A method for predicting the breeding behavior of Chinese sturgeon

Abstract

The application discloses a Chinese sturgeon breeding behavior prediction method, which comprises the following steps: step 1, obtaining and uniformly formatting Chinese sturgeon breeding field monitoring data; step 2, calculating Chinese sturgeon breeding behavior evaluation indexes; step 3, constructing a Chinese sturgeon breeding behavior evaluation model coupled with multiple factors; step 4, identifying key factors and threshold intervals affecting Chinese sturgeon breeding behavior; and step 5, predicting the evolution law of Chinese sturgeon breeding activities under the comprehensive action of multiple factors and putting forward breeding field operation scheduling and protection and restoration strategies. The application solves the problems of the existing Chinese sturgeon breeding monitoring and evaluation means, such as difficulty in unifying multiple source data, difficulty in quantitatively evaluating breeding behavior, difficulty in identifying key factor thresholds, and insufficient prediction and decision support.

Description

Technical Field

[0001] This invention relates to the field of aquatic biodiversity conservation and behavior monitoring technology, and in particular to a method for predicting the reproductive behavior of the Chinese sturgeon. Background Technology

[0002] Chinese sturgeon ( Acipenser sinensis As a typical flagship and indicator species in the Yangtze River Basin, the Chinese sturgeon's reproductive activities are highly sensitive to river connectivity, hydrodynamic patterns, and key habitat conditions. The natural reproductive process of the Chinese sturgeon has been significantly constrained by multiple factors, including habitat fragmentation caused by the construction of water conservancy projects, shrinkage of natural spawning grounds, and alterations in hydrodynamic processes. To maintain and improve the reproductive effectiveness of the Chinese sturgeon, artificial natural breeding grounds have been constructed and managed in key river sections in recent years. These grounds create relatively suitable hydrodynamic and microhabitat conditions to promote reproductive behaviors such as parent fish aggregation, companionship, and spawning, and based on this, protection and restoration measures are implemented. However, whether the operation of breeding grounds generates an effective reproductive behavioral response, how the time period and spatial range of key behaviors change, and how reproductive effectiveness evolves under different inflow and disturbance conditions still urgently require support from scientific, continuous, and quantifiable monitoring and assessment methods.

[0003] Current monitoring of Chinese sturgeon reproduction relies heavily on egg collection, manual patrols, shore-based observations, and localized acoustic and video aids. However, this approach suffers from limited coverage, insufficient continuity, and low quantification. Firstly, key breeding grounds often experience significant variations in water depth, strong turbulence, and high turbidity, making traditional optical observations difficult. Single methods struggle to simultaneously and continuously record multiple areas, including deep water zones, channel sections, and areas near nets, resulting in incomplete capture of crucial processes such as parent fish gathering, residence, and companionship. Secondly, existing assessments often remain at the level of phenomenon description or single-factor correlation, lacking methods to unify the timing and spatial benchmarks of multi-source monitoring data and create a standardized sample library for direct modeling. Furthermore, there is a lack of reproductive behavior evaluation and prediction models that couple multiple factors, including hydrology, hydrodynamics, habitat conditions, and human disturbance, into a unified framework. This makes it difficult to identify key factors influencing reproductive behavior and spawning-related activities, as well as their threshold ranges. It is even more challenging to predict the evolution of reproductive activities under multiple scenarios and develop operable operational scheduling and protection / restoration strategies.

[0004] Therefore, there is an urgent need to propose a method for predicting the reproductive behavior of Chinese sturgeon in wild breeding farms. This method should enable unified formatting of monitoring data from multiple channels, calculation of reproductive behavior indicators and construction of observation vectors, multi-factor coupled modeling, identification of key factor thresholds and scenario prediction decision support, so as to provide technical support for the scientific management of breeding farms and the improvement of reproductive effectiveness. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a method for predicting the reproductive behavior of Chinese sturgeon, which solves the problems of existing Chinese sturgeon reproductive monitoring and assessment methods, such as difficulty in unifying multi-source data, difficulty in quantitatively evaluating reproductive behavior, difficulty in identifying key factor thresholds, and insufficient support for prediction and decision-making.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for predicting the reproductive behavior of the Chinese sturgeon includes the following steps: Step 1: Acquisition and standardized formatting of monitoring data from Chinese sturgeon breeding farms; Step 2: Calculation of evaluation indicators for the reproductive behavior of Chinese sturgeon; Step 3: Construct a multi-factor coupled evaluation model for the reproductive behavior of Chinese sturgeon; Step 4: Identify key factors and threshold ranges affecting the reproductive behavior of Chinese sturgeon; Step 5: Predict the evolution of the reproductive activities of Chinese sturgeon under the combined effects of multiple factors, and propose strategies for the operation, scheduling, protection and restoration of breeding farms.

