The study focuses on Lake Balkhash, the terminal lake within the endorheic Ile-Balkhash basin (Figure 1). The upper parts of the catchment lie within western China; the lower parts and the lake itself are situated in south-eastern Kazakhstan. The lake occupies an average area of −17,000 km² (Pueppke et al., 2018a), while the entire catchment spans 115,000 km² (De Boer et al., 2021). Lake Balkhash has an average width of 30 km, reaching a maximum width of −74 km in the western section (Sala et al., 2016). Lake Balkhash is divided into distinct western and eastern sub-basins by the approximately 3.5 km wide Uzynaral strait (Figure 1). The western part of the lake contains fresh water and is shallow, with depths ranging from 3 to 11 m (Mischke, 2020; Myrzakhmetov et al., 2022). The eastern lake sub-basin is saline and substantially deeper, reaching depths of 26 m.

Inflow to the lake is dominated by the Ile River (or Ili River) (De Boer et al., 2021), which flows into the western sub-basin through a large (8,000 km²) delta system. The Ile River originates from the partially glaciated high altitude Chinese Tien Shan mountains and flows westward through south-eastern Kazakhstan (Tang et al., 2019). The Ile River delta serves as a natural hydrological regulator through short-lived wet and dry periods and plays a critical role in the ecological balance of the lake system (Dostay et al., 2012; Myrzakhmetov et al., 2022). The eastern lake sub-basin is fed by smaller rivers such as the Karatal, Aksu and Lepsy, which derive from the largely glacierised northern Tien Shan (Zhetysu Alatau) in eastern Kazakhstan (Kaldybayev et al., 2016). A small amount of freshwater enters the eastern sub-basin from the Ile-recharged western part. The waters of the eastern part of the lake are more saline than the western sub-basin due to a net reduction in water flow (−24.28%) compared to the average between 1936 and 1969 (Mukhitdinov et al., 2020); in particular, the Ayagoz River, which used to enter the lake from the north, as well as inflow from the Alakol lake system to the east, have no surface inflow today (Myrzakhmetov et al., 2022). The difference in water characteristics of the western and eastern Balkhash sub-basins has led to the nickname “One lake with two waters” (Kezer and Matsuyama, 2006; Huang et al., 2022).

Lake Balkhash receives its largest inflow over the warmer months (May to August). Peak discharge typically occurs in July and mostly results from glacial and snow melt (Petr, 1992; Sala et al., 2016; Mischke, 2020). Based on available average runoff data (1924–72), the Ile River has largest mean runoff with 17.4 km³/yr, accounting for −70–80% of total inflow to Lake Balkhash. Inflow from the Karatal, Aksu, Lepsy and Ayagoz Rivers to the eastern sub-basin comprises 20%–30% of total inflow (Petr, 1992; Kezer and Matsuyama, 2006). These eastern tributaries are also the major source of sediments (Petr, 1992). The variation in inflow and sediment input from the different tributaries has led to the division of the modern Balkhash basin into five distinct sub basins/zones, i.e., 1) West-South, 2) West-North, 3) Middle, 4) Lepsy Karatal, and 5) Kentubek (Figure 1). The depth of these basins increases from west to east, with lake bottoms at 334, 333, 327, 326, and 316 m asl, respectively corresponding to current water depths of c. 10, 11, 15, 16, and >20 m (Myrzakhmetov et al., 2022).

The Ile-Balkhash catchment experiences a continental climate and diverse geomorphology, spanning highest glaciated mountains of Khan Tengri peak rising to 6995 m asl, and terminating in the desert at 340 m asl with lowest point in the Lake Balkhash (Cao et al., 2022). The extreme climate supports desert and steppe vegetation near the lake, with deciduous trees and coniferous forests, alpine meadows and tundra vegetation with increasing altitude (Mischke et al., 2020). Summers in this region are typically hot, with temperatures frequently exceeding 40°C, while winters can be extremely cold, with subzero temperatures (Ahmed et al., 2022). These extreme temperature oscillations are also reflected in the mean surface water temperatures, which fluctuate between c. 0°C and 24°C from December to July, respectively (Barinova et al., 2017). This marked temperature variation has important implications for the water cycle, especially in terms of evaporation and ice cover on the lake. Precipitation is highly inconsistent in the Ile-Balkhash basin, with a general spring peak in rainfall and annual averages of 100–200 mm (Ahmed et al., 2022). Extended dry spells during the summer are common, leading to increased evapotranspiration rates of 8%–15% (Chub, 2000). The continental climate, in particular the high summer temperatures and irregular precipitation patterns, is a significant driver of water balance in the lake.

