Loess-palaeosol sequences (LPS) play an important role in reconstructing Quaternary climates and environments. They provide information on past climate conditions by comparison with Marine Isotope Stages (MIS) and pollen records (e.g. Antoine et al., 2019; Rousseau et al., 2020; Wolf et al., 2021) or by analyses of mollusc assemblages (e.g. Moine et al., 2005; Richter et al., 2020), on loess source areas (Ujvari et al., 2012; Ujvari et al., 2014), on past vegetation cover (e.g. Hatté et al., 1998, Hatté et al., 2013; Trigui et al., 2019) and former soil formation conditions (e.g. Buggle et al., 2011).

For such environmental reconstructions, precise chronologies are essential. Loess as windblown sediment is considered as ideal material for optically stimulated luminescence (OSL) dating (e.g. Fuchs et al., 2013; Lomax et al., 2019). However, quartz luminescence ages are often underestimated in high dose ranges, explained by an aberrant behaviour of the laboratory generated dose response curve, which restricts quartz OSL dating to around 50–65 ka in loess when using fine to medium sized grains (Chapot et al., 2012; Timar-Gabor and Wintle, 2013; Avram et al., 2020). Upper age limits in feldspars are much higher (e.g. Avram et al., 2020), but anomalous fading may lead to an unwanted loss of the infrared stimulated luminescence (IRSL) signal (Spooner, 1994) and thus to underestimated IRSL ages. In the recent decade however, feldspar luminescence protocols have been developed which decrease or even overcome the problem of fading in feldspar dating. These protocols use a first lower temperature IRSL which empties luminescence traps prone to fading, followed by a second or consecutive IRSL steps at elevated temperatures, which sample the luminescence of more stable traps (Thomsen et al., 2008; Buylaert et al., 2009; Li and Li, 2011, 2012; Thiel et al., 2011; Zhang and Li, 2020). This measurement procedure is usually termed post-IR-IRSL (pIRIR) dating. Because there seems to be a negative relation between first and second stimulation temperature and fading, and a positive relation between first and second stimulation temperature and residual signals (e.g. Kars et al., 2014), a careful consideration of these two problems needs to be carried out. Younger samples, which usually show smaller fading rates, but in which residuals can contribute largely to the palaeodose, are often measured using lower first and second stimulation temperatures (e.g. Reimann and Tsukamoto, 2012), whereas for older samples, the opposite is the case (Zhang and Li, 2020). For these reasons, there is so far no general protocol for feldspar samples, and therefore, a wide range of protocols and approaches exist and their application is dependent on the individual local geoconditions.

LPS in north-eastern Armenia provide excellent conditions for reconstructing palaeoenvironmental conditions. They contain multiple sequences of loess and pedocomplexes indicating that they represent several glacial-interglacial cycles. Furthermore, several tephra layers within the loess sections provide the possibility of reconstructing past volcanic eruptions and for using these tephra layers as chronological marker horizons. Studies on the chronology of north-eastern Armenian loess sections are rare so far. Wolf et al. (2016) present five pIRIR ages and two OSL ages of the loess section Sevkar, which is located close to the section Achajur studied in this paper. Together with studies of Trigui et al. (2019) and Richter et al. (2020) it is shown that during glacials, dry conditions promoted shrub-steppe and loess accumulation in the area, whereas interglacials were accompanied by forest- and high-grass-steppe and landscape stability with soil formation. The aim of this paper is to refine the chronology of LPS in the Sevkar loess area for the last glacial-interglacial cycle, to enable a correlation between the loess sections Sevkar and Achajur, and to test the reliability of the obtained luminescence ages. Therefore, three different protocols were applied on different mineral fractions, dating a total of 13 samples. The quartz fine grain fraction was measured using a 1) SAR protocol (Murray and Wintle, 2000; 2003) in order to provide reliable ages for the younger part of the loess section. Furthermore, two different pIRIR protocols, namely a 2) pIRIR-225 protocol (Buylaert et al., 2009) and a 3) pIRIR-290 protocol (Thiel et al., 2011; Li and Li, 2012), were applied to the polymineral fine grain fraction. The aim of this approach is to test which protocol yields the most reliable ages, especially for the older part (>50 ka) of the section, in order to apply this protocol to further sections from the Sevkar loess area. For evaluating the reliability of these polymineral feldspar ages, a comparison with the quartz ages from the younger part of the section, stratigraphic information and comparison with regional palaeoclimatic records is used as age control.

