Threshold-Driven Ecological Risk Transitions Revealed by High-Frequency Dissolved Oxygen Dynamics in a Shallow Eutrophic Lake

Jiaqi Tang; Jianbo Chang

doi:10.20944/preprints202606.0789.v1

Submitted:

09 June 2026

Posted:

10 June 2026

You are already at the latest version

Abstract

Ecological process interruption in shallow lakes often manifests as algal blooms, but multiple daily disturbances may accumulate before bloom occurrence. To characterize ecosystem dynamics, dissolved oxygen (DO), pH, water temperature, and electrical conductivity (EC) were selected as axis variables and monitored at 5-min intervals in a restored shallow urban lake in Wuhan, China. A total of 112,896 observations collected over 392 days were analyzed. Because external inputs were minimal except for rainfall, observed dynamics primarily reflected internal ecological processes. DO was identified as the most sensitive indicator of ecological instability due to its direct coupling with ecosystem metabolism. Piecewise logistic regression revealed a critical DO threshold of 3.62 mg L^-1, below which ecological risk increased sharply. The frequency, duration, and cumulative exposure of threshold exceedance provided effective early-warning signals of ecosystem deterioration. These results demonstrate that high-frequency monitoring of key environmental variables can reveal ecological transitions overlooked by conventional low-frequency assessments and provide a practical framework for early detection of hypoxia-related ecological risk in shallow lakes.

Keywords:

high-frequency data

;

dissolved oxygen threshold

;

ecological risk

;

shallow freshwater lake

Subject:

Environmental and Earth Sciences - Ecology

1. Introduction

Lake ecosystems are increasingly characterized by rapid metabolic fluctuations, nonlinear ecological transitions, and short-lived ecological stress events driven by eutrophication and climate warming. However, most ecological assessment frameworks still rely on weekly or monthly observations and therefore implicitly assume gradual ecosystem change. Traditional water-quality assessment commonly relies on integrated water-quality indices derived from a limited set of physicochemical variables and infrequent observations, which may fail to capture rapid ecological processes and short-term ecosystem instability [1]. This mismatch limits the ability to detect ecological thresholds and early instability in eutrophic shallow lakes. Conventional low-frequency monitoring often fails to capture diel fluctuations in dissolved oxygen (DO) and pH, thereby underestimating the intensity, duration, and recurrence of hypoxic stress events [2]. As a result, ecological deterioration may remain undetected until substantial degradation has already occurred, reducing opportunities for timely intervention and for preventing ecological risk [3,4,5].

High-frequency monitoring provides an opportunity to move beyond state-based assessments and to develop process-oriented indicators that resolve ecosystem metabolism and short-term instability [6,7]. In eutrophic shallow lakes, DO and pH exhibit pronounced diel oscillations driven by the balance between photosynthesis and community respiration [8,9]. These cycles repeatedly expose aquatic organisms to alternating periods of nocturnal oxygen stress and daytime recovery, suggesting that ecological risk may accumulate through recurrent daily metabolic interruptions rather than through isolated disturbance events [10,11]. Here, we use the term “axis parameters” to denote the four core variables—DO, pH, temperature, and electrical conductivity (EC)—that characterize variable-specific dynamical rhythms in shallow-lake systems. In this study, “axis parameters” does not refer to the figure's coordinate axes; rather, it emphasizes that these variables exhibit distinct temporal patterns that can be tracked through high-frequency monitoring to capture the process information underlying stability loss. Accordingly, monitoring these axis parameters at high frequency is expected to better capture the development of instability than relying solely on traditional state-based assessments. While DO directly reflects ecosystem metabolic balance and oxygen stress, pH serves as an indicator of biological production and carbon dynamics. Temperature regulates metabolic intensity and oxygen demand, whereas electrical conductivity (EC) reflects hydro-chemical conditions and external environmental influences. Together, these four environmental variables provide complementary information on ecosystem functioning and instability. When monitored at high frequency, they enable the quantification of nocturnal oxygen depletion, daytime recovery capacity, diel amplitude, and threshold-approaching trajectories that are typically overlooked by conventional monitoring programs.

Ecological deterioration in eutrophic shallow lakes may therefore emerge not from a single abrupt disturbance but from the cumulative effect of many small daily metabolic interruptions, including recurrent nocturnal oxygen depletion, incomplete daytime recovery, and repeated metabolic imbalances. Ignoring these accumulation processes may limit the ability of conventional assessment and prediction frameworks to identify approaching ecological transitions. Recent studies have emphasized the importance of identifying early-warning signals prior to regime shifts in aquatic ecosystems, although translating theoretical indicators into practical monitoring tools remains challenging [12]. To address this gap, this study developed a high-frequency ecological indicator framework for identifying metabolic instability, ecological thresholds, and ecological risk transitions in a eutrophic shallow lake. Using 112,896 observations collected at 5-min intervals over 392 monitoring days, we integrated diel metabolic analysis, nonlinear threshold detection, principal component analysis (PCA), and partial least squares structural equation modeling (PLS-SEM) to investigate the ecological roles of DO, pH, temperature, and EC in describing ecosystem metabolism and instability. Specifically, the study aimed to (i) identify ecological thresholds associated with risk occurrence, (ii) evaluate the ecological roles of the four environmental variables in characterizing ecosystem metabolism and instability, and (iii) clarify how repeated daily metabolic interruptions contribute to ecological risk transitions and ecological early-warning signals.

2. Materials and Methods

2.1. Study Area

This study was conducted in a shallow lake, Lijiajiao Lake, located in Wuhan City, Hubei Province, China (Figure 1), centered at approximately 30°37′N, 114°04′E. The lake is a typical shallow freshwater lake on the Jianghan Plain of Hubei Province, with a surface area of about 0.5 km² and an average depth of 1.5–2.0 m. The lake lacks major natural inflow or outflow rivers. It therefore functions as a relatively enclosed hydrological system with low water-exchange capacity, a typical feature of small urban and peri-urban lakes that increases their vulnerability to eutrophication [13]. Due to its long history of eutrophication, it is a pilot site under China’s one-lake-one-policy restoration framework during this research, which aims to improve lake water quality through ecosystem restoration.

2.2. Data Collection

2.2.1. Automatic Sampling of Axis Parameters

A floating-platform automatic monitoring station was self-assembled and equipped with multi-parameter sensors for continuous observation in Lijiajiao Lake. The monitoring system continuously measured the axis parameters, i.e., dissolved oxygen (DO), pH, water temperature, and electrical conductivity (EC), via multi-parameter sensors. Data were recorded at 5-min intervals from August 1, 2020 to December 31, 2021.

