The suitability of a low temperature post-IR IRSL signal 1 for dating alluvial and colluvial “ cut and fill ” sequences 2 in the Great Karoo , South Africa

11 Alluvial and colluvial “cut and fill” deposits preserved in valleys of the Karoo, 12 South Africa, reflect basin-scale adjustments in fluvial process-regime. Such 13 deposits in the Wilgerbosch catchment have previously proven difficult to date 14 using radiocarbon (C) and optically stimulated luminescence (OSL) methods. In 15 this paper, we test the suitability of K-feldspar post-IR infrared (pIRIR) methods 16 on 19 samples from Africanders Kloof, a low-order tributary of the Sundays River. 17 Using three carefully screened quartz OSL ages, radiocarbon dating and site 18 stratigraphic considerations we argue that the pIRIR170 protocol can be used to 19 produce reliable age estimates. Fading rates for the pIRIR170 signal are 20 Final version published in Quaternary Geochronology available here: https://www.sciencedirect.com/science/article/pii/S1871101420300133 2 consistently low (average g2days: 0.81 ± 0.58). The pIRIR170 residuals are dose 21 dependent (r=0.58); but are consistently low as a proportion (e.g. 1-4%) of 22 sample equivalent dose (De). Despite the water-lain depositional context, single 23 aliquot De distributions tend toward normality (for 11/19 samples) irrespective of 24 aliquot size (2 mm or micro-aliquots containing 2-30 grains) with only a few 25 statistical outliers per sample (max. n=3) and overdispersion (OD) ranging from 26 1.6-30% excluding the two youngest (late Holocene) samples (OD: 37-87%). The 27 resulting pIRIR170 ages are in the correct stratigraphic order and illustrate the 28 potential of pIRIR170 luminescence dating to investigate the timing, processes and 29 drivers of fluvial system adjustments across the Karoo. 30


166
Grain size terminology follows the Udden-Wentworth (1922) scale. Some of the stratigraphic relations 167 between these fills have been revised (from Oldknow and Hooke, 2017) in light of subsequent field 168 observations. OSL prefix codes have been changed from LV-(as reported in Oldknow and Hooke, 2017) to Final version published in Quaternary Geochronology available here: https://www.sciencedirect.com/science/article/pii/S1871101420300133 9 The basin is infilled by at least four extensive terrace fills in the higher order 171 tributaries which have been interpreted as evidence of basin-wide cyclic cut and  The alluvial succession and locations of pIRIR samples are provided in Fig. 3. 177 The sampling scheme ensured complete representation of the stratigraphy with a 178 view to developing a chronology of cut and fill at the specified sites. The revealed that this calcrete is "rhizogenic" (Wright, 1990;Oldknow, 2016) and 198 formed in association with a phreatic root system accessing an elevated water 199 table (Wright et al., 2005).

200
T3 is only partially exposed at outcrop AK11 due to an inset deposit up to 2.4 m 201 thick (T4). T3 consists of thick units of inversely graded silty sand (unit F1) and 202 sandy silt (unit F2); buried by matrix-supported gravels (unit G). Oldknow and 203 Hooke (2017) interpreted these facies as indicative of slopewash rather than 204 channel or overbank sedimentation.

205
In contrast, T4 at profile AK11 is characterized by normally-graded, horizontally-206 bedded units including matrix-supported gravels (units A and D), clast-supported 207 cobbles (unit B), silty sands (unit C); and a relatively thick (~1 m) homogeneous 208 sand unit (unit E) from which a quartz OSL age (AK11-1: 17 ± 2.5 ka) was 209 obtained (Oldknow and Hooke, 2017). These deposits are interpreted as a 210 migrating single-thread channel. In particular, the elevation and thickness of high-211 energy palaeo-flood deposits (unit B) corresponds to debris flow facies (e.g. AK8 212 unit D) in close proximity to an eroding dolerite tor (see Fig. 1C). On this basis,

213
Oldknow and Hooke (2017) proposed that the aggradation recorded at T4 (AK11; 214 this study) must have resulted from a phase of strong slope-channel connectivity (AK-8), implying that both the OSL and pIRIR ages for samples AK8-1 and AK11-216 1 ages should be concordant with one another.

217
T5 is characterized by well-sorted sandy silt (e.g. AK13 unit B2) or organic-rich 218 silty sands containing plant macrofossils (e.g. AK15 unit B). It is interspersed with 219 coarser gravel and sand units with inverse grading in the Wilgerbosch River.

220
These facies indicate valley-floor wetland systems with low-energy channels that 221 later incised due to land-use changes introduced during the European incursion 222 (Boardman, 2014). An AMS 14 C age (P-37289; 0.44 ± 0.04 cal kyr BP, outcrop 223 WGR-1, Fig. 2) was obtained from fossilized Juncus stems and provides a 224 maximum age for the incision of these wetlands (Oldknow and Hooke, 2017).

225
Rowntree (2013) and Boardman (2014) demonstrated gully erosion in this area 226 was active at the start of the 20 th century. Therefore, new pIRIR ages from AK13-227 1 and AK15-1 should bracket or overlap with this radiocarbon age.

