Bullard Laboratories, Department of Earrh Sciences, University of Cambridge, Madingley Rise, Madingley Road, Cambridge CB3 OEZ

Transcription

1 Geophys. J. R. astr. Soc. (1983) 12, A North Sea-southem Norway seismic crustal profile Bruce R. CaSSell * Bullard Laboratories, Department of Earrh Sciences, University of Cambridge, Madingley Rise, Madingley Road, Cambridge CB3 OEZ SVein Mykkeltveit NTNF~NORSAR, Post BOX 51, N-2007 Kjeller, Norway Reidar Kanestrqhn Seismological Observatory, University of Bergen. Allegt. 41, N-SO14 Bergen V, Norway Eystein s. Husebye NTNFINORSAR, Post Box 51, N-2007 Kjeller, Norway Received 1982 September 7;in original form 1982 May 12 Summary. In this paper results are presented from a seismic refraction experiment (CANOBE) carried out in southern Norway. Ten explosions, fired at sea, were recorded on land by shifting 13 recording instruments along a profile with an average station spacing of 5 km. The main line runs in a northeasterly direction from the south coast at Lista along the western margin of the Oslo Graben into the NORSAR array, with a total length of 5 15 km. A separate leg runs across the Graben just north of Oslo, for the first time allowing a direct comparison of seismic records in the Graben with those in the adjacent Precambrian shield area. The laterally varying crustal structure along the profile is examined by modelling of travel times and amplitudes of P-wave arrivals. The Moho, which is the major discontinuity in the lowermost crust, appears to sink beneath the coast from 27 km on the seaward side to 34 km onshore. The two-dimensional modelling procedure adopted proves invaluable in explaining the characteristic amplitude pattern observed in this area. Beyond the coastal area our picture of the crust is that of a relatively homogeneous one, as expected for a shield area. There are no indications of significant discontinuities in the crust along the first 300 km of the profile, although from the large Pg amplitudes within this distance range we infer a strong velocity gradient in the lower crust. Two structural models are proposed for the Oslo Graben where the Moho appears to be elevated to between 25 and 29 km. 1 Introduction During the last decade numerous seismic profiling investigations have been carried out in Fennoscandia and a relatively large number of crustal models have been derived. There is *Now at: Western Geophysical Company of America, 455 London Road, Isleworth, Middlesex TW7 5AB. 25

2 734 B. R. CasseZZ et al. considerable inconsistency concerning the number of crustal layers and the associated velocity distributions as reported by various authors. The reason for this is probably a combination of real crustal variations, relatively poor sampling densities and the non-uniqueness of the inversion of refraction seismic data. Recently. Bungum, Pirhonen & Husebye (1980), using a spectral ratio technique, published crustal thicknesses beneath 1 1 permanent seismograph stations in Fennoscandia. This study was combined with a review of previous refraction surveys, and the major results was that the two types of Moho depth estimates agreed well for profiling lengths exceeding 300 km. The area of this study is southern Norway where several other profiling surveys have been undertaken as indicated in Fig. 1. Earthquake recordings at the NORSAR array, which is located partly within the Oslo Graben, have prompted a number of inversion experiments IF Precambrian rocks Mare Gneiss Region Precambrian and Cambro Silurian rocks within the Caledonian zone 0 N2 Figure 1. Simplified geological map of southern Norway with CANOBE shot points and recording legs. Resillts from previous profiles are reported by Sellevoll & Warrick (1971) and Kanestrdm & Nedland (1975) for the Flora-Asnes and Fedje-Grimstad profiles, Kanestrdm & Haugland (1971) for the 3-4 profile, Tryti & Sellevoll (1977) for profiles in the Oslo Graben, Weigel, Hjelme & Sellevoll(1970) for the Skagerrak profile and Mykkeltveit (1980) for the Arsund-Otta profile. Also shown is the NORSAR array siting area with the original 22 subarrays, each comprising six short-period instruments. Five of the subarrays (bold rings) recorded the CANOBE shots.

