Role of Geophysical Testing in Geotechnical Site Characterization

. The lecture attempts to highlight the insights late Victor De Mello provided on some key areas. Considering the increasing role of the geophysical methods in the geotechnical site characterization, the writer focuses on the use of in-hole geophysical methods when assessing, both in field and in laboratory, the parameters depicting the soil state and its stiffness at small strain. With this aim the writer draws the attention to seismic transversal (S) and longitudinal (P) body waves generated both in field, during in-hole tests, and in laboratory using piezocristals. Within this framework the following issues are discussed:


Introduction
Considering the growing importance of the geophysical methods [Stokoe (2011)] for the geotechnical site characterization, this paper focuses on the in-hole techniques, such as cross-hole (CH) and down-hole (DH) tests which, if properly instrumented and performed, can provide reliable values for compression (V p ) and shear (V s ) waves velocity.
When it is only requested the knowledge of V s , reference will also be made to seismic cone penetration tests (S-CPTU) and to seismic Marchetti's dilatometer (S-DMT), equipped to provide a reliable measure of V s in DH-mode. The main features of CH and DH tests are shown in Fig. 1, while Fig. 2 highlights the seismic waves that can be propagated in situ, during CH and DH tests, and in laboratory by means of bender elements (BE).
The generated seismic waves are classified according to the propagation direction (first capital letter) and to the polarization plane (second capital letter). Figure 2 shows also the soil stiffness at very small shear strain (10 -6 £ g £ 10 -5 ), see Rahtje et al. (2004) and Cox (2006), that can be computed from the seismic waves velocity, being: G 0 = shear modulus at very small strain, M 0 = constrained modulus at very small strain and r t = bulk soil mass density.
The following aspects, relative to the use of in-hole measured seismic body wave velocities in geotechnical design, are discussed: 1. Stiffness at very small strain: G 0 = f (V s ) and M 0 = f(V p ) (Applicable to V p propagated through dry soils or at least having S r < 90%). 2. Distinction between fully saturated and near to saturated soils ® f (V p ).
3. Assessment of undisturbed samples quality ® f [V s (Field) vs. V s (Lab.)]. 4. Evaluation of in situ void ratio e 0 by means of Foti et al. (2002) approach f (V p and V s ). 5. Susceptibility of coarse grained saturated soil to cyclic liquefaction ® f (V s ). The above topics are only loosely interconnected, thus each subject matter is detailed in a specific section with dedicated closing remarks.

Stiffness at Very Small Strain and its Anisotropy
The use of seismic waves velocity allows to evaluate, in situ and in laboratory, the shear modulus G 0 = r t V s 2 and the constrained modulus M 0 = r t V p 2 . G 0 is representative of the very initial portion of the soil stress-strain curve (Fig. 3), which, upon loading is linear and in unloading state exhibits a recoverable strain, including a minor amount of the delayed viscous component.
The g t l for non rocky-like materials usually ranges between 10 -5 and 10 -4 , see Fig. 4a which reports also the volumetric threshold shear strain (g t v ), see [Dobry et al. (1982) and Vucetic (1994)]. The g t v corresponds to the point where a soil element, subject to constant mean effective stress (p'), under the action of shear stress increase, during drained loading starts exhibiting plastic strain, whereas, under undrained loading a pore pressure excess is generated.
It can be therefore assumed that G 0 represents the initial tangent shear stiffness of a given geomaterial applicable to both static and dynamic problems, with possibly minor differences due to the strain rate effects involved in two different loadings modes, see Fig. 4b.
This figure, after Menq (2003), reports an example of the normalized shear modulus [G/G 0 = f.(g ³ g t l )] degradation curve as function of the shear strain g, pointing out the difference between monotonic and cyclic loadings. As such, G 0 plays a role in the numerical analyses involving complex constitutive soil models, allowing separating elastic from total strains.
Another important function of G 0 (F) measured in the field is to allow for the correction of the laboratory determined modulus degradation curve G(g) for disturbance effect. The procedure, see Fig. 5, is based on the available field and laboratory extensive data base, proposed by Ishihara (1996) and makes reference to the following empirical formula: being G 0 (F) = shear modulus at very small strain (g £ g t l ) from in situ seismic tests, G 0 (L) = shear modulus at very small strain (g £ g t l ) measured in laboratory, G(L) = shear modulus measured in laboratory at the given value of g ³ g t l , G(F) = corrected field value corresponding to the same value of g likewise G(L) and C r = correction factor depending on the sample quality and type. In his work, Ishihara (1996) provides C r values as function of the strain level for different kinds of sampling techniques including reconstituted specimens.
