Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Geophys. J. Int. (2001) 147, 543–561 An integrated geophysical analysis of the upper crust of the southern Kenya rift Silas M. Simiyu1 and G. Randy Keller2 1 2 Geothermal Project, The Kenya Electricity Generating Company, PO Box 1202, Naivasha, Kenya. E-mail: [email protected] Department of Geological Sciences, University of Texas at El Paso, El Paso, Texas 79968, USA. E-mail: [email protected] Accepted 2001 June 18. Received 2001 June 13; in original form 1999 May 27 SUMMARY Previous interpretations of seismic data collected by the Kenya Rift International Seismic Project (KRISP) experiments indicate the presence of crustal thickening within the rift valley area beneath the Kenya dome, an uplift centred on the southern part of the Kenya rift. North of the dome, these interpretations show thinning of the crust and an increase in crustal extension. To the south near the Kenya /Tanzania border, crustal thinning associated with the rift is modest. Our study was aimed at further investigating crustal structure from this dome southwards via a detailed analysis focused on upper crustal structure. We used results from surface geological mapping, drill hole data from water wells and geothermal exploration wells, KRISP 85 seismic data for a profile across the rift, KRISP 85 and 90 seismic data for a profile along the rift axis and KRISP 94 seismic data for a profile crossing southernmost Kenya to constrain gravity modelling and construction of integrated models of crustal structure. Our integrated analysis produced the following results concerning the structure and evolution of the southern Kenya rift: (1) the graben master faults are consistently located along the western margin of the rift valley, and there is no evidence for half-graben polarity reversals for a distance of about 300 km; (2) there is no axial (north–south) crustal symmetry with respect to the apex of the Kenya dome, and the crustal thickness may be as much related to pre-rift crustal type and thickness as it is to crustal thickening and modification by magmatic processes; (3) the pre-existing lithospheric contrast between the Archaean and Proterozoic basement terranes played a significant role in the location and structural geometry of the rift; (4) south of latitude 1uS, low velocities and densities observed under the western flank of the rift probably represent reworked Archaean Tanzanian craton; (5) magmatic modification of the upper crust is modest except near the major Quaternary volcanic centres that produce a series of isolated axial gravity highs; (6) the other element of the axial gravity high is an intrarift horst block that extends along the axis of most of the rift valley. Key words: continental rifting, crustal structure, gravity, refraction seismology. INTRODUCTION The East African rift zone is the classic example of an active continental rift, and its eastern arm extends from southernmost Ethiopia, through Kenya and into Tanzania and is variously called the Eastern, Gregory or Kenya rift, the term we will employ. The southern portion of the Kenya rift (Fig. 1) has been the target of a number of recent geophysical and geological investigations, many of which are summarized in two special volumes (Prodehl et al. 1994, 1997). This is due in part to the Kenya Rift International Seismic Project (KRISP), which has focused attention on the region, but it is also the result of the exploration efforts of the Kenya Electricity Generating # 2001 RAS Company. These studies have created a large new database, which has not been fully exploited to elucidate the structure of the upper crust. Here, new gravity measurements and drill hole data from geothermal areas are combined with the KRISP seismic data, data from water wells and geological constraints in an integrated analysis of upper crustal structure. Our specific goals were to investigate basement structures beneath the rift valley and the extent of magmatic modification of the upper crust. Our focus on the upper crust is different from the previous studies motivated by the KRISP effort that have targeted lithospheric-scale structure. The combination of all of these data allows us to pursue a level of detail that has not been 543 544 S. M. Simiyu and G. R. Keller 340E 380E 10 N NA I Mt Elgon L. Baringo L AN ELDORET L. Bogoria Mt Kenya KISUMU 0 Aberdare Range Menengai Mf L. Victoria 0 z Eburru L. Naivasha KENYA Ol Ok Kf Sf Longonot 10 S Suswa NA I NAIROBI TANZANIA L. Magadi 20 S AN L Tanzania Craton Chyulu Range L. Natron Kilimanjaro Tanzania craton Rift volcanics and sediments Off–axis major volcanoes Mobile belt quartzites with reworked Archean Mobile belt gneisses and other metasediments Outline of the Kenya dome NAI - Nyanega-Athi-Ikutha shear zone & ANL - Aswa-Nandi-Loita shear zone Volcanic ranges Rift axis caldera volcanoes Figure 1. Tectonic map of the southern Kenya rift showing the surface distribution of the major lithologic units (after Smith 1994). Mfz: Mau fault zone; Sf: Sattima fault; Kf: Kinangop fault zone; Ol: Oloololo Escarpment; Ok: Olkaria volcanic centre possible previously. Our approach included simultaneous 2-D modelling of first arrivals from KRISP seismic lines in the area (Fig. 2) and 2.5-D modelling of gravity profiles along these lines and across key features where drill holes provided important constraints. In addition, we used updated gravity maps to provide a view of the lateral extent of features revealed in the modelling. Throughout this process, all supporting data available were employed as constraints. Our integrated approach produced geophysical models with some details that cannot be defended on the basis of a single data set. Finally, we produced interpretative crustal-scale cross-sections along the three profiles where we had the most constraints. Geological setting Baker (1986), Baker et al. (1988), Smith & Mosley (1993) and Smith (1994) provide recent reviews of the geology and evolution of the southern Kenya rift region (Fig. 1). Besides the rift and associated volcanics, two major geological provinces are present. The Archaean Tanzania craton to the west is characterized by exposures of granitoid gneisses and metavolcanics. The northeastern boundary of this craton is the Aswa–Nandi– Loita NW–SE-trending shear zone (Fig. 1), and in southern- most Kenya, geophysical data place the boundary between the Oloololo escarpment and Lake Magadi (Birt et al. 1997; Simiyu & Keller 1998). Between these areas, the boundary is not visible due to volcanic cover. Rocks of the Neo-Proterozoic Mozambique belt are well exposed east of the rift valley and consist mainly of gneisses, schists and migmatites that contain the Nyangea–Athi–Ikutha shear zone that crosses the rift valley (Fig. 1). Near the Kenya–Tanzania border, this mobile belt contains many areas composed of highly fractured quartzites and gneisses, which may have been overthrust onto a reworked portion of the Archaean craton (Smith 1994). Recent geological studies (Shackleton 1986; Smith & Mosley 1993; Stern 1994) suggest that the southern Kenya rift developed entirely above thick cratonic lithosphere and that the Mozambique belt rocks seen on the surface are thrust over a reworked craton that extends at least 100 km east of the craton /mobile belt outcrop boundary. Smith & Mosley (1993) discussed the evidence for Archaean crust underlying surface exposures of the Mozambique belt, which include zones of thrust faulting containing slices of Archaean greenstones, gneissic xenoliths in overlying Tertiary lavas east of the mapped boundary that yield Archaean Nd model ages, and a variety of metamorphic indicators. Smith (1994) suggested that the # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure L. Baringo ELDORET L. Bogoria 0 CHE o KISUMU Menengai Mt Kenya SOL ELM VIS Eburru NAI L. Naivasha Olkaria OLO EWA NTU MOR TANZANIA Longonot SUS MAR Suswa NAIROBI MAG KAJ o 2 S Shotpoint KRISP 85 } KRISP 90 KRISP 94 KRISP 85 (EW) KIB L. Natron Kilimanjaro Figure 2. Index map of the southern Kenya rift showing the location of the KRISP 85, KRISP 90 and KRISP 94 seismic refraction lines used in this study. The locations of the shotpoints used in this study are indicated by stars and are named as follows: Victoria (VIS), Oloololo (OLO), Morijo (MOR), Magadi (MAG), Kajiado (KAJ), Ewaso Ngiro (EWA), Ntulelei (NTU), Suswa (SUS), Margaret (MAR), Naivasha (NAI), Elmenteita (ELM), Solai (SOL), Chepkererat (CHE) and Baringo (BAR). margin for the reworked craton trends broadly NW–SE and that its northern limit is marked by the Aswa–Nandi–Loita shear zone (Fig. 1). He also argued that in the Magadi– Oloololo area (Fig. 1), the Archaean basement is buried by gravitationally collapsed nappe structures. The rift valley is a 50–70 km wide volcanic filled depression bounded by major normal faults (e.g. Baker et al. 1972). There has been considerable discussion about the role of the boundary (suture or shear zone) between the Tanzania craton and the Mozambique belt controlling the location of the rift at both the regional and the local level (Nyblade & Pollack 1992; Smith & Mosley 1993; Smith 1994; Hetzel & Strecker 1994; Birt et al. 1997; Tesha et al. 1997; Simiyu & Keller 1998). These authors all feel that this boundary exerts some level of control on the location of the rift valley. However, without seismic reflection profiles, we cannot trace structural features from the surface to depth. The highest portion of the rift valley (y1.9 km) is the Nakuru–Naivasha–Suswa region (Fig. 1), which contains three large volcanoes with calderas (Menengai, Longonot and Suswa) and large volcanic fields at Olkaria and Eburru (Fig. 1). These volcanoes are located near the axis of the rift valley, but basalt– trachyandesite volcanic cones and flows occur in the intervening low areas. Exposed volcanic rocks include the Mau and Kinangop tuffs, which have been dated at 2.6–3.68 Ma (Baker et al. 1988). The Kinangop tuffs /ignimbrites are exposed by the eastern boundary fault while the Mau tuffs are exposed along the western fault scarp. Deep drilling at Olkaria and Eburru has shown that the Mau tuffs are downthrown to