The formation of the Namib Sand Sea inferred from the spatial pattern of magnetic rock fragments

The formation of the Namib Sand Sea inferred from the spatial pattern of magnetic rock fragments

Earth and Planetary Science Letters 395 (2014) 168–172 Contents lists available at ScienceDirect Earth and Planetary Science Letters www.elsevier.co...

962KB Sizes 2 Downloads 56 Views

Earth and Planetary Science Letters 395 (2014) 168–172

Contents lists available at ScienceDirect

Earth and Planetary Science Letters www.elsevier.com/locate/epsl

The formation of the Namib Sand Sea inferred from the spatial pattern of magnetic rock fragments Andreas U. Gehring a,∗ , Nima Riahi b , Jessica Kind a , Bjarne S.G. Almqvist c , Peter G. Weidler d a

Institute of Geophysics, ETH Zurich, 8092 Zurich, Switzerland Geological Institute, ETH Zurich, 8092 Zurich, Switzerland c Department of Earth Sciences, Uppsala University, 75236 Uppsala, Sweden d Institute of Functional Interfaces, Karlsruhe Institute of Technology, 76131 Karlsruhe, Germany b

a r t i c l e

i n f o

Article history: Received 16 January 2014 Received in revised form 12 March 2014 Accepted 17 March 2014 Available online 12 April 2014 Editor: J. Lynch-Stieglitz Keywords: hyperarid climate environmental magnetism erosion regime sand mixing statistical test

a b s t r a c t The Namib Sand Sea on the west coast of Namibia is one of the world’s oldest desert region and based on cosmogenic dating it has likely existed since the earlier Pleistocene. Among the possible sand sources, geomorphological and petrographic evidence points towards the Orange River catchment as the most prominent one. Little is known about the dynamics of transport and mixing of the sand during the desert formation and this is because the information about the Namib Sand Sea generally rests upon study sites at its edges. Here, we present a statistical analysis of magnetic components in sand samples collected along a south to north transect through the desert and at two inland sites. The magnetic components are rock fragments mainly of basaltic origin. Their statistically uniform distribution in the Namib Sand Sea indicates no significant sand source other than the Orange River and thus a predominant northward direction of the sand transport. A northward transport and the absence of a magnetic trend along the transect suggests mixing of the sand prior to its deposition in the Namib Sand Sea, most likely during river transport and under high current conditions along the shoreline. Finally, the uniform magnetic pattern provides compelling evidence for a stable erosion regime in the Orange River catchment with a steady release of magnetic components at least since the Pleistocene. © 2014 Elsevier B.V. All rights reserved.

1. Introduction The Namib desert, which stretches about 1000 km along the coast of south western Africa, is a classical example of a costal desert (Laity, 2009). The Orange River is the largest river in southern Africa. It arises in the Drakensberg mountains in the Lesotho highlands and transports sediments over a distance of 2300 km to the South Atlantic Ocean. Since the Cenozoic era the Orange River and its tributary, the Vaal River, have formed the drainage system with the largest sediment discharge in southern Africa (e.g., Bluck et al., 2007). The Orange River has been considered to be the main sand source of the Namib Sand Sea, which is the central dune region of the Namib desert near the Tropics of Capicorn (e.g., Rogers, 1977; Lancaster, 1989). The Namib Sand Sea, also termed the Namib erg, covers an area of about 34 000 km2 between Lüderitz in the south and the Kuiseb River in the north, having the Great Escarpment as its eastern margin (Fig. 1). It contains a variety of dune morphologies; these are mainly linear dunes, with star dunes on the eastern margin and crescentic dunes along

*

Corresponding author.

http://dx.doi.org/10.1016/j.epsl.2014.03.041 0012-821X/© 2014 Elsevier B.V. All rights reserved.

