Method and system for obtaining fish habitat suitability

Abstract

This invention provides a method and system for obtaining fish habitat suitability, belonging to the field of river ecology and water environment protection technology. The method includes: using environmental DNA technology for water sampling and fish species identification; screening habitat factors significantly related to fish distribution; using a random forest model to analyze the relationship between the number of operational taxa for each fish species and habitat factors, constructing single-factor suitability curves and determining weights; and finally establishing a weighted habitat suitability model to generate habitat suitability for different fish species. This invention achieves objective quantification and spatial visualization of the functional structure of fish habitats, providing a precise and efficient technical means for large-scale habitat assessment and ecological protection planning.

Description

Technical Field

[0001] This invention belongs to the field of river ecology and water environment protection technology, specifically relating to a method and system for obtaining fish habitat suitability. Background Technology

[0002] River habitats possess diverse aquatic habitats, providing essential environmental conditions for fish reproduction, foraging, and overwintering. In recent years, due to human activities such as water conservancy projects, invasive alien species, overfishing, groundwater extraction, floodplain barriers, point and non-point source pollution, and climate change, the habitats of many native fish species and traditionally dominant economic fish have undergone significant changes, manifesting as habitat loss, fragmentation, and homogenization. This has led to genetic isolation and a decline in species richness and diversity. A healthy habitat is fundamental for fish survival and reproduction. Therefore, rapidly identifying and quantifying the spatial distribution of suitable habitats for different ecological function groups over large areas without impacting fish populations is crucial for developing targeted river ecological protection and restoration strategies.

[0003] However, existing methods mainly obtain species composition data through traditional fish catch surveys, and then conduct habitat assessments based on expert experience or single-species models. This method is time-consuming and labor-intensive, has a limited survey scope, causes significant disturbance to fish populations, and even with multiple catches at each sampling point, it is difficult to ensure that all fish species distributed at that sampling point are collected evenly each time. Furthermore, it is difficult to reveal the community's response mechanism to habitat changes in real time, resulting in assessment results that cannot accurately reflect the differences in ecological functions and fail to meet the practical needs of refined management of river ecosystems. Summary of the Invention

[0004] To address the above technical problems, this invention introduces environmental DNA (eDNA) technology for non-destructive biological monitoring and innovatively combines functional traits with cluster analysis to classify complex fish communities into functional groups with clear ecological significance. This method can rapidly and accurately identify the optimal habitats for different functional groups at the whole-basin scale, providing more scientific and precise technical support for the ecological management of cascade reservoirs and the protection of endemic fish habitats. This invention proposes a method for obtaining fish habitat suitability, including the following steps: To achieve the above-mentioned objective, the present invention provides a method for obtaining fish habitat suitability, comprising the following steps: Sampling points were set up along the target water area to collect water samples. The fish species composition in the water was identified based on environmental DNA technology, and candidate habitat factor data were collected at each sampling point simultaneously. The identified fish species were annotated, and operational taxonomic units were divided based on sequence similarity. The values ​​of the operational taxonomic units (OTUs) for each fish species were statistically analyzed. The correlation between candidate habitat factors was analyzed, and candidate habitat factors with collinearity were eliminated to obtain target habitat factors. A random forest model was constructed with the OTU value of fish species as the dependent variable and the target habitat factors as independent variables. The importance of each target habitat factor to the OTU value was analyzed, and the relative weight of each target habitat factor in the habitat suitability model was calculated. Based on the partial effect analysis of the random forest model, a scoring rule for the adaptation index (AD) of fish species corresponding to each target habitat factor was established. The measured values ​​of the target habitat factors at the sampling points are converted into the corresponding fitness index (AD) according to the grading and scoring rules. The habitat suitability of the corresponding fish species at the sampling point is obtained by weighted calculation based on the relative weights of each target habitat factor.

[0005] Preferably, the identification of fish species composition in water bodies based on environmental DNA technology specifically includes: extracting environmental DNA from water samples after vacuum filtration, performing PCR amplification using fish-specific primers, and performing high-throughput sequencing and bioinformatics analysis on the amplification products to complete the identification of fish species composition.

[0006] Preferably, the step of dividing the operation classification units based on sequence similarity specifically involves using the UPARSE algorithm to divide the operation classification units at a sequence similarity threshold of 99%.

[0007] Preferably, the analysis of the correlation between candidate habitat factors and the elimination of candidate habitat factors with collinearity specifically involves: using Pearson correlation analysis to calculate the correlation coefficient r between candidate habitat factors, and using |r|≥0.7 as the collinearity judgment threshold to eliminate candidate habitat factors with strong collinearity; the candidate habitat factors include at least one of water temperature, water depth, flow velocity, pH value, dissolved oxygen, transparency, altitude, conductivity, and river width.

[0008] Preferably, the analysis of the importance of each target habitat factor to the OTU value and the calculation of relative weights specifically include: When constructing a random forest model, the number of decision trees is set to 500-2000. The percentage increase in mean squared error (%IncMSE) is used to determine the importance of each target habitat factor. The feature importance score of each target habitat factor is obtained, and the ratio of the feature importance score of each target habitat factor to the total score is used as its relative weight.