3 S. Norway seismic profile 735 aimed at detailed mapping of crustal/lithospheric structures beneath the array (eg Berteussen 1977a, b; Aki, Christoffersson & Husebye 1977; Christoffersson & Husebye 1979; Troitskiy, Husebye & Nokolaev 1981). Gravity observations have also been used for inferring crustal structure (e.g. Ramberg & Smithson 1975; Husebye, England & Ramberg 1978). The results obtained from the various surveys studies mentioned above exhibit occasional discrepancies which cannot be ascribed to genuine structural differences as the areas investigated were partly overlapping. Consequently, as a Cambridge group was carrying out a refraction experiment across the North Sea in the summer of 1980, we took the opportunity to expand the shot-firing scheme in order to obtain a densely sampled land profile (CANOBE) in southern Norway. The profile was orientated in a north-easterly direction, away from the line of shots off the southern coast and running parallel to the western border of the Oslo Graben as shown in Fig. 1. At various stages the CANOBE main line crosses the three profiles Fedje-Grimstad, 3-4 and Flora-Bsnes (Sellevoll & Wari-ick 1971 ; Kanestrdm & Nedland 1975; Kanestrdm & Haugland 1971) before extending into the NORSAR array area. With the aim of establishing a direct comparison of travel times between the shield area and the adjoining Oslo Graben where intrusive rocks of Permian ages are prevalent, an additional leg was arranged to run within the Graben. Previous refraction studies (Tryti & Sellevoll 1977) suggest a three-layered crust within the Graben. In this paper we present an interpretation of the high quality records obtained along the densely sampled CANOBE profile. Synthesis of P-wave amplitudes and travel times, assuming a laterally varying structure, constitutes the basic tool for determining the crustal structure and thickness in the coastal areas of southern Norway and the general features of the Moho near and in the Oslo Graben. These results are then compared with and discussed in the light of past crustal studies in the CANOBE surveying area. 2 Field work and preparation of the data The CANOBE project, its name derived from the participating institutions Cambridge University, NORSAR and Bergen University, took place between 1980 July 26 and August 4. A total of 13 recording instruments were available, consisting of six Cambridge SCRAPS (Seismic Cassette Recording Apparatus, see Cassell 1981) five 3-component MARS 66 stations from Bergen (Berckhemer 1970) and two Sprengnether DR-100 digital recorders from NORSAR. At the same time as the CANOBE experiment a Cambridge North Sea refraction project was being carried out, with shots fired along a profile between Leith in Scotland and the southern coast of Norway. Four of these explosions (N2-N5 in Fig. 1) were located to the east of the Central Graben in the North Sea, and were used for the land-recording in Norway. In addition the Royal Norwegian Navy provided several tons of AMATOL in the form of torpedo warheads and a coast guard ship to fire shots Hl-H6. The recording scheme consisted of seven legs (Fig. I), one per shot, each comprising 13 mobile stations. Legs 1-3 had a sensor spacing of 4 km, legs 4-6 a 6 km spacing and leg 7 a spacing of 5 km. The relatively short station intervals for legs 1-3 were chosen for dense sampling of the inner cusp of the expected triplication of the travel-time curve associated with the crust-mantle boundary. Leg 7 transected the Oslo Graben just north of Oslo while the main line ran across the Precambrian rock to the west, extending into the NORSAR array. Five NORSAR subarrays (OIA, 02C, 03C, 04C, 06C) recorded the shots continuously, thus enabling the main line distance range to be extended to about 515 kin. During shots N2, N3, N4 and N5 the 13 field stations occupied permanent positions along leg 1. For subsequent shots Hl-H6 at the same position as N5, the stations were moved along the profile, covering one leg per shot. The N2, N3 and N4 shots had charges of 500kg TNT while shot N5 was increased to 750 kg TNT. These were buoyed charges with known

4 736 B. R. Cassell et al. Table I. Coordinates and explosion times for the CANOBE seismic field experiment between 1980 July 27-August 4. Shot Date Time Shot Point Coordinates Shot Water Charge 1980 GM 1.a t. Long. Depth Depth kg H M S (0 1 ") A (O ' ")E TNT 57 N? N4 N5 HI H2 H1 H4 115 H Zi8 31R no in R.n n.o no 12.n 5' n.n detonation depths. Hl-H6 were 825 kg TNT charges (Table 1) and were all detonated at the sea bottom. The distance from the shot position N5 to the first recording station was 73 km. Cambridge shot times are accurate to 0.01 s. Shots H1 -H6, fired by the Norwegian Navy, were timed electrically with an accuracy of s. The timing for the field stations was based on the coded MSF signal (60kHz) transmitted from Rugby, UK. Unfortunately, reception was bad during the shot windows for H5 and H6, resulting in the loss of some records. After each shot records were played back in order to monitor gain settings. A considerable amount of time was invested in field reconnaissance which resulted in a smoothly run operation and recordings with excellent signal-to-noise ratios. The task of digitizing the analogue records was distributed amongst the participants using a common sampling rate of 200 Hz and standardized tape formats. A topographical correction was applied to the data, and the time delays in water for the suspended-from-buoys charges N2-N5 were calculated using a mean water velocity of km s-l, and subtracted from the travel times. In this way all stations are projected to sea-level and all shot positions are projected to the sea bottom. The travel-time errors accumulated during the preparation of the data and in the topographical correction are discussed in Cassell (1981) where the total error per record is estimated to be less than a tenth of a second. The initial record sections were filtered between 0.2 and 15 Hz. Fig. 2 shows the record section corresponding to the main line for shots N5 and H1 -H5, reduced by 8 km s-l. The NORSAR array records, from 450 km onwards, have been low pass filtered at 4.75 Hz sampling rate 20 Hz. Apart from these NORSAR records, the section presented contains the SCRAP, MARS and DR-100 data. True amplitudes are multiplied by distance in the record sections. The signal polarities were the same for all recording instruments. The spectra for typical records in this section are found to have dominant frequencies within Hz in % R25 R R Interpretation of the CANOBE data Although the quality of the data and the relatively dense sensor spacing along the profile provide an excellent basis for interpretation we are somewhat encumbered by lack of observations for the first 70 km of the profile and by the absence of a reversed coverage. This last point leaves us without direct information concerning possible dips of structures beneath the profile. On the other hand, previous profiling results and seismological Moho thickness estimates favour as a first approximation a standard crust of 33 km thickness. In consequence, the first part of the interpretation is carried out based on lateral homogeneity for obtaining the velocity-depth distribution beneath the profile. The final interpretation of