In a given soil G 0 and M 0 are controlled by the effective ambient stresses and by the current value of void ratio, reflecting the state of the material.
With reference to the seismic waves propagation and to their computed moduli, the following empirical relations, experimentally validated, [Roesler (1979), Lewis (1990), Lee & Stokoe (1986), Weston (1996)], allow exploring how the current soil state affects V s , hence G 0 and V p , thus M 0 respectively: Soils and Rocks, São Paulo, 35 (2) ]ÖF(e) (2.1) ]ÖF(e) (3.1) being C s , C p , C G , C M = experimental material constant, na, nb = experimental stress exponent, F(e) = experimental void ratio function, p a = reference stress = 98.1 kPa, s' a = effective stress in the direction of wave propagation and s' b = effective stress on polarization plane. Note: In case of V s , na ¹ nb, while for V p , na = nb.
In the everyday practice, G 0 and M 0 are considered as isotropic elastic body stiffness making simpler also to assess the Young E 0 and bulk B 0 modulus assuming the value of Poisson coefficient of the soil skeleton n' 0 . With this respect it is worth mentioning that, as confirmed by laboratory tests, n' 0 at strain level not exceeding the linear threshold, ranges between 0.15 and 0.25, typically exhibiting a trend to decrease with increasing the confining stresses [Hoque (1996), Weston (1996)].
However, the lesson learnt from the propagation of seismic waves in situ and in laboratory [Lee & Stokoe (1986), Lee (1993), Bellotti et al. (1996), Fioravante (2000, Kuwano & Jardine (2002), Giretti et al. (2012)] has demonstrated that in the presence of the level-ground the soil behavior, at very small strain (g £ g t l ), can be better approximated by the cross-anisotropic (= transversally isotropic) linear elastic half-space, with the vertical axis (z) of symmetry and the horizontal plane (xy) of isotropy [Love (1927)]. The relationship, broadly describing the stressstrain behavior of such body, requires determining five independent elastic material constants, see the stiffness matrix in Fig. 6.
For the plane body waves generated on the vertical (xz) or horizontal (yx) planes, White (1965) derived three equations expressing the velocities in terms of five independent material constants of the cross-anisotropic halfspace, see Stokoe et al. (1991) and Lee (1993).
The difference in velocities of V p and V s propagating on zx and xy or yx planes, coinciding with the principal stresses directions respectively, reflect the material initial anisotropy.
Dealing with the initial elastic anisotropy (g £ g t l ) of the non rocky-like geomaterials, two components of different phenomenological nature should be distinguished: • Fabric or structural anisotropy exhibited by the soil under isotropic state of stress. • Stress induced anisotropy disclosed even by a soil with isotropic fabric when subject to anisotropic stress state.
Referring to the level ground, i.e. geostatic stress state, the stress induced anisotropy is governed by the magnitude of earth pressure coefficient at rest K 0 , hence by the soil depositional and post-depositional history.
The initial anisotropy can be quantified in the field measuring, during CH tests, V s (HH) on the isotropy plane and V s (VH) along the symmetry axis plane.
The same measurements have been carried out at the copper mine tailings at Zelazny Most (Poland) site yielding initial anisotropy values in terms of Vs(HH)/V s (VH) ratio ranging between 0.92 and 1.12.
Unfortunately, the five independent constants of the cross-anisotropic geomaterials cannot be determined in situ. Four of them: G 0 (HH), G 0 (VH), M 0 (H) and M 0 (V) can be assessed from the corresponding shear and compression waves measurable in CH tests.
