The presence of discontinuities and faults inside a rock mass has a significant influence on the way of propagation of seismic waves and on their amplification effects on the ground level  . The situation will further complicate if one or more underground caves (anthropogenic or natural) are present in the rock mass , , , , . The site effect evaluation appears to be complex because of these conditions, so that it cannot be conducted with one-dimensional numerical code. On the contrary, two-dimensional or three-dimensional software, such as UDEC , are more appropriate to conduct this kind of analysis because it is possible to model both the presence of an underground cave and the discontinuous and anisotropic characteristics of the geotechnical model. Previous studies mainly referred to models with homogeneous and isotropic rocks, without considering rock mass with faults and underground karstic caves. The local seismic response (LSR) analysis of these models (especially in the professional practice) is usually conducted on 1D models, in which only the stratigraphic variation is considered. In the past few years, studies on LSR have been increasing significantly in literature. They aim to evaluate the influence of reliefs, valleys, faults and underground caves , , , , , , as well as the soil–structure interaction , . However, even more recent studies are based on homogeneous and isotropic subsoil models and are mainly conducted with finite element method (FEM), finite difference method (FDM), and boundary element method (BEM) calculation software which are not suitable in the case of discretely fractured and/or stratified rock masses, unlike the discontinuous deformation analysis (DDA) ,  and distinct element method (DEM), the latter used in the present case study , . Studies about LSR on real and/or parametric cases ,  were conducted on homogeneous and isotropic subsoil, varying the diameter and the depth of the cave. These studies highlighted the strong influence of these two variables over the amplification factor (FA), mostly in the frequency domain (Fig. 1). Overall, a decrease in the FA factor is observed as the cave depth (H) increases and, considering the same depth, as the cave diameter (D) decreases. Moreover, previous studies underline that diameter’s influence on LSR decreases dramatically for D ≤ 5.6 m, that is, for diameters comparable to case study. Furthermore, the effects of amplification/deamplification of the ground motion above a cave are significantly different, depending on whether the Vp or Vs volume waves hit the cave. In the case of Vs waves, the amplification effects on the surface are concentrated on the cave sides, while a deamplification occurs above the cavity. On the other hand, in the case of Vp waves, the amplification of the ground motion is concentrated at the centre of the cavern and decrease sharply, moving to its boundaries.
The UDEC numerical program was used to simulate the LSR of an anisotropic and discontinuous geotechnical model of a real case study, in order to analyse the effects of the simultaneous presence of underground caves, discontinuities and lithostratigraphic variations.
2 Geostructural and morphological layout of the area
The case study area is part of the extensive Apulian karst area of the Murge. In particular, it is located in the northwest neighbourhood of the town of Turi (Bari, Italy). From the geological and stratigraphic point of view, the rocks that constitute the substratum can be traced back to the ‘Calcari di Bari’ Mesozoic formation. This is a succession of whitish micritic limestones, dolomite limestones and greyish dolomite with local small inclusions and/or grey leading to black coloured laminations (Fig. 2). These limestones have a thickness of several hundred metres. In the study area, they are stratified with layers of 1-m thickness (standard deviation, 0.5 m), position D = 330°, i = 5° (standard deviation, 1°) and I = 60°. The tectonic structures found in the area are attributable to two phases, characterised by different tension fields. The first phase had a compressive nature and generated a smooth anticlinal fold, whose axis are oriented approximately northwest–southeast. The fold lies a few hundred metres south of the study area which is, therefore, placed on the northeast side of this anticline. On the other hand, the second phase generated two normal faults defining a small graben, with northeast–southwest oriented axis, in which the case study area falls.
However, it must be said that the two normal faults are not directly detectable on the surface, as they are inter-formational and with minimal vertical component of dip-slip.