[0009] Preferably, before using the %IncMSE index to determine the importance of influence, the method further includes performing a permutation importance test on the random forest model: for each target habitat factor, the independent variable is randomly permuted n times, and the random forest model is rerun after each permutation. The model error after permutation is calculated, and the original undisturbed %IncMSE is compared with the %IncMSE distribution after n permutations. The significance p-value is calculated. If the p-value is <0.05, the target habitat factor is retained; otherwise, it is removed. Where n≥200.

[0010] Preferably, the establishment of the scoring rules specifically involves: drawing a partial effect map of each target habitat factor on the OTU value based on the random forest model; using the OTU peak value of the partial effect curve as a benchmark, dividing the quantile interval of the OTU value into at least 5 gradients, corresponding to different levels of fitness index AD; and establishing the correspondence between the target habitat factor value range and the fitness index AD.

[0011] Preferably, the scoring rules are as follows: Using the peak value of the partial effect curve OTU as the 100% baseline, the quantile intervals of OTU values ​​≥ 80%, 60%-80%, 40%-60%, 20%-40%, and < 20% of the peak value were assigned scores of 0.9, 0.7, 0.5, 0.3, and 0.1 respectively, which were used as the adaptation index AD of the corresponding target habitat factor.

[0012] Preferably, the formula for the weighted calculation is: ; In the formula, Fish species i At the sampling point j The habitat suitability at the location, where m is the number of target habitat factors. - The relative weights of each target habitat factor. - AD represents the fitness index corresponding to each target habitat factor.

[0013] Preferably, the method further includes the step of generating a spatial distribution visualization map of fish species habitat suitability based on the habitat suitability calculation results of each sampling point and combined with the geospatial data of the target water area.

[0014] Another aspect of the present invention provides a system for obtaining fish habitat suitability, the system being used to implement the method for obtaining fish habitat suitability, comprising: The sampling and data acquisition module is used to set up sampling points along the target water area, collect water samples, and simultaneously collect candidate habitat factor data at each sampling point. The species identification and OTU statistics module is used to identify the fish species composition of water samples based on environmental DNA technology, annotate the identified fish species, divide the operational taxonomic units based on sequence similarity, and count the OTU values ​​of each fish species. The factor screening and weight calculation module is used to analyze the correlation between candidate habitat factors, eliminate candidate habitat factors with collinearity, and obtain target habitat factors; construct a random forest model, with the OTU value of fish species as the dependent variable and the target habitat factors as the independent variables, to analyze the importance of each target habitat factor on the OTU value, and calculate the relative weight of each target habitat factor in the habitat suitability model. The fitness rule construction module is used for partial effect analysis based on the random forest model to establish the classification and scoring rules of the fitness index (AD) of fish species corresponding to each target habitat factor. The habitat suitability calculation module is used to convert the measured values ​​of target habitat factors at the sampling point into the corresponding fitness index (AD) according to the grading and scoring rules. Combined with the relative weights of each target habitat factor, the habitat suitability of the corresponding fish species at the sampling point is obtained through weighted calculation.

[0015] Preferably, the species identification and OTU statistics module is also used to perform vacuum filtration, environmental DNA extraction, PCR amplification, high-throughput sequencing and bioinformatics analysis on water samples to complete the identification of fish species composition; and to use the UPARSE algorithm to divide the operational classification units at a sequence similarity threshold of 99%.

[0016] Preferably, the factor screening and weight calculation module is further used to calculate the correlation coefficient r between candidate habitat factors using Pearson correlation analysis, using |r|≥0.7 as the collinearity judgment threshold, and eliminating candidate habitat factors with strong collinearity; and to use the percentage increase in mean squared error %IncMSE index to determine the influence importance of each target habitat factor, obtain the characteristic importance score of each target habitat factor, and use the ratio of the characteristic importance score of each target habitat factor to the total score as its relative weight.

[0017] Preferably, the factor screening and weight calculation module is also used to perform a permutation importance test on the random forest model: the independent variable corresponding to each target habitat factor is randomly permuted n times, the random forest model is rerun after each shuffle, the model error after shuffling is calculated, the original undisturbed %IncMSE is compared with the %IncMSE distribution after n shuffles, and the significance p-value is calculated; if the p-value < 0.05, the target habitat factor is retained, otherwise it is removed.

[0018] Preferably, the fitness rule construction module is further used to draw a partial effect map of each target habitat factor on the OTU value based on the random forest model, and divide the quantile interval of the OTU value into at least 5 gradients based on the OTU peak value of the partial effect curve, which correspond to different levels of fitness index AD, and establish the correspondence between the target habitat factor value range and the fitness index AD.

[0019] Preferably, it also includes a visualization output module, which is used to generate a spatial distribution visualization map of the habitat suitability of fish species based on the habitat suitability calculation results of each sampling point and combined with the geospatial data of the target water area.

[0020] Another aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for obtaining fish habitat suitability.

[0021] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for obtaining fish habitat suitability.

