J. Phys.

Earth,36,255-266,1988

A SPECTRALAPPROACH TO MOHO DEPTHS ESTIMATION FROM GRAVITY MEASUREMENTS IN EPIRUS (NW GREECE)
Gerasimos-Akis TsersNrrs,l' * John DRerolouros,2 and Klisthenis DrMnnlq.ols2 | furthquake Research Institute, The Universityof Tokyo, Bunkyo-ku, Tokyo I 13, Japan ? Department of Geophysics, Athens (lniversity, Panepistimiopolis Athens157 84, Greece (Received January7, 1988;Revised November16, 1988) We compute crustalthickness in Epirus(NE Greece), from spectral analysis of the Bouguergravity field. Moho depths wereestablished from the slopeof the log-power radialspectrum at the lowerendof the wavenumber bandand werefound to vary from 3l to 38km. This is in agreement with the results obtained by otherinvestigators employing seismological methodsof analysis.

lntroduction

A two-dimensional set of data representing a potential field such as the gravity field, can be thought of as a superposition of fields due to a subsurfacedistribution of sources.Consequently, linear methods such as two-dimensional filtering should be of use in separating the various components according to their wavenumber spectrum. Spectral methods of gravity data processingare being increasinglyemployed in recent years, owing to their elegancein handling a large amount of data and their built-in mechanism for sigaal-to-noise ratio enhancement (BuerrecHenyy.l and LEu. 1975). By Fourier transformation, the gravity field data can be representedby twodimensional Fourier seriesconsisting of various frequencieswhich characterizethe anomalies.Thb amplitude and phase relationship among thesefrequencieshas been used extensivelyby many workers for the interpretation of gravity data, particulary in the case of downward continuation and source depth estimation (Srncron and GReNr,1970; Tnsnrr et al.,l97l; GnrnN, 1972H ; IHN eral.,1976;plrtet at.,1979:

Nscr et al.,1983;Bossand SnNcuere, 1984). Representing the Moho by a variable interface,h(x), and considering a
x Present address: AthensUniversity,Panepistimiopolis Athens 15784, Greece. 255

256

G.-A. Tsrt-nNus,J. Duropour.os, and K. Dtutrnllols

constant density contrast, P., the Fouriei transform of the corresponding gravity field can be written (Mssnn and PsnsnspN,1982):
f"o

g-ikx.-lftlrtx)6* / g ( k ) : ( 2 G P ^- l H I J -*

(1)

where, k is the wavenumber(:liwavelength), and G is the internationalgravity constant. Expanding the exponential and taking into account only the first order term which is a very good approximation for the small wavenumbers that we are in, the aboverelation can be written: interested /g(k):2cPos-itxtn"771li Q)

indicating a direct connection between the computed gravity spectra and the unknown mean depth, ho, of the interface, perturbated by the term Ah(k). At the long wavelengthsin which we are interested the non-linear effect of 1982),and the relation betweengravity /h(k) is negligible (Mrsrne and PsonnsEN, spectra and depth is almost log-linear. Thus, the slope of the graph provides an estimateof the depth. For the case of two-dimensional data, it is convenient to work with the absolute value of the radial spectrum of all the partial waves falling within a frequency range, as will be explained in section 3. 2. Tectonicsand Geologyof the Area

The area of Epirus lies along the North-Western margin of the Greek mainland, extendsinto SW Albania, and is situated along the Apullian and Aegean blocks (Fie. l). The rocks presentin the area aredivided into the lonian depositionalzone and its subzone (Gavrovo). The Adriatic-Ionian zone begins from Tirana in Albania It is made up of rocks and extendsalong the study area down to NW Peloponnese. ranging in age from Triassic (evaporitic facies) to upper Neocen (I.F.P., 1966). These are mainly pelagic limestonesfollowed by silts, sandstones,and conglomerates of typical flysch facies of Miocene age. The subzone of Gavrovo is exposedeast of the Ionian zone and consists of thick, mainly neritic limestones protruding through the flysch. Upper Eocene limestoneslie disconformably upon Cretaceouslimestones accumulated From Eocenetime until the end of the Tertiary, detrital sediments in the areL reaching a few thousand meters above the Triassic evaporites. Significanttectonic activity started in the subzoneof Gavrovo in the Oligocene,and in the Ionian zone in the Middle Miocene (I.F.P., 1966).This activity createdfolds and thrust faults that generally strike NW-SE (Fig. l). The SW migration of deformation and fault plane solutions from a few moderate earthquakesnear the Ionian coast (Dnarorouros and Dsrmests, 1974;

