Method and system for analyzing hydrological-ecological multi-element change coupling of rivers and lakes

Abstract

The application discloses a kind of river lake hydrology-ecological multi-element change coupling analysis method and system.The method comprises: obtaining the historical monitoring data of water resources elements, water environment elements and water ecological elements of river lake, and river-lake water level difference sequence;Based on river-lake water level difference sequence, sensitivity analysis is carried out on candidate index, and key river-lake health evaluation index set is constructed;Coupling water resources, water environment and water ecological simulation sub-module to build river-lake health evaluation model;Based on historical monitoring data, a multi-dimensional joint distribution model is constructed and random scenario data is generated;Random scenario data and index set are input into the model to solve multi-objective optimization problem, and health evaluation value is obtained and comprehensive weight is determined.The application solves the problems of static identification of traditional evaluation method index and one-way coupling process, and improves the accuracy and physical reasonableness of evaluation result.

Description

Technical Field

[0001] This invention relates to the field of hydrological monitoring and river and lake assessment technology, and in particular to a method and system for coupled analysis of hydrological and ecological multi-element changes in lakes connected to rivers. Background Technology

[0002] As crucial links in the river-lake system, interconnected lakes and estuaries exhibit complex and intertwined relationships among their hydrological processes, water conservancy projects, water quality, and aquatic biological systems. In-depth analysis of the response mechanisms of these multiple factors during their evolution is of significant technical guiding importance for accurately identifying the evolutionary patterns of river and lake health, optimizing watershed water resource allocation, and developing refined ecological restoration strategies.

[0003] Currently, health assessments of complex water systems typically employ distributed hydrological models or water quality simulation tools for multi-scenario simulations, combined with mathematical tools such as the analytic hierarchy process (AHP) or entropy weighting method to construct a comprehensive evaluation index system. In practical applications, the identification and selection of relevant evaluation indicators are often based on long-term statistical data for static analysis over all time periods. Furthermore, existing multi-factor coupled simulation schemes mostly adopt a forward-driven approach, using the calculation results of the hydrological module as the boundary input for the water environment or aquatic ecosystem module. This approach struggles to accurately depict the nonlinear feedback effects of changes in bottom habitat elements on upstream hydrodynamic processes when dealing with interconnected lakes and rivers experiencing significant hydrological fluctuations. Moreover, in large-scale stochastic health assessments, existing indicator weighting logic often relies on fixed weights or highly manual judgment methods, facing technical limitations such as difficulty in converging computational logic and deviations between evaluation results and physical reality when processing massive evaluation matrices.

[0004] Therefore, in complex and alternating hydrological environments, improving the accuracy of multi-factor coupled evaluation systems in depicting dynamic physical processes and achieving self-consistency and automated processing of evaluation logic under massive stochastic scenarios are common technical problems that urgently need to be solved in the field of river and lake health assessment. It is necessary to study a technical method that can improve the robustness and accuracy of multi-factor coupled analysis to support the scientific characterization of the health status of complex water systems. Summary of the Invention

[0005] Therefore, this application discloses the following technical solution:

[0006] A method for coupled analysis of hydrological and ecological changes in lakes connected to rivers includes:

[0007] To obtain historical monitoring data on water resources, water environment, and water ecology elements of lakes connected to the Yangtze River, as well as the water level difference sequence of the lakes and rivers;

[0008] Sensitivity analysis was performed on candidate indicators based on the river and lake water level difference sequence to construct a set of key river and lake health evaluation indicators;

[0009] A river and lake health assessment model is constructed by coupling water resources, water environment and water ecology simulation sub-modules. The water ecology simulation sub-module achieves iterative feedback with the water resources simulation sub-module through physical feedback variables.

[0010] A multidimensional joint distribution model is constructed based on historical monitoring data, and multiple sets of random scenario data are generated based on the multidimensional joint distribution model.

[0011] Random scenario data and a set of key river and lake health evaluation indicators are input into the river and lake health evaluation model. The river and lake health evaluation values ​​under each scenario are obtained by solving a multi-objective optimization problem, and the comprehensive weight of each indicator is determined based on the river and lake health evaluation values.

[0012] Optionally, before constructing a set of key river and lake health evaluation indicators by performing sensitivity analysis on candidate indicators based on the river and lake water level difference series, the following steps are also included:

[0013] Meta-analysis was used to quantitatively synthesize multiple pre-set research data to identify candidate indicators for river and lake health assessment and to construct an initial indicator set.

[0014] Optionally, before performing sensitivity analysis on candidate metrics, the following may also be included:

[0015] Calculate the correlation matrix among the indicators in the initial indicator set, and use the correlation matrix as input for principal component analysis to perform eigenvalue decomposition;

[0016] Principal components are extracted based on the preset cumulative variance contribution rate threshold, and the original index with the largest absolute value of load in each principal component is selected as the representative index. After removing redundant indexes, a simplified index set is obtained.

[0017] Optionally, coupling water resources, water environment, and water ecology simulation sub-modules to construct a river and lake health assessment model includes:

[0018] A water resources simulation submodule with runoff and water level simulation functions is constructed based on the hydrological response unit;

[0019] Construct a water environment simulation submodule that includes nutrient cycling and water quality parameter simulation functions;

[0020] Construct a water ecology simulation submodule that includes functions for simulating the growth dynamics of algae and vegetation.

[0021] Optionally, the physical feedback variable is the comprehensive Manning roughness coefficient, and the iterative feedback process includes:

[0022] Extract the submerged vegetation coverage and emergent vegetation coverage of shallow water areas of Tongjiang Lake from the water ecology simulation submodule;

[0023] The comprehensive Manning roughness coefficient is updated by adding the preset bare bed roughness, the product of the incremental contribution of submerged vegetation to the roughness and the submerged vegetation coverage, and the product of the incremental contribution of emergent vegetation to the roughness and the emergent vegetation coverage.

[0024] The updated integrated Manning roughness coefficient is injected into the water resources simulation submodule to correct the relationship between water level and flow rate, and is used as the input parameter for the next iteration step.

[0025] Optionally, when updating the overall Manning roughness coefficient, the following may also be included:

[0026] Determine whether the calculated value obtained by addition exceeds the preset physical upper limit constraint;

[0027] If the value is exceeded, the overall Manning roughness coefficient will be set to the value of the physical upper limit constraint to restrict the roughness within a reasonable physical range.

[0028] Optionally, a multidimensional joint distribution model can be constructed based on historical monitoring data, including:

[0029] Marginal distributions were fitted to water resources, water environment and water ecology elements in historical monitoring data, and the optimal marginal distribution for each element was selected from multiple candidate distribution families based on the Akaike information criterion and goodness-of-fit test.

[0030] Based on the optimal marginal distribution, a probability integral transformation is performed on historical monitoring data to obtain uniformly distributed pseudo-observations.

[0031] Optionally, the health assessment values ​​of rivers and lakes under various scenarios can be obtained by solving a multi-objective optimization problem, including:

[0032] Random scenario data is input into the river and lake health assessment model. The objective function is to maximize the health scores of three sub-items: water resources, water environment, and water ecology. A multi-objective genetic algorithm is used to solve the problem in parallel.

[0033] During the solution process, individuals are selected by non-dominated sorting and crowding distance to obtain a Pareto optimal solution set that satisfies the preset algebra, and the solution closest to the ideal point in the solution set is determined as the sub-item health score under the corresponding scenario.

[0034] Optionally, submerged vegetation cover and emergent vegetation cover can be extracted from the aquatic ecosystem simulation submodule, including:

[0035] Simulated values ​​were obtained by numerical simulation using the growth dynamics equations of aquatic vegetation.

[0036] Normalized Difference Vegetation Index (NDVI) or Phytoplankton Index (PDI) are extracted from satellite remote sensing images to obtain remote sensing values;

[0037] The simulated values ​​were assimilated and corrected using remote sensing extracted values ​​to obtain the final vegetation cover field.

