Rare fish nondestructive monitoring method and system based on water body eDNA sample recognition

Abstract

The invention relates to the technical field of fish monitoring, in particular to a rare fish nondestructive monitoring method and system based on water eDNA sample recognition. A hotspot model is constructed through historical eDNA sampling distribution, sampling hotspot transfer prediction is achieved, and the optimal sampling point position is determined; high-throughput sequencing and a base sequence recognition decision tree are combined, the sequence rare degree is judged, and valuable and rare fishes are retrieved; further coupling the water flow dynamic model and the eDNA migration state space model, and tracing the potential migration area or inhabitation area of the rare fish; and a dynamic population community graph is constructed based on the time sequence community change, and the seasonal scale trend is analyzed. According to the method, accurate nondestructive monitoring of seasonal distribution rules and community structure trends of the rare fishes is realized on the basis of efficient collection and accurate identification of the eDNA of the water body, and a reliable scientific basis is provided for resource protection of the rare fishes in different water areas.

Description

Technical Field

[0001] This invention relates to the field of fish detection technology, and in particular to a non-destructive monitoring method and system for rare fish species based on aquatic eDNA sample identification. Background Technology

[0002] With the increasing demand for aquatic ecological protection and biodiversity assessment, and given the diverse range of rare fish species, including the Chinese sturgeon, Yangtze sturgeon, red-lipped barbel, Songjiang perch, Chinese mud carp, and white-lipped barbel, monitoring of these species is gradually shifting from destructive or semi-destructive methods such as traditional fishing sampling, tagging and recapture, and acoustic detection to non-destructive monitoring technologies centered on aquatic environmental DNA (eDNA). This includes, but is not limited to, shed scales, fish mucus, excrement, or body tissue. eDNA technology identifies the presence of target species by collecting genetic material released by organisms in the water, offering significant advantages in reducing ecological disturbance and expanding monitoring coverage. Furthermore, as rare fish are endangered species, their natural protection and resource management require high priority. Therefore, monitoring and tracking their seasonal migrations is crucial to reducing the risk of human interference with their survival and reproduction, and providing a scientific basis for delineating ecological red lines, nature reserves, and fishing ban areas at different times of the year.

[0003] Current technologies for monitoring rare fish species typically compare eDNA samples with known genetic data from gene databases. This method heavily relies on manual experience and lacks in-depth analysis of the rarity of eDNA base sequence patterns. This leads to significant identification errors in determining the presence of rare fish species in different waters, reducing the rationality and accuracy of subsequent seasonal migration monitoring and resource conservation efforts. Secondly, existing technologies only output static "presence" results, making it difficult to trace potential habitats or migration routes of rare fish species based on sample distribution, and even more difficult to characterize the dynamic size changes of populations in different seasons. Some methods rely on single sequence matching or simple threshold determination, which lacks the ability to identify rare base patterns and is susceptible to background noise and interference from non-target species, affecting the reliability and interpretability of monitoring results. In addition, existing monitoring technologies require the deployment of a large number of monitoring tools underwater, which will undoubtedly interfere with and damage the natural ecology and living environment of underwater fish. For example, when fish encounter underwater cameras, they may exhibit abnormal phenomena such as fright or stress, affecting the normal activities, reproduction and habitat quality of fish. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a non-destructive monitoring method and system for rare fish species based on aquatic eDNA sample identification.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification, comprising the following steps: S01: Obtain the historical collection distribution coordinates of eDNA samples from different water bodies in the target monitoring area, construct a hotspot model based on the historical collection distribution coordinates, analyze the historical state of the hotspot model to express the spatial sampling jump of the hotspots and predict the hotspot transfer of the current sampling distribution of the last collection, and locate the optimal sampling point. S02: Collect water eDNA samples at the optimal sampling point and perform high-throughput sequencing to obtain the actual base sequence set. Construct a base sequence recognition decision tree driven by known pattern rules to identify the rarity of the sequence pattern of the actual base sequence set. Based on the sequence rarity, import the sequence set into the gene coding database for comparison and retrieval of existing rare fish species in the target monitoring water area, and output the recognition results. S03: Construct and couple the forward hydrodynamic model with the migration state space model of water body eDNA samples as they change with different hydrodynamic environments, form the Hamiltonian observation coupling equation, import historical seasonal hydrodynamic monitoring parameters into the Hamiltonian observation coupling equation to perform fuzzy inversion solution of the initial migration area of ​​the sample, and trace the potential migration area or potential habitat of existing rare fish as the first non-destructive monitoring result output; S04: Establish a dynamic population community map of existing rare fish species by analyzing temporal community image changes. Based on the first non-destructive monitoring results, update the local abundance changes of the dynamic population community according to seasonal temporal increments of the global migration network. Analyze the seasonal scale trend of rare fish species as the output of the second non-destructive monitoring results.

[0006] Preferably, step S01 specifically includes the following steps: Obtain the target monitoring area and water body sampling logs, and extract multiple historical collection distribution coordinates of different water body eDNA samples collected at anchor points at continuous historical moments in the target monitoring area from the water body sampling logs; A Kriging interpolation architecture is constructed. Based on historical time anchor points, sampling hotspots are fitted, drawn, and reconstructed within the Kriging interpolation architecture for each corresponding historical data collection distribution coordinate. This generates several historical sampling distribution hotspot models over a continuous time series, and the number of hotspots in each historical sampling distribution hotspot model is counted. The spatial hash misalignment value between the changes in the layout and number of sampling hotspots expressed by the historical sampling distribution hotspot model corresponding to anchor points at adjacent historical moments is calculated using a hash algorithm. Based on the spatial hash misalignment value, the sampling jump rate of the water body eDNA sample in the target monitoring water area is determined. The established operating strategy of the eDNA sampling device and the sampling distribution hotspot model of the last collection are obtained and defined as the current sampling distribution hotspot model. Based on the established operating strategy, the adjacent sampling step length when the eDNA sampling device arrives at each historical collection distribution coordinate is extracted. Based on the adjacent sampling step size and the preset spatiotemporal condition intensity index, a spatiotemporal kernel function containing an exponential decay time kernel and a Gaussian space kernel is constructed based on the sampling jump rate. The maximum likelihood method is introduced. Based on the spatiotemporal condition intensity index, the spatiotemporal kernel function is used in the maximum likelihood method to maintain and update the historical state of the current sampling distribution hotspot model until the maintenance triggers the calculation of the likelihood increment of the current hotspot state, so as to obtain the hotspot transfer prediction probability of water body eDNA samples. Only coordinates of hotspot shift prediction probabilities greater than a preset probability threshold are located and extracted in the target monitoring water area, marked as the optimal sampling points, and uploaded to the data decision terminal of the sampling equipment.