The Ile-Balkhash basin mainly comprised of sedimentary rocks such as carbonates, silicates, and alluvial deposits (Shen et al., 2021). Carbonate rocks are found extensively across the basin, especially near the upper part of the Ile River and the northern shore of Lake Balkhash. Silicate rocks are primarily located in the central part of the basin, along the Ile River, Aksu River, Lepsy River, and Karatal River. Igneous and metamorphic rocks are also present throughout the basin but are more common to the northern shoreline of the lake, characterized by a tectonically disturbed terrain as high as with 20–30 m asl. The western shoreline of the Lake Balkhash is dominated by clay deposits whereas, the south of the lake shoreline is dominated by flat sandy alluvial and wind-blown sediments deposits, mostly composed of siliceous materials (Sala et al., 2016; Shen et al., 2021). Thus, the overall configuration of the lake shoreline suggests that the lake receives substantial sediment input from the tributary rivers, which drain an actively uplifting catchment containing significant fine-grained unconsolidated sediments, as well as from aeolian transport by the dominant westerly winds which entrain sediment while sweeping across the loess steppe and dune landscapes (Sala et al., 2016; Fitzsimmons et al., 2020; Myrzakhmetov et al., 2022). The high sediment input, particularly in the south of the lake, impacts the turbidity of the water. The spatially diverse climate and topography results in complex interactions between precipitation, runoff, evaporation, accumulation of sediments, salts, and pollutants, ultimately affecting the lake ecosystem as a whole (Klein et al., 2014).

The primary and secondary datasets relevant to this study include remote sensing datasets, precipitation (P), temperature (T), the wind variables for direction (Wd) and speed (Ws), water level (WL) and landuse landcover (LULC).

The aim of this study is to identify, characterize, and explain turbidity variations over space and time at Lake Balkhash. To achieve these objectives, we undertook an iterative approach to analysing the datasets (Figure 2). To begin with, we examined the annual (9M) and seasonal (3M) turbidity data represented by the Normalized Difference Turbidity Index (NDTI). Season-specific analysis offers insight into water quantity and quality as it moves through the hydrological system. We simultaneously interrogate the annual (9M) and seasonal (3M) variability of meteorological (e.g., P, T and Ws/Wd) and anthropogenic influenced variables (e.g., WL and LULC) to match the data consistency to that of satellite imageries obtained for turbidity analysis. In order to better understand spatial variability within the large lake basin, we collected −2000 random point samples from time-series data for NDTI, precipitation (P) and temperature (T) (Supplementary Figure S2). Finally, we performed a time-series trend analysis using linear regression and Pearson correlation analysis on both NDTI and the climatic and anthropogenically influenced variables to investigate the socio-climatic effects on water surface dynamics and quality. The idea of obtaining −2000 random sample points for P, T and NDTI in this study is to select a representative subset of data points from a larger dataset. This method ensures equal opportunity for each data point to be chosen, thus minimizing bias and accurately reflecting the entire data population.

Here, we first present the trend analysis of precipitation (P), temperature (T), the wind variables for direction (Wd) and speed (Ws), over the 1991–2022 period at both annual and seasonal scales. Overall, we observe a trend of significant correlation for temperature but the long-term averaged precipitation and wind data do not appear to yield any correlation through time. The details of these observations are further elaborated in the subsequent sub-sections.

Overall, we observe a trend of significant correlation for water level (WL), with increasing over the 1991–2022 period at both annual and seasonal scales. Similarly, the different LULC variables around the Lake Balkhash shows significant correlation with time over the study period. The data in detail are presented in the following sub-sections. The details of these observations are further elaborated in the subsequent sub-sections.

