Ecological Risk Assessment of Storm-Flood Processes in Shallow Urban Lakes Based on Resilience Theory

Abstract

Urban shallow lakes are sentinel ecosystems whose stability is increasingly threatened by acute, sediment-laden storm floods. While chronic nutrient loading has been extensively studied, rapid risk assessment tools for short-pulse disturbances are still missing. Our aim was to develop a resilience-based, process-linked framework that couples depth-averaged hydrodynamics, advection-diffusion sediment transport and light-driven macrophyte habitat suitability to quantify hour-scale ecological risk and week-scale recovery. The ecological risk model integrates a depth-averaged hydrodynamic module, an advection–diffusion sediment transport routine, and species-specific light-suitability functions. We tested the model against field observations from Xinglong Lake (Chengdu, China) under 5-year and 50-year design storms. Ecological risk exhibited a clear west-to-east gradient. Under the 5-year storm, high-risk cells (complete inhibition) formed a narrow band at the eastern inlet and overlapped 82% with the SSC > 0.1 kg m⁻³ plume at 6 h; several western macrophyte beds returned to “suitable” status by 72 h. In contrast, the 50-year event pushed R > 0.9 over all macrophyte beds, with slow recovery after 192 h. Lake-scale risk peaked above 80% within 24 h for both return periods, but residual risk remained elevated in the 50-year scenario owing to the larger spatial footprint. The study provides a transferable early-warning tool for lake managers to decide when to trigger low-cost interventions and species-specific resilience rankings to guide targeted vegetation protection in shallow urban lakes worldwide.

1. Introduction

Urban shallow lakes typically exhibit characteristics such as relatively fragile ecosystems, low pollution load capacity [1], and unstable soil-water interfaces [2]. As stabilisers of urban ecosystems, shallow lakes provide critical ecological functions. To maintain ecological stability, comprehensive pollution source control measures are usually implemented. However, the progression of global climate change has altered local climatic characteristics, leading to frequent extreme storm events in urban areas [3]. Storm floods transport large quantities of pollutants into urban lakes, causing water quality deterioration and submerged macrophyte decline [4], thereby triggering regime shifts from clear-water to turbid-water states and severely degrading ecosystem functions [5].

To prevent undesirable regime shifts in lake ecosystems (e.g., abrupt transitions from clear-water to turbid-water states), scientifically assessing and precisely predicting ecological risks is critical. While chronic nutrient loading has been extensively studied, storm-driven acute disturbances remain under-represented in current risk indices. To address this gap, we distinguish acute storm stress (hours to days) from chronic nutrient stress (months to years), following the disturbance regime framework for aquatic systems [6] and empirical evidence that acute sediment pulses can override chronic nutrient effects on submerged plants [7]. This distinction is crucial because acute events can push ecosystems past tipping points even under low nutrient levels. For instance, a single rainstorm event delivering substantial sediment loads to a lake can trigger abrupt ecosystem regime shifts through the rapid increases in SSC during storms and attenuate underwater light, which is the primary controlling factor on macrophyte growth [8]; reduce photosynthetic oxygen evolution; and drive rapid vegetation collapse [9,10].

Consequently, lake managers lack rapid, mechanistic tools to predict when an incoming storm will push submerged macrophytes beyond their tolerance threshold and how long recovery will take. Existing models either (i) omit hydrodynamics [11], (ii) treat sediment as a passive tracer, or (iii) use static habitat indices that cannot capture transient light limitation [12,13]. This gap hampers timely, cost-effective interventions such as pre-storm drawdown or inflow diversion.

Here we hypothesise that: hour-scale peaks in suspended-sediment concentration (SSC) and water depth jointly determine acute ecological risk to submerged macrophytes; species-specific resilience can be quantified by integrating light-attenuation thresholds with cumulative exposure time; and a process-based model forced by forecast rainfall can provide operational early-warning at ≤24 h lead time.

The aim of this study was therefore to develop and validate a resilience-based, process-linked risk framework that couples open-source 2-D hydrodynamics, advection-diffusion sediment transport and light-driven habitat suitability to predict hour-scale risk and week-scale recovery of submerged macrophytes during storm floods. The framework is validated using field observations from Xinglong Lake, Chengdu, under two design storms (5-year and 50-year return periods).

