Teleseismic receiver functions from a ten station network deployed in northeast India region sampling the Shillong plateau, Mikir Hills, Himalayan foredeep and the Himalayan convergence zone, are analyzed to obtain the crustal structure in this seismically active but less studied region. The Shillong plateau and Mikir hills, away from the convergent margins, reveal remarkably simple crust with thickness (∼35 km) and Poisson's ratio (∼0.25), akin to the Indian shield values. A surprisingly thin crust for the uplifted Shillong plateau may be explained invoking presence of an uncompensated crust that popped up in response to tectonic forces. In contrast, crustal signatures from Assam valley suggest a thicker crust and higher Poisson's ratio with evidences for a dipping Moho. Predictably, the crust is much thicker and complicated in the eastern Himalaya further north, with values in excess of 50 km.
 The northeast India is an acknowledged region of tectonic complexity and diversity that mainly stems from its evolutionary history and subsequent modifications to its structure imposed by an ongoing collision in the North and the Indo-Burmese convergence to the East. This is indeed testified by the surface geology and sustained seismicity. The region is comprised of distinct geological units, i.e., the Himalayan frontal arc, the highly folded Indo-Burman mountain ranges, the Brahmaputra alluvium in the Assam valley, the Shillong plateau and the Mikir hills (Figure 1). The Shillong plateau, believed to be uplifting even at present is the site of the 1897 Great Assam earthquake of M > 8.0 [Oldham, 1899]. The Assam valley is a narrow E-W trending feature, filled up by sediments brought by the Brahmaputra river system.
 The prominent geological units of the northeast India region remain geophysically less studied owing to the inaccessibility of this terrain. Studies related to the crust-mantle configuration of this region are largely from travel time modeling [De and Kayal, 1990; Baruah, 1995; Sitaram et al., 2001]. Tomographic images of the upper mantle in Shillong plateau and Assam valley regions reveal presence of lateral velocity inhomogeneities and low velocity layers in the crust [Kayal and Zhao, 1998]. The depth to the Moho in the foredeep region is found to vary between 42–45 km [Baruah, 1995; Sitaram et al., 2001]. However, the Moho depth in this region remains largely uncertain due to absence of well constrained Moho phases besides the usual ambiguities associated with travel time modeling.
 Knowledge of crust-mantle structure in this region is scarce and assumes importance in view of its complexity and proximity to two convergent margins. Earthquakes in the Shillong plateau occur at enigmatically deep depths, with a few events close to 50 km, in the upper most mantle, [Chen and Molnar, 1990; Kayal, 2001] suggesting presence of a cold mantle in this region. A question of topical interest is whether the Shillong plateau and Mikir hills, now separated by the NW-SE trending Kopili lineament, were contiguous terrains in the geologic past? Continuity of shield-like crust-mantle structure in the foredeep region [Gupta et al., 1977] is another topic of geo-tectonic interest.
 In the present study, the receiver function technique is applied to teleseismic waveforms in the epicentral distance range of 30–100°, from a ten station network in northeast India (Figure 1 and Table 1), to obtain estimates of crustal thickness and its average composition. With the exception of a permanent station at SHL operated by the India Meteorological Department, all the other stations are temporary and deployed by the National Geophysical Research Institute. All the stations except ZIR are equipped with broadband sensors.
Table 1. Station Locations With Their Tectonic Affiliation, the Number of Receiver Functions and the Average Vp Used to Derive the Crustal Parametersa
The beginning of operation of the stations is indicated adjacent to the station codes. For the Himalayan stations the zM and σ values are given over a range since they cannot be constrained owing to absence of multiples, though the Pms phase is well observed. Note that the values for station SJA are poorly resolved due to bad quality of Pms. At stations TEZ and JPA the results are relatively ambiguous due to possible deviations from 1D structure.
