Variations in Sediment Grain Size from a Lake in the Tianshan Mountain of Central Asia: Implications for Paleoprecipitation Reconstruction

: The Tianshan Mountain is the largest mountain range in Central Asia, and the source area of many river systems. Changes in precipitation result in signiﬁcant alterations to regional hydrological processes. Lake sediment from the Tian Shan representative of the last 90 years was chosen as the object of this research study. The grain-size data were used in conjunction with instrumental data to provide a method for determining changes in paleoprecipitation. The results showed the three-point moving average curve of the silty fraction content with a size of 16 to 32 µ m to be signiﬁcantly consistent with the curve of total precipitation from April to September since 1950. The total content of clay and ﬁne-silty fraction (0–16 µ m) was clearly consistent with the monthly precipitation in July. The total precipitation from April to September showed a signiﬁcant downward trend from 1930 to 1975, and then an overall increasing trend beginning in 1975, which may have been inﬂuenced by the North Atlantic Oscillation. The change in precipitation reconstructed by the grain size of lake sediments was signiﬁcantly di ﬀ erent from the high-resolution gridded datasets (Climatic Research Unit Time-Series version 4.04) because of the lack of data from meteorological stations in China before 1950. The conclusions of this study are signiﬁcant for evaluating the validity of climatic research unit (CRU) data in arid areas of Western China. In addition, the results of this study serve as a bridge between modern instrumental records and long time-scale paleoclimate research and provide important reference values for future reconstructions of long time-scale paleoclimate.


Introduction
Located in the hinterland of Eurasia, the arid region of Central Asia has a typical continental arid climate and one of the most vulnerable terrestrial ecosystems [1]. A comprehensive understanding of the environmental issues resulting from climate change and human activities in Central Asia is of great significance for the ecological protection and improvement of the region, national security, and sustainable development of the social economy [2][3][4]. Over the past 100 years, air temperature in the arid region of Central Asia, which is mainly controlled by westerly winds, has clearly shown an increasing trend [5,6]. Annual precipitation in this region also shows an overall increasing trend, but with spatial differences [7,8]. Under the combined effect of climate change and human activities,  83.25 • E), the average annual temperature within this grid for the period 1960 to 2019 was 1.4 • C. The hottest month was July, with an average temperature of 14.7 • C; the coldest period was January, with an average temperature of −16.0 • C. The total annual precipitation was 159.4 mm; the average monthly precipitation was 25.6 mm, with the maximum precipitation occurring in May. The closest meteorological station to Lake Ta-Lung-Chi is Bayanbulak Station with an elevation of 2458 m. The average annual temperature recorded at Bayanbulak Station from 1960 to 2019 was 3.2 • C. The hottest month was July, with an average temperature of 18.3 • C; the coldest period was January, with an average temperature of −19.6 • C. The average precipitation for the period was 283.7 mm; the average monthly precipitation was 71.4 mm, with the maximum occurring in July. where Lake Ta-Lung-Chi is located. Monthly precipitation and monthly average temperature data for Lake Ta-Lung-Chi are from the climatic research unit (CRU) (Harris et al., 2020) while data for the Bayanbulak Station are from the Meteorological Data Center of China Meteorological Administration. (b) Lake Ta-Lung-Chi water system. The base map of (a) was derived from a SRTM30PLUS color-encoded shaded relief world topography (approximately 4 km) GeoTIFF image [36]. The base map of (b) was derived from global topographic data at 1 arc-second (~30 m) horizontal resolution (NASADEM) from NASA Land Processes Distributed Active Archive Center (LP DAAC) Distribution Server hosted at the USGS Figure 1. Geographical map of the study area. (a) Location of the Tian Shan where Lake Ta-Lung-Chi is located. Monthly precipitation and monthly average temperature data for Lake Ta-Lung-Chi are from the climatic research unit (CRU) (Harris et al., 2020) while data for the Bayanbulak Station are from the Meteorological Data Center of China Meteorological Administration. (b) Lake Ta-Lung-Chi water system. The base map of (a) was derived from a SRTM30PLUS color-encoded shaded relief world topography (approximately 4 km) GeoTIFF image [36]. The base map of (b) was derived from global topographic data at 1 arc-second (~30 m) horizontal resolution (NASADEM) from NASA Land Processes Distributed Active Archive Center (LP DAAC) Distribution Server hosted at the USGS Earth Resources Observation and Science (EROS) Center [37]. (c) Vegetation around the lake. (d) Core sediment TLC01 and field sub-sampling.