5 S. Norway seismic profile 737 Q z u) 0 t z Y Figure 2. Main line record section. Amplitudes are multiplied by distance and the seismograms are filtered between 0.2 and 15 Hz. The records beyond 450 km were recorded by the NORSAR array and have been passed through a 4.75 Hz low pass fiter. Dots indicate first arrival picks. The continuation of the crustal phase to the ongin is not clear although an intercept time of 1.1 s indicates a travel-time delay in the upper crust.

6 738 B. R. Cassell et al. the CANOBE main line and the Oslo Graben leg, however, was carried out on the basis of 2-D seismogram synthesis. In the latter case, constraints can be placed on the final model by amplitudes as well as travel times of primary and multiple arrivals. The interpretation of the coastal record section originating from the N2, N3 and N4 shots will be presented separately (Mykkeltveit & Cassell 1983). Suffice it here to mention that these data indicate the presence of two predominant phases, Pg and P,,, both appearing as first arrivals with apparent velocities of 6.3 and 7.8 km s-', respectively. 3.1 THE MAIN LINE RECORD SECTION A first inspection of the main line record section in Fig. 2 shows two clear first arrival traveltime branches with apparent velocities of 6.3 and 8.1 km s-l. The linear 6.3 km s-l crustal phase (Pg) appears in the first record in the section and continues to a distance of approximately 240 km where its amplitude rapidly decays. The P,, phase, which is linear from 150 to 270 km, begins to undulate and shows considerable offsets from 270 km onwards. One of the most conspicuous first arrival offsets between neighbouring traces amounts to 0.4s and occurs at approximately 370km. The first arrivals in the NORSAR records ( km) are continuous within each subarray although there are differences in arrival times from one subarray to the next. This is particularly pronounced in the records from subarray 02C (Fig. 1) around 490 km which are delayed by nearly 0.8 s. This offset is also apparent in the secondary arrivals at 13 and 19s in the NORSAR records (Fig. 2). Apart from these late arrivals, which can be ascribed to multiple reflections within the crust, there appear to be no identifiable secondary arrivals in the latter portion of the record section. So, in addition to the primary phases, the most pronounced feature of the record section in Fig. 2 is the branch of Moho reflections (PMP). Other phases, possibly revealing discontinuities like the Conrad within the crust, have not been identified. 3.2 INTERPRETATION OF THE MAIN LINE SECTION ASSUMING LATERAL HOMOGENEITY To get a preliminary idea of the structure beneath the profile, linear travel-time branches were drawn through the first arrivals which were interpreted assuming lateral homogeneity. These branches were located in the undulating portion of the first arrivals in such a way as to obtain a by-eye best fit. By extrapolating the Pg branch to 0 km we find a positive intercept time of 1.1 s, which was accounted for by introducing a surface layer of relatively low velocity as shown in Fig. 3. Our main concern at this stage of the interpretation was to gain some basic knowledge on the gross features of the crust-mantle transition including Moho depth, so the rather arbitrary surface layer introduced merely serves the purpose of making the total travel times attain the observed values. For example, information on the sequence of sedimentary layers near the N5 shotpoint were obtained independently from the Norwegian Petroleum Directorate and subsequently used in the final 2-D modelling below. Starting with Pg and P,, velocities taken from the record section and Moho depths in the expected range of km, theoretical time-distance curves were computed until a satisfactory fit with the data was achieved. A prominent feature in the record section is the large amplitudes of secondary arrivals at about 186 km, apparently part of the PMP branch. While modelling the data on the basis of travel times a velocity-depth configuration was established which produces a focusing effect near the outer cusp in the triplication (Fig. 3). The outer cusp is made to terminate at 250km by introducing a velocity gradient of s-' in the depth interval km, and an increasing velocity gradient down to 32.5 km where a velocity Of 8.1 km sc1 is reached. As can be seen in the figure, the

