The conceptual workflow (Fig. 5) that leads to the evaluation of the expected seismic action at Nafplio town started with the construction of the preliminary geological model based on existing literature and the stratigraphy of the previously available surveys. The interpreted geological cross-sections of the historical town centre represented the geological model of reference that has been validated using a large set of in-situ geophysical survey results, i.e., by fitting site resonance frequency (f0) with the main peak of the amplification function obtained by the 1D linear elastic modelling of the corresponding site stratigraphy.
The validated engineering geological model has been represented in 3D, which allows observing the spatial relationships between different geological bodies with marked differences in physical and mechanical properties. The reliability of local seismic response analyses is indeed directly related to the accuracy of the engineering geological reconstruction of the near-surface subsoil.
A 2D numerical modelling of seismic motion has been performed with an equivalent linear approach (finite elements) with LSR 2D (by Stacec) on 4 selected engineering geological cross-sections to extract the time histories and elastic response spectra (5% damping) on 124 control points distributed on the cross sections.
By analysing the values of the amplification index along the modelled cross-sections and the corresponding engineering geological model of the subsoil, a total of 6 zones with peculiar effects of expected seismic amplification were recognised and mapped. An average elastic response spectrum has been computed for each zone and each analysed return period (i.e., 50, 475 and 2000 years).
4.1. Engineering-geological modelling
The reconstruction of geological and engineering geological models of the Nafplio area has been achieved using several previous investigations and the data collected from the on-site geological and engineering geological surveying, as well as previous scientific literature (Karastathis et al. 2010).
Regarding the previously available data, the following types of investigations were considered:
-
Geophysical investigations: data of a seismic reflection campaign conducted by the Hellenic Survey of Geology & Mineral Exploration (HSGME) with the morphological and stratigraphic reconstruction of the seabed of the Gulf of Nafplio, thanks also a bathymetric map at a scale of 1:5000. Through the seismic reflection, the geological structures below the seabed were also outlined with the identification of 3 unconformities, which define four seismic units. HSGME also conducted, both in the coastal area and the hinterland of Nafplio, a series of seismic reflection surveys to detect the depth of the bedrock and the possible presence of faults. HSGME also carried out 4 cross-hole investigations, 2 of which are close to the town of Nafplio. Results show the presence of a loose deposit, consisting of clay and silt, with S-wave velocities of less than 400 m/s. The velocity of the P-waves is instead higher than 1500 m/s, probably also due to the saturation of the deposit for marine intrusion;
-
Boreholes: a total of 21 boreholes, 17 offshore, in the port area, and 4 on land, (performed by Triton Consulting Engineers S.A., Geomichaniki, HSGME) have been consulted. The same stratigraphic sequence characterised the analysed boreholes, with differences in the thickness of the lithotypes: off the harbour to the northwest, between the coast and Bourtzi island, the terrigenous component is prevalent, while in the northern sector of the offshore of the ancient town, the terrigenous component is lower than the lithoid one. All the stratigraphies obtained from the boreholes were used to better constrain the preliminary geological model.
Together with the information derived from previous studies and the technical reports used to reconstruct the geological model of the plain and the gulf in front of Nafplio, two detailed geological and geotechnical field surveys were carried out in July 2020 and July 2021, mainly focused on the reconstruction of the geological model of the Acronafplia and Palamidi reliefs and distribution of the recent deposits. The on-site survey at Acronafplia revealed that the hill is composed of heavily fractured Cretaceous limestone, with flysch outcropping in the depression towards Palamidi ridge. The survey along the ridge’s southern and western sides confirmed susceptibility to rock falls and coastal erosion leading to shoreline retreat. The N/NW side of Palamidi features a thrust with an NW vergence, where an inverted stratigraphic succession crops out. The thrust zone shows complex deformation, with fractured limestone, flysch, and striated calcite surfaces. The area between the ridges exhibits repeated outcrops, and a normal fault has lowered the Acronafplia peninsula. The northern slope of Palamidi is surrounded by a Pleistocene gravel deposit, which is in erosive contact with Eocene flysch and extends to the coast along the W-SW slope.
