The Karamaidan loess section (Figure 1) is located at the northern edge of the Tajik depression, c. 40 km east of the Tajik capital Dushanbe, at the foothills of the Zarafshan mountain range, between the Tianshan range to the north and the Pamir Mountains to the east. The region is characterized by a mild temperate continental climate, with an average annual temperature of 10.8°C, peaking at 23.1°C in July, and the lowest average monthly temperature of −1.7°C in January (Fick and Hijmans, 2017). The region receives 730 mm of annual precipitation, which overwhelmingly falls in the spring months.

The Karamaidan site is a >100 m-thick loess-paleosol package that preserves semi-continuous accumulation of sediment since at least the Brunhes-Matuyama boundary (Forster and Heller, 1994; 1997; Zhou et al., 1995). The current profile was exposed by mass slumping detached by earthquakes in 1930 and 1943 (Gorshkov, 1935; 1949; Kuhtikova, 1966). The value of the site as a loess-paleosol sequence (LPS) has been known since Soviet times (Dodonov, 1991); subsequent studies described the pedo- and magneto-stratigraphy (Dodonov, 1991; Forster and Heller, 1994; Bronger et al., 1995; Shackleton, 1995; Bronger, 2003). More recently, variability in trapped-charge characteristics of quartz from the sequence was investigated as potential indicators of shifts in sediment source down the sequence (Dave et al., 2022). As yet, however, there have been no studies of the chemical composition or weathering intensity of the Karamaidan sequence; absolute dating of the sequence to confirm the proposed correlation with Quaternary glacial-interglacial cycles (Shackleton, 1995) is likewise lacking.

In this study, we examine the uppermost 16 m of the Karamaidan LPS. This subsection spans the near-surface Holocene to modern soil, primary loess intersected by a weak soil, and a well-developed soil at the base, which has been interpreted based on magnetostratigraphic correlation to span the last full glacial cycle from oxygen isotope stages (OIS) 1–5 (Forster and Heller, 1994). We focus on this subsection since it covers the greatest variability in geochemical weathering intensities possible in the smallest possible depth range, and this time period can be more reliably dated directly using luminescence techniques.

Since the profile is exposed as a cliff-face, sampling took place by abseil. Stratigraphic and sampling depths were continuously checked with respect to a datum fixed at the top of the profile. The profile was excavated back to avoid sampling of sediment washed down or blown onto the surface. Tools used for sampling were cleaned before the collection of each sample to avoid cross-contamination. We undertook continuous sampling at 10 cm resolution, starting at 0.80 m depth to avoid recent reworking. We collected c. 300 g of bulk material in airtight plastic bags for multiple analyses, 153 bulk samples in total. We collected samples for laboratory analysis of magnetic susceptibility at 10 cm intervals in 8 cm³ plastic cubes from 0.8–16 m depth. Sampling for luminescence dating was undertaken in duplicate at 1 m intervals using stainless steel light-proof tubes (4 cm diameter, 10 cm length); an additional c. 300 g was collected from surrounding the sample tubes for dose-rate analysis.

We subsampled the bulk sediment samples in the laboratory for the X-ray fluorescence (XRF) elemental and loss on ignition (LOI) analyses. We took approximately 10 g of material from each bulk sample bag, taking precaution to avoid cross-contamination between samples by cleaning work surfaces with demineralized water and then isopropyl alcohol, covering the work surface with aluminum foil heated to 100 °C overnight, and using nitrile gloves and sterilized spatulas.

Magnetic susceptibility (MS) measurements are widely used in loess-palaeosol studies to assess pedogenesis, weathering intensity, sediment sources, and rainfall, making it a valuable tool for reconstructing environmental and paleoclimatic changes (e.g., Banerjee et al., 1981; Maher, 1986; Kukla et al., 1988; Maher and Thompson, 1995; Dearing, 1999; Antoine et al., 2001; Balsam et al., 2011). Mechanisms and paths by which soils show MS enhancement are related to enrichment of fine-grained ferrimagnetic and paramagnetic minerals by a wide range of pedogenic processes. At Karamaidan Forster and Heller (1997) specifically documented that MS susceptibility enhancement along the loess-paleosols profile is controlled in a linear fashion by the increasing amount of ferrimagnetic minerals and clay mineral content (paramagnetic phase), indicating that MS is a reliable proxy for weathering in our study.

We measured the magnetic susceptibility of the Karamaidan sediments both in situ and in the laboratory. In the field, we used a Bartington MS3 meter combined with an MS2K field sensor and Trimble Nomad minicomputer. The Bartington probe was positioned directly on the cleaned profile every 5 cm down profile. We compensated for thermally induced drift due to environmental temperature variations by calibrating each measurement individually to air temperature. In the laboratory, we measured the volume-specific, low-frequency magnetic susceptibility at the Laboratory of Paleomagnetism of Universidade Estadual Paulista (UNESPmag, Rio Claro, Brazil) using a Bartington MS3 meter equipped with a MS2B sensor set at 4.6 kHz.