7 - X t- 40.q I i velocity km/s TT-T--'-T 5.a G Q. a i T - V ~ zm.0 z DISTANCE (KM) Figure 3. Travel time-distance curves for the laterally homogeneous velocity-depth model (inserted). The portion of the velocity-depth function corresponding to unrecorded arrivals in the first 70 km of the record section is dashed. Crosses indicate observed arrivals. Note the discrepancy between observed and theoretical subcritical PMP which indicates an elevated Moho beneath the first part of the profile CQNOBE DISTANCE Figure 4. Synthetic seismograms computed by the reflectivity method for the model in Fig. 3. The source signal has a dominant frequency of 3 Hz and four extrema. The dot indicates the outer cusp in thepmpsurface layer multiple caused by the artificial first order discontinuity. The strong PMP amplitudes are produced by the velocity gradients in the crust. Amplitudes are multiplied by distance. IN KM

8 740 B. R. Cassell et al. computed subcritical reflection travel times do not agree with the secondary arrivals between 70 and 1 10 km. The implications of this discrepancy will be discussed later on, as at present we confine ourselves to demonstrating the effect of the transition zone above the Moho on the PMP amplitudes. Now, applying the reflectivity method (Fuchs & Muller 1971) to the model in Fig. 3, the above gradient is sufficient to reproduce large amplitudes in the retrograde travel-time branch of the observations (Fig. 4). The theoretical amplitudes near the critical distance are large in comparison with the observed ones. However, further attempts to model these amplitudes in the context of laterally homogeneous models were not considered due to the mentioned travel-time discrepancies. The synthetic seismograms also contain multiple reflections of the Pg and PMP waves off the bottom of the surface layer. Although similar arrivals can be seen in the record section, the theoretical ones arrive approximately 1 s too early. Since the observed multiples have longer travel times, this suggests that they are associated with free surface reflections, which are not included in the synthetic section of Fig INTERPRETATION OF THE MAIN LINE SECTION ASSUMING A LATERALLY VARYING STRUCTURE From the results of previous profiling experiments and seismological studies in southern Norway (Fig. I) crustal structure and Moho depth are expected to vary especially around the endpoints of the CANOBE profile. It is essential at this stage to deduce a model for the first 200 km of the profile for which PMP observations are available in addition to the Pg and P,, phases. Close examination of Fig. 5 reveals that strong P MP amplitudes are confined to two distinct distance intervals. The first is around 11 5 km and is considered to be near the critical distance where relatively large amplitudes are expected. The second is around 185km and is limited to a few seismograms. The latter extreme P MP amplitudes can be attributed to fucusing effects caused by waves touching a caustic (e.g. see Cerveny, Molotkov & Psencik 1977). While modelling these large amplitudes, however, the relatively early PMP subcritical arrivals (Fig. 3) must be accounted for. These suggest a thinner crust as compared to the laterally homogeneous model of the previous sections although the travel times of the P,, arrivals from 150 km onwards must still be retained. The calculations for the laterally varying models are performed using Cassell's (1982) box method (based on zero-order ray theory) where physical medium parameters are defined at 1 x 1 km grid points. A model is presented in Fig. 6 which produces good agreement between the theoretical and observed travel times. The structure beneath the source is known to contain a 2.62 km thick stack of sedimentary layers with a mean velocity of 3.26 km s-'. The surface layer along the remaining part of the profile is taken to be 3 km thick with a velocity of 5.5 km s-'. These values are modified from the interpretation in the previous section but are not uniquely determined from the observational data available. The main features in the model consist of a Moho with a depth of km offshore which increases to 33 km over a distance of 30 km beneath the coastline. The velocity gradient associated with the Moho transition gradually fades away after 200km as shown in Fig. 10. In the coastal area the main discontinuity, represented by a velocity jump from 6.8 to 7.5 km s-', occurs at a depth of km and is overlying a velocity gradient zone reaching a velocity of 8.1 km s-l at 34 km. This discontinuity provides the reflection coefficient necessary to satisfy the observed PMP amplitudes while the underlying velocity gradient causes the refracted rays to be turned around after a relatively short distance. In this region the Moho should be regarded as a transition zone in which its depth is not clearly defined. Beyond 110 km the Moho transition is assumed to change into a first-order discontinuity at a depth of 34 km

9 S. Norway seismic profire 741 Figure 5. Arrival picks in the main line record section. Amplitude maxima in the secondary arrivals occur at 115 and 185 km.

