The Curie point depth distribution in the central to southern Kenya Rift is investigated using aeromagnetic data and the spatial variations of these depths were mapped for the first time. We produced the analytical signal map from residual aeromagnetic data of the area, which clearly defined the tectonic provinces (Tanzanian Craton, Mozambique belt and Rift Valley) that were better resolved in the southern area. For spectral analysis, we selected radially averaged spectra within 100km x 100km blocks with a 50% overlap. The result from this analysis masked the rift floor. Based on size of the tectonic provinces, we used 50km x 50km block (that defined rift floor) and 25km x 25km (that recognized narrow thermal structures in the central Kenya Rift). We obtained average depths of 15km in rift valley, 45km along Mozambique belt and 52km in the Tanzanian Craton. In addition, three shallow and isolated Curie point depths were determined along the rift valley at Lakes Magadi (15.3 and 14.8km); Naivasha (14km) and Baringo (17.2km). We have interpreted these shallow Curie point depth areas as magmatic intrusions reflecting regional thermal structure in central and southern Kenya rift that is mainly controlled by the rifting tectonic (asthenospheric upwelling). Our analysis of the earthquake data revealed that the areas with shallow Curie point depth are characterized by shallow focal depth of swarm earthquake activities. Therefore, we safely conclude that the earthquakes are results of movement of or within magmatic intrusions and indicate rifting tectonics, at least, along these zones.
Key words: Aeromagnetic data, curie isotherm, spectral methods, thermal structure, seismic velocity, seismicity, Kenya Rift
1. Introduction
Heat flow measurements in near earth surface provide an estimate to the crustal thermal structure (Tanaka and Ishikawa, 2005). For this reason, the spatial variation of temperature as an approach to study the thermal field structure in the earth's crust has received considerable attention in the literature (e.g. Trifona, et.a al., 2009 and references therein). Indeed, the high density of seismic events (earthquakes) within the crust and the correlation of maximum focal depth with surface heat flow suggest that the temperature may be a major factor controlling the mechanical strength (Sibson, 1982). Although this explanation is generally acceptable, heat flow values are from relatively low depths and their measurements are geographically unevenly distributed making them insufficient to define regional thermal structures particularly in areas that cover several (varied) tectonic settings. In order to ameliorate this limitation, over the last years, magnetic spectral analysis has been applied to estimate curie point depths from the earlier works by Spector and Grant, 1970; Bhattacharyya and Leu, 1975. Similar works were extended to cover territories such Japan (e.g. Okubo et al., 1989), USA (e.g. Blakely, 1988), Greece (e.g. Stampolidis and Tsokas, 2002), Portugal (e.g. Okubo et al., 2003), central Europe (Chiozzi, 2005) etc. It is notable that such works have received no attention in Africa. The method allows determination of the bottom of ensemble of magnetic sources which could be either the depth at which ferromagnetic minerals pass to a paramagnetic state under the effect of increasing temperature or the depth of compositional change where magnetic rocks are replaced with non-magnetic. The obtained bottom depth of the magnetic sources is assumed to be the Curie point depth. The thermal conditions implied by the heat flow data and basal depth of the magnetic sources have different limitations and sample sizes and thus provide complementary information (Tanaka and Ishikawa, 2005). Never the less, transforming regional magnetic data sets to the Fourier domain and analyzing their spectra allows us to obtain information about deep magnetic discontinuities associated with the regional temperature distribution.
The present study area covers the Central to southern Kenya rift, which is part of the active East African Rift System (Fig. 1). The crustal structure of this part of the rift is relatively well known from studies on KRISP refraction profiles ((KRISP Working Party, 1991; Birt et al., 1997; Prodehl et al., 1997). The Moho depth beneath the rift axis increases from ~33 to ~35 km from Lakes Baringo to Magadi. From an E-W profile crossing Lake Magadi, Birt et al. (1997) found a mid-crustal boundary at ~12 km where the P-wave velocity is ~6.44 km s−1. They also evidenced a thick lower crustal layer of high velocity (7.1-7.2 km s−1), probably related, beneath the rift, to magmatic under plating that could explain the small amount of crustal thinning observed in this region.