[0007] Preferably, step S02 specifically includes the following steps: By deploying eDNA sampling equipment at the optimal sampling point, several sets of water eDNA samples of the target monitoring water area are collected. The water eDNA samples are then subjected to high-throughput sequence detection using a biological detection method to obtain the actual base sequence set of each water eDNA sample. Based on the knowledge graph of marine fishes, the system retrieves the base sequence string templates of different rare fish species that tend towards empirically standardized vectors, as well as the regulatory genomes corresponding to each base sequence string template. At the same time, the system retrieves the known pattern rules of the eDNA base sequences of rare fish species through graph retrieval. Among them, the known pattern rules include base coding rules and base pairing rules. By analyzing the driving potential of the genome, decision projection based on known pattern rules is performed on each base sequence string template. Base sequence recognition branches are divided for different rare fish species, and base sequence recognition decision trees are generated. The base sequence recognition decision trees are used to make decision recognition on the sequence patterns of actual base sequence groups to determine the sequence rarity of water body eDNA samples. If the sequence rarity is greater than the preset sequence rarity threshold, it indicates that the water body eDNA sample belongs to the eDNA fragment of a rare fish species, and the gene coding database of rare fish species is obtained. The gene coding sequence retrieval code is written into the identified water body eDNA sample using gene editing technology, and the edited water body eDNA sample is then sent to the gene coding database of rare fish for comparison and retrieval. During the comparison and retrieval process, the degree of matching between the retrieval code and the prefix index code of the known gene code corresponding to each rare fish is calculated one by one; Only the rare fish species corresponding to the known gene codes with the highest contrast fit are extracted, and these are identified as the existing rare fish species in the target monitoring water area. The identification results of the water body eDNA samples are output and uploaded to the non-destructive monitoring terminal.

[0008] Preferably, the step of performing decision projection on each base sequence string template based on known pattern rules through the analysis of the driving potential of the genome, dividing the base sequence recognition branches of different rare fish, generating a base sequence recognition decision tree, and using the base sequence recognition decision tree to make decision recognition on the sequence pattern of the actual base sequence group to determine the sequence rarity of the water body eDNA sample, specifically includes the following steps: A pattern evaluation system was introduced to calculate the driving plasticity of each regulatory genome based on known pattern rules. The potential of known sequence patterns to drive base sequence string templates to a high extent was determined as high driving potential, and the potential of known sequence patterns to drive base sequence string templates to a low extent was determined as low driving potential. A slanted decision plane for prior constraints is established using known pattern rules. Each base sequence string template is projected onto the slanted decision plane based on high and low driving potential, resulting in projection values ​​and projection vectors. Based on the projection values, a recursive partitioning operation of leaf nodes for high-throughput site characterization is performed on the base sequence patterns of different rare fish according to the projection vector, generating a base sequence recognition decision tree driven by known pattern rules. The frequency of the actual base sequence group was characterized by fitting the K-mer frequency framework, and the actual K-mer frequency map of the water eDNA sample was generated. The sample pattern site characterization and corresponding site frequency intensity value of the actual base sequence group were extracted by analyzing the actual K-mer frequency map. Based on the site frequency intensity value injected into the base sequence recognition decision tree, the branch factor node performs traversal recognition of the sample pattern site representation, and obtains the pattern decision response value of the leaf node item corresponding to each high-throughput site representation. If the pattern decision response value is greater than the preset response threshold, the leaf node item corresponding to the pattern decision response value is highlighted and marked as a type 1 leaf node item; if the pattern decision response value is less than the preset response threshold, the traversal and dwell of the leaf node item is ignored or skipped and marked as a type 2 leaf node item. The pattern recognition ratio between Class I leaf node items and Class II leaf node items is calculated based on the number of nodes, and the sequence rarity of the water body eDNA sample is determined according to the pattern recognition ratio.

[0009] Preferably, step S03 specifically includes the following steps: Obtain a marine geomorphological map of the target monitoring water area and a fixed deployment layout of underwater monitoring outposts. Based on the fixed deployment layout, divide the marine geomorphological map into N sub-non-destructive monitoring areas. The flow characteristics and potential energy of the target monitoring water area are obtained. A positive flow dynamic model is constructed based on the flow characteristics and potential energy. The past flow monitoring parameters of each sub-non-destructive monitoring area during the historical seasonal period are extracted through the water monitoring log. The past flow monitoring parameters include the flow rate, velocity, direction and wave energy of the flow. A wandering state space model is established to represent the random drift of water eDNA samples driven by different water flow environment conditions that change with time. The Hamilton algorithm is introduced to dynamically and collaboratively couple and transform the wandering state space with the forward hydrodynamic model, forming the Hamilton observation coupling equation for hydrodynamic-driven eDNA sample drift. Based on past water flow monitoring parameters, the Hamilton cost of water flow environmental variables is constructed. The Hamilton cost is then introduced into the Hamilton observation coupling equation to perform inverse integration from the terminal time of the optimal sampling point to the fuzzy time of the initial drift, and a series of global source value functions on the monitoring time series are obtained. By calculating the covariate matrix, the inverse vector and inverse gradient of a series of global source tracing value functions are determined. Based on the inverse vector and inverse gradient, the fuzzy inversion trajectory of the eDNA sample that caused the past water flow monitoring parameters to drift randomly to the optimal sampling point is reconstructed and extracted to generate the source tracing probability of the water eDNA sample reaching each sub-non-destructive monitoring area. If the source tracing probability does not exceed the maximum source tracing probability threshold, then the marine geomorphological area corresponding to the sub-non-destructive monitoring area with that source tracing probability is excluded. If the source tracing probability exceeds the maximum source tracing probability threshold, the marine geomorphological area corresponding to the sub-non-destructive monitoring area of ​​that source tracing probability is determined as the potential migration area or potential habitat area of ​​the existing rare fish belonging to the water body eDNA sample, and the first non-destructive monitoring result is output.