The overall spatio-temporal trend of NDTI at Lake Balkhash for the observation period is provided in Figure 8. Averaged NDTI values range from −0.149 to −0.044 and therefore correspond to slightly turbid conditions. Time-series analyses of the annual NDTI trend show an overall decrease in turbidity through time (r = −0.865; p = 0.000), with short-lived peaks and troughs in NDTI values (Figure 9A; Table 2). The period 1991–1996 experienced the highest turbidity values, with a peak of −0.059 in 1996. Other short-lived peaks occurred in 2001 (−0.091) and 2010–2011 (−0.090 and −0.091 respectively), and smaller turbidity peaks in 2017 and 2019.

We also observe differences in turbidity between the seasons (Figure 9A). Turbidity during the spring season fluctuates more substantially between the years than during summer and fall; spring NDTI values indicate high turbid water. There is a significant and strong trend of decreasing turbidity for each of the seasons (spring r = −0.620; p = 0.002; summer r = −0.815; p = 0.000; fall r = −0.856; p = 0.000), which underscores the annual time-series trend (Figure 9A; Table 2). The trend of decreasing turbidity is strongest for the fall and summer seasons. Less turbid water is likely to be the product of a range of factors including reduced runoff into the lake, changes in precipitation patterns, land use, or improved water quality management practices.

We summarise the spatial variability of both annual and seasonal NDTI values derived from the different zones in Figure 9B. We observe the lowest annual average NDTI in the central Zone III; turbidity increases towards the eastern and western extremities of the lake (Figure 9B; Table 3). This spatial trend is echoed by NDTI values for the summer and fall months. The spring season experiences relatively higher ordinates of turbidity and a less pronounced spatial trend (Figure 9C; Table 3).

To further assess the dominant factors affecting lake turbidity in an arid climate, we investigated potential correlation between the five exploratory variables, i.e., P, T, Ws, WL, LULC, and NDTI variation. On an inter-annual basis, NDTI displays a negative correlation with most of the variables. We find no significant relationship between NDTI and P (r = −0.074, p = 0.371), and between NDTI and Ws (all 4 stations) (Table 6). However, we observe a moderate negative association between NDTI and T (r = −0.532, p = 0.010). We also find a significant negative correlation between NDTI and WL (r = −0.696, p = 0.001) (Table 6). These correlations suggest that higher turbidity is associated with lower temperatures and lower water levels and vice versa.

Similarly, we conducted statistical trend analysis to establish the correlation between the major LULC classes and NDTI (Table 6) in the present study area. Examining the major LULC classes around the lake, we observe a strong positive correlation between NDTI and bareland (r = 0.764, p = 0.000) and a moderate positive correlation with grassland (r = 0.558, p = 0.011). However, the percentage statistics of these two variables indicate a relative decrease in the areal extent (Figure 7; Supplementary Table S2). On the other hand, shrubland (r = −0.771; p = 0.000), and wetland (r = −0.665; p = 0.000), which dominate the southern part of Lake Balkhash, exhibit a negative correlation with NDTI at the annual scale (Table 6). Whereas, the statistics of these variables reveal an increase in the areal percentage over time (Supplementary Table S2). An overall assessment suggests that an increase in shrubland and wetland, with a relative decrease in grassland and bareland in the vicinity of Lake Balkhash, has a significant impact on reducing sediment availability to the lake turbidity (Table 6). Also, the other LULC variables such as cropland, forests, urban land, and snow and ice as distributed in the basin exhibit strong and negative correlation with NDTI, with their relatively low and indirect contribution to turbidity dynamics.

We further repeated this statistical analysis for the available seasonal datasets to better understand the trends in variability specifically for water levels and temperature with the NDTI (Table 6). We observe a strong negative correlation between NDTI and both T (r = −0.659; −0.564 and p = 0.001; 0.006) and WL (r = −0.605; −0.606 and p = 0.003; 0.003) during the spring and summer seasons. However, during fall, NDTI only yields a statistically significant correlation with WL and not with T (r = −0.578 and p = 0.006) (Table 6). This result suggests that WL fluctuations may have a more pronounced impact on turbidity, particularly during fall. These findings emphasize the influence of T and WL on turbidity dynamics at Lake Balkhash.

