Spatial Heterogeneity and Temporal Variation of Water Levels in Dongting Lake

Abstract

To quantify the spatiotemporal patterns of the water-level variations in the study area, we conducted cluster analysis of the temporally varying measurements across multiple hydrological stations. The temporal trends and change points were analyzed, followed by IHA-RVA quantification of the water-level alterations before and after change points. Cluster analysis demonstrated the following. (1) Hydrological stations segregate into two distinct groups at the Euclidean distance threshold d = 5, and into three clusters at d = 4, confirming the pronounced west–east heterogeneity in the lake. (2) The hydrological alteration degrees exhibit considerable variation across the lake’s sub-lakes (Qili, Muping, South Dongting, East Dongting), with marked heterogeneity persisting even among representative monitoring stations within individual sub-lakes. The water-level regimes in Qili Lake can be partitioned into two distinct periods, before and after the change point, exhibiting the highest hydrological alteration degree across the lake. Representative stations of the other sub-lakes fall into three periods. During the first phase of hydrological alteration, Zhouwenmiao, Jinshi, and Chenglingji exhibit moderate alteration. Throughout the second alteration phase, all the representative stations consistently exhibit moderate alteration, although significant heterogeneity emerges across hydrological indicators among the sub-lakes. (3) Downstream of Yangliutan station, the longitudinal profile exhibits terraced morphology, segmented into three distinct levels by two hydraulic knickpoints. This geomorphic configuration primarily controls both the localized stage reductions and the maintenance of elevated upstream water levels during dry seasons. Confronting the persistent dry-season stage declines at Yingtian Station, enhanced monitoring and conservation of terraced transition zones in South Dongting Lake must be prioritized, with implementation of the zoned control principle for water-level governance and lake management. This study establishes a scientific foundation for the protection and governance of Dongting Lake, thereby advancing sustainable utilization of its water resources.

1. Introduction

The lake stage constitutes a pivotal regulatory factor governing lacustrine ecosystem integrity, regional water security, and socio-economic development dynamics. As a core functional modulator, its fluctuations directly impact ecological health, water resource sustainability, and economic productivity within lake basins. Global lake stages exhibited heterogeneous spatiotemporal patterns during the 21st century [1,2], primarily driven by coupled climatic and anthropogenic forcing mechanisms [3], with significant inter-regional heterogeneity observed across major global lake basins. For instance, across the Tibetan Plateau, increased precipitation constituted the primary driver of lake water storage gains, with glacial ablation and permafrost degradation contributing significantly to these changes [4]. Rising temperatures alter cryospheric processes—including snowfall/rainfall patterns, ice sheet dynamics, groundwater recharge, and low-order stream runoff—in cold regions, driving hydrological regime shifts that elevate water levels [5]. The hydrological stage fluctuations in Van Lake during recent decades have been primarily driven by seasonal hydro-meteorological variability. However, in regions such as Lake Urmia, while climate change constitutes the fundamental driver [6], contemporary agricultural water withdrawals represent the predominant determinant of lake water volumes, exerting decisive control over hydrological persistence [6]. Similarly, approximately 65% of the water-level variations in Lake Poopó were attributable to climate change, with the remaining 35% originating from irrigation management practices [7]. Research demonstrates that precipitation trends and anthropogenic interventions—such as reservoir regulation and agricultural water withdrawals—constitute the primary drivers of lake stage fluctuations and declines [8]. Unregulated lakes generally exhibit less pronounced stage trends and reduced storage variation amplitudes compared to their human-regulated counterparts [9]. Artificially managed reservoirs demonstrate significantly greater seasonal stage variations (mean Δh = 0.86 m) than natural water bodies (mean Δh = 0.22 m) [10].

Persistent alterations in lake stages trigger significant ecological impacts and water management challenges. Elevated water levels may confer ecological benefits, including reduced landscape fragmentation and enhanced hydrological connectivity [11], often accompanied by a decline in comprehensive water pollution index (CWPI) values [11]. However, water-level fluctuations exert multifaceted impacts on aquatic ecosystems: the stage height exhibits significant negative correlations with the total nitrogen (TN) and total phosphorus concentrations [12]. During low-water periods, elevated total phosphorus (TP) concentrations are frequently observed [13], correlating with increased chlorophyll a levels, organic carbon accumulation, and reduced water transparency [14]. Furthermore, the water quality responses to stage variations demonstrate intra-annual heterogeneity across the hydrological year [15]. Simultaneously, the water level governs the plant community distribution as a dominant controlling factor [16]. Wetland’s spatial extent—characterized by longitudinal reach and inundation duration—typically contracts with rising stages [17], potentially compromising functional integrity and habitat value. Seasonal water-level fluctuations significantly modulate the antibiotic distribution in aqueous and sedimentary matrices [18], while concurrently serving as a key driver of fish biomass and species composition in lacustrine ecosystems [19]. Water-level regulation constitutes a critical tool for addressing these challenges. Research demonstrates that proactive regulation strategies, when integrated with revegetation efforts, effectively establish waterbird habitats and promote submerged macrophyte recovery [20,21], potentially serving as a key mechanism to stabilize or mitigate wetland contraction [17].

Under climate change, intensive anthropogenic activities and distinctive fluvial–lacustrine interactions have shaped the unique hydrological evolution patterns in Dongting Lake. Dongting Lake, China’s second-largest freshwater lake and a critical flood-regulation reservoir for the Yangtze River, receives inflow via three south-bank distributaries (Songzi, Taiping, Ouchi), converges runoff from its watershed, and discharges back to the Yangtze mainstream at Chenglingji. Consequently, river–lake interactions profoundly govern its hydro-morphological evolution. Hydrological changes in the Yangtze River profoundly impact its river-connected lakes [22]. Alterations in water–sediment dynamics are predominantly driven by intensive anthropogenic interventions, including the Jingjiang River cutoffs (1967–1972), Gezhouba Dam (1981), Three Gorges Dam (TGD, 2003), and upstream reservoir clusters [23]. Channel scour along the Yangtze mainstem has induced declining low-water stages, with significant incision-reducing stages at equivalent low-flow discharges [24]. This process has driven a regime shift from high-flow/high-stage to medium-flow/high-stage conditions [25,26]. Consequently, alterations in the Yangtze mainstem have modified the flow–sediment partitioning at its three distributaries, significantly reducing the runoff and sediment flux (particularly sediment) entering Dongting Lake [27]. The current stage reductions at equivalent discharges [28] have diminished the Yangtze River’s capacity to alleviate droughts in Dongting Lake through replenishment [29]. The Three Gorges Project constituted a pivotal watershed, inducing a shift from sedimentation to erosion in Dongting Lake’s bed [30]. This transition triggered significant stage declines with spatiotemporal heterogeneity in magnitude [31]. Regulated management may extend the lake’s functional longevity [32]. Furthermore, land reclamation via lake enclosure, polder consolidation, and polder-to-floodplain restoration in the Dongting Lake area [33] reflects long-term anthropogenic interventions in lacustrine drainage patterns. Concurrently, lakebed alterations induced by human activities—including water abstraction, navigation channel modification, and sand mining [34,35]—trigger cascade adjustments to lake stages.

Consequently, Dongting Lake has undergone substantial stage variations over recent decades. Owing to its vast spatial extent and hydrodynamic complexity, significant heterogeneity in the water levels exists across sub-lake regions, manifesting in a characteristic “high-water lake, low-water river” phenomenon—where lacustrine conditions dominate at high stages while fluvial morphologies prevail during low stages. Although prior studies have analyzed the stage variations in Dongting Lake, their reliance on limited hydrological monitoring stations failed to achieve comprehensive spatial coverage across all the sub-lake regions—particularly western sub-lakes and marginal wetlands—and inadequately elucidated the causal mechanisms underlying stage heterogeneity. This study analyzes long-term stage records from multiple hydrological stations across Dongting Lake. Applying agglomerative hierarchical clustering (AHC) [36,37], we classify the monitoring stations into hydrologically coherent groups and identify the primary drivers of longitudinal stage differences during dry seasons. Simultaneously, we employ the Indicators of Hydrologic Alteration–Range of Variability Approach (IHA-RVA) methodology [38,39] to quantitatively assess the hydrological alterations across sub-lake regions. This approach is extensively applied to evaluate anthropogenic impacts on riverine hydrological regimes, with its 33 indicators comprehensively encompassing the key hydrological characteristics within five functional groups. A comprehensive understanding of the spatial heterogeneity in the lake stages and temporal variation patterns across Dongting Lake constitutes the scientific foundation for advancing knowledge of river–lake hydrological connectivity, optimizing water resources management, and maintaining wetland ecological integrity. This understanding is critical for achieving sustainable conservation and utilization of the lake’s water resources.