2. Materials and Methods

2.1. Study Area

Xinglong Lake, located in Chengdu Tianfu New Area, covers a water area of over 4500 mu. Its upstream Luxi River originates from the western slope of Changsong Mountain, Longquanyi District, Chengdu, serves as a natural mountain stream characterised by large seasonal water level and flow variations and high sediment concentration. The climate is a transitional zone from humid hilly areas to plains, frequently experiencing extreme storms. After ecological restoration in October 2020, submerged macrophyte coverage reached 70%, including Vallisneria natans, Hydrilla verticillata and Ceratophyllum demersum (Figure 1). The normal water level is 464 m, with an average depth of 2.38 m and maximum depth of 9.02 m. During non-flood seasons, water transparency exceeds 3 m, forming a clear-water stable ecosystem. On July 15, 2021, a severe storm caused upstream sediment-laden floods to displace ~20% of lake volume, affecting 80% of the lake area and creating turbid conditions. Transparency recovered after 7 days, but extensive macrophyte mortality occurred.

2.2. Hydro-Sediment Model Construction

2.2.1. Hydrodynamic Governing Equations

Sediment transport and erosion/deposition calculations are based on hydrodynamic modelling. Considering clear flow boundaries and large width-to-depth ratios, 2D shallow-water equations are adopted:

$${\displaystyle \frac{\partial \zeta }{\partial t}}+{\displaystyle \frac{\partial p}{\partial x}}+{\displaystyle \frac{\partial q}{\partial y}} = {\displaystyle \frac{\partial d}{\partial t}}$$

(1)

$${\displaystyle \frac{\partial p}{\partial t}}+{\displaystyle \frac{\partial }{\partial x}}\left({\displaystyle \frac{p^2}{h}}\right)+{\displaystyle \frac{\partial }{\partial y}}\left({\displaystyle \frac{pq}{h}}\right)+g\mathrm{h}{\displaystyle \frac{\partial \xi }{\partial x}}+{\displaystyle \frac{gp\sqrt{p^2{+q}^2}}{C^2·h^2}}-{\displaystyle \frac{1}{\rho }}\left[{\displaystyle \frac{\partial }{\partial x}}\left({h\tau }_{xx}\right)+{\displaystyle \frac{\partial }{\partial y}}\left({h\tau }_{xy}\right)\right]-{\Omega }_q-{fVV}_x+{\displaystyle \frac{h}{\rho }}{\displaystyle \frac{\partial }{\partial x}}\left(P_a\right)=0$$

(2)

$${\displaystyle \frac{\partial q}{\partial t}}+{\displaystyle \frac{\partial }{\partial y}}\left({\displaystyle \frac{q^2}{h}}\right)+{\displaystyle \frac{\partial }{\partial x}}\left({\displaystyle \frac{pq}{h}}\right)+g\mathrm{h}{\displaystyle \frac{\partial \xi }{\partial y}}+{\displaystyle \frac{gp\sqrt{p^2{+q}^2}}{C^2·h^2}}-{\displaystyle \frac{1}{\rho }}\left[{\displaystyle \frac{\partial }{\partial y}}\left({h\tau }_{yy}\right)+{\displaystyle \frac{\partial }{\partial x}}\left({h\tau }_{xy}\right)\right]-{\Omega }_p-{fVV}_y+{\displaystyle \frac{h}{\rho }}{\displaystyle \frac{\partial }{\partial y}}\left(P_a\right)=0$$

(3)

where ζ is the water level, d is the time-varying water depth, and h is the water depth, h = ζ − d, m; p, q are the single-breadth flow in the x and y directions, m²/s; C is the Chézy coefficient, m^1/2/s; ρ is the density of water, kg/m³; τ_xx, τ_xy, τ_yy are effective shear stress components; Ω is Koch force coefficient; f is wind resistance coefficient; V, V_x, V_y are wind speed and wind velocity components in x and i directions, m/s.

2.2.2. Sediment Transport Model

Fine sediment transport (diameter < 0.06 mm) is modelled using the advection-diffusion equation:

$${\displaystyle \frac{\partial \rho ’}{\partial t}}+u{\displaystyle \frac{\partial \rho ’}{\partial x}}+v{\displaystyle \frac{\partial \rho ’}{\partial y}}={\displaystyle \frac{1}{h}}{\displaystyle \frac{\partial }{\partial x}}\left({hD}_x{\displaystyle \frac{\partial \rho ’}{\partial x}}\right)+{\displaystyle \frac{1}{h}}{\displaystyle \frac{\partial }{\partial y}}\left({hD}_y{\displaystyle \frac{\partial \rho ’}{\partial y}}\right)+\Delta S$$

(4)

where ρ’ is sediment mass density averaged over water depth; D_x, D_y are diffusion coefficient in x and y directions; ∆S is the source and sink terms indicating scour and siltation.

