River health evaluation method and system fusing eDNA and unmanned aerial vehicle remote sensing

Abstract

This invention proposes a river health assessment system and method integrating eDNA and UAV remote sensing. The method includes: First, collecting eDNA water samples from upstream and downstream of river-blocking structures and ordinary river sections, and obtaining fish species composition, relative abundance, and functional groups through sequencing and comparison to generate river section-specific data units; Second, calculating improved community fragmentation characteristic values ​​based on upstream and downstream community differences, and generating fish community connectivity fragmentation characteristics after comparison with thresholds, thereby determining the UAV verification range; Then, identifying shoreline habitats within the verification range using a pre-trained model to obtain quantity supply components, continuous supply components, and supply degree; Finally, outputting the type of fish health deficiency in the river section and restoration directions. This invention expands river health assessment from comprehensive scoring to a river section-level diagnostic method with cause identification and engineering guidance capabilities.

Description

Technical Field

[0001] This invention belongs to the field of river health assessment, and particularly relates to a method and system for river health assessment that integrates eDNA and UAV remote sensing. Background Technology

[0002] River health assessment is a crucial foundational task in river and lake management, the river chief system assessment, aquatic ecological restoration, and the construction of "happy rivers and lakes." Its assessment scope has gradually expanded from simply meeting water quality standards to encompassing multiple dimensions, including river morphology, hydrological processes, biological integrity, and social service functions. Taking the Class A river health assessment in Guangdong Province as an example, the assessment typically requires a comprehensive evaluation based on criteria such as "basin," "water," "biology," and "social service functions," considering indicators such as longitudinal connectivity, natural shoreline conditions, riverbank width, degree of illegal development and utilization of the shoreline, ecological flow, water quality, bottom sediment, fish, benthic animals, waterbirds, aquatic plants, flood control, water supply, and public satisfaction. In current engineering practices, surveys of aquatic organisms such as fish rely heavily on manual fishing, electrofishing, on-site identification, or comparison with historical data. These methods suffer from long survey cycles, significant disturbances, and high rates of missed detection for rare and invasive species. Spatial conditions such as shoreline natural conditions, shoreline vegetation cover, hard revetments, backwaters, shoals, reservoir bays, tributary inlets, and nearshore passages rely heavily on manual surveys, single image interpretations, or conventional shoreline index statistics. These methods fail to continuously reflect the nearshore space supply required for fish habitat, foraging, reproduction, and passage at the river section scale. In recent years, environmental DNA technology has been able to rapidly obtain information on fish species composition, relative abundance, and functional taxa through water sample testing, while UAV remote sensing can efficiently acquire spatial images of river channels and shorelines. These technologies have respectively improved upon the low efficiency of traditional biological surveys and the high subjectivity of shoreline surveys. However, in current river health assessments, these two types of data are often assigned to different indicators. eDNA testing results are mainly used for fish population indices, species lists, or identification of invasive species, while UAV imagery is mainly used for shoreline vegetation coverage, riverbank stability, or shoreline development and utilization. There is a lack of continuous diagnostic relationships between the two regarding the causes of abnormal fish community conditions. In other words, while existing methods can conclude that fish indicators in a certain river section are low, longitudinal connectivity is insufficient, or the natural condition of the shoreline is weak, it is difficult to further determine whether changes in fish communities are caused by the disruption of upstream and downstream biological connectivity due to dam structures, insufficient supply of shoreline habitats such as aquatic plant belts, shoals, backwaters, naturally vegetated shorelines, and gravel beaches, or a combination of both. Especially in rivers like the Wujiang River, which have multiple dam structures, uneven distribution of shoreline vegetation, and significant spatial differences in fish communities, if the method of "monitoring multiple indicators separately, assigning scores separately, and diagnosing based on the lowest score" is still used, the evaluation results are likely to remain at the level of indicator description, making it difficult to provide clear river section locations and causes for fish passage facility renovation, dam ecological scheduling, shoreline microhabitat restoration, and riparian zone restoration.The existing Wujiang River evaluation project has divided the Wujiang River into 7 evaluation sections and conducted a comprehensive evaluation based on 22 indicators of 4 criteria layers for Class A rivers. The results show that the Wujiang River is generally a Class II healthy river, but there are still problems such as insufficient longitudinal connectivity, sediment pollution, large variability in flow process, uneven distribution of shoreline vegetation, and invasive alien species. This indicates that simply obtaining a comprehensive score cannot fully support the detailed diagnosis and targeted restoration at the river section level.

[0003] Therefore, there is an urgent need for a method that can incorporate fish community results reflected by eDNA, upstream and downstream spatial relationships of dam structures, and shoreline habitat supply conditions reflected by UAV remote sensing into the same diagnostic chain. This would enable river health assessment to not only determine the health status of river sections but also explain the main sources of fish community anomalies and translate the diagnostic results into restoration guidelines for specific river sections, specific dam structures, and specific shoreline spatial units. Summary of the Invention

[0004] The purpose of this invention is to propose a method and system for river health assessment that integrates eDNA and UAV remote sensing, in order to solve the above-mentioned problems.

[0005] To achieve the above objectives, a method for river health assessment integrating eDNA and UAV remote sensing is provided in a first aspect of the present invention, the method comprising: S1. Obtain eDNA water samples from upstream and downstream of the dam structure and from ordinary river sections within the target evaluation river section. Sequencing and comparison are used to obtain fish species composition, relative abundance, and functional group labels, generating river section-specific eDNA fish community data units. Each sampling point corresponds to one river section-specific eDNA fish community data unit. Each river segment-based eDNA fish community data unit includes a river segment number, a river-blocking structure number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field. S2. Based on the eDNA fish community data units of the upstream and downstream sides of the same dam structure, an improved community fragmentation feature value is calculated by combining the corresponding differences in species and relative abundance. This community fragmentation feature value is used to represent the degree of fragmentation of fish communities on both sides of the same dam structure, and is compared with a preset judgment threshold to generate fish community connectivity fragmentation features. The fish community connectivity fragmentation features are recorded with the dam structure as the recording object, including the river segment number, dam structure number, upstream and downstream community difference summary, community fragmentation feature value, degree of fragmentation, affected fish functional groups, and shoreline habitat type to be verified. S3. Determine the target shoreline verification range of UAV remote sensing based on the fish community connectivity and discontinuity characteristics; within the target shoreline verification range, input the UAV images collected by the UAV into the pre-trained shoreline habitat recognition model to identify the target river segment image blocks and generate UAV remote sensing shoreline habitat supply information; the UAV remote sensing shoreline habitat supply information includes quantity supply component, continuous supply component and shoreline habitat supply degree. S4. Based on the information on shoreline habitat supply obtained from the UAV remote sensing, output the types of health deficiencies of fish in the river section and the directions for restoration.

[0006] Furthermore, the fish functional group labels include flowing water fish, migratory fish, local sensitive fish, still water adapted fish, benthic fish, aquatic plant dependent fish, and alien fish; The relative abundance is specifically: Using all valid sequence readings of fish within the same sampling point as a benchmark, the valid sequence readings of a certain fish species are converted into its proportion in the valid sequences of fish at that sampling point, and the current proportion is written into the relative abundance field.

[0007] Furthermore, the improved community fragmentation characteristic value is calculated by combining the corresponding differences in species and relative abundance, specifically as follows: The eDNA fish community data units of the river section upstream and downstream of the same dam structure are compared in pairs to generate a species control list on both sides. The relative abundance difference of each fish species is calculated, and the corresponding connectivity-sensitive weights are introduced according to the fish functional group labels for weighting. Combined with the reservoir bay direction labels, the community breakage feature value is obtained after normalization, and then compared with the preset judgment threshold to determine the degree of breakage. The bay orientation label is generated based on the enrichment of upstream still-water adapted fish or alien fish and downstream flowing-water fish or local sensitive fish. The connectivity sensitivity weights enable flowing water fish, migratory fish, and locally sensitive fish to make a greater contribution to fracture detection.

[0008] Furthermore, the species comparison list on both sides records the common species detected on both sides, the different species detected on the upstream side, and the different species detected on the downstream side, and simultaneously retains the functional group label corresponding to each fish species or reliable classification level.

[0009] Furthermore, the river segment number and the dammed structure number are derived from the river segment-based eDNA fish community data unit; the upstream and downstream community difference summary is formed by comparing the upstream and downstream fish species composition field and relative abundance field of the same dammed structure; the community fragmentation feature value is formed by the upstream and downstream relative abundance difference, connectivity sensitivity weight, and reservoir bay orientation label; the fragmentation degree is formed by the community fragmentation feature value corresponding to the judgment threshold in the project configuration file; the affected fish functional groups are determined by the functional group label field and its upstream and downstream change direction; the shoreline habitat type to be verified is generated corresponding to the affected fish functional groups.

[0010] Furthermore, the cluster fracture characteristic value of each river-blocking structure is compared with the preset low-level judgment threshold, medium-level judgment threshold and high-level judgment threshold to generate a fracture degree label. When it falls into the low-level judgment range, it is recorded as low fracture degree; when it falls into the medium-level judgment range, it is recorded as medium fracture degree; and when it falls into the high-level judgment range, it is recorded as high fracture degree. Then there is, The determination of the target shoreline verification range by UAV remote sensing based on the connectivity and discontinuity characteristics of the fish community is specifically as follows: If the degree of fracture is as described, then the near-shore continuous channel covering the upstream side, downstream side and both sides of the river-blocking structure is covered. If it is of the medium degree of fracture, then the nearshore shoreline zone related to the functional groups of affected fish on both the upstream and downstream sides will be covered; If the degree of fracture is low, then the nearshore area within the same river segment that corresponds to the habitat type of the shoreline to be verified is covered.

[0011] Furthermore, the pre-trained shoreline habitat identification model adopts an encoder-decoder semantic segmentation structure, with the input being an orthophoto patch of the target river segment and the output being a pixel-by-pixel shoreline habitat category, resulting in a trained model with fixed parameters. After the pre-trained shoreline habitat recognition model completes pixel-by-pixel recognition, the following steps are performed: The boundary line between the water body and the land area is extracted from the identification results as the shoreline baseline, and a nearshore analysis zone is generated along the shoreline baseline. Natural vegetation shorelines, aquatic plant belts, shoals, gravel beaches, backwater bays, slow-flowing reservoir bays, channel-type nearshore spaces, tributary confluences, hard revetments, and bridge pier disturbance areas are converted into shoreline habitat units to locate specific missing or insufficient shoreline units. The quantity supply component is obtained by weighting the supply ratio of the shoreline habitat type to be verified according to the demand weights in the pre-set fish functional group-shoreline habitat demand table; the continuous supply component is obtained by the proportion of continuous habitat segments on the shoreline baseline in the target river section; the shoreline habitat supply degree is obtained by integrating the quantity supply component, the continuous supply component, and the functional group continuous demand weights.

