Near-fault seismic site response through observed and simulated data of the 2009 L’Aquila (central Italy) Mw 6.1 earthquake

Contents

Introduction

Case Study #1: Western L’Aquila Basin
The Aterno River Valley Strong-Motion Array
Geological, Geotechnical and Geophysical Data
Ground Motion Record Analysis
Dynamic Modeling
Peak Ground Acceleration Amplification Factor
Observed vs. Numerical Peak Ground Acceleration
Excess Pore-Water Pressure and Liquefaction
Observed vs. Numerical H/V and V/H Response Spectral Ratios
Conclusions
References

Case Study #2: Historical L'Aquila City center
Geological, Geotechnical and Geophysical Data
Ground Motion Record Analysis
Dynamic Modeling
Peak Ground Acceleration
Observed vs. Numerical H/V Response Spectral Ratio
Excess Pore-Water Pressure and Nonlinear Behavior
Conclusions
References

Introduction

During the night of 6 April 2009, at 01:32 GMT, an earthquake hit the L’Aquila Basin, in central Italy. The earthquake caused damage to between 3000 and 11000 buildings in the medieval center. Several buildings also collapsed. Three hundred and nine people died and more than 1500 were injured. About 65000 people (out of a population of 72000) had to abandon their homes. The Mw 6.1 main shock occurred with an epicenter at 42.3400° N, 13.3800 °E, approximately 90 km NE of Rome, near the city of L'Aquila. The earthquake was caused by movement on a NW-SE trending normal fault that was defined to strike 140° and dip 50° to the SW. Most co-seismic slip occurred in a small rupture area of approximately 16 km along strike and nearly 8 km in the dip direction, in a depth range of 4-10 km [41]. The main shock and many aftershocks of the 2009 L’Aquila seismic sequence were recorded by the near-fault stations Colle Grilli (AQG), Fiume Aterno (AQA), Centro Valle (AQV), Il Moro (AQM), Aquil Park Int. (AQK) that belong to the Italian Strong Motion Network (RAN) [20,51] and Aquila Castello (AQU) that belong to the National Institute of Geophysics and Volcanology (INGV). The stations are situated in the Upper Aterno River Valley and in the medieval City of L’Aquila. Later, several groups and institutions made a great quantity of data available as a result of geological, geotechnical and geophysical investigations [2,17,46], whereas various authors performed different types of research studies.

In particular, De Luca et al. [10] found evidence of low-frequency amplification at 0.6 Hz in the City of L’Aquila through analysis of earthquake and ambient noise data. They also performed 2D numerical modeling by means of both finite and boundary element methods, which allowed them to relate the low-frequency amplification to the presence of a sedimentary basin about 250 m deep.

Puglia et al. [34] performed spectral ratio analyses for the permanent seismic stations of the Upper Aterno River Valley array, which were based on earthquake and noise recordings and 1D elastic equivalent-linear modeling. They demonstrated a fundamental frequency shift from 3 Hz to 1.5 Hz at seismic station AQV during the 2009 Mw 6.1 main shock and suggested that one possible explanation was the nonlinear behavior of soil.

Chioccarelli and Iervolino [8] proposed that frequency change with time is not directly related to nonlinear shear modulus behavior, but to peculiar phases of the seismic signal whose frequency is more related to source effect such as directivity and flings.

Lanzo and Pagliaroli [26] estimated the seismic site effects in the Aterno River Valley through standard spectral ratio analysis of near-fault strong motion records by using seismic station AQG as reference site. They also performed 1D numerical modeling, but did not clarify the observed frequency change with magnitude level at seismic station AQV.

Del Monaco et al. [13] carried out microtremor recordings in historical Downtown L’Aquila and applied the Nakamura method [33], which revealed maximum amplification at the frequency ranges of 0.4-0.7 Hz and 3-15 Hz.

