Toxic element pollution ecological risk evaluation method based on soil nematode ecological index

Abstract

The invention provides a toxic element pollution ecological risk evaluation method based on soil nematode ecological indexes, and belongs to the technical field of toxic element pollution ecological risk evaluation. The technical problem to be solved is to provide a toxic element pollution ecological risk evaluation method based on soil nematode ecological indexes. According to the technical scheme, multi-period and multi-section sampling is conducted on soil polluted by toxic elements of different degrees in a region, and the sampled soil is used for soil potential toxic element content detection, soil physicochemical property detection and soil nematode community structure analysis; respectively calculating characteristic indexes of potential toxic element pollution in the soil, wherein the characteristic indexes comprise a pollution factor, a Nemero index, a pollution load index and a potential ecological risk index; analyzing a soil nematode community structure, and respectively calculating evaluation indexes of the soil nematode community structure; the method is applied to toxic element pollution ecological risk evaluation.

Description

Technical Field

[0001] The invention provides a method for evaluating ecological risk of toxic element pollution based on soil nematode ecological index, belonging to the technical field of ecological risk evaluation of toxic element pollution. Background Art

[0002] In recent years, the problem of soil pollution by potentially toxic elements has become prominent, which can threaten the stability of the ecosystem and human health through the coupled cascade reaction of soil-food-environment-health. It is necessary to attach great importance to the prevention and control and early warning of soil pollution by potentially toxic elements, actively carry out soil pollution ecological risk assessment work, and establish more effective ecological risk assessment methods.

[0003] At present, the evaluation methods for the ecological risks of pollution by potential toxic elements often assess the pollution level and ecological risks of potential toxic elements through the content of potential toxic elements in the soil, such as single factor pollution index, geoaccumulation index, Nemerow index, pollution load index and potential ecological risk index, all of which are based on the content of potential toxic elements in the soil for risk calculation, and rarely consider or do not consider the biological toxicity of potential toxic elements and their impact on the structure of biological communities. The existing methods cannot capture all ecological risks and threats at the ecosystem level, and sometimes even overestimate the risks. In addition, traditional physical and chemical indicators have problems such as lag, accumulation, single pollutant assessment limitations, cost-effectiveness issues and limitations in monitoring unknown chemicals in soil pollution monitoring and assessment, which makes it impossible to fully, timely and accurately reflect the ecological effects and comprehensive risks of soil pollution.

[0004] Studies have found that soil nematodes are sensitive to environmental changes such as soil pollution by potentially toxic elements. They are rich in species, have short generation times, occupy multiple nutritional niches and are easy to separate and identify. They are considered to be a good indicator animal group. Studies have shown that soil pollution by potentially toxic elements can significantly change the community structure of soil nematodes. Using soil animals to monitor and assess environmental quality is more comprehensive and accurate than traditional soil potentially toxic elements and soil physical and chemical indicators. New developments in soil animal biotechnology will help establish assessment indicators for soil and environmental health.

[0005] However, soil ecosystems are quite complex. Current research mainly focuses on the toxicological research and changes in community structure of animals such as soil nematodes under potential toxic element pollution, and the community structure of soil nematodes is easily affected by environmental factors. Although studies have found nematode ecological indices that are significantly correlated with soil potential toxic element pollution, the dose-response relationship between the two in the wild environment is not clear (not a linear relationship), and no relevant research has been found to use soil nematodes to establish a systematic ecological risk model for potential toxic element pollution. The community structure of soil nematodes will be affected by a variety of environmental factors other than the stress of potential toxic elements, such as soil properties, climatic conditions, crops, etc., and there are often multiple high concentrations of potential toxic elements at the same time. The mutual influence between potential toxic elements and potential toxic elements, and between potential toxic elements and various environmental factors makes it impossible to find and determine the accurate dose-response relationship between potential toxic elements and soil nematodes from complex environmental factors, and thus it is impossible to judge the ecological risk of potential toxic element pollution. Summary of the invention

[0006] In order to overcome the deficiencies in the prior art, the present invention aims to solve the technical problem of providing an ecological risk assessment method for toxic element pollution based on soil nematode ecological index.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: an ecological risk assessment method of toxic element pollution based on soil nematode ecological index, comprising the following assessment steps:

[0008] Step 1: Sampling the soil in the region that is contaminated by potentially toxic elements to varying degrees is carried out at multiple time periods and multiple locations, and the sampled soil is used for testing the content of potentially toxic elements in the soil, testing the physical and chemical properties of the soil, and analyzing the structure of the soil nematode community;