Materials and Methods
In May 2018, a sedimentary core with 50-cm length (TLC01) (42.447775 • , 83.285435 • ) was obtained from Lake Ta-Lung-Chi at a depth of 5.5 m using a gravity corer (UWITEC, Mondsee, Austria) fitted with a 60 mm internal diameter Perspex tube. The color of the core sediment is dark brown with no bedding structure. The core log was shown in Figure 2. The core was sampled in-situ at 1.0 cm intervals, and a total of 50 samples were obtained. The samples were analyzed at the Key Laboratory of Lakes and the Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 6 of 17 Figure 2. The vertical distribution of 137 Cs specific activity in Lake Ta-Lung-Chi sediments, a mountain lake in the Tian Shan (left), and the established chronological record for sediment core TLC01 (right). Different colors represent the percentage (%) of clay (gray)/silt (blue)/sand (white) versus depth.

Sediment Grain Size
The grain size distributions of sediments in core TLC01 ranged from 0.28 to 1905.46 μm and displayed clear bimodal or multimodal styles ( Figure 3). Sediment size analysis of the entire TLC01 core using an algorithm of end-member modelling analysis (EMMA) [38,39] showed that there are four relatively independent end-element components (EM1, EM2, EM3, and EM4); however, these four end-element components do not adequately express the entire granularity data sequence ( Figure  4). These results indicate that the grain size of Lake Ta-Lung-Chi sediments reflect abundant environmental information. The 137 Cs specific activity was detected by an Ortec HPGe GWL series well-type, coaxial low background intrinsic germanium detectors (EG&G ORTEC, Oak Ridge, TN, USA). The procedure for measuring sediment grain size involved placing a small amount of sediment sample (about 0.3 g) into a 100 mL beaker, adding 20 mL of distilled water, and 10 mL of 10% hydrogen peroxide (H 2 O 2 ). The sample was then heated and brought to a boil on a hot plate throughout which a wash bottle was used to continuously clean the beaker walls from substances deposited from the reaction foam. Once the sample was fully reacted following the complete decomposition of excess hydrogen peroxide, 10 mL of 10% hydrochloric acid was added and the beaker was removed upon boiling. Then, 100 mL of distilled water was added and left overnight, after which the distilled water was removed, the excess hydrochloric acid was washed off, and the sample was neutralized. Next, 20 mL of distilled water and 10 mL of potassium hexametaphosphate with a concentration of 0.05 mol/L were added to the sample and the beaker was placed in an ultrasonic cleaner and agitated for 15 min. Agitated samples were tested with a British Malvern Mastersizer 2000 laser particle size analyzer with a relative error of less than 1%.
The temperature (TMP) and precipitation (PRE) data were derived from Climatic Research Unit (CRU) Time-Series (TS) version 4.04 of high-resolution gridded data of month-by-month variation in climate (cru_ts_4.04) of the University of East Anglia in the United Kingdom, with a time range from 1901 to 2019 [22]. The monitoring data were obtained from the Meteorological Data Center of China Meteorological Administration. Sediment size analysis of the entire TLC01 core using an algorithm of end-member modelling analysis (EMMA) [38,39], which was used to extract geo-meaningful end members. The HYSPLIT model [40] was used to determine the possible source of water vapor. Global Data Assimilation System (GDAS) meteorological data (GDAS1) were downloaded from the National Centers for Environmental Prediction [41] for use in the HYSPLIT model, and the daily backward parcel trajectory for 48 h from April 2019 to September 2019 was calculated using MeteoInfo (TrajStat package) [42].