[0022] The present invention has the following beneficial effects: 1. This invention achieves non-destructive and highly sensitive monitoring of fish species based on environmental DNA technology, avoiding the disturbance to fish populations caused by traditional fish catch surveys, and significantly improving survey efficiency and species detection rate. It is especially suitable for monitoring rare and endangered fish species and large-scale watershed surveys, and solves the technical bottlenecks of low data accuracy and limited coverage of traditional methods.

[0023] 2. This invention uses the random forest algorithm to analyze the nonlinear response relationship between habitat factors and fish OTU values. It objectively determines the weight of habitat factors based on feature importance indicators and screens significant influencing factors through ranking importance tests. This completely eliminates the reliance of traditional models on expert experience and greatly improves the objectivity, prediction accuracy and generalization ability of the model.

[0024] 3. This invention constructs a single-factor suitability grading and scoring rule based on partial effect analysis, which accurately characterizes the nonlinear response law of fish to changes in habitat factors. Combined with a weighted model, it realizes the quantitative calculation of habitat suitability and can generate a spatial visualization distribution map of habitat suitability. This provides a refined and quantitative scientific basis for the ecological scheduling of cascade reservoirs, the protection and restoration of fish habitats, and the ecological planning of watersheds.

[0025] 4. The method of this invention is standardized and highly repeatable, applicable to habitat suitability assessment for different types of water bodies and different fish species, and has broad engineering application value and ecological protection significance. Attached Figure Description

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] Figure 1 This represents the Pearson correlation coefficient among environmental factors.

[0028] Figure 2 This is a screenshot of part of the program code used to calculate the partial effect data.

[0029] Figure 3 This is a partial effect diagram of the variable on the change in the number of fish OTUs.

[0030] Figure 4 This is a flowchart of the method of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the described embodiments of this disclosure without creative effort are within the scope of protection of this disclosure.

[0032] Example 1: Reference Figure 1 This embodiment provides a method for obtaining fish habitat suitability, including the following steps: S1. Sample Collection and Species Identification: Sampling points were set up along the target water area to collect water samples. Environmental DNA was extracted from the water samples after vacuum filtration. PCR amplification was performed using fish-specific primers. The amplified products were then subjected to high-throughput sequencing and bioinformatics analysis to identify the fish species composition. The fish-specific primers were mitochondrial 12S rRNA primers, with the forward primer sequence being 5'-GTYGGTAAAWCTCGTGCCAGC-3' and the reverse primer sequence being 5'-CATAGTGGGGTATCTAATCCYAGTTTG-3'. Candidate habitat factor data were collected simultaneously from each sampling point. S2, Operational taxonomic unit statistics: The identified fish species are annotated, and the UPARSE algorithm is used to divide the operational taxonomic units at a 99% sequence similarity threshold. The operational taxonomic unit (OTU) values ​​corresponding to each fish species are counted. S3, Habitat Factor Screening and Weight Calculation: Pearson correlation analysis was used to calculate the correlation coefficient r among candidate habitat factors. Using |r|≥0.7 as the collinearity threshold, candidate habitat factors exhibiting strong collinearity were eliminated, resulting in m habitat factors selected for model construction. A random forest model was constructed, with the OTU value of fish species as the dependent variable and the target habitat factors as independent variables. The importance of each target habitat factor to the OTU value was analyzed, and the relative weights of each target habitat factor in the Habitat Suitability (HSI) model were calculated. ; S4, Single-factor fitness rule construction: Based on the partial effect analysis of the random forest model, establish the grading and scoring rules for the fitness index AD of fish species corresponding to each target habitat factor. The specific scoring rules are as follows: Based on the random forest model, a partial effect map of each target habitat factor on the OTU value is drawn. Taking the OTU peak value of the partial effect curve as the benchmark, the quantile interval of the OTU value is divided into at least 5 gradients, which correspond to different levels of fitness index AD. The correspondence between the target habitat factor value range and the fitness index AD is established.

[0033] In this embodiment, the specific values ​​of habitat factors measured at sampling point j are converted into the fitness index of fish species i corresponding to each habitat factor according to the classification and scoring rules established in step 4: ; S5, Quantitative calculation of habitat suitability: The measured values ​​of the target habitat factors at the sampling points are converted into the corresponding fitness index AD according to the grading and scoring rules. Combined with the relative weights of each target habitat factor, the habitat suitability of the corresponding fish species at the sampling point is obtained by weighted calculation.

[0034] The formula for weighted calculation is: ; In the formula, Fish species i At the sampling point j The habitat suitability at the location, where m is the number of target habitat factors. - The relative weights of each target habitat factor. - AD represents the fitness index corresponding to each target habitat factor.

[0035] S6, Visualization: Based on the habitat suitability calculation results of each sampling point, combined with the geospatial data of the target water area, a spatial distribution visualization map of the habitat suitability of fish species is generated.

[0036] Further, in step S3, the candidate habitat factors include at least one of water temperature, water depth, flow rate, pH value, dissolved oxygen, transparency, altitude, electrical conductivity, and river width.

[0037] Furthermore, in step S3, when constructing the random forest model, the number of decision trees is set to 500-2000. The percentage increase in mean squared error (%IncMSE) index is used to determine the importance of each target habitat factor. The feature importance score of each target habitat factor is obtained, and the ratio of the feature importance score of each target habitat factor to the total score is used as its relative weight.

