Modern distribution of Cladocera in lakes of northern China and western Mongolia and its environmental implications

Highlights

• •The first large-scale biogeographical investigation of cladocerans in lakes in northern China and western Mongolia.
• •Lake level, nutrient, and salinity are the main controls of the structure of modern cladoceran communities.
• •Subfossil cladoceran-inferred lake level reconstructions are consistent with lake areas extracted from Landsat images.

Abstract

Cladocerans are widely distributed in aquatic ecosystems and are sensitive to environmental changes; hence, they are commonly used to trace ecological changes in aquatic environments. However, the distribution of cladocerans in northern China is poorly known. We investigated the occurrence of cladocerans in 136 water bodies (including lakes and reservoirs), with varying hydro-chemical and climatic gradients, in northern China and western Mongolia. We identified 38 cladoceran taxa in the surface sediments of 88 different water bodies, while the remaining lakes, mainly saline, had no cladocerans. The spatial distribution characteristics of cladoceran species and the environmental factors determining their community composition were investigated using canonical correspondence analysis. The results demonstrate that water depth, trophic status, and water salinity were the top three variables influencing the composition of cladoceran taxa. Transfer functions between the surface sediment cladocerans and lake level and nutrient (NH₄) levels in 62 lakes (26 lakes lacked environmental data) were established via multivariate statistical analyses. Using the subfossil cladocerans of Jili Lake, lake level and nutrient values were reconstructed for the past 100 years. The results agree overall with the observational data, and they reveal that the changes in the cladoceran community reflected mainly lake level fluctuations before 1988 CE, and mainly nutrient changes thereafter. Our research clarifies, for the first time, the environmental significance of modern Cladocera in the lakes of northern China and western Mongolia, and it provides a foundation for the future reconstruction of long-term changes in the ecological environment of lakes.

Graphical abstract

1. Introduction

Lakes are important components of global aquatic ecosystems, and they possess distinct feedback mechanisms and relatively closed natural conditions, making them sensitive to global climate changes [[1], [2], [3]]. In recent years, many lake ecosystems have undergone varying degrees of degradation, characterized by reduced water capacity, homogenization of biological communities, a sharp decrease in species diversity, disruption of their ecological balance, and a decline in ecosystem services [4]. In arid and semi-arid regions, the compound effects of increasing water body mineralization, drought, and dwindling water resources are not only accelerating lake eutrophication but also promoting salinization, leading to the disruption of ecological balances [5,6]. Therefore, exploring the evolutionary processes of the lacustrine environments and their driving factors is crucial for the ecological protection and sustainable development of lakes [7,8]. Zooplankton, which are a critical component of aquatic ecosystems, play a key role in the cycling of materials and in energy flow. Characterized by a vast population size, small individual size, and short lifespan, zooplankton effectively bridge trophic levels and can provide reliable information on the dynamics of aquatic ecosystems [[9], [10], [11]].

Cladocera, also known as water fleas, are a significant component of zooplankton. Their sensitivity to environmental changes makes them reliable indicators of aquatic ecological health [12]. These organisms possess a transparent, chitinous exoskeleton that includes various anatomical features, such as a head shield, abdominal shell flaps, postabdomen, and tail spines, along with an ephippium. After animal death, these chitinous remnants are deposited in lake sediments, making them valuable for studying lacustrine paleoenvironments on long timescales [13]. Cladoceran research includes qualitative descriptions of the evolution of lake biota evolution and their productivity [[14], [15], [16], [17], [18]], and paleolimnological reconstruction, considering multiple stressors such as climate change and human activities. Limnological variables affecting cladocerans include temperature [[19], [20], [21]], lake level [[22], [23], [24], [25], [26]], salinity [27,28], nutrients [29,30], pH [31], fish predation [32], plant coverage [33], and chlorophyll content [34]. Modern cladoceran ecology is well-established in Europe and America and is widely applied in ecology and paleoecology [31,[35], [36], [37]]. In China, research on modern cladoceran communities is relatively scarce. Dong et al. [33] and Zhu et al. [38] conducted a statistical analysis of cladoceran assemblages in surface sediments in the middle and lower reaches of the Yangtze River and Yunnan province, and they noted that the cladoceran communities were influenced by macrophyte density and site altitude. Li et al. [39] analyzed 40 lakes in the Badain Jaran Desert and found that most salt lakes were devoid of cladocerans. It has been confirmed that cladocerans reflect changes in lake ecology, with regional differences. However, there is a lack of comprehensive studies on cladoceran in northern China, a region characterized by arid and semi-arid climates. We would like to emphasize two important aspects of the study area. First, these dry areas differ from the lakes where cladoceran research has already been conducted, meaning that the composition and ecological significance of Cladocera may be different. Second, lakes in arid areas have important ecological and resource functions, given that they are primary surface water resource, and hence they have an especially high ecological value.

