Sample preparation for limestone rocks from Coal and Lake Valley included the following steps:

The outer secondary carbonate coatings observed on the surfaces of some limestone rocks were filed away with a file or Stylo-style Dremel tool. These coatings did not cover the entire rock surface but were patchy and thin (generally < 5 mm thick).

This approach to sample preparation for limestone gravel-sized rocks requires the following assumptions:

During beach ridge formation, light penetrated the outer 2 mm or more of the limestone surface to bleach the signals from detrital quartz and feldspar minerals.

Volcanic rocks sampled from Cave Valley were prepared using the following approaches. Polymineral grains (i.e., samples that had undergone no mineral separation) were extracted from volcanic gravel-sized rocks in two ways: by coring and slicing (the traditional method), in addition to removal of the entire outer ∼ 1 mm layer from all sides of each rock. This second approach maximized the amount of material that could be used for De measurement:

The outer carbonate coating on the volcanic rocks was removed by setting them in a bath of high-concentration (36 %) HCl acid.

Our sample preparation protocol for volcanic gravels requires the following assumptions:

Minerals extracted by coring and slicing require that at least one side (the top or bottom) of the rock was exposed to sunlight long enough prior to burial to yield accurate ages.

The equivalent dose (De) was measured using variations of the single aliquot regenerative dose (SAR) protocol (Wallinga et al., 2000; Murray and Wintle, 2000). This protocol measures the sensitivity-corrected natural signal (Ln/Tn) of each grain/aliquot, followed by sensitivity-corrected regenerative dose signals (Lx/Tx) measured after a series of increasing laboratory doses (i.e., the dose response curve). The protocol included the measurement of a repeat-dose point (i.e., one regenerative dose point is measured twice) to measure the recycling ratio, and a zero-dose point (i.e. Lx/Tx is measured after no dose is given) to measure recuperation. Aliquots/grains were rejected from further analysis if the recycling ratio was beyond 10 % of unity and if recuperation was greater than 5 % of the sensitivity-corrected natural signal. Aliquots/grains were also rejected if their natural test dose signals were less than three times the standard deviation of the background signal.

We applied varying SAR protocols during this study to measure the infrared signal measured at 50 °C (IR₅₀ signal) as well as post-infrared infrared (pIRIR) signals measured at a range of temperatures. These protocols are shown in Tables 2 and S2. The IR₅₀ protocol includes low-temperature (160 °C) preheats prior to measurement of the regenerative and test doses, as these have been found in the past to reduce or remove recuperation and improve dose recovery test results (e.g., Neudorf et al., 2015). The pIRIR₂₉₀ protocol was adapted from Thiel et al. (2011). Because preheats must be equal to, or higher than pIRIR stimulation temperatures (Murray et al., 2009), this protocol includes a high preheat of 320 °C.

In this study, SAR protocols were tested prior to De measurement using dose recovery tests (Aitken, 1994; Roberts et al., 1999). During these tests the sample was bleached using IR or pIRIR stimulation, administered a known laboratory dose to the sample, then measured using the SAR protocol to be tested. The IR₅₀ SAR protocol was tested because, though it can exhibit anomalous fading, it depletes faster during light exposure in nature than pIRIR signals (Thomsen et al., 2008). pIRIR SAR protocols were also tested, as these exhibit lower fading rates, and may be more appropriate for samples with high anomalous fading rates (Thomsen et al., 2008; Buylaert et al., 2009).

Polymineral grains measured at the single-grain (or micro-hole) level in single-grain discs were first bleached simultaneously for 1000 s using IR diodes, then stimulated individually (in each hole) for 2 s using the IR laser. It was found that this second bleach with the IR laser was necessary as the shadowing effects of the holes in the single-grain discs prevented full depletion of the IR₅₀ signal in all grains by the IR diodes. A similar approach has been adopted for single K-feldspar grains by Feathers et al. (2019). Because the available grain size fraction from the rocks was significantly smaller (63–250 µm) than the width of the single-grain disc holes (300 µm) in the reader, it is likely that several (up to 5) grains within each disc hole contributed to the IRSL signal, and the results should be viewed as small multi-grain aliquot data, or “micro-hole” data as termed by Berger et al. (2013).

For the pIRIR protocols, the pIRIR signal was depleted using a pIRIR Ln measurement (Table 2) in order to target the pIRIR traps and avoid leaving a slowly bleaching part of the signal that can remain if the sample is only exposed to daylight or a solar simulator for a finite period of time (e.g., Li and Li, 2011). A disadvantage of this approach is that the bleaching method (the pIRIR Ln measurement) can lead to sensitivity changes that may lead to inaccurate dose recovery test results. For all SAR protocols tested, the protocol was accepted if the ratio of the measured-to-given dose was within 10 % of unity and few aliquots or grains were rejected from analysis due to dim signals, high recuperation or recycling ratios that are not within 10 % of unity. However, these results should be viewed with caution until further experiments can be conducted.