[0012] Furthermore, based on the drone-sensed shoreline habitat supply information, the output of the types of fish health deficiencies and restoration directions for the river section specifically includes: Based on the aforementioned shoreline habitat units, diagnostic objects are established according to river segment number, dam number, upstream location attributes, downstream location attributes, and affected fish functional groups; Based on the diagnosed object, the corresponding quantity supply component, continuous supply component, and shoreline habitat supply degree are obtained, and the shoreline habitat supply gap is determined according to the shoreline habitat supply degree. For the insufficient supply of shoreline habitat, a fracture degree guidance coefficient is obtained and compared with the connectivity pressure determination threshold in the project configuration file to generate a connectivity pressure label and the corresponding merged longitudinal connectivity pressure. By merging the shoreline habitat supply gap and the longitudinal connectivity pressure and deducting the overlapping contribution of the two pressure sources, the total pressure is kept within a stable proportional scale to obtain the short-board pressure value. This value is then compared with a preset level threshold to determine the short-board pressure level and repair priority. Based on the fault degree guidance coefficient, shoreline habitat supply, quantity supply component, continuous supply component, and short-board pressure value, the types of short-boards in fish health in river sections are output, including longitudinal connectivity fault type, shoreline habitat insufficient supply type, and barrier-habitat composite stress type. Different restoration strategies are proposed based on the types of health deficiencies of fish in different river sections.

[0013] Furthermore, the repair direction includes: The longitudinal connectivity disruption type of short-term bottleneck corresponds to the installation of fish passage facilities, improvement of fish passage connectivity, and ecological scheduling of sluice gates and dams; the shoreline habitat shortage type of short-term bottleneck is determined by the type of missing shoreline habitat, and corresponds to the restoration of aquatic plant belts, the creation of slow-flowing backwater bays, the restoration of natural vegetation shorelines, the restoration of gravel beaches, the restoration of shallow beaches, the ecological transformation of hard revetments, and the dredging of channel-type nearshore spaces; the barrier-habitat composite stress type of short-term bottleneck corresponds to the combined implementation of connectivity improvement and shoreline habitat restoration.

[0014] A second aspect of the invention provides a river health assessment system integrating eDNA and UAV remote sensing, the system comprising: The eDNA fish community processing module is used to acquire eDNA water samples from upstream and downstream of the river-blocking structures and ordinary river sections within the target evaluation river section. After sequencing and comparison, fish species composition, relative abundance, and functional group labels are obtained, generating river section-specific eDNA fish community data units. Each sampling point corresponds to one river section-specific eDNA fish community data unit. Each river segment-based eDNA fish community data unit includes a river segment number, a river-blocking structure number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field. The connectivity and discontinuity identification module is used to calculate improved community discontinuity feature values ​​based on the eDNA fish community data units of the upstream and downstream sides of the same river-blocking structure, by combining the corresponding differences in species and relative abundance. These community discontinuity feature values ​​are used to represent the degree of discontinuity of fish communities on both sides of the same river-blocking structure, and are compared with a preset judgment threshold to generate fish community connectivity and discontinuity features. The fish community connectivity and discontinuity features are recorded with the river-blocking structure as the recording object, including the river segment number, the river-blocking structure number, the summary of differences between upstream and downstream communities, the community discontinuity feature value, the degree of discontinuity, the functional groups of affected fish, and the shoreline habitat type to be verified. The UAV remote sensing shoreline interpretation and shoreline habitat supply calculation module is used to determine the target shoreline verification range of UAV remote sensing based on the fish community connectivity and discontinuity characteristics; within the target shoreline verification range, the UAV-collected images are input into a pre-trained shoreline habitat recognition model to identify the target river segment image blocks and generate UAV remote sensing shoreline habitat supply information; the UAV remote sensing shoreline habitat supply information includes quantity supply component, continuous supply component, and shoreline habitat supply degree. The shortcoming diagnosis and repair direction generation module is used to output the type of shortcoming in the health of fish in the river section and the repair direction based on the habitat supply information of the shoreline sensed by the UAV.

[0015] The beneficial technical effects of the present invention are at least as follows: The main innovation of this invention lies in organizing eDNA fish community detection and UAV remote sensing shoreline habitat identification into a seamless diagnostic process, focusing on the core issue of "river section-level identification and restoration guidance for the causes of abnormal fish communities," rather than simply superimposing the two as parallel data sources.

[0016] Specifically, this invention first binds eDNA detection results with the evaluated river section, the number of the dammed structure, the upstream or downstream positional relationship, fish species composition, relative abundance, and functional group labels to form river section-specific fish community data that supports upstream-downstream comparison. Then, based on differences in fish communities on both sides of the same dammed structure, changes in connectivity-sensitive functional groups, and upstream reservoir bay enrichment characteristics, it determines the connectivity fault characteristics of the fish community and further identifies the affected fish functional groups and the shoreline habitat types to be verified. Finally, it uses these fault characteristics to limit the verification scope and identification objects of UAV remote sensing, enabling the UAV orthophotos to focus on extracting natural vegetation shorelines corresponding to the affected fish functional groups. The shoreline spatial units include aquatic plant belts, shoals, gravel beaches, backwater bays, slow-flowing reservoir bays, channel-type nearshore spaces, hard revetments, and bridge pier disturbance areas. This information forms shoreline habitat supply information for specific fish functional groups. Finally, the longitudinal connectivity pressure and shoreline habitat supply gap are combined and judged in the same diagnostic object, outputting the shortcomings of longitudinal connectivity failure, insufficient shoreline habitat supply, and barrier-habitat composite stress. Correspondingly, restoration directions are formed for fish passage facility construction, fish passage connectivity improvement, dam ecological scheduling, aquatic plant belt restoration, slow-flowing backwater bay creation, natural vegetation shoreline restoration, gravel beach restoration, shoal restoration, ecological transformation of hard revetments, and channel-type nearshore space dredging. Compared to traditional methods, this invention does not rely solely on static judgments based on the number of dams, the presence or absence of fish passage facilities, fish population indices, or shoreline vegetation coverage. Instead, it uses "abnormalities in fish functional groups shown by eDNA" as the triggering condition and identification direction for UAV remote sensing shoreline interpretation, and "shoreline habitat supply identified by UAVs" as spatial evidence explaining the causes of fish anomalies. This forms a closed diagnostic chain of "biological results—barrier evidence—spatial supply—shortcoming type—restoration direction." Through this design, this invention enhances the intrinsic connection between fish indicators, longitudinal connectivity indicators, and shoreline indicators based on the existing river health assessment system. It overcomes the problems of fragmented application of eDNA and UAV remote sensing in existing technologies, separate scoring of multi-source data, difficulty in distinguishing the causes of fish degradation, and difficulty in accurately corresponding river section restoration measures. This expands river health assessment from comprehensive scoring to a river section-level diagnostic method with cause identification and engineering guidance capabilities.

[0017] This invention is not simply about digitizing manual evaluation rules, nor is it merely about outputting river health management recommendations. The system actually processes eDNA sequence alignment results with spatial coordinates of sampling points and batch information, spatial records of the evaluated river section and dam structures, UAV orthophotos and their semantic segmentation results, fish functional groups-shoreline habitat requirements tables, and project configuration files. Through data binding, pairwise comparison, image interpretation, scaling calculations, and diagnostic object generation, the system obtains structured results that can be used in river health evaluation reports, restoration plan development, or spatial layer systems. Therefore, the technical effects of this invention are reflected in improving the spatial registration accuracy of multi-source ecological monitoring data, reducing the uncertainty of relying solely on human experience to determine the causes of fish anomalies, and enabling restoration efforts to be tailored from comprehensive scores to specific river sections, dam structures, and shoreline spatial units. Attached Figure Description

[0018] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0019] Figure 1 This is a flowchart of a river health assessment method integrating eDNA and UAV remote sensing, according to Embodiment 1 of the present invention.

[0020] Figure 2 This is a schematic diagram showing the relative abundance of major fish taxa within each river section / habitat unit in Embodiment 2 of the present invention. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0022] Example 1

[0023] like Figure 1 As shown in the embodiment of the present invention, the method for river health assessment integrating eDNA and UAV remote sensing includes: S1. Obtain eDNA water samples from upstream and downstream of the dam structure and from ordinary river sections within the target evaluation river section. Sequencing and comparison are used to obtain fish species composition, relative abundance, and functional group labels, generating river section-specific eDNA fish community data units. Each sampling point corresponds to one river section-specific eDNA fish community data unit. Each river segment-based eDNA fish community data unit includes a river segment number, a river-blocking structure number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field.

[0024] Specifically, in this embodiment, the system first forms a segmented eDNA fish community data unit. This data unit is used to carry the fish community detection results within the same evaluation segment, the upstream and downstream correspondence of the dam structure, and the fish functional group information, enabling direct comparison of fish community differences on both sides of the same dam structure. The basic data received by the system includes the evaluation segment division results, the list of dam structures, field-collected eDNA water samples, and a fish reference database. The evaluation segment division results are formed in the early stage of river health assessment, recording the segment number and the start and end range of the segment; the list of dam structures is confirmed by the water conservancy project ledger, on-site survey records, and UAV patrol results, recording the dam structure number, the segment to which it belongs, and its spatial location; the eDNA water samples are obtained by on-site water sampling, with the sampling object being the river water body; the fish reference database is formed by compiling local historical fish lists, existing survey reports, publicly available fish barcode sequences, and expert verification data, and is used to identify the sequencing sequences as fish species or reliable taxonomic levels.

[0025] To ensure the clarity of the source of subsequent fields, the system simultaneously establishes a sampling record table, an engineering spatial table, and a fish ecological attribute table upon receiving basic data. The sampling record table includes at least the water sample number, sampling point number, sampling time, sampling coordinates, sampling water layer, sampling volume, parallel sample number, negative control number, positive control number, sequencing batch number, and laboratory test report number. The engineering spatial table includes at least the river segment number, the starting and ending station numbers or spatial range of the river segment, the dam number, the coordinates of the dam, the upstream determination range, and the downstream determination range. The fish ecological attribute table includes at least the fish name or reliable taxonomic level, local occurrence records, functional group label, whether it is an introduced fish species, and the data source used for label determination. All the above tables use a unified coordinate system and a unified numbering rule within the same evaluation period, enabling the matching of eDNA results, dam locations, and subsequent UAV imagery according to the same river segment index.