[0038] Further, in step S3, the specific calculation process of relative weights is as follows: the OTU values ​​of m fish species i are used as dependent variables, and habitat factors are used as independent variables in the random forest model. The random forest model is constructed using the "RandomForest" package in R, with the number of trees set to 1000. The %IncMSE index is used to determine the relative importance of each independent variable to the dependent variable. The "importance" function is used to obtain the feature importance score of the model, and the feature importance score of each habitat factor is used as the weight value for weighting.

[0039] Furthermore, before using the %IncMSE index to determine the relative importance of each independent variable to the dependent variable in S3, the following steps are included: using the "rfPermute" package to perform a permutation importance test on the random forest model. For each habitat factor, the independent variable is randomly permuted multiple times. After each permutation, the random forest model is rerun, and the model error is calculated. The original, unpermuted %IncMSE distribution is compared with the %IncMSE distribution after multiple permutations. The p-value is calculated as: p = (number of times %IncMSE after permutation ≥ original %IncMSE + 1) / (299 + 1). If the p-value < 0.05, the corresponding habitat factor in the random forest model is retained; otherwise, the corresponding habitat factor is removed. Here, n ≥ 200. Using the %IncMSE index to determine the relative importance of each independent variable to the dependent variable includes removing factors with low %IncMSE values ​​and insignificant changes in response.

[0040] Furthermore, the scoring rules are as follows: taking the peak value of the partial effect curve OTU as 100% benchmark, the quantile intervals of OTU values ​​≥ 80%, 60%-80%, 40%-60%, 20%-40%, and < 20% of the peak value are respectively assigned scores of 0.9, 0.7, 0.5, 0.3, and 0.1, which are used as the adaptation index AD of the corresponding target habitat factor.

[0041] Example 2: See Figure 1-4 A method for obtaining fish habitat suitability includes the following steps: Step 1: Use eDNA technology to sample and investigate fish. The evaluation was conducted in accordance with the "Technical Guidelines for Freshwater Biological Evaluation Based on Environmental DNA" (T / CSES82-2023) and the "Environmental DNA Macrobarcoding Method for Freshwater Biological Monitoring" (T / CSES81-2023).

[0042] Based on differences in topography, geomorphology, and hydrology, a cross-section (sampling point) was set up every 1 km along the upstream of Reservoir A. At each sampling point, 1000 mL of surface water (approximately 0.3 m deep) was collected using a water sampler and placed in a sealable wide-mouth bottle. Three parallel water samples were collected from each cross-section (sampling point) and stored under cold storage.

[0043] All samples were vacuum filtered within 24 hours using a 0.45 μm mixed cellulose membrane (MCE, Whatman). One negative control was included for each filtration. After filtration, the membranes were immediately frozen and brought back to the laboratory. DNA was extracted from the membranes using a DNA extraction kit (Mobio, Mobio Laboratories, USA), and the DNA solution was stored at -20°C. A blank membrane was used as a negative control to assess for contamination during DNA extraction. Candidate habitat parameters were simultaneously measured. These parameters included: dissolved oxygen (mg / L, DO), water transparency (cm, TR), conductivity (μS / cm, CO), river width at the sampling section (m, RW), water depth at the sampling point (m, WD), flow velocity at the sampling point (m / s, CV), water temperature (°C, WT), altitude (m, AL), and pH.

[0044] An eDNA reference database was established based on fish taxonomic information, fish species information from historical surveys in the Jinsha River, and fish mitochondrial 12S rRNA sequence information from the NCBI nucleotide database.

[0045] PCR amplification of environmental DNA samples was performed using mitochondrial 12S RNA forward and reverse primers: 5'-GTYGGTAAAWCTCGTGCCAGC-3' and 5'-CATAGTGGGGTATCTAATCCYAGTTTG-3'.

[0046] The total reaction volume of 25 μL included 5 μL of 5×PCR Buffer, 5 μL of 5×GC Buffer, 2 μL of dNTPs (2.5 mmol / L), 1 μL of forward primer (10 pmol / μL), 1 μL of reverse primer (10 pmol / μL), 2 μL of DNA template, 0.25 μL of Q5 DNA Polymerase, and 8.75 μL of sterile double-distilled water. Reaction conditions: 98℃ pre-denaturation for 2 min; 98℃ denaturation for 15 sec, 55℃ annealing for 30 sec, 72℃ extension for 30 sec (30 cycles); final extension at 72℃ for 5 min. Sterile water was used as a template for each PCR amplification as a negative control.

[0047] The raw sequencing data underwent preliminary processing steps in the QIIME pipeline. Operational taxonomic units (OTUs) were segmented using the UPARSE algorithm at a similarity threshold of 99%.

[0048] Fish were annotated using the eDNA reference database and the MitoFish database. The number of species, OTUs, and sequence reads at different taxonomic levels were counted, and the relative abundance of species sequences was calculated.

