Abstract
Central Asia is located at the confluence of large-scale atmospheric circulation systems. However, the number of Holocene climate records is still low in most parts of this region and insufficient to allow detailed discussion and comparisons to disentangle the complex climate history and interplays between the different climatic systems. Here, we present the first stalagmite record from arid Central Asia (south-western Kyrgyzstan) by using δ18O, δ13C, and micro x-ray fluorescence (µXRF)-sulfur data spanning the last 5000 years. The cave hosting stalagmite Uluu-2 is ideally suited to identify past shifts in seasonal variations in precipitation in this part of the world. Comparison of instrumental and paleo-isotopic studies demonstrates that the Uluu-2 speleothem isotope composition faithfully records climate changes and responds to shifts in the proportion of moisture derived from mid-latitude Westerlies during the winter/spring season. The reconstructions suggest that the area was characterized by a dry climate from 4700 to 3900 yr BP, interrupted by a wet episode around 4200 yr BP. Further drier conditions also occurred between 4000 and 3500 yr BP. Wetter conditions were re-established at ca. 2500 yr BP, after another dry episode between 3000 and 2500 yr BP. With the exception of two short dry events (1150 and 1300 yr BP), the period after 1700 yr BP shows moderate to wet conditions. Regional comparisons suggest that the strength and position of the Westerly winds control climatic shifts in arid Central Asia, leading to complex local responses.
Keywords
Introduction
The impact of global warming on current and future water resources is attracting increasing attention. Virtually no other region on Earth is thought to be affected as strongly by future changes in water supply than arid Central Asia (CA; Chase et al., 2000; De Martino et al., 2005). A projected increase in temperature and changes in the interannual variability of rainfall may bring further environmental and economic stress to a region, which is characterized by population growth and a high vulnerability to natural disasters (De Martino et al., 2005; Held and Soden, 2006). In the light of anthropogenic forcing of climate and associated surface processes (IPCC, 2013), it is necessary to understand changes in the response of the natural climate system to long-term forcing mechanisms, such as insolation and the interaction between atmospheric and oceanic circulation systems and their relationship with short-term variations on decadal to centennial or millennial timescales. Furthermore, understanding teleconnections is crucial to decipher climate relationships over long distances that reflect organized patterns of atmospheric and oceanic circulation. Because of their effect on large-scale synoptic patterns, mountain systems in CA modulate global-scale climate variability, but at the same time are highly sensitive to such variations (Sorg et al., 2012). Different climate archives have been used for understanding the history and mechanisms of climate change in CA during the Holocene (e.g. lakes (Boomer et al., 2009; Mathis et al., 2014; Mischke et al., 2010; Ricketts et al., 2001; Sorrel et al., 2007), tree-rings (Esper et al., 2007), ice cores (Aizen et al., 2006; Thompson et al., 1997)) especially in the analyses of long-term changes in zonal and meridonal atmospheric circulation patterns and its teleconnections, such as the North Atlantic Oscillation (NAO; Aizen et al., 2006; Lauterbach et al., 2014) or the El Niño Southern Oscillation (Barlow et al., 2002). The increasing number of climate archives from CA has revealed significant spatial differences in the Holocene climate history between specific regions throughout CA. However, the number of Holocene climate records is still low in many parts of CA and insufficient to allow detailed discussion and comparisons to disentangle the complex climate history and interplays between the different climatic systems.
Here, we present evidence from a stalagmite from Uluu-Too cave located in the core region of arid CA between the foothills of the Tien Shan and the intramontane Fergana Valley in south-western Kyrgyzstan. Our results will help to understand hydrological changes in the Central Asian mountains and to unravel the forcing mechanisms behind precipitation variations that have occurred during the last 5000 years.
Study site
Cave settings
The studied Uluu-Too cave (or Onyxsovaya, named after the mined Onyx mineral) lies within folded Upper Devonian to Lower Carboniferous limestones of the Uluu-Too Mountains on the southern rim of the Fergana Valley in Kyrgyzstan (Figure 1). This small mountain range with monolithic character is situated to the west of the city of Osh and close to the Uzbek border (40°23′N, 72°21′E; Figure 1).

(a) Map of the Asian continent, showing topographic conditions and principal atmospheric trajectories. Strengths and compositions of these atmospheric regimes have a direct effect on the amount of rainfall as well as influencing where the moisture comes from and the pathway that it follows. Brown dashed line indicates current monsoon limits, adapted from Holmes et al. (2009); (b), (c) position of the Uluu-Too cave (yellow and black star), on the southern rim of the Fergana Basin. The cave is located at 40°23′N, 72°21′E and at about 1490 m above sea level, in Paleozoic limestones of the Uluu-Too Mountains west of the city Osh close to the Uzbek border.
The narrow cave entrance lies on the southern mountain slope at about 1490 m a.s.l. with a horizontal length of approximately 100 m in a northern direction and ends in two poorly ventilated cave chambers (Dudashvili and Dudashvilli, 2012). These chambers are overlain by 30–50 m of karstified limestone bedrock. The area around the cave is characterized by typical steppe vegetation, with montane grasslands and shrublands, which change into bare rocks. The thin karstified bedrock above the cave promotes fast infiltration of seasonal rainfall and the marginal vegetation cover reacts in a sensitive fashion to changing moisture conditions.