Stochastic equations for sediment-flow interactions to represent deposition rates:

$$S_D{ = w}_s{c}_h{p}_d$$

(5)

where $w_s$ is sinking speed; $c_h$ is nearshore sediment concentration; $p_d$ is deposition probability; $p_d = 1-{\displaystyle \frac{{\tau }_b}{{\tau }_{cd}}}{, \tau }_{b }\le { \tau }_{cd}, {\tau }_b$ is the bed bottom shear stress, and ${\tau }_{cd}$ is the critical shear stress for deposition.

Erosion is categorised into two cases depending on the characteristics of the bottom bed, with the following erosion rate equations for fully consolidated bottom beds in dense chambers:

$$S_E = E\left({\displaystyle \frac{{\tau }_b}{{\tau }_{\mathit{ce}}}}-1\right){,\tau }_b>{\tau }_{\mathit{ce}}$$

(6)

For loose, partially consolidated bottom beds, there is:

$${ S}_E = Eexp\left[{\alpha (\tau }_b-{\tau }_{\mathit{ce}}{)}^{1/2}\right]{,\tau }_b>{\tau }_{\mathit{ce}}$$

(7)

where E is bottom bed erosion per unit area per unit time, (kg/m²/s); ${\tau }_{ce}$ is critical shear stress for erosion, (N/M²); ${\tau }_b$ is bed shear stress, (N/M²); N is erosion index; α is topographic slope coefficient (m/$\sqrt{N}$).

2.2.3. Model Boundary Conditions

This paper aims to study the change pattern of lake suspended sediment content throughout its recovery process, covering a range of rainstorm intensities from low to extreme, specifically designed with return periods of five years (5a) and fifty years (50a), as two typical rainfall scenarios. The rainstorm design formula used in the Chengdu city centre:

$$i={\displaystyle \frac{44.594\times \left(1+0.651lgP\right)}{{\left(t+27.346\right)}^{0.953\left[{\left(lgP\right)}^{-0.017}\right]}}}$$

(8)

where i is rainfall intensity, mm/min; t is rainfall duration, min; p is rainstorm return period, year.

The rainfall duration is set to 24 h, and the design flood process for each scenario is shown. The initial water level of the model is set to 464 m, the sand content is given according to the experimental measured data, and the initial sediment concentration is 0.0023 kg/m³.

2.3. Resilience-Based Ecological Risk Assessment Model

2.3.1. Theoretical Basis

Resilience theory frames ecosystem stability as the capacity to resist, absorb and recover from disturbance, emphasising process over static state [14,15]. Because urban lakes are periodically subjected to sediment-laden storm floods, this perspective is well-suited to assess macrophyte vulnerability across return periods.

Submerged macrophytes are controlled primarily by underwater light availability. The relative light intensity (I_h/I₀) is therefore widely used as a habitat-suitability proxy [16,17]. Storm-driven increases in SSC and water depth rapidly reduce I_h/I₀, triggering macrophyte decline. Accordingly, we adopt I_h/I₀, as the evaluation index to quantify temporal changes in macrophyte performance and to estimate lake-wide ecological risk during storm events.

2.3.2. Experiments

During storm-flood events, the relative underwater light intensity is governed primarily by water depth and turbidity, expressed as:

$$I_{h }{= I}_0{e}^{-kh}$$

(9)

where k is the vertical light attenuation coefficient (m⁻¹) and h is water depth (m). A strong linear relationship exists between k and turbidity, which was determined experimentally.

(1) Relationship between turbidity and suspended-sediment concentration

Forty-eight water samples were collected from different locations and times in the lake. Each sample was analysed for suspended-sediment concentration (C, mg/L) and turbidity (T, NTU). Linear regression yielded:

$$T = 0.4845C + 0.223$$

(10)

where T is turbidity, NTU; C is suspended-sediment concentration, mg/L.