However, for M 0 values, to ensure that the V p propagation is entirely controlled by the soil skeleton compressibility, such approach is limited to materials that are either dry or with a satisfactorily low degree of saturation 1 . In these circumstances, the basic studies for crossanisotropic materials have been mostly carried out in laboratory, testing mainly on reconstituted soil specimens. Three different methodologies have been employed so far: • Using exclusively the static stress-strain laboratory probing [Hoque (1996), Hoque & Tatsuoka (1998)], however requiring a simplified assumption to assess the fifth independent cross-anisotropy body parameter. • As above, combining the results of static probing, with the dynamic measurements of seismic waves velocity using bender elements. This methodology has allowed Kuwano & Jardine (2002) to determine all the five independent material constants. • Using solely seismic waves generated in large calibration chambers [Lee (1985), Lee & Stokoe (1986), Stokoe et al. (1991), Bellotti et al. (1996), Giretti et al. (2012)], as well as in triaxial apparatuses, see Fioravante (2000), all the above five independent material parameters can be determined. In these tests, usually carried out under biaxial confinement, it can be determined the fifth independent material parameter, even though generating, in the anisotropy plane, the V p and V s waves at the angle Q with respect to the axis of symmetry (z). Lee (1985) and Lee & Stokoe (1986) have pointed out that the propagation of planar waves in zx and zy planes, and not along the principal stress directions, uncouples the velocity surface (= the front of the wave normal) from the overlapping wave surface (energy ray path). On the other hand, as observed by Stokoe et al. 1991 andLee 1993, for dry silica sands the resulting discrepancy is sufficiently small and leads to minor corrections of the measured ray velocity to obtain the phase velocity.
In the following are given some examples of seismic tests carried out in a large calibration chamber housing specimen 1.2 m in diameter and 1.5 m in height and instrumented with miniature geophones, see Fig. 7. The adopted geophones arrangement allows the generation, under biaxial confinement, of P and S waves in three orthogonal principal stress directions xyz in Fig. 7 as well as of the oblique waves P(Q), S(Q) inclined at an angle of 45°as regard the axis of symmetry (z), with the oblique shear waves S(45°) polarized in a vertical plane.
Details of the tests experimental setup can be found in Lo Presti & O'Neill (1991) and Bellotti et al. (1996). In the second work, it is also illustrated the trial and error computation procedure used to estimate, with the aid of P(45°V) and S(45°V), the fifth independent parameter C 13 of the stiffness matrix of Fig. 6.
Hereafter are summarized some examples of seismic tests results performed in CC on dry pluvially deposited TS-Ticino river (Bellotti et al. 1996) and KS-Calcareous Kenya beach  sands; the same test sands were employed by Fioravante (2000) to investigate, in a triaxial apparatus, the elastic anisotropy. The test sands characteristic are depicted in Table 1. Tables 2 and 3 show the moduli ratio G 0 (HH)/G 0 (VH) and M 0 (H)/M 0 (V) as obtained from CC seismic tests in dry TS and KS.
More details respectively for TS and KS, can be found in the works by Bellotti et al. (1996) and Giretti et al. (2012).
To sum up, the seismic waves velocity measurement, in situ and in laboratory, plays a central role in the evaluation of the soil stiffness at very small strain and of its anisotropy.
The main issues significant to the engineering applications are: • G 0 corresponding to the initial tangent shear modulus for both static and dynamic loading. • Knowing G 0 , elastic and plastic strains can be separated.  • G 0 inferred from V s measured in the field offers the possibility to correct the laboratory G vs. g degradation curves accounting for disturbance effects. • The generation of S(HH) and S(VH) waves in field and in laboratory consent to estimate the material initial anisotropy. • Although so far limited to laboratory testing on reconstituted specimens, the generation of seismic waves, alone or in combination with static probing, carried out in the triaxial apparatus consent to study the basic behavior of the elastic cross-anisotropic geomaterials.

Fully Saturated vs. Near-To-Saturated Soils
In the last two decades many laboratory and field experiments have proved that the compression wave propagation is an extremely sensitive tool to distinguish fully from near to saturated soils: [Ishihara et al. (1998) Figures 8 and 9 show the results of laboratory experiments aimed at exploring the dependence of V p on the saturation degree. The results confirm the extreme sensitivity of the P-wave velocity to even small deviations from the full saturation, occurring when V p exceeds 1450 to 1500 m/s, and correspond to the compression wave in water velocity.       from high tides. This figure shows V p (H) as well as V s (HV) and V s (HH) resulting from CH tests, together with the relevant lagoon soil profile. The V p profile highlights the presence, below the sea bottom, of an unsaturated soil zone, » 12 m thick, due to marsh gas.
The capacity of V p to detect the presence in the subsoil of near to saturated spots, plays a crucial role in evaluating the susceptibility of coarse grained soils to cyclic and monotonic liquefaction during undrained loading [Ishihara et al. (1998); Grozic et al. (1999Grozic et al. ( , 2000; Ishihara et al. (2004), Lee et al. (2005)]. Figure 11 displays the cyclic resistance ratio (CRR) obtained from undrained triaxial tests of the near to saturated Toyoura sand, normalized with respect to the CRR of the same sand at full saturation, see Ishihara et al. (1998) and Tsukamoto et al. (2001).