Gaudiosi et al. [19] focused their attention on a cross-section of the Upper Aterno River Valley, analyzed the strong and weak motion data by means of SSR (standard spectral ratio) technique and compared the results with those obtained from 2D numerical modeling that used both finite element and finite difference methods. They found evidence of a strong amplification at seismic station AQV, which was related to the constructive interference of S and surface waves. They also noticed nonlinearity in soil behavior.

Nunziata and Costanzo [33] applied a hybrid method (modal summation plus finite difference) to poorly defined geologic models of the Upper Aterno River Valley and Downtown L’Aquila and validated the numerical results with recordings at stations AQG, AQA, AQV, AQK and AQU. It resulted that: the alluvial sediments filling the Upper Aterno River Valley caused amplification at the frequency range of 2 and 7 Hz; L’Aquila breccia and sand, which outcrop in historical L’Aquila City, caused response spectral amplification at the frequency range of 0.6 and 7 Hz.

Considering the results of mentioned surveys, in this book I will try to present the outcomes of my research (36, 37), which is about the seismic response of sediments that fill L’Aquila Basin, by using two case studies: 1) Western L’Aquila Basin; 2) Historical L'Aquila City center. In particular, I will analyze the April 2009 recorded ground motions and compare the observed data with those revealed from finite element nonlinear dynamic models.

Case Study #1: Western L’Aquila Basin

The Aterno River Valley Strong-Motion Array

The strong-motion array in the upper Aterno River Valley (Figure 1), which is situated in Western L’Aquila Basin, is a part of the Italian Strong Motion Network. It was installed in 1994 to investigate the seismic site effects along a line transversal to the valley [4]. The array originally consisted of seven stations and was modified over the years. It is currently composed of six digital strong-motion instruments, which are named Colle Grilli (AQG), Fiume Aterno (AQA), Centro Valle (AQV), Il Moro (AQM), Ferriera (AQF) and Monte Pettino (AQP).

illustration not visible in this excerpt

Fig. 1. Study area map showing the Aterno River Valley strong-motion array and the epicenter of 2009 Mw 6.1 main shock. Map © OpenStreetMap contributors, http://www.openstreetmap.org/copyright/en

These stations are notably near the hanging wall of Mt. Pettino master extensional fault [15]. The array is approximately 2 km long, with a 250 m station spacing. Stations AQG, AQM and AQP are placed on Meso-Cenozoic carbonate rock outcrops; AQA and AQV are on the Aterno River alluvial deposits; and AQF is on the Mt. Pettino debris flow and pediment alluvial deposits. With the exception of AQF, which recorded only one event of the 2009 seismic sequence, the stations and the records that are considered for the analyses and subsequently used as the input motion in the 2D dynamic model, are summarized in Table 1. Stations AQF and AQP did not record the main shock because of a power supply failure. Station AQM recorded the main shock and was affected by saturation (PGA = 1 g) [38], but the data were not formally released because the recorded motion could be disturbed by earthquake-induced damage to the seismometer protective box [3].

Table 1

Analyzed events (recorded in the period 1998-2009) and maximum acceleration (PGA) according to the magnitude levels (ML and MW) for the examined seismic stations.

ML is the– local magnitude and MW is the– moment magnitude.