the coast (Livingstone et al., 2010). The unconsolidated eolian sand of the Namib erg is termed the Sossus Sand Formation and lies separated by an unconformity on the Tsondab sandstone (Miller, 2008). This terrestrial sandstone formation is of Miocene age and consists of eolian dunes associated with fluvial and playa sediments (Ward, 1988). The deposition of the Sossus Sand Formation may have begun during the latest Pliocene about 2 million years ago (Goudie and Eckhardt, 1999). Fossil remains in the Tsondab sandstones, however, indicate that the desertification started before ≈16 million years ago (Senut et al., 2009). Chronological control of the Sossus Sand Formation is challenging, because the dynamics of the sand transport and mixing can critically affect the results of the different dating methods. Optical stimulated luminescence data yielded variable dune deposition ages of about 104 yr and several 104 of years, suggesting that dunes in the Namib Sand Sea could be formed after the last global climate change in the latest Pleistocene (Bristow et al., 2007) or the Last Glacial Maximum (Stone and Thomas, 2013). In a recent paper, Vermeesch et al. (2010) have reported cosmogenic nuclei data and zircon U–Pb age spectra that confirm the Orange River catchment as the exclusive source for the near-coastal Namib erg and indicate a residence time of the sand of more than a million years.

A.U. Gehring et al. / Earth and Planetary Science Letters 395 (2014) 168–172

169

delta (e.g., Stone, 2013, and references therein). It has been argued that the redness of inland dunes is caused by reworked material from the Tsondab sandstone (e.g., Besler, 1996). A magnetomineralogical study on sand samples from a reddish inland dune at the Sossusvlei site showed that the magnetization of the sand material is carried by the high susceptibility minerals magnetite and maghemite in basaltic fragments (Gehring et al., 2009). Rock fragments are generally valuable indicators for the provenance of the sand, and they are independent of the dune colors (e.g., Garzanti et al., 2012). Consequently, the spatial distribution of magnetic rock fragments could provide constraints for the interplay between source, transport and mixing of eolian sands during the formation of the Namib erg. Here we report the statistics of the magnetic pattern of sand samples from a south-to-north transect and from two inland sites of the Namib erg in order to test this assumption. 2. Methodology

Fig. 1. Moderate Resolution Imaging Spectroradiometer (MODIS) image of the Namib erg. Red dots represent sampling sites along the transect between Lüderitz and Walvis Bay, subdivided into the three subgroups T1, T2, and T3. Green squares and triangles mark the inland sites near the Gababeb Research Center and at Sossusvlei.

These authors have suggested two possibilities to explain the discrepancy of the time constraint of the Namib erg. First, the dunes were active throughout this time and the sand grains have been recycled many times on their journey to the north. Alternatively, the sand was buried for an extensive period of time during which relatively humid conditions alternated with dry conditions that fostered increased dune mobility and sand mixing. A striking characteristic of the Namib erg is its color gradient from whitish and yellow-brownish near the coast to much redder sand on the eastern edge (Fig. 1). A magnetic study showed that this variation originates from fine grained secondary iron oxides, such as goethite and/or hematite, associated with clay coatings of the sand grains (Walden et al., 2000). Detailed sand color analysis along a west-east profile in the northern part of the Namib erg, using a combination of remote sensing and laboratory measurements on sand samples, has provided evidence that the sand arises from the mixing of at least two distinct sources (White et al., 2007). Generally, three sand sources have been considered: (i) reworked Tsondab sandstone, (ii) river-transported material from the Great Escarpment, and (iii) wind-derived material from the Orange River