[0010] Preferably, step S04 specifically includes the following steps: Multiple time-series images of existing rare fish communities were captured by underwater monitoring outposts during historical seasonal periods in various potential migration areas or potential habitats, obtained through water monitoring logs. Using community features in the image as given nodes and position changes as temporal connecting edges, a dynamic population community map of existing rare fish in each potential migration area or potential habitat area is constructed, and the historical seasonal period is divided into several seasonal sub-segments with equal intervals. A k-cluster maintenance algorithm based on incremental update mechanism is introduced to maintain the dynamic population community graph of each seasonal sub-segment by enumerating k-clusters. During the maintenance process, if the Mahalanobis distance between adjacent k-clusters is greater than the preset Mahalanobis distance, a cluster adjacency graph of local community seasonal changes is constructed based on the adjacent k-clusters, resulting in several seasonal regional cluster adjacency graph blocks. Based on the first non-destructive monitoring results, we can obtain the seasonal migration distribution network of existing rare fish species that are active or inhabit the target monitoring waters from historical periods to the current period, and at the same time, we can obtain the population survival characteristics of existing rare fish species through big data. Based on the population survival characteristics, we traverse different cluster adjacency graphs to find fish community sets under the continuous seasonal time sequence, and obtain the community connectivity components on each seasonal sub-segment. Using seasonal time series as the batch dynamic cache step size, the connected components of each community are incrementally updated based on the edge insertion-edge deletion situation field formed by the seasonal migration distribution network, so as to obtain the local community structure of existing rare fish under the seasonal expansion and decline under migration changes. The dynamic relative abundance gradient of existing rare fish species is obtained by monitoring the density clustering increment rendering of local community structure expression, and the seasonal size trend of existing rare fish species is determined based on the dynamic relative abundance gradient.

[0011] A second aspect of the present invention provides a non-destructive monitoring system for rare fish species based on aquatic eDNA sample identification. The system includes: a memory, a processor, and a communication interface. The memory includes a program for a non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification. The communication interface is used for data connection communication between the memory and the processor. When the processor executes the program for the non-destructive monitoring method for rare fish species, it implements the steps of any one of the non-destructive monitoring method steps for rare fish species described in the present invention.

[0012] This invention addresses the technical deficiencies in the prior art, and its beneficial technical effects are as follows: By modeling hotspots and analyzing spatial jumps in historical eDNA sampling distribution, dynamic shift prediction of sampling hotspots is achieved, thus accurately locating optimal sampling points without increasing sampling frequency, significantly improving sampling efficiency and representativeness. By introducing a base sequence recognition decision tree driven by known pattern rules, sequence rarity identification and gene database comparison of base sequences obtained from high-throughput sequencing are performed, enabling accurate identification of rare fish species in target waters amidst complex background noise, improving the sensitivity and reliability of identification. Coupled with a hydrodynamic model and a migratory state-space model of eDNA sample diffusion with water flow, a Hamiltonian observation equation is constructed and combined with historical seasonal hydrological parameters for fuzzy inversion, enabling effective tracing of potential migration or habitat areas for rare fish species. A dynamic population community map is constructed through temporal community image changes, continuously updating and analyzing the local abundance and seasonal size trends of rare fish species, achieving multi-scale, long-term population dynamic monitoring. This invention enables efficient collection and accurate identification of aquatic eDNA samples, thereby facilitating non-destructive monitoring of the seasonal migration areas and scale changes of rare fish species in target waters. This provides an efficient and reliable monitoring method for the protection and scientific management of rare fish resources. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0014] Figure 1 A flowchart of the first method for non-destructive monitoring of rare fish species based on aquatic eDNA sample identification is shown. Figure 2 A flowchart of the second method for non-destructive monitoring of rare fish species based on aquatic eDNA sample identification is shown. Figure 3 A system framework diagram of a non-destructive monitoring system for rare fish species based on aquatic eDNA sample identification is shown. Detailed Implementation

[0015] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0017] The first aspect of this invention provides a non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification, such as... Figure 1 As shown, it includes the following steps: S01: Obtain the historical collection distribution coordinates of eDNA samples from different water bodies in the target monitoring area, construct a hotspot model based on the historical collection distribution coordinates, analyze the historical state of the hotspot model to express the spatial sampling jump of the hotspots and predict the hotspot transfer of the current sampling distribution of the last collection, and locate the optimal sampling point. S02: Collect water eDNA samples at the optimal sampling point and perform high-throughput sequencing to obtain the actual base sequence set. Construct a base sequence recognition decision tree driven by known pattern rules to identify the rarity of the sequence pattern of the actual base sequence set. Based on the sequence rarity, import the sequence set into the gene coding database for comparison and retrieval of existing rare fish species in the target monitoring water area, and output the recognition results. S03: Construct and couple the forward hydrodynamic model with the migration state space model of water body eDNA samples as they change with different hydrodynamic environments, form the Hamiltonian observation coupling equation, import historical seasonal hydrodynamic monitoring parameters into the Hamiltonian observation coupling equation to perform fuzzy inversion solution of the initial migration area of ​​the sample, and trace the potential migration area or potential habitat of existing rare fish as the first non-destructive monitoring result output; S04: Establish a dynamic population community map of existing rare fish species by analyzing temporal community image changes. Based on the first non-destructive monitoring results, update the local abundance changes of the dynamic population community according to seasonal temporal increments of the global migration network. Analyze the seasonal scale trend of rare fish species as the output of the second non-destructive monitoring results.

[0018] Preferably, step S01 specifically includes the following steps: Obtain the target monitoring area and water body sampling logs, and extract multiple historical collection distribution coordinates of different water body eDNA samples collected at anchor points at continuous historical moments in the target monitoring area from the water body sampling logs; A Kriging interpolation architecture is constructed. Based on historical time anchor points, sampling hotspots are fitted, drawn, and reconstructed within the Kriging interpolation architecture for each corresponding historical data collection distribution coordinate. This generates several historical sampling distribution hotspot models over a continuous time series, and the number of hotspots in each historical sampling distribution hotspot model is counted. The spatial hash misalignment value between the changes in the layout and number of sampling hotspots expressed by the historical sampling distribution hotspot model corresponding to anchor points at adjacent historical moments is calculated using a hash algorithm. Based on the spatial hash misalignment value, the sampling jump rate of the water body eDNA sample in the target monitoring water area is determined. The established operating strategy of the eDNA sampling device and the sampling distribution hotspot model of the last collection are obtained and defined as the current sampling distribution hotspot model. Based on the established operating strategy, the adjacent sampling step length when the eDNA sampling device arrives at each historical collection distribution coordinate is extracted. Based on the adjacent sampling step size and the preset spatiotemporal condition intensity index, a spatiotemporal kernel function containing an exponential decay time kernel and a Gaussian space kernel is constructed based on the sampling jump rate. The maximum likelihood method is introduced. Based on the spatiotemporal condition intensity index, the spatiotemporal kernel function is used in the maximum likelihood method to maintain and update the historical state of the current sampling distribution hotspot model until the maintenance triggers the calculation of the likelihood increment of the current hotspot state, so as to obtain the hotspot transfer prediction probability of water body eDNA samples. Only coordinates of hotspot shift prediction probabilities greater than a preset probability threshold are located and extracted in the target monitoring water area, marked as the optimal sampling points, and uploaded to the data decision terminal of the sampling equipment.