A meteorological monitoring system for cave temperature and humidity, in service since October 2011, indicates a constant cave air temperature of 15.4°C (Figure 2). This is close to the mean annual air temperatures measured outside the cave (15.1°C) and the mean annual air temperature of 15.2°C recorded between 1959 and 1997 at the Osh meteorological station (875 m) in the Fergana Valley (http://www.ncdc.noaa.gov). Cave humidity during the recent observation period has been constant by 100% (Figure 2).

Monitoring data of temperature, relative humidity, and atmospheric pressure inside and outside the Uluu-Too cave. Monitoring data of temperature and relative humidity from October 2011 to September 2014 indicate a constant cave air temperature of 15.4°C. This is close to the mean annual air temperatures measured outside the cave (15.1°C) and the mean annual air temperature of 15.2°C recorded between 1959 and 1997 at the Osh meteorological station (875 m a.s.l.). In-cave humidity during the observation period has been constant at 100%. Humidity outside the cave is characterized with dry conditions during the summer/autumn months in comparison with wetter conditions during the winter/spring period. Atmospheric pressure outside the cave indicates an annual cycle of low pressure during the summer in contrast to high pressure during the winter period.
Climate
Rainfall in this region is highly seasonal because of the interaction between the Siberian High and the Westerlies system (and in some more southerly regions the limits of the Indian monsoon) (Figure 1; Aizen et al., 2001; Chen et al., 2008). During winter and early spring, the southwest branch of the Siberian High extends southward, accompanied by a southward shift of the Westerlies, which carry moisture-laden air from the Atlantic, Mediterranean, and Caspian Sea regions (maximum precipitation in March, nearest station Osh and Tashkent, Figure 3). The reverse applies in late spring and summer, when the Siberian High retreats northward and interrupts the influx of moisture into this area, resulting in a very pronounced seasonality in precipitation (see Figure 3). Furthermore, annual reconstruction of monthly air mass trajectories illustrates the remote provenance of moisture between November and June mostly driven by the Westerlies system. In contrast, air mass movements during summer and early autumn (July to October) are limited to the regional scale (Figure 3; Draxler and Hess, 1998).

Map with climate diagrams from meteorological stations (number 6: precipitation data from Osh via GHCNv2 and temperature data from Osh (http://www.ncdc.noaa.gov)) under the influence of the Westerlies atmospheric circulation system (IAEA, 1992; Morris et al., 2005; Peterson and Vose, 1997). The yellow star indicates Uluu-Too cave. Monthly mean of air temperature (red line), precipitation (blue bars), and oxygen isotope values of precipitation (green line) from meteorological stations in the Westerlies region (Eastern Mediterranean and Central Asia) show a robust positive correlation between air temperature and δ18O of precipitation, independent from altitude, geographic location, moisture source, and shifts in precipitation seasonality. The numbers behind each site show typical minima (winter) and maxima (summer) δ18O precipitation values. Under the same seasonal air temperature conditions, it is apparent that the δ18O of winter/spring precipitation became progressively lower eastward from the Mediterranean and Caspian Sea region to Central Asia, illustrating the ‘continental effect’ (progressive depletion of heavy isotopes with increasing distance from the coast as a consequence of the gradual rainout of the air masses entering the continent; Rozanski et al., 1982). (Map) Monthly reconstruction of air mass trajectories for the year 2014 (NOAA HYSPLIT MODEL using GDAS Meteorological data (Draxler and Hess, 1998; http://ready.arl.noaa.gov/HYSPLIT.php); starting date was set to every 15th of each month for the duration of 300 h backward) illustrates the remote provenance of moisture between November and June. In contrast, air mass movements during summer and early autumn (July to October) are limited to regional scale within few kilometers.
The long-term average annual air temperature (precipitation) at the meteorological station in Osh is 15.2°C (358 mm), with January and July means of about −3°C (35 mm) and 26°C (12 mm), respectively (http://www.ncdc.noaa.gov). Monitoring data of temperature, humidity, and atmospheric pressure outside the cave clearly reflect this seasonality, with low pressure during the summer months (~840 hPa) and high-pressure influence during the winter period (~860 hPa; Figure 2). Temperature and humidity show the same pattern, but with reverse scale. High temperatures correspond with low humidity during the summer period and vice versa during winter (Figure 2). Furthermore, the significantly reduced precipitation and associated heating of the desert lowlands cause local to regional advection, which is responsible for numerous violent dust storms (Schettler et al., 2014; Sorrel et al., 2007).
Materials and methods
The 176-mm-long, broken (cave conservation regulations) stalagmite Uluu-2 was retrieved in October 2011 as a whole sample at the end of the cave around 100 m from the cave entrance. Data for cave temperature and humidity were collected using a HumiLog ‘rugged’ logger (Driesen and Kern, Germany). Outside meteorological parameters (temperature, humidity, and atmospheric pressure) were measured with a HumiBaroLog ‘rugged’ logger (Driesen and Kern, Germany). A total of 24 samples were dated at the University of Minnesota using 230Th, in order to establish a precise chronology. For analytical details and standards used, we refer to Cheng et al. (2013) (Table 1).