In this paper variations in the magnetic basement are estimated using magnetic spectral analysis. Results are compared with known tectonic provinces in terms of crustal structure and seismicity to infer regional thermal structure and driving mechanism of seismotectonics. A discussion is provided on choice of the size of block for magnetic spectral analysis versus size of tectonic provinces.
2. Regional tectonic setting and geology
The Kenya Rift (also known as Gregory Rift) is considered the archetype of a continental rift in the initial stage of continental breakup (Chorowicz, 2005). It forms part of the East African Rift System (EARS) (Fig. 1), which comprises a series of rift zones stretching more than 3000 km from the Afar triple junction in the North to the Zambezi river in southern Africa. The rift system bifurcates around the Tanzanian Craton, which coincides with the uplifted East African plateau. The Kenya Rift transects the Kenya Dome, which itself is superimposed on the eastern margin of the East African plateau and is located close to the boundary of the Tanzanian Craton and the Pan-African Mozambique shear belt (Baker et al., 1972). Kenya Rift in Northern Tanzania widens into a broad depression and becomes indistinctly defined.
Fig.1(a). Shows location of Kenya Rift and key features: ASWA-Nandi-Loita and Aberdare Detachment including KRISP seismic line G and axial line. Insert is map of Africa showing East African Rift System (EARS) and Kenya Rift. Fig.1(b). SRTM DEM Showing location study area, tectonic structures marked by rift escarpment, major geothermal systems (black dots are volcanoes, light blue dots are hypersaline hotsprings and white dot is geothermal system under exploitation) and red dots are background seismicity (catalogue from year 1993-1996 and 1999-2001).
According to Baker (1987) there are three major stages involved in the evolution of EARS, which are accompanied by intense alkaline volcanism in its tectonic development: (1) the pre-rift stage (30-12Mya), forming deformation and minor faulting (2) the half-graben stage (12-4 Mya), forming of the main boundary faults, and (3) the graben stage(<4 Mya) with an increase and inward migration of faulting. The volcanism and rifting of the Kenya Rift began 40 -45 Ma and ~ 25Ma ago in the north and propagated to the south. According to Atmaoui and Hollnack (2003), with time, there is a north - south rift propagation and a shift in volcanism and tectonic activities from the rift margins towards the axial part of the rift floor, which is accompanied by high fault density. Therefore, the tectonic activity is mainly concentrated in the axial part of the central (lake Baringo area) to southern Kenya rift (Lake Magadi Area). They also investigated the tectonic extension direction of southern Kenya rift (Lake Magadi) using structural analysis of open fractures and size characteristics of faults and indicated an E-W to ESE-WNW normal faulting extension direction as against the NE-SW extensional stress proposed by Bosworth et al. (1992). Neotectonics studies of the Rift reveals four sets of faults: normal N-S fault, dextral NW-SE fault, strike slip ENE-WSW fault and sinistral NE-SW fault (Le Turdu et al., 1999) and these fault systems are deforming under E-W normal extension and are potential sources of earthquakes probably related to magmatic intrusion (Kuria et al., 2010).
Geologically, the rift is comprised of Cenozoic volcanic and sedimentary rocks. Baker et al. (1971) grouped the Cenozoic rocks into four litho-stratigraphic units, mainly: Miocene basalts, Miocene phonolites, Pliocene and Quaternary volcanic rocks and sedimentary rocks. The crustal structures of the Kenya rift have been investigated using integrated seismic, magnetic and gravity data . A 5km- deep sediment, volcanic filled basin and thinning crust of about 8km in a 100km- wide zone beneath the rift valley were discovered by Braile et al. (1994) using a 2D seismic velocity model. An integrated seismic-refraction/teleseismic survey by Keller et al. (1994a), undertaken to study the deep structure beneath the Kenya rift revealed that the rift is associated with sharply defined lithospheric thinning and very low upper mantle velocities down to depths of over 150 km. Mechie et al.(1994) after a detail study of the structures of the rift from south to north showed crustal variations along the rift axis from 35km in the south beneath the Kenya dome around Lake Naivasha to 20km in the north beneath the Turkana region. This variation in crustal thickness along the rift axis has been attributed to tertiary rifting episode. Most of the Kenya Rift International Seismic Project (KRISP) investigations revealed increase in the crustal extension to the north and the presence of low primary velocities anomalies which are possibly caused by magma rising from below and get trapped in the uppermost mantle (Keller et al., 1994b; Mechie et al., 1994).