(2) Light-attenuation coefficient in sediment-laden water

Seven treatment levels were prepared: clear water and turbidities of 30, 60, 90, 120, 150 and 180 NTU. Under clear-sky conditions at solar noon, underwater light intensity was measured in situ at 0.1 m depth intervals. The attenuation coefficient k was calculated for each treatment and regressed against T, giving:

$$k = 0.0005T + 0.0176$$

(11)

Substituting Equation (11) into Equation (9) yields the predictive relationship:

$${\displaystyle \frac{I_{\mathrm{h}}}{I_0}}{=e}^{-\left(0.005T + 0.0176\right)\mathrm{h}}$$

(12)

2.3.3. Light-Suitability Curves for Macrophyte Growth Status

Because plant growth status is difficult to quantify with crisp numerical indices, fuzzy-mathematics evaluation is well suited to this task. We therefore adopted a fuzzy-logic approach to construct light-suitability curves for macrophyte performance. The procedure was as follows:

(1) For each macrophyte species, a suitability function (µ) was developed that relates growth status to relative underwater light intensity (I_h/I₀). The function was chosen to range between 0 and 1, where µ→1 indicates optimal growth and µ→0 indicates severe inhibition.

(2) Each species was assigned to three qualitative classes: suitable for growth, moderately inhibited, and completely inhibited; “completely inhibited” refers to when net growth rate falls below 0 and plants begin to senesce or die. The corresponding ranges of I_h/I₀ for each class were determined from field surveys and controlled laboratory experiments.

The resulting suitability functions for the three dominant submerged species planted in Xinglong Lake [4,18]—Vallisneria natans, Hydrilla verticillata and Ceratophyllum demersum—are illustrated in Figure 2.

2.3.4. Ecological Risk Assessment Model

This research developed an ecological risk evaluation system that uses submerged macrophyte growth status as the characteristic indicator for shallow lakes. First, the lakebed was partitioned by macrophyte species. Each species-specific zone was then subdivided into equal-area grid cells, denoted a_i, b_i, …, n_i, where a, b, …, n indicate the macrophyte species and i = 1, 2, 3, …, N. For each cell, the time series of growth-suitability values (fitted as a continuous curve f(t)) was extracted from the onset of the storm until lake turbidity returned to background levels. The cumulative ecological risk of a single cell is calculated as:

$$r ={\displaystyle \frac{1-{{\displaystyle \mathsf{\int }}}_{t_0}^{t_n}f\left(t\right)dt}{t_n}}$$

(13)

where r is the cell-level ecological risk (dimensionless, 0 ≤ r ≤ 1), and t₀→t_n is the full duration from storm initiation to turbidity recovery.

The lake-wide ecological risk is then aggregated as:

$${R=\omega }_1{\displaystyle \frac{{{\displaystyle \mathsf{\sum }}}_1^i{a}_i}{N_a}}{+\omega }_2{\displaystyle \frac{{{\displaystyle \mathsf{\sum }}}_1^i{b}_i}{N_b}}{+\dots \omega }_n{\displaystyle \frac{{{\displaystyle \mathsf{\sum }}}_1^i{n}_i}{N_n}}$$

(14)

where R is the overall ecological risk for the lake; N is the number of grid cells occupied by species s; and ω is the proportion of species s in the total vegetated area.

2.4. Uncertainty and Sensitivity Analysis Methods

To assess uncertainty in our predictions and identify influential parameters, we employed Monte Carlo simulation and sensitivity analysis.

2.4.1. Monte Carlo Simulation

We conducted Monte Carlo simulations (n = 1000) to quantify uncertainty in ecological risk assessments for both storm scenarios. Parameter distributions were established based on literature values and experimental data, including normal distributions for hydrodynamic parameters (Manning’s roughness, sediment settling velocity) and triangular distributions for ecological parameters (species-specific light thresholds). For each simulation run, we calculated 90% confidence intervals (5th–95th percentiles) for ecological risk at both lake-wide and species-specific levels.

2.4.2. Sensitivity Analysis

We employed Sobol’s variance-based sensitivity analysis to identify the most influential parameters affecting ecological risk predictions. For each parameter θᵢ, we calculated first-order sensitivity indices (Sᵢ) representing their direct contribution to output variance:

Sᵢ = Vᵢ/V

(15)

where Vᵢ is the variance attributed to parameter θᵢ and V is the total variance. We also calculated second-order indices (Sᵢⱼ) to quantify interaction effects between parameter pairs.

Key parameters examined included light attenuation coefficient (k), species-specific light thresholds, sediment settling velocity, critical SSC thresholds, bed roughness, and sediment input rates. Parameters with Sᵢ > 0.1 were classified as highly influential, while those with 0.05 < Sᵢ < 0.1 were considered moderately influential. We focused our analysis on lake-wide ecological risk magnitude and spatial extent of high-risk zones.