[0019] It should be noted that traditional eDNA sampling typically relies on fixed sampling sites selected based on human experience. However, due to variations in water flow characteristics in different target water bodies, the location of fish eDNA samples is not fixed. This makes it difficult to quickly and frequently capture eDNA samples from different fish species at a single fixed location, reducing sampling efficiency. To address this, this method uses the distribution coordinate data of previous sampling sites to infer and analyze the patterns of spatial location changes in eDNA samples collected from different water bodies. The nonlinear rate changes caused by the water flow characteristics at different times result in uncertainty and random walk-like discreteness in the spatial location transfer frequency of eDNA samples in the water flow. Traditional spatial inference is difficult to capture the inflection points of the sample changes with the water flow. To address this, this method uses a Kriging interpolation architecture to interpolate and render the historical collection distribution coordinates into hotspot models at different connection times. This provides a global bird's-eye view of the spatial movement layout of eDNA samples. Subsequently, a hash algorithm calculates the spatial hash misalignment value to quantify the temporal jump mileage of the hotspot size and distribution in two adjacent hotspot models, i.e., the sampling jump rate. This sampling jump rate reveals the prior probability guidance of water eDNA samples driven by different water flow characteristics. Compared with traditional spatial pattern methods, this ensures more accurate subsequent transfer probability inference and significantly improves the accuracy of water eDNA sampling point planning and decision-making.

[0020] It should be noted that for the current sampling work, the most valuable reference is the sample distribution results from the most recent historical collection. Therefore, this method uses the sampling distribution hotspot model from the last collection as the termination reference for the current collection. The adjacent sampling step size reflects the time span of spatial transfer of eDNA samples driven by water flow characteristics. Based on its preset spatiotemporal condition intensity index, it characterizes the spatiotemporal excitation dependence strength between the previous time-series hotspot model and the next time-series hotspot model, enabling the sample's time-space transfer inference to possess a self-excited term with a precise static distribution. The spatiotemporal kernel function constructed based on the sampling jump rate controls or restricts the time scale and spatial propagation range of the random drift of hotspots driven by water flow characteristics. This establishes the hotspot transfer occurrence rate under the influence of historical events, avoids misjudging random noise as self-excited propagation of hotspots, and improves the robustness and interpretability of transfer inference. If the predicted probability of hotspot transfer is greater than the preset probability threshold, it indicates that the probability of water eDNA samples in the current target monitoring area being drifted to a certain point by water flow characteristics tends to be maximized, indicating that the maximum probability of capturing water eDNA samples at that precise point can be guaranteed. This method enables the analysis of temporal hotspot changes in water eDNA samples based on historical sample collection and distribution data. By maintaining and updating the current sampling distribution hotspots using historical sampling status based on the analysis of sampling jump results, it is possible to infer the possible locations where water eDNA samples can be accurately collected with high reliability. This avoids the blind and random sampling caused by the reliance on human experience in traditional methods, which leads to a high probability of missed sample collection and improves sampling efficiency.

[0021] Preferably, step S02 specifically includes the following steps: By deploying eDNA sampling equipment at the optimal sampling point, several sets of water eDNA samples of the target monitoring water area are collected. The water eDNA samples are then subjected to high-throughput sequence detection using a biological detection method to obtain the actual base sequence set of each water eDNA sample. Based on the knowledge graph of marine fishes, the system retrieves the base sequence string templates of different rare fish species that tend towards empirically standardized vectors, as well as the regulatory genomes corresponding to each base sequence string template. At the same time, the system retrieves the known pattern rules of the eDNA base sequences of rare fish species through graph retrieval. Among them, the known pattern rules include base coding rules and base pairing rules. By analyzing the driving potential of the genome, decision projection based on known pattern rules is performed on each base sequence string template. Base sequence recognition branches are divided for different rare fish species, and base sequence recognition decision trees are generated. The base sequence recognition decision trees are used to make decision recognition on the sequence patterns of actual base sequence groups to determine the sequence rarity of water body eDNA samples. If the sequence rarity is greater than the preset sequence rarity threshold, it indicates that the water body eDNA sample belongs to the eDNA fragment of a rare fish species, and the gene coding database of rare fish species is obtained. The gene coding sequence retrieval code is written into the identified water body eDNA sample using gene editing technology, and the edited water body eDNA sample is then sent to the gene coding database of rare fish for comparison and retrieval. During the comparison and retrieval process, the degree of matching between the retrieval code and the prefix index code of the known gene code corresponding to each rare fish is calculated one by one; Only the rare fish species corresponding to the known gene codes with the highest contrast fit are extracted, and these are identified as the existing rare fish species in the target monitoring water area. The identification results of the water body eDNA samples are output and uploaded to the non-destructive monitoring terminal.

[0022] It should be noted that aquatic eDNA samples from different fish species possess unique base sequence patterns. For example, the mucus of the Chinese sturgeon is rich in COI (cytochrome oxidase I), and its third codon site exhibits high variation, resulting in a base sequence pattern characterized by: [conserved primer region]—[highly species-specific variation region]—[conserved region]. Therefore, the base sequence pattern can be used to analyze whether an aquatic eDNA sample belongs to a rare fish species. This method uses the well-known pattern rules of rare fish eDNA base sequences as constraints to perform high-throughput site decision-based projection partitioning on standard base sequence text data (base sequence string templates) of different rare fish species. This constructs a tree structure capable of determining the base sequence pattern of eDNA samples from different sources regarding rare fish species—a base sequence identification decision tree. Finally, based on the reasonable tendency of the decision results (sequence rarity), it identifies whether the aquatic eDNA sample belongs to a rare fish species, achieving the effect of rarity identification of aquatic eDNA samples. Contrast fit measures the degree of similarity, agreement, or overlap between the gene codes of collected rare fish eDNA samples and known gene codes stored in the database. This method uses a decision tree structure driven by rare fish base sequence pattern rules to identify the sequence rarity of collected samples. Based on the sequence rarity results, it uses a gene coding database to identify the corresponding rare fish species, achieving effective identification of rare fish species belonging to different water bodies' eDNA samples in the target monitoring area. Compared to the cumbersome steps of traditional high-throughput sequencing requiring manual base sequence difference analysis and one-by-one comparison with gene databases, this method eliminates many inefficient processes, improves the identification efficiency of water body eDNA samples, and optimizes the identification accuracy of rare fish in different water bodies.

