Sand, hearths, lithics and a bit of bioturbation: Site formation processes at Umhlatuzana rockshelter, South Africa

Umhlatuzana rockshelter is known for its continuous record of Middle and Later Stone Age lithic assemblages. This study presents multiproxy geoarchaeological data (micromorphology, X‐ray diffraction and scanning electron microscopy with energy‐dispersive spectroscopy) to reconstruct the depositional and post‐depositional history of the site. Although the Stone Age deposits macroscopically appear homogeneous, micromorphological analysis reveals the existence of primary, unaltered depositional microlayering throughout the sequence. Sediments related to combustion activities on‐site are observed in both the Holocene and Pleistocene deposits. Post‐depositional geochemical alterations result in the formation of several phosphatic minerals that significantly affect the site's preservation conditions. One of those is vashegyite, a rare magnesium phosphate mineral related to acidic and moist sedimentary environments. Bioturbation features are prominent at the microscale, but sediment mixing does not seem to affect the vertical distribution of the artifacts. The observation of horizontal microlayering in both the Pleistocene and Holocene illuminates the dominant mechanism of sedimentation throughout the site's 70,000‐year occupational history. It moreover shows that the lithics can be analysed as coherent assemblages.

. Sites such as Umhlatuzana with potentially continuous, undisturbed sections throughout the transitional period are required to fuel archaeological discussions on the transition from the MSA to the LSA.This makes Umhlatuzana an important site to document this technological reorganisation in Southern Africa.The site was first excavated in the 1980s by Jonathan Kaplan and yielded a rich chronocultural sequence representing potential pre-Still Bay, Still Bay, Howiesons Poort, late MSA, transitional MSA/LSA and Robberg technocomplexes (Kaplan, 1990;Lombard et al., 2010).
The seemingly continuous sedimentary sequence and the high find density provide a unique opportunity to study technological developments and human behavioural evolution.It also provides an important point of comparison for nearby sites like Border and Sibhudu Caves that encompass comprehensive sequences lacking some of the technocultural manifestations represented at Umhlatuzana.
The stratigraphy of the site is characterised by clearly delineated sedimentary packages within the Holocene, but the Pleistocene sediments appear very homogeneous and lack visible boundaries to subdivide the lower part of the sequence.Kaplan (1990) suggested that because of a large-scale slumping event, Pleistocene sediments were post-depositionally displaced.This would impact the integrity of the site's archaeological assemblages.As such, the archaeological community has been hesitant to incorporate the site's archaeological material into the discourse on the MSA to LSA transition, for example, Mitchell (2002, p. 120) in his overview of Southern Africa states: 'Umhlatuzana is unique in featuring several unifacial points, similar to earlier MSA examples at the same site, but this again raises the concern that the occurrences here are mixed'.
In 2018 and 2019, our team conducted renewed excavations to clarify the site's formation processes (Sifogeorgaki et al., 2020).We employed high-resolution excavation techniques supported by threedimensional (3D) spatial measurements to document the archaeological occurrences.Sifogeorgaki et al. (2020) and Reidsma et al. (2021) have shown that geogenic sedimentation of sand derived from the local bedrock provided most of the sedimentary input and refuted the slumping event proposed by Kaplan (1990)-based on the 3D distribution of artefacts and geochemical analyses, respectively.These geoarchaeological studies also furthered our understanding of other aspects of the site's history, including the poor preservation of organic materials, such as bones and charcoal, in the Pleistocene deposits.
The current study builds upon the recent stratigraphic and geoarchaeological studies (Reidsma et al., 2021;Sifogeorgaki, 2020, Sifogeorgaki et al., 2023;Sifogeorgaki & Dusseldorp, 2022), to provide a holistic assessment of the different formation processes that affected the sequence over the last 70,000 years.Specifically, we aim to address the following questions: 1. How much did human activities contribute to the formation of the sequence?Are there indications that the nature of these activities varies diachronically?How do the site's taphonomic processes affect the preservation of anthropogenic and biogenic material?2. Are there indications for major depositional hiatuses?Can micromorphology help distinguish whether the higher find density zones in the Pleistocene deposits are a result of anthropogenic (e.g., higher occupation intensity) or geogenic (e.g., lower geogenic sedimentation rate) reasons? 3. How did post-depositional processes (e.g., bioturbation) affect the integrity of the archaeological materials and features?Are the anthropogenic finds located within stratigraphically secure contexts?
We characterise the structure of the sequence by micromorphological analysis and document the sedimentary process as well as post-depositional alterations to the sediment fabrics.We support these observations with additional strands of evidence such as high-resolution techniques such as scanning electron microscopy with energy-dispersive spectroscopy (SEM-EDS) and X-ray diffraction (XRD).Micromorphological analysis is crucial because it provides microscopic documentation of depositional and post-depositional processes that occur in an archaeological site (e.g., Branch et al., 2005;Courty et al., 1989;Goldberg & Macphail, 2006;Karkanas & Goldberg, 2018).At Umhlatuzana, the micromorphological samples derive from across the entirety of the sequence.Indeed, although the Pleistocene deposits at Umhlatuzana appear macroscopically uniform, micromorphology may reveal components and structural characteristics invisible to the naked eye, including microstructures that indicate sediment mixing due to bioturbation, trampling, slumping and so forth (Goldberg et al., 2009;Karkanas et al., 2015;Kooistra & Pulleman, 2010;Kovda & Mermut, 2018;Miller et al., 2010).the site while small grassland patches occur on nearby plateaus (Mucina & Rutherford, 2006).
The excavated deposits of Umhlatuzana represent Marine Isotope Stages (MIS) 5-1.Regional paleoenvironmental data suggest that sea surface temperatures (SSTs) were around 20-22°C during the Late Pleistocene, while sometimes falling to 19°C during MIS 4 (Simon et al., 2013;Ziegler et al., 2013).Slightly colder environments than today during MIS 4 are also suggested by paleobotanical work at the nearby site of Sibhudu (Bruch et al., 2012).Colder periods may locally be correlated to increased precipitation (Chase, 2010;Clark, 2013), although other studies connect them to periods of aridification (e.g., d 'Errico et al., 2017).
Drier conditions (decrease in precipitation) are argued to characterise MIS 3 (Baker et al., 2014;Esteban et al., 2020).From late MIS 2 and continuing in MIS 1 (Holocene), SSTs increased to 22-24°C (Simon et al., 2013;Ziegler et al., 2013).Dashed line to distinguish between Group H and P deposits.Higher moisture units are present in the lower/southern part of the profile.The colour of the northern part is affected by shading from the safety scaffolding installed at the site.(d) Western view of UMH.(e) Upper hearth feature exposed in the western profile that is characterised by rubified sediments (R-Unit H4), underlying a charcoal-rich layer (CH-Unit H3) and an ash-rich layer (A-Unit H2).A second ash-rich layer is underlying the upper hearth feature.BC, Border Cave; EBC, Eland Bay Cave; GR, Grassridge rockshelter; PP, Pinnacle Point; SB, Sibhudu Cave; UBB, Umbeli Belli.highway.Subsequently, Jonathan Kaplan undertook an excavation in 1985 opening a trench of 6 m 2 , (squares J2, J3, J4, K2, K3, K4) reaching bedrock at a depth of around 2.5 m in four of them (Kaplan, 1989(Kaplan, , 1990) (Figure 1b).He retrieved and described an abundance of MSA, LSA and some Iron Age finds.He reported a sequence with homogeneous Pleistocene deposits of uncertain stratigraphic integrity.By contrast, the Holocene layers exhibit clear stratification and are characterised by the presence of combustion features (hearths) and sediments (ash, charcoal) (Kaplan, 1990).The sequence was initially radiocarbon-dated using charcoal fragments (Kaplan, 1990;recalibrated in Sifogeorgaki et al., 2020 using OxCal, using the SHCal13 curve;Bronk Ramsey, 1995;Hogg et al., 2013).As a large part of the sequence does not have any preserved charcoal remains and the lower part of the sequence extends beyond the limit of the radiocarbon technique, there are gaps in the site's radiocarbon chronology.Additional optically stimulated luminescence (OSL) dates were obtained for the lower part of the Pleistocene sediments (Jacobs & Roberts, 2008;Lombard et al., 2010).A summary of the dating results to date is provided in Sifogeorgaki et al. (2020, tab. 1; relevant dates are indicated in the profile drawing in Figure 2).

| Research history
We report on fieldwork in 2018 and 2019 in which three subsquares (L2a, L2b, L3a) were excavated to a depth of between 2.04 and 2.40 m (Sifogeorgaki et al., 2020) (Figure 1b).The finds retrieved by the renewed excavations were piece-plotted using a Robotic Total Station.The newly excavated western profile was documented in detail and sampled for various analyses.Geoarchaeological and paleoenvironmental results were presented by Sifogeorgaki et al. (2020), Reidsma et al. (2021) and Sifogeorgaki and Dusseldorp (2022), while a study of raw material categories was published by Sifogeorgaki et al. (2023).

