Tufas indicate prolonged periods of water availability linked to human occupation in the southern Kalahari

Detailed, well-dated palaeoclimate and archaeological records are critical for understanding the impact of environmental change on human evolution. Ga-Mohana Hill, in the southern Kalahari, South Africa, preserves a Pleistocene archaeological sequence. Relict tufas at the site are evidence of past flowing streams, waterfalls, and shallow pools. Here, we use laser ablation screening to target material suitable for uranium-thorium dating. We obtained 33 ages covering the last 110 thousand years (ka) and identify five tufa formation episodes at 114–100 ka, 73–48 ka, 44–32 ka, 15–6 ka, and ~3 ka. Three tufa episodes are coincident with the archaeological units at Ga-Mohana Hill dating to ~105 ka, ~31 ka, and ~15 ka. Based on our data and the coincidence of dated layers from other local records, we argue that in the southern Kalahari, from ~240 ka to ~71 ka wet phases and human occupation are coupled, but by ~20 ka during the Last Glacial Maximum (LGM), they are decoupled.

This statement is required for submission and will appear in the published article if the submission is accepted. Please make sure it is accurate.

Unfunded studies
Enter: The author(s) received no specific funding for this work.  South Africa, preserves a Pleistocene archaeological sequence. Relict tufas at the site are evidence of 29 past flowing streams, waterfalls, and shallow pools. Here, we report an extensive dating programme 30 of the tufas. Using laser ablation screening to target material suitable for uranium-thorium dating, we 31 obtained 33 ages covering the last 110 thousand years (ka). We identify four tufa formation episodes 32 are challenging materials for dating due to detrital contamination and generally low uranium 75 concentrations [31,32] and so samples were pre-screened using laser ablation inductively coupled 76 plasma mass spectrometry (LA-ICP-MS) to target optimal zones for study. This method has been 77 used previously for dating speleothems [33], but to the best of our knowledge, this is the first time it 78 has been applied to tufas. We obtained 33 U-Th age estimates and identify four wet periods, 79 providing a record of localized climate change linked to a dated record of human occupation.  Ga-Mohana Hill has spiritual significance for the local communities, with visits to the shelter 90 deliberate and rare [34]. Out of respect for this and as part of our on-going engagement with these 91 communities, we adopted a low-impact sampling approach, with targeted samples carefully chosen 92 after extensive survey of the 6 km area around the shelter. During this pedestrian survey, the field 93 occurrences, positions and types of tufa were identified and mapped using a roaming Geographic 94 Positioning System. A total of twenty-nine tufa hand samples were collected from the ~ 1 km 2 Ga-95 Mohana hillside sampling all five tufa morphologies recognised. Eighteen hand samples were 96 collected using a geological hammer, mallet and chisel, marking the way-up on each sample with an 97 arrow using permanent marker. Material sampled from the outer layer of cascade tufas returned to the host rock dolomite, in order to try and constrain the onset of preserved tufa formation. We 100 used a modified Makita cordless hand drill fitted with Pomeroy Model SW-3 Miniature Water 101 Swivel and a custom made Pomerory 1.5" ID diamond-tipped core barrel. A total of eleven small 102 cores were collected, 8 cm in length on average, from in-situ mound tufas, and both in-situ and ex-103 situ cascade tufas (S1 Table, S1-S5 Figs). The cores were set in epoxy resin and then halved 104 lengthways with a diamond rock saw and polished. Thin sections were made from a sub-set of 105 fourteen samples, representative of all the morphology types, for characterisation using a Zeiss 106 AXIO polarising light microscope (S1 Table,