233
Incomplete bleaching is most likely to be a challenge for samples collected from 234 coarser (higher energy) channel deposits coupled with the shortest transportation 235 distances (e.g. AK8-1). Samples collected from units previously subjected to 236 waterlogging (e.g. AK12-6 to AK12-9; AK13-1) may also have been affected by variability in water content and external environmental dose. Indeed, a further 238 advantage of using pIRIR methods in this context is the partial mitigation of these 239 factors via the dose rate contribution from internal K and Rb.

241
The three ages previously reported (two quartz OSL, one 14 C) from the 242 Wilgerbosch sequence offer a key benchmark against which to test pIRIR age 243 reliability, especially the potential issue of bleaching. However, given the 244 problems with quartz OSL in this area (Boardman et al., 2005;Oldknow, 2016), 245 additional screening of the published OSL ages (samples AK4-1 and AK11-1, 246 originally LV-509 and LV-515 in Oldknow and Hooke, 2017) as well as analysis 247 of a new OSL sample (AK8-1) was undertaken to more thoroughly investigate the 248 purity of the quartz. K-feldspar was then extracted from the same samples, dated 249 with a low temperature pIRIR protocol and compared to the quartz OSL results.    Table S2a) produced 43 De estimates on 1 mm aliquots. Data analyses and 298 integration limits used were the same as specified previously.
Final version published in Quaternary Geochronology available here: https://www.sciencedirect.com/science/article/pii/S1871101420300133 15 As such types of non-UV emitting feldspar are potentially present given the 304 geological context at Wilgerbosch (section 2.1), a series of alternative non-305 standard luminescence tests to screen for feldspar contamination were carried 306 out. 1) Thermal quenching (cf. Wintle, 1975). A pure quartz sample exhibits 307 "thermal quenching" with increased stimulation temperature (Duller et al., 1995) 308 while feldspar exhibits "thermal assistance" (McKeever et al., 1997). A modified 309 SAR protocol whereby the stimulation temperature was varied (from 50-325 o C at 310 20 o C intervals) was applied to 8 aliquots of each sample (Table S3). The (1) I= I0 / (1 + C exp (-W / kT) 315 Where W is the thermal activation energy, I is the OSL intensity, I0 is the initial 316 OSL intensity, k is the Boltzmann constant and C is a constant. Pure quartz 317 should exhibit a W value of around 0.51 eV (Wintle, 1975;Shen et al., 2007) with 318 W values ~0.2 eV reported for feldspar (Poolton et al., 2002;Shen et al., 2007).

319
2) The ratio of IR intensity to TL intensity at 110 o C (IR/TL herein) employed by Li 320 et al. (2002). Most feldspars exhibit TL at 110 o C, but unlike quartz do not exhibit 321 a single peak; rather they manifest as the rising limb (thermal assistance) of 322 higher temperature TL peaks (McKeever et al., 1997). As only some feldspars 323 emit UV under IR stimulation, but most feldspars exhibit thermal assistance (post 324 110 O C), the IR/TL ratio is more sensitive to polymineral mixtures than the IR-depletion test alone (Duller, 2003). Mauz and Lang (2004) demonstrated that 326 IR/TL ratios approaching 0 are indicative of pure quartz, while IR/TL ratios closer 327 to 1 indicate feldspar contamination. They cautioned however that as the IR/TL 328 ratio only considers signal intensity at a specific temperature (110 o C), 329 interpretation of the IR/TL ratio should be coupled with qualitative appraisal of the 330 shape of the TL peak. Thus, we define two additional aliquot rejection criteria: 1)

331
Aliquots which demonstrate thermal assistance (rather than thermal quenching)

337
K-feldspar grains were extracted by treating each of the 19 samples with 10 % 338 v/v HCl and 32 % v/v of H2O2 to remove carbonates and organics, respectively.

339
Wet sieving isolated the coarsest fraction possible (controlled by sample amount) 340 and was usually 180-or 212-250 µm (Table S1). Density separation utilizing 341 sodium polytungstate provided the 2.53-2.58 g cm -3 K-rich feldspar fractions, 342 which were then etched in 10% HF for 10 minutes. 2 mm aliquots (several 343 hundred grains) and micro-aliquots (2-30 grains) were analysed using a pIRIR 344 protocol comprising a low temperature (50 o C) initial stimulation, followed by a 345 high temperature (pIRIR) stimulation (Table S2b). IRSL for the 50 o C was 346 measured for 100 s; IRSL for the high temperature stimulation was measured for Final version published in Quaternary Geochronology available here: https://www.sciencedirect.com/science/article/pii/S1871101420300133 17 300 s. Samples were preheated for 60 s at a temperature set 30 o C above pIRIR 348 stimulation temperature following Roberts (2012).  Dose recovery tests (herein DRTs) and additional residual dose measurements 363 were performed to assess the robustness of the SAR protocol (section 4.5).

