, 1%,
2%, 3%, 5%, 7%, 10%, 15%, 20%, and is a not a function of M or Rrup .
 958 REZAEIAN ET AL.

382 In Figure 17, the model of Eurocode 8 (CEN 2004) is superimposed on the plots of our
383 proposed model for the same 11 damping ratios. This model is applicable to periods ranging
384 roughly from 0.2 s to 6 s with unity imposed at very low and very high periods and should not
385 be applied to β values resulting in a DSF smaller than 0.55. It is not a function of M or Rrup .
386 Observe that for β < 3% this model tends to be relatively low. This might be expected
387 because this model was based on the work of Bommer et al. (2000), focusing on high damp-
388 ing ratios. The model compares best with our prediction of the DSF for T ¼ 1 s and high
389 damping ratios.
390 As previously mentioned, many recent studies have not developed a functional form for
391 the DSF and instead have presented their results in the form of tables for various magnitude–
392 distance bins. This makes a direct comparison difficult. But as an example, in Figure 18, our
393 model is compared to that of Cameron and Green (2007). Cameron and Green (2007) tabu-
394 lated their results at select periods between 0.05 s to 10 s for β ¼ 2%, 7%, 10%, 15%, 20%,
395 30%, 40%, 50% for three magnitude bins and for rock and soil conditions. For β ¼ 1%, in
396 addition to magnitude, the DSF values also depended on two distance bins. In Figure 18,
397 magnitude bins of 6–7 and 7þ are selected for Cameron and Green (2007) when our model is
398 plotted for 6.5 and 7.5. For 1% damping ratio, the distance bin of 0–50 km is selected

Figure 17. The proposed model and the model by Eurocode 8 (CEN 2004) are plotted for all
11 damping ratios from 0.5% to 30%. The model by Eurocode 8 (CEN 2004) is not a function
of M or Rrup . In this figure, we assume very low and very high periods, where the model by
Eurocode 8 is equal to unity, to be 0.01 s and 10 s (10 s is the value specified in Eurocode 8).
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 959

Figure 18. The proposed model is plotted for all 11 damping ratios from 0.5% to 30%. Cameron
and Green (2007) is plotted for β ¼ 1%, 2%, 7%, 10%, 15%, 20%, 30% for western U.S. rock site
conditions at select periods and interpolated in between. Values from magnitude bin of 6–7 are
plotted for M ¼ 6.5 and those from magnitude bin of 7þ are plotted for M ¼ 7.5. For β ≥ 2%;
their model is independent of distance, but for β ¼ 1% the distance bin of 0–50 km is plotted.

399 (our model is plotted for 5 km and 10 km). In all figures, the rock site conditions are selected
400 for Cameron and Green (2007); as they demonstrated, there is little difference between rock
401 and soil conditions for the western U.S. region (i.e., shallow crustal earthquakes in active
402 tectonic regions). With the exception of low damping ratios at long periods and magnitude of
403 7.5, their model is in good agreement with our model. The discrepancies could be due to the
404 wide range of magnitude and distance bins. As one can see in Figures 10 and 11, there is a
405 good amount of variation within these bins, particularly at low damping levels.
406 Additional comparisons to existing models (e.g., Stafford et al. 2008) are presented in
407 Rezaeian et al. (2012).

408 CONCLUSIONS
409 We developed a new predictive model for a damping scaling factor (DSF), which can be
410 used to scale elastic pseudo-spectral acceleration (PSA) values predicted at a 5% damping
411 ratio to PSA values at damping ratios other than 5%. We conducted a comprehensive litera-
412 ture review and employed an extensive database of recorded ground motions from shallow
413 crustal earthquakes in active tectonic regions (i.e., the NGA-West2 project database). The
414 general trends of the DSF with potential predictor variables (i.e., damping ratio, spectral
 960 REZAEIAN ET AL.