[0049] Step 2: Annotate the fish species, and use the UPARSE algorithm based on the annotations to divide the operational taxonomic units at a similarity threshold of 99%, and count the OTU values ​​of different functional groups; Representative sequences from each OTU were selected, and homologous sequences were searched in the established database using BLAST. Species information that met the set parameters was used as the species annotation information for that sequence. The number of fish species, OTUs, and sequence reads at different taxonomic levels were counted, and the relative abundance of fish species sequences was calculated to obtain a fish relative abundance matrix. Each data point represents the abundance of fish in the r-th column at the t-th row sampling point. The alignment software package used was BLAST+2.6.0, with alignment parameters of sequence consistency and coverage both greater than 80%, and e-value < 1e-5. Results showed that the CytbF / CytbR primers annotated 26 fish species, and the 16SF / 16SR primers also annotated 26 fish species. There were both differences and overlaps in the fish species annotated by the two primer pairs. Examples include: silver carp, common carp, silver carp, Taihu silverfish, schizothorax, bream, topmouth culter, striped loach, goby, long-snout catfish, *Gnaphalium affine*, Yunnan scaly barbel, Chinese barbel, grass carp, bighead carp, Nile tilapia, small-scaled loach, catfish, spiny loach, snakehead, broadfin barbel, yellow catfish, redfin culter, high-bodied bitterling, Chinese carp, and large-scaled loach. Fish species are shown in Table 1.

[0050] Table 1 Fish species

[0051] Step 3: Screening for habitat-sensitive factors The Pearson correlation coefficients (r) among candidate habitat factors for all samples from different sampling points were analyzed. Candidate habitat factors included dissolved oxygen (DO), transparency (TR), conductivity (CO), river width (RW), water depth (WD), flow velocity (CV), water temperature (WT), altitude (AL), and pH. This involved transforming the LOG values ​​of all candidate habitat factors from Table 2 for all samples from different sampling points and then using the `cor()` function to calculate an n1×n2 correlation coefficient matrix. The correlation coefficient matrix consisted of n2 LOG-transformed candidate habitat factors from n1 sampling points. The correlation coefficient ranged from -1 to +1, with positive values ​​indicating positive correlation and negative values ​​indicating negative correlation. |r| ≥ 0.7 was defined as a strong correlation. A color-coded matrix was created using the `corrplot` package, with red indicating positive correlation and blue indicating negative correlation. The intensity of the color indicated the strength of the correlation, and the r values ​​were labeled in the squares. The Pearson correlation coefficients are shown below. Figure 2 As shown.

[0052] Table 2 Candidate habitat factors for all samples from different sampling points

[0053] Based on a predetermined correlation coefficient threshold (|r|>0.7) to determine collinearity, a high positive correlation was found between transparency (TR) and conductivity (CO) (r=0.767). Similarly, river width (RW) was highly positively correlated with water depth (WD), river width (RW) with transparency (TR), and river width (RW) with conductivity (CO). The correlation coefficients were: CV-WD:r=-0.716, WD-RW:r=0.766, RW-TR:r=0.726, RW-CO:r=0.705, TR-CO:r=0.767. Given that transparency more directly reflects the impact of water turbidity on the light environment, and water depth more directly characterizes the ecological effect of water volume on fish than river width, river width (RW) and conductivity (CO) were removed for further analysis based on the ecological significance of environmental factors. Finally, CV (current velocity), AL (altitude), WD (water depth), WT (water temperature), dissolved oxygen (DO), pH (acidity / alkalinity), and transparency (TR) were selected as habitat factors for subsequent model construction.

[0054] Step 4: Construction of fitness curve Random forest was used to determine the weights of each habitat factor. The OTU values ​​of fish species i (i = Schizothorax) at each sampling point (which need to be log-transformed before inputting into the model, approximating fish resource abundance) were used as the dependent variable. The data of the habitat factor independent variables selected in step 3 at the corresponding sampling points were log-transformed and used as independent variables (the original data of the 7 environmental variables selected from Table 2 all need to be log-transformed before inputting into the model) and incorporated into the random forest model. The random forest model was constructed using the "RandomForest" package in R, with 1000 trees. Although the model splits nodes based on indicators such as the Gini coefficient during construction, the %IncMSE (mean squared error increase percentage) indicator was used to determine the relative importance of each independent variable to the dependent variable when evaluating feature importance. The "importance" function was used to obtain the model's feature importance score, and the feature importance scores of each habitat factor were used as weights for weighting. After model calculation, the coefficient of determination (COP) was used... R 2 The percentage increase in mean squared error (%IncMSE) is used to judge the prediction accuracy of the random forest model, ensuring the reliability of the importance of each variable feature.

[0055] Based on the variable importance (%IncMSE) obtained from the random forest model analysis, the "rfPermute" package was used to perform a permutation importance test on the random forest model. With nrep=299, each environmental variable was randomly permuted 299 times. After each permutation, the random forest model was rerun, and the model error was calculated. The distribution of the original, unpermuted %IncMSE was compared with the distribution of the %IncMSE after each 299 permutations. The p-value was calculated as (number of times the permuted %IncMSE ≥ the original %IncMSE + 1) / (299 + 1). If the p-value was small (e.g., <0.05), it indicated that the importance of the variable was unlikely to be randomly generated and was statistically significant.