Moho Depths Estimation from Gravity Measurements

257

Ittl
7:
;:..1 _d-/-.)

li;f7i7.t-'g'",4

Fig. l. Tectonic setting of the general area, illustrating the tectonic zones encountered.

258

G.-A. Tsu-eNrts,J. Duropoulos,

and K. Dtutrnllpls

Pnpnzrcnos et al., 1984), together with studies of late Cenozoic deformation (MEncmn et al., 1979),suggest that the tectonic activity resultsfrom the collision of the Northern Ionian sea floor, and continuing North-Westward convergence of the Aegeanregion (LE PIcHoNand which is affected by the tectonicdevelopment 1980). AucnrrnR, 1979;BuncHFrELD, The tectonic framework of the area is mainly compressive, characterized by numerous zones of thrust faulting striking parallel to the coast (Fig. l), with earthquakes showinga predominantcompressive stress component. However, some seismic events have been attributed to normal faulting (Dnarorouros and DEtln.r.srs, 1974;McKENZTE, 1978)and a recent microearthquake surveyin the area (KINc et al., 1983)revealedthat the derived fault plane solutionsexhibit a varietyof focal mechanisms, which are not consistent with simple zonesof shorteningor extension. Additionally, a number of in situ stressmeasurements in the area (PlqutN el al., 1984),indicatethe presence of a dominant E-W extensional stress regime. 3. Method of Analysis The Bouguermap of Epirus, preparedby Mnnrs (1977),has beenusedin the presentinvestigation. The areawas surveyed gravimetricallyduring the yearsl97l1973(Fig. 2), with a surveydensityof 0.075stations/km2. The data wereuniformly reducedto Bougueranomalies at meansealevel,and topographicreductions from 0 to 166.7 km werecomputedon a spherical earth with a uniform density of 2.67g/cm3. The whole map was digitized and the data were gridded using a modified version of an interpolation algorithm written by SwlrN (1976)which usesfinite differenceequationsto produce a systemwith minimum curvature.The resulting grid was then contouredand inconsistent data were identifiedby the tight circular produced contours they and were removedfrom the data file. Eight squaregrids, measuring160x l60km each (32x32 data points) and overlappingby 501were usedfor the calculationof Moho depthsacrossa N-S direction. For each square grid, the following operations were performed. First, we removedthe regionalfield'by fitting a plane surfaceto all the data by the method of least squares.The obtained regional field was found to be describedby the expression: 'f(x, y): -3.985x* 1.068y+0.827 (3)

where x and y are the two axesof the grids. The residualgravity field valuesfor each block were then transformedfrom the spacedomain to the frequencydomain by meansof a Fast Fourier Transform (FFT). Applying an FFT algorithm to the abovedata subgridswill obviouslycausea

Moho Depths Estimation from Gravity Measurements

259

_i'n

2tr30',
,, r-40:t5/
I

\\.:\r:t{\,i((rj;
irl:N\N

;, )/ t;;i, !: ",'li,,i, [,

.rN\\\\N
\i il

-,rj"-ANi
teTs).