| Umhlatuzana rockshelter stratigraphic sequence
Based on macroscopic observations a clear stratigraphic division was made with the basal ~2 m of deposit correlated to the Late Pleistocene, dubbed Group P. The upper ~50 cm of deposits are correlated to the Holocene, called Group H (Holocene) (Reidsma et al., 2021;Sifogeorgaki et al., 2020) (Figures 1c and 2).The boundary of the two groups is diffused but clear and well-defined.The temporal correlation of the Group P-H boundary is based on radiocarbon dates obtained by Kaplan (1990); the precise chronology of the deposits in this part of the sequence is unclear and the possibility of a hiatus or a major stratigraphical change in this zone was suggested by previous studies (Reidsma et al., 2021;Sifogeorgaki et al., 2020).More detailed descriptions of the units' sedimentological and geochemical properties are presented in Sifogeorgaki et al. (2020) and Reidsma et al. (2021), respectively.The main stratigraphic attributes of the sequence are summarised as follows: 1. Group P deposits appear homogeneous in colour apart from the lower south area where the deposits have a darker colour due to higher moisture content (Sifogeorgaki & Dusseldorp, 2022).This is attributed to seeping water flow from the rockshelter's wall.The boundaries of the Group P units were primarily determined based on changes in find density (Sifogeorgaki et al., 2020).Distinct stratigraphic units with sharp contacts are present within Group H (Sifogeorgaki et al., 2020).The Group H units are rich in ash and charcoal.For example, Units H9 and H7 correspond to partially cemented, well-defined ash layers, Unit H5 to charcoal-rich layers and Units H6 and H3-H4 to in situ hearths.The ash Unit H9 is ~10 cm thick and extends throughout the western, northern and eastern profiles.The hearths are composed of three distinctive layers: an uppermost layer with ash remains (H2, H6a), a layer dominated by charcoal and charred material (H3, H6b) and a lowermost layer of rubified sand (H4, H6c) (Figures 1d and 2).
F I G U R E 2 Stratigraphic drawing of the west profile of squares L3a, L2b and L2a indicating the names of the stratigraphic units (as defined in Sifogeorgaki et al., 2020), the higher and lower find density zones, the micromorphological and XRD sample locations.Radiocarbon (back) and optically stimulated luminescence (brown) dates based on Kaplan (1990) and Lombard et al. (2010) are indicated on the right.Note that the location of micromorphology samples UMH1 and UMH2 is schematic since they derive from the west profile of square K2.UMH, Umhlatuzana rockshelter; XRD, X-ray diffraction.
2. Group H deposits generally demonstrate better bone and charcoal preservation (Reidsma et al., 2021;Sifogeorgaki et al., 2020).Within Group P, the higher moisture sediments have relatively better charcoal preservation, while the lower moisture sediments have relatively better bone preservation.
3. The overall find density within Group P is higher than in Group H (Sifogeorgaki et al., 2020).In addition, within Group P there are further variations in find density.The Group P units were defined based on their relative find density (higher-lower find density).
4. The Pleistocene deposits were divided into relatively higher (ZH1-ZH4) and lower (ZL1-ZL3) find density zones.The higher and lower find density zones could be attributed to higher and lower occupation intensities during the Pleistocene and/or changes in the sedimentation rate (Reidsma et al., 2021;Sifogeorgaki et al., 2020).

Most of the sediment analyses show different geochemical
and sedimentological patterns in Groups P and H. Group H is characterised by higher pH values, higher organic matter content (e.g., loss on ignition analysis) and higher CaO content.Group P is characterised by higher percentages of clay and higher SiO 2 content (Reidsma et al., 2021;Sifogeorgaki et al., 2020).
6.The available radiocarbon dates obtained by Kaplan (1990) suggest a depositional hiatus may have existed between the Group P and H deposits.The uppermost Pleistocene yielded an uncalibrated date of 13,400 ± 120 (Pta-4226), while the overlying Holocene yielded a date of 9180 ± 90 (Pta-4307).However, bioturbation may have affected the dates as below Pta-4226, a date of 9870 ± 90 (Pta-4631) was obtained.
Kaplan suggested that because of a large slumping event, the Pleistocene sediments were post-depositionally displaced 10-15 cm towards the rockshelter dripline (Kaplan, 1990).He argued that the lowermost high-moisture deposits acted as a slip plane for the uppermost lower moisture deposits that resulted in a single rotational slipping event in which both of the units maintained their sediment cohesion.The following observations were used to support this proposal: i. the colour difference of the sediments (higher moisturedark brown, lower moisture-brown) and ii. the stone artefact sequence and raw material composition from the material derived from the higher and lower moisture units.In particular, he considered the vertical distribution of the hollow-based points, which suggested that there is a 10-15 cm offset between the high-and-low moisture sediments, with hollow-based points occurring lower down the sequence in the lower moisture units (located closer to the shelter edge) than in the higher moisture units (located more to the back of the shelter) (Kaplan, 1990, p. 6).
Stratigraphic evidence from the renewed excavations however does not support Kaplan's inferred large-scale sediment movement (Reidsma et al., 2021;Sifogeorgaki et al., 2020).Importantly, the 3D find distribution indicates that the Pleistocene deposits demonstrate discreet higher and lower find density layers that extend across the alleged slip plane (Sifogeorgaki et al., 2020).In the case of a slumping event, we would expect disturbances to this layering.Moreover, the geochemistry of the Pleistocene sediments (XRF, pH, magnetic susceptibility) shows gradual changes (Reidsma et al., 2021), and preliminary micromorphological results did not find evidence indicating large-scale movement (Reidsma et al., 2021).The colour difference between the sediments is only related to moisture differences (Sifogeorgaki & Dusseldorp, 2022).

| Micromorphological analysis
We obtained micromorphological samples from the western profile that was revealed after the 2018 and 2019 excavations (Figures 1 and 2).
We were only able to collect samples from the western profile because the northern profile was damaged due to vandalism, the eastern profile was poorly visible due to tree roots and vandalism and the southern profile was stepped (also see Sifogeorgaki et al., 2020, pp. 555-556).
We collected a total of 35 micromorphology samples mostly using plaster of Paris, except for several samples from the Group H deposits where we used Kubiena tins (Goldberg & MacPhail, 2006;Stoops & Nicosia, 2017).For the current study, we present 19 micromorphology samples covering the entire stratigraphic sequence (Figure 2).We sampled most of the stratigraphic layers of the Group H deposits using a single, 22-cm-long block (Sample UMH22).We took additional Group H samples from hearth features (Units 2-4) and the uppermost Unit H1.Samples UMH1 and UMH2 were taken from the profile of square K2 (before the 2018-2019 excavation), while the rest of the samples were derived from profiles L3a and L2 (after the 2018-2019 excavation).
The samples were prepared at the laboratories of the Cultural Heritage Agency of the Netherlands in Amersfoort.Initially, the blocks were air-dried and afterwards oven-dried at 25-40°C.The samples were subsequently impregnated with a mixture of 2:1 polyester resin:acetone.The impregnated blocks were then irradiated with gamma radiation at the Synergy Health Ede B. V. company to harden the resin.The blocks were cut into ~3-cm-thick slabs and were documented using a digital SLR (Nikon D3400) with a standard lens.We produced 31 soil-thin sections of 30 μm thickness and 11 × 7.5 cm from these blocks (for context see Table 1).
We studied the micromorphology samples under the microscope and through high-resolution thin section scans using a flat-bed slide scanner.We observed the thin sections in plane-polarised and cross-polarised light (PPL, XPL), using a Euromex IS.1053-PLPOLi polarisation microscope with ×10/23 mm eyepiece and trinocular, and magnifications from 5 to ×40.Additional images were taken at the Weiner Laboratory for Archaeological Science of the American School of Classical Studies at Athens using a Leica DM2700P polarisation microscope with magnifications from 1.25 to ×50.We analysed and described the thin sections following standard micromorphological literature (Stoops, 2003;Stoops et al., 2010;Stoops & Nicosia, 2017).
To facilitate interpretations, we present the micromorphological results using the concept of facies.Facies are referred to as parts throughout the sediment sequence that demonstrate similar characteristics (e.g., composition, structure, sedimentation) and are distinct from the surrounding deposits (Karkanas & Goldberg, 2018 and references therein).They are the result of specific depositional or post-depositional processes and are used to document and comprehend the structure and site formation processes at the site.
In addition, a geoarchaeological study based on facies assists in the comparison between sites and can help link formation processes to regional paleoenvironmental changes.
T A B L E 1 Provenience of micromorphological samples and information on the distribution of selected contents (bone, charcoal, iron oxides, phytoliths), as well as the abundance of the facies and subfacies encountered in the thin sections.
The samples in which quartz grains appear fragmented due to sample preparation are indicated with light blue.