The Achajur loess section within the Sevkar loess area is located approximately 75 km north of Lake Sevan in north-eastern Armenia (Tavush region) (Figure 1). The section is part of a larger area of loess deposits stretching along the north-eastern foothills of the Southern Caucasus, which forms a NW-SE stretching boundary to the Kura depression. Some of the loess deposits of the Sevkar area form dune-like ridges and show a strong inclination of loess units and soil horizons (Wolf et al., 2016). Furthermore, most of the loess sections contain prominent tephra layers, derived from volcanoes in Armenia, Georgia and Turkey (Neill et al., 2013; Sherriff et al., 2019). The nearby Aghstev River drains the Southern Caucasus and joins the Kura River in Azerbaijan, which in turn drains into the Caspian Sea.

At present, annual precipitation in the study area ranges between 450 and 550 mm with a maximum in May and a minimum in winter. The mean annual temperature averages ca. 11°C (Wolf et al., 2016). The most common type of vegetation of the adjacent Kura Basin is steppe vegetation in lower areas, associated with patches of oak forests on the slopes of the Southern Caucasus.

The section Achajur is located at 41°0′17″N, 45°9′22″E and 891 m a.s.l. The composite profile is derived from compiling four sub-sections and reaches a total thickness of 24 m. It comprises three palaeosol complexes (P-3, P-2, P-1) and the Holocene soil (P-0), intercalated by the loess units L-3, L-2 and L-1 (Figure 2). This indicates that the section most probably reaches down to at least the third last glacial.

A short stratigraphic description deals with the section (Figure 2) from the bottom to the top: Above a blackish horizon that marks the upper part of pedocomplex P-3, sub-sequence 3 represents a succession of loess deposits (L-3), ending with the formation of pedocomplex P-2. Loess unit L-3 comprises another weakly developed brownish soil (L-3/S1). The lower part of pedocomplex P-2 (P-2/S2) exhibits a greyish-brown soil with large (∼10 cm) calcium-carbonate enrichments. Above at least two layers of relocated material, which are characterised by small rounded clay-pebbles (<6 mm), the upper part of pedocomplex P-2 (P-2/S1) reveals another strong dark-brown to reddish-brown soil containing big calcium-carbonate enrichments and concretions. Sub-sequence 3 ends with a greyish colluvial layer. Sub-sequence 2 is characterised by the deposition of thick layers of loess material (L-2) bearing features of strong relocation such as rock fragments or isolated clay-pebbles. Based on field observations, no prominent interstadial palaeosols seem to be present in this unit. A pale bluish tephra layer interrupts unit L-2, while a couple of relocated greyish tephra-bearing layers are admixed above. Weakly weathered palaeosols were found below the bluish tephra (L-2/S2) as well as in the upper part (L-2/S1). Sub-sequence 2 ends with pedocomplex P-1 that shows a succession of three well-developed palaeosols. While the lowermost palaeosol (P-1/S3) is characterized by a ∼60 cm thick strongly weathered brown horizon with large calcium-carbonate enrichments, the overlying palaeosols (P-1/S1 and P-1/S2) reveal a blackish color and crumbly structure and can be classified as chernozems. The uppermost sub-sequence 1 is characterised by a succession of loess deposits (L-1) with interposed pale greyish-brown palaeosols (L-1/S1-S4). At the top of the section, a reddish-brown soil (P-0/S2) with small calcium-carbonate concretions is present and is covered by a sequence of colluvial deposits.

For a further characterisation of the loess section, field magnetic susceptibility (MS) measurements were carried out using a portable Bartington MS2F probe (580 Hz), taking measurements at an interval of 5 cm. Often, MS values represent the degree of soil formation, with high MS values in palaeosols due to the formation of iron oxides during soil formation, and low values in pure loess (Evans and Heller, 1994). In the Sevkar loess area however, high values of magnetic minerals within the loess units, probably related to tephric material, hamper a clear distinction between loess layers and soil horizons (Wolf et al., 2016). Due to this, the loess layers can yield higher MS values than some of the palaeosols.

