Imprints of volcanism in the upper mantle beneath the NW Deccan volcanic province

Continental flood basalt provinces are classic examples of voluminous outpouring of lava over large areas within short durations of time. Conventionally, continental flood basalts are linked to mantle plumes, primarily based on tracking the time-progressive chain of volcanic ridges connecting a continental flood basalt with a hotspot (Morgan, 1972). However, alternate views attributing the generation of flood basalts to shallow upper-mantle processes exist (Anderson, 1994, 2005; Foulger and Natland, 2003). The Deccan volcanic province of India, with its large, well-preserved outcrop area, represents an excellent example of a continental flood basalt province (Fig. 1) for which the genesis is largely attributed to the passage of the Indian plate over the Reunion hotspot. The plume-lithosphere interaction is understood to have (1) resulted in the separation of Seychelles from western India at 65 Ma, causing widespread magmatism in west-central India, and (2) left a trail of volcanic ridges leading all the way to the hotspot (Morgan, 1972). On a global scale, seismic evidence for mantle plumes, including Reunion, comes in the form of mapping low-velocity conduits in the upper mantle (Zhao, 2007), but similar evidence is not so obvious for continental flood basalts. Thus, unequivocal seismic evidence for a plume-lithospheric interaction leading to thermal and/or compositional anomalies in the lithosphere beneath the Deccan volcanic province has been hitherto elusive. Numerical modeling suggests that thermal anomalies can be preserved up to 200 m.y. (Campbell, 2007), although Ziegler and Cloetingh (2004) estimated that after 60 m.y., ∼65% of a deep-seated thermal anomaly would be dissipated. Based on geochemical analysis of mantle xenolith–bearing alkalic lavas, Sen et al. (2009) proposed a thin and metasomatized lithosphere beneath Kutch in the northwestern Deccan volcanic province (Fig. 1). Thus, the compositional and thermal imprints of the 65 Ma Deccan event can be expected to be still preserved in the upper mantle beneath the Deccan volcanic province. Regional-scale broadband teleseismic receiver function studies in the south-central Deccan volcanic province (Fig. 1), which experienced the main effusive phase of Deccan volcanism, revealed the absence of anomalous signatures associated with a mantle plume in the upper mantle beneath the south-central Deccan volcanic province (Kumar and Mohan, 2005). In contrast, a regional P-wave tomography study (Kennett and Widiyantoro, 1999) provided evidence for possible imprints of a plume head in terms of an ∼200-km-wide cylindrical zone of low velocities, which extend from shallow depth down to a more extensive low-velocity zone deeper than 200 km in the upper mantle beneath northern Cambay (north of the northwestern Deccan volcanic province), centered around the Mer Mundwara alkaline complexes (Fig. 1). However, the P-wave tomograms lose resolution in the northwestern Deccan volcanic province encompassing the Saurashtra horst, the Cambay and Kutch rifts, the western part of the Narmada rift, and offshore the Gulf of Cambay. Thus, the northwestern Deccan volcanic province is a seismically uncharacterized area between two regions exhibiting contrasting seismic signatures.

A plume-lithosphere interaction is expected to result in continental stretching, lithospheric thinning, and thermal and compositional changes leading to reduction in the seismic velocities of the upper mantle (White and McKenzie, 1989; Anderson et al., 1992) and apparent deepening of the upper-mantle discontinuities. These mantle discontinuities correspond to phase transitions of mantle minerals and are sensitive to temperature, composition, and mantle fabric (Bina and Helffrich, 1994). The 410 km and 660 km discontinuities have positive and negative Clapeyron slopes, respectively, and so they are anticorrelated, providing a diagnostic of the temperature conditions in the mantle transition zone. The thickness of the mantle transition zone could be reduced by ∼20–30 km for an increase in temperature of ∼200 °C (Helffrich, 2000). Mineralogical investigations have revealed that variations in the iron (Irifune and Isshiki, 1998) and aluminum (Weidner and Wang, 1998) content in the mantle, as well as the presence of hydrogen (Yusa and Inoue, 1997; Yusa et al., 2000), can affect the depth, impedance contrast, and sharpness of the seismic discontinuities (Gilbert et al., 2003). In regions with a normal mantle transition zone, variations in the depths to the mantle discontinuities may be attributed to lateral variations in the seismic velocities/composition in the shallow mantle above 410 km discontinuity (Chevrot et al., 1999). The time lags of the P₄₁₀s and P₆₆₀s mantle phases show strong correlation with upper-mantle structure above 410 km depth (Bostock, 1996). A thick cold lithosphere with a high-velocity root would result in shorter time lag, while lithospheric thinning or low velocities in the shallow upper mantle would be expected to increase the time lag. Mapping of the mantle discontinuities would therefore provide important clues toward understanding mantle structure, which has a bearing on the mechanism for the eruption of the Deccan basalts. We investigated and compared the upper-mantle structure beneath the northwestern Deccan volcanic province and south-central Deccan volcanic province through receiver function analysis in an attempt to understand the genesis of the Deccan volcanism.