Despite the vast evidence that the central and southern Kenya rift is a region of high geodynamic activity expressed by recent volcanism, geothermal activity and a high rate of seismicity (Ibs-von Seht et al., 2001) no efforts have been made to understand the thermal structure of the area using aeromagnetic data. It is on this premise that this project was undertaken.
3. Aeromagnetic data
The aeromagnetic dataset used in this research is part of the African Magnetic Mapping Project (AMMP) which compiled airborne magnetic data covering most parts of Africa (Barrit, 1993). The aeromagnetic data of the study area was acquired in the year 1987 by Compagnie Générale de Géophysique, a French Seismic Acquisition and Processing Services Company with line spacing of 2km and flight direction of 90° running in W-E direction at a flying height of 2896 m above mean sea level. The magnetic data was pre-processed by AMMP that included leveling correction and regional field removal and then creating a 1km by 1 km cell size with a projection system of AMMP grid.
3.1 Analytic signal analysis
The analytic signal is not only useful in delineating geological contacts (e.g. Kuria et al., 2010), but also the structural interpretation of aeromagnetic data since the 1980s (e.g. Roest et al., 1992). In addition, analytical signal is an alternative to reduction to the pole (which exaggerates N-S structures) at low magnetic latitudes i.e. near equator. This primarily because analytic signal outlines magnetic sources independent of the direction of magnetization ((Roest et al., 1992).
The AMMP 1km by 1km data available was imported into a database of Oasis montaj (Geosoft software) for further processing. The database consists of lines and channel data. Each line covers a distance of 615km running in W - E direction. A subset of these data was created covering central and southern Kenya rift and adjacent areas. The residual data were interpolated using the minimum-curvature method (Briggs, 1974) from the original survey spacing onto a 0.5km grid and contoured (Fig.2a). In order to delineate the tectonic provinces in our study area, we used the gridded data sets of the geomagnetic anomaly field to determine the analytical signal of the total field (Roest et. al., 1992) for our application. The results from analytic signals shows three distinct anomaly interpreted from west as Tanzanian Craton, Rift Valley and Mozambique Belt (Fig.2b).
3.2 Spectral analysis
The spectral analysis techniques are used to separate influences of different body parameters in the observed magnetic anomaly field as a method of determining Curie point depths. The signal from the top surface of a magnetized body dominates the signal from the bottom at all wavelengths which makes the inverse problem more complicated (Blakely, 1995).
Magnetic data, from which the effect of the main field and external current systems are removed, contains information down to the depth where rocks lose their magnetization either due to compositional or temperature changes (e.g. Rajaram, et al., 2009). Analyzing the long wavelength part of the magnetic data can provide information about this depth. Ravat et al. (2007) discussed several methods used to estimate the depth at which rocks lose their magnetization from the azimuthally averaged Fourier spectra summarized as: the centroid method (Bhattacharyya and Leu, 1975; Okubo et al., 1985; Ibrahim et al., 2005), the spectral peak method (Spector and Grant, 1970; Blakely, 1988; Connard et al., 1983; Ross et al., 2006), the power law corrections (Pilkington and Todoeschuck, 1993; Maus and Dimri, 1995) etc. For noisy data the spectral method may be the only way to determine the depth as the other direct methods will have problems dealing with white noise (Odegard and Dickson, 2004).
Fig.2(a). Map of the geomagnetic anomaly field (ΔT)of the covering central to southern Kenya rift: Tanzania Craton is characterized by relatively quiet geomagnetic field - several broad mainly negative anomalies are observed on western part with amplitudes of -1.5 to -75 nT. The rift floor has intensive positive anomalies (40 - 90 nT), which are high frequency sources. Patches of low magnetic anomalies (-130 to -240 nT) mark location of the Mozambique Belt. Fig.2(b). Map of the analytic signal of the total magnetic anomalies (from AMMP 1km x 1km cell size) correlated with Tanzania Craton, Mozambique Belt and Rift Valley. Notice the sharp boundaries between these tectonic provinces that define or constrain the three 50km wide blocks used for radially averaged spectral analysis.
