A method for optimizing station of marine fishery resources investigation based on space-time coupling model

Abstract

The application discloses a marine fishery resource investigation station adaptive optimization method based on a space-time coupling model. The method couples a Bayesian hierarchical species distribution model and a regional marine model system to construct a space-time continuous species distribution dynamic benchmark field. Ecological stratification is carried out based on time, space and environmental dimensions. When the environmental variable changes significantly, the habitat suitability index is dynamically reconstructed and the station layout is re-optimized to ensure that the key ecological hotspot area is covered and the gradient zone is maintained. The method has the advantages of fast environmental response, strong spatial representativeness and high adaptability, and is suitable for resource assessment, ecological protection and biodiversity monitoring scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of marine fishery resource monitoring and investigation, in particular to a station layout optimization method combining ecological modeling and marine numerical simulation, specifically to a marine fishery resource investigation station adaptive optimization method based on a space-time coupling model, belonging to the cross-technology category of marine ecological monitoring, fishery resource assessment and investigation design optimization. BACKGROUND

[0002] In the current marine fishery resource assessment and investigation work, the scientificity and dynamic response capability of station layout directly affect the representativeness and accuracy of fishery resource monitoring. Traditional fishery resource investigation usually relies on fixed or random station layout methods, among which simple random sampling, systematic sampling, cluster sampling and stratified sampling methods are widely used. Especially in China, stratified sampling is commonly used, which is divided according to geographical features such as water depth, bottom type, estuary and geographical location, etc. These methods have the advantage of simple operation, but still have significant shortcomings in ecological sensitive area identification, resource hotspot coverage and dynamic response to environmental changes.

[0003] Specifically, several main problems of current investigation design include:

[0004] 1. Weak ability to depict resource spatial heterogeneity: The existing stratification standard mainly depends on geographical factors, and fails to reflect the coupling mechanism of resource aggregation (hot spot) area and environmental gradient, resulting in the possibility of missing ecological hot spot area and affecting the representativeness of the investigation.

[0005] 2. Lack of ecological process consideration: The traditional division method fails to reflect the driving effect of environmental factor (such as temperature, salinity) gradient change on species distribution, so as to truly reflect the coordination mechanism between resource and environment.

[0006] 3. Static simulation distribution assumption deviates from actual ecological dynamics: When simulating the "real" distribution, Kriging interpolation or statistical modeling method is often used, but most of them are based on historical static variables (geographical location, water depth, etc.), assuming that the resource distribution is stable in space and time. This method ignores the influence of seasonal variation, climate events (such as El Nino) and sudden environmental disturbance on resource distribution, resulting in the lack of foresight and adaptability of the generated distribution field.

[0007] 4. Lack of rapid response mechanism to climate change and sudden events: When the marine environment undergoes abnormal changes (such as sudden temperature rise, pollution events), the existing station layout method is difficult to adjust in time, causing the investigation to lag behind the ecological response, affecting the timeliness and accuracy of resource assessment.

[0008] In addition, under the background of increasing demand for multi-species surveys, traditional methods are difficult to simultaneously consider populations at different life stages and ecological needs, especially during critical ecological periods such as spawning, feeding, and overwintering, and cannot accurately identify population aggregation areas and effectively deploy stations.

[0009] Therefore, it is urgent to establish a survey station optimization method with dynamic modeling capability, ecological driving mechanism interpretation, and adaptive response capability, which can fully integrate environmental variables and species behavior, design a survey program based on the spatial-temporal-environmental three-dimensional characteristics of species distribution, improve survey efficiency and scientificity, and meet the high-quality development requirements of modern marine ecological protection and resource management. SUMMARY

[0010] The purpose of the present application is to overcome the problems of strong staticity in existing fishery resource survey station design, insufficient coupling of ecological processes, and lag response to environmental disturbance, and to propose a station optimization method with adaptive ability, which can dynamically adjust the survey layout according to environmental changes, and take into account ecological representativeness and operational efficiency. To achieve the above purpose, the present application is based on the deep coupling of ecological modeling and physical simulation, and constructs a dynamic benchmark field of species distribution with continuous space-time, on the basis of which a multi-dimensional ecological region division system is designed, and a dynamic stationing mechanism is introduced combined with real-time environmental monitoring results, thereby improving the scientificity and sensitivity of fishery resource survey.