[0038] A multi-factor hydrological-ecological change coupling analysis system for lakes connected to rivers includes:

[0039] At least one processor; and

[0040] Memory that is communicatively connected to at least one processor;

[0041] The memory stores instructions that can be executed by a processor. When these instructions are executed by the processor, they implement the hydrological-ecological multi-element change coupling analysis method for lakes and rivers described in this invention.

[0042] Through the above technical solutions, this invention solves the problems of static indicator identification and unidirectional coupling process in traditional evaluation methods, thereby improving the accuracy and physical rationality of the evaluation results. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of the process for a multi-element hydrological-ecological coupling analysis method for lakes connected to rivers, provided in this application.

[0045] Figure 2 This is a schematic diagram of the process for constructing a river and lake health assessment model using the coupled water resources, water environment, and water ecology simulation sub-modules provided in this application.

[0046] Figure 3 This is a schematic diagram of the iterative feedback process provided in this application.

[0047] Figure 4 This is a schematic diagram of the process for extracting submerged vegetation coverage and emergent vegetation coverage from the aquatic ecosystem simulation submodule provided in this application.

[0048] Figure 5 This is a schematic diagram of the process for calculating the weights of indicators within the category provided in this application. Detailed Implementation

[0049] Example 1

[0050] This embodiment provides a method for coupled analysis of hydrological and ecological multi-factor changes in lakes connected to rivers, such as... Figure 1 As shown, the method includes the following steps:

[0051] Step 101: Obtain historical monitoring data on water resources, water environment, and water ecology elements of the lakes connected to the Yangtze River, as well as the water level difference sequence of the lakes.

[0052] Specifically, water resource elements can include indicators reflecting the spatiotemporal distribution characteristics of water volume, such as annual runoff, average monthly water level, and water level fluctuations of lakes connected to the Yangtze River. Water environment elements can include indicators reflecting the physicochemical properties of water bodies, such as annual average total phosphorus concentration, annual average total nitrogen concentration, permanganate index, ammonia nitrogen concentration, and dissolved oxygen concentration. Water ecological elements can include indicators reflecting the survival status of organisms and habitat quality, such as annual average chlorophyll a concentration, algal biomass, water temperature, and water quality transparency. Historical monitoring data can be obtained from representative hydrological stations, water quality monitoring stations, and meteorological stations in the study area. It is generally recommended to select long-term daily or monthly data spans of no less than 30 years to ensure the robustness of subsequent statistical analyses. The river-lake water level difference series specifically refers to the set of differences between the water levels of representative monitoring stations within the lakes connected to the Yangtze River and the water levels at the control sections of the navigable main stream. This series can characterize the complex dynamic backwater or inflow relationship between lakes and rivers.

[0053] Step 102: Perform sensitivity analysis on candidate indicators based on the river and lake water level difference sequence to construct a set of key river and lake health evaluation indicators.

[0054] Specifically, it can also involve identifying candidate indicators for river and lake health assessment based on historical monitoring data, performing sensitivity analysis on the candidate indicators using a pre-constructed river and lake health assessment model, and constructing a set of key river and lake health assessment indicators.

[0055] In this embodiment, a candidate indicator library composed of multi-dimensional elements is constructed using acquired historical monitoring data, and the physical conditions of lakes connected to the Yangtze River at different times are identified using the river-lake water level difference sequence. Sensitivity analysis is used to quantify the contribution of each candidate indicator to the system's health output under different physical conditions. Specifically, by analyzing the response degree of indicators under different conditions such as lacustrine, riverine, and transitional phases, indicators with weak impact on the evaluation results or high information redundancy under certain conditions are eliminated. The final set of key river and lake health evaluation indicators should have a reasonable hierarchical structure, clear physical meaning, and the ability to cover the core characteristics of water resources, water environment, and aquatic ecology. This step can effectively simplify the indicator system and reduce the dimensionality and complexity of subsequent coupled model calculations.

[0056] Step 103: Couple water resources, water environment and water ecology simulation sub-modules to construct a river and lake health assessment model. The water ecology simulation sub-module achieves iterative feedback with the water resources simulation sub-module through physical feedback variables.

[0057] In this embodiment, the river and lake health assessment model is built on a distributed hydrological model architecture based on hydrological response units. It establishes sub-modules for simulating runoff processes (water resources), pollutant transport and transformation (water environment), and aquatic vegetation growth and ecological responses (water ecology). To overcome the limitations of unidirectional driving forces in traditional models, this embodiment introduces physical feedback variables.

[0058] Specifically, the physical feedback variable can be implemented using the comprehensive Manning roughness coefficient. During the calculation process, the aquatic ecosystem simulation submodule dynamically updates the comprehensive Manning roughness coefficient of the calculation unit based on the simulated aquatic vegetation distribution status, and feeds the updated coefficient back to the water resources simulation submodule in real time. The water resources simulation submodule then corrects the relationship curve between water level and flow rate, thereby affecting the simulated boundary conditions of the aquatic environment and aquatic ecosystem. Through a closed-loop iterative feedback mechanism, the nonlinear interactions between hydrodynamics, the aquatic environment, and aquatic vegetation in connected lakes can be more realistically depicted.

[0059] Step 104: Construct a multidimensional joint distribution model based on historical monitoring data, and generate multiple sets of random scenario data based on the multidimensional joint distribution model.

[0060] Specifically, the multidimensional joint distribution model is used to characterize the complex relationships and dependencies among water resources, water environment, and water ecology elements. First, the optimal probability distribution of each individual element is determined through marginal distribution fitting. Based on this, a three-dimensional or multidimensional probabilistic model reflecting the asymmetric dependencies among multiple elements is constructed using connection functions. In this embodiment, a structured hierarchical dependency modeling approach is preferred, using water resources as the root node for conditional probability decomposition. Based on the constructed multidimensional joint distribution model, Monte Carlo sampling techniques are used to perform large-scale stochastic simulations, generating a massive set of stochastic scenario data reflecting potential future scenarios with multi-element coupling characteristics. This stochastic scenario data can provide rich uncertainty input samples for subsequent health assessments, covering a variety of typical states from normal conditions to extreme drought or flooding.

[0061] Step 105: Input the random scenario data and the key river and lake health evaluation index set into the river and lake health evaluation model, obtain the river and lake health evaluation value under each scenario by solving the multi-objective optimization problem, and determine the comprehensive weight of each index based on the river and lake health evaluation value.

[0062] In this embodiment, massive amounts of randomly generated scenario data are used as an external driver and input into the pre-constructed river and lake health assessment model. The multi-objective optimization problem uses the criterion that each key indicator should fall within a preset reasonable health threshold range to find the optimal solution under the coordinated development of various factors. Through parallel computation, the river and lake health assessment value corresponding to each set of random scenarios is obtained. This assessment value is typically a continuous value between 0 and 1, representing the quality of the lake's health status.

[0063] Based on this, and combining simulation results and evaluation outputs under various scenarios, a weighting method that integrates subjective and objective factors is adopted to determine the comprehensive weight of each key indicator in the evaluation system. In specific calculations, the comprehensive weights are used to weight and sum the evaluation results of each sub-item to obtain the final comprehensive health index.

[0064] Optionally, the determination of the comprehensive weights adopts a game-theoretic fusion approach combining subjective and objective weights. Specifically, with the goal of minimizing the deviation between the subjective and objective weight vectors, the optimal combination coefficients are determined by solving the Nash equilibrium condition, ensuring that the fused comprehensive weights simultaneously take into account expert experience and objective data information. This dynamic weight determination method enables the evaluation results to remain adaptive in changing environments.

[0065] According to one aspect of this application, a method for coupled analysis of hydrological and ecological multi-factor changes in lakes connected to rivers may further include the following steps:

[0066] To obtain historical monitoring data on water resources, water environment, and water ecology elements of lakes connected to the Yangtze River, as well as the water level difference sequence of the lakes and rivers;

[0067] Candidate indicators for river and lake health assessment were identified based on historical monitoring data.