depths of more than 2.5 km by major N–S-trending faults (Omenda 1994). Limuru trachytes that have been dated at 1.68–2.01 Ma # 2001 RAS, GJI 147, 543–561 (Baker et al. 1988) are exposed to the southeast of Suswa and are thought to underlie most of the inner graben of the rift floor. Plateau trachytes that have been dated at 0.98–1.23 Ma overlie the Limuru trachytes. The plateau trachytes are exposed in the southern part of Suswa and as horst blocks in the Eburru– Elmenteita areas (Baker et al. 1988). Overlying these trachytes are Quaternary volcanics that include basalts, trachytes, rhyolites, tuffs and pyroclastics. These materials form a thickness of about 1.5 km within the graben in the Olkaria–Eburru area (Omenda 1998) Late Quaternary volcanic activity within the rift valley was concentrated in the axial region of the rift valley and resulted in the formation of caldera volcanoes and volcanic cones (Fig. 1). Menengai, which is the northernmost volcano in the Nakuru– Naivasha basin, is located about 60 km north of Eburru and consists almost exclusively of strongly peralkaline oversaturated trachytes (Clarke & Woodhall 1987). Eburru is dominated on the surface by pantellerites, and pantelleritic trachytes, but trachytes are more abundant with depth. Syenite intrusives were encountered at the bottom of some of the holes drilled in the volcanic complex for geothermal exploration. The Elmenteita and Ndabibi basaltic fields are located northeast and south of Suswa, respectively, and both lie along the N–S axial fault system (Fig. 1). Previous gravity studies The southern portion of the Kenya rift has been the focus of many gravity studies, and gravity anomalies across the region reflect the classic pattern of a long-wavelength minimum due to mantle structure and shorter-wavelength anomalies due to crustal structure (Fig. 3). These studies are summarized by Swain et al. (1994) and provided an invaluable database that laid the foundation for our study. Bouguer gravity (mGal) BAR Suswa * * ** * * * * * * * * * Observed Gravity Estimated regional –160 * –180 * ** * *** * –200 Menengai * *** * * * * ** ** ** * * * * * * * *** * * Olkaria ** * ** ** * * * * South North (a) Bouguer gravity (mGal) 36 o E Mt Elgon 545 0 o o o 0.5 S 1.0 S East West –160 Regional (Swain, 1992) Regional (Fairhead, 1976) Olkaria –180 Observed gravity –200 (b) Regional Searle (1970) 0 50 100 150 200 DISTANCE (KM) Figure 3. Observed gravity profiles and interpreted regional anomaly profiles. (a) Axial (N–S) gravity profile along the Kenya rift valley through Menengai, Olkaria and Suswa volcanoes (after Simiyu 1990). (b) E–W gravity profile crossing the rift valley and passing through Olkaria (after Simiyu 1990). 546 S. M. Simiyu and G. R. Keller Swain (1992) re-interpreted the axial gravity high from Lake Baringo to Suswa using the KRISP 85 seismic results as constraints. He showed that the axial high in the southern Kenya rift is sometimes associated with volcanoes, which produce local maxima. He also showed that a minor densification of the basement by dyke injection could explain much of the axial high. The KRISP 90 and 94 results were not available at the time he did his analysis, and he did not have access to the new gravity data and drill hole information presented here. However, we used his analysis as a starting point for the gravity portion of this study. Most of the earlier studies concentrated on the axial high anomaly that is superimposed on three different wavelength negative anomalies: (1) the primary regional negative anomaly associated with the East African Plateau; (2) the secondary regional negative anomaly associated with the mantle anomaly centred on the Kenya dome; (3) local anomalies associated with upper crustal features (e.g. Simiyu & Keller 1997). Attempts to separate these anomalies using different methods [smoothing (Searle 1970), tying to outcrops of Precambrian basement (Baker & Wohlenberg 1971), a combination of these approaches including balancing of the thickness and density of the rift volcanics (Fairhead 1976), and linear filters (Swain 1992)] have yielded different residual anomaly wavelengths and amplitudes (Fig. 3) and thus differing models of crustal structure. These studies lacked seismic and drill hole control, and gravity data coverage was sparse in many key areas at the time. In an effort to integrate recent seismic studies with regional gravity anomalies, Simiyu & Keller (1997) looked at the entire East African Plateau region crossing both the eastern and western arms of the East African rift. Their models of deep (150 km) structure provide the regional framework for this study. Simiyu & Keller (1998) analysed KRISP 94 seismic data and gravity readings to derive a detailed model of the upper crust at Lake Magadi, where the KRISP 94 line crossed the rift valley. A goal of this paper is to extend this detailed analysis northwards as far as the Menengai region. Previous seismic studies In the first major seismic study of the crust of the southern Kenya rift, Henry et al. (1990) interpreted the KRISP 85 seismic refraction profiles, one of which partially followed the axis of the rift from Lake Baringo to Lake Magadi (Fig. 2). The overall data quality was fair, and the lower crustal phases were not reversed on this profile. Interpretation of these data showed that there were significant differences in crustal structure between the northern and southern parts of the rift valley. The rift fill thickness was found to vary between 1.5 and 5 km and to be underlain by basement material of velocity 6.05 km sx1. They found no evidence for an extensive axial mafic intrusion in the crust suggested by some of the early studies. The model of Henry et al. (1990) shows the presence of highvelocity material lying directly beneath the Menengai gravity high. However, this model does not explain the corresponding increase in the axial positive Bouguer gravity anomalies in the area between Elmenteita and Suswa, where several thousand metres of Cenozoic volcanics of relatively low density (inferred from their low seismic velocity of 2.8–5.1 km sx1) blanket the area. Their data were sparse south of the Elmenteita region, and they did not suggest the presence of any crustal highvelocity /density bodies beneath this area, although it contains a large positive gravity anomaly. The KRISP 90 (Fig. 2) experiment was aimed at studying the variation in crustal and uppermost mantle structure across and along the axis of the Kenya rift. The analysis of the axial profile by Mechie et al. (1994) and Keller et al. (1994a) was of particular significance to this study. This analysis shows that the crystalline upper crust, interpreted to be the Precambrian basement, has velocities of 6.1–6.3 km sx1 and is 5–6 km below surface in the Menengai–Suswa region. Mechie et al. (1994) converted their velocity structure to densities using standard relationships and showed that the regional Bouguer anomaly can be explained well by southwards crustal thickening along the rift axis. However they did not attempt to fit shortwavelength upper crustal anomalies in the Elmenteita, Naivasha and Suswa areas. In this paper, we attempt to complete the picture of the upper crust by modelling the local gravity anomalies using well data and seismic constraints. In 1994, an experiment was carried out across the southern Kenya rift extending from Lake Victoria to the Indian Ocean (Fig. 2). The KRISP 94 experiment showed only slight crustal thinning under the rift at Lake Magadi, with the Moho being about 35 km deep (Birt et al. 1997). However, Thybo et al. (2000) studied the lower crust and concluded that significant underplating occurred during rifting and that the amount of thinning of the pre-rift crust was probably significant. East of the rift valley, the crust thickens to 42 km in the Chyulu Hills area with increased crustal velocities (Novak et al. 1997). Based on the KRISP 94 results, a detailed interpretation of the upper crustal structure near Lake Magadi was presented by Simiyu & Keller (1998) and their results showed that the rift axial high anomaly was caused by a basement horst structure at its axis. SEISMIC INTERPRETATION Seismic and gravity data tend to be complementary due to the general relationship between seismic velocity and density. We chose to model seismic data first and then use these results as constraints for the integrated modelling of gravity profiles. The three seismic profiles modelled are shown in Fig. 2. They are the KRISP 85-90 axial profile, the KRISP 85 profile that crossed the rift valley just north of Mt Suswa, and the KRISP 94 profile that crossed the rift valley at Lake Magadi. Although picking first arrivals for these data (shown in Henry et al. 1990; Mechie et al. 1994; Birt et al. 1997) was generally straightforward, we fine-tuned this process using a commercial seismic processing package. Frequency filtering, deconvolution, trace normalization and other amplitude enhancement techniques were employed with some success using this interactive program, which includes an algorithm for the automatic picking of first arrivals. In order to be as objective as possible, this algorithm was utilized to pick the Ps (direct wave) and Pg (head /diving wave travelling in the crystalline upper crust) first arrivals that we used for modelling the velocity structure. These traveltime picks were confirmed manually and found typically to differ from handpicked arrivals by less than 0.05 s. Record sections and the first arrival traveltime picks were used to determine apparent velocities for the Ps and Pg phases that were used in the starting models. These velocities were determined by least-squares straight-line fitting in which the traveltimes were weighted by the reciprocal of the observational error estimates. The starting model for the shallow velocity structure was based on previous velocity results from KRISP 85, KRISP 90 and KRISP 94 (Mechie et al. 1994; Henry et al. # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure sediments and volcanics. The KRISP 90 data were then used to model the basement structure, with the exception that the data from the 1985 Magadi shotpoint were used to reverse the KRISP 90 Naivasha shotpoint. The Baringo (85,90), Naivasha (90) and Magadi (85) shotpoints and the KRISP 94 EW model at Magadi (Birt et al. 1997; Simiyu & Keller 1998) provided the most rigorous constraints on the velocity model. The model shows that the depth to Precambrian basement along the axis of the rift varies from about 3.5 km beneath Lake Baringo to a maximum of 6 km below Lake Naivasha and back to about 3 km to the south at Lake Magadi (Fig. 4). The rift fill is modelled as being composed of three layers. The first layer has a velocity of 2.8 km sx1 and extends from model coordinates 0–270 km, which is just south of Suswa (Fig. 4). The second layer (average velocity of 4.2 km sx1) is 3 km thick in the north at Lake Baringo and thins southwards to 1 km beneath Lake Magadi. The velocity of this layer is variable, ranging from 3.9 km sx1 beneath Lake Baringo to 4.6 km sx1 beneath Lake Magadi, with a localized low-velocity area beneath Lake Naivasha (3.6 km sx1). 