[0009] Step 2: Detect the content of potentially toxic elements in the soil;

[0010] Step 3: Calculate the characteristic indexes of potential toxic element pollution in the soil, including: pollution factor, Nemerow index, pollution load index, and potential ecological risk index;

[0011] Step 4: Analyze the soil nematode community structure and calculate the evaluation indicators of the soil nematode community structure, including: nematode abundance, proportion of different soil nematodes, Shannon-Wiener diversity index, soil nematode evenness index, nematode maturity index, nematode channel index, enrichment index, and structure index;

[0012] Step 5: Use Bayesian kernel machine regression to analyze the dose-response relationship between the content of soil potential toxic elements, soil nematode community structure, and pollution characteristic index;

[0013] Step 6: Based on Bayesian kernel machine regression analysis, the nematode ecological index and pollution characteristic evaluation index sensitive to potential toxic elements are obtained, and multiple computational models are used to construct ecological risk prediction models of potential toxic element pollution based on soil nematode ecological index. The computational models used include: linear regression model, ridge regression model, random forest model, gradient boosting regression tree model, decision tree model, and extreme random tree model;

[0014] Step 7: Evaluate the constructed ecological risk prediction models and select the best fitting model.

[0015] The potentially toxic elements sampled in the contaminated soil in step one specifically include Zn, Cu, Cr, Cd, Ni, Mn, Hg, Pb, As, Sb, and Se.

[0016] The specific method for detecting the content of potentially toxic elements in the soil in step 2 is:

[0017] Weigh 0.2 g of air-dried soil sample that has passed through a 100-mesh sieve into a polytetrafluoroethylene digestion tube TF-36, add 10 mL of hydrochloric acid, and insert it into the digestion hole;

[0018] Set the digestion instrument temperature to 100°C, heat until a small amount of sample remains, then remove and cool;

[0019] Add 5 mL of nitric acid, 5 mL of hydrofluoric acid, and 1 mL of perchloric acid and cover;

[0020] Set the digestion instrument temperature to 150°C, digest for 120 min, and open the lid after cooling;

[0021] Raise the temperature to 200℃ and wait until the thick white smoke disappears;

[0022] Steam until the digestion solution becomes viscous and repeat the above digestion process;

[0023] Remove the digestion tube and cool it, add 2mL of hydrochloric acid to dissolve the soluble residue, rinse the inner wall with deionized water, transfer the entire amount to a 50mL volumetric flask, and shake well;

[0024] Inductively coupled plasma mass spectrometry was used for detection, in which Hg was not driven out of acid during digestion to prevent Hg volatilization, and atomic absorption spectroscopy was used to detect the content of potentially toxic elements in the soil.

[0025] The specific method for calculating the potential toxic element pollution characteristic index in step 3 is:

[0026] In order to assess the degree of potential toxic element contamination, the contamination factor CF is used for evaluation. The calculation formula of this factor is:

[0027]

[0028] Where: Ci represents the concentration of potentially toxic elements at sampling point element i, C o is the background value;

[0029] The grading criteria for the pollution factor CF are defined as:

[0030] When CF < 1, it indicates mild soil pollution;

[0031] When 1 ≤ CF < 3, it indicates moderate soil pollution;

[0032] When 3 ≤ CF < 6, it indicates high soil pollution;

[0033] When CF > 6, it indicates extremely high soil pollution;

[0034] To evaluate the comprehensive index of combined potentially toxic element pollution, the Nemerow index NSPI is used for evaluation. The calculation formula of this index is:

[0035]

[0036] Among them: CF imax is the maximum value among all single pollution indices CF of a certain potentially toxic element in a certain area; CF iave is the average value of all single pollution indices CF of a certain potentially toxic element in a certain area;

[0037] The grading criteria for the Nemerow index NSPI are defined as:

[0038] When NSPI ≤ 0.7, the pollution of potentially toxic elements in this area is within the safe range;

[0039] When 0.7 < NSPI ≤ 1.0, the pollution of potentially toxic elements in this area is within the warning range;

[0040] When 1.0 < NSPI ≤ 2.0, the pollution of potentially toxic elements in this area is at a slight pollution level;

[0041] When 2.0 < NSPI ≤ 3.0, the pollution of potentially toxic elements in this area is at a moderate pollution level;