[0011] To achieve the above goal, the present application provides a marine fishery resource survey station adaptive optimization method based on a spatio-temporal coupling model, and the technical scheme is as follows:

[0012] In one possible implementation, a marine fishery resource survey station adaptive optimization method based on a spatio-temporal coupling model is provided, aiming to improve the spatial representativeness and environmental response capability of fishery resource survey. The method includes the following three core steps:

[0013] (1) Dynamic benchmark field construction: This step integrates historical survey data and multi-source environmental factors (such as temperature, salinity, chlorophyll concentration, etc.) by deep coupling Bayesian hierarchical species distribution model (INLA) and regional ocean model system (ROMS) through data assimilation technology, and establishes a dynamic benchmark field of species distribution with physical consistency and ecological significance. The field has high spatial and temporal resolution and can dynamically reflect the trend of resource density, which is the basis for subsequent layering and optimization operations.

[0014] (2) Three-dimensional ecological region stratification design: In order to make the sampling strategy reflect the characteristics of ecological processes, the time dimension (such as life stage division), the space dimension (such as hot spot region identification based on Getis-Ord Gi statistics) and the environmental dimension (such as temperature and salinity gradient identification) are comprehensively divided into ecological regions. Among them, the time dimension focuses on the key stages of the target species such as spawning period, foraging period and wintering period; the space dimension divides the resource aggregation degree by means of significance Z value; the environmental dimension identifies the ecological sensitive physical boundary through Sobel operator and Thorpe scale.

[0015] (3) Adaptive station optimization mechanism: When the actual environmental conditions are significantly different from the model construction period (such as temperature difference exceeding ±1.5℃), the benchmark field reconstruction and station adjustment are automatically triggered. The optimization goal is to ensure that the survey station covers more than 80% of the resource hot spot area, and at least 3 stations are arranged in each environmental gradient zone to enhance the spatial representativeness and ecological adaptability of the data. Station selection is completed by stratified random sampling, and the candidate point set is initialized on the pre-defined grid (such as 30'x30'). The evaluation index includes relative estimation error and relative deviation.

[0016] In one possible implementation, the Bayesian hierarchical model is composed of two sub-models:

[0017] Species presence / absence sub-model: used to predict the probability of the target species appearing in different regions, based on binomial distribution and log link function;

[0018] Species conditional abundance sub-model: used to estimate the species density in the area where the species has appeared, using Poisson distribution and log link function;

[0019] Spatial autocorrelation is modeled by Gaussian Markov random field to ensure the spatial continuity and reasonableness of the prediction results.

[0020] In one possible implementation, different mechanisms are introduced in the data assimilation link to handle heterogeneous variables:

[0021] Temperature and salinity are assimilated in real time by particle filtering technology;

[0022] Chlorophyll concentration is biophysical coupled assimilated by combining NPZD ecological model to improve the ecological accuracy of habitat prediction.

[0023] In one possible implementation, the three-dimensional stratification design is guided by ecological processes:

[0024] The spawning period of the time dimension is determined by the gonadal development index or the cumulative temperature threshold method;

[0025] The foraging period is determined by combining stomach content analysis;

[0026] The overwintering period is identified based on the overwintering cluster behavior.

[0027] In terms of spatial dimension (Getis-Ord) Spatial statistical methods (statistics) are used to identify hotspots in species distribution. By calculating the statistical significance of each spatial unit with its neighboring units, the study area is divided into core, transition, and background zones.

[0028]

[0029] in, Let i be the spatial weights of positions i and j. S is the population mean, and S is the standard deviation. According to... Size, dividing the area into hotspot core areas ( >1.96), transition zone (-1.96≤ ≤1.96) and background area ( <-1.96), enabling a quantitative description of spatial heterogeneity of resources.

[0030] In terms of environmental dimensions, the gradient fields of temperature and salinity are calculated using the Sobel operator, and the abrupt change points of the mixed layer depth are located using the Thorpe scaling method to identify gradient zones with ecological transition significance.

[0031] In one possible implementation, adaptive optimization is achieved through the following process:

[0032] Temperature is the primary monitoring variable, and a deviation threshold of ±1.5℃ is set as the trigger condition.

[0033] The updated environmental field is obtained using the ROMS model and then input into the INLA model to predict resource distribution.

[0034] Recalculate the Habitat Suitability Index (HSI) and reassess the ecological zone delineation accordingly;

[0035] A stratified random sampling strategy is adopted to select stations from the candidate point set (such as a 30′×30′ grid) to meet the dual constraints of "hotspot coverage + gradient representativeness".

[0036] In one possible implementation, the technical configuration of the ROMS model includes:

[0037] The horizontal resolution is 1-3 kilometers;

[0038] The vertical terrain is divided into 20-30 layers following the coordinates;

[0039] The atmospheric forcing data source was ERA5 reanalysis data with a temporal resolution better than 6 hours.