2. Materials and Methods

Dongting Lake is situated on the mid-reach Yangtze alluvial plain, extending from 111°52′ E to 113°7′ E longitude and 28°42′ N to 29°38′ N latitude (Figure 1). As the Yangtze basin’s second-largest river-connected lake (2680 km²), Dongting Lake receives inflows through the Three Outlets (Songzi, Taiping, Ouchi) and integrates four principal tributaries—the Xiang, Zi, Yuan, and Li Rivers—establishing a hydrologically complex system. Dongting Lake is a large shallow lake with highly irregular shorelines dominated by artificial dikes (>85% of perimeter), resulting from historical reclamation. Spatially partitioned into the East, South, and West Dongting Lake, the system exhibits significant heterogeneity: East Dongting has the largest area and deepest bathymetry, while the South and West sections are shallower, resulting in divergent hydrological regimes across sub-regions.

Dongting Lake has developed a comprehensive hydrometric monitoring network, with hydrological stations and water-level gauges strategically deployed at all the critical control points and cross-sections throughout the lake system. Integration of automation systems, information platforms, and smart monitoring technologies has substantially improved the real-time tracking accuracy and temporal resolution for Dongting Lake’s hydrological dynamics, enhancing flood-risk mitigation and water allocation efficiency. The spatial distribution of 21 gauging stations within Dongting Lake, based on station location data, is mapped in Figure 1. The hydrometric stations in Dongting Lake exhibit a uniform spatial distribution, deployed along major river channels and lake basin margins. Each sub-lake section can select multiple stations as representative monitoring points.

The selection of ten representative hydrological stations was guided by three essential criteria: (1) long-term data continuity to ensure robust temporal analysis, (2) spatial coverage uniformity across sub-lakes to capture regional hydrological heterogeneity, and (3) distinct upstream influences reflecting diverse hydrodynamic drivers. This approach achieved a comprehensive representation of the water-level characteristics throughout Dongting Lake’s sub-lake system.

(1) West Dongting Lake: Situated in the western lake basin, this sub-lake comprises Qili Lake and Muping Lake. It receives inflows from the Yuan, Li, Songzi, and Hudu rivers, collectively contributing >50% of Dongting Lake’s total inflow. Muping Lake regulates these flows before downstream discharge. Jinshi Station monitors inflows upstream of Qili Lake’s inlet, while Shiguishan Station gauges outflows at its outlet. These paired stations serve as representative monitoring proxies for Qili Lake’s hydrological processes. Nanzui Hydrological Station—a national benchmark facility—documents the water and sediment flux dynamics entering South Dongting Lake via the northern outlet of West Dongting Lake, receiving inflows from the Songzi, Hudu, Li, and Yuan rivers. Xiaohezui Hydrological Station monitors the corresponding fluxes at the southern outlet. Zhouwenmiao Water-Level Station operates at the confluence of the Yuan River and Muping Lake. Thus, Nanzui, Xiaohezui, and Zhouwenmiao stations function as representative monitoring sites for Muping Lake’s hydrological characteristics

(2) South Dongting Lake: Situated in the southern lake basin, this sub-lake comprises numerous interconnected lakes of varying sizes. It receives inflows from West Dongting Lake to the west and connects with the Zi and Xiang rivers to the south, functioning as a hydraulic transition zone between East and West Dongting Lake. Three hydrometric stations monitor key boundaries of South Dongting Lake: Yuanjiang Station tracks the water levels along the western shoreline; Yangliutan Station quantifies the water levels at the confluence of the Zi River’s northern branch; and Yingtian Station records the water levels on the Xiang River. These three stations function as representative hydrometric sites for South Dongting Lake’s hydrological processes.

(3) East Dongting Lake: Positioned at the basin’s outlet, this sub-lake receives all of Dongting Lake’s integrated inflows. It discharges into the Yangtze River via Chenglingji and constitutes the largest sub-basin by surface area. Lujiao Station monitors the water levels along East Dongting Lake’s eastern shoreline, while Chenglingji Station regulates the outflow dynamics at the Yangtze confluence. These two stations function as representative monitoring sites for East Dongting Lake’s hydrological processes.

Data from 21 intra-lake hydrometric stations were statistically compiled, as detailed in

Table 1. Hydrometric station metadata for Dongting Lake.

No.	Station Name	Time Series	Monitoring Parameters	Remarks
1	Chenglingji	1953–2024	Water level, Discharge	Representative station for East Dongting Lake
2	Lujiao	1952–2024	Water level	Representative station for East Dongting Lake
3	Yingtian	1953–2024	Water level	Representative station for South Dongting Lake
4	Xiangyin	1949–2024	Water level
5	Yangliutan	1953–2024	Water level	Representative station for South Dongting Lake
6	Yangdi	1965–2022	Water level, Discharge
7	Shatou	1956–2024	Water level, Discharge
8	Ganxigang	2002–2022	Water level, Discharge
9	Yiyang	2017–2024	Water level
10	Caowei	1968–2024	Water level, Discharge
11	Xiaohezui	1955–2024	Water level, Discharge	Representative station for West Dongting Lake
12	Yuanjiang	1951–2024	Water level	Representative station for South Dongting Lake
13	Zhouwenmiao	1968–2024	Water level
14	Changde	1959–2024	Water level
15	Nanzui	1955–2024	Water level, Discharge	Representative station for West Dongting Lake
16	Xiaojiawan	2017–2024	Water level
17	Baibengkou	2017–2024	Water level
18	Anxiang	1955–2024	Water level, Discharge
19	Haozigang	2017–2024	Water level
20	Shiguishan	1958–2024	Water level, Discharge	Representative station for West Dongting Lake
21	Jinshi	1959–2024	Water level, Discharge	Representative station for West Dongting Lake

. Hydrometric records for key control sections in Dongting Lake exhibit reasonable completeness. Data since the 1950s–1960s originate from the Hunan Hydrology and Water Resources Survey Center and the Middle Yangtze River Hydrology and Water Resources Survey Bureau, Hydrology Bureau of Yangtze River Water Resources, while supplementary water-level records (2017–2024) derive from the Hunan Hydrology and Water Resources Survey Center.

To characterize the seasonal water-level distribution patterns across Dongting Lake, three coordinated hydrological surveys were conducted in the South and West sub-lakes during distinct hydrological periods (

Table 2. Water-level measurements and remote sensing datasets.

Survey Date	Measured Stage at Chenglingji (m)	Landsat 8 Acquisition Date	Path	Row	Image-Derived Stage (m)
8 December 2022	17.56	19 December 2021	124	40	18.18
		7 December 2021	123	40	18.27
26 April 2022	24.06	27 March 2017	124	40	24.07
		22 August 2015	123	40	24.51
26 June 2024	29.33	20 July 2018	124	40	29.41
		1 August 2019	123	40	29.21

). Water-level measurements were conducted via vessel-based field surveys, with all the data referenced to a unified elevation datum consistent with hydrometric station benchmarks. Landsat 8 imagery was concurrently integrated for comparative analysis. Where contemporaneous imagery was unavailable during water-level surveys, scenes with analogous hydrological conditions were selected from the 2013–2023 Landsat 8 archive for systematic paired assessment.