35 ± 1
0.25 ± 0.01
35 ± 1
0.23 ± 0.01
34 ± 4
0.25 ± 0.02
35 ± 3 (41)
0.29 ± 0.02 (0.25)
35 ± 3 (41)
0.28 ± 0.02 (0.25)
2. Receiver Functions
 The method we follow to obtain receiver functions involves rotation of the Z, N and E components into a ray coordinate system. This essentially decomposes the wave field into its P, SV and SH components. The converted phases are isolated from the coda by deconvolving the P from the SV component by simple spectral division in the frequency domain using a water level stabilization. In addition, low pass filtering with a gauss function limits the frequency band to enhance the desired signals, especially from deeper interfaces. In this study, we use a gauss value of 2 Hz. In order to make receiver functions with different slowness values comparable, and to distinguish multiples from converted phases, a moveout correction is applied separately for the converted phases and multiples. This correction is made with reference to a slowness value of 6.4s/°. The receiver functions for stations sorted by slowness/back azimuth are shown in Figure 2. The P-to-s conversion times from the Moho (Pms) picked from the stacks after moveout correction for converted phases are listed in Table 1 for all the stations. The receiver functions are also moveout corrected for multiples, to enable picking of the delay times of the first (Ppms) and second (Psms) multiples from the Moho. These times for stations at which they are clearly observed are shown in Table 1. Using all the observed receiver functions at a station, the Moho depth (zM) and Poisson's ratio (σ) are estimated by performing a grid search over a wide range of zM and σ values, to find the optimal values that maximize the summation of amplitudes of the converted and multiple phases [Zhu and Kanamori, 2000]. However, we refrain from applying this method to stations where the multiples are poorly registered. In such cases the P-to-s conversion times alone were used to obtain the crustal thickness values using either a range of Poisson's ratios or a normal value of 0.25, assuming appropriate Vp values (see Table 1). It may be noted that the estimates of Poisson's ratio have very small dependence on the assumed Vp values unlike the Moho depths that are directly related to it.
3. Crustal Structure
3.1. Shillong Plateau and Mikir Hills
 It can be seen from Figure 2a that the stations on the Shillong plateau and Mikir Hills situated away from the convergent margins, depict simple waveforms with excellent Moho conversions and clear multiples. The results of the grid search method yield Poisson's ratios of 0.25, 0.23 and 0.25 for NGB, SHL and DMK respectively (Figure 3), using an average Vp of 6.4 km/s obtained from earlier studies [De and Kayal, 1990; Baruah, 1995]. At HMN, a Moho conversion time of 3.85s translated into depth for a normal crustal composition (σ = 0.25), results in a crustal thickness of 33 km.
 A surprisingly thin crust (∼35 km) for the uplifted Shillong plateau can be explained invoking presence of an uncompensated crust (evidenced by a +100 mgal isostatic anomaly) that popped up in response to tectonic forces [Rao and Kumar, 1997] and subsequently got eroded. Interestingly, detailed modeling of the anomalous gravity field under the plateau yields crustal thickness values in the range 41–46 km [Verma and Mukhopadhyay, 1977]. Though the model with a thicker crust seems to explain the gravity anomalies better, these authors concede that due to possible absence of isostatic equilibrium, the gravity observations beneath Shillong plateau could be related more to presence of a high density crust rather than an anomalously thick one, lending support to our estimate of a thinner crust.
 The estimates of Moho depth and Poisson's ratio for this region are similar to those found for the Indian shield [e.g., Kumar et al., 2001], supporting the northeastern extension of the Indian shield like crust as suggested earlier from surface wave analysis restricted to the Gangetic plains [Gupta et al., 1977]. Earlier estimates of Moho depth in Shillong Plateau and Mikir hills region from travel time analysis reveal a much thicker crust in the range of 41.5 km to 54 km [De and Kayal, 1990; Baruah, 1995]. Given the presence of lateral heterogeneity in this region, as evidenced in the tomographic images [Kayal and Zhao, 1998], the observed variance in travel time estimates and receiver function results is understandable. In addition, results from a local network operated on the Shillong plateau [Kayal, 2001] have delineated zones of intense seismic activity constrained within 10–30 km and mild activity restricted to 40–50 km depth, with a possible aseismic layer in between. The transition zone at 30–40 km could be due to change in rheology marking the crust mantle boundary. This observation further supports a 35 km crust beneath the plateau.
 It is interesting to observe that HMN on Shillong plateau and DMK on Mikir hills, separated by the Kopili lineament, besides sharing similar crustal thickness values seem to be underlain by similar velocity structure. Their underlying geology is also same and they share a common Archaean basement as other parts of the Indian shield. These results support an existing geological hypothesis [Nandy, 2001] that the Shillong plateau and Mikir hills were probably together in the geologic past.