We observe clear spatial and temporal variability in the NDTI of Lake Balkhash waters over the time period 1991–2022. Since the lake is situated in an arid climatic setting (Pueppke et al., 2018a), we selected the five exploratory variables i.e., P, T, Wd/s, WL and LULC as those most likely to elucidate the controls on lake turbidity at annual and seasonal timescales. We group these variables into two categories: a) entirely natural factors such as precipitation, temperature and wind, and b) parameters reflecting a coupling between climate and anthropogenic influence. We assume water level to correspond to the latter category, since Lake Balkhash is a regulated system and its flow is almost regulated by dams and diversions (Dostay et al., 2012; Pueppke et al., 2018a).

We find minimal correlation between NDTI and precipitation (Table 6), and attribute this to the weak effect of direct precipitation over the lake in the study area (Duan et al., 2020; Salnikov et al., 2023). This is despite of an overall trend of increasing precipitation (Figure 3A) over the observation period reported across the river basin (Matsuyama and Kezer, 2009; Propastin, 2012; Yu et al., 2021), in particular in the spring and summer seasons (Duan et al., 2020). Our results clearly suggest that local precipitation over the lake has minimal impact on turbidity variation, most likely relating to the small absolute amount of rainfall as well as high rates of evaporation loss from large water bodies (Dostay et al., 2012; Pueppke et al., 2018a; Yu et al., 2021; Albarqouni et al., 2022) arid and semi-arid climatic settings. At the catchment scale, precipitation is much higher in the headwater regions, and is particularly pronounced in the climatically diverse Ile-Balkhash catchment; catchment-scale rainfall and snow has an indirect impact on surface runoff and stream discharge upstream region (Petr, 1992; Göransson et al., 2013; Yu et al., 2021). The impact of catchment-scale precipitation is evidenced by large floods following extreme rainfall events; such events controls a considerable part of the sediment load being transported into dryland lakes such as Lake Balkhash, so driving short-lived turbidity fluctuations (Karthe, 2018; Duan et al., 2020). Unfortunately, the scope of our study prevented the investigation of specific events such as floods and the quantification of NDTI resulting from such events.

Although impacts of wind strength has been reported to impact turbidity in arid zone water bodies (Wu et al., 2013; Ouni et al., 2019), we observe no statistical correlation between turbidity and wind regime parameters at Lake Balkhash (Table 6). It is also been reported that western part of the lake observed less frequent strong winds than in the eastern part (Ivkina, 2022). This may be explained by the local topographic context of the Lake Balkhash coastline and surrounds. The western and northern lake shorelines are up to 30 m above the base lake level (Figure 1), with rocky uneven terrain; this relatively high topographic difference may act as a natural barrier to air flow, preventing strong winds and associated storm surges reaching inland and causing turbidity within the water (Ivkina, 2022). On the other hand the deeper, narrower, and dissected eastern sub-basin prevents the generation of sizeable waves and therefore there is no or limited wind-driven turbidity in that region (Sala et al., 2020; Ivkina, 2022). The one place where wind regimes may have an influence on turbidity is in the southern part of the lake, which has a very low relief shoreline, with abundance of loose alluvial and aeolian sediments and slightly stronger wind speeds with a N-NE orientation (Issanova et al., 2015; Sala et al., 2020). Under these conditions, winds may disturb the lake bottom sediments in the shallow southern part, partly contributing to turbidity in zones I and II (Petr, 1992; Ivkina, 2022).