The V p capability to map the saturation surface position in the subsoil, finds many important applications in the engineered constructions experiencing complex hydraulic regime, variable in time and space.
A typical example is the second world largest copper tailings storage disposal, whose peculiar features can be inferred from Fig. 12. At this Polish site, in Zelazny Most, since 1993 CH tests are being carried out periodically on the pond beaches, to map the position of the saturation line in the tailings, [Jamiolkowski et al. (2010)]. Figure 13 shows the location of 9 CH tests performed during the 2011 campaign.
Soils and Rocks, São Paulo, 35 (2)  Figures 14 through 16 display the depth position of the saturation line in the tailings, as determined based on the V p measured in CH tests at variable distance from dam crest for cross-sections in correspondence of the West, North and East dams respectively. As the figures show, the measured V p value, allows recognizing the presence of saturated tailings at a depth below which V p remains greater than 1450 to 1500 m/s. Moreover, the profiles of V p vs. depth show also the presence, in the tailings, of the perched water horizons. See Figs. 15 and 16 where the perched water horizons are labeled with the symbol P H .
From the above one can deduce that: • The measured V p is an extremely sensitive tool to distinguish in situ and in laboratory fully (S r @ 100%) from near to saturated (90% £ S r < 100%) state; see Fig. 9 after Tsukamoto et al. (2001)  • This V p feature represents a simple and reliable tool to map the distribution of fully and nearly to saturated soil deposits in situ.
• In the last decade there have been many attempts to correlate Skempton's (1954) pore pressure coefficient B measured in laboratory against the velocity of the compression wave, [Kokusho (2000), Tsukamoto et al.

Quality Assessment of Undisturbed Samples
In case of homogeneous low permeability clays, quality undisturbed samples can be evaluated in laboratory measuring the sample suction p s immediately after its retrieval from the ground, [Skempton (1961), Chandler et al. (2011)]. This approach is quite complex, see Chandler et al. (2011) and time consuming thus not routinely employed. Moreover, it is restricted to homogeneous fine grained soils able to preserve high suction after zeroing of the total in situ stress as results of sample retrieval.
This prompts to develop some easier semi-empirical criteria to assess undisturbed samples quality.    With this respect, a widely used criterion has been proposed by Lunne et al. (1997; for fine grained soils in terms of De/e 0 ratio, being: • De = reduction of the void ratio during one dimensional recompression of undisturbed specimen to in situ vertical effective stress s' v0 existing at a depth from which the sample were retrieved. • e 0 = in situ void ratio. The other criterion, applicable to both coarse and fine grained soils [Sasitharan et al. (1994); Landon et al. (2007) De Groot et al. (2011Fioravante et al. 2012)] is based on the comparison of normalized shear wave velocity V s1 (L) measured on laboratory specimens with that measured in the field V s1 (F) by means of one of the methods recalled in Fig. 1.
The values of V s1 (F) and V s1 (L) are computed by means of the formula 4, a somehow simplified version than Eq. 2.1, considering that the separate values of exponents na and nb are difficult to measure and therefore rarely available: where V s (F) = shear wave velocity measured in the field at the same depth the sample has been retrieved, V s (L) = shear wave velocity measured in laboratory on the specimen reconsolidated to the best estimate of the in situ geostatic stresses at the same depth the sample has been retrieved, p a = reference stress = 98.1 kPa, s' a = effective stress in the wave propagation direction, s' b = effective stress on the plane of the wave polarization, ns = stress exponent na+nb, pertinent to V s1 (F), s' v0 = effective vertical stress at the sampling depth, s' h0 = effective horizontal stress at the sampling depth, ns = stress exponent na+nb, pertinent to V s1 (L). The closer V s1 (L)/V s1 (F) ratio is to unity, the better the quality of undisturbed sample.
This ratio can also be used to estimate the mechanical characteristics of the specimens reconstituted in laboratory that the soil, in undisturbed state, should have in situ.
