Controls on sand ramp formation in southern Namibia

Sand ramps have the potential to provide rich palaeoenvironmental information in dryland regions where proxy records are typically scarce. However, current knowledge of the geomorphic controls and processes of sand ramp formation is limited. This study provides a data‐rich examination of the key factors controlling sand ramp formation. The location and morphology of 75 sand ramps in southern Namibia are examined. The sediments and chronologies of 10 sand ramps are studied in detail using 51 OSL dates and 83 grain‐size and LOI samples. Heavy mineral assemblages are used to determine the provenance of 10 samples and OSL sensitivity is used to explore geomorphic processes of eight samples.


Introduction
Studies of Quaternary environmental change in dryland regions are often hampered by a paucity of preserved proxy records (Thomas and Burrough, 2012). The arid climate inhibits the preservation of organic proxies typically used for palaeoenvironmental reconstruction, whilst ubiquitous 'geoproxy' records (e.g. dunes) can be difficult to interpret in terms of past climatic controls (Chase and Meadows, 2007, Chase, 2009, Stone and Thomas, 2008, Thomas and Burrough, 2012. This is particularly the case in the most arid parts of Namibia, including the Namib Sand Sea, where the analysis of geoproxy archives from within the sand sea has been relatively restricted and sometimes difficult to interpret in palaeoclimatic terms (Lancaster, 2002;Livingstone et al., 2010, Stone, 2013. Geochemical records, from speleothems (Geyh and Heine, 2014), tufa (Stone et al., 2010a), and especially hyrax middens (Chase et al, , 2010Lim et al. 2016), found in neighbouring Great Escarpment locations have provided alternative data sources that offer more high-resolution data sets.
There are however other potential untapped sources of data pertaining to changing environmental conditions in the region. Sand ramps are relatively widespread landforms in drylands which have the potential to provide valuable information on past sediment dynamics (Lancaster and Tchakerian, 1996) but, at present, they are globally under-investigated. Formed when aeolian sediment is trapped against a topographic barrier that also actively contributes to sediment accumulation, sand ramps contain varying proportions of aeolian, hillslope and fluvial sediments (Bateman et al., 2012;Lancaster and Tchakerian, 1996;Thomas et al., 1997). As such, they occupy the continuum between purely aeolian topographic features (e.g. climbing/falling dunes) and purely hillslope deposits (e.g. alluvial fans, talus cones) ( Figure 1). As the sediments are derived from different geomorphic processes (that may result from different environmental regimes) and the topographic barrier protects these sediments from remobilisation by subsequent aeolian transport, sand ramps have the potential to provide long and detailed palaeoenvironmental records which can be independently dated using OSL. However, relatively little is known in detail about the geomorphic controls and processes that govern sand ramp formation and their interdigitated sedimentary units, and whether variations in climatic forcing are clearly recorded. As such their current utility as palaeoenvironmental archives is limited.
The sand ramps in these investigations embrace a range of geomorphological contexts and sedimentological differences (Table 1). Sand ramps require a suitable accommodation space (i.e. a hillslope against which aeolian and colluvial sediments can accumulate) and contain a combination of aeolian and colluvial deposits, but in varying proportions. Both climbing (up wind of a topographic barrier) and falling (downwind) sand ramps have been described. Falling ramps are typically smaller, positioned higher on mountain slopes with steeper slope angles, and with better sorted, finer sediments (Lancaster and Tchakerian, 1996). Gullies have been observed in many sand ramp contexts. These run parallel to the axis of the sand ramp body or are found separating the ramp from the hillslope against which it has accumulated (Lancaster and Tchakerian 1996;Thomas et al., 1997;Bertram, 2003;Bateman et al., 2012). Both gully types are assumed to post-date accumulation of the main sand ramp body and are generally interpreted to be formed by fluvial activity (Bertram, 2003;Lancaster and Tchakerian 1996), although aeolian scour has also been suggested as the origin of the latter type (Lancaster and Tchakerian, 1996). The origin of sediments within a sand ramp has important implications for the antecedent conditions needed for formation. Colluvial, talus and fluvial sediments must be derived locally from the adjacent topographic obstacle (Lancaster and Tchakerian, 1996;Pease and Tchakerian, 2003;Turner and Makhlouf, 2002). Aeolian units are also suggested to be locally-sourced from neighbouring ephemeral channels or proximal upwind playas or dunefields (Bertram, 2003;Bateman et al. 2012, Pease andTchakerian, 2003), but may also incorporate further-travelled sediment (e.g. Thomas et al., 1997).
Chronometric data are available for some sand ramps, giving indications of the timing and age of accumulation (Table 1). However, some studies provide complex or conflicting data from the same location. TL and IRSL ages from sand units bracketing stone layers in Mojave ramps showed some to represent up to 10 ka breaks in aeolian unit accumulation, while others indicated no significant temporal breaks in accumulation (Rendell and Sheffer, 1996). The Soldier Mountain sand ramp (Mojave) was originally dated to c.10-22 ka using quartz and feldspar TL and feldspar IRSL dating (Rendell and Sheffer, 1996) and the sediments were interpreted as containing six palaeosols (Lancaster and Tchakerian, 1996). A more recent investigation of this feature, utilising SAR OSL dating, suggested rapid accumulation between c.11.6-10.3 ka. No evidence of palaeosol development was found and stone layers were not associated with temporal hiatuses. Rates of modern stone movement on the surface could not be reconciled with the rapid emplacement of stone layers indicated from OSL dating. Deposition of this sand ramp was therefore reinterpreted as opportunistic and event-driven with stone layers suggested to be deposited by periodic small streams distributing the stones across the sand ramp (Bateman et al., 2012).
This review suggests that sand ramps are complex and varied features, differing in formation history, and in the relative contributions from different sedimentary processes to the final landform. However, most investigations to date have been conducted on single features (e.g. Thomas et al. 1997;Bateman et al. 2012;Telfer et al. 2012), such that it is unclear how representative of wider conditions their conclusions are. While the overall processes of ramp accumulation (aeolian, fluvial and slope processes) are largely agreed upon, there are components of ramp systems that are currently poorly investigated or explained. Understanding the timescale and mechanism of sand ramp formation has important implications for palaeoenvironmental reconstructions with opportunistic, single-event driven accumulation providing limited regional information. Meanwhile, the provenance of the aeolian sediment is important for interpreting transport histories (Pease and Tchkarian, 2003) and understanding the environmental significance of the deposit.
This study aims to better understand the controls and mechanisms of sand ramp formation and to evaluate their potential as a palaeoenvironmental archive. This is achieved by conducting: (i) a systematic regional survey of the locations and broad morphologies of sand ramps in south-west Namibia and (ii) a detailed examination of the morphology, sedimentology, sediment provenance and chronology of a morphologically varied subset of these sand ramps. The combination of heavy mineral analysis and luminescence characteristics (sensitivity) offers an important novel insight into provenance of the aeolian component of sand ramp sediments.

