Method for evaluating fish habitat and ecological flow regulation and control based on data-driven fuzzy model

Abstract

The invention discloses a method for evaluating fish habitat and ecological flow regulation and control based on a data-driven fuzzy model, belongs to the field of water conservancy and hydropower engineering environmental protection and ecological hydraulics, and particularly relates to a method for evaluating fish habitat and ecological flow regulation and control. According to the method, a hydraulic parameter membership function is automatically deduced from on-site fish behavior observation data through a fuzzy C-means clustering (FCM) algorithm, a fuzzy rule base is constructed, a habitat suitability index (HSI) model under the synergistic effect of multiple factors such as water depth and flow velocity is established, and an environment flow threshold value based on weighted available area (WUA) maximization is output. According to the method, the defects that a traditional habitat model excessively depends on expert experience and multi-factor interaction processing is insufficient are overcome, objective quantification of model parameters and accurate capture of a complex nonlinear relation are achieved, quantifiable and species-specific decision support is provided for cascade reservoir ecological scheduling, and the method has the advantages of being high in practicability and easy to popularize. And hydroelectric benefits and ecological protection requirements are effectively balanced.

Description

Technical Field

[0001] This invention belongs to the field of environmental protection and ecohydraulics in water conservancy and hydropower engineering, and specifically relates to a method for evaluating fish habitats and regulating ecological flow. Background Technology

[0002] River habitats are the fundamental carriers for aquatic organisms to complete their life cycles, and their hydrological conditions, such as flow pulses and water temperature cycles, directly drive key behaviors of fish, such as spawning, foraging, and migration. The intensive development of cascade hydropower stations alters natural flow patterns, disrupts river connectivity, and slows water flow, leading to sedimentation, dissolved oxygen stratification, and water temperature homogenization, severely compressing fish spawning grounds. Therefore, quantifying the impact of hydropower development on fish habitats and the environmental flow threshold during the breeding season is a core scientific issue for mitigating the cumulative ecological effects of hydropower development.

[0003] Habitat Suitability Index (HSI) models are the mainstream tool for fish habitat assessment, and their accuracy depends on the quantitative mapping between habitat factors and fish preferences. However, traditional habitat suitability curves are not well-suited for handling multi-factor interactions, and the models neglect the synergistic effects of various environmental parameters, which may lead to systematic biases in habitat suitability assessment results. Secondly, they lack sensitivity to data uncertainty; curves fitted under sparse sample conditions are easily disturbed and struggle to accurately capture the potential nonlinear characteristics of habitat preferences.

[0004] To overcome the aforementioned limitations, fuzzy logic methods have been widely applied to HSI modeling in recent years, significantly improving the objectivity of statistical interpretation of experimental data. Although fuzzy logic-based models can effectively handle the continuity and fuzziness of habitat factors, their core components (boundary setting of membership functions and rule base construction) heavily rely on expert experience, thus posing a risk of introducing subjective bias. Therefore, optimizing the parameter construction mechanism of fuzzy logic models (such as reducing reliance on subjective experience) has become a new direction in habitat modeling research, aiming to achieve more accurate assessments of habitat quality and environmental traffic.

[0005] In recent years, machine learning methods have made significant progress in the field of aquatic ecology. Machine learning can identify complex nonlinear data relationships and accurately predict data trends. However, habitat and biological response models directly built based on machine learning algorithms are essentially black-box mathematical models, lacking the ability to truly explain the various physical mechanisms in aquatic ecosystems. Therefore, there is an urgent need to develop a method that integrates data-driven approaches and mechanistic analysis to achieve accurate and interpretable habitat assessment and flow management. Summary of the Invention

[0006] This invention proposes a data-driven fuzzy model-based method for evaluating fish habitats and ecological flow regulation. This method automatically derives the membership function of habitat parameters by coupling fuzzy logic and machine learning optimization algorithms, constructs a multi-factor synergistic HSI model, and outputs an environmental flow threshold based on maximizing habitat suitability.

[0007] A method for evaluating fish habitats based on a data-driven fuzzy model is characterized by the following steps:

[0008] S1, collect on-site hydraulic parameters such as flow velocity and water depth during the breeding season of the target fish, as well as the distribution data of the frequency of occurrence of the target fish.