10 P" a+-. ~ +) m- m 0 00 I I 1 I I I I I I I I I v I l- a ; 52.s 78-0 IwV1ODEL I 1 I 4, I coast DISTANCE (KM) 0- observed I I I I I I I I I Figure 6. Comparison of observed travel times and theoretical ones computed for model 1 in Fig. 10. The vertical radii of the symbols indicating the observed times represent the reading errors. K denotes the Moho depth derived from Kongsberg seismograph station data. Downloaded from at Pennsylvania State University on Marc

11 S. Norway seismic profile 743 with a P,, velocity of 8.1 km s-'. The upper mantle velocity gradient is approximately s-l. The PMP travel-time branch is assumed to represent two types of arrivals: (1) reflections off the elevated Moho and (2) diving rays which are 'turned around' by the strong velocity gradient (0.07 s-l) in the depth interval of km. As the initial angles of the diving rays in the crust increase from the horizontal these rays reach a maximal distance at 250 km. From this point the ray surfacing distances decrease with increasing initial angles, forming a retrograde travel-time branch which terminates as the diving rays begin to reflect off the elevated Moho (Fig. 6). The large amplitude values associated with the inner point of this retrograde travel-time branch are shown in Fig. 7 along with the total distribution of amplitudes computed by the box method for the proposed laterally varying structure. The PMP amplitude-distance curve exhibits two pronounced maxima corresponding to the two arrival types mentioned above. Of special interest is the second maximum, associated with the caustic. Synthetic seismograms were calculated for Model 1 (Fig. 8). Note that zero-order ray theory as used here is not exact near caustics or critical points. Nevertheless, the amplitude distribution in the synthetic section qualitatively justifies the choice of the model. The wavelets of the theoretical rays which have touched the caustic have not been subjected to a 90" phase shift and it is not possible to ascertain whether this applies to the observed wavelets (Choy & Richards 1975). Amplitude ratios (PMP/Pg and PMP/P,) of observed and theoretical seismograms are presented in Fig. 9. Each curve exhibits distinct peaks corresponding to the critical point of the PMP travel-time branch and the caustic. The first theoretical amplitude maximum is located at 95 km whereas the observed maximum is found to be at approximately 115 km. This discrepancy is not surprising as zero-order ray theory does not include wave effects in the critical region beyond the critical point. The reflected and head waves interfere with each other in this region and may cause a shift of the maximum amplitudes by several tens of kilometres. Cerveny & Ravindra (1971) have quantified the extent of the interference zone, given velocities, interface depth and pulse duration. For the structure found here this zone is found to be over 30 km wide. A similar phenomenon applies Figure 7. Amplitude-distance curves for models 1 and 2 in Fig. 10. The dot indicates the outer cusp of the Pg travel-time branch. The loop near the dot is an artifact of the box method. Curves 1, 2 and 3 correspond to the surface multiples in Fig. 11. The dashed P,, corresponds to head waves while the solid P, corresponds to refracted waves.

12 10 h m v co -r X I- I h 27.5 I: x Y I 35.0 t- a. w Q DISTANCE (KM) MODEL I I 1 I I I I I 1 I I I I I I Figure 8. Synthetic seismograms for the box method (Cassell 1982). The source wavelet has a dominant frequency 3.1 Hz, and amplitudes are multiplied by distance. Note the large amplitude at 176 km resulting from the velocity gradient in the lower crust. Downloaded from at Pennsylvania State University on Marc