Samples for XRF chemical analysis were first air-dried at 40 °C and then homogenized by sieving and splitting. We then ground the samples in a planetary ball mill in agate jars at the Max Planck Institute for Chemistry (Mainz, Germany).

XRF analyses were performed at the Institute of Geosciences, Johannes Gutenberg University (Mainz, Germany). XRF analysis was undertaken using a MagiXPro spectrometer from Philips and Panalytical, containing an Rh tube as X-ray source with a maximum power of 4 kW (60 kV, 126 mA). We prepared glass beads for XRF measurement by diluting the sample with Li-tetraborate (Spectromelt A10, Merck) dry flux at a ratio of 1:13 (0.4000 g ± 0.0050 g sample and 5.2000 g ± 0.0050 g dry flux). The flux had previously been dried at 450°C for 2 h and placed in a desiccator for cooling. The samples were mixed and homogenized in porcelain crucibles with a glass stirring rod, then poured into platinum crucibles using the resonance properties of porcelain to minimize material loss. We then melted the samples at approximately 1,050°C with an automatic Vulcan fan burner (HD Elektronik und Elektrotechnik GmbH) and poured the melt into platinum molds to form glass beads. Element concentrations are expressed as a weight percentage (wt%) of the respective oxides and were recalculated on a volatile-free basis. From our geochemical data (Supplementary Material SM-1), generated as element oxides (wt%), we calculated the 12 WIs (Table 1).339457720088

We determined the LOI in the same laboratory as the XRF analyses, heating approximately 0.5–1.0 g of subsample in a muffle furnace at 1,000 °C ± 20 °C for 3 h. Sample mass was precisely measured to 4th decimal places before and after heating.

At the LABECA (Cultural and Environmental Heritage Laboratory) of the University of Pisa, we determined the CaCO3 content by CO2 measurement using the gasometric method (Leone et al., 1988), from which we calculated calcite-bound CaO (CaO calcite) and silicate-bound “corrected CaO” (CaO*).

Samples were prepared for luminescence dating under red-light conditions at the Institute of Geosciences, Johannes Gutenberg University, Mainz (Germany). Due to the lack of sufficient >63 µm size grains, we prepared fine-grained (4–11 µm) quartz from all 12 samples. We treated the <63 µm sieved sediment fraction with 10% HCl and 10% H₂O₂ to remove carbonates and organics respectively, after which we washed the samples with 0.1 N sodium oxalate to remove clays. Each successive chemical pre-treatment step was followed by multiple washing with deionized water. The fine-grained (4–11 µm) size fraction was settled out using Stokes Law. We then etched the 4–11 µm polymineral fraction with 37% hexafluorosilicic acid for 7 days, followed by a wash with 10% HCl, to obtain the fine-grained quartz fraction. This material was then suspended in water and placed onto 10 mm diameter stainless steel cups to obtain a uniform layer of fine-grained quartz for measurement of equivalent dose.

Equivalent-dose (D_e) measurements on fine-grained quartz were made at the Max Planck Institute for Chemistry (Mainz) on an automated Risø TL-DA-20 reader with ⁹⁰Sr/⁹⁰Y beta source (Thomsen et al., 2006) and EMI 9235QA photomultiplier tube. We undertook optically stimulated luminescence (OSL) dating of fine-grained quartz on 20-24 aliquots for all samples using the Double-Single aliquot regenerative (DSAR) protocol (Banerjee et al., 2001; Jain et al., 2005; Supplementary Table SM-2). All pIR290 De-measurements were conducted on a Riso TL/OSL DA-20 automated reader at the Max Planck Institute for Evolutionary Anthropology (Leipzig). The feldspar-signal stimulation was delivered with IR light-emitting diodes transmitting at 870 nm (145 mW/cm²) and a Schott BG-39 and Corning 7-59 filter combination was used to detect in the blue-violet wavelength region. For beta-irradiation a calibrated ⁹⁰Sr/⁹⁰Y beta source with a dose-rate of ∼0.12 Gy/s was used. We describe the measurement protocols in more detail and provide data from the preliminary tests and quality controls of the DSAR protocol in the Supplementary Material SM-2).