1990; Birt et al. 1997; Simiyu & Keller 1998) and knowledge of the geology of the area. The modelling of the velocity structure was carried out using the MacRay ray tracing package (Luetgert 1992). For refracted phases to be generated, layers were assigned positive velocity–depth gradients, typically of around 0.01 km sx1 kmx1. The starting model was adjusted by manual iteration until the observed traveltimes closely matched the calculated traveltimes. The average difference between calculated and observed traveltimes for the three models is about 0.06 s, which is on the order of the picking error. Axial seismic model There were numerous sources along the axial profile (Fig. 2), so the model derived from these data (Fig. 4) is well constrained except for a small area south of Suswa where there is a data coverage gap. The KRISP 85 shots, most of which only produced Ps arrivals, were modelled first as they were more closely spaced and most useful for constraining the rift fill of 3 NORTH Predicted Observed BAR CHE SOL 547 SOUTH T–X/6.00 2 1 0 ELM NAI MAG –1 0 40 80 120 160 200 240 280 0 40 80 120 160 200 240 280 Depth (km) 0 3 6 9 12 15 Distance (km) NORTH CHE SOL BAR Depth (km) 0 Kenya Dome ELM NAI 4.5 3.6 4.2 2.8 5.3 5.0 6.2 6.3 6 4.6 4.4 4.9 5.0 6.0 6.0 SOUTH 2.8 MAG 3.8 2.8 3.9 3 2.8 5.6 5.5 6.0 6.4 Precambrian Basement 9 6.6 6.2 6.3 12 6.45 6.7 6.7 6.45 6.70 15 0 40 80 120 160 200 240 280 Distance (km) Figure 4. Upper crustal seismic velocity model (lower diagram) along the rift valley axis extending from Lake Baringo to Lake Magadi (Fig. 2). The numbers are P-wave velocities in km sx1. Shotpoint locations are shown by stars (Fig. 2). The middle diagram shows the ray coverage provided by the observed data and the velocity model. The upper diagram shows the fit between the observed traveltimes and those predicted by ray-trace calculations for the velocity model. # 2001 RAS, GJI 147, 543–561 548 S. M. Simiyu and G. R. Keller The third and lowermost layer of the rift fill is thin at Baringo. It thickens around Chepkererat, where it has a velocity of 5.0 km sx1. The velocity increases to 5.1 km sx1 beneath Lake Naivasha, to about 5.4 km sx1 beneath the Suswa volcano, and to 5.6 km sx1 beneath Magadi. A 5.6 km sx1 velocity layer on the KRISP 94 EW profile was interpreted to represent fractured Precambrian basement (Simiyu & Keller 1998). The velocity of the uppermost portion of the crystalline crust varies from 6.0 to 6.2 km sx1 at Baringo, increasing to a maximum of 6.4 km sx1 between Lake Naivasha and Suswa Volcano. Suswa seismic model The Suswa profile (Fig. 2) crosses the rift valley just north of the Suswa volcano, which is capped by one of the largest calderas in the rift valley. The original seismic profile was designed as a cross-line that intersected the axial profile and is in an area that has the highest gravity values along the entire southern rift valley. The model (Fig. 5) is well constrained for the volcanics within the rift valley because two shots produced energy transmission effective enough to probe depths down to T– X/ 6.00 3 WEST about 6 km. Constraints from the N–S axial line also helped in determining the starting model parameters. The western rift flank volcanic thickness and basement velocities are also well constrained by the Ntulelei (NTU) and Ewaso Nyiro (EWA) shots, which are reversed. This model includes a y1 km thick low-velocity (2.8 km sx1) surface layer that thickens to 2.2 km just east of the Ntulelei (NTU) shotpoint at the western rift boundary fault. Further east (Fig. 5), this surface layer thins to less than 0.2 km at the Suswa (SUS) shotpoint. At the Mt Margaret (MAR) shotpoint, this layer thickens to 0.5 km. The next layer has a velocity of about 4 km sx1. Its thickness is 1.2 km west of Ntulelei (NTU), increasing to a maximum of about 2.0 km at the Suswa (SUS) shotpoint, then decreasing to less than 1.7 km towards the eastern part of the rift floor at the Mt Margaret (MAR) shotpoint. The lowermost layer of the rift fill has velocities ranging from 5.0 to 5.5 km sx1. The layer we interpret as crystalline basement has a velocity of 5.8 km sx1 west of the rift valley. This low value possibly indicates the presence of quartzitic basement rocks, and based on the axial line results, the velocity of this layer increases towards the rift axis. EAST Predicted Observed 2 1 0 EWA NTU SUS MAR –1 20 0 40 60 Rift 80 100 Valley Depth (km) 0 3 Precambrian Basement 6 9 12 WEST 15 EAST 20 0 40 Distance (km) 60 NTU –3 EWA 100 80 MAR SUS 2.8 Depth (km) 3.8 5.0 0 5.8 5.8 3.8 5.2 6.0 3 2.7 3.9 5.0 3.9 4.5 5.2 5.0 5.5 6.0 Precambrian Basement 6.0 6 9 12 15 0 20 40 60 100 80 Distance (km) Figure 5. Shallow seismic velocity model (lower diagram) along the KRISP 85 EW profile that extends from Ewaso Nyiro through Suswa to Mt Margaret (Fig. 2). The numbers are P-wave velocities in km sx1. Shotpoint locations are shown by stars (Fig. 2). The middle diagram shows the ray coverage provided by the observed data and the velocity model. The upper diagram shows the fit between the observed traveltimes and those predicted by ray-trace calculations for the velocity model. # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure velocity of the crystalline crust from an average of 5.9 km sx1 on the western flank to 6.0 km sx1 beneath the rift and 6.3 km sx1 on the eastern flank. Magadi seismic model The 1994 KRISP profile crossed the rift valley near Lake Magadi (Fig. 2). Details of the seismic data and modelling results are discussed by Birt et al. (1997), Simiyu & Keller (1998) and Thybo et al. (2000). The most significant aspect of this model (Fig. 6) is a 4 km thick layer of relatively low-velocity (5.6 km sx1) material overlying a normal (6.0 km sx1) basement. This layer starts from the eastern rift boundary and extends to the western limit of the model. Beneath the rift, the layer extends to a depth of 5.5 km. The derived geometries of the rift basin show that the sedimentary and volcanic fill is deepest (3.3 km) to the west. The fill is composed of a 2.5 km sx1 surface layer underlain by a 4.2–4.9 km sx1 layer at the base. Below the Magadi shotpoint, the model includes a basement upwarp suggesting a buried horst block separating two sub-basins. The eastern basin is shallower than the western one and does not contain the 2.5 km sx1 layer found at the surface in the western basin. From west to east, the model shows a gradual increase in the 2 WEST T–X/ 6.00 1 0 549 GRAVITY ANALYSIS AND MODELLING Our first task in the gravity analysis was to determine the average density of the topography for the area and to use this value to calculate Bouguer anomaly values for the data base that we constructed by combining our new data and published data provided by Leicester University. We then used these data to construct anomaly maps and integrated crustal models. Density determinations In order to remove the effects of topography, Bouguer and terrain corrections made routinely as part of gravity data reduction require knowledge of the average density of the rocks that constitute the topographic relief of the surveyed area. The success of these corrections depends on determining the best Predicted Observed EAST MOR MAG KAJ –1 –2 200 240 Depth (km) –3 280 320 360 320 360 Rift Valley 0 3 6 9 12 15 200 240 WEST –3 Depth (km) 0 5.0 MOR 280 Distance (km) Nguruman Escarpment EAST 2.5 MAG 5.6 5.5 4.2 5.8 5.9 3 KAJ 3.9 5.9 4.2 4.9 4.9 6.0 6.0 9 6.2 6.3 5.7 5.6 6 6.1 4.9 Precambrian Basement 6.25 6.3 6.45 12 15 200 240 280 320 360 Distance (km) Figure 6. Shallow seismic velocity model (lower diagram) along the KRISP 94 EW line G crossing the Lake Magadi area (Fig. 2). The numbers are P-wave velocities in km sx1. Shotpoint locations are shown by stars (Fig. 2). The middle diagram shows the ray coverage provided by the observed data and the velocity model. The upper diagram shows the fit between the observed traveltimes and those predicted by ray-trace calculations for the velocity model. # 2001 RAS, GJI 147, 543–561 550 S. M. Simiyu and G. R. Keller average density possible for these rocks. No one density value is completely satisfactory for a large area, but there are a number of techniques available to estimate the density of rocks in situ (e.g. Nettleton 1939). The underlying goal in these techniques is to minimize the correlation between Bouguer gravity anomalies and elevation. The approaches we employed were as follows. (1) Traverses were made across topographic features that were lithologically similar to their surroundings, and a series of gravity profiles using various Bouguer reduction densities were plotted. A density between 2.5r103 and 2.6r103 kg mx3 produced the minimum correlation with topography along these traverses. (2) Bouguer gravity values were calculated for different densities and were plotted against elevation. A value of 2.55r 103 kg mx3 produced the least correlation between Bouguer gravity and elevation. (3) More than 2000 density measurements were made on samples from drill holes to a depth of about 3000 m in the geothermal areas. Shallow drill holes outside the rift show densities of less than 2.4r103 