3. Methodology

3.1. AHC Method

AHC is a classical clustering algorithm that constructs hierarchical cluster structures through similarity or distance metrics between data points. A key advantage of AHC lies in its provision of hierarchical structural information (dendrograms) and inherent flexibility without requiring predefined cluster numbers. AHC is particularly suited for exploratory data analysis, visually revealing inherent data hierarchies, with established applications in bioinformatics, text mining, and social network analysis [40,41]. In contrast, while K-means demonstrates computational efficiency, it requires predefined cluster numbers (K) and exhibits bias toward spherical clusters, with the results being sensitive to initialization. DBSCAN excels at detecting arbitrarily shaped clusters and exhibits noise robustness, yet it necessitates neighborhood parameter tuning and provides no hierarchical output. AHC offers unique value in revealing hierarchical data relationships and enabling flexible analytical perspectives.

This bottom-up approach initiates by treating each sample as an individual micro-cluster. During iterative merging steps, the pairwise distances between existing clusters are calculated using selected distance metrics and linkage criteria. The two closest clusters are identified and merged into a new macro-cluster. This process repeats until predefined termination conditions are met, progressively building a hierarchical cluster tree that elucidates data aggregation pathways. The dendrogram serves as a critical tool for interpreting hierarchical data structures and determining final cluster counts, providing a graphical representation of the AHC process. In dendrograms, the x-axis typically represents the data point indices, while the y-axis denotes the cophenetic distance (dissimilarity). Branches depict cluster mergers, with their height (y-axis value) indicating the dissimilarity at which merging occurs. Higher branch positions correspond to greater dissimilarity between merged clusters. By truncating the dendrogram at distinct heights, varying numbers of clusters can be obtained. A defining feature of this method is its elimination of predefined cluster count requirements. It generates a complete hierarchical structure where dendrogram cutting at specified distance thresholds—representing merge dissimilarity—yields clusters of varying granularity.

When clusters contain multiple data points, defining the distance between two clusters requires specifying an inter-cluster linkage criterion, which fundamentally governs the merging behavior in hierarchical clustering. The selection of linkage criteria critically governs the cluster morphology and merging behavior. The most prevalent linkage criteria include the single, complete, average, centroid, and Ward’s methods. Given the high-dimensional data characteristics in this study, the average linkage criterion was selected, defining the distance between two clusters as the mean of all the pairwise distances between points across clusters. This criterion typically yields superior performance in generating relatively balanced cluster sizes, as empirically validated in our analysis.

When applying AHC to multi-site datasets with heterogeneous characteristics and quality, data preprocessing constitutes a critical step for ensuring result reliability and interpretability. In this study, data from distinct monitoring sites were integrated into a unified dataset, ensuring complete cases with no missing values. Min–max normalization was applied to each variable across the consolidated dataset (post-site merging), preserving inherent inter-site level differences while standardizing measurement scales.

3.2. Trend Analysis and Change Point Detection Methods

The Mann–Kendall (MK) trend test is a nonparametric statistical method for detecting monotonic (increasing/decreasing) trends in time series data. Its core principle evaluates the relative magnitudes of all the pairwise data points to determine trend direction significance [42,43]. This method requires no assumption of specific distributions (e.g., normality), exhibits robustness to outliers, and effectively handles missing values. Widely applied in hydrometeorological analyses—including temperature, precipitation, and streamflow—this study employs the MK test to quantify the long-term trends in the annual mean water levels across lake basin stations.

Numerous methods exist for change-point detection in time series, with broad applications across disciplines. This study employs four widely adopted techniques—the Mann–Kendall test, cumulative anomaly method, sliding t-test, and Pettitt method—to detect abrupt changes in the annual mean water levels at Dongting Lake’s hydrological stations. Given their extensive documentation in the hydrological literature, methodological details are omitted for brevity [43,44,45,46].

The four complementary methods address distinct change detection challenges: the Mann–Kendall test captures gradual trend reversals with nonparametric robustness but exhibits limited sensitivity to multiple breakpoints; the cumulative anomaly method intuitively visualizes persistent shifts without distributional assumptions yet lacks statistical validation; the moving t-test quantifies localized abrupt changes with p-values but requires normality compliance and optimal window sizing; while the Pettitt test excels at identifying single abrupt events with high statistical power but disregards gradual transitions. By requiring consensus across ≥2 methods (±2-year tolerance), we mitigate individual limitations while leveraging synergistic strengths for robust breakpoint identification in nonstationary hydrological series.

3.3. IHA-RVA Framework

The Indicators of Hydrologic Alteration–Range of Variability Approach (IHA-RVA) is an integrated assessment framework that quantifies hydrological regime changes and associated ecological impacts in river systems. This method characterizes natural flow regimes and measures alterations induced by anthropogenic activities (e.g., dam construction, water diversion) [38,39]. The IHA component utilizes 33 daily-flow metrics across five parameter groups: monthly magnitudes, extreme flow magnitudes/durations/timing, high/low-flow pulse frequency/duration, and flow change rates/frequency. Building upon IHA, the RVA quantifies ecological deviations by establishing an ecologically acceptable variation threshold based on the natural range of hydrological indicators during a baseline period. The alteration degrees are assessed by comparing the frequency of the post-impact indicator values falling within this threshold range [47].

This study applies the IHA-RVA framework to analyze the water levels at multiple hydrological stations in Dongting Lake, with relevant indicators detailed in

Table 3. Water-level indicators derived from the indicators of the IHA framework.

Group	Description	Stage Metrics	Count
1	Monthly stage characteristics	Mean stage for each calendar month	12
2	Annual extreme stages	Annual minimum/maximum 1 d, 3 d, 7 d, 30 d, and 90 d stages; base flow index	11
3	Timing of annual extremes	Julian date of annual minimum/maximum stage occurrence	2
4	Frequency and duration of high/low-stage events	Annual count of high/low-stage pulses; mean duration of high/low-stage events	4
5	Stage change dynamics	Mean rates of stage rise/fall; number of stage reversals	3

. Given that Dongting Lake experiences no flow cessation events and zero-stage conditions never occur, the zero-stage days metric was excluded; thus, 32 hydrological alteration indicators were selected for analysis.

The RVA calculation is as follows:

$$D_i={\displaystyle \frac{(N_{io}-N_{ie})}{N_{ie}}}\times 100\%$$

(1)

$$N_{ie}=r{N}_{\mathrm{T}}$$

(2)

where D_i denotes the alteration degree of the i-th flow metric; N_io represents the observed post-impact years with IHA values within RVA bounds; N_ie indicates the expected post-impact years within RVA bounds; r signifies the pre-impact proportion of IHA values within RVA thresholds (r = 50%); and N corresponds to the total years in the post-impact hydrological series. Hydrological alterations are classified as follows: low (0 ≤ D_i < 33%), moderate (33% ≤ D_i < 67%), or high (67% ≤ D_i ≤ 100%).

The integrated alteration degree provides a comprehensive assessment of the hydrological shifts across all the indicators before and after anthropogenic disturbances, delivering a single quantifiable value that visually represents the magnitude of the hydrological modification. The system-level alteration magnitude (D₀) is computed through ecological-weight-based aggregation of individual metric alteration degrees (D_i), with the calculation procedure described as follows:

$$D_0=\sqrt{{\displaystyle \frac{1}{33}}{\displaystyle \sum _{i=1}^{33}{D}_i^2}}$$

(3)

The alteration degree for individual metrics is denoted as D_i, while the system-level alteration (D₀) is categorized into low (0 ≤ D₀ < 33%), moderate (33% ≤ D₀ < 67%), or high (67% ≤ D₀ ≤ 100%).