More significantly, we observe a clearer and stronger connection between NDTI and temperature (T) as well as water level (WL). These parameters likely govern most of the turbidity fluctuations at annual and seasonal time scales. We propose that temperature is one of the dominant factor in this region, controlling process-feedbacks for turbidity in Lake Balkhash. Several studies have documented increasing temperatures in this region (Propastin, 2012; Qi et al., 2020), resulting in high evaporation rates (Abdrahimov et al., 2020) and exacerbating discharge patterns through an increase in water level (Qi et al., 2020; Moldakhanova et al., 2023). It is important to note that the major contributing rivers - Ile, Karatal, Lepsy and Aksu - are fed by glaciers and snowmelt, and are consequently highly sensitive to temperature-driven impacts (Kaldybayev et al., 2016; Sala et al., 2016; Mischke, 2020; Cao et al., 2022) on runoff and the seasonal distribution of water. During winter, the lake is mostly frozen; the lake surface starts melting in the early March, with the onset of spring (Petr, 1992). This phenomenon elevates seasonal turbidity variation in the lake due to the loss of lake ice cover and large temperature fluctuations (Ivkina, 2022). Rising temperatures also affect spring snowmelt in the upstream reaches (Duan et al., 2020; Zhang Q. et al., 2022), increasing water levels as well as sediment input to the lake (Karthe, 2018; Panyushkina et al., 2018). Frequent event-based flooding due to local snow melt in the spring (Petr, 1992; Yapiyev et al., 2017; Salnikov et al., 2023) also contributes to the lake turbidity. Although, both spring and summer are dominated by high water inflow from the upstream regions, the lake water level is ultimately controlled by regulated discharge from the Kapshagay dam which also controls sediment transport downstream (Pueppke et al., 2018a; Panyushkina et al., 2018; Mukhitdinov et al., 2020). Nevertheless, high temperatures combined high water inflow moderates the input of sediment to Lake Balkhash in the spring, reducing spring turbidity relative to summer (Sala et al., 2016; Panyushkina et al., 2018). By fall, as temperatures decrease, there is reduced water inflow into the lake (Petr, 1992; Propastin, 2012), which further decreases the sediment contributions and subsequent turbidity.

Human induced LULC dynamics also have a major impact in driving the water turbidity in inland lakes from local to regional scale (Li et al., 2014; Baltodano et al., 2022; Zhang L. et al., 2022). The different LULC pattern observed across the basin has a significant bearing in driving the turbidity of a lake. For instance, the expansion of grassland or shrubland areas around the lake serves as a protective cover and the root systems effectively bind the soil, reducing erosion caused by wind and water (De Baets et al., 2011; Zhang et al., 2013). Furthermore, they intercept rainfall, minimising soil disturbance and consequently reducing surface runoff carrying sediments into water bodies (Mohamadi and Kavian, 2015). This natural sediment trapping mechanism not only preserves soil integrity but also improves water quality by limiting the influx of pollutants into lakes, thereby safeguarding aquatic ecosystems and maintaining ecological balance. In our study, we observed an increase in shrubland areas (Figure 7B) within the basin, which has likely contributed to soil erosion control and, consequently, might have helped in reducing the lake water turbidity. The Ile-Balkhash basin has also been marked by an expansion of wetland areas, mostly associated with the southeastern part of the lake (Cao et al., 2022). These wetlands serve as natural filters, trapping sediment carried by runoff before it reaches the lake (Johnston, 1991; Mereta et al., 2020), thereby decreasing turbidity levels. In summary, the increase in wetland areas around Lake Balkhash can act as a vital mechanism for mitigating turbidity and promoting the overall health and ecological balance of inland lakes (Cao et al., 2022). Conversely, barren landscapes are prone to erosion, leading to sediment runoff and positively contribute to water turbidity. A change in barren areas in the region can significantly contribute to turbidity dynamics in nearby water bodies (Li and Xia, 2023). With vegetation replacing barren areas, soil erosion is minimized as roots bind the soil and vegetation intercepts rainfall, reducing sediment distribution along the lake river system in the Ile -Balkhash basin. Several studies put forward a strong link between the rapid urbanization and industrial areas with the detriment of river and lake water quality (Li et al., 2014; Pueppke et al., 2018b; Duan et al., 2020; Baltodano et al., 2022). However, in the present study reveals that although there is strong and negative correlation, of cropland, forests, urban land, and snow and ice majorly distributed in the upstream region, has an indirect control in turbidity variation in Lake Balkhash (Pueppke et al., 2018a; Duan et al., 2020). Therefore, these variables singularly or in combination impact the water quality of the associated water resources (Tahiru et al., 2020).