[0009] S2, based on the distribution data of target fish under different water depths and flow velocities, fit the Gaussian membership function of flow velocity and water depth using the fuzzy C-means clustering (FCM) algorithm;

[0010] S3. Construct a fuzzy rule base, divide the hydraulic parameters flow velocity and water depth into multi-level linguistic variables, and use triangular and trapezoidal membership functions to describe the comprehensive suitability index (HSI).

[0011] S4. Based on the membership function, the degree of satisfaction of the parameters with the fuzzy rules is quantified. The weighted average method is used to output the fuzzy set of habitat suitability. The centroid method is used to defuzzify the membership function and obtain the 0-1 standardized comprehensive suitability index HSI value, which is used to evaluate fish habitats.

[0012] The data-driven fuzzy model-based ecological flow regulation method is characterized by:

[0013] S1, Based on the HSI value obtained in step S4 of claim 1, the weighted usable area (WUA) is calculated using the grid accumulation method, and the functional relationship between WUA and flow rate is established as follows:

[0014] ;

[0015] Where: HSIi is the comprehensive suitability index of the i-th grid, Ai is the area of ​​the i-th grid, n is the number of environmental factors affecting habitat suitability considered in the simulation, and Q is the flow rate;

[0016] S2. Based on the WUA maximization principle, determine the environmental flow threshold during the target fish's breeding season, and generate ecological scheduling suggestions accordingly to carry out ecological flow regulation.

[0017] A system for implementing the method of claim 1, characterized in that the system comprises the following modules:

[0018] The data acquisition module is used to acquire on-site hydraulic parameters and fish behavior data;

[0019] The data processing module is used to run the FCM algorithm to optimize the membership function and build a fuzzy rule base, and to run the weighted average method and centroid method to calculate the Habitat Suitability Index (HSI).

[0020] A system for implementing the method of claim 2, characterized in that the system comprises the following modules:

[0021] The data acquisition module is used to acquire on-site hydraulic parameters and fish behavior data;

[0022] The data processing module is used to run the FCM algorithm to optimize the membership function and build a fuzzy rule base, and to run the weighted average method and centroid method to calculate the Habitat Suitability Index (HSI).

[0023] The model calculation module is used to perform calculations for HSI and WUA;

[0024] The flow control module is used to output environmental flow thresholds and generate ecological scheduling suggestions.

[0025] In step S2, the fuzzy C-means clustering (FCM) algorithm minimizes the objective function. The method optimizes parameters, where c represents the number of clusters, n represents the number of samples, m>1 represents the fuzzy weight index (usually set to m=2), uij represents the membership strength of the j-th sample xj relative to the i-th cluster center vi, and ||xj-vi||2 represents the Euclidean distance between sample xj and cluster center vi. This method replaces the traditional expert-based parameter setting method.

[0026] Step S3 constructs a fuzzy rule base, classifying flow velocity and water depth into three levels of linguistic variables: low, medium, and high. Through five levels of fuzzy classification—specifically, extremely low, low, medium, high, and extremely high—the comprehensive fuzzy rule base HSI value is output. The fuzzy rule base takes the form: if the flow velocity is A and the water depth is B, then the suitability is C.

[0027] The flow velocity and water depth under different flow conditions are simulated by a hydrodynamic model, and the simulation results are input into the fuzzy logic model to predict the spatial variation of HSI and WUA.

[0028] This invention combines fuzzy logic with the fuzzy C-means clustering (FCM) algorithm in machine learning, overcoming key limitations of traditional habitat models. Traditional Habitat Suitability Index (HSI) models heavily rely on expert experience to define thresholds for suitability curves, a process that often introduces subjective bias and struggles to handle the continuity and interactions between environmental factors. This invention, through the FCM algorithm, directly derives the membership functions of flow velocity and water depth from field observation data, achieving objectivity and quantification of model parameters. The membership functions identified by machine learning smoothly reflect the gradual changes in fish preferences, avoiding unnatural phenomena such as abrupt changes in suitability caused by rigid threshold divisions in traditional methods.

[0029] Compared to purely "black box" machine learning models (such as neural networks), the fuzzy model employed in this invention improves prediction accuracy while retaining good interpretability. Researchers can understand the model's decision-making logic by examining the fuzzy rule base and membership functions, for example: "When the flow rate is 'medium' and the water depth is 'high,' the habitat suitability is 'high.'" This characteristic is crucial for ecological management because it provides a physical explanation different from purely data-driven models, enabling managers to understand and trust the model's recommendations, thereby developing more targeted ecological management strategies.