The analysis of the bibliographic sources, the available technical reports and the data collected during the two field surveys allowed the definition of a detailed and updated geological map concerning the versions available in the literature (Fig. 2). Furthermore, 9 geological cross-sections were conceptualised to constrain the geological model in depth and for better reconstructing the lateral relations of the various geological bodies (Fig. 6).
The critical analysis of both previous data and field surveys has allowed the reconstruction of an engineering geological map (Fig. 7) and nine engineering geological sections (Saroglou et al. 2021), highlighting that outcropping geological formations can be regrouped into three main lithotechnical units as bedrock, recent soft soil materials, and man-made ground.
4.2. Single-station seismic ambient noise measurements
A total of 99 single-station seismic ambient noise measurements were carried out to validate the geological model between 2020 and 2021. The survey area covered the different zones identified in the Nafplio area (Fig. 7), each representing peculiar geological characteristics as identified by engineering geological modelling. The single-station seismic ambient noise measurements were performed using two different kinds of sensors: i) LE-3D/5 s three-component seismometers (0.2 Hz eigenfrequency) together with REFTEK 130-01 dataloggers; ii) SL06 24-bit digitisers with built-in SS20 three-component velocimeter (2.0 Hz eigenfrequency) by SARA Electronic Instruments. Seismic ambient noise was recorded for 1 hour at each measurement site at a sampling frequency of 250 Hz. and elaborated by GEOPSY (Wathelet et al. 2020) with the following processing parameters:
- a 40 s windows with overlaps of 25% for each time series;
- tapering to 5% and smoothing of the spectrum for each window by the algorithm of Konno and Omachi (1998) with the b parameter = 40;
− 0.2–20 Hz frequency range of the analysis.
The horizontal-to-vertical noise spectral ratio (HVNR) method, here utilised to elaborate the seismic ambient noise measurements, is commonly employed to assess the 1D stratigraphic resonance of locations where there is a significant contrast between the low S-wave velocity subsoil and the seismic bedrock beneath it (Haghshenas et al. 2008). In HVNR processing, the Fourier transforms (FFTs) of the horizontal ground motion components are averaged and subsequently divided by the FFT of the vertical component. Additionally, Geopsy software was employed to analyse the polarization of the HVNR peak. The average HVNR function was computed in 10° increments from 0° to 180° and reconstructed on the horizontal plane as a function of azimuth.
The HVNR functions, obtained from the measurements carried out on the recent deposits of the Nafplio coastal plain, are characterised by a marked peak having frequency values of 2–5 Hz in the historical town centre (Fig. 8, left panel) and 1–2 Hz in the modern town area (Fig. 8, middle panel). These values can be associated with the stratigraphic resonance within the coastal plain deposits since the HVNR peaks present the characteristic “eye-shape” (Castellaro and Mulargia 2009) of the Fourier amplitude spectra, i.e., horizontal components increase, and vertical component decreases, and no prevalent direction of polarization is noticed. In addition, the increase of the frequency resonance found from the measurements performed on the coastal plain (Fig. 7) can be referred to a decrease in the deposit thickness moving from the modern town to the historical centre area in agreement with the reconstructed subsoil model (Fig. 6). On the other hand, the HVNR functions obtained from the seismic ambient noise measurements performed on the Acronafplia relief do not show any significant peak (Fig. 8, right panel), strengthening the hypothesis of the presence of the limestone seismic bedrock.
4.3. Engineering geological model validation
A first phase of numerical modelling of seismic wave propagation was performed to validate and calibrate the reconstructed engineering geological model. An iterative procedure (Moscatelli et al. 2021; Iannucci et al. 2022) based on the comparison between the amplification functions obtained by 1D numerical simulations of seismic wave propagation performed with the Strata software (Kottke and Rathje 2008), and the HVNR curves, obtained from the single-station seismic ambient noise measurements, were used as explained in the following.