3. Results

3.1. Results of Lake Hydrodynamics Under Different Scenarios

Peak inflow arrived at 6 h. Under both scenarios, depth exceeded 6 m in the east and south, while macrophyte dominated central and western zones remained below 5 m (Figure 3). Peak depth rose by 0.71 m (5a) and 0.93 m (50a); the 0.22 m difference attenuated gradually after rainfall ceased.

Flow velocities were generally <0.05 m/s. Floodwater entering through the two eastern inlets rapidly decelerated, exerting limited momentum on the main lake. Along the primary flow axis, southern transects were faster than northern ones, whereas the eastern extremity became nearly stagnant, preventing significant bed shear. In the 50a case, propagation was quicker, yet high-velocity zones remained confined to the inlets and their confluence; central and eastern sectors showed only minor differences between scenarios.

3.2. Spatial Distribution of Total Suspended Sediment Concentration in Lakes

Four representative instants—6 h, 24 h, 72 h, and 192 h—were selected to examine the spatial distribution of total suspended-sediment concentration (SSC); the model results are presented in Figure 4 and Figure 5. The SSC pattern closely mirrors the flow-field distribution, exhibiting a strong correlation with current velocity. In the inlet vicinity, where velocities and turbulence are highest, lake SSC deviates only marginally from the inflow concentration (0.146 kg/m³), indicating limited additional resuspension from the bed. The dominant source of elevated SSC is thus the advection of sediment-laden floodwater. SSC declines with transport distance; during the rainfall period the 50a scenario spreads sediment markedly faster and farther along the main flow axis than the 5a scenario, underscoring the overriding influence of inflow sediment load. Between 6 h and 24 h, the high-SSC domain expands rapidly, yet the sediment plume does not reach the eastern outlet in either scenario.

After rainfall cessation, sediment continues to advance a short distance along the primary flow path, but flocculation and settling progressively reduce concentrations (Figure 6). At 72 h, in the 5a scenario only the outlet and central lake retain SSC > 0.1 kg/m³, whereas in the 50a scenario SSC > 0.1 kg/m³ persists not only in these regions but also along the southern and eastern corridors where flow remains enhanced. By 192 h, the 5a scenario shows residual SSC ≈ 0.05 kg/m³ near the inlets and the lake centre, while all other areas have returned to background levels. In contrast, the 50a scenario still exhibits SSC > 0.1 kg/m³ over parts of the southern lake; inlet and central concentrations are similar to those in the 5a case, but the affected area is larger, so the lake as a whole remains under sediment influence.

3.3. Ecological Risk Assessment Results

The resilience-based model was applied in real time to every vegetated zone in the lake. The spatial patterns of ecological risk are tightly coupled to sediment-transport pathways, displaying a clear “west-to-east” gradient of increasing risk (Figure 7).

Under the 5a scenario (Figure 8), high-risk zones (complete inhibition) form a narrow band adjacent to the eastern inflow. The overlap between these zones and the high-SSC (>0.1 kg/m³) area at 6 h reaches 82%. By 72 h, macrophyte beds in the western lake—situated away from the main sediment route—have reverted to the “suitable growth” category.

In the 50a scenario (Figure 9), the larger flood peak forces sediment plumes over topographic barriers (Figure 5). Vallisneria natans meadows in the southern lake experience complete inhibition within 24 h, with most cells showing limited recovery (<20% vegetated grid cells return to pre-storm R value) by 192 h, a prolonged stress response. This confirms that topography-controlled velocity decay is the primary driver of risk heterogeneity.

Dynamic risk curves for the three macrophyte species are presented in Figure 10, and lake-scale risk trajectories are shown in Figure 11.

For the 50a case, Vallisneria natans carries the lowest risk because its long, vertically adjustable leaves provide a wider tolerance to reduced light. Nevertheless, mechanical stress from water-level rise (>7 m on the eastern side) drives all species into complete inhibition. Hydrilla verticillata and Ceratophyllum demersum are almost entirely suppressed; their overall risk exceeds that of Vallisneria natans by more than 5%, and only scattered patches recover after 192 h. In contrast, the 5a case produces only modest depth increases, so post-peak risk is governed primarily by SSC distribution. Consequently, Ceratophyllum demersum beds—situated in zones less affected by high sediment loads—exhibit the lowest risk.

Lake-scale risk peaks within 24 h in both scenarios, exceeding 80%. Subsequent sediment settling and turbidity reduction drive risk downward, but recovery remains slow because the elevated water column prevents adequate light restoration in affected macrophyte zones.