110
The aphanitic micrite layers free from detritus and inclusions, identified in thin section, were 111 primary targets for U-Th dating. However, these layers tend to be fine, undulating and laterally 112 variable, and so while visual evaluation of the tufas is an important first step in identifying suitable 113 material to target for U-Th dating, it is not sufficient considering the complexity of the tufas on a 114 microscale. We employed an additional pre-screening step, using laser ablation inductively coupled 115 plasma mass spectrometry (LA-ICP-MS), to measure and image the U and Th concentrations and 116 distributions along transects within the tufa samples. This allowed us to target layers with 117 sufficiently high levels of 238 U, and low levels of 232 Th, i.e. detrital thorium, as these are the best 118 targets for producing reliable age data [35]. 119 120 Tufa U and Th concentrations and distributions were collected for 16 samples using laser-121 ablation with an Applied Spectra RESOlution SE 193nm ArFexcimer laser-ablation system coupled 122 to an Agilent 7700x Quadrupole ICP-MS at the University of Melbourne, following the protocols 123 outlined in Woodhead et al.[36]. High-resolution images (3200 dpi) of the samples were captured 124 using a flat-bed scanner, used to reference the co-ordinate system of the laser cell using GeoStar 125 software (Norris Software). Between 6 and 12 parallel lines per sample, set 62m apart, were chosen 126 perpendicular to the growth layers. Pre-ablation was performed twice using a 60m spot size and 127 stage translation speed of 150m/s. accompanied by little to no 232 Th were targeted for U-Th sampling. These samples were then 137 chemically processed for U-Th dating following protocols described in Hellstrom[35,40].  [40,41]. Isotope-ratio 149 measurements for 230 Th/ 238 U and 234 U/ 238 U were calculated using an internally standardised parallel 150 ion-counter procedure and calibrated against the secular equilibrium standard, HU-1. Reproducibility 151 was monitored using a second in-house standard (YB-1). An a priori estimate of 1.5 ± 1.5 for the 152 initial 230 Th/ 232 Th was applied to all the samples in order to correct for the inherent detrital 153 component [35]. With this initial value and its uncertainty, corrected ages for all samples were

219
Of 21 sampled tufas, 16 were subjected to the LA-ICP-MS pre-screening process (S1 Table, (Table 1, S2 Table). Cascade, rim pool and terrace tufas exhibited high success rates; 86% 228 of cascade samples (18 of 21), 100% of rim pools (7 of 7) and 100% of terrace breccias sampled (7 229 of 7) yielded resolvable U-Th age estimates, while only one of four dome samples returned a reliable 230 age. Reliable ages tended to be unresolvable on samples with very low 230 Th / 232 Th ratios (e.g. 230 Th 231 / 232 Th <7) indicating a significant detrital component (S3 Table). It was not possible to resolve 232 reliable or precise ages for any of the barrage samples, three dome and two cascade samples, 233 however some of the corrected ages for these samples may provide a useful upper limit age estimate, 234 i.e. the corrected age plus the associated 2σ uncertainty (S3 Table).  The tufa ages span the last interglacial cycle, from 110.6 ± 3.0 ka through to 3.0 ± 0.9 ka 250 (Table 1, Fig 3). The ages are clustered, suggesting episodic growth over this time, with at least four 251 intervals of tufa formation at Ga-Mohana Hill identified at approximately 114-100 ka, 73-48 ka, 44-252 32 ka, and 15-2 ka (Fig 3). The 2σ uncertainties associated with the ages are small; most samples are 253 associated with errors of <3 ka (on average approximately 1 ka) except for two samples, GHS-5 and 254 GHS-6.3, which have an uncertainty of 4.9 ka (49%) and 4.2 ka (7.9%) respectively. These larger 255 errors are due to a high detrital thorium component (Table 1) (E) tufa U-Th age data with 2σ error bars presented in Table 1. 262

263
The ages for the timing of human occupation at Ga-Mohana Hill coincides with three of the 264 tufa forming intervals during MIS 5d, late MIS 3, and late MIS 2, indicating contemporaneous 265 human activity and tufa precipitation at Ga-Mohana during those periods (Fig 3). The age certainty 266 for the interval of tufa formation that overlaps with the MIS 2 occupation at Ga-Mohana Hill is less 267 secure than the other intervals as it has a large error associated with it. The human occupation falls 268 within the 2σ uncertainty of the tufa age. 269

270
Comparison to global records proxies to consider potential forcing factors (Fig 3). There is no clear glacial/interglacial partitioning 273 of tufa formation episodes, as evidenced by comparing our data to the LR04 d 18 O benthic stack[46] 274 (Fig 3). This adds to growing evidence that the wet/dry, interglacial/glacial dichotomy through which 275 much of southern African palaeoclimates has traditionally been viewed is overly simplistic [9, 49-276 51]. While tufas in the northern hemisphere are typically associated with interglacial climate 277 conditions[52-55], our record suggests tufa formation was semi-continuous across MIS 4 and MIS 3; 278 similarly anomalous tufa growth is reported from other sites locally [29] and globally [30,56]. This 279 suggests that tufa formation is neither restricted to interglacial periods, nor is it a simple product of 280 changing global climate states. 281