2.2.2. Manual Sampling of Total Phosphorus (TP) and Treatment

Total phosphorus (TP) was monitored via monthly manual sampling at the same location as the automatic monitoring station to ensure comparability of results. Surface water was collected at a depth of 0.5 m into pre-acid-washed polyethylene bottles (500 mL). Samples were acidified to pH < 2, stored at 4°C, and analyzed in the laboratory. TP concentrations were determined using the molybdate spectrophotometric method (GB/T 11893-1989), consistent with widely used colorimetric methods [14]. Quality control procedures included field duplicates (relative deviation ≤10%), field blanks, and laboratory control samples (error ≤±5%), in accordance with QA/QC guidelines for nutrient analysis in freshwater systems [15].

2.3. Data Sorting

After quality control, excluding values outside the sensor measurement range and removing logically inconsistent observations, hourly and daily averages were computed and used for subsequent processing and analyses [16].

Strict filtering and temporal alignment procedures were applied to handle the large dataset and occasional missing data resulting from sensor maintenance [17]. First, for each parameter, 5-min records were grouped by date. Days containing fewer than 288 records (<100% coverage of a full day at 5-min intervals) were excluded, and all observations from those days were removed. Next, the intersection of complete sampling days across the four parameters was extracted. After filtering, 392 complete days (112,896 observations) from August 2020 to December 2021 were retained.

2.4. Analytical Methods

2.4.1. Diel and Seasonal Variations of Axis Parameters

The diel fluctuations and seasonal variability in the four axis parameters, i.e., DO, pH, temperature, and EC, were characterized using high-frequency data analysis, which provides essential insights into diel ecological processes that traditional low-frequency sampling cannot capture [18,19]. The four parameters were categorized by day and plotted as long-term diel fluctuation diagrams to identify short-term ecological stress events [20,21]. A DO heatmap (“time × date”) was constructed to capture seasonal drift and low-oxygen zone formation. For each season, a representative day was selected to generate a 3D surface diagram illustrating the diel coupling among Temp, pH, and DO. Similar approaches have been widely used in studies of lake metabolism and hypoxia [9,22].

2.4.2. Identification of Key Ecological Indicators

Relationships among DO, pH, temperature, and EC were evaluated using Spearman’s rank correlation, with Pearson’s correlation applied as a robustness check. In addition, the dataset was divided into four seasons, and diel variation curves were compared across seasons to assess each parameter's responsiveness to short-term ecosystem processes. Seasonal diel patterns provide insight into how temperature, primary productivity, and ecosystem metabolism influence environmental dynamics [23,24]. Parameters exhibiting strong associations with other environmental variables, together with pronounced diel and seasonal variability, were considered more suitable indicators of ecosystem-state fluctuations and ecological risk transitions.

2.4.3. Piecewise Logistic Regression and Threshold Detection

A piecewise logistic regression model was used to determine the DO threshold associated with the occurrence of ecological risk. Ecological risk occurrence was represented as a binary response variable (1 = ecological risk event; 0 = non-risk condition). Candidate thresholds were generated from the observed DO distribution, and separate logistic regressions were fitted above and below each candidate threshold. The optimal threshold was determined by maximizing the total log-likelihood. To ensure model stability, candidate thresholds were excluded if they had insufficient sample size, produced single-category outcomes (all 0s or all 1s), or yielded non-convergent models. All valid thresholds were subsequently used to construct a threshold–log-likelihood curve to identify the most pronounced transition point. Threshold-based approaches are widely used for detecting ecological tipping points, regime shifts, and oxygen-related ecological disturbances [3,25,26,27]. The identified threshold was incorporated into annual diel DO variation plots to evaluate its correspondence with observed temporal dynamics and ecological risk events.

To further characterize oxygen-stress accumulation and recovery dynamics around the identified threshold, three threshold-based indicators were calculated during the ecological transition period: low-oxygen duration, oxygen-stress exposure index (OSEI), and daytime recovery rate. Low-oxygen duration was calculated as the cumulative daily duration during which DO remained below the identified threshold of 3.62 mg L⁻¹. OSEI was calculated as the cumulative oxygen deficit below the threshold:

OSEI = Σ(3.62 − DOᵢ) × Δt, for DOᵢ < 3.62 mg L⁻¹,

where DOᵢ represents each 5-min dissolved oxygen observation and Δt is the sampling interval expressed in hours. This metric integrates both the intensity and duration of oxygen stress.

Daytime recovery rate was used to quantify oxygen recovery capacity and was calculated as the mean positive 5-min DO increment during daytime periods (06:00–18:00), converted to an hourly rate (mg L⁻¹ h⁻¹). Temporal changes in these indicators were examined to evaluate the progression of oxygen stress and recovery dynamics before and after the ecological transition. All analyses were performed in MATLAB R2024a.

2.4.4. PCA Composite Evaluation Model

PCA was used to identify the dominant environmental gradients underlying high-frequency variability in DO, pH, temperature, and EC. Loading structures of retained components were used to evaluate the relative contributions of environmental indicators to lake metabolic dynamics and ecological variability. Composite indicator weights were subsequently calculated based on absolute loadings weighted by the variance explained by each principal component. PCA and composite weighting were conducted in MATLAB R2024a [28].

2.4.5. Exploratory PLS-SEM Analysis Based on PCA-Identified Latent Structure

To validate the dominant ecological gradient revealed by the preceding principal component analysis (PCA) and to further identify causal relationships among variables, this study constructed a partial least squares structural equation modeling (PLS-SEM) model. This model is suitable for environmental systems with small sample sizes, complex latent constructs, and non-normal distributions [29,30].

The PLS-SEM model included three latent variables: Water_Quality (TP and the proportion of pH > 8), DO_Condition (DO_mean, DO_below3.5_ratio, DO_std, and DO_daynight_diff), and Eco_Risk (ecological risk occurrence, 0/1). Reflective measurement models were specified for all constructs, and path coefficients and indicator loadings were used to evaluate relationships among variables. High-frequency DO and pH observations were aggregated into monthly metrics to match the monthly TP measurements. Given the limited sample size (n = 12), the analysis was intended primarily to examine hypothesized ecological pathways rather than provide strict inferential tests [31,32]. The model was estimated using MATLAB R2024a.