Core Dating
Variations in 137 Cs specific activity with depth in the lake sediment core are shown in Figure 2. There is significant accumulation of 137 Cs in the sediment core starting at a depth of 35 cm (0.51 Bq/kg) and peaking at 20 cm (14.60 Bq/kg). Based on the distribution of 137 Cs in lake sediments within the northern hemisphere, it is believed that the occurrence of 137 Cs residual layers corresponds to the start of global nuclear testing in 1954 [43], and the main peak at 20 cm may correspond to the Chernobyl nuclear leak in the former Soviet Union in 1986 [44,45]. The accumulation peak occurring between these depths corresponds to the global 137 Cs scattering peak (30 cm) date to 1963 [46,47], and the sedimentation rate of sediments in core TLC01 was calculated based on this year. The average sedimentation rate in Lake Ta-Lung-Chi is about 0.55 cm/year. With the average sedimentation rate, the age for the bottom of the sediment is about 1928 (Figure 2).

Sediment Grain Size
The grain size distributions of sediments in core TLC01 ranged from 0.28 to 1905.46 µm and displayed clear bimodal or multimodal styles ( Figure 3). Sediment size analysis of the entire TLC01 core using an algorithm of end-member modelling analysis (EMMA) [38,39] showed that there are four relatively independent end-element components (EM1, EM2, EM3, and EM4); however, these four end-element components do not adequately express the entire granularity data sequence ( Figure 4). These results indicate that the grain size of Lake Ta-Lung-Chi sediments reflect abundant environmental information.     . Grain-size data using the algorithm of end-member modelling analysis (EMMA) [39].
Appl. Sci. 2020, 10, x FOR PEER REVIEW 9 of 17 Figure 5. Grain size distribution and nonlinear trend fitting in sediment core TLC01 from Lake Ta-Lung-Chi. The blue triangles for the change points with the method of Piecewise linear fitting and trend changing points [51,52].
The trend depicted by component 4.0-16.0 μm in sediment core TLC01 was generally consistent with changes in component 16.0-32.0 μm, but opposite to changes noted in component >64.0 μm. The median particle size was consistent with the average particle size and with changes in component >64.0 μm. Following the analysis of Piecewise linear fitting and trend changing points [51,52], the vertical trends noted for each particle size could be divided into two stages, but the trend changing points differed between particle sizes ( Figure 5). The trend changing points for the median diameter, mean value, silty fraction (16-32 μm), and sandy fraction (>64 μm) all occurred at a depth of 23 cm. The trend changing points for clay fraction (<4 μm) and fine-silty fraction (4-16 μm) occurred at a 21cm depth. The trend changing point for coarse-silty fraction (32-64 μm) was at 32 cm, which differed from the other grain sizes. Taking the vertical change in the fine-silty fraction content of 4-16 μm as an example, with increasing depth below the trend changing point at 21 cm, the content tended to increase. However, above the trend changing point, the content decreased with increasing depth.

Discussion
Although EMMA is a very effective grain-size research tool [39,[53][54][55][56], through the result, the end-member model cannot effectively extract the grain size components affected by a single geological action (Figure 4). The grain size of the lake sediments was affected by the combined effects of multiple external stresses, or the strength of the same force has large fluctuations. Although the end member model has wide applicability [57,58], it is not an effective method to conduct research in this region. Existing studies have shown that the size characteristics of lake sediments are sensitive to the regional climatic environment [59][60][61]. In humid-semi-humid regions, on inter-annual and 10year scales, coarse-grained sediments indicate wet years with heavy rainfall while fine-grained sediments indicate dry years with low rainfall [60]. During the process of lake sedimentation, changes in rainfall affecting the intensity of surface runoff can also determine to a considerable extent the The median particle size was consistent with the average particle size and with changes in component >64.0 µm. Following the analysis of Piecewise linear fitting and trend changing points [51,52], the vertical trends noted for each particle size could be divided into two stages, but the trend changing points differed between particle sizes ( Figure 5). The trend changing points for the median diameter, mean value, silty fraction (16-32 µm), and sandy fraction (>64 µm) all occurred at a depth of 23 cm. The trend changing points for clay fraction (<4 µm) and fine-silty fraction (4-16 µm) occurred at a 21-cm depth. The trend changing point for coarse-silty fraction (32-64 µm) was at 32 cm, which differed from the other grain sizes. Taking the vertical change in the fine-silty fraction content of 4-16 µm as an example, with increasing depth below the trend changing point at 21 cm, the content tended to increase. However, above the trend changing point, the content decreased with increasing depth.