4. Results

4.1. Hierarchical Clustering of Water Levels

The monthly mean water levels at 21 hydrometric stations within the lake basin were plotted, as shown in Figure 2. The plotted data reveal consistent water-level trends across all the stations. Except for Changde Hydrological Station, which peaked in June, the other stations consistently reached maximum levels in July and minimum values in January. Stations were classified as high-amplitude or low-amplitude types based on the intra-annual water-level variations. High-amplitude stations include Chenglingji and Lujiao (East Dongting), and Yingtian and Xiangyin (South Dongting). Low-amplitude stations are exclusively distributed in the central–western South Dongting and West Dongting sub-lakes. Notably, low-amplitude stations maintained persistently high water levels (>25 m year-round) during dry seasons, while high-amplitude stations exhibited significantly lower water levels during dry seasons. The spatial heterogeneity in the dry-season water levels across Dongting Lake may primarily result from the differential lake basin topography. To mitigate dry-season water-level declines in Dongting Lake, a phased regulation strategy should be implemented in lake management practices.

AHC was applied to classify the water levels from 21 hydrometric stations in Dongting Lake. The multidimensional dataset included (a) monthly mean water levels, (b) annual maximum/minimum water levels, (c) monthly extreme water levels, and (d) monthly water-level fluctuations. The resultant clustering patterns are presented in Figure 3. At a Euclidean distance threshold > 5, the stations bifurcate into two clusters: Group 1 (Chenglingji, Lujiao, Yingtian, Xiangyin) demonstrates high hydrological coherence due to connectivity via the broad Xiangjiang Floodway and Yangtze River mainstem, while Group 2 comprises all the remaining stations. Within Group 1, Chenglingji and Lujiao—located in East Dongting Lake—exhibit stronger hydraulic influence from the Yangtze River mainstem, resulting in higher intra-cluster hydrological homogeneity. Yingtian and Xiangyin exhibit higher hydrological similarity due to their co-location in eastern South Dongting Lake, where a greater distance from the Yangtze River mainstem (>35 km) reduces the hydraulic connectivity.

At a Euclidean distance threshold of d = 4, Group 2 subdivides into two subgroups: Group 2-1 and Group 2-2, as shown in Figure 3. Within Cluster Group 2-1, Shiguishan, Jinshi, and Anxiang stations exhibit high hydrological similarity. This likely results from their co-location in northern West Dongting Lake, where floodways hydraulically connect the Lishui and Songzi Rivers, establishing strong hydrodynamic linkages. Within Cluster Group 2-2, Yangliutan station demonstrates distinctiveness from the other sites due to its location at the Zi River–South Dongting Lake confluence, where dual hydraulic influences (river inflow and lacustrine backwater) drive complex stage dynamics. Changde and Zhouwenmiao stations are located on the upper Yuan River. Baibangkou and Xiaojiawan stations occupy the Li River flood channel between Qili Lake and Muping Lake. Haozigang station lies along the Songhu floodway upstream of Muping Lake. All five stations are situated in the upper river network and exhibit high stage similarity. Xiaohezui, Nanzui, Yuanjiang, and Caowei stations are located at the confluence of Muping Lake and South Dongting Lake, demonstrating strong hydraulic connectivity. Yiyang, Shatou, Ganxigang, and Yangdi stations within the Zishui River system exhibit high stage similarity. Despite the significant spatial separation, Yangdi and Caowei stations exhibit high stage coherence, likely attributable to their shared positioning along constricted upper thalweg segments of curved lake channels.

4.2. Trend Detection and Change-Point Analysis

Based on the spatial variability in the lake stage dynamics across Dongting Lake, the system is partitioned into four hydrologically distinct zones: Qili Lake, Muping Lake, South Dongting Lake, and East Dongting Lake. Benchmark hydrological stations representing each zone are as follows: Jinshi and Shiguishan for Qili Lake; Nanzui, Xiaohezui, and Zhouwenmiao for Muping Lake; Yuanjiang, Yangliutan, and Yingtian for South Dongting Lake; and Lujiao and Chenglingji for East Dongting Lake.

The annual mean stage series of benchmark stations in Qili Lake—Jinshi (1959–2024) and Shiguishan (1958–2024)—are plotted in Figure 4a. Both stations exhibit declining trends at a rate of 0.03 m/year (Jinshi: slope = −0.0327 m/year, p < 0.001; Shiguishan: slope = −0.0340 m/year, p < 0.001). Persistently low stages have characterized Jinshi and Shiguishan stations over the past three years. In 2023, these stations registered their respective record-low stages of 28.7 m and 28.1 m—statistically significant minima observed throughout their full monitoring periods.

Among the three benchmark stations in Muping Lake, Zhouwenmiao—situated upstream at the Yuanjiang River inflow—exhibits the highest stage levels relative to Nanzui and Xiaohezui. These two downstream stations display interwoven stage dynamics. As derived from the regression model in Figure 4b, three stations—Nanzui, Xiaohezu, and Zhouwenmiao—exhibited declining water-level trends with mean annual decreases of 0.009 m, 0.007 m, and 0.006 m, respectively. Mann–Kendall trend tests confirmed statistically significant declines at Nanzui (slope = −0.0078 m/year, p < 0.001) and Xiaohezu (slope = −0.0063 m/year, p < 0.001), while Zhouwenmiao showed no statistically significant trend (slope = −0.0053 m/year, p = 0.182). During major flood events in 1998 and 2020, Nanzui, Xiaohezui, and Zhouwenmiao stations recorded their two highest historical stage maxima. Conversely, persistently low annual mean stages characterized the last triennium, culminating in record-low minima across all the stations in 2023.

South Dongting Lake’s extensive spatial domain drives significant hydrological heterogeneity between eastern and western sectors. The benchmark stations Yuanjiang, Yangliutan, and Yingtian—spatially distributed across the western, south–central, and eastern lake regions with inter-station distances > 25 km—collectively capture the regional stage regimes, serving as representative monitoring nodes for lake-wide hydrological characterization. Figure 4c presents the annual mean stage variations (1954–2024) at key South Dongting Lake stations: Yuanjiang, Yangliutan, and Yingtian. Analysis reveals minimal stage fluctuations at Yuanjiang and Yangliutan, with Yuanjiang exhibiting a marginal declining trend (slope = −0.0055 m/year, p = 0.182) and Yangliutan showing no statistically significant trend (slope= −0.0010 m/year, p = 0.9052). Yingtian station exhibited a statistically significant declining water-level trend, with a mean annual decrease rate of 0.016 m/year derived from the regression model (Mann–Kendall slope = −0.0144 m/year, p = 0.0104). During the 1954 and 1998 flood events, all the benchmark stations recorded their two highest historical stage maxima. In the 2020 flood event, Yuanjiang and Yangliutan stations registered their third-highest stages within full-period records, whereas Yingtian exhibited only moderately elevated stages. Following the commencement of reservoir filling in 2003, sustained stage declines have been documented across all the monitoring stations. This trend intensified during consecutive drought years (2021–2023). In 2023, control stations registered record-low stages.

Figure 4d depicts the annual mean stage variations at key East Dongting Lake control stations: Lujiao and Chenglingji. The fitted curves reveal statistically significant increasing trends at both stations, with Chenglingji showing a higher rate of increase (0.013 m/year) than Lujiao (0.004 m/year). The water-level difference between the two stations progressively decreases. Mann–Kendall trend tests revealed a statistically significant increasing water-level trend at Chenglingji station (slope = 0.0149 m/year, p = 0.0063), while Lujiao station exhibited no significant trend (slope = 0.0059 m/year, p = 0.214). Owing to their proximity to the Yangtze River mainstream, both Lujiao and Chenglingji stations are highly susceptible to hydrological influences from its dynamic flow regime. Persistent stage declines characterized the 2021–2023 period; this culminated in 2023 in record-low stages at both stations.

Overall, statistically significant stage declines have been observed across the Dongting Lake stations since the Three Gorges Reservoir impoundment, with an increased frequency of extremely low stages, culminating in all the stations simultaneously registering historical minima in 2023.

Based on the Mann–Kendall test results (

Table 4. Trend detection results for representative stations.

