Introduction

1.1 Rationale

Understanding Earth's climate is one of the most important and urgent tasks science is fa-cing today. An accurate prediction of future climate shifts due to anthropogenic and natural impacts on atmospheric temperature and circulation is paramount for long-term planning of political and economic measures to secure and promote man's welfare in a changing en-vironment. Physical circulation models, which could ultimately be able to simulate the non-linear effects of changes in climate forcing in a way precise enough for these purpo-ses, integrally depend on palaeoclimate datasets to serve as either boundary conditions or evaluation benchmarks (Kohfeld & Harrison, 2000). One important palaeo-dataset for the evaluation of climate system models is the record of past mountain glaciations.

While studies in soil development in formerly glaciated mountain areas, e.g. considering horizon thickness, clay mineralogy, iron and aluminium chemistry, or feldspar weathering indices can be used to establish a relative chronology (Baeumler, 2001a), numerical ages can only be obtained using physical dating methods, e.g. radiocarbon dating, or lumines-cence dating (Bradley, 1999). The applicability of these methods in studies of glacial histo-ry, however, is limited.

Radiocarbon dating at present can provide reasonable ages only up to ~40 ka, and it is limited to preserved carbon-containing materials, which in arid regions may be few or absent. In addition, radiocarbon ages can provide no more than age limits for a great number of glacial deposits, because the radiocarbon age may not correspond to the glacial event, but reflect organisms dying long before, or long after it, respectively.

Luminescence dating requires fine-sandy to silty material well radiated during transport, but totally shielded from sunlight ever since deposition. Such material is not common in glacial deposits, and as the preservation of aeolian and fluvial sediments frequently is scarce in mountainous areas, chronologies obtained with luminescence dating are often fragmentary and leave room for different interpretations (e.g. Richards et al., 2000a, Kamp et al., 2003).

Surface exposure dating (SED) using the accumulation of in-situ cosmogenic nuclides (e.g.

3He, ¹⁰Be, ¹⁴C, ²⁶Al, ²¹Ne, ³⁶Cl) instead provides a way of dating rock surfaces directly, if

these have been freed from deep shielding (>3 m of rock cover) by a short-lived geologic event. Up to now, ¹⁰Be surface exposure dating may be considered the most advanced and most widespread SED method. By employing ¹⁰Be surface exposure dating in formerly glaciated catchments containing quartz rich rocks, complete glacial chronologies can be inferred without depending on the presence of buried organic material or the outcropping of suitable sediments for luminescence dating. Complete SED chronologies in turn can provide crucial information on palaeoclimate in the region, especially when combined with information derived from soil development investigations (Cerling & Craig, 1994, Fabel &

Harbor, 1999, Gosse & Phillips, 2001).

This work is part of an effort to establish ¹⁰Be surface exposure dating as an important new tool in the pedogeographical and palaeoecological research activities at the Institute of Soil Science and Soil Geography at the University of Bayreuth, in collaboration with the Paul Scherrer Institute at the ETH Zurich.

1.2

Be surface exposure dating

In-situ cosmogenic ¹⁰Be is continually produced within the upper one to three meters of the lithosphere by interaction of particles from the secondary cosmic radiation with the O and Si atoms of the quartz mineral lattice. The production rate of ¹⁰Be in the mineral depends on the amount of cosmic radiation reaching the sample, which can be predicted by using empirical measurements of cosmic ray activity in the atmosphere, along with calibration measurements of ¹⁰Be in rock surfaces with an independently known exposure age (Gosse

& Phillips, 2001). In-situ produced ¹⁰Be in quartz is locked in the mineral grid and there-fore accumulates, its accumulation only limited by radioactive decay of the nuclide. The nuclide concentration N at a rock surface therefore is a function of its production rate P, the exposure time of the surface t, and its decay constant λ.

The standard physical model for ¹⁰Be surface exposure dating is a flat, even, infinite rock surface z [g cm^-2] = 0, exposed to a full sky of cosmic radiation since a point in time t0 [a]

= 0 (Nishiizumi et al., 1993). The cosmogenic nuclide production rate at the surface is P [atoms g^-1 a^-1], which below the surface decreases exponentially with the attenuation length Λ [g cm^-2] of the cosmic rays. No other way of production of ¹⁰Be, e.g. by α-particles from U decay, is allowed in the model. Nuclides formed in the rock are completely retained and lost only by radioactive decay or surface erosion. Radioactive decay of the nuclide depends

on its concentration N [atoms g^-1] in the rock and its decay constant λ [a^-1]. In case of linear erosion with an erosion rate ε [g cm^-2 a^-1], the resulting standard production equation used in ¹⁰Be surface exposure dating is

,

or, resolved for t, the standard exposure age calculation equation,



Production in this model is simplified. The production rate P in fact has to be calculated separately for three production mechanisms (by neutrons, capture of slow negative muons, and reactions of fast muons) which are characterized by different values of Λ, and it has to be calculated as a product of the global standard production rate at sea level, high latitude, P0, the local scaling factor S, and a set of correction factors f used to account for model weaknesses. These weaknesses are 1) the shielding of a part of the sky by topographic objects, 2) the shielding effect of surface inclination, 3) the shielding of the surface by overlying matter, like snow or vegetation, 4) the shielding effect of the finite thickness of the sample, 5) the neutron-scattering effects of the three-dimensional form of the sampled object, 6) the time-dependency of the production rate due to changes in the local magnetic field coordinates (dipole wobble) and strength (dipole moment), and 7) the time dependency of the production rate due to tectonic uplift or downlift of the sample surface.