282
The principal conditions required for tufa formation are sufficient effective precipitation to 283 recharge the aquifers and CaCO3 supersaturation of those waters [26,28,55]. Productive soil and 284 vegetation cover is necessary to enhance the pCO2 of the percolating waters, and moderate 285 temperatures which balance productivity, moisture and evaporation, are important secondary 286 requirements [26,30]. Tufa formation is thus sensitive to multiple environmental parameters, but 287 ultimately provides direct evidence of fresh water and associated productivity on the landscape. Our 288 record indicates that these conditions were met during the time intervals presented here in the 289 southern Kalahari over the last ~110 ka. 290

291
The limiting factor for tufa formation in semi-arid, low latitude regions is water 292 availability [28,57]. The spatial and temporal variability of rainfall in this southern Kalahari region is 293 poorly constrained, but is thought to be modulated by summer insolation, with increased 294 precipitation corresponding to insolation maxima [58]. However, there is no simple correlation 295 between Ga-Mohana tufa formation and insolation. Based on the mean summer insolation curve for 296 27°S[48] (Fig 3), tufa formation during the 114-100 ka and 44-32 ka intervals coincide with 297 increasing summer insolation, while tufa formation during 73-48 ka is variable, and at a minimum 298 during the most recent 15-2 ka episode. It has been suggested that direct insolation forcing has 299 played a lesser role over the last ~50 ka due to lower amplitude changes related to declining 300 eccentricity [59], and that after ~70 ka, high latitude changes may have had a greater influence on 301 southern African hydroclimate [60]. suggesting that SST is not the driving mechanism for increased rainfall in this region. While warmer 308 SSTs coupled with a negative Southern Oscillation Index is suggested as contributing to higher 309 rainfall during the 114-100 ka interval [19], it is likely that the primary driving mechanism for 310 rainfall in this region has varied over time [59]. 311 312 Comparison to regional records 313 We compare the record of tufa formation intervals at Ga-Mohana Hill with other 314 palaeoenvironmental records at nearby Kathu Pan and Wonderwerk Cave (Fig 4). These three sites 315 all occur within ~60 km of each other and are likely to have experienced the same climate systems. providing an opportunity to consider the relationships between wet periods and evidence for human 340 occupation in this region of the southern Kalahari (Fig 4) After ~71 ka, the timing of human occupation and wet periods are decoupled (Fig 4) that no single factor explains the timing of the past wet, tufa-forming periods, and that hydrological 385 dynamics in the southern Kalahari were influenced by multiple factors operating at various scales. 386 The tufa intervals represent a southern Kalahari environment characterised by a positive hydrological 387 balance and mild temperatures favourable for productive vegetation and soils. Our results challenge 388 global generalisations of past climate change, in accordance with other studies that highlight the 389 necessity for regionally specific models [10,73,74]. 390

391
In the southern Kalahari, early human population distributions appear to have been 392 modulated by water availability before ~71 ka but not after. Despite evidence for wetter conditions, 393 archaeological deposits dating to MIS 4 and the early part of MIS 3 have not yet been identified in 394 the punctuated record of human occupation at Ga-Mohana Hill, nor at nearby Kathu Pan or 395 Wonderwerk Cave. Future work is required to determine whether this absence of evidence is evidence of absence, or whether issues with site formation, site visibility, and/or dating challenges 397 explain why no archaeological deposits have yet been identified. This work will be key for further 398 testing hypotheses that link early human population distribution patterns to water availability, 399 potential refugia conditions, and interglacial/glacial cycling [4][5][6][7][8]. 400

401
The time interval corresponding to MIS 2 provides little coherence with respect to the 402 relationship between water availability and human occupation. The three records considered here do 403 not agree on whether conditions were wetter or drier during the LGM and humans appeared to have 404 occupied the region through both the LGM and late glacial. Others have highlighted that the 405 palaeoenvironmental record for MIS 2 across the Kalahari Basin and surrounding regions is 406 complex, documenting a high degree of spatial and temporal variability [10]. This lack of coherence 407 may be in part due to the variable responses of palaeoenvironmental proxies to temperature and 408 water availability changes, and potentially lags in responses. A shift in seasonality may also play a 409 role, with some proxies responding to seasonality changes for precipitation, as evidenced at Kathu Identifying the timing and nature of human occupation in the Kalahari Desert is critical for 418 understanding the emergence of our ability to adapt to new and extreme environments 2 . For a long 419 time, the Kalahari Desert has been considered too arid for early human populations to persist, and 420 evidence for occupation was assumed to represent wetter periods. Until now, a rarity of integrated palaeoenvironmental and archaeological records has largely prevented adequate testing of these 422