No.	Hydrological Station	Z-Value	p-Value	Slope	Trend (α = 0.05)	Lake Region
1	Jinshi	−7.24	4.46 × 10−13	−0.0327	Decreasing	Qili Lake
2	Shiguishan	−7.95	2.00 × 10−15	−0.0340	Decreasing
3	Zhouwenmiao	−1.34	0.1816	−0.0053	No trend	Muping Lake
4	Nanzui	−3.42	0.0006	−0.0078	Decreasing
5	Xiaohezui	−2.93	0.0034	−0.0063	Decreasing
6	Yuanjiang	−2.49	0.0128	−0.0055	Decreasing	South Dongting Lake
7	Yangliutan	−0.12	0.9052	−0.0010	No trend
8	Yingtian	−2.56	0.0104	−0.0144	Decreasing
9	Lujiao	1.24	0.2135	0.0059	No trend	East Dongting Lake
10	Chenglingji	2.72	0.0063	0.0149	Increasing

), the annual mean stages at all the stations except Zhouwenmiao, Yangliutan, and Lujiao exhibited statistically significant trends (p < 0.05). Among the seven stations with significant trends, Chenglingji demonstrated an increasing trend (Z = +2.72, p = 0.0063), while the other six stations (Jinshi, Shiguishan, Nanzui, Xiaohezui, Yuanjiang, Yingtian) showed declining trends, with Z-values ranging from −3.42 to −7.95 (p ≤ 0.0128).

Abrupt change points in the annual mean stages were identified using four detection methodologies—Mann–Kendall test, cumulative anomaly analysis, sliding t-test, and Pettitt method—with the results presented in

Table 5. Change-point years of key monitoring stations.

Lake Region	Hydrological Station	Methods				Change Point Year(s)
		M-K Test	CUSUM	Moving t-Test	Pettitt’s Test
Qili Lake	Jinshi	1990	1990	1983, 2003	1991	1990
	Shiguishan	1990	1990	1983, 2003	1990
Muping Lake	Zhouwenmiao	2003	2003	1983, 2003	2003	1983, 2003
	Nanzui	2003	2003	1983, 2004	2003
	Xiaohezui	-	2003	1983, 2003	-
South Dongting Lake	Yuanjiang	-	2003	1983, 2003	2003	1983, 2003
	Yangliutan	-	1979, 2003	1983, 2003	2003
	Yingtian	-	2003	1983, 2003	2003
East Dongting Lake	Lujiao	-	1980, 2005	1983, 2005	-	1980, 2005
	Chenglingji	-	1980, 2005	1979, 2005	1979

. The key findings include a single change point in 1991 for Qili Lake; dual change points in 1983 and 2003 for Muping Lake and South Dongting Lake, respectively; and two change points in 1980 and 2005 for East Dongting Lake.

4.3. Water-Level Alteration Degree

Based on the IHA-RVA (Indicators of Hydrologic Alteration–Range of Variability Approach) results from representative stations in Dongting Lake’s subregions, a heatmap depicting the degrees of hydrologic alteration was generated (Figure 5). Significant heterogeneity in the hydrologic alteration degrees exists among Dongting Lake’s sub-lakes (Qili, Muping, South Dongting, and East Dongting). Substantial differences were also observed between benchmark stations within individual sub-lakes, reflecting complex localized hydrological responses.

Qili Lake, positioned as the uppermost sub-lake of Dongting Lake, exhibits a distinct hydrological regime evolution. Abrupt change analysis partitioned its timeline exclusively into two periods: the natural baseline period (P0: 1958–1989) and the anthropogenically altered period (P1: 1990–2024). Jinshi and Shiguishan stations demonstrated the highest hydrologic alteration degrees across Dongting Lake (63% and 65%, respectively). Significant changes occurred in the monthly mean stages, extreme water levels, low-stage event frequency, and flow reversal counts—with the minimum stages and flow direction reversals reaching 100% alteration, indicating complete hydrological regime shifts.

The water levels at representative stations in Muping Lake, South Dongting Lake, and East Dongting Lake were partitioned into three distinct hydrological periods: T0, T1, and T2. For Muping Lake and South Dongting Lake, the period before 1983 (T0) is designated as the natural baseline period. The interval 1983–2002 represents the first phase of hydrological alteration (T1), while 2003–2024 constitutes the second alteration phase (T2). For East Dongting Lake, the period prior to 1980 (T0) is defined as the natural baseline period, with 1980–2004 marking the first alteration phase (T1), followed by 2005–2024 as the second alteration phase (T2).

During the T1 period, Yuanjiang, Yangliutan, Yingtian, and Lujiao stations exhibited low hydrologic alteration (≤33%), whereas Zhouwenmiao, Nanzui, Xiaohezui, and Chenglingji stations demonstrated moderate alteration (34–66%) according to the IHA-RVA classification standards. Spatially, Zhouwenmiao station at Dongting Lake’s upstream inflow and Chenglingji at its outflow exhibited higher hydrologic alteration (38% and 42%, respectively). Yangliutan showed minimal alteration (22%) owing to its mid-lake position in South Dongting Lake, where the lake’s regulatory capacity buffered upstream influences. The remaining stations demonstrated alteration degrees ranging from 31% to 34%. Moderate-to-high hydrologic alteration (34–100%) predominantly occurred in the monthly mean stages during February–April and October, as well as Group 2 minimum stages (1/3/7-day minima) across stations. At Chenglingji and Lujiao stations, the minimum stages across all the durations exhibited moderate-to-high alteration, primarily driven by backwater effects from the Yangtze River mainstream.

During the T2 period, the hydrologic alteration patterns in Dongting Lake differed from those in T1, with all the benchmark stations across the sub-lakes exhibiting moderate alteration (34–66%). However, the parameter-specific alteration profiles varied significantly among the sub-lakes, indicating divergent hydrological response mechanisms despite comparable overall alteration magnitudes. For Muping Lake and South Dongting Lake, moderate-to-high hydrologic alteration predominantly occurred in the monthly mean stages during April–June and October–November, with low-stage event frequency and low-stage duration. Additionally, the multi-day minimum stages (1/3/7/90-day minima) exhibited significant alteration at select stations. At Yingtian station, the multi-day minimum stages across all standard durations (1/3/7-day minima) reached 100% hydrologic alteration. East Dongting Lake exhibited distinct patterns of moderate-to-high hydrologic alteration parameters compared to Muping and South Dongting lakes, with concentrations in the May–July and October monthly mean stages, maximum stages of different durations, and duration of high-stage events.

Statistical counts of the high, moderate, and low hydrologic alteration parameters across stations and periods are presented in Figure 6. Jinshi and Shiguishan stations exhibited 11 and 14 high-alteration parameters, respectively, with 13 and 10 moderate-alteration parameters. During the T1 period, Chenglingji showed seven high-alteration parameters, Yangliutan had zero, and other stations ranged from one to two. Zhouwenmiao peaked at 12 moderate-alteration parameters, while Yangliutan reached a minimum of 4, with the others at 6–9. In the T2 period, Yingtian demonstrated the maximum high-alteration parameters (7), while other stations had 2–3. The moderate-alteration parameters ranged from 7 to 13 across all the stations during T2.

Significant monthly variations in the hydrologic alteration were observed across Dongting Lake’s stations. To characterize the intra-annual stage dynamics, the monthly mean stage deviations from baseline periods are visualized in Figure 7. Jinshi and Shiguishan stations in Qili Lake demonstrated strongly synchronized monthly stage declines, with the minimum reduction magnitude occurring in July and the maximum reduction in October during the observation period. During the T1 period, the water-level changes at six representative stations in Muping Lake and South Dongting Lake were relatively similar. The maximum water-level increase occurred in July, with increases gradually increasing from upstream to downstream stations, ranging from 0.77 to 1.15 m. April and November were troughs, showing significant water-level declines. Lujiao and Chenglingji stations exhibited slight stage decreases in April but showed increases across the other months. At Lujiao, the stage increases exceeded 1.00 m during January–March and July–September. Chenglingji demonstrated marginally greater increases than Lujiao, with its March increase peaking at 1.96 m.