[0007] Preferably, step 1 specifically involves: dividing the monitoring water area in the wild breeding ground of Chinese sturgeon into zones and identifying the suspected spawning core area in the deep water, key channel sections, and areas adjacent to the barrier net; acquiring monitoring data on target activities and environmental factors through multiple channels; and converting the monitoring data into standardized data records in a unified format.

[0008] Preferably, the monitoring data includes: a) Target activity data: underwater acoustic echo and target extraction data, imaging sonar target imaging data, positioning or trajectory data, and underwater video verification data; b) Environmental factor data: water level, flow rate, water depth, flow velocity, flow direction, turbulence index, water temperature, dissolved oxygen, turbidity, and the amount of shipping or noise intensity.

[0009] Preferably, the standardized data record in the unified format includes: time, spatial coordinates, partition number, data source device identifier and its unit; the conversion method includes: unifying time synchronization, spatial reference, statistical time period Δt, and dimensions, and removing or marking invalid echo segments and abnormal segments that occur during the acquisition process as not to participate in the statistics, thereby forming a standardized data sample set that can be directly used for the index calculation in step 2 and the model construction in step 3.

[0010] Preferably, step 2 specifically involves: calculating evaluation indicators for the reproductive behavior of Chinese sturgeon based on the target activity data and environmental factor data obtained in step 1; the indicators include aggregation intensity indicators, residence time indicators, kinematic indicators, vertical activity intensity indicators, and suspected spawning index, and the multi-source indicators are fused into an observation vector through normalization or standardization processing, which is used to analyze the spatiotemporal distribution patterns of reproductive-related behaviors and spawning-related activities of Chinese sturgeon, such as swimming, residence, aggregation and companionship, and suspected spawning.

[0011] Preferably, the reproductive behavior evaluation indicators and their definitions are as follows: (1) Aggregation intensity, defined as the ratio of the number of targets within the core monitoring area to the area of ​​the region within a unit time period:

[0012] In the formula: The intensity of aggregation; Time period The number of targets extracted from the core monitoring area; The core monitoring area; (2) Dwell time, defined as the cumulative duration of an individual's stay within the core monitoring area:

[0013] In the formula: For length of stay; , The first The moment of entering and leaving the core area; This refers to the number of times the user entered the core area within the statistical period. (3) Vertical activity intensity, defined as the average absolute rate of change of individual depth time series data over the time period:

[0014] In the formula: Vertical activity intensity; For the first Depth values ​​at each sampling time; For the first The timestamp of each sampling moment; This represents the number of sampling points within the time period. This represents the change in depth between adjacent time points; () represents the time interval between adjacent samplings; the unit of VAI is "depth unit / time unit" (such as m / s or m / min).

[0015] (4) Suspected spawning index, defined as a weighted combination of aggregation intensity, dwell time, vertical activity intensity, and video evidence characteristics:

[0016] In the formula: The index is suspected to be the spawning index. They are respectively for , , Normalized metrics (such as linear normalization to [0,1] or Z-score standardization followed by mapping); The value is a normalized value of behavioral evidence features extracted from image sonar or video verification (such as the intensity of escort pursuit, group morphological change characteristics, frequency of key actions, etc.); The weighting coefficients for each indicator satisfy the following conditions: And it is acceptable .

[0017] Preferably, step 3 specifically involves: constructing a Chinese sturgeon reproductive behavior evaluation model using a time-series probability model coupled with multiple factors. The model takes the observation vector formed in step 2 as input, and uses hydrological factors, hydrodynamic factors, and habitat condition factors as covariates. It outputs the occurrence probability of each reproductive behavior category, the behavior category determination result, and the spatiotemporal range of suspected spawning activities. The reproductive behavior categories include swimming, loitering, gathering and swimming together, and suspected spawning. The model estimates the parameters through maximum likelihood estimation or the EM algorithm and realizes continuous identification and evaluation of reproductive behavior on a time scale.

[0018] Preferably, the time-series probability model is a non-homogeneous hidden Markov model (NH-HMM), which divides the monitoring data into preset time periods. Discretized At any given time; let the hidden state variable be... Indicates the first The observation vector corresponding to the reproductive behavior category (such as swimming, loitering, gregarious companionship, egg-laying related behaviors, etc.) at each time point. The evaluation indicators constructed in step 2 consist of (e.g.) ), covariate vector For multiple environmental and disturbance factors (such as flow rate, water level, flow velocity, flow direction, turbulence, propulsion facility operating conditions, water temperature, dissolved oxygen, turbidity, noise, or ship intensity) acquired synchronously within the same time period, the initial state distribution of the model is as follows:

[0019] In the formula: Initially in state The probability of; This represents the hidden state in the first time period; This represents the total number of hidden states. For state index; The observation probability (emission probability) of the model is defined as:

[0020] In the formula: In the state The observed index vector The probability of; For the first The observation vector for each time period, with dimension [ ]. ; For state The observed distribution function can be selected as a multivariate Gaussian distribution or a mixture thereof; For the observed distribution parameters, if a Gaussian distribution is taken, then... ,in It is a mean vector. It is the covariance matrix; It is indexed by time; The state transition probabilities of the model are driven by covariates and change over time, and are presented in a multinomial Logit form.