kg mx3 up to depths of 2 km, increasing to more than 2.5r103 kg mx3 as the depth increases. The mean drill core density value is 2.5r103x2.6r103 kg mx3, and in situ density measurements carried out by the Kenya Electricity Generating Company (KenGen) in areas of geothermal exploration show basement density values that range from 2.58r103x2.88r103 kg mx3. Based on these results, we chose a density of 2.55r 103 kg mx3 for our reductions. We believe that our density analysis is a significant contribution, but the use of the traditional 2.67r103 kg mx3 value would not greatly alter our results. The Bouguer anomaly map (Fig. 7) constructed using this density value indicates that the regional negative anomaly over the Kenya rift between 1uN and 3uS is complex and displaced slightly to the west of the rift valley. This map is dominated by this regional gravity low due to deep structures, which are not of interest in this study but which have recently been analysed by Simiyu & Keller (1997). In order to delineate major upper crustal features of interest in this study, a variety of bandpass filters were applied to the Bouguer anomaly values. The map for wavelengths of 30–150 km (Fig. 8) was the one we found most useful in our analysis. This map does not include long-wavelength (>150 km) features associated with deep mantle anomalies under the Kenya dome (e.g. Green et al. 1991; Achauer 1992; Achauer et al. 1994; Slack et al. 1994). Wavelengths of less than 30 km were also removed since such short-wavelength anomalies are caused by features too local to be of interest in this study. We feel that this map best depicts the features targeted in this study. In particular, the axial gravity high is very well defined as a narrow (y20 km) feature extending from the Lake Baringo area to Lake Naivasha. In the Baringo area, this high corresponds to the combined effects of a horst block bounding the Baringo basin to the west in which Precambrian rocks are exposed at the surface (Swain et al. 1981; Smith & Mosley 1993; Maguire et al. 1994) and of the anomaly associated with the Silali volcanic centre. South of Lake Naivasha, this anomaly widens to a maximum of 40 km at Suswa and then dies out 30 km south of this volcano. At Lake Magadi and further south, the rift valley occurs within a broad gravity low that is probably due to one or more of the following effects: (1) more low-density fill is present than to the north; (2) the rift beneath Lake Magadi developed on a different crustal type; or (3) the crust at Lake Magadi has not been magmatically affected in the same way as in the sectors to the north. At the equator, a gravity low follows the Nyanza rift eastwards to Lake Victoria, which implies that it has a simple structure with limited magmatic modification of the crust. East of the rift valley at Magadi, a positive gravity anomaly begins at Kajiado (2.1uS, 36.7uE) in an area coincident with outcrops of the Turoka marble (Shackleton 1986). Integrated gravity models Our next step was to extend the seismic results by employing gravity data for which the geographical coverage was extensive. Our approach was to model the total Bouguer anomaly, and we used the models of Simiyu & Keller (1997) to fit the longerwavelength anomalies and then modelled detailed local profiles. This approach has the advantage of eliminating the need to perform mathematical or visual regional–residual separation, which can be a highly subjective process. A 2.5-D forward modelling program based on the technique of Cady (1980) was employed. In this study, densities and depths to particular interfaces were constrained using seismic and drilling data to the maximum extent possible. The 2.5-D approach allows bodies of various strike lengths to be included. For most of the bodies in the models, strike lengths were long relative to the depth, and therefore, the strike distance was not very important. However for localized intrusives, such as those in the Menengai–Suswa area, calculated gravity was sensitive to the strike lengths. Five gravity profiles were constructed, and their locations (Fig. 9) were governed by data coverage and the availability of seismic and drill hole constraints. The first of the four E–W profiles is located along y 0.2uS, passing through the Menengai volcano near Nakuru. The second profile is located at y 0.5uS, crossing the Mau escarpment to the west and extending through Naivasha to Mt Kenya. Further south of Lake Naivasha, the third profile follows a KRISP 85 seismic line, crossing the rift valley near Suswa. The fourth profile follows KRISP 94 line G along the Kenya–Tanzania border. The last profile is a north–south profile along the KRISP 85-90 axial line and extends from Lake Baringo to Lake Magadi. The analysis of these profiles focuses on upper crustal structure and the rift fill. An attempt was made to determine and separate the gravity anomalies related to basement topography from those related to magmatic intrusions within the basement. In the seismic models, the uppermost layer was only constrained directly near the shotpoints. However, geological and drill hole data and gravity anomalies show that this layer of tuffs, sediments and other volcaniclastics can vary rapidly in thickness due to faulting and topography. In order to fit both gravity and seismic data, it was necessary to adjust the thickness of this layer using drill hole data as constraints. The starting gravity model in each case was based on the seismic results, drill hole and geological data, and densities from previous studies (Henry et al. 1990; Hay et al. 1995; Simiyu & Keller 1997). The density of the uppermost portion of the crust was allowed to vary laterally based on typical seismic velocity–density conversions (Barton 1986). It was assumed that since the cover rocks are several kilometres thick, individual dykes within the basement will not produce detectable gravity anomalies but could increase the overall basement density (e.g. Swain 1992). # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure 35˚E 37˚E 36˚E 38˚E 1˚N -13 34˚E 1˚N 551 0 0 8 -1 50 0 0 -140 -15 -16 -1 -1 70 00 -2 -190 0˚ 0˚ -23 0 -17 0 0 -210 1˚S -180 -200 -23 -160 0 -210 -180 -170 -22 -190 -130 1˚S -1 -1 20 10 -1 50 -1 80 -1 60 -160 -170 90 -1 -170 0 -17 00 -1 2˚S -1 80 2˚S -15 0 0 -17 -160 -150 3˚S 34˚E 35˚E 36˚E 3˚S 38˚E 37˚E Figure 7. Bouguer gravity anomaly map of the southern Kenya rift using a reduction density of 2.55r103 kg mx3. The contour interval is 5 mGal. The dashed white lines show the approximate limits of the rift valley. 34˚E 1˚N 35˚E 37˚E 36˚E 38˚E 1˚N 0 10 -50 -10 -1 0 20 0 -10 -20 0˚ 0 20 10 -20 10 20 0 -20 -10 0˚ 0 -10 0 10 1˚S -2 20 0 -1 10 10 0 0 -10 0 20 -1 0 1˚S 0 10 10 -10 0 2˚S 0 2˚S 10 20 0 -1 0 0 3˚S 34˚E 35˚E 36˚E 37˚E 3˚S 38˚E Figure 8. Bandpass filtered gravity map of the southern Kenya rift. Wavelengths passed by the filter were 30–150 km. The contour interval is 5 mGal. The dashed white lines show the approximate limits of the rift valley. # 2001 RAS, GJI 147, 543–561 552 S. M. Simiyu and G. R. Keller o 1 N Mt. Elgon M Kenya L. Baringo ELDORET L. Bogoria 0 o Equator regional profile CHE SOL Menengai KISUMU A Eburru 1 S B' L. Naivasha ELM B VIS o A' Mt. Kenya D Olkaria OLO EWA Longonot C C' MAR NTU SUS Mt Suswa NAIROBI MOR L. Magadi o 2 S Tanzania KAJ D' KIB M' Southern regional profile o 34 E CHN L Natron o o 36 E 38 E Figure 9. Index map showing locations of the gravity profiles that were modelled in this study (A–Ak to M–Mk). Shotpoints from seismic profiles are shown as stars (Fig. 2) to indicate the close correspondence between the seismic and gravity models. Also shown are locations of the regional gravity profiles that were modelled by Simiyu & Keller (1997). In order to emphasize crustal features, the density contrast between the rifted and cratonic mantle is large relative to what the seismic velocities suggest (Ravat et al. 1999), so that the depth of the models could be limited to 100 km. By extending the models to 150 km or more based on the teleseismic results (e.g. Achauer et al. 1994; Slack et al. 1994), this density contrast could be reduced. In the seismic models, a depth of 0 indicates sea level, but in the gravity models a depth of 0 corresponds to the lowest elevation of the rift valley floor on a particular profile. Profile A–Ak (Figs 9 and 10) is located near the triple junction between the northern Kenya, Nyanza and southern Kenya rifts. This profile and those that follow sample large topographic variations across the rift in which the surface rocks have widely varying densities based on surface and drill hole density measurements. The profile starts 10 km west of the surface contact between the Quaternary volcanics and the Mozambique belt rocks. Further east it crosses upper Miocene phonolites and then lower Miocene mixed lavas. At the 90 km point, the profile crosses into the rift valley at the northern end of the Mau escarpment and fault zone (Fig. 1). East of this fault system, the profile crosses the rift valley and mainly Pleistocene to Recent pyroclastics. Near the eastern flank of the rift valley, the profile crosses an area with 1–2 km of low-density volcanic deposits and wind-deposited lapilli tuffs (Williams 1978). A hole drilled to a depth of 475 m in this area encountered volcanic soils and tuffs (Mwaura 1990). Further east, the profile crosses the Mount Kenya volcanic edifice and a gravity low that was modelled to be the result of at least 2.5 km of Pleistocene trachytes and tuffs. The gravity high at Menengai, with a 40 mGal amplitude and 35 km width was modelled by a 20 km wide intrusive body at a depth of 4 km (Fig. 10). This anomaly is bounded by gravity gradients that we interpret as being caused by sharp contacts between the dense intrusion and thick low-density tuff and ash. Henry et al. (1990) noted a traveltime advance at Menengai that suggests a fairly large high-velocity body, and the seismic model for the axial line (Fig. 4) indicates high basement velocities in this area. Profile B–Bk (Figs 9 and 11) extends from Mau Narok across the Mau fault zone, the Olkaria geothermal field south of Lake Naivasha, the Aberdare Range