Eight samples for luminescence dating (Gi538-545) were taken in the L-1 loess layer, and three samples in palaeosols (Gi546 and Gi548) and intercalated loess layers (Gi547) of the pedocomplex P-1. Further two samples (Gi549 and Gi550) are derived from the younger part of the L-2 loess, above the bluish tephra layer at 15 m. At Achajur, it was not possible to take samples below the tephra, since these layers were heavily disturbed by erosional processes, and therefore, a stratigraphical assignment was not possible.

Table 5 summarises radionuclide concentrations and resulting dose rates for the quartz and the polymineral fine grain fraction. Table 6 and Figure 5 show all luminescence ages, obtained from quartz and from uncorrected polymineral fine grain D_e values. It is observed that between 35 and 55 ka, ages of the three protocols and mineral fractions agree very well. Assuming that the quartz ages are reliable in this age range, this argues for negligible fading of the polymineral fine grain ages for samples <55 ka, and for the correctness of the residual subtraction procedure of the polymineral fraction. The quartz ages are younger than the polymineral ages from ca. 60 ka onward, but agree within errors for samples up to ca. 85 ka. Sample Gi549 below the P-1 pedocomplex shows a discrepancy between the quartz and the polymineral fine grain fraction, most likely caused by an underestimation of the quartz fine grain fraction. Underestimation of the quartz fine grain fraction from ca. 60 ka onward have been shown for loess samples from Eastern Europe (e.g. Timar-Gabor et al., 2011, 2015a; Kreutzer et al., 2012a; Constantin et al., 2015). The reason for the underestimation is not yet totally understood, but is probably linked to differences in natural and laboratory generated growth curves (Chapot et al., 2012; Timar-Gabor and Wintle, 2013; Timar-Gabor et al., 2015b; Avram et al., 2020). All polymineral ages, measured with the pIRIR-225 and pIRIR-290 protocol agree within errors, and the mean ratio (pIRIR-290/pIRIR-225) between the two datasets is 1.03. This suggests that there is no significant difference in fading between the two protocols for our samples.

Results of the field magnetic susceptibility measurements are not standardized to mass or volume. However, a comparison with results from other loess areas measured with the same device shows that maximum MS values of the Achajur section are at least 10 times the maximum values in eastern German loess sections (e.g. Meszner et al., 2011) or 100 times the maximum values of loess sections in central Spain (Wolf et al., 2018). Most probably, this is due to a strong admixture of magnetic minerals derived from volcanic clastics such as tephra material or weathered volcanic bedrock.

The sequence of the L-2 loess gives a good impression about how the incorporation of volcanic material affects the magnetic susceptibility. Lowest values of ∼300 × 10⁻⁶ SI-units are linked to the bluish tephra (∼15 m), possibly indicating rhyolitic volcanism. Three meters above this tephra numerous thinner and darker tephra layers as well as volcanic rock fragments originating from neighbouring slopes are present. Magnetic susceptibility shows background values in the pure loess of about 1700 × 10⁻⁶ SI-units, while in loess with volcanic contaminations, values rise to ca. 2700 × 10⁻⁶ SI-units. This clearly illustrates the importance of the respective proportion of volcanic clasts within loess deposits for MS results. Furthermore, it shows that a simple scheme of low MS in loess and pedogenic enrichment of MS in palaeosols is not applicable to the Achajur section.