3. Results

3.1. Diel and Seasonal Variations of Axis Parameters

The pH and DO displayed pronounced diel oscillations (Figure 2a,b), while temperature showed a unimodal diel pattern with small amplitudes (Figure 2c). Daytime increases in DO and pH corresponded to intensified photosynthesis, whereas nighttime declines reflected enhanced respiration. EC exhibited the most stable diel pattern and the smallest diel amplitude (Figure 2d).

The DO decreased during early-day periods in warm seasons and expanded into broader low-DO zones in summer, followed by increased DO during subsequent seasons and relatively higher levels in winter(Figure 3).

The 3D surface plot revealed a consistent positive association between DO and pH across seasons, with elevated DO concentrations generally corresponding to higher pH values. In contrast, temperature exhibited relatively limited diel variation and primarily reflected seasonal differences. (Figure 4).

3.2. Selection of Key Ecological Indicators

The diel dynamics and seasonal variability of DO, pH, temperature, and EC were shown in Figure 5, Figure 6, Figure 7 and Figure 8. The DO and pH showed pronounced diel oscillations across seasons, whereas temperature mainly reflected seasonal variation, and EC remained comparatively stable within each day. DO concentrations generally increased during daytime and decreased during nighttime, and pH followed a similar diel pattern. The magnitude of diel fluctuations was greater during warm seasons, particularly in summer, when extended periods of low nighttime DO were observed. In contrast, temperature showed clear seasonal differences with relatively smooth diel variation, while EC exhibited no obvious diel pattern and showed the smallest within-day fluctuations throughout the monitoring period.

Spearman correlation showed a strong positive relationship between DO and pH (ρ = 0.891, p < 0.001), which was consistent with Pearson correlation results (r = 0.893). Temperature was negatively correlated with both DO (ρ = −0.587, p < 0.001; r = −0.647) and pH (ρ = −0.384, p < 0.001; r = −0.434). In contrast, EC exhibited relatively weak correlations with the other parameters (e.g., ρ = −0.140 with temperature and ρ = 0.195 with pH), indicating lower correlation coefficients than those observed among DO, pH, and temperature (Figure 9 and Figure 10).

3.3. Threshold Dynamics and Progressive Deterioration of Oxygen Recovery

Since DO was identified as the most informative parameter for describing ecosystem instability and ecological risk based on the indicator-screening results (Section 3.2), it was selected for threshold detection and ecological transition analysis. The threshold–log-likelihood analysis identified an optimal DO threshold of 3.62 mg L^-1 (Figure 11a). This threshold corresponded to the maximum log-likelihood and therefore represented the most probable point of ecological transition. A piecewise logistic regression revealed a clear nonlinear relationship between daily minimum DO and the occurrence of ecological risk (Figure 11b). Seasonal diel DO variations further showed that DO generally remained above 3.62 mg L^-1 during spring and winter, whereas below-threshold conditions occurred more frequently and persisted for longer durations during summer and autumn (Figure 12). This seasonal pattern coincided with periods characterized by elevated ecological risk and oxygen instability. In August 2021, diel DO fluctuations declined rapidly after August 4, accompanied by reduced diel amplitude, intensified nocturnal oxygen depletion, and increasingly frequent low-DO episodes. Although several threshold exceedances occurred before the ecological transition, the system initially retained the capacity to recover above the threshold during daytime. As oxygen-stress events accumulated, daytime recovery progressively weakened, and low-oxygen duration increased. After August 6, recovery was insufficient to offset nocturnal oxygen losses, resulting in a persistent low-oxygen state (Figure 13).

Quantitative indicators further revealed substantial changes in oxygen-stress exposure and recovery dynamics during the transition period (Figure 14). Low-oxygen duration remained close to zero before August 2 but increased rapidly thereafter, reaching more than 20 h d^-1 by August 4 and approaching continuous exposure (24 h d^-1) after August 6 (Figure 14a). Concurrently, the oxygen-stress exposure index (OSEI) increased sharply from 2.36 mg L^-1 h d^-1 on August 2 to more than 60 mg L^-1 h d^-1 after August 6, indicating a rapid increase in cumulative oxygen deficit (Figure 14b). In contrast, daytime recovery rate declined progressively throughout the transition period, decreasing from approximately 0.75 mg L^-1 h^-1 in early August to below 0.50 mg L^-1 h^-1 after August 6 (Figure 14c). These temporal changes coincided with the establishment of persistent low-oxygen conditions following the ecological transition.

3.4 PCA of Axis Parameters

After Z-score standardization, PCA on DO, temperature, pH, and EC showed that PC1 and PC2 explained 83.7% of total variance (PC1: 57.9%; PC2: 25.8%). Loadings indicated PC1 was dominated by positive loadings of DO and pH and a negative loading of temperature (Figure 15a). PC2 was dominated by EC (Figure 15a). Composite weights based on absolute loadings and variance contributions yielded DO = 0.28, pH = 0.27, temperature = 0.25, and EC = 0.20 (Figure 15b).

3.5 PLS-SEM Path Analysis

The pH_above8_ratio was the dominant indicator of Water_Quality (Figure 16), whereas the DO_mean and DO_below3.5_ratio contributed most strongly to DO_condition. The loading results (Figure 17) similarly indicated strong loadings of the pH_above8_ratio (−0.996), DO_mean (0.953), and DO_below3.5_ratio (−0.801), whereas DO_daynight_diff showed a comparatively weak loading (0.176). The cross-loading matrix demonstrated acceptable discriminant validity of the measurement model according to commonly used PLS-SEM criteria [29,33] (Figure 18). The structural model revealed a strong negative effect of Water_Quality on DO_Condition (−0.710) and a strong direct effect of DO_Condition on Eco_Risk (−0.727), whereas the direct effect of Water_Quality on Eco_Risk was comparatively weak (−0.167) (Figure 19). The path-effect analysis further indicated a Water_Quality → DO_Condition → Eco_Risk pathway.

4. Discussion

4.1. Ecological Roles of Four High-Frequency Environmental Parameters

The four monitored axis parameters exhibited distinct ecological functions and together provided complementary information for describing short-term ecosystem dynamics. Unlike conventional water-quality indicators that are often measured at low temporal frequency, dissolved oxygen (DO), pH, temperature, and electrical conductivity (EC) can be continuously monitored and respond rapidly to environmental change. Their high-frequency observations therefore provide direct insight into ecosystem processes operating at diel timescales [2,6,17].