The Deccan basalts occupy an area of 1.5 × 10⁶ km² in the central and western parts of India and the adjoining Arabian Sea. The northwestern Deccan volcanic province hosts the basaltic outcrop of the Saurashtra horst bounded by the east-west–trending Kutch rift in the north, the NNW-SSE–trending Cambay rift to the east, and the east-west–trending Narmada rift to the southeast (Fig. 1). The rifts in NW India formed due to reactivation of faults of Precambrian age, as a consequence of the northward drift of the Indian plate after its breakup from the Gondwana supercontinent in Late Triassic–Early Jurassic (Biswas, 1987). The Cambay basin is a 400-km-long, NNW-SSE–trending graben, with an average width of 50 km, that consists of a 2- to 4-km-thick complete sequence of Tertiary sediments overlying the Tertiary Deccan basement, below which lie Mesozoic rocks (Chowdhary, 2004). The Kutch rift is filled with Tertiary–Holocene sediments, within which occur horsts built of Mesozoic and Tertiary sedimentary rocks intruded by alkalic and lamprophyre bodies. The Deccan Traps of the northwestern Deccan volcanic province are predominantly composed of tholeiitic basalts with minor volumes of alkalic, felsic, and ultramafic intrusive rocks, which are exposed mostly along the rift zones in the northwestern Deccan volcanic province. The Mer Mundwara alkaline complexes near North Cambay are perceived to represent the earliest pre-Deccan phase of volcanism, based on ³He/⁴He ratios (Basu et al., 1993). Rare mantle-derived xenoliths along with alkalic and tholeiitic intrusions are found in the Kutch basin (Sen et al., 2009).

The crustal structure of the Deccan volcanic province has been well studied through several deep seismic sounding profiles cutting across the province (Kaila and Krishna, 1992). The average crustal thickness and P velocity in the province is observed to be 38 km and 6.4 km/s, respectively (Kaila and Krishna, 1992). The deep seismic sounding profile data acquired earlier (Kaila and Krishna, 1992) was reprocessed, remodeled, and reinterpreted (Rao and Tewari, 2005; Dixit et al., 2010) to yield a more precise structure along two profiles (Fig. 1), a 160-km-long, E-W profile from Navibunder to Amreli, cutting across the northwestern Deccan volcanic province in Saurashtra (Rao and Tewari, 2005), and a 230-km-long, N-S profile from Mehmadabad to Billimora, along the Cambay rift and across the Narmada rift (Dixit et al., 2010). In the reinterpreted section, the crustal thickness is found to vary from 37 km to 33 km from west to east in the northwestern Deccan volcanic province (Rao and Tewari, 2005) and from 31 to 33 km in the Cambay rift to 37 km farther south of the Cambay rift (Dixit et al., 2010). The crustal thickness in the south-central Deccan volcanic province is largely constrained to be 38 km through two roughly 200-km-long, E-W deep seismic sounding profiles near Koyna (Fig. 2B), ∼200 km south of Mumbai (Kaila and Krishna, 1992).

From 2004 to 2009, the Indian Institute of Technology Bombay (IITB) has instrumented the northwestern Deccan volcanic province in a phased manner through a portable network of broadband stations (Fig. 2A). In this study, the data from the five IITB stations were supplemented by 11 yr of data (1997–2008) recorded by the permanent broadband station at BHUJ, operated by the India Meteorological Department. Three stations are located on the west coast of Saurashtra, while the other three stations were deployed to the west of the Cambay rift in order to image the upper mantle beneath the northwestern Deccan volcanic province, in view of the fact that the usable earthquake sources are mostly located along eastern azimuths (Fig. 2A).

