Estimation of Strong Response Spectra from Weak Motion Data for Earthquakes in Taiwan Yi-Ben Tsai and Hsiao-Ling Chao Institute of Geophysics, National Central University Chung-Li, Taiwan ABSTRACT We have developed and tested an empirical approach for rapid estimation of strong ground motion response spectra that are useful for making timely damage assessment following a strong earthquake. This approach takes advantage of the unique features of two operational strong ground motion networks in Taiwan: the Taiwan Rapid Earthquake Information Release System (TREIRS), currently consisting of about 80 real-time digital accelerographic stations and the dense Taiwan Strong Motion Instrumentation Program (TSMIP) network, currently consisting of about 650 free-field digital accelerographic stations. The TREIRS can quickly provide complete ground acceleration time histories at its stations after a strong earthquake, whereas the TSMIP network can provide dense spatial coverage of populated areas at about 5-km grid spacings. By this approach we can make rapid estimation of dense spatial patterns of ground motion response spectra within minutes after a strong earthquake. In this approach, we first calculate the response spectra from recorded acceleration time histories at the TREIRS stations immediately after a strong earthquake. We then use pre-established inter-station spectral ratio functions between a TSMIP station and a nearby TREIRS station to obtain the corresponding response spectra for individual TSMIP stations. This empirical approach is based on the fact that for a given station the response spectral shape, after removal of source and propagation path effects, is highly repeatable from earthquake to earthquake. We use an average response spectral shape of regional stations to represent the source and propagation path effects. This regionally averaged spectral shape is next used to normalize the response spectral shape at individual stations. The resultant spectral shape represents primarily the local site response effects. We can obtain a smooth local site response spectral shape for each station by averaging over many earthquakes. We have obtained an average local site response spectral shape for each of the 650 free-field stations by using recordings of harmless ground motion from more frequent, moderate earthquakes. Once a strong earthquake takes place, we can use the actual recording at a TREIRS station to calculate the response spectrum. We then use this spectrum as a regional reference to estimate the response spectra at nearby free-field stations. This empirical approach was tested satisfactorily on the recordings of the 1999 Chi-Chi, Taiwan earthquake sequence. 1 Introduction Estimation of strong ground motions from large earthquakes is essential for seismic hazard assessement and loss estimation. Here we develop an empirical approach to estimate strong ground motion response spectra from a large earthquake by using much more frequent weak motion data. The approach uses the response spectral shape normalized by the MSA (Mean Spectral Acceleration), instead of by PGA (Peak Ground Acceleration). Figure 1 shows the average response spectral shapes and their dispersions of four California earthquakes. On top we show the results from normalizing by PGA, whereas at the bottom we show the results from normalizing by MSA. It is noted that the dispersion is not uniform over the frequency band of interest when normalized by PGA. In the following analysis, we use the MSA-normalized spectral shape as our working ground motion parameter. Strong Ground Motion Data from the Chi-Chi