Age dating results from Uluu-2 speleothem.
Bold values are the result of the 2-sigma calculations.
d234U = ([234U/238U]activity – 1) × 1000.
d234Uinitial was calculated based on 230Th age (T), that is, d234Uinitial = d234Umeasured × el234×T. Corrected 230Th ages assume the initial 230Th/232Th atomic ratio of 4.4 ± 2.2 × 10–6. Those are the values for a material at secular equilibrium, with the bulk earth 232Th/238U value of 3.8. The errors are arbitrarily assumed to be 50%.
BP stands for ‘Before Present’ where the ‘Present’ is defined as the year AD 1950.
Sample preparation for stable isotope analyses was carried out at the Helmholtz Centre Potsdam GFZ, using a ‘Sherline 5410’ vertical milling machine with a special stage and a micrometer-drive to allow precise movement of the speleothem block, for stepwise milling at a resolution of 0.1 mm. Oxygen and carbon isotope analyses (precision of δ18O and δ13C on reference material is <0.06‰) were carried out on 1806 samples using an automated carbonate-extraction system (KIEL IV) interfaced with a MAT253 IRMS (ThermoFisher Scientific). All isotopic ratios were expressed in the delta notation relative to VPDB.
Optical microscopy, SEM imaging, and x-ray diffraction (XRD) were used to verify the mineral structure of the samples. Micro x-ray fluorescence (µXRF) scanning was performed using an EAGLE III XL µXRF system spectrometer (Röntgenanalytik, Germany) at GFZ, applying a 40-kV tube voltage, a 300-µA tube current, a 123-µm spot size, and a 100-µm step size. Element intensities of this nondestructive method are expressed in counts per second (cps). The consistency of replicate profiles of the same scanning line demonstrates the good repeatability of the µXRF measurements. Sulfur intensities have been used relative to calcium (S*1000/Ca) to minimize physical measurement effects.
Water samples for stable isotope analyses of oxygen (δ18O) and deuterium (δD) were collected between October 2011 and September 2014 from various locations within the cave and were stored in sealed air-tight polypropylene bottles until analysis at the Alfred Wegner Institute for Marine and Polar Research in Potsdam. A Picarro water isotope analyzer (Cavity Ring-Down Spectrometer – L2120i) was used for determination of hydrogen and oxygen isotope composition, following Meyer et al. (2000). Analytical precision of SMOW was better than 0.8‰ for δD and 0.1‰ for δ18O. Spectral analysis of Uluu-2 δ18O time-series was performed using the AnalySeries software package (Paillard et al., 1996) with a Blackman–Tukey spectral window. Considering the resolution and age uncertainties (avg. 38 years) of our record, we disregard spectral peaks below 50 years.
Results
Uluu-2 stalagmite
The Uluu-2 stalagmite is characterized by sequences of light beige to brownish gray laminations with large variations in thickness and laminae color (Figure 4). Microscopic observations and SEM diagrams on several selected thin sections, as well as XRD results, show the Uluu-2 speleothem consists of calcite. In general, laminations are thicker and darker in the lower part, changing into thinner and brighter in the upper part.

Chronology and lithological units: (a) age model of the Uluu-2 stalagmite including a total of 24 samples (see Table 1); red bars depict errors of 230Th dates (2σ), and (b) image of the 176-mm-long Uluu-2 stalagmite with lithological units and locations of dated samples.
Based on thickness and lamination color and the resolution of the age model, we divided the speleothem into three lithological units (Figure 4). Unit 1 (0–20 mm) is characterized by a bright thick sequence without appreciable brownish gray layers. Unit 2 (20–110 mm) consists of finely laminated successions of brownish gray and light beige layers. Unit 3 (110–176 mm) is composed of several light beige episodes interrupted by dark gray brownish layers of variable thickness.
Chronology and growth rates
The age model yielded 24 precise U-series ages, all in stratigraphic order, which were then interpolated using the COPRA software (Breitenbach et al., 2012) to establish an ~4700-year-long chronology (Figure 4), with chronological uncertainties ranging between 11 and 66 years around an average of 38 years (Table 1).
Growth rates in stalagmites have been related to water supply and/or the temperature of drip water (Fleitmann et al., 2004; Genty et al., 2003) with higher rainfall and/or warmer temperatures resulting in higher growth rates. However, stalagmite growth rates are strongly influenced by pCO2-level in soil and cave air, and saturation state of the infiltrating water. CO2 degassing of drip water upon entering low-pCO2 cave voids leads to oversaturation and carbonate precipitation. Thus, periods with either strong CO2-gradients between infiltrating water and cave air, or low drip rates will be mirrored by increased carbonate precipitation (Baldini, 2010). The Uluu-2 age model allows constructing a highly resolved series of growth rates. Growth rates vary between ~0.0069 and ~0.327 mm/yr with an average of ~0.094, with significant changes at 100–120 mm. The upper section between 20 and 100 mm shows more than double mean growth rates (~0.07 mm/yr) than the lower part (0.03 mm/yr).