The paper presents the detailed analysis of cladoceran assemblages in the surface sediments of 136 water bodies across northern China and western Mongolia. The author’s intension was to determine the spatial distribution of cladocerans and identify the key environmental factors influencing their assemblages. The geographical range of this study spans 13 provinces in northern China and one province in western Mongolia, extending from the eastern regions to the northwestern inland region. These areas are known for their significant natural variability, and they host numerous lakes that are vital for climate regulation, water supply, flood control, and the provision of other ecological services. The specific objectives of the presented work are: (1) to conduct a comprehensive biogeographical survey of Cladocera communities and the physicochemical properties of these water bodies, and to establish transfer functions between the cladoceran community and key environmental factors such as water and nutrient levels; (2) to utilize Jili Lake as a case study for the cladoceran-based quantitative reconstruction of the history of lake level and nutrient changes and to validate the reconstruction results.

2. Materials and methods

2.1. Regional settings

The study was conducted at 136 lakes and reservoirs in northern China and western Mongolia (81.37°–132.37°E, 33.03°–49.48°N) (Fig. 1A), at altitudes ranging from 0 to 2860 m, and covering diverse terrains such as mountains, hills, basins, and plains. Most of these water bodies are located in the temperate zone of the Northern Hemisphere, with the eastern part having a monsoon climate and the western part having a continental climate. The climatic differences within the regions are significant, with the mean annual air temperature (MAT) from 1970 to 2000 CE varying between −5.7 and 15 °C, and the mean annual precipitation (MAP) varying from 398 mm upwards. The meteorological data for the sampling points were derived from the WorldClim 2.1 global gridded interpolation meteorological reanalysis dataset [40], using temperature and precipitation values at the grid points (resolution 30″×30″) as representative climate parameters. Analysis of the collected water samples revealed substantial differences in physicochemical characteristics across these regions, with water depths ranging from 0.05 to 77 m, salinity (in terms of total dissolved solids, TDS) from 0.03 to 155.5 g/L, and pH from 7.1 to 10.5 (Table 1).

Fig. 1.

Locations of the lake surface and core sediment samping sites. (A) Locations of the surface sediment sampling sites in northern China and western Mongolia. They comprise 40 samples from Li et al. [39] and 96 samples collected in the present study. The red star indicates the location of core JL18–01-A from Jili Lake; (B) location of Jili Lake and Ulungur Lake. The canal is the result of a project to channel the Irtysh River into Ulungur Lake. Red arrows indicate the direction of rivers. The base map is based on the standard map of China: GS(2019)1823. The topographic map is from the National Platform for Common GeoSpatial Information Services (https://map.tianditu.gov.cn/).

Table 1.

Summary statistics of environmental variables in the 136 water bodies.

Variable	Mean	Maximum	Minimum	Standard deviation
Lon (°E)	109.83	132.37	81.37	11.34
Lat (°N)	41.77	49.48	33.03	3.48
Alt (m)	952.7	2860	0	536.7
MAT (°C)	5	15	−5.7	4.1
MST (°C)	20.2	26.7	9	3.3
MAP (mm)	398	877	75	158
AirT (°C)	25.4	32.8	16	3.2
Depth (m)	3.8	77	0.05	10.5
WaterT (°C)	25.1	34	0.6	6.1
TDS (g/L)	29.1	155.5	0.03	47.4
pH	9.1	10.5	7.1	0.6
DO (mg/L)	8.6	20.6	0.5	3.5
NH4 (mg/L)*	1.03	9.49	0	1.78
NO3 (mg/L)*	0.45	5.41	0	0.88
PO4 (mg/L)*	0.55	7.31	0	1.24

Note: Lon, longitude; Lat, latitude; Alt, altitude; MAT, mean annual air temperature; MST, mean summer air temperature; MAP, mean annual precipitation; TDS, total dissolved solids. Meteorological data (1970–2000 CE) are from WorldClim 2.1 [40]. * “Zero” represents data values that were either extremely low or even below the instrumental background. The 128 samples data of environmental variables from Li et al. [39] and Xiang et al. [46,47], remaining samples data from this study.

We conducted a case study at Jili Lake, and applied modern ecology to validate the environmental reconstruction performance. We reconstructed changes of the lake level and nutrient state over the past century, based on the subfossil cladocerans of a sediment core from the lake [41]. Jili Lake (46.85°–47°N, 87.33°–87.53°E) is an inland freshwater lake, located in a low-altitude basin (486 m) in the arid northwest China. Intensive human activities have had a significant impact on the development of the lake and its aquatic communities. The lake covers an area of ∼74 km², with a maximum depth of 16.6 m and average depth of 9.9 m [42]. The Ulungur River passes through Jili Lake, and the water of Jili Lake enters Ulungur Lake via the Kuyigat River (Fig. 1B) [43]. The lake hosts a total of 145 phytoplankton species, and 100 zooplankton species, comprising 27 protozoan species, 54 rotifer species, 15 cladoceran species, and 4 copepod species [44]. Fish farming in Jili Lake dates back to the establishment of the Fuhai fishery in 1958. Furthermore, with the continuous input of nutrients, the nutritional status of Jili Lake has changed from oligotrophic to mesotrophic [44]. In its natural state, the lake level of Jili Lake is 2 m higher than that of Ulungur Lake [41]. In October 1987, the regional government completed a project to channel the Irtysh River into Ulungur Lake, located in the northeastern part near the Irtysh River [45]. This project introduced freshwater from the Irtysh River into Ulungur Lake [43], which made it possible for the lake level in Ulungur Lake to increase quickly and surpass even that of Jili Lake within a short period of time. As a result, the saline water from Ulungur Lake can now flow into Jili Lake.