| SEM-EDS
Thin sections 35B and 27B were further investigated at the Wiener Laboratory using a Jeol ISM-IT300LV SEM with an Oxford x-act EDS to elucidate the nature and composition of specific phosphate minerals.We analysed the thin sections uncoated using a low vacuum at 15 kV accelerating voltage.The working distance was 10 mm.

| XRD analysis
We analysed a total of six samples using XRD.Samples H1, D2, A14, A11 and A13 are sediment samples, originally collected for the study of Reidsma et al. (2021).We selected these samples to illuminate the mineralogical composition of layers associated with specific micromorphological samples.In addition, we analysed a sample of the secondary minerals encountered at the higher moisture Group P units (Sample 6718).This sample was pure and did not contain clastic sediment or other remains.The exact location of all samples is shown in Figure 2.
The sediments were dried and hand-ground using a mortar and a pestle.The sample holder was filled using the back-loading sampling technique.The samples were analysed in a Bruker D2 PHASER with an LYNXEYE/XE-T detector at the Wiener Laboratory with the following settings: tube Cu, 30 kV voltage and 10 mA current.The 2θ was set from 4°to 64°with an increment of 0.02 and a time step of 1 s.The processing and interpretation of the XRD results were made using DIFRAC.EV software.

| Micromorphology
In some of the thin sections (samples UMH9, UMH22, UMH26, UMH27, UMH28, UMH35), the quartz grains appear fragmented due to the settings of the grinding machine.Other components in those slides (e.g., bone fragments) were unaffected.
A description of the contents observed in the thin sections is presented in Table 1, while Table 2 documents how the different components are distributed within the samples.We present the main facies and subfacies identified in Figure 3.The facies are discussed below, while an overview of their distribution and frequency within the samples is presented in Table 2.

| Sand deposits (SDs)
A large proportion of the Group P deposits are comprised of facies SD (Table 1 and Figures 4-6).Macroscopically, these deposits appear homogeneous, separable only into layers with relatively higher and lower densities of stone fragments.The boundaries of these stratigraphic layers are not obvious on either the impregnated blocks or the thin sections.
Burnt remains in the form of aggregates mostly consisting of fine matter are present within the sediment matrix.The aggregates show signs of rolling through combustion remains (CRs) such as ash and charcoal (Figure 7e).The presence of submillimetre charcoal particles within the aggregates was confirmed by testing through incident light and dark field (metallographic microscope).In addition, we observe large amounts of black, silt-sized particles possibly related to charred material and charred bone (UMH15).The amount of charcoal and charred material varies across the thin sections (e.g., less in UMH3B, see Table 1).
The SD facies is divided into three subfacies: SDl (layered), SDa (altered) and SDb (bioturbated).The approximate boundaries of the subfacies were defined on the high-resolution scans.The abundance of the three subfacies is presented in Table 1.It is observed that the higher moisture Group P sediments have the highest proportion of SDb.The micromorphological characteristics of the three subfacies are provided below:

SDl
Remnants of finely laminated sediment were observed in the Group P deposits (subfacies SDl).The sand grains, silt grains and fine organic matter demonstrate a banded to linear distribution pattern with a horizontal orientation that is intercalated with layers with a more massive appearance of relatively unsorted sand (Figure 4).In addition to the horizontal distribution and orientation of finer fractions, the elongated rock fragments also frequently demonstrated a moderately expressed horizontal orientation (Figure 4).SDl was observed in almost every sample in various abundance (see Table 1).SDl is present throughout the sedimentary sequence except for samples UMH35A, B and C, which corresponds to the lowermost high moisture deposits (Units P16 and P17).The SDl facies seem particularly developed in the uppermost Group H UMH30 sample representing Unit H1 (Figure 4a,b).

SDa
The coarse material that constitutes the groundmass of SDa comprises mostly fine sand composed of subangular to subrounded quartz grains.
The fine material has a brown colour and medium to high interference colours in XPL (Figure 5).The voids within these deposits are mostly complex packing voids and are randomly arranged.Larger voids are occasionally filled with silt and clay pedofeatures, mainly loose infillings and internal hypocoating.Locally, the microfacies show slightly higher void and lower fine fraction proportion and some redder coloration (Figure 5c,d).SDa was observed in most of the samples in various proportions and appears to be dominant in all of the Group P layers with the exception of lowermost layers P16 and P17 (Table 1).

SDb
The SDb subfacies represents clearly distinguishable bioturbation features.The dominant bioturbation features observed include voids, infilled burrows, insect excrements and root tissue remains (Figure 6).
Loose sediment of the infilled channels often creates a crescent pattern (Figure 6b).In addition, channels are also associated with external and internal hypocoating that has a high proportion of fine material and T A B L E 2 Main components observed in the micromorphological samples.

Components of Umhlatuzana micromorphological thin sections Components Description
Sand-sized silicate grains (a) Subangular unsorted fine-to very coarse sand-sized quartz grains dominate the sediments of Umhlatuzana.The dominant source of the quartz minerals is the bedrock sandstone since they appear to have the same shape and size as the bedrock sandstone grains.(b) Very few surrounded fine-to sand-sized glauconite grains that demonstrate green colour in PPL, parallel extinction and high birefringence levels.(c) Very few subrounded to subangular fine-to coarse sand-sized chert fragments are observed.These are also components of the bedrock sandstone and originate from the granular disaggregation of the bedrock.
Gravel-sized rock fragments Very fine-to very coarse gravel-sized rock fragments that are frequently horizontally oriented are common in the Umhlatuzana deposits.The following petrologies are observed, the distribution of which varies across the sequence: (a) Arenite sandstone fragments with sharp boundaries are present throughout the sequence.The arenite sandstone is poorly sorted with grain sizes ranging from ~50 to ~500 μm.The mineralogy of the sandstone consists of predominantly subrounded to subangular quartz grains (~95%) that are bound by siliceous cement.Very few rock fragments (mostly chert) and iron oxide grains are also present while no feldspar grains were observed.(b) Elongated angular hornfels fragments with approximately a 6:1 length-to-width ratio are present.Overall, the hornfels pieces have planar-parallel fabrics, variation in colour (greyish, yellowish-white, greenish, black, etc.) and often exhibit a 'spotted' appearance.The mineral grains in the hornfel sections are not easily distinguishable as their premetamorphosed composition would have been pelitic (clay-rich) sediments.(c) Quartz fragments exhibit an angular shape, a wave type of extinction and low birefringence colours.Secondary granoblastic quartz veins are present in some quartz fragments deriving from the lowermost sample UMH35.(d) Dolerite (aka diabase) comprises a small (~2%) proportion of the total rock fragments.It is a mafic igneous rock with an interstitial (between porphyritic and aphanitic) texture mostly containing interlocking, lath-shaped plagioclase phenocrystals.(e) A total of two siltstones were identified in the micromorphological thin sections (sample UMH_3).They are characterised by interlocking silt-sized quartz crystals and iron-rich minerals.(f) One silcrete fragment was present in a thin section of UMH_3.It is characterised by cryptocrystalline and chalcedonic silica cement.More information on the petrographic and geochemical characteristics as well as the distribution of the rock fragments can be found in Sifogeorgaki et al. (2023) Organic silt and humified organic matter Silt-and clay-sized materials are present as coatings and bridges between the grains and often form sand-sized aggregates.Those aggregates sometimes include silt-sized burning residues such as charcoal and ash.The fine material often appears to be of a darker red colour.The organic silt material appears enriched on the areas connected to high bioturbation activity.

Clay and silt
Fine clay matter is present within the fine material.The presence of clay and mica minerals is apparent from the speckled b-fabric of the fine mass under XPL.Silt-sized mica particles are in the form of colourless lath-like crystallites with second-order interference colours.Silt of a nonorganic origin is potentially present as a result of wind-blown silt particles and/or the in situ weathering of rock fragments (e.g., hornfels).
Iron oxides Some red clay coatings and iron-bearing materials were documented in the samples.Under TL, the opaque materials demonstrate a red, yellow or greyish colour suggesting components such as haematite, goethite and magnetite and were mostly found in two forms: (a) Sand-to gravel-sized iron subrounded nodules of quartz grains embedded in a dark red matrix.(b) Sand-to gravel-sized angular and subangular iron oxide grains or fragments with distinct boundaries.

Charcoal
Silt-to centimetre-sized charcoal fragments appear subangular and, in many cases, fragmented.Some of the charcoal particles contain large vessels concentrated in a ring shape (ring-porous wood).Other charcoal fragments have evenly scattered vessels (diffuse-porous wood).In some cases, vessels are not observed.All types of charcoal often have randomly developed fragmentation.In some cases of ring-porous wood, the cracks are developed parallel or vertical to the orientation of the vessels.Charcoal fragments are observed throughout the sequence.

Ash remains
The ash remains to have a pale-yellow colour (both in PPL and XPL) and appears isotropic (undifferentiated b-fabric).This indicates that the ash calcite converted to apatite (see below).Ash remains are observed only in the Group H deposits.