Studies on the saturation behavior of fine-grained quartz suggest that this grain-size fraction does not saturate at high doses under laboratory conditions, and therefore tends to underestimate the true depositional age of samples beyond c. 60 ka (Timar-Gabor et al., 2017; Dave et al., 2023a). Therefore, for such samples we additionally measured the D_e for fine-grained polymineral aliquots using the elevated temperature post-Infrared-Infrared-Stimulated-Luminescence (pIRIR) signal, to circumvent the issue of anomalous fading (Thiel et al., 2011). Additional information on these measurements are provided in the Supplementary Material.

We evaluated the dose rates for the fine-grained quartz and polymineral samples from radionuclide concentrations of U, Th and ⁴ K, determined using high resolution germanium gamma spectrometry performed at the Felsenkeller, VKTA in Dresden, Germany, converted into beta and gamma dose rates using published rates (Guérin et al., 2011). Dose-rate attenuation by moisture was assumed to be 10% ± 5% for all samples following published average water content values for Central Asian loess (Fitzsimmons et al., 2017; 2018; Dave et al., 2023a). An alpha efficiency value of 0.038 ± 0.002 for fine-grained quartz and 0.086 ± 0.004 for polymineral fine grains was used for all dose rate calculations (Rees-Jones, 1995). The cosmic-ray dose rate contribution was calculated from published formulae (Prescott and Hutton, 1994). A detailed account of the luminescence dating protocols can be found in the Supplementary Material SM-2.

Our magnetic susceptibility (k) results are illustrated in Figure 2. In situ measurements, which were made at 5 cm resolution, are shown as a solid red-brown line. Lab-measured k values, measured at 10 cm resolution, are provided as white dots. We observe substantial overlap of field and lab-derived k-values in the upper part of the profile and a divergence in values below 11.45 m in pedocomplex PS2. Laboratory values are slightly higher than field measurements, particularly in the depth range 12.90–13.60 m. The increase in k-values measured in the field is inversely correlated with carbonate content. Since the in situ measurements were made at higher resolution, all subsequent reference to magnetic susceptibility values derives from this dataset.

We observe a slight decrease in k-values in the HS down to PL1 at 2.0 m. This likely reflects an increase in recent oxidation and weathering which increases magnetic susceptibility; values decrease into the zone of calcareous enrichment below 80 cm depth. Magnetic susceptibility remains relatively constant within PL1 down to 11.45 m; we observe no appreciable change in k-values within the weakly developed paleosol PS1. We notice a decreasing trend of carbonate content within the sediments from 1.30 m down to 7.50 m, which is not reflected by k values. Magnetic susceptibility intensifies approximately 50 cm above the strong palaeosol contact as observed by eye at 11.45 m, which corresponds to a carbonate-poor layer. Susceptibility then decreases down to c. 13.00 m to k-values like those of primary loess. K-values increase from 13.0 to 13.75 m and likely correlate with the clay enrichment observed at this depth. Magnetic susceptibility remains high down to 14.70 m, with peak k-values occurring at the same depth as clayey horizons at 14.70 and 15.10 m. Below 14.70 m, k-values decrease to intermediate values, corresponding, according to field observations, to a horizon containing less clay and iron oxides than the one immediately above it.

Volatile components such as water and carbon dioxide are found in Karamaidan loess sediments in the form of carbonates, organic matter, and hydrated mineralogical phases (e.g., clays). The presence of calcium carbonate is particularly interesting when studying the weathering of loess since it is typically enriched within the lower “C” horizons of semi-arid palaeosols (Sprafke and Obreht, 2016). In this study, we measure element oxides (XRF), LOI and CaCO₃ (gasometry) and from these we calculate CaO (calcite-bound), and CaO* (silicate-bound) stoichiometrically. Table 2 summarizes these results (see SM-1 for the complete dataset).

Our measurements of LOI indicate an average value (μ) of 12.31 ± 3.03 wt%. The highest LOI values, reaching up to 21.38 wt%, are found within the PS2 pedocomplex at a depth of 12.20–13.40 m. We also observe LOI values up to 16.45 wt% in the transitional area between HS and PL1, spanning approximately 0.80–2.50 m. The maximum LOI values decrease in the following order: PS2>HS>PL1>PS1>PL2. Conversely, the lowest LOI values are found in PS2 (5.41 wt%) and increase in the following order: PS2<PL2<PL1<HS<PS1. This highlights the variability of LOI within the pedocomplex PS2 (see Figure 3).

Curves with similar trends are found in the results for the element oxides (Figure 2). The suggested concentrations of element oxides, obtained using XRF, also display variability between the primary loess and paleosol subhorizons. The average measured carbonate content (μCaCO₃) at Karamaidan is 15.53 ± 6.64 wt%. The highest and lowest values of μCaCO₃ are both found within the PS2 pedocomplex, measuring 0.07 and 0.82 wt% respectively. PS2 exhibits the highest average values for most element oxides, while HS has the lowest values, although there are some exceptions. For example.