and the southern slopes of Mt Kenya. West of the rift valley, a 23 mGal amplitude gravity low is located on the Mau escarpment and is modelled as being the result of a 2.5 km thick layer of Mau tuffs with a density of 2.2r103 kg mx3 (Fig. 11). Nyblade & Pollack (1992) argue that the gravity lows west of the rift valley in this region could reflect the presence of local crustal thickening due to the suture contact between the Archaean Tanzania craton and the Mozambique belt. This particular anomaly is only about 30 km wide, but Tesha et al. (1997) found wider negative anomalies to the south that correlated with this possible suture. # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure A A' –50 Bouguer Anomaly (mGals) 553 Tie to Axial Line Calculated Observed –100 Mt. Kenya Menengai –150 MFZ –200 KFZ –250 0 WEST 50 100 150 200 250 2.4 2.90 2.3 2.74 2.3 300 350 EAST 300 350 2.4 Depth (km) 0 2.70 2.85 20 40 2.70 2.85 2.85 3.05 60 3.12 3.26 3.26 80 100 0 50 100 150 200 250 Distance (km) Figure 10. Integrated gravity model of Earth structure along profile A–Ak, which crosses the rift valley at Menengai volcano (Fig. 9). Numbers indicate density valuesr103 kg mx3. In order to emphasize crustal features, the density contrast in the mantle is very large to limit the vertical extent of the model. MFZ: Mau fault zone; KFZ: Kinangop fault zone. gravity high beneath the Olkaria geothermal field has been modelled as being the result of a 9 km wide intrusion that rises to within 3.5–4.0 km of the Earth’s surface. Profile C–Ck (Figs 9 and 12) starts at Ewaso Nyiro in the west, crosses the Mau escarpment, passes through Suswa caldera and extends to Thika, which is 30 km north of Nairobi. A unique feature of this profile is that gravity values on the western flank of the rift valley are uniform (about x200 mGal) and lower than on the eastern side (Fig. 12). Local gravity lows coincide with exposures of tuffs that have densities as low as 2.0r103 kg mx3. East of Ewaso Nyiro, a negative anomaly can be attributed to 1.5 km of tuffs where a well drilled nearby encountered continuous tuffs and limited trachyte intercalation to the bottom at 597 m (Mwaura 1990). The steep east rift valley boundary fault with a throw of more than 3.5 km is located at model coordinate 120 km. The thickness of volcanic fill in the rift valley decreases eastwards from more than 3.0 km on the western margin against the Mau fault zone to 1.5 km against the Kinangop fault zone at the eastern margin. Wells drilled to a depth of about 3 km on the western part of the rift floor have never reached the Precambrian basement or encountered any form of sedimentary rocks (Roessner & Strecker 1997). The thickness of the volcanics suggests that the master fault for the Naivasha basin region is on the western side of the rift valley. On the eastern margin, the Kinangop tuffs form a 1.0 km thick layer (Roessner & Strecker 1997). Further east, south of Mt Kenya, a gravity low has been modelled as being due to a 2.0 km thick layer of volcanics. The B' B Bouguer Anomaly (mGals) –50 Tie to Axial Line Calculated Observed –100 Mt. Kenya Olkaria volcanic field –150 KFZ MFZ –200 –250 0 WEST 50 2.3 2.3 Depth (km) 100 150 2.90 0 2.4 250 300 2.3 350 EAST 2.4 2.70 2.74 2.70 2.85 20 200 2.85 40 60 3.05 3.26 3.26 3.12 80 100 0 50 100 150 200 250 300 350 Distance (km) Figure 11. Integrated gravity model of Earth structure along profile B–Bk, which crosses the rift valley at the Olkaria volcanic field (Fig. 9). Numbers indicate density valuesr103 kg mx3. In order to emphasize crustal features, the density contrast in the mantle is very large to limit the vertical extent of the model. MFZ: Mau fault zone; KFZ: Kinangop fault zone. # 2001 RAS, GJI 147, 543–561 554 S. M. Simiyu and G. R. Keller C –50 C' Tie to Axial Line Bouguer Anomaly (mGals) Calculated Observed –100 Nairobi Suswa –150 –200 –250 0 WEST 50 2.3 Depth (km) 0 100 2.3 150 200 300 2.90 2.3 2.74 2.4 350 EAST 2.4 2.70 2.68 20 250 2.85 2.85 40 3.05 60 3.26 3.26 3.12 80 100 0 50 100 150 200 250 300 350 Distance (km) Figure 12. Integrated gravity model of Earth structure along profile C–Ck, which crosses the rift valley at Suswa volcano (Fig. 9). Numbers indicate density valuesr103 kg mx3. In order to emphasize crustal features, the density contrast in the mantle is very large to limit the vertical extent of the model. On the eastern shoulder of the rift in the Nairobi–Thika area, a local low in the much broader regional positive anomaly coincides with low-density surface volcanics that thicken from longitude 37.20uE towards the rift. Water wells drilled in the area east of Nairobi (Mwaura 1990) have reached Mozambique gneissic basement at a depth of 530 m, after going through Kerichwa Valley tuffs (80 m thick), Nairobi phonolites (100 m thick), Athi tuffs (320 m thick) and Kapiti phonolites (30 m thick). All these lavas and tuffs have densities less than 2.6r103 kg mx3. To the west of Nairobi, on the escarpment bounding the rift valley, the south Kinangop tuffs form a layer about 1.5 km thick. Within the rift, the positive anomaly is 60 km wide and has an amplitude of 30 mGal. The rift valley area has been modelled as a graben whose fill varies in thickness from 2.7 km in the west to 1.5 km in the east and whose density is 2.4r103 kg mx3. Beneath the Suswa caldera, the gravity high is due to an 18 km wide body at a depth of about 2.5 km with a density of 2.90r103 kg mx3. A difference between this model and models for profiles to the north (Menengai and Naivasha) is that the Tanzania craton is present under a significant portion of the western flank of the rift valley. Thus, based on the work of Tesha et al. (1997), which showed 2.68r103 kg mx3 to be a good choice for the density of the upper crust of the Tanzania craton, we used this lower value to model the upper crust across the western flank. Profile D–Dk (Figs 9 and 13) is an extension of the model presented by Simiyu & Keller (1998) and crosses the southern portion of the rift valley before it splays out in the area of Lake Natron in northern Tanzania (Fig. 1). The location of this profile away from the thick volcanic pile associated with the Kenya dome makes it more favourable for analysing the details of basement structures underlying the Kenya dome that are otherwise obscured by these volcanics. The profile starts within the Archaean Tanzanian craton east of Lake Victoria. At the Oloololo escarpment (model coordinate 100 km), there is a geological, structural and topographic boundary separating surface exposures of the Archaean Tanzania craton and the Neo-Proterozoic Mozambique belt (Fig. 13). East of the escarpment, the profile crosses the flat Mara plains, which are covered by flood trachytes and phonolites of the Isuria Group. Further east the profile crosses the Loita Hills and the major bounding fault of the western part of the rift valley, the Nguruman escarpment (Crossley 1979). The Loita Hills and the Nguruman escarpment are composed of massive, highly fractured quartzite and gneisses of Proterozoic age (Crossley 1979). These quartzites are thought to extend west to the craton–mobile belt boundary beneath the Mara plains (Smith 1994). The profile crosses the rift at coordinate 250 km, where the rift valley appears to have developed within the mobile belt. East of the rift around Kajiado, the profile crosses a band of marble surrounded by some discontinuous quartzites and gneisses (Shackleton 1986). Further east of Kajiado, there are minor layers of volcanics from the Chyulu Hills volcanic complex. Unlike the three profiles to the north, there is only a small short-wavelength gravity high in the rift valley area. This anomaly is flanked by two minima with the western one having the larger amplitude. Our modelling of these anomalies shows that the rift fill is about 3.5 km thick against the Nguruman escarpment to the west and only 2.0 km thick to the east. The basin is divided into two parts by a horst structure that runs along its axis, thus creating the small gravity high. The crystalline upper crust on the western flank has been modelled in two parts: an upper 4 km thick layer with a low density of 2.60r103 kg mx3 (VP=5.6 km sx1; Simiyu & Keller 1998) and a lower layer that is 11 km thick with a typical basement density (2.68r103 kg mx3; VP=5.9 km sx1; Simiyu & Keller 1998). East of the rift, the upper crust has been modelled with a 2.71r103 kg mx3 density based on velocity values (VP=6.3 km sx1; Birt et al. 1997). The lower crust has been modelled using the KRISP 94 seismic results (Birt et al. 1997) as constraints. From west to east, there is little crustal thinning under the rift valley. However, east of the rift, there is crustal thickening from y35 km # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure Bouguer Anomaly (mGals) D D' –50 Calculated Observed –100 Oloololo Escarpment –150 Tie to Axial Line –200 –250 WEST Tanzania Craton Reworked Zone Lake Magadi EAST Rift Valley Mozambique Belt 2.60 2.90 0 2.71 2.68 Depth (km) 20 555 2.85 2.86 2.85 2.88 40 60 3.26 3.26 3.15 80 100 Depth (km) 50 250 NE Lake 2.4 2.5 Morijo Magadi 150 0 2.60 2.65 Kajiado 2.71 2.68 2.73 2.74 15 200 450 2.60 5 10 350 250 2.5 2.75 300 350 Distance (km) Figure 13. Integrated gravity model of Earth structure along profile D–Dk, which crosses the rift valley at Lake Magadi following KRISP 94 seismic line G (Fig. 9). Numbers indicate density valuesr103 kg mx3. In order to emphasize crustal features, the density contrast in the mantle is very large in order to limit the vertical extent of the model. A detailed upper crustal model (15 km depth) for the region outlined by the dashed box in the upper diagram is shown in the lower diagram. NE: Nguruman Escarpment. beneath the rift to about 40 km at the end of the profile. West of the rift, the lower crust is modelled with a density of 2.85 r 103 kg mx3 (VP = 6.7 km sx1; Birt et al. 1997). The lower crustal densities increase from 2.86 r 103 kg mx3 (VP = 6.9 km sx1; Birt et al. 1997) beneath the rift to 2.88 r 103 kg mx3 (VP = 7.2 km sx1; Novak et al. 1997) beneath the eastern flank. Using mantle densities suggested by Hay et al. (1995), the broad