2.2. Sample collection and laboratory analysis

Surface sediment samples from the lakes and reservoirs from various regions of northern China and western Mongolia were collected in 2019, 2021, and 2023. Most of the samples were taken from the center of the lakes using a small grab sampler, collecting the top 1–3 cm of sediment, as well as several samples from the top of drilling cores. In total, samples from 136 sites were used for Cladocera analysis, comprising 40 samples collected Li et al. [39] and 96 samples from the present study (Fig. 1A; Table S1). On-site measurements of physicochemical parameters of the lakes/reservoirs, including water depth, TDS, pH, instantaneous water temperature (WaterT), instantaneous air temperature (AirT), and dissolved oxygen (DO), were conducted using portable water quality analyzers (AZ86031 and YSI). Water chemistry parameters like nitrate nitrogen (NO₃), ammonia nitrogen (NH₄), and phosphate phosphorus (PO₄) were measured at the Key Laboratory of Western China’s Environmental Systems (Ministry of Education), Lanzhou University. At 88 sampling points, we only have climate data (MAT, MST, MAP) and water depth information, with some physicochemical data missing due to limitations during field work. Only 62 water bodies have a complete set of environmental data.

Additionally, a 355-cm-long core (JL18–01-A) was obtained from the central part of Jili Lake, at a water depth of 16.6 m using a piston-corer [41]. The core was divided into 1-cm intervals for subsequent analysis, and 11 samples of organic matter from the core were sent to Beta Analytic Laboratory in Florida, USA for AMS ¹⁴C dating [41]. Profiles of ²¹⁰Pb and ¹³⁷Cs and the identification of subfossil cladoceran compositions were determined by Hu et al. [41].

The pretreatment and identification of cladoceran samples were completed at the Key Laboratory of Western China’s Environmental Systems, Lanzhou University, following standard methods [48,49]. Around 0.2–0.3 g of dry sample or 0.5–1.0 cm³ of wet sample was treated with 50–60 ml of 10% KOH solution and heated on a hotplate at 60 °C for ∼45 min. After cooling, the samples were washed through a sieve with a mesh size of 38 µm, transferred to a centrifuge tube, and left to settle for 24 h. The supernatant was then removed and the sample adjusted to a known volume, after which 1–2 drops of ethanol were added for long-term preservation. For slide preparation, the samples were shaken and an appropriate amount was removed using a pipette, placed on a microscope slide, spread evenly with a fine needle, and a few drops of staining agent (Safranine) added. After the water had almost evaporated, a mounting medium was added and a cover slip emplaced. Identifications of cladoceran species were performed under an optical microscope at 200× or 400× magnification [48,50,51]. A minimum of 100 individuals per sample were identified, and well-preserved, representative cladocerans were photographed. The number of head shields, shell valves, postabdomens, tail spines, and ephippia were all recorded. Due to the uneven preservation of different parts of cladocerans, the most abundant body part was selected for analysis [52,53]. To identify different compositional groups of cladoceran assemblages, we used a constrained hierarchical clustering method, constrained incremental sum of squares (CONISS), implemented in Tilia 2.6.1 [54].

2.3. Statistical analysis

2.3.1. Ordination

Prior to the data analysis, environmental variables, including MAT, MST, MAP, AirT, depth, WaterT, TDS, DO, and cladoceran composition data were square root-transformed, while nutrient data (NO₃, NH₄, PO₄) were subjected to [log₁₀(x + 1)] transformation [33]. Detrended correspondence analysis (DCA) was applied to the species data. If the length of the first axis is < 3 SD units, linear redundancy analysis (RDA) is used, and if the length is > 4 SD units, canonical correspondence analysis (CCA) is deemed more appropriate for analyzing the relationship between species composition and environmental factors; while if the length is between 3 and 4 SD, both methods are suitable [55]. To address the issue of collinearity among environmental factors, an analysis of variance inflation factors (VIFs) was used to eliminate some of the environmental variables. Environmental factor selection was conducted using forward selection, and the significance of environmental variables was tested using the Monte Carlo method (with 999 unrestricted permutations). Ordination analyses and graphing were performed using Canoco 5.0 software [55]. Key environmental factors were selected based on RDA or CCA ordination results for independent testing. The importance and significance of variables were assessed using Monte-Carlo permutation tests, with p < 0.05 considered statistically significant.

