Interpretation of sedimentary subpopulations extracted from grain size distributions in loess deposits at the Sea of Azov, Russia

Abstract

Loess-palaeosol sequences in Eastern Europe, especially those in the Azov region, are among the most sensitive terrestrial archives for identification of past aeolian dynamics and Quaternary palaeoclimatic reconstruction. Grain size analyses of loess sediments are widely used to interpret these transport mechanisms and palaeoclimatic changes based on granulometric parameters and statistical decomposition methods. This topic is therefore of growing interest in Earth sciences and has been a major focus of sedimentary studies. Here, we present the results of unmixing grain size distributions from a loess-palaeosol section at the Sea of Azov, Russia, by jointly applying the standard deviation method and end-member modelling. The results indicate that the two methods can produce similar grain size decompositions but that end-member modelling has advantages in terms of quantitative and objective characteristics. In addition, three main loess subpopulations or end-members (EMs) with mode sizes of 8 μm, 18 μm and 32 μm, which represent distinct aerodynamic environments, are identified from the grain size distribution in the Azov region. Thereinto, EM1 with a mode size of 8 μm is the integrated result of combining atmospheric circulation with other environmental processes. EM2 with a mode size of 18 μm is inferred to represent continuous background dust under non-dust storm conditions. EM3 with a mode size of 32 μm is a fraction transported in short-term, low-altitude suspension clouds during dust storm outbreaks. Of the three EMs, EM1 and EM2 have multiple origins due to their complex formations, whereas EM3 is primarily derived from the alluvial plains of different rivers flowing into the Sea of Azov.

Introduction

Loess and loess-like sediments are widespread in Europe, composed primarily of silt-sized particles and formed predominantly by the accumulation of aeolian dust (Stuut et al., 2009, Smalley et al., 2011, Lehmkuhl et al., 2018). Their deposition and characteristics are integrated results of the interplay between sediment properties (e.g., grain size distribution, magnetic characteristics and mineralogical composition), environmental conditions (e.g., atmospheric circulation pattern, topography, vegetation cover) and post-depositional alteration (e.g., pedogenesis, physical and chemical weathering, aggregate formation). As these characteristics are closely associated with climatic factors, loess-palaeosol sequences are regarded as excellent terrestrial archives for palaeoclimate and palaeoenvironment reconstructions over the last million years in Europe (Marković et al., 2015, Lehmkuhl et al., 2016, Schaetzl et al., 2018).

However, deciphering such information that relates loess-palaeosol sequences to climatic factors in a quantitative way is indispensable for an appropriate interpretation of climate proxies. The grain size distribution, as an important sedimentary characteristic of loess sediments, is a powerful proxy for process identification and palaeoenvironment reconstruction (Vandenberghe, 2013). It is a valuable tool to analyse the composition of sediments and provides insights into distinguishing potential sediment provenances, different transport dynamics and specific depositional environments (Tsoar and Pye, 1987, Pye, 1995, Sun et al., 2002, Vandenberghe, 2013, Újvári et al., 2016, Vandenberghe et al., 2018). Hence, variations of loess grain size distribution and numerous granulometric proxies are extensively applied in Quaternary loess research to reconstruct palaeoclimatic changes on different time scales (Porter and An, 1995, An, 2000, Ding et al., 2002, Rousseau et al., 2002, Nugteren and Vandenberghe, 2004, Újvári et al., 2016), such as the mean, mode and median grain size, various fractions of grain size (An et al., 1991, Ding et al., 2002, Ding et al., 2005, Antoine et al., 2013), the U-ratio (Vandenberghe et al., 1997, Nugteren and Vandenberghe, 2004, Nugteren et al., 2004) and the grain size index (Rousseau et al., 2002, Rousseau et al., 2011, Antoine et al., 2009a, Antoine et al., 2009b).

However, as mentioned above, loess sediments are complex functions of many variables and are also mixtures of several subpopulations derived from multiple sources and/or transported by different mechanisms and subsequently deposited under specific sedimentary environments (Prins et al., 2007, Weltje and Prins, 2007, Vandenberghe, 2013). In addition, the grain size distributions of loess sediments have a quite narrow range as they are subjected to intense sorting in the sedimentation process (Tsoar and Pye, 1987). As a consequence, these fixed grain size ratios and defined relations of certain fractions are less informative and over-simplistic in sedimentary and palaeoclimatic interpretation, allowing only a general comparison of influencing processes that contributed to loess sedimentation (Újvári et al., 2016, Schulte et al., 2018, Varga et al., 2019). Deeper genetic information from the grain size distribution of loess sediments needs to be further deciphered by the application of more complex statistical approaches.

In general, there are two dominant methods that have been widely employed to identify the subpopulations composing the loess grain size: the standard deviation method and statistical decomposition analysis. The former method is primarily based on comparison of each grain-size class with its respective standard deviation for all grain size data, permitting the identification of the highest variability intervals of grain size along a whole sedimentary sequence (Boulay et al., 2003). The latter statistical decomposition approach, however, can be further divided into parametric decomposition methods, such as the Weibull-distribution fitting (Sun et al., 2002, Sun, 2004, Wang et al., 2017) and lognormal-distribution fitting (Qin et al., 2005, Lin et al., 2016), and non-parametric decomposition methods, such as the end-member modelling method (Weltje, 1997, Prins and Vriend, 2007, Dietze et al., 2012, Vandenberghe, 2013, Paterson and Heslop, 2015, Yu et al., 2016, Dietze and Dietze, 2019). Although theoretically different, these statistical decomposition approaches in practice produce similar results, leading to identical sedimentary interpretations (Liu et al., 2016, Dietze and Dietze, 2019, Varga et al., 2019). In this paper, we apply the standard deviation method and end-member modelling in combination.

Due to prevailing loess accumulation and unique sedimentary environments, loess-palaeosol sequences in Eastern Europe, especially the north-eastern Azov region of the East European Plain, are becoming one of the most promising areas for research on regional palaeoclimate fluctuations (Velichko, 1990, Velichko et al., 2011). Large numbers of studies based on multi-disciplinary approaches to loess-palaeosol sequences have been performed at the Sea of Azov (Velichko et al., 2009a, Velichko et al., 2009b, Matishov et al., 2013, Liang et al., 2016, Chen et al., 2018a, Chen et al., 2018b, Konstantinov et al., 2018, Panin et al., 2018). In this study, we present grain size distributions derived from the upper part of a loess-palaeosol sequence in the Azov region of the southern East European Plain and apply for the first time an end-member modelling method to decompose these grain size data. The main purpose of this paper is to quantify different subpopulations in the loess grain size distribution of the Azov region and to relate these identified subpopulations to potential sediment origins, possible sediment transport processes and different conditions of deposition or re-deposition.