Study site
The study was conducted in south-west Namibia between 23°-28°S and 15°-18.5°E (Figure 2). This area encompasses the transitional zone between the Great Escarpment, where sand supply is limited, and the hyper-arid sandy coastal plain where sediment supply is high but topographic obstacles are rare. In the intermediate zone of the western foreland of the Great Escarpment both sand supply and topographic obstacles, especially inselbergs, are sufficient to lead to the presence of numerous sand ramps in a variety of accommodation spaces (Bertram, 2003). Exposed lithologies are of mixed erodibility with sandstones, limestones and shales prevalent in the region (Atlas of Namibia Project, 2002; Figure 2 inset).
Western Namibia has an arid to hyperarid climate, primarily due to a dominant subtropical atmospheric high pressure regime and the limited incursion of moisture from easterly sources. This is enhanced in the coastal zone by upwelling of the cold Benguela current, which further cools the atmosphere inhibiting precipitation other than fog development. Consequently a steep W-E precipitation gradient (barely above 0 to 200mm), high inter-and intra-annual variability are key features of the regional climate (Lancaster et al., 1984;Nicholson, 2000). Wind regimes are bimodal. Prevailing winds from the S-SW dominate for most of the year reaching maximum strength (up to 14 ms -1 at the coast, reducing inland) in early Austral summer (September-November) (Lancaster, 1985). E to NE katabatic, mountain-to-plain winds are frequent in Austral winter (June-August) (Tyson and Seely, 1980). When regional pressure gradients are normal to the coast these can develop into strong easterly 'berg' winds. These frequently reach velocities of 14-17 ms -1 and can account for up to 65% of total sand transport on the northern and eastern margins of the Namib Sand Sea, although their strength rapidly decreases towards the west (Lancaster et al., 1984;Lancaster, 1985). The range of geological and morphological settings, E-W precipitation gradient, and wind dynamics in the study area provide prime test conditions for assessing the controls on sand ramp formation.

Sand Ramp identification and selection
Sand ramps were primarily identified using Google Earth™ satellite imagery, much with a 15-30 m (derived from Landsat imagery) or finer (CNES/Astrium and Digitalglobe) resolution.
The identification of sand ramps was an iterative process. The aerial morphology of previously studied ramps (Bertram, 2003;Lancaster and Tchakerian 1996;Bateman et al., 2012) was initially used as a reference training set and all potential ramps in the study area were identified. Features were identified by first locating potential accommodation spaces (hillslopes and topographic barriers) using systematic scanning and secondly by searching for accumulations of sand against these mountains. The simple regional geomorphology of inselbergs within a plain and low vegetation cover meant potential accommodation spaces and sand accumulations were relatively easy to identify. Ground-truthing was conducted on a sub-set of the potential sand ramps. This highlighted that some features identified aerially were only thin sand drapes at ground level whilst some features were easy to identify on the ground but overlooked in aerial surveys ( Figure 3). Following this, 57 sand ramps were removed from the inventory based on ground level information or unclear aerial morphology whilst 6 previously unidentified features were added. When present, gullies gave a good indication of sand ramp thickness. Of the 75 remaining sand ramps in the inventory approximately 70% were verified in the field. The remaining 30% were inaccessible at ground level but had clear sand ramp aerial morphology.
Aerial imagery and field observations revealed that whilst sand ramp morphology exists along a continuum, there are four recognizable broad groupings of these features (Table 2). A (nested) classification system was developed to test whether morphological characteristics could be used as a predictor of the accumulation history (and maximum age) of the sand ramps.
Sand ramps were initially classified based on the presence/absence of morphological features such that: class 1 are connected to the hillslope with a unconsolidated sand surface and no secondary dune features on the surface; class 2 are disconnected from the hillslope by a gully, often with a vegetated surface but no secondary dune features; class 3 are disconnected from the hillslope by a gully, often with an indurated surface that is well-vegetated and overlying secondary dune features and class 4 are disconnected from the hillslopes by a large gully, frequently with a duricrust surface and presence of overlying secondary dunes. Following this classification, a relationship between size and morphology emerged with size increasing with morphological class (Table 2). We therefore hypothesise that size (related to available accommodation space) and increasing morphological complexity will be positively correlated with the length of the accumulation history preserved.
Ten sand ramps were selected for detailed investigation in the field. Sites were chosen to reflect a variety of the identified characteristics (three are class 1, two class 2, two class 3 and three class 4) ( Figure 4). The majority of these features had natural exposures, formed by gullying, which allowed sedimentary structures to be observed in detail. In these cases, OSL samples were taken by hammering opaque tubes into cleaned faces. Samples were taken to provide depositional ages for sedimentary units: (i) to bracket changes in sedimentary features (e.g. stone layers) and (ii) where the sediment was relatively homogenous, a resolution of ~1 m was used to assess accumulation history. For large exposures (e.g. Sarama) sampling was restricted by accessibility and resolution reduced accordingly. When gullies were not available (e.g. Neuhof-1) hand-dug pits were used to examine the sedimentology of the top ~2 m of sand ramps sediments and samples were taken to bracket changes in sedimentary features or at a resolution of ~0.5 m (sampling resolution was reduced due to the smaller size of exposures). The dunes overlaying class 3 and 4 sand ramps were sampled to assess their relationship with the sediments of the main sand ramp. Samples were taken from hand augured holes up to 5 m deep using a Dormer™ portable auger adapted for OSL sampling (Telfer and Thomas, 2007). Samples were taken at ~1 m resolution although depth of sampling was frequently limited to the top ~2 m due to poor sand cohesion.