The south-to-north transect through the Namib Sand Sea in this study follows the predominant wind trajectories, and, therefore, the major directions of sand transport (Lancaster, 1989). This sampling disposition provides a continuous spatial record that complements magnetic, petrographic, and geochemical data collected from the edges of the Namib Sand Sea (Walden et al., 2000; Vermeesch et al., 2010; Garzanti et al., 2012). A set of 113 sand samples were collected in 250 ml plastic flasks in roughly equal intervals along a transect between Lüderitz and Walvis Bay (Fig. 1). The sampling sites were determined by passages through the desert and are independent of the type, height, size and shape of the dunes. Sampling position within the dune was not specially considered. Thus, the samples represent a random collection in a south-to-north spatial frame. In addition, sand samples were taken at two inland sites near the Gobabeb Research Station and at the Sossusvlei including dune 45, respectively (Fig. 1). It is worth noting that the accessibility of the Namib desert permits no E to W sampling, i.e., from the coast to the inland, in a similar spatial resolution as for the S to N transect. The magnetic content of the samples was determined by susceptibility measurements using an AGICO KLY-2 Kappabridge on 3.2 cm3 material taken from each flask. The ordering temperatures of the magnetic carriers of a total of 12 samples throughout the transect and 6 samples from the inland sites were determined on magnetic separates with the Kappabridge coupled to a thermoelement. The average heating/cooling rate was 12 ◦ C min−1 and the sample was kept in a steady argon gas flow to prevent oxidation during the measurements. In addition mineralogical and morphological properties of magnetic separates were investigated by powder X-ray diffractometry (XRD: Bruker D8, θ –θ geometry, Lynxeye stripe detector (PSD), CuKα radiation) and scanning electron microscopy (SEM: FEI Quanta200FEG instrument). Backscatter electron (BSE) images were obtained at analytical conditions of the SEM with a working distance of 10 mm (Gehring et al., 2009). The collected sand samples were divided into five groups, according to their geographic location: the first three groups were the southern, center, and northern thirds of the transect (termed T1, T2, and T3) and were used to test for north-south trends in the magnetic susceptibility. The fourth and fifth group, respectively, consisted of samples from the Gobabeb and the Sossusvlei sites and were employed to compare inland locations with the coastal transect. The grouping is shown in Fig. 1. To assess the degree of mixing in the Namib erg, two null-hypotheses pertaining to the geographically defined groups were statistically tested: (1) the magnetic susceptibility variation is the same for all groups, and (2) the average magnetic susceptibility is the same for all groups. For the statistical evaluation the test by Brown and Forsythe (1974)

170

A.U. Gehring et al. / Earth and Planetary Science Letters 395 (2014) 168–172

Fig. 2. (a) SEM micrograph of magnetic components separated from the Namib sand, (b) backscattered electron image of magnetic rock fragment from a sand sample of T3 with sprinkled, tiny salt particles at the surface arrowed, and larger whitish components enriched in iron with variable shapes in the grayish matrix, (c) magnetic rock fragment of the Sossusvlei, and (d) enlarged surface area with clay coatings covering an iron-enriched region with skeletal shape.

and the unbalanced one-way analysis of variance (ANOVA) were applied (Montgomery, 2001). 3. Results The magnetic separates consist of dark particles and they make up less than 10% of the dune material. These particles have no appreciable chromophore effect for the sand. In the field, however, these dark particles can be visible as flow-mark-like shades at the brink or the slip face of the dunes. Plagioclase, clinopyroxene, and quartz are the main mineral components of the separates as detected by XRD. This indicates that the magnetic fraction contains aphanitic, volcanic and metamorphic rock fragments. Under the SEM the dark fraction consists of sub-rounded to rounded rock fragments with grain sizes generally in the fine to medium sand range (Fig. 2a). Surfaces of rock fragments from sites near the coast are sprinkled with tiny sodium chloride salt crystals (Fig. 2b). Moreover, BSE images show whitish areas at the surface of the grains, clearly contrasting with the grayish grain matrix (Figs. 2b, c). The brightness of the spots is indicative of a relative iron-rich phase and their skeletal shape is characteristic for magnetite (Fe3 O4 ) in basalts (e.g., Haggerty, 1976). These basaltic rock fragments are major magnetic components in the samples from the transect and the inland sites (Figs. 2b, c). At higher magnification the grains reveal coatings of clay minerals (Fig. 2d). It has been shown that these coatings can also contain ochreous goethite (α -FeOOH) and/or reddish hematite (α -Fe2 O3 ) and that these iron oxides are responsible for the coloration of the dunes (e.g., Walden et al., 2000). These iron phases at the surface of the sand grains are most likely weathering products. The plots of magnetic susceptibility X versus temperature exhibit a steep slope at about 580 ◦ C, indicative of the Curie temperature (T C ) of magnetite, and, therefore, this high-susceptibility mineral can be considered as the major magnetic carrier in the