Sediments exposed at site CV21P1 reveal two main units including a massive, cohesive blocky silt (Unit 1 – clayey silt) overlain by a weakly bedded gravel with medium-coarse sandy matrix (Unit 2 – sandy beach gravel) (Fig. 3a, Table 3). Gravel clasts rarely exceed 40 mm along their intermediate (b-axis) dimension and lithologies are dominated by limestones derived from outcropping Devonian carbonate shelf deposits to the west (Fig. S5). These sediments are interpreted to represent pluvial lake highstand beach gravels overlying fine-grained sediments that were ponded in a lagoon behind the beach ridge shortly before rising waters deposited the gravels over them (Wriston and Adams, 2020) (Fig. 4). In this interpretation, the gravels and underlying lagoonal deposits are contemporaneous (cf. Adams and Wesnousky, 1998). Therefore, the age of radiocarbon dated shell from the lagoonal deposits (15 873–16 281 cal yr BP) is interpreted to represent the age of the beach gravels that were deposited there during the pluvial Lake Coal highstand.

Sediments exposed in excavation pits at sites CA21P1 and CA21P2 include gravel with silty sand matrix. Sedimentary structures were not visible due to the loose silt that quickly covers exposed sections. Both beach ridge gravels are dominated by volcanic rock lithologies derived from outcropping Miocene and Oligocene flows and tuffaceous sedimentary rock to the west and south, with more felsic varieties appearing more common at CA21P1 (Figs. 3b, S6, Table 3). Gravel clasts were subangular to subrounded with their longest dimensions commonly under 10 cm.

Sediments exposed in the excavation pit of site LK21P1 include gravel with silty sand matrix. Limestone dominates the rock lithologies, which is derived from neighbouring Devonian lithified carbonate shelf deposits (Fig. S7). Gravel clast sizes were similar to those found in Cave Valley. Samples collected from Lake Valley were not datable (see below), so further discussion of the site sedimentology is not reported here.

Twenty-one limestone rocks from site CV21P1 were processed for IR signal testing and De measurement (Table 4). Nineteen of these were clasts excavated within the beach ridge at ∼ 1 m depth, and two were cobbles collected from the ground surface (i.e., modern samples). Minerals extracted from seven of these gravels passed dose recovery tests either at the multi-grain or single-grain level, but only three of these (rocks #2, 10, and 18) had sufficient material left for De measurement and enough aliquots/grains that passed SAR rejection criteria. Minerals extracted from the two cobbles collected from the ground surface (rocks #2M and 3M, Table 4) did not have a detectable IR or OSL signal.

The dose recovery test results for Rocks 2 (63–90 µm), 10 (125–180 µm) and 18 (180–250 µm) are plotted in Fig. S9, where a laboratory dose of ∼ 21 Gy was recovered. Measured σb values were 22 %, 18 % and 17 % for Rocks 2, 10 and 18, respectively, where σb refers to measurement uncertainties that are attributed to instrument reproducibility and grain-to-grain variations in signal properties (Galbraith and Roberts, 2012). Measured-to-given dose ratios were 0.99 ± 0.02, 0.91 ± 0.03 and 0.96 ± 0.03 for Rocks 2, 10 and 18, respectively suggesting that the IR₅₀ SAR protocol is suitable for these samples, and thus these samples were chosen for dating. The 90–125 µm grain size fraction from Rock 2 was also dated in the following sections, but there was not enough material to perform a dose recovery test prior to dating. Thus, a σb value of 22 % has been assumed for this grain size fraction.

Fading measurements for Rocks 2, 10 and 18 were made on the same grains (or small multi-grain aliquots) used for De determination. These included single-grain (i.e., “micro-hole”) measurements of the 63–90 µm fraction from Rock 2 using single-grain discs, and single-grain measurements of the 125–180 µm fraction from Rock 10 and the 180–250 µm fraction from Rock 18. This allows each grain (or small multi-grain aliquot in the case of the 63–90 µm fraction) to be corrected for its own fading rate, which can vary significantly from grain to grain or aliquot to aliquot. An average fading rate was also measured from four 2 mm diameter multi-grain aliquots from the 90–125 µm grain size fraction from Rock 2.

Representative single-grain/small aliquot IR₅₀ signals and fading plots are shown in Fig. S12. The fading rate, or g-value, of each grain/small aliquot was measured using multiple prompt and delayed signal measurements. The weighted mean g-value for Rock 2 (285 micro-hole measurements of the 63–90 µm fraction) is 2.55 ± 0.52 % per decade, which is slightly lower but within 2 standard deviations of that measured from the 90–125 µm grain size fraction from the same rock (3.19 ± 0.15 % per decade). The weighted mean g-value of all single-grain measurements made for Rocks 10 and 18 is 2.65 ± 1.71 and 1.99 ± 0.79 % per decade, respectively.

ICP-MS/AES measurements of U, Th, K and Rb were conducted on crushed and milled subsamples of Rocks 2, 10 and 18 as well as a bulk sample of the gravel matrix (Table S8 in the Supplement). All radionuclide contents for the limestone gravels are low relative to values typically obtained for bulk non-carbonaceous sediment, and this is expected given the relatively low concentration of silicate minerals within the limestone. The gravel matrix yielded values that are slightly above those of the rocks, reflecting a higher silicate mineral content.