Sedimentological analysis
Field-logging of exposures was undertaken, including broad-descriptions of sediment composition in different units. Material from the ends of OSL sample tubes were used for sedimentological analysis. Particle size analysis of the <2 mm size fraction was conducted using a Malvern Laser Granulometer Hydro Mu 2000 using wet dispersion and ultrasonic diffusion (pilot tests indicated pretreatment to remove organics and carbonates was unnecessary). Statistics were calculated using the Folk and Ward (1957) formula and the GRADISTAT™ program (Blott and Pye, 2001). The >2 mm fraction was not measured. Percentage carbonate and organic content were calculated using sequential loss on ignition at 550 °C and 950 °C (Heiri et al., 2001).
Analysis of heavy mineral assemblages provides constraints on the mineralogy of the sediment source and therefore can be used to identify the most likely sediment provenance (Morton and Hallsworth, 1999). Heavy mineral analysis was conducted at the University of Milano-Bicocca. Heavy minerals were separated using sodium polytungstate (density ~2.90 g/cm 3 ). After mounting with Canada balsam, ≥200 transparent heavy-mineral grains were point-counted to obtain real volume percentages (Galehouse, 1971). Heavy-mineral concentration was calculated as the weight percentage of total (HMC) and transparent (tHMC) heavy minerals relative to the bulk-sediment sample (Garzanti and Andò, 2007). Variance between sand ramps samples and samples from the Namib Sand Sea and major Namibian ephemeral rivers (see Garzanti et al., 2012 for sample details) was assessed using un-scaled principle component analysis (PCA) (see Wold et al., 1987).

OSL sample preparation and measurement
All samples were prepared under subdued orange light at the Oxford Luminescence Dating Laboratory. The quartz fraction was isolated to 90-125 µm for dating and180-210 µm for sensitivity experimets, using, treatment with H 2 O 2 and HCl, density separation using sodium polytungstate to 2.58 -2.72 g/cm -3 wet sieving and a 60 minute etch in 40% HF. 14 samples were additionally treated with H 2 SiF 6 for 7-14 days to remove persistent feldspar contamination.
All luminescence measurements were conducted using a Risø TL/OSL-DA-15 automated readers (Bøtter-Jensen et al., 2003) fitted with either blue (470 nm) or green (525 nm) LED array. Infra-red (IR) simulation was conducted using a 870 nm laser diode array. Luminescence was measured using an EMI9235QA photomultiplier fitted with 7 mm of Hoya U-340 filters. Beta irradiation was administered using 90Sr/90Y beta sources calibrated relative to the National Physical Laboratory, Teddington Hotspot 800 60Co γ-source (Armitage and Bailey, 2005).
Typically, 24 small aliquots (~500 grains, Duller, 2008) were measured using a modified Single Aliquot Regeneration (SAR) protocol (Wintle and Murray, 2006). Preheats (PH1-260 °C for 10 s and PH2 -220 °C for 10 s) were determined following preheat plateau tests on recovered doses from a sub-set of samples (Sup. Info.1). OSL was measured at 130 °C for 40 s or 100 s (blue or green diodes respectively). Each SAR cycle was followed by a 280°C optical bleach for 40 s or 100 s. A repeated dose (recycling ratio; RR point) was included to monitor sensitivity change and an IR-depletion ratio point (Duller, 2003) was included to test for feldspar contamination. Recuperation was monitored using a zero point. Samples were rejected if repeat RR or IR points were >10% of unity, recuperation was >5% of the natural signal or samples had a fast ratio <20 (Durcan and Duller, 2011). Dose recovery tests (DRTs) were performed on a subset of samples. Before irradiation to magnitudes similar to the equivalent dose, discs were twice bleached for 1000 s at 20 °C separated by a 10,000 s pause. All measured samples recovered the given dose within 10% (Sup. Info. 1) Dose response curves (DRCs) were fitted with the most appropriate of a single saturating exponential (SSE), saturating exponential plus linear (SEPL) or double saturating exponential (DSE) function. Individual D e estimates were derived using interpolation of the natural onto this curve. Uncertainties were calculated using 1000 Monte Carlo fits of the curve and propagated with a 2% measurement error. Sample D e s were calculated using the Central Age Model (CAM) (Galbraith, 1999).
Radionuclide (232Th, 238U and 40K) concentrations of all samples were measured using ICP-MS (232Th and 238U) and ICP-OES (40K). Where available, gamma spectrometry was used to determine the gamma contribution. Radionuclide concentrations were calculated using the window method (Aitken, 1985). Radioactivity was calculated using the Liritzis et al. (2013) conversion factors. Based on the aridity of the modern climate, values used in previous studies in the region (e.g. Bristow et al., 2005, 2007, Stone and Thomas, 2008, Stone et al., 2010b and the position of samples (either free draining slopes with evidence of fluvial activity or dunes) a water value of 5 ± 3% was used for all samples. Beta attenuation was corrected for using Guérin et al. (2012) and the etch depth values of Bell (1979). Cosmic dose was calculated using the formula of Prescott and Hutton (1994). Total dose rates were calculated using DRAC (Durcan et al., 2015).