About 1000 high-quality teleseismic waveforms in the distance range 30°–95°, recorded by six broadband stations (Fig. 2A), were used in this study. On an average, nearly 100 waveforms were used at each station, with the exception of station MANK. The station codes and the number of waveforms used at each station are BHUJ—363, SONT—195, TANA—136, KAND—125, MOKR—113, and MANK—51. The locations of the piercing points (Fig. 2A) at 520 km depth provide an estimate of coverage of the study region. In general, the piercing points at 410 km and 660 km are offset by 118 km and 216 km, respectively, from the stations, for an event at an epicentral distance of 67°. The events recorded at BHUJ, MOKR, and MANK stations have travel paths that sample the mantle beneath the Kutch rift and the Saurashtra Peninsula, while those recorded by stations KAND, TANA, and SONT in eastern Saurashtra have travel paths that sample offshore the Gulf of Cambay, the Cambay and Narmada rifts, and a segment of the Aravalli craton and the south-central Deccan volcanic province.

About 1400 high-quality waveforms recorded at the 11 station broadband network in the south-central Deccan volcanic province (Fig. 2B) were used to construct a revised composite image for the south-central Deccan volcanic province for comparison with the northwestern Deccan volcanic province. In a prior study, the south-central Deccan volcanic province was imaged using 900 waveforms from six stations (Kumar and Mohan, 2005); we supplement this imaging with the addition of over 500 waveforms primarily from the permanent stations PUNE, KARD, BOM, and the Koyna network (Fig. 2B). The piercing point plot (Fig. 2B) indicates good coverage for resolving the deep structure beneath the south-central Deccan volcanic province.

The receiver function technique is popularly used to explore the crustal and upper-mantle structure beneath a receiver site (e.g., Vinnik, 1977; Burdick and Langston, 1977; Langston, 1979; Owens et al., 1984; Kind and Vinnik, 1988; Zandt et al., 1995). Teleseismic P waveforms recorded at three-component seismic stations contain information on the earthquake source, Earth's structure in the vicinity of both source and receiver, and mantle propagation effects. P waves propagating through the mantle and incident from below upon the crust and upper mantle beneath a seismic station generate P-to-s converted waves at discontinuities beneath the station as well as free surface reverberations associated with the discontinuities. Removal of the effects of source and mantle path from the teleseismic body waveforms produces the receiver functions, which are signals composed of primarily P-to-s conversions and associated multiple reverberations, which show the relative response of Earth's structure beneath the receiver. Waveforms of teleseismic earthquakes in the epicentral range of 30° to 95° were extracted and quality checked automatically using the vertical components to retain those with a signal-to-noise ratio (SNR) ≥ 2.5. Subsequently, the data were further sifted based on visual inspection of all the three components. The receiver functions were computed by first decomposing the observed wave field into its P, SV, and SH components. This involves rotation of the Z, N, and E components into a ray coordinate system using optimized back azimuths and incidence angles (Vinnik, 1977). Deconvolving the P component from the SV component through spectral division using a water-level stabilization produces the radial receiver function. The water level was globally set at 0.001 for all the traces after trials with several values. The Ps signals from the mantle discontinuities were enhanced by applying a Gaussian filter centered around 0.5 Hz. The time difference in the arrival of the direct P wave and the converted phase (Ps) for a range of slowness/incidence angle values enables us to estimate the depth to the causative discontinuity based on a velocity model. The Ps conversions that are sensitive to shear velocity contrasts have much stronger amplitude on the horizontal component than on the vertical. In order to qualitatively assess the layering beneath each station, the receiver functions in the slowness range 4 s/° to 9 s/° were binned, stacked, and plotted in intervals of 0.2 s/° with respect to the delay time. Because the converted phases and multiples exhibit an opposite moveout with respect to epicentral distance, a moveout correction for a reference epicentral distance of 67° was applied, which makes receiver functions at different slowness values comparable and enables us to distinguish converted phases from multiples. The summation trace was obtained by sorting the receiver functions by slowness (or back azimuth), smoothing the successive traces, and stacking.