:i

J-

ls7.s6,

is 5 mgal. gravitymapof theinvestigated Contourinterval area. Fig. 2. Bouguer pronounced Gibbs phenomenonleading to great edge effects.In order to make transitionzoneswider, we usedthe following procedure(BnerrlcHARYYAand LEu, We assume that the residualfield vanishesat the points locateda distanceof

G.-A. TsneNns, J. Dnnopout-os, and K. Dnntntllts

r-- - r'+ -. 41"

|
I I
I I I I

(-

ru+8 D a r A - J
}O{ + }'-a rFa -iFa

I
<N DArA -----)

1 I I I I
I

I I

I
I

I
I I I

I I I
I I

I
I

L Fig. 3. Data expansion scheme.

four units of the data spacingfrom the boundary of eachsubgrid.Inclusion of these data points for an NxN block points resultsin (N+8)x(N+8) non-equispaced (Fig. 3). Next, bicubic splinesurfaces were fitted to the data by employing the finite (INouE, 1986) in such a way that the residual field and the element technique continuity of the first and secondderivativesare maintainedat eachone of the data points. These surfaces, were used to generatethe "tapered" data set. The Fourier transform of each data grid results in a set of Real R,n and Imaginary ,l* amplitudes by means of which the field values given at the grid points (x, y) can be representedby the sum:

s(x,y): I I nf, os{2n I(DXN) (kx+ my)\ " km + 1| sin{2n I@XN) (kx+ my))
where DX is the grid interval. Equation 4 can be written as follows:

(4)

s(x,y):

T;

(kx*my)- Ph} C\cos{2nl@xN)

(5)

where Pl is the appropriate phaseangle and

ch:{(Gh)'+(t!,)r}'t'
is the amplitude of a partial field wave, with wavelength DX'Nl(k2+m2)tt' frequencyF h:(k2 * m27rrz.

(6) and

Moho Depths Estimation from Gravity Measurements

261

t l v A lt,lU t t'l8tlS t c / $I t
. !:J:: , ibl: .;::::

Fig. 4. Schematiccomputation of the radial spectrum for the northern 160x 160km block. Spectrum matrix coefficients Cl are also shown. In order to calculate the radial spectrum for each data grid, we calculate first the 2-D power spectrum:

P(1,i): R(1,J)2+ Ig, n'.

0)

The radial spectrumis calculatedby superposing the 2-D spectrumwith a numberof concentric point of the matrix ringsthat arecentered at [1, 1],(upperleft P(1,J), which is the lowestfrequency componentof the data set(meanvalue),and with radialdistances described in Fig. 4. Next, the squares of the amplitudes of the partial wavesoccurringbetween two consecutive cycles are averaged. The scheme for computingradial spectrafor the northern 160x 160km data grid is illustratedin Fig. 4. The resultingvaluesform the radial spectrumof the anomalous field underconsideration, and are normalized to the valueCo(1,l). By this technique,we transform the current two-dimensional problem to a onedimensional one, assuming no preferredorientationin the iesponse. If the logarithmsof sucha radial spectrum are plotted againstfrequency, one

G.-A. TsnrrNrn, J. DnexoPoulos,and K. Dmlrntlpls

t

by one or more straight of points which may well be represented often finds a series lines. From the slope of the lines and the standard error of fit we calculate the correspondingdepth and error depth to the appropriate density discontinuity for eachdata grid. 4. of the Results Discussion

of the northern and southern160x 160km parts of Figure 5 showsthe spectra the map respectively.The obtained 3-D representationsof the encountered that the problem due to the preferredorientation is minimal. suggest wavenumbers radial spectrumof the northern block is shown in Fig. 6. The corresponding of this plot correspond to two different density disconThe two linear segments 160-40km while the second rangesfrom wavelengths tinuities.The first segment segmentrangesfrom 40 to 20km. Beyond the 20 km wavelength,the amplitudes in the to randomnoisecontained which corresponds foim an almostwhite spectrum data. discontinuity, (which is attributed to the To estimatethe depth to the deepest Moho), the central axis of the area was scannedby 8 overlappinggrids of the maximum allowed grid size 32x32 data points (160x l60km), allowing the radial spectraare greaterthan 40 km. The corresponding detectionof wavelengths depictedin Fig. 7, while the deriveddepthsto the Moho with the corresponding that the average in Fig. 8. This latter figure suggests .rio, are graphicallypresented Moho depthsin the areavary from a minimum of 32km to about 39km. A recent

Orr{7r,

wavenumbers of the encountered representation Fig. 5. Three-dimensional the northern (A), and the southern(B) blocks'

Moho Depths Estimation from Gravity Measurements

263

ltAvENUl.lBERS I (CYCLES/KI'|)
t

km

Fig. 6. Radial spectrum for the northern block.