2.3.2. Development of a model for quantitative paleoenvironmental reconstruction

The most significant environmental variables were selected for developing transfer functions. Detrended canonical correspondence analysis (DCCA) was then used to determine whether the cladoceran surface sediment training set was suitable for linear or nonlinear model development. The response pattern of cladocerans to environmental variables was assessed based on the effective gradient length of the environmental variable on the first principal axis of the DCCA. If the gradient length exceeded 2.4 SD, a nonlinear weighted averaging model was used to establish the transfer function between cladoceran and environmental variables; otherwise, a linear model was used [56]. C2 software [57] was used for transfer function selection. For nonlinear models, weighted averaging (WA), weighted averaging partial least squares (WA-PLS), and the modern analogue technique (MAT) were evaluated to determine the optimal model and components. Random forest (RF) is a machine-learning approach based on regression tree analysis [58], which was performed with “randomForest” [59] in R software. Transfer functions were tested on the modern cladoceran dataset; 60% of the dataset was used for calibration and 40% was used to evaluate the transfer functions, thereby preventing circular reasoning [60].

2.3.3. Model performance

The performance of the transfer functions was evaluated via error testing and other statistical measures. Leave-one-out cross-validation was used to estimate error values, to assess the accuracy of the model and reconstruction results. By removing samples one-by-one and retraining the model, the transfer function was validated using the remaining n-1 samples. Root Mean Square Error (RMSE) and Root Mean Square Error of Prediction (RMSEP) were calculated to assess the error, and the coefficient of determination (R²) was used to measure the correlation between measured and predicted values. A higher R² value indicates better model performance. Significance testing was used to determine the statistical relevance of the environmental reconstruction. A reconstructed result is deemed reliable only if an environmental factor explains the variation in cladoceran composition better than a random variable. The reconstruction is deemed statistically significant when the explained variance in the observed environmental variable calibration set exceeds 95%. This testing process was conducted using the “PalaeoSig 2.1–3” package in R software [61], employing the “randomTF” function [62].

2.4. Landsat and lake area mapping

Landsat satellite images provide multispectral imagery of the Earth’s surface for ∼30 years since its launch (1989–2021), with 30 m resolution, from the Geospatial Data Cloud (GDC) (https://www.gscloud.cn/). These images were used to extract the interannual trends of lake areas. The areal extents of Ulungur Lake in 1960 and 1970 were obtained using the Chinese lake dataset (available from http://www.tpdc.ac.cn/en/, last accessed on Aug. 9, 2021). Landsat surface reflectance images were acquired using Landsat 4–5 TM and Landsat 8 OLI. Ulungur Lake and Jili Lake are plain lakes, mainly recharged by precipitation and glacial meltwater from the Altai Mountains [63]. The peak in recharge occurs from July to September when the lake reaches its annual maximum area [64]; therefore, we focused on this period in this study. Cloud cover made it impossible to obtain usable images of all the lakes in all years, meaning that some spatial data were missing for certain lakes in certain years.

Only Landsat images from July to September were retrieved from the GDC platform to avoid seasonal changes in lake areas. Landsat images with relatively low cloud coverage were processed using a “cloudless” algorithm to obtain suitable lake images. A total of 20 images from 1989 to 2021 were used. The normalized difference water index (NDWI) is a widely used method for extracting the boundaries of water bodies [65]. The formula for calculating NDWI is as follows (1):

$$\text{NDWI}=\frac{\text{Green}-\text{NIR}}{\text{Green}+\text{NIR}}$$ (1)

where Green and NIR represent the Green and near-infrared bands of Landsat images, respectively. NDWI has the range of [−1, 1].

3. Results

3.1. Cladoceran assemblages in the modern surface lake-sediments

Among the various lake types of northern China and western Mongolia (Fig. 1A), diverse environments provide habitats for planktonic cladocerans, resulting in notable differences in cladoceran assemblages across different regions. After identifying cladocerans in 136 surface sediment samples, it was found that 48 water bodies lacked cladocerans. In total, we identified 38 taxa, including 24 genera and 34 species in 88 surface sediment samples. Among them, 9 taxa had a relative percentage > 40% and occurrence > 10, including Alona guttata (max. 100% & 75 occurrences), Chydorus sphaericus (max. 89.8% & 70 occurrences), Bosmina longirostris (max. 91.9% & 43 occurrences), Alona rectangula (max. 68.5% & 61 occurrences), Daphnia longispina (max. 80% & 25 occurrences), Sida crystallina (max. 93.8% & 14 occurrences), Daphnia pulex (max. 70% & 12 occurrences), Ceriodaphnia spp. (max. 47.1% & 14 occurences), and Disparalona rostrata (max. 48% & 22 occurrences). Then was conducted sequential T-test analysis of regime shifts algorithm (STARS) between TDS and the cladoceran composition, which showed a breakpoint at ∼2 g/L. Cladocerans were generally absent at higher levels of TDS (above 6 g/L) (Fig. 2A). Within the depths range from 1 to 20 m, the total concentration of cladocerans was relatively high (Fig. 2B). Regression analysis indicated that the total concentration of cladocerans has a negative correlation with TDS and a positive correlation with depth (Fig. S1A, B), and there was a negative correlation between TDS and depth (Fig. S1C). There was a weaker correlation between the total cladoceran concentration and nutrient, as demonstrated by NH₄ (Fig. 2C).