[0068] A river and lake health assessment model is constructed by coupling water resources, water environment and water ecology simulation sub-modules. The water ecology simulation sub-module achieves iterative feedback with the water resources simulation sub-module through physical feedback variables.

[0069] Based on the river and lake water level difference sequence, sensitivity analysis was performed on candidate indicators using a river and lake health assessment model to construct a set of key river and lake health assessment indicators.

[0070] A multidimensional joint distribution model is constructed based on historical monitoring data, and multiple sets of random scenario data are generated based on the multidimensional joint distribution model.

[0071] Random scenario data and a set of key river and lake health evaluation indicators are input into the river and lake health evaluation model. The river and lake health evaluation values ​​under each scenario are obtained by solving a multi-objective optimization problem, and the comprehensive weight of each indicator is determined based on the river and lake health evaluation values.

[0072] Example 2

[0073] Based on the above Example 1, the process of indicator identification and situation partitioning screening is further described in detail.

[0074] In one possible implementation, before constructing a set of key river and lake health evaluation indicators by performing sensitivity analysis on candidate indicators based on the river and lake water level difference sequence, the method further includes: using meta-analysis to perform quantitative comprehensive analysis on multiple pre-set research data, identifying candidate indicators for river and lake health evaluation and constructing an initial indicator set (i.e., the initial set of key river and lake health evaluation indicators).

[0075] Specifically, meta-analysis refers to the systematic and quantitative comprehensive analysis of collected research data on multiple similar topics using statistical methods. In this embodiment, firstly, literature related to river and lake health assessment is retrieved from domestic and international academic databases according to a pre-set retrieval strategy, and inclusion and exclusion criteria are established. For example, inclusion criteria can be set as studies containing evaluation examples of lakes connected to rivers, and evaluation indicators with clear physical definitions that include water resources, water environment, or water ecology dimensions. The names, frequencies, and criterion layers of evaluation indicators used in each independent study are extracted, and the indicators are classified and summarized. Indicators with high frequency and general applicability are identified as candidate indicators, thereby constructing an initial indicator set.

[0076] The quantitative review approach described above can increase the sample size of indicators, resolve inconsistencies that may exist in the results of a single study, and improve the scientific rigor and credibility of the indicator system.

[0077] In one possible implementation, prior to performing sensitivity analysis on the candidate metrics, the method further includes:

[0078] Calculate the correlation matrix among the indicators in the initial indicator set, and use the correlation matrix as input for principal component analysis to perform eigenvalue decomposition;

[0079] Principal components are extracted based on the preset cumulative variance contribution rate threshold, and the original index with the largest absolute value of load in each principal component is selected as the representative index. After removing redundant indexes, a simplified index set (i.e., a simplified key river and lake health evaluation index set) is obtained.

[0080] In this embodiment, lateral redundancy is eliminated from the initial indicator set. First, the historical sequence data of each indicator is standardized, and the Pearson correlation coefficient between each pair of indicators is calculated to construct a correlation matrix. The correlation matrix characterizes the degree of linear correlation between indicators. The correlation matrix is ​​used as input for principal component analysis, and eigenvalues ​​and corresponding eigenvectors are obtained by solving the characteristic equation. The eigenvalues ​​are arranged in descending order, and the cumulative variance contribution rate of each principal component is calculated.

[0081] When the cumulative variance contribution rate reaches a preset threshold, such as 85%, the first p' principal components are retained. For each retained principal component, the principal component loading between it and each original index is calculated, and the index with the largest absolute value of the loading is selected as the representative index of that dimension. If there are multiple indices with insignificant differences in loading values, such as a difference of less than 0.05, the index with clearer physical meaning and easier data acquisition is retained first, and the remaining redundant indices with high collinearity are removed.

[0082] This step reduces the dimensionality of interrelated indicators into a set of independent and simplified indicators, preserving the original information to the maximum extent while minimizing indicator overlap.

[0083] In one optional embodiment, sensitivity analysis is performed on candidate indicators based on the river and lake water level difference sequence to construct a set of key river and lake health evaluation indicators, including the following process:

[0084] The river and lake water level difference sequence is compared with a preset threshold to divide the study period into different hydrological situation zones.

[0085] Within each situation zone, range standardization is applied to each indicator in the simplified indicator set.

[0086] The standardized index values ​​are used as the input perturbation of the river and lake health assessment model. Sensitivity sampling calculation is performed in the standardized index space to obtain the mean absolute basic effect and standard deviation of each index under different situation zones.

[0087] The mean absolute basic effect is summed with the standard deviation to obtain the comprehensive sensitivity score of each indicator in each situation zone.

[0088] Based on the proportion of each situation zone in the total time, the comprehensive sensitivity score is weighted, and after being sorted in descending order of weight, a preset number of indicators are extracted to construct a set of key river and lake health evaluation indicators.

[0089] In this embodiment, Morris (Morris Global Sensitivity Analysis) dynamic sensitivity screening based on situational partitioning is performed. Specifically, preset thresholds include a high threshold and a low threshold, determined based on the mean and standard deviation of multi-year river-lake water level difference sequences. A water level difference greater than the high threshold is defined as a lacustrine region, a water level difference less than a negative value of the low threshold is defined as a fluvial region, and a region in between is defined as a transitional situational region. Within each defined situational region, to eliminate the influence of dimensional differences between indicators, a range standardization formula is used for processing:

[0090] ξ * =(ξ-ξ min ) / (ξ max -ξ min );

[0091] Where, ξ * ξ represents the standardized index value; ξ represents the original index observation value; min This is the historical minimum value of the indicator within the corresponding situation zone; ξ max This represents the historical maximum value of the indicator within the corresponding situation zone.

[0092] In this embodiment, the sensitivity sampling calculation is performed based on a pre-calibrated river and lake health assessment model (see the description of offline model calibration in Embodiment 6). Specifically, the standardized values ​​of each indicator in the simplified indicator set are used as inputs to the river and lake health assessment model. The Morris sampling strategy is used to perturb each indicator value one by one, and the change in the health score output by the river and lake health assessment model is recorded. The basic effect is then calculated based on this change.

[0093] Within the standardized indicator space, a Morris global sensitivity analysis was performed using a one-to-one factor sampling strategy. The basic effects of each indicator were obtained by calculating the changes in the output of the river and lake health assessment model caused by small perturbations in the indicators. Furthermore, the mean absolute basic effect and standard deviation of each indicator within each situation zone were statistically analyzed. The mean absolute basic effect measures the average strength of the indicator's influence on the output, while the standard deviation measures the nonlinearity of the indicator and its interaction effects with other indicators. The sum of these two values ​​yielded the comprehensive sensitivity score.

[0094] S i，l =μ i，l * +σ i，l ;

[0095] Among them, S i，l The overall sensitivity score for the i-th indicator in the l-th situation zone; μ i，l * σ represents the mean absolute fundamental effect of the i-th indicator in the l-th situation zone; i，l Let be the basic effect standard deviation of the i-th indicator in the l-th situation zone.

[0096] Based on this, a full-situation weighted average is calculated by combining the time proportion of each situation zone:

[0097] S i =∑(p l *S i，l );

[0098] Among them, S i The full-situation weighted sensitivity of the indicator; p l ∑ represents the proportion of the l-th situation zone to the total duration of the historical records; ∑ represents the summation over all situation zones.

[0099] All indicators are categorized by S iThe indicators are sorted in descending order of their sensitivity scores, and the top q indicators or those with sensitivity scores greater than a preset sensitivity threshold are retained. In this embodiment, the q value can be determined based on the cumulative sensitivity contribution rate: the weighted sensitivity scores of each indicator after descending order are summed one by one, and when the sum reaches a preset proportion of the total score of all indicators, such as 90%, the corresponding number of indicators is q. Those skilled in the art can also determine a suitable q value by setting an absolute sensitivity threshold or by expert experience according to actual evaluation needs.