Thus, despite the apparent simplicity of the production equation, a standard procedure for calculating ¹⁰Be exposure ages still has not been agreed on. The differences are concerning 1) the scaling factors used to derive the local ¹⁰Be production rate in quartz from the standardized ¹⁰Be production rate in quartz at sea level in high latitude (SLHL), 2) the standardized production rate itself, 3) the complexity of treatment of the production by different production mechanisms, and 4) the set of correction factors used.

The interpretation of ¹⁰Be exposure ages is also still problematic. Calculated exposure ages up to now are considered only within the uncertainties resulting from the errors of the measured concentrations. Rigorous error analysis is often put aside (Gosse & Phillips,

2001). Secondly, deriving a moraine age from surface exposure ages of a selection of erratic boulders has in some cases proven to be a more difficult task than thought at first.

On the one hand, erratic boulders deposited on a moraine may contain ¹⁰Be inherited from a previous period of exposure, leading to an overestimation of the moraine age; on the other hand, erratic boulders might have been broken free from a larger block, or might have been cleared from sediment cover long after deposition of the moraine, leading to an underestimation of the moraine age (Owen et al., 2003a, b). Several models have been pro-posed to derive a moraine age from a distribution of erratic boulder exposure ages (Zreda et al., 1994; Hallet & Putkonen, 1994; Shanahan & Zreda, 2000; Putkonen & Swanson, 2003), but all of them are based on linear moraine degradation, which can explain uni-modal distributions of exposure ages only. However, bi- or even polyuni-modal distributions are frequently observed (e.g. Owen et al., 2003a, b) and have to be interpreted.

In this work, I introduce TEBESEA (acronym for TEn BEryllium Surface Exposure Ages), a program I devised for the calculation of ¹⁰Be surface exposure ages of erratic boulders with fully propagated errors, and I employ this program 1) to evaluate the current calcula-tion procedures in the light of the standard in-situ cosmogenic ¹⁰Be production rate calibra-tion studies published up to now, 2) to compare them in the context of our dating studies in Nepal and Central Asia, and 3) to estimate the influence of the variable correction factors on exposure ages. Further, I discuss in detail both error propagation and interpretative model use in deriving moraine ages from ¹⁰Be exposure ages of erratic boulders, in order to understand how moraine ages are best determined from erratic boulder exposure ages, and how exact those ages can safely be considered at present.

1.3 Palaeoglaciations of the Nepal Himalaya

Since the late 1990s, a lot of effort is spent in defining new glacial chronologies for the Nepal Himalaya using optically stimulated luminescence (OSL) and in-situ cosmogenic nuclide dating techniques (e.g. Richards et al., 2000b, Asahi et al., 2003, Finkel et al., 2003). These studies provide a new foundation for the discussion about past climatic conditions in the Himalaya as a whole, which is mainly about whether the past glaciations have been triggered during warm stages, in connection with an enhanced Indian monsoon, or during cold stages, in connection with a strengthening of the westerly circulation (Benn

& Owen, 1998, Bush, 2000, Fort, 2000).

In this work, ¹⁰Be surface exposure dating (SED) of erratic boulders is applied to confirm and complement the results of former soil geographic studies at two sites in the central Nepal Himalaya, the Macha Khola Valley (Zech et al., 2003), and the Langtang Valley (Baeumler et al., 1996, 1997, Baeumler, 2001a). Results are compared with other SED and OSL dating studies in order to evaluate, to which extent glacial advances in different regions of central Nepal have been synchronous.

1.4 Palaeoglaciations of the Pamir

A lot of effort is presently spent in defining numerical glacial chronologies all over High Asia, ranging from the mountain ranges of Central Asia in the northwest to the southeastern margin of the Tibetan plateau (e.g. Owen et al., 2002a, Owen et al., 2003c, Gillespie et al., 2003), a region that was extensively glaciated in the past and is considered a key locality for the understanding of the world's climate (Benn & Owen, 1998).

However, there still is no consensus about the timing of glaciations in the different parts of the region and its implications for past climate change (Zheng et al., 2002, Ono et al., 2004, He et al., 2004).

In this study ¹⁰Be SED is used to reconstruct the glacial history of the north-western part of High Asia, namely the Central Asian mountains between the Turkestan and Alay Ranges of south-western Kyrgyzstan, and the south-central Pamir plateau of eastern Tajikistan.