Overall, exponents np and ns, the former pertinent to V p , vary within a relatively narrow range (0.22 to 0.25) in case of fine grained soils and uniform sands but tend to increase in coarse gravelly sand and sandy gravel as the uniformity C u coefficient increases [Weston (1996)], see Fig. 17. This figure adapted after the quoted work by Weston, with the support of some writer's data, gives the stress exponents n S and n G from V S and G 0 respectively determined experimentally in laboratory tests on the reconstituted specimens.
The quality evaluation of three examples based on V S1 (L)/V S1 (F) ratio is hereafter presented.
The first example deals with undisturbed samples of sandy gravel 600 mm in height (H s ) and 300 mm in diameter (D s ) retrieved on the Sicilian shore of Messina Strait by means of the freezing technique ], see Fig. 18.  Figure 19 shows the comparison between V S1 (F) measured during CH test and V S1 (L) obtained from bender element (BE) tests.
In the examined case, the characteristics of the generated shear waves during BE tests were as follows: The second example refers to undisturbed samples of fine to medium sand retrieved by means of freezing, see Fig. 20 at the Tyrrhenian shore close to Gioia Tauro, in Southern Italy. Table 4 reports the values of V s1 (L)/V s1 (F) ratio as obtained for the tested undisturbed samples. Again in this case, V s (L) has been measured by means of BE tests while V s1 (F) was obtained from CH test whose results 2 . The resulting values of V s1 -ratio, probably except for the one from a 24.5 m depth, confirm the tested samples high quality.
The third example deals with the undisturbed sampling of very uniform stiff to hard OC clay, see Fig. 21, retrieved at the Porto Empedocle site on the Eastern Sicilian Coast. In this case, besides using the available V s1 ratio, the quality of undisturbed samples has been evaluated from suction measured by means of Ridley & Burland (1993) transducer, carried out soon after the samples retrieval, see Chandler et al. (2010) and also referring to the Lunne et al. (1997Lunne et al. ( , 2007 criterion based on the ratio of De/e 0 measured in oedometer tests. Table 5 shows the comparison for a number of Porto Empedocle clay samples between V s1 (L)/V s1 (F) and De/e 0 ratios together with the ratio of p s /p' 0 , being: p s = measured suction in the sample, p' 0 = the best estimate for mean in situ effective stress at the sampling depth. In the case in hand, all the three used approaches indicate the excellent quality of tested samples.
The information collected by De Groot et (2011) supports the idea that both ratios, V s1 (L)/V s1 (F) and De/e 0 , as shown in Fig. 22 are useful and complementary tools when evaluating undisturbed samples quality.
Basically, based on the above the following comments apply: • V s1 (F) reflects in situ soil state, fabric, aging and particles bonding. • V s1 (L) has to be assessed on specimens reconsolidated to the best estimate of in situ geostatic stresses.
Soils and Rocks, São Paulo, 35 (2)  • The main uncertainty in determining V s1 (L) is linked to an appropriate selection of the laboratory horizontal consolidation stress. • The closer V s1 (L)/V s1 (F) is to one, the better the quality of the specimen tested in laboratory. • Unlike other methods for the assessment of undisturbed samples quality (e.g. suction measurements or the comparison of the void ratio reduction after the specimen 1-D 12 Soils and Rocks, São Paulo, 35 (2)

Evaluation of In Situ Void Ratio
The geomaterials in situ porosity n 0 and void ration e 0 are important state parameters, crucial for a thorough site characterization when working out many geotechnical boundary value problems.
The assessment of n 0 or e 0 , while routinely determined via laboratory tests on undisturbed samples of fine grained soils, results by far more complex and expensive when dealing with coarse grained soils in which undisturbed sampling [(Yoshimi et al. (1978), Hofmann (1997, Yoshimi (2000), Huang et al. (2008)] is still far to become a common practice.
To overcome this restraint, several empirical correlations have been proposed based on various penetration tests [Schmertmann (1978), Skempton (1986), Cubrinovski & Ishihara (1999), Jamiolkowski et al. (2001)] and in situ relative density (D R ), which, in combination with laboratory determined maximum (e max ) and minimum (e min ) void ratio allow estimating, in first approximation, the e 0 .
In this circumstance, the researchers and practitioners attention was drawn by Foti et al. (2002) work who, within the frame of Biot (1956) linear theory of poroelasticity, has developed a procedure to compute in situ e 0 or n 0 via inversion of the seismic waves V p and V s measured in the in-hole geophysical tests.