Cave drip water
Analysis of four drip water samples collected in October 2011, May 2012, April 2013, and September 2014 follows the isotopic composition of winter/spring versus summer/autumn rainfall. Seasonal comparison between October 2011 (−8.7‰) and September 2014 (−7.7‰) versus May 2012 (−11.6‰) and April 2013 (−11.3‰) reveals different isotopic compositions, with lower values in April and May, which seemingly track the isotopic seasonality of rainfall (Figure 3).
Variations in stable isotopes (δ18O, δ13C) and µXRF-sulfur signals
The full range of δ18O in the 4700-year-long Uluu-2 record is 5.2‰, ranging from −8.5‰ to −13.7‰ with a mean of −10.8‰ (Figure 5). Carbon-isotope and S/Ca values indicate a similar trend, ranging between −2‰ and −9‰ for δ13C, and 2 and 0.3 for S/Ca. Variations in laminae color and δ18O, δ13C, and S/Ca ratios show strong similarities. Dark sequences correspond with more positive δ18O and δ13C values and bright zones reflect zones with more negative values. The same trend is shown for S/Ca values; higher values correspond with darker, more grayish color zones.

Lithological units and Uluu-2 stable isotope (δ18O, δ13C) and µXRF-sulfur analyses spanning the last 5000 years. Above: oxygen isotope data (blue line) compared with carbon isotope data (green line). Black horizontal bars are U/Th ages with 2σ error ranges. Below: oxygen isotope data (blue line, 7-point running mean) between AD 1200 and AD −2000 compared with the S/Ca µXRF scanning profile (yellow line, 7-point running mean).
Discussion
How do Uluu-2 stable isotopes and µXRF-S/Ca ratio changes translate into climate signals?
Speleothem isotope modeling studies have shown that the variation in kinetic isotope fractionation, differences in drip rate, and further cave processes are not responsible for the large range observed in the Uluu-2 δ18O record (5.2‰; Deininger et al., 2012; Dreybrodt, 2008; Mühlinghaus et al., 2009). It is more likely that changes in precipitation characteristics above the cave are responsible for fluctuations in speleothem δ18O values.
By evaluating isotopes of precipitation in the context of climate parameters in Asia, the amount effect is the dominant factor on monsoonal East Asia, while temperature mainly controls the δ18O values of precipitation in arid CA in the seasonal cycle (Araguas-Araguas et al., 1998; Yao et al., 2013). In addition, large-scale circulation changes or a shift in the balance of different moisture sources and transport trajectories can result in isotopic shifts over time (Cheng et al., 2012; Dansgaard, 1964; Rozanski et al., 1982). Today, annual rainfall in the Uluu-Too cave region is dominated by Westerlies-controlled transport of moisture from the Mediterranean and Caspian Sea regions to CA. Meteorological stations along a transect from the Eastern Mediterranean and CA (Figure 3) reveal a robust positive correlation between air temperature and δ18O of precipitation. All stations record extreme seasonal changes in the proportion of winter/spring versus summer/autumn precipitation (Figure 3) with typical minima in δ18O during winter and maxima in the summer months. If these seasonal controls also determine interannual variations in the isotopic composition of rainfall, then temperature could be a major factor explaining the reconstructed oxygen isotopic variability. However, we assume that the temperature-related effects on centennial to decadal scale at Uluu-2 are likely to be minimal because changes in the temperature-dependent fractionation between calcite and water are small, ~−0.25‰/°C (Friedman and O’Neil, 1977), and insufficient to explain the large range in δ18O Uluu-2 over the last 4700 years (5.2‰) (Figure 5). Furthermore, the constant cave temperature supports this assumption of minor effect (Figure 2). If the temperature effect is likely to remain minimal on interannual δ18O Uluu-2 composition, the linkage between climate anomalies and reconstructed δ18O Uluu-2 variability seems controlled by (1) moisture sources and transport trajectories, (2) seasonality changes, and (3) snowmelt:
Re-calculation of monthly air mass trajectories illustrates the remote provenance of moisture between November and June mostly driven by the Westerlies system (from Atlantic). In contrast, air mass movements during summer and early autumn (July to October) are limited to the regional scale within few kilometers (Figure 3; Draxler and Hess, 1998). Under the same air temperature seasonal conditions, the δ18O of winter/spring precipitation became progressively lower eastward from the Mediterranean and Caspian Sea region to CA (e.g. December data: Ras Muneef: −6.5‰ in contrast to Bishkek: −17‰), illustrating the progressive depletion of heavy isotopes with increasing distance from the coast as a consequence of the gradual rainout of the air masses (transport via Westerlies) entering the continent ‘continental effect’ (Rozanski et al., 1982).
Seasonality changes could alter the δ18O values significantly. Wet winters/springs, for example, would result in a shift toward substantially lower δ18O values in contrast to dry summers/autumns with substantially higher δ18O values. Intensified winter/spring precipitation tends to increase the winter/spring contribution to the overall recharge budget and hence results in a decrease of the annual mean oxygen isotopic ratio.