[0040] In a possible implementation, the method is suitable for marine fishery resource assessment, ecological protection zone planning, marine biodiversity monitoring and other application scenarios, and has strong popularization value and practical significance.

[0041] Based on the above technical scheme, the "marine fishery resource survey station adaptive optimization method based on a space-time coupling model" provided by the application introduces three technical innovations of physical-ecological coupling, ecological process layering and real-time response mechanism in the field of fishery resource survey design, systematically solves the key problems of station distribution static, poor ecological representativeness and environmental response lag in the traditional fishery survey design method, and has significant technical progress and practical application value.

[0042] Firstly, the dynamic reference field construction method proposed by the application breaks through the limitation of relying on static sample data or geographical factors for species distribution modeling, and realizes the deep coupling of Regional Ocean Model System (ROMS) and Bayesian Hierarchical Species Distribution Model (INLA) for the first time. By introducing data assimilation mechanism, integrating satellite remote sensing, buoy observation and historical survey data, and strengthening the consistency and spatial continuity of parameters through filtering and ecological model, a species distribution prediction field with high resolution, strong timeliness and complete ecological logic is generated. This dynamic reference field can not only be used for current survey station optimization, but also can serve for resource change warning and trend simulation in future period, significantly improving the foresight of resource management.

[0043] Secondly, the three-dimensional ecological layering system constructed by the application realizes the fundamental change of sampling design from "geographical driving" to "ecological driving". In the time dimension, the monitoring window is dynamically divided according to the life history stage of the species, which enhances the sampling representativeness of key periods such as spawning period, foraging period and wintering period; in the spatial dimension, the Getis-Ord statistic method is introduced to identify resource hotspots, scientifically delimit high aggregation area and background area, and improve the ability of station to capture spatial heterogeneity of resources; in the environmental dimension, the ecological boundary is identified based on temperature-salinity gradient and mixed layer structure, which makes up for the deficiency of traditional methods in identifying physical-ecological transition zone. The coordinated design of three-dimensional layering provides an ecological basis for station optimization, ensuring the systematicness and integrity of sampling results.

[0044] Again, the adaptive station optimization mechanism introduced by the present application has dynamic perception and real-time response capability. When external environmental conditions such as temperature deviate significantly from the model set value (such as more than ±1.5℃), the system can automatically reconstruct the dynamic reference field and adjust the sampling layout, ensuring that the survey design and the ecological actuality evolve synchronously. The optimization algorithm considers the dual constraints of ecological hotspot coverage and environmental gradient representation, and further improves the accuracy and stability of resource density estimation by screening the optimal station layout scheme from the candidate grid point set through stratified random sampling. In actual simulation, the optimization mechanism outperforms the traditional equidistant layout method in monitoring index error control under the premise of reducing the number of stations, showing higher efficiency and reliability.

[0045] In addition, the present application has good versatility and scalability. The method framework is not dependent on specific regional parameter settings and can adapt to different sea areas, different target species and different time scales of fishery survey tasks; the modularized model structure and standardized data input facilitate the docking with existing business numerical prediction systems, and can be widely applied in national or regional fishery resource assessment, ecological protection zoning and marine ranching planning scenarios.

[0046] In summary, the present application first integrates ecological modeling methods (such as Bayesian hierarchical species distribution models) and marine numerical simulation systems (such as ROMS-ecological coupled models), and proposes a multi-species, multi-time period and multi-ecological gradient adaptive optimization method for fishery resource survey station. Compared with the existing station design methods based on spatial uniformity or empirical regional division, the present application establishes a station optimization process with real-time response capability through dynamic reference field driven suitability assessment, spatial hotspot identification and ecological zoning modeling, breaking through the technical bottlenecks of traditional methods such as “static division, single-scale consideration and low environmental sensitivity”. The method has high ecological representativeness, environmental adaptability and algorithm scalability, and can be widely applied to different sea areas and multi-target species survey tasks, improving the survey efficiency while enhancing the scientificity and accuracy of resource assessment, and has important theoretical value and application prospect. DETAILED DESCRIPTION

[0047] To better understand the adaptive optimization method for marine fishery resource survey station based on spatio-temporal coupled model proposed by the present application, the technical implementation process of the present application will be described in detail in combination with typical application scenarios.

[0048] It should be understood that the core idea of the present application is to integrate ecological process modeling with physical ocean simulation, to build a dynamic distribution benchmark field based on historical data and real-time assimilation results, and to dynamically optimize the spatio-temporal configuration structure of the survey station by combining a multi-dimensional ecological zoning method and an environmental response mechanism. This method aims to improve the scientificity, representativeness and adaptability of the design of fishery survey stations, especially in scenarios where the distribution of target populations is significantly affected by climate and environment, the spatial pattern is complex, and the survey resources are limited.