The formula by Foti et al. (2002), applicable to fully saturated soils only is reported here below: where r s = soil particles mass density, r f = pore fluid mass density, B f = bulk modulus of pore fluid, n s = Poisson ratio of soil skeleton. Since its publication this formula has been calibrated against laboratory tests results carried out on good quality undisturbed samples of fine grained geomaterials [Foti & Lancellotta (2004), Arroyo et al. (2007), (Jamiolkowski et al. (2009)], yielding, overall, satisfactory results.
In the following are compared, and when appropriate commented, three examples of void ratio e 0 computed from seismic waves velocity measured in CH tests and those obtained in laboratory on high quality undisturbed samples.
The first examples, see Fig. 23, compares n 0 values measured in laboratory on high quality undisturbed samples of soft lightly OC Pisa clay with those computed from V p and V s .
The second example in Fig. 24, compares the e 0 measured in laboratory on the undisturbed samples of sandy gravel retrieved by means of freezing, at Messina Strait and those computed from the V p and V s measured in the CH test located nearby the in-hole from which the frozen samples have been retrieved. The e 0 computed values on average result to be 10 to 15 percent lower than those determined in laboratory . The reasons for this difference can be attributed to a combination of the following factors: uncertainties involved in the accuracy of measured  V p and V s ; the large disparity between the volume of the undisturbed specimen tested in laboratory and the volume of soils involved in waves propagation during CH testing associated with the spatial variability of the sandy gravel deposit in question.
The third examples in Fig. 25, displays the comparison between e 0 measured in laboratory on undisturbed frozen samples of fine to medium sand retrieved at Gioa Tauro site, with those computed from the V p and V s measured in the CH test located in the vicinity of the sampling in-hole. In this case, the agreement between e 0 values measured and computed is satisfactory.
However, as to the reliability of the in situ void ratio, as computed from V p and V s measured in the state of the art CH tests, not all the experimental evidences, collected so far by the writer, have yielded satisfactory comparisons with the laboratory determined e 0 . Figure   A few similar examples have raised the issue of the accuracy and reliability of in situ void ratio computed from V p and V s . This subject has been addressed by Foti (2003) who has investigated the error propagation of the measured seismic waves velocities in the porosity computed by means of Foti et al. (2002) formula.
As it can be expected, dealing with an inverse problem, the reliability of the computed e 0 or n 0 is very sensitive to the accuracy of the measured key input parameters, V p and, to a less extent, V s .  non rock like geomaterials, the error affects the computed porosity. It can be observed that within the range of the considered V p and V s , the error on the measured seismic wave velocity amplifies, by three times that of the computed porosity. Moreover, Lai & Crempien (2012), investigating the stability of the inversion procedure to compute the porosity after the formula by Foti et al. (2002), have pointed out that there are combinations of V p , V s and pair with the soil skeleton Poisson ratio n' 0 can be solved only in terms of complex numbers.
However, within the range of CH tests data base covered by the Author (100 £ V s = 550 m/s; 1500 £ V p = 3500 m/s) in combination with 0.15 £ n' 0 £ 0.25, the use of Foti et al. (2002) formula has, so far, yielded a solution in terms of real numbers. This holds also for the data reported in Fig. 26, where the Foti's formula, although well posed, has yielded results conflicting with the comprehensive and reliable set of e 0 values determined in laboratory [Chandler et al. (2011)].
The evidence that the error on measured seismic waves velocity for the range of V p and V s considered in Fig. 27, amplifies by three times the error on the computed n 0 , has triggered the attempt to explore the intrinsic variability of V p and V s measured during 9 state-of-the-art CH tests recently carried out at the Zelazny Most site copper tailings, mentioned in Section 3 of this paper.
The following testing program has therefore been set up: • In each CHT, at 1 m intervals, the seismic waves (V p and V s ) velocity measurements have been repeated 10 times and the obtained values stored. • In each in-hole a survey of the deviation from the verticality and of its azimuth has been carried in both in down-hole and up-hole modes repeating all the measurements three times at depth intervals of 3 m. The bulk of the collected data will be used for the statistical and probabilistic evaluation of how the combination of the two independent variables, time and distance, affect the accuracy of measured V p and V s in the high quality CH tests.
The following preliminary information arising from the above tests can, currently, be anticipated:  ery 1 m, computed from the data gathered after the tenfold replications of the waves propagation.