Bone
Bone fragments have a variety of sizes (sand to centimetre fragments), shapes (rounded to angular), colours (pale white, yellow, orange, red) and structures (compact, spongy).In some cases, Haversian canals are clearly visible.Most bone fragments show characteristics of heating such as cracks and dark orange colours.Bone fragments are present throughout the sequence.
Secondary minerals (a) Secondary amorphous apatite was identified due to its moderate positive relief, and isotropy or very low interference colours under XPL.The apatite replaced various Ca-rich materials such as ash deposits, bone fragments and dung.Secondary apatite was observed locally in Group H and lower moisture Group P deposits.(b) Vashegyite characterised by pale colours and is almost isotropic with an undifferentiated b-fabric.The vashegyite is present in the higher moisture Group P deposits.The mineral is present in the form of millimetre-to centimetre-sized nodules, aggregates and crusts.(c) Secondary authigenic gypsum zones are present in Group H deposits associated with ash (Continues) | 219 often has a laminated or crescent pattern.Other bioturbation features include chambers infilled with organic ellipsoid excrements of a very fine to fine sand size and chambers associated with root penetration (Figure 6).The roots often appear partly decomposed and the root chamber is sometimes infilled with fine material and/or ellipsoid excrements.This facies is present throughout the sequence and is most dominant in the lowermost layers P16 and P17.8).Facies CR was only observed in Group H samples (Table 1).

Ash deposits (CRa)
CRa are deposits rich in ash remains.The ash remains appear isotropic, decalcified and the original contents appear to have been transformed into or replaced by mostly amorphous apatite (carbonate hydroxyapatite or dahlite).The amorphous apatite was first identified under a petrographic microscope due to its moderate positive relief, light grey Components of Umhlatuzana micromorphological thin sections Components Description remains.The gypsum crystals appear elongated, demonstrating random and parallel intergrowth, random orientation, a columnar to lensoidal shape and low interference colours.(d) Calcite crystals were observed in the thin section of UMH2 and demonstrate a columnar shape and form spear-and rosette-like aggregates.
Tooth fragments Two single microfaunal tooth fragments were identified in the thin sections.
Plant remains (a) Opal phytoliths of a silt to very fine sand size were observed.The vast majority was angular with a rectangular or elongated shape, in some cases containing a central diffuse inclusion.Phytoliths occurred as part of the groundmass and were observed throughout the sequence.(b) Cross-and longitudinal sections of root tissue remains were observed.In PPL, the roots appeared pale yellow to yellow, and in XPL, the root remains demonstrated high interference colours.The roots often appeared partly decomposed.(c) Cross-sections of leaves were present only in the uppermost thin section of UMH30.The epidermis, mesophyll and palisade were all visible.(d) Round grass stems of a fine sand size were observed with preserved epidermis, veins, parenchymatic tissue and central pith.The grass stems are observed in both Group H and P thin sections.

Dung
Dung containing plant fibres that appear loose and finely layered are observed in part of the Group H thin sections.
The plant fibres are isotropic, which indicates that they have been converted to apatite.This is also supported by the total absence of spherulites, since spherulites are unstable in a high phosphatic environment (see Brochier et al., 1992;Canti, 1999;Shahack-Gross, 2011).The dung has an orange, brown, dark brown and black colour.
The darker colours suggest heating.The coprolites often appear fragmented, which might be the result of heating or drying.

Microfaunal excrement
Microfaunal excrement is identified in association with bioturbation features.They include organic ellipsoid excrements that have a very fine to fine sand size.
F I G U R E 3 Overview of facies and subfacies observed on the micromorphology thin sections of Umhlatuzana rockshelter.
to beige amorphous appearance and isotropy (or undifferentiated bfabric; Karkanas & Goldberg, 2010b) under XPL (Figure 5).This assessment was confirmed by XRD analyses (see Section 4.3).The ash residues are dispersed within frequent to common sand-sized quartz grains, sand-to centimetre-sized charcoal and heated bone fragments, phytoliths as well as humified/charred plant material (Figure 5).Zones of secondary authigenic gypsum (CaSO 4 •2H 2 O) are sporadically distributed within these deposits (Figure 8).The crystals demonstrate random and parallel intergrowth and random orientations.Secondary calcite crystals (CaCO 3 ) were observed in the thin section of UMH2.The calcite was formed in sand-sized channels (Figure 8a,b) as well as on a bone fragment (Figure 10c,d).
Charcoal-rich (CRc) The CRc subfacies exhibits higher proportions of charcoal and is characterised by a random orientation and distribution (Figure 9).
Other combustion residues such as heated bones, phytoliths and fragments of fat-derived char or wood charcoal are present.In some samples (e.g., UMH1), the charcoal fragments and other charred material, sand grains and ash remains show a banded distribution pattern with strongly expressed parallel orientation of the elongated charcoal fragments (Figure 9c).Part of these layers were also identified in the field (e.g., Unit H6), while in other cases they are smaller (centimetre) scale occurrences (e.g., in sample UMH22B).CRc facies are only present in Group H deposits (Table 1).

Dung (CRd)
Subfacies CRd is characterised by the presence of fibrous coprolites (Figure 11).The excrements contain plant fibres that often appear loose and finely layered, while in other cases they feature a convolute microstructure.The plant fibres are partially impregnated with apatite (Figure 11c).In some parts, the dung is characterised by dark brown to blackish colours and shows signs of burning (e.g., plant tissues are finely fragmented, charred) (Figure 11).Spherulites are absent.The lack of spherulites was also confirmed through the phytolith analysts (I.Esteban-Alama, personal communication, 23 November 2022).The microstructure appears vermicular and the thin planar voids that pervade the dung appear to be created as a result of heating and or drying, rather than secondary microfaunal activity.Vesicles, channels and chambers with clear boundaries are also present and seem to be caused by bioturbation (i.e., microfaunal activity, root penetration).The overall structure of CRd suggests that the dung remains are of large herbivores (Brönnimann, Ismail-Meyer, et al., 2017;Brönnimann, Pümpin et al., 2017).Facies CRd was only observed in Group H deposits (Table 1).

| XRD
The XRD results are presented in

SDs
The quartz sand grains forming most of the sedimentary matrix derive from ongoing rockshelter wall attrition.This is demonstrated by granulometric analysis reported by Sifogeorgaki et al. (2020), which shows that the grain size in the deposits matches with average grain size of Natal Group sandstone rocks.This is a common site formation process at sandstone rockshelters (e.g., Karkanas & Goldberg, 2010a;Martínez et al., 2013;Miller et al., 2013Miller et al., , 2016;;de la Peña et al., 2018).The organic silt originates from the decomposition of various organic remains that were transported to the site by natural, anthropogenic and biogenic processes.Natural processes include wind-blown plant remains, anthropogenic processes include the biological material transported to the site by humans, while biogenic processes include material brought to the site by animals (e.g., for nesting purposes), as well as biogenic remains such as herbivore and microfaunal excrements (Stolt & Lindbo, 2010).
SDl.This subfacies represents primary features, remnants of original depositional fabric.The interspersed fine material lenses can be a result of i. fine matter deposition through decomposition of organic remains, ii.exposed surfaces, iii.gentle sheetwash, iv.input of fine dust and v. anthropogenic activities or a combination of the above processes.Small-scale aeolian activity occurring within the rockshelter has resulted in the creation of aggregates consisting of fine matter.The coated quartz grains that result in a chitonic coarse to fine (c/f) distribution are possibly the result of this aeolian sedimentation mechanism (e.g., Figure 5).Gentle sheetwash and illuviation of the deposits could also result in this chitonic c/f distribution as well as the fine laminaes observed, although the rockshelter is currently well-protected from running water.The water observed in the higher moisture part of the sequence is not the result of sheetwash but rather originates from the rockshelter wall (Reidsma  et al., 2021;Sifogeorgaki & Dusseldorp, 2022;Sifogeorgaki et al., 2020).The SDl subfacies are present throughout the sequence and are very strongly expressed in the uppermost micromorphological sample (UMH30).This suggests that this mechanism of sedimentation has been constantly active during the accumulation of the sedimentary sequence.The presence of these unaltered features in parts of the Group P units (e.g., UMH5, UMH28, also see Table 1) implies that post-depositional reworking was mostly modest in intensity during the deposition of the facies.The horizontal aspect of this layering indicates that the archaeological materials collected in horizontal spits do not crosscut the original depositional environment.SDl is absent in the lowermost higher moisture units (P16, P17) in which the sediments appear highly bioturbated (also see discussion SDb).SDl is found in both higher and lower find density zones.
However, more samples from ZL3 need to be examined to discern patterns related to find density.
SDa. SDa facies comprise sediments with a low primary structural development.Sda-composed deposits demonstrate variations in coarse material, fine material and void proportions, the distribution of which does not show distinct patterning.The groundmass of this facies is expected to be reworked, but unlike facies SDb that is clearly reworked through biogenic activities, Sda could be affected also by geogenic and anthropogenic processes.The granular and relatively homogeneous character of the groundmass is similar to that of other South African rockshelters such as Sibhudu Cave (Goldberg et al., 2009), Elands Bay Cave (Miller et al., 2016) and Umbeli Belli (Bader et al., 2016(Bader et al., , 2018;;Sifogeorgaki et al., 2023).The absence of distinctive occupational layers or well-defined features is common in sandstone rockshelter deposits with poor preservation conditions (Cosgrove, 1995;Lowe et al., 2018;Martínez et al., 2013;Parenti et al., 2018;Ward et al., 2006).
The bioturbation apparent in the thin sections is mostly the result of worm, insect and plant activity: The crescent-like infilling pattern of the burrows together with their millimetre to centimetre size is characteristic of mesofaunal worm activity (Kooistra & Pulleman, 2010).Other types of burrows infilled with loose sediments could be associated with antlion activity since antlion pit traps were visible in the rockshelter.The documented insect excrements are derived from micro-and mesofauna.Channels connected to root activity are present throughout the sequence and in some cases root tissue remains are still preserved.The higher moisture units P16 and P17 appear to be the most intensely bioturbated parts of the section.This is possibly due to the higher water content that enhances bioturbation activity, especially that of worms and plants (Kooistra & Pulleman, 2010;Ruiz et al., 2021;Stoops & Schaefer, 2010).