Fig. 2.

Response model of the total cladoceran concentration in the surface sediments of the study sites, compared with (A) TDS with the breakpoint calculated using Sequential T-test Analysis of Regime Shifts algorithm (STARS, red line) (B) depth, and (C) NH₄. Note that the cladoceran concentration is significantly correlated with depth and TDS (p < 0.01), but not with NH₄ (p > 0.05).

A plot of cladoceran assemblages, sorted according to water depth, revealed obvious variations in cladoceran composition with increasing lake level (Fig. 3). Based on CONISS cluster analysis results, the assemblages were divided into four groups:

Fig. 3.

Surface sediment cladocerans in lakes and reservoirs in northern China and western Mongolia, in order of sample water depth. Sample point numbers are listed in Table S1.

Group I (water depths: 77 to 16.6 m): This group has a large depth gradient and a high proportion of planktonic taxa, dominated by Bosminidae, especially B. longirostris (0–91.9%). Other species with higher proportions in this group than in the others include Bosmina coregoni (0–23.1%), Bosmina longispina (0–15.5%), Eurycercus spp. (0–20%) and Alonella nana (0–40%).

Group II (water depths: 16.6 m to 7.3 m): B. longirostris (0–50.5%) remains dominant in this group, and there is a slight increase in C. sphaericus (0–5.6%). D. longispina (0–0.3%) and Ceriodaphnia spp. (0–0.1%) occur less frequently. The diversity of littoral species in this group is extremely low, with only 12 taxa.

Group III (water depths: 7.3 m to 0.92 m): This group has a smaller depth range. Cladoceran species diversity rapidly increases, and the total concentration fluctuates but is generally high. In addition to the continued dominance of B. longirostris (0–15.1%), the proportions of C. sphaericus (0–25.5%), A. guttata (0–23.5%), and A. rectangula (0–6.6%) increase significantly. S. crystallina (10 occurrences), Leydigia leydigi (12 occurrences), Graptoleberis testudinaria (14 occurrences), and D. rostrata (16 occurrences) occur more frequently than in the other groups.

Group IV (water depths: 0.92 m to 0.13 m): In this shallow water group, the total cladoceran concentration is low and the community structure is simpler. Planktonic species decrease significantly and there is the predominance of littoral species like C. sphaericus (0–21.1%), A. guttata (0–41.7%), and A. rectangula (0–21.7%). Alona intermedia (0–26.7%), Oxyurella tenuicaudis (0–33.3%), Leydigia acanthocercoides (0–16.7%), and Monospilus dispar (0–36.4%) are more abundant than in the other groups.

3.2. Relationship between cladoceran compositions and environmental factors

We conducted a constrained ordination analysis of a dataset comprising the cladoceran data from the surface sediments of 62 water bodies, and the complete set of environmental data. From this dataset, 26 cladoceran taxa were selected, each occurring in at least 2% of all samples and having a maximum percentage of at least 2%. Due to multicollinearity among the environmental factors, a variance inflation factor (VIF) analysis was performed on 13 environmental variables. After removing Altitude, which had high VIFs, the remaining 12 variables with VIFs < 10 were selected for CCA analysis. The percentages of these species were then square-root-transformed. DCA of the cladoceran compositions data revealed that the gradient of the first axis spanned 4.2 SD. This finding indicated that CCA, a unimodal response model, would be appropriate [66]. Based on forward selection in CCA and individual testing of the significance of each variable, four significant environmental variables were identified. Water depth was the most significant factor affecting the cladoceran assemblages, explaining the largest amount variance (9.6%) in the cladoceran dataset, with a small p-value (p = 0.002). NH₄, TDS, and AirT were secondary factors, explaining 5.0%, 3.6%, and 2.8% of the variance, respectively (Table 2). The results of the CCA based on this subset of environmental variables are plotted in Fig. 4 as an ordination diagram. Most of the cladoceran taxa have significant correlations with environmental variables (Fig. 4). Axes 1 and 2 of the CCA explain 10.9% and 6.1% of the total variance in cladoceran assemblages, respectively (Fig. 4), with corresponding eigenvalues of 0.29 and 0.16. This suggests that the first axis represents a stronger environmental gradient. The positive direction of the first axis was mainly associated with variables associated with water depth, while the negative direction was associated with TDS. The negative direction of the second axis is mainly associated with environmental variables related to NH₄.

Table 2.

Summary statistics of the CCA of cladoceran taxa and environmental variables.

	Separated CCA		CCA with forward selection
	Expl. Var (%)	P-value	Expl. Var (%)	P-value
Depth	9.6	0.002**	9.6	0.002**
NH4	5.1	0.002**	5.0	0.002**
TDS	6.3	0.002**	3.6	0.006**
AirT	2.7	0.068	2.8	0.018*
DO	5.1	0.002**	2.0	0.122
PO4	4.3	0.004**	1.8	0.16
NO3	1.7	0.428	1.9	0.126
MAP	5	0.002**	1.7	0.2
pH	5	0.002**	1.3	0.446
MST	2.9	0.024*	1.1	0.7
MAT	3.1	0.046*	1.5	0.35
WaterT	2.4	0.086	1.2	0.572

Significance levels: ^⁎⁎⁎p ≤ 0.001; ^⁎⁎p ≤ 0.01; * p ≤ 0.05.