[0049] The present embodiment takes a specific sea area as the object, relies on real fishery survey data and environmental observation data, and gradually explains the whole process of model construction, data processing, ecological zoning and station optimization to show the applicability and effectiveness of the present application in actual survey tasks. However, it should be noted that the following examples are only preferred ways of the present application, and any replacement and adjustment of the model structure, algorithm parameters or data sources without departing from the essence of the present application belongs to the protection scope of the present application.

[0050] The following describes the adaptive optimization method of marine fishery resource survey stations based on the spatio-temporal coupling model proposed by the present application in combination with actual marine fishery survey scenarios. The present embodiment takes a key fishing ground in the East China Sea as an example, and the specific process is as follows:

[0051] In the survey preparation stage, first, fishery resource survey data covering the spawning period, foraging period and overwintering period are collected, and environmental observation data in the corresponding time period are obtained, including sea surface temperature, salinity, chlorophyll a concentration, wind speed and other multi-source information. Based on this data, the input data set of the coupling prediction model is constructed to provide data support for subsequent modeling and optimization.

[0052] In terms of model construction, the Integrated Nested Laplace Approximation (INLA) method is combined with the Regional Ocean Modeling System (ROMS) to construct an ecological-physical integrated species distribution prediction model. The INLA method is used to approximate the Bayesian inference in the latent Gaussian model, aiming to realize fast and accurate calculation of the posterior joint distribution. The model assumes as follows:

[0053] Let the observation vector be y=(y1,…,yn), the Gaussian random field be x=(x1,…,xn), the hyperparameter be θ=(θ1,…,θk) (k∈N), μ(θ) and Q(θ) represent the mean vector and precision matrix respectively, and the joint distribution satisfies:

[0054]

[0055] The linear prediction term satisfies:

[0056]

[0057] where, is the intercept, is the coefficient of the covariate , represents the random effect with covariate as the domain, is the error term. In the present invention, is mainly used to express the spatial (or spatio-temporal) structure, which is usually quantified by the spatial autocorrelation through Gaussian Markov Random Field (GMRF) modeling.

[0058] The INLA model structure includes two sub-models: one is the binomial distribution model constructed for the presence / absence of species (Presence-Absence), and the other is the Poisson distribution model for the modeling of abundance under the condition of presence, both of which use the logarithmic link function. At the same time, in order to consider the zero inflation problem and spatial heterogeneity in the sampling process, the model introduces a spatial random term to enhance the credibility of the prediction.

[0059] The environmental covariates of the above model come from the ROMS output field. The spatial resolution of the ROMS model is set to 1-3 km, and the vertical direction is divided into 20-30 layers using terrain-following coordinates (sigma coordinates), with emphasis on strengthening the thermocline and surface area. The initial state of the physical field is set by global reanalysis data, and the boundary conditions are provided by global models such as HYCOM or GLORYS; atmospheric forcing data is provided by the ERA5 dataset, with a time resolution of 6 hours. Chlorophyll concentration is introduced into the NPZD ecological model as an ecological process indicator, and the ROMS and NPZD modules jointly construct a physical-ecological integrated simulation system.

[0060] In the data assimilation process, the particle filter algorithm is used for physical field correction of temperature and salinity variables; while the chlorophyll concentration is coupled and assimilated through the fusion of the NPZD model and remote sensing data. Short-term forecast analysis is performed every 6 hours, and the physical reasonableness of the frontal and mixed layer changes is judged according to the assimilation increment field.

[0061] Based on the above modeling and assimilation results, a high spatio-temporal resolution dynamic benchmark field of species distribution is generated. Further, combined with the predicted resource abundance A and its extreme value, the habitat suitability index (Habitat Suitability Index, HSI) is calculated:

[0062]

[0063] where, and The minimum and maximum of resource abundance in the prediction area, respectively. HSI is used to measure the resource suitability of different grid cells and serves as the basis for subsequent ecological evaluation and optimization.

[0064] After the benchmark field construction is completed, the three-dimensional ecological stratification phase is entered. First, in the time dimension, the whole year is divided into spawning period, feeding period and overwintering period according to the life history stages of the target species. The identification methods include: determining the spawning period by gonad development grading or cumulative temperature threshold method; identifying the feeding period by combining with the peak time of plankton; determining the overwintering period according to the winter population aggregation behavior.

[0065] In the spatial dimension, Getis-Ord Gi* method is used to identify resource hot spot areas, and the significance value of each grid and its adjacent unit is calculated

[0066]

[0067] Among them, is the spatial weight of position i and j, is the overall mean, and S is the standard deviation (S is the standard deviation of the global attribute variable). According to size, the area is divided into hot core area ( >1.96), transition zone (-1.96≤ ≤1.96) and background area ( <-1.96), realizing the quantitative description of resource spatial heterogeneity.