Luminescence sensitivity tests
The sensitivity of quartz (counts produced per given dose) can provide insights into the environmental history of a sediment sample. This is based on three observations: (i) quartz is sensitised in nature and the laboratory through repeated cycles of signal accumulation and bleaching without heating (Fitzsimmons et al., 2010;Moska and Murray, 2006;Pietsch et al., 2008), (ii) sensitivity increases significantly after heating to 500 °C in the laboratory (e.g Bøtter-Jensen et al., 1995;Poolton et al., 2000) and (iii) sensitivity changes due to heating to 500°C are dependent on initial sample sensitivity, with already-sensitised quartz undergoing a subdued sensitivity increase (Moska and Murray, 2006). Therefore, grains that have undergone repeated burial and bleaching cycles in nature, including through aeolian processes, produce more luminescence for a given dose than freshly weathered quartz (Fitzsimmons et al., 2010;Pietsch et al., 2008;Sawakuchi et al., 2011a) and are less likely to undergo very large sensitivity changes after heating (Moska and Murray, 2006). This approach is particularly relevant for sand ramps that result from multiple depositional processes including colluvium that may be relatively freshly weathered from the topographic barrier and aeolian sediment that may be derived from local and distant sources. OSL sensitivity tests were therefore applied to different grain-size fractions to investigate whether material may have been sourced from more freshly weathered or distant sources.
Sensitivity test were conducted using the SAR protocol and conditions described above. Aliquots were twice bleached at 20 °C for 1000 s with a 10,000 s pause between bleaches. The response to a 50 Gy given dose (normalised by weight) was measured seven times. The first four measurements were conducted to measure initial sensitivity and change with dose cycle. Prior to the fifth measurement aliquots were heated to 500 °C for 10 s and measured twice to assess the impact of annealing. The final measurement was an IR-depletion point (Duller 2003). Measurements were conducted on ten medium aliquots of 90-125 µm quartz (~2000 grains; Duller 2008) and ten large aliquots 180-210 µm quartz (~1000 grains; Duller 2008) of 8 samples. Dune samples NAM07/3/2 and NAM07/4/12 from linear dunes of the Namib Sand Sea (see Stone et al., 2010b for details) and NAM14/10/3 from dunes overlaying the Samara sand ramp were used as examples of aeolian material (assumed to undergone a number of cycles of sedimentation) from the region. These were compared with sand ramp body samples AUS13/1/1 (Aus), NAM14/1/2 (Rosh Pinah), NAM14/7/3 (Neuhof-1), NAM14/8/3 (Neuhof-2) and NAM14/12/1 (Sandkop).

Sand ramp location and morphology
In total, 75 sand ramps were confirmed within the study region (Figure 2 & Sup. Info. 2). ~90% are climbing features as identified by their position in relation to topography and their morphology (see Lancaster and Tchakerian 1996;Chojnacki et al. 2010;Goudie 2013, Ellwein et al. 2015. 74% are predominantly east facing and 17% are west facing (Figure 5 & Sup. Info 2). Accumulation occurs in sheltered alcoves or along elongate mountain fronts aligned perpendicular to the wind with 95% of sand ramps accumulated against inselbergs or inselberg complexes and 5% against the edge of the Great Escarpment (Sup. Info 2 & Figure 5). All sand ramps are located either within 4 km of a large ephemeral river channel or within 5.5 km of a dune field (Sup. Info 2 & Figure 5).
Based on aerial morphology and field analysis, 33 sand ramps were classified as class 1, 23 as class 2, and 15 as class 3 (Sup. Info 2). As the identification of class 4 sand ramps requires sedimentary information, only Aus, Samara, Jagkop and an additional examined but unstudied sand ramp were grouped in this category (Sup. Info 2). The dunes overlaying the class 3 and 4 sands are perpendicular to the orientation of the sand ramp in 75% of the ramps and orientation is clearly modified by topography in 55% of cases. Figure 6 depicts the sediment logs from the exposures and augured profiles within the sand ramps and from the overlaying dunes. Sup. Info. 3 gives detailed sediment properties of the <2 mm sizefraction for samples within the sand-rich units.

Sand ramp sediments
A number of broad sediment characteristics are observed within the 10 sand ramps. These provide insight into depositional processes and are described as a series of 'units'. Unit 1 is the dominant sediment in the main bodies of all ten sand ramps. This is a matrix of moderately-well to poorly sorted medium sands with varying proportions of sub-rounded to angular gravels. This unit is split into loose unconsolidated sediments, found at the class 1-3 sand ramps (Unit 1a) and consolidated sediments observed at the class 3 and 4 sand ramps (Unit 1b), and suggests a dominance of aeolian sediments with minor components of gravel clasts provided by slope processes. Unit 2 represents situations where slope processes have dominated. Units comprise semi-continuous horizontal layers of cobble-sized material with occasional boulders, and are found interdigitated within the sediments at Aus (class 4), Jagkop (class 4), Samara (class 4) and Neuhof-1 (class 1) sand ramps. Unit 2 sediments are also found as a surface unit on parts of Aus (class 4), Jagkop (class 4), Samara (class 4), Neuhof-1 (class 1) Simplon (class 1) and Sandkop (class 3) ramps. Unit 3 comprises calcrete duricrust units, which are interpreted as pedogenic; these are found on the surface of class 4 ramps (Samara, Jagkop and Aus). There are less-well developed calcrete nodule layers found at depth within the Keerwader ramp. Unit 4 represents the purely aeolian sediments found in the dunes that overlay class 3 and 4 sand ramps. This comprises moderately-well to poorly sorted medium sand without any larger clasts, which alongside the morphology of the features can be unambiguously assigned to aeolian deposition.