Furthermore, we suggest that a high proportional contribution of water derived from snowmelt, after relatively long and wet winters with high amounts of snowfall, can further lead a shift toward more negative δ18O Uluu-2 values. Constant temperature and humidity in the cave as well as the little amount of annual rainfall suggest that longer snow cover, which would have effects on the isotopic composition, is probably negligible.
Stalagmite carbonate δ13C is influenced by the δ13C of CO2 in the soil lying above the cave, the amount of CO2 available in the soil, soil microbial productivity, CO2 degassing of drip water, and infiltration rates (Baker et al., 1997; Breitenbach et al., 2015; Genty et al., 2003; Hendy, 1971). The δ13C of soil CO2 is mainly controlled by vegetation type and vegetation density above the cave, with lower values when the plant cover is predominantly of the C3 type and higher values when C4 plants predominate. Drier conditions in the past would have enhanced evaporation, restricted plant growth, and reduced the volume of CO2 in the soil. In contrast, wetter conditions would lead to reduced evaporation, increased vegetation growth, and increased volume of CO2 in the soil. Drier (wetter) conditions would also lower (enhance) drip rates in the cave, thus allowing for prolonged (reduced) CO2 degassing and 13C-enrichment (depletion) in the fluid. Under conditions of limited (prolonged) exchange between hostrock limestone and infiltrating water, less (more) carbon would contribute from hostrock (δ13CUluu-Too = 1.5‰) rather than soil CO2 (δ13C of C3 soil CO2 is ca. −23‰; McDermott, 2004) because of the lower (higher) levels of soil CO2 and the likelihood of dissolution under more closed-system (open-system) than open-system (closed-system) dynamics. Higher δ13C values in Uluu-2 are interpreted as indicative of drier/warmer conditions, while lower values reflect wetter/cooler conditions at the Uluu-Too region.
Several ‘Hendy tests’ (variability of δ18O and δ13C along an individual laminae; Hendy, 1971) performed at different depths over the entire profile (0.8, 2.5, 3, 5.5, and 9 cm) show only minor δ18O variations (<0.5‰) and support near-equilibrium conditions during deposition. However, several workers have questioned its reliability for paleoclimatic research (Dorale and Liu, 2009) and advocate replication test. However, this was not possible because of cave conservation. Instead, to further corroborate our interpretation of the Uluu-2 δ18O record as indicator for moisture variations, we compare it with the sulfur-to-calcite (S/Ca) ratio obtained with µXRF. In general, variations in the sulfur concentrations in speleothems may reflect atmospheric sulfur variability because of sea spray, dust, volcanic activity, and/or industrial pollution (Frisia et al., 2005; Wynn et al., 2010). Sulfur aerosols in CA are mainly derived from gypsum sourced from the Caspian and Aral Sea regions (Kreutz and Sholkovitz, 2000). Enhanced sulfur concentrations in Uluu-2 are therefore likely to reflect increased atmospheric dust transport under drier conditions (Figure 5). Modern variations in dust characteristics on the northern slope of the Tien Shan reveal a complex pattern in seasonality, depending on local and hemispheric circulation dynamics, outwash of atmospheric dust, and the seasonal frequency in dust storms (Schettler et al., 2014) with higher event variability during the summer months (JJAS) and very low dust concentrations during winter months (Schettler et al., 2013, 2014). Low dust concentrations during times of enhanced wind strength (Westerlies) are likely due to enhanced moisture supply, dense vegetation, and/or agglomerated soils. To a remarkable degree, high S/Ca ratios correspond to higher δ18O values (Figure 5).
These correspondences provide an independent evidence of a broad negative relation between rainfall (δ18O) and dust input (sulfur), which supports our expectations that the Uluu-2 δ18O record reflect interannual fluctuations in mean annual precipitation, in particular the winter/spring rainfall contribution. On interannual timescales, extended winter/spring seasons would lead to abnormal wetter conditions and probably to shortened summer/autumn periods and/or denser vegetation, associated with low sulfur input, and vice versa.
Each individual proxy (δ18O, δ13C, and S/Ca, growth rate) shows internal indications that the Uluu-2 δ18O speleothem record can be used as proxy to indicate moisture variation and to reconstruct environmental changes over the last 5000 years. Under this assumption, the Uluu-2 stalagmite δ18O record can be interpreted as reflecting centennial variations in mean annual isotopic composition of recharging precipitation related to interannual/decadal variability in air mass dynamics, most probably during an intensified Westerlies regime, and vice versa.
Regional paleoclimate perspective on the Uluu-2 record
When compared with regional paleoclimate records, the Uluu-2 δ18O record provides insights into precipitation variability at centennial timescales and during abrupt late-Holocene climate events, which are poorly documented by available climate reconstructions from this region (Solomina and Alverson, 2004). The spatial patterns of Holocene moisture evolution in arid CA are very complex and show significant differences in climate development between the western (Kazakhstan, Tajikistan, Uzbekistan, Kyrgyzstan) and the eastern (Mongolia, NW China) parts of CA (Chen et al., 2008; Herzschuh, 2006; Rudaya et al., 2009).