[0100] By using the above methods, it is possible to identify indicators that have a significant driving force on the health status of the system under different hydrological conditions, and the constructed key river and lake health evaluation indicator set has stronger dynamic response characteristics.

[0101] By using the difference in water levels between lakes and rivers to define hydrological situation zones, and independently performing range standardization sensitivity sampling within each zone, evaluation indicators can be adaptively identified as lacustrine-river facies alternate. This solves the problem of indicator identification being disconnected from physical conditions and improves the response accuracy to dynamic evolution processes.

[0102] Example 3

[0103] Based on the above embodiment 1, this embodiment further describes in detail the model construction and dynamic physical feedback process.

[0104] In one possible implementation, water resources, water environment, and aquatic ecosystem simulation sub-modules are coupled to construct a river and lake health assessment model, such as... Figure 2 As shown, it includes the following steps:

[0105] A water resources simulation submodule with runoff and water level simulation functions is constructed based on the hydrological response unit;

[0106] Construct a water environment simulation submodule that includes nutrient cycling and water quality parameter simulation functions;

[0107] Construct a water ecology simulation submodule that includes functions for simulating the growth dynamics of algae and vegetation.

[0108] Specifically, the water resources simulation submodule, the water environment simulation submodule, and the water ecology simulation submodule are natively located within the same model framework. Each module shares the spatial discretization scheme for hydrological response units. By dividing the watershed into multiple sub-watersheds and further subdividing each sub-watershed into hydrological response units based on land use, soil type, and slope combination, refined modeling of regions with different geographical characteristics is achieved.

[0109] The water resources simulation submodule calculates processes such as surface runoff, lateral flow, shallow groundwater recirculation, and deep seepage to quantitatively describe watershed runoff generation and confluence, as well as water level fluctuations in connected lakes and rivers. Key parameters include the number of runoff curves, baseflow decay coefficient, and soil evaporation compensation factor.

[0110] The water environment simulation submodule, based on runoff simulation, further simulates the migration process of nutrients such as nitrogen and phosphorus in the soil profile and their chemical kinetic transformation after entering the river channel. Key parameters include sediment oxygen demand rate, algal nitrogen content, maximum algal growth rate, nitrogen half-saturation constant, and phosphorus half-saturation constant.

[0111] The aquatic ecosystem simulation submodule utilizes algae and vegetation growth kinetic equations to dynamically simulate the biomass evolution of aquatic vegetation in lakes connected to rivers, based on environmental driving factors such as light, water temperature, nutrient concentration, and water depth. Specifically, aquatic vegetation can include submerged and emergent vegetation growing in the shallow waters of lakes.

[0112] In one optional embodiment, the physical feedback variable is the comprehensive Manning roughness coefficient, and the iterative feedback process is as follows: Figure 3 As shown, it includes:

[0113] Extract the submerged vegetation coverage and emergent vegetation coverage of shallow water areas of Tongjiang Lake from the water ecology simulation submodule;

[0114] The comprehensive Manning roughness coefficient is updated by adding the preset bare bed roughness, the product of the incremental contribution of submerged vegetation to the roughness and the submerged vegetation coverage, and the product of the incremental contribution of emergent vegetation to the roughness and the emergent vegetation coverage.

[0115] The updated integrated Manning roughness coefficient is injected into the water resources simulation submodule to correct the relationship curve between water level and flow rate, and is used as the input parameter for the next iteration step.

[0116] Specifically, the aquatic ecosystem simulation submodule calculates vegetation cover based on the simulated vegetation biomass. In this embodiment, a quantitative feedback logic is established between physical feedback variables and vegetation distribution status to calculate the comprehensive Manning roughness coefficient.

[0117] n comp =n bed +Δ n_SAV *φ SAV +Δ n_EAV *φ SAV ;

[0118] Where, n comp The updated composite Manning roughness coefficient; n bed The preset surface roughness; Δ n_SAV φ represents the incremental contribution of submerged vegetation per unit coverage to the roughness. SAV Submerged vegetation coverage; Δ n_EAV φ represents the incremental contribution of emergent vegetation to roughness per unit coverage. SAV To measure the coverage of emergent vegetation.

[0119] In this embodiment, n bed Set to 0.025, Δ n_SAV Set to 0.05, Δ n_EAV The calculated comprehensive Manning roughness coefficient is set to 0.08. This coefficient is then injected into the Manning formula of the water resources simulation submodule, directly correcting the water level and flow response relationship in each calculation unit. This process achieves physical feedback from the ecological module to the water resources module, overcoming the limitation of traditional methods where roughness, as a static parameter, is difficult to respond to seasonal vegetation growth.

[0120] One possible implementation, such as Figure 4 As shown, the method for extracting submerged vegetation cover and emergent vegetation cover from the aquatic ecosystem simulation submodule includes the following process:

[0121] Simulated values ​​were obtained by numerical simulation using the growth dynamics equations of aquatic vegetation.

[0122] Normalized Difference Vegetation Index (NDVI) or Phytoplankton Index (PDI) are extracted from satellite remote sensing images to obtain remote sensing values;

[0123] The simulated values ​​were assimilated and corrected using remote sensing extracted values ​​to obtain the final vegetation cover field.

[0124] In this embodiment, a multi-source data assimilation mechanism is introduced to improve the accuracy of the feedback. The aquatic vegetation growth dynamics equation is used to provide continuous simulated vegetation cover values ​​within each calculation step. Simultaneously, observational data acquired by satellite sensors are used to calculate the normalized vegetation index or phytoplankton index, thereby identifying the spatial distribution and cover intensity of vegetation within the lake.

[0125] For example, remote sensing values ​​are acquired using medium-to-high resolution satellite imagery on a monthly basis. The remote sensing values ​​at the observation time are then injected into the model using a data assimilation algorithm (an ensemble Kalman filter method is used in this embodiment) to correct errors in the simulated values ​​from the same period.

[0126] Specifically, the ensemble Kalman filtering method uses simulated values ​​as the forecast field and remotely sensed values ​​as the observation field, achieving optimal fusion of the forecast and observation fields by calculating the Kalman gain matrix. Those skilled in the art can also employ other data assimilation methods capable of achieving the same function.

[0127] During periods without remote sensing observation, the simulation process continues using the corrected model parameters to obtain a final vegetation cover field that is spatiotemporally continuous and has higher accuracy.

[0128] In an optional embodiment, updating the composite Manning roughness coefficient further includes:

[0129] Determine whether the calculated value obtained by addition exceeds the preset physical upper limit constraint;

[0130] If it exceeds the limit, the overall Manning roughness coefficient will be set to the value of the physical upper limit constraint to restrict the roughness within a reasonable physical range;

[0131] If the sum is not exceeded, the calculated value obtained by addition will be directly used as the comprehensive Manning roughness coefficient.

[0132] In some embodiments, when the vegetation cover in a lake is high, the shading effect between different vegetation communities causes the roughness to no longer increase linearly with vegetation cover. To ensure the stability of the hydrodynamic simulation, a physical upper limit constraint is set:

[0133] n comp =min(n bed +Δ n_SAV *φ SAV +Δ n_EAV *φ SAV n max );

[0134] Where, n max This is the preset physical upper limit for the overall roughness of the lake.

[0135] The specific value of the physical upper limit constraint can be determined based on the physical morphological characteristics of the lakebed and vegetation. In this embodiment, it is preferably set to 0.15. By using this constraint, abnormally high roughness values ​​can be avoided in densely vegetated areas, ensuring that the simulation results of physical quantities such as water level and flow velocity are always within the reasonable range for engineering applications.

[0136] Regarding model coupling, a quantitative feedback loop between vegetation cover and Manning roughness is established, upgrading the original unidirectional driven model into a mutually feeding system with physical feedback channels. This improvement corrects the logical defect in the traditional model that roughness does not change with habitat, ensures the physical rationality of the water level-discharge relationship correction, and uses the relative residual criterion to guarantee the numerical convergence of multi-factor iterative simulation.