Discussion
Although EMMA is a very effective grain-size research tool [39,[53][54][55][56], through the result, the end-member model cannot effectively extract the grain size components affected by a single geological action (Figure 4). The grain size of the lake sediments was affected by the combined effects of multiple external stresses, or the strength of the same force has large fluctuations. Although the end member model has wide applicability [57,58], it is not an effective method to conduct research in this region. Existing studies have shown that the size characteristics of lake sediments are sensitive to the regional climatic environment [59][60][61]. In humid-semi-humid regions, on inter-annual and 10-year scales, coarse-grained sediments indicate wet years with heavy rainfall while fine-grained sediments indicate dry years with low rainfall [60]. During the process of lake sedimentation, changes in rainfall affecting the intensity of surface runoff can also determine to a considerable extent the amounts of coarse and terrigenous clastic materials entering the lake. In years with heavy rainfall, the capacity of surface runoff for erosion and transport is enhanced, and the sediment particle size in the runoff is increased; in dry years with low rainfall, surface runoff is low, making it difficult to transport coarse particulate matter to the lake, and the sediment particle size in the runoff decreases [62,63]. In arid and semi-arid regions, during humid periods, precipitation is high, lake water levels are high, and the sediment particles are coarse [64]. In summary, grain size of lake sediments can provide environmental information about lake runoff, which indirectly reflected the variation in the regional climate [61]. Studies on grain size in arid regions can also reveal the frequency and intensity of sandstorms that have occurred in the past [65,66], e.g., lakes in the Tibetan Plateau [59], Lake Chaiwopu, Xinjiang Province, China [67], Lop Nur, Tarim Basin, northwestern China [68], San Juan Mountains, Colorado [69], and Lake Hongjiannao, Shaanxi Province, China [70], all of which indicate that the debris transported by wind in arid areas is also a source of sediments to lakes. Through this research on Lake Ta-Lung-Chi in the Tian Shan, it was found that lake sediment size data can reflect regional climate change information.
Because the CRU data are based on data from existing meteorological monitoring stations, it was possible to reconstruct a complete set of high-resolution, monthly average surface climate data for the period 1901 to 2019 covering an area of 0.25 • × 0.25 • that includes all land types [22]. The weather station serving the area where Lak Ta-Lung-Chi is located in the middle of the Tian Shan is shown in Figure 1 [22]. The earliest record of meteorological stations in China related to the study of Lake Ta-Lung-Chi is from 1951. Therefore, it is necessary to verify whether the CRU precipitation data from before 1951 [22] are credible. Through this study, it was found that the three-point moving average curve of the silty fraction content (16-32 µm) was significantly consistent with the curve of total precipitation from April to September since 1950 ( Figure 6). The results suggested that the grain size of Lake Ta-Lung-Chi in the Tianshan Mountains sensitively recorded the information of precipitation changes. The proxies that are sensitive to precipitation changes in lake sediments are different in different regions, for example, biomarker compounds [34,71,72], Rb/Sr ratios [73], and magnetic susceptibility [74] are sensitive to changes in precipitation in some regions. In addition, the environmental information reflected by the grain size of lake sediments in different regions is also different, for example, it can reflect wind intensity and dust transport [75,76], riverine input [77] and lake water level changes [78]; however, it can be used as a useful indicator for quantitative reconstruction of long-term paleoprecipitation in the region of Lake Ta-Lung-Chi in Tianshan Mountains. The total precipitation from April to September showed a significant downward trend from 1930 to 1975, after which an overall increasing trend began. What caused this change? The study area is located in the mid-latitude zone, which is affected by westerlies, and the change in the North Atlantic Oscillation (NOA) is closely related to westerly intensity [79,80]. From the Figure 6, it was an interesting phenomenon that the precipitation change in the study region may be affected by the NAO, which showed that there is a certain correspondence between the Monthly North Atlantic Oscillation Index (station-based, December, January, and February) [81,82] and the precipitation curve reconstructed since 1930 in this study. The North Atlantic Oscillation Index is positively correlated with the westerly wind intensity [83], indicating that more water vapor is brought to the study area with increasing westerly winds, which may result in more precipitation. However, existing studies have shown that the precipitation in Central Asia and NAO have an anti-phased relationship on interannual to multi-centennial time scales [19,84,85]. The impact mechanism of the North Atlantic Oscillation on climate change in the Tianshan Mountains and even Central Asia may be different at different time scales, which is worthy of further discussion in the future. From the Figure 6, 1951 years ago, the reconstructed precipitation from April to September are inconsistent with the change from the CRU data, which was mainly because the earliest instrumental records of weather stations of China started at 1951 [86] (Figure 1), and thus the data before 1951 in this region were mainly calculated based on data from weather stations in the western section of the Tian Shan (within Kyrgyzstan, Figure 1). By comparing the data from our study area to data from a meteorological station (Chonkyzylsu, 42.20° N, 78.19° E; 1883-1996) in the Tian Shan [86], it can be seen that there is a large difference in precipitation between central Tian Shan and the region of Lake Ta-Lung-Chi in western Tian Shan. What is the reason for this difference? The main influencing factor From the Figure 6, 1951 years ago, the reconstructed precipitation from April to September are inconsistent with the change from the CRU data, which was mainly because the earliest instrumental records of weather stations of China started at 1951 [86] (Figure 1), and thus the data before 1951 in this region were mainly calculated based on data from weather stations in the western section of the Tian Shan (within Kyrgyzstan, Figure 1). By comparing the data from our study area to data from a meteorological station (Chonkyzylsu, 42.20 • N, 78.19 • E; 1883-1996) in the Tian Shan [86], it can be seen that there is a large difference in precipitation between central Tian Shan and the region of Lake Ta-Lung-Chi in western Tian Shan. What is the reason for this difference? The main influencing factor of precipitation is the source of water vapor. The HYSPLIT analysis showed that water vapor sources in the region of Lake Ta-Lung-Chi (42.45 • N, 83.29 • E) from April to September are not only affected by westerlies blowing from west to east, but also in part by winds blowing from east to west (8.33%) and from north to south (20.56%), which infers a local effect to water vapor transmission (Figure 7). At Chonkyzylsu station (42.20 • N, 78.19 • E), apart from the 10.56% of water vapor transmitted from north to south, the rest of the water vapor is transmitted from west to east (Figure 7). There are obvious differences between the two regions, which may be one of the reasons for the differences noted in precipitation.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 12 of 17 of precipitation is the source of water vapor. The HYSPLIT analysis showed that water vapor sources in the region of Lake Ta-Lung-Chi (42.45° N, 83.29° E) from April to September are not only affected by westerlies blowing from west to east, but also in part by winds blowing from east to west (8.33%) and from north to south (20.56%), which infers a local effect to water vapor transmission (Figure 7). At Chonkyzylsu station (42.20° N, 78.19° E), apart from the 10.56% of water vapor transmitted from north to south, the rest of the water vapor is transmitted from west to east (Figure 7). There are obvious differences between the two regions, which may be one of the reasons for the differences noted in precipitation. Chonkyzylsu station (42.20° N, 78.19° E) using the HYSPLIT [40] and MeteoInfo (TrajStat package) [42] models. The base map was derived from SRTM 90 m Digital Elevation Database v4.1 [87]. The red lines suggested the cluster trajectories of water vapor source for the site of Chonkyzylsu Station, and the yellow for Lake Ta-Lung-Chi.