[0042] When NSPI > 3.0, the pollution of potentially toxic elements in this area is at a severe pollution level;

[0043] According to the toxicity of each potentially toxic element, the combined potentially toxic element pollution load of the study area is evaluated. The pollution load index PLI is used for evaluation. The calculation formula of this index is:

[0044]

[0045] Among them, CF imis the maximum value among all single pollution indices CF of a certain potentially toxic element, PLI i is the pollution load index of a certain sampling point, and n is the number of potentially toxic elements to be evaluated, PLI zone is the pollution load index of a certain sampling area;

[0046] Define the grading standard of the pollution load index PLI as:

[0047] When PLI ≤ 1, it is pollution-free;

[0048] When 1 < PLI ≤ 2, it is lightly polluted;

[0049] When 2 < PLI ≤ 3, it is moderately polluted;

[0050] When PLI > 3, it is highly polluted;

[0051] To evaluate the synergistic effect among multiple potentially toxic elements and the potential degree of ecological harm, the potential ecological risk index RI is used for evaluation. The calculation formula of this index is:

[0052] E i = T i × CF i ;

[0053]

[0054] Among them: E i is the ecological harm coefficient of a certain potentially toxic element in a certain sampling point, and T i is the toxicity response coefficient of a certain potentially toxic element;

[0055] Grade the potential ecological risk index, and the grading rule is:

[0056] When E i < 40 or RI < 150, it is defined as a low risk;

[0057] When 40 ≤ E i < 80 or 150 ≤ RI < 300, it is defined as a medium risk;

[0058] When 80 ≤ E i < 160 or 300 ≤ RI < 600, it is defined as a relatively high risk;

[0059] When 160 ≤ E i < 320 or RI ≥ 600, it is defined as a high risk;

[0060] When E i ≥ 320, it is defined as an extremely high risk.

[0061] The specific method for calculating the evaluation index of the soil nematode community structure in the fourth step described above is:

[0062] When calculating nematode abundance, it was expressed as the total number of nematodes per 100 g of dry soil;

[0063] The percentage of plant-parasitic nematodes, bacteria-feeding nematodes, fungi-feeding nematodes, and predatory omnivorous nematodes in the total number of nematodes was calculated for each trophic group in 100 g of dry soil;

[0064] Calculate the Shannon-Wiener diversity index H, the calculation formula is:

[0065] H=-∑P i (InP i );

[0066] Among them, P i is the proportion of individuals in the i-th taxonomic unit to the total number of nematodes;

[0067] In order to illustrate the spatial characteristics of nematode distribution in the soil, the soil nematode uniformity E was calculated using the following formula:

[0068] E = H / InS;

[0069] The maturity index MI of free-living nematodes was calculated using the following formula:

[0070] MI=∑v i ×f i ;

[0071] Among them, v i is the cp value assigned according to the different life strategies of free-living nematodes in ecological succession, f i is the ratio of the number of individuals of the i-th nematode to the total number of individuals in the community;

[0072] The plant parasitic nematode maturity index (PPI) was calculated using the following formula:

[0073] PPI=∑v i ×f i ;

[0074] Among them, v i is the cp value of plant parasitic nematodes, f i is the ratio of the number of individuals of the i-th nematode to the total number of individuals in the community;

[0075] The ratio of the number of bacteria-feeding nematodes to that of micro-feeding nematodes was calculated, and the formula for calculating the nematode channel index NCR was:

[0076] NCR = Ba / (Ba + Fu);

[0077] In order to assess the response of the food web to available resources, the enrichment index EI was calculated using the following formula:

[0078] EI = 100 × e / (e + b);

[0079] Among them, according to the cp value of free-living nematodes, b represents the basic component in the food web, referring to the two major groups Ba2 and Fu2; e represents the enriched component in the food web, referring to the two groups Ba1 and Fu2;

[0080] The calculation formulas for b and e values ​​are:

[0081] b=∑k b n b ;

[0082] e=∑k e n e ;

[0083] Where: k b , k e is the weighted value corresponding to each group, n b 、n e is the abundance of each group;

[0084] In order to reflect the changes in soil food web structure during human disturbance or ecological restoration, the structural index SI was calculated using the following formula:

[0085] SI = 100 × s / (s + b);

[0086] Where: s represents the structural components in the food web, including Ba3-Ba5, Fu3-Fu5, Om3-Om5, and Ca3-Ca5 groups;

[0087] The calculation formula of s value is:

[0088] s=∑k s n s ;

[0089] Where: k s is the weighted value corresponding to each group, n s is the abundance of each group.