CRs
This facies includes hearths that are preserved in situ and mostly refers to the H2-H4 hearths that can be traced along the west, north and eastern profiles.The hearth layering at Umhlatuzana is typical for in situ combustion features (Aldeias et al., 2016;Ferro-Vázquez et al., 2022;Goldberg et al., 2017;Mentzer, 2014;Puech et al., 2021;Schiegl et al., 1996;Stahlschmidt et al., 2020).The lowermost brown layer corresponds to rubified sediment, the middle layer to heated organic material and the uppermost layer to ash remains (Figures 1e and 11a).Bioturbation features such as burrows are observed but overall, the intact nature of these features suggests the combustion activity was performed in situ with limited subsequent disturbance.Fragmented charcoal and bone fragments are indicative of trampling.The H2-H4 hearths represent either a single event or several short-term combustion activities (Karkanas, 2021;Mentzer, 2014).Through micromorphological analysis, it became apparent that remnants of in situ hearths also exist within Units H5 and H8 (Table 1).Those were not identified during excavation and are heavily bioturbated.
CR ash.When the ash layers are not part of the CR, they are most likely redeposited, probably the result of anthropogenic activities such as dumping, trampling, raking out or hearth clearance (e.g., Bentsen, 2014;Karkanas et al., 2015;Marcazzan et al., 2022;Mentzer, 2017;Miller et al., 2010;Puech et al., 2021;Théry-Parisot et al., 2010).This is the case for Units H9 and H7.The former is inferred to be the result of continuous and extensive burning events as it can be traced in the western, northern and eastern profiles and has a thickness that sometimes reaches 15 cm.
CR charcoal.The CRc layers studied here show signs of redeposition because of their chaotic structure.This redeposition can be caused by natural or anthropogenic processes, such as raking out (Meignen T A B L E 3 XRD results., 2007;Miller et al., 2010;Weiner et al., 1998).Since Umhlatuzana appears largely protected from natural processes, we propose that the reworking is mainly a result of human activities.The CRc layers observed were mostly part of Unit H5, which was reported to have a much darker colour (Sifogeorgaki et al., 2020), probably due to the high content of charcoal remains.
CR dung.CRd sediments comprise burnt herbivore dung remains and seem to be part of the in situ hearth structures (CR).In particular, CRd is part of the middle hearth layer that consists of burned organic material.The CRd subfacies corresponds to Unit H4 of the hearth structure documented during fieldwork.CRd is also present in the newly identified hearths within Units H5 and H8.

Higher moisture post-depositional alterations
Vashegyite is created in highly acidic environments and is associated with the presence of guano deposits (Audra et al., 2019;Onac et al., 2006).At Umhlatuzana, the high acidity of the deposits could be attributed to guano, but could also be related to animal excrement (e.g., from rock hyrax: Procavia capensis) (Chase et al., 2012;Prinsloo, 2007).Traces of vashegyite at Umhlatuzana are identified throughout the higher moisture units: in lowermost Units P16-P17 (micromorphology analysis) and also in Units P9 and P14 (XRD analysis).Wet sediment conditions are required for vashegyite formation (Audra et al., 2019;Giurgiu et al., 2013;Onac et al., 2006).
This suggests that the higher moisture present in the lower southern sediments (observed during the excavation) is probably a long-term feature, rather than a recent change in the depositional environment due to, for example, excavations or the N2 highway construction.
Vashegyite nodular forms were created by long-term alteration.This is another indication that vashegyite replacement started very soon after the material was deposited.

| Discontinuity in sedimentation
The main source of geogenic sediments is the rockshelter bedrock, either through granular disintegration of the rockshelter or through the situ decay of tabular sandstone clasts.The processes behind bedrock weathering are likely salt-induced spalling, a very common process in siliceous rockshelters (e.g., Goldberg et al., 2009;Karkanas & Goldberg, 2010a;Miller et al., 2013Miller et al., , 2016)).Any visible stratigraphic changes are therefore primarily a result of anthropogenic input.Still, the development of primary features that are remnants of the original depositional fabric (facies SDl) in the Group P deposits is poor.At the same time, the samples exhibit a high degree of bioturbation, which results in this unconsolidated and homogeneous appearance of the Pleistocene sequence.The fact that no clear stratigraphic boundaries were observed macroscopically or microscopically supports a consistent type of geogenic sedimentation at the site.Major geogenic events such as extensive rockfalls are not expected but short-term, minor changes and low-energy erosional episodes cannot be excluded.
We do not observe clay-rich compacted surfaces as evidence of long-term exposure, abrupt changes in compaction or other data that would indicate the presence of long-term depositional hiatuses within the Group P sequence.However, the boundary between the lowermost Group H unit H10 and the uppermost Group P unit P1 appears diffuse when observed on the micromorphological blocks or under the thin sections.In addition, macroscopic, microscopic and geochemical differences between Unit H10 and Unit P1 are profound: Macroscopically, Unit H10 is darker in colour than Unit P1 (for a complete sedimentary description see Sifogeorgaki et al., 2020).
Microscopically, Unit H9 contains a higher concentration of bone fragments and fine organic matter when compared to Unit P1 (Table 1).
Geochemically, the sediments of Group H demonstrate an abrupt increase at this boundary in elements that connect to combustion activities such as CaO, MgO and Sr (Reidsma et al., 2021).
Radiocarbon results by Kaplan (1990) indicate a potential chronological hiatus of 4220 radiocarbon years.Nevertheless, Kaplan's results were based on charcoal samples that could have been transported vertically by post-depositional processes such as bioturbation.The absence of secure chronological data makes it difficult to determine whether there is a gap in deposition between Group P and Group H.In view of the available data, we propose that a hiatus combined with more intensive burning activities in the Holocene may explain the difference between Group P and Group H.
The sharp stratigraphic boundaries within the Group H deposits (e.g., the upper and lower boundary of the ash layer, Unit H9) indicate that there is a possibility of sediment removal and thus the existence of one or more hiatuses.This is also supported by a more than 6000year gap between Pta-4226 (Unit P1) and Pta-4307 (Unit H10) (Kaplan, 1990).The sediment removal is unlikely to be a natural occurrence (e.g., erosion) because the rockshelter is well-protected.It is, however, possible that it is a result of hearth construction activities such as levelling and raking out by mid-Holocene people.Such activities have been reported in other South African sites such as Grassridge rockshelter (Ames et al., 2020), Sibhudu Cave (Goldberg et al., 2009) and Pinnacle Point (Karkanas et al., 2015).
Unlike in other South African rockshelters (Miller et al., 2016;de la Peña et al., 2018), we did not observe significant sediment input of an aeolian origin (wind-blown silts and sands).This is supported by the micromorphological analysis as well as granulometry results indicating that the grain size distribution is consistent throughout the sequence and the grain size is similar to the grains of the bedrock quartz arenite (Sifogeorgaki et al., 2020).There is, however, evidence of aeolian activity during the deposition of the sediments.The rounded sand-sized aggregates that contain silt-sized residues of charcoal and ash were probably created due to rolling within the site as a result of wind action or trampling under wet conditions (Figure 6e) (Dowey et al., 2017;Rentzel et al., 2017;Wilson, 1992).
Finally, there were no indications of well-developed soil horizons, the absence of which can be the result of sediment mixing, intensive human presence, the absence of vegetation on the rockshelter floor and/or low rates of pedogenesis.Other buried surfaces or slumped deposits were not observed.