Earthquake Sequence When the M7.6 Chi-Chi, Taiwan earthquake of September 21, 1999 struck, Taiwan was instrumented by a very dense strong ground-motion network to record the main shock and many aftershocks. Figure 2 shows the locations of the free-field strong motion stations as well as the epicenters of the nine earthquakes in the Chi-Chi sequence. Figure 3 shows the distribution of the four types of recording site classification: Site B for rock sites, Site C and D for terrace and gravel sites, and Site E for soft soil sites. Local Site Response of Ground Motion Figure 4 shows the Fourier amplitude spectra normalized at high-frequency band for the horizontal and vertical components. It is evident that the local site effects are most pronounced over the frequency band from 0.1 to 5 Hz on the horizontal component. This is also shown in Figure 5 where the average response spectral shapes for the four site classes from the Chi-Chi earthquakes are distinctively different. In addition to local site effects, the response spectral shapes also depend on the source effects, as shown in Figure 6 which shows the average response spectral shapes for the nine earthquakes in the Chi-Chi sequence. Estimation of Strong Motion Response Spectra from Weak Motion Data We next obtain the average response spectral shape at each station from the eight smaller earthquakes in the sequence. We then use one or two stations at each of the seven regions as a reference station. Spectra of the reference station are multiplied by the spectral ratio between the target and reference stations which are obtained from the weak motion records. Figures 7 and 8 show the comparison between the actual 2 and estimated response spectra for the Chi-Chi main shock at some of the stations in the TCU and CHY regions. From these figures we can see that actual ground motion at majority of stations was satisfactorily estimated in terms of both spectral amplitude and shape. We also noticed that the ground motions at stations west of the Chelungpu fault were consistently overestimated. This is because we use a reference station on the hanging wall of the Chelungpu fault where the ground motion was significantly greater than the footwall stations west of the fault. Finally, Figure 9 shows the average and dispersion of the ratio between the observed and estimated response spectra in the seven regions of Taiwan. It can be seen that five of the seven regions are very accurately estimated. On the other hand, the ground motions were underestimated in TAP and TCU regions. In addition, the dispersion in TCU region is significantly greater than the rest. In conclusion, ground motion response spectra from the M7.6 Chi-Chi earthquake can be estimated satisfactorily from the response spectra of smaller earthquakes. With a combination of the dense free-field network and the rapid earthquake reporting system operated by the CWB, it is possible to obtain spatial patterns of ground motion response spectra only in a few minutes following a strong earthquake. 3 Figure 1. Mean Response spectral shapes and their dispersions for four California Earthquakes. Top: Normalized by PGA; Bottom: Normalized by MSA 4 00 30 30 60 60 Kilometers Kilometers 3 3 11 44 77 66 22 55 8 8 99 TAP TCU ILA HWA CHY KAU TTN 11 ChiChi ChiChi Earthquake Earthquake 2~7 2~7 ChiChi ChiChi Aftershock Aftershock 88 Chiayi Chiayi Earthquake Earthquake 99 Rayli Rayli Earthquake Earthquake 震央 震央 車籠埔斷層 車籠埔斷層 觸發之測站 觸發之測站 未觸發之測站 未觸發之測站 Figure 2. Free-field strong motion stations and the nine earthquakes in the Chi-Chi sequence. 5 Figure 3. Site classification of the free-field strong motion stations in Taiwan. 