HPGe measurements of the limestone yielded results that are in range of those measured using ICP-MS/AES, while measurements made from the gravel matrix deviate somewhat from the ICP-MS/AES results. To check the calibration of our HPGe detector, measurements were made on a standard prepared by Murray et al. (2015) that was measured by 23 different laboratories (Table S9, Fig. S16A and B). Our results are within 20 % of those previously published.

Radionuclide activity ratios for the U and Th series in the limestone and gravel matrix samples are reported in Table S10 and plotted in Figs. S16C–F and S17. The gravel matrix sample from CV21P1 shows elevated Pb-210 activity relative to U-238 and Ra-226, while the Ra-226 / U-238 ratio approximates unity and the Ra-224 / Ac-228 ratio falls within 20 % of unity. Given that the gravel matrix sample was collected from loose, porous medium-coarse sandy gravels ∼ 1 m below the surface, we interpret the elevated Pb-210 values to reflect the translocation of atmospheric Pb-210 that has leached down to the level of sampling during rain events (Murray, 1996). The Ra-226 / U-238 and Ra-224 / Ac-228 ratios from the gravel matrix otherwise suggest that the deposit approximates secular equilibrium. Radionuclide activity ratios calculated from the limestone rocks are well within 20 % of unity, suggesting that our assumption that any U-disequilibrium that may have existed within the limestone during its formation will have corrected itself since Devonian times (Sect. 3.2.1).

Alpha, beta and gamma dose rates have been calculated for all limestone rocks as well as the gravel matrix from site CV21P1 (Table S13). Dose rates were calculated from both the ICP-MS/AES and HPGe measured radionuclide concentrations. As expected for carbonate-rich materials, all dose rates are low, ranging from ∼ 1.2 to ∼ 1.9 Gy ka⁻¹. HPGe measurements from limestone rocks yielded alpha, beta and total dose rates within the range of results obtained from ICP-MS/AES analysis. HPGe measurements of the gravel matrix underestimate the ICP-MS/AES beta and gamma results leading to a ∼ 24 % reduction in the total dose rate.

Dose rates were calculated for the outermost ∼ 2 mm of each limestone taking into account the measured alpha, beta and gamma dose rates from each rock, beta and gamma dose rates from the surrounding bulk gravel matrix as well as cosmic rays from outer space. Dose rates for each rock were modelled using the approach of Riedesel and Autzen (2020), which incorporates experimentally derived attenuation factors for granite, assuming that the rock is shaped like a sphere. This approach yields total dose rates that are lower than those calculated using earlier approaches that assume laterally infinite beta and gamma dose rates for both the rock and surrounding sediment (e.g., Jenkins et al., 2018) (see Sect. S8 for a comparison). To check that the beta and gamma attenuation factors of Riedesel and Autzen (2020) are appropriate for our limestone samples, the elemental concentrations of 4 representative limestone subsamples were measured by XRF and used to calculate attenuation factors using Geant4 simulations. These simulations generated attenuation factors that matched those published by Riedesel and Autzen (2020) for granite (Martin Autzen, personal communication, 2024), and so we apply them to our dose rate models in our study.

Modelled alpha, beta and gamma dose rates with depth into Rocks 2, 10 and 18 are shown in Fig. 5. These are based on radionuclide concentrations measured by ICP-MS/AES (Sect. 4.3.3). Similar dose rate with depth models were also calculated using the radionuclide contents determined using HPGe. The total dose rate used for calculating the age of detrital mineral grains within the outer 2 mm layer of each limestone rock is an average value highlighted in green in Fig. 5.

The De and luminescence ages were measured from polymineral grains from Rocks 2, 10 and 18; these rocks passed dose recovery tests (Sect. 4.3.1). The De was measured from two grain sizes from Rock 2 (63–90 and 90–125 µm fractions), the 125–180 µm fraction from Rock 10, and the 180–250 µm fraction from Rock 18. De was measured using the IR₅₀ SAR protocol in Table 2 for single-grains, and each grain (or small multi-grain aliquot) was corrected for its own fading rate using the correction model of Huntley and Lamothe (2001) for all samples except for the 90–125 µm fraction from Rock 2. All aliquots from the 90–125 µm fraction of Rock 2 were corrected for fading using an average fading rate measured from medium sized multi-grain aliquots (Sect. 4.3.2). As Fig. S13 shows, single-grain fading corrections have the effect of magnifying single-grain age errors for older grains within the distribution, and this is attributed to the limited precision with which we can measure single-grain fading rates.

The fading-corrected aliquot age distributions are shown in Fig. 6 where the cumulative distribution plot is superimposed on a kernel density estimate (KDE) curve. Overdispersion values (OD) for the fading-corrected data are reported in Table 5. These have been calculated using the central dose model of Galbraith et al. (1999) after taking into account the measured σb value in dose recovery tests (Sect. 4.3.1) and record the level of spread in the data that can be attributed to environmental factors during burial, such as incomplete re-setting of grain signals by sunlight or heterogeneities in the dose rate field that lead to grain-to-grain variations in acquired dose (Galbraith and Roberts, 2012).