Here, we calculate the Curie isotherm depths using spectral peak method (Anand and Rajaram, 2007; Rajaram et al., 2009).
In the Fourier domain, the power density spectrum (Blakely, 1995) of the observed magnetic field (ΔTxy) is given by:
(1)
where ФΔT and ФM are power density spectra of the observed total field anomaly and magnetization respectively, Θm and Θf are orientations of magnetization and regional field, Cm is proportionality constant, k is the wave number ( , kx and ky are the wave numbers in the x and y direction), Zt and Zc are respectively depth to the top and bottom of the magnetic sources. The radial averages of Θm and Θf in Eq. (1) are constants and all other terms are radially symmetric. This 2D power spectrum is averaged within concentric rings about the origin, which transforms Eq. (1) into 1D spectrum
Ф∆T|k|=AФM|k|exp(-2|k|Zt)(1-exp(-|k|(Zc-Zt)))²
(2)
Ф∆T|k|=Bexp(-2|k|Zt)(1-exp(-|k|(Zc-Zt)))²
where A is a constant that depends on the orientations of magnetization and regional field. If the magnetization (M(x,y)) is random and uncorrelated, ФM(kx,ky) is a constant and Eq. (2) takes the form
(3)
This one dimensional power spectrum of the potential field due to a prismatic body has a broad spectrum whose peak location is a function of depth to top and bottom surfaces, and whose amplitude is determined by its magnetization. Taking the logarithm on both sides of Eq. (3) yields
log Ф∆T|k|= logB - 2|k|Zt + 2 log 1-exp(-|k|(Zc-Zt)))
(4)
At wavelengths less than about twice the thickness of the layer, the curve of the above equation can be equated to a straight line with slope equal to −2Zt. Thus the depth to the top of the sources can be estimated from the power-density spectrum (Blakely, 1995) of the observed field (ΔTxy) by radially averaging the spectrum within rings concentric about |k|=0, and fitting a straight line through the high wave number part of the radially averaged spectrum. The important assumption made in averaging around the origin is that the statistical properties of M(x,y) are not directionally dependent, implying M(x,y) is random. This is particularly true for large regions as the one considered in the present scenario, where the geological formations range in age from Archean (within craton) to Recent (Cenozoic Rift).
It is very difficult to estimate the depth to the bottom of the magnetic sources, Zc, as the spectra in the Fourier domain are dominated at all wavelengths by the contribution from the shallower parts. The spectrum of a bottomless prism peaks at zero wave number (Bhattacharyya, 1966). The limited depth extent of the body leads to a maximum in the power spectrum and when a significant spectral maximum does occur, indicating that the source bottoms are detectable, the wave-number of this maximum Kmax is related to the depth to the bottom of the source (Curie isotherm), Zc and the depth to the top of the source, Zt, by the following relation (Blakely, 1995):
Kmax=(log Zc - logZt)/(Zc - Zt)
(5)
A well-defined spectral peak exists only if the dimensions of the selected windows are large enough to detect the spectral signature of the source bottoms. Estimates of the thickness of the magnetized portion of the earth's crust suggest that there are two types of lower boundaries of the layer of the magnetized rocks. The first type of boundary corresponds to vertical changes in crustal composition and the second type, where high temperatures at depth cause the rocks to lose their ferromagnetic properties (Connard et al., 1983). In very low heat flow regions (eg. Craton areas), the Curie isotherm may lie deeper than the Moho and because mantle rocks are non-magnetic, the depth to bottom in those regions corresponds to Moho (Saad, 1969) rather than the actual Curie depth.
For the purpose of the spectral analysis, we created a subset of 100km by 100 km blocks from the magnetic database for the central to southern Kenya rift. The blocks were made with an overlap of 50%, sampling part of the Tanzania Craton together with Kenya Rift on the western block and Kenya rift and Mozambique belt on the eastern block. The two blocks (eastern and western) covered a stretch of 150 km, this is the width of the available dataset. For each 100km by 100 km block the residual data were interpolated, using the minimum-curvature method (Briggs, 1974), from the original survey spacing of 2km onto a 0.5-km grid and colour contoured using histogram equalization at illumination and declination angles of 45°. Before transforming to wave number domain (applying the forward FFT), the data was detrended to remove the long wavelengths by taking out the first-order surface that best fits the data in a least square sense (Connard et al., 1983). We then tapered the map data in order to smooth gradually the intensity to zero at the edges of map data. A 10 point Hamming window was used. In addition, we calculated a 200km upward continuation using Fourier domain technique (Blakely, 1995) to remove or minimize the effects of the shallow sources and noise in grids.