---J-l 0.5

0,5

fig. Z. Radial spectrum for the eight grids used for Moho depths estimation.

r
I

264

G.-A. TsnrnNrIs. J. Dnexopoulos. and K. Drurrntlors 123456.78 Irllllll

r--20KM-i
IT IF

to V

tft
r 30
40

!ze

S T l | OA YR I I

a=

Points errors. along(N-S)Epiruswith associated Fig. 8. Moho depths traverse .., grids the scan. used for x 160km of the 160 1,2,' 8 represent thecenters investigation in the area employingthe analysisof P, and P,' waves(Knnacosrns, a crustalthickness of the order of 40 km, while an applicationof a 1982), suggested resultedin a crustal thickness singlestation deconvolutionto long-periodP-waves 1988).Theseresultsseemto be in of about 35km (TsnrnNusand DRAKopouLos, good agreement with the resultsobtainedfrom the presentanalysis. 5. Accuracyof the Results in the resultsmight be found in the effectthat every One causeof discrepancy gravity field, thus the spectral unit of buried masscontributesto the total observed will alwaysoccur. overlap betweenthe effectof deeperand shallowersources caused by digitization. errors Care also has to be taken to minimize aliasing Obviously, closer digitization has the effect of incorporating high frequency components,resulting in the decreaseof the slope of the log spectral curve (Sencron,1968). One difficulty with this refinementoccurs becausethe higher wavenumber hard to and it becomes components smallermagnitudes tend towards increasingly separatethem from noise. The overall effect is that the derived depths will be Increasingdigitization on the other hand leadsto aliasingerrors, underestimated. hencean optimization of digitization spacingis essential. 6. Conclusion

geophysical investigationin the area of Epirus (NW As part of an extensive from the existinggravity data. Greece),'we attemptedto derivethe crustalthickness This was achievedby employing spectraltechniquesof analysisof the Bouguer gravity map of the area.