Their location and orientation stemmed from the correlations between the stratigraphic reconstruction through different borings performed and the electrical tomographies acquired from previous studies. An ephemeral watercourse has set itself, with a south south east - north north west course, along the tectono-karstic depression of the graben, which is locally named as ‘lama’. This depression plays a role in drainage and vehiculation of rainwater. However, the water flow occurs underground (and not on the surface), because of the lithological nature of the rocks in the area and their high permeability (because of fracturing and karst), especially in the most superficial layers of the rock mass. As a result of the carbonatic nature of the rock mass, the landscape morphology of the area is essentially related to the karst phenomenon. It manifests itself both with epigeal morphologies, such as the dolina located mostly south of the study area, and with hypogean morphologies, such as the sinkhole ‘capovento’ located at the west side of the ‘Don Bosco’ nursery school. Other hypogean karst forms have been found a few hundred metres to north north east from the study area, such as the cave of ‘S. Oronzo’ and an adjacent small cave located a few decimetres underneath the surface.
3 In situ investigation and geomechanical model of the rock mass
The widespread presence of underground karst caves in the Murge area, the proximity of the S. Oronzo’s cave and the structural arrangement of the area suggest that the ‘capovento’ sinkhole is not isolated, but connected to an underground karstic cavern. This assumption is supported by resistivity values found in previous indirect investigations, provided by the technical department of the municipality of Turi (BA), such as electric tomographies with Wenner–Schlumberger arrays (Fig. 3).
The analysis of the n.1 tomography (the most representative one) showed a low resistivity anomaly, flanked by high resistivity areas, beneath the sinkhole ‘capovento’. The trend of resistivity is confirmed by the graben structural layout, in the centre of which the karstic sinkhole ‘capovento’ develops. The normal faults, positioned on the sides of the graben, are identified by the sharp contrast between high resistivity values on the sides and low values in the centre. Finally, low resistivity values can be explained by two evidences. The first one is the underground water circulation, which is channelled along the tectono-karstic depression ‘lama’, at a depth of about 10–14 m from the surface, along directives from southeast to northwest. The second one is the possible presence of underground karst caves filled with residual clay, known in the area as ‘red soils’. Mechanical surveys, with continuous core drilling and core destruction, were performed with the purpose of clearly identifying the presence of a karstic cave connected to the sinkhole and to potentially reconstruct its geometry (Fig. 4).
In particular, 25 core destruction borings (diameter Φ = 60 mm and depth of 6 m from the ground level) and 2 rotation drills with continuous coring (diameter Φ = 101 mm and depth of 15 m from the ground level) were performed. Both the core destruction borings and the rotation drills with continuous coring had the purpose of finding the hypogeal caves and determining their geometric extension (Fig. 5). Moreover, those with continuous coring have allowed to collect six samples for geotechnical tests (two in the S1 core drill and four in the S2 one) and to evaluate the rock quality designation (RQD) index and its variation along boring verticals (Fig. 6). Furthermore, the planimetric and altimetric characteristics of the nearby S. Oronzo’s cave were considered for the reconstruction of the geometry of the hypogeal caves, which have been identified with the geological surveys.
The S. Oronzo’s cave, indeed, is a karstic one of interlayer type with a planimetric development (91.00 m), which is predominant compared to the altimetric one (20.50 m). Besides, it is articulated on two overlapping karst levels. The deepest level, northwest–southeast oriented, presents cave bottom altitudes between 215.50 and 209.50 m a.s.l., while the more superficial one, oriented north–south and northwest–southeast, presents cave bottom altitudes between 218.00 and 214.00 m a.s.l.
Also, the cave connected to the vertical sinkhole ‘capovento’, adjacent to the ‘Don Bosco’ nursery school, is of the interlayer type, with a prevalently sub-horizontal development on two overlapping karst levels. The development direction of the deepest level is from SE to NW, with a total height of about L = 1.50 m and cave bottom altitudes between 226.90 and 222.30 m a.s.l. The most superficial karst level consists of two ramifications that start from the sinkhole ‘capovento’ and develop one in the northwest direction and the other in the north north east direction. The cave has a height of about L = 4.00 m in correspondence with the sinkhole ‘capovento’ and barely more than L = 1.00 m in correspondence of the two branches that start from it; moreover, the altitudes of the cave bottom are between 225.20 and 234.90 m a.s.l. In the geomechanical model section, used for the LSR evaluation, the compact and not altered limestone was distinguished from the altered and karstic one, depending on the alteration level and the karst increase (Fig. 7).