13 S. Norway seismic profile ( DISTANCE IN KM Figure 9. Amplitude ratios for the observed (solid) and calculated (dashed) seismograms for model 1 in Fig. 10. The black triangles indicate the critical distance for the observed PMP reflection and the amplitude maximum in the observed PMP branch. The horizontal distance between observed and theoretical peaks is a result of the infinite frequency approximation in the ray method. to the amplitudes near the caustic, which are exaggerated by zero-order ray theory. The relatively high theoretical amplitudes past 220 km are considered to be caused by the highfrequency approximation. Real wave amplitudes near endpoints in travel-time curves do not decay as abruptly as the ray amplitudes modelled here. As described in the previous section, the P, arrivals begin to undulate at distances greater than 270km. The interpretation of structures in this distance range is ambiguous as we cannot determine whether the travel-time perturbations are caused by Moho topography or by lateral crustal variations or by a combination of both. The most pronounced feature is the large offset at 380km (Fig. 2), which is associated with a rather abrupt change in apparent velocity. This can be ascribed to a change in Moho dip, assuming a constant P, velocity of 8.1 km s-', implying that the Moho has a small up-dip between 110 and 330km after which its depth increases relatively rapidly before it begins to rise again. An up-dip of 1" would produce an increase in apparent velocity of 0.1 km s-' compared to a horizontal Moho. At large ranges the P, diving rays, in contrast to P, head waves, will have propagated at greater depths below the Moho, sampling higher velocities, causing an increased apparent velocity at the surface. This suggests that the Moho in the last 80 km beneath the profile is nearly horizontal. In Fig. 10 two possible structural configurations are presented which satisfy the P, traveltime data. The modelling was done with the criterion of fitting computed travel-time curves to the trend of the two apparent velocities in the P, arrivals. These are a model with a varying Moho (model 1) and one with a varying surface layer thickness (model 2). When assuming crustal velocities to account for P, travel-time variations we cannot determine in which part of the crust these variations occur. Hence it is only possible to quantify the mean velocity value. For illustrative purposes, however, the crustal variations are represented by variations in surface layer thickness. The travel-time delay caused by the decreased mean crustal velocity has the same effect on the P, arrivals as does the varying Moho in model

14 746 B. R. Cassell et al. IN KF O!Si=INCE a d ZMa.0 ZA AQQ Figure 10. Velocity-depth distributions for models 1 and 2. The first-order discontinuity in the crust after 340 km for model 2 has not been observed and serves only to signify a decrease in the mean crustal velocity in that area. Model 1 has a mean crustal velocity of 6.46 km s-l in the last 180 km of the profile as opposed to the mean velocity of 6.16 km s-l achieved by thickening the surface layer (model 2). Significant constraints are imposed on the interpretation when attempting to account for the late arrivals which occur between 16 and 20s in the last 150km of the main line (Fig. 1 1). These arrivals appear to be surface multiples corresponding to the Pg, PMP and P, phases of the first arrivals. The conspicuous travel-time delays at NORSAR subarray 02C (Fig. I), located approximately 30km to the north-west of the extrapolated profile line, were disregarded in the correlation of the travel-time branches. These delays are not surprising as the associated path legs coincide with a relative low velocity part of NORSAR (Christoffersson & Husebye 1979; Berteussen 1977b). When calculating multiple ray paths for model 1, we find that the corresponding travel times fall approximately 1 s short of the observed arrivals. This indicates that the mean crustal velocity is too high as these waves travel only in the crust. The required delay is achieved for model 2 in Fig. 10 by introducing a Moho depth at 30km and a mean crustal velocity of 6.16 km s-. It is not possible to determine accurately the arrival times of the observations near the cusp in the multiple travel-time curve although the presence of larger amplitudes in that region is evident. The discontinuous Moho in Fig. 10 (model I), which has been modelled assuming as normal crust, can be smoothed out by decreasing the mean velocity of the crust in the last 180km of the model. In this way the observed travel times for both the P, arrivals and the presumed multiple arrivals are satisfied by introducing a surface layer which thickens from a distance of 300 km to the end of the profile (model 2). It has not been possible to correlate the P, multiple travel-time branch while at the same time leaving the multiple Pg branch in place. This could imply that the true P,, multiples are too weak to be detected and that the faint arrivals at 13s in the last records in the section are associated with other wave types. In Fig. 12 synthetic seismograms are calculated for model 2. The increasing P, amplitudes with distance are in general agreement with those in the observations, although the

15 LP L

16 748 B. R. Cassell et al. I,,a, I J" I m m m t

17 S. Norway seismic profile 749 2c h Y, v - 03 X I- I 15 1c DISTANCE (KM) Figure 13. Oslo Graben records and travel time picks (dots). The secondary arrival (line) has not been interpreted.

18 750 B. R. Cassell et al. n I o otncrved I u DISTANCE IN KM t ELEVATED MOHO Mow S t- w Q a CANOBE - OSLO GRABEN Figure 14. Oslo Graben models consistent with P, travel times. The Oslo Graben first arrivals are approximately 0.8 s earlier than the main line first arrivals at the same distance (upper diagram). The dotted area represents the mafic intrusion.