[0023] Preferably, the step of performing decision projection on each base sequence string template based on known pattern rules through the analysis of the driving potential of the genome, dividing the base sequence recognition branches of different rare fish, generating a base sequence recognition decision tree, and using the base sequence recognition decision tree to make decision recognition on the sequence pattern of the actual base sequence group to determine the sequence rarity of the water body eDNA sample, specifically includes the following steps: A pattern evaluation system was introduced to calculate the driving plasticity of each regulatory genome based on known pattern rules. The potential of known sequence patterns to drive base sequence string templates to a high extent was determined as high driving potential, and the potential of known sequence patterns to drive base sequence string templates to a low extent was determined as low driving potential. A slanted decision plane for prior constraints is established using known pattern rules. Each base sequence string template is projected onto the slanted decision plane based on high and low driving potential, resulting in projection values ​​and projection vectors. Based on the projection values, a recursive partitioning operation of leaf nodes for high-throughput site characterization is performed on the base sequence patterns of different rare fish according to the projection vector, generating a base sequence recognition decision tree driven by known pattern rules. The frequency of the actual base sequence group was characterized by fitting the K-mer frequency framework, and the actual K-mer frequency map of the water eDNA sample was generated. The sample pattern site characterization and corresponding site frequency intensity value of the actual base sequence group were extracted by analyzing the actual K-mer frequency map. Based on the site frequency intensity value injected into the base sequence recognition decision tree, the branch factor node performs traversal recognition of the sample pattern site representation, and obtains the pattern decision response value of the leaf node item corresponding to each high-throughput site representation. If the pattern decision response value is greater than the preset response threshold, the leaf node item corresponding to the pattern decision response value is highlighted and marked as a type 1 leaf node item; if the pattern decision response value is less than the preset response threshold, the traversal and dwell of the leaf node item is ignored or skipped and marked as a type 2 leaf node item. The pattern recognition ratio between Class I leaf node items and Class II leaf node items is calculated based on the number of nodes, and the sequence rarity of the water body eDNA sample is determined according to the pattern recognition ratio.

[0024] It should be noted that regarding the construction of the identification decision tree, the driving plasticity of each regulatory genome based on the known pattern rules is evaluated. Since the regulatory genomes corresponding to the base sequence string templates of different rare fish species have already been shaped according to the known pattern rules, the known pattern rules here have a sequence pattern driving potential assessment for the regulatory genome, namely high driving potential and low driving potential. This makes the pattern decision domains for forming the left and right subtrees of the tree structure quantify the pattern decision scale for different eDNA sample base sequences, making the subsequent pattern rarity identification of base sequences more accurate. By establishing a slanted decision plane, the known pattern rules can be anchored as a pre-threshold constraint for the identification decision tree structure, so that the decision split direction of the eDNA sample base sequence pattern strictly follows the driving linearity of this rule. Compared with the traditional decision tree structure, the identification decision tree of this method applies this rule constraint, ensuring the rationality and interpretability of the sequence rarity division and identification. Subsequently, each base sequence string template is projected onto the oblique decision plane based on high and low driving potential, achieving pattern decision-making guidance based on the joint features of multi-base sequences of rare fish. Based on the systematic guidance of the projection results, high-throughput site representation leaf nodes for different rare fish base sequence patterns are recursively split, significantly improving the sequence discrimination ability of complex eDNA samples. The K-mer frequency map, represented by a bar chart or histogram, is a visual frequency chart that decomposes the base sequence into fixed-length (K) subsequences (K-mers) and counts their frequency of occurrence. The K-mer frequency map provides representative pattern information of the aquatic eDNA samples, namely, the sample pattern site representation and site frequency intensity values. The site frequency intensity values ​​are injected into the branch factor nodes to await the response of the leaf nodes. If a leaf node responds, a decision tree chain response is formed: rare fish base sequence pattern (root node) - sample pattern site representation (branch factor node) - high-throughput site representation (leaf node), achieving the effect of verifying the rarity of actual base sequences using prior data of rare fish.

[0025] It should be noted that if the pattern decision response value is greater than the preset response threshold, it indicates that a high-throughput site in the actual base sequence set, driven by known pattern rules, highly corresponds to the rarity sequence structure of rare fish species on the corresponding leaf node. Therefore, the leaf node item corresponding to this pattern decision response value is explicitly highlighted and labeled as a Class I leaf node item. Conversely, if the response value is lower, it indicates a lower response, so the traversal and dwell of this leaf node item is ignored or skipped, and it is labeled as a Class II leaf node item. The node ratio between the two types reveals the proportion of sequence rarity bias, thereby enabling the identification of the sequence rarity of the actual base sequence set.

[0026] Preferably, S03, as Figure 2 As shown, the specific steps include: Obtain a marine geomorphological map of the target monitoring water area and a fixed deployment layout of underwater monitoring outposts. Based on the fixed deployment layout, divide the marine geomorphological map into N sub-non-destructive monitoring areas. The flow characteristics and potential energy of the target monitoring water area are obtained. A positive flow dynamic model is constructed based on the flow characteristics and potential energy. The past flow monitoring parameters of each sub-non-destructive monitoring area during the historical seasonal period are extracted through the water monitoring log. The past flow monitoring parameters include the flow rate, velocity, direction and wave energy of the flow. A wandering state space model is established to represent the random drift of water eDNA samples driven by different water flow environment conditions that change with time. The Hamilton algorithm is introduced to dynamically and collaboratively couple and transform the wandering state space with the forward hydrodynamic model, forming the Hamilton observation coupling equation for hydrodynamic-driven eDNA sample drift. Based on past water flow monitoring parameters, the Hamilton cost of water flow environmental variables is constructed. The Hamilton cost is then introduced into the Hamilton observation coupling equation to perform inverse integration from the terminal time of the optimal sampling point to the fuzzy time of the initial drift, and a series of global source value functions on the monitoring time series are obtained. By calculating the covariate matrix, the inverse vector and inverse gradient of a series of global source tracing value functions are determined. Based on the inverse vector and inverse gradient, the fuzzy inversion trajectory of the eDNA sample that caused the past water flow monitoring parameters to drift randomly to the optimal sampling point is reconstructed and extracted to generate the source tracing probability of the water eDNA sample reaching each sub-non-destructive monitoring area. If the source tracing probability does not exceed the maximum source tracing probability threshold, then the marine geomorphological area corresponding to the sub-non-destructive monitoring area with that source tracing probability is excluded. If the source tracing probability exceeds the maximum source tracing probability threshold, the marine geomorphological area corresponding to the sub-non-destructive monitoring area of ​​that source tracing probability is determined as the potential migration area or potential habitat area of ​​the existing rare fish belonging to the water body eDNA sample, and the first non-destructive monitoring result is output.