Given that most rock dating studies derive ages from rock (primarily granite, sandstones, quartzites, and volcanics) slices, rather than single-grains or small aliquots, the relationship between single-grain/aliquot age distribution shapes, OD, and depositional process for rocks is not known and has not been examined in the same way these relationships have been examined for traditional sediment dating studies (e.g., King et al., 2014). We provide preliminary interpretations of these data that should be solidified by future testing. The OD of the 90-125 µm fraction of Rock 2 is 30 % and it shows a symmetrical KDE curve, which may indicate that most grains have been sufficiently bleached prior to burial (cf. Arnold and Roberts, 2009, Table 4 for quartz). The OD of all other samples are much higher, ranging from 66 % to 99 %, and may record incomplete bleaching as well as scatter that is the result of beta microdosimetry effects. KDE curves for the fading-corrected data are nearly symmetrical for Rock 10, possibly indicating that most scatter is due to beta microdosimetry effects (Mayya et al., 2006), while those of Rocks 2 and 18 show a slight positive skew that may record beta microdosimetry effects as well as incomplete bleaching of grains.

Statistical models have been developed for calculating a sample De value from a distribution of single-grain De values (e.g., Galbraith et al., 1999; Galbraith and Roberts, 2012; Guérin et al., 2017). These models are based on assumptions regarding the depositional history and composition of the sample analysed. Single-grain ages in this study were calculated using the minimum dose (MDM) and central dose (CDM) models of Galbraith et al. (1999) as well as the more recently developed average dose (ADM) model of Guérin et al. (2017). A description of the assumptions behind each model are outlined in Sect. S9.

All calculated ages are summarized in Table 5 and plotted in Figs. 6 and 7 alongside the radiocarbon age for the freshwater shell collected from the lagoonal silts (Unit 1). The radiocarbon age plots closest to the CDM ages for limestone Rocks 2 (90–125 µm, ICP-MS/AES) and 18, overlapping at 1σ, and agrees with CDM ages calculated for Rock 10 at 2σ. At 2σ the radiocarbon age also overlaps with MDM ages for Rock 2 (63–90 µm), as well as ADM ages for Rock 2 (90–125 µm, ICP-MS/AES) and Rock 18 (Table 5).

Figure 7 shows the relationship between ages determined using ICP-MS/AES dose rates and those determined using HPGe dose rates. These results must be viewed with caution as the HPGe measurements could not be made from each individual limestone sample, but rather were made from one milled sample that combined limestone pieces from multiple rocks from the site. The method of dose rate determination impacted the oldest ages of those compared, where the ICP-MS/AES method resulted in slightly younger age estimates relative to the HPGe method, however ages derived from both dose rate measurement methods agree within 1σ.

Preliminary dose recovery tests were conducted on 3 mm diameter multi-grain aliquots from all volcanic rocks from Cave Valley that exhibited an IR signal. These were all collected from site CA21P1. Pieces of each rock sample were ground and sieved into 32–63 µm grain size fractions prior to measurement. Each aliquot was bleached using a pIRIR Ln measurement (Table 2), then administered a ∼ 40 Gy beta dose prior to measurement using SAR. The pIRIR protocols tested included the 180, 225 and 290 °C protocols (Tables 2, S2), and results obtained from both the IR₅₀ and pIRIR signals from each protocol were plotted (three aliquots per rock) (Fig. S10) and tabulated (Table S6). Despite some outlying values of the measured-to-given dose, most aliquots passed the dose recovery test suggesting that, after signal sensitisation during the Ln measurement, feldspars in these rocks were suitable for SAR.

After assessment of volcanic rock fading rates (Sect. 4.4.2) and rock characteristics (i.e., the composition of Rock 14, below), additional dose recovery tests were conducted on Rocks 4, 7, 11, 12, 13 and 18 using only the PIRIR₂₉₀ protocol, as this was the only protocol to yield acceptable fading rates (see Sect. 4.4.2). These tests were conducted on the ground and suspension-settled 32–63 µm grain size fraction (24 aliquots per sample) and were used to estimate a σb value. IR₅₀ and pIRIR₂₉₀ signals from all rocks passed dose recovery tests, yielding measured-to-give dose ratios within 10 % of unity and σb values equal to 12 ± 4 % or less (Fig. S11, Table S6). Cutting and coring of Rock 14 revealed that most of this rock was composed of cryptocrystalline quartz (chert). This chert does not have a luminescence signal, so Rock 14 was not processed further.