Fig. 3. Changes of magnetic susceptibility during heating/cooling cycle measured in an argon atmosphere for magnetic separates that (a) are magnetite-rich and (b) contain a mixture of magnetite and maghemite. The dashed line indicates the Curie temperature of magnetite.

rock fragments (Fig. 3a, b). The magnetic susceptibility curves provide no evidence for significant amounts of goethite or hematite, which have Curie temperatures at about 120 ◦ C and 680 ◦ C, respectively (e.g., Gehring and Heller, 1989; Morrish, 1994). Therefore, the contribution of these low-susceptibility magnetic phases is negligible for the magnetization of the sand. The behavior of heating and cooling curves in the X–T measurements, however, can be schematically subdivided into two groups, which show no systematic spatial distribution. In the first group the contribution of magnetite to the decay in the heating curve is dominant and the heating and cooling curves are similar (Fig. 3a). In the second group the heating curve has a higher magnetic susceptibility compared to the cooling curve (Fig. 3b). Furthermore, there is a

A.U. Gehring et al. / Earth and Planetary Science Letters 395 (2014) 168–172

171

that the five groups have the same mean magnetic susceptibility cannot be rejected (p-value 0.17). 4. Discussion

Fig. 4. Measured magnetic susceptibility for the five groups of samples shown in Fig. 1. The data points were dithered horizontally for better visualization.

considerable decay above 580 ◦ C and the bifurcation of the heating and the cooling curves indicates a magnetic phase with a slightly higher T C than magnetite (Fig. 3b). Moreover, this behavior suggests that the magnetic phase with T C above 580 ◦ C is metastable and converts into a phase with lower magnetic susceptibility than that of magnetite. Considering this behavior, the second magnetic phase can be assigned to maghemite (γ -Fe2 O3 ), which converts to hematite at higher temperature (e.g., Gehring and Hofmeister, 1994). A maghemite with stability above 600 ◦ C has been reported from sand samples at the Sossusvlei site and was interpreted as an alteration product during slow oxidation of magnetite in basaltic rock fragments under arid conditions (Gehring et al., 2009). Taking the above, the magnetic response of the Namib sand is mainly due to basaltic rock fragments and this agrees well with petrographical results reported by Garzanti et al. (2012). The 113 measured magnetic mass susceptibility values from the southern, center, and northern transect groups (T1, T2, T3) are in the range (0.4–11.9) × 10−3 . Two samples with magnetic susceptibilities above 8 × 10−3 in group T2 were considered as extreme outliers and excluded from further analysis. Fig. 4 gives a graphic overview of the magnetic susceptibilities of the remaining 111 samples from the transect groups. We statistically test two null-hypotheses, regarding the three transect groups: (1) the variability in magnetic susceptibility is the same for all groups and (2) the average magnetic susceptibility is the same for all groups. Depending on the data, different instead of same may be suited to address the first null-hypothesis. The distributions in Fig. 4 are not strictly Gaussian and there are a few distinctly high values. An appropriate test in this case is the Brown–Forsythe test (Brown and Forsythe, 1974), which defines variation as absolute deviations within a group from the group median and tests for equality of that measure using a variant of the F-statistic. By this measure of variability we find that the null-hypothesis cannot be rejected at 95% confidence (p-value 0.70), i.e., the group variabilities are not significantly different from each other. The second hypothesis is tested using an unbalanced one-way ANOVA (Montgomery, 2001), which tests for equality of the mean values of the groups assuming the variability is the same. This hypothesis also fails to be rejected at 95% significance (p-value 0.27), i.e., the group averages are not significantly different. To assess the mixing between coastal and inland areas we investigated the inland samples at two sites (groups Gobabeb and Sossusvlei) for comparison with the three transect groups and repeated the above statistical hypothesis tests (Fig. 4). The first test shows that the five groups, three from the transect and two from inland, have the same variability and the null-hypothesis cannot be rejected at 95% confidence (p-value 0.84). Also the null-hypothesis