Considering that the numerical code used for the LSR analysis is a DEM one , , , , , it was necessary to parameterise the geomechanical model. Therefore, laboratory tests on the six specimens taken from the mechanical borings (conducted at the ‘M. Salvati’ laboratory–DICAR Department, Politecnico di Bari) and geomechanical measurement in situ were conducted. The collected data allowed to classify the rock mass with the Bieniawski RMR’76 indexes  and to estimate the friction angle and the cohesion values. Moreover, the GSI values were obtained (Table 1) through the correlations between the geomechanical classification indexes , :
Geomechanical classification of the altered and the compact calcareous rock mass.
|Rock mass||R1 (compressive strength)||7||49||φb = 35°||49|
|Altered and karstic limestone||R(RQD) 2||6||cb = 0.15 MPa|
|R3 (spacing of discontinuities)||15|
|R4 (joint conditions)||11|
|R5 (water conditions)||10|
|Rock mass||R1 (compressive strength)||7||62||φb = 40°||62|
|Compact and not altered limestone||R(RQD) 2||13||cb = 0.20 MPa|
|R3 (spacing of discontinuities)||15|
|R4 (joint conditions)||17|
|R5 (water conditions)||10|
In addition, the Hoek–Brown curvilinear strength criterion  was implemented using the values of the mechanical parameters obtained from the laboratory tests and the GSI classification indexes. This criterion considers the strong non-linearity of the strength envelope, which is typical of the rock masses, and estimates the representative parameters, even by means of the Mohr–Coulomb linearisation. The physical–mechanical parameters used for the rocks and the discontinuities of the geomechanical model are summarised in Table 2.
Mechanical parameters of rocks and discontinuities (based on laboratory tests) of the geomechanical model.
|VEGETAL SOIL||AND “RED SOILS”|
|ALTERED AND KARSTIC LIMESTONE from laboratory tests||ALTERED ROCK MASS from Hoek-Brown linearized criterion|
|COMPACT AND NOT ALTERED LIMESTONE from laboratory tests||COMPACT ROCK MASS from Hoek-Brown linearized criterion|
|DISCONTINUITY – ALTERED AND KARSTIC LIMESTONE|
|DISCONTINUITY – COMPACT AND NOT ALTERED LIMESTONE|
4 Seismic characterization of the case study area and choice of the seismic input
In order to assess the LSR of the case study, the seismic input consists of two accelerograms, respectively, for the X-component and Z-component of seismic input motion. Therefore, according to the seismic hazard of the case study area (Table 3) and the Technical Standards for Italian Constructions (Norme Tecniche per le Costruzioni Italiane ), the accelerograms were extracted from the European Strong-motion Database   present in the REXEL 3.5 software .
Relevant input data for the accelerograms search with the REXEL 3.5 software.
|Target Spectrum||ag (g)||Geographic ED50||Coordinates||Site class||Topographic category||Nominal Life||Functional Type||Limit State||M||R(km)|
|Lat. N||Long. E|
|Italian Buil- ding Code 2008||0.074||17.0222||40.9185||A||T1||50 years||III||SLV||6-7||0-100|
These accelerograms are natural and spectrum compatible; moreover, the focal mechanism (normal fault) and the bedrock lithology (carbonatic rocks) were compatible with those of the case study area (Table 4).
Main data of the seismic events extracted from the database for the X-acceleration (above) and the Z-acceleration (below).
|Waveform ID||Earthquake ID||Station ID||Earthquake Name||Date||Mw||Fault Mechanism||Epicentral Distance (km)||EC8 Site class|
|Waveform ID||Earthquake ID||Station ID||Earthquake Name||Date||Mw||Fault Mechanism||Epicentral Distance (km)||EC8 Site class|
Subsequently, the accelerograms previously extracted, which are the most spectrum-compatible and also those that have a scale factor closer to the average value (SFmean) of the respective combinations of seven earthquakes, were processed with SeismoSignal software  to obtain the X-velocity and Z-velocity which were used to define the seismic input for the UDEC numerical simulation (Fig. 8). The seismic waves velocities Vs = 2821 m/s and Vp = 4607 m/s in the rock mass were inferred from previous seismic refraction traverse, provided by the technical department of the municipality of Turi (BA).