[0021] In the formula: For covariates From state under conditions Transition to state The probability of; and The first With the The hidden state of a time period; For the first The covariate vector for each time period has dimensions of . This includes water level, flow rate, flow velocity, flow direction, turbulence characteristics, water temperature, dissolved oxygen, turbidity, key topographic water depth or bottom sediment characteristics, and the intensity of ship activity or noise level; For the base transfer term; This is a vector of covariate coefficients; Indicates transpose; The state index is used for summation; Model parameters By using maximum likelihood estimation and the EM (Baum–Welch) algorithm, the posterior probability of each state at each time step is calculated:

[0022] In the formula: For the first Each time period belongs to the state. The posterior probability; For observation vectors; For covariates; The total number of time periods is used to determine the behavior category.

[0023] In the formula: For the first The behavioral category determination results for each time period; This indicates the state index that maximizes the posterior probability. The positioning system is in the Spatial location given in a time period Combined with posterior probability, using a threshold Extract the corresponding status of egg-laying related behaviors Spatiotemporal range:

[0024] In the formula: A collection of the spatiotemporal scope of activities related to spawning; For the first Spatial location at a given time period (can be planar coordinates and water depth); The hidden state number corresponding to oviposition-related behaviors; The threshold is used to extract the set of continuous time periods and corresponding spatial locations from the posterior probability curve.

[0025] Preferably, step 4 specifically involves: to address the potential multicollinearity of covariates in the model from step 3 and improve model prediction accuracy, firstly, correlation tests and multicollinearity diagnoses are performed on the covariates to eliminate redundant factors. Based on this, a set of candidate models containing different combinations of covariates, interaction terms, and nonlinear terms is constructed, and the optimal model is selected using the Akaike Information Criterion (AIC). Based on the optimal model, key factors influencing the probability of reproductive behavior (especially the probability of suspected spawning status) are identified according to covariate coefficients, marginal effects, or posterior probability change rates. The threshold range of these key factors is determined using response curve inflection points, piecewise regression, or probability thresholding methods, thereby quantifying the direction and intensity of the key factors' influence on reproductive behavior. The specific process is as follows: 4.1) Construction of candidate models: Based on the covariates determined in step 3 and their possible interactions, multiple candidate models are constructed. The candidate models differ in terms of variable selection and complexity. 4.2) Model Fitting and Parameter Estimation: Fit each candidate model and calculate the log-likelihood function value of the model. and related parameters; 4.3) Calculation of AIC value: The AIC value of each candidate model is calculated according to the Akaike Information Criterion formula:

[0026] In the formula: The number of model parameters; This represents the log-likelihood function value of the model; 4.4) Model comparison and selection: Compare the AIC values ​​of all candidate models. The smaller the AIC value, the better the model fit and the lower the complexity. Select the model with the smallest AIC value as the optimal model. 4.5) Determination of key factors and threshold intervals: By analyzing the covariates and their regression coefficients or marginal effects in the optimal model, the key factors that have a significant impact on the probability of suspected spawning status or the suspected spawning index are identified, and their threshold intervals are determined based on the response relationship.

[0027] Preferably, step 5 specifically involves: based on the key factors and their threshold ranges determined in step 4, setting up multi-factor scenario combinations and using different hydrological and hydrodynamic conditions, habitat conditions, and disturbance intensities as model inputs; using the optimal model obtained in step 3 to simulate and predict the probability of occurrence of each reproductive behavior category and the suspected spawning index; plotting the response relationship curves between key factors and the probability and index of suspected spawning states; and outputting the time period, duration, and spatial hotspot distribution of spawning-related activities, thereby revealing the evolutionary pattern of Chinese sturgeon reproductive activities under the combined effects of multiple factors and providing a quantitative basis for breeding farm decision-making; the breeding farm operation scheduling and protection and restoration strategies include: a) Hydrodynamic and ecological management strategies for breeding farms: Based on the predicted suspected spawning windows and hotspot areas, optimize the operation of flow propulsion facilities or hydrodynamic organization methods to maintain favorable flow velocity, water depth and turbulence conditions in key water areas in order to promote the aggregation and swimming of Chinese sturgeon and the occurrence of spawning-related behaviors. b) Mitigation measures for shipping and noise interference: During suspected spawning windows, implement navigation and speed restrictions, noise restrictions or set up temporary quiet zones in key waters, and take control measures to avoid suspected spawning hotspots to reduce human interference; c) Risk warning and patrol response measures: When high concentration intensity, long dwell time, and high vertical activity intensity occur in the vicinity of the barrier net and exceed the threshold range, a risk warning of the barrier net being blocked is triggered, and on-site patrol, diversion or facility adjustment measures are carried out. d) Key microhabitat maintenance and restoration measures: carry out maintenance and cleaning, obstacle removal or micro-topography optimization of key deep trench / substrate microhabitats to improve the suitability of spawning microhabitats and ensure the smooth progress of spawning activities.