The Ps converted phases generated at a subsurface interface exhibit variations in their time lags depending on the epicentral distance (or equivalently, slowness), where the Ps phases arrive earlier for larger epicentral distances and vice versa. With increasing conversion depths, the slope of the delay time curve becomes larger. In contrast, multiple phases arrive earlier when the slowness is larger. A moveout correction assuming all the energy is from conversions aligns the Ps conversions parallel to the P wave while the multiples are inclined. However, a moveout correction assuming that all the energy is produced by multiples would produce an opposite effect. In order to distinguish Ps conversions from multiples and also to effectively stack the receiver functions at different slowness values, moveout corrections were dynamically applied to the SV seismograms and were predicted accurately by assuming a one-dimensional reference model and a constant velocity ratio. Following the work of Yuan et al. (1997), who were the first to introduce the concept of moveout correction in receiver functions, the one-dimensional IASP91 model (Kennett and Engdahl, 1991) was used for this correction using a slowness value of 6.4 s/°. This chosen reference slowness corresponds to an epicentral distance of ∼67° in the IASP91 standard Earth velocity model, which is close to the median epicentral distance of the events used for receiver function analysis (30° to 95°). With increasing depth of conversion, the moveout increases, and the errors of stacking without moveout correction get larger, resulting in the disappearance of the Ps conversions from the 410 km and 660 km discontinuities in the summation trace. Stacking after moveout correction for Ps enhances the amplitude of the corresponding Ps conversions, which enables the depth to the causative interface to be estimated reliably.

The uncertainties in the time lags of the P₄₁₀s and P₆₆₀s phases observed on the receiver function summation trace were estimated through bootstrap resampling technique (Efron and Tibshirani, 1986; Dueker and Sheehan, 1998). In this scheme, a single receiver function summation trace is generated by summation of those traces decided by a random number generator, where the number of traces for summation remains the same as the total number of receiver functions. The trace numbers being randomly generated are not in a sequence but a string of numbers, where some numbers are repeated while the others are omitted to keep the total number the same. This procedure is repeated 200 times, by stacking a totally different sequence of traces corresponding to randomly generated trace numbers. The mean and standard deviations of the times of converted phases are then calculated from the values picked from the distribution of the 200 realizations. If most of the observed traces are coherent, then the standard deviation of the times is small and vice versa.

In total, ∼1000 receiver functions were computed to image the upper mantle down to a depth of 700 km beneath the northwestern Deccan volcanic province. To study the lateral variations in the mantle transition zone discontinuities and also the pervasive effects of volcanism, the receiver functions were analyzed individually at all the stations. To enhance the signals from the mantle discontinuities, the receiver functions were binned and stacked in narrow slowness bins for each station. The P₄₁₀s and P₆₆₀s phases were identified and timed on the summation trace obtained by summation of all the receiver functions corresponding to each station. Receiver function images plotted as a function of (1) slowness and (2) back azimuth for all six stations are shown in Figure 3. The receiver functions were averaged in slowness bins of 0.2 s/° and azimuthal bins of 20° to enable identification of the dominant trends in the variation of delays as a function of ray parameter and back azimuth. The P-to-s conversions from the mantle discontinuities are seen on the receiver function images (Fig. 3) at all six stations as distinct features linearly aligned and delayed relative to time lags of 44.1 s and 68.1 s predicted by the IASP91 standard Earth velocity model for a slowness value of 6.4 s/°. The delays are apparent in the summation traces of slowness and back azimuth at each station. However, the receiver function plots (Fig. 3) indicate variation in delay time and amplitude of both the P₄₁₀s and P₆₆₀s phases with slowness and back azimuth. The amplitude variation with back azimuth (Fig. 3) is primarily due to averaging of a large number of receiver functions in each azimuthal bin in the eastern azimuths (0°–180°) compared to the western azimuths (180°–360°), where the data are sparse, resulting in higher amplitudes. The bootstrap time lags of P₄₁₀s and P₆₆₀s at BHUJ, SONT, and TANA are 45.35 s ± 0.22 s, 68.96 s ± 0.09 s; 44.86 s ± 0.13 s, 68.51 s ± 0.52 s; and 44.98 s ± 0.14 s, 68.07 s ± 0.42 s, respectively. Similarly, delays of ∼1 s are observed for both the mantle phases at stations KAND (45.44 s ± 0.27 s, 69.16 s ± 0.51 s), MOKR (45.36 s ± 0.19 s, 69.27 s ± 0.37 s), and MANK (45.22 s ± 0.42 s, 69.20 s ± 0.31 s). Considering the bootstrap uncertainties, the P₄₁₀s values are delayed at all stations, while four out of six stations exhibit P₆₆₀s delays. The stations SONT and TANA exhibit reduced P₆₆₀s delays, possibly due to rays sampling the high-velocity upper-mantle region beneath the Aravalli craton in the northeast, which possibly masks the delays beneath these stations. Similar variation in delays with back azimuth are also observed at SONT and TANA, where the delays appear to be marginally increasing from NE (0°–40°) to ESE (80°–120°), corresponding to rays sampling the Aravalli craton and the Gulf of Cambay, respectively. Similarly, delays (>1 s) are observed at BHUJ in the NE sector (0°–60°) relative to the other azimuths. The variability in the arrival times of the P₄₁₀s and P₆₆₀s phases with slowness and back azimuths suggests lateral variation in seismic velocities and/or short-scale variations in the discontinuity topography, which possibly masks the correlation between the mantle phases in certain slownesses and back azimuths. Busse (1981) and Gilbert et al. (2003) proposed that convective flow in the upper and lower mantle affects the depths of each discontinuity separately, which also can lead to uncorrelated topography. However, a thermal/compositional anomaly extending into the transition zone would show up in terms of strong anticorrelation between the mantle phases, which is not the case here, since the mantle transition zone time lag (P₆₆₀s – P₄₁₀s) at most stations is observed to vary between 23.6 and 24 s, i.e., close to the global average of 24 s.