| Preservation conditions
Umhlatuzana rockshelter is characterised by generally poor preservation of organic material like bone and charcoal.Geochemical investigations (Reidsma et al., 2021;Sifogeorgaki et al., 2020) have demonstrated variable pH values throughout the sequence.The Holocene deposits have pH values as high as 9.1, the higher moisture Group P units have pH values as low as 4.6, while the lower moisture Group P units have pH values close to neutral.This pH variability is correlated with bone preservation, with more bone remains being found in sediments with a higher pH.
We observe that the intra-site taphonomic variations are connected to post-depositional processes.The upper stratigraphic units (Group H) show better charcoal and bone preservation conditions than the underlying Group P deposits.This can be the result of progressive decomposition with depth and/or the increase of sediment alkalinity due to more intensive combustion activities (Reidsma et al., 2021).The progressive decomposition is apparent in the micromorphological study with the uppermost layers (H1) demonstrating the preservation of plant tissue, the lower Group H units (e.g., H8-H10) yielding more frequent bone and charcoal fragments and the Group P units containing very few organic remains (Table 1).It is likely that the combustion materials acted as a chemical buffer to keep the pH values relatively high and prevented the dissolution of bone and other organic material.As an indication, ashes deriving from fresh wood can have a pH of up to 13 (Etiegni & Champbell, 1991) and wood charcoal was observed in the micromorphological thin sections.An increase in pH values near the layers where ash deposition occurred has been also reported in other siliceous rockshelters (e.g., Miller et al., 2013;de Sousa et al., 2017).
The percolating groundwater present in the higher moisture units results in local changes of the sediment chemistry (and potentially structure) and as a result differing preservation conditions across the Group P deposits.This has been reported in other sites of a similar geoarchaeological setting (e.g., Karkanas et al., 2000;McAdams et al., 2019;Miller et al., 2016;Morley et al., 2017;Stewart et al., 2012).For example, at Elands Bay Cave (Western Cape Province, South Africa) bones are present in the whole sequence with the exception of a wetting front, in which bones are almost completely absent (Miller et al., 2016).In extreme cases, this can also result in altered stratigraphic layering due to the postdepositional reduction of sediment volumes within parts of an archaeological sequence (Karkanas et al., 2000).At Umhlatuzana, we argue that the higher presence of bone in the drier parts of the sequence (see Table 1; Group P, square L2a) is not linked to anthropogenic and biogenic activities but, rather, to post-depositional chemical diagenesis.

| Post-depositional mineral formations
Chemical transformation processes are generally the main factor accounting for the destruction of archaeological material such as bone and CRs.Understanding secondary mineral formation can shed further light on past sedimentary conditions like water availability, oxidation conditions and sediment chemistry (Karkanas & Goldberg, 2010a;Poch et al., 2010;Reidsma et al., 2021;Weiner, 2010).The authigenic mineral formation is direct evidence of the geochemical processes at play.At Umhlatuzana, chemical diagenesis led to the formation of two discreet phosphatic mineral types.In the higher moisture sediments of Group P, a phosphatic mineral rich in aluminium is formed, while in Group H deposits apatite has replaced the vast majority of bone and ash remains.Iron impregnation features were also encountered and may be the result of biomineralization or groundwater action (Bazylinski & Moskowitz, 1997;Ghiorse & Ehrlich, 1992;Lindbo et al., 2010).
In acidic conditions, phosphate mineral phases such as crandallite (Ca), montogomeryte (Al) and taranakite (K, Na, Al, Fe) can be formed (Berna et al., 2004;Karkanas, 2010Karkanas, , 2017;;Karkanas & Goldberg, 2010a;Miller et al., 2016).The exact phosphatic mineral formed depends on both the pH and the ions present in the solution (Karkanas, 2010).Within the higher moisture units (Group P), vashegyite has completely replaced bone remains and also impregnates the deposits in the form of microcrystalline nodules and crusts.Vashegyite is an extremely rare phosphatic mineral that is created in highly acidic and wet environments, within guano-rich deposits (Audra et al., 2019;Giurgiu et al., 2013;Onac et al., 2006).At Umhlatuzana, the phosphate can originate from guano and other excrement (e.g., deriving from rock hyrax middens; Chase et al., 2012), but bone dissolution could have contributed as well during the later stages of phosphate diagenesis (Karkanas & Goldberg, 2010b).
The fact that the original bone structure is often maintained and that vashegyite nodules were formed after the initial bone replacement indicates that the dissolution of the bone and its replacement by vashegyite happened not too long after burial.These observations point to the fact that this part of the sequence has always been characterised by higher moisture content since we did not observe features indicating leaching of the deposits (e.g., cappings, clay translocation), which indicates a low-energy environment.
Evidence for vashegyite crusts is also visible in portable X-ray fluorescence (p-XRF) analysis of Umhlatuzana lithics (Sifogeorgaki et al., 2023).Most of the lithics excavated in the high moisture sediments demonstrate a cemented crust of phosphatic minerals.
The p-XRF analysis of the cleaned lithics yielded higher phosphate and aluminium readings in the assemblage related to the higher moisture sediments.This shows how diagenetic processes at the site also affect the archaeological assemblage: Elemental signals of aluminium content have been used to determine and trace raw materials (e.g., Fiers et al., 2019;Nash et al., 2013;Pengilley, 2021).
Thus, it is possible that the high aluminium content would be interpreted incorrectly as a change in raw material source.
In the Group H deposits, material rich in Ca-like ash and bone fragments has been exclusively replaced by apatite, while others appear in the initial stages of recrystallisation.Apatite is a phosphate mineral that forms in moderate alkaline conditions.Moreover, secondary gypsum was observed in Group H. Gypsum was likely formed due to calcium from the ash and sulphur deriving from the dung-rich layers.The existence of SIFOGEORGAKI ET AL.
| 231 gypsum also points to a low-moisture regime of the sediments (Herries, 2006;Poch et al., 2010).Within the Southern African geoarchaeological context, the presence of gypsum points to a variety of depositional environments.At sites close to the shoreline, it has been suggested that part of the sulphur that led to gypsum formation has a marine origin (e.g., Miller et al., 2016).At Pinnacle Point 13B, secondary gypsum observed in the sequence was attributed to the decay of organic material like guano (Karkanas & Goldberg, 2010a).
The iron-bearing materials are assumed to be formed before deposition due to their pronounced differences in texture compared to the surrounding sediments.Furthermore, they appear to have a partially geogenic (e.g., inclusions in the bedrock sandstone) and partially anthropogenic origin (demonstrated by fragments of ochre in thin sections and excavated larger pieces).

| Stratigraphic integrity of Umhlatuzana
Macroscopic observations and micromorphological analysis indicate small-scale sediment mixing due to bioturbation.Some degree of mixing is thus expected but, in this study, we focus on determining the nature and degree of mixing in the excavated area of the site.One of the most prominent results of this study is the observation of primary depositional structures (SDl facies) throughout the seemingly homogeneous lowermoisture Group P deposits.This implies that if bioturbation was not uniformly intensive, otherwise SDl remnants of primary features would not be preserved.Additionally, if extensive bioturbation was at play, the lithics observed within the thin sections would not have the tendency to be horizontally oriented (see subsection SDl in Section 4.1.1).We, therefore, concluded that bioturbation processes caused microscale obliteration of the stratigraphy, which results in millimetre-to, at cases, centimetre-scale movement of the sedimentary matrix.Larger sediment inclusions such as pieces of rock could be re-oriented or moved in a millimetre scale only.
Fabric features indicating large-scale sediment movement-like slickensides and evidence of slumping and liquefaction (e.g., Blokhuis et al., 1990;Jeong & Cheong, 2005;Karkanas & Goldberg, 2013;Morley et al., 2017) were not observed.The Group P units were mainly differentiated by their distinctive and fluctuating find densities and did not show evidence of vertical or lateral dislocation.This observation supports previous findings that the Umhlatuzana sequence did not undergo large-scale sediment movement (Reidsma et al., 2021;Sifogeorgaki et al., 2020).
We thus suggest that artefacts and clasts are situated in close proximity to their original position.Overall, the lithic assemblage integrity is intact and its technological analysis will provide secure archaeological interpretations.