A seismostratigraphy was derived from the engineering geological cross-sections of Nafplio in correspondence with the sites where the measurements were performed. Each seismostratigraphy contains a log of lithotechnical units (subsoil column) composed of a soft soil multilayer, with relative thickness supposed in the site, a seismic bedrock infinite half-space placed at a validated depth and a value, or a range of values associated with each unit based on geotechnical data already available. The numerical simulations were performed for each soil column by a linear elastic approach, varying firstly the values of unit weight volume (γn) and shear wave velocity (Vs) within the ranges defined for each lithotechnical unit and secondly the thickness of the lithotechnical units up to obtain a correspondence between the amplification function (i.e., the ratio of the FFT amplitude spectra at the subsoil surface and at the outcropping seismic bedrock) obtained by the numerical modelling and the HVNR peak from the seismic ambient noise measurements. Finally, the cross-sections were modified according to the thickness of the lithotechnical units obtained by the validation and calibration process.
Table 2 reports the physical and geotechnical parameters of each lithotechnical unit obtained by the validation and calibration procedure, which were used in the subsequent 2D numerical modelling to assess the expected seismic shaking. Based on available literature data, dynamic parameters of the soils, such as decay curves for damping and shear stiffness modulus, were attributed to the multilayer.
Table 2
Parameters of the lithotechnical units obtained by the validating and calibrating process and used for the subsequent 2D numerical modelling.
Litothecnical unit
|
𝛾n (kN/m3)
|
Vs (m/s)
|
Decay curve
|
Man-made fill
|
18
|
150
|
elastic linear D = 2%
|
Clayey sands
|
20
|
170–240
|
Anastasiadis et al. (2001)
|
Marly-clayey flysch
|
19
|
300
|
National Seismic Service of Italy (2003)
|
Loose gravels
|
19
|
250
|
Lo Presti et al. (2002)
|
Sandy gravel
|
20
|
420
|
Lo Presti et al. (2002)
|
Conglomerates
|
23
|
750
|
elastic linear D = 2%
|
Bedrock (limestones)
|
23
|
1100
|
elastic linear D = 0.5%
|
4.4. 3D engineering geological modelling
The validated geological engineering model was exploited for the extrapolation of virtual stratigraphic boreholes at 358 points located along the 9 cross-sections previously validated. The spacing of the virtual boreholes, set equal to 50 m, locally descends to 10 m to increase the resolutive accuracy at complex geological structures. The geophysical-stratigraphic data collection belonging to a total of 406 points was stored in a geodatabase and processed for the generation of a 3D solid model (Song et al. 2019). The relational geodatabase constitutes a digital record collection model of georeferenced data that is stored in a multiple Excel worksheet (Ciampi et al. 2021). Stratigraphic parameters associated with the drillings (elevation and lithotechnical units) were interpolated using the inverse distance weighting technique, employing 4 neighbouring points and an exponent of 2, for generating a voxel-based solid model (Liu et al. 2021). A high-fidelity filter was adopted to honour the value of the control point variable (Ciampi et al. 2022). The domain of the data-driven block model consists of a 3D mesh with a node resolution of 10 m (x) x 10 m (y) x 0.2 m (z). The model features 153 x 123 x 106 voxels in three dimensions and spans from − 66 to 144 m a.s.l., covering an area of about 130 ha. A 4D modelling, performed based on the temporal evolution of coastal deposits inferred from 1960 and 2020 aerial photos, provides a volumetric assessment of geomorphological processes and erosional dynamics over time. The resulting digital models paint data-driven 3D geologic structures and capture lateral-vertical variations of lithotechnical units, employing a vertical exaggeration factor of 1.5 to highlight lithologic transitions (Fig. 9).