[0056] As shown in Table 3, the analysis results indicate that all seven selected environmental variables have a significant impact on OTU numbers in *Schizothorax schizothorax*, providing key environmental factors for constructing a habitat suitability (HSI) model. The specific screening principle was to eliminate factors with a weak impact on OTU numbers (i.e., those with low %IncMSE values ​​and insignificant changes). Based on this, the combinations of habitat factors used to construct the HSI model for each group were determined as shown in Table 4: Altitude (AL), water depth (WD), transparency (TR), and current velocity (CV) are the most significant variables affecting the spatial variation of *Schizothorax schizothorax* OTU numbers.

[0057] Table 3 Analysis Results

[0058] Table 4 Weights of Key Environmental Variables

[0059] After determining the combination of habitat factors, the relative weight of each habitat factor in the HSI model is calculated based on its %IncMSE value in the corresponding functional group. The final weight assignments for fish species 1 are shown in Table 4.

[0060] The logarithmic transformation values ​​of the abundance data of fish species i (i = Schizothorax) at each sampling point are used as the y-axis variable, and the logarithmic transformation values ​​of the habitat factor data at the corresponding sampling points are used as the x-axis variable. The partial effect data are calculated using the "pdp" package, and the partial effect plot is drawn using the "ggplot2" package. Using the peak value of the partial effect curve as the upper limit, the partial effect curve is divided into multiple quantile intervals distributed from high to low. Each quantile interval is assigned a fitness index AD from high to low, thus obtaining the fitness index AD corresponding to the numerical range of each habitat factor data. The fish species abundance and environmental factor dataset was compiled using Excel (as shown in Table 3, where 1-10 are the 10 sampling points of the dam). In R software 4.4.1, the model data was imported (as shown in Table 3, the OTU data in column 3 after log transformation, and the data for flow velocity, water depth, transparency, and altitude after log transformation). The partial effect data were calculated using the "pdp" package, and the graphs were plotted using the "ggplot2" package (partial code shown in the image). Figure 2 As shown), the partial effect diagram of the variable changes in the number of OTUs for fish species i was finally obtained. Figure 3 ).

[0061] Based on the highest OTU values ​​(using the actual values ​​after log transformation) of the partial effect plots of major environmental factors affecting each fish species, a grading system was established. Partial dependencies based on significantly correlated factors were obtained from the partial effect plots, and the marginal effects of each dominant factor on the probability of functional group occurrence were analyzed. Using the quantile threshold method, the continuous environmental gradient was divided into five fitness levels: "least suitable – unsuitable – moderately suitable – relatively suitable – highly suitable". Using the peak value of the partial effect curve as a benchmark, five quantile intervals with OTU abundance ≥ 80% of the peak, 60%–80%, 40%–60%, 20%–40%, and <20% were assigned fitness values ​​of 0.9, 0.7, 0.5, 0.3, and 0.1, respectively.

[0062] The specific standards are as follows: (1) If the number of OTUs is 80% or more of the highest value, the suitability score is 0.9; (2) If the number of OTUs is between 60% and 80% of the highest value, the suitability score is 0.7; (3) If the number of OTUs is between 40% and 60% of the highest value, the suitability score is 0.5; (4) If the number of OTUs is between 20% and 40% of the highest value, the suitability score is 0.3; (5) If the number of OTUs is less than 20% of the maximum value, the suitability score is 0.1.

[0063] Table 5. Suitability values ​​of environmental factors

[0064] Subsequently, the fitness value based on the partial effect diagram grading was used as the fitness index AD for the schizothorax.

[0065] Step 5: Generate the habitat adaptability index of the schizothorax at sampling point j. The specific values ​​of habitat factors measured at the sampling points are converted into single-factor fitness values ​​for fish species i corresponding to each habitat factor, based on the grading and scoring rules established in step 4. , , , , , ; Based on the weights of each variable for fish species i obtained in step 4, the habitat suitability of fish species i (Schizothorax) at sampling point j is calculated using a weighted average method. ; ; In the formula, i The number indicating the species of fish. j This indicates the number of the sampling point.

[0066] Example 3: This embodiment provides a system for obtaining fish habitat suitability, the system being used to implement the method for obtaining fish habitat suitability, including: The sampling and data acquisition module is used to set up sampling points along the target water area, collect water samples, and simultaneously collect candidate habitat factor data at each sampling point. The species identification and OTU statistics module is used to identify the fish species composition of water samples based on environmental DNA technology, annotate the identified fish species, divide the operational taxonomic units based on sequence similarity, and count the OTU values ​​of each fish species. The factor screening and weight calculation module is used to analyze the correlation between candidate habitat factors, eliminate candidate habitat factors with collinearity, and obtain target habitat factors; construct a random forest model, with the OTU value of fish species as the dependent variable and the target habitat factors as the independent variables, to analyze the importance of each target habitat factor on the OTU value, and calculate the relative weight of each target habitat factor in the habitat suitability model. The fitness rule construction module is used for partial effect analysis based on the random forest model to establish the classification and scoring rules of the fitness index (AD) of fish species corresponding to each target habitat factor. The habitat suitability calculation module is used to convert the measured values ​​of target habitat factors at the sampling point into the corresponding fitness index (AD) according to the grading and scoring rules. Combined with the relative weights of each target habitat factor, the habitat suitability of the corresponding fish species at the sampling point is obtained through weighted calculation.

