SEISMIC VELOCITY STRUCTURE AND SEISMOTECTONICS OF THE SHILLONG MASSIF, NORTH EASTERN INDIA

Abstract

Shillong massif in North Eastern India is an element of a complex tectonic and geologic province. It is essential to under stand the structure and seismicity of each part of this province to unravel the active tectonic processes operating in this region as a whole. In order to understand the structural fabric of the massif and the tectonic processes which are presently active there the following analyses were carried out. 1) The upper crustal velocity structure of the Shillong massif was determined using the microearthquake data recorded by a short aperture local seismic network. 2) The microearthquakes were relocated using the above velocity model and the new epicentral map shows better definition of seismically active zones. The composite fault plane solutions were determined for different spatial clusters of these microearth quakes. 3) The resultant seismicity, fault plane solutions and velocity structure as well as other available geological and geophysical informations were used for proposing a tectonic model of the Shillong massif and further evidence on the possible focal mecha nism of the 1897 earthquake. The choice of a method for velocity analysis depended upon the limitations imposed by the small number of recording stations (=5) in the array. Initially we tried to use the simultaneous inversion technique developed by Pavlis and Booker (1980) and Spencer and Gubbins (1980). However, the results of this analysis became unstable for our data. We, therefore, resorted to a more traditional approach of using Wadati (1933) and Riznichenko (1958) methods in conjunction. This was originally a graphical method of determination of velocity. However, we automated it to take advantage of the computers. We further examined the avail able methods of inversion of the results of the analysis with Riznichenko method to obtain the layer velocities and thicknesses of the medium. They are the backstripping method of Nicholson and Simpson (1985) and the Bune's method (Bune et al. , 1960). A modified version of the Bune*s method developed by us gives reasonably accurate estimates of the velocity structure of the medium. Detailed examination of these methods were carried out with the help of simulated data to check the following: a) Relationship of the velocity obtained by Riznichenko method with the actual velocity structure. b) Accuracy, stability and sensitivity of the backstripping and the modified Bune's methods. c) Resolving power of the array, i.e. whether special type of velocity structures like low velocity layers and thin layers can be resolved using the available array configuration. It was concluded that the modified Bune's method gives reasonably accurate (s within 2%) estimate of crustal velocity structure even with data from a small array. The phase data for a total of 1138 earthquakes were read. Out of these 670 could be located. For 112 of these earthquakes, both the P and the S phase readings from all the 5 stations were

available with reading accuracy better than 0.2 and 0.5 sec respectively. Moreover these 112 earthquakes occurred within the array. Therefore, the epicentral location, which were required in Riznichenko (1958) method, could be estimated sufficiently accu rately even when the model used for obtaining these locations was different from the one finally discovered. These 112 earthquakes were used for velocity analysis. The interpreted velocity model of the upper crust of the Shillong massif consists of a relatively homogeneous top layer of thickness of 11.5 km with the P and the S wave velocities of 5.9 and 3.4 km/sec respectively. The P to S wave velocity ratio and the Poisson's ratio for this layer are 1.71 and 0.24 respec tively. The scatter in the derived vertical travel time versus depth in 11.5 km to 26 km depth range shows that the material in this range is considerably heterogeneous. An analysis of the observed data as well as model simulations allows the conclusion that the heterogeneity under the array can be explained in terms of a stack of layers having low and high velocity material. The velocity and thickness of individual layers could not be resolved using the available data set. However, this does not constitute a unique explanation. Significant amount of lateral heterogeneity and/or dipping layers may be alternative models to explain the observations. The average P and S wave velocities of this mate rial were estimated to be 6.3 and 3.5 km/sec respectively. The P to S wave velocity ratio and the Poisson's ratio were 1.79 and 0.27 respectively. The bulk (K) and the shear (y) moduli for the two layers were estimated using the average density of the top 45 km as estimated by Verma and Mukhopadhyay (1977) and for the density of material as expected from the relation between velo city and density as given by Ludwig et al. (1970) and Dobrin (1976). The microearthquakes (670 in number) were relocated using the above estimated velocity model. Below 26 km we used the layer velocities estimated by Kharshiing (1985) for locating the earth quakes occurring below that depth level. There was a definite improvement in the location in comparison to that obtained by an initial assumed velocity model as the root mean square error (errog) in travel time decreased considerably. For example, the percentage of earthquakes with erros less than 0.25 sec increased from about 31% to about 50% when the microearthquakes were relo cated using the revised velocity model. They show a complex pattern of seismicity of which an L-shaped zone of intense acti vity is the most prominent. The N-S section of this zone encom passes the surface trace of the Dudhani fault. The E-W trend parallels that of the Dauki fault and is offset from it by an amount of 25-30 km toward north. There are a number of other linear trends of epicenters observed in this area. They trend in NW-SE and NE-SW directions respectively. The depth sections show that barring one or two clusters the microearthquakes are usually dispersed in depth and occur upto a depth of 45 km. However, most of the earthquakes occur in the top 0-14 to 0-25 km which leads us to conclude that the base of the seismogenic zone as defined by Anderson et al. (1983) and Sibson (1986) lies within the 14-25 km depth range in the massif. As vi large magnitude earthquakes usually occur at the base of the zone of crustal earthquakes we come to the conclusion that the 1897 great Assam earthquake occurred within this depth range. A number of composite fault plane solutions for various epicentral clusters were drawn. Although the solutions show both thrust and strike-slip mechanism their P axis orientations are remarkably similar. The P axes for all of them have a strike roughly in N-S direction and plunge of less than 20*. This obser vation matches with the P axis orientation for earthquakes near the southern side of the massif as obtained by Chandra (1981). The nodal planes have steep dip (>30*) and strike in E-W direc tion for thrust mechanism and NW-SE for strike slip mechanism. We augmented the microearthquake data with the seismicity data reported by the National Oceanographic and Atmospheric Administration (NOAA) in the Earthquake Data Reports (EDR) and with other microearthquake data obtained by the Department of Earth Sciences, University of Roorkee during 1979 and 1982 (Khattri et al., 1988, 1989a). The resultant map of seismicity, fault plane solution, the estimated velocity structure as well as other available geophysical and geological informations were used for proposing a tectonic model of this region. On the basis of this analysis we propose that the Shillong massif is overthrusting toward south over the Bengal basin along a hidden fault. The overthrusting occurs along a shallow dipping thrust fault hidden below the massif. During 1897 rupture along this plane caused the great Assam earthquake. That is why no vii primary surface fault that can account for this earthquake is observed. We propose that movement along this plane causes the overly ing material containing the massif to buckle upward. This may explain the coseismic uplifting of the massif during this earth quake inferred by Molnar (1987) from analysis of Oldham's (1899) observations. It may also explain the horst like uprising of the massif inferred from geological evidences (Murthy, 1970; Auden, 1972; Desikachar, 1974; GSI, 1974). The steep nodal planes obtained by the fault plane solutions and the observed steeply dipping small faults in this region are proposed to have formed due to upward buckling of the massif. They may also be splay branches of the main thrust plane underlying the massif. The diffused pattern shown by microearthquakes seen in the massif may be explained as caused by minor movements along these faults. The dominant tectonic movement occurs in N-S direction resulting in a N-S orientation of the maximum compressive stress axis (P axis). Most probably the main thrust plane lies within the depth range 14-25 km. Increase in heterogeneity in 11.5-26 km depth compared to the top 11.5 km is possibly the effect of fold ing and thrusting in that zone. That is, we propose that such heterogeneity is the consequence of the tectonic activity in that zone.