19 S. Norway seismic profile 75 1 theoretical ones are not continuous throughout all the records. This inconsistency is caused by unavoidable lateral velocity gradients in the laterally varying model. The multiple arrivals, however, have much smaller relative amplitudes than those in the observations. In a few cases, some of the theoretical records lack arrivals due to difficulties in finding rays to each receiver location. 3.4 THE OSLO GRABEN DATA Due to difficulties concerning time signal reception while recording shot H6 along the final leg across the Oslo Graben only five stations produced useful records. These five records, however, show a considerable decrease of approximately 0.8 s in P, travel times in the Oslo Graben compared to records at the same distances along the main line (Figs 13 and 14). The first arrivals are relatively weak compared to much stronger secondary arrivals which appear 0.5 s later. This secondary phase, which has an apparent velocity of 8.96 km s-l, may imply vertical velocity variations beneath part of the profile, but the data at hand are insufficient for any detailed interpretation. Two models (Fig. 14) are proposed which fit the Oslo Graben first arrivals equally well and both of which have the same structure as the main line models up to a distance of 300 km. The bottom model in Fig. 14 consists of a Moho rising from 32 km to approximately 24 km in the distance range km. The mean crustal velocity is 6.38 km s-'. Other studies have revealed evidence for the intrusion of a relatively dense, high velocity body in the lower crust (Husebye et az. 1978; Ramberg 1976). In the second model in Fig. 14 a 7 km thick region with a velocity of 7.1 km s-' is located on top of a 29 km deep Moho and the mean crustal velocity is now increased to 6.52 km s-'. 4 Discussion and conclusions In this study we have demonstrated that the use of amplitude information as a supplement to travel-time data is essential in delineating earth structure. The outstanding example here is the extremely strong amplitudes in the PMP branch, confined to a relatively small distance interval around 180km, which appear to be created by a strong velocity gradient in the lower crust. In this respect, the calculations of amplitudes for 2-D media proved very important compared to the 1-D amplitude calculations by the reflectivity method. Moreover, with a less dense station spacing along the first part of the profile these high amplitude arrivals might not have been recorded and a different interpretation would have been made. The Moho is found to be km deep beneath the first 80 km of the profile before it dips downward over a distance of 30 km beneath the coast and subsequently levels out at a depth of approximately 34 km. The interpretation of the record section presented is only approximate for the last 300 km of the main line. Two structural models were proposed for this distance range. In order to keep the results presented here consistent with Kongsberg seismograph station data (Bungum et al find a Moho depth of 34km at Kongsberg - see Figs 1 and 6), preference should be given to model 1 in Fig. 10. In this model the Moho becomes slightly shallower as it approaches the margin of the Oslo Graben from the southwest. After a distance of 3 10 km along the profile it appears to sink to 36 km from where, over a distance of 180 km, it rises to a depth of approximately 35 km beneath the NORSAR array. This Moho depth is in general agreement with Berteussen's (1977a) Moho depth values which are based on spectral ratio analysis of long-period P-waves recorded at NORSAR. On the other hand, model 2 of Fig. 10 resulted from an attempt to correlate the strong secondary arrivals in the last 100 km of profile, interpreted as multiple surface reflections of the Pg, PMP and P, phases. The travel-time delays in the crust, which are required for the

20 752 B. R. Cassell et al. correlation of these multiple phases, suggest a decrease in mean crustal velocity by as much as 0.3 km s- and an elevation of the Moho by 5 km in comparison with model 1. Still, the synthetic section (Fig. 12) for model 2 was not particularly successful in matching the amplitudes of multiple phases, and the possibility remains that the phases in question have not been correctly identified. So, in order to confirm with previous investigations, we adopt model 1 as representative for the main profile. No evidence is found for significant discontinuities within the crust, for instance like those found by Mykkeltveit (1980) for the Arsund-Otta profile (Fig. 1) transecting the M$re Gneiss Region. The irregular variations in the P, travel-time branch of the profile along the margin of the Oslo Graben could conceivably be caused by rays travelling along paths which deviate from the vertical plane between source and receivers. This possibility, however, could not be confirmed as there are no pronounced amplitudes in the available horizontal component records. Small-scale travel-time variations in the P, phase have been ignored as it is not certain whether they originate in Moho undulations or in lateral crustal variations. Also, the lack of identifiable secondary phases in the first few seconds after the P, arrivals throughout the latter half of the main line records suggests that there are no major sub-moho discontinuities beneath the central 300km of the profile. Two models (Fig. 14) were proposed for the Oslo Graben from records obtained along a separate leg. Both of these models account for the 0.8 s observed decrease in travel time of the P,, phase relative to the main line further west, but we do not have sufficient data to constrain our interpretation. Nevertheless, a thinning from the Precambrian to the west into the Graben is evident from our data. Despite the high quality of the data and the relatively dense station spacing along the CANOBE line it is not possible to develop a unique model without additional data. In order to obtain better crustal control it would be necessary to have additional shotpoints along the profile and reversed coverage. This is crucial in the presence of considerable lateral crustal variations and applies especially to the Oslo Graben area. This general area has been subjected to several lithospheric 3-D inversion experiments using teleseismic data recorded at NORSAR, and lateral variations in the crustal velocity have been found which roughly correspond to the graben surface contours (Christoffersson & Husebye 1979). Specific results on Moho depth variations near and inside the Graben, however, are hardly obtainable by such experiments. Apparently, reflection profiling of the COCOW type with sufficient resolution for short wavelength structural details may offer the only means to unravel the geological complexities of this area. Acknowledgments We wish to thank all the participants who, by their effort, contributed to the success of the field part of the CANOBE project. Funds were made available by an H. 0. Wood grant (administered by the Carnegie Institute of Washington), the Norwegian Petroleum Directorate, Statoil and the Royal Norwegian Council for Scientific and Industrial Research. We are grateful to our colleagues aboard the NERC vessel RRS John Murray for firing the N shots, and we are indeed indebted to the Royal Norwegian Navy for handling every aspect of the explosion programme for the H shots. The successful execution of the explosion programmes by both ships and crews involved contributed greatly to the success of the CANOBE experiment. References Aki, K., Christoffersson, A. & Husebye, E. S., Determination of the three-dimensional seismic structure of the lithosphere, J. geophys. Res., 82,