[0068] In the environmental dimension, in order to identify the key ecological gradient zone, Sobel operator is used to calculate the horizontal gradient field of temperature and salinity, and Thorpe scale method is used to analyze the sudden change of mixed layer depth, so as to define the position of ecological transition zone.

[0069] On the basis of the above three-dimensional partition results, the application further designs an adaptive station optimization mechanism. When the deviation of sea surface temperature during the planned investigation period and the benchmark value during model construction exceeds ±1.5℃, the dynamic reconstruction process is triggered. The system calls INLA and ROMS model for rapid reestimation, updates HSI distribution and ecological stratification structure, and adjusts the station layout under the following two optimization constraints: first, the layout area should cover more than 80% of the hot spot grid; second, at least 3 investigation stations should be arranged in each environmental gradient zone.

[0070] ​​The final station optimization is realized by a hierarchical random sampling algorithm. First, a set of candidate grid points of 30' x 30' is established based on the ecological regional boundary, and under the premise of meeting the above constraints, grid points with higher HSI values and more representative environmental gradient are preferentially selected for stationing. By introducing the resource index (such as CPUE) as the sampling weight, and comparing the relative estimation error (Relative Estimation Error) and the relative bias (Relative Bias) under different station number conditions, the comprehensive advantages of the optimization scheme in precision, efficiency and cost are verified.

[0071] Application Example One:

[0072] In one specific embodiment, the East China Sea is taken as the survey area to carry out the practical application and technical verification of the method of the present application.

[0073] First, data preparation and model initialization are performed. The survey team has collected fishery resource survey data covering different seasons in the past five years, including the spatial distribution, abundance observation, etc. of the target species in the spawning period, feeding period and wintering period, as well as multi-source environmental observation data in the same period, including sea surface temperature (SST), salinity (SSS), chlorophyll-a concentration, wind speed and wind stress. After pre-processing in a unified format, all data are used as model inputs.

[0074] Subsequently, a regional ocean numerical model system (ROMS) is constructed and initialized, with a horizontal resolution of 1 to 3 kilometers and a vertical division of 20 to 30 layers, and a sigma coordinate system is used to adapt to complex bottom shapes. The boundary conditions of ROMS are provided by the HYCOM global reanalysis product, and the atmospheric forcing field is provided by the ERA5 dataset, with a time resolution controlled within 6 hours. The NPZD model is used to simulate key ecological factors such as chlorophyll concentration, and a physical-ecological coupled model system with ecological dynamics characteristics is constructed.

[0075] Next, dynamic reference field construction is performed. Particle filtering method is used for assimilation of temperature and salinity fields, and NPZD model is used for biological-physical coupled assimilation of chlorophyll concentration. Bayesian hierarchical species distribution model is constructed by integrating nested Laplace approximation (INLA) algorithm, wherein:

[0076] The first layer is the occurrence probability sub-model, which uses binomial distribution and logarithmic link function;

[0077] The second layer is the conditional abundance sub-model, which uses Poisson distribution and logarithmic link function;

[0078] Both layers introduce Gaussian Markov random field to describe spatial correlation.

[0079] After the model inference is completed, the maximum value of the predicted resource abundance is used. and minimum value Calculate the Habitat Suitability Index (HSI) for each predicted location:

[0080]

[0081] HSI is used as a quantitative indicator to measure the habitat suitability of a target species under a specific grid, providing a basis for subsequent regional delineation.

[0082] In the three-dimensional ecological stratification design phase, the survey area was divided into three dimensions: time, space, and environment. In the time dimension, based on the ecological habits of the target species and historical observation data, the year was divided into the spawning period (based on gonadal development level and accumulated temperature), the foraging period (combined with peak zooplankton biomass), and the overwintering period (referring to winter gregarious behavior characteristics). In the spatial dimension, the Getis-Ord model was adopted. Statistical methods are used to identify resource hotspots. By calculating the local standardized score Z, the study area is divided into hotspot core areas (…). >1.96), transition zone (-1.96 ≤ ≤ 1.96) and background area ( <-1.96). In terms of environmental dimensions, the temperature-salinity gradient is calculated using the Sobel operator, and the Thorpe scaling method is used to identify abrupt changes in the depth of the mixed layer in order to define the boundaries of the ecological gradient.