[0026] Furthermore, during sampling deployment, the system uses the evaluation river segment as the basic organizational unit and sets up ordinary river segment points within each evaluation river segment to record the background state of fish communities in that river segment that are not directly affected by the hydraulic disturbance upstream and downstream of the current dam structure. Ordinary river segment points are placed in locations that can represent the fish activity status of that river segment, specifically in slow-flowing areas near the natural shoreline, the edge of the main channel, near aquatic vegetation zones, or near existing biological monitoring sections. When a dam structure exists within the river segment, ordinary river segment points are placed in locations that avoid direct impoundment, discharge disturbance, and direct mixing effects from tributaries. For river segments with dam structures, the system simultaneously sets up upstream and downstream sampling points around the same dam structure, creating a paired relationship between the two sampling points in the data structure. The upstream sampling point is selected at the junction of the main stream and the near-shore slow flow or in a stable area of ​​water exchange in the reservoir area, while the downstream sampling point is selected in the near-shore area after the outflow from the dam has stabilized. The degree of shoreline disturbance and water exchange conditions of the two sampling points are kept comparable.

[0027] For example, in the implementation scenario of a river section where a hydropower station dam is located, the upstream sampling point is set at the junction of the main stream and the slow-moving near-bank upstream of the dam, the downstream sampling point is set at the near-bank where the water flow has stabilized downstream of the dam, and the ordinary river section points are set at locations far away from the direct hydraulic disturbance of the dam within the same evaluation river section. When there is a tributary confluence or obvious sewage outlet near the dam, the upstream, downstream, and ordinary river section points are all set at locations that avoid direct mixing effects, so that the collected water samples can respectively represent the fish community status on both sides of the dam structure and the background fish community status of the river section.

[0028] During on-site sampling, staff used water samplers or cleaned sampling containers to collect representative water samples, and simultaneously recorded the river section number, dam number, sampling point location type, and corresponding relationship in the sampling record. The sampling point location type was fixed and recorded as one of three states: upstream of the dam, downstream of the dam, or ordinary river section. For parallel water samples from the same sampling point, the laboratory merged them into the fish community results for that sampling point after completing the testing. After sampling, the water samples were filtered through a membrane to obtain environmental DNA vectors. The membrane was then cryopreserved and sent to the laboratory for DNA extraction, fish macrobarcode amplification, and high-throughput sequencing. The same filtration, extraction, amplification, and sequencing procedures were used for each sampling point to ensure that fish community results from different sampling points could be included in the same comparison system.

[0029] Furthermore, to mitigate the risk of insufficient disclosure due to sampling contamination, batch variations in sequencing, or incidental detections, the system performs quality control before generating fish community results. Sequences that consistently appear in the negative control are excluded as suspected contamination sequences. Sequencing batches that fail to meet the expected detection criteria in the positive control are marked as requiring verification. For parallel samples from the same sampling point, only classification results meeting the minimum valid readings, minimum number of replicates, or minimum relative abundance conditions specified in the project configuration file are included in the official community results. Results that do not meet the quality control conditions are not directly used for breakage assessment but are recorded in the data unit as pending verification or low-confidence detection, avoiding the direct interpretation of incidental sequence readings as true differences in fish distribution.

[0030] Further, after sequencing, the system performs quality screening on the sequences, generates fish species composition, and retains valid sequences that can be stably classified as fish. Contaminating sequences, low-quality sequences, and sequences that cannot be classified as fish in the negative control are marked by the system and excluded from the fish community data unit. Valid sequences are compared in the fish reference database. Sequences that can be stably identified at the species level are written into the species name field; those that can be stably identified at the genus or family level are written into the reliable taxonomic level field. For closely related fish with small barcode differences, the system retains stable identification results and assigns corresponding functional group labels based on the local fish ecological habits table. In implementation scenarios where closely related Cyprinidae fish can only be stably identified at the genus level, the system records them as undetermined species of that genus and fills in the functional group label according to the ecological attributes stably corresponding to that genus in local data.

[0031] Furthermore, the system then generates a relative abundance field. This field is derived from the statistical concept of relative frequency, using the proportion of a certain type of observation within the same set of observations to represent its community occupancy level.

[0032] Understandably, when applying this concept to eDNA fish community processing, this implementation uses all valid sequence reads of fish within the same sampling point as a benchmark. The valid sequence reads of a particular fish species are converted into its proportion within the valid sequences of fish at that sampling point, and this proportion is written into the relative abundance field. This processing allows for comparison of fish community composition on the same scale even if the total sequencing reads differ between different sampling points. In the sequencing alignment results of one sampling point, there are 5,000 valid sequence reads for all fish species, including 3,200 for species A, 800 for species B, and 1,000 for the remaining species. The system records the relative abundance of species A as 64% and the relative abundance of species B as 16%. When subsequently comparing fish communities upstream and downstream of the same dam structure, the system uses the relative abundance field for comparison, rather than the raw sequencing reads.

[0033] Furthermore, relative abundance is stored internally as a dimensionless scale value between zero and one; the percentage representation in the manual is for ease of understanding only. If all valid fish sequence readings at a sampling point are zero, or if the valid sequence readings are below the minimum valid sample condition specified in the project configuration file, the system does not calculate the relative abundance of that sampling point. Instead, it marks the sampling point as invalid and requires resampling or manual verification. If a fish species is detected only on the upstream or downstream side, the system records the relative abundance of that species as zero on the other side and retains the undetected mark. This processing ensures that subsequent upstream-downstream comparisons, weighting, and stress calculations all use the same scale, preventing the mixing of raw readings and scale values.

[0034] Furthermore, in the functional group labeling stage, the system calls a pre-organized table of local fish ecological habits and writes functional group labels for each fish species or reliable taxonomic level. Functional group labels include flowing-water fish, migratory fish, locally sensitive fish, still-water adapted fish, benthic fish, aquatic plant dependent fish, and introduced fish. Flowing-water fish, migratory fish, and locally sensitive fish are used to subsequently determine longitudinal connectivity breaks; benthic fish and aquatic plant dependent fish are used to subsequently correlate with habitat supply information from UAV remote sensing of the shoreline; still-water adapted fish and introduced fish are used to identify community substitution or enrichment phenomena on the reservoir bay side upstream of the dam structure.

[0035] At a sampling point upstream of a river-blocking structure, the system recorded a relatively high abundance of still-water adapted fish, while at a sampling point downstream, the relative abundance of flowing-water fish was higher. The system recorded this phenomenon in the data units of both sampling points and used it for connectivity fault feature identification in the next step. At a typical river section, the system recorded a relatively low abundance of aquatic plant dependent fish. The system recorded this phenomenon in the data unit of this typical river section and correlated it with the identification results of aquatic plant zones and slow-flowing backwaters in subsequent shoreline habitat supply analysis.

[0036] Furthermore, the local fish ecological habits table is compiled into a versioned file based on historical fish lists, local aquatic ecological survey data, publicly available literature, and expert verification records before the evaluation begins. The system locks this version within the same evaluation cycle. When a fish taxonomic unit possesses multiple ecological attributes, the system writes the primary functional group according to the preset priority in the project configuration file, while retaining the auxiliary functional group field. For classification results where only genus or family levels can be identified and the functional group is unstable, the system does not forcibly assign highly sensitive labels but instead writes them into functional groups to be verified, reducing their contribution or requiring manual review during subsequent fault feature generation. This process avoids a lack of definitive basis for subsequent connectivity fault judgments and shoreline-habitat matching due to unclear functional group origins or arbitrary label changes.

[0037] The system ultimately integrates the location relationships of sampling points, fish species composition, relative abundance, and functional group labels into a fixed-structure river segment-based eDNA fish community data unit. Each sampling point corresponds to a data unit, which includes a river segment number, a dam number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field. For ordinary river segment points, the system records the ordinary river segment point in the sampling point location type field; for sampling points upstream and downstream of a dam, the system establishes a pairing relationship through the same dam number, enabling direct retrieval and comparison of data from both sides of the same dam. This implementation outputs the river segment-based eDNA fish community data unit as input for subsequent identification of fish community connectivity discontinuity features.

[0038] S2. Based on the eDNA fish community data units of the upstream and downstream sides of the same dam structure, an improved community fragmentation feature value is calculated by combining the corresponding differences in species and relative abundance. This community fragmentation feature value is used to represent the degree of fragmentation of fish communities on both sides of the same dam structure, and is compared with a preset judgment threshold to generate fish community connectivity fragmentation features. The fish community connectivity fragmentation features are recorded with the dam structure as the recording object, including the river segment number, dam structure number, upstream and downstream community difference summary, community fragmentation feature value, degree of fragmentation, affected fish functional groups, and shoreline habitat type to be verified.

[0039] Specifically, the data includes fields for the dam structure number, sampling point location type, fish species composition, relative abundance of fish species, and fish functional group label. The system aggregates data units according to the dam structure number. Data units with the same dam structure number and sampling point location type (upstream) are written to the upstream community archive; data units with the same dam structure number and sampling point location type (downstream) are written to the downstream community archive; and data units with sampling point location type (ordinary river section) within the same evaluated river are written to the background community archive. The upstream and downstream community archives form paired comparison objects on both sides of the same dam structure. The background community archive records the common occurrence status and relative abundance fluctuation range of each fish functional group in the ordinary river section.

[0040] Understandably, this aggregation process directly uses the species composition, relative abundance, and functional group labels generated in the previous step, without re-performing eDNA sequencing and alignment, or reclassifying fish functional groups. This allows the processing object of this step to focus on identifying differences in fish communities upstream and downstream of the same dam structure.

[0041] Furthermore, the system reads the fish species composition fields from the upstream and downstream sides of each paired community file and generates a species comparison list for both sides. The species comparison list records the common species detected on both sides, the differential species detected on the upstream side, and the differential species detected on the downstream side, and simultaneously retains the functional group label corresponding to each fish species or reliable taxonomic level.

[0042] For reliable taxonomic levels recorded in the previous step at the genus or family level, the system uses this taxonomic level for comparison and employs the functional group labels already written in the previous step. The system then reads the relative abundance fields of each fish species on the upstream and downstream sides, forming an upstream relative abundance table and a downstream relative abundance table; when a fish species appears only in the community archive on one side, the system records it as undetected on the other side and incorporates this status into subsequent difference assessments.

[0043] Through this process, whether a species is detected, the difference in relative abundance between the two sides, and the functional group attributes of fish are integrated into the same comparison record, enabling the system to further determine whether the difference is concentrated in the common fish community or in flowing fish, migratory fish, local sensitive fish, still water adapted fish, or alien fish.

[0044] Furthermore, the system employs an improved community fragmentation eigenvalue to represent the degree of fragmentation between fish communities on either side of the same dam structure. This eigenvalue is initially derived from the Bray-Curtis community difference method in ecology, whose basic idea is to reflect the degree of difference between two communities by using the proportion of the difference in species abundance between the two communities to the overall abundance scale of the two communities.