[0090] The specific method of using Bayesian kernel machine regression in step 5 to analyze the dose-response relationship between the content of soil potential toxic elements, soil nematode community structure, and pollution characteristic index is:

[0091] By analyzing the dose-response relationship among the content of potentially toxic elements in soil, the structure of soil nematode communities, and the pollution characteristic index, the nematode ecological index sensitive to potentially toxic elements and the pollution characteristic evaluation index were determined. The calculation formula is:

[0092] Y i =h(z i1 , ..., zim )+βx i +e i ;

[0093] Where: Y i is the soil nematode ecological index of the i-th sample; z i1 , ..., z im are the m potentially toxic elements in soil corresponding to the i-th sample; x i is a group of potential confounding factors; h (·) is the exposure-response function; β is the covariate x i The effect of i is the residual.

[0094] The specific method for constructing the ecological risk prediction model of potential toxic element pollution based on soil nematode ecological index in step 6 is:

[0095] The data set was randomly divided into two parts, 70% for model training and 30% for model testing. The parameters of each model were set to the default parameters.

[0096] The linear regression model is used to build a prediction model. The expression of the linear regression model is:

[0097]

[0098] Among them, n is the dimension of the sampling point, m is the number of the sampling point, θ0 is the intercept of the regression equation, θ j is the coefficient corresponding to the feature in the regression equation, X ij is the feature vector of the sampling point, y i is the actual value, is the predicted value, L is the value of the loss function;

[0099] The ridge regression model is used to build a prediction model. The expression of the ridge regression model is:

[0100]

[0101] Where: n is the dimension of the sampling point; m is the number of the sampling point; θ0 is the intercept of the regression equation; θ j is the coefficient corresponding to the feature in the regression equation; X ij is the feature vector of the sampling point; y i is the actual value; is the predicted value; α is the structural characteristic coefficient; L is the value of the loss function;

[0102] A random forest model was used to construct a prediction model, with the nematode ecological index as the predictor variable and ecological risk as the outcome variable, implemented using the randomForest package in R software;

[0103] A gradient boosting regression tree model was used to construct a prediction model, with the nematode ecological index as the predictor variable and ecological risk as the outcome variable, implemented using the caret package in R software;

[0104] A decision tree model was used to construct a prediction model, with the nematode ecological index as the predictor variable and ecological risk as the outcome variable, which was implemented using the rpart package in R software;

[0105] The extreme random tree model was used to construct the prediction model, with the nematode ecological index as the predictor variable and the ecological risk as the outcome variable, which was implemented using the extraTrees package in R software.

[0106] The specific method for selecting the best fitting model in step seven is:

[0107] The mean absolute error (MAE), root mean square error (RMSE), and goodness of fit (R) were used. 2 To evaluate the accuracy of each prediction model, the calculation formula is:

[0108]

[0109] Where m is the number of sample sites, y i is the real value of the output label, is the predicted value of the output label, is the average of the output labels;

[0110] According to the calculation results, the smaller the MAE and RMSE, the higher the prediction accuracy. 2 Used to evaluate the fitting accuracy of modeling, validation set R 2 Used to evaluate prediction accuracy and model generalization ability.

[0111] The beneficial effects of the present invention compared with the prior art are as follows: the present invention uses Bayesian kernel machine regression to analyze the dose-response relationship among the content of potentially toxic elements in soil, the structure of soil nematode community, and the pollution characteristic index, and obtains the nematode ecological index and the pollution characteristic index sensitive to the potentially toxic elements. The nematode ecological index sensitive to the potentially toxic elements is used as a predictor variable, and the Nemerow index, the pollution load index, and the potential ecological risk index are used as outcome variables. A variety of machine learning methods are used to construct an ecological risk assessment model for toxic element pollution based on the soil nematode ecological index, and model verification and accuracy evaluation are performed to select the most suitable model, thereby achieving the purpose of using the soil nematode ecological index to evaluate the ecological risk of potentially toxic elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] The present invention will be further described below in conjunction with the accompanying drawings:

[0113] Figure 1The present invention is a flowchart of the steps of the method for ecological risk assessment of toxic element pollution. DETAILED DESCRIPTION

[0114] like Figure 1 As shown, the present invention evaluates the risk of ecological pollution by potential toxic elements based on the detection and analysis of soil nematode ecological index. The specific method for risk assessment includes the