gravity low associated with deep structure was modelled primarily by a mantle anomaly (3.15r103 kg mx3) beneath the rift valley. However, west of the rift valley, the combination of the reduced density of the upper crust and slight crustal thickening, probably due to Mozambique belt collision tectonics, also explains some of the regional minimum. The axial profile (M–Mk, Figs 9 and 14) was constructed along the KRISP 85 and KRISP 90 seismic lines and extends from Lake Baringo at 1uN southwards to Lake Magadi. The significance of this profile is that it follows the axis of the rift and the gravity axial high as closely as possible and that it ties all of the E–W profiles together. The profile starts at Lake Baringo within the Baringo–Bogoria sub-basin characterized by recent volcanics and lake sediments (Mugisha et al. 1997). South of Lake Bogoria at model coordinate 50 km (Fig. 14), the profile crosses the proposed Metkei–Marmanet accommodation zone, which is also within the Nyangea–Athi–Ikutha shear zone axis (Smith 1994). South of this accommodation zone, the Nakuru–Naivasha sub-basin extends to Suswa. Seismic # 2001 RAS, GJI 147, 543–561 data (Henry et al. 1990; Mechie et al. 1994) show this basin to have the greatest thickness of rift fill (6 km) beneath the Lake Naivasha area (Fig. 4). The profile crosses areas with deep drill hole data between Lake Elmenteita and Suswa that are underlain by a large thickness of tuffs, trachytes, rhyolites and phonolites. South of Suswa, there is another proposed accommodation zone (Bosworth 1987), and then the profile passes into the Magadi–Natron sub-basin. In this basin, the profile is within the Aswa–Nandi–Loita shear zone, which is a highly faulted and fractured zone of massive quartzitic basement rocks covered by a thin layer of rift-related volcanics (Smith 1994). The main features of the profile (Fig. 14) are the broad negative Bouguer anomaly with a wavelength >350 km and an amplitude of >50 mGal and a series of positive gravity anomalies superimposed on it. These gravity highs (Fig. 8) are aligned along the axis of the rift at the Menengai, Eburru, Olkaria and Suswa volcanoes. Following Simiyu & Keller (1997) and the seismic constraints, the broad gravity low was modelled as resulting from two factors: (1) the mantle is less dense at the apex of the Kenya dome because it is hotter compared to the northern and southern sectors, and (2) a large thickness (1–5 km) of low-density (2.3r103 kg mx3) tuffs and lavas fill the graben associated with the rift valley. The gravity highs were modelled as resulting from a number of volcanic centres underlain by discrete mafic bodies with a density of 2.90r103 kg mx3. These bodies are interpreted to represent upper-level mafic plutons (Omenda 1998). 556 S. M. Simiyu and G. R. Keller M' M Calculated Observed Equator regional profile –100 (mGals) Bouguer Anomaly 0 C-C' B-B' A-A' D-D' –200 SOUTH NORTH 0 Menengai Lake Baringo 0 Depth (km) 200 100 2.70 300 Olkaria 400 Lake Magadi 2.90 Suswa 2.74 2.70 2.70 2.85 20 40 3.05 60 3.15 3.15 3.12 80 100 0 200 Distance (km) 100 300 400 Figure 14. Integrated gravity model of Earth structure along the axis of the rift valley (M–Mk) from Lake Baringo to Lake Natron in northern Tanzania (Fig. 9). Numbers indicate density valuesr103 kg mx3. In order to emphasize crustal features, the density contrast in the mantle is very large to limit the vertical extent of the model. Ties to the E–W profiles are shown. pretation of geological, seismic refraction, gravity and drill hole data. Key points addressed by our analysis are discussed below. DISCUSSION The results of this study show that the southern Kenya rift has a crustal structure that is equally as complex as its tectonic history. To synthesize our results and illustrate the main features we have constructed three schematic geological crosssections (Figs 15–17), and one interpreted map of the major units that crop out or subcrop beneath the volcanic cover (Fig. 18) was constructed. These diagrams represent our integrated inter- Elev. (km) WEST 4 2 Rift basin geometry Because of the deep drill holes associated with geothermal exploration, the best-constrained cross-section can be drawn through the Olkaria region (Fig. 15). In this area, Bosworth et al. (1986) and Bosworth (1987) attributed the sinuous course EAST Rift Valley DH Mau Escarpment Olkaria Kinangop–Kijabe plateau OW305 OW709 OW30 OW704 OW–304 DH DH DH DH D Depth (km) 0 SL 2 4 Upper crust Mozambique Belt 6 8 10 12 14 0 10 20 30 40 50 60 70 80 90 100 Distance (km) Pleistocene–Recent pyroclastics and sediments Pliocene Mau and Kinangop tuffs Precambrian metamorphic basement (upper crust) Plateau trachytes intercalated with tuffs Miocene lavas (thickness estimated) Intrusive body OW–304 Deep geothermal drillholes DH Shallow drillhole Figure 15. An interpreted geological cross-section across the rift valley in the Lake Naivasha–Mt Longonot area. # 2001 RAS, GJI 147, 543–561 Elev. (km) S. Kenya rift upper crustal structure WEST 3 EAST Nguruman Rift Valley Escarpment DH L. Magadi DH 557 DH Kajiado DH DH (SL) 0 Depth (km) 3 Zone of tectonic reworking Mozambique belt 6 9 12 Tanzanian Craton? 15 120 160 200 240 280 320 360 Distance (km) Recent volcaniclastics, and sediments Crystalline upper crust (Precambrian basement) Pliocene / Miocene volcanics and sediments Pleistocene–Recent pyroclastics Mozambique belt metasediments Archean crust Highly fractured quartzitic Mozambique belt rocks Plateau trachytes, DH Drillhole Figure 16. An interpreted geological cross-section across the rift valley in the Lake Magadi area (modified from Simiyu & Keller 1998). crossing the rift valley north of Suswa volcano (Fig. 5). The upper surface of the Pliocene Mau and Kinangop volcanics is exposed along the western boundary of the rift valley, where it is offset by more than 1000 m by the Mau fault zone (Roessner & Strecker 1997). The depth of the Pliocene volcanics beneath the inner graben in drill holes in the Olkaria geothermal field shows pronounced subsidence of this unit to a depth of 1500 m. The overall implication of this result is that the southern and central Kenya rift valley contains a series of asymmetric basins that are all bounded to the west by master faults. In addition, intrarift horst blocks occur near the axis of the rift valley in several locations. To the north of Menengai, the Kerio basin Elev. (km) of the rift valley (Fig. 1) to the en echelon arrangement of faults defining a series of asymmetric half-grabens (basins) that change polarity at accommodation zones. In their model, extension is accommodated along major listric detachments of lithospheric extent that alternate in polarity along the rift. They suggested that the major detachment fault in the Nakuru– Naivasha sub-basin is the Sattima fault (Fig. 1), which lies along the eastern margin of the rift valley, the Aberdare Range. Our results show the graben to be deepest in front of the western margin of the rift valley at the Mau escarpment (Fig. 15). This rift basin geometry is also clearly shown on the cross-section of the Suswa area (Fig. 16) and the seismic model for the profile NORTH 3 L. Baringo Menengai Eburru DH DH Olkaria SOUTH L. Magadi Suswa DH DH DH (SL) 0 Depth (km) 3 6 9 12 15 0 40 80 120 160 240 200 280 Distance (km) Pleistocene–Recent pyroclastics and sediments Miocene lavas (thickness estimated) Intrusive bodies in the upper crust Plateau trachytes intercalated with tuffs Highly fractured quartzitic Precambrian basement Mid–lower crust Pliocene Mau and Kinangop tuffs Metamorphic granitoid Precambrian basement DH Figure 17. An interpreted geological cross-section along the rift axis from Lake Baringo to Lake Magadi. # 2001 RAS, GJI 147, 543–561 Drillhole 558 S. M. Simiyu and G. R. Keller o o 34 E Mt Elgon 38 E ? NA L ? ? I AN ELDORET L. Baringo L. Bogoria ? ? KISUMU Aberdare Range o Mt Kenya Menengai L. Victoria 0 ? L. Naivasha ? ? Suswa NA I ? NAIROBI L. Magadi o yu ng L e Archean Tanzania craton ? Unsure of the buried geology Off—axismajor volcanoes lu Ra AN L. Natron Volcanic ranges 2S Ch Tanzania Craton Kilimanjaro Zone with thrusts, reworked craton and mobile belt rocks Mobile belt gneisses and other metasediments Rift axis caldera volcanoes Outline of the apex the Kenya dome NIA & ANL - Nyangea—Athi-Ikutha &Aswa—Nandi—Loita shear zones Figure 18. Interpreted map of the major units that crop out or subcrop beneath the volcanic cover in the southern rift valley region. Boundaries are based on gravity anomaly characteristics and geophysical modelling together with drilling and geological information (Crossley 1979; Shackleton 1986; Smith & Mosley 1993; Maguire et al. 1994; Smith 1994; Stern 1994). west of Lake Baringo (Fig. 1) is bounded by a master fault on the west and by a horst block on the east that is, in turn, bounded on the east by the relatively shallow Baringo basin (Swain et al. 1981; Maguire et al. 1994). Thus, there is a lack of polarity reversals in rift basins for a distance of about 300 km along the rift valley. This observation contrasts with the western branch of the East African rift system, where polarity reversals are well documented (Rosendahl 1987). The axial gravity maxima and magmatic modification of the crust Our new data and improved density analysis produce a particularly clear picture of the axial gravity high (Fig. 8). As summarized by Swain et al. (1994), early workers interpreted the axial gravity high in the rift valley to be the result of a crustal-scale dyke. The results of KRISP have shown that no continuous feature of this magnitude exists. However, this is not to say that magmatic modification of the crust is negligible. Obviously, the volume and nature of volcanic activity in the region require considerable modification of the crust. However, the nature and extent of this modification is an important issue. As discussed by Keller et al. (1994b), Hay et al. (1995) and Thybo et al. (2000), a significant amount of underplating / intrusion near the base of the crust (Moho) has occurred under the region between Lake Baringo and Lake Magadi (Fig. 14). However, this material extends across the entire rift