[0045] Furthermore, in this implementation, when applying this idea to upstream and downstream comparisons of eDNA fish communities, the relative abundance field generated in the previous step is used as the basis for community abundance, ensuring that differences in total sequencing volume at different sampling points do not directly affect the comparison results. Simultaneously, connectivity-sensitive weights are introduced to give higher contributions from flowing-water fish, migratory fish, and locally sensitive fish in breakage assessment. A reservoir-bay orientation label is also introduced to reflect the typical barrier scenario of enrichment of still-water adapted fish or introduced fish on the upstream side of the dam structure, and retention of flowing-water fish or locally sensitive fish on the downstream side. The system first calculates the difference between the upstream and downstream relative abundance for each fish species or reliable taxonomic level. Then, based on the functional group label of the species, the connectivity-sensitive weights are invoked to convert the difference magnitude into a community difference contribution with ecological indicative significance. The system then summarizes the community difference contributions of all fish species to obtain the functionally weighted difference of paired community profiles.

[0046] The connectivity sensitivity weights are determined by the local fish ecological habits table and project configuration file, and are written into the system before the evaluation begins and remain fixed throughout the evaluation. Flowing-water fish, migratory fish, and locally sensitive fish correspond to the high-sensitivity level; benthic fish and aquatic plant-dependent fish correspond to the medium-sensitivity level; and still-water adapted fish, introduced fish, and other fish correspond to the basic level. This weighting is based on engineering facts from river longitudinal connectivity evaluation: for fish species heavily dependent on flow continuity, migratory channels, and natural river connectivity, relative abundance differences on both sides of a dam structure have a stronger indicative significance of the barrier effect; changes in still-water adapted fish and introduced fish are more used to identify the direction of upstream reservoir enrichment. When generating community fault characteristic values, the system uses these fixed weights to distinguish the upstream and downstream differences in the contributions of different functional groups, ensuring that the same relative abundance difference has different diagnostic significance in different functional groups.

[0047] The bay formation direction label is generated by the system based on two types of phenomena. The first type is when the relative abundance of still-water adapted fish or introduced fish species is higher on the upstream side than on the downstream side, meeting the bay enrichment criteria in the project configuration file. The second type is when downstream flowing-water fish, migratory fish, or locally sensitive fish species maintain a high detection rate, meeting the downstream retention criteria in the project configuration file. When both types of phenomena occur simultaneously, the system writes the bay formation direction label to a high-level state; when only one type of phenomenon occurs, it writes to a mid-level state; and when no clear bay formation direction is formed on either side, it writes to a basic state. The system then converts the bay formation direction label into a direction correction value according to the correction intensity in the project configuration file, and this correction value, along with the function-weighted difference value, participates in the generation of community fracture characteristic values. The correction intensity is determined before the evaluation begins based on background community archives, local fish ecological behavior tables, and experience from existing river health evaluation projects, and remains fixed within the same evaluation period.

[0048] Furthermore, when generating community fragmentation characteristic values, the system also generates a community size benchmark for normalization. This benchmark is determined by the relative abundance and connectivity-sensitive weights of all participating fish species on both the upstream and downstream sides, and a normalization benchmark term consistent with the intensity of the reservoir bay orientation correction is added. The system combines the function-weighted difference and orientation correction, and then performs scale unification processing with the community size benchmark to obtain community fragmentation characteristic values ​​that can be compared between different dam structures. The higher the characteristic value, the more concentrated the differences in fish communities on both sides of the same dam structure are, the more concentrated they are, reflecting the weakening of longitudinally connected sensitive groups and the enrichment of upstream reservoir bays; the lower the characteristic value, the weaker the differences in fish communities on both sides. The inputs throughout the generation process all come from the data units in the previous step and the project configuration file fixed before the evaluation began. The generated results are used for subsequent fragmentation degree determination.

[0049] The aforementioned functional weighted difference, direction correction, community size benchmark, and community fracture characteristic value are all calculated using dimensionless proportions or dimensionless weights.

[0050] The connectivity sensitivity weight, direction correction strength, and judgment threshold are all given in the project configuration file. Their values ​​remain unchanged within the same evaluation period and cannot be directly added to unit quantities such as the original sequence readings, area, and length. If there are no valid fish community results upstream and downstream of the same dam structure, or if the community size baseline required for normalization is zero, the system does not generate community breakage characteristic values. Instead, it records the dam structure as having insufficient data and outputs a prompt for supplementary sampling or manual verification. If only one side has valid results, the system can generate a single-sided abnormal record, but its breakage degree must be marked as low confidence and not directly used as a high-breakage conclusion.

[0051] For example, in the evaluation record of a hydroelectric dam, after the system read the upstream and downstream community files, it found that: migratory fish were not detected on the upstream side, but their relative abundance was 35% on the downstream side; flowing-water fish had a relative abundance of 10% on the upstream side and 35% on the downstream side; still-water adapted fish had a relative abundance of 55% on the upstream side and 20% on the downstream side; and invasive fish had a relative abundance of 35% on the upstream side and 10% on the downstream side.

[0052] The system calls connectivity sensitivity weights based on functional group labels, taking the upstream and downstream differences of migratory and flowing fish as high-sensitivity difference contributions, and the upstream and downstream differences of still-water adapted fish and alien fish as basic difference contributions; at the same time, based on the enrichment of still-water adapted fish and alien fish on the upstream side and the retention of migratory and flowing fish on the downstream side, the reservoir bay orientation label is written into the high-level state.

[0053] The system generated community fracture characteristic values ​​based on this, placing them within the high-level determination range and recording the fracture degree of the river-blocking structure as high. The affected fish functional groups were recorded as migratory and flowing-water fish. This record was subsequently used to define the scope of the UAV remote sensing interpretation and key shoreline habitat types for the next processing stage.

[0054] Furthermore, the system combines background community archives to determine the range of fracture severity. Background community archives consist of data units from ordinary river sections, recording the common detection status and relative abundance fluctuation range of fish functional groups in the same evaluated river section. At the start of the evaluation, the system sets low, medium, and high thresholds based on the background community archives, local fish ecological behavior tables, and experience from existing river health assessment projects, and writes these thresholds into the project configuration file. The system compares the community fracture characteristic value of each dam structure with the aforementioned thresholds; values ​​falling within the low threshold are recorded as low fracture severity, those within the medium threshold as medium fracture severity, and those within the high threshold as high fracture severity. This determination method brings the eDNA community differences of multiple dam structures within the same river to a unified scale and allows natural fluctuations in ordinary river sections to participate in the fracture severity classification.

[0055] Understandably, the low, median, and high thresholds are not adjusted temporarily after the diagnostic results are generated, but rather determined and solidified before the evaluation begins based on the natural fluctuation range of the background community archive, local fish ecological habits tables, and project experience. The system saves the threshold version number, activation time, and applicable river type; when the evaluation object, sampling season, or sequencing process changes significantly, the system creates a new configuration version instead of directly overwriting the original threshold. Therefore, the determination of the degree of community fragmentation has a traceable parameter source, avoiding arbitrary interpretation of the same feature value in different records.

[0056] Furthermore, while determining the degree of fragmentation, the system extracts the functional groups of affected fish. The system first reads the detection status and relative abundance changes of flowing-water fish, migratory fish, and native-sensitive fish in paired community files, recording the functional groups that contribute significantly to the community fragmentation characteristic values ​​and are sensitively related to vertical connectivity as the main affected functional groups. Then, it reads the upstream and downstream changes of benthic fish and aquatic plant-dependent fish, recording them as functional groups to be verified for shoreline habitat supply. Based on the affected fish functional groups, the system generates shoreline habitat types to be verified: when the main affected functional group is migratory fish, the shoreline habitat type is recorded as continuous natural shoreline, shoals, and channel-type nearshore spaces; when the functional group is aquatic plant-dependent fish, the shoreline habitat type is recorded as aquatic plant belts, slow-flowing backwaters, and naturally vegetated shorelines; when the functional group is benthic fish, the shoreline habitat type is recorded as gravel beaches, shoals, and the proportion of hard revetments. Therefore, the fish functional group anomalies identified by eDNA are transformed into shoreline spatial units that need to be identified in the next stage of UAV remote sensing.

[0057] The system ultimately outputs fish community connectivity discontinuity features. These features, recorded using river-blocking structures as the object, include river segment number, river-blocking structure number, upstream and downstream community difference summary, community discontinuity feature value, discontinuity degree, affected fish functional groups, and shoreline habitat type to be verified. The river segment number and river-blocking structure number are derived from the previous step's segmented eDNA fish community data unit; the upstream and downstream community difference summary is formed by comparing the upstream and downstream fish species composition and relative abundance fields of the same river-blocking structure; the community discontinuity feature value is formed by the upstream and downstream relative abundance difference, connectivity sensitivity weight, and reservoir bay orientation label; the discontinuity degree is formed by the community discontinuity feature value corresponding to the judgment threshold in the project configuration file; the affected fish functional groups are determined by the functional group label field and their upstream and downstream change directions; and the shoreline habitat type to be verified is generated corresponding to the affected fish functional groups. This output serves as the input for the next stage of generating UAV remote sensing shoreline habitat supply information, where the discontinuity location limits the UAV remote sensing interpretation range, and the affected fish functional groups and the shoreline habitat type to be verified limit the shoreline spatial units that need to be identified by UAV remote sensing.

[0058] S3. Determine the target shoreline verification range of UAV remote sensing based on the fish community connectivity and discontinuity characteristics; within the target shoreline verification range, input the UAV images collected by the UAV into the pre-trained shoreline habitat recognition model to identify the target river segment image blocks and generate UAV remote sensing shoreline habitat supply information; the UAV remote sensing shoreline habitat supply information includes quantity supply component, continuous supply component and shoreline habitat supply degree.

[0059] Specifically, the system receives the fish community connectivity and discontinuity features output in step two, and reads the river segment number, the river-blocking structure number, the summary of upstream and downstream community differences, the community discontinuity feature value, the degree of discontinuity, the functional groups of affected fish, and the shoreline habitat type to be verified. The river segment number comes from the assessment river segment division results in the river health assessment, and is used to determine the assessment river segment covered by UAV remote sensing imagery; the river-blocking structure number comes from the paired community archives already formed in step two, and is used to locate the corresponding upstream and downstream shorelines; the upstream and downstream community difference summary comes from the comparison results of the eDNA fish species composition field, relative abundance field, and functional group label field, and is used to determine the focus of image interpretation; the community fragmentation characteristic value and fragmentation degree come from the community fragmentation judgment results in step two, and are used to determine the coverage intensity of the shoreline verification range; the affected fish functional groups come from the identification results of changes in flowing fish, migratory fish, local sensitive fish, benthic fish, and aquatic plant dependent fish in step two, and are used to invoke the corresponding fish habitat requirements; the shoreline habitat type to be verified comes from the shoreline verification objects generated in step two based on the affected fish functional groups, and is used to limit the shoreline spatial units that need to be identified in the UAV imagery.

[0060] The system writes the above fields into the UAV remote sensing mission record, so that the eDNA biological fragmentation results are transformed into the remote sensing interpretation range, interpretation object and supply calculation object for this stage.

