Rock surface luminescence dating of gravel determines the age of a glacial outburst megaflood, Glacial Lake Missoula, Montana, USA

Geologic records of large floods supply important data with which to determine flood magnitudes, recurrence intervals, associated hazards, and climate variability (Kochel and Baker, 1982). Boulder deposits (Balbas et al., 2017) and sand and gravel sequences in Glacial Lake Columbia (Atwater, 1987) and the Channeled Scabland (CS) in Washington State (USA) document late Pleistocene outburst floods of Glacial Lake Missoula (GLM) between ca. 18.2 and 14 ka (Fig. 1; O’Connor et al., 2020). Catastrophic drainage events formed giant gravel bars in the lake basin, including on Cayuse Hill, near the head of Alberton Gorge (Fig. 2; Smith, 2006). To date, GLM flood chronology is only known from deposits in the CS.

Paleoflood chronologies successfully developed from rock surfaces in the CS and elsewhere have used in situ cosmogenic nuclide (CN) dating of exposed boulder surfaces (Balbas et al., 2017), flood-scoured bedrock (Lamb et al., 2014), and flood-flipped boulders (Fujioka et al., 2015). Recently, a new method based on optically stimulated luminescence (OSL) has been successfully applied to constrain the burial ages of in-place rocks in surficial deposits (Chapot et al., 2012; Sohbati et al., 2012b; Simkins et al., 2013; Souza et al., 2021) and till (Rades et al., 2018), but it has not yet been applied to paleoflood deposits. An advantage of rock surface optical methods over those applied to sand and silt is that surface bleaching is documented by luminescence-depth profiling. The rock surface luminescence method raises for the first time the possibility of direct measurement of the burial chronology of gravel eroded and transported during paleofloods.

We applied rock surface luminescence burial dating to gravel deposited during highstage paleoflood(s); namely, draining of GLM. The ultimate goals were to obtain burial ages for high-energy gravel deposits that represent one or more large flood events in a basin that underwent multiple lake drainages and, more broadly, to gain insights into optical dating of paleoflood deposits.

GLM formed by impoundment of the Clark Fork River where the Purcell lobe of the Cordilleran ice sheet terminated in a canyon near the present border of Idaho and Montana in the northwestern United States (Fig. 1A). Repeated total or partial failure of the ice dam caused the lake to release water onto the CS multiple times (Pardee, 1942; Waitt, 1985). As shown by giant bed forms within GLM and the CS, the earlier drainage events were catastrophic megafloods, whereas later floods were smaller (Smith, 2006; Alho et al., 2010; O’Connor et al., 2020). Megaflood deposits within the GLM basin include eddy bars and flow expansion bars (Figs. 1B and 1C). These bars contain >10-m-tall crossstratification and open-work gravel fabric of grain-supported gravel that lack a sand-sized matrix. These features are characteristic of sedimentation of gravel deposited from suspension by high-velocity currents.

For this study, we sampled a giant gravel bar near the lee-side crest of the bedrockcored Cayuse Hill in three locations in a quarry (Fig. 2). The bar is one of five bars in the Alberton Gorge, ~170 m above flood-eroded bedrock in the valley floor, and ~250 m below the maximum GLM water level reached by the glacial lake (Fig. 2A; O’Connor et al., 2020). Cross-stratification shows that the lee face prograded ~300 m downstream during one or more megaflooding events (Fig. 2). Bedrock east of the gravel deposit has ~10-m-wide, internally drained scabland depressions that suggest erosion by high-velocity currents.

As the hill was isolated from nearby mountains at high GLM water levels, sediment could only have been supplied to the site by floods and not by shoreline transport. The large amount of gravel in this deposit must have accumulated during one or more megafloods as water was forced over the hill to drain into the head of Alberton Gorge where the upper Clark Fork River tributaries converge (Fig. 1A). The cobbles are feldspathic siltite or fine- to mediumgrained sub-feldspathic and feldspathic sandstone and quartzite derived from Proterozoic Belt Supergroup that crops out upstream of the bar (Lonn et al., 2010).