Among the four parameters, DO showed the strongest ecological responsiveness. Seasonal diel curves revealed pronounced fluctuations between daytime oxygen production and nighttime oxygen depletion, while correlation analysis and PCA further identified DO as the most influential variable. Because DO integrates photosynthesis, respiration, organic matter decomposition, and atmospheric exchange, it directly reflects ecosystem metabolic balance and ecological condition [8,9,18]. The threshold analysis further demonstrated that changes in DO were closely associated with the occurrence of ecological risk, supporting its role as the most informative indicator of ecosystem instability.

The pH exhibited diel dynamics closely coupled with DO and showed the strongest positive correlation among all parameter pairs. This relationship reflects the shared influence of photosynthesis and respiration on oxygen and carbon cycling within the lake ecosystem [34]. In contrast, temperature primarily acted as a metabolic driver, regulating biological activity, oxygen demand, and oxygen solubility [35,36,37]. EC exhibited comparatively weak diel variability and weaker coupling with other variables, suggesting that it primarily reflects hydrochemical background conditions rather than short-term metabolic processes [38,39].

Collectively, the four parameters represent complementary dimensions of ecosystem functioning, with DO and pH characterizing metabolic processes, temperature regulating metabolic intensity, and EC describing hydrochemical conditions. However, the results consistently demonstrated that DO occupied a central position linking environmental variability to the development of ecological risk. This finding provides the ecological basis for the subsequent threshold analysis and highlights the value of high-frequency DO monitoring for ecological risk assessment and early warning in eutrophic shallow lakes.

4.2 Threshold Dynamics and Daily Accumulation of Oxygen Stress

Threshold analysis identified a critical dissolved oxygen concentration of 3.62 mg L-1, below which ecological risk increased rapidly. The ecological relevance of this threshold is supported by previous studies. A global synthesis of 872 experiments involving 206 species demonstrated that many aquatic organisms exhibit lethal or sublethal responses when dissolved oxygen concentrations decline to 2–5 mg L⁻¹ [40]. Importantly, high-frequency observations showed that threshold exceedance was not instantaneous. Instead, it was preceded by a sequence of recurrent daily metabolic interruptions: intensified nocturnal oxygen depletion, reduced diel amplitude, and incomplete daytime oxygen recovery. These patterns developed progressively over multiple days rather than emerging abruptly when DO fell below the threshold.

Ecologically, each nocturnal hypoxia episode can be interpreted as a small-scale interruption of normal metabolic functioning. While any single interruption may appear transient, repeated occurrences can increase physiological stress, reduce recovery efficiency, and weaken ecosystem resilience. Similar resilience-based assessment frameworks have emphasized that ecosystem vulnerability is often determined not only by disturbance intensity but also by the capacity for recovery following repeated stress events [41]. As these interruptions accumulate, the system becomes increasingly vulnerable to threshold transitions and subsequent ecological deterioration [42,43,44,45]. This daily accumulation perspective also reframes management targets: preventing large-scale transitions requires reducing the mechanism that permits repeated daily oxygen-stress buildup, rather than focusing only on static nutrient or biomass snapshots.

The transition-period analysis provides quantitative support for this accumulation process. Following the initial threshold exceedances, low-oxygen duration increased rapidly from near-zero to near-continuous exposure, while oxygen-stress exposure increased more than 20-fold within several days. At the same time, daytime recovery rate declined progressively, indicating a reduced capacity for daytime oxygen replenishment. These changes occurred concurrently with the establishment of persistent low-oxygen conditions after August 6. Rather than being associated with a single threshold-crossing event, the ecological transition was preceded by a period characterized by increasing oxygen-stress exposure and progressively weakened recovery dynamics. Similar patterns have been reported in ecosystems approaching ecological deterioration, where repeated stress events and reduced recovery capacity contribute to declining resilience and increased vulnerability to regime shifts [46].

In this sense, DO-based prediction gains reliability not only from statistical threshold detection, but also from the concordance between threshold timing and the observed trajectory of oxygen-stress accumulation and recovery deterioration. The progressive increase in low-oxygen duration and oxygen-stress exposure, together with declining daytime recovery rates, provides a process-based interpretation of ecosystem instability before ecological transition. Such temporal shifts are broadly consistent with previous observations that increasing instability and altered ecosystem dynamics may precede ecological deterioration [47].

4.3 Ecological Mechanisms Underlying the Central Role of Dissolved Oxygen

PCA indicated that DO and pH form the dominant environmental gradient, while temperature and EC contribute as additional axes with different functional roles. The dominance of DO and pH suggests that short-term ecosystem variability in this eutrophic shallow lake was primarily governed by metabolic processes. DO reflects the metabolic balance between photosynthesis and respiration and therefore captures the ecosystem’s recovery capacity and trajectory of instability. pH responds sensitively to photosynthetic carbon uptake and serves as an indicator of biological production and algal activity [34]. Temperature primarily modulates metabolic intensity and thus the propensity for oxygen instability [35,36,37], whereas EC largely represents background hydro-chemical conditions and external influences with weaker short-term coupling to metabolic dynamics [38,39].

PLS-SEM further supported a mechanistic pathway whereby water-quality pressure influences ecological risk primarily through the deterioration of DO conditions. The strong effect from Water_Quality to DO_Condition and the dominant direct effect from DO_Condition to Eco_Risk indicate that oxygen dynamics occupy a central mediating position between environmental stress and ecological response. This hierarchical structure is consistent with the PCA results, which identified a dominant environmental gradient characterized by coupled variation in DO- and pH-related parameters.

The central role of DO is ecologically reasonable because dissolved oxygen integrates multiple fast-responding ecosystem processes, including primary production, community respiration, and oxygen consumption associated with the decomposition of organic matter [8,34]. In eutrophic shallow lakes, enhanced primary productivity can simultaneously elevate pH and increase daytime oxygen concentrations, whereas nighttime respiration may result in substantial oxygen depletion before sunrise. Such coupled fluctuations in pH and DO are widely recognized as characteristic features of metabolically active shallow-lake ecosystems [48].

The strong contributions of DO_mean and DO_below3.5_ratio in the PLS-SEM model further suggest that both overall oxygen status and the duration of low-oxygen exposure are important determinants of ecological risk. Previous studies have shown that prolonged hypoxia can affect fish, zooplankton, benthic organisms, and microbial communities, often reducing ecosystem resilience and increasing vulnerability to ecological disturbance [49,50,51,52]. These observations are consistent with the threshold analysis presented in this study, which identified a marked increase in ecological risk below a DO concentration of approximately 3.62 mg L⁻¹.