[0061] The system generates the target shoreline verification range based on the river section number and the number of the dam structure.

[0062] Among them, for records with a high degree of fracture, the system includes the upstream and downstream sides of the river-blocking structure and the near-shore continuous channel between the two sides in the key image range; For records with a medium degree of fracturing, the system includes the nearshore shoreline zone related to the functional groups of affected fish on both the upstream and downstream sides within the image range; For records with low degree of fracturing, the system includes nearshore areas within the same river segment that directly correspond to the habitat type of the shoreline to be verified within the image range.

[0063] Among them, the community fault characteristic value is used to refine the continuous coverage of the shoreline verification range. The higher the characteristic value, the more the system selects the image range that emphasizes the continuous coverage of the upstream reservoir bay, the downstream flow restoration area, and the near-shore passage space on both sides of the dam.

[0064] The upstream and downstream community difference summary further determines the focus of image recognition: when the summary shows an abundance of still-water adapted fish and invasive fish species in the upstream side, the system prioritizes the recognition of upstream bays, slow-flowing zones, backwaters, and areas near hard revetments; when the summary shows preservation of flowing and migratory fish species in the downstream side and a weakening of corresponding groups in the upstream side, the system prioritizes continuous natural shorelines, shoals, channel-type nearshore spaces, and gravel beaches; when the summary shows significant changes in aquatic plant-dependent fish and benthic fish species, the system prioritizes the proportion of aquatic plant zones, slow-flowing backwaters, gravel beaches, and hard revetments. Through this processing, all output fields from step two are incorporated into the image task generation and object determination process in this stage.

[0065] Furthermore, to ensure that eDNA results and UAV imagery correspond to the same diagnostic subject, the system also performs a spatiotemporal consistency check when generating the target shoreline verification area. Among these checks, Spatially, the system unifies the coordinates of sampling points, dam structures, river sections, and image blocks into the same coordinate system, and confirms that the sampling points are located within the upstream and downstream determination range of the corresponding river section or dam structure. Temporally, the system reads the sampling date, the date of UAV image acquisition, and hydrological records. If the interval between the two exceeds the time window allowed by the project configuration file, or if events such as floods, flood discharges, dredging, or engineering construction occur during the period that significantly change the distribution of fish or the morphology of the shoreline, the system marks the diagnostic object as having reduced spatiotemporal matching confidence and retains the verification mark in the output record.

[0066] Furthermore, the system organizes UAV image acquisition according to the target shoreline verification scope. The UAV, equipped with a visible light camera, flies along the river channel, covering the upstream and downstream sides of the dammed structures, the nearshore shoreline zone, and adjacent nearshore waters. After image acquisition, the system generates orthophotos through aerial survey stitching and crops them into target river segment image blocks according to river segment number, dammed structure number, and upstream and downstream location attributes. Each image block retains image coordinate information, its river segment number, dammed structure number, and location attributes, enabling the identified shoreline habitat units to be written back into the fish community connectivity break feature record corresponding to step two. Image preprocessing includes orthorectification, stitching, cropping, and coordinate registration. The processed images unfold around the water body, shoreline, nearshore vegetation, and nearshore topography, allowing subsequent identification results to directly correspond to the spatial location of fish community anomalies.

[0067] Furthermore, before the UAV imagery enters the recognition model, the system performs image quality control. Quality control includes at least checking whether the image resolution meets the minimum spatial scale required for shoreline habitat type identification, checking whether forward overlap, lateral overlap, orthophoto stitching error, and coordinate registration error meet the project configuration file, and checking whether strong reflections, shadows, cloud shadows, turbid water surfaces, and motion blur render shoreline boundaries indistinguishable. Image patches that do not meet the quality requirements are not directly used in shoreline habitat provision calculations but are marked as requiring re-flying, manual verification, or confidence reduction. This quality control ensures that the UAV remote sensing results have actionable input conditions, rather than simply relying on the image recognition model in a general way.

[0068] Furthermore, the system invokes a pre-trained shoreline habitat recognition model with fixed parameters to identify image patches of the target river segment. This model employs an encoder-decoder semantic segmentation structure, taking as input orthophoto patches of the target river segment and outputting pixel-by-pixel shoreline habitat categories. The model's encoder layer consists of four levels of convolutional feature extraction layers, each comprising two convolutions, normalization, and nonlinear activation. Downsampling is used to progressively expand the receptive field, extracting features such as land-water boundaries, shoreline textures, vegetation patches, aquatic plant textures, shallow water color variations, and hard revetment edges. The model's intermediate layers incorporate multi-scale dilated convolutional units to identify larger-scale shoreline structures such as backwater bays, slow-flowing reservoir bays, continuous natural shorelines, and channel-type nearshore spaces. The model's decoder layer uses upsampling and skip layers to merge high-level semantic features with shallow boundary features, restoring the boundaries of aquatic plant areas, shallow waters, gravel beaches, and hard revetment edges to their original spatial positions in the orthophoto image. The model output employs a pixel-by-pixel classification layer, outputting categories for water bodies, natural vegetation shorelines, aquatic plant belts, shoals, gravel beaches, backwater bays, slow-flowing reservoir bays, channel-type nearshore spaces, tributary inlets, hard revetments, and bridge pier disturbance areas. The model training samples are derived from historical UAV orthophotos of river channels and manually labeled shoreline habitat samples. The manual labels categorize fish according to their habitat, foraging, reproduction, and passage needs, ensuring that the model output directly corresponds to the functional groups of affected fish species.

[0069] The shoreline habitat identification model was trained, validated, and version-fixed before being used in this evaluation. The system saves the model version number, training sample source, category label definition, validation set accuracy record, and applicable image resolution range. The model outputs the category and category confidence score for each pixel. When the average confidence score of a shoreline habitat unit is lower than the minimum confidence score requirement in the project configuration file, or when there is significant confusion between adjacent category boundaries, the system marks the unit as a low-confidence identification result and initiates a manual review or conservative statistical process. The system does not directly use low-confidence identification results as the sole basis for identifying high-pressure weaknesses, thus avoiding the black-box model output determining the repair direction without verification.

[0070] Furthermore, after the model completes pixel-by-pixel recognition, the system organizes the output results into shoreline habitat units. The system extracts the boundary line between water and land as the shoreline baseline from the recognition results and generates a nearshore analysis zone along this baseline. Subsequently, natural vegetation shorelines, aquatic plant zones, shoals, gravel beaches, backwater bays, reservoir bay slow-flow areas, channel-type nearshore spaces, tributary inlets, hard revetments, and bridge pier disturbance areas are converted into vectorized shoreline habitat units. Area-type units are recorded based on their area proportion within the nearshore analysis zone, linear shoreline units are recorded based on their length proportion within the shoreline length, and local disturbance units are recorded based on their influence range proportion. The system uniformly writes these spatial records into the shoreline habitat unit supply proportion field and binds them to the river segment number, dam number, upstream location attribute, and downstream location attribute. This processing converts the UAV image classification map into a supply field that directly corresponds to the functional group requirements of fish.

[0071] The supply proportions of shoreline habitat units are all dimensionless ratios. The proportion of area-type units is obtained by comparing the area of ​​that type of unit with the total area of ​​the target nearshore analysis zone; the proportion of linear shoreline units is obtained by comparing the length of that type of shoreline with the total length of the target shoreline; and the proportion of locally disturbed units is obtained by comparing their affected length or affected area with the corresponding analysis baseline. The proportions of different baselines are not directly added together. Only when fish functional groups—shoreline habitat demand tables clearly belong to the same supply dimension and have undergone weight transformation are they included in the quantity supply component. If the total area of ​​the target nearshore analysis zone or the total length of the target shoreline is zero, the system does not calculate the supply proportion but instead marks the image cropping range or shoreline baseline extraction result as an anomaly.

[0072] The system extracts targeted supply information from image recognition results based on the affected fish functional groups and the types of shoreline habitats to be verified. For records of affected fish functional groups that are migratory and flowing fish, the system extracts the supply ratios for continuous natural shorelines, shoals, channel-type nearshore spaces, and gravel beaches. For records of affected fish functional groups that depend on aquatic plants, the system extracts the supply ratios for aquatic plant belts, slow-flowing backwaters, and shorelines with natural vegetation. For records of affected fish functional groups that are benthic, the system extracts the supply ratios for gravel beaches, shoals, and hard revetments. For records of upstream still-water adapted fish and introduced fish, the system extracts the supply ratios for slow-flowing bays, backwaters, and hard revetments. Each affected fish functional group corresponds to a set of shoreline habitat requirements, derived from a fish functional group-shoreline habitat requirement table. This requirement table is compiled from local fish ecological habits data, river health assessment experience, and aquatic ecological restoration engineering experience, and remains fixed within the same assessment period.

[0073] Furthermore, the system generates shoreline habitat supply for affected fish functional groups. The algorithm for this supply is derived from the weighted overlay method in ecological suitability assessment, the basic idea of ​​which is to weight and summarize multiple habitat factors according to their correlation with the needs of the target organisms. In this implementation, when using this method with UAV remote sensing shoreline identification results, a quantitative supply component is first formed from the UAV identification results and the fish functional group-shoreline habitat demand table. This component represents the combined result of the proportion of suitable shoreline habitat supply for each type of river segment and the demand weight of that functional group. Then, a continuous supply component is formed by segmenting the shoreline baseline, representing the continuous distribution level of suitable habitats along the shoreline. Subsequently, the continuous demand weight of the functional group is read from the fish functional group-shoreline habitat demand table to represent the degree of dependence of the affected fish functional group on shoreline continuity. The system then merges the quantitative supply component and the continuous supply component according to the continuous demand weight of the functional group to form the shoreline habitat supply.

[0074] Understandably, this process reflects two types of relationships in the scenario of this invention: anomalies in fish functional groups determine the types of shoreline habitats that need to be checked, and the ecological habits of fish functional groups determine the contribution of shoreline continuity to the supply level. The degree of fracturing and community fracturing characteristic values ​​output in step two are used to generate a fracturing degree guiding coefficient. This guiding coefficient is output along with the supply information in this step, for the next step to combine the longitudinal connectivity pressure and the shoreline habitat supply gap for judgment.

[0075] The fracture severity guidance coefficient is generated by the system in this step based on the fracture severity label and community fracture characteristic values ​​from step two. The system first reads the basic guidance values ​​corresponding to low, medium, and high fracture severity in the project configuration file, then reads the position of the community fracture characteristic values ​​within their respective threshold ranges, refining the basic guidance values ​​so that different river-blocking structures belonging to the same high fracture severity can still generate different longitudinal connectivity pressures according to fracture strength. The fracture severity guidance coefficient is a dimensionless proportional value between zero and one. When the output data from step two is insufficient, there are unilateral anomalies, or low-confidence fracture records, the system synchronously marks the guidance coefficient as a low-confidence state. This state is retained in the next step when generating the short-plate pressure value, but it is not used as a deterministic basis for high pressure.