3.4. Uncertainty and Sensitivity Analysis Results

3.4.1. Confidence Intervals for Ecological Risk Assessment

Monte Carlo simulations produced 90% confidence intervals for ecological risk assessments under both storm scenarios. Based on our model outputs, the lake-wide ecological risk for the 5a scenario was 0.85 [0.80–0.89], while the 50a scenario yielded 0.96 [0.93–0.98]. Notably, uncertainty was not spatially uniform; higher uncertainty was observed in transition zones between low- and high-risk areas, particularly along the sediment plume boundaries.

Species-specific risk assessments showed varying levels of uncertainty. Vallisneria natans exhibited moderate risk levels under the 5a scenario (approximately 0.87–0.90) and higher risk under the 50a scenario (approximately 0.93–0.96). Hydrilla verticillata showed the most dramatic difference between scenarios, with the 5a scenario producing distinctly lower risk values (approximately 0.73–0.79) compared to other species. Ceratophyllum demersum demonstrated the highest overall risk values across both scenarios, maintaining risk levels near 0.95–0.98 under the 50a scenario. This means that Ceratophyllum should often be given priority for restoration. However, it is worth noting that when the storm surge reaches 50 years, the recovery of Vallisneria natans also needs to be considered, and even guiding the water level to drop earlier can be considered.

3.4.2. Sensitivity of Risk Predictions to Key Parameters

Sensitivity analysis revealed that the light attenuation coefficient (k) was the most influential parameter for ecological risk predictions (Si = 0.42), followed by species-specific light thresholds (Si = 0.28) and sediment settling velocity (Si = 0.15). For recovery period predictions, the critical SSC thresholds emerged as the dominant parameters (Si = 0.53), with substantial interaction effects between SSC thresholds and light attenuation (Sij = 0.21).

The spatial extent of high-risk zones was most sensitive to the rate of sediment input during peak flow (Si = 0.37) and the critical settling velocity (Si = 0.33). Interestingly, bed roughness showed low direct influence (Si = 0.07) but significant interaction effects, indicating its importance through interactions with other parameters.

4. Discussion

4.1. Applicability of the Resilience-Based Ecological Risk Assessment Framework

Traditional lake ecological risk assessments rely on snapshot measurements of water-quality parameters (e.g., TP, TN, Chl-a) and biological indicators [19,20]. Such static approaches struggle to capture the non-linear dynamics of ecosystems exposed to short, intense disturbances. Carpenter et al. (2001) stressed that steady-state frameworks ignore threshold effects and lagged responses characteristic of extreme events [21]. By integrating resilience theory, this research developed a process-based, dynamic risk framework that quantifies ecosystem trajectories across the full disturbance–recovery cycle.

Building on the resilience theory framework introduced in Section 2.3.1, our time-integral model explicitly quantifies the cumulative deviation of macrophyte communities from normal growth during disturbance events. This approach directly operationalizes Walker et al.’s (2004) [14] dimensions of resistance and recovery into measurable parameters. Our time-integral model $r = 1-{{\displaystyle \mathsf{\int }}}_{t_0}^{t_n}f\left(t\right)d$/t_n explicitly quantifies the cumulative deviation of macrophyte communities from normal growth, aligning with the resilience tenet of “system absorption and function maintenance”. Scheffer et al. (2001) demonstrated pronounced hysteresis between clear-water and turbid states in shallow lakes; once a regime shift occurs, a stronger reversal is needed to return to the original state [21]. The incomplete recovery of Vallisneria meadows in the 50a storm scenario after 192 h corroborates this prediction. Compared with the long-term, nutrient-based risk assessment of Zhang et al. (2016) [4], our resilience framework resolves acute, hourly stress responses, supporting early-warning capabilities.

Methodologically, Forbes & Calow (2002) noted that conventional probabilistic risk assessment, though capable of handling uncertainty, is constrained by dose–response relationships ill-suited to complex ecological processes [22]. Our resilience framework couples a mechanistic physical model (hydrodynamics–sediment transport) with an ecological response model (macrophyte light suitability), delivering an end-to-end quantification from driving forces to ecological effects.

In practice, the resilience-based approach offers three distinct advantages:

• Temporal resolution: identifying transient responses (e.g., risk peaks within 24 h) that static methods overlook.
• Multi-scale integration: linking hourly stress to weekly recovery.
• Scenario analysis: providing quantitative support for adaptive management [23].