Sand ramp ages
Details of the luminescence ages are displayed in Table 3. The 10 dated sand ramps indicate that sand ramp and dune accumulation in southern Namibia occurred episodically for >200 ka with the majority of dated activity occurring between ~80-12 ka ( Figure 6). Holocene dates are mostly obtained from dunes, near surface sediments or slumped gullies. Material from the main body of class 1 and 2 sand ramps, and the overlying dunes on Klein Aus (class 3), Samara (class 4) and Aus (class 4) provide the simplest data to interpret. The remaining dataset contains three more complicated attributes which require further explanation.
First, for some young samples ages were obtained from pilot studies of <10 aliquots. These are used to give indicative ages, and are displayed on Figure 6 as (< age). Pilot analysis of gully samples from Rosh Pinah (NAM13/1/1 and NAM13/1/2) and Klein Aus (NAM14/15/1 and NAM14/15/3) are younger than expected and show greater than expected scatter in D e values indicating these sediments have been reworked, most likely due to slumping. Dune samples from Sandkop (NAM14/13/1 and NAM14/13/2) contained very high levels of feldspar contamination. Pilot studies indicated these samples were very young (<0.1 ka) but no further analysis was conducted.
Second, a number of gully and dune samples from class 2, 3 and 4 sand ramps are approaching luminescence saturation (~ 200 Gy in this dataset). These ages are therefore quoted as minimum ages (> age).
Third, there are a number of samples with high (>30 %) to very high (124%) overdispersion values that require interpretation alongside their field setting, and caution to be applied when their age estimates are considered. These ages are displayed in italics in Figure 6. 124% overdispersion in sample NAM14/8/1 (Nauhof -2) is linked to very high concentrations of plant roots in the sediment body from which the sample was taken. Therefore, overdispersion is most likely a result of bioturbation. 76% overdispersion in NAM14/5/3 (Keerwader) may result from beta microdosimetry caused by the calcrete nodules which surround this sample. Relatively high overdispersion is also observed in Neuhof-1 samples NAM14/7/1 (43%), NAM14/7/2 (34%) and NAM14/7/3 (32%) and Samara dune sample NAM14/10/3 (38%). This may be due to the heterogeneous nature of the Neuhof -1 substrate and the position of NAM14/10/3 above a calcrete layer creating beta heterogeneity. For Aus dune sample AUS13/6/1 (overdisperson of 40%) very low luminescence sensitivity is a significant contributor to the observed scatter.

Relationship between age and sand ramp morphology
The oldest sediment ages were obtained from the sediments of class 4 sand ramps and indicate accumulation >200 ka. The main bodies of class 2 and 3 sand ramps, aside from Klein Aus, date to 12->75 ka whilst all class 1 dates are younger than 40 ka ( Figure 6). This suggests some relationship between sand ramp size and complexity with age but only to the extent that the two end members, class 4 and class 1, are the oldest and youngest features respectively. However, as class 1 ages are derived from the top 2 m of sediment or from toe sediments their relative youth in comparison to class 2 and class 3 features may be a function of sampling strategy. The erroneously young ages obtained from the gullies of Klein Aus and Rosh Pinah are thought to be due to recent slumping as sediments at both these locations struggled to hold a face. Dates from Neuhof-1 and Rosh Pinah (class 1) were only obtained from the top 2 m of sediment wherein they show some stratigraphic constancy with dates from class 2 and 3 sand ramps ( Figure 6). Near surface sediments of most dunes (class 3 & 4) and sand ramp bodies (class 1 & 2) show activity within the last 2 ka ( Figure 6).

Relationship between age and sand ramp stratigraphy
Duricrust layers (Unit 3) are associated with a hiatus in sedimentation of >~70 ka (Jagkop, Samara) and calcrete nodules at Keerwader are associated with a hiatus of 13-20 ka ( Figure 6). Consolidated sediments (Unit 1a) were deposited >~90 ka. The chronological significance of stone layers (Unit 2) is site specific, at Jagkop and Samara clast layers could represent a long-term hiatus in deposition whilst at Neuhof-1 the stone layer represents a maximum hiatus of 7 ka (~4 ka weighted mean). Unconsolidated sand ramp sediments (Units 1b and 4) are dated to between 0.2 ->75 ka with no clear relationship between sedimentary features and accumulation rate ( Figure 6).