[0076] The quantity supply component is obtained by the system reading the shoreline habitat types corresponding to the affected fish functional groups and summing the supply proportions of each type according to the demand weights in the fish functional group-shoreline habitat demand table. The continuous supply component is generated by the system on the shoreline baseline: the system divides the target river section shoreline into continuous segments and reads whether there are shoreline habitat types required by the affected fish functional groups in the adjacent areas of each segment; when adjacent segments have corresponding habitat units continuously, the system records it as a continuous habitat segment; when the corresponding habitat unit is interrupted by hard revetments, bridge pier disturbance areas, or lack of suitable nearshore space, the system records it as an interrupted segment; the continuous supply component is formed by the proportion of continuous habitat segments in the target river section shoreline. The continuity demand weight of functional groups is determined by the fish functional group-shoreline habitat demand table, with migratory fish and flowing fish corresponding to higher continuity demand weights, aquatic plant dependent fish corresponding to intermediate continuity demand weights, and benthic fish corresponding to basic continuity demand weights. The system integrates the quantity supply component and the continuous supply component according to the continuous demand weight to obtain the shoreline habitat supply degree. The higher the continuous demand weight, the more the system emphasizes the continuous distribution of suitable habitats along the shoreline, and the lower the continuous demand weight, the more the system emphasizes the supply ratio of suitable habitat units.

[0077] Furthermore, the quantity supply component, continuous supply component, functional group continuous demand weight, and shoreline habitat supply are all dimensionless proportional values, with internal values ​​ranging from zero to one. When expressed as percentages in the text, they correspond one-to-one with the internal proportional values. The demand weights in the fish functional group-shoreline habitat demand table are normalized under the same functional group, ensuring that the sum of the demand weights for multiple habitats corresponding to the same functional group remains within the same proportional scale. When a certain type of demand habitat cannot be reliably identified, the system records the supply proportion of that type as zero and retains the reason for not identifying it, rather than deleting the type and re-normalizing. This avoids overestimating the shoreline habitat supply due to the omission of missing items.

[0078] In one implementation record, step two outputs that the affected fish functional groups of a certain river-blocking structure are aquatic plant dependent fish, and the shoreline habitat types to be verified are aquatic plant belts, slow-flowing backwater bays, and natural vegetation shorelines. After acquiring UAV orthophotos of the target river section, the shoreline habitat identification model identifies the supply ratio of aquatic plant belts as 12%, slow-flowing backwater bays as 8%, and natural vegetation shorelines as 36%. In the fish functional group-shoreline habitat demand table, the demand weight of aquatic plant dependent fish for aquatic plant belts is high and converted to 60%, the demand weight for slow-flowing backwater bays is medium and converted to 30%, and the demand weight for natural vegetation shorelines is low and converted to 10%. Based on this, the system forms a quantity supply component, recorded as 13.2%. The system counts continuous habitat segments along the shoreline baseline and obtains that the shoreline proportion that can continuously support aquatic plant dependent fish is 9%, forming a continuous supply component. The fish functional group-shoreshore habitat demand table records the continuous demand weight of aquatic plant-dependent fish as 50%. The system combines the 13.2% quantity supply component and the 9% continuous supply component according to this continuous demand weight to generate the shoreline habitat supply degree for aquatic plant-dependent fish in this river segment, with a result of 11.1%. This supply degree proceeds to the next step to determine whether the fish health deficiency in this river segment is due to insufficient shoreline habitat supply or complex stress.

[0079] The system ultimately outputs drone-based remote sensing information on shoreline habitat supply. This supply information is indexed by river segment number and dam number, and includes the target shoreline verification range, upstream or downstream location attributes, affected fish functional groups, shoreline habitat types to be verified, shoreline habitat identification results, supply proportion of each shoreline habitat unit, quantity supply component, continuous supply component, functional group continuity demand weight, discontinuity guidance coefficient, and shoreline habitat supply degree. The river segment number, dam structure number, fracture degree, community fracture characteristic value, upstream and downstream community difference summary, affected fish functional groups, and shoreline habitat type to be verified are derived from the fish community connectivity fracture characteristics output in step two; the target shoreline verification range is determined by fracture location, fracture degree, and community fracture characteristic value; shoreline habitat identification results are derived from UAV orthophotos and shoreline habitat identification models; the supply ratio of each shoreline habitat unit is derived from the spatial organization of the identification results; the quantity supply component is generated by the supply ratio and demand weight; the continuous supply component is derived from the shoreline baseline segment statistics; the functional group continuity demand weight is derived from the fish functional group-shoreline habitat demand table; the fracture degree guiding coefficient is derived from the fracture results in step two; the shoreline habitat supply degree is jointly generated by the quantity supply component, the continuous supply component, and the functional group continuity demand weight.

[0080] S4. Based on the information on shoreline habitat supply obtained from the UAV remote sensing, output the types of health deficiencies of fish in the river section and the directions for restoration.

[0081] Specifically, the system receives the shoreline habitat supply information output by UAV remote sensing in step three, and reads the river section number, the number of the river-blocking structure, the target shoreline verification range, the upstream location attribute, the downstream location attribute, the functional groups of affected fish, the type of shoreline habitat to be verified, the shoreline habitat identification result, the supply ratio of each shoreline habitat unit, the quantity supply component, the continuous supply component, the continuous demand weight of functional groups, the fault degree guidance coefficient, and the shoreline habitat supply degree. The river segment number and the number of the dam structure are derived from the supply information index in step three, used to identify the diagnostic targets for this stage; the target shoreline verification scope and location attributes are derived from the interpretation results of UAV orthophotos, used to determine the spatial location of the shortfall; the affected fish functional groups are derived from the fish community connectivity and discontinuity characteristics in step two, used to determine the fish ecological needs corresponding to the current diagnostic targets; the shoreline habitat types to be verified are derived from the fish functional group-shoreline habitat demand matching results in step three, used to limit the shoreline spatial units participating in the shortfall assessment; the shoreline habitat identification results and the supply ratio of each shoreline habitat unit are derived from UAV orthophotos and the shoreline habitat identification model, used to locate specific missing or insufficient shoreline units; the quantity supply components are derived from each The combined result of the shoreline habitat unit supply ratio and the fish habitat demand weight is used to determine the overall supply level of suitable shoreline habitats. The continuous supply component, derived from the statistical results of shoreline baseline segments, is used to determine the continuity of suitable habitats along the shoreline. The functional group continuity demand weight, derived from the fish functional group-shoreline habitat demand table, is used to determine the contribution of the continuous supply component to the shoreline habitat supply. The fault degree guiding coefficient, derived from the community fault characteristic value and fault degree in step two, is used to represent the longitudinal connectivity pressure. The shoreline habitat supply is derived from the fusion result of the quantitative supply component, continuous supply component, and functional group continuity demand weight in step three, and is used to represent the comprehensive support level of the target river segment for the affected fish functional groups. The system writes the above fields into the river segment fish health deficiency diagnosis record, so that the eDNA community connectivity fault evidence and the UAV remote sensing shoreline supply evidence are matched in the same diagnostic object.

[0082] Furthermore, the system establishes diagnostic objects based on river segment number, dam number, upstream location attributes, downstream location attributes, and affected fish functional groups. Each diagnostic object corresponds to a specific river segment, a specific dam structure, a specific shoreline side, and a specific affected fish functional group.

[0083] In cases where the same dam structure simultaneously affects both migratory fish and aquatic plant-dependent fish, the system generates separate diagnostic targets for migratory fish and aquatic plant-dependent fish. The migratory fish diagnostic targets utilize supply information from continuous natural shorelines, shoals, and channel-type nearshore spaces, while the aquatic plant-dependent fish diagnostic targets utilize supply information from aquatic plant belts, slow-flowing backwaters, and naturally vegetated shorelines. Based on the habitat type of the shoreline to be verified, the system extracts corresponding spatial units from the shoreline habitat identification results in step three and reads the supply ratio, spatial distribution, and continuous segments of these spatial units within the target shoreline verification area.

[0084] Therefore, each shortcoming judgment is linked to the actual shoreline spatial unit already identified in step three, and the diagnostic object can be located in a specific river section, a specific dam structure, and a specific shoreline side.

[0085] Furthermore, the system determines whether shoreline habitat supply has become a bottleneck. The system reads the quantity supply component, the continuous supply component, and the shoreline habitat supply level, and compares the shoreline habitat supply level with the insufficient supply threshold in the project configuration file. When the supply level falls within the insufficient threshold range, the system marks the diagnostic object as having insufficient shoreline habitat supply; when the supply level falls within the supportable threshold range, the system marks the diagnostic object as having supportable shoreline habitat supply. The system continues to read the status of the quantity supply component and the continuous supply component to further analyze the reasons for the deficiency. When the quantity supply component falls into a low range, the system records it as insufficient habitat quantity; when the continuous supply component falls into a low range, the system records it as insufficient habitat continuity; when both the quantity supply component and the continuous supply component fall into a low range, the system records it as insufficient quantity and insufficient continuity.

[0086] The detailed results directly lead to the restoration-oriented generation process. Insufficient quantity corresponds to the expansion of suitable habitat area or length, insufficient continuity corresponds to the connection of scattered habitat segments, and insufficient quantity and continuity correspond to the combined implementation of microhabitat restoration and continuity reconstruction.

[0087] Furthermore, the system reads the fault severity guidance coefficient to determine the role of longitudinal connectivity pressure in the current diagnostic object. The fault severity guidance coefficient is derived from the shoreline habitat supply information obtained from UAV remote sensing in step three, and its value is formed based on the community fault characteristic value and fault severity obtained in step two. The system compares the fault severity guidance coefficient with the connectivity pressure judgment threshold in the project configuration file. When it enters the high range, it is recorded as significant longitudinal connectivity pressure; when it enters the middle range, it is recorded as present longitudinal connectivity pressure; and when it enters the low range, it is recorded as weak longitudinal connectivity pressure. For migratory fish, flowing water fish, and native sensitive fish, the system uses the connectivity pressure label as the priority basis for determining the shortcoming type. For aquatic plant dependent fish and benthic fish, the system simultaneously reads the shoreline habitat supply degree, quantity supply component, and continuous supply component, so that the shortcoming judgment can reflect the correspondence between fish functional groups and shoreline spatial supply.

[0088] Furthermore, the system generates a river segment fish health shortfall pressure value, which is used to combine longitudinal connectivity pressure and shoreline habitat supply gap at the same diagnostic scale, and to generate shortfall pressure levels and remediation priorities. The algorithm for this pressure value is derived from the probabilistic and synthesis method in fuzzy logic, which is often used to combine two factors leading to the same adverse outcome into a single overall pressure level.