Receiver functions stacked with respect to the piercing points are much more effective in characterizing the lateral variations in the depths to the mantle discontinuities compared to the single-station stacks. Therefore, receiver functions from the entire array were binned separately based on the locations of the piercing points at 520 km depth. Since the location of the piercing points at 410 (660) km depth is 118 (216) km offset from the station for an event at an epicentral distance of 67°, a depth of 520 km, which lies almost midway between 410 km and 660 km, was chosen. The piercing points (Fig. 2A) for all the events were obtained by ray tracing using the IASP91 model based on the slowness of the P wave. The receiver functions for which piercing points were located in a 2° (longitude) × 1° (latitude) cell were binned after moveout correction and stacked. These stacked traces were sorted latitude-wise from 20°N–25°N along 70°E–72°E and 72°E–74°E longitudes and plotted along with the mean and standard deviation (Fig. 4). The receiver function plots (Fig. 4) indicate strong and consistent delays of ∼1 s for P₄₁₀s along both 70°E–72°E and 72°E–74°E longitudes for all latitudes between 20°N and 25°N. The P₄₁₀s delays appear to increase toward north in the longitudinal range 70°E–72°E, corresponding to the Kutch rift, while they appear to decrease to the north in the 72°E–74°E range, possibly due to rays sampling the Aravalli craton. Large delays are seen in the lower latitudes in the longitudinal range 72°E–74°E, which correspond to the southward extension of the Cambay rift into the Gulf of Cambay. Similarly, the P₆₆₀s phases, although weak in amplitude, are also consistently delayed, albeit by varying amounts along several latitudes. The P₆₆₀s phases are clearly identifiable and delayed in the 23°N–25°N latitude range in both longitudinal ranges, correlating very well with the delays in the P₄₁₀s phase. Although the P₆₆₀s phases along certain latitudes are not clearly visible, the general trend of the P₆₆₀s arrival appears to correlate with the P₄₁₀s phase. However, there does not appear to be any indication of anticorrelation between the mantle phases, with the average mantle transition zone time lag being close to 24 s as predicted by the IASP91 reference model.

In northwestern India, being a geologically complex terrain, ray paths from source to receiver traverse diverse tectonic regions, which include rifts and cratons with varying subsurface velocity characteristics, resulting in variations in the time lags of the mantle converted phases. However, to obtain a representative picture of the entire study region, a composite image (Fig. 5) for the whole region was constructed by combining the receiver functions at all stations in the northwestern Deccan volcanic province. To facilitate identification of converted phases from the mantle discontinuities, the receiver functions were moveout corrected for converted phases (Fig. 5). The bootstrap time lags of P₄₁₀s and P₆₆₀s phases for the northwestern Deccan volcanic province are 45.13 s ± 0.09 s and 68.96 s ± 0.08 s, respectively (Fig. 5). The composite image validates the observations at individual stations indicating an overall residual time lag of ∼1 s relative to IASP91 and a transition zone time lag of 23.83 s. The composite image of the northwestern Deccan volcanic province clearly indicates that P₄₁₀s and P₆₆₀s phases exhibit correlated delays, implying that the factors responsible for the delays are constrained to be in the upper mantle above the 410 km discontinuity.