[0030] Instruction manual illustrations

[0031] Figure 1 This is a statistical chart showing the frequency of target fish species appearing in different flow velocity ranges in Example 1.

[0032] Figure 2 This is a statistical chart showing the frequency of the target fish species appearing in different water depth ranges in Example 1.

[0033] Figure 3 The flow velocity membership function diagram is shown in Example 1.

[0034] Figure 4 This is a membership function diagram of water depth in Example 1.

[0035] Figure 5 This is a chart showing the overall suitability HSI value for Example 1.

[0036] Figure 6 This is a graph showing the results of the comprehensive habitat suitability index in Example 1.

[0037] Figure 7 This is a comparison chart of actual fish fry density and WUA (Whole Fish Scale). Detailed Implementation

[0038] Example 1: The controlled catchment area at the dam site of a certain reservoir is 235,400 km². 2 The average annual flow rate is 1590 m³ / h. 3 / s. The reservoir's tail end has a spawning ground for round-mouthed copper fish. This section of the river has wide, winding banks and complex hydrological conditions, making it suitable for fish habitats.

[0039] The specific steps for assessing fish habitat and regulating ecological flow in this reservoir are as follows:

[0040] S1, Data Acquisition:

[0041] The data collection period was June-July 2022, the peak breeding season, and included the flow velocity and water depth of the control flow at the reservoir dam site. The flow velocity ranged from 0 to 1.79 m / s, and the water depth ranged from 0 to 22.42 m.

[0042] The target fish is the round-mouthed bronze fish;

[0043] The frequency distribution data of the target fish, the round-mouthed copper fish, at different flow velocities and water depths is shown in the figure below. Figure 1 and Figure 2 .

[0044] S2, Data Processing: Based on the distribution data of fish occurrence frequency under different water depths and flow velocities collected on-site, the fuzzy C-means clustering (FCM) algorithm is used to obtain the Gaussian membership function of flow velocity and water depth; the Gaussian membership function is used to describe the fitness of flow velocity and water depth.

[0045] Based on 129 sets of data collected on-site, the FCM algorithm was used to obtain the membership functions of flow velocity and water depth, as follows: Figure 3 and Figure 4 As can be seen from the figure, the spawning current velocity range of the round-mouthed copper fish is 0.2-1.7m / s, with the optimal range being 0.6-1.2m / s; the spawning water depth range is 1-11m, with the optimal range being 4-8m.

[0046] Fuzzy C-means clustering algorithm is used for automatic identification of fuzzy membership function. The FCM parameters are selected as shown in Table 2 below:

[0047] Table 2 Cluster number c Initial cluster centers Fuzziness index m Maximum number of iterations Termination threshold Convergence threshold η 5 （0.5，0.5） 2 500 <![CDATA[1×10 −5 ]]> 0.001

[0048] S3, Fuzzy Rule Base Construction: Flow velocity and water depth are divided into three levels of linguistic variables: low, medium, and high. Triangular and trapezoidal membership functions are used, described by five fuzzy subsets: range (VL), difference (L), medium (M), good (H), and very good (VH). Based on the subsets of each fuzzy membership function, fuzzy logic reasoning rules are determined according to research by fish and ecological experts and the indication experience of local fishermen for the spawning season of the round-mouthed copper fish. Based on existing research on the spawning habitat of the round-mouthed copper fish, the fuzzy rule base shown in Table 1 is finally formulated, containing 25 fuzzy rules. For example, the rule description is: if the flow velocity is medium and the water depth is high, then the suitability is high.

[0049] Table 1

[0050] S4 integrates the membership functions of flow velocity and water depth with the fuzzy rule base, and outputs the integrated multi-rule fuzzy rule base using the weighted average method. The membership function is then defuzzified using the centroid method to obtain the integrated suitability HSI value.

[0051] S5, the hydraulic parameter distribution of flow rates from 1190 to 3870 m³ / s in June and July was simulated using the River2D hydrodynamic model; the comprehensive habitat suitability index (HSI) was then imported into the hydrodynamic model, and the final results of the comprehensive habitat suitability index for the reservoir were obtained as follows: Figure 6 As shown.

[0052] S6, calculate the HSI value of each grid and sum them to obtain WUA.

[0053] The results showed that WUA was significantly correlated with the measured fish fry density (r=0.707, p<0.01), verifying the reliability of the model in this invention. Figure 7 As shown.