[0089] The underlying idea is that when two pressure sources coexist, the total pressure is synthesized using a probability-based approach, and the overlapping contributions of the two pressure sources are deducted to maintain the total pressure within a stable proportional scale. In this implementation, when the method is applied to a river health diagnosis scenario, the fracture degree guiding coefficient is used as the longitudinal connectivity pressure, and the supply gap calculated from the shoreline habitat supply is used as the spatial habitat pressure. The supply gap is obtained by subtracting the shoreline habitat supply by 100%; the lower the supply, the higher the supply gap. The system first reads the fracture degree guiding coefficient, then the supply gap, and then merges the two types of pressure according to the probability-based synthesis rules, deducting the overlapping contributions of the two types of pressure to obtain the short-board pressure value. The fracture degree guiding coefficient, shoreline habitat supply, and supply gap are all proportional diagnostic parameters, and the generated short-board pressure value is also a proportional diagnostic parameter. The system compares the short-board pressure value with the short-board pressure level threshold in the project configuration file, generating high-pressure, medium-pressure, or low-pressure labels; high-pressure labels correspond to priority repair, medium-pressure labels correspond to inclusion in recent repair, and low-pressure labels correspond to inclusion in routine maintenance.

[0090] Furthermore, the thresholds for short-plate pressure level, insufficient supply judgment threshold, and connectivity pressure judgment threshold are all provided by the project configuration file before the evaluation begins. These thresholds can be derived from historical evaluation records of the same watershed, background fluctuation ranges of ordinary river sections, local river health evaluation technical requirements, or project expert verification results. The system saves the threshold source, version number, and applicable conditions, and retains the specific threshold version used for judgment in the output record. Short-plate pressure values ​​are used for ranking and assisting in determining remediation priorities, but do not directly replace engineering design review. When low-confidence markers are present in eDNA data, image recognition results, or hydrological condition records, the system simultaneously outputs a confidence status next to the short-plate pressure level, indicating that manual review is required during the remediation plan preparation stage.

[0091] Understandably, in one implementation record, the affected fish functional group corresponding to a certain dam structure is aquatic plant dependent fish. The shoreline habitat supply output in step three is 11.1%, and the fracture degree guidance coefficient is 50%. The system first converts the shoreline habitat supply into a supply gap, obtaining 88.9%. Then, it probabilistically combines the 50% longitudinal connectivity pressure with the 88.9% supply gap, and deducts the overlapping contribution of the two, obtaining a short-board pressure value of approximately 94.4%. The short-board pressure value of this diagnosed object enters the high-pressure range, and the system records its repair priority as priority repair. At the same time, the shoreline habitat supply of this diagnosed object enters the insufficient judgment range, and the continuous supply component is also at a low level. The system focuses on recording the short-board of this diagnosed object as insufficient shoreline habitat supply, while retaining the state of existence of longitudinal connectivity pressure. In another implementation record, the affected fish functional group is migratory fish. The fracture degree guidance coefficient enters the high range, and the shoreline habitat supply is within the supportable judgment range. The system focuses on recording the short-board of this diagnosed object as longitudinal connectivity fracture, and determines the repair priority based on the short-board pressure value. In one implementation record, the fault degree guidance coefficient entered the high range, and the shoreline habitat supply entered the insufficient judgment range. The system recorded the diagnosed object as barrier-habitat composite stress and determined the restoration priority based on the short board pressure value.

[0092] Furthermore, the system outputs the type of fish health bottleneck in a river segment based on the fault degree guidance coefficient, shoreline habitat supply, quantity supply component, continuous supply component, and bottleneck pressure value. When the fault degree guidance coefficient enters a high range and the shoreline habitat supply is within a supportable range, the system outputs a longitudinally connected fault type bottleneck; when the shoreline habitat supply enters a deficient range and the fault degree guidance coefficient is in a low or medium range, the system outputs a shoreline habitat supply deficiency type bottleneck; when the fault degree guidance coefficient enters a high range and the shoreline habitat supply enters a deficient range, the system outputs a barrier-habitat combined stress type bottleneck. For shoreline habitat supply deficiency type bottlenecks, the system continues to read the quantity supply component and continuous supply component to generate secondary bottleneck labels: quantity deficiency type, continuous deficiency type, or quantity-continuous combined deficiency type. For barriers and habitat-related stress-related short-term hazards, the system determines the restoration sequence based on the functional groups of affected fish species: migratory and flowing-water fish are prioritized for connectivity improvement, with simultaneous verification of nearshore channel space; aquatic plant-dependent and benthic fish are prioritized for microhabitat restoration, with simultaneous improvement of upstream and downstream shoreline continuity. Short-term hazard stress values ​​are used to rank the severity and determine restoration priorities within the same type of hazard, enabling engineering implementation to proceed in order of stress level for similar hazards.

[0093] Furthermore, the system generates restoration directions based on the type of shortcoming, its stress value, restoration priority, and the type of shoreline habitat to be verified. For shortcomings with longitudinal connectivity disruptions, the restoration directions include the installation of fish passage facilities, improvement of fish passage connectivity, and ecological scheduling of dams and sluices. The restoration location is determined based on the dam structure number, upstream location attributes, and downstream location attributes. For shortcomings with insufficient shoreline habitat supply, the restoration directions are determined by the type of missing shoreline habitat. For aquatic plant-dependent fish, the directions are restoration of aquatic plant belts, creation of slow-flowing backwater bays, and restoration of natural vegetation shorelines. For benthic fish, the directions are restoration of gravel beaches, shallow waters, and ecological transformation of hard revetments. For migratory and flowing fish, the directions are restoration of continuous natural shorelines, shallow waters, and unblocking of nearshore passageways. For shortcomings with combined barrier-habitat stress, the restoration directions are a combination of connectivity improvement and shoreline habitat restoration. When generating restoration directions, the system locates the specific restoration objects within the target shoreline verification range and their upstream or downstream location attributes, and associates them with the shoreline habitat identification results from step three, so that the restoration directions can be applied to specific river sections, specific river-blocking structures, and specific shoreline spatial units.

[0094] Finally, this stage outputs the types of fish health deficiencies in river sections and their restoration directions. The types of fish health deficiencies are indexed by the river section number, the number of the dam structure, and the functional groups of affected fish species. These include longitudinal connectivity disruption deficiencies, shoreline habitat insufficiency deficiencies, and barrier-habitat combined stress deficiencies. When shoreline habitat insufficiency is identified, secondary labels such as insufficient quantity, continuous insufficiency, or a combination of both are added, along with the stress level of the deficiencies. Restoration directions are generated based on the target shoreline verification area, upstream or downstream location attributes, the type of shoreline habitat to be verified, the functional groups of affected fish species, and restoration priorities. The output includes engineering directions such as fish passage facility installation, fish channel connectivity improvement, dam ecological management, aquatic plant restoration, slow-flowing backwater creation, natural vegetation shoreline restoration, gravel beach restoration, shoal restoration, ecological transformation of hard revetments, and dredging of nearshore passageways. This output transforms eDNA evidence of fish community disruption and UAV remote sensing evidence of shoreline habitat supply into diagnostic results that can be used for river health assessment reports and river section restoration plan development.

[0095] This invention also provides a river health assessment system integrating eDNA and UAV remote sensing, the system comprising: The eDNA fish community processing module is used to acquire eDNA water samples from upstream and downstream of the river-blocking structures and ordinary river sections within the target evaluation river section. After sequencing and comparison, fish species composition, relative abundance, and functional group labels are obtained, generating river section-specific eDNA fish community data units. Each sampling point corresponds to one river section-specific eDNA fish community data unit. Each river segment-based eDNA fish community data unit includes a river segment number, a river-blocking structure number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field. The connectivity and discontinuity identification module is used to calculate improved community discontinuity feature values ​​based on the eDNA fish community data units of the upstream and downstream sides of the same river-blocking structure, by combining the corresponding differences in species and relative abundance. These community discontinuity feature values ​​are used to represent the degree of discontinuity of fish communities on both sides of the same river-blocking structure, and are compared with a preset judgment threshold to generate fish community connectivity and discontinuity features. The fish community connectivity and discontinuity features are recorded with the river-blocking structure as the recording object, including the river segment number, the river-blocking structure number, the summary of differences between upstream and downstream communities, the community discontinuity feature value, the degree of discontinuity, the functional groups of affected fish, and the shoreline habitat type to be verified. The UAV remote sensing shoreline interpretation and shoreline habitat supply calculation module is used to determine the target shoreline verification range of UAV remote sensing based on the fish community connectivity and discontinuity characteristics; within the target shoreline verification range, the UAV-collected images are input into a pre-trained shoreline habitat recognition model to identify the target river segment image blocks and generate UAV remote sensing shoreline habitat supply information; the UAV remote sensing shoreline habitat supply information includes quantity supply component, continuous supply component, and shoreline habitat supply degree. The shortcoming diagnosis and repair direction generation module is used to output the type of shortcoming in the health of fish in the river section and the repair direction based on the habitat supply information of the shoreline sensed by the UAV.

[0096] Example 2 To illustrate the data processing closed loop of "binding eDNA fish community results to spatial units, forming community difference summaries, and further supporting habitat verification" in this invention, this embodiment was tested on a set of quality-controlled river eDNA samples and corresponding river segment / habitat unit data. The test data included reading fields for 174 eDNA samples, 6 types of river segment / habitat units, and 66 fish taxonomic units. The system read the spatial unit number, sample number, total allocated reading, number of taxonomic units detected in the sample, and readings for each fish taxonomic unit for each sample. Then, it summarized the fish readings according to spatial units and calculated the relative reading abundance of each fish taxonomic unit within that spatial unit, using all fish readings within the same spatial unit as a benchmark. This testing process corresponds to the relative abundance generation in step one, the spatial object aggregation in step two, and the community difference summary generation. In actual engineering, the spatial unit can be determined by the UAV shoreline habitat identification results, target shoreline verification range, and river segment number in step three of this invention. The summarized test results are shown in Table 1 below.

[0097] Table 1

[0098] As shown in the table above, there are significant differences in the fish community composition based on eDNA across different river sections / habitat units. For example, *Cyprinus scabra* had the highest relative abundance of readings in the downstream main stream unit, reaching 33.3%; *Carangophytum commune* had the highest relative abundance of readings in the upstream tributary unit, reaching 24.2%; and *Carangophytum rubrum* had a relative abundance of 36.9% in the slow-flowing bay unit. These results indicate that by binding eDNA detection results with spatial units, the system can generate a recalculated summary of community differences and further trigger subsequent verification of habitat supply conditions in the corresponding spatial units. Figure 2 The relative reading abundance composition of major fish taxa within each river segment / habitat unit was further illustrated; from Figure 2As can be seen, the dominant fish taxa and their relative abundance distributions are not the same in different spatial units. The system can bind similar community difference summaries with the upstream and downstream sides of the dam structure, the target shoreline verification range, and the shoreline habitat units identified by UAVs: when the relative abundance of the target functional group in a certain spatial unit decreases, the dominant taxa is replaced, or the community structure is significantly different from that of the adjacent spatial units, the system can generate the affected fish functional groups and the shoreline habitat types to be verified according to step two, and then calculate the shoreline habitat supply by calling the UAV remote sensing results according to step three.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or units may be electrical, mechanical, or other forms.