In order to compare the structures beneath the northwestern Deccan volcanic province and south-central Deccan volcanic province, a revised composite receiver function image for the south-central Deccan volcanic province was reconstructed using processing techniques identical to those used for obtaining the composite image for the northwestern Deccan volcanic province. About 1400 high-quality waveforms from 11 stations, including new additional data of ∼500 receiver functions from permanent stations and Koyna network in the south-central Deccan volcanic province (Fig. 2B), were used to update and revise the composite receiver function image for the south-central Deccan volcanic province that was obtained earlier using 900 receiver functions from six stations (Kumar and Mohan, 2005). The image for the south-central Deccan volcanic province (Fig. 5) exhibits sharp and clear conversions of the P₄₁₀s and P₆₆₀s phases at 44.17 s ± 0.15 s and 68.07 s ± 0.12 s, respectively, similar to the IASP91 times. The composite receiver function images of the northwestern and south-central Deccan volcanic province are observed to be distinctly different.

The IASP91 velocity model predicts time lags of 44.1 s and 68.1 s for Ps conversions from 410 km and 660 km discontinuities for a slowness of 6.4 s/°. Global compilations of P₄₁₀s and P₆₆₀s time lags indicate that they are 1–2 s early or 2–3 s delayed relative to IASP91 predicted time lags at stations on Precambrian platforms and hot spots and regions that evidenced recent magmatic activity, respectively (Chevrot et al., 1999). The observed average delay of ∼1 s in the northwestern Deccan volcanic province falls in between the extreme values expected over typical Precambrian cratons and hotspots, respectively. Such delays assume significance in context of the Deccan volcanism since the Deccan episode is 65 m.y. old, and the Deccan volcanic province does not presently lie above a hotspot. Maximum delays are observed in the Gulf of Cambay, which is a confluence of the Cambay and the Narmada rifts. The transition zone time lag for all stations in the northwestern Deccan volcanic province averages 23.83 s, which is marginally smaller than the global average of 24 s, suggesting a normal mantle transition zone, thereby attributing the delays to the region above the 410 km discontinuity. The anomalous delays in the northwestern Deccan volcanic province are comparable to those observed over provinces associated with Cenozoic or Mesozoic volcanism. For instance, a delay of ∼1 s is observed for the 250 Ma Emeishan Mesozoic traps (Chevrot et al., 1999), while it is larger, ∼1.5 s, for the Yellowstone hotspot (Fee and Dueker, 2004). Similar delays (∼1.3 s) are observed at stations in the Arabian Shield, which are interpreted to be due to low shear-wave velocities of a hot upper mantle associated with Cenozoic volcanism (Kumar et al., 2002). A prominent exception is the south-central Deccan volcanic province, encompassing the large Deccan outcrop area where the P₄₁₀s and P₆₆₀s time lags are 44.17 s ± 0.15 s and 68.07 s ± 0.12 s, i.e., close to normal values predicted by IASP91 (Fig. 5). Thus, the northwest and south-central segments of the Deccan volcanic province are observed to be characterized by contrasting upper-mantle seismic signatures.

The anomalous delays observed could be due to a combination of several factors, which include velocity reduction due to lithospheric thinning coupled with thermal/compositional changes and/or topographical variations of the mantle discontinuities. The variation in the crustal thickness in the northwestern Deccan volcanic province is too small to significantly affect the overall average delay of 1 s from the mantle discontinuities. However, a 10 km variation in the topography of the 410 km discontinuity can account for a delay of 1 s. The topographic variations, although a possibility, may not be the primary effect causing the delays, since the observed average time lags of both P₄₁₀s and P₆₆₀s phases are to a major extent correlated in the single-station image (Fig. 3), and piercing point stacks (Fig. 4), as well as in the composite image (Fig. 5). Importantly, the observed mantle transition zone time lag is close to the global average of 24 s, implying absence of any strong anomaly within the mantle transition zone that might cause apparent deepening of the 410 km discontinuity. Although, short-scale topography in the mantle discontinuities cannot be ruled out, topographic variations in the 410 km discontinuity were neither observed in the mantle images at six broadband stations in the south-central Deccan volcanic province (Kumar and Mohan, 2005) nor anywhere else in the Indian Shield (Saul et al., 2000). In the absence of significant topographic variations of the mantle discontinuities, the possible cause for systematic delays would be reduction in seismic velocities in the upper mantle. Such correlations between the depths to mantle discontinuities and upper-mantle heterogeneity have also been observed in global studies (Chevrot et al., 1999).