The lack of significant changes in the magnetic susceptibility between the various sample groups suggests that the magnetic minerals are equally admixed in the sand dunes. The magnetic minerals are constituents of rock fragments, and, therefore, they can be taken as proxy for the magnetic properties of the source rocks. Given this, the relative amount of the magnetic components reflects the erosion rate of the rock formation in the area from which the magnetic fragments originate. Geomorphological arguments, U/Pb dating and petrographic results provide clear evidence that the Orange River is the primary sand source for the dunes along the transect (Lancaster, 1989; Vermeesch et al., 2010; Garzanti et al., 2012). Therefore, it can be supposed that changes in the magnetic content of the sand correspond to changes of the erosion rate in the catchment of the Orange River, assuming constant dwell time of the sand fraction in the estuary. Given this the homogeneous magnetic pattern along the transect indicates a stable erosion regime in the source area because variations would be preserved in the magnetic pattern from S to N due to the predominant northwards wind trajectories in the Namib erg. This suggests that the transport through the Namib erg has no significant effect on the degree of mixing of the magnetic components. Hence, it is most likely that the magnetic components were already mixed prior to accumulation in the Namib erg. Taking the Orange River as source, the path of the sand mixing can be considered a multi-stage process. After long fluvial transport detrital material reached the estuary of the Orange River, where high-energy currents in the delta and further offshore separated the detrital material into gravel-to-clay fractions (Bluck et al., 2007). The magnetic components were transported together with the sand fraction by currents along the shore, resulting in sand accumulations in bays north of the Orange River estuary (Corbett, 1993). The sand fraction exposed by sea-level fluctuation was captured by southeast trade winds and carried on-land (Lancaster, 1989; Vermeesch et al., 2010). As mentioned above, the eolian transport through the Namib erg has no significant effect on the degree of mixing; therefore, the sand was mainly mixed in aquatic systems, most likely by river and offshore currents. With this in mind the magnetic data provide no evidence for intense reworking cycles of the sand within the dunes on their northward journey, which has been suggested as a possible explanation of the cosmogenic nuclides pattern of the Namib erg (Vermeesch et al., 2010). It has been postulated that the additional sand sources for the inland dunes could be the Tsondab sandstone and/or the ephemeral rivers with catchment in the Great Escarpment, e.g., Tsauchab River (e.g., Besler, 1996; White et al., 2007). The uniform statistical distribution of magnetic susceptibility along the transect and at the two inland sites argues against a significant input from an additional source, because the magnetic properties of the rock formations in the drainage area of these ephemeral rivers and those of the Orange River are different, mainly with respect to the supply of basaltic detritus (e.g., Schlüter, 2006). The Tsondab sandstone, however, may be an additional source, if it can be assumed (1) that the Orange River was also the major sand source and (2) the erosion regime in the drainage area, which fed the Tsondab sandstone during middle to late Miocene, was similar to that during the formation of the Namib erg (see Koncurek et al., 1999). Considering the Orange River catchment as the exclusive source, the magnetic pattern of the Namib sand can also be constrained in the context of Pleistocene climate. The onset of an arid environment, as a requirement for the development of the Namib erg