[0101] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for river health assessment integrating eDNA and UAV remote sensing, characterized in that the method... include: S1. Obtain eDNA water samples from upstream and downstream of the dam structure and from ordinary river sections within the target evaluation river section. Sequencing and comparison are used to obtain fish species composition, relative abundance, and functional group labels, generating river section-specific eDNA fish community data units. Each sampling point corresponds to one river section-specific eDNA fish community data unit. Each river segment-based eDNA fish community data unit includes a river segment number, a river-blocking structure number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field. S2. Based on the eDNA fish community data units of the upstream and downstream sides of the same dam structure, an improved community fragmentation feature value is calculated by combining the corresponding differences in species and relative abundance. This community fragmentation feature value is used to represent the degree of fragmentation of fish communities on both sides of the same dam structure, and is compared with a preset judgment threshold to generate fish community connectivity fragmentation features. The fish community connectivity fragmentation features are recorded with the dam structure as the recording object, including the river segment number, dam structure number, upstream and downstream community difference summary, community fragmentation feature value, degree of fragmentation, affected fish functional groups, and shoreline habitat type to be verified. S3. Determine the target shoreline verification range of UAV remote sensing based on the fish community connectivity and discontinuity characteristics; within the target shoreline verification range, input the UAV images collected by the UAV into the pre-trained shoreline habitat recognition model to identify the target river segment image blocks and generate UAV remote sensing shoreline habitat supply information; the UAV remote sensing shoreline habitat supply information includes quantity supply component, continuous supply component and shoreline habitat supply degree. S4. Based on the information on shoreline habitat supply obtained from the UAV remote sensing, output the types of health deficiencies of fish in the river section and the directions for restoration.

2. The river health assessment method integrating eDNA and UAV remote sensing according to claim 1, characterized in that, The functional group labels for fish include flowing water fish, migratory fish, local sensitive fish, still water adapted fish, benthic fish, aquatic plant dependent fish, and alien fish; The relative abundance is specifically: Using all valid sequence readings of fish within the same sampling point as a benchmark, the valid sequence readings of a certain fish species are converted into its proportion in the valid sequences of fish at that sampling point, and the current proportion is written into the relative abundance field.

3. The river health assessment method integrating eDNA and UAV remote sensing according to claim 1, characterized in that, The improved community fragmentation characteristic value is calculated by combining the corresponding differences in species and relative abundance, specifically as follows: The eDNA fish community data units of the river section upstream and downstream of the same dam structure are compared in pairs to generate a species control list on both sides. The relative abundance difference of each fish species is calculated, and the corresponding connectivity-sensitive weights are introduced according to the fish functional group labels for weighting. Combined with the reservoir bay direction labels, the community breakage feature value is obtained after normalization, and then compared with the preset judgment threshold to determine the degree of breakage. The bay orientation label is generated based on the enrichment of upstream still-water adapted fish or alien fish and downstream flowing-water fish or local sensitive fish. The connectivity sensitivity weights enable flowing water fish, migratory fish, and locally sensitive fish to make a greater contribution to fracture detection.

4. The river health assessment method integrating eDNA and UAV remote sensing according to claim 3, characterized in that, The species comparison list on both sides records common species detected on both sides, differential species detected on the upstream side, and differential species detected on the downstream side, and simultaneously retains the functional group label corresponding to each fish species or reliable classification level.

5. The river health assessment method integrating eDNA and UAV remote sensing according to claim 1, characterized in that, The river segment number and the dam structure number are derived from the river segment-based eDNA fish community data unit; the upstream and downstream community difference summary is formed by comparing the upstream and downstream fish species composition field and relative abundance field of the same dam structure; the community fragmentation feature value is formed by the upstream and downstream relative abundance difference, connectivity sensitivity weight, and reservoir bay orientation label; the fragmentation degree is formed by the community fragmentation feature value corresponding to the judgment threshold in the project configuration file; the affected fish functional groups are determined by the functional group label field and its upstream and downstream change direction; the shoreline habitat type to be verified is generated corresponding to the affected fish functional groups.

6. The river health assessment method integrating eDNA and UAV remote sensing according to claim 1, characterized in that, The cluster fracture characteristic value of each river-blocking structure is compared with the preset low-level judgment threshold, medium-level judgment threshold and high-level judgment threshold to generate a fracture degree label. When it falls into the low-level judgment range, it is recorded as low fracture degree; when it falls into the medium-level judgment range, it is recorded as medium fracture degree; when it falls into the high-level judgment range, it is recorded as high fracture degree. Then there is, The determination of the target shoreline verification range by UAV remote sensing based on the connectivity and discontinuity characteristics of the fish community is specifically as follows: If the degree of fracture is as described, then the near-shore continuous channel covering the upstream side, downstream side and both sides of the river-blocking structure is covered. If it is of the medium degree of fracture, then the nearshore shoreline zone related to the functional groups of affected fish on both the upstream and downstream sides will be covered; If the degree of fracture is low, then the nearshore area within the same river segment that corresponds to the habitat type of the shoreline to be verified is covered.

7. The river health assessment method integrating eDNA and UAV remote sensing according to claim 1, characterized in that, The pre-trained shoreline habitat identification model adopts an encoder-decoder semantic segmentation structure. The input is an orthophoto patch of the target river segment, and the output is a pixel-by-pixel shoreline habitat category. The model is trained and the parameters are fixed. After the pre-trained shoreline habitat recognition model completes pixel-by-pixel recognition, the following steps are performed: The boundary line between the water body and the land area is extracted from the identification results as the shoreline baseline, and a nearshore analysis zone is generated along the shoreline baseline. Natural vegetation shorelines, aquatic plant belts, shoals, gravel beaches, backwater bays, slow-flowing reservoir bays, channel-type nearshore spaces, tributary confluences, hard revetments, and bridge pier disturbance areas are converted into shoreline habitat units to locate specific missing or insufficient shoreline units. The quantity supply component is obtained by weighting and summing the supply proportion of the shoreline habitat type to be verified according to the demand weight in the pre-set fish functional group-shoreline habitat demand table; The continuous supply component is obtained by the proportion of continuous habitat segments on the shoreline baseline in the target river section shoreline; the shoreline habitat supply degree is obtained by the fusion of the quantity supply component, the continuous supply component, and the functional group continuity demand weight.

8. The river health assessment method integrating eDNA and UAV remote sensing according to claim 7, characterized in that, Based on the shoreline habitat supply information obtained from the UAV remote sensing, the system outputs the types of fish health deficiencies and restoration directions for the river section, specifically as follows: Based on the aforementioned shoreline habitat units, diagnostic objects are established according to river segment number, dam number, upstream location attributes, downstream location attributes, and affected fish functional groups; Based on the diagnosed object, the corresponding quantity supply component, continuous supply component, and shoreline habitat supply degree are obtained, and the shoreline habitat supply gap is determined according to the shoreline habitat supply degree. For the supply of shoreline habitats with insufficient supply, the degree of fracture guidance coefficient is obtained and compared with the connectivity pressure judgment threshold in the project configuration file to generate connectivity pressure labels and corresponding merged longitudinal connectivity pressures. By merging the shoreline habitat supply gap and the longitudinal connectivity pressure and deducting the overlapping contribution of the two pressure sources, the total pressure is kept within a stable proportional scale to obtain the short-board pressure value. This value is then compared with a preset level threshold to determine the short-board pressure level and repair priority. Based on the fault degree guidance coefficient, shoreline habitat supply, quantity supply component, continuous supply component, and short-board pressure value, the types of short-boards in fish health in river sections are output, including longitudinal connectivity fault type, shoreline habitat insufficient supply type, and barrier-habitat composite stress type. Different restoration strategies are proposed based on the types of health deficiencies of fish in different river sections.

9. The river health assessment method integrating eDNA and UAV remote sensing according to claim 8, characterized in that, The repair targets include: The longitudinal connectivity disruption type of short-term bottleneck corresponds to the installation of fish passage facilities, improvement of fish passage connectivity, and ecological scheduling of sluice gates and dams; the shoreline habitat shortage type of short-term bottleneck is determined by the type of missing shoreline habitat, and corresponds to the restoration of aquatic plant belts, the creation of slow-flowing backwater bays, the restoration of natural vegetation shorelines, the restoration of gravel beaches, the restoration of shallow beaches, the ecological transformation of hard revetments, and the dredging of channel-type nearshore spaces; the barrier-habitat composite stress type of short-term bottleneck corresponds to the combined implementation of connectivity improvement and shoreline habitat restoration.

10. A river health assessment system integrating eDNA and UAV remote sensing, characterized in that, The system includes The eDNA fish community processing module is used to acquire eDNA water samples from upstream and downstream of the river-blocking structures and ordinary river sections within the target evaluation river section. After sequencing and comparison, fish species composition, relative abundance, and functional group labels are obtained, generating river section-specific eDNA fish community data units. Each sampling point corresponds to one river section-specific eDNA fish community data unit. Each river segment-based eDNA fish community data unit includes a river segment number, a river-blocking structure number, a sampling point location type, a fish species composition field, a fish species relative abundance field, and a fish functional group label field. The connectivity and discontinuity identification module is used to calculate improved community discontinuity feature values ​​based on the eDNA fish community data units of the upstream and downstream sides of the same river-blocking structure, by combining the corresponding differences in species and relative abundance. These community discontinuity feature values ​​are used to represent the degree of discontinuity of fish communities on both sides of the same river-blocking structure, and are compared with a preset judgment threshold to generate fish community connectivity and discontinuity features. The fish community connectivity and discontinuity features are recorded with the river-blocking structure as the recording object, including the river segment number, the river-blocking structure number, the summary of differences between upstream and downstream communities, the community discontinuity feature value, the degree of discontinuity, the functional groups of affected fish, and the shoreline habitat type to be verified. The UAV remote sensing shoreline interpretation and shoreline habitat supply calculation module is used to determine the target shoreline verification range of UAV remote sensing based on the fish community connectivity and discontinuity characteristics; within the target shoreline verification range, the UAV-collected images are input into a pre-trained shoreline habitat recognition model to identify the target river segment image blocks and generate UAV remote sensing shoreline habitat supply information; the UAV remote sensing shoreline habitat supply information includes quantity supply component, continuous supply component, and shoreline habitat supply degree. The shortcoming diagnosis and repair direction generation module is used to output the type of shortcoming in the health of fish in the river section and the repair direction based on the habitat supply information of the shoreline remote sensing by the UAV.

Images