• μMgO is highest in PL2 (>PS1>PL1) and lowest in HS (2.89 ± 0.08 and 2.36 ± 0.06 wt% resp. ly);

When compared with the Global Average Loess values (GAL (Újvári et al., 2008)), the samples from Karamaidan exhibit lower SiO₂ values (51.92 ± 4.80 vs. 70.71 wt%, a decrease of approximately 27%) and higher CaO values (11.24 ± 4.24 vs. 6.67 wt%, an increase of approximately 68%). The values of Al₂O₃ (12.38 ± 1.31 wt%), Fe₂O_3(T) (5.84 ± 0.46 wt%), MnO (0.10 ± 0.01 wt%), and MgO (2.69 ± 0.22 wt%) are above the GAL values. Na₂O (1.30 ± 0.11 wt%) is approximately 23% below the GAL values, while K₂O (2.20 ± 0.23 wt%) and TiO₂ (0.65 ± 0.07 wt%) are slightly below average (a decrease of around 1%–18%). In general, the soils (Holocene + paleosols) at Karamaidan exhibit higher values of SiO₂ (53.20 vs. 50.80 wt%), Al₂O₃ (12.86 vs. 11.97 wt%), and Fe₂O_3(T) (5.34 vs. 4.96 wt%), and lower values of MgO (2.58 vs. 2.79 wt%), Na₂O (1.26 vs. 1.34 wt%), and CaO (10.00 vs. 12.31 wt%) (SM-1).

The results of our chemical analysis are summarized in bivariate x/y diagrams (Harker, 1909) (Figure 4A–F) in order to visualize compositional changes related to weathering and pedogenesis (e.g., Újvári et al., 2008; Schatz et al., 2015). The Al₂O₃/SiO₂ diagram shows a positive linear relationship between the two components irrespective of stratigraphic unit (Figure 4A). Samples from the Holocene/recent and weakly developed soils (HS and PS1) cluster in the same region as the primary loess (PL1 and PL2), with the latter trending slightly towards PS2. Samples from the strongly developed soil pedocomplex PS2 yield a wider range of values and generally lower SiO₂ values than the other units as aluminum content increases. Similar trends and relationships are observed in the Fe₂O_3(T)/TiO₂ and TiO₂/SiO₂ diagrams (Figures 4C, D). The Karamaidan samples, and particularly those from PS2, generally yield lower SiO₂ but higher Al₂O₃ values than GAL with values closer to those of the Upper Continental Crust (UCC). Both iron oxide and titanium oxide values are generally higher at Karamaidan than both GAL and UCC.

Basic elements (Σbases = Ca+Mg+Na+K) are displayed in Figure 4E and indicate an inverse linear relationship with TiO₂. The PS2 pedocomplex is more depleted in the bases than the other stratigraphic units and yields similar or slightly lower values than both UCC and GAL. By contrast, the graph K₂O/Al₂O₃ (Figure 4B) indicates a positive linear relationship but two distinct trends, one for the primary loess and more weakly developed soils (HS, PL1, PS1, and PL2), and another for the PS2 pedocomplex. This trend is even clearer in the K₂O/Al₂O₃ vs. Na₂O/Al₂O₃. plot (Figure 4F) in which PS2 samples result in lower ratios values than HS, PL1, PS1, and PL2.

In Figure 5 we plot the 12 selected WIs calculated from our XRF data. We calculated Pearson correlation coefficients (r) for the different WIs (Table 3), and for WIs vs. Magnetic Susceptibility (Table 4). Certain weathering indices (WIs) exhibit contrasting calculation methods, whereby a higher weathering intensity corresponds to a lower numerical value, while others follow the opposite pattern (Table 4). In general, we notice consistent patterns across the WI plots. However, there are two indices, CPA, and FENG, that show lower correlation values (r < 0.72). Interestingly, all WIs reach their highest points of weathering intensity within the PS2 pedocomplex. Additionally, the zones of carbonate enrichment show a decrease in weathering intensity. It is also worth noting that there is no increase in weathering intensity within the weakly developed soil PS1 for any of the WIs calculated. Magnetic susceptibility likewise increases only in PS2 and HS, and not in PS1.

The results of the equivalent dose measurements, dose rate and resulting luminescence ages from fine-grained quartz and polyminerals are presented in Table 5. The combined quartz OSL–polymineral pIRIR luminescence chronology is illustrated against depth and stratigraphy in Figure 2 and compared against other published sites in Figure 6. Our quartz OSL chronology indicates ages between c. 11 to 35 ka within the depth range 1–8 m. Below 8 m, our preliminary pIRIR ages suggest ages exceeding 60 ka. Of the 12 samples dated using luminescence, eight from the top 8 m were dated using fine grained quartz. The quartz OSL signals in these samples exhibit a fast exponential decay suitable for dating. Below 8 m depth, we observe saturation of the fine-grain quartz signal (See Supplementary Material S2 and Table 5 for more details) consistent with fine-grained loessic quartz of this age elsewhere (Timar-Gabor et al., 2017).