[0054] This embodiment uses Habitat Suitability Index (HSI) simulation and spatial analysis to reveal the changing characteristics of the spawning habitat of *Cyprinus circinus* under different hydrological conditions. Based on the flow velocity and water depth data collected in June and July 2022, the flow velocity ranged from 0 to 1.15 m / s in June, with an average of 0.44 m / s; the flow velocity increased significantly in July, ranging from 0 to 1.79 m / s, with an average of 0.68 m / s. This change indicates that the hydrological conditions in July are more dynamic, possibly influenced by rainfall events or upstream reservoir operations. The increase in flow velocity in July brings it closer to the optimal flow velocity range for *Cyprinus circinus* spawning (0.6-1.2 m / s), suggesting that the hydraulic conditions during this period are more conducive to stimulating fish spawning behavior. Spatially, the suitable flow velocity in June is mainly distributed in the inlet and outlet sections of the studied river segment, while the flow velocity in most areas in July is suitable for *Cyprinus circinus* spawning.

[0055] The water depth ranged from 0 to 20.71 m in June and from 0 to 22.42 m in July. The increase in water depth is related to the arrival of the flood season and the increase in runoff in the basin. Spatially, the water depth in most areas of the studied river section exceeded the suitable water depth range for the round-mouthed copper fish, with only the shallow areas in the north and south of the studied river section showing relatively high water depth suitability.

[0056] The Habitat Suitability Index (HSI) also showed a significant difference between June and July. The HSI value ranged from 0.25 in June to 0.31 in July, indicating an overall improvement in habitat quality. This improvement is mainly related to the optimization of flow velocity parameters, especially in July. The increased outflow brought flow velocity conditions in more areas into the preferred range of *Agropyron cristatum*, resulting in a higher proportion of river areas being assessed as "moderate" to "highly suitable." Overall, conditions in July were closer to the ecological optimum range for *Agropyron cristatum*, and its overall habitat suitability was higher than in June. This result highlights the importance of hydraulic parameters in assessing breeding habitats and also demonstrates that the data-driven fuzzy model constructed in this study can effectively capture the positive impact of environmental factor changes on habitat quality.

Claims

1. A method for evaluating fish habitats based on a data-driven fuzzy model, characterized in that, Includes the following steps: S1, collect on-site hydraulic parameters such as flow velocity and water depth during the breeding season of the target fish, as well as the distribution data of the frequency of occurrence of the target fish. S2, based on the distribution data of the target fish under different water depths and flow velocities, fit the Gaussian membership function of the flow velocity and water depth using the fuzzy C-means clustering algorithm; S3. Construct a fuzzy rule base, divide the hydraulic parameters flow velocity and water depth into multi-level linguistic variables, and use triangular and trapezoidal membership functions to describe the comprehensive suitability index; S4. Based on the membership function, the degree of satisfaction of the parameters with the fuzzy rules is quantified. The weighted average method is used to output the fuzzy set of habitat suitability. The centroid method is used to defuzzify the membership function and obtain the 0-1 standardized comprehensive suitability index HSI value, which is used to evaluate fish habitats.

2. A data-driven fuzzy model-based ecological flow regulation method, characterized by: S1, Based on the HSI value obtained in step S4 of claim 1, the weighted usable area (WUA) is calculated using the grid accumulation method, and the functional relationship between WUA and flow rate is established as follows: ； Where: HSIi is the comprehensive suitability index of the i-th grid, Ai is the area of ​​the i-th grid, n is the number of environmental factors affecting habitat suitability considered in the simulation, and Q is the flow rate; S2. Based on the WUA maximization principle, determine the environmental flow threshold during the target fish's breeding season, and generate ecological scheduling suggestions accordingly to carry out ecological flow regulation.

3. A system for implementing the method of claim 1, characterized in that, The system includes the following modules: The data acquisition module is used to acquire on-site hydraulic parameters and fish behavior data; The data processing module is used to run the FCM algorithm to optimize the membership function and build a fuzzy rule base, and to run the weighted average method and centroid method to calculate the habitat suitability index.

4. A system for implementing the method of claim 2, characterized in that, The system includes the following modules: The data acquisition module is used to acquire on-site hydraulic parameters and fish behavior data; The data processing module is used to run the FCM algorithm to optimize the membership function and build a fuzzy rule base, and to run the weighted average method and centroid method to calculate the habitat suitability index. The model calculation module is used to perform calculations for HSI and WUA; The flow control module is used to output environmental flow thresholds and generate ecological scheduling suggestions.

Images