There are several possible explanations (Thybo, 2006) for the origin of the low seismic velocities, which are primarily manifestations of thermal/chemical anomalies associated with regional/local causes. Sensitivity analysis reveals that a 1 s delay in P₄₁₀s time lag relative to IASP91 corresponds to ∼1% reduction in shear velocities over a depth range of 400 km of the IASP91 velocity model (Singh and Kumar, 2009). Attributing the 1 s delay to a 200-km-thick anomalous zone of elevated temperature at a depth of 100 km in the upper mantle would require a 5% reduction in shear velocities. According to laboratory measurements of the temperature derivatives of wave speeds in olivine (Isaak, 1992), a 200–300 K temperature increase in the upper mantle would reduce S-wave velocities by ∼2%. Taking the commonly accepted value for the shear velocity gradient with temperature of –0.35 ms⁻¹ K⁻¹ (Sumino and Anderson, 1982), an increase in temperature by 100 K results in a velocity reduction of 0.767%. Thus, explaining a 1 s delay would require an unrealistic increase in temperature, thereby ruling out the possibility of only temperature effects. Further, Ziegler and Cloetingh (2004) estimated that after 60 m.y., only ∼35% of a deep-seated thermal anomaly would remain. Thus, after 65 m.y., the lithosphere beneath the Deccan volcanic province is expected to retain a significantly reduced thermal imprint of the volcanic episode, which also contributes to reduction in the seismic velocities. The effect of lithospheric thinning was estimated based on the model proposed by Yuan et al. (2006), which is a variation of the IASP91 model and incorporates a low-velocity zone, and we found that a thinning of 50 km results in a delay of only 0.384 s. Thus, a thermal anomaly coupled with lithospheric thinning cannot adequately explain the observed delays unless compositional heterogeneities are also invoked. Compositional anomalies may originate from variations in the degree of depletion due to magmatic activity and metasomatism (Thybo, 2006). Seismic velocities are considerably decreased by the presence of even small amounts (<1%–2%) of melts or fluids (Mavko, 1980). Geochemical evidence (Sen et al., 2009) suggests that the lithosphere beneath Kutch in the northwestern Deccan volcanic province was subjected to carbonatite metasomatism. The presence of carbonate reduces the solidus temperature, resulting in melting and development of low-velocity zones (Presnall and Gudfinnsson, 2005). Hence, it is likely that lithospheric thinning coupled with compositional and reduced thermal effects are responsible for the low velocities in the shallow upper mantle beneath the northwestern Deccan volcanic province.

In order to assess feasible models for the genesis of Deccan volcanism, we analyzed the results from the northwestern and south-central Deccan volcanic province in conjunction with the earlier results obtained through P-wave tomography in northwestern India (Kennett and Widiyantoro, 1999). The plume model relates Deccan volcanism to the Reunion hotspot (Morgan, 1972; Richards et al., 1989; White and McKenzie, 1989; Courtillot et al., 1999), which is assumed to rise from the core-mantle boundary (Courtillot et al., 2003; Zhao, 2007). The flood basalts are understood to be generated mainly by decompressional melting of abnormally hot mantle brought to the base of the lithosphere by plumes (White and McKenzie, 1995). Previous evidence for the possible imprints of a plume was in terms of a 200-km-wide, low-speed, P-wave anomaly (relative to the AK135 model; Kennett et al., 1995) centered beneath North Cambay, extending from shallow (Fig. 6) to deeper than 200 km depths in the upper mantle (Kennett and Widiyantoro, 1999). This anomaly appears to continue, albeit weakly, to the south, beneath the northwestern Deccan volcanic province, as seen from Figure 6. Our results in the northwestern Deccan volcanic province suggest a much stronger low-velocity anomaly in the upper mantle as revealed by the ∼1 s delay transition zone conversions relative to IASP91 model. Such a delay translates to ∼1.2 s relative to the AK135 model used as a reference model in the P-wave tomographic studies. In contrast, the south-central Deccan volcanic province, south of the Narmada rift, is characterized by much smaller delays (Fig. 6), indicating a largely unperturbed lithosphere. The contrasting signatures beneath the northwestern Deccan volcanic province and south-central Deccan volcanic province are addressed through the geological framework supported by results from numerical modeling.