We measured the pIRIR signal from polymineral fine-grain to determine the depositional age for samples from below 8 m depth. Investigations of the elevated-temperature pIRIR protocol on feldspar yield low fading rates (Buylaert et al., 2008; Thomsen et al., 2008; Thiel et al., 2011). We argue that since our samples were obtained from aeolian deposits which experienced ample exposure to sunlight prior to burial, that our samples are well bleached and therefore that residual signal should be negligible in our samples. We observe significant uncertainties in our pIRIR ages, particularly for the samples at 10 and 11 m depth. The cause for such large uncertainties remains unknown and further tests are required to identify the likely cause of such uncertainties and reliability of the ages. These tests are ongoing and beyond the scope of this study; we herewith consider our pIRIR chronology as a preliminary framework only.

There is an increasing balance of evidence indicating spatial variability in loess stratigraphy relating to spatial differences in accumulation rates and palaeo-topography, particularly in piedmont regions (e.g., Fitzsimmons, 2017; Dave et al., 2023a), although in aggregate it may still be possible to make regional correlations and identify “representative” type sections (e.g., Marković et al., 2015a). Even over scales of tens to hundreds of meters, variability in stratigraphy may be observed (e.g., Fitzsimmons et al., 2013; Fenn et al., 2021). Furthermore, although the thick loess sequences exposed in the Tajik Depression have long been recognized for their potential as archives down to at least the Brunhes-Matuyama boundary (e.g., Dodonov et al., 2006; Yang et al., 2006, Jia et al., 2018; Dodonov et al., 2006, Lü et al., 2020), few sections have been investigated in high resolution and with robust dating over the last full glacial cycle. It is therefore relevant to place the last full glacial cycle stratigraphy of the Karamaidan LPS in the context of previous studies and other sites in the Tajik Depression before proceeding with our interrogation of indications of weathering intensity through time.

In Figure 6 we compare and correlate our (chrono-)stratigraphy at Karamaidan with published work for the last full glacial cycle. The stratigraphy of the Karamaidan LPS which has previously been described (Dodonov, 1991; Bronger et al., 1995) was investigated on a part of the section c. 450 m north of where we worked. We note that the weakly developed paleosol observed in our study (PS1) was not noted by Bronger et al. (1995), and the HS is much thinner at the northern locality, although the depth of the strong PS2 paleosol is similar between the two sites. Nearby sites with published stratigraphy and/or chronologies spanning the last full glacial cycle include Darai Kalon, c. 73 km southeast of Karamaidan (Dodonov et al., 2006), Tian 2021-Fakhrobod, c. 55 km SW of Karamaidan (Tian et al., 2021), KP, c. 27 km west of Karamaidan (Yang et al., 2006) and Hoalin, c. 63 km south east of Karamaidan (chronology but not stratigraphy available: Wang et al., 2016). Apart from Hoalin, where no stratigraphic description is published, and KP, where only 5 m was exposed and no subsurface paleosol was identified, all studies describe a strongly developed pedocomplex comparable with our PS2, at upper depths varying from 11.50 m (Bronger et al., 1995), c. 18.50 m (Dodonov et al., 2006), and c. 16 m (Tien et al., 2021). Our weak paleosol (PS1) is reported only by Tian et al. (2021), at a slightly deeper depth than at Karamaidan. Our identification of PS1 is based on field observations of change in coloration, supported by geochemistry (an increase in CaCO₃ and decrease in SiO₂, Al₂O₃, K₂O, Na₂O, Fe₂O_3(T), TiO₂; Figure 3) although it is not distinguishable by magnetic analyses. The presence of a weakly developed and barely distinguishable paleosol at some but not all localities in the Tajik Depression suggests potential factors such as regionally discontinuous pedogenesis or accumulation, subtle amelioration of climatic conditions influencing weathering and pedogenesis, or variations in the level of detail in observations at specific sites.