illustration not visible in this excerpt

Geological, Geotechnical and Geophysical Data

The geologic setting of Western L’Aquila Basin (Figures 2 and 3) consists of a Quaternary sediment-filled alluvial valley, which lies over a Meso-Cenozoic carbonate bedrock [15,45,47]. The Quaternary deposits are mostly represented by granular soils and reach a maximum thickness of approximately 50 meters near the station AQV. The carbonate bedrock, which includes flinty limestone (Corniola Formation) and intercalations of reef-slope facies detrital limestone (Maiolica Formation, Calcarenites and calcareous breccias with Fucoids), is structured as an asymmetric graben. It is variously dislocated by a series of northwest-southeast and northeast-southwest normal faults and a back-thrust; the latter brings the Maiolica Formation over the Miocene arenaceous-pelitic turbidites. Geological and geophysical data, which are obtained from the borehole logs and in situ seismic tests (Figures 4 and 5) [4,16,25,44,46], allow us to subdivide the investigated lithotypes into 10 geotechnical units and define the 2D subsurface model along the Aterno River Valley strong-motion array (Figure 3). Among the Quaternary deposits, it is possible to recognize seven geotechnical units (A, B1, B2, C, D1, D2 and E in Figures 3, 4 and 5) and three units (F1, F2 and F3 in Figures 3, 4 and 5) in the Meso-Cenozoic carbonate bedrock according to the soil and rock geotechnical properties and seismic wave velocity. Boreholes S4, S3, S5 and S1 also helped to define the water table in the alluvial deposits at 3.8, 5.1, 5.3 and 14.8 meters under the ground level, respectively (Figure 4). The geotechnical units are summarized in Table 2 with the geotechnical and geophysical features that are obtained from laboratory and in situ tests.

illustration not visible in this excerpt

Fig. 2. Geologic map of the studied area and the surrounding areas. Geological units. Quaternary units: Ri – anthropogenic fill material (Holocene); Cl – colluvium: fine-grained deposits (Holocene); Dt – debris slope deposits (Holocene); Al – Aterno River alluvium (Holocene); Cd – Mt. Pettino debris flow and pediment alluvial deposits: dense and poor to well-cemented calcareous gravel with tephra horizons; Cda – micro-karst calcareous gravel (upper-middle Pleistocene); At – Vetoio Stream terraced alluvium: gravel, sand and clayey-sandy silt with tephra horizons (upper-middle Pleistocene); Br – L’Aquila breccia: dense and poor to well-cemented calcareous gravel (middle Pleistocene). Meso-Cenozoic carbonate units: Sc – detrital Scaglia Formation: mudstone and fine grainstone (Eocene -Cenomanian); Ma – detrital Maiolica Formation, Calcarenite and calcareous breccias with fucoids: mudstone, fine grainstone and oolitic limestone (Cenomanian-upper Tithonian); Co – detrital Corniola Formation: mudstone and fine grainstone, which are sometimes dolomitized (middle Lias). Syn-orogenic unit: Mo – pelitic-arenaceous turbidites (middle Miocene). 1 – dip direction and angle; 2 – horizontal layers; 3 – fault; 4 – overthrust fault; 5 – fault dip; 6 – fault rock; 7 – alluvial fan; 8 – borehole; 9 – seismic station; 10 – geologic cross-section line. Modified after [47].

illustration not visible in this excerpt

Fig. 3. Geologic cross section of the studied area (its line is positioned in Figure 2). Geotechnical units. Quaternary units: anthropogenic fill material and colluvium: A – gravelly silty clay, clayey silt with gravel; Aterno River alluvium: B1 – silty clayey gravel with sand, poorly graded gravel, sand and silt with gravel; some level of clay with gravel; C – sandy silty clay with gravel, sandy silt with gravel, clay with sand, silty clay, silty clayey gravel with sand; B2 – poorly graded silty clayey gravel, well graded coarse sand; some level of gravelly silty clay; D1 – sandy silt and clayey silt with sand and gravel; poorly graded silty clayey gravel with sand, silty sand, sandy silt; D2 – sandy silt, silt with clay and sand sometimes with gravel, sandy silty clay sometimes with gravel, alternations of sandy silt-silty sand-silty gravel with sand; Mt. Pettino debris flow and pediment alluvial deposits: E – dense and poor to well-cemented calcareous gravel with tephra horizons. Meso-Cenozoic carbonate bedrock: F1 – highly fractured; F2 – moderately fractured; F3 – slightly fractured. Syn-orogenic unit: G – pelitic-arenaceous turbidites. 1 – back-thrust fault; 2 – normal fault; 3 – seismic station; 4 – borehole; 5 – fractured and weathered limestone.