172

A.U. Gehring et al. / Earth and Planetary Science Letters 395 (2014) 168–172

in the early Pleistocene, is well documented by pollen records, which show an abrupt change from grass to semi-desert pollen about 2.2 my ago (Dupont, 2006). The change to arid conditions along the coast has been explained by cooler sea surface water as a result of the intensification of the Benguela current, which impeded the onshore precipitation and strengthened the trade winds (e.g., Etourneau et al., 2009). Moreover, the global shift to cooler temperatures from the Pliocene to the Pleistocene epoch caused periglacial conditions in the high Drakensberg mountains at altitudes over 3000 m (Mills et al., 2009). Such conditions led to pronounced physical weathering and the release of debris from the Karoo Supergroup, which mainly consists of clastic sedimentary rocks and of flood basalts with over 1000 meters in thickness (Schlüter, 2006; Miller, 2008). Consequently, an erosion regime in the Drakensberg mountains that steadily supplies volcanic material to the Orange River, accompanied by stable conditions in the sand accumulation area during the Pleistocene, can explain the relatively high content of basaltic fragments and their distribution throughout the Namib erg. Moreover, such a balance between the source and the sink of the sand also agrees well with the long residence time of the sand in the Namib erg that has been inferred from cosmogenic nuclei data. 5. Conclusion In a novel approach we analyzed the statistics of the magnetic pattern in the Namib Sand Sea using samples from a south-tonorth transect and two inland sites. The statistical uniformity of the magnetic pattern gives an insight into the source, transport and mixing of the sand. There is no significant sand source other than the Orange River and that in turn indicates a northwards direction of the sand transport. The absence of a magnetic pattern suggests that the sand was mixed prior to its accumulation in the erg, most likely during fluvial transport and under high current conditions along the shoreline. In addition, the statistically uniform distribution of the magnetic components indicates a stable erosion regime in the catchment of the Orange River at least over two million years, if an admixture of material of the Tsondab sandstone is excluded. If, however, there is a contribution of the Tsondab sandstone to the Sossus Sand Formation, a much longer period of stability of the erosion regime in the Orange River catchment – probably since the Miocene – could be postulated. Acknowledgements The authors would like to thank Stefan Schultheiss, TRICON Ltd., Windhoek, and Jo Henschel, Gobabeb Research Center, for logistic support during the field campaign and Bill Lowrie for critically reading the manuscript. Appendix A. Supplementary material Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/j.epsl.2014.03.041. References Besler, H., 1996. The Tsondab sandstone in Namibia and its significance for the Namib erg. S. Afr. J. Geol. 99, 77–87.