From this new grid, the 2-D DFT transform was computed using the algorithm of Press et al. (1992), and the power spectra calculated. Once the power spectrum is estimated, the next step is to calculate the radial power spectrum. This is a radially averaged 2-D spectrum and is estimated by averaging the power spectrum over a set of concentric rings about k=0 with increasing radius (Naidu and Mathew, 1998). Finally, the radial power spectrum is normalized with respect to the value of the first harmonic, the natural logarithm calculated, and then the new data are plotted vs. averaged wavenumber (Fig. 3). This procedure was followed for each of the 10 blocks.
The typical radially averaged power spectrum of all the 10 blocks selected blocks show maxima implying that the source bottom is detectable. After computing the radially averaged power spectrum of the aeromagnetic data best-fit lines were drawn using linear fits on the steepest segment of the spectra (Fig. 3). A change in gradient in all the spectral plots occurs around 0.4cycles/km. The higher wave number for the steepest part of the spectra has therefore been constrained to lie within 0.4cycles/km. The slope of the straight-line segment gives the depth to the top of the deep-seated magnetic sources.
Once the depth to the top of the top (Zt) is calculated, the depth to the bottom (Zc) is estimated using Eq. 5 and solving for Zc. Block 3 did not yield a well defined spectra peak and was avoided for further calculation. The location of the centre of the 10 blocks used for calculation of the power spectrum is marked with (+) on the map of the depth to the bottom of the ensemble of magnetic sources derived from aeromagnetic data together with Curie point depth for each block (Fig. 4a). The results of the 10 blocks showed a minim curie point values of 38 kilometer and a maximum of 47 for the selected window (100 by 100 km block). The deeper Curie points are located on the southern part of the Kenya Rift around Lake Magadi whereas the shallow Curie point depths occur in the central part around Lake Naivasha. These results did not correspond with the known tectonic provinces i.e. Tanzanian Craton, Kenya Rift valley and Mozambique Belt. For this reason, the study area was divided into 50 km contiguous blocks corresponding to the average width of the Kenya rift valley (Kabede, 1989; Baker et al., 1972). This allowed sampling 50km of the Tanzanian Craton, 50km of the rift valley and 50km of the Mozambique belt. For each new 50km by 50km blocks (grid) the radially averaged power spectrum was calculated and the bottom of the magnetic ensemble determined following the method described for 100km by 100km blocks. The centre of each block was contoured and spatial distribution of the Curie point depth presented (Fig.4b). The depth estimates to the bottom of the ensemble of magnetic sources indicate shallow sources on the western side compared to the eastern side of the rift. The shallower depth values were observed in Block 2 and 5 (at 15.3 and 14.8km respectively) near Lake Magadi area. Similarly, shallow depths were observed within central Kenya Rift (Block 11) at depths of about 14km close to Lake Naivasha, a location that marks the geothermal fields in Kenya, which are currently being exploited for geothermal energy. The Curie point within Mozambique Belt is 46.5km (Block 18) and the Curie point values were deepest (56.5km) on the western part of the rift corresponding to location of the Tanzanian Craton near Kenya dome.
Fig.3. Radially averaged power spectrum of aeromagnetic anomalies of some of the representative blocks. Kmax represents the maximum wavenumber from which depth to the bottom of the magnetic sources is estimated. The calculated depth to the top of the deepest layer, in kilometers, is noted along side the spectra. Examples are as follow: Fig 4(a). Rift Valley Block 3; Fig. 4(b). Mozambique Belt Block 18; and Fig. 4(c). Tanzanian Craton Block 10.