Sediment provenance
The results of heavy mineral analyses are displayed in Figure 7 and can be used to assess whether the aeolian components of sand ramp sediments are derived from local or far-travelled sources.
All Aus samples contain a high garnet concentration and a secondary amphibole component ( Figure  7b). This matches the Mesoproterozoic rocks of the Namaqua Metamorphic Complex which outcrops in the Aus region and consist mainly of garnet/kyanite to garnet-sillimanite paragneisses with subordinate orthogneisses and amphibolites (Becker et al. 2006). At Neuhof the sediments of Neuhof-1 (NAM14/6/2, NAM14/7/3) and Neuhof-2 (NAM14/8/3) and the ephemeral river channel at the base of Neuhof-1 (NAM14/6/R) are similar, all dominated by a very high amphibole content and a lesser epidote content ( Figure 7B). This shows a strong agreement with the Neuhof Formation described in this area (Becker et al. 2006). The presence of kyanite in NAM14/7/3 and NAM14/8/3 (Figure 7) suggests a significant contribution from locally exposed amphibole-facies metasediments. At Jakop, dune (NAM14/4/2) and sand ramp (NAM14/3/2) sediments are similar but not identical. Both contain significant pyroxene suggesting an affinity to the sands of the Orange River and the Namib Sand Sea (Garzanti et al., 2012) (Figure 7B). However, heavy mineral concentration is low and the significant presence of amphibole, garnet, epidote and ZTR is akin to the local Sinclair Group geology ( Figure 7B). The absence of pyroxenes at Aus and Neuhof indicates zero to negligible supply of sediment from the Namib Sand Sea (Garzanti et al., 2012) (Figure 7B).
Comparison of sand ramp sediments with samples from the Namib Sand Sea and the major ephemeral rivers of southern Namibia (Garzanti et al. 2012) using PCA analysis shows Neuhof and Aus are distinct from each other and from the Namib Sand Sea and major river systems ( Figure 7C). Neuhof-1 and the associated channel are closely related, whilst Jagkop shows similarities with the Namib Sand Sea and the major ephemeral rivers ( Figure 7C). Overall these results indicate a local origin for the sand ramp sediments and suggest that the river sediments at Neuhof have contributed material to the sand ramp.
Luminescence sensitivity tests on Unit 1 sediments from five of the sand ramps provide some additional insights into the provenance and depositional processes associated with these sediments.
Aeolian reference samples (NAM07/3/2, NAM07/4/12 and NAM14/10/3) show negligible sensitivity change in both size fractions after heating to 500°C. This indicates they have been significantly sensitised in nature by repeated burial cycles (Fitzsimmons et al., 2010;Moska and Murray, 2006;Pietsch et al., 2008). However, these reference samples also highlight a likely artefact relating to grain size and aliquot dimensions. Following initial measurement, the 90-125 µm size fraction is more sensitive than the 180-210 µm fraction (Figure 8a). This is most likely a product of an absolute increase in the number of luminescing grains in the smaller size fraction as the use of different aliquot sizes was not quite sufficient to normalise the number of grains. (medium aliquots of 90-125 µm size grains likely to contain ~2000 grains compared to ~1000 grains in large aliquots of 180-210 µm (Duller, 2008)). This artefact is common to both aeolian reference and sand ramp samples. Therefore, the magnitude of sensitivity difference between grain-sizes in the aeolian reference samples provides a baseline to compare with the sand ramp sample data.
The 180-210 µm fraction from the Aus sand ramp sample (AUS13/1/1) demonstrates a large increase in luminescence signal intensity following heating (Figure 8b). This may partly reflect the different provenance of Aus sediments and the aeolian reference samples (Sawakuchi et al., 2011a(Sawakuchi et al., , 2011b Figure 7), but also suggests that the sediment has undergone fewer luminescence dosing and bleaching cycles. The sensitivity change after heating of the 90-125 µm size fraction is far more subdued ( Figure 8B). In addition, the 90-125 µm size fraction is an order of magnitude more sensitive than the 180-210 µm fraction when responding to a given dose of 50 Gy ( Figure 8A). This is well above the baseline difference observed in aeolian samples ( Figure 8A) and indicates a real difference in sensitivity between the size fractions.
A notable sensitivity change after heating to 500°C is also observed in the 180-210 µm size fraction of the Sandkop sample (NAM14/12/1) with some difference in sensitivity between the grainsizes also observed (Figure 8). The observations from other sand ramps are less conclusive. This preliminary dataset highlights the potential of this technique as a provenancing tool but also demonstrates the variability of sand ramp sediments and the need for a larger dataset before definitive conclusions can be drawn for all sand ramps.

Controls on sand ramp formation
Data on the locations and sediments of confirmed sand ramps (Figure 2, Figure 5) indicate that four factors govern sand ramp formation in southern Namibia:

an appropriate accommodation space;
3. directionally-persistent winds with sufficient energy for aeolian sand transport; 4. an arid to semi-arid climate but with sufficient seasonal, or longer term, variability to promote differing geomorphic and depositional processes.
All four factors need to be met for sand ramp accumulation to occur (Figure 9). Factors 1 and 2 are local and determine where sand ramps form in a region where factors 3 and 4 are met. Within southern Namibia the bimodal wind direction and local perturbations of factor 3 also have a local influence on the distribution of sand ramps. Therefore, the pattern of sand ramp (non) occurrence in southern Namibia can be explained by factors 1, 2 and 3 ( Figure 9).
All sand ramps are within 5.5 km downwind of sand sources in the form of river channels or dune fields (Sup. Info 2 & Figure 5). Heavy mineral analysis indicates the sediments of the Aus, Neuhof-1, Neuhof 2 and to some extent Jagkop sand ramps are sourced from the local geology (Figure 7). The Jagkop sand ramp sediment also shows some affinity with the Namib Sand Sea, reflecting its location on the margin of the the Namib Sand Sea. For the Neuhof sand ramps the affinity to the local stream indicates that this is a potential sediment source (Figure 7).
Ramps do not accumulate against larger mountain complexes with significant run-off generation or when there are no sheltered accommodation spaces perpendicular to the wind (Figure 9). Thus, suitable accommodation spaces must be subject to minimal aeolian and alluvial erosion. Typically, these are small concave facets on inselbergs/small inselberg complexes or along the flanks of elongate inselbergs (Sup. Info 2, Figure 2). Local topographic perturbations in wind dynamics influence aeolian transport potential (Xianwan et al., 1999) and thus sand ramps do not form in accommodation spaces where winds are disrupted by topographic obstacles, despite the occurrence of available sediment sources (Sup. Info 2 & Figure 2, Figure 9).
Over 70% of sand ramps in this study are east-facing. All of these ramps are climbing features and all morphological classes are represented. >80% of west-facing sand ramps are small, class 1 features and ~10% are falling features. West facing class 3 or 4 ramps are not observed (Sup. Info 2 & Figure  5). This suggests that seasonal easterly winds are the primary creator of sand ramps in southern Namibia whilst the prevailing SW winds also have a small influence. Wind measurements are sparse on the eastern margin of the Namib Sand Sea (Livingstone et al., 2010). Nevertheless, available data agree with the distribution of sand ramps. SW winds are capable of reaching transport velocities ~2% of the time but wind speeds rarely exceed 8 ms -1 . Easterly winds only reach transport velocities ~0.5% of the time but can reach speeds of 14-17 ms -1 and thus total sand transport potential is greater (Lancaster et al., 1984;Lancaster, 1985). Satellite imagery indicates that there are several west facing accommodation spaces that currently do not contain sand ramps (e.g. Figure 9).