Due to the lack of organic material or tephra within GLM sediments, deposit ages have been limited to those obtained from quartz OSL and K-feldspar infrared stimulated luminescence (IRSL) on sandy glaciolacustrine deposits and buried periglacial features (Hanson et al., 2012; Smith et al., 2018). Initial filling of the lake to >85% of its maximum depth occurred by 20.9 ± 0.9 ka (Smith et al., 2018). Other than sand carried by traction currents during lake drawdowns (Hanson et al., 2012), no megaflood deposits have yet been dated in the former lake basin. Ice-rafted boulders at the highest flooding water depths recognized in the CS suggest that the largest late Pleistocene outburst paleofloods occurred at 18.2 ± 1.5 ka, based on ¹⁰Be exposure dating of four boulders at two locations (Fig. 1A; Balbas et al., 2017; O’Connor et al., 2020).

On a rock surface, decreasing penetration of light with depth resets the luminescence signal to different amounts below the exposed surface (Fig. 3A; Habermann et al., 2000; Sohbati et al., 2011). This provides an advantage for rock surface luminescence dating over conventional OSL dating of sand grains because rocks preserve a record of their bleaching history that can be deciphered from the shape of the luminescence signal with depth profile beneath the exposed rock surface (Sohbati et al., 2012b; Freiesleben et al., 2015). In a buried rock that was sufficiently bleached prior to deposition, this luminescence-depth profile has a characteristic sigmoidal shape starting from finite values close to the surface and rising toward saturation at depth, where bleaching is negligible (Sohbati et al., 2012a; Freiesleben et al., 2015). The finite luminescence signal close to the surface is acquired during burial and can be measured to derive a burial age (Fig. 3A; Sohbati et al., 2015; Jenkins et al., 2018).

We sampled 38 cobbles in low-light conditions from three pits dug at different distances along the progradation direction of the gravel bar on Cayuse Hill (Fig. 2A). Each pit was ~1.5 m below a recently excavated floor of an active gravel quarry (Fig. 2B). Preferentially selected cobbles were rounded or subrounded and had an oblate shape. These clasts would have tended to rest on one of their flat sides, promoting preferential daylight exposure on the opposite side prior to entrainment. We also collected seven dose rate samples of the surrounding matrix. In the laboratory, each tabular side of the cobbles was cored under low-level amber light using a coring bit mounted in a drill press. Cores were sectioned into thin slices for luminescence measurements. Dose rates were measured for each of the cobbles identified as well bleached, and thus suitable for dating, and for each matrix sample.

All of the luminescence measurements were done using a single-aliquot regenerative-dose protocol (Murray and Wintle, 2003) for equivalent dose (D_e) measurements (Table S1 in the Supplemental Material¹). Eight cobbles collected in 2018 CE from site 1 (Fig. 2A) were analyzed using the IRSL signal measured at 50 °C (Fig. S1). Thirty (30) cobbles sampled in 2019 from sites 2 and 3 were screened by measuring relative sensitivities to infrared and blue light stimulations; 17 of the 30 cobbles were identified as suitable for dating by pulsed OSL (POSL) signal measured after infrared stimulations; the remaining cobbles were measured by IRSL or conventional OSL (see the Supplemental Material).

Effective dose rate at the surface of each target cobble was calculated by modeling the contributions of beta and gamma dose rates from the cobble itself and the surrounding matrix, including the effect of grain sizes in the cobbles and the porosity of the matrix (see the Supplemental Material; Fig. S3; Table S2). Water content at the site, after withdrawal of the lake, was assumed to be negligible given the deposit’s high permeability and water-well records.

Luminescence-depth profiles were measured from the natural sensitivity-corrected signals (L_n) and the subsequent response to a test dose (T_n) with depth from the two tabular sides of 36 cobbles, and one side of two cobbles. Profiles of 24 of the 38 cobbles had consistent L_n/T_n ratios from the surface to the depth measured, showing no light exposure before burial; their luminescence signals were in field saturation. Another 11 cobbles showed reduced L_n/T_n values near a surface but no near-surface plateaus, indicating that these cobbles were exposed to light before burial but not enough to be bleached (Fig. S2). Three cobbles, C-3, C-4, and C-1-11, showed profiles with sigmoidal shapes typical for well-bleached rock surfaces, with small values near the surface, where the ratio of the predicted preburial profile to the observed postburial profile was <5%, rising toward saturation at depth (Fig. 3B). The luminescence signals from these surfaces were acquired during burial and thus correlate to their burial ages. The IRSL signal measured from the exterior slices from multiple cores of bleached surfaces of cobbles C-3 and C-4 and the post-IR POSL signal from the bleached surfaces of cobbles C-1-11 were used for estimation of D_e (Table 1).