Because DO integrates and responds rapidly to these ecosystem processes, changes in DO behavior can emerge earlier than substantial changes become detectable in many conventional low-frequency indicators. The cross-loading patterns observed in the measurement model further suggest that water-quality variation and oxygen dynamics were not fully independent. Similar coupling has been reported in eutrophic lakes, where nutrient enrichment alters ecosystem metabolism, amplifies diel oxygen fluctuations, and increases the occurrence of low-oxygen events [37,53]. Phosphorus plays a particularly important role in regulating primary productivity and metabolic responses in aquatic ecosystems, thereby influencing oxygen dynamics and ecosystem stability [54]. Consequently, ecological deterioration may develop through the cumulative effects of recurring oxygen stress rather than through abrupt changes in conventional water-quality variables.

This hierarchical mediation is consistent with the daily accumulation interpretation: water-quality pressure promotes conditions that degrade DO, and repeated DO deterioration then increases the probability of ecological transition. Rather than representing a single disturbance event, ecological risk may arise from the accumulation of many small metabolic disruptions that occur during daily ecosystem operations.

Overall, the combined results from PCA and PLS-SEM strengthen the reliability of DO-based prediction by linking statistical thresholds and pre-transition patterns to an ecologically interpretable process chain, in which DO conditions mediate the transfer of environmental stress into ecological risk. This pathway is broadly consistent with the nutrient enrichment–oxygen depletion–ecological deterioration sequence reported in eutrophic lake ecosystems [6]. Together, these findings highlight the potential of high-frequency DO monitoring to provide early warning of ecological risk and improve ecosystem assessment in shallow eutrophic lakes.

5. Conclusions

The results of this study suggest that ecological assessment of eutrophic shallow lakes should place greater emphasis on high-frequency monitoring of axis parameters and the cumulative effects of repeated daily metabolic stress. Unlike conventional assessments that rely primarily on infrequent measurements of nutrient concentrations or algal biomass, the four axis parameters examined in this study (DO, pH, temperature, and EC) directly reflect ecosystem functioning and its short-term variability. Their continuous observations provide a more direct and objective representation of ecosystem metabolic dynamics and ecological change.

By integrating diel fluctuation analysis, threshold detection, PCA, and PLS-SEM, this study demonstrated how high-frequency axis parameters can be used to identify ecological thresholds, quantify ecosystem instability, and reveal the pathways linking environmental stress to ecological risk. The transition-period analysis further showed that ecological deterioration was accompanied by progressive declines in daily minimum DO and daytime recovery capacity, together with increasing low-oxygen duration, indicating that persistent hypoxia emerged through the accumulation of repeated oxygen-stress events rather than from a single threshold-crossing event. Particular attention should be given to indicators describing nocturnal oxygen depletion, daytime oxygen recovery, and diel oxygen amplitude, as these metrics directly characterize ecosystem metabolic balance and recovery capacity. Integrating high-frequency monitoring approaches with analyses of environmental parameters provides a practical framework for detecting early metabolic imbalances, evaluating ecosystem resilience, identifying ecological thresholds, and improving ecological risk forecasting and adaptive management in shallow eutrophic lakes.

Author Contributions

J. Tang: data acquisition; TP laboratory/experimental data processing; data processing; high-frequency monitoring analysis; PCA modeling; threshold modeling; PLS-SEM construction; statistical analysis; visualization; and manuscript drafting. J. Chang: conceptual framework and research logic development; supervision; result interpretation; manuscript revision; and project administration. Both authors reviewed and approved the final manuscript.

Funding

This research was funded by the Hubei Provincial Technology Innovation Major Project, Grant Number 2019ACA154, China.

Data Availability Statement

Due to restrictions imposed by local environmental management authorities, the raw high-frequency sensor data and manual TP records cannot be made publicly available. Processed datasets and analysis scripts supporting the findings of this study can be obtained from the corresponding author upon reasonable request.

Acknowledgments

The authors gratefully acknowledge all colleagues involved in the long-term ecological monitoring program of Lijiajiao Lake. We particularly thank Prof. Xiaoning Liu, Miss Yue Deng, and Mr. Xingjian Li for their assistance with field sampling and laboratory analyses of total phosphorus (TP) during the study period. Their contributions provided important support for data acquisition and quality assurance. We also thank colleagues at the Institute of Hydroecology for their assistance with monitoring-system maintenance and field operations.

Conflicts of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Abbreviations

The following abbreviations are used in this manuscript:

OSEI	oxygen-stress exposure index

References

Anastasopoulos, P.; Akratos, C.S. A review on water quality indices. Hydroecol. Eng. 2025, 2, 10003. [Google Scholar] [CrossRef]
Rozemeijer, J.; Jordan, P.; Hooijboer, A.; Kronvang, B.; Glendell, M.; Hensley, R.; Rinke, K.; Stutter, M.; Bieroza, M.; Turner, R.; Mellander, P.E.; Thorburn, P.; Cassidy, R.; Appels, J.; Ouwerkerk, K.; Rode, M. Best practice in high-frequency water quality monitoring for improved management and assessment: A novel decision workflow. Environ. Monit. Assess. 2025, 197, 353. [Google Scholar] [CrossRef] [PubMed]
Carpenter, S.R.; Cole, J.J.; Pace, M.L.; Batt, R.; Brock, W.A.; Cline, T.; Coloso, J.; Hodgson, J.R.; Kitchell, J.F.; Seekell, D.A.; et al. Early warnings of regime shifts. Science 2011, 334, 1070–1074. [Google Scholar] [CrossRef]
Raymond, P.A.; Zappa, C.J.; Butman, D.; Bott, T.L.; Potter, J.; Mulholland, P.; Laursen, A.E.; McDowell, W.H.; Newbold, D. Implications of short-term variability for long-term trends in river metabolism. Proc. Natl. Acad. Sci. USA 2012, 109, 8563–8568. [Google Scholar] [CrossRef]
Gsell, A.S.; Özkundakci, D.; Selmeczy, G.B.; Huser, B.J.; Hansson, L.-A.; Adrian, R. On the sampling frequency for early warning indicators of lake eutrophication. J. Ecol. 2023, 111, 1952–1969. [Google Scholar] [CrossRef]
Wilhelm, S.; Adrian, R. Impact of summer warming on the thermal characteristics of a polymictic lake and consequences for oxygen dynamics. Freshw. Biol. 2008, 53, 226–237. [Google Scholar] [CrossRef]
Coloso, J.J.; Cole, J.J.; Pace, M.L. Difficulty in discerning drivers of lake ecosystem metabolism with high-frequency data. Ecosystems 2011, 14, 935–948. [Google Scholar] [CrossRef]
Odum, H.T. Primary production in flowing waters. Limnol. Oceanogr. 1956, 1, 102–117. [Google Scholar] [CrossRef]
Solomon, C.T.; Bruesewitz, D.A.; Richardson, D.C.; Rose, K.C.; Van de Bogert, M.C.; Hanson, P.C.; Kratz, T.K.; Larget, B.; Adrian, R.; Pace, M.L. Ecosystem respiration: Drivers of daily variation and background respiration in lakes. Limnol. Oceanogr. 2013, 58, 849–866. [Google Scholar] [CrossRef]
Altieri, A.H.; Gedan, K.B. Climate change and dead zones. Glob. Change Biol. 2015, 21, 1395–1406. [Google Scholar] [CrossRef]
Dugener, M.J.; Lesht, B.M.; McKinney, P.; Hawley, N. Recurringly hypoxic: Bottom water oxygen depletion is linked to temperature and precipitation in a Great Lakes estuary. Hydrobiology 2023, 2, 356–372. [Google Scholar] [CrossRef]
Stelzer, J.A.; Olden, J.D.; Dudgeon, D.; Vander Zanden, M.J. Early warning signals of regime shifts for aquatic systems: Can experiments help to bridge the gap between theory and real-world application? Ecol. Complex. 2021, 47, 100944. [Google Scholar] [CrossRef]
Jeppesen, E.; Søndergaard, M.; Meerhoff, M.; Lauridsen, T.L.; Jensen, J.P. Shallow lake restoration by nutrient loading reduction—some recent findings and challenges ahead. Hydrobiologia 2005, 584, 239–252. [Google Scholar] [CrossRef]
Murphy, J.; Riley, J.P. A modified single solution method for the determination of phosphate in natural waters. Anal. Chim. Acta 1962, 27, 31–36. [Google Scholar] [CrossRef]
APHA; AWWA; WEF. Standard Methods for the Examination of Water and Wastewater, 23rd ed.; American Public Health Association: Washington, DC, USA, 2017. [Google Scholar]
Peeters, F.; Atamanchuk, D.; Tengberg, A.; Encinas-Fernández, J.; Hofmann, H. Comparison of lake metabolic rates estimated from a diel CO2- and the common diel O2-technique. PLoS ONE 2016, 11, e0168393. [Google Scholar] [CrossRef]
Winslow, L.A.; Zwart, J.A.; Batt, R.D.; Dugan, H.A.; Woolway, R.I.; Corman, J.R.; Hanson, P.C.; Read, J.S. LakeMetabolizer: An R package for estimating lake metabolism from free-water oxygen using diverse statistical models. Inland Waters 2016, 6, 622–636. [Google Scholar] [CrossRef]
Staehr, P.A.; Bade, D.; Van de Bogert, M.; Koch, G.R.; Williamson, C.; Hanson, P.; Read, J.S. Lake metabolism and the diel oxygen technique: State of the science. Limnol. Oceanogr. Methods 2010, 8, 628–644. [Google Scholar] [CrossRef]
Rose, K.C.; et al. Annual patterns of high-frequency lake data from the global GLEON network. Inland Waters 2017, 7, 417–433. [Google Scholar] [CrossRef]
Giling, D.P.; Reyyer, J.; Imberger, J.; Stockwell, J.D.; Chiu, C.Y.; Loken, L.C.; Solomon, C.T. Delving deeper: Metabolic processes in the metalimnion of deep lakes. Limnol. Oceanogr. 2017, 62, 1652–1665. [Google Scholar] [CrossRef]
Stefanidis, K.; Gkelis, S.; Parlamas, P.; Papastergiadou, E. Differentiation in aquatic metabolism between littoral and pelagic habitats in a shallow Mediterranean lake. Water 2019, 11, 287. [Google Scholar] [CrossRef]
Hu, Z.; Liu, R.; Zhang, Y.; Ding, Y.; Liu, Y.; Zhou, Y. Temporal dynamics and drivers of ecosystem metabolism in a subtropical shallow lake. PLoS ONE 2015, 10, e0124601. [Google Scholar] [CrossRef]
Martinsen, K.T.; Brothers, S.M.; Andersen, M.R.; Sand-Jensen, K.; Staehr, P.A. Ecosystem metabolism and gradients of temperature, oxygen, and dissolved inorganic carbon in the littoral zone of a macrophyte-dominated lake. J. Geophys. Res. Biogeosci. 2022, 127, e2021JG006502. [Google Scholar] [CrossRef]
Kraemer, B.M.; Pilla, R.M.; Woolway, R.I.; Anneville, O.; Ban, S.; Colom-Montero, W.; Devlin, S.P.; Dokulil, M.T.; Gaiser, E.E.; Hambright, K.D.; Hessen, D.O.; Jöhnk, K.D.; Keller, W.; Knoll, L.B.; Leavitt, P.R.; Lepori, F.; Luger, M.S.; Maberly, S.C.; Müller-Navarra, D.C.; Paterson, A.M.; Pierson, D.C.; Richardson, D.C.; Rogora, M.; Rusak, J.A.; Sadro, S.; Salmaso, N.; Schmid, M.; Silow, E.A.; Sommaruga, R.; Stelzer, J.A.A.; Straile, D.; Thiery, W.; Timofeyev, M.A.; Verburg, P.; Weyhenmeyer, G.A.; Adrian, R. Climate change drives widespread shifts in lake thermal habitat. Nat. Clim. Chang. 2021, 11, 521–529. [Google Scholar] [CrossRef]