The reconstructions based on the Uluu-2 record suggests that the area was characterized by a dry climate prior to 3900 yr BP, interrupted by a wet episode around 4200 yr BP. Drier conditions recurred between 4000 and 3500 yr BP (Figure 6), followed by a wet interval that lasted 500 years until 3000 yr BP. The re-establishment of wetter conditions occurred at ca. 2500 yr BP after another 500-year-long dry spell. With the exception of two dry events (1150 and 1300 yr BP), the period after 1700 yr BP shows moderate to wet conditions.

Uluu-2 oxygen isotope data (5-point running mean) with other selected regional climate records over the last ~5000 years: (a) δ15N of Lake Son-Kul as a proxy for winter precipitation (Lauterbach, personal communication) and/or snowfall intensity (Lauterbach et al., 2014); (b) Lake Karakul pollen samples’ principal component analysis (PCA1) and δ18Oostr data (Mischke et al., 2010) reflecting changes in the amount of moisture (PCA1) and long-term changes in the isotopic composition of rainfall (δ18Oostr) (Mischke, personal communication) (Taft et al., 2014); (c) Aral Sea lake-level fluctuations (Boroffka, 2010); (d) δ18Oostr data from Lake Issyk-Kul reflecting changes between evaporation and water input; and (e) leaf wax carbon isotope record from Lake Karakuli (Aichner et al., 2015).
Other regional climate records from CA are in line with this reconstruction (Figure 6), demonstrating generally similar patterns of Holocene environmental and climate change. We verified the regional significance of our stalagmite profile as precipitation-sensitive record by comparing the δ18O data with Aral Sea lake level of the last 4700 years (Figure 6), inferred from geomorphic and archeological evidence (Boomer et al., 2009; Boroffka, 2010). To a remarkable degree, lower isotope values coincide with higher lake level, and vice versa. The results confirm that Uluu-2 δ18O is a robust indicator of regional climate and its large-scale sensitivities. Interestingly, the dry period 4500–3900 yr BP is not reflected in the Aral lake-level record, which can be explained by dating uncertainties, sampling resolution, and/or meltwater input in Lake Aral at higher air temperatures. Pollen records from Hovsgol Lake (Prokopenko et al., 2007) indicate that a pronounced warming may have occurred around 5500 to 4000 yr BP.
However, similar to our Uluu-2 record, in Westerlies-dominated arid China, Bosten Lake experienced shallow lake level during this period (Wünnemann et al., 2006). A very pronounced drought event is also recorded at Lake Balikun (China) (An et al., 2012) and Lake Son-Kul/Kyrgyzstan (Lauterbach et al., 2014) between 4300 and 3800 yr BP (Figure 6). However, the missing indication for this assumed wet period in Lake Aral in other regional records shows that site-specific hydrological conditions, internal thresholds, the sensitivity of proxies, temporal resolution, and dating accuracy are of crucial importance for identifying such climate fluctuations. Uluu-2 climate reconstructions toward drier conditions set in between 4000 to 3500 yr BP. This broadly coincides with low lake levels and relatively dry conditions in the Aral Sea region (Figure 6; Boomer et al., 2000; Boroffka, 2010), distinct phases of reduced winter precipitation in Central Kyrgyzstan (Huang et al., 2014; Lauterbach et al., 2014), drier conditions at Lake Karakuli (Aichner et al., 2015), pollen data from Lake Karakul (Mischke et al., 2010), and evaporation versus water input data reported from Lake Issyk-Kul (Ricketts et al., 2001) (Figure 6). Furthermore, the onset of drier conditions in the Aral Sea region (Boomer et al., 2000; Boroffka, 2010) and significant water level decrease at Lake Van (Reimer et al., 2009) all correspond with the dry event observed at 3500 yr BP at Uluu-Too. An adjacent distinct wetter episode between 3500 and 3000 yr BP in the Uluu-2 record is also shown in records from Lake-Son-Kul (Lauterbach et al., 2014), Lake Karakul (Mischke et al., 2010; Taft et al., 2014), Lake Issyk-Kul (core IK97 11P) (Ricketts et al., 2001), Lake Karakuli (Aichner et al., 2015), and higher water levels of the Aral Sea (Boroffka, 2010) (Figure 6). Drier conditions between 3000 and 2500 BP agree with a significant drop in water levels of Lake Issyk-Kul (Ferronskii et al., 2003), Aral Sea (Boroffka, 2010), Bosten Lake (Mischke and Wünnemann, 2006; Wünnemann et al., 2006), as well as with reduced water inflow (δ18OOstra), pollen indication from Lake Karakul (Mischke et al., 2010; Taft et al., 2014) and reduced winter precipitation at Lake Son-Kul (Lauterbach et al., 2014).