There are multiple ways to assess the degree of rock and sediment weathering, some of which have been successfully applied to LPS. In Figure 4 we investigate compositional changes due to pedogenesis using Harcker diagrams. In Figure 4 a, c and e, we observe the relative enrichment of both Al and Fe during pedogenesis (e.g., Schatz et al., 2015), whereby alkali and alkaline-earth metals such as Na, K, Ca and Mg (Σbases) are mobilized through the breakdown of feldspars. The depletion of alkali metals can be seen most clearly in Figure 4F, which shows the mobilization of more labile components, typically derived from plagioclase and K-feldspar, with increasing chemical weathering. In Figure 4, it is evident that the Karamaidan samples exhibit a decrease in K and Na concentrations when compared to UCC TiO₂. Additionally, the higher concentrations of Al₂O₃ and Fe₂O_3(T) observed at Karamaidan indicate a more pronounced weathering intensity in the Tajik samples when compared to the Tokaj and Tian-Fakhborod sites included in our comparative analysis. In Figure 7B we observe distinctions in K content between loess and paleosol sediments for both Tokaj and Karamaidan in the K₂O/Al₂O₃, plot. By contrast there is a greater similarity between loess and paleosol chemistry at Tian-Fakhrobod, which may suggest weaker weathering and pedogenesis at that site.

The degree of chemical weathering is additionally illustrated using a ternary A-CN-K diagram (A=Al₂O₃, CN=CaO* + Na₂O, K=K₂O, Figure 8A), (Nesbitt and Young, 1984). Plotted against the Post-Archaean Shale (PAAS, Taylor and McLennan, 1995)–the reference composition for weathered continental sediments–the Karamaidan sediments are clearly less weathered. Our samples trend almost parallel to the A-CN line, which follows the meteoric alteration trend of plagioclase (Nesbitt and Young, 1984), irrespective of loess or paleosol attribution. Our data reflect progressive Al enrichment and reduction of Ca and Na with increasing weathering intensity. Interestingly the Karamaidan samples terminate below UCC values, unlike other LPS where values lie between UCC and PAAS; this could be related to compositional variations due to the heterogeneity of the source material and will be addressed in future studies. In Figure 8B, the A-CNK-FM diagram (A=Al₂O₃, CNK= CaO* + Na₂O+ K₂O, FM = Fe₂O_3(T) + MgO) includes the mafic mineral component of the system (Nesbitt and Young, 1989), and suggests that samples from the PS2 pedocomplex are depleted in Fe and Mg compared with the primary loess (PL1 and PL2) and PS1. The less well-developed HS soil lies midway between the two series and may support the hypothesis that PS1 represents an early or very weak stage of pedogenesis.

When we review the WI comparison (Figure 5) we observe that all 12 WIs indicate an increase weathering degree in the pedocomplex PS2, first noticeable c. 50 cm above the paleosol contact. This trend agrees with the χ variation, with reduced magnetic susceptibility in more highly calcareous layers. We observe a series of weathering “spikes” in the meter below the weakly developed PS1, in the WIs which are calculated using CaO*. The WIs correlating closely with each other (Tables 3, 4), except for the FENG and CPA indices, although these two correlate well with each other. Consequently, most of the WIs are interchangeable in any assessment of weathering intensity at Karamaidan.

As suggested by Schatz et al. (2015), we compare and correlate different weathering indices with a direct weathering proxy to determine the most sensitive index for variations in weathering intensity. While both low field magnetic susceptibility (χ) and grain size distribution are commonly used as indicators of weathering intensity (e.g., Marković et al., 2008; Buggle et al., 2009; Terhorst et al., 2014; Schatz et al., 2015), only magnetic susceptibility can be directly linked to weathering (e.g., Forster and Heller, 1997) and has been used in this study to evaluate the weathering indices (Table 4). We find that the FENG (r = 0.92) and CPA (r = 0.84) WIs provide the best correlation with magnetic susceptibility at Karamaidan. This result echoes that found by Schatz et al. (2015) for the Tokaj section in Hungary, and therefore we consider first FENG, then CPA, as the most appropriate WI for our site.

The FENG index suggests that zones of lower values, which are generally richer in carbonates, indicate areas that have undergone a less leaching, or accumulation of carbonates leached from overlying sediment. Higher FENG values correspond to more intensely leached sediment which contain less carbonate (Feng, 1997). At Karamaidan, FENG values decrease downwards from the HS soil to the primary loess unit PL1. The lowest values occur within PL1 (c. 5.5 m). The highest FENG weathering intensity occurs just above PS1 (at c. 7 m) and in PS2 (at c. 12 m, 13.6 m, and 15 m), in zones of carbonate leaching (Figure 2).