Timescale of sand ramp accumulation
OSL dates confirm provisional dating by Bertram (2003) and indicate that sand ramps have been present in the Namibian landscape for over 100 ka (Figure 6). This demonstrates that sand ramps are not solely features of the Last Glacial and that they can act as long term sediment stores. Akin to Telfer et al (2012) and Ellwien et al (2015) deposition is episodic for most ramps with only Simplon (class 1) potentially representing a single, rapid phase of deposition as described by Bateman et al (2012) and Thomas et al (1997) (Figure 6). Chronologies vary between sand ramps reflecting local accommodation space availability, sediment supply and wind dynamics but common periods of activity are apparent ( Figure 6).

Significance of sand ramp morphology
There is some relationship between sand ramp morphological class and age but only to the extent that the large and sedimentologically complex class 4 sand ramps are of greatest antiquity whilst younger ages are found from the smallest, class 1 features ( Figure 6). However, basal, head gully dates are not available for the class 1 sand ramps so a true comparison with other classes cannot be made. The lack of a clear relationship between morphology and age indicates that sand ramps do not follow a common evolution of increasing size and complexity over time and therefore, asides from the class 4 sand ramps with significant duricrust formation, morphological class cannot be used as a predictor of age. Instead, the size and morphological complexity of the sand ramp is more likely to be determined by the size and shape of the accommodation space.
Akin to the sand ramps in this study, large free-form dunes are frequently superimposed by smaller dunes (Lancaster, 2009;Livingstone et al., 2010). The mechanism controlling superimposition is debated with large scale changes in wind regime and climate (Warren and Allison, 1998), seasonal or annual variation in wind (Bristow et al., 2007), landscape patterning and self-organisation (Dong et al., 2009;Ewing et al., 2006) and stabilisation of the underlying dune due to climatic change (Dong et al., 2004) all suggested as potential drivers. There is some evidence that the size of the dune determines the quantity and complexity of superimposed features with secondary features unable to form where the primary dune is below a certain size (Al-Masrahy and Mountney, 2013; Breed and Grow, 1979). Thus, secondary dunes may simply be a function of airflow fluctuation created when the flanks of the dunes are large enough to form a planar surface (Lancaster, 1988). In this study superimposed dunes only form on sand ramps with a length or width >1 km with the number of secondary features increasing with sand ramp size (Sup. Info. 2). Therefore, for these sand ramps, superimposed dune features are likely to represent the opportunistic filling of the secondary accommodation space created by the planar flank of the ramp.
The main bodies of the sampled class 3 and 4 sand ramps are semi-consolidated to consolidated (Unit 1b) and date to >60 ka. The superimposed dunes of class 3 and 4 ramps (Unit 4) and the main bodies of class 1 and 2 ramps are unconsolidated (Unit 1a) and have episodic chronologies from present to >75 ka ( Figure 6). This suggests that after sand ramps reach a certain size (~1 km in length or width in this study) aeolian deposition becomes focused on the superimposed dunes, enabling the underlying sand ramp sediments to stabilize. Therefore, class 3 and 4 ramps may preserve initial accumulation with little reworking. The small size of class 1 and 2 sand ramps inhibits the formation of secondary features and thus, unless protected by a talus mantle, surface sediments are vulnerable to aeolian reworking. The parallels between dune and class 1 and 2 chronologies ( Figure 6) suggests both are responding to the same environmental forcing. In this way, the size of the accommodation space, and thus the size of the sand ramp, controls the accumulation history of the sand ramp.
Accommodation space may also control the formation of the gullies which separate the head of sand ramps from topographic obstacles. Steep slopes promote reverse flow and thus aeolian erosion (Tsoar 1983;Pye and Tsoar 1990;Qian et al. 2011a;Qian et al. 2011b) whilst the accommodation space also channels run-off to promote either continued erosion downslope or erosion at the head of the sand ramp. All gullies now show evidence of flow and fluvial erosion but, depending on the accommodation space, aeolian activity may have played a role in initiation at some sites.
Gullies at the head of the sand ramps protect the sediments from downslope fluvial erosion but also provide a mechanism for sediment to be removed from the sand ramp via aeolian erosion and slumping. This is exemplified by the erroneously young ages obtained from the gully of Klein Aus and from the shallow gully of Rosh Pinah ( Figure 5). Therefore, if gullies have been present throughout sand ramp accumulation, episodic deposition at Keerwader and Neuhof-2 may be explained by repeated periods of aeolian erosion and subsequent accommodation space availability and accumulation. This may explain some correspondence between the dune and sand ramp records.

Significance of sedimentary features
The consolidation of sediments (Unit 1b) and the formation of duricrust layers (Unit 3) is associated with sand ramp stability whilst the significance of stone layers and the chronology of unconsolidated sediments is site specific (Figure 6). Well-developed calcretes (such as those at Samara, Jagkop and Aus) require lengthy development times under stable conditions with little sediment input (Wright, 2007). At Samara, calcrete formation is associated with a ~70 ka hiatus in accumulation whilst ages from consolidated sediments below duricrusts at Aus and Jagkop are >100 ka indicating long term sand ramp stability. Calcretes increase in complexity with time (Netterberg, 1969). Simple calcrete nodules at Keerwader are associated with a 13-20 ka hiatus, indicating medium term stability. Akin to Bateman et al. (2012) and Rendell & Sheffer (1996) buried stone layers (Unit 2) are not always associated with long accumulation hiatuses. At Neuhof-1 the stone layer represents between c.1.6-7.1 ka (mean 4.3 ka) possibly suggesting event driven distribution (cf Bateman et al., 2012) whilst a longer period of stability is possible for Jagkop ( Figure 6). Unconsolidated sand ramp sediments (Unit 1a) do not necessarily represent recent activity with ages ranging between 0.2 ->75 ka with no clear relationship between sedimentary features and chronology.