Moho Depths Estimation from Gravity Measurements

265

A series of 8 overlapping grids of maximum allowed grid size (160 x 160 km), were used in an attempt to increase resolution while retaining the capacity to discern deeper sources. For each grid, Moho depths were established from the slope of the log-power radial spectrum at the lower end of the wavenumber band. crustal thickness in the area was found to vary from 3l to 39km, and this result is in good agreement with the results obtained employing seismologic al data. REFERENCES BHltracsanYYA, B. K. and L. LEu, Spectralanalysisof gravity and magneticanomaliesdue to two-dimensional structures,Geophysics,40, 993-1013, 1g75. Bosp,P. K. and B. J. SENcuprn, Spectral analysesof offshore magnetic data for computing depths to the basementrocks. Mar. Geophys.Res., 6,245_256, lgg4. BuncHrIELo, B., Evolution tectonique du systemAlpin en Mediterranee.poinconnement et ecrasement. Bull. Soc. Geol. Fr., 19, 437460, 1990. DrlrrrRre,ors, K., G.-A. TSELENTTS, and K. THeN.Lssoules, A basic program for 2-D spectral analysis of gravity data and source-depthestimation, Comput. Geosci.,13, 549-560, 1987. Dnaropouros, J. K. and N. D. Drrtsesls, On the mechanismof someearthquakesin the area of western Greeceand the stressproducing them, Proc. l3th Eur. Seism.Com. Special Issue of rechn. and Ec. studies, BucharestD-ser. Geoph.prosp.,10, 169-192, 1974. GRTEN, A. G., Magnetic profile analysis, Geophys'. J. R. Astron soc. London 30, 393-403. 1972. HluN, A., E' G. KtNo, and D. C. Mrsune, Depth estimation of magneticsourcesby means of Fourier amplitude spectra.Geophys.prospect., 24, 2g7_30g,1g76. INoue, H.' A least squares smooth fitting for irregularly spaced data-Finite element approach using the cubic B-spline basis, Geophysics, sl,205l-2066, 19g6. INstIrurn FneNcnrs ou Prrnor-E (LF.p.), Etude Geologiquede t'Epire,I.G.M.E., Athens, 1966. K,{n,{cosrAs, G., Crustal structure from seismologicalinvestigationsin the area of Greece. Ph.D. Diss. Seismological Lab., ThessalonikiUniversity, 133pp., 19g2. KING, G., G.-A. TssrnNtts, P. MotN.m, S. Roncxrn, H. SrNvu,lL,c. Surr-snls,and J. Srocr, Microearthquake seismicityand active tectonics of northwestern Greece,Earth planet. Sci. Lett., 6, 279288, 1983. Lr PrcsoN, X. and J. ANcErmn, The Hellenic arc and rrench system: A key to the neotectonic evolution of the E. Mediterranean arca; Tectonophysics, ffi, 142, 1979. MaKnrs, J., Geophysical Investigations oJ-the Hellenides, Geophys. Eiyzel-schr., vol. 34, Hamburg University, 126 pp., 1977. McKENzrE, D. P., Active tectonics of the Alpine-Himalayan belt: The Aegean sea and surrounding regions, Geophys. J. R. Astron Soc., 55, 217_254,lg7g. MEncrrR, J., N. D. Dpusassrs, A. Geur*n, J. JeRnrcr, F. Lruullr, M. SssnrpR, and D. SonEr, La neotectoniquede l'arc Aegeen, Rev. Geor.Dyn. Geogr.phys.,2,67-92, rg./g. MISun.l, D' C. and L. B. PnoEnsrN,Statistical analysis of potential fields from subsurface -265, L992. reliefs, Geoexploration,19, 247 Npcr, J. G., P. K. AcneweL, and K. N. N. Reo, Three dimensionalmodel of the Koyna area of Maharashtra State (India)-based on the spectral analysis of aeromagnetic data, Geophysics,48, 964-974, 1983.

G.-A. TsnrsNrrs.J. Duropoulos.

and K. Dnflrnllnrs

Plr, P. c., K. K. KnuneuA, and P. UNuxnrsHNAN, Two examples of spectralapproachto sourcedepth estimationin Gravity and Magnetics,Pageoph, 17,772-783, 1979. Pepa,z.e,cnos, 8., A. KrRATzr, P. HaraonrarrRrou, and A. Rocca, Seismic faults in the Aegean area, Tectonophysics, 106, 7l-85, 1984. P.l,qun, C., J. Brovrr, and C. ANcEuoIs, Tectonic stresses on the boundary of the Aegean domain. In situ measurements by overcoring,Tectonophysics,ll0,145-150,1984. Stncton, A., Spectral analysis of aeromagneticmaps, Ph.D. thesis,University of Toronto, 1968. SrnctoR, A. and F. S. Gur.rr, Statistical models for interpreting aeromagnetic data, Geophysics, 25, 293-302, 197 0. Sw.LIN,C. J., A Fortran IV program for interpolating irregular spaced data using the differenceequations for minimum curvature, Comput. Geosci.,1,231-240, 1976. TREITEL, S.,W. G. CLuurNt, and R. K. K.lul, The spectraldetermination of depthsto buried magnetic basementrocks, Geophys. J. R. Astron: 9oc.,24, 411'/,28, 1971. TsnLnutts,G.-A. and J. Dnaropouroso Crustal structure in W. Greeceas obtained from the analysis of surfacewaves, Pure Appl. Geophys.,1988(in press).