Bestelmeyer, B.T.; et al. Analysis of abrupt transitions in ecological systems. Ecosphere 2011, 2, 1–26. [Google Scholar] [CrossRef]
Andersen, T.; Carstensen, J.; Hernandez-Garcia, E.; Duarte, C.M. Ecological thresholds and regime shifts. Trends Ecol. Evol. 2009, 24, 49–57. [Google Scholar] [CrossRef] [PubMed]
Dakos, V.; Carpenter, S.R.; Brock, W.A.; Ellison, A.M.; Guttal, V.; Ives, A.R.; Kéfi, S.; Livina, V.; Seekell, D.A.; van Nes, E.H.; Scheffer, M. Methods for detecting early warnings of critical transitions in time series are illustrated using simulated ecological data. PLoS ONE 2012, 7, e41010. [Google Scholar] [CrossRef]
Begum, S.; Firdous, S.; Naeem, Z.; Gul-E-Saba Chaudhry, A.; S., A.; F., Z.; S., K.; S., A.; M., S.; Y.Y.; Tengku Muhammad, T.S. Combined multivariate statistical techniques and Water Quality Index (WQI) to evaluate spatial variation in water quality. Trop. Life Sci. Res. 2023, 34, 129–149. [CrossRef]
Hair, J.F.; Ringle, C.M.; Sarstedt, M. PLS-SEM: Indeed, a silver bullet. J. Mark. Theory Pract. 2011, 19, 139–152. [Google Scholar] [CrossRef]
Sánchez, G. PLS Path Modeling with R; Trowchez Editions: Berkeley, CA, USA, 2013; Available online: http://www.gastonsanchez.com/PLS_Path_Modeling_with_R.pdf (accessed on 7 June 2026).
Grace, J.B. Structural Equation Modeling and Natural Systems; Cambridge University Press: Cambridge, UK, 2006. [Google Scholar] [CrossRef]
Kim, D.; Hwang, S.; Shin, H.; Hwang, G. An estimated structural equation model to assess the effects of land use on water quality and benthic macroinvertebrates in streams of the Nam-Han River system, South Korea. Environ. Sci. Pollut. Res. 2020, 27, 9393–9407. [Google Scholar] [CrossRef]
Fornell, C.; Larcker, D.F. Evaluating structural equation models with unobservable variables and measurement error. J. Mark. Res. 1981, 18, 39–50. [Google Scholar] [CrossRef]
Cole, J.J.; Pace, M.L.; Carpenter, S.R.; Kitchell, J.F. Persistence of net heterotrophy in lakes during nutrient addition and food web manipulations. Limnol. Oceanogr. 2000, 45, 1718–1730. [Google Scholar] [CrossRef]
Weiss, R.F. The solubility of oxygen in water and seawater. Deep-Sea Res. Oceanogr. Abstr. 1970, 17, 721–735. [Google Scholar] [CrossRef]
Hanson, P.C.; Carpenter, S.R.; Kimura, N.; Wu, C.H.; Cornelius, S.P.; Kratz, T.K. Evaluation of metabolism models for free-water dissolved oxygen methods in lakes. Limnol. Oceanogr. Methods 2008, 6, 454–465. [Google Scholar] [CrossRef]
Woolway, R.I.; Kraemer, B.M.; Lenters, J.D.; Merchant, C.J.; O’Reilly, C.M.; Sharma, S. Global lake responses to climate change. Nat. Rev. Earth Environ. 2020, 1, 388–403. [Google Scholar] [CrossRef]
Wetzel, R.G. Limnology: Lake and River Ecosystems, 3rd ed.; Academic Press: San Diego, CA, USA, 2001. [Google Scholar]
Wang, W.; Li, W.; Xue, M.; Gu, X.; Ye, C.; Jiao, Y.; Liu, B.; Han, Y.; Tong, Y.; Zhang, X. Spatial–temporal characteristics and influencing factors of lake water and groundwater chemistry in Hulun Lake, Northeast China. Water 2023, 15, 937. [Google Scholar] [CrossRef]
Vaquer-Sunyer, R.; Duarte, C.M. Thresholds of hypoxia for marine biodiversity. Proc. Natl. Acad. Sci. USA 2008, 105, 15452–15457. [Google Scholar] [CrossRef]
Peng, X.; Wen, S.; He, Y.; Wu, S. A framework for watershed flood resilience in the context of climate change: Concept, assessment, and application. Hydroecol. Eng. 2025, 2, 10007. [Google Scholar] [CrossRef]
Ortiz, J.D.; Chandra, S.; Schladow, S.G.; McKenzie, V.J.; Kaylor, M.J. Detecting changes in resilience indicators prior to a critical transition in lakes. Ecosphere 2020, 11, e03200. [Google Scholar] [CrossRef]
Paerl, H.W.; Xu, H.; Hall, N.S.; Rossignol, K.L.; Joyner, A.R.; Fulton, R.S. Nutrient–oxygen dynamics during cyanobacterial blooms. Harmful Algae 2016, 54, 14–27. [Google Scholar] [CrossRef]
Jane, S.F.; Winslow, L.A.; Remucal, C.K.; et al. Widespread deoxygenation of temperate lakes. Nature 2021, 594, 66–70. [Google Scholar] [CrossRef] [PubMed]
Kong, X.; Xu, G.; Zhou, Q.; Yu, X.; Shi, X. Synergistic effects of warming and internal nutrient loading on cyanobacterial blooms in eutrophicated lakes. Environ. Sci. Technol. 2023, 57, 1638–1649. [Google Scholar] [CrossRef]
Carpenter, S.R.; Brock, W.A. Stochastic dynamics of phycocyanin in years of contrasting phosphorus load. Ecosphere 2024, 15, e4903. [Google Scholar] [CrossRef]
Clements, C.F.; Ozgul, A.; Benton, T.G. Early warning signals of recovery in complex systems. Nat. Commun. 2019, 10, 1681. [Google Scholar] [CrossRef]
Khan, H.; Laas, A.; Marcé, R.; Obrador, B. Major effects of alkalinity on the relationship between metabolism and dissolved inorganic carbon dynamics in lakes. Ecosystems 2020, 23, 1566–1580. [Google Scholar] [CrossRef]
Diaz, R.J.; Rosenberg, R. Spreading dead zones and consequences for marine ecosystems. Science 2008, 321, 926–929. [Google Scholar] [CrossRef] [PubMed]
Friedrich, J.; Janssen, F.; Aleynik, D.; Bange, H.W.; et al. Investigating hypoxia in aquatic environments: Diverse approaches to addressing a complex phenomenon. Biogeosciences 2014, 11, 1215–1259. [Google Scholar] [CrossRef]
McCullough, I.M.; et al. Diel oxygen dynamics reveal shifts in metabolic regimes in lakes across gradients of eutrophication. Limnol. Oceanogr. Lett. 2018, 3, 473–481. [Google Scholar] [CrossRef]
Breitburg, D.; Levin, L.A.; Oschlies, A.; et al. Declining oxygen in the global ocean and coastal waters. Science 2018, 359, eaam7240. [Google Scholar] [CrossRef]
Xu, Z.; Xu, Y.-J. Determination of trophic state changes with diel dissolved oxygen: A case study in a shallow lake. Water Environ. Res. 2015, 87, 1970–1979. [Google Scholar] [CrossRef]
Liu, J.; Yu, Y.; Liu, M.; Liu, X. A review of phosphorous in fluvial floodplains: Source or sink? Hydroecol. Eng. 2025, 2, 10001. [Google Scholar] [CrossRef]