Measurements of anomalous fading were conducted on the 32–63 µm fractions tested above. As expected, fading rates were the highest for the IR₅₀ signal (up to ∼ 40 % per decade), and generally decreased with increasing pIRIR signal temperature (Fig. S15). The pIRIR₂₉₀ signal yielded an average g-value of 3.3 ± 0.7 % per decade, with individual aliquot fading rates varying from 0 to ∼ 12 % per decade, while the pIRIR₁₈₀ and IR₅₀ signals yielded average g-values of 10.3 ± 0.9 and 23.1 ± 1.8 % per decade, respectively. We chose the highest temperature pIRIR protocol tested (pIRIR₂₉₀) to minimise fading as much as possible, and thus additional fading tests using the pIRIR₂₂₅ protocol were not pursued.

ICP-MS/AES and HPGe measurements for volcanic Rocks 4, 7, 11, 12, 13 and 18 are shown in Table S11. As expected, radionuclide concentrations are much higher than those observed in limestone (Sect. 4.3.3). The HPGe results in Table S11 agree with the ICP-MS/AES results within 1σ for U and K, and within 2σ for Th. Radionuclide activity ratios for U in the gravel matrix collected from site CA21P1 show elevated levels of Pb-210 relative to U-238 that is likely atmospheric and sourced from recent rain events (Table S12, Fig. S18) (Murray, 1996). All other activity ratios fall within 20 % of unity, suggesting that disequilibrium in the U and Th decay chains is limited or negligible.

The modelled dose rates for the dated volcanic rocks from Cave Valley are shown in Fig. 8, and calculated dose rates for both rocks and the gravel matrix are tabulated in Table S14. Calculated dose rates for the gravel matrix from site CA21P1 are high, with a total dose rate equal to ∼ 5 Gy ka⁻¹ (Table S14). Dose rates calculated for all volcanic rocks using ICP-MS/AES radionuclide concentrations are also high, ranging from ∼ 6.2 to ∼ 7.4 Gy ka⁻¹ (Table S14).

The outer ∼ 1 mm layer (dremeled 32–63 µm fraction) of Rocks 4, 7, 11, 12, 13 and 18 was measured for De using the pIRIR₂₉₀ protocol for multi-grain aliquots in Table 2. Measurements were made on 3 mm diameter multi-grain aliquots, each containing ∼ 2600 grains. Aliquot De values (not corrected for anomalous fading) were divided by the average calculated dose rate within 1 mm of the gravel surface to obtain ages for both the IR₅₀ and pIRIR₂₉₀ signals. The pIRIR₂₉₀ ages were corrected for fading using the model of Huntley and Lamothe (2001) using g-values calculated from four multi-grain aliquot fading measurements from each rock (Table S7).

As expected, IR₅₀ uncorrected ages are significantly younger than pIRIR₂₉₀ uncorrected ages (Figs. 9 and 10) and this is attributed to the high rate of fading of the IR₅₀ signal as well as the lower bleaching rate of the pIRIR₂₉₀ signal. Fading-corrected aliquot pIRIR₂₉₀ age distributions for the dremeled outer rock surfaces are plotted in Figs. 9 and 10. Weighted mean surface ages for each rock were calculated using CDM and are shown in Figs. 9 and 10 as well as in Table 6.

De was measured from rock core slices using the same pIRIR₂₉₀ SAR protocol that was used for volcanic rock surfaces (Figs. S20, S21). Some slices could not be measured due to limited sample material or dim signals and aliquots that failed SAR aliquot rejection criteria. Due to irregular rock surfaces and heterogeneities in rock hardness, slice thicknesses varied. As expected, IR₅₀ De values consistently underestimate the pIRIR₂₉₀ De values in all rocks, and that includes those De values measured from the surface (i.e., top- or bottom-most) slice. Several rock pIRIR₂₉₀ De profiles (Rock 4, Rock 7 core 1, Rock 11, Rock 13, Rock 18 core 2) rise with depth into the rock at either the top or bottom side. For example, the Rock 7, core 1 pIRIR De profile starts at its lowest point at the bottom surface of the rock (left), then rises with depth into the rock (Fig. S20B). The IR₅₀ De profiles of the same rocks typically rise in a similar pattern, but at a much lower rate, yielding much more subdued (flatter) profiles. Similar patterns have been observed in IR₅₀ and pIRIR luminescence-depth profiles reported elsewhere (e.g., Sohbati et al., 2015; Freiesleben et al., 2015; Jenkins et al., 2018).

Aliquot slice De values were divided by depth-attenuated dose rates calculated in Sect. 4.4.4 to derive age-depth profiles (Figs. 11–13). Age-depth profiles generally show similar patterns to those observed in the De profiles; here we plot them using a logarithmic scale on the y-axis so that the IR₅₀ age-depth profile shape is vertically enhanced. The uncorrected IR₅₀ and pIRIR₂₉₀ age-depth profiles, as well as the fading-corrected pIRIR₂₉₀ age-depth profiles further support our interpretation above that gravels experienced short-term, heterogeneous light exposure to their surfaces prior to burial. Clear, plateaus that intersect the rock surface suggestive of a long-term bleaching event prior to gravel burial (such as that shown in Fig. 1) are absent.

