Analysis And Simulation Of Earthquake Ground Motion Records

Earthquakes occur every day, everywhere since the earth has existed. The main reason for earthquakes is the main tectonic borders of convergent or divergent plate’s movements. Due to the convection currents in the mantle, plates move with respect to each other on the lithosphere like sailing on the sea. A large number of the earthquakes (about 90%) and volcanic eruptions occur in a zone called the ‘RING of FIRE’ in the Pacific Ocean.

There are series of oceanic trenches and volcanic belts due to the plate movements. Another of the most seismically active regions (5-6% of the total number of earthquakes) is the Alpide belt, which ranges from the Mediterranean in the west to the Himalayas in the east). Turkey and Iran are located in the western part of the Alpide belt while northern India is on its eastern part.

Earthquakes are recorded all over the world by seismometers that are electronic systems with a mass pendulum arrangement and record ground motion in two horizontal (NS, EW) and one vertical (Z) direction. The record of a seismometer is called a ‘seismogram’, which is a kind of signature of the earth. It is also the ground motion of the earth like an electrocardiogram (EKG) of heart. To reduce the effect of the earthquakes on the man-made structures and life as well as to protect ourselves from the natural disaster, we need to understand detail features of the seismograms. It is hard to understand the details of earthquake mechanism from its signatures… It is a long and hilly road for earth scientist…

In the beginning of the 1900s, seismologists started to record earthquakes. Meanwhile, seismologists improved methods by using mathematics and physics to extract the earth structures from analyzing ground motions (seismograms). After a long way, simulation of earthquake ground motion was started at the end of 19th century. The aim of the earthquake simulation studies is to define the character of ground motion as close to the real one to reduce the damage mitigations before, during, and after an earthquake.

Analyses And Simulation Of Ground Motions For Engineering Applications

In the last 20 years, the availability of high-quality ground motion records of earthquakes increased all over the world. Therefore, seismic hazard studies have become important for earthquake engineering applications. The most important key for assessing and mitigating earthquake disasters is the prediction of ground motion as close to reality as possible. It is the first step to reduce casualties and collapse buildings before an earthquake for damage assessment. Ground motion prediction generally has two processes as shown in Figure 1.

First, simulate ground motion on seismic or engineering bedrock. For this aim, we use source information and crustal velocity structure. These are called source and path effects. After determining waveform on seismic/engineering bedrock, we can calculate surface motion by multiplying bedrock motion with site amplification factor at each frequency in the frequency domain for evaluating of soil effect in sediments. Finally, surface motion in the time domain is obtained by applying inverse Fourier transform.

Figure 1. Schematic illustration of wave propagation through seismic bedrock and soil surface (modified from http://seismo.geology.upatras.gr/MICROZON-THEORY1.htm)

The amplitude of ground motion is controlled by three major effects: source, path, and site effects. Among them, site effects have sometimes played a principal role in damage to buildings due to local geological and soil conditions. The important characteristics of ground motions especially amplitude, frequency content, and duration can be affected by local site conditions. Heavy damage because of local site conditions is seen several large earthquakes such as 1985 Michoacan (Mexico), 1995 Kobe (Japan), 1999 Kocaeli and Duzce (Turkey), 2014 Iquique (Chile) and Nagano (Japan), 2015 Gorkha (Nepal), 2016 Kumamoto (Japan). Validating site effects is inevitable to estimate earthquake ground motion and disaster mitigation.

One of the important points for local seismic hazard studies is the definition of 1D velocity structure from the surface to the seismic or engineering bedrock. This is necessary to know site responses to estimate ground motions impact of the buildings. The local geology significantly modifies ground motion characters and controls the irregular distribution of damage observed during large earthquakes. Therefore, shallow low-velocity layers are responsible for the variation of earthquake ground motion amplification in an area. Heterogeneity of the soil structures, velocity impedance differences between layers, resonant effects, irregular topography of the layers beneath a basin, the effect of the surface topography, nonlinear soil behavior, fault geometry and lateral variation of S-wave velocity (Vs) causes variations of earthquake ground motion amplification.

Most of these effects cannot be determined easily and comprehensive studies are needed. On the other hand, it is important to estimate Vs structure in near-surface layers for estimating strong motion characteristics during an earthquake. One-dimensional (1D) soil profile can be obtained by different geophysical methods, using earthquake data or ambient noise recording to retrieve the vertical soil structure as well as borehole logging. The 1D assumption of the soil structure is widely accepted and easy to implement. Array exploration of microtremors has been gaining much popularity in 1D Vs profiling because estimation of Vs structure requires only a simple circular array with a few seismometers (Figure 2). If the microtremors are recorded by vertical sensors, they are often regarded to have the dispersive characteristics of Rayleigh waves. Additionally, surface wave group velocity dispersion curves from ground motions of an earthquake can be used to determine the horizontal 1D velocity structures for deep sediments and earth’s crust from focal layer to the engineering bedrock or seismic bedrock beneath a site. If there is no detailed 3D velocity information for a region, determining approximate 1D velocity structure helps to generate more reliable simulation results for seismic hazard studies.