Moreover, suitability functions were derived primarily from laboratory experiments and literature, and long-term field validation is needed. The current framework focuses on light stress alone, neglecting interactions with nutrients, temperature, pH, etc. Future work should broaden the indicator set and further test the framework’s generality and robustness.

4.2. Mechanisms by Which Hydro-Sediment Processes Govern Ecological Risk in Shallow Lakes

The coupling between hydrodynamic–sediment transport and ecological risk in shallow lakes is complex and inherently interdisciplinary, bridging hydraulics, sediment dynamics, and ecology [24]. Kristensen et al. (1992) demonstrated that hydrodynamic conditions are the dominant control on sediment resuspension and associated nutrient release in shallow systems [25]. Our numerical simulations reveal a complete causal chain, flow-field forcing → sediment diffusion → light attenuation → plant stress, that quantitatively dissects how storm floods impact lake ecosystems.

(1) Lake morphology as the primary control.

Xinglong Lake’s elongated east–west geometry and dual-inlet layout generate a pronounced confluence effect, consistent with Håkanson’s (2005) theory on morphometric controls of lake hydrodynamics [26]. The resulting velocity field is spatially heterogeneous: mainstream corridors reach 0.05–0.08 m/s, whereas embayments and deeper zones fall below 0.02 m/s. This velocity gradient dictates both sediment transport capacity and settling patterns, forming a high-concentration transport corridor along the main flow and low-concentration retention zones laterally. Similar sediment segregation has been observed in Lake Ontario by Hamilton & Mitchell (1996), confirming that morphology-driven hydrodynamic zonation is a universal feature of shallow lakes [27].

(2) Spatiotemporal evolution of SSC.

SSC evolution follows clear physical laws:

Early phase (0–6 h): high concentrations (0.146 kg/m³) are confined to the inlets, with gradients up to 0.146–0.020 kg m⁻³/km.

Propagation phase (6–24 h): Sediment plumes advance along the main flow at 0.03–0.05 m/s.

Post-rain phase (24–192 h): SSC decays exponentially; the half-life is 48–72 h, matching flocculation–settling patterns reported by Droppo (2001) [28].

In the 50a scenario, the plume breaches topographic constraints, entering the previously sheltered southern shallows—an overflow effect that markedly expands the ecological risk footprint.

(3) Water-level rise as a co-driver.

Storm-induced depth increase not only attenuates light exponentially (Beer–Lambert law) but also alters hydraulic residence time and bed-shear distribution. A rise from 2.38 m to 3.31 m reduces I_h/I₀ by about 28% even under clear-water conditions. Vadeboncoeur et al. (2003) noted that depth effects are often underestimated in turbid water, combined depth and turbidity can depress photosynthetic rates by >50% [29].

(4) Particle size and residence time.

The suspended fraction here is dominated by particles <0.06 mm with settling velocities of 0.1–0.5 mm/s, requiring 24–48 h for complete deposition under quiescent conditions. Wind-induced resuspension and thermal stratification, however, can prolong suspension. Kristensen et al. (1992) showed that even light winds (2–3 m/s) maintain fine sediments in suspension [25], explaining why SSC in parts of Xinglong Lake remained above background after 192 h in the 50a scenario.

(5) Non-linear amplification of risk.

The two return-period storms trigger markedly different responses. Under the 5a scenario, risk is governed primarily by SSC and remains spatially limited. In contrast, the 50a scenario couples depth and SSC effects, producing a disproportionate increase in both risk magnitude and spatial extent consistent with ecosystem threshold theory [30]. The abrupt shift of Vallisneria zones from suitable to complete inhibition corroborates Carpenter et al.’s (2011) prediction of abrupt ecological transitions once disturbance intensity exceeds system capacity [31].

These mechanistic insights provide a scientific basis for lake management: regulating inflow distribution, optimising inlet layout, and implementing pre-storm drawdown can all curtail sediment transport and reduce ecological risk. Identifying ecologically vulnerable areas controlled by topography further supports targeted vegetation protection strategies.

4.3. Comparative Analysis with Conventional Ecological Risk Assessment Approaches

Conventional lake ecological risk assessments fall into three main categories: indicator-species methods, ecosystem-health indices, and composite water-quality indices. All differ markedly from the present study in temporal scale, assessment variables, and intended application [32].