One of the major ambitions of Quaternary science is the quantitative reconstruction of palaeoclimate parameters. This is certainly the case in investigations of loess; only recently have tools become available to achieve this end, in the form of organic biomarker geochemistry (Peterse et al., 2014; Schreuder et al., 2016), clumped isotopes in pedogenic carbonates (Újvári et al., 2019; Prud’homme et al., 2021), stable isotope geochemistry from earthworm calcite granules (Prudhomme et al., 2021; Prud’homme et al., 2018), and D47 composition of mollusk shells as a quantitative temperature proxy (Újvári et al., 2021). A less tested approach for loess sediments is the use of transfer functions based on elemental WIs. Such transfer functions require calibration to regional conditions and are widely used in palaeoclimate reconstruction in soils outside of the loess realm (Jiamao et al., 1996; Sheldon et al., 2002; Maher et al., 2003; Retallack, 2005; 2009; Sheldon, 2006; Nordt et al., 2007; Sheldon and Tabor, 2009; Gallagher and Sheldon, 2013; Wang et al., 2013). However, the application of WI-derived transfer functions to loess sediments has been met with skepticism since algorithms are based on soil weathering indices that are fundamentally affected by the complex interplay between soil, climate, biota, parent material chemistry and time, and which are poorly understood (Ujvari, 2014). Early attempts to make use of these transfer functions in loess made use of calibrations from regions far removed from the likely environment in which the loess accumulated (Schatz et al., 2015). More recently, however, Wang et al. (2023) developed new transfer functions for mean annual precipitation (MAP) and temperature (MAT) in loess derived from calibration datasets from the Chinese loess plateau and surrounding regions.

Here we investigate the applicability of the Wang et al. (2023) transfer functions to the Central Asian context, and compare these results against “global” MAP (Sheldon et al., 2002) and MAT (Sheldon, 2006; Gallagher and Sheldon, 2013) transfer functions, with respect to present day values (Figure 9). The present-day climate in the Karamaidan area, sitting above 1,600 m, is semi-arid to dry subhumid continental climate, with a notable temperature range, with warm summers and cold winters, and moderate precipitations. The average annual temperature of Karamaidan is approximately 10.8 °C (with a monthly minimum of −1.7 °C in January and maximum of 23.1 °C in July), and the mean annual precipitation is approximately 730 mm (with a monthly minimum of 3 mm in August and maximum of 143 mm in April) (Figure 9B).

In Figure 9 it is evident that the Sheldon (2006) MAP values correlate closely with one another (SM-3), whereas the Wang et al. (2023) reconstruction produces an entirely different trend. The latter shows less variation in the upper portion, down to a depth of c. 11 m with an average precipitation of c. 590 mm. At depths greater than 11 m, in PS2, the Wang transfer function shows a significant increase with an average of 665 mm (and a peak of c. 720 mm), without particularly reflecting fluctuations caused by stratigraphic variations, unlike the other MAP functions that suggest substantial fluctuations in rainfall in PS2, indicating variations from extremely less rainy periods (down to c. 350 mm) to extremely rainy periods (over 800 mm). This trend seems at odds with the stratigraphic, geochemical, and magnetic susceptibility results, which suggest a substantial and persistent increase in weathering during the PS2 pedogenic phase - presumably linked to higher precipitation - compared to other periods. Apart from the marked variations in PS2, the mean values calculated for the three Sheldon MAP functions show averages of c. 515 mm in the first 11 m depth and c. 611 mm in the lower portion (PS2 pedocomplex). Lower values then, but not far from those calculated with Wang. Today’s average annual rainfall values at Karamaidan are around 730 mm (Figure 9 a and b, WorldClim v.2 Data; Fick and Hijmans, 2017) and are generally higher than those calculated, indicating a possible gradual but distinct change in the rainfall regime during the LGM. As far as MAT is concerned, we can observe in Figure 9A a wide variation in absolute value of the calculated values with the different functions between the different curves, although each curve presents minimal variations in temperature in almost all cases. Greater variations in the calculated temperatures are recorded for almost all curves (Wang and XRF2-MAT, albeit with inverse trends to each other) at c. 11 m depth, suggesting a marked change in the local climatic regime. The mean temperature values according to Sheldon’s functions are generally higher (XRF1-MAT c. 15°C; XRF2-MAT c. 22°C; XRF3-MAT c. 12.5 °C) than those by the Wang’s function (c. 11°C). In general, XRF2-MAT shows higher temperature values up to 11 m and then shows slightly lower values in PS2 (average values go from 22.7°C to 22.1°C); while for the other functions they all show lower values in the upper portion of the profile and varying degrees of reduction in PS2. In particular, Wang shows in the upper portions of the profile, more recent, temperatures comparable to today’s for Karamaidan (10.8 °C present day MAT measured data vs. c. 11°C calculated according to Wang). This might suggest a higher reliability, at least for estimating MAT, of the values calculated using Wang’s function. Such a hypothesis seems confirmed considering that MAT according to Wang is obtained from the WI CPA, which we have already assessed as well suited to describe Karamaidan. Observing Wang’s MAT values for the lower portion of the PS2 profile we note an increase in temperatures with an average c. 12 °C in the loess and max. c. 13°C in the palaeosol, an increase consistent with the milder interglacial conditions already proposed by other authors (Bronger, Shackleton) as being responsible for the pedogenesis of this stratigraphic unit.