[0137] Example 4

[0138] The construction of the multidimensional joint distribution model and the process of scenario generation are further explained in detail.

[0139] In one possible implementation, a multidimensional joint distribution model is constructed based on historical monitoring data, including the following steps:

[0140] Marginal distributions were fitted to water resources, water environment and water ecology elements in historical monitoring data, and the optimal marginal distribution for each element was selected from multiple candidate distribution families based on the Akaike information criterion and goodness-of-fit test.

[0141] Based on the optimal marginal distribution, a probability integral transformation is performed on historical monitoring data to obtain uniformly distributed pseudo-observations.

[0142] Specifically, historical monitoring data can include multi-year annual runoff sequences, monthly average water level sequences, annual average total phosphorus concentration sequences, and annual average chlorophyll a concentration sequences. The pre-defined candidate distribution families can include the generalized extreme value distribution (GEV), the log-normal distribution (Log-Normal), the Gamma distribution, the Weibull distribution, and the normal distribution.

[0143] In this embodiment, parameter estimation is performed on the historical time series of each element. The maximum likelihood estimation (MLE) method is used to fit the parameters of each candidate distribution family. After fitting, the Kolmogorov-Smirnov (KS) test is used to verify the goodness of fit of each distribution family. A significance level of 0.05 is set; when the p-value obtained from the test is greater than 0.05, the distribution family is considered to have passed the goodness of fit test.

[0144] For the distribution family that passes the test, the optimal marginal distribution is determined using the Akaike information criterion:

[0145] AIC = -2 * L(θ) + 2 * m;

[0146] Where AIC is the Akaike Information Criterion value; L(θ) is the maximum likelihood function value; and m is the number of parameters in the distribution model. The distribution family with the smallest AIC value is selected as the optimal marginal distribution for this element.

[0147] After determining the optimal marginal distribution of each element, a probability integral transformation is performed on the original historical monitoring data based on the cumulative distribution function. This maps the physical quantity sequence to the 0~1 interval, resulting in uniformly distributed pseudo-observations.

[0148] This step can eliminate the dimensional and distribution differences between different physical quantities, providing standardized data input for the subsequent construction of a joint distribution model.

[0149] One possible implementation involves constructing a multidimensional joint distribution model, specifically including any of the following implementation methods:

[0150] The first implementation method is to construct a C-vine Copula model with a multi-layered tree structure, decompose the multidimensional joint distribution into multiple binary function pairs, and independently match the optimal distribution family for each binary function pair based on the Akaike information criterion.

[0151] The second implementation method involves constructing a simple three-dimensional Copula model and selecting the distribution family with the smallest Akaike information criterion from the preset candidate three-dimensional distribution families as the multidimensional joint distribution model.

[0152] In a preferred embodiment, the C-vine Copula model is constructed using Implementation Method 1. Specifically, with water resource elements as the root node, the three-dimensional joint distribution is decomposed into a hierarchical structure of binary Copula pairs:

[0153] f(x1,x2,x3)=f1(x1)*f2(x2)*f3(x3)*c 12 (u1, u2; θ) 12 )*c 13 (u1, u3; θ) 13 )*c 23|1 (F 2|1 (u2|u1), F 3|1 (u3|u1); θ 23|1 );

[0154] Where f(x1, x2, x3) is the three-dimensional joint density function; f1(x1), f2(x2), and f3(x3) are the edge density functions of each element; c 12 c 13 and c 23|1 Here are the bivariate Copula density functions at different levels; u1, u2, and u3 are pseudo-observations; θ 12 θ 13 and θ 23|1 For the corresponding associated parameters; F 2|1 and F 3|1 This is the conditional distribution function.

[0155] The conditional distribution function is calculated using the following formula:

[0156] ;

[0157] Among them, F j|1 The value of j is the conditional distribution function, where j takes the value 2 or 3; Let be the partial derivative of the bivariate Copula function with respect to u1.

[0158] In this embodiment, for each level of binary Copula pair, independent selections are made from the Clayton, Gumbel, Frank, Normal, and Student-t distribution families, and the optimal distribution family for each level is selected according to the AIC criterion. This hierarchical decomposition method can flexibly characterize the complex asymmetric dependency structure and tail-related characteristics among water resources, environmental, and ecological elements.

[0159] In another implementation, a simple three-dimensional Copula model can be constructed using implementation method two. From a pre-defined candidate three-dimensional distribution family, the fitting parameters are estimated using maximum pseudo-likelihood, and the distribution family with the smallest AIC value is selected.

[0160] After constructing the multidimensional joint distribution model, Monte Carlo sampling techniques are used to generate massive amounts of random scenario data. Specifically, samples are extracted from the constructed joint distribution model using the conditional Rosenblatt inverse transform, with the sample size set to be greater than 10,000 groups. Through the optimal marginal distribution inverse transform of each element, the sampled probability values ​​are restored to random scenario data with physical meaning, forming the input scenario matrix for the subsequent river and lake health assessment model.

[0161] The joint modeling based on C-vine Copula further refined the asymmetric dependencies among multiple factors and captured the tail correlation characteristics under extreme scenarios. Experiments show that this integrated approach can effectively identify key drivers of water system health degradation and provide quantitatively reliable decision support for the coordinated development of regional socio-economic and ecological environments.

[0162] Example 5

[0163] Based on the above embodiment 1, the process of multi-objective optimization solution, two-layer weighted iterative architecture and automatic weighting promotion is further described in detail.

[0164] In one possible implementation, obtaining the river and lake health assessment values ​​under various scenarios by solving a multi-objective optimization problem includes the following steps:

[0165] Random scenario data is input into the river and lake health assessment model. The objective function is to maximize the health scores of three sub-items: water resources, water environment, and water ecology. A multi-objective genetic algorithm is used to solve the problem in parallel.

[0166] During the solution process, individuals are selected by non-dominated sorting and crowding distance to obtain a Pareto optimal solution set that satisfies the preset algebra. The solution in the solution set that is closest to the preset ideal point is determined as the sub-item health score under the corresponding scenario, where the ideal point is the coordinate point where the sub-item health score of each dimension takes the maximum value.

[0167] Specifically, the sub-scores for health include scores for water resources, water environment, and water ecology. In this embodiment, a dimensionless health scoring function is pre-constructed for the key indicators of each category, mapping the physical quantity indicators to a score range between 0 and 1.

[0168] The health scoring function is constructed based on the grade thresholds of various water quality indicators specified in the National Surface Water Environmental Quality Standards. Specifically, for negative indicators (the higher the value, the worse the health condition, such as total phosphorus concentration and total nitrogen concentration), the following piecewise linear scoring function is used:

[0169] When the indicator value is less than or equal to the Class I water standard limit, the score is 1;

[0170] When the indicator value is greater than or equal to the Class V water standard limit, the score is 0;

[0171] Scores are determined by linear interpolation between the limits for Class I and Class V.

[0172] For positive indicators (higher values ​​indicate better health, such as dissolved oxygen concentration and water transparency), a reverse piecewise linear scoring function is used:

[0173] When the indicator value is greater than or equal to the Class I water standard limit, the score is 1;

[0174] When the indicator value is less than or equal to the Class V water standard limit, the score is 0;

[0175] The value is determined by linear interpolation between the two limits.

[0176] For water resource indicators that are not applicable to water quality standard classification, such as annual runoff, the multi-year average is used as a reference, and a scoring function is constructed using the guarantee rate curve: when the indicator value reaches or exceeds the multi-year average, the score is 1, and when it is lower than the multi-year average, it decreases linearly according to the ratio to the average.