Fig. 4.

CCA ordination biplots of relationships between environmental factors with (A) cladoceran species and (B) 62 samples in northern China and western Mongolia. Arrows indicate environmental factors and triangles represent cladoceran taxa (see Table S2 for taxa codes).

Therefore, the results from the cladoceran training set of the surface sediments of 62 water bodies indicate that water depth and nutrient are key factors determining the species composition and spatial distribution of cladocerans. This provided a basis for the further development and analysis of cladoceran-water depth and cladoceran-nutrient transfer functions.

3.3. Performance of the transfer functions

The CCA analysis of training sets and environmental factors showed that depth was the most significant environmental factor, and nutrients the secondary factor, influencing cladoceran assemblages in northern China and western Mongolia. Comparative evaluation of the WA, WA-PLS, MAT, and RF models (Tables 3, 4) revealed that for water depth, the transfer function derived using WMAT was superior. This model had the highest correlation coefficient (R²_jack = 0.61) and the RMSEP was 10.81. For nutrients, (WA_Cla), established from the cladoceran dataset of the surface sediments in 62 water bodies, emerged as the best model, having the highest correlation coefficient (R²_jack = 0.41) and a predictive error (RMSEP) of 0.71. Results of leave-one-out cross-validation indicated a good linear correlation between the estimated values of water depth reconstructed using the WMAT model and the observed values, as well as between the nutrient estimates reconstructed using the WA_Cla model and the observed values (Fig. 5). This suggests that the surface sediment cladoceran dataset can reliably be used for quantitative reconstruction based on subfossil cladoceran data, and it is thus a new quantitative method in regional paleoenvironmental construction.

Table 3.

Test results of the transfer function based on a dataset from northern China and western Mongolia based on 62 water bodies, using the weighted averaging (WA), weighted-average partial least squares (WAPLS), and the modern analogue technique (MAT) models.

Environmental factor	Model	R2_jack	AveBias_Jack	MaxBias_Jack	RMSEP
Water depth	WA_Inv	0.488331	−0.40002	30.719	10.486
	WA_Cla	0.549238	−0.95907	20.2943	12.8932
	WATOL_Inv	0.170433	1.51123	54.0035	13.2766
	WATOL_Cla	0.246162	2.00789	53.2203	12.8594
	WA-PLS (1)	0.497499	−0.15401	29.303	10.5427
	WA-PLS (2)	0.475812	−0.86789	28.5622	11.3046
	WA-PLS (3)	0.29494	−1.53495	28.6067	15.7351
	WA-PLS (4)	0.201212	−2.06042	28.9018	19.2686
	WA-PLS (5)	0.180176	−2.1575	27.8982	20.3492
	MAT	0.487781	2.06621	56.935	12.0335
	WMAT	0.609285	1.76667	48.7579	10.8102
Nutrient	WA_Inv	0.388611	−0.00574	1.04696	0.520762
	WA_Cla	0.406747	−0.0145	1.00551	0.70536
	WATOL_Inv	0.301153	0.036516	1.14974	0.561545
	WATOL_Cla	0.316288	0.06383	1.0325	0.720634
	WA-PLS (1)	0.387103	0.01042	1.06317	0.521656
	WA-PLS (2)	0.300526	0.008253	0.987221	0.595051
	WA-PLS (3)	0.235497	0.018134	0.850994	0.683313
	WA-PLS (4)	0.192258	0.018299	0.895831	0.756405
	WA-PLS (5)	0.173105	0.02874	0.955974	0.808871
	MAT	0.308209	−0.07794	0.756372	0.565269
	WMAT	0.326216	−0.07038	0.704228	0.560478

Table 4.

R², Root mean squared error (RMSE) of 62 water bodies uisng the random forest (RF) model.

Environmental factor	R2	RMSE
Water depth	0.2456	3.750627
Nutrient	0.1906	0.558599

Fig. 5.

Scatter plots of observed (A) depth and (B) NH₄predicted depth and NH₄based on the modern analogue technique (MAT) and weighted averaging (WA), respectively.

3.4. Reconstructions of lake level and nutrient state since ∼1930 CE

Significance testing demonstrated that the lake level and nutrient level reconstruction, based on 62 surface sediment training sets, accounted for over 95% of the variance in all random variables (Fig. 6). This suggests that the selected MAT and WA models are useful for reconstructing both lake levels and nutrient levels. The cladoceran-inferred lake level and nutrient level reconstructions spanning the last ∼100 years are shown in Fig. 7. The reconstructed lake level was moderately variable during 1926–1960 CE, with lower levels between 1960 and 1988 CE. Subsequently, from 1988 to 2016 CE, there was a notable increase in both lake level and nutrient concentration.

Fig. 6.