[0028] The beneficial effects of this invention are as follows: 1. This invention addresses the need for predicting the reproductive behavior of Chinese sturgeon in wild breeding farms. It acquires target activity and environmental factor data through multiple channels and performs unified time synchronization, unified coordinate benchmark, unified statistical time period Δt, and dimension processing to form a data sample library with consistent structure that can be directly calculated and modeled.

[0029] 2. Based on multi-source monitoring data, this invention establishes a reproductive behavior evaluation index system including aggregation intensity, residence time, vertical activity intensity, and suspected spawning index. It also constructs a reproductive behavior evaluation model (preferably a non-homogeneous hidden Markov model NH-HMM) that couples hydrology, hydrodynamics, habitat conditions, and interference factors to achieve automatic identification of reproductive-related behaviors such as swimming, residence, aggregation and companionship, and suspected spawning, as well as quantitative extraction of the spatiotemporal range of spawning-related activities. At the same time, AIC screening is used to identify key factors and threshold ranges that can influence the occurrence of reproductive behavior and spawning-related activities.

[0030] 3. This invention can predict the time period, duration, and spatial hotspot distribution of spawning-related activities under multiple scenarios, thereby forming strategies directly corresponding to the operation of the breeding farm, such as hydrodynamic organization optimization, navigation restriction and noise interference control, risk warning and patrol of the adjacent area of ​​the net blocking, and maintenance and restoration of key micro-habitats. This provides quantifiable and verifiable technical basis for the refined operation scheduling and protection and restoration of the breeding farm, and improves the scientific nature of the operation and breeding effectiveness of the Chinese sturgeon breeding farm.

[0031] 4. This invention overcomes the shortcomings of existing Chinese sturgeon breeding monitoring and assessment methods, such as difficulty in unifying multi-source data, difficulty in quantitatively evaluating breeding behavior, difficulty in identifying key factor thresholds, and insufficient support for prediction and decision-making. It achieves continuous identification of breeding behavior categories and prediction of occurrence probability, identifies key factors affecting the occurrence of breeding behavior and spawning-related activities and their threshold ranges, and reveals the evolution law of breeding activities under the comprehensive effect of multiple factors through multi-scenario simulation prediction, forming an operable breeding farm operation scheduling and protection and restoration strategy, providing technical support for the refined management of breeding farms and the improvement of breeding effectiveness. Attached Figure Description

[0032] Figure 1 This is a flowchart of the method steps of the present invention; Figure 2 A schematic diagram showing the monitoring water area zones and equipment deployment; Figure 3 Correlation analysis diagram of multifactor covariate matrix; Figure 4 Evaluate the forest graph for optimal model parameters; Figure 5 The curves showing the relationship between spawning probability and flow velocity U are presented. Figure 6 The curves showing the relationship between spawning-related probability and turbulent TI response; Figure 7 The response curve of spawning-related probability to shipping volume P; Figure 8 The curves show the relationship between spawning probability and water temperature T. Detailed Implementation

[0033] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0034] Example 1: As Figures 1 to 7 As shown, this invention provides a method for predicting the reproductive behavior of Chinese sturgeon. This method aims to identify reproductive-related behaviors and predict spawning-related activities in key waters of breeding farms. First, it completes zonal positioning in the monitored waters of the breeding farm, acquiring target activity data and environmental factor data through multiple channels. These data are then processed using unified time synchronization, a unified coordinate benchmark, a unified statistical time period Δt, and unified dimensions to form a standardized data sample library. Based on this, evaluation indicators such as aggregation intensity, residence time, vertical activity intensity, and suspected spawning index are calculated according to Δt. A reproductive behavior evaluation model coupling multiple factors such as hydrology, hydrodynamics, habitat conditions, and human disturbance is constructed. Key factors and threshold ranges affecting reproductive behavior are screened and identified. The optimal model is used to predict the evolution of reproductive activities under different scenarios, thus providing a quantitative basis for the operation scheduling and protection and restoration of breeding farms. The above-mentioned method proposed in this invention can provide technical support for the scientific management and improvement of reproductive effectiveness in wild Chinese sturgeon breeding farms. The steps are as follows: Step 1: Acquisition and standardized formatting of monitoring data from Chinese sturgeon breeding farms; Step 2: Calculation of evaluation indicators for the reproductive behavior of Chinese sturgeon; Step 3: Construct a multi-factor coupled evaluation model for the reproductive behavior of Chinese sturgeon; Step 4: Identify key factors and threshold ranges affecting the reproductive behavior of Chinese sturgeon; Step 5: Predict the evolution of the reproductive activities of Chinese sturgeon under the combined effects of multiple factors, and propose strategies for the operation, scheduling, protection and restoration of breeding farms.