[0027] It should be noted that aquatic eDNA samples drift randomly under the influence of water flow characteristics. The current sampling location is not the initial location where rare fish individuals generated the aquatic eDNA sample. Current technologies lack the ability to trace the possible migration locations of rare fish from the collection location of aquatic eDNA samples using water flow characteristic data as a variable perturbation. This makes it impossible to monitor the seasonal spatial distribution of rare fish, thus hindering the exploration and protection of rare fish in target waters. To address this, this method establishes a positive hydrodynamic model that describes water movement by using water flow characteristics and potential energy. It introduces long-term stable statistical characteristics at the seasonal scale to define the physical driving basis of material transport in the water body, effectively visualizing the time-varying and periodic nature of the water flow system. This provides a realistic and reliable physical constraint mathematical platform for the random drift of aquatic eDNA samples, avoiding tracing errors based solely on statistical or empirical assumptions. Simultaneously, a state-space model of the random drift of aquatic eDNA samples under different water flow environment driving conditions that change with time is established. By coupling the wandering state-space model with the forward hydrodynamic model, the directional transport characteristics of eDNA samples driven by hydrodynamics can be characterized, forming a Hamiltonian observation coupling equation for hydrodynamic-driven eDNA sample drift. This equation enables the reverse extrapolation of the nonlinear and abrupt drift behavior of eDNA samples as water flow characteristics change, enhancing the reverse description capability. Historical seasonal water flow parameters are transformed into energy constraints and path costs of the water flow, i.e., Hamiltonian costs, and incorporated into the equation for the reverse calculation from "observed sampling points" to "unknown source areas." Each global source tracing value function reflects the rationality of the drift path under the physical constraints of water flow characteristics at a certain time. Finally, a family of reverse drift trajectories that water eDNA samples may experience under water flow drive is reconstructed, transforming the abstract source tracing value function into an interpretable spatial path, quantifying the contribution of each sub-region to the source of water eDNA samples, and significantly enhancing the stability and robustness of the sample source tracing results.

[0028] It should be noted that if the tracing probability does not exceed the maximum tracing probability threshold, it indicates that the probability of the aquatic eDNA sample originating from that sub-nondestructive monitoring area is low. This sub-area is not the original source of the sample, suggesting that the rare fish species are unlikely to migrate seasonally within this area. Therefore, the marine geomorphological area corresponding to the sub-nondestructive monitoring area with this tracing probability is excluded. Conversely, if the tracing probability exceeds the maximum tracing probability threshold, it indicates that the aquatic eDNA sample is highly likely to originate from that sub-nondestructive monitoring area. This suggests that the rare fish species are highly likely to exhibit seasonal behaviors such as migration, activity, or habitat within this area. Therefore, it is determined to be a potential migration area or potential habitat area for the existing rare fish species to which the aquatic eDNA sample belongs. This method enables the establishment of an inversion inference mechanism for samples drifting randomly with the water flow, coupled with a hydrodynamic model and a wandering state-space model. Seasonal historical water flow data is imported into this mechanism as inversion variables to trace the origin of aquatic eDNA samples from observation sampling points. This allows for the high-reliability localization of the migration and habitat of rare fish species in aquatic eDNA samples. Furthermore, this method does not require extensive underwater equipment for monitoring, thus avoiding damage to the habitat of rare fish species and achieving non-destructive monitoring of the seasonal migration of rare fish species.

[0029] Preferably, step S04 specifically includes the following steps: Multiple time-series images of existing rare fish communities were captured by underwater monitoring outposts during historical seasonal periods in various potential migration areas or potential habitats, obtained through water monitoring logs. Using community features in the image as given nodes and position changes as temporal connecting edges, a dynamic population community map of existing rare fish in each potential migration area or potential habitat area is constructed, and the historical seasonal period is divided into several seasonal sub-segments with equal intervals. A k-cluster maintenance algorithm based on incremental update mechanism is introduced to maintain the dynamic population community graph of each seasonal sub-segment by enumerating k-clusters. During the maintenance process, if the Mahalanobis distance between adjacent k-clusters is greater than the preset Mahalanobis distance, a cluster adjacency graph of local community seasonal changes is constructed based on the adjacent k-clusters, resulting in several seasonal regional cluster adjacency graph blocks. Based on the first non-destructive monitoring results, we can obtain the seasonal migration distribution network of existing rare fish species that are active or inhabit the target monitoring waters from historical periods to the current period, and at the same time, we can obtain the population survival characteristics of existing rare fish species through big data. Based on the population survival characteristics, we traverse different cluster adjacency graphs to find fish community sets under the continuous seasonal time sequence, and obtain the community connectivity components on each seasonal sub-segment. Using seasonal time series as the batch dynamic cache step size, the connected components of each community are incrementally updated based on the edge insertion-edge deletion situation field formed by the seasonal migration distribution network, so as to obtain the local community structure of existing rare fish under the seasonal expansion and decline under migration changes. The dynamic relative abundance gradient of existing rare fish species is obtained by monitoring the density clustering increment rendering of local community structure expression, and the seasonal size trend of existing rare fish species is determined based on the dynamic relative abundance gradient.

[0030] It should be noted that the population survival characteristics of rare fish species lead to local changes in their community structure during seasonal migrations. Existing monitoring methods typically require extensive underwater equipment for deep monitoring of these migration or habitat areas, which may interfere with and disrupt the ecological environment in which these rare fish thrive. Furthermore, existing methods often provide static distribution analysis of community structure, which is insufficient to meet the needs of seasonal community updates, resulting in errors and significant inaccuracies in monitoring the scale of seasonal migrations of rare fish. Therefore, this method maps temporal community images to a directed graph representation of dynamic population communities. Community features are defined as predetermined nodes, and the temporal migration and change of community features are set as temporal connecting edges, thereby revealing the incremental units of local community structure corresponding to different potential areas of rare fish populations. The seasonal sub-segments serve as a key update step size with controllable and maintainable dynamic increments, accurately capturing the micro-migration trends of local communities and improving the fine-grained quantification of the seasonal activity scale of rare fish. Subsequently, based on incremental updates, k-cluster maintenance identifies highly cohesive fish substructures (k-clusters) in each seasonal sub-segment, and maintains the k-cluster set that evolves over time in real time. The k-clusters characterize the tight community structure of the seasonal migration behavior of real rare fish populations. When inter-cluster differences are significant, a cluster adjacency graph of local community seasonal changes is constructed, and Mahalanobis distance is used to determine inter-cluster differences. Mahalanobis distance introduces a perception mechanism for the joint changes of multiple features, thereby capturing temporal abrupt changes in community structure. The seasonal migration distribution network is a dynamic migration channel network structure that changes with the seasons, i.e., the internal mesh layout changes of the migration channel architecture, responding to the temporal migration of potential migration channels or habitats. Among these, population survival characteristics provide ecological prior conditions for community evolution; that is, using population survival characteristics to potentially constrain the temporal incremental potential field of community characteristics effectively prevents connectivity inference errors that do not conform to ecological laws, aligning the dynamic changes in community structure with real fish survival mechanisms, and improving the ecological interpretability of seasonal scale monitoring results.