The spatial variability of NDTI values is dominated by inflow from the Ile River as the largest contributor to the Lake Balkhash (Petr, 1992; Mukhitdinov et al., 2020; De Boer et al., 2021), specifically to zones I and II in the western sub-basin. The Ile River is also responsible for most of the sediment influx responsible for its turbidity, giving the western sub-basin a yellow-gray color (visibility 5–10 m) and silty-sandy lake bottom deposits (Petr, 1992; Krupa et al., 2020; Mischke, 2020). The shallow western sub-basin, and its large surface area, are additional possible factors contributing to the high turbidity in that part of the lake (Ivkina, 2022; Myrzakhmetov et al., 2022). The eastern sub-basin, in particular zones IV and V, receives significantly lower sediment and water input (Petr, 1992; Mischke, 2020); it is also deeper with smaller surface area relative to water volume, and therefore has greater potential for clearer water (Ivkina, 2022). The central zone III in the eastern sub-basin experiences the lowest turbidity, since it is least affected by sediment inflow from the rivers. In summary, our study indicates a strong connection between water level and temperature as key drivers of turbidity at Lake Balkhash. Additional factors may contribute to water quality fluctuations which are subject to the influence of human activities and have not been considered in this study. The increase in human population in the Ile-Balkhash basin in the last several decades (Duan et al., 2020) has led to increased water consumption and changes in land use practices (Mukhitdinov et al., 2020) with an ultimate loss of natural land cover (Mischke, 2020; Huang et al., 2022). In addition, poor management of waste and sewage disposal from the municipal and industrial point sources, are also likely to have a localized impact on lake water quality (Karthe et al., 2015; Karthe, 2018; Mukhitdinov et al., 2020).

Monitoring and managing turbidity variations in waterbodies is essential for keeping the ecosystem healthy and for maintaining the services that we get from them. The time series analysis of the NDTI for Lake Balkhash reveals a significant reduction in water turbidity, signifying an enhancement in water clarity over the past 3 decades (Tilekova et al., 2016). The observed decrease in turbidity over 3 decades in Lake Balkhash can have several potential impacts on the ecology and ecosystem services (Huang et al., 2022). Clearer water conditions in the lake system has led to reduced mineralization and increased penetration of sunlight, which promotes the growth of aquatic vegetation and altering habitat structure (Krupa et al., 2014; Barinova et al., 2017). This, in turn, has affected the distribution and abundance of aquatic organisms and may lead to shifts in species composition (Pueppke S. et al., 2018). Some studies have reported the increase in zooplanktons, phytoplankton’s and fish spawning behavior and reproductive success of some species over the other in the Lake Balkhash (Tilekova et al., 2016; Pueppke S. et al., 2018). Overall, the changes in turbidity levels can influence trophic interactions and ecosystem functioning, potentially disrupting food webs and altering nutrient cycling processes (Imentai et al., 2015; Krupa et al., 2020). Therefore, highlighting the importance of continuous environmental monitoring and in-depth studies for effective management of turbidity is crucial for preserving the health of Lake Balkhash’s ecosystem.

This study obtains valuable insights into the likely drivers of water quality variability and trends in Lake Balkhash. However, it is important to note that our study relies entirely on infrequent Earth observation datasets, and as such is limited by data gaps and may not adequately capture seasonal and short-term turbidity events. The accuracy of our interpretations can only be obtained by increasing the spatial and temporal resolution of the available datasets. In addition, it is important to recognize the significance of local factors specific to Lake Balkhash, including shoreline and lake basin characteristics, human land use practices within the basin, and inflows and sediment input from individual tributaries. These local influences can have a significant impact on turbidity levels within the lake.

The calibration and validation of our turbidity calculations against ground-based measurements are essential steps in ensuring the reliability of satellite-based assessments. The relatively remote setting and limited infrastructure at Lake Balkhash currently limits the opportunity for generating in-situ turbidity datasets over the observed time period both at spatial and temporal scale (Barinova et al., 2017; Krupa et al., 2020). However, several studies have make use of the remote sensing based indices (e.g., NDTI) which found to have strong correlation with the in-situ observation to highlight the usefulness of the satellite imageries in water resource management in various climatic settings (Baughman et al., 2015; Bid and Siddique, 2019; Elhag et al., 2019; Kaplan et al., 2019). Putting this forward, future efforts should be directed toward collecting and integrating ground-based data for an observed period to improve the accuracy and reliability of turbidity forecasting using remote sensing observations at Lake Balkhash.