Attempts to apply a light bleaching model to the luminescence signal depth profiles to predict pre-burial profile shapes (e.g., Freiesleben et al., 2015; Khasawneh et al., 2019) were unsuccessful; due to the scatter in the data and relative paucity of data points from each rock, model results were overly sensitive to parameter starting values. Therefore, plateaus were detected in the data using the statistical test for homogeneity of Galbraith (2003) following the approach of Gliganic et al. (2021). Using this approach, we identified statistically consistent populations of pIRIR₂₉₀ ages within each slice and rejected outlying values. Then slice ages in each depth profile (each calculated using CDM) were systematically tested against each other to identify plateaus within each age-depth profile (Figs. 11–13). Plateau ages were then calculated by applying CDM to all statistically consistent groups of ages (Table 6).

Weighted mean pIRIR₂₉₀ ages for the dremeled outer ∼ 1 mm layer of each volcanic rock (32–63 µm grain size fraction) are summarized in Table 6 and Fig. 14a. The pIRIR₂₉₀ fading-corrected ages range from ∼ 3 to ∼ 17 ka, where all rocks except for Rock 13 dates to the mid-late Holocene between ∼ 3 and 6 ka. The CDM weighted mean of all volcanic rock surface ages, excluding the low-precision, high outlying value from Rock 13, is 3.7 ± 0.4 ka.

Rock surface ages were calculated from the outer-most slices of each rock core (125–250 µm sized grains from the top, bottom or sides of the rocks), as well as from statistically consistent slice ages (the plateaus) within each age-depth profile (Table 6, Fig. 14b). Again, fading-corrected surface slice ages vary widely with pIRIR₂₉₀ ages ranging from ∼ 3 to ∼ 20 ka. The rock surface slice ages are generally equal to, or older than the dremeled surface ages discussed in Section 4.4.7 above, suggesting that rock slices may contain unbleached, residual signal as a result of irregular cut rock surfaces. Rock top and bottom age values show significant variability within and between rocks suggesting that sun exposure was not uniform across rock surfaces prior to burial and/or the rocks experienced phases of re-mobilization prior to final emplacement.

Weighted mean ages calculated from age-depth profile plateaus range from ∼ 9 to ∼ 20 ka and show consistency from core to core in Rocks 7 and 18 where more than one core was taken (Table 6, Fig. 14b). The CDM weighted mean age of all plateau ages is 12.2 ± 0.9 ka.

Luminescence ages derived from limestone gravels from Coal Valley varied significantly from rock to rock and even showed dependence on measured grain size within the same rock (compare ages for grain size fractions 63–90 µm and 90–125 µm for Rock 2, Table 5). This variability is likely attributed to non-unform bleaching of rock surfaces, variability of rock light transmission properties (Ou et al., 2018), differences in chemical or physical weathering, as well as beta microdosimetry effects that result from non-uniform rock composition (Meyer et al., 2018).

Calculated CDM ages were most likely to agree with the independent age control, while ADM ages tended to overestimate the expected age (Table 5, Fig. 7). The congruency between the CDM ages of Rocks 2 (90–125 µm fraction), 18 and 10 suggest that their surfaces were reasonably well bleached. The anomalously old age obtained from the 63–90 µm grain size fraction from Rock 2 is more difficult to reconcile but may be indicative of incomplete bleaching of finer grains that were more susceptible to clumping than their coarser grain counterparts. MDM ages straddle the radiocarbon age estimate of the deposit, again with the 63–90 µm fraction from Rock 2 yielding the oldest calculated MDM age (Table 5, Fig. 7). MDM ages that appear to post-date the time of beach ridge formation may not be derived from grains inside the limestone at all, but rather, despite our efforts to remove outer carbonate coatings, may be contaminating grains that became cemented onto the limestone surface as part of a pedogenic carbonate coating.

The main challenge dating limestone clasts relate to the limited quantity of detrital grains available for dating (Table 4, column 2). Limestone formations in Nevada can vary in their detrital sediment content but are commonly associated with, or interbedded with, sandstones and shales (e.g., Rowley et al., 2017; Hurtubise and Bray, 1988). The concentration of detrital sediment available within limestone rocks for dating varies from location to location, and this necessitates the need for testing for viable luminescence signals from each site, perhaps with some guidance from local geological maps. Quartz signals from polymineral grain extracts exhibited no fast component in any of the limestone gravels tested, so further analysis was restricted to feldspar. Grains that do have an IR signal, however, were typically well suited to SAR, and had moderate fading rates that are relatively easy to correct for, even at the single-grain level.

For limestone gravels, it is preferable to apply the single-grain approach as opposed to measuring larger multi-grain aliquots, primarily because the single-grain approach requires relatively small amounts of material, but also because it produces a high-resolution age distribution for each sample that can be examined for evidence for incomplete bleaching of grains or other sources of scatter. Past research shows that it is common for rocks to experience only partial sun exposure prior to burial on one or more sides (e.g., Meyer et al., 2018; Smith et al., 2023), so single-grain dating, and perhaps the MDM model, should allow us to target grains that are more likely date the most recent bleaching event. However, we must keep in mind that the relationship between single-grain age distribution shapes, overdispersion (OD), rock transport and depositional history, and chemical and physical weathering of limestone rock surfaces, has yet to be examined in detail, and should be the focus of future research.