415 period, duration, earthquake magnitude, distance, and site conditions) were examined. In
416 addition to the damping ratio and the spectral period, the predictor variables in our proposed
417 model are magnitude and distance. Duration is a variable that strongly influences the DSF,
418 however, inclusion of magnitude and distance in the model captures most of the dependence
419 on duration. We also found that the standard deviation of lnðDSFÞ varies systematically with
420 the damping ratio. The final models for the median of DSF and its logarithmic standard
421 deviation are presented in Equations 3 and 4. The model parameters are calculated for
422 21 spectral periods ranging between 0.01 s to 10 s. These parameters are given in Tables 1
423 and 2, respectively, for the RotD50 and GMrotI50 horizontal components of ground motion.
424 Minor differences exist between the models’ coefficients for these two components. Recent
425 studies are available that present factors to convert from RotD50 and GMrotI50 to other
426 definitions of the horizontal component of motion (see Boore 2010, or Bozorgnia et al.
427 2012). We compared our proposed model with (1) computed DSF values from the database
428 of recorded ground motions (we saw close agreement between the model and data), and
429 (2) selected existing models in the literature (i.e., Idriss 1993, Abrahamson and Silva
430 1996, Newmark and Hall 1982, CEN 2004, and Cameron and Green 2007). Based on
431 the comprehensiveness of the database used, detailed analyses of residuals, and its wide
432 range of applicability, we recommend using the damping scaling model developed under
433 the current study. Our proposed model is applicable to shallow crustal earthquakes in active
434 tectonic regions for damping ratios from 0.5% to 30%, moment magnitudes between 4.5 and
435 8.0, and distances of less than 200 km. This model is developed based on observed spectral
436 ordinates; therefore, it is independent of any specific GMPE for PSA.
437 Near-fault effects, such as directivity, can cause velocity pulses and may change the
438 effectiveness of damping. Such effects are incorporated into our model implicitly by the
439 inclusion of near-fault records in the ground motion database. However, separate examina-
440 tion of near-fault effects is a subject of future research. Extension of the model to the vertical
441 component of ground motion is the subject of a follow-up paper.

442 ACKNOWLEDGMENTS
443 This study was sponsored by the Pacific Earthquake Engineering Research Center
444 (PEER) and funded by the California Earthquake Authority, California Department of
445 Transportation, the Pacific Gas & Electric Company, and the U.S. Geological Survey.
446 Any opinions, findings, and conclusions or recommendations expressed in this material
447 are those of the authors and do not necessarily reflect those of the above mentioned agencies.
448 We would like to thank all the NGA model developers and supporting researchers for their
449 assistance and valuable feedback throughout this project. Efforts and cooperation of Dr. Tim
450 Ancheta in development of the database are gratefully acknowledged. The valuable feedback
451 from Dr. Nicolas Luco, Dr. Julian Bommer, and anonymous reviewers of this paper are
452 appreciated.

453 REFERENCES
454 Abrahamson, N. A., and Silva, W. J., 1996. Section 4: Spectral scaling for other damping values,
455 in Empirical Ground Motion Models, Report, Brookhaven National Laboratory, Upton, NY.
456 Akkar, S., and Bommer, J. J., 2006. Influence of long-period filter cut-off on elastic spectral
457 displacements, Earthquake Engineering and Structural Dynamics 35, 1145–1165.
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 961