illustration not visible in this excerpt

Fig. 4. Borehole logs of the investigated area (their position is in Figure 2). Geotechnical units. Quaternary units: anthropogenic fill material and colluvium: A – gravelly silty clay, clayey silt with gravel; Aterno River alluvium: B1 – silty clayey gravel with sand, poorly graded gravel, sand and silt with gravel; some level of clay with gravel; C – sandy silty clay with gravel, sandy silt with gravel, clay with sand, silty clay, silty clayey gravel with sand; B2 – poorly graded silty clayey gravel, well graded coarse sand; some level of gravelly silty clay; D1 – sandy silt and clayey silt with sand and gravel; poorly graded silty clayey gravel with sand, silty sand, sandy silt; D2 – sandy silt, silt with clay and sand sometimes with gravel, sandy silty clay sometimes with gravel, alternations of sandy silt-silty sand-silty gravel with sand. Meso-Cenozoic carbonate bedrock: F1 – highly fractured; F1(2) – highly fractured and weathered; F2 – moderately fractured; F3 – slightly fractured. Wt – water table.

illustration not visible in this excerpt

Fig. 5. Borehole logs with P and S wave velocity profiles (their position is in Figure 2). Geotechnical units. Quaternary units: anthropogenic fill material and colluvium: A – gravelly silty clay, clayey silt with gravel; Aterno River alluvium: B1 – silty clayey gravel with sand, poorly graded gravel, sand and silt with gravel; some level of clay with gravel; C – sandy silty clay with gravel, sandy silt with gravel, clay with sand, silty clay, silty clayey gravel with sand; B2 – poorly graded silty clayey gravel, well graded coarse sand; some level of gravelly silty clay; D1 – sandy silt and clayey silt with sand and gravel; poorly graded silty clayey gravel with sand, silty sand, sandy silt; D2 – sandy silt, silt with clay and sand sometimes with gravel, sandy silty clay sometimes with gravel, alternation of sandy silt-silty sand-silty gravel with sand. Meso-Cenozoic carbonate bedrock: F1 – highly fractured; F1(2) – highly fractured and weathered; F2 – moderately fractured.

Table 2

Mean geotechnical and geophysical features of the units. γ – unit weight; c – cohesion; φ – friction angle; VS – S-wave velocity; VP – P-wave velocity.

illustration not visible in this excerpt

Ground Motion Record Analysis

The strong- and weak-motion data are analyzed for 312 signals, including (the three-component acceleration waveforms), which were recorded using seismometers at the stations AQG, AQA, AQV, AQM and AQP during the events with magnitude Mw ranging from 1.6 to 6.1, in the period 1998-2009 (Table 1). The corrected acceleration time-histories are downloaded from the Italian Accelerometric Archive web site [28,46], viewed and analyzed to compute the observed maximum acceleration and velocity (Table 1) and the 5 % damping elastic acceleration response spectra [9,32]. The PGA (Peak Ground Acceleration) amplification factor of the recorded 2009 events with a magnitude range of 4.1 to 6.1 is computed based on the ratio between the recorded PGA at the site and the reference site during the same event, assuming that AQG and AQP are the reference stations. Nonetheless, these rock sites are not free of local effects [1] and are considered the ‘‘reference’’ sites to highlight the PGA amplification at the two soil stations AQA and AQV. (We are well aware of the limitations that this assumption implies).

Finally, the single station H/V (Horizontal-to-Vertical response spectral ratio) (10,18,21,27,29,40,42,43) and V/H (Vertical-to-Horizontal response spectral ratio) methods are used to evaluate the site effects at stations AQG, AQA, AQV and AQM.

Dynamic Modeling

To simulate the seismic site response at the four stations AQG, AQA, AQV and AQM, 2D dynamic analyses are performed on three parts (Figures 6a, b, c) of the geologic cross-section displayed in Figure 3, using QUAKE/W [35], (which is a software based on finite element formulations with a direct integration scheme in the time domain). The mesh pattern of the models consists of quadrangular and triangular elements (with global sizes of 7.5 m (Figures 6a and b) and 5 m (Figure 6c), respectively). The initial static analyses are performed to determine the initial state of stress in the ground before starting the dynamic analyses. The initial pore-water pressure conditions are computed by specifying the water table, which is revealed from the boreholes in Figure 4.