Fig. 4(a). Contour map shows fictitious bottom depth (in Km) of magnetically active layer determined from 100km x 100km blocks, distinctly masking the crustal provinces. Fig.4b. Contour map showing bottom depth (in km) of magnetically active layer of central to southern Kenya rift from 50km x 50km blocks. Map shows structure that clearly reflect tectonic provinces (Tanzania Craton, Rift valley, Mozambique) used to constrain the size of blocks. Shallow and isolated depths are interpreted as magmatic intrusions that drive tectonics within this area.
4. Earthquake data
The maximum focal depth of earthquakes has long been known to provide fundamental properties of the upper part of the lithosphere, in particular the transition from brittle faulting to plastic flow in the crust, or a change in the frictional sliding process (e.g. Sibson, 1982). In order to understand the properties of the crust and possibly infer the rifting tectonics, the earthquake focal depth distribution was evaluated and correlated with the Curie point depths considering that the thermal state (indicated by CPD) and mineral composition of the continental lithosphere are factors controlling the first order its strength (Chen and Molnar, 1983). For this purpose, we used earthquake data for period between October 1993 and August 1996 (Hollnack and Stangl, 1998) and between November 1999 and December 2001 (Ibs-von Seht et al., 2001; Kianji, 2003), which represents the most complete data set for the study area. For more than 2000 recorded events in the 1993-1996 database only 435 local earthquakes could be localized within the study area with local magnitudes of up to 5. For 1999-2001 databases a total of 603 earthquakes were considered. Similar to the Curie point depth, we gridded the focal depths of the earthquake data to determine the spatial variations of these depths over the study. The result is shown in Fig. 5. The results show three isolated areas characterized by shallow earthquake focal depths namely; Lakes Baringo, Naivasha and Magadi. The trend of the earthquake sources is similar to that of the Kenya rift in its central to southern segment: in the central area the rift of oriented NNE-SSW and then deflected to run NNW-SSE prior to becoming broad depression in the northern Tanzania and indistinctly defined. On the southern end of the Kenya rift, an arm trending generally W-E branches off and this marks the location of the Pangani Rift of Tanzania rift (Chorowicz and Sorlien, 1992). Deep earthquake sources characterize the western and eastern parts of the rift. From the foregoing, it is apparent that there is qualitative spatial relationship between shallow curie points and shallow focal depth which provides information about the crustal strength of the crust.
In order to quantitatively determine the correlation between the Curie point depth and earthquake focal depth. We established an area defined by coordinate latitudes 0.3° North and 2° and longitudes 35° and 36.6° corresponding to the area covered by the Curie point depth. We then gridded the Curie point depth data at a line spacing of 0.1 by 0.1 and obtained a total of 408 grid node points. We then gridded the earthquake data with the same grid spacing and obtained the same number of grid node points. On each grid node, we read the grid value of Curie point depth and earthquake focal depth. We made a scatter plot of the earthquake focal depth versus Curie point depth and obtained the results shown in Fig.6. The results show largely that there is a relationship between the shallow earthquakes at a depth of 5 - 10 km and shallow Curie point depths possibly corresponding to earthquakes located within the rift. This is followed by a rather non-linear relationship between the two parameters that generally characterize earthquakes at depth ranging from 10 to 30 km and are characteristic of deep curie points. The shallow earthquakes within the Tanzania Craton and Mozambique could possibly be related to shallow feeder dykes.
5. Discussion
An attempt has been made to reconstruct the thermal structure of the central and southern Kenya Rift by computing the Curie point depths and comparing the results with the focal depth of the earthquakes. In order to determine the thermal structure, the depth to bottom of the magnetic sources from the regional aeromagnetic data has been determined using spectral methods. Initially blocks 100km by 100km with a 50% overlap were selected. It was found that significant spectra maxima existed for all 10 blocks except for Block 3. Similar observations were made by Rajaram et al. (2009) who suggested that lack of a well defined spectral peak could be due to the large wavelength part of the spectrum extending beyond the dimensions of the selected block. The map showing the distribution of the Curie points determined from 100km by 100km reflected a structure of tectonic provinces different from that expected for the central to southern Kenya rift. We selected smaller blocks of 50 km x 50 km corresponding to the width of the three anomalies determined from the analytical signal of the aeromagnetic data for this study area. These magnetic anomalies marked on the ground by location of the Tanzanian Craton on the western side, rift in the central section and parts of the mobile belt on the eastern side. It is noted that other previous studies have used similarly small blocks i.e. 64km by 64km (Espinosa-Cardeña and Campos-Enriquez, 2008); and even smaller blocks varying from 9 to 20 km (e.g. Dolmaz1, et al., 2005) to carrying out spectral analysis.