During the T2 period, three benchmark stations in Muping Lake and Yuanjiang/Yangliutan stations in South Dongting Lake exhibited coherent stage variation patterns. Minimal fluctuations characterized January–June, with Zhouwenmiao showing slight increases in January, March, and May, while the other stations experienced reductions of ≤0.30 m. Pronounced stage declines occurred during July–December, peaking in July–October and attenuating through November–December. Lujiao station exhibited significant stage reductions across all the months, diverging from the upstream hydrological stations. During January–April, the reductions ranged from 0.84 to 1.32 m, while the July–December reductions exceeded 1.00 m, peaking at 2.32 m in October. At Yingtian station in East Dongting Lake, the stages showed modest declines during January–April but minor increases in May–June. Chenglingji exhibited consistent increases (0.18–0.35 m) throughout January–June. Both stations demonstrated synchronized stage variations during July–December, with the September–October declines exceeding 2.25 m.

The temporal differentials of the extreme stages between distinct periods were quantified for each hydrological station and are visualized in Figure 8. For extremely low stages, Jinshi and Shiguishan stations exhibited stage reductions exceeding 1.00 m across all the durations. Four stations—Zhouwenmiao, Nanzui, Xiaohezui (West Dongting), and Yuanjiang (South Dongting)—demonstrated sustained declines with greater reductions during T2 than T1, though the magnitudes remained below 1.00 m. Yangliutan exhibited relatively minor stage reductions among the observed stations, with greater declines during T1 than T2. At Yingtian station, the 1- to 30-day stages during T1 showed increases below 0.10 m, while the 90-day levels exhibited minimal change. In contrast, T2 demonstrated declines exceeding 1.50 m across all the minimum stage metrics (1/3/7/30/90-day minima). At Lujiao station in East Dongting Lake, the 1- to 90-day stages during T1 showed significant increases (magnitude ≈ 1.30 m), while T2 exhibited marked declines. Chenglingji demonstrated substantial T1 increases across all the minimum stage metrics (1/3/7/30/90-day minima; range: 1.25–1.55 m), whereas the T2 increases were constrained below 0.30 m.

For extremely high stages, both representative stations in Qili Lake exhibited pronounced declines across all the hydrological metrics, with selected minimum stage metrics declining by >2.00 m. Shiguishan demonstrated marginally greater reductions than Jinshi. Across representative stations in Muping, West Dongting, and South Dongting lakes, the maximum stage metrics (1/3/7/30/90-day maxima) universally exhibited increases of varying magnitudes during the T1 period. Zhouwenmiao, Nanzui, Xiaohezui, Yuanjiang, and Yangliutan stations exhibited similar stage variation patterns. Significant 1–7 d stage increases (0.48–1.02 m) occurred, whereas the 30 d and 90 d rises were comparatively smaller. Specifically, Zhouwenmiao and Nanzui showed 90-day increases of 0.03 m and 0.14 m, respectively. Yangliutan demonstrated 1–7 d rises exceeding 1.20 m, with 30 d and 90 d increases of 0.77 m and 0.63 m, respectively, greater than those at upstream hydrological stations. Yingtian exhibited stage increases of 1.11–1.54 m across all the minimum stage metrics (1/3/7/30/90-day minima), peaking at the 7-day duration. Both Lujiao and Chenglingji showed greater increases than Yingtian, with maximum rises at the 30-day duration (2.15 m and 2.39 m, respectively). Their minimum increases exceeded 1.52 m.

During flood seasons, the stages are predominantly governed by upstream inflows. The documented reductions in the extreme stages and flood-season water levels at representative stations across Muping, South Dongting, and East Dongting lakes during the T2 period demonstrate significant flood-regulation effects from hydraulic structures. Concurrently, the markedly diminished stage reversal frequencies further attest to their regulatory efficacy. During dry seasons, upstream lake stages are primarily controlled by upstream inflows and lakebed topography. In the T1 period, the concurrent reductions in the extremely low stages and elevations of the mean stages—despite declining inflow rates—indicate sedimentation in the upstream basin. In contrast, the declines in both the extremely low and mean stages during T2, occurring amid moderated inflow reduction rates, suggest basin scour.

Flood-season stage reductions benefit Dongting Lake’s flood control capacity, whereas dry-season declines substantially impair lake functionality. Using the 7-day annual minimum stage levels as the representative metric, Figure 9 presents the temporal variations in these values across hydrological stations, enabling further analysis of the dry-season stage dynamics within the lake. The 7-day minimum stages at Zhouwenmiao, Xiaohezui, Nanju, Yuanjiang, and Yangliutan stations in the upstream lake area show no statistically significant trends. All the stations, however, recorded their lowest values within the observation period during 2023–2024, with the 2022–2024 period collectively representing the minimum levels observed since the TGD’s impoundment. This persistent occurrence of extreme lows may indicate a transition toward normalized conditions, necessitating heightened scientific attention. The water levels at Xiaohezui, Yuanjiang, and Nanju stations show hydrologic convergence, indicating enhanced connectivity. At Yingtian, Lujiao, and Chenglingji stations, statistically significant declining trends emerged after the TGD impoundment. During the 2023–2024 extreme drought, period-specific minima occurred at all three stations: Yingtian reached its recorded-period minimum, Lujiao registered levels unseen since 1973, and Chenglingji attained its lowest stage since 2000. The dry-season stage differentials between Lujiao and Chenglingji stations have narrowed significantly in recent years, with concurrent reductions in the water-surface gradients. Notably, the stage difference between Yingtian and Lujiao decreased abruptly from >0.50 m (pre-2011) to <0.08 m, indicating enhanced hydraulic connectivity between these sites—likely associated with intensive sand mining in the Xiangjiang River channel. The declining dry-season stages at Yingtian may amplify the hydraulic head differences across the step-transition zone in eastern South Dongting Lake, potentially triggering headward erosion. This necessitates enhanced monitoring and stabilization measures for the lake’s step-transition zones.

5. Discussion

As China’s second-largest freshwater lake, Dongting Lake covers an extensive area but exhibits highly irregular morphology. Influenced by multiple inflowing rivers, its hydrological regime displays significant complexity, with pronounced spatial heterogeneity among distinct sub-basins. This study employed qualitative and quantitative approaches to analyze the water-level heterogeneity and spatiotemporal variations in Dongting Lake. We elucidated the drivers of spatial water-level disparities, investigated the cascading effects on water quality and ecosystem dynamics, and proposed lake management and conservation strategies.

5.1. Analysis of Spatial Heterogeneity in Water-Level Processes

Dongting Lake’s water levels result from integrated hydrological drivers, including upstream inflows, lakebed erosion–deposition dynamics, Yangtze River backwater effects, and anthropogenic interventions. Spatially, Dongting Lake exhibits a distinct west-high–east-low stage distribution pattern, directly linked to lakebed topography and erosion–deposition dynamics. Situated on the Yangtze–Four Rivers alluvial plain, the lake experiences intense sedimentation—particularly from the Yangtze—resulting in higher sedimentation rates upstream than downstream. This differential sedimentation elevates the upstream lakebed elevation relative to downstream zones, consequently amplifying the spatial stage differences. The 2022 bathymetric map of Dongting Lake (Figure 10a) indicates severe prior sediment accumulation, with primary deposition concentrated in central–western East Dongting Lake, northwestern South Dongting Lake, most areas of Muping and Qili Lakes, and the Song-Li Floodway connecting them. Sandbar elevations in Qili Lake exceed 32 m, while sedimentation in Muping and South Dongting lakes is more extensive than in East Dongting Lake. The Xiang River channel in eastern Dongting Lake exhibits sub-10 m topography, demonstrating pronounced topographic contrasts.