The contents of clay fraction (<4 μm) and fine-silty fraction (4-16 μm) particles in Lake Ta-Lung-Chi sediments show a significant positive linear correlation (R 2 = 0.85, p < 0.0001), indicating that these components are influenced by the same factors. Furthermore, it was found that the total content of clay and fine-silty fraction (0-16 μm) particles are clearly consistent with the monthly precipitation in July ( Figure 6). However, during the 1940s and since 2000, the relationship between this particle content (0-16 μm) and July precipitation has been more complicated. Through previous research, it has been found that the sources of lake sediments in arid areas are not limited to surface runoff within the basin, and that atmospheric dust also represents an important source of lake sediments. In addition to this, studies on grain size in arid regions can reveal the frequency and intensity of sandstorms that have occurred in the past [65,66], e.g., lakes in the Tibetan Plateau [59], Lake Chaiwopu, Xinjiang Province, China [67], Lop Nur, Tarim Basin, northwestern China [68], San Juan Mountains, Colorado [69], and Lake Hongjiannao, Shaanxi Province, China [70], all of which show that the debris transported by wind in arid areas is also a source of lake sediments. A large amount of modern dustfall grain size data shows that the modal grain size of modern atmospheric dustfall materials is about 20 μm [69,88]. Therefore, some of the land-based materials in Lake Ta-Lung-Chi sediments less than 16 μm in size may also come from atmospheric dustfall, which may be the reason for the inconsistency observed between the variation curve of sediment content smaller than 16 μm and precipitation in the study area in July.  [40] and MeteoInfo (TrajStat package) [42] models. The base map was derived from SRTM 90 m Digital Elevation Database v4.1 [87]. The red lines suggested the cluster trajectories of water vapor source for the site of Chonkyzylsu Station, and the yellow for Lake Ta-Lung-Chi.

Conclusions
The contents of clay fraction (<4 µm) and fine-silty fraction (4-16 µm) particles in Lake Ta-Lung-Chi sediments show a significant positive linear correlation (R 2 = 0.85, p < 0.0001), indicating that these components are influenced by the same factors. Furthermore, it was found that the total content of clay and fine-silty fraction (0-16 µm) particles are clearly consistent with the monthly precipitation in July ( Figure 6). However, during the 1940s and since 2000, the relationship between this particle content (0-16 µm) and July precipitation has been more complicated. Through previous research, it has been found that the sources of lake sediments in arid areas are not limited to surface runoff within the basin, and that atmospheric dust also represents an important source of lake sediments. In addition to this, studies on grain size in arid regions can reveal the frequency and intensity of sandstorms that have occurred in the past [65,66], e.g., lakes in the Tibetan Plateau [59], Lake Chaiwopu, Xinjiang Province, China [67], Lop Nur, Tarim Basin, northwestern China [68], San Juan Mountains, Colorado [69], and Lake Hongjiannao, Shaanxi Province, China [70], all of which show that the debris transported by wind in arid areas is also a source of lake sediments. A large amount of modern dustfall grain size data shows that the modal grain size of modern atmospheric dustfall materials is about 20 µm [69,88]. Therefore, some of the land-based materials in Lake Ta-Lung-Chi sediments less than 16 µm in size may also come from atmospheric dustfall, which may be the reason for the inconsistency observed between the variation curve of sediment content smaller than 16 µm and precipitation in the study area in July.

Conclusions
By analyzing the characteristics of particle size in the sediments from Lake Ta-Lung-Chi in the Tian Shan, the possibility of using sediment grain size to reconstruct past changes of regional precipitation was discussed. The following conclusions are drawn.
The grain size of lake sediments in the Tianshan Mountains sensitively reflects the changes in precipitation in the basin, and it can be used as a useful indicator for quantitative reconstruction of long-term paleoprecipitation. The three-point moving average curve of silty fraction (16-32 µm) content is significantly consistent with the total precipitation curve from April to September since 1950. The total content of clay and fine-silty fraction (0-16 µm) is clearly consistent with the monthly precipitation in July.
The reconstructed precipitation (1950 years ago) with grain size of lake sediments was significantly different from the CRU database, which was due to the lack of meteorological data from monitoring stations in China before 1950, The total precipitation from April to September showed a significant downward trend from 1930 to 1975, and an overall increasing trend began in 1975.