[0031] It should be noted that this method seeks fish community evolution paths that conform to the survival characteristics of the population within a continuous seasonal timeline. This integrates local cluster structures into stable communities at the seasonal scale. By analyzing community connectivity components, the appearance, disappearance, or aggregation relationships of rare fish individuals within the community can be clearly identified, achieving a concrete visualization of the sustained scale of seasonal migration. Finally, based on edge insertion-deletion situational field incremental community updates, with the season as the buffer step size, the method describes the emergence (edge ​​insertion), habitat decline, or emigration (edge ​​deletion) of migration paths. The visualization of local community structure vividly portrays the seasonal fish dynamics of increased community aggregation or decreased community density as migration changes occur, accurately capturing the relative abundance of rare fish during seasonal migrations. This method enables incremental observation of local community structure changes in rare fish under network alterations in seasonal migration channels, thereby monitoring the dynamic scale of seasonal migrations of rare fish and providing accurate and reliable monitoring data for resource conservation and trajectory tracking of rare fish.

[0032] A second aspect of this invention provides a non-destructive monitoring system for rare fish species based on aquatic eDNA sample identification, such as... Figure 3 As shown, the system includes: a memory 301, a processor 302, and a communication interface 303. The memory 301 includes a program for a non-destructive monitoring method for rare fish based on water eDNA sample identification. The communication interface 303 is used for data connection and communication between the memory 301 and the processor 302. When the program for the non-destructive monitoring method for rare fish is executed by the processor 302, it implements any of the steps of the non-destructive monitoring method for rare fish described above.

[0033] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification, characterized in that, Includes the following steps: S01: Obtain the historical collection distribution coordinates of eDNA samples from different water bodies in the target monitoring area, construct a hotspot model based on the historical collection distribution coordinates, analyze the historical state of the hotspot model to express the spatial sampling jump of the hotspots and predict the hotspot transfer of the current sampling distribution of the last collection, and locate the optimal sampling point. S02: Collect water eDNA samples at the optimal sampling point and perform high-throughput sequencing to obtain the actual base sequence set. Construct a base sequence recognition decision tree driven by known pattern rules to identify the rarity of the sequence pattern of the actual base sequence set. Based on the sequence rarity, import the sequence set into the gene coding database for comparison and retrieval of existing rare fish species in the target monitoring water area, and output the recognition results. S03: Construct and couple the forward hydrodynamic model with the migration state space model of water body eDNA samples as they change with different hydrodynamic environments, form the Hamiltonian observation coupling equation, import historical seasonal hydrodynamic monitoring parameters into the Hamiltonian observation coupling equation to perform fuzzy inversion solution of the initial migration area of ​​the sample, and trace the potential migration area or potential habitat of existing rare fish as the first non-destructive monitoring result output; S04: Establish a dynamic population community map of existing rare fish species by analyzing temporal community image changes. Based on the first non-destructive monitoring results, update the local abundance changes of the dynamic population community according to seasonal temporal increments of the global migration network. Analyze the seasonal scale trend of rare fish species as the output of the second non-destructive monitoring results.

2. The non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification according to claim 1, characterized in that, S01 specifically includes the following steps: Obtain the target monitoring area and water body sampling logs, and extract multiple historical collection distribution coordinates of different water body eDNA samples collected at anchor points at continuous historical moments in the target monitoring area from the water body sampling logs; A Kriging interpolation architecture is constructed. Based on historical time anchor points, sampling hotspots are fitted, drawn, and reconstructed within the Kriging interpolation architecture for each corresponding historical data collection distribution coordinate. This generates several historical sampling distribution hotspot models over a continuous time series, and the number of hotspots in each historical sampling distribution hotspot model is counted. The spatial hash misalignment value between the changes in the layout and number of sampling hotspots expressed by the historical sampling distribution hotspot model corresponding to anchor points at adjacent historical moments is calculated using a hash algorithm. Based on the spatial hash misalignment value, the sampling jump rate of the water body eDNA sample in the target monitoring water area is determined. The established operating strategy of the eDNA sampling device and the sampling distribution hotspot model of the last collection are obtained and defined as the current sampling distribution hotspot model. Based on the established operating strategy, the adjacent sampling step length when the eDNA sampling device arrives at each historical collection distribution coordinate is extracted. Based on the adjacent sampling step size and the preset spatiotemporal condition intensity index, a spatiotemporal kernel function containing an exponential decay time kernel and a Gaussian space kernel is constructed based on the sampling jump rate. The maximum likelihood method is introduced. Based on the spatiotemporal condition intensity index, the spatiotemporal kernel function is used in the maximum likelihood method to maintain and update the historical state of the current sampling distribution hotspot model until the maintenance triggers the calculation of the likelihood increment of the current hotspot state, so as to obtain the hotspot transfer prediction probability of water body eDNA samples. Only coordinates of hotspot shift prediction probabilities greater than a preset probability threshold are located and extracted in the target monitoring water area, marked as the optimal sampling points, and uploaded to the data decision terminal of the sampling equipment.

3. The non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification according to claim 1, characterized in that, S02 specifically includes the following steps: By deploying eDNA sampling equipment at the optimal sampling point, several sets of water eDNA samples of the target monitoring water area are collected. The water eDNA samples are then subjected to high-throughput sequence detection using a biological detection method to obtain the actual base sequence set of each water eDNA sample. Based on the knowledge graph of marine fishes, the system retrieves the base sequence string templates of different rare fish species that tend towards empirically standardized vectors, as well as the regulatory genomes corresponding to each base sequence string template. At the same time, the system retrieves the known pattern rules of the eDNA base sequences of rare fish species through graph retrieval. Among them, the known pattern rules include base coding rules and base pairing rules. By analyzing the driving potential of the genome, decision projection based on known pattern rules is performed on each base sequence string template. Base sequence recognition branches are divided for different rare fish species, and base sequence recognition decision trees are generated. The base sequence recognition decision trees are used to make decision recognition on the sequence patterns of actual base sequence groups to determine the sequence rarity of water body eDNA samples. If the sequence rarity is greater than the preset sequence rarity threshold, it indicates that the water body eDNA sample belongs to the eDNA fragment of a rare fish species, and the gene coding database of rare fish species is obtained. The gene coding sequence retrieval code is written into the identified water body eDNA sample using gene editing technology, and the edited water body eDNA sample is then sent to the gene coding database of rare fish for comparison and retrieval. During the comparison and retrieval process, the degree of matching between the retrieval code and the prefix index code of the known gene code corresponding to each rare fish is calculated one by one; Only the rare fish species corresponding to the known gene codes with the highest contrast fit are extracted, and these are identified as the existing rare fish species in the target monitoring water area. The identification results of the water body eDNA samples are output and uploaded to the non-destructive monitoring terminal.