458 Akkar, S., and Bommer, J., 2007. Prediction of elastic displacement response spectra in Europe
459 and the Middle East, Earthquake Engineering and Structural Dynamics 36, 1275–1301.
460 Ancheta, T., Darragh, R., Silva, W., Abrahamson, N., Atkinson, G., Boore, D., Bozorgnia, Y.,
461 Campbell, K., Chiou, B., Graves, R., Idriss, I. M., Stewart, J., Shantz, T., and Youngs, R.,
462 2012a. PEER NGA-West2 Database: A database of ground motions recorded in shallow
463 crustal earthquakes in active tectonic regimes, in Proceedings, 15th World Conference on
464 Earthquake Engineering, Lisbon, Portugal, Paper Number 5599.
465 Ancheta, T., Darragh, R., Stewart, J., Seyhan, E., Silva, W., Chiou, B., Woodell, K., Graves, R.,
466 Kottke, A., and Boore, D., 2012b. PEER NGA-West2 Database, PEER Report, Pacific
467 Earthquake Engineering Research Center, Berkeley, CA.
468 Ashour, S. A., 1987. Elastic Seismic Response of Buildings with Supplemental Damping, Ph.D.
469 Thesis, Department of Civil Engineering, University of Michigan.
470 Atkinson, G. M., and Pierre, J. R., 2004. Ground motion response spectra in eastern North
471 America for different critical damping values, Seismological Research Letters 75, 541–545.
472 Berge-Thierry, C., Cotton, F., Scotti, O., Griot-Pommera, D. A., and Fukushima, Y., 2003.
473 New empirical response spectral attenuation laws for moderate European earthquakes, Journal
474 of Earthquake Engineering 7, 193–222.
475 Bommer, J., Elnashai, A. S., Chlimintzas, G. O., and Lee, D., 1998. Review and Development of
476 Response Spectra for Displacement-Based Design, ESEE Research Report No. 98–3, Imperial
477 College, London.
478 Bommer, J., Elnashai, A. S., and Weir, A. G., 2000. Compatible acceleration and displacement
479 spectra for seismic design codes, in Proceedings, 12th World Conf. on Earthquake Engineer-
480 ing, Auckland, New Zealand, New Zealand Society for Earthquake Engineering, Inc.,
481 Wellington, New Zealand, Paper No. 207.
482 Bommer, J., and Mendis, R., 2005. Scaling of spectral displacement ordinates with damping
483 ratios, Earthquake Engineering and Structural Dynamics 34, 145–165.
484 Bommer, J. J., Stafford, P. J., and Alarcon, J. E., 2009. Empirical equations for the prediction of
485 the significant, bracketed and uniform duration of earthquake ground motion, Bulletin of the
486 Seismological Society of America 99, 3217–3233.
487 Boore, D. M., 2010. Orientation-independent, non geometric-mean measures of seismic intensity
488 from two horizontal components of motion, Bulletin of the Seismological Society of America
489 100, 1830–1835.
490 Boore, D. M., Joyner, W. B., and Fumal, T. E., 1993. Estimation of Response Spectra and Peak
491 Accelerations from Western North American Earthquakes: An Interim Report, U.S. Geologi-
492 cal Survey Open-File Report 93–509.
493 Boore, D. M., Watson-Lamprey, J., and Abrahamson, N. A., 2006. Orientation-independent mea-
494 sures of ground motion, Bulletin of the Seismological Society of America 96, 1502–1511.
495 Bozorgnia, Y., and Campbell, K. W., 2004. Engineering characterization of ground motion, in
496 Earthquake Engineering: From Engineering Seismology to Performance-Based Engineering
497 (Bozorgnia and Bertero, Eds.), CRC Press, Boca Raton, FL.
498 Bozorgnia, Y., Abrahamson, N. A., Campbell, K. W., Rowshandel, B., and Shantz, T., 2012.
499 NGA-West2: A comprehensive research program to update ground motion prediction equa-
500 tions for shallow crustal earthquakes in active tectonic regions, Proceedings, 15th World Con-
501 ference on Earthquake Engineering, Lisbon, Portugal. Paper Number 2572.
502 Cameron, W. I., and Green, I., 2007. Damping correction factors for horizontal ground motion
503 response spectra, Bulletin of the Seismological Society of America 97, 934–960.
 962 REZAEIAN ET AL.