Luminescence signals suitable for dating are most likely to be found in volcanic rocks that are intermediate to felsic in composition and contain a higher concentration of K-rich feldspars. Such types of rock are common in Lincoln County, with intermediate silicic ash flow tuffs and other tuffaceous sedimentary rocks covering more of Nevada than any other type of rock (Crafford, 2007). Intermediate and felsic volcanic rocks are best represented at site CA21P1 in our study.

Volcanic rock surface ages and age-depth profiles from site CA21P1 suggest heterogeneous light exposure on rock surfaces. The variability in rock surface ages, where clast tops often appear younger or older than clast bottoms, and the variability between age-depth profiles from within the same rock (Table 6, Fig. 14), are consistent with observations made at other sites (e.g., Rades et al., 2018; Souza et al., 2019; Smith et al., 2023) and are indicative of incomplete bleaching of all rock sides prior to burial. Bleaching rates in rocks are dependent on rock surface aspect, light transmission of the rock, daylight spectrum, intensity, and duration of exposure (Ou et al., 2018; Smedley et al., 2021; Furhmann et al., 2022).

Variability in age-depth profiles may also be attributed to rock composition. Rocks that are light in color, and fine-grained, with a homogeneous mineral composition over the scale of analysis, are anticipated to be better and more uniformly bleached than rocks that are dark in color, coarse-grained, and have a heterogeneous composition over the scale of analysis (Ou et al., 2018; Meyer et al., 2018). Heterogeneities in the dose rate field within gravels and near gravel surfaces could also lead to micro-beta dosimetry effects that are not accounted for in our dose rate models.

The oldest ages of the dataset from Cave Valley approach ∼ 20 ka (Rock 13) (Table 6). This supports the inference that the beach ridge at site CA21P1 formed no earlier than Marine Isotope Stage (MIS) 2. However, an unexpected finding of this study was the large number of anomalously young ages obtained from volcanic gravel surfaces from site CA21P1. Gravel surface ages obtained from pIRIR₂₉₀ signals suggest that many gravel surfaces were light exposed long after the last pluvial lake highstand in Cave Valley, and even as recently as the late Holocene (Table 6, Fig. 14). Given that our sampling depth at this site was less than 0.5 m below the present-day surface, this must be due, in part, to (i) bioturbation, and possibly (ii) post-burial light exposure of gravels in what were initially open-work gravels near the beach ridge surface after beach ridge formation.

Soils in dust-influenced arid and semi-arid regions of the American west are thought to form by accretionary processes where dust influxes increase soil volume and inflationary strain (McFadden, 2013). In this accretion-inflationary mode of profile development (also known as AIP), dust becomes trapped by vegetation, desert pavement stones, or irregular, rocky surfaces, then translocates down the soil profile thickening the Bt and Bk horizons. Increases in dust flux during the Holocene have been linked to a 20 cm rise in ancient Pleistocene desert pavement surfaces on lava flows in the Cima Volcanic Field in the Mohave Desert, California (McFadden et al., 1987). Thermoluminescence (TL) ages of ∼ 12–13.5 and 5 ka were obtained from Bwk and Av soil horizons, respectively, that are situated on 560 ka lava flows. These ages were attributed to the continuous supply of aeolian materials to the soils during the Late Pleistocene, with an increase in dust flux during the early to middle Holocene (McFadden et al., 1998).

Lake level reconstructions in our study area suggest that Cave Lake receded from its highstand ∼ 18 000–20 000 years ago (Duke and King, 2014), consistent with the plateau age of Rock 13 (Fig. 14). Initially these gravels may have been open-work gravels where large pore spaces allowed light penetration to some depth below the beach surface, allowing for the bleaching of near-surface rocks after ∼ 18 000 years ago. As the climate conditions became more arid from 14 600 to 12 900 years ago (Rhode and Adams, 2016), and again after the Younger Dryas, dust accumulation would have facilitated the establishment of soil and shrubland vegetation (t1 in Fig. 15). Root penetration and burrowing animals would have created a zone of bioturbation below the surface. Given that roots of Great Basin desertscrub species (Spaulding, 1985) can extend several feet below the surface to access water at depth, this depth of disturbance likely extended up to a meter or more and likely pushed gravels at the surface down to the depth of sampling. Plateau ages from inside Rocks 4, 7, 11 and 18 (Fig. 14b) coincide with the desiccation of lakes in the region between ∼ 14 600 and ∼ 9400 years ago (Rhode and Adams, 2016; Duke and Young, 2018) and may record burial of these gravels by newly exposed and wind-transported playa silt and sand. The mid-late Holocene ages derived from rock surfaces (Fig. 14a and b) likely record more recent temporary heterogeneous light exposure of the rocks as they are re-mobilised in the bioturbation zone of the soil.