Results of statistical significance testing of (A) lake level (depth) and (B) nutrient level (NH₄) reconstructions of Jili Lake using the modern analogue technique (MAT) and weighted averaging (WA), respectively. Gray histograms indicate the proportion of the variance in the subfossil cladoceran records explained by random variables (999 times), with the red dotted line indicating the 95% quantile; the black dotted line indicates the proportion of the variance explained by the first axis of a PCA, and the black solid line indicates the proportion of the variance explained by the reconstructed depth and NH₄.

Fig. 7.

Modern analogue technique (MAT)–based lake level and weighted averaging (WA)–based nutrient level (NH₄) reconstructions and associated bootstrap uncertainty estimates for Jili Lake.

Based on statistical methods to analyze the habitat significance of the cladoceran communities, CCA roughly categorized species according to the dominant environmental factors influencing them (Fig. 4). The species’ optimum ordering was further refined to accurately identify representative cladoceran genera for water depth and nutrients (Fig. 8). In the deeper lakes, B. longispina, B. coregoni, and B. longirostris performed better, with optimal depths of 70.1 m, 54.7 m, and 12.4 m, respectively; and A. intermedia (optimum depth: 2.1 m), and L. acanthocercoides (optimum depth: 12.4 m) have the highest proportions in shallow lakes (Fig. 8A). The results show that in nutrient gradient, M. dispar (optimum value: 29.9 µg/L), A. intermedia (optimum value: 27.1 µg/L), and B. longirostris (optimum value: 27 µg/L) occur under conditions of higher nutrient levels; in oligotrophic lakes, B. longispina (optimum value: 19.2 µg/L) and Camptocercus rectirostris (optimum value: 18.8 µg/L) were more common (Fig. 8B).

Fig. 8.

Optima and tolerances of (A) depth and (B) NH₄ for selected cladoceran taxa based on weighted averaging (WA) regression. The orange lines indicate the mean WA environmental optima of the cladoceran taxa, and the purple lines indicate their WA tolerances.

The areas of Ulungur Lake and Jili Lake were extracted from remote sensing images (Fig. 9), and due to the interconnectedness of the two lakes, the obtained changes in area are consistent. Compared to 1960–1970 CE, the two lakes areas were increased during 1989–2021 CE.

Fig. 9.

Comparison of records of NH₄ (B) and water depth (C) based on cladocerans in core JL18–01-A with environmental records from Jili Lake and Ulungur Lake since 1960 CE: (A) Annual fish yield in Jili Lake [67]; (D) lake level in Ulungur Lake [43]; (G) lake level in Jili Lake [43]; (E-F) lake areas of Ulungur Lake and Jili Lake, respectively.

4. Discussion

4.1. Spatial distribution of cladoceran taxa along the lake level and salinity gradients, and identification of indicator species

In northern China and western Mongolia, water depth and TDS are important environmental controls on aquatic organisms, including Cladocera [68], chironomids [69], diatoms [70], and Pediastrum [46,71]. In this study, water depth was the most significant variable explaining the distribution and abundance of cladoceran taxa (Fig. 4), which is consistent with the conclusions of Amsinck et al. [22] and Wang et al. [72].

In our dataset, most of the cladoceran species are found in freshwater lakes and were absent in many lakes in the Badain Jaran Desert [39] and central-west Inner Mongolia, especially those with a high salinity. Studies of lakes in northwest China showed large reductions in biodiversity and food chain length, indicating loss of ecological functioning, with increasing salinity [73]. Although fish [74] and macrophyte abundances [33] also determine the structure of cladoceran communities, in high salinity lakes, due to the lack of fish predation of zooplankton, there are usually only two trophic levels: zooplankton and phytoplankton [75].

Bosmina and Daphnia, previously interpreted as planktonic species [[76], [77], [78]], have been further observed in our study to predominantly inhabit deep water (depths > 50 m), particularly species like B. coregoni and B. longispina [25]. Compared to these species, A. intermedia and L. acanthocercoides inhabit shallow lakes, and 11 species have optimal depths of < 5 m (Fig. 8A). A. guttata, C. sphaericus, and A. rectangula are widespread in the 88 lakes with cladocerans, and they are species characteristic of littoral habitats and are associated with submerged macrophytes [51]. In the deep lakes of northeast China, the cladoceran concentration was high, and the community was species rich. Bosmina was dominant with a high species diversity, including B. longirostris, B. coregoni, and B. longispina, which are less common in other lakes. In the freshwater lakes of the northwest China, species such as B. longirostris, C. sphaericus, and A. guttata are predominant. C. sphaericus is known as a facultative-planktonic species that can tolerate harsh conditions [34]. The subfossil cladoceran communities in the core samples from Jili Lake over the past hundred years also include these species [41].