| Umhlatuzana rockshelter site use
Evidence of fire use is present throughout the Umhlatuzana sequence and includes components of charcoal, ash remnants, charred material, heated bone, burned dung and heated rocks.
Micromorphological analyses confirm the macroscopic observations that hearth features contain stratified combustion material (also see Sifogeorgaki et al., 2020).Micromorphology also detects specific combustion materials used for hearth construction that were not macroscopically identified.One such material is the presence of burnt dung that we can assume to be used as some type of fuel.
Evidence for burning activities is prominent in the Holocene layers and includes individual hearths that appear intact and in situ (CR) as well as layers rich in redeposited combustion material such as ash (CRa) and charcoal (CRc).The latter are most likely the result of material dumping, trampling and/or raking out (Marcazzan et al., 2022;Miller et al., 2010;Stahlschmidt et al., 2020).Although combustion-derived residues are more dominant in Group H deposits, they are also present in Group P deposits.The micromorphological analysis facilitated the identification of microscopic, burned materials within the Pleistocene.One outstanding question is whether fire-related activities had the same intensity within the Holocene and Pleistocene but due to post-depositional processes most of the Group P CRs are not preserved.We observe a lateral patterning in charcoal preservation of Group P deposits with the samples connected to lower moisture units (UMH27 and UMH28) demonstrating lower charcoal counts.This suggests that preservation conditions vary, enhancing the argument that CRs within the Group P deposits are unequally preserved.
Based on archaeological evidence, the southern African MSA is characterised by intense fire-related activities (Bentsen, 2014;Esteban et al., 2018;Haaland et al., 2017;Karkanas et al., 2015;Morrissey, 2022).When comparing to nearby sites, we observe that, when overall preservation conditions are good (e.g., Sibhudu Cave: Goldberg et al., 2009, Border Cave: Backwell et al., 2018, 2022), the MSA layers are characterised by high intensity of burning activities while, when the preservation is bad (e.g., Umbeli Belli: Bader et al., 2018;Sifogeorgaki et al., 2023), indications for burning activities are not preserved.We thus suggest that the less prominent combustion features at the Group P deposits are a result of poor preservation conditions.
Previously, researchers evaluated the higher and lower find density layers as the result of distinct occupation intensity pulses during the Pleistocene (Reidsma et al., 2021;Sifogeorgaki et al., 2020).
The alternative explanation suggests that the higher and lower find density layers can be explained by inconsistent geogenic sedimentation rates (Reidsma et al., 2021;Sifogeorgaki et al., 2020).Micromorphological analysis does not provide additional evidence to support one of the two hypotheses.The main limitation is the lack of micromorphological samples from ZL4. Future OSL and find density analyses may shed additional light on this question.

| CONCLUSION
The current study sought to clarify the depositional and postdepositional history of Umhlatuzana rockshelter, as well as answer long-standing questions relating to the site's stratigraphic integrity and anthropogenic inputs.This was achieved by using multiproxy geoarchaeological data (micromorphology, SEM-EDS, XRD).The application of micromorphology allowed for a link between the macro-and microscale within the deposits.Micromorphological analysis was successfully integrated with other geoarchaeological data.This allowed us to identify the stratigraphic sequence-forming processes and to infer the geogenic, biogenic and anthropogenic factors that have driven accumulation and post-depositional changes at the Umhlatuzana rockshelter.Coming back to the initial research questions, the following conclusions can be drawn: 1. We show that primary, unaltered depositional fabrics are present in both the Group P (facies SDl) and Group H deposits (layers dominated by material related to combustion activities-facies CR).Bioturbation processes caused millimetre-to, at cases, centimetre-scale movement of the sedimentary matrix.Larger pieces of rock could be reoriented or moved on a millimetre scale only, meaning that artefacts and clasts are located in close proximity to their original position.The overall lithic assemblage integrity is high and its technological analysis will provide secure archaeological interpretations.
2. We observe that human activities greatly contributed to the formation of the sequence through the accumulation of anthropogenic inputs (mainly lithics and CRs).The fact that the Pleistocene deposits contain microscopic CRs indicates that burning activities took place at Umhlatuzana rockshelter throughout its occupation history (MSA, LSA, Iron Age).We conclude that due to poor preservation conditions within the Pleistocene deposits, the CRs are not as prominent as they are in the Holocene deposits.Nevertheless, we suggest that burning activities during the Pleistocene may have been as profound and intense in the Pleistocene as they were during the Holocene.To assess this statement, we need additional data that could be derived from further excavating the site.
3. There are no indications of major geogenic depositional hiatuses within the sequence as observed from the micromorphological analysis.However, based on the chronological framework (also see Figure 2) and other geoarchaeological data, we conclude that there might be a gap in deposition.Combined with more intensive burning activities during the Holocene, this may explain the macroscopical and geochemical differences between the Group P and H deposits.To distinguish whether the higher find density zones in the Pleistocene deposits are a result of higher occupation intensity or lower geogenic sedimentation rate, additional dating and archaeological analyses need to be conducted.
All in all, stratigraphic assessment is a major factor in the interpretation of the material culture of a site.The original assessment is rarely put to the test.At Umhlatuzana, the stratigraphic context turned out to be much more intact than previously inferred.
The results of this study are part of a larger research effort that includes chemical analyses, lithic studies, paleoenvironmental proxies (e.g., phytoliths, lipids) and dating.The ultimate goal is to integrate all of the above to achieve an accurate interpretation of past human behaviour at Umhlatuzana rockshelter.
Umhlatuzana rockshelter was discovered by Dr Rodney Maud in 1982 during the construction of the N3 Johannesburg-Durban F I G U R E 1 (a) Elevation map of Southern Africa indicating the location of Umhlatuzana rockshelter (UMH) and other sites mentioned in the text.Map published in Reidsma et al. (2021).(b) Umhlatuzana site map illustrating the excavation squares of the 1985 (K2-K4, J2-J4) and 2018-2019 (L3a, L2b, L2a) fieldwork campaigns.Figure published in Sifogeorgaki et al. (2020).(c) West profile of squares L3a, L2b and L2a.
4.1.2| CRsThe CR facies refers to combustion layers that have characteristics of being formed as part of an in situ combustion feature.The CRs include typical layering of a lowermost rubified layer, an intermediate layer rich in burned material (bone, charcoal, dung, etc.) and an uppermost ash layer.CR also commonly includes horizontally oriented bone and charcoal fragments (Figure9c).The bone remains appear fragmented and are orange-red, indicative of heating (Figure10) (e.g.,Clark & Ligouis, 2010;Hanson & Cain, 2007).The CR facies consists of subfacies CRa (uppermost ash-rich layer), CRc (intermediate layer rich in burned material) and occasionally CRd (intermediate layer and or lowermost rubified layer).For instance, at the thin section of UMH1 (associated with hearths H2-H4), the CR consists of a lowermost layer of rubified sediment, an intermediate CRd layer and an uppermost CRa layer.The contacts between the layers that make up the CR facies are sharp both in the macro-and microscale.The ash and dung remains present appear to be replaced by isotropic phosphate minerals (also see Section 'Carcoalrich CRc'; Figures 7 and

F
I G U R E 4 Remnants of layered sand facies, SDl.The quartz sand grains (dominant), silt grains (few) and fine organic matter (common) demonstrate distinctive banded linear distribution patterns.The coarse to fine (c/f) distribution appears chitonic and single-spaced fine to equal enaulic.(a) Photomicrograph of SDl subfacies in the Group H deposits.Banded distribution of subangular quartz sand grains and fine organic matter (arrow) horizontally oriented.The presence of partially accommodated vesicle (V) with rough walls (thin section of UMH30, PPL).(b) Same as (a) in XPL.(c) Remnants of subfacies SDl in the Group P deposits.The quartz grains and fine matter demonstrate a weakly developed banded distribution of a horizontal orientation (arrow).Siltstone (Ss) and silcrete (Sc) stone fragments also show a horizontal orientation (thin section of UMH3B, PPL).(d) Same as (c) in XPL.PPL, plane-polarised light; SDl, sand deposits layered; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.
are characterised by a phosphatic mineral, which is present in the form of millimetre-to centimetre-sized nodules, aggregates and crusts.Optical properties include a light beige colour and undifferentiated b-fabric (Figure 12).This phosphate mineral was identified as vashegyite (Al 11 (PO 4 ) 9 (OH) 6 •38H 2 O) using SEM-EDS and XRD analyses (see Sections 4.2 and 4.3).Occasionally, bone material seems to have been replaced by vashegyite.Composite nodules are formed with vashegyite coatings surrounding rock or altered bone cores, often containing quartz grain impurities.Vashegyite also appears as external and internal hypocoating of inclusions and voids.Vashegyite nodular forms were created in multiple alteration episodes.One example is the nodule in Figure 12a,b.The bone fragment was replaced by vashegyite and subsequent processes created vashegyite-rich coating.No other authigenic mineral was observed in association with the HMA occurrences.The HMA facies are exclusively concentrated within the higher moisture units (