The burial ages of cobble surfaces were calculated by dividing the average D_e values by the effective dose rates (Table 1). We obtained a post-IR POSL age of 21.7 ± 1.1 ka for C-1-11, which was collected from site 2, the farthest down-transport site on the gravel bar (Fig. 2A). Cobbles C-3 and C-4 from site 1 provided uncorrected IRSL ages of 12.2 ± 0.7 ka and 13.1 ± 1.0 ka, respectively. Due to instability of the feldspar IRSL signals, the ages must be corrected for signal fading. We followed two different approaches for fading correction (see the Supplemental Material). The g-value (anomalous fading) corrections (Auclair et al., 2003) yielded ages of 16.5 ± 1.0 ka and 19.8 ± 1.7 ka for C-3 and C-4, respectively, and corrected ages derived from field-to-laboratory saturation (FLS) ratios (Rades et al., 2018) were 16.5 ± 0.9 ka and 18.5 ± 1.4 ka for C-3 and C-4, respectively (Table 1). In the case of C-4, it appears that the g-value corrected age is older than the FLS ratio corrected age. This may be attributed to an overestimation of the IRSL age for this particular sample with a high g value (>4% per decade; Reimann et al., 2011). Consequently, the FLS ratio corrected ages were used for comparison with the independent age control and for deriving the average luminescence rock surface burial age.

The corresponding IRSL ages of C-3 and C-4 determined by different correction methods are similar within the error limits, increasing our confidence in their reliability. These two cobbles, collected from near the middle of the bar, were likely deposited simultaneously by the same flood event. Cobble C-1-11, from farther downstream, has an older burial age than the others but is consistent with the 20.9 ± 0.9 ka age for initial transgression of GLM (Smith et al., 2018). This suggests that C-1-11 was likely buried at 21.7 ± 1.1 ka and then entrained in a megaflood during which no bleaching occurred.

A comparison between the average luminescence age of C-3 and C-4 of 17.5 ± 1.0 ka and the ¹⁰Be CN exposure age of 18.2 ± 1.5 ka on four boulders deposited during the largest megaflood (determined by Balbas et al., 2017; see also Fig. 1) suggests that the luminescence rock surface chronology agrees with the independent ¹⁰Be exposure age control within the error limits. This agreement shows for the first time that megaflooding within GLM can be temporally correlated with megaflood deposits in the CS. This highlights the potential of rock surface luminescence burial dating in providing chronologies for paleoflood events.

The high likelihood of sediment erosion combined with the low likelihood of bleaching during a catastrophic flood event make it challenging to apply traditional OSL dating to sandy flood deposits. Measuring bleaching in gravel solves this problem. Application of luminescence dating to gravelly flood deposits would be like that of mass movements, in that clasts previously exposed to daylight (and so bleached at the surface) are entrained in the body of the deposits during the event and buried after transportation. We thus deduce that this technique can be equally successful in dating similar highenergy deposits and mass movements such as debris flows, rockfalls, and landslides (Chapot et al., 2012).

While the small proportion of well-bleached cobbles suitable for dating (8%) compared to that of partially or unbleached cobbles may be regarded as a disadvantage with this technique, the screening procedure to select the well-bleached cobbles is efficient (Jenkins et al., 2018). Moreover, new advancements in instrumentation based on the infrared photoluminescence signal from feldspar (Prasad et al., 2017) promise rapid in situ assessment of luminescence-depth profiles by imaging in the field (Sellwood et al., 2022). Such a field instrument may enable selection of cobbles that were well bleached before burial and therefore increase laboratory efficiency. Portable OSL readers (Sanderson and Murphy, 2010) may also be useful for detecting bleached rock surfaces. Sample selection with the use of these instruments may enable generation of larger age distributions, providing sedimentary process information that is currently unobtainable.

The landowner D. Rehbein graciously allowed access to, and excavated, the sites. Montana Technological University (Butte, Montana) supplied partial support for shipping samples and travel to the Technical University of Denmark Risø campus for L.N. Smith. Suggestions by A. Balbas, Y. Gavillot, and two anonymous reviewers improved previous versions of the manuscript. Nordic Laboratory for Luminescence Dating staff at Risø, including V. Hansen, W. Thompson, and H. Olesen, provided help and guidance.