5 2d distinct elements numerical static and dynamic analysis
To evaluate the LSR of the case study, two discontinuous models with deformable blocks were designed using the UDEC 2D distinct element numerical code. The constitutive model used for rock blocks, discontinuities and faults was the Mohr–Coulomb plasticity one. This model was chosen both because it is one of the most efficient computationally plasticity models and the analysis concerns general rock mechanics issues (e.g. slope stability and underground excavation). Before starting the dynamic analysis, aimed at the LSR evaluation, it was crucial to properly define the boundary conditions and the calculation options:
- ––presence of a ‘graben’ adjacent to the school building and within which the two overlapping underground caves are located;
- ––presence of a tectono-karstic palaeochannel ‘lama’, located above the underground caves and flanked by the school building, filled by vegetal soil and/or landfill with a maximum thickness of 3-m;
- ––loads transmitted by the school building foundation (isolated plinth type) on the substratum;
- ––use of viscous boundaries at the base of the model and on the three sides of the foundation excavation and free field conditions on the sides of the model;
- ––Rayleigh damping equal to 5% of the critical damping, applied to the natural vibration frequency of the system;
- ––seismic waves incidence angle θ = 0°;
- ––seismic waves velocity Vs = 2821 m/s and Vp = 4607 m/s in the rock mass.
The static stress-deformation analysis of the rock mass showed that the model has tensile plasticisation points that tend to cluster on the roof of the upper karst cavity (Fig. 9). These yielding conditions, accompanied by downward displacements of the rock blocks of the karst cavity roof, are even more evident in the results of the dynamic analysis.
6 The local seismic response of the case study model
According to previous studies , , and general considerations described above, the LSR simulation of the case study was finally assessed considering two hypotheses: the absence and presence of the two overlapping karst caves. For the simulation, the model was extended for a length of 25–30 m on both the sides of karst caves, the boundary conditions listed above were applied and, in addition, 18 control points were placed, 9 at the base of the model, and the others 9 (aligned with the first ones) on the ground level (Fig. 10). The amplification/ deamplification (FA) and the most amplified frequencies on the ground level of the model were analysed to estimate the effects of LSR on the X-accelerations component. More specifically, in each control point on the ground level of the model, the amplification/deamplification (FA) has been calculated using two methods (Table 5):
FA, calculated in the absence and presence of karst caves, as ratio of the maximum X-acceleration and ratio of the Fourier’s amplitudes spectrum in three ranges of frequency.
|Output Point ID|
|Without caves||With caves||Without caves||With caves||Without caves||With caves||Without caves||With caves|
b) ratio between the values of the Fourier’s amplitudes spectrum in the points at the top of the model and in the corresponding ones at the base for three ranges of frequencies normally used in previous RSL studies , : 0.2–2.0, 2.0–3.8 and 3.8–5.6 Hz. These frequencies are comparable with the computed natural ones of the existing buildings in the case study area, ranging roughly between 2.21 and 3.10 Hz.
The maximum values of FA, for both the calculation hypotheses (with or without the caves) and for the two calculation methods implemented, are found on the right edge of the caves at no. 8 control point, close to the edge of the school building foundations (Fig. 10).
Furthermore, the FA values obtained with both calculation methods showed that, in general, in the control points at the ground level of the model, there is a deamplification of the seismic motion (Table 5). The only exception is represented by the point no. 8 in the hypothesis without caves, where the FA = 1.672 (ratio between the max X-accelerations) would highlight a local effect of seismic amplification. Furthermore, between the three frequency ranges analysed in the simulations, those that are most amplified are 2.0–3.8 Hz (without caves) and 3.8–5.6 Hz (with caves).