• Figure 30 exemplifies how the variables uncertainties, travel time and travel distance, individually considered, affect the standard deviation and covariance of the measured V p . Figure 30 highlights the important evidence that, at least in the examined case, the uncertainty linked to the variable travel distance has a more significant impact than the travel time on the measured seismic waves velocity in CH tests reliability.
Thanks to its solid theoretical background, the formula by Foti et al. (2002) allows assessing e 0 and n 0 with the consistency most demanding engineering applications require, remarking that the hardware and software employed in CH and DH tests will be improved.
This work by Foti et al. (2002), offers a valid opportunity to estimate the porosity and the void ratio in situ of fully saturated soils from seismic body waves velocity measured in the field. However, when using this formula, which is yet to be validated, the following points should be considered: Soils and Rocks, São Paulo, 35 (2)  • Dealing with the solution of an inverse problem, the computed value of porosity or void ratio is significantly affected by the accuracy and reliability of the measured seismic waves velocity. The above is especially relevant as regard the compression wave [Foti (2003)]. • However, the above issue, crucial when dealing with liquefaction and flow failure problems, becomes less significant in other engineering applications for which Foti et al. (2002) procedure, represents a step forward compared with the empirical correlations reliability between D R and penetration tests results, used in common practice. • A properly arranged and interpreted CH test is the most suitable mean to obtain independent, accurate V p and V s measurements to be used as input in the Foti et al. (2002) formula. • As to Poisson coefficient n 0 to be adopted when computing the porosity or the void ratio from V p and V s , it should be considered that the strains associated with the propagation of seismic waves is of the order of 10 -6 at the best up 10 -5 . At this strain level, the results of the large data base collected from the drained triaxial and plain strain tests with internal strains measurement, suggest values of n 0 in the range between 0.15 and 0.25. • The porosity and the void ratio computed using Foti et al. (2002) procedure, can be further enhanced if the uncertainties involved in assessing the picking arrival time and travel distance of V p and V s are accounted for.

Susceptibility of Coarse Grained Soils to Liquefaction
Since the pioneering work by Andrus & Stokoe (2000), the empirical approach to assess the susceptibility of sandy soils to cyclic liquefaction, based on the V s measured in field, has been used in parallel with more conventional methods based on penetration tests results (SPT, CPTU, DMT). Figure 31 shows the correlation of V s1 vs. the cyclic stress ratio (CSR) valid for an earthquake of M W = 7.5 magnitude based on the analysis of the collected case records at locations where the cyclic liquefaction has been observed.
A comprehensive discussion and enhancement of the V s procedure to estimate to what extent the coarse grained soil deposit is prone to liquefaction can be found in the book by Idriss & Boulanger (2008), who, in their discussion, raise the issue, already pointed out by Liu & Mitchell (2006), that V s exhibits a lower sensitivity to variation of D R in situ if compared to penetration tests.
The writer, referring to a large data base of more than 650 CPT DMT and seismic tests carried out in CC's on a variety of pluvially deposited dry sands, has attempted to explore the V s1 response to D R changes as compared to those 18 Soils and Rocks, São Paulo, 35 (2): xxx-xxx, May-August, 2012. Jamiolkowski Figure 31 -V s -based liquefaction susceptibility, Andrus & Stokoe (2000).  [Fioravante (2001)] and for the siliceous Ticino river sand [Bellotti et al. (1996), Jamiolkowski et al. (2001)] are shown in Tables 6 and 7.
Comparing the results reported in Table 6 with those in Table 7 it can be confirmed the minor sensitivity of V s1 to D R changes with respect to those of q c and K D . It is worthy to recall the readers' attention, that this difference is even more pronounced if the different range of D R considered in the compilation of Tables 6 and 7 is accounted for. The brief mention to V s used to assess the susceptibility of sandy soils to cyclic liquefaction allows the following comments: The CC tests results on two dry sands confirm the lower capability of shear waves to respond to D R changes if compared to the CPT-q c and the DMT-K D . This happens despite V s , similarly to q c and K D , is function of in situ void ratio and effective stresses. Moreover, differently from all the penetration tests, since V s measurements are less invasive than penetration tests, are more prone to be affected by some depositional and post-depositional phenomena as aging, cementation and cyclic pre-straining.
In the light of the above, the use of V s should continue to evaluate the liquefaction potential, although subject to further laboratory and field validations. The current state of such method development offers the advantage of an easy application in gravelly soils where the feasibility and reliability of the approaches based on penetration tests, in many circumstances, appear questionable.