Figure 2.  Microtremor array measurements. Credit: Ozlem Karagoz

An example: Simulation of the 24 May 2014 Gokceada (North Aegean Sea) Earthquake (Mw 6.9, NW Turkey)

Turkey is located between the three main tectonic plates: Eurasia, Arabia, and Africa. The result of the Eurasian-Arabian continental collision in the east and extensional regime in the Aegean, the Anatolia Plate escapes to the west between North and East Anatolian strike-slip fault systems similar to a watermelon seed being squeezed between two fingers as shown in Figure 3.

The North Anatolian Fault Zone (NAFZ) is one of the significant right-lateral strike-slip faults on the Earth. It is about is about 1,200 km-long between the Eurasian and Anatolian plates. It begins from eastern Turkey, cuts the Sea of Marmara in roughly east-west direction and then extends to the Aegean Sea in the west The NAFZ has a uniform slip-rate of ~25 mm/yr and releases the accumulated seismic energy with large earthquakes (M>7).

Our study area is located western part of the Marmara Region in northwestern Turkey. Hence the Marmara Region has the smallest area among the seven geographical regions of Turkey (Aegean, Black Sea, Central Anatolia, Mediterranean, Eastern Anatolia and Southeastern Anatolia regions), it covers a rapidly growing part of the country and encompasses the main financial and industrial centers, including Istanbul which is one of the most populated cities in the world. Marmara Region is suffered from destructive earthquakes and was selected as Supersites that are principle study regions for natural hazards in the world. Historical records show that destructive earthquakes have frequently been visited the region. In the last century,   9 August 1912 Murefte (Mw 7.3) earthquake occurred in the west while 17 August 1999 Kocaeli (Mw 7.4) occurred in the eastern parts of the region (Figure 3). The Gokceada earthquake (24 May 2014, Mw 6.9) also affected the west of the region.

Figure 3. Main tectonic units and fault systems in Turkey. Red rectangle shows the study area. EAFZ: East Anatolian Fault Zone, NAFZ: North Anatolian Fault Zone. Significant earthquakes occurred in the last century are shown by yellow circles. Credit: Ozlem Karagoz

Our deterministic numerical simulation well regenerated the high-frequency parts of the waveforms at most of the stations (Figure 4). This study provides the 1D S-wave velocity structural models at the strong motion stations to calculate reliable synthetic ground motions of the 2014 Gokceada earthquake. These subsurface models can be effectively used in prediction of strong ground motion due to future large earthquake along the NAFZ in the Marmara Sea.

Figure 4. Mainshock and aftershocks (M≥2, 24-25 May 2014) of the 2014 Gokceada earthquake from the KOERI catalog and observation sites used in this study. The focal depth of the mainshock is given as 21.2 km in the KOERI catalog. The focal mechanism solution of the mainshock is from Pinar (2014). The red line shows the North Anatolian Fault Zone. AFAD strong ground motion stations and microtremor array sites are shown by red triangles and yellow circles, respectively. The ruptured area given by Pinar (2014) is shown by a black rectangle on the depth section. The black line in the upper figure is its projection on the map. Republished with permission from Springer from: 10.1007/s10518-017-0207-6.

The importance of preparation of a proper 1D deep velocity structure beneath engineering bedrock to simulate low frequencies is emphasized by 1D ground motion simulation of the 2014 Gokceada earthquake. The local site effects were mostly successfully generated on the synthetic high-frequency seismograms (Figure 5). On the other hand, the 1D simulation cannot generate sufficient results at special sites. Sites with a thick sediment basin with low seismic velocity beneath stations may cause slow and large surface waves, which cannot be generated in the 1D simulation, because of generation of locally trapped waves.

Figure 5. Comparison of the observed (black) and simulated (red) velocity waveforms of the Gokceada earthquake (top), the location of stations in the map (bottom). Comparison of the amplitude spectra of the waveforms is on the top-right. Credit: Ozlem Karagoz

These findings are described in the article entitled Broadband ground-motion simulation of the 24 May 2014 Gokceada (North Aegean Sea) earthquake (Mw 6.9) in NW Turkey considering local soil effects, published in the journal Bulletin of Earthquake Engineering. This work was led by Ozlem Karagoz from the Tokyo Institute of Technology (formerly), Çanakkale Onsekiz Mart University (currently).