Comparison of dating methods for paleoglacial reconstruction in Central Asia

Reconstruction of former Central Asian glaciers extents can provide valuable information about past atmospheric circulation variations. These extents, often marked by terminal moraines, need to be chronologically constrained. Cosmogenic nuclide exposure (CNE) dating is widely used to directly date moraines. In addition, there is increasing interest on using optically stimulated luminescence (OSL) techniques for dating glacial landforms. This thesis focuses on the methodological aspects of directly dating glacial landforms to perform paleoglacial reconstructions in Central Asia, with an emphasis on OSL dating. For OSL dating of sediments from glacial settings, it is important to measure the luminescence signal at the single grain scale, because the sediments are likely affected by partial bleaching due to short light exposure during glacial or glaciofluvial transport. The use of an Electron Multiplying Charges Coupled Device (EMCCD)-based imaging system for single grain OSL measurements would offer larger flexibility in light stimulation and sediment type, compared to the current Single Grain Risø reader. An automated image processing procedure has been developed to compensate for sample carrier displacement over repeated measurements and for attributing pixels to each grain for signal integration when using this imaging system. However, significant cross talk contamination, demonstrated by laboratory and simulation experiments, prohibits accurate single grain luminescence measurements. Preliminary experiments using a basic image processing algorithm show good potential for software correction solutions. Paleoglacial reconstructions conducted in the Altai Mountains, Central Asia, using both CNE and OSL dating demonstrate that luminescence measurements of glaciofluvial sediments performed at the multi-grain scale result in large age overestimates, and that single grain measurements allow for more accurate dating of glacial landforms. However, uncertainties remain that are related to the model used for extracting equivalent doses for well-bleached grains and to fading corrections when using feldspar minerals. The timing of glaciation can be inferred from scattered CNE moraine boulder ages if most of the ages are concentrated within a few thousand years, with only few ages clearly older or younger. Overall, combining CNE and OSL techniques for dating a glacial landform is a powerful approach for producing robust glacial chronologies, despite uncertainties inherent to each technique. Paleoglacial reconstructions from the Altai Mountains indicate Marine Isotope Stage (MIS) 2 and MIS 4/late MIS 5 local Last Glacial Maximums. In Central Asia, in addition to a regional MIS 2 glaciation, previous studies indicate a period of major glacial advances during MIS 3 that is out of phase with global ice volume records. However, most MIS 3 glacial chronologies from Central Asia are based on too few or too heavily scattered CNE data sets, or on OSL or Electron Spin Resonance (ESR) ages for which partial bleaching has not been properly investigated. Hence, at this stage, chronological evidence is insufficient to demonstrate a regional MIS 3 glaciation in Central Asia. Surge-related glacial features identified in the Russian Altai also highlight the importance of conducting detailed geomorphology and sedimentology studies to understand former ice dynamics, which is essential for inferring appropriate paleoclimate information from paleoglacial reconstructions.


Reconstruction of former extent and timing of Central
Asian glaciers can provide valuable information about past atmospheric circulation variations. Further understanding of paleoenvironmental conditions in this area is of particular interest because several studies indicate that periods of extent and retreat of Central Asian glaciers were out of phase with global ice volume records and alpine glaciers from most formerly glaciated areas. However, the robustness of paleoglacial reconstructions largely rests on the accuracy of chronological constraints of glacial landforms, most often moraines. Glacial chronologies derived from moraines and glacio-fluvial deposits are largely based on cosmogenic nuclide exposure (CNE) and optically stimulated luminescence (OSL) dating methods. Large scattering of CNE moraine boulders ages, due to geomorphological processes (e.g. inheritance and post-exposure), often complicates the assignment of a glacial timing associated with moraine formation. Low sensitivity quartz commonly observed in high mountain environment and partially bleached sediments due to short light exposure during glacial or glaciofluvial transport most often hamper the dating of glacio-fluvial deposits using OSL techniques. This thesis focuses on the methodological aspects of directly dating glacial landforms using CNE and OSL techniques, with an emphasis on OSL, in order to improve our understanding of glaciation pattern in Central Asia.