504 Comite Europeen de Normalisation (CEN), 2004. Eurocode 8: Design of Structures for
505 Earthquake Resistance—Part 1: General Rules, Seismic Actions and Rules for Buildings,
506 EN 1998-1: 2004, Brussels.
507 Chopra, A. K., 2012. Dynamics of Structures: Theory and Applications to Earthquake Engineer-
508 ing, 4th Edition, Prentice Hall: Upper Saddle River, NJ.
509 Faccioli, E., Paolucci, R., and Rey, J., 2004. Displacement spectra for long periods, Earthquake
510 Spectra 20, 347–376.
511 Federal Emergency Management Agency (FEMA 440) 2005. Improvement of Nonlinear Static
512 Seismic Analysis Procedures, Washington, D.C.
513 Hatzigeorgiou, G. D., 2010. Damping modification factors for SDOF systems subjected to near-
514 fault, far-fault and artificial earthquakes, Earthquake Engineering and Structural Dynamics
515 39, 1239–1258.
516 Idriss, I. M., 1993. Procedures for Selecting Earthquake Ground Motions at Rock Sites, Nat. Inst.
517 of Stan. and Tech., Rept. NIST GCR 93–625.
518 Iwan, W. D., and Gates, N. C., 1979. Estimating earthquake response of simple hysteretic struc-
519 tures, Journal of the Engineering Mechanics Division, ASCE 94, 391–405.
520 Kempton, J. J., and Stewart, J. P., 2006. Prediction equations for significant duration of
521 earthquake ground motions considering site and near-source effects, Earthquake Spectra
522 22, 985–1013.
523 Lin, Y. Y., and Chang, K. C., 2003. Study on damping reduction factor for buildings under earth-
524 quake ground motions, Journal of Structural Engineering, ASCE 129, 206–214.
525 Lin, Y. Y., and Chang, K. C., 2004. Effects of site classes on damping reduction factors, Journal
526 of Structural Engineering, ASCE 130, 1667–1675.
527 Lin, Y. Y., Miranda, E., and Chang, K. C., 2005. Evaluation of damping reduction factors for
528 estimating elastic response of structures with high damping, Earthquake Engineering and
529 Structural Dynamics 34, 1427–1443.
530 McGuire, R. K., Silva, W. J., and Costantino, C. J., 2001. Section 4.9: Estimation of spectra for
531 other dampings, in Technical Basis for Revision of Regulatory Guidance on Design Ground
532 Motions: Hazard- and Risk-Consistent Ground Motion Spectra Guidelines, Report, NUREG/
533 CR-6728. Prepared for Division of Engineering Technology, Office of Nuclear Regulatory
534 Research, U.S. Nuclear Regulatory Commission, Washington, D.C.
535 Mendis, R., and Bommer, J. J., 2006. Modification of the Eurocode 8 damping reduction factors
536 for displacement spectra, in Proceedings, First European Conf. on Earthquake Engineering
537 and Seismology, Geneva, Switzerland, Paper No. 1203.
538 Naeim, F., and Kircher, C. A., 2001. On the damping adjustment factors for earthquake response
539 spectra, The Structural Design of Tall Buildings 10, 361–369.
540 Newmark, N. M., and Hall, W. J., 1982. Earthquake Spectra and Design, Monograph,
541 Earthquake Engineering Research Institute, Berkeley, CA.
542 Pacific Earthquake Engineering Research Center (PEER), 2010. Guidelines for Performance-
543 Based Seismic Design of Tall Buildings, PEER Report 2010/05, November 2010. Prepared
544 by the TBI Guidelines Working Group, University of California, Berkeley.
545 Pacific Earthquake Engineering Research Center/Applied Technology Council (PEER/ATC72-1),
546 2010. Modeling and Acceptance Criteria for Seismic Design and Analysis of Tall Buildings,
547 Applied Technology Council, Redwood City, CA.
548 Power, M., Chiou, B., Abrahamson, N. A., Roblee, C., Bozorgnia, Y., and Shantz, T., 2008.
549 An introduction to NGA, Earthquake Spectra 24, 3–21.
 DSF FOR ELASTIC RESPONSE SPECTRA FOR SHALLOW CRUSTAL EARTHQUAKES 963

550 Priestley, M. J. N., 2003. Myths and Fallacies in Earthquake Engineering, Revisited, The Mallet
551 Milne Lecture, IUSS Press, Istituto Universitario di Studi Superiori di Pavia, Pavia, Italy.
552 Regulatory Guide 1.61 Revision 1, 2007. Damping Values for Seismic Design of Nuclear Power
553 Plants, U.S. Nuclear Regulatory Commission, Washington, D.C.
554 Rezaeian, S., Bozorgnia, Y., Idriss, I. M., Campbell, K., Abrahamson, N., and Silva, W.,
555 2012. Spectral Damping Scaling Factors for Shallow Crustal Earthquakes in Active Tectonic
556 Regions. PEER Report 2012/102, Pacific Earthquake Engineering Research Center,
557 Berkeley, CA.
558 Rosenblueth, E., 1980. Characteristics of earthquakes, Design of Earthquake Resistant Structures
559 (E. Rosenblueth, Ed.), Wiley & Sons, New York.
560 Stafford, P. J., Mendis, R., and Bommer, J. J., 2008. Dependence of damping correction factors
561 for response spectra on duration and numbers of cycles, Journal of Structural Engineering,
562 ASCE 134, 1364–1373.
563 Tolis, S. V., and Faccioli, E., 1999. Displacement design spectra, Journal of Earthquake
564 Engineering 3, 107–125.
565 Trifunac, M. D., and Lee, V. W. 1989. Empirical models for scaling pseudo relative velocity
566 spectra of strong earthquake accelerations in terms of magnitude, distance, site intensity
567 and recording site conditions, Soil Dynamics and Earthquake Engineering 8, 126–144.
568 Vanmarcke, E. H., 1976. Structural response to earthquakes, in Seismic Risk and Engineering
569 Decisions (C. Lomnitz and E. Rosenblueth, Eds.), Elsevier Publishing Co., New York.
570 Wu, J., and Hanson, R. D., 1989. Study of inelastic spectra with high damping, Journal of Struc-
571 tural Engineering, ASCE 115, 1412–1431.
572 (Received 5 October 2012; accepted 9 March 2013)