[0067] Preferably, the species identification and OTU statistics module is also used to perform vacuum filtration, environmental DNA extraction, PCR amplification, high-throughput sequencing and bioinformatics analysis on water samples to complete the identification of fish species composition; and to use the UPARSE algorithm to divide the operational classification units at a sequence similarity threshold of 99%.

[0068] Preferably, the factor screening and weight calculation module is further used to calculate the correlation coefficient r between candidate habitat factors using Pearson correlation analysis, using |r|≥0.7 as the collinearity judgment threshold, and eliminating candidate habitat factors with strong collinearity; and to use the percentage increase in mean squared error %IncMSE index to determine the influence importance of each target habitat factor, obtain the characteristic importance score of each target habitat factor, and use the ratio of the characteristic importance score of each target habitat factor to the total score as its relative weight.

[0069] Preferably, the factor screening and weight calculation module is also used to perform a permutation importance test on the random forest model: the independent variable corresponding to each target habitat factor is randomly permuted n times, the random forest model is rerun after each shuffle, the model error after shuffling is calculated, the original undisturbed %IncMSE is compared with the %IncMSE distribution after n shuffles, and the significance p-value is calculated; if the p-value < 0.05, the target habitat factor is retained, otherwise it is removed.

[0070] Preferably, the fitness rule construction module is further used to draw a partial effect map of each target habitat factor on the OTU value based on the random forest model, and divide the quantile interval of the OTU value into at least 5 gradients based on the OTU peak value of the partial effect curve, which correspond to different levels of fitness index AD, and establish the correspondence between the target habitat factor value range and the fitness index AD.

[0071] Preferably, it also includes a visualization output module, which is used to generate a spatial distribution visualization map of the habitat suitability of fish species based on the habitat suitability calculation results of each sampling point and combined with the geospatial data of the target water area.

[0072] Example 4: This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method for obtaining fish habitat suitability.

[0073] Example 5: This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0074] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this invention. It should be understood that the above descriptions are merely specific embodiments of this invention and are not intended to limit the scope of protection of this invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for obtaining fish habitat suitability, characterized in that, Includes the following steps: Sampling points were set up along the target water area to collect water samples. The fish species composition in the water was identified based on environmental DNA technology, and candidate habitat factor data were collected at each sampling point simultaneously. The identified fish species were annotated, and operational taxonomic units were divided based on sequence similarity. The values ​​of the operational taxonomic units (OTUs) for each fish species were statistically analyzed. The correlation between candidate habitat factors was analyzed, and candidate habitat factors with collinearity were eliminated to obtain target habitat factors. A random forest model was constructed with the OTU value of fish species as the dependent variable and the target habitat factors as independent variables. The importance of each target habitat factor to the OTU value was analyzed, and the relative weight of each target habitat factor in the habitat suitability model was calculated. Based on the partial effect analysis of the random forest model, a scoring rule for the adaptation index (AD) of fish species corresponding to each target habitat factor was established. The measured values ​​of the target habitat factors at the sampling points are converted into the corresponding fitness index (AD) according to the grading and scoring rules. The habitat suitability of the corresponding fish species at the sampling point is obtained by weighted calculation based on the relative weights of each target habitat factor.

2. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The identification of fish species composition in aquatic bodies based on environmental DNA technology specifically includes: extracting environmental DNA from water samples after vacuum filtration, performing PCR amplification using fish-specific primers, and performing high-throughput sequencing and bioinformatics analysis on the amplification products to complete the identification of fish species composition.

3. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The operation classification unit based on sequence similarity is specifically defined as follows: the UPARSE algorithm is used to divide the operation classification unit at a sequence similarity threshold of 99%.

4. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The analysis of the correlation between candidate habitat factors and the elimination of candidate habitat factors with collinearity specifically involves: using Pearson correlation analysis to calculate the correlation coefficient r between candidate habitat factors, and using |r|≥0.7 as the threshold for collinearity judgment to eliminate candidate habitat factors with strong collinearity; the candidate habitat factors include at least one of water temperature, water depth, flow velocity, pH value, dissolved oxygen, transparency, altitude, electrical conductivity, and river width.

5. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The analysis of the importance of each target habitat factor to the OTU value and the calculation of relative weights specifically include: When constructing a random forest model, the number of decision trees is set to 500-2000. The percentage increase in mean squared error (%IncMSE) is used to determine the importance of each target habitat factor. The feature importance score of each target habitat factor is obtained, and the ratio of the feature importance score of each target habitat factor to the total score is used as its relative weight.