[0035] In step 1, the method for acquiring and uniformly formatting monitoring data from Chinese sturgeon breeding farms is as follows: First, the monitoring area is divided into zones to identify key water areas such as suspected spawning core areas in deep water, key channel sections, and areas adjacent to the netting. Target activity data and environmental factor data are then acquired through multiple channels in these water areas. Target activity data preferably includes echo intensity, target intensity (TS), target density, and vertical distribution data obtained by the EY80 scientific fish finder; target 3D position and trajectory data obtained by the VPS 3D positioning system; target imaging and entry / exit channel characteristic data obtained by the DIDSON imaging sonar; and behavioral verification video data obtained by underwater camera equipment. Environmental factor data preferably includes water level, flow rate, water depth, flow velocity, flow direction, turbulence indicators, water temperature, dissolved oxygen, turbidity, and the number of ships or noise intensity. Furthermore, the aforementioned multi-source data is uniformly converted into standardized data records. These standardized data records include at least time, spatial coordinates, partition numbers, data source device identifiers, and their units. The conversion includes unifying time synchronization, spatial reference, statistical time period Δt, and dimensions. Invalid echo segments and abnormal segments that occur during the acquisition process are removed or marked as not participating in the statistics, thereby forming a standardized data sample library that can be directly used for index calculation in step 2 and model construction in step 3.

[0036] In step 2, the calculation method for the Chinese sturgeon reproductive behavior evaluation index is as follows: Based on the target activity data and environmental factor data obtained in step 1, the data are aligned and statistically analyzed according to a unified time period Δt to calculate the Chinese sturgeon reproductive behavior evaluation index. The index includes aggregation intensity (GI), residence time (RT), vertical activity intensity (VAI), and suspected spawning index (SI). The definitions of each evaluation index are as follows: (1) Aggregation intensity, defined as the ratio of the number of targets within the core monitoring area to the area of ​​the region within a unit time period:

[0037] In the formula: The intensity of aggregation; Time period The number of targets extracted from the core monitoring area (which can be obtained from image sonar target extraction); The area of ​​the core monitoring region.

[0038] (2) Dwell time, defined as the cumulative duration of an individual's stay within the core monitoring area:

[0039] In the formula: For length of stay; , The first The moment of entering and leaving the core area; This represents the number of times the data entered the core area within the statistical period.

[0040] (3) Vertical activity intensity, defined as the average absolute rate of change of individual depth time series data over the time period:

[0041] In the formula: Vertical activity intensity; For the first Depth values ​​at each sampling time; For the first The timestamp of each sampling moment; This represents the number of sampling points within the time period. This represents the change in depth between adjacent time points; The interval between adjacent sampling is 0. The unit of VAI is "depth unit / time unit" (e.g., m / s or m / min).

[0042] (4) Suspected spawning index, defined as a weighted combination of aggregation intensity, dwell time, vertical activity intensity, and video evidence characteristics:

[0043] In the formula: The index is suspected to be the spawning index. They are respectively for , , Normalized metrics (such as linear normalization to [0,1] or Z-score standardization followed by mapping); The value is a normalized value of behavioral evidence features extracted from image sonar or video verification (such as the intensity of escort pursuit, group morphological change characteristics, frequency of key actions, etc.); The weighting coefficients for each indicator satisfy the following conditions: And it is acceptable .

[0044] In step 3, the method for constructing a multi-factor coupled reproductive behavior evaluation model is as follows: Using the observation vector formed in step 2 as input, and hydrological factors, hydrodynamic factors, habitat conditions, and human disturbance factors as covariates, a time-series probability model is constructed to output the occurrence probability and judgment result of each reproductive behavior category, achieving continuous identification and evaluation of reproductive behavior over a time scale. The preferred time-series probability model is a non-homogeneous hidden Markov model (NH-HMM): monitoring data is processed according to time periods. Discretized Set the hidden state Indicates the first Reproductive behavior categories for each time period (including at least swimming, loitering, gregarious companionship, and spawning-related behaviors), observation vector Depend on , , , Composed of indicators, covariates The model consists of parameters such as flow rate, water level, flow velocity, flow direction, turbulence intensity, operating conditions of the flow propulsion facility, water temperature, dissolved oxygen, turbidity, and navigation or noise intensity. The model estimates parameters through maximum likelihood estimation and uses the EM algorithm to obtain the posterior probability of spawning-related behavioral states. (Egg Laying Status), and combined with the VPS spatial location, output the time period and spatial hotspot distribution of egg laying-related activities (see...). Figure 6 ).

[0045] In step 4, the key factor and threshold interval identification method is as follows: Based on the posterior probability or suspected spawning index of spawning-related status obtained in step 3, a candidate model set containing different covariate combinations is constructed. First, correlation testing and collinearity diagnosis are performed to eliminate redundant factors. Then, each candidate model is fitted and the AIC value is calculated. The model with the smallest AIC is selected as the optimal model. Based on the optimal model, key factors are determined according to regression coefficients, marginal effects, or inflection points of response curves. Threshold intervals are determined through piecewise regression or probability thresholding (see...). Figure 5 The formula for AIC is:

[0046] In the formula: The number of model parameters; This represents the log-likelihood function value of the model.