Finally, it should be observed that the UDEC dynamic analysis is conducted with a linearly elastic behaviour 12 of the rocks. The values of the seismic FA on the ground level of the model are, therefore, lower or at the most equal to those that would be obtained with a non-linear analysis, because the degradation of the stiffness and shear modules of the rocks, especially of the vegetal and/ or landfill soil, increases the seismic amplification on the ground level , .
7 Concluding remarks
Preliminarily to the case study LSR, the stress–strain analysis of the rock mass (Fig. 9) showed that the model presents, already in static conditions, but even more in dynamic ones, tensile plasticisation points that tend to cluster on the roof of the upper karst cavity. Moreover, in dynamic conditions, the downward displacements of the rock blocks of the karst cave roof are compatible with a local collapse of the same. The results of the modelling are in perfect agreement with the real occurrence of a sinkhole that, in the recent past, brought to the light the underlying karst cave in the case study area.
The LSR numerical simulation evaluated by the values of FA obtained with two distinct calculation methods shows a generalised deamplification of seismic motion on the ground level of the model, both in the absence and in the presence of hypogean karst caves. The only exception to this trend is represented by point no. 8 in the hypothesis without caves, where an FA = 1.672 (ratio between the max X-accelerations) should highlight a local effect of seismic amplification. The presence of the lower karst cave, filled by ‘red soils’ clay, and the thin vegetal soil that fills the paleochannel ‘lama’ do not seem to influence the model’s LSR.
The generalised deamplification for the horizontal component of the ground level seismic motion is in agreement with what was reported in the previous studies , ,  about the relevant effect of seismic energy absorption operated by the layer discontinuities. Moreover, the effect of seismic deamplification at the top and in axis with the underground caves, already highlighted in the literature for cavities with diameter D > 5.6 m in homogeneous and isotropic substrate models , is confirmed by the minimum FA values reached at the point no. 5, with both calculation methods, even for a cave with D < 5.6 m.
Thanks to the Municipality of Turi for having made available the previous geophysical data and for the authorisation to publish the results of this research.
Pyrak-Nolte L. J. Myer L. R. Cook Neville G. W. - Transmission of Seismic Waves Across Single Natural Fractures. Journal of Geophysical Research vol. 95 n° B6 8617-8638 June 10 1990.
Ghosh A. Hsiung S. - Effects of tilted and faulted strata on seismic ground motion. U.S. Nuclear Regulatory Commission Contract NRC-02-07-006 Center for Nuclear Waste Regulatory Analyses San Antonio Texas 2011.
Rodriguez-Castellanos A. Sánchez-Sesma F. J. Luzón F. and Martin R. - Multiple scattering of elastic waves by subsurface fractures and cavities. in Bulletin of the Seismological Society of America vol. 96 n. 4A 1359-1374 2006.
Smerzini C. Avilés J. Sánchez‐Sesma F. J. and Paolucci R. - Analytical solutions for the seismic response of underground structures under SH wave propagation. AIP Conference Proceedings 1020 674 https://doi.org/10.1063/1.29639002008.
Smerzini C. Avilés J. Paolucci R. and Sánchez-Sesma F. J. - Effect of underground cavities on surface earthquake ground motion under SH wave propagation. in Earthquake Engineering and Structural Dynamics n.38 1441-1460 2009.
Verrucci L Lanzo G Pagliaroli A Sano` T. - Effects of cavities on seismic ground response. In: Procceding. Second international conference on performance-based design in earthquake geotechnical engineering May 28-30 2012 Taormina (Italy) CDROM paper #1.15 2012.
Sanò T. - Risposta sismica locale in aree con presenza di cavità. in Geologia dell’Ambiente - Supplemento al n.2/2013 55-61 2013.
Itasca Consulting Group Inc. “UDEC Version 4.0 User’s Guide.” Minneapolis Minnesota: Itasca Consulting Group Inc. 2011.
Scotto di Santolo A. de Silva F. Calcaterra D. Silvestri F. - Site effects due to the presence of cavity near the cliffs. 6th International Conference on Earthquake Geotechnical Engineering 1-4 November 2015 Christchurch New Zealand 2015.