[0177] The decision variables for a multi-objective optimization problem are the control values ​​of each indicator, and the constraints include:

[0178] The values ​​of each key indicator must not exceed their physical feasible region, meaning each indicator value must be between the minimum and maximum values ​​recorded in history. The water level and flow rate output by the water resources simulation submodule must satisfy the water balance equation; the nutrient concentration output by the water environment simulation submodule must satisfy mass conservation; and the vegetation cover output by the water ecology simulation submodule must be between 0 and 1. These constraints are implemented by assigning a penalty term to individuals that do not meet the conditions during fitness evaluation.

[0179] A non-dominated sorting genetic algorithm II is used as the solver to generate the offspring population through simulated binary crossover and polynomial mutation. Those skilled in the art can determine the operating parameters, such as population size, crossover probability, and mutation probability, based on the dimensionality of the decision variables and the number of objective functions, using conventional hyperparameter tuning methods. In this embodiment, the population size is set to 200, the simulated binary crossover probability is set to 0.9, and the polynomial mutation probability is set to 1 / n, where n is the number of decision variables.

[0180] In each iteration, the dominance level of individuals is determined through non-dominated ranking, and the distribution diversity of the solution set is maintained using crowding distance. The calculation terminates when the number of iterations reaches 500, resulting in a Pareto optimal solution set composed of multiple non-dominated solutions. To determine the optimal sub-health score from the solution set, the Euclidean distance from each individual in the solution space to the ideal point (the coordinate point where the score of each dimension is 1) is calculated, and the solution with the smallest distance is selected as the evaluation result under this stochastic scenario.

[0181] One possible implementation involves determining the comprehensive weight of each indicator based on the river and lake health assessment value, and employing a two-layer weight calculation architecture to eliminate circular dependencies. This architecture uses a hierarchical iterative approach, combining inter-category weights and intra-category indicator weights, including:

[0182] The overall weight is divided into inter-category weight and intra-category indicator weight;

[0183] In the initial solution phase, the weights between categories are set to a preset proportional value, and the initial health evaluation value is calculated based on the sub-item health scores;

[0184] Based on the initial health assessment value and the key river and lake health assessment indicator values ​​output by the river and lake health assessment model for each random scenario, the weights of the indicators within each category are calculated, and the weights between categories are updated according to the degree of variation of the health scores of each category.

[0185] The health assessment value is recalculated based on the updated inter-category weights, and the outer iteration is performed until the assessment value between two consecutive iterations meets the preset relative residual criterion.

[0186] In this embodiment, the inter-category weights reflect the relative importance of water resources, environment, and ecology, while the intra-category indicator weights reflect the weight allocation of different key indicators within the same category. Since calculating the evaluation value requires weights, and the determination of objective weights depends on the distribution of evaluation values ​​and indicator values, this embodiment introduces a two-layer iterative architecture to eliminate circular dependencies. In the initial stage, the inter-category weights for water resources, water environment, and water ecology are all set to 1 / 3, and an initial health evaluation value is calculated accordingly. Subsequent weight calculation steps are then performed based on this evaluation value. After obtaining the intra-category indicator weights, the coefficient of variation is calculated based on the weighted sub-item scores for each category.

[0187] Specifically, the key indicator values ​​under each category are weighted and summed using the intra-category indicator weights to obtain the weighted sub-scores for each category under each scenario; the coefficient of variation of each category's weighted sub-score is calculated across all scenarios; the coefficients of variation for each category are normalized, and the normalized coefficients of variation are used as the updated inter-category weights. The logic is that categories with greater variability contribute more to distinguishing health status in different scenarios and should be assigned greater weights.

[0188] For example, the coefficient of variation (CV) of each category of health score is calculated under all random scenarios. c c=1, 2, 3 correspond to water resources, water environment, and water ecology, respectively. The normalized coefficients of variation for each category are used as the updated inter-category weights.

[0189] W c =CV c / Σ(CV c' );

[0190] Among them, W c This represents the updated weight of the c-th category, CV. c CV is the ratio of the standard deviation to the mean of the score for the c-th category. c' This represents the coefficient of variation for all categories involved in the normalization calculation.

[0191] Based on this, the health evaluation value is recalculated and the process proceeds to the next round of outer layer iteration.

[0192] Optionally, when performing outer iterations, the relative residual criterion is set to 10. -3 Furthermore, the outer iteration also has a maximum number of iterations, which is set to 100. The iteration stops when either the relative residual criterion or the maximum number of iterations is met.

[0193] After each iteration, convergence is evaluated by calculating the relative residuals:

[0194] ε k =max{1≤i≤N}(|H i，k -H i，k-1 | / max(H i，k-1 ,ε0));

[0195] Where, ε k H represents the relative residual of the k-th iteration; i，k H represents the health assessment value of the i-th scenario after the k-th iteration; i，k-1 Let be the health evaluation value of the i-th scenario after the (k-1)-th iteration; N be the total number of scenarios; ε0 be a preset minimum value used to prevent division by zero errors, which is set to 10 in this embodiment. -10 ;max is the maximum value function, max{1≤i≤N} represents maximizing the value over all scenarios. When ε k Less than the preset relative residual criterion (e.g., 10) -3 The iteration stops when the number of iterations reaches 100, and the final health evaluation result is output.

[0196] One possible implementation involves calculating the weights of indicators within each category, such as... Figure 5 As shown, the process includes the following:

[0197] Extract the key indicator value matrix corresponding to a large number of random scenarios, and use a clustering algorithm to divide the key indicator value matrix into a preset number of categories to determine the representative scenarios corresponding to each category.

[0198] Obtain the representative scenario weight vectors obtained by experts using the Analytic Hierarchy Process (AHP) for each representative scenario;

[0199] Calculate the Euclidean distance from each random scenario to each representative scenario, and determine the fuzzy membership degree of each random scenario relative to each representative scenario based on the Euclidean distance;

[0200] The subjective weight of each random scenario is obtained by weighting and summing the weight vectors of each representative scenario using fuzzy membership degrees. This subjective weight is then fused with the objective weights calculated by the entropy weight method based on the key indicator value matrix using game theory to obtain the in-category indicator weights corresponding to each random scenario.

[0201] Optionally, when determining the representative scenario and calculating the fuzzy membership degree, the following may also be included:

[0202] The optimal number of clusters for the clustering algorithm is determined by maximizing the silhouette coefficient in order to select representative scenarios;

[0203] When calculating the denominator of fuzzy membership, a pre-defined modified minimum term is introduced to avoid division by zero errors.

[0204] Due to the sheer number of random scenarios, it is difficult for experts to make subjective judgments on each one individually. Therefore, this embodiment employs a clustering-based weight generalization method. First, the standardized key indicator value matrix is ​​subjected to K-medoids clustering, and the optimal number of clusters is determined by maximizing the silhouette coefficient SC(k). Representative scenarios are selected as the central elements (medoid points) of each cluster to ensure they are genuine sample points for expert understanding. Experts construct a judgment matrix for a small number of representative scenarios, and the weight vector of the representative scenarios is obtained by solving the matrix. The fuzzy membership degree between each random scenario s and each representative scenario c is calculated.

[0205] λ sc =(1 / (d sc +ε) 2 ) / ∑(1 / (d sc' +ε) 2 );

[0206] Where, λ sc Let d be the fuzzy membership degree of scenario s to scenario c; sc Let be the Euclidean distance between scenario s and representative scenario c; ε is the preset modified minima, set to 10. -10 ; ∑ represents the summation over all categories c', d sc' This represents the distance between the random scenario s and all other representative scenarios c' traversed by the summation symbol ∑.

[0207] The calculated fuzzy membership degrees are used as weight coefficients to weight and sum the representative scenario weight vectors of experts, yielding the subjective weight vector corresponding to that scenario. Based on this, the objective weights of the scenario are calculated using the entropy weight method, and then fused using the Nash equilibrium formula from game theory to obtain the final in-category index weights.