The Wang et al. transfer functions demonstrates a more realistic trend in precipitation and mean annual temperature in Central Asian loess-paleosol sequences compared to alternative transfer functions. This is supported by observed correlations with other lines of evidence, indicating its credibility in capturing relative changes in climate. However, it is important to acknowledge that the transfer function may not accurately estimate absolute present-day values in the study area. Discrepancies between the reconstructed values and actual climate data highlight potential limitations in absolute value estimation. These discrepancies may stem from various factors, including the complex interactions of climate, soil, biota, parent material chemistry, and time, which influence the algorithm’s weathering indices. Further scientific validation and investigation are necessary to assess the function’s broader applicability and its ability to provide accurate absolute values in other regions. Gathering robust data from diverse sources will be crucial to refine and validate the algorithm’s performance, enabling a comprehensive evaluation of its realism and suitability for broader usage in Central Asian loess-paleosol sequences.

Our luminescence-based chronostratigraphy strongly suggests that at Karamaidan, the PS2 pedocomplex formed during the OIS 5 interglacial, the weak PS1 paleosol during the OIS3 interstadial, and the two primary loess units accumulated during the cold OIS2 and OIS4 stadials. Our WI results indicate a higher degree of weathering responsible for developing the PS2 pedocomplex, which we can now confirm as representing the OIS5 interglacial phase. Finally, our quantitative reconstruction of palaeoclimate parameters MAP and MAT, based on the loess-relevant Wang et al. transfer functions derived from the CPA index which we have established is suitable to the Karamaidan context, gives us the first insight into past climate conditions through time in the Tajik Depression.

Based on the OSL dating and the results of the WIs and soil paleoclimate transfer functions, the PS2 pedocomplex at Karamaidan likely developed during OIS 5. This pedocomplex exhibited higher mean temperatures (approximately 13°C), but lower precipitation (665 mm) compared to present-day mean annual conditions (10.8°C and 730 mm). The temperature fluctuations down the profile appear to be around ± c. 1.00°C during stadials and interstadial, while temperatures during interglacial OIS 5 increase of c. 2°C. Precipitation varies along the profile by about ±50 mm on an average of about 600 mm per year during the stadial/interstadial phase, while during OIS 5 it increases by about 100 mm on the same average. The subsequent deposition of primary loess at Karamaidan was characterized by minimal weathering and pedogenesis, and increasing accumulation rates, as indicated by the presence of “Sommerprossen” (small, roundish or irregularly shaped spots or patches of a different color) increasing upward in the profile. The overall accumulation reached approximately 9 m over the age range of approximately 30–70 ka. During the deposition of the primary loess, the calculated MAT was approximately 11°C on average, which closely aligns with the present-day MAT of 10.8°C. However, MAP was estimated to be around 600 mm, about 130 mm lower than the present-day recorded values of 730 mm and 120 mm lower than calculated precipitation during PS2 accumulation/pedogenesis. Consequently, primary loess deposition at Karamaidan experienced a reduction in both temperatures and precipitations by approximately 15% and 17% respectively, compared to the preceding PS2 pedocomplex. These conditions appear to contradict the hypothesis that there was during the OIS 3 an increase in precipitation favored by the northward expansion of the Asian monsoon and glacial expansion in the Tian Shan, but the spatio-temporal inhomogeneity and asynchrony of these events has recently been emphasized by other authors (Dave et al., 2023a). Of course it might also mean that the Wang’s transfer functions, being based on data recorded form Chinese LPSS, are not sensitive enough for the local context. The weak paleosol PS1, likely ascribable to OIS 3, represents a principle of pedogenesis under conditions of high dust accumulation, as shown by the presence of numerous “Sommersprossen” in the parent material. This weakly developed soil suggests that the interstadial conditions in the region were not as pronounced as in Europe or China (e.g., Fitzsimmons et al., 2012; Marković et al., 2015b). Furthermore, PS1 is likely an accretionary soil, indicating the availability of sediment in the region during that time. The presence of PS1 subdivides the loess flow into two pulses, one before and one after its deposition. However, dating these pulses is challenging due to the uncertainty of the results. Nevertheless, they are comparable to depositional events in neighboring Kazakhstan, such as Remizovka (Fitzsimmons et al., 2018). No significant variations in MAT and MAP were detected during this period. Regarding the HS, Wang’s paleoclimate transfer functions suggest a MAT of approximately 11 °C and a MAP of about 600 mm. These findings indicate that during the deposition of the HS at Karamaidan, slightly higher temperatures and lower precipitation levels were present compared to the conditions observed today.