[0083] Based on the three-dimensional zoning results, an adaptive station location optimization mechanism was implemented. Temperature was used as the key environmental variable, with a trigger threshold set at ±1.5℃. When the environmental deviation between the prediction survey period and the modeling period exceeded the threshold, the system automatically reconstructed the dynamic baseline field and ecological stratification structure, and re-estimated the posterior distribution of species distribution using the INLA model. The optimized station placement followed the following dual constraints:

[0084] First, ensure that over 80% of the ecological hotspot grids are covered;

[0085] Second, at least three survey stations should be set up in each environmental gradient zone.

[0086] Based on this, a 30′×30′ spatial grid candidate point set is constructed, and the final survey station is selected through a stratified random sampling algorithm. Priority is given to grid locations with strong ecological representativeness and high resource density, while controlling the number and distribution of stations to balance survey accuracy and cost-effectiveness.

[0087] To evaluate the effectiveness of the optimization scheme, an index system for survey accuracy evaluation was constructed. The relative estimation error (REE) and the relative bias (RB) were set as evaluation indexes, and the proposed scheme was compared with the traditional equidistant layout method. The simulation results show that, under the same number of stations, the REE and RB of the optimization scheme are reduced by more than 15% on average, and the accuracy in hot spot areas and the coverage rate of ecological gradient zones are better.

[0088] This example also verifies the practicality of the method in the application scenarios of red tide response, interannual variation of fishery resources warning, and dynamic delineation of ecological red line, and reflects its wide adaptability and engineering feasibility in data-driven ecological monitoring.

[0089] Application Example Two:

[0090] In another specific embodiment, the method of the present application is applied to the nearshore fishing ground in the southern Yellow Sea, which is an important habitat and foraging ground for many economic fish species (such as small fish and yellow perch), and has been frequently disturbed by sudden red tide events in recent years, with significant fluctuations in resource distribution. To achieve dynamic monitoring and efficient survey station layout of key populations in this sea area, the adaptive optimization method based on the spatio-temporal coupling model described in the present application is used for pilot application.

[0091] First, synchronous collection of survey data and environmental data was carried out. The research team collected fishery resource survey data of the same period (May to August) in the past three years, including trawl density, catch biomass of target fish, and information on plankton distribution, and obtained high-frequency environmental observation data collected by nearshore buoys and unmanned aerial vehicle remote sensing systems. The collected environmental variables include sea surface temperature (SST), salinity, chlorophyll-a concentration, turbidity, wind field and tidal current data, etc.

[0092] Based on the above data, a ROMS numerical model of the southern Yellow Sea region was constructed. To adapt to the complex topography and shelf dynamic characteristics of the nearshore area, the model grid horizontal resolution was set to 1 km, the vertical direction used a 24-layer sigma coordinate system, and the upper mixing layer was encrypted. The boundary data was derived from the regional nested North Yellow Sea HYCOM sub-model, and the 6-hour resolution ERA5 data provided by CMEMS was used for atmospheric forcing input. The NPZD module was tightly coupled with the physical field to describe the evolution process of chlorophyll concentration and nutrient salt, and to provide dynamic boundary conditions for the ecological model.

[0093] After the initialization of the numerical model, the particle filter was introduced for data assimilation of sea temperature and salinity, and the biological field variables (chlorophyll-a) and NPZD output were co-corrected to generate realistic physical-ecological coupled state fields. On this basis, the INLA method was used to construct a hierarchical Bayesian distribution model, which combined environmental variables and historical species distribution data to obtain a probabilistic spatial distribution prediction of the target fish species in the region.

[0094] In particular, to cope with the abnormal disturbance of sea surface temperature and chlorophyll caused by red tide, the adaptive station arrangement threshold in this embodiment is set as follows: the sea surface temperature deviates from the baseline state by more than 1.2℃ or the chlorophyll concentration anomaly increases by more than 50%. When the trigger condition is met, the system automatically performs dynamic baseline field reconstruction and re-executes habitat suitability index (HSI) calculation and ecological region reclassification.

[0095] In the ecological stratification aspect, the time dimension is based on the feeding intensity peak and spawning activity of the target species to determine the key monitoring period; the spatial dimension uses the Gi statistical method to identify high-density habitat patches near the coast, and the results show that the resource hotspots have a significant linear aggregation feature, and the core area is concentrated near the estuary outflow zone and the tidal mixing front; the environmental dimension identifies the positions of the temperature-salinity front and the nutrient salt mutation zone through the Sobel operator to construct stratification boundaries with ecological indication significance.