364
Bleaching was achieved by exposing samples to natural UK daylight for 48 hours.

365
The administered beta dose was matched to the sample De and measured using 366 the protocol outlined in Table S2b. Additionally, an "added dose" approach to  water content (Aitken, 1985) and HF etching (Bell, 1979). For K-feldspar, the 395 internal K and Rb dose rates assume grain K and Rb concentrations of 12.5 ± 396 0.5% and 400 ± 100 ppm respectively (Huntley and Baril, 1997;Huntley and 397 Hancock, 2001). Cosmic dose rates were determined using the profile elevation 398 (metres above sea-level, herein masl) and the measured sample depth (Prescott 399 and Hutton, 1994) with a 5% relative uncertainty.

406
Taking the calcrete as a surrogate of the maximum upper limit of the 407 palaeowater-table (section 2), an average burial water content of 15 ± 5% 408 (Oldknow and Hooke, 2017) was used in age calculations for any samples below 409 it (AK12-6 to AK12-9). For samples above the calcrete, the measured water 410 content was used. A 5% change in water content results in a ~2000 year age 411 difference for samples aged ~50 ka and ~1200 years for samples aged ~30 ka.

412
Final version published in Quaternary Geochronology available here: https://www.sciencedirect.com/science/article/pii/S1871101420300133 414 Table 1  is not reached for >75% of aliquots and for 9 aliquots, the TL observed at 110 o C 422 does not occur as a single peak; but as a thermally assisted rising limb of a 423 higher temperature peak indicative of feldspar (Fig. S4b). Six of these nine 424 aliquots also exhibit IR/TL values >0.25 that are consistently higher than AK4-1 425 and AK8-1 (Table 1) Table 1 438 Range and mean values of selected OSL-based parameters (thermal activation energy W; IR/TL ratio) to 439 assess quartz purity.   Table 2 450 Results of quartz OSL analyses and revised ages.     Figure 5 shows the pIRIR170 luminescence behaviour and De distributions.

475
Results are expanded to include pIRIR170 data from a fourth sample (AK15-1) 476 which comes from a stratigraphic horizon coeval with radiocarbon age P-37289 477 (section 2). Despite the larger De of samples AK8-1 and AK11-1 (up to 63.6 ± 1.2 Gy) compared to AK15-1 and AK4-1 (up to 41.1 ± 1.2 Gy), the IRSL signals are 479 not especially bright (max. ~7000 counts). This could be a consequence of less

534
Sample AK4-1 exhibits an OD of 17%. Removal of a single low outlier (19 Gy) 535 reduces OD (7.3%) and moves the CAM pIRIR170 age from 8.2 ± 0.5 ka to 8.6 ± 536 0.3 ka (Table 4). AK8-1 is rather more dispersed (OD: 30%) considering its (2 537 mm) aliquot size, but removal of two high outliers (89 and 151 Gy) brings its 538 pIRIR170 age from 18 ± 1.5 ka to 16.1 ± 0.4 ka, which is in better agreement (1 σ 539 Final version published in Quaternary Geochronology available here: https://www.sciencedirect.com/science/article/pii/S1871101420300133 26 error) with the OSL age (Figure 4, Tables 2 and 4). Sample AK11-1 exhibits very 540 low OD (6.4%). 541 Figure 6 demonstrates that more than one hour of sunlight exposure is required 542 to reduce the pIRIR170 signal of sample AK12-8 to less than 33% of its original  Table S1). This pIRIR170 residual age is four 548 times greater than that of AK15-1 (Table 4) and 25% the magnitude of age AK8-1; 549 yet both of these ages are concordant with independent age control. This is 550 considered further in section 4.5.

607
Internal checks of pIRIR170 protocol performance carried out on 2 mm aliquots. Note: (i) dose recovery ratios 608 (DRR) are average values using all measured (n=3) aliquots; (ii) AK8-2 and AK12-X were subjected to two 609 DRTs using low and high administered doses; (iii) AK12-X was used during preliminary testing and not 610 dated.

611
*Refers to pIRIR225 measurements given for comparative purposes.

708
We now consider the performance of the pIRIR170 protocol on the basis of 709 internal checks (Table 3). In general, the high aliquot acceptance rate and good equally not be serious, but a full single grain analysis would be required to 729 assess the degree of grain-to-grain signal variation (Rhodes, 2015). What is clear

742
Sample AK8-1 is the only example of a 2 mm aliquot pIRIR170 age exhibiting high 743 OD (~30%) most likely due to inclusion of partially bleached grains (section 5.1).

744
Indeed, although several high De outliers were observed for other samples (e.g.

37
The second possible extrinsic contributor to OD is bioturbation (Jankowski et al.,

756
The observation that De scatter increases down to the micro-aliquot scale (i.e.

757
<10 grains, e.g. AK13-1) may indicate that sample OD is incorporating the effects 758 of intrinsic factors (i.e. variations in K content, fading and microdosimetry), but an 759 analysis of single grains would be required to assess their relative contributions 760 (Rhodes, 2015).

761
In summary, sample OD tends to largely reflect few individual outliers, the 762 presence of which makes minimal difference to the resulting CAM pIRIR170 ages.