4. The non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification according to claim 3, characterized in that, The process involves analyzing the driving potential of the genome to perform decision projection on each base sequence string template based on known pattern rules, dividing the base sequence recognition branches for different rare fish species, generating a base sequence recognition decision tree, and using the base sequence recognition decision tree to make decisions on the sequence patterns of actual base sequence groups to determine the sequence rarity of water body eDNA samples. Specifically, this includes the following steps: A pattern evaluation system was introduced to calculate the driving plasticity of each regulatory genome based on known pattern rules. The potential of known sequence patterns to drive base sequence string templates to a high extent was determined as high driving potential, and the potential of known sequence patterns to drive base sequence string templates to a low extent was determined as low driving potential. A slanted decision plane for prior constraints is established using known pattern rules. Each base sequence string template is projected onto the slanted decision plane based on high and low driving potential, resulting in projection values ​​and projection vectors. Based on the projection values, a recursive partitioning operation of leaf nodes for high-throughput site characterization is performed on the base sequence patterns of different rare fish according to the projection vector, generating a base sequence recognition decision tree driven by known pattern rules. The frequency of the actual base sequence group was characterized by fitting the K-mer frequency framework, and the actual K-mer frequency map of the water eDNA sample was generated. The sample pattern site characterization and corresponding site frequency intensity value of the actual base sequence group were extracted by analyzing the actual K-mer frequency map. Based on the site frequency intensity value injected into the base sequence recognition decision tree, the branch factor node performs traversal recognition of the sample pattern site representation, and obtains the pattern decision response value of the leaf node item corresponding to each high-throughput site representation. If the pattern decision response value is greater than the preset response threshold, the leaf node item corresponding to the pattern decision response value is highlighted and marked as a type 1 leaf node item; if the pattern decision response value is less than the preset response threshold, the traversal and dwell of the leaf node item is ignored or skipped and marked as a type 2 leaf node item. The pattern recognition ratio between Class I leaf node items and Class II leaf node items is calculated based on the number of nodes, and the sequence rarity of the water body eDNA sample is determined according to the pattern recognition ratio.

5. The non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification according to claim 1, characterized in that, S03 specifically includes the following steps: Obtain a marine geomorphological map of the target monitoring water area and a fixed deployment layout of underwater monitoring outposts. Based on the fixed deployment layout, divide the marine geomorphological map into N sub-non-destructive monitoring areas. The flow characteristics and potential energy of the target monitoring water area are obtained. A positive flow dynamic model is constructed based on the flow characteristics and potential energy. The past flow monitoring parameters of each sub-non-destructive monitoring area during the historical seasonal period are extracted through the water monitoring log. The past flow monitoring parameters include the flow rate, velocity, direction and wave energy of the flow. A wandering state space model is established to represent the random drift of water eDNA samples driven by different water flow environment conditions that change with time. The Hamilton algorithm is introduced to dynamically and collaboratively couple and transform the wandering state space with the forward hydrodynamic model, forming the Hamilton observation coupling equation for hydrodynamic-driven eDNA sample drift. Based on past water flow monitoring parameters, the Hamilton cost of water flow environmental variables is constructed. The Hamilton cost is then introduced into the Hamilton observation coupling equation to perform inverse integration from the terminal time of the optimal sampling point to the fuzzy time of the initial drift, and a series of global source value functions on the monitoring time series are obtained. By calculating the covariate matrix, the inverse vector and inverse gradient of a series of global source tracing value functions are determined. Based on the inverse vector and inverse gradient, the fuzzy inversion trajectory of the eDNA sample that caused the past water flow monitoring parameters to drift randomly to the optimal sampling point is reconstructed and extracted to generate the source tracing probability of the water eDNA sample reaching each sub-non-destructive monitoring area. If the source tracing probability does not exceed the maximum source tracing probability threshold, then the marine geomorphological area corresponding to the sub-non-destructive monitoring area with that source tracing probability is excluded. If the source tracing probability exceeds the maximum source tracing probability threshold, the marine geomorphological area corresponding to the sub-non-destructive monitoring area of ​​that source tracing probability is determined as the potential migration area or potential habitat area of ​​the existing rare fish belonging to the water body eDNA sample, and the first non-destructive monitoring result is output.

6. The non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification according to claim 1, characterized in that, S04 specifically includes the following steps: Multiple time-series images of existing rare fish communities were captured by underwater monitoring outposts during historical seasonal periods in various potential migration areas or potential habitats, obtained through water monitoring logs. Using community features in the image as given nodes and position changes as temporal connecting edges, a dynamic population community map of existing rare fish in each potential migration area or potential habitat area is constructed, and the historical seasonal period is divided into several seasonal sub-segments with equal intervals. A k-cluster maintenance algorithm based on incremental update mechanism is introduced to maintain the dynamic population community graph of each seasonal sub-segment by enumerating k-clusters. During the maintenance process, if the Mahalanobis distance between adjacent k-clusters is greater than the preset Mahalanobis distance, a cluster adjacency graph of local community seasonal changes is constructed based on the adjacent k-clusters, resulting in several seasonal regional cluster adjacency graph blocks. Based on the first non-destructive monitoring results, we can obtain the seasonal migration distribution network of existing rare fish species that are active or inhabit the target monitoring waters from historical periods to the current period, and at the same time, we can obtain the population survival characteristics of existing rare fish species through big data. Based on the population survival characteristics, we traverse different cluster adjacency graphs to find fish community sets under the continuous seasonal time sequence, and obtain the community connectivity components on each seasonal sub-segment. Using seasonal time series as the batch dynamic cache step size, the connected components of each community are incrementally updated based on the edge insertion-edge deletion situation field formed by the seasonal migration distribution network, so as to obtain the local community structure of existing rare fish under the seasonal expansion and decline under migration changes. The dynamic relative abundance gradient of existing rare fish species is obtained by monitoring the density clustering increment rendering of local community structure expression, and the seasonal size trend of existing rare fish species is determined based on the dynamic relative abundance gradient.

7. A non-destructive monitoring system for rare fish species based on aquatic eDNA sample identification, characterized in that, The system includes: a memory, a processor, and a communication interface. The memory includes a program for a non-destructive monitoring method for rare fish species based on aquatic eDNA sample identification. The communication interface is used for data connection and communication between the memory and the processor. When the processor executes the program for the non-destructive monitoring method for rare fish species, it implements the steps of the non-destructive monitoring method for rare fish species as described in any one of claims 1-6.

Images