Tectonically, northwestern India has experienced Mesozoic crustal extension, which resulted in continental stretching and rifting, leading to the formation of aborted rifts (Biswas, 1987). On the other hand, the south-central Deccan volcanic province, encompassing the northern extension of the Dharwar craton, is strongly influenced by west-coast rifting, leading to the formation of rift shoulders in terms of the Western Ghats. The seismic-velocity structure of the south-central Deccan volcanic province is similar to that of the south Indian craton (Kumar et al., 2001), while the northwestern Deccan volcanic province, encompassing the Cambay rift and Saurashtra, is characterized by crustal thinning and high-velocity signatures (Vp > 7.1 km/s) in the lower crust (Kaila and Krishna, 1992; Rao and Tewari, 2005). Drill holes in Cambay rift reveal >3-km-thick basalts at a depth of 3–5 km beneath the Tertiary sediments (Roy, 1991), indicating postvolcanic subsidence. Evidence of high-density bodies, possibly mafic intrusions, related to Deccan–pre-Deccan volcanism, is seen in all the rifts in the northwestern Deccan volcanic province. Sen et al. (2009) argued for a pre-Deccan rifting and lithospheric thinning episode beneath Kutch in the northwestern Deccan volcanic province followed by carbonatite metasomatism of the lithosphere and alkalic volcanism during the emplacement of the Deccan volcanics. Earlier, a thick unperturbed lithosphere beneath the south-central Deccan volcanic province was suggested through receiver function studies (Kumar and Mohan, 2005), as well as through P-wave tomographic studies (Ramesh et al., 1993). The contrasting seismic signatures of the northwestern and south-central parts of the Deccan volcanic province possibly reflect the variations in the shear-wave velocities with depth, coupled with the variations in lithospheric thickness.

An explanation for the enormous volumes of melt (1.5 × 10⁶ km³) on the surface in addition to unknown amounts emplaced within the crust and lithosphere would need a thermal source over which rifting occurred (White and McKenzie, 1989). Such a possibility would lead to thermal/ compositional changes in the upper mantle, which, if preserved, would lead to lowered upper-mantle velocities. A deep-rooted, large-scale thermal anomaly akin to a plume head would be expected to influence the characteristics of the entire Deccan volcanic province, unlike the contrasting signatures in the northwestern and south-central Deccan volcanic province. Clues toward explaining this paradox are provided by numerical modeling of the interaction between a plume and a moving lithospheric plate of nonuniform thickness (Manglik and Christensen, 2006), which indicates that a thick root tends to deflect an impinging plume head to regions of thin lithosphere, where preexisting rifts facilitate effusion of basalts. Further, a mantle plume impinging under a craton would be prevented from decompression melting and result only in transient uplift until the plate moves away from the plume. However, a stretched and thinned lithosphere traversing over a mature plume can result in flood basalts, since the large extent of the thermal anomaly becomes significant owing to the fact that the volume of melt generated beneath rifts is sensitive to small variations in mantle temperature (White and McKenzie, 1989). A similar scenario for the Deccan volcanic province implies that relative to the cratonic south-central Deccan volcanic province, the stretched and rifted northwestern Deccan volcanic province is underlain by a thinner lithosphere, which provides a favorable setting for facilitating the eruption of flood basalts through decompressional melting beneath the rift systems of the northwestern part of the province. The imprints of the volcanic episode in terms of thermal/compositional changes together with lithospheric thinning lead to lowered upper-mantle velocities beneath the northwestern Deccan volcanic province, resulting in the contrasting seismic signatures observed for the northwestern and south-central parts of the Deccan volcanic province. Further, the pattern of melting predicted by numerical modeling for nonuniform lithospheric thickness does not vary for reduced dimensions of the plume head (Manglik and Christensen, 2006). So, it is likely that the relative difference in lithospheric thickness between the northwestern and south-central Deccan volcanic province governs the pattern and volume of eruption.

Seismological evidence in terms of average delays of ∼1 s for P₄₁₀s and P₆₆₀s phases relative to the arrival times predicted by IASP91 suggests that the upper mantle beneath the northwestern Deccan volcanic province is significantly perturbed, while the transition zone is close to normal. The perturbations are due to reduced seismic velocities in the upper mantle above the 410 km discontinuity, associated with thermal/compositional anomalies coupled with the possibility of lithospheric thinning. Contrasting seismic signatures over the rifted and stretched lithosphere of the northwestern Deccan volcanic province and the undisturbed craton of the south-central Deccan volcanic province seem to be linked to the flow pattern of the upwelling mantle material, which is governed by variations in the lithospheric thicknesses. Our results suggest that the lithospheric architecture of the northwestern Deccan volcanic province coupled with the reactivation of preexisting rift systems appears to have facilitated the eruption of Deccan basalts, for which source signatures are still retained in the upper mantle.

This work is sponsored by the deep continental studies program of the Department of Science and Technology, India. We thank India Meteorological Department for providing the data recorded at station BHUJ. The work done at NGRI has been performed under the Supra institutional project of NGRI SIP 0012-28 (M.R.K.).