Considering the polymineral fine grain ages as reliable, the LPS Achajur shows a chronology in line with the expected stratigraphy, with the prominent pedocomplex P-1 falling into MIS5, and the overlying loess unit L-1 falling into MIS4 and 3. The MIS2 does not seem to be present at Achajur. The loess unit below the P-1 pedocomplex should probably fall into MIS6, according to stratigraphic expectations, which ends at around 130 ka (Lisiecki and Raymo, 2005). Likewise, at Lake Van in eastern Anatolia, full interglacial conditions assigned to MIS5 are established at ca. 129 ka, following an initial warming phase from ca. 131 ka onwards (Pickarski et al., 2015). The ages of sample Gi550, i.e. 117 ± 9 (pIRIR-290) and 123 ± 8 (pIRIR-225) underestimate this stratigraphic expectation, and they are also slightly younger than the overlying sample Gi549, which in turn agrees with the stratigraphic expectation. Possibly, the age discrepancy can partially be explained by an inaccurate estimation of the dose rate, since both samples and both protocols show agreeing palaeodoses. O’Gorman et al. (2021) for example showed that tephras can have very low internal potassium contents in IR-sensitive grains, e.g. due to a dominance of sanidine which is typical in rhyolitic rocks. The internal K-content of 12.5 ± 0.5% (Huntley and Baril, 1997) used for all our samples may thus be an overestimate in those samples with a large tephric contribution. Nonetheless, the effect on the pIRIR age would be small. For example, a theoretical internal K-content of 5% would only increase the age of sample Gi550 by 0.5%.

A further explanation could be a stronger rate of signal loss due to fading for sample Gi550 due to a thin (relocated?) tephra layer between sample Gi549 and Gi550. Since volcanic feldspar is known to fade to a higher degree than non-volcanic feldspar (Wintle, 1973), this tephra could indeed be responsible for a higher fading rate in sample Gi550. For the pIRIR-225 protocol, the sample showed a measured g-value of 1.8%, higher than for other samples for which g-values were measured. At this stage of research, we can only speculate on the underestimation of sample Gi550 and will thus treat the ages of this sample as most likely underestimated in the following.

With the resulting luminescence ages, the Achajur section provides a similar chronostratigraphy as the nearby LPS Sevkar, published by Wolf et al. (2016). They also present a “too young” age of 110 ± 17 ka for the loess unit below the P-1, and MIS4 to MIS3 ages for the loess unit above the P-1 palaeosol complex. Again, at this stage of research, we consider the age below the P-1 at the Sevkar section as true age underestimate. A similar chronostratigraphy is also recognised in LPS from northern Iran, with high loess deposition rates during MIS4, a three-fold palaeosol assigned to MIS5 and further loess deposition during MIS6 (Lauer et al., 2017; Kehl et al., 2021). In the latter two studies, a non-fading corrected pIRIR-290 protocol (Toshan loess section, Lauer et al., 2017) and a fading-corrrected pIRIR-225 protocol (Neka-Abelou, Kehl et al., 2021) were applied to polymineral fine grain samples. Similar to our results at Achajur, some age uncertainties, i.e. a potential age underestimation at Toshan and potential age overestimation at Neka-Abelou are noted for the MIS5 and MIS6 deposits. This demonstrates that–despite the huge advances made since the development of pIRIR protocols–dating of loess samples >100 ka still remains challenging. In contrast to the LPS at Achajur, the Iranian loess section Toshan also comprises units of significant loess deposition younger than 35 ka (MIS3-MIS2) (Lauer et al., 2017). At Neka Abelou however, the MIS2 is also not fully represented, with a youngest age of 28 ± 2 ka relatively near to the surface (Kehl et al., 2021).

In the following, the chronostratigraphy is compared with the well dated palynological records of Lake Van in Turkey (Litt et al., 2014; Stockhecke et al., 2014) and Lake Urmia in Iran (Djamali et al., 2008). Further comparisons are made with the studies of Richter et al. (2020) and Trigui et al. (2019), who reconstructed palaeoenvironments directly in the Sevkar loess area by gastropod analyses and leaf wax n-alkane analyses, respectively.

In conclusion, the LPS Achajur is generally in line with the expected global stratigraphy and with nearby pollen records of Lake Van and Lake Urmia. It indicates dry glacials promoting loess deposition and wetter interglacials and interstadials with soil formation. The LGM is not represented in the section, either due to prior deposition and later erosion or due to a lack of loess formation. With exception of the lowermost sample, which is located near an overlying tephra, polymineral fine grain samples measured with a pIRIR-225 and pIRIR-290 protocol do not seem to require a fading correction. Quartz ages older than 60 ka appear slightly underestimated, especially for the MIS6 loess layer. Future research will concentrate on the relation between soil and tephra layers and an explanation for the lacking LGM loess layers.