Bluck, B.J., Ward, J.D., Cartwright, J., Swart, R., 2007. The Orange River, southern Africa: an extreme example of a wave-dominated sediment dispersal system in the Southern Atlantic Ocean. J. Geol. Soc. 164, 341–351. Bristow, C.S., Duller, G.A., Lancaster, N., 2007. Age and dynamics of linear dunes in the Namib Desert. Geology 35, 555–558. http://dx.doi.org/10.1130/G23369A. Brown, M.B., Forsythe, A.B., 1974. Robust tests for equality of variances. J. Am. Stat. Assoc. 69, 364–367. Corbett, I., 1993. The modern and ancient pattern of sandflow through the southern Namib deflation basin. Spec. Pub. Int. Ass. Sediment. 16, 45–60. Dupont, L.M., 2006. Late Pliocene vegetation and climate in Namibia (southwestern Africa) derived from palynology of ODP Site 1082. Geochem. Geophys. Geosyst. 7, Q05007. http://dx.doi.org/10.1029/2005GC001208. Etourneau, J., Martinez, P., Blanz, T., Schneider, R., 2009. Pliocene–Pleistocene variability of upwelling activity, productivity, and nutrient cycling in the Benguela region. Geology 37, 871–874. http://dx.doi.org/10.1130/G25733A.1. Garzanti, E., Andò, S., Vezzoli, G., Lustrino, M., Boni, M., Vermeesch, P., 2012. Petrology of the Namib Sand Sea: long-distance transport and compositional variability in the wind-displaced Orange delta. Earth-Sci. Rev. 112, 173–189. Gehring, A.U., Heller, F., 1989. Timing of natural remanent magnetization in ferriferous limestones from the Swiss Jura mountains. Earth Planet. Sci. Lett. 93, 261–272. Gehring, A.U., Hofmeister, A.M., 1994. The transformation of lepidocrocite during heating: a magnetic and spectroscopic study. Clays Clay Miner. 42, 409–415. Gehring, A.U., Fischer, H., Louvel, M., Kunze, K., Weidler, P.G., 2009. High temperature stability of natural maghemite: a magnetic and spectroscopic study. Geophys. J. Int. 179, 1361–1371. http://dx.doi.org/10.1111/j.1365-246X.2009.04348.x. Goudie, A.S., Eckhardt, F., 1999. The evolution of the morphological framework of the central Namib desert, Namibia since the Early Cretaceous. Geogr. Ann. 81A, 443–458. Haggerty, S.E., 1976. Oxidation of opaque oxides in basalts. In: Rumble, D. (Ed.), Oxide minerals. In: Short Course Notes, vol. 3. Min. Soc. Am, pp. 1–20. Koncurek, G., Lancaster, N., Carr, M., Frank, A., 1999. Tsondab sandstone formation: preliminary bedform reconstruction and comparison to modern Namib Sand Sea dunes. J. Afr. Earth Sci. 29, 629–642. Laity, J., 2009. Deserts and Desert Environments. Wiley–Blackwell, Oxford. Lancaster, N., 1989. The Namib Sand Sea: Dunes, Forms, Processes, and Sediments. Balkema, Rotterdam. Livingstone, I., Bristow, C., Bryant, R.G., Bullard, J., White, K., Wiggs, G.F.S., Baas, A.C.W., Bateman, M.D., Thomas, D.S.G., 2010. The Namib Sand Sea digital database of Aeolian dunes and key forcing variables. Aeolian Res. 2, 93–104. Miller, R.M., 2008. The Geology of Namibia, 3 volumes. Ministry of Mines and Energy, Windhoek. Mills, S., Grab, S.W., Carr, S.J., 2009. Recognition and paleoclimatic implications of late Quaternary niche glaciation in eastern Lesotho. J. Quart. Sci. 24, 647–663. Montgomery, D.C., 2001. Design and Analysis of Experiments. Wiley, New York. Morrish, A.H., 1994. Canted Antiferromagnetism: Hematite. World Scientific Publishing, Singapore. Rogers, J., 1977. Sedimentation on the continental margin off the Orange River and the Namib Desert. Geol. Survey/Univ. Cape Town Marine Geosci. Group Bull. 7, 162. Schlüter, T., 2006. Geological Atlas of Africa. Springer, Berlin. Senut, B., Pickford, M., Ségalen, L., 2009. Neogene desertification of Africa. C. R. Geosci. 341, 591–602. Stone, A.E.C., 2013. Age and dynamics of the Namib Sand Sea: a review of chronological evidence and possible landscape development models. J. Afr. Earth Sci. 82, 70–87. Stone, A.E.C., Thomas, D.S.G., 2013. Casting new light on the late Quaternary environmental and palaeohydrological change in the Namib desert: a review of the application of optically stimulated luminescence in the region. J. Arid Environ. 93, 40–58. Vermeesch, P., Fenton, C.R., Kober, F., Wiggs, G.F.S., Bristow, C.S., Xu, S., 2010. Sand residence times of one million years in the Namib Sand Sea from cosmogenic nuclides. Nat. Geosci. 3, 862–865. http://dx.doi.org/10.1038/NGEO985. Walden, J., White, K.H., Kilcoyne, S.H., Bentley, P.M., 2000. Analysis of iron oxide assemblage within Namib dune sediments using high field remanence measurements (9T) and Mössbauer analysis. J. Quart. Sci. 15, 185–195. Ward, J.D., 1988. Eolian, fluvial and pan (playa) facies of the Tertiary Tsondab sandstone formation in the central Namib desert, Namibia. Sediment. Geol. 55, 143–162. White, K., Walden, J., Gurney, S.D., 2007. Spectral properties, iron oxide content and provenance of Namib dune sands. Geomorphology 86, 219–229. http://dx.doi.org/10.1016/j.geomorph.2006.08.014.