Figure 1. The Lijiajiao Lake. The main panel shows the delineated lake boundary and the high-frequency data sampling site used in this study. Insets indicate the geographical location of the study area within China and Hubei Province.

Figure 2. Diel variation patterns of the four axis parameters in Lijiajiao Lake are illustrated with high-frequency data at 5-min intervals representing one complete 24-h cycle. (a) DO, (b) pH, (c) temperature, and (d) EC.

Figure 3. Heatmaps of DO diel dynamics in Lijiajiao Lake during the monitoring period, highlighting pronounced seasonal differences of DO concentration (mg L⁻¹) in diel behaviors.

Figure 4. Coordinated diel variations among DO, pH, and temperature during representative periods of different seasons in Lijiajiao Lake. The x-axis represents water temperature (°C), the y-axis represents time of day (h), and the z-axis represents DO concentration (mg L⁻¹). Colors correspond to pH values. The surfaces illustrate how diel timing relates to concurrent changes in DO and pH under different thermal conditions.

Figure 5. Seasonal diel fluctuations of DO occur during spring, summer, autumn, and winter. The x-axis denotes time (5-min intervals), and the y-axis denotes DO concentration (mg L^-1).

Figure 6. Seasonal diel fluctuations of pH occur during spring, summer, autumn, and winter. The x-axis denotes time, and the y-axis denotes pH.

Figure 7. Seasonal diel fluctuations of water temperature occur during spring, summer, autumn, and winter. The x-axis denotes time, and the y-axis denotes water temperature (°C).

Figure 8. Seasonal diel fluctuations of EC occur during spring, summer, autumn, and winter. The x-axis denotes time, and the y-axis denotes EC (μS cm⁻¹).

Figure 9. Spearman correlation coefficient matrix among DO, pH, temperature, and EC. Color intensity indicates the magnitude and direction of Spearman correlation coefficients. A significant positive association was observed between DO and pH (ρ = 0.891, p < 0.001). Temperature showed significant negative correlations with DO and pH, whereas EC exhibited comparatively weak correlations with the other parameters.

Figure 10. Pearson correlation coefficient matrix among DO, pH, temperature, and EC. Color intensity indicates the magnitude and direction of Pearson correlation coefficients. DO and pH were strongly and positively correlated (r = 0.893, p < 0.001). Temperature showed significant negative correlations with both DO and pH. In contrast, EC exhibited weaker correlations with the other parameters compared with temperature and the pH–DO pair.

Figure 11. Identification of a dissolved oxygen threshold associated with ecological risk transitions in Lijiajiao Lake. (a) Threshold–log-likelihood curve obtained from piecewise logistic regression showing the maximum log-likelihood occurred at DO = 3.62 mg L⁻¹, identifying the most probable ecological threshold. (b) Piecewise logistic relationship between daily minimum dissolved oxygen and ecological risk occurrence, indicating a nonlinear ecological transition associated with oxygen depletion.

Figure 12. Seasonal diel DO variation with the ecological threshold curves in Lijiajiao Lake, with the identified ecological threshold of 3.62 mg L⁻¹ indicated by the dashed orange line. Blue lines represent individual diel DO cycles recorded at 5-min intervals, and the shaded areas indicate periods when DO fell below the ecological threshold.

Figure 13. Continuous DO time series during the ecological risk escalation period in August 2021. The red dashed line indicates the first threshold crossing (DO < 3.62 mg L^-1), after which recurrent nocturnal oxygen depletion, reduced diel amplitude, and incomplete daytime recovery became increasingly pronounced.

Figure 14. Quantitative indicators of oxygen-stress accumulation and recovery dynamics during the ecological transition period. (a) Daily low-oxygen duration below the identified dissolved oxygen threshold of 3.62 mg L-1. (b) Oxygen-stress exposure index (OSEI), representing the cumulative oxygen deficit below the threshold. (c) Daytime recovery rate, representing the rate of daytime oxygen recovery. The vertical dashed line indicates August 6, when persistent low-oxygen conditions became established.

Figure 15. PCA loadings and relative contributions of DO, pH, temperature (Temp), and EC. (a) Loadings of PC1 and PC2. (b) Combined parameter weights derived from PCA loadings and explained variance.

Figure 16. Indicator importance values in the PLS-SEM model for Lijiajiao Lake. Bars represent the absolute contributions of each indicator to its corresponding latent construct (Water_Quality, DO_Condition, and Eco_Risk). Within the Water_Quality construct, pH_above8_ratio showed the highest contribution. Within the DO_Condition construct, DO_mean and DO_below3.5_ratio contributed most strongly, whereas DO_daynight_diff showed a relatively weak contribution.

Figure 17. Indicator loadings on latent constructs in the PLS-SEM model for Lijiajiao Lake. The Water_Quality construct was dominated by pH_above8_ratio, which showed a strong negative loading (−0.996), whereas TP exhibited a relatively weak positive loading (0.232). Within the DO_Condition construct, DO_mean showed the strongest positive loading (0.953), while DO_below3.5_ratio and DO_std exhibited strong negative loadings (−0.801 and −0.571, respectively). DO_daynight_diff showed comparatively weak loading (0.176). The Eco_Risk construct was represented by a single indicator with a loading value of 1.00.

Figure 18. Cross-loading matrix of the observed indicators on latent constructs in the PLS-SEM model for Lijiajiao Lake. Color intensity represents the magnitude and direction of the loading values. Indicators associated with DO_Condition and Eco_Risk generally showed the strongest loadings on their respective latent variables. In contrast, several Water_Quality indicators exhibited noticeable cross-loadings on other latent constructs, indicating weaker separation between Water_Quality and the oxygen-related dimensions.

Figure 19. Path coefficient matrix in the PLS-SEM model for Lijiajiao Lake. The heatmap illustrates the direction and magnitude of relationships among latent variables. Water_Quality showed a strong negative effect on DO_Condition (−0.710). DO_Condition exerted the strongest direct effect on Eco_Risk (−0.727). In contrast, the direct effect of Water_Quality on Eco_Risk was comparatively weak (−0.167).

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.