Indicator-species approaches typically use zooplankton or fish as sentinels, focusing on quarterly or annual changes in population density and community structure. The zooplankton-based system developed by Jeppesen et al. (2000), for example, captures long-term lake trajectories but operates on a multi-month to multi-year cycle, making it insensitive to the instantaneous impacts of extreme events [33]. By contrast, our macrophyte-centred framework resolves hourly dynamics and identifies risk peaks within 24 h, improving temporal resolution by two to three orders of magnitude.

Ecosystem-health evaluations, such as the ecosystem-service valuation proposed by Costanza et al. (1997), rely on steady-state assumptions to gauge ecological integrity [30]. Yet Scheffer & Carpenter (2003) emphasised that shallow lakes exhibit pronounced bistability; linear, equilibrium models can therefore substantially underestimate the effects of extreme disturbances [34]. The dynamic resilience framework this research present tracks the entire disturbance–recovery trajectory, explicitly overcoming the steady-state limitation.

Composite water-quality indices, exemplified by the Trophic State Index (TSI), offer rapid assessment through multi-parameter weighting [35]. They suffer, however, from two drawbacks: (i) indicators are mainly chemical, providing only indirect insight into ecological processes, and (ii) reliance on instantaneous snapshots precludes accounting for cumulative impacts. Our study uses ecological function as the endpoint and quantifies cumulative risk via temporal integration, offering superior mechanistic transparency and timeliness compared with traditional indices.

Regarding applicability, conventional methods are tailored to chronic pressures such as eutrophication, whereas climate change demands tools for acute risk. Our framework is purpose-built for short, intense events, enabling quantitative evaluation of both impact magnitude and recovery time across storm intensities. By coupling physical-process (hydrodynamic-sediment transport) and ecological-response (macrophyte light suitability) models, we construct a complete causal chain, yielding stronger mechanistic insight and predictive power than traditional correlative analyses. We compared our resilience index R with Trophic State Index, and TSI failed to detect any risk 24 h after the 50a storm because TP remained <30 µg L⁻¹. In contrast, R peaked at 0.96, reflecting acute light limitation. This demonstrates the higher temporal resolution and sensitivity of our framework for short-pulse disturbances.

Nevertheless, the proposed method entails higher model complexity and data demands, employs a narrower indicator set, and requires broader validation for generalisability. While traditional techniques are limited in addressing acute disturbances, their simplicity and maturity remain valuable for routine monitoring. Future efforts should merge the strengths of both paradigms to develop an integrated assessment system that incorporates multiple timescales and multifactor interactions, including inflow variability, wind forcing, and nutrient fluxes.

5. Conclusions

To elucidate how storm-flood events influence shallow-lake ecosystems, we developed an innovative ecological risk assessment framework grounded in resilience theory. By mechanistically coupling hydrodynamics, sediment transport, and ecological responses, the model simultaneously resolves lake hydro- and eco-dynamics across multiple temporal scales, thereby realistically portraying the evolution of ecological risk during storm disturbances. Simulations of macrophyte risk patterns in Xinglong Lake under contrasting storm scenarios reveal the following key insights:

Distinct risk regimes emerge for different return periods. A 5a storm primarily affects the inflow area, whereas a 50a storm amplifies risk, breaches topographic constraints, and propagates impacts across the entire southern macrophyte zone.

Spatial heterogeneity is pronounced and governed by sediment-transport pathways. High-risk zones exhibit a “west-to-east” expansion driven by the prevailing flow field.

Species-specific resilience differs markedly. Vallisneria natans, owing to its adjustable canopy and elongated leaves, demonstrates superior resistance; Hydrilla verticillata and Ceratophyllum demersum are almost completely suppressed under elevated suspended-sediment concentrations. These findings provide scientific guidance for macrophyte selection in lake-restoration projects, underscoring the importance of prioritising resilient species and optimising community composition.

We identified a complete causal chain, flow-field forcing → sediment diffusion → light attenuation → plant stress, that clarifies how lake morphology dictates flow patterns, which in turn control sediment routing and concentration distributions. When storm intensity exceeds topographic thresholds, ecological risk exhibits non-linear amplification. Management interventions (e.g., reconfiguring inlets, redistributing inflows, pre-storm drawdown) can effectively alter hydro-sediment pathways and mitigate risk. The framework is transferable to other shallow lakes under diverse disturbance scenarios, offering robust support for ecosystem management and engineering measures aimed at enhancing ecological resilience. By integrating physical processes with ecological responses, the model delivers valuable insights for adaptive management strategies in an era of increasing extreme weather events under climate change.