Aral Sea lake-level changes between 2500 and 780 yr BP correspond well with reconstructed moisture dynamics at Uluu-Too (Figure 6). The onset of wetter conditions found in our record at 2100 yr BP also coincides with moisture evolution in the Tien Shan. Reconstructed high-stands of Lake Issyk-Kul around 1800 yr BP (Ricketts et al., 2001) and Lake Manas (Rhodes et al., 1996) as well as wet conditions at Lake Balikun (An et al., 2012) and enhanced winter precipitation in Central Kyrgyzstan (Huang et al., 2014; Lauterbach et al., 2014) correspond with the lowest Uluu-2 δ18O values over the entire record. This reported wet phase is in agreement with pollen analyses from Lake Karakul (Mischke et al., 2010) and macrophyte productivity reconstruction from Lake Karakuli (Figure 6; Aichner et al., 2015).
Global forcing mechanisms on Central Asian climate
The interplay between the main atmospheric circulation systems in central Asia – (i) the mid-latitude Westerlies, (ii) the Siberian High, and a partly/or marginal – (iii) the Indian Summer Monsoon (ISM), lead to complex interannual climate patterns (Aizen et al., 2001; Chen et al., 2008).
Modern air mass trajectories (Figure 3) show that the annual rainfall is controlled by the mid-latitude Westerlies transport of moisture from the Mediterranean and Caspian Sea region to CA during winter and spring (Figure 3; Aizen et al., 2001). Flow tracks from the Siberian High deliver cool and dry air during winter/spring and seem to have only a minimal contribution to annual rainfall. However, variation of the Siberian High intensity seems crucial for moisture transport to the Uluu-Too region. Comparison of the Uluu-2 record with two Siberian High reconstructions for the last 250 years (tree-ring reconstruction and GISP2 K+ ions; D’Arrigo et al., 2005; Mayewski et al., 1997) suggests that the coupling of the Siberian High with the mid-latitude Westerlies system likely contributed significantly to the observed climate variability in the study area (Figure 7). In contrast, the marginal rainfall during the summer months in the Uluu-Too region and the lack of marine salts in Muztag Ata ice cores (Northwest Tibet; Seong et al., 2009) supports that the mid-latitude Westerlies provide the main fraction of present-day precipitation while the ISM is only of minor importance. Variation in strength, duration, and main tracks of the Westerlies and its related movement of the polar front (Machalett et al., 2008) could have influenced the amount of winter and spring moisture which has reached the Uluu-Too region during the last 5000 years. Comparison of the Uluu-2 record with a reconstruction of the NAO index for the last 250 years (Luterbacher et al., 2001) shows that moisture changes in the Uluu-Too region exhibit a similar trend with the reconstructed NAO index. This correspondence corroborates the inferred North Atlantic influence on CA precipitation via the Westerlies. Modern climate data comparisons between arid CA and the NAO during 1950 and 2015 indicate the impact of the NAO leading to more winter/spring precipitation and cooler summer temperature (Figure 8). Interestingly, comparisons with temperature reconstructions from tree rings (summer indicator) in Kyrgyzstan (Esper et al., 2007) and the Northern Hemisphere (Moberg et al., 2005) for the last 250 years also match our reconstructed winter/spring data from Uluu-Too (Figure 7). Similarities between cooling events at Lake Karakuli (Figure 6; Aichner et al., 2015), northern Atlantic ice drift events (Figure 9; Bond et al., 2001), and strengthening phases of the Siberian High (Figure 7; D’Arrigo et al., 2005; Mayewski et al., 1997) are also seen in the Uluu-Too record and support our assumption that the relatively wet episodes were mainly sustained by winter/spring precipitation derived from mid-latitude Westerlies. Earlier studies (Aizen et al., 2006; Lei et al., 2014) suggested that during negative NAO modes the main pathway for westerly winds is located further south, with wetter conditions in the Mediterranean and enhanced amounts of moisture transport into Central Asian regions of the same latitude. Modeling studies propose a more complex interplay between the Eurasian and Pacific circulation systems, leading to generally higher moisture flux into Central-Southwestern Asia during episodes of intensified westerly winds and positive NAO modes (Syed et al., 2010), without inferring re-positioning of the Westerlies tracks. Interestingly, lower winter precipitation reported from the Mediterranean (Figure 9; Fletcher et al., 2012) and contrary more rainfall in the Scandinavian region (Figure 9; Bjune et al., 2005) during episodes of higher winter/spring precipitation at Uluu-Too follow this interpretation that a positive NAO mode could drive moisture delivery through CA.

Comparison with Uluu-2 δ18O data and Northern Hemisphere climate records over the last 250 years: (a) Northern Hemisphere temperature anomaly reconstructions (Moberg et al., 2005); (b) reconstruction of the North Atlantic Oscillation index (Luterbacher et al., 2001); (c) tree-ring-based global reconstruction of the Siberian High Index (D’Arrigo et al., 2005); (d) reconstruction of the Siberian High inferred from the Greenland GISP2 potassium concentrations (Mayewski et al., 1997); and (e) West Tien Shan tree-ring data (Esper et al., 2007).