[0096] In the final optimization of station arrangement, under the double constraints of "80% hotspot coverage" and "≥3 stations per gradient zone", the optimal station arrangement scheme is selected from the initialized 30'x30' grid through stratified random sampling method. Compared with the traditional uniform interval arrangement scheme, it is found that under the same number of stations, the representativeness of the target species habitat hot zone is improved by 18%, the resource amount error is reduced by 12%, and the system can complete adaptive update of station arrangement structure within 48 hours after the occurrence of sudden red tide.

[0097] As can be seen, the present application shows good sensitivity and response ability in small-scale, high-dynamic nearshore ecosystems, and can effectively improve the scientificity and responsiveness of fishery surveys, providing important technical support for fishery resource monitoring and protection in the face of environmental emergencies.

[0098] Application Example Three:

[0099] In another specific embodiment, the present application is applied to the Taiwan Strait region of China, which is significantly affected by multi-scale ocean circulation (such as monsoon flow, coastal current, strait exchange flow), has strong physical field instability, and has complex fishery resources and significant seasonal and regional differentiation in distribution. In order to carry out cross-season, multi-target species survey tasks, while coping with the challenges of long survey period and high prediction uncertainty, this embodiment uses the adaptive optimization method of survey station based on the spatio-temporal coupling model proposed by the present application to improve the scientificity, foresight and robustness of station arrangement.

[0100] The implementation period covers from April to October, including spring, summer, and autumn. The target species are large pelagic fish and cephalopods. First, a ROMS regional model covering the entire Taiwan Strait is constructed, with a horizontal resolution of 3 km. The double-nested mechanism is adopted, with the outer boundary connected to HYCOM and the inner sub-region used for high-precision prediction in key sea areas. The vertical direction uses a 30-layer sigma coordinate system, focusing on enhancing the vertical accuracy of the middle and upper layers to the thermocline region. The atmospheric forcing field is derived from ERA5 products, with a 3-hour resolution. The NPZD advanced module is selected for the biogeochemical module, and the sediment nutrient release mechanism is introduced to simulate the impact of bottom nitrogen and phosphorus load on chlorophyll and plankton.

[0101] To cope with seasonal mutations that may occur during long-term investigation tasks (such as summer monsoon outbreaks and vertical disturbances of water masses caused by tropical cyclones), a multi-source fusion strategy is set for environmental variable assimilation: particle filtering method is used to fuse remote sensing data (such as MODIS, AVHRR) and Argo profile data for sea surface temperature and salinity fields, and the NPZD model is combined with shipboard sampling data for rolling updates of plankton and chlorophyll concentration. The assimilation system runs once a day, outputting 3-7 day prediction products.

[0102] At the ecological model level, INLA models are used to model each target species, building a multi-species, hierarchical joint distribution framework. The INLA model structure introduces an inter-species spatial random effect covariance term on the standard GAM framework, enabling it to capture spatial collaborative variation patterns among coexisting populations. The prediction results of each species are superimposed, and a comprehensive resource suitability index (Multi-species HSI) is formed through weighted averaging, which serves as the basis for comprehensive station optimization, ensuring spatial representativeness of multi-target species investigation.

[0103] The calculation of habitat suitability index (HSI) still follows the standard formula:

[0104]

[0105] where, represents the predicted abundance of the mth species at location i, and are the maximum and minimum abundances of the species at all grid points, respectively. The comprehensive index is calculated by:

[0106]

[0107] where, W m is the weight of the mth species, set according to its ecological status, economic importance, and seasonal dynamics.

[0108] In the optimization of station position, considering the long investigation period and unstable meteorological window, more stringent station arrangement constraints are set:

[0109] Coverage constraint: the cumulative coverage of the HSI>0.6 area should reach more than 90%;

[0110] Redundancy constraint: two sets of redundant station positions should be arranged in the core hotspot area for replacing in severe sea conditions;

[0111] Transmission constraint: the optimization result should adapt to the ship maneuvering path and fuel budget, and the spatial distribution scheme with the shortest route should be selected preferentially.

[0112] The optimization algorithm introduces a spatial clustering-hierarchical sampling joint method: firstly, K-means is used to preliminarily cluster the candidate points based on the ecological stratification result, and then weighted hierarchical sampling is performed in each ecological sub-area to ensure the balance between representativeness and workability. The comparison between the optimized station positions and the original voyage station arrangement shows that, under the condition of reducing the total number of stations by 12%, the average sampling time is saved by 15%, and the species abundance estimation error is controlled within ±10%.

[0113] The present application realizes high-precision adaptive layout of investigation station positions in the complex, dynamic and multi-species coexistence area of the Taiwan Strait in China, and has strong system adaptability, spatial generalization ability and multi-objective balance ability, and is especially suitable for the needs of standardization station design of future national long-term fishery resource monitoring plan, R / V investigation voyage overall optimization and related international joint investigation projects.