6 Horizontal Mean Normalized Fourier Spectrum 100 10 1 EQ 1 All Sites (396) class-B (52) class-C (58) class-D (187) class-E (99) 0.1 0.01 0.1 1 10 100 10 100 Frequency(Hz) Vertical Mean Normalized Fourier Spectrum 100 10 1 EQ 1 All Sites (396) class-B (52) class-C (58) class-D (187) class-E (99) 0.1 0.01 0.1 1 Frequency(Hz) Figure 4. Average Fourier amplitude spectral shapes for the four types of site from the Chi-Chi main shock. Top: Horizontal component; Bottom: Vertical component. 7 Spectral Acceleration (g) 10 V 1 0.1 0.1 1 10 100 Frequency (Hz) Spectral Acceleration (g) 10 EW 1 0.1 0.1 1 10 100 Frequency (Hz) Spectral Acceleration (g) 10 NS 1 EQ 1 All sites (396) class-B (52) class-C (58) class-D (187) class-E (99) 0.1 0.1 1 10 100 Frequency (Hz) Figure 5. Average response spectral shapes for the four types of site from the Chi-Chi main shock. Top: Vertical component; Middle: EW component; Bottom: NS component. 8 Spectral Acceleration (g) 10 V 1 0.1 0.1 1 10 100 Frequency (Hz) Spectral Acceleration (g) 10 EW 1 0.1 0.1 1 10 100 Frequency (Hz) Spectral Acceleration (g) 10 EQ1,Mw=7.6 EQ2,Mw=6.3 EQ3,Mw=6.4 EQ4,Mw=6.4 EQ5,Mw=6.4 EQ6,Mw=6.4 EQ7,Mw=6.5 EQ8,Mw=5.8 EQ9,Mw=5.7 NS 1 0.1 0.1 1 10 100 Frequency (Hz) Figure 6. Average response spectral shapes for the nine earthquakes in the Chi-Chi sequence. Top: Vertical component; Middle: EW component; Bottom: NS component. 9 0921 Chichi TCU Horizontal-component TCU039 TCU017 TCU006 1 0.1 0.01 0.1 1 10 Frequency (Hz) 1 Sa (g) Sa (g) Sa (g) 1 0.1 0.01 0.1 100 1 10 Frequency (Hz) 0.1 0.01 0.1 100 1 10 Frequency (Hz) TCU059 TCU072 1 Sa (g) Sa (g) 1 0.1 0.01 0.1 1 10 Frequency (Hz) 0.1 0.01 0.1 100 TCU070 100 1 Sa (g) Sa (g) 1 10 Frequency (Hz) TCU074 1 0.1 0.01 0.1 1 10 Frequency (Hz) 0.1 0.01 0.1 100 TCU048 1 10 Frequency (Hz) 100 TCU089 1 Sa (g) 1 0.1 1 10 Frequency (Hz) TCU122 TCU110 0.1 1 10 Frequency (Hz) 100 TCU082 0.1 0.01 0.1 1 10 Frequency (Hz) 1 1 Sa (g) 1 0.01 0.1 0.1 0.01 0.1 100 Sa (g) 0.01 0.1 Sa (g) Sa (g) 100 1 10 Frequency (Hz) 100 TREIRS stations 0.1 0.01 0.1 1 10 Frequency (Hz) 100 Calculated Observed Figure 7. Comparison of the observed with estimated response spectral in the TCU area for the Chi-Chi main shock. 10 100 0921 Chichi CHY Horizontal-component CHY004 CHY002 CHY080 1 0.1 0.01 0.1 1 10 Frequency (Hz) 100 Sa (g) 1 Sa (g) Sa (g) 1 0.1 0.01 0.1 1 10 Frequency (Hz) 0.1 0.1 100 1 10 Frequency (Hz) CHY012 CHY046 1 Sa (g) Sa (g) 1 0.1 0.01 0.1 1 10 Frequency (Hz) 0.1 0.01 0.1 100 CHY058 100 1 Sa (g) Sa (g) 1 10 Frequency (Hz) CHY014 1 0.1 0.01 0.1 1 10 Frequency (Hz) 0.1 0.01 0.1 100 CHY067 100 Sa (g) 1 0.1 1 10 Frequency (Hz) 0.01 0.1 CHY070 Sa (g) Sa (g) 1 1 0.1 1 10 Frequency (Hz) 100 1 10 Frequency (Hz) CHY019 CHY022 1 0.01 0.1 0.1 100 Sa (g) 0.01 0.1 1 10 Frequency (Hz) CHY079 1 Sa (g) 100 0.1 0.01 0.1 1 10 Frequency (Hz) 100 TREIRS stations 0.1 0.01 0.1 1 10 Frequency (Hz) 100 Calculated Observed Figure 8. Comparison of the observed with estimated response spectra in the CHY area for the Chi-Chi main shock. 11 100 2 2 ILA-Horizontal 1 ln (OBS/CAL) ln (OBS/CAL) TAP-Horizontal 0 -1 -2 0.1 1 Frequency (Hz) 10 1 0 -1 -2 0.1 100 10 100 10 100 10 100 HWA-Horizontal TCU-Horizontal 1 ln (OBS/CAL) ln (OBS/CAL) Frequency (Hz) 2 2 0 -1 -2 0.1 1 Frequency (Hz) 10 1 0 -1 -2 0.1 100 2 1 Frequency (Hz) 2 CHY-Horizontal TTN-Horizontal 1 ln (OBS/CAL) ln (OBS/CAL) 1 0 -1 -2 0.1 1 Frequency (Hz) 10 100 1 0 -1 -2 0.1 1 Frequency (Hz) 2 ln (OBS/CAL) KAU-Horizontal 1 0 -1 -2 0.1 1 Frequency (Hz) 10 100 Figure 9. Ratio between the observed and estimated response spectra for the seven areas in Taiwan. 12