Chenglingji station at Dongting Lake’s outflow and Xiaohezui station at the West–East Dongting confluence exhibit a mean annual stage difference of 5.22 m. Their July mean stages measure 28.23 m and 30.19 m, respectively, yielding a monthly difference of 1.96 m. This phenomenon arises from mutual backwater effects between the Yangtze River mainstream and Dongting Lake during flood seasons, inducing backwater rises at all the lake stations. Consequently, the stage difference between Chenglingji and Xiaohezui stations is significantly reduced in flood periods. Chenglingji and Xiaohezui stations exhibit January mean stages of 18.34 m and 26.73 m, respectively, yielding a maximum monthly stage difference of 8.39 m—the annual peak. During this period, the diminished flows in both the Yangtze River and Dongting Lake attenuate the mutual backwater effects. Compared to July, Chenglingji shows a more pronounced stage decline, while Xiaohezui maintains relatively higher stages. This contrast drives greater stage differences during dry seasons.

To investigate why upstream stations like Xiaohezui maintain relatively high stages during dry seasons, the longitudinal stage variations from Zhouwenmiao to Yingtian were mapped using measured stage data across Dongting Lake’s sub-regions, as depicted in Figure 11. Figure 11a presents the measured longitudinal water levels during dry seasons. Upstream of Yangliutan station, the levels exhibit a gentle gradient with an average slope of 0.018‰. Downstream, two segments display abrupt declines: Segment 1 (24.53–23.20 m), with a slope of 0.296‰, and Segment 2 (22.54–18.22 m), with a steep slope of 2.070‰. Remote sensing imagery and measured topography near the measurement points (Figure 10b–e) indicate diminished upstream inflows and reduced downstream stages during this period. Sandbars are extensively exposed across the region, with the channels between sandbars completely desiccated. Adjacent sandbars interconnect via expansive silt flats, while the drainage network relies exclusively on narrow inlets (20–60 m wide) for connectivity. Pronounced anthropogenic interventions are evident. Downstream of Yangliutan station, the lakebed exhibits a terraced topography segmented by two hydraulic drops into three distinct steps. This geomorphic configuration fundamentally drives the abrupt stage declines and sustains the elevated upstream stages during dry seasons.

When the Chenglingji stage reaches 24.06 m (Figure 11b), the longitudinal stage profile from Xiaohezui to Yingtian shows a smooth gradient without abrupt declines. This occurs because high water levels submerge silt flats between partially exposed sandbars, significantly expanding the water surface area and reducing the topographic control on stage variations. The hydraulic slopes upstream and downstream of Yangliutan are 0.036‰ and 0.086‰, respectively, indicating marginally steeper gradients downstream. When the Chenglingji stage reaches 29.33 m (Figure 11c), water levels in Muping Lake and South Dongting Lake both exceed 30 m. Under these high-stage conditions, most sandbars are submerged, resulting in a gentle longitudinal water-surface profile with minimal topographic influence on the stage distribution.

Beyond topographic controls, Dongting Lake receives inflows from the Three Yangtze Outflows and Four Major Tributaries. Hydrological stations within specific river channels exhibit stage variations strongly correlated with the discharge from their respective rivers—a contributing factor to the spatial stage heterogeneity across the lake. Analysis identifies 1990 as an abrupt stage-shift year for Qili Lake, demarcating two distinct hydrological periods. Unlike Muping, South Dongting, and East Dongting lakes, Qili Lake’s stage dynamics—being the uppermost sub-lake—are governed by inflows from the Lishui and Songzi rivers combined with downstream backwater effects. Furthermore, its connection to Muping Lake via the 50 km Song-Li Floodway is characterized by abundant sandbars that significantly impair hydraulic connectivity. Both Muping Lake and South Dongting Lake exhibited stage transitions in 1983 and 2003, demonstrating strong temporal synchronization. In contrast, East Dongting Lake showed divergent transition years, primarily attributable to its position in the lake system’s outflow section—where Lujiao and Chenglingji stations neighbor the Yangtze River—resulting in stronger influences from mainstream Yangtze flows and river–lake confluence dynamics.

5.2. Impacts on Lake Water Quality and Ecosystems

The water quality responses to water-level fluctuations in Dongting Lake exhibit significant spatiotemporal heterogeneity and staged characteristics across the hydrological year. The variations in water-quality parameters during distinct hydrological stages (e.g., low-flow, rising, and high-flow periods) clearly differentiate limnological and riverine phases, highlighting the dominant regulatory role of water-level dynamics in shaping intra-annual water quality patterns [15]. Spatially, the water quality exhibits significant heterogeneity, with the eastern Dongting Lake region demonstrating the most severe pollution. The concentrations of total nitrogen, total phosphorus, permanganate index, and chlorophyll a in this area are significantly higher than in other lake regions [12,48]. The operation of large-scale hydraulic projects such as the Three Gorges Dam has fundamentally altered key drivers of water-quality dynamics. Pre-project, the water-level fluctuations and suspended sediment load were primary water–sediment interaction drivers for total phosphorus (TP), with contribution rates of 11% and 9%, respectively. Post-project, the sediment load contribution decreased sharply to 2%, while lake-internal factors—including increased tributary TP inputs and sand mining activities—replaced Yangtze River influences as the dominant contributors to TP variations in Dongting Lake [49,50,51]. The water volume, water-level fluctuations, and nutrient inputs collectively constitute key drivers governing the spatiotemporal dynamics of water quality [52,53]. Under intensifying climate change and anthropogenic activities (e.g., water regulation, sand mining) [51,54], alterations in the environmental water allocation represent a primary factor driving the synergistic dynamics between lake water levels and water quality [54].

Dongting Lake, a critically important waterbird wintering site and migratory corridor in East Asia, exhibits significantly positive correlations between its water and marsh areas and waterbird abundance and species diversity. The water-level fluctuations in these wetlands serve as a primary driver regulating habitat availability for avian species [55]. Seasonal water-level fluctuations critically regulate waterbird aggregation patterns. Sustained high water levels during mid-winter constrain shallow-water habitats, reducing suitable habitat area and triggering functional turnover in waterbird communities [56]. The early dry season induced by the Three Gorges Dam impoundment has fundamentally altered the wetland hydrological regimes [57]. By shortening the inundation periods and advancing flood recession [58], this process significantly reshapes vegetation patterns and ecological processes: directly driving the most conspicuous change—expansion of dry-season vegetation and contraction of aquatic zones [59]—while modulating the distribution of keystone plants (e.g., Carex breviculmis, Phragmites australis) through altered submersion conditions [60]. Their areal changes exhibit significant sensitivity to key hydrological parameters, including the mean and minimum water levels during high-flow periods, as well as the mean, minimum, and maximum water levels during low-flow periods [61]. Water-level fluctuations further indirectly affect migratory bird habitat suitability by altering the vegetation distribution across elevation gradients [62]. For example, the wintering habitats of the Siberian Crane exhibit high sensitivity to short-term water-level changes: both rises and falls may cause habitat loss, while moderate fluctuations can create more extensive and concentrated habitats [63]. Anthropogenic hydrological degradation—such as downstream lake-level decline and intensified droughts induced by the Three Gorges Dam [64]—promotes terrestrial vegetation encroachment into lacustrine zones. This process synergistically suppresses wetland vegetation alongside elevated nutrient levels [65], propagating impacts through food chains to waterbirds: water-level declines directly reduce Carex biomass [66], the primary food source for herbivorous waterbirds, ultimately constraining the waterbird carrying capacity in wetlands. Furthermore, hydrological changes exhibit species-specific impacts, regulating the interannual distribution of the Lesser White-Fronted Goose (Anser erythropus) through altered food conditions [67], while the Bean Goose (Anser fabalis) shows no significant response [68]. Critically, even minor water-level changes (ΔWL < 0.2 m) may trigger irreversible ecological consequences [69]. Reservoir impoundment upstream—particularly during slow-filling phases—intensifies the competition among ecological objectives [64]. Collectively, the inflow volume and water level function as pivotal regulators [70]. Alterations in their magnitude, frequency, and duration profoundly reconfigure the wetland ecosystem structure and function through both direct pathways (e.g., vegetation distribution, submersion conditions) and indirect pathways (e.g., water quality parameters influencing phytoplankton dynamics [71] and seed bank density [72]).