The three-dimensional geological model reproduces the subsurface geological architecture as well as the spatial relationships between the various geological units. In contrast to conventional 2D representation techniques, which omit significant portions of the spatial domain, the 3D model empowers anatomical restitution of entire subsurface volumes through solid reconstruction, paving an opportunity for overall interpretation and qualitative/quantitative analysis of the geological bodies’ spatiotemporal distribution (Royer et al. 2015; Antonielli et al. 2023). A volumetric computation in post-processing operations reveals the deposition of about 0.8 million m3 of filling material in the last 60 years, testifying to advancing littoral dynamics driven by coastal anthropisation over time (Ouillon 2018). Besides, dynamic extraction of 2D engineering geological cross-sections from the data-driven solid model encourages numerical modelling of the local seismic response (Fig. 10).
4.5. Numerical modelling of the local seismic response
To evaluate the seismic response of the studied area, a subsequent phase of numerical modelling was conducted using the LSR 2D software (STACEC s.r.l.). This tool enables two-dimensional modelling of seismic motion propagation via an equivalent linear analysis, applying finite element methods in the time domain under total stress conditions and incorporating a viscoelastic rheological behaviour based on the Kelvin–Voigt model. Physical, geophysical, and decay parameters (such as shear stiffness and damping degradation curves), designated to the geological-technical units (refer to Table 1), were assigned to the multilayer structure, and the preselected seismic inputs were applied to the infinite half-space representing the seismic bedrock. Additionally, the model was discretised into triangular meshes with an approximately 2 m resolution grid. This grid was automatically generated by the software to achieve optimal resolution, considering the varying S-wave velocities of the layers and a maximum analysis frequency of 20 Hz. Free-field conditions (i.e., damping in the X and Z directions) were applied on the lateral borders, along with a kinematic constraint on vertical motion at the base of the domain within the infinite half-space. Numerical simulations were conducted for all reconstructed engineering geological cross-sections using one of the 3 sets of seven time histories. According to ICMS (2008) guidelines, these histories were selected for the number of inputs required for numerical simulations, with 50, 475, and 2000-year return periods. The ESM database and the Rexel software were used, with a spectro-compatibility period range of 0.15-2s and tolerances of 10% lower and 30% upper.
This numerical analysis made it possible to determine the following for each control point along the cross-sections (indicated by red dots in Fig. 11) and for each analysed return period: the Fourier amplitude spectrum, the amplification factor (AF) according to ICMS (2008) standards, and the elastic response spectrum (with 5% damping).
The results from the 2D modelling, expressed as horizontal acceleration, indicate higher values in the soft soils along the valley edge, likely due to diffraction and reflection of seismic waves caused by a non-horizontal, shallow bedrock. An incremental rise in acceleration was observed at the top of elevations, such as Acronafplia Hill. This increase can be attributed to the ridge effect, where the focusing of seismic waves leads to amplification at the peak of the ridge.
4.6. Local seismic response from recorded earthquakes
For the present study, a temporary seismic net of 4 seismic stations composed of 3-component velocimeters (2.0 Hz eigenfrequency) by SARA Electronics Instruments was also installed in a continuous acquisition mode, distributed in peculiar places of the Nafplio town chosen based on the results derived from HVNR analysis (see section 4.2). This seismic net allowed the recording of regional and local earthquakes to be used to define the amplification functions by the spectral ratios to reference (SSR) (Borcherdt 1970) and horizontal-to-vertical earthquake spectral ratio (HVER). The stations (see Fig. 7 for location) were active in the period 17–25 September 2022 and recorded more than fifty events (1 < M < 5) out of about 160 recorded by the Hellenic Seismic Network (https://bbnet.gein.noa.gr/).