Evangelista L. Landolfi L. d’Onofrio A. Silvestri F. - The influence of the 3D morphology and cavity network on the seismic response of Castelnuovo hill to the 2009 Abruzzo earthquake. Bulletin of Earthquake Engineering 14 33633387 DOI: 10.1007/s10518-016-0011-8 2016.
Piro A. de Silva F. Parisi F. Silvestri F. Scotto di Santolo A. - Interazione sismica tra terreno cavità e struttura: un caso di studio nel centro storico di Sant’Agata de’ Goti. XXVI Convegno Nazionale di Geotecnica Rome Italy June 2017.
Fu X. Sheng Q. Zhang Y. Chen J. and Leng X. - Extension of the discontinuous deformation analysis method to simulate seismic response of a large rock cavern complex. Int. J. Geomech. 17(5):E4016008 DOI: 10.1061/(ASCE)GM.1943-5622.0000712 (2016b).
Hsiung S. M. Shi G. H. - Simulation of earthquake effects on underground excavations using discontinuous deformation analysis (DDA) [A]. In: Rock Mechanics for Industry [C] 1413-1420 1999.
Ahola M.P. Chowdhury A.H. - Effect of seismic load on a proposed nuclear waste repository at Yucca Mountain. Inter. Jour. of Rock Mech. and Min. Science 34 (3-4) ISSN 0148-9062 1997.
Barla G. Monacis G. Perino A. Hatzor Y. H. - Distinct Element Modelling in Static and Dynamic Conditions with Application to an Underground Archaeological Site. Rock Mechanics and Rock Engineering 43 (6) 877-890 2010.
Ciaranfi N. Pieri P. Ricchetti G. - Carta Geologica delle Murge e del Salento. Mem. Soc. Geol. It. 42 1988.
Cundall P.A. Hart R. D. - Development of generalized 2-D and 3-D distinct element programs for modeling jointed rock. Itasca Consulting Group U.S. Army Corps of Engineers Misc. Paper SL-85-1 1985.
Marcato G. Silvano S. Zabuski L. - Modellazione di ammassi rocciosi instabili con il metodo degli elementi distinti. Giornale di Geologia Applicata 2 87-92 DOI: 10.1474/GGA.2005-02.0-12.0038 2005.
Bieniawski Z. T. - Engineering classification of jointed rock masses. Trans S. Afr. Inst. Civ. Engrs 15 335-344 1973.
Bruno G. - Caratterizzazione geomeccanica per la progettazione ingegneristica. Flaccovio Dario Editore Palermo ISBN 978-88-579-0150-3 2012.
Hoek E. Kaiser P. K. Bawden W. F. - Support of underground excavations in hard rock. A.A. Balkema Ed. Rotterdam 1998.
Hoek E. & Brown E. T. - Empirical strength criterion for rock masses. J. Geotech. Engng Div. ASCE 106 (GT9) 1013-1035 1980.
Ministero Infrastrutture e Trasporti DM 17 Gennaio 2018 - Aggiornamento delle «Norme tecniche per le costruzioni». Gazzetta Ufficiale della Repubblica Italiana 42 2018.
Ambraseys N. Smit P. Sigbjornsson R. Suhadolc P. and Margaris B. - Internet-Site for European Strong-Motion Data European Commission Research-Directorate General Environment and Climate Programme 2002.
Ambraseys N. N. Douglas J. Rinaldis D. Berge-Thierry C. Suhadolc P. Costa G. Sigbjornsson R. Smit P. - Dissemination of European strong-motion data Vol. 2 CD-ROM Collection Engineering and Physical Sciences Research Council UK 2004.
Iervolino I. Galasso C. Cosenza E. - REXEL: computer aided record selection for code-based seismic structural analysis. in Bulletin of Earthquake Engineering n.8 339-362 2010.
SeismoSignal_2018 rel.1 - Earthquake Engineering Software Solution. SEISMOSOFT Ltd. Pavia 2018.
Barbero M. Barla G. Demarie G. V. - Applicazione del Metodo degli Elementi Distinti alla dinamica di mezzi discontinui. Rivista Italiana di Geotecnica 3 2004.