[0208] This paper utilizes a game theory-based combinatorial weighting method to fuse subjective and objective weights. Specifically, let the subjective weight vector be w. sub The objective weight vector is w obj The fused weight vector w comb :

[0209] w comb =β1w sub +β2w obj ,

[0210] Where β1 and β2 are the combination coefficients to be determined. A multi-objective game model is constructed with the objective of minimizing the sum of squared deviations between the merged weights and each individual weight. Based on the Nash equilibrium condition, by differentiating each objective function and setting it to zero, a system of linear equations is obtained:

[0211] ;

[0212] Where · denotes the vector dot product operation, and the superscript T indicates transpose. Solving the above system of linear equations yields β1 and β2, which are then normalized:

[0213] β1 * =β1 / (β1+β2),

[0214] β2 * =β2 / (β1+β2).

[0215] The normalized combination coefficient β1 * and β2 * Substituting the values, we obtain the final in-category indicator weights w. comb =β1 * w sub +β2 * w obj .

[0216] One possible implementation involves using a multi-objective genetic algorithm for parallel solution, including:

[0217] The massive random scenario data is divided into multiple computation task subsets; the computation task subsets are distributed to multiple computing nodes in a pre-set computing cluster, and each computing node independently performs optimization and solution, and the results are aggregated to obtain the health evaluation results of rivers and lakes in the study area, that is, the sub-item health scores under each scenario.

[0218] Considering the computational demands of handling massive scenarios, this embodiment employs a task-level parallel computing strategy. A distributed computing framework is used to evenly distribute N sets of random scenarios across the nodes of the computing cluster. Each computing node has independent processor and memory resources, enabling parallel execution of the aforementioned NSGA-II optimization and weight calculation processes. After each node completes its calculation, the evaluation values ​​and weight combinations are aggregated and sent to the master node to obtain the river and lake health evaluation results for all scenarios. This improves the efficiency of processing massive data matrices and shortens the time cycle of coupled analysis.

[0219] A two-layer weighted iterative architecture is adopted, which eliminates the logical loop between health evaluation values ​​and indicator weights through equal weight initialization between categories and residual iteration. At the same time, typical samples are extracted using the K-medoids clustering algorithm, and the subjective opinions of experts are automatically extended to massive scenarios by combining the fuzzy membership formula, which shortens the manual weighting cycle that originally required months to computer processing in minutes.

[0220] Example 6

[0221] Based on the above embodiment 1, the offline determination method of hydrological situation zoning threshold, the calibration and verification process of model parameters, and the hardware implementation of the analysis system are further described in detail.

[0222] In one possible implementation, the preset threshold and the model parameters of the river and lake health assessment model are determined through the following offline construction method:

[0223] (1) Based on the historical observation data of the daily average lake water level and the main stream water level in the study area, the multi-year daily average and standard deviation of the daily average lake-river water level difference sequence are calculated. The sum of the multi-year daily average and standard deviation is determined as the high threshold of the lacustrine region, and the absolute value of the difference between the multi-year daily average and standard deviation is determined as the low threshold of the river region, which are used as the preset thresholds.

[0224] Specifically, it is recommended that the historical observation data be selected over a period of no less than 30 years to cover the complete dry and wet water cycles.

[0225] In this embodiment, the daily average water level sequence h of representative water level stations of the lake is first extracted. lake And the daily average water level sequence h of the corresponding control section of the main stream. river Calculate the difference in river and lake water levels Δh at each moment:

[0226] Δh=h lake -h river .

[0227] Based on this, the partition threshold is determined through statistical calculation:

[0228] Δh bar=(1 / T')*∑ t=1 T' (Δh t );

[0229] Where, Δh bar The value represents the multi-year daily average of the river and lake water level difference series, where T' is the total number of observation days, and Δh is the value of the river and lake water level difference series. t Let be the water level difference on day t.

[0230] σ Δh =sqrt((1 / (T'-1))*∑ t=1 T' (Δh t -Δh bar ) 2 );

[0231] Where, σ Δh Let be the standard deviation of the river / lake water level difference sequence, and sqrt represents the square root operation.

[0232] Based on the above statistical results, the preset threshold is determined as follows:

[0233] Δ high =Δh bar +σ Δh ;

[0234] Δ low =|Δh bar -σ Δh |;

[0235] Where, Δ high For the high threshold of the lacustrine region, Δ low This is the low threshold value for the riverine facies region.

[0236] The above calculations can adaptively determine the boundaries of situation zones by utilizing historical hydrological evolution patterns.

[0237] When the real-time water level difference is greater than Δ high At that time, it was determined that the current situation was in the lacustrine phase, at which point the lake water level was significantly affected by the backwater effect of the downstream main stream;

[0238] When the real-time water level difference is less than negative Δ low At this time, the current situation is determined to be in the river phase, at which point the lake exhibits strong outflow or backflow characteristics.

[0239] (2) Divide the historical measured data into calibration period and verification period according to time, and set a warm-up period before the calibration period; during the calibration period, generate multiple sets of parameter combinations in the parameter space through Latin hypercube sampling, and gradually narrow the parameter uncertainty interval according to the degree of matching between the simulation results and the observation data, until the calibrated model parameters in the river and lake health assessment model are obtained.

[0240] Specifically, historical measured data are typically divided into a calibration period and a validation period in a 7:3 ratio. To eliminate the interference of initial conditions on simulation accuracy, a warm-up period of no less than two years is set before the calibration period. In this embodiment, the parameter optimization is performed using the SUFI-2 (Sequential Uncertainty Fitting Algorithm) parameter uncertainty analysis algorithm.

[0241] Key sensitive parameters in the hydrology, water environment, and aquatic ecology sub-modules, such as runoff curve number CN2 and maximum algal growth rate μ, are targeted. max First, set the initial range of values. Then, use the Latin hypercube sampling method to generate multiple parameter combinations within the defined parameter space and execute simulations. After each simulation, calculate the objective function, such as the Nash efficiency coefficient (NSE), between the simulated and observed values.

[0242] NSE=1-[∑ t=1 T' (O t -S t ) 2 / ∑ t=1 T' (O t -O bar ) 2 ];

[0243] Where NSE is the Nash efficiency coefficient, O t S is the measured value at time t. t Let O be the simulated value at time t. bar This represents the average of the measured values.

[0244] Based on the feedback from the NSE (Negative Spectrum Error), the parameter uncertainty range is gradually narrowed. When the simulation result envelope can cover 95% of the observation data uncertainty band (95PPU), and the NSE during the validation period is >0.5, the parameter calibration is deemed qualified, and the calibrated model parameters are output, thereby ensuring that the coupled model can accurately reflect the physical laws of the study area.

[0245] According to one aspect of this application, a method for coupled analysis of hydrological and ecological multi-factor changes in lakes connected to rivers may further include the following steps:

[0246] Step S1: Collect literature, use meta-analysis to identify river and lake health evaluation indicators, construct an initial set of river and lake health evaluation indicators, calculate the sensitivity of river and lake health evaluation indicators, screen river and lake health evaluation indicators based on sensitivity, and construct a set of key river and lake health evaluation indicators.

[0247] Step S2: Using the SWAT model (distributed watershed hydrological model), construct water resources simulation model, water environment simulation model and water ecology simulation model respectively, and couple the water resources simulation model, water environment simulation model and water ecology simulation model to construct a river and lake health assessment model;

[0248] Step S3: Collect water resources, water environment and water ecology data of the study area, fit the distribution of water resources, water environment and water ecology respectively, construct a three-dimensional joint distribution, and obtain a large number of scenarios by random sampling.

[0249] Step S4: Input the set of key river and lake health evaluation indicators under massive scenarios into the river and lake health evaluation model, use the optimization algorithm to solve the model, obtain the river and lake health evaluation of the study area under massive scenarios, and calculate the weight combination corresponding to the massive scenarios based on the river and lake health evaluation of the study area under massive scenarios and the key river and lake health evaluation indicator values.