Table 3
Earthquake selected for the cluster 1, 1 < M < 2.5 until to 150 km with main sources W-NW7
DATE
|
TIME (UTC)
|
LAT
|
LONG
|
DEPTH (Km)
|
Ml
|
17/09/2022
|
12:50:51
|
38.166
|
21.6463
|
19.2
|
2.1
|
17/09/2022
|
17:51:55
|
36.9699
|
21.351
|
15.9
|
2.5
|
19/09/2022
|
8:58:33
|
37.3755
|
23.2983
|
79.4
|
2.1
|
20/09/2022
|
9:35:43
|
37.4725
|
22.2821
|
16
|
1.7
|
20/09/2022
|
11:12:59
|
38.3043
|
23.3153
|
7.8
|
2
|
20/09/2022
|
11:50:45
|
37.6204
|
23.2512
|
15.5
|
2
|
20/09/2022
|
17:58:29
|
37.3801
|
23.3395
|
22.1
|
1.7
|
20/09/2022
|
21:39:56
|
37.6749
|
22.6474
|
63.6
|
1.7
|
20/09/2022
|
23:12:45
|
38.1715
|
21.6463
|
20
|
2
|
21/09/2022
|
0:47:15
|
37.4162
|
23.3208
|
14.9
|
1.5
|
21/09/2022
|
4:37:16
|
37.4382
|
23.1404
|
16.1
|
1.6
|
21/09/2022
|
7:10:34
|
37.408
|
23.4206
|
19.1
|
1.6
|
21/09/2022
|
14:02:18
|
37.462
|
21.9868
|
10.4
|
2.5
|
24/09/2022
|
5:28:00
|
37.9555
|
22.5893
|
10.3
|
1.3
|
24/09/2022
|
6:08:19
|
37.9697
|
22.5851
|
10.9
|
1.5
|
24/09/2022
|
9:28:37
|
37.8223
|
22.7637
|
7.4
|
1.4
|
24/09/2022
|
19:59:39
|
37.5577
|
23.5703
|
11.2
|
1.3
|
25/09/2022
|
3:05:57
|
38.237
|
22.8543
|
12.8
|
1.1
|
25/09/2022
|
6:38:07
|
37.9628
|
22.5993
|
13.7
|
2.3
|
Table 4
Earthquake selected for the cluster 2, 2.6 < M < 5 until 500 km with main sources W and S.
DATE
|
TIME (UTC)
|
LAT
|
LONG
|
DEPTH (Km)
|
Ml
|
17/09/2022
|
18:41:41
|
35.5728
|
25.8261
|
15.7
|
4.2
|
18/09/2022
|
1:49:07
|
35.554
|
25.8073
|
18.9
|
2.7
|
18/09/2022
|
14:45:29
|
38.156
|
21.6591
|
26.7
|
2.8
|
19/09/2022
|
3:31:37
|
35.0775
|
26.424
|
10.4
|
4.2
|
19/09/2022
|
10:17:38
|
37.9399
|
20.2263
|
17.5
|
2.8
|
19/09/2022
|
11:40:35
|
37.9065
|
20.0867
|
9.4
|
2.7
|
19/09/2022
|
22:43:20
|
37.8085
|
20.3641
|
6.8
|
3.1
|
20/09/2022
|
1:31:27
|
39.4048
|
20.7069
|
15.7
|
2.9
|
20/09/2022
|
7:02:02
|
37.8745
|
20.2469
|
10.7
|
2.6
|
20/09/2022
|
11:53:50
|
35.5733
|
25.8257
|
18.2
|
3.4
|
20/09/2022
|
16:42:40
|
37.131
|
21.0818
|
5.8
|
3.2
|
21/09/2022
|
6:27:04
|
37.9166
|
20.159
|
26.4
|
2.8
|
21/09/2022
|
9:59:58
|
38.3418
|
22.1169
|
11.4
|
2.7
|
21/09/2022
|
17:33:31
|
39.0202
|
23.5368
|
11.8
|
2.9
|
23/09/2022
|
0:00:03
|
37.847
|
20.0583
|
10.9
|
2.8
|
23/09/2022
|
3:43:14
|
36.2256
|
21.3721
|
13.2
|
2.7
|
23/09/2022
|
7:30:11
|
36.2622
|
21.2622
|
23.5
|
2.8
|
23/09/2022
|
14:15:52
|
35.6708
|
27.5226
|
25.7
|
3.9
|
23/09/2022
|
22:20:38
|
34.8898
|
26.7325
|
6.5
|
3.4
|
24/09/2022
|
13:21:46
|
36.2732
|
21.427
|
16.9
|
2.6
|
24/09/2022
|
15:41:34
|
36.2494
|
21.3959
|
14.3
|
2.6
|
24/09/2022
|
18:16:20
|
36.2439
|
21.4197
|
28.8
|
3.2
|
25/09/2022
|
2:04:08
|
39.537
|
25.145
|
13.7
|
2.6
|
Among the total recorded earthquakes, 42 signals only were considered suitable to be processed and were divided into two different clusters magnitude-distance (from Nafplio) (Fig. 12): cluster 1, 1 < M < 2.5 (19) until 150 km with main sources W-NW (Table 3) and cluster 2, 2.6 < M < 5 (23) until 500 km with main sources W and S (Table 4); for each cluster the following elaborations were firstly performed: |