5.3. Lake Management and Conservation Strategies

Dongting Lake, situated on the middle Yangtze alluvial plain, exhibits profound anthropogenic influences. Extensive reclamation [33] has fragmented the lake spatially and reduced the basin area, diminishing the flood-regulation capacity. Upstream hydraulic structures alter the intra-annual runoff distribution and decrease the sediment influx. Yangtze channel engineering accelerates the natural decay of the Three Outflows, reducing their flow–sediment distribution ratios. Operation of the TGD releases sediment-deficient flows, inevitably altering the erosion–deposition dynamics in Dongting Lake and the Yangtze River [23,73,74]. Recent socioeconomic development has escalated human demands on Dongting Lake’s functions, with navigation projects, water diversion systems, and sand mining activities projected to significantly alter the lake’s hydrological dynamics.

To address the spatial heterogeneity in Dongting Lake’s water-level dynamics and resolve water resource management and ecological challenges, we recommend implementing a three-tiered zoning control system using four hydraulic control projects to regulate three lake partitions (West/South/East Dongting), with the engineering locations detailed in Figure 12. The Chenglingji water control project serves as the primary hub at Dongting Lake’s outlet, regulating lake-wide water levels; the Leishishan project functions as a secondary hub at the South–East Dongting confluence, controlling water levels in both sub-lakes; while the Nanzui and Xiaohezui projects operate as tertiary hubs at the South–West Dongting junction, managing West Dongting’s water levels. Hydraulic projects at all levels require holistic integration of flood control, ecological integrity, water allocation, and navigation needs through scientifically formulated operation schemes to maximize the synergistic benefits. As a critical flood-regulation lake in the Yangtze Basin, Dongting Lake plays a vital role in watershed flood control. Adhering to the principle of regulating dry-season flows without flood control intervention, the floodgates remain fully open during high-flow periods, while hub operations elevate the lake levels in dry seasons to mitigate water scarcity, enhance irrigation and water supply reliability, and meet navigable depth requirements.

At the national level, the Yangtze Protection Law serves as the overarching legal framework prohibiting reclamation of hydraulically connected lakes, while Hunan Province’s complementary Dongting Lake Protection Regulation strengthens judicial accountability for wetland destruction and illegal sand mining activities. Hydrological authorities spearhead systematic ecological restoration initiatives—including returning reclaimed land to restore natural lake areas, dredging the Four-River Estuary channels, and reconstructing submerged vegetation belts for carbon sequestration and water purification—while implementing a Total Phosphorus Reduction Initiative targeting key pollution sources through agricultural fertilizer reduction, urban–rural wastewater plant upgrades, and comprehensive sediment dredging to holistically improve water quality. Biodiversity conservation prioritizes keystone species’ habitat restoration, exemplified by establishing elk dispersal corridors and constructing finless porpoise ex situ reserves, supported by intelligent monitoring networks integrating satellite remote sensing and ground-based IoT sensors to achieve real-time oversight. The policy’s core objective balances flood control, water supply, and ecological needs, driving Dongting Lake’s transition from engineered management to process-based restoration for sustainable transformation.

This study compiled a comprehensive dataset of daily water-level measurements from 21 hydrometric stations across Dongting Lake. This multi-station network captures the spatial heterogeneity in the water-level dynamics across distinct lake regions, with all the data updated through 2024 to characterize contemporary hydrological variations. Integrating repeated field measurements with the latest bathymetric data, this study reveals the underlying mechanisms driving the upstream–downstream water-level gradients in Dongting Lake during dry seasons, significantly enhancing the mechanistic understanding of spatiotemporal hydrological dynamics. This study has the following limitations. (1) Dongting Lake’s water-level evolution results from the synergistic effects of climate change, anthropogenic activities, and topographic alterations. While we revealed fundamental causes of spatial heterogeneity, we did not fully quantify these driving factors. Future work should compile meteorological, hydrological, and bathymetric datasets spanning pre- and post-alteration periods to comprehensively quantify their contributions. (2) Future studies should conduct quantitative assessments of the impacts of the proposed three-tiered hydraulic complexes on the water allocation efficiency and aquatic ecosystem integrity, utilizing integrated modeling approaches to evaluate multi-dimensional effects under varying hydrological regimes. (3) Future work could leverage functional data analysis (e.g., Kokoszka & Reimherr, 2017; Fan & Reimherr, 2017 [75,76]) to model the water-level dynamics as continuous curves, capturing nuanced spatiotemporal patterns beyond discrete metrics. (4) Notably, Dongting Lake’s stage dynamics persistently adjust to synergistic pressures from anthropogenic interventions, climate change, and evolving Yangtze–lake hydrodynamic interactions. More observations and analyses are required to further evaluate future water-level variations.

6. Conclusions

This study conducted a cluster analysis of hydrological stations using multi-temporal measured stage datasets. Change-point detection methods were employed to assess the trends and abrupt shifts at representative stations, followed by a comprehensive evaluation of hydrological alterations via IHA-RVA (Indicators of Hydrologic Alteration–Range of Variability Approach) before/after change points, elucidating the mechanisms driving spatial stage disparities. The key conclusions are summarized below.

(1) AHC was applied to stage data from 21 hydrological stations in Dongting Lake. At a Euclidean distance threshold > 5, the stations diverged into two primary clusters. Chenglingji, Lujiao, Yingtian, and Xiangyin (Group 1) exhibited high similarity, interconnected via the broad Xiangjiang floodway and the Yangtze River mainstem. The remaining stations formed Group 2. When reducing the Euclidean distance threshold to 4, three clusters emerged, with Group 2 subdivided into Group 2-1 and Group 2-2. Jinshi, Shiguishan, and Anxiang were clustered within a single subgroup under this resolution.

(2) Significant heterogeneity exists in the hydrological alteration degrees among Dongting Lake’s sub-lakes (Qili, Muping, South Dongting, and East Dongting), with variations observed even between representative stations within each sub-lake. Qili Lake exhibited an abrupt hydrological transition, dividing its record into two periods, where Jinshi and Shiguishan stations showed alteration degrees of 63% and 65%, respectively. Other sub-lakes were categorized into three periods. During T1, East Dongting experienced the most substantial alterations. Lujiao and Chenglingji exhibited 32% and 52% alteration degrees. Muping Lake followed with 34%, 34%, and 38% alterations at Nanju, Xiaohezui, and Zhouwenmiao stations. South Dongting showed minimal changes, with Yuanjiang, Yangliutan, and Yingtian at 31%, 22%, and 31%. Medium-to-high alterations predominantly occurred in the February–April and October mean stages, along with the Group 2 minimum stages. At Chenglingji and Lujiao, the minimum stages across durations showed medium-to-high alterations, primarily driven by Yangtze mainstream influences. During T2, all the stations demonstrated moderate alterations. In Muping and South Dongting, medium-to-high alterations concentrated in the April–June and October–November mean stages, low-stage frequency/duration, and annual minimum stages across durations, with Yingtian reaching high alteration in minimum stages. East Dongting’s medium-to-high alterations focused on the May–July and October mean stages, annual maximum stages, and high-stage duration.

(3) Given the pronounced spatial heterogeneity in Dongting Lake’s stage distribution, implementing zonation-based control principles is essential to sustain the dry-season water levels across lake subregions. We recommend implementing a three-tiered zoning control system for Dongting Lake using four strategically positioned hydraulic control projects to regulate three distinct partitions. Downstream of Yangliutan Station, the lakebed exhibits a terraced morphology demarcated by two hydraulic drops, forming three distinct topographic steps. This geomorphic configuration primarily drives the abrupt stage declines while enabling upstream areas to sustain elevated stages during dry periods. The declining dry-season stages at Yingtian station amplify the hydraulic head differences across the step-transition zone in eastern South Dongting Lake, potentially triggering headward erosion. This necessitates enhanced monitoring and stabilization measures for these critical geomorphic interfaces.