F
I G U R E 5 Altered sand deposits, SDa.Overall, the subfacies has a separated granular chitonic microstructure.(a) Photomicrograph of SDa subfacies in the Group P deposits.Chitonic to single-spaced fine enaulic c/f distribution of subangular quartz sand grains (Q) and fine organic matter.The presence of rounded coarse sand-sized aggregates (arrow).In terms of component abundance, the coarse fraction is frequent, the fine fraction is common and the voids are few (thin section of UMH4A, PPL).(b) Same as (a) in XPL.Notice the strong interference colours of the clay coatings and aggregates.(c) Fabric of subfacies SDa in the Group P deposits.The smaller fraction is coating the quartz grains (Q) resulting in a chitonic c/f distribution.This sample has a higher void abundance than (a) with dominant coarse fraction, little fine fraction and frequent voids.Notice loose organic excrements (arrow) that make up a high proportion of the fine fraction (thin section of UMH12B, PPL).(d) Same as (c) in XPL.c/f, coarse to fine; PPL, plane-polarised light; SDa, sand deposits altered; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.F I G U R E 6 Bioturbation SDb.(a) Cross-section of burrow (BR) that forms a passage feature with a distinct circular pattern.Dolerite (D) fragment exhibiting hypocoating (thin section scan of UMH15).(b) Longitudinal sections of channels caused by microfaunal burrowing.The burrows are infilled with loose sediment consisting of quartz and clay.They often have a crescent pattern.The upper channel demonstrates external crescent hypocoating.Orange-coloured bone fragments (B) are also present as are hornfels (H) and sandstone (S) rock fragments (thin section scan of UMH28).(c) Highly bioturbated deposits.This thin section scan corresponds to high moisture unit P16 in which most of the groundmass seems affected by bioturbation features such as burrows (BR).Often the loose infilling sediments within the burrows demonstrate a crescent pattern (arrow).Secondary vashegyite (V) has been present throughout the unit.Quartz (Q) and sandstone (S) fragments are common (thin section scan of UMH35A).(d) Loosely packed concentration of organic ellipsoid excrements (E) within the groundmass.Quartz grains (Q) are coated by amorphous organic material demonstrating chitonic c/f distribution (thin section of UMH4B, PPL).(e) Same as (d) in XPL.Notice strong interference colours of clay aggregates.(f) Loose organic ellipsoid excrements (E) making up most of the groundmass.Hornfels (H) fragment demonstrates typic external hypocoating (C).The presence of quartz grains (Q) (thin section of UMH35A, PPL).(g) Same as (f) in XPL.Strong interference colours of clay coating and excrements.c/f, coarse to fine; PPL, plane-polarised light; SDb, sand deposits bioturbated; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.

F
I G U R E 7 Subfacies CRa.CRa is characterised by a weakly separated angular blocky microstructure, accommodating planes and the presence of intrapedal channels.(a) Thin section scan of sample UMH23.Transformed ash remnants (A) with compound layering are visible in the upper and middle parts of the sample.Partially altered bone fragments (B) and charcoal (CH) fragments are visible throughout the sample.Bioturbation features (BF) include channels infilled with a groundmass characterised by a double-spaced enaulic microstructure.(b) Photomicrograph of CRa subfacies.Quartz (Q) grains and ash remnants (A) demonstrating a double-spaced porphyric c/f distribution.The fine fraction contains microcharcoal fragments (CH) that are indicated with arrows.Bioturbation features (BFs) in the form of a channel infilled with loose sediment are visible in the upper part of the picture (thin section of UMH23, PPL).(c) Same as (b) in XPL.Notice the speckled b-fabric of the fine mass due to clay and mica minerals.(d) Facies CRa.Ash (A) and quartz grains (Q) demonstrate a double-spaced to single-spaced porphyric c/f distribution (thin section of UMH2, PPL).(e) Facies SD.Rounded sand-sized aggregate (A) containing silt-sized residues of charcoal (arrow) and ash remains.Sand-sized quartz gains (Q) exhibiting external hypocoating (thin section of UMH3A, PPL).The presence of granular excrements possibly composed of ash-rich sediment.Notice the similarity in the groundmass of ash-rich facies CRa and SD.c/f, coarse to fine; CRa, combustion remain ash; PPL, plane-polarised light; SD, sand deposits; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.F I G U R E 8 Secondary mineral formation.(a) Secondary calcite crystals (CCs) formed in a channel with a hypocoating of humified/ oxidised organic material (arrow).The crystals have a columnar shape and form spear-and rosette-like aggregates (arrow).Groundmass of subangular quartz grains in brown fine matter with a close porphyric c/f distribution.Parts of the fine material have been impregnated with iron (thin section of UMH2, PPL).(b) Same as (a) in XPL.Notice the high birefringence of calcite and the speckled b-fabric of the fine mass that corresponds to clay and or mica particles.(c) Subfacies CRa with groundmass of subangular quartz grains and brown matrix that includes ash remnants (A) in a chitonic to single-spaced enaulic distribution.Zones of secondary gypsum (G, arrow) are sporadically distributed within the compound packing voids (thin section of UMH2, PPL).(d) Same as (c) in PPL.Notice the isotropic character of secondarily formed apatite that replaced ash and other organic remains of the fine fraction (A).(e) Detail of gypsum crystals (G, arrow) and quartz grains (Q).The elongated crystals demonstrate random and parallel intergrowth, random orientation, a columnar or lensoidal shape and an average size of 150 μm (thin section of UMH2, PPL).(f) Same as (e) in XPL.c/f, coarse to fine; PPL, plane-polarised light; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.

F
I G U R E 9 Charcoal rich, CRc.The CRc subfacies usually demonstrates a single-spaced porphyric c/f distribution with the ash remains comprising the fine fraction.(a) Thin section scan of sample UMH22B illustrating facies CRc.Charcoal (CH) appears fragmented with random breakage patterns.Sand-and silt-sized charcoal fragments are present in the micromass.Bioturbation features (BFs) are visible throughout the thin section and include a channel with weakly visible crescent infilling (arrow).Root tissue remains (R) can be recognised in cross-section.Ash residues (A) are visible and constitute part of the micromass, especially in the upper part of the thin section.(b) This charcoal fragment shows elongated lenses of pores.It is possible that the elongated shape is a result of the thin section being cut at an angle (thin section of UMH13, PPL).(c) Horizontally oriented charcoal (CH) fragments that suggest the in situ formation of the CR facies.Single-spaced porphyric c/f distribution, ash remnants (A) comprise a large part of the groundmass.Evident bioturbation features that potentially are the reason the charcoal appears fragmented (thin section of UMH1, PPL).(d) The charcoal (CH) fragment belongs to a diffused-porous wood species with vessels scattered evenly.The micromass demonstrates a single-spaced porphyric distribution.An elongated bone fragment (B) is present and is characterised by orange-brown colours, which are indicative of heating (thin section of UMH23, PPL).(e) Same as (d) in XPL.c/f, coarse to fine; CRc, combustion remains charcoal; PPL, plane-polarised light; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.

F
I G U R E 10 Bone fragments.(a) Dark orange bone fragment exhibiting fragmentation with an elongated angular blocky structure.The colour and fragmentation are indicative of heating.Sediments exhibit chitonic c/f distribution (thin section of UMH12B, PPL).(b) Same as (a) in XPL.(c) Spongy bone fragment with a pale yellow colour (thin section of UMH2, PPL).(d) Same as (c) in XPL.Notice the randomly distributed secondary needle-shaped calcite crystals that exhibit high interference colours forming on the bone.c/f, coarse to fine; PPL, plane-polarised light; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.

F
I G U R E 11 (a) Thin section scan of sample UMH22A.Ash remnants (A) belonging to facies CRa are present in the upper part of the section.Dung-rich (D) CRd deposits are in the middle part of the section.The dung layer has a brown to dark brown colour.The lower part of the thin section consists of CRa deposits characterised by planar voids.The planar fractions could be the result of desiccation, trampling, or other anthropogenic activities.Bone (B) and charcoal (CH) are present throughout the sample.Bioturbation features (BF) mostly caused by burrowing are visible throughout the sample.(b) Microphotograph of the CRd facies.CRd is composed of coprolites (dominant), quartz sand grains (few) and an occasional presence of charcoal and bone fragments (very few).The plant fibrous material within the dung remains is visible, appears black in colour and has a highly separated angular platy microstructure.Quartz grains are embedded within the dung remains (thin section of UMH22A, PPL).(c) Same as (b) in XPL.Notice the speckled b-fabric of the dung due to clay and mica minerals.(d) Orange to dark brown dung remains with loosely layered plant fibres.The thin planar voids seem to be the result of heating or drying.The channel (ch) is possibly related to microfaunal activity.(e) Dung remains that do not show signals of heating (thin section of UMH30, PPL).CRa, combustion remains ash; PPL, plane-polarised light; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.F I G U R E 12 (a) Micrograph of a vashegyite (V) nodule within a groundmass dominated by quartz grains in a double-spaced fine enaulic chitonic c/f distribution.The vashegyite core (cr) is a bone fragment replaced by vashegyite that shows remnants of the spongy structure (cancellous) of the parent bone.The core has external hypocoating (ct) that consists of a vashegyite fine fraction and quartz medium sand grains (thin section of UMH35B, PPL).(b) Same as (a) in XPL.Note the low birefringence and isotropic character of the vashegyite.(c) Detail of vashegyite (V), core (cr) and coating (ct) (thin section of UMH35B, PPL).(d) Same as (c) in XPL.c/f, coarse to fine; PPL, plane-polarised light; UMH, Umhlatuzana rockshelter; XPL, cross-polarised light.