-
Trim of the continuous records using the onset time of events.
-
Fast Fourier Transform (FFT) calculation for each component of all selected earthquakes;
-
application of the smoothing function by Konno and Ohmachi (1998) on the FFT spectra (considering one single window for each earthquake);
-
computation of the HVER function as the ratio between the quadratic mean of the two horizontal (H) FFT spectra and the vertical (V) FFT spectrum for each earthquake;
-
computation of the mean HVER function, with associated standard deviation, for each site by averaging the HVER calculated for all earthquakes of the cluster (Fig. 13).
By comparing HVNR and HVER for the 4 sites distributed in Nafplio town, it has been confirmed that an absence of amplification effects results in the case of outcropping bedrock, i.e., for SS1 (A in Fig. 13) situated close to the town centre; for the SS2 (B in Fig. 13) station the HVER analysis reveals weak amplification effects at frequencies lower than 2 Hz, probably related to the location of the station at the top of a ridge. Moreover, in the case of sites SS3 (C in Fig. 13) and SS4, localised downtown, HVNR and HVER are in good agreement and confirm evidence of amplification effects at resonance frequencies of about 2.5 Hz for SS3 (C in Fig. 13) and 7 Hz for SS4 (D in Fig. 13). These results highlight the different local geological conditions of the subsoil and the deepening of the seismic bedrock moving towards the northern sector of the town.
Based on the HVER analysis, the SS1 station was chosen as the reference site and to better constrain the local seismic response analysis, the selected earthquake signals were used to define the amplification functions by computing the standard spectral ratio (SSR - Borcherdt 1970) according to the following steps:
-
standard spectral ratios SSR were computed for each measurement site as the ratio between homologues components (E-W and N-S) of the earthquake recorded in SS2, SS3, SS4 and in the reference site SS1;
-
the mean SSR function was computed, with associated upper and lower standard deviation limits, for each measurement site by averaging the SSR calculated for all earthquakes of the cluster (Fig. 14).
Results from SSR analysis were compared with the outputs from 1D and 2D numerical modelling under elastic conditions to validate the amplification effects observed in Nafplio town. In general, it is worthy to observe (Fig. 14):
-
a good agreement in the main amplification frequencies at all measurement sites has been obtained except for SS3, cluster 2 (D), where some problems in raw data were present;
-
Amplification amplitudes from SSR functions are generally higher than the ones from the modelled functions;
-
For the SS4 measurement site, the SSR functions also show amplification frequencies different from the main mode, not evidenced by the outputs of 1D and 2D numerical modelling, probably due to minor seismic impedance contrasts related to the local subsoil layering.