Palaeoenvironmental significance of sand ramps
Due to the number of geomorphic processes operating on sand ramps, accumulation is unlikely to be a binary process of solely aeolian or solely colluvial deposition followed by preservation. This is typified by unit 1 which indicates a mix of both aeolian and colluvial processes operating within the same unit. In addition, the difference in luminescence sensitivity between size fractions of AUS13/1/1 suggests that the 90-125 µm fraction has undergone several charge accumulation and bleaching cycles (i.e. an aeolian sediment history) before being deposited whilst the 180-210 µm fraction is more likely to be freshly weathered material (i.e. colluvium) (see Moska & Murray 2006;Pietsch et al. 2008;Fitzsimmons et al. 2010). The presence of both populations in a single sample indicates interplay between the two depositional mechanisms. This is supported by the poor sorting of the <2 mm fraction, the presence of angular clasts amongst the aeolian sands and the frequent absence of bedding structures in aeolian units (Lancaster & Tchakerian 1996;Bertram, 2003).
Akin to the sand ramps of the Mojave (Pease and Tchakarian, 2003), sediment supply to the Aus and Neuhof sand ramps is of local origin with the heavy mineral assemblages reflecting the local lithology (Figure 7, Becker et al. 2006). At Neuhof-1 the heavy mineral suite from the sand ramp and the neighbouring ephemeral river channel are nearly identical, confirming the potential of the river channel as a sediment source. At Japkop the heavy mineral assemblage closely resembles the local lithology but with 15-30% contribution from the dunes of the Namib Sand Sea which are immediately to the south and east of the mountain. All studied sand ramps are within 4 km of river channels or within 5.5 km of dune fields (Figure 5), suggesting these are the most likely sediment sources. The local origin of the sediments suggests that periods of colluvial and aeolian deposition both reflect local environmental conditions and if the accommodation space is available, local sediment supply is expected to be the controlling factor in sand ramp activity.
Individual sand ramps therefore reflect local conditions wherein sediment availability coincided with an available accommodation space. However, due to the environmental controls determining sediment dynamics, analysing multiple sand ramps in unison may provide a record of regional environmental conditions. When multiple sand ramps are analysed the local, and temporally variable, influence of accommodation space is reduced meaning periods of aeolian accumulation observed across multiple sand ramps are likely to be environmentally significant. If factors controlling sediment supply are understood, the simultaneous analysis of multiple sand ramps with geomorphically similar sediment sources may provide relatively long and stable records of Quaternary sediment dynamics, palaeoenvironmental conditions and potentially palaeoclimatic conditions in regions typically devoid of detailed Quaternary records

Conclusions
Sand ramps are abundant along the eastern margin of the Namib Sand Sea and are present, although less common, in other locations in south west Namibia with 75 sand ramps identified in the study region ( Figure 2). These sand ramps range from 150-3000 m in length with varying morphology and sedimentology. OSL dating indicates episodic sand ramp activity between 0.2 and >200 ka ( Figure 6) although more recent ages may represent surface reworking. This confirms preliminary dates from Bertram (2003) and indicates that sand ramps are not solely Last Glacial features but existed in the Namibian landscape during the Last Interglacial, and likely before.
Analysing where sand ramps do and do not form across the varying geology and climatic gradients of the study area allowed the conditions needed for sand ramp formation to be assessed. These controls are proposed to be: 1. an adequate supply of sediment with significant aeolian transport potential; 2. an appropriate accommodation space; 3. directionally-persistent winds with sufficient energy for aeolian sand transport; 4. an arid to semi-arid climate but with sufficient seasonal, or longer term, variability to promote differing geomorphic and depositional processes Four classes of sand ramp morphology were identified. These closely link size to complexity and indicate that the available accommodation space, and the ultimate size a sand ramp can grow to, play a large role in the morphology of the sand ramp and potentially its sensitivity to environmental change. Sediments of large sand ramps are able to stabilise as aeolian activity is refocused on secondary dune features which are able to form on the planar surface of the sand ramp. Dunes are unable to form on smaller sand ramps, leaving the surface vulnerable to reworking unless it is protected by talus.
The stratigraphy of individual sand ramps reflects local conditions wherein sediment availability coincided with available accommodation space or non-stabilised sediments were reworked.
Accommodation space availability offers a purely local control on sand ramp formation whereas sediment supply provides a regional environmental picture with ephemeral rivers representing the likely sediment source for most of the studied sand ramps. Therefore, the primary record of individual sand ramps is local and is influenced by non-climatic controlling factors including: the size of the sand ramp, the availability of the accommodation space and the sediment supply. However, when multiple sand ramps are analysed in unison with sedimentological and morphological context, regionally common periods of accumulation can be identified. From this regional patterns of sediment dynamics, which are likely to reflect past regional climatic changes, can emerge. This article is protected by copyright. All rights reserved. Aeolain activity controlled by sediment supply. Associated with low sealevel during glacial periods. Colluvium and palaeosols develop during interglacials. # -Number of studied sand ramps L -Length T -Thickness C or F -climbing or falling. n/m -not mentioned in text * Unspecified number of sand ramps were observed and described. 5 sites (including complexes) were studied in detail.

Study & location # L (m) T (m) C/F Sedimentology
This article is protected by copyright. All rights reserved. Overlying dunes present.
This article is protected by copyright. All rights reserved.  (Galbraith, 1999