In order to deal with partially bleached sediments, it is essential to measure the luminescence signal at the single grain scale. In line with this need, this thesis contributes to the development of a new Electron Multiplying Charges Coupled Device (EMCCD)-based imaging system for single grain luminescence dating. Such a system would offer a larger flexibility in term of stimulation source, grain fraction and mineral type compared to the current and wellestablished laser-based single grain attachment scanning system. However, three main issues hamper accurate single grain measurements when using an EMCCD-based imaging system. These issues relate to the sample carrier mo-  tion over repeated measurements, the absence of an automatic and reproducible method to assign pixels per grain for individual signal integration, and the contamination of individual signals by signal emission from neighbouring grains (cross talk). An automated image processing procedure has been developed to compensate for sample carrier displacement and to attribute pixels cluster for each grain to allow automatic signal integration . While laboratory measurements and simulations demonstrate that significant cross talk effects prevent the use of this system to perform routine single grain measurements, preliminary experiments using a basic image processing algorithm suggests good potential for a software correction solution , encouraging further work in this direction.
Palaeoglacial reconstructions studies were conducted in the Altai Mountains in Central Asia (Russia and China; Gribenski et al. 2016, in review) to investigate uncertainties related to CNE and OSL dating techniques and their effects on the reliability of glacial chronologies. Resulting chronological data demonstrate the importance of conducting luminescence measurements at the single grain scale for sediments deposited in glacial environments. Measurements carried out with multi-grain aliquots were shown to yield large age overestimates of glacial landform age, due to partial bleaching effects. However, determining accurate glacial landform age from single grain luminescence measurements remains challenging. The ages inferred from single grain measurements can vary significantly depending on the statistical sub-sampling technique used for extracting luminescence data associated with well-bleached grains. In addition, fading corrections, associated with the use of the feldspar mineral as a dosimeter, increase the final uncertainty. CNE dates of moraine boulders revealed important inheritance effects, in contrast to previous datacompilation analyses and model-based studies that suggest a pre-dominance of post exposure effects. In general, CNE ages of moraine boulders exhibit significant scattering, however glaciation timing can be extracted from CNE age populations that are composed of a majority of ages concentrated within a few thousand years and a small number of older/younger outliers. Glaciation periods inferred from CNE ages ranging over a few thousand years are considered acceptable, as this time range can reflect still stands oscillations of the ice margin. The presented paleoglacial reconstruction studies show that combining CNE and OSL methods to directly date glacial landforms is a powerful approach for producing reliable glacial chronologies, in particular for pre-LGM glacial deposits for which source of uncertainties are typically larger when dated using a single technique. Consistent CNE and single grain OSL ages strengthens our confidence in the chronologies despite uncertainties related to each dating method individually.
The palaeoglacial reconstructions conducted in the Altai Mountains in this project clearly constrain glacial advances to the MIS 2 and are in line with the glaciation timing indicated by most of the well-constrained glacial chronologies published in Central Asia region. In addition, our study also provides robust evidence for a local maximum extent beyond the MIS 2 glacial extent (Kanas Valley, China; Gribenski et al., in review), which occurred somewhere during the MIS 4/late MIS 5 and so predated the global LGM. However, too few palaeoglacial reconstructions have been conducted in the Altai area to attribute a regional character to this pre-LGM glacial event. In general, abundant geomorphological evidence indicates earlier Last Glacial glaciation(s) exceeding the MIS 2 glacial extent in Central Asia. In particular, a period of major glacial advances during MIS 3 in Central Asia has been suggested by several studies, which in addition to be out of phase with global ice volume records, corresponds to a period of relative global warmth. However, detailed analysis of chronological data set associated with major MIS 3 glacial advances in Central Asia indicate that most of the reconstructions are based on few or heavily and evenly scattered CNE ages of moraine boulders, or on OSL or Electron Spin Resonance (ESR) data for which potential partial bleaching of the samples has not been thoroughly investigated. Therefore, at this stage, available chronological data do not present a compelling case for a widespread MIS 3 glaciation in Central Asia. Future detailed geomorphological and chronological studies of former glacial extents in this region are warranted to provide tighter constraints on the pre-LGM glacial activity in Central Asia.
Finally, this thesis highlights the importance of detailed geomorphological and sedimentological investigations to understand former ice dynamics, which is crucial for proper paleoclimate inferences. Indeed, numerous non-climatic factors such as local topography, debris cover, or geological substrate can influence glacier expansion, interfering with the interpretation of climatically induced glacier extent changes. As such, geomorphological and sedimentological data indicate that large outermost moraine belts marking prominent advances of the Chagan Uzun Glacier in the Russian Altai, and dated at ~MIS 2, were formed during fast flowing or surge-like event. Based on these interpretations, it would be inappropriate to reconstruct ELA from the outer moraine belts, as fast flowing/surge-like behaviour reflects a glacier not in equilibrium with climate. Although understanding palaeo ice dynamics might be challenging, detailed geomorphological and sedimentological investigations can help to identify non-climatically driven (or not mainly) paleo glacier expansions, and so prevent inappropriate ELA reconstructions and paleoclimate inferences.