Fig. 6. Scatter plot of the earthquake focal depth and Curie point depth obtained from grid notes covering same locations. The first section shows a linear relationship: increasing earthquake depth with increase in Curie point this possibly corresponds to seismically active zones.
Fig.5. Map of the focal depth distribution for central to southern Kenya rift. The shallow focal depth (up to 0-6km) around Lakes Magadi, Naivashs and Bogoria marks the clustering of earthquakes (swarm in Fig. 1) and correlates to the shallow Curie point depth (Fig.4.(b)) marking seismically active regions.
We found that significant spectral maxima exist for smaller blocks (50 km x 50 km). In addition, it was noted that when the blocks size was increased to 100km x 100 km the spectral maxima remained stable but for the Tanzanian Craton and Mozambique Belt but masked the narrow rift zone, 30-60km wide (Kabede, 1989; Baker et al., 1972). These small blocks used in this study (50km 50 km to 100km x 100 km) largely contrast those for other areas which have attained spectral maxima at 440km x 440km to 550km to 550km (Rajaram, et al., 2009). The calculated Curie point depth as expected are shallow within parts of the rift system (i.e. central and southern rift) separated by deep Curie points that is similar to those of the Mozambique Belt and Tanzanian Craton. Although the aeromagnetic datasets are utilized to calculate the Curie point depth is spectral estimates, the depths are comparable to the well resolved lithospheric structure from KRISP projects (Marita and Keller 2007; Simiyu and Keller, 2001).
A thick magnetic crust suggests stable continental regions while thin magnetic crust accounts for tectonically active regions also associated with higher heat flow (Rajaram, et al., 2009). From the curie depth estimated using aeromagnetic data, we find that the magnetic crust in the southern western part off Lake Magadi (Block1: 47km), immediately north of Lake Magadi (Block 10: 51.5km) and north eastern side of Lake Bogoria (Block 18: 47km) suggesting that these regions are quite tectonically stable. The axial part of the rift floor in the Central Kenya Rift (Lake Bogoria) and the Southern Kenya Rift (Lake Magadi) are associated with thin magnetic crust. The other areas adjacent to the rift and particularly the eastern part of rift correspond to location of the Mozambique belt have intermediate thicknesses characterized by deep earthquakes.
The shallow Curie point depths were determined in Block 2 and 5 (at 15.3 and 14.8km respectively) near Lake Magadi area, Block 11 (14km) around Lake Naivasha area and Block 17 (17.2km) for Baringo area. We have interpreted these shallow Curie point depth areas as magmatic intrusions namely Magadi, Olkaria and Eburru, and Bogoria within rift floor respectively. These areas are also characterized by shallow focal depth of swarm earthquake activities (Young et al. (1991); Tongue et al., 1992 Ibs-von Seht et al. (2001)) and such swarms in regions of active volcanism is commonly related to eruptions and magma movements (e.g. Karpin and Thurber, 1987). Therefore, considering that in volcanically active regions around the world, earthquake swarms associated with contemporaneous crustal deformation are often inferred to be the result of subsurface magma movements (e.g. Wright et al., 2006) we safely conclude that the shallow earthquakes are results of movement of or within magmatic intrusions and indicate rifting tectonics along, at least, along these zones. Indeed, Baer et al. (2008) and Calais, et al. (2008) observed such magma-driven earthquake swarms caused by dyking in southern Kenya rift in Tanzania near Lake Natron.