During the Middle Holocene Drought (∼ 8–4 ka) (Wriston, 2009; Steponaitis et al., 2015), dust influx to the beach ridge soils likely intensified (cf. McFadden et al., 1992). We infer that accretion-inflationary processes contributed to volume expansion of the soil raising its surface above the original Late-Pleistocene open-work gravel surface level (t2, Fig. 15). This implies that gravel clasts sampled at ∼ 38–44 cm depth in this study, if not bioturbated, would have been several centimetres closer to the original open-work gravel surface prior to accretion. The youngest volcanic rock surface ages cluster between ∼ 2.8 and ∼ 6.4 ka. This suggests that the number of surface gravel clasts pushed down to the depth of sampling decreased substantially after ∼ 3 ka as the A horizon thickened and the substrate became less permeable. This implies that by the late-Holocene the soil surface and the zone of subsurface bioturbation had risen above the level of sampling (t2, Fig. 15).

All fading-corrected pIRIR₂₉₀ plateau ages from volcanic rocks from site CA21P1 are finite, with no evidence of saturation (Figs. 11–13 and 14b). In previous studies where cobble-sized rocks are dated, the highest plateau identified near the centre of the rock luminescence-depth profile is typically interpreted to record “saturation” (Fig. 1). “Saturation” refers to the area where bleaching has never occurred, and the luminescence signal has reached an equilibrium, where signal loss due to fading (for feldspars) is in equilibrium with signal gained via irradiation. However, in some cases, if the rocks dated are small enough, and if light exposure is long enough, light can penetrate the entire thickness of the rock, leading to a flattening of the plateau. Lehmann et al. (2018) showed that the inflection point of the IR₅₀ luminescence-depth curve migrates to 3–4 mm depth after 20 to 140 years of sun exposure for coarse grained orthogneiss. Ou et al. (2018) found that after 91 days of sun exposure, the pIRIR₂₂₅ signal was bleached to half of its initial intensity at a depth of ∼ 1.8 mm for a dark grey, fine grained indurated sedimentary greywacke. Furhmann et al. (2022) depleted the pIRIR₂₂₅ signal to a depth of over 2 mm in granite exposed to the sun for 108 d. Evidence for internal bleaching of IR₅₀ signals in rocks of thicknesses ranging from ∼ 15 to ∼ 23 mm has been shown by Souza et al. (2019) for samples collected from a modern beach and a ∼ 2000-year-old sandy beach ridge in Denmark. pIRIR signals are known to bleach at a slower rate than IR₅₀ signals, but nonetheless, this begs the question as to whether the Cave Valley beach gravel rocks are small enough such that the pIRIR₂₉₀ signal was completely depleted throughout their thickness in the swash or near-shore zone of the lake environment prior to burial.

Despite the susceptibility of gravels to re-working in the soil zone, the age range of our rock plateau ages suggests that pIRIR₂₉₀ signals inside gravel rocks may serve as more reliable geochronometers for the time of beach ridge formation and initial soil development at the late-Pleistocene-Holocene transition than the signals measured at the rock surfaces. Such a scenario could occur if gravels in the swash zone of the lake, or emplaced at or near the surface of an abandoned beach ridge, were sun-exposed for long enough periods to completely deplete their pIRIR₂₉₀ signals throughout their thickness prior to final burial (cf. Souza et al., 2019). In this scenario, the pIRIR₂₉₀ signal that accumulates at the centre of the gravels during burial may be less prone to depletion during subsequent brief periods of sun exposure during bioturbation events, which preferentially depletes the signal near the surface of exposed rock surfaces.

The ∼ 18 000–20 000 year old Cave Lake highstand is thought to have preceded the Lake Coal highstand 15 873–16 281 cal yr BP (Duke and King, 2014; Wriston and Adams, 2020), however despite this, most Cave Lake rock surface and internal plateau luminescence ages post-date the highstand of Lake Coal as well as the model-equivalent rock surface ages from the Coal Valley beach ridge (Tables 5 and 6). This is likely due in part to the depth of sampling; at Cave Valley the sampling depth (< 0.5 m) was less than that at Coal Valley (1 m) and clearly within the zone of the A/B soil horizon, thus increasing our chances of sampling bioturbated material. When sampling pluvial lake gravel beach ridges, efforts should be made to sample at depths of ∼ 1 m or greater. Natural exposures of beach ridge sediments such as those sampled at Coal Valley are rare in the Great Basin, and so deeper sampling may require the use of mechanical equipment.

Dose rates for Cave Valley were calculated assuming that the present geochemistry of the samples have been consistent during the burial history of the beach ridge. This assumption is problematic given the shallow (< 0.5 m) sampling depth within soils that would have developed shortly after beach ridge abandonment and continued to evolve during the Holocene. Accretion-inflationary soil profiles are dynamic, characterized by cumulic growth as well as pedogenic modification. As the soil develops, it increasingly influences infiltration rates, depths of water movement, and thus, rates and processes of carbonate translocation and accumulation for the entire soil (McFadden et al., 1992, 1998). Future luminescence sampling should ideally focus on primary beach ridge sediments below any soil development until pedogenic effects on time-averaged dose rates are better understood.