valley and a portion of the flanks and is the cause of only long-wavelength gravity anomalies (Figs 10–13). Swain (1992) showed that the early KRISP results suggested some densification (distributed intrusions) of the upper crust in the vicinity of the rift valley axis, and as we have shown, subsequent KRISP data are consistent with this interpretation. However, a small level of densification is not enough to explain the axial gravity high, and our analysis and those of Swain (1992), Maguire et al. (1994) and Simiyu & Keller (1998) show that intrarift structures (horst blocks) play a significant role in creating the axial gravity high. Details revealed by the new gravity data presented here show that the magnitude of the axial high changes significantly along its extent. Our seismic and gravity modelling show that a series of local gravity highs are coincident with the major volcanic centres (Figs 8 and 14). The individual anomalies associated with each centre sometimes have half-wavelengths greater than their distance apart so that the anomalies merge. Away from these centres, the axial high can be explained by the combination of an intrarift horst and modest magmatic modification of the upper crust. Thus, the previous perception of a simple continuous axial positive gravity anomaly was the result of sparse data and the alignment of closely spaced volcanic centres across which gravity profiles were constructed. Apart from the intrusions at volcanic centres, the whole crustal section of the area from Elmenteita to Suswa shows evidence for increased density that may result from magmatic injection by dyke swarms as evidenced by drill hole data. This has had the effect of raising the upper crustal density by about 1–2 per cent (from 2.70 to 2.74r103 kg mx3, Fig. 14), thus effectively counteracting the effect of the low-density rift fill. This interpretation is consistent with a southward increase in gravity values from Elmenteita to Suswa even though the area is covered by 6 km of relatively low-density volcanic deposits dominated by ash-flow tuffs, air-fall tuffs and minor phonolites with trachytes. Crustal genesis and the role of older structures Analysis of the gravity maps and modelled seismic and gravity profiles shows different signatures that have led us to classify the crust in the region into broad zones: the Archaean craton, the mobile belt with reworked craton and the Proterozoic mobile belt (Fig. 18). Regional relationships (Smith 1994) and the seismic and gravity models (Fig. 16 and Birt et al. 1997) show that the mobile belt /reworked craton layer is about 4 km thick and ends near the western margin of the rift valley. Stern (1994) proposed an alternative Precambrian collision model that suggests that more than one orogenic event involving smaller land masses occurred in the southern Kenya area. This suggestion implies that there is not a single suture contact zone between the Archaean craton and the mobile belt but rather a wide zone of imbricate thrusting and suturing. Gravity profiles constructed during this study do not clearly show the characteristic suture signature at the surface contact between the Archaean craton and the mobile belt that is seen in northern Tanzania (Tesha et al. 1997) and north of the equator (Nyblade & Pollack 1992). This may imply that the reworked and buried craton of Smith & Mosley (1993) is actually part of the small land masses that were involved in the multicollision events of Stern (1994). # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure The Kenya rift occurs in an area between contrasting lithosphere types (craton and mobile belt) that are expected to have different mechanical strengths. In the event of applied stresses, failure is likely to occur at such a boundary. Another source of weakness is the likelihood that during Mozambique belt collisional tectonics, there was crustal thickening near the suture as evidenced in gravity studies (Nyblade & Pollack 1992; Tesha et al. 1997). Smith & Mosley (1993) and Hetzel & Strecker (1994) have suggested that the overall rift trend closely follows large- and small-scale basement structures. This study provides further evidence that localization of rifting was controlled by the existing contact between the Archaean craton and the Proterozoic mobile belt. The Kenya dome gravity low The Kenya dome is located at the confluence of structures that resulted from a variety of tectonic events. The area has been affected by (1) thrusting related to one or more collisional events during the formation of the Mozambique belt, (2) deformation associated with its location between the Nyangea–Athi–Ikutha shear zone to the north and the Aswa–Nandi–Loita shear zone to the south (Fig. 18), which are major NW–SE-trending zones of ductile shear and thrusting in the underlying Precambrian basement, and (3) sub-Miocene uplift and doming followed by magmatism and N–S-trending rift faulting. All these events must have some influence on the gravity anomalies observed today. The Bouguer gravity map, bandpass filtered maps and the modelled gravity and seismic profiles show that the gravity low associated with the Kenya dome is the result of several features. At the apex of the Kenya dome, Cenozoic low-density tuffaceous volcanic rocks reach thicknesses of 2 km on the rift flanks (Baker et al. 1988) and 6 km within the rift valley (Henry et al. 1990). Thus some of the relief of the dome is in fact due to volcanic construction. The bandpass filtered map (Fig. 8) shows that part of the gravity low is contributed by low-density material on the Mau escarpment and in the Aberdare Range. This low-density material is associated with a number of small volcanic centres, but since they are closely spaced, their anomalies merge and appear as a single long-wavelength anomaly. An example of this phenomenon can be seen on the eastern rift flank where the Aberdare Range volcanic centres form a series of closely linked silicic volcanoes (Smith 1994). The overall effect of these is to create a long-wavelength gravity low that merges with the adjacent low due to the rift fill. Crustal thickening by about 5 km due partly to underplating /intrusion near the Moho (Keller et al. 1994b; Hay et al. 1995) and pre-existing thickened crust caused by Precambrian collisional tectonics (Nyblade & Pollack 1992; Tesha et al. 1997) also contribute to the gravity low. On a regional basis, crustal thickening provides Airy-type (Banks & Swain 1978) compensation for the dome. In the area west of Lake Magadi, the uppermost Precambrian crust is a relatively low-density layer that was probably structurally emplaced, adding to the crust from above. The extent of this layer shown in Fig. 18 is based on the persistence of a characteristic gravity minimum that also contributes to the Kenya dome gravity low. The low-density mantle beneath the apex of the dome is the major contributor to the long-wavelength component of the gravity low. Simiyu & Keller (1997) showed that the uppermost # 2001 RAS, GJI 147, 543–561 559 mantle density varies along the rift by 0.03r103 kg mx3, decreasing from the Lake Baringo region to a minimum at the apex of the Kenya dome (Lake Naivasha). South of Naivasha, the density increases towards Lake Magadi (Fig. 14). The implication of this density variation is that the mantle is hotter beneath the apex of the Kenya dome and provides part of the isostatic compensation for the Kenya dome via a Pratt-type mechanism (Banks & Swain 1978). Crustal symmetry with respect to the Kenya dome The KRISP 90 axial profile showed a dramatic increase in crustal thickness from the north (Lake Turkana) to the Kenya dome (Mechie et al. 1994) that is the result of the thickening of a high-velocity layer at the base of the crust. This layer was interpreted to be due to magmatic underplating /intrusion at the base of the crust (Keller et al. 1994b; Hay et al. 1995). If the crustal thickening at the apex of the dome is only due to underplating as a result of recent magmatic activity, then the crust should also thin southwards away from the dome. However, this study and the analysis of Birt et al. (1997) and Simiyu & Keller (1997) show that the crustal thickness does not decrease greatly to the south away from the apex of the dome and that the low-velocity /density mantle does not reach the base of the crust as happens to the north. However, a highvelocity layer at the base of the crust extends for more than 200 km outside the rift in the Lake Magadi area (Birt et al. 1997). This suggests that the relatively thick crust beneath the Kenya dome may be only partly due to underplating of highvelocity material. We believe that the most likely explanation for areas of thick crust not due to underplating is pre-existing variations in crustal thickness resulting from the formation of the Mozambique belt. These pre-existing zones of thickened crust may have created an ideal weak zone for rift propagation. CONCLUSIONS Our integrated interpretation of the crustal structure of the southern Kenya rift shows that the graben master faults are all located along the western flank of the rift valley, and there are no half-graben polarity reversals. There is no symmetry in crustal thickness variations with respect to the apex of the Kenya dome, and the existing crustal thickness is in part due to pre-rift crustal type and thickness in addition to underplating / intrusion and extension. The pre-existing lithospheric contrast between the Archaean and Proterozoic basement terranes played a significant role in the location and structural geometry of the rift. Low upper crustal velocities and densities observed under the rift flank west of Magadi probably represent material added to the top of the Archaean Tanzania craton by tectonic transport. Magmatic modification of the upper crust is modest except near the major Quaternary volcanic centres. An intrarift horst block that extends along at least most of the rift axis and this series of volcanic centres create gravity maxima that together produce the axial gravity high. ACKNOWLEDGMENTS We would like to acknowledge the help and input provided by our colleagues in the KRISP Working group. We also wish to thank Dr Jeff Karson of the Division of Earth and Ocean Science for his constructive review of the original manuscript as 560 S. M. Simiyu and G. R. Keller well as the reviewers of this paper, who provided many helpful and constructive comments. The Kenya Electricity Generating Company helped support the gravity surveying. The seismic data analysed in this study were largely gathered using equipment provided by the Incorporated Research Institutions for Seismology. Major funding was provided by NSF grant EAR-9316868. Additional support was provided by Conoco Inc and NASA through its support of the Pan-American Center for Earth and Environmental Studies at the University of Texas at El Paso. REFERENCES Achauer, U., 1992. A study of the Kenya rift using delay-time tomographic analysis and gravity modeling, Tectonophysics, 209, 197–207. Achauer, U. et al., 1994. New ideas on the Kenya rift based on the inversion of the combined data set of the 1985 and 1989 /90 seismic tomography experiments, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C. Keller, G.R. & Khan, M.A., Tectonophysics, 236, 305–330. Baker, B.H., 1986. Tectonics and volcanism of the southern Kenya rift valley and its influence on sedimentation, in Sedimentation in the African Rifts, eds Frostick, L.E., Renaut, R.W., Reid, I. & Tiercelin, J.