Previous works on heat flow measurement recorded a mean value of 103 + 52 (sd) mW m-2 in the Kenya rift comes from data collected by Ebinger et al., (1991). The heat flow in the Mozambique belt averages 64 mW m-2 (sd=19 mW m-2; se = 4 mW m-2) is poorly constrained in the rift valley (Nyblade, 1997; Nyblade et al., 1990). In Nyblade et al. ( 1990) and Nyblade (1997) slightly lower heat flow values averaging 33 mW m-2 (sd=10 mW m-2; se=3 mW m-2) were measured in the Tanzanian Craton. A comparison of the heat flow measurements and our estimated Curie point depth indicate that high heat flows characterize shallow curie points i.e 15.3 km (Lake Magadi), 14 (Lake Naivasha area) and 17.2 km (Lake Baringo). For Mozambique belt the Curie point depth was estimated at 46.5 km, similar to those estimated for Tanzanian Craton (47km). Deep Curie point depths (53.5 to 56.5 km) were estimated for structural feature separating Lakes Magai and Naivasha. This feature marks Aswa Shear Zone (Chorowicz, 2005) that indicates a thickened crust influencing the rift tectonic between the two lakes.
In the continental regions the Moho marks a magnetic boundary, as the mantle material is non magnetic (Wasilewski and Mayhew, 1992) and therefore the Curie isotherm should lie either within the crust (Shallower than Moho) or should coincide with the Moho. It follows therefore that if the Curie depth matches Moho depth then the Curie depth represents a compositional boundary rather than thermal boundary. It has been established that magnetite with Curie temperature of 580ËšC is the dominant magnetic mineral in the lower crust (e.g. Frost and Shive, 1986) that accounts for the long wavelength induced magnetization. Although hematite has much higher Curie temperature, its magnetization is weak and hence cannot give rise to long wavelengths at aeromagnetic altitudes (Rajaram, et al., 2009). A comparison of the Moho depths from the Central Kenya Rift and Southern Kenya Rift (Keller et al., 1994a,b; Prodehl, et al., 1997 with depth to the bottom of magnetic crust from the present study, it can be seen that the bottom of the magnetic crust lies above the Moho depth and could represent a thermal boundary rather than a petrological or compositional boundary. Thus, the depth to the bottom of the ensemble of magnetic sources derived in the present study may be called Curie isotherm depth. However, within the Tanzanian Craton and Mozambique Belt the depths to the bottom of the ensemble is greater than that of the Moho. It has however been noted that in some cases the uppermost part of the mantle may be magnetic (Toft and Arkani-Hamed, 1992) and this large depth to the bottom of the magnetic crust observed in these cases can possibly be due to the fact that these regions have low heat flow and low vertical geothermal gradient (Eppelbaum and Pilchin, 2006).
The Curie isotherm depth is related to the heat flow (e.g. Blakely, 1988) in a given region and higher heat flow results in lower Curie depth and vice versa. Considering this fact, it can be inferred that the deep Curie point depth area separating the Central and Southern Kenya Rift is a region with low heat flow suggestive of stable continental shield regions where rift is quite narrow. This area marks the location of the Aswa shear zone. While the Lakes Bogoria, Naivasha (e.g. Hochstein and Kagiri, 1997) and Magadi are associated with relatively high heat flow suggesting that these regions are tectonically active (e.g. Kuria et al., 2010) and mark areas of high geothermal energy potential. It worthwhile to note that Lake Bogoria and Magadi are characterized by geysers and hot springs together with swarm earthquakes (e.g. Young et al., 1991; Ibs-von Seht., 2001; 2008) that take place along the geothermal gradient zone in which the thermal stress is concentrated.
6. Conclusion
The following conclusions can be drawn from this study:
The regions of Central and southern Kenya Rift are separated by Aswa Shear zone (of thickened crust and possibly very narrow rift valley) that influences tectonic evolution of the Rift Valley.
Curie point depths follow structural trends of major tectonic provinces (Craton, Mozambique belt and Rift Valley) and reflect regional thermal structures.
For regions with low heat flow (Mozambique belt and Tanzanian Craton), the depth of the Curie point is greater than that of Moho discontinuity i.e. averaged 45km and 52km respectively.
The Curie point depth is related to magmatic intrusion with the largest intrusion at Olkaria, followed by Lake Magadi that broadens into northern Tanzania and narrowed small intrusions at Lake Bogoria area. These areas are characterized by shallow focal depth and therefore these earthquakes are results of movement of or within magmatic intrusions and indicate rifting tectonics, at least, along these zones.
Although there is consensus on choice of window size for average radial spectral analysis, our results strongly suggest that size of the tectonic provinces forms a good basis.