21 S, Norway seismic profile 753 Berckhemer, H., Eine Magnetbandapparatur fuer seismische Tiefensondierung, J. Geophys., 36, Berteussen, K.-A., 1977a. Moho depth determinations based on spectral ratio analysis of NORSAR long period P-waves, Phys. Earth planet. int., 15, Berteussen, K.-A., 1977b. Direct measurements of crustal P-velocities in the NORSAR area, fire appl. Geophys., 115, Bungum, H., Pirhonen, S. E. & Husebye, E. S., Crustal thicknesses in Fennoscandia, Geophys. J. R. astr. SOC., 53, Cassell, B. R., Synthetic seismograms in laterally varying media and comparison with observations in southern Norway, PhD thesis, University of Cambridge. Cassell, B. R., A method for calculating synthetic seismograms in laterally varying media, Geophys. J. R. astr. Soc., 69, Cerveny, V., Molotkov, I. A. & Psencik, I., Ray Method in Seimology, Univerzita Karlova, Praha. Cerveny, V. & Ravindra, R., Theory of Seismic Head Waves, University of Toronto Press. Choy, G. L. & Richards, P. G., Pulse distortion and Hilbert transform in multiply reflected and refracted body waves, Bull. seism. SOC. Am., 65, Christoffersson, A. & Husebye, E. S., On 3-D inversion of P-wave time residuals: Option for geological modeling, J. geophys. Res., 84, Fuchs, K. & Miiller, G., Computation of synthetic seismograms with the reflectivity method and comparison with observations, Geophys. J. R. astr. Soc., 23, Husebye, E. S., England, P. C. & Ramberg, I. B., The ideal-body concept in interpretation of the Oslo Rift gravity data and their correlation with seismic observations, in Tectonics and Geophysics of Continental Rifts, pp , eds Ramberg, I. B. & Neumann, E.-R., Reidel, Dordrecht. Kanestrdm, R. & Haugland, K., Profile section 3-4, in Deep Seismic Sounding in Northern Europe, pp , ed. Vogel, A,, Swedish Natural Science Research Council, Stockholm. Kanestrdm, R. & Nedland, S., Crustal structure of southern Norway: a reinterpretation of the 1965 Seismic Experiment. Publication No. 117 in the Norwegian Geotraverse Project, in Seismic investigations of the Crust and Moho in Southern Norway, ed. KanestrQm, R., University of Bergen, Seismological Observatory. Mykkeltveit, S., A seismic profile in southern Norway, Pure appl. Geophys., 118, Mykkeltveit, S. & Cassell, B. R., Crustal structure in the North Sea near Lista, southern Norway, in preparation. Ramberg, I. B., Gravity interpretation of the Oslo Graben and associated igneous rocks, Norg. geol. Unders. No Ramberg, I. B. & Smithson, S. B., Geophysical interpretation of crustal structure along the southeastern coast of Norway and Skagerrak, Bull. geol. Soc. Am., 86, Sellevoll, M. A. & Warrick, R. E., A refraction study of the crustal structure of southern Norway, Bull. seism. Soc. Am., 61, Troitskiy, P., Husebye, E. S. & Nikolaev, A,, Lithospheric studies based on holographic principles, Nature, 294, Tryti, J. & Sellevoll, M. A,, Seismic crustal study of the Oslo Rift, &re appl. Geophys., 115, Weigel, W., Hjelme, J. & Sellevoll, M. A., A refraction profile through the Skagerrak from northern Jutland to Southern Norway, Geod. hst. Meddr. Kbh., 45, 1-28.