[0047] In step 5, the multi-scenario prediction and strategy proposal method is as follows: Based on the key factors and threshold ranges determined in step 4, a multi-factor scenario combination is set and input into the optimal model to predict the response relationship between the probability of spawning-related states / suspected spawning index and scenario changes. Suspected spawning windows and spatial hotspots are output, and strategies such as hydrodynamic and ecological scheduling of breeding grounds, mitigation of shipping and noise interference, risk warning and patrol management, and maintenance and restoration of key microhabitats are proposed accordingly (see...). Figure 7 ).

[0048] Table 1. Data file variables for the reproductive behavior evaluation model

[0049] Example 2: Taking a monitoring cycle of a Chinese sturgeon wild breeding ground downstream of the Gezhouba Hydropower Station as an example (see...) Figure 2 ), take time period Δt =10min, continuous monitoring obtained T=420 time periods of data. First, DIDSON and VPS were used to extract target counts, individual location and depth changes in the core monitoring area, and GI, RT, VAI and SI were calculated according to equations (1)–(4), and the multi-source indicators were combined into an observation vector; second, NH-HMM was used to identify four types of latent states: "swimming, dwelling, gathering and accompanying, and spawning-related behaviors", and the posterior probability of spawning-related states was obtained. (Egg-laying status) and set a threshold. A value ≥0.60 is considered a spawning-related status, denoted as Status=1; otherwise, it is 0. Then, using Status as the dependent variable, a candidate multivariate model is constructed, and correlation tests and AIC screening are performed (see...). Figure 3 (Compared with Table 3), the optimal model was obtained and the evaluation results of key factor parameters were output (see Table 3). Figure 4 (See Table 4); Finally, plot the response curves of the key factors and give the threshold intervals (see Table 4). Figure 5 – Figure 7 ), used to guide the scheduling, protection, and restoration of breeding farms: 1. Sample data variables and value range: Record the hydrodynamic, hydrological, water quality, and disturbance factors (Q, U, TI, D, T, DO, TU, P, PF) corresponding to each time period, as well as GI, RT, VAI, and SI calculated from the monitoring data, and use them as... The threshold determination yields the Status (Table 2).

[0050] 2. Collinearity diagnosis and factor screening: Correlation analysis was performed on candidate covariates (see...) Figure 3 It was found that flow rate Q and flow velocity U are strongly correlated (|r|>0.7). To avoid multicollinearity, flow velocity U and turbulence TI, which are more directly explainable to the behavior mechanism, were retained in the candidate models, and Q was included as an alternative in the extended model.

[0051] 3. AIC Selection of Optimal Model: A set of candidate models with different combinations of factors was constructed, and the AIC was calculated to obtain the optimal model (Table 3). The results show that the model with flow velocity U, turbulence TI, shipping volume P, and water temperature T as independent variables has the lowest AIC and is the optimal model.

[0052] 4. Direction and significance of influence of key factors: Evaluation results based on optimal model parameters (Table 4). Figure 4 It can be seen that flow velocity U and turbulence TI have a significant positive effect on the probability of spawning-related states (P<0.01), shipping volume P has a highly significant negative effect (P<0.001), and water temperature T has a significant negative effect (P<0.05).

[0053] 5. Threshold Range and Management Implications: Table 4 shows that for every 1 m / s increase in flow velocity U, the probability of spawning-related states increases by approximately 3.60 times; increased turbulence TI significantly increases the probability of spawning-related states occurring; for every 1 ship / hour increase in shipping volume P, the probability of spawning-related states decreases to 0.81; increased water temperature T has a significant negative effect on the probability of spawning-related states occurring. The response curves of U, TI, P, T, and the suspected spawning index SI were plotted (…). Figure 5 – Figure 7 When U enters the favorable range and TI is at a medium-high level, while P is below the interference threshold, the probability of spawning-related states increases significantly. Based on this, the operating conditions of propulsion facilities can be optimized during the suspected spawning window, shipping and noise interference can be reduced, and high-risk warnings and patrols can be triggered in the vicinity of the net.

[0054] Table 2 Sample Data File Variables

[0055] Table 3 Model Selection Based on Akaike Information Criterion (AIC)

[0056] Note:

[0057] For model weights, These are the optimal model weights.

[0058] Table 4 Parameter Evaluation of the Optimal Model (U+TI+P+T)

[0059] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for predicting the reproductive behavior of the Chinese sturgeon, characterized in that: It includes the following steps: Step 1: Acquisition and standardized formatting of monitoring data from Chinese sturgeon breeding farms; Step 2: Calculation of evaluation indicators for the reproductive behavior of Chinese sturgeon; Step 3: Construct a multi-factor coupled evaluation model for the reproductive behavior of Chinese sturgeon; Step 4: Identify key factors and threshold ranges affecting the reproductive behavior of Chinese sturgeon; Step 5: Predict the evolution of the reproductive activities of Chinese sturgeon under the combined effects of multiple factors, and propose strategies for the operation, scheduling, protection and restoration of breeding farms.

2. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 1, characterized in that: Step 1 specifically involves dividing the monitoring waters in the wild breeding grounds of Chinese sturgeon into zones and identifying the suspected spawning core area in the deep water, key channel sections, and areas adjacent to the barrier nets. Monitoring data on target activities and environmental factors are obtained through multiple channels, and the monitoring data is converted into standardized data records in a unified format.

3. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 2, characterized in that: The monitoring data includes: a) Target activity data: underwater acoustic echo and target extraction data, imaging sonar target imaging data, positioning or trajectory data, and underwater video verification data; b) Environmental factor data: water level, flow rate, water depth, flow velocity, flow direction, turbulence index, water temperature, dissolved oxygen, turbidity, and the amount of shipping or noise intensity.

4. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 2, characterized in that: The standardized data record in the unified format includes: time, spatial coordinates, partition number, data source device identifier and its unit; The conversion method includes: unifying the time, spatial reference, statistical time period Δt, and dimensions, and removing or marking invalid echo segments and abnormal segments that occur during the acquisition process so as to form a standardized data sample set that can be directly used for the index calculation in step 2 and the model construction in step 3.

5. The method for predicting the reproductive behavior of Chinese sturgeon according to claim 1, characterized in that: Step 2 specifically involves: based on the target activity data and environmental factor data obtained in Step 1, calculating evaluation indicators for the reproductive behavior of Chinese sturgeon; the indicators include aggregation intensity indicators, residence time indicators, kinematic indicators, vertical activity intensity indicators, and suspected spawning index, and the multi-source indicators are fused into an observation vector through normalization or standardization processing, which is used to analyze the spatiotemporal distribution patterns of reproductive-related behaviors and spawning-related activities of Chinese sturgeon, such as swimming, residence, aggregation and companionship, and suspected spawning.

6. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 5, characterized in that: The evaluation indicators for reproductive behavior and their definitions are as follows: (1) Aggregation intensity, defined as the ratio of the number of targets within the core monitoring area to the area of ​​the region within a unit time period: In the formula: The intensity of aggregation; Time period The number of targets extracted from the core monitoring area; The core monitoring area; (2) Dwell time, defined as the cumulative duration of an individual's stay within the core monitoring area: In the formula: For length of stay; , The first The moment of entering and leaving the core area; This refers to the number of times the user entered the core area within the statistical period. (3) Vertical activity intensity, defined as the average absolute rate of change of individual depth time series data over the time period: In the formula: This refers to the intensity of vertical activity. For the first Depth values ​​at each sampling time; For the first The timestamp of each sampling moment; This represents the number of sampling points within the time period. This represents the change in depth between adjacent time points; () represents the time interval between adjacent sampling sessions; (4) Suspected spawning index, defined as a weighted combination of aggregation intensity, dwell time, vertical activity intensity, and video evidence characteristics: In the formula: The index is suspected to be the spawning index. They are respectively for , , The indicators after normalization; The normalized value of behavioral evidence features extracted from image sonar or video verification; The weighting coefficients for each indicator satisfy... And it is acceptable .

7. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 1, characterized in that: Step 3 specifically involves constructing a Chinese sturgeon reproductive behavior evaluation model using a time-series probability model coupled with multiple factors. The model takes the observation vector formed in Step 2 as input, and uses hydrological factors, hydrodynamic factors, and habitat condition factors as covariates. It outputs the probability of occurrence of each reproductive behavior category, the behavior category determination result, and the spatiotemporal range of suspected spawning activities. The reproductive behavior categories include swimming, loitering, gathering and swimming together, and suspected spawning. The model estimates the parameters through maximum likelihood estimation or the EM algorithm and realizes continuous identification and evaluation of reproductive behavior on a time scale.

8. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 7, characterized in that: The time-series probability model is a non-homogeneous hidden Markov model, which divides the monitoring data into preset time periods. Discretized At any given time; let the hidden state variable be... Indicates the first The reproductive behavior category corresponding to each time point, observation vector The evaluation indicators constructed in step 2 constitute the covariate vector. The initial state distribution of the model is as follows, considering the multi-factor environmental and interference factors acquired synchronously within the same time period: In the formula: Initially in state The probability of; This is the hidden state in the first time period; This represents the total number of hidden states. For state index; The observation probability of the model is defined as: In the formula: In the state The observed index vector The probability of; For the first The observation vector for each time period, with dimension [ ]. ; For state The observed distribution function can be selected as a multivariate Gaussian distribution or a mixture thereof; For the observed distribution parameters, if a Gaussian distribution is taken, then... ,in It is the mean vector. It is the covariance matrix; It is indexed by time; The state transition probabilities of the model are driven by covariates and change over time, and are presented in a multinomial Logit form. In the formula: For covariates From state under conditions Transition to state The probability of; and The first With the The hidden state of a time period; For the first The covariate vector for each time period has dimensions of . This includes water level, flow rate, flow velocity, flow direction, turbulence characteristics, water temperature, dissolved oxygen, turbidity, key topographic water depth or bottom sediment characteristics, and the intensity of ship activity or noise level; For the base transfer term; This is a vector of covariate coefficients; Indicates transpose; The state index is used for summation; Model parameters By using maximum likelihood estimation and the EM (Baum–Welch) algorithm, the posterior probability of each state at each time step is calculated. In the formula: For the first Each time period belongs to the state. The posterior probability; For observation vectors; For covariates; The total number of time periods is used to determine the behavior category. In the formula: For the first The behavioral category determination results for each time period; This indicates the state index that maximizes the posterior probability. The positioning system in the Spatial location given in a time period Combined with posterior probability, using a threshold Extract the corresponding status of egg-laying related behaviors Spatiotemporal range: In the formula: A set of spatiotemporal scopes for activities related to spawning; For the first Spatial location within a time period; The hidden state number corresponding to oviposition-related behaviors; The threshold is used to extract the set of continuous time periods and corresponding spatial locations from the posterior probability curve.

9. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 1, characterized in that: Step 4 specifically involves: To address the potential multicollinearity of covariates in the model from Step 3 and improve prediction accuracy, firstly, correlation and collinearity tests are performed on the covariates to eliminate redundant factors. Based on this, a set of candidate models containing different combinations of covariates, interaction terms, and nonlinear terms is constructed, and the optimal model is selected using the Akaike Information Criterion. Based on the optimal model, key factors influencing the probability of reproductive behavior are identified according to covariate coefficients, marginal effects, or posterior probability change rates. Threshold intervals for these key factors are determined using response curve inflection points, piecewise regression, or probability thresholding methods, thereby quantifying the direction and intensity of their influence on reproductive behavior. The specific process is as follows: 4.1) Construction of candidate models: Based on the covariates determined in step 3 and their possible interactions, multiple candidate models are constructed. The candidate models differ in terms of variable selection and complexity. 4.2) Model Fitting and Parameter Estimation: Fit each candidate model and calculate the log-likelihood function value of the model. and related parameters; 4.3) Calculation of AIC value: The AIC value of each candidate model is calculated according to the Akaike Information Criterion formula: In the formula: The number of model parameters; This represents the log-likelihood function value of the model; 4.4) Model comparison and selection: Compare the AIC values ​​of all candidate models. The smaller the AIC value, the better the model fit and the lower the complexity. Select the model with the smallest AIC value as the optimal model. 4.5) Determination of key factors and threshold intervals: By analyzing the covariates and their regression coefficients or marginal effects in the optimal model, the key factors that have a significant impact on the probability of suspected spawning status or the suspected spawning index are identified, and their threshold intervals are determined based on the response relationship.

10. The method for predicting the reproductive behavior of the Chinese sturgeon according to claim 1, characterized in that: Step 5 specifically involves: based on the key factors and their threshold ranges determined in Step 4, setting up multi-factor scenario combinations and using different hydrological and hydrodynamic conditions, habitat conditions, and disturbance intensities as model inputs; using the optimal model obtained in Step 3 to simulate and predict the probability of occurrence of each reproductive behavior category and the suspected spawning index; plotting the response relationship curves between key factors and the probability and index of suspected spawning states; and outputting the time period, duration, and spatial hotspot distribution of spawning-related activities, thereby revealing the evolutionary pattern of Chinese sturgeon reproductive activities under the combined effects of multiple factors and providing a quantitative basis for breeding farm decision-making; the breeding farm operation scheduling and protection and restoration strategies include: a) Hydrodynamic and ecological management strategies for breeding farms: Based on the predicted suspected spawning windows and hotspot areas, optimize the operation of flow propulsion facilities or hydrodynamic organization methods to maintain favorable flow velocity, water depth and turbulence conditions in key water areas in order to promote the aggregation and swimming of Chinese sturgeon and the occurrence of spawning-related behaviors. b) Mitigation measures for shipping and noise interference: During suspected spawning windows, implement navigation and speed restrictions, noise restrictions or set up temporary quiet zones in key waters, and take control measures to avoid suspected spawning hotspots to reduce human interference; c) Risk warning and patrol response measures: When high concentration intensity, long dwell time, and high vertical activity intensity occur in the vicinity of the barrier net and exceed the threshold range, a risk warning of the barrier net being blocked is triggered, and on-site patrol, diversion or facility adjustment measures are carried out. d) Key microhabitat maintenance and restoration measures: carry out maintenance and cleaning, obstacle removal or micro-topography optimization of key deep trench / substrate microhabitats to improve the suitability of spawning microhabitats and ensure the smooth progress of spawning activities.

Images