-J., Geol. Soc. Lond. Spec. Publ., 25, 45–57. Baker, B.H. & Wohlenberg, J., 1971. Structure and evolution of the Kenya Rift Valley, Nature, 229, 538–542. Baker, B.H., Mohr, P.A. & Williams, L.A.J., 1972. Geology of the eastern rift system of Africa, Geol. Soc. Am. Spec. Pap., 136, 1–67. Baker, B.H., Mitchell, J.G. & Williams, L.A.J., 1988. Stratigraphy, geochronology and volcano-tectonic evolution of the KedongNaivasha-Kinangop region, Gregory rift valley, Kenya, J. geol. Soc. Lond., 145, 107–116. Banks, R.J. & Swain, C.J., 1978. The isostatic compensation of East Africa, Proc. R. Soc. Lond., A364, 331–352. Barton, P.J., 1986. The relationship between seismic velocity and density in the continental crust—a useful constraint?, Geophys. J. R. astr. Soc., 87, 195–208. Birt, C.S., Maguire, P.K.H., Khan, M.A., Thybo, H., Keller, G.R. & Patel, J., 1997. The influence of pre-existing structures on the evolution of the southern Kenya rift valley–evidence from seismic and gravity studies, Tectonophysics, 278, 211–242. Bosworth, W., 1987. Off-axis volcanism in the Gregory rift, East Africa: implications for models of continental rifting, Geology, 15, 397–400. Bosworth, W., Lambiase, J. & Keisler, R., 1986. A new look at Gregory’s rift: the structural style of continental rifting, EOS, Trans. Am. geophys. Un., 67, 577–582. Cady, W.J., 1980. Calculation of gravity and magnetic anomalies of finite-length right polygonal prisms, Geophysics, 45, 1507–1512. Clarke, M.C.G. & Woodhall, D.G., 1987. Geological and geothermal mapping, BGS Exploration for Geothermal Energy Project Progress Rept, BGSKEN /2, Rept to the Ministry of Energy, Kenya. Crossley, R., 1979. The Cenozoic stratigraphy and structure of the western part of the rift valley in southern Kenya, J. geol. Soc. Lond., 136, 393–405. Fairhead, J.D., 1976. The structure of the lithosphere beneath the Eastern rift, East Africa, deduced from gravity studies, Tectonophysics, 30, 269–298. Green, W.V., Achauer, U. & Meyer, R.P., 1991. A three-dimensional seismic image of the crust and upper mantle beneath the Kenya rift, Nature, 354, 199–201. Hay, D.E., Wendlandt, R.F. & Keller, G.R., 1995. The origin of Kenya rift plateau-type flood phonolites: integrated petrologic and geophysical constraints on the evolution of the crust and upper mantle beneath the Kenya rift, J. geophys. Res., 100, 5435–5444. Henry, W.J., Mechie, J., Maguire, P.K.H., Khan, M.A., Prodehl, C., Keller, G.R. & Patel, J., 1990. A seismic investigation of the Kenya Rift Valley, Geophys. J. Int., 100, 107–130. Hetzel, R. & Strecker, M.R., 1994. Late Mozambique belt structures in western Kenya and their influence on the evolution of the Cenozoic Kenya rift, J. struct. Geol., 16, 189–201. Keller, G.R., Mechie, J., Braile, L.W., Mooney, W.D. & Prodehl, C., 1994a. Seismic structure of the uppermost mantle beneath the Kenya rift, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 201–210. Keller, G.R. et al., 1994b. The East African rift system in light of KRISP 90, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 453–464. Luetgert, J.H., 1992. MacRay—interactive two-dimensional seismic raytracing for the Macintosh, USGS Open File Rept, 92–356. Maguire, P.K.H., Swain, C.J., Masotti, R. & Khan, M.A., 1994. A crustal and uppermost mantle cross-sectional model of the Kenya rift derived from seismic and gravity data, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 217–249. Mechie, J., Keller, G.R., Prodehl, C., Gaciri, S., Braile, L.W., Mooney, W.D., Gajewski, D. & Sandmeier, K.J., 1994. Crustal structure beneath the Kenya rift from axial profile data, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 179–199. Mugisha, F., Ebinger, C.R., Strecker, M. & Pope, D., 1997. Two-stage rifting in the Kenya rift: implications for half-graben models, in Stress and Stress Release in the Lithosphere, eds Fuchs, K., Altherr, R., Muller, B. & Prodehl, C., Tectonophysics, 278, 63–81. Mwaura, M.N., 1990. Bore holes drilled for groundwater in the southern rift, Kenya, Geology, Maintenance, Productivity and Water Quality, Rept to the Ministry of Water Development, Nairobi, Kenya. Nettleton, L.L., 1939. Determination of density for reduction of gravimeter observations, Geophysics, 4, 176–183. Novak, O., Prodehl, C., Jacob, B. & Okoth, W., 1997. Crustal structure of the Chyulu Hills, southeastern Kenya, in Stress and Stress Release in the Lithosphere, eds Fuchs, K., Altherr, R., Muller, B. & Prodehl, C., Tectonophysics, 278, 186–210. Nyblade, A.A. & Pollack, H.N., 1992. A gravity model for the lithosphere in western Kenya and northeastern Tanzania, Tectonophysics, 212, 257–267. Omenda, P.A., 1994. The geological structure and reservoir characteristics of the Olkaria West geothermal field, Proc. Stanford Geothermal. Reservoir Eng. Workshop, 19, 125–130. Omenda, P.A., 1998. The geological structure of the Olkaria geothermal field, Geothermics, 19, 125–130. Prodehl, C., Keller, G.R. & Khan, M.A., eds, 1994. Crustal and upper mantle structure of the Kenya Rift, Tectonophysics, 236, xxx–xxx. Prodehl, C., Ritter, J., Mechie, J., Keller, G.R., Khan, M.A., Fuchs, K., Nyambok, I.O. & Obel, J.D., 1997. The KRISP 94 lithospheric investigation of southern Kenya-the experiments and their main results, in Stress and Stress Release in the Lithosphere, eds Fuchs, K., Altherr, R., Muller, B. & Prodehl, C., Tectonophysics, 278, 121–148. Ravat, D., Lu, Z. & Braile, L.W., 1999. Velocity-density relationships and modeling the lithospheric density variations of the Kenya rift, Tectonophysics, 302, 225–240. Roessner, S. & Strecker, M.R., 1997. Late Cenozoic tectonics and denudation in the Central Kenya Rift: quantification of long-term denudation rates, in Stress and Stress Release in the Lithosphere, eds Fuchs, K., Altherr, R., Muller, B. & Prodehl, C., Tectonophysics, 278, 83–94. Rosendahl, B.R., 1987. Architecture of continental rifts with special reference to East Africa, Ann. Rev. Earth planet. Sci., 15, 445–503. # 2001 RAS, GJI 147, 543–561 S. Kenya rift upper crustal structure Searle, R.C., 1970. Evidence from gravity anomalies for thinning of the lithosphere beneath the Rift Valley in Kenya, Geophys. J. R. astr. Soc., 21, 13–31. Shackleton, R.M., 1986. Precambrian collision tectonics in Africa, in Collision Tectonics, eds Coward, M.P. & Ries, A.C., Geol. Soc. Spec. Publ., 19, 329–349. Simiyu, S.M., 1990. Geophysical interpretation of the Eburru geothermal field, UNU GTP, 216, 1–39. Simiyu, S.M. & Keller, G.R., 1997. Integrated geophysical analysis of the East African Plateau from gravity and recent seismic studies, in Stress and Stress Release in the Lithosphere, eds Fuchs, K., Altherr, R., Muller, B. & Prodehl, C., Tectonophysics, 278, 291–314. Simiyu, S.M. & Keller, G.R., 1998. Upper crustal structure in the vicinity of Lake Magadi in the Kenya Rift Valley, J. African Earth Sci., 27, 359–371. Slack, P.D. et al., 1994. Attenuation and velocity of P-waves in the mantle beneath the East African rift, Kenya, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 179–199. Smith, M., 1994. Stratigraphic and structural constraints on mechanisms of active rifting in the Gregory rift, Kenya, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 3–22. Smith, M. & Mosley, P.N., 1993. Crustal heterogeneity and basement influence on the development of the Kenya rift, East Africa, Tectonics, 12, 591–606. # 2001 RAS, GJI 147, 543–561 561 Stern, R.J., 1994. Arc assembly and continental collision in the NeoProterozoic East African orogen: implications for the consolidation of Gondwanaland, Ann. Rev. Earth planet. Sci., 22, 319–351. Swain, C.J., 1992. The Kenya rift axial gravity high: a re-interpretation, Tectonophysics, 204, 59–70. Swain, C.J., Khan, M.A., Wilton, T.J., Maguire, P.K.H. & Griffiths, D.H., 1981. Seismic and gravity surveys in the Lake Baringo-Tugen Hills area, Kenya rift valley, J. geol. Soc. Lond., 138, 93–102. Swain, C.J., Maguire, P.K.H. & Khan, M.A., 1994. Geophysical experiments and models of the Kenya rift before 1989, in Crustal and Upper Mantle Structure of the Kenya Rift, eds Prodehl, C., Keller, G.R. & Khan, M.A., Tectonophysics, 236, 23–32. Tesha, A.L., Nyblade, A.A., Keller, G.R. & Doser, D.I., 1997. Rift localization in suture-thickened crust: evidence from Bouguer gravity anomalies in northeastern Tanzanian, East Africa, in Stress and Stress Release in the Lithosphere, eds Fuchs, K., Altherr, R., Muller, B. & Prodehl, C., Tectonophysics, 278, 315–328. Thybo, H., Maguire, P.K.H., Birt. C. & Perchuc, E., 2000. Seismic reflectivity of the lower crust and magmatic underplating beneath the Kenya rift, Geophys. Res. Lett., 27, 2745–2749. Williams, L.A.J., 1978. Character of Quaternary volcanism in the Gregory Rift Valley, in Geological Background to Fossil Man, pp. 55–69, ed. Bishop, W.W., Geological Society of London, London.