Modern climate data comparisons between arid Central Asia and the averaged NAO (1950 and 2015) using the KNMI Climate Explorer (http://climexp.knmi.nl). Seasonal precipitation analyses show strongest impact between November and January and May to July for temperature, respectively: (a) correlation between November and January averaged NOAA CPC NAO data (Climate Prediction Center) and the time-series dataset for precipitations (CRU TS3.23) produced by the Climatic Research Unit (CRU) at the University of East Anglia between 1950 and 2014 (University of East Anglia Climatic Research Unit et al., 2008) and (b) correlation between May and July averaged NOAA CPC NAO data (Climate Prediction Center) and the NOAA National Center for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) surface temperature between 1950 and 2015 (Kalnay et al., 1996).

Uluu-2 oxygen isotope data with other selected records: (a) North Atlantic hematite grains indicating northern hemispheric cooling events (Bond et al., 2001); (b) variation in forest composition (percentage of temperate forest taxa) in the western Mediterranean as proxy for Westerlies-influenced winter precipitation (Fletcher et al., 2012); (c) reconstructed percentage of winter precipitation reflects changes in the intensity of the Westerlies regime (present-day reference is 100%; Bjune et al., 2005); (d) power spectrum on Uluu-2 oxygen isotope data for the period 1100 and 4500 yr BP. Spectral analysis of Uluu-2 speleothem record reveals several periodicities that are significant at the 95% confidence interval (gray dashed line) that closely match the periodicities found in tree-ring ∆14C record (Stuiver and Braziunas, 1993); and (e) δ18O speleothem record from Indian Central Himalaya (Sainji cave; Kotlia et al., 2015).
However, that mechanism would be contrary to our correlation within the last 250 years with more winter/spring rainfall at the Uluu-Too region to the negative mode of NAO (Figure 7).
A strengthened Siberian High would push the mid-latitude Westerlies pathways further to the south, resulting in comparably drier conditions in the northern Central Asian regions (e.g. Tien Shan) but wetter conditions in south-western CA (Pamir) (Lei et al., 2014). Recently, a similar mechanism has been invoked to explain the current moisture gradient in the Pamirs, with lower winter/spring precipitation during the positive NAO mode and more northward-oriented Westerlies, while the central Pamirs receive high winter precipitation (Aichner et al., 2015). Based on our data, we argue that the wetter episodes recorded in our speleothem data are controlled by increased winter/spring precipitation caused by strengthened Westerlies. Reconstructed wetter periods are simultaneous within the Central Asian realm (Figure 6) indicating common climate variations within the last 5000 years. Variations in the North Atlantic conditions and Siberian High both appear to align with isotopic variability in our speleothem record, including intensity and pathways of the westerly winds.
On millennial to centennial timescales, our reconstructed regional moisture variability from Uluu-Too show similar long-term climate patterns found in the Central Himalaya. Interestingly, a region influenced by the ISM (Figure 9; Kotlia et al., 2015). These complex interactions between high- and low-latitude atmospheric circulation systems can also be seen in spectral data from the Uluu-2 record. Several periodicities that are significant at the 95% confidence interval (56, 88, 104, 128, 150, 172, 208, 280, 512, and 1500 years; Figure 9) match closely the periodicities found in the tree-ring ∆14C record (Stuiver and Braziunas, 1993), emphasizing links between high- and low-latitude hydrology on millennial to centennial timescales. Still, the oceanic and atmospheric interactions over Eurasia remain poorly understood and need additional data from well-dated high-resolution proxy reconstructions to unravel the relationships between the large-scale atmospheric circulation systems of CA.
Conclusion
A U-series dated high-resolution speleothem record from Uluu-Too cave from the southern rim of the Fergana Valley/Kyrgyzstan documents environmental changes during the last 5000 years with notable excursions around 3500 yr BP, 2500 yr BP and between 1500 and 1000 yr BP. The location of the cave is ideally suited to identify past shifts in seasonal variations in winter/spring precipitation and provide evidence that the transition between regions of summer versus winter/spring precipitation could have been a key factor for local climate in the past in CA.
The reconstructed centennial to decadal scale moisture evolution in arid CA can be linked with changes in the intensity and pathways of the mid-latitude Westerlies, with substantial impact on winter/spring precipitation. This is regionally synchronous with other Westerlies-dominated paleoclimate records, revealing a hemispheric teleconnection between the North Atlantic and the Central Asian climate system and its effect on winter/spring precipitation in arid CA. We find evidence for a significant influence of the North Atlantic NAO on CA winter/spring precipitation.
However, understanding the underlying climate mechanism in full detail requires further investigations and more archives from the hydrologically sensitive region of arid CA.
Footnotes
Acknowledgements
We are grateful to H Meyer for water isotope analysis, B Richert for µXRF scanning, and R Naumann for XRD measurements. We thank H Oberhänsli and two anonymous reviewers for their constructive comments, which helped us to improve the manuscript.
Funding
This research was supported by the Helmholtz Centre Potsdam (GFZ), the Global Change Observatory of Central Asia (GCO-CA), the PROGRESS research cluster at the University of Potsdam (CW and MS), and the Central Asian Climate Dynamics Project (CADY) which is funded by Germany’s Federal Ministry of Education and Research (BMBF). HC received support from NSFC grants 41230524 and NBRPC 2013CB955902.