6. The method for obtaining fish habitat suitability according to claim 5, characterized in that, Before using the %IncMSE index to determine the importance of influence, a permutation importance test is also performed on the random forest model: the independent variable corresponding to each target habitat factor is randomly permuted n times, and the random forest model is rerun after each shuffle. The model error after shuffling is calculated, and the original undisturbed %IncMSE is compared with the %IncMSE distribution after n shuffles. The significance p-value is calculated. If the p-value is <0.05, the target habitat factor is retained; otherwise, it is removed.

7. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The establishment of the scoring rules is specifically as follows: Based on the random forest model, a partial effect map of each target habitat factor on the OTU value is drawn. Taking the OTU peak value of the partial effect curve as the benchmark, the quantile interval of the OTU value is divided into at least 5 gradients, which correspond to different levels of fitness index AD. The correspondence between the target habitat factor value range and the fitness index AD is established.

8. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The specific rules for grading and assigning scores are as follows: Using the peak value of the partial effect curve OTU as the 100% baseline, the quantile intervals of OTU values ​​≥ 80%, 60%-80%, 40%-60%, 20%-40%, and < 20% of the peak value were assigned scores of 0.9, 0.7, 0.5, 0.3, and 0.1 respectively, which were used as the adaptation index AD of the corresponding target habitat factor.

9. The method for obtaining fish habitat suitability according to claim 1, characterized in that, The formula for the weighted calculation is: ； In the formula, Fish species i At the sampling point j The habitat suitability at the location, where m is the number of target habitat factors. - The relative weights of each target habitat factor. - AD represents the fitness index corresponding to each target habitat factor.

10. The method for obtaining fish habitat suitability according to claim 1, characterized in that, It also includes the step of generating a spatial distribution visualization map of fish species habitat suitability based on the habitat suitability calculation results of each sampling point and combined with the geospatial data of the target water area.

11. A system for obtaining fish habitat suitability, characterized in that, The system is used to implement the method for obtaining fish habitat suitability as described in any one of claims 1-10, including: The sampling and data acquisition module is used to set up sampling points along the target water area, collect water samples, and simultaneously collect candidate habitat factor data at each sampling point. The species identification and OTU statistics module is used to identify the fish species composition of water samples based on environmental DNA technology, annotate the identified fish species, divide the operational taxonomic units based on sequence similarity, and count the OTU values ​​of each fish species. The factor screening and weight calculation module is used to analyze the correlation between candidate habitat factors, eliminate candidate habitat factors with collinearity, and obtain target habitat factors; construct a random forest model, with the OTU value of fish species as the dependent variable and the target habitat factors as the independent variables, to analyze the importance of each target habitat factor on the OTU value, and calculate the relative weight of each target habitat factor in the habitat suitability model. The fitness rule construction module is used for partial effect analysis based on the random forest model to establish the classification and scoring rules of the fitness index (AD) of fish species corresponding to each target habitat factor. The habitat suitability calculation module is used to convert the measured values ​​of target habitat factors at the sampling point into the corresponding fitness index (AD) according to the grading and scoring rules. Combined with the relative weights of each target habitat factor, the habitat suitability of the corresponding fish species at the sampling point is obtained through weighted calculation.

12. The system for obtaining fish habitat suitability according to claim 11, characterized in that, The species identification and OTU statistics module is also used to perform vacuum filtration, environmental DNA extraction, PCR amplification, high-throughput sequencing and bioinformatics analysis on water samples to complete the identification of fish species composition; and uses the UPARSE algorithm to divide operational taxonomic units at a 99% sequence similarity threshold.

13. The system for obtaining fish habitat suitability according to claim 11, characterized in that, The factor screening and weight calculation module is also used to calculate the correlation coefficient r between candidate habitat factors using Pearson correlation analysis, and to remove candidate habitat factors with strong collinearity by using |r|≥0.7 as the collinearity judgment threshold; and to use the percentage increase in mean squared error %IncMSE index to determine the influence importance of each target habitat factor, obtain the characteristic importance score of each target habitat factor, and use the ratio of the characteristic importance score of each target habitat factor to the total score as its relative weight.

14. The system for obtaining fish habitat suitability according to claim 13, characterized in that, The factor selection and weight calculation module is also used to perform a permutation importance test on the random forest model: the independent variable corresponding to each target habitat factor is randomly permuted n times, and the random forest model is rerun after each permutation. The model error after permutation is calculated, and the original undisturbed %IncMSE is compared with the %IncMSE distribution after n permutations. The significance p-value is calculated. If the p-value is <0.05, the target habitat factor is retained; otherwise, it is removed. Where n≥200.

15. The system for obtaining fish habitat suitability according to claim 11, characterized in that, The fitness rule construction module is also used to draw a partial effect map of each target habitat factor on the OTU value based on the random forest model. Taking the OTU peak value of the partial effect curve as the benchmark, the quantile interval of the OTU value is divided into at least 5 gradients, which correspond to different levels of fitness index AD, and the correspondence between the target habitat factor value range and the fitness index AD is established.

16. The system for obtaining fish habitat suitability according to claim 11, characterized in that, It also includes a visualization output module, which is used to generate a spatial distribution visualization map of the habitat suitability of fish species based on the habitat suitability calculation results of each sampling point and combined with the geospatial data of the target water area.

17. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 10.

18. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 10.

Images