4.2. Effects of trophic status on cladoceran assemblages

Planktonic cladocerans are very sensitive to changes in water nutrient levels [38,[79], [80], [81]]. With increasing lake eutrophication there is a succession of the dominant cladoceran species from oligotrophic littoral species to pollution-tolerant and eutrophic planktonic species [29]. The CCA results showed that NH₄ is the secondary environmental factor, significantly affecting the cladoceran communities (p < 0.01) (Table 2), and M. dispar, A. intermedia, B. longirostris, and D. rostrata are primarily eutrophic species (Fig. 8B). The presence of B. longirostris indicates a high lake trophic state [82]. Luoto et al. [37] also noted that B. longirostris and D. rostrata are eutrophic species. By contrast, species such as B. longispina and Camptocercus rectirostris have lower optimal values for NH₄ [79]. Kamenik et al. [83] found that B. longirostris was a typical inhabitant of nutrient-rich lakes, although it can be found in oligotrophic lakes in southwest China [72]. Additionally, in a subtropical shallow freshwater lake, B. longispina was significantly negatively correlated with TP, while C. sphaericus was significantly positively correlated with TP [81].

4.3. Lake level variations of Jili lake over the past century

The increased use of water for irrigation and the construction of multiple dams on the Ulungur River since 1960 CE have led to a sharp reduction in the quantity of inflowing water, and consequently the lake level of Jili Lake has decreased substantially [43]. The reconstructed lake levels inferred from subfossil cladocerans for this period (1960–1988 CE) are the lowest, which is consistent with the measured data (Fig. 9F, G). The proportion of Bosmina was at its lowest during this period, with an increase in littoral species [41].

After 1988 CE, the Cladocera-based lake level in Jili Lake closely matches the measured lake level, indicating the reliability of our reconstruction. After the project channeling the Irtysh River into Ulungur Lake in 1988, which diverted water from the Irtysh River to Ulungur Lake, the lake levels in both lakes began to rise [45]. The Cladocera-based lake level also indicates an increase in lake level, but it may overestimate its true magnitude. This overestimation could be caused by an increase in both the lake nutrient level and salinity. The channeling project make the saline water from Ulungur Lake can now flow into Jili Lake, thereby accelerating the salinization and eutrophication of the lake [41,43,84]. Additionally, human activities, particularly intensive fish farming in Jili Lake, have led to eutrophication, greatly affecting the cladoceran assemblages. The cladoceran community was dominated by the eutrophic species B. longirostris during this phase, which is similar to changes observed in several lakes in the middle and lower reaches of the Yangtze River that are influenced by human activities [80].

We acknowledge that several uncertainties exist in our study: (1) the relatively large range of the measured water depths in the modern surface samples may have led to a greater amplitude of the reconstructed lake levels; (2) multiple environmental factors simultaneously impacted the cladoceran communities, as indicated by the CCA results, which show that four factors significantly affect our modern surface sediment samples, with depth being the most important factor. Despite these uncertainties, our results are valuable given the current scarcity of cladoceran training sets in China.

5. Conclusion

We established a modern surface sediment cladoceran training set for northern China and western Mongolia. This dataset shows that cladoceran concentrations are primarily influenced by lake level and salinity. Environments with very shallow water and high salinity are unfavorable for cladoceran survival. Cladoceran communities respond to variations in water depth and nutrient level. Application of this training set to reconstruct lake levels and nutrient status in Jili Lake showed a close resemblance between the predicted lake level and actual measurements, with lake levels rising after 1988 CE, accompanied by higher nutrient levels. Our findings provide a valuable modern dataset on cladoceran indicators across a large region in northern China and western Mongolia, offering a reliable methodology for reconstructing changes in lake levels and nutrient dynamics. They highlight the importance of considering both natural and anthropogenic factors in the management and conservation of lake ecosystems in arid and semi-arid regions. This approach not only aids in understanding the historical changes in lake ecosystems but also contributes to predicting future trends and informed decision-making for sustainable ecosystem management.

CRediT authorship contribution statement

Wenjia Wang: Writing – original draft, Formal analysis, Software, Data curation. Lixiong Xiang: Writing – original draft, Investigation, Formal analysis, Data curation, Funding acquisition. Yuan Li: Writing – original draft, Formal analysis, Data curation, Funding acquisition, Conceptualization. Qi Suo: Writing – review & editing, Data curation. Xueqing Song: Writing – review & editing, Data curation. Yong Zhang: Writing – review & editing, Software, Data curation. Chong Huang: Writing – review & editing, Software, Investigation. Farqan Muhammad: Writing – review & editing, Data curation. Tao Wang: Data curation. Xiuxiu Ren: Data curation. Shengqian Chen: Data curation. Guangjie Chen: Writing – review & editing, Conceptualization. Xiaozhong Huang: Writing – review & editing, Supervision, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no conflicts of interest in this work.

Biographies

Wenjia Wang received her master's degree in Physical Geography from the College of Earth and Environmental Sciences at Lanzhou University in 2023. Her research interest focuses on micropaleontology, especially cladoceran and pollen assemblages analysis and its environmental implications analysis.

Xiaozhong Huang (BRID: 09586.00.96028) graduated from Lanzhou University in 2006 with a Ph.D. degree in Physical Geography. He then began his research career at Lanzhou University and earned the position of professor at the College of Earth and Environmental Sciences, Lanzhou University in 2018. His current research interest includes past environmental changes and human activities, palynology and paleoecology.