3.2. Assam Valley
 In the Assam valley, stations TEZ and JPA show clear P-to-s conversions from the Moho, traceable close to 5s (Figure 2a). This delay time when converted into depth corresponds to 41 km (Table 1), similar to the results from earlier seismological studies [Baruah, 1995; Sitaram et al., 2001]. For station TEZ, results from the grid search method yield an abnormal Poisson's ratio of 0.29 (Figure 3), in conformity with presence of multiples close to 15s and 20s (Figure 2a). The lack of strong multiples, variation of Pms with back-azimuth (Figure 2b) and presence of considerable energy on the transverse component together suggest presence of dipping interfaces and/or crustal anisotropy. Since a dipping Moho would considerably influence the estimates of the crustal parameters, 2D modelling of receiver functions would lead to better estimates. However, accepting a value of 0.29 would drastically reduce the crustal thickness to 35 km, for the observed Moho conversion time. It is interesting to note that in the vicinity of Tezpur, at least three low velocity zones (LVZ's) in the crust were revealed in the tomographic images that have been well correlated with prolific seismic activity [Kayal and Zhao, 1998]. If these LVZ's are due to fluid filled zones that are common in a seismogenic crust, a high σ value of 0.29 seems tenable. Further, a Poisson's ratio of 0.28 found at JPA (Figure 3) argues in favour of a higher than normal σ value for the Assam valley. Moreover, a crustal thickness of ∼35 km obtained using this high σ value, is in excellent agreement with the results from gravity modeling which indicate a crustal thickness in the range 31.4–33.7 for a mean density of 2.84 g/cc [Verma and Mukhopadhyay, 1977].
 Of the stations in the sub-Himalaya, BKD and SJA in the foothills of the Himalayan convergence zone are sited on loose sediments. At these stations, the Moho conversions are poor (Figure 2), in spite of the large number of good quality waveforms used. The Pms times at these stations are close to 7s and 9s respectively. In addition, large conversions from the base of the sediments are seen close to 1s for BKD and 1.6s for SJA (Figure 2a). Receiver function stacks weighted by back-azimuth (Figure 2b) bring out these phases in a more convincing manner. Contrastingly, stations RUP and ZIR, where the shallow sedimentary layer is absent (as evident from absence of significant conversions prior to Pms) show clear Moho conversions at 5.70s and 4.75s, devoid of multiples. At SJA and BKD reverberations due to sediments seem to mask the Moho conversions. Absence of strong Moho conversions could also be due to other reasons that include a highly disturbed crust owing to the stations lying within the collision zone, a gradational Moho and/or reduced mantle velocities immediately following the Moho.
 The crustal thickness values for the sub-Himalayan stations close to the Main Boundary Thrust (MBT) (Table 1) show a large variation. While the estimate at SJA remains largely uncertain due to an ambiguous Pms phase, stations BKD, RUP and ZIR indicate values centered on 53, 45 and 37 km respectively for a Poisson's ratio of 0.27 (Table 1). These values are close to an estimate of 49 km along the MBT [Sitaram et al., 2001] in the study region. Further, their study also revealed presence of a layer, interpreted as a boundary between the upper and lower crust, at a depth of 25 km in the MBT region. Receiver functions do lend support to this observation in terms of a converted phase from an intra-crustal boundary (Figures 2a and 2b) seen at 3.47s at BKD, 2s at RUP and ZIR. Converted into depths, these times correspond to 27 km and 16 km respectively, for a Poisson's ratio of 0.25 and Vp of 6.0 km/s for the upper crust.
 The current uplift of the Shillong plateau has been attributed to compressive forces due to Himalayan collision in the North and Burmese subduction in the East, resulting in a pop up of the plateau along pre-existing steep faults [Rao and Kumar, 1997]. A variant of this mechanism invokes repeated failures due to great earthquakes like the Assam earthquake of 1897 along a south dipping fault resulting in uplift [Bilham and England, 2001]. A thin crust beneath the elevated Shillong plateau together with a normal Poisson's ratio, although apparently surprising, can be explained by an uncompensated crust that it is merely uplifting in response to the prevailing tectonic forces while preserving its shield like character. Contrastingly, in the Assam valley, the receiver functions show indications of a dipping Moho with weaker multiple energy. This observation suggests that the Indian plate might begin to bend in this region before it underthrusts the Himalaya at a shallow angle. The dipping nature of the Moho coupled with weaker multiples could result in ambiguous crustal parameters shown in Table 1 for stations TEZ and JPA, in spite of large amount of good quality data. While a thinner crust due to a high Poisson's ratio of 0.29 agrees with the gravity observations, the thicker (41 km) alternative matches with the estimates from travel time modelling [Sitaram et al., 2001] and fits into the general framework of northward thickening of the crust with values reaching 65 ± 5 km beneath southern Tibet [Huang et al., 2000]. In this context, the crustal thickness values in the sub-Himalaya though variable, are generally larger than 45 km commensurate with the models of thickening in a collision environment. The local variations in crustal parameters may partly be due to block movements dominated by vertical tectonics, which is generally attributed to isostatic behaviour, in the backdrop of large scale thrusting.
 The financial support from Department of Science and Technology, India is gratefully acknowledged.