[0114] In summary, the present application provides a marine fishery resource investigation station adaptive optimization method based on a space-time coupling model, which fully integrates marine physical processes and biological and ecological characteristics, realizes multi-dimensional ecological stratification by constructing a high-precision dynamic reference field, and carries out adaptive station arrangement based on an environmental trigger mechanism, thereby effectively improving the scientificity, systematicness and real-time response ability of fishery resource investigation. The method overcomes the limitations of traditional station design relying on static assumptions and poor adaptability to environmental changes, and can realize flexible deployment and dynamic optimization in multi-species, multi-region and multi-scale application scenarios, and has good universality and engineering implementability.

[0115] The technical solutions described in the present application are not only suitable for fishery resource assessment and dynamic monitoring, but also can be widely applied in the fields of ecological protection area planning, marine environmental impact assessment, fishery management early warning system construction and the like, and have significant application prospect and popularization value.

[0116] It should be understood by those skilled in the art that any equivalent replacement or modification of specific structure, model details and parameter setting without departing from the essential spirit of the present application should be considered to fall within the protection scope of the present application.

Claims

1. A method for optimizing the station of marine fishery resources survey based on a spatio-temporal coupling model, characterized in that, Comprising the following steps: (1) Dynamic baseline field construction: coupling the integrated nested Laplace approximation (INLA) algorithm Bayesian hierarchical species distribution model with the regional ocean model system (ROMS), integrating historical fishery survey data and environmental variables through data assimilation technology to generate a spatiotemporally continuous dynamic baseline field of species distribution, wherein the Bayesian hierarchical species distribution model comprises: species presence / absence sub-model: predicting the presence probability based on the binomial distribution and the log link function; species conditional abundance sub-model: predicting the abundance value based on the Poisson distribution and the log link function; the model quantifies spatial autocorrelation through a Gaussian Markov random field; combining the predicted resource abundance with its extreme value, the habitat suitability index is calculated; (2) Three-dimensional hierarchical design: ecological zoning of the survey area through three directions of time dimension, space dimension and environmental dimension; the time dimension divides the ecological period according to the activity characteristics of the target species at different life stages; the spatial dimension uses the significant characteristics of resource distribution to identify hotspots and demarcate resource aggregation areas; the environmental dimension determines the position of the ecological gradient zone according to the gradient change characteristics of temperature, salinity and other environmental factors, thereby forming a three-dimensional hierarchical structure with ecological representativeness; (3) Adaptive station optimization: when the environmental variable changes exceed the preset threshold, automatically reconstruct the dynamic baseline field, recalculate the habitat suitability index, and reevaluate the ecological regionalization accordingly, and then select the station from the candidate point set using the hierarchical random sampling strategy under the constraint condition that more than 80% of the ecological hotspot areas are covered and at least 3 stations are arranged in each environmental gradient zone, to realize the adjustment of station layout.

2. The method of claim 1, wherein, The data assimilation technology in step (1) comprises: particle filtering is used for temperature and salinity assimilation; NPZD ecological model is used for biological-physical coupled assimilation of chlorophyll concentration.

3. The method of claim 1, wherein, The life stage division of the time dimension in step (2) comprises: spawning period: identified by gonadal development characteristics or cumulative temperature threshold method; feeding period: determined by combining plankton biomass change; overwintering period: based on winter clustering behavior.

4. The method of claim 1, wherein, The hotspot analysis of the spatial dimension in step (2) employs Getis-Ord A statistical method comprises: Core region: > 1.96; Transition Zone: -1.96 < x < 1.96 ≤1.96; Context: <-1.

96.

5. The method of claim 1, wherein, The gradient zone identification of the environmental dimension in step (2) comprises: using Sobel operator to calculate temperature and salinity gradient field; using Thorpe scale method to locate the mixed layer depth mutation point.

6. The method of claim 1, wherein, The environmental variable change trigger condition in step (3) is: the average deviation of temperature between the planned survey period and the model construction time exceeds ±1.5℃.

7. The method of claim 1, wherein, The station optimization in step (3) is realized by hierarchical random sampling, specifically comprising: initializing a 30′×30′ grid of candidate point set; using resource amount index as an indicator, the accuracy is evaluated by relative estimation error and relative bias.

8. The method of claim 1, wherein, The configuration of the regional ocean model system (ROMS) comprises: horizontal resolution 1-3 km, vertical 20-30 layers terrain-following coordinates; atmospheric forcing field uses ERA5 reanalysis data, time resolution ≤6 hours.

9. The method according to any one of claims 1-8 for use in marine fishery resource assessment, ecological protection zone planning or biodiversity monitoring.

Images