[0250] In this embodiment, generalized indicators representing water resources, water environment, and water ecology are introduced. From three levels—structural framework design, selection of underlying indicators, and simplification of the indicator system—an indicator system for river and lake health assessment with a reasonable hierarchical structure, clear physical meaning, and an appropriate number of indicators is established. A weighting method combining subjective and objective approaches is proposed, considering the primary and secondary relationships of indicators at different stages of environmental change. Health regulation belongs to a complex collaborative regulation problem of a coupled and mutually reinforcing system of water resources, water environment, and water ecology. Furthermore, the regulation targets will also dynamically change with changes in the external environment. Based on the constructed indicator system, the boundaries, dimensions, constraints, and evolutionary characteristics of the regulation targets of different subjects at different stages are clarified, along with their evolutionary characteristics at different spatiotemporal scales. A mathematical description method for collaborative regulation targets is established based on multiple characterization factors.

[0251] According to another aspect of this application, a multi-factor hydrological-ecological change coupling analysis system for lakes connected to rivers is provided, comprising:

[0252] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the processor, which, when executed by the processor, implement the hydrological-ecological multi-element change coupling analysis method for lakes and rivers as described in the embodiments of the present invention.

[0253] In this embodiment, the analysis system can be implemented based on a high-performance computing cluster or workstation. The processor can be a central processing unit (CPU) or a graphics processing unit (GPU). The memory includes random access memory (RAM) and non-volatile memory, used to load monitoring data, store massive scenario matrices, and cache intermediate variables during model iteration.

[0254] The processor exchanges data with memory via a bus. When executing instructions, the processor first reads historical monitoring data and performs edge distribution fitting; the generated random scenario data is stored in a predetermined data buffer. During the multi-objective optimization solution phase, the processor schedules multiple cores to calculate the health score for each scenario in parallel. After each iteration, the processor calculates the relative residual and determines whether the convergence condition is met. If met, the final health assessment value combined with weights is output to the user interface or storage device.

[0255] This hardware architecture ensures the efficient and stable execution of the multi-factor coupled analysis process. After the processor completes its execution, it outputs the final river and lake health evaluation value, the comprehensive weight of each indicator, and the sub-scores under each scenario to the user interface for visualization, or stores them in the database for use by the decision support system.

[0256] For example, the method of the present invention was applied to a typical lake-connected area for verification. The results show that, compared with the traditional evaluation method using static indicators throughout the entire time period, the situational zoning dynamic indicator screening scheme of the present invention can identify key driving indicators with significantly different responses during the lacustrine and fluvial phases. After introducing the vegetation-roughness feedback mechanism, the Nash efficiency coefficient of water level simulation in the shallow water area of ​​the lake is improved compared with the non-feedback scheme. The two-layer weighted iterative architecture can stably converge within a preset number of iterations when handling random scenarios on a scale of tens of thousands. The above verification results show that the method of the present invention has significant improvements in the dynamic adaptability of evaluation indicators, the accuracy of physical process characterization, and the automated processing of large-scale calculations. Those skilled in the art will understand that the specific improvement may vary depending on the hydrological characteristics of the study area, data quality, and parameter configuration.

[0257] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for coupled analysis of hydrological and ecological multi-factor changes in lakes connected to rivers, characterized in that, include: To obtain historical monitoring data on water resources, water environment, and water ecology elements of lakes connected to the Yangtze River, as well as the water level difference sequence of the lakes and rivers; Sensitivity analysis was performed on candidate indicators based on the river and lake water level difference sequence to construct a set of key river and lake health evaluation indicators; A river and lake health assessment model is constructed by coupling water resources, water environment and water ecology simulation sub-modules. The water ecology simulation sub-module achieves iterative feedback with the water resources simulation sub-module through physical feedback variables. A multidimensional joint distribution model is constructed based on historical monitoring data, and multiple sets of random scenario data are generated accordingly. Random scenario data and a set of key river and lake health evaluation indicators are input into the river and lake health evaluation model. The river and lake health evaluation values ​​under each scenario are obtained by solving a multi-objective optimization problem, and the comprehensive weight of each indicator is determined accordingly.

2. The method according to claim 1, characterized in that, Before constructing a set of key river and lake health assessment indicators, the following steps are also included: Sensitivity analysis of candidate indicators based on river and lake water level difference sequences is performed. Meta-analysis was used to quantitatively synthesize multiple pre-set research data to identify candidate indicators for river and lake health assessment and to construct an initial indicator set.

3. The method according to claim 2, characterized in that, Before performing sensitivity analysis on candidate metrics, the following steps are also included: Calculate the correlation matrix among the indicators in the initial indicator set and use it as input for principal component analysis to perform eigenvalue decomposition; Principal components are extracted based on the preset cumulative variance contribution rate threshold, and the original index with the largest absolute value of load in each principal component is selected as the representative index. After removing redundant indexes, a simplified index set is obtained.

4. The method according to claim 1, characterized in that, Coupled with water resources, water environment, and water ecology simulation sub-modules, a river and lake health assessment model is constructed, including: A water resources simulation submodule with runoff and water level simulation functions is constructed based on the hydrological response unit; Construct a water environment simulation submodule that includes nutrient cycling and water quality parameter simulation functions; Construct a water ecology simulation submodule that includes functions for simulating the growth dynamics of algae and vegetation.

5. The method according to claim 1, characterized in that, The physical feedback variable is the comprehensive Manning roughness coefficient, and the iterative feedback includes: Extract the submerged vegetation coverage and emergent vegetation coverage of shallow water areas of Tongjiang Lake from the water ecology simulation submodule; The comprehensive Manning roughness coefficient is updated by adding the preset bare bed roughness, the product of the incremental contribution of submerged vegetation to the roughness and the submerged vegetation coverage, and the product of the incremental contribution of emergent vegetation to the roughness and the emergent vegetation coverage. The updated integrated Manning roughness coefficient is injected into the water resources simulation submodule to correct the relationship between water level and flow rate, and is used as the input parameter for the next iteration step.

6. The method according to claim 5, characterized in that, The update of the overall Manning roughness coefficient also includes: Determine whether the calculated value obtained by addition exceeds the preset physical upper limit constraint; If the value is exceeded, the overall Manning roughness coefficient will be set to the value of the physical upper limit constraint to restrict the roughness within a reasonable physical range.

7. The method according to claim 1, characterized in that, A multidimensional joint distribution model was constructed based on historical monitoring data, including: Marginal distributions were fitted to water resources, water environment and water ecology elements in historical monitoring data, and the optimal marginal distribution for each element was selected from multiple candidate distribution families based on the Akaike information criterion and goodness-of-fit test. Based on the optimal marginal distribution, a probability integral transformation is performed on historical monitoring data to obtain uniformly distributed pseudo-observations.

8. The method according to claim 1, characterized in that, The health assessment values ​​of rivers and lakes under various scenarios are obtained by solving a multi-objective optimization problem, including: Random scenario data is input into the river and lake health assessment model. The objective function is to maximize the health scores of three sub-items: water resources, water environment, and water ecology. A multi-objective genetic algorithm is used to solve the problem in parallel. During the solution process, individuals are selected by non-dominated sorting and crowding distance to obtain a Pareto optimal solution set that satisfies the preset algebra, and the solution closest to the ideal point in the solution set is determined as the sub-item health score under the corresponding scenario.

9. The method according to claim 5, characterized in that, Submerged vegetation cover and emergent vegetation cover were extracted from the aquatic ecosystem simulation submodule, including: Simulated values ​​were obtained by numerical simulation using the growth dynamics equations of aquatic vegetation. Normalized Difference Vegetation Index (NDVI) or Phytoplankton Index (PDI) are extracted from satellite remote sensing images to obtain remote sensing values; The simulated values ​​were assimilated and corrected using remote sensing extracted values ​​to obtain the final vegetation cover field.

10. A multi-element hydrological-ecological change coupling analysis system for lakes connected to rivers, characterized in that, include: At least one processor; as well as Memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by a processor, which, when executed by the processor, implement the hydrological-ecological multi-element change coupling analysis method for lakes and rivers as described in any one of claims 1 to 9.

Images