The lower boundary where the movement is fixed in both the x and y directions is positioned: for the model in Figure 6a, (the lower boundary is on the interface) between the Meso-Cenozoic fractured carbonate bedrock and the Miocene turbidites (Figure 3); for the models in Figures 6b and c, (the lower boundary is on the contact) between the Quaternary deposits and the Meso-Cenozoic carbonate bedrock (Figure 3).

To simulate the observed non-linear effective stress behavior (6,11,14,24,48,49), 122 nonlinear dynamic analyses based on the MFS (Martin-Finn-Seed) pore-water pressure model (7,24,50) are executed for each model (Figures 6a, b, c). The MFS model is based on the concept that the pore-pressure, which is generated during undrained loading, is related to the volumetric strain that would occur for the same stress increment under drained loading according to the following expression:

illustration not visible in this excerpt

where Er is known as the rebound modulus (Figure 7a) and Δεvd is the incremental volumetric strain that would occur under drained loading conditions. The expression for the incremental volumetric strain, which was developed by Martin [24], is the following:

illustration not visible in this excerpt

where γ is the dynamic shear strain amplitude in the cycle and εvd is the plastic or accumulated volumetric strain (Figure 7b). C1, C2, C3 and C4 are curve-fitting constants and in this sense are material properties.

The geotechnical unit properties that are used as input parameters in the analyses are presented in Table 3. The rebound modulus Er [24] and the MFS pore-water pressure functions (7,50) that are estimated for different materials are shown in Figure 7.

Ten-second windows of the horizontal and vertical acceleration time-histories are recorded by AQG, AQM and AQP during the 1998-2009 events with Mw ranging between 1.6 and 6.1 (Table 1), and the time windows are scaled and used as the input motion records. In particular, during each analysis the models (Figures 6a, b and c) are simultaneously subjected to the horizontal and vertical accelerations of the earthquake. Finally, the computed acceleration response spectra are normalized to calculate the numerical H/V and V/H response spectral ratios.

illustration not visible in this excerpt

Fig. 6. Model setup of the material zones (A, B1, B2, C, D1, D2, E, F1, and F1(2) in Figures 3, 4 and 5), the boundary conditions and the mesh patterns that were used to perform the 2D dynamic modeling. AQG,…, AQM are the– seismic stations.

Table 3

Geotechnical and dynamic properties of the units used as input parameters in the numerical analyses.

illustration not visible in this excerpt

illustration not visible in this excerpt

Fig. 7. a) Variation of the rebound modulus with the y-effective stress, which was obtained for silica sand at a relative density of 45% (Martin et al., 1975); b) Estimated variation of the plastic volumetric strain of the modeled materials with a number of shear strain cycles.

Peak Ground Acceleration Amplification Factor

For the recorded 2009 Mw 4.1-6.1 events, the horizontal (Figure 8a) and vertical (Figure 8b) PGA amplification factors display a maximum amplification at station AQV (on the thickest part of the alluvium), which can be related to the strong impedance ratios [22] that occur in the alluvium and between the alluvium and the carbonate bedrock (Figure 5e). The horizontal PGA amplification factor slightly decreases when the earthquake magnitude increases for stations AQG, AQA and AQV (Figures 9a, b, and c), which is related to the nonlinear behavior of the fractured-weathered rock (F1(2) in Figures 4 and 5a) and the alluvial deposits. The excess pore-water pressure growth with increasing energy level makes the alluvium and the fractured-weathered rock less stiff and more dissipative, which reduces the PGA amplification. The PGA amplification factor for the AQM horizontal component (Figure 9d) and the AQG, AQA, AQV and AQM vertical components (Figures 9e, f, g and h) increases when the earthquake magnitude increases, which does not appear to be particularly influenced by the nonlinear behavior.

illustration not visible in this excerpt

Fig. 8. Horizontal (a) and vertical (b) PGA amplification factors computed at seismic stations AQG, AQA, AQV, AQM, AQF and AQP using AQG and AQP as the reference sites for the 2009 events with magnitude MW 4.1-6.1. MW is the– moment magnitude and PGA is the– peak ground acceleration. 1 and 2 – AQF = (accepting the value) computed for the recorded MW 5.6 event; the amplification factors are estimated.

illustration not visible in this excerpt

Fig. 9. Horizontal (a, b, c, d) and vertical (e, f, g, h) PGA amplification factors computed at seismic stations AQG, AQA, AQV and AQM as a function of the magnitude MW. MW is the– moment magnitude and PGA is the– peak ground acceleration.

Observed vs. Numerical Peak Ground Acceleration

To understand the peak values of acceleration that may occur at the seismic stations, during an earthquake such as the 2009 Mw 6.1 main shock, 102 nonlinear dynamic simulations are performed. We assume 10-second windows of the horizontal and vertical acceleration time-histories, which are recorded by AQG during the 2009 Mw 6.1 main shock and by AQG, AQM, AQP during the 2009 Mw 4.4-5.6 aftershocks, as the input motion records. The horizontal and vertical records are scaled to the peaks of 0.2 g and 0.1 g, respectively.

Table 4 presents a comparison between the observed horizontal maximum acceleration and the computed value for stations AQG, AQA, AQV and AQM. The values are notably consistent. The simulations reveal that during an event such as the 2009 Mw 6.1 main shock, the expected maximum acceleration is usually higher at the stations on soil and can vary at each station because the shape and the frequency content of the input seismic signal vary.

The 2D simulations have also allowed us to draw contour maps of the horizontal acceleration (Figure 10a) and the velocity (Figure 10b) and obtain the relative acceleration and velocity vectors, which display a clear seismic wave amplification at station AQV.

Table 4

Observed during the 2009 MW 6.1 L’Aquila earthquake and computed (mean +/- 1 standard deviation) maximum acceleration at seismic stations AQG, AQA, AQV and AQM. NS denotes the– north-south component, and WE denotes the– west-east component.

illustration not visible in this excerpt

illustration not visible in this excerpt

Fig. 10. 2D simulation of the 2009 MW 6.1 L’Aquila earthquake using the model displayed in Figure 6b: a) X-acceleration contour map and the relative acceleration vectors (blue arrows) at 2.56 seconds of the nonlinear dynamic analysis; vector length magnification: 37 times; b) X-velocity contour map and the relative velocity vectors (black arrows) at 3.25 seconds of the nonlinear dynamic analysis; vector length magnification: 71 times. AQG,…, AQM are the– seismic stations.

Excess Pore-Water Pressure and Liquefaction

The 2D nonlinear dynamic analyses based on the MFS (Martin-Finn-Seed) pore-water pressure model [24] provided insights on the excess pore-water pressure that was generated under dynamic loading. In particular, the simulation of 2009 Mw 6.1 main shock revealed a maximum excess pore-water pressure of 400 kPa near the interface alluvium-carbonate bedrock and next to station AQV (Figure 11a). Based on the initial state of stress and the pore-water pressure changes, which were computed during the shaking, it was possible to evaluate the potential liquefaction zones (Figure 11b) in the units C, D1 and D2 (Figures 3, 4, 5 and 6). Moreover, the time-histories of excess pore-water pressure, which were computed at different depths under the seismic stations AQG, AQA and AQV (Figures 12a, b and c, respectively), exhibited a clear increase that reached the highest values after 2 to 3 seconds of shaking. The maximum excess pore-water pressure values, which range between 50 and 300 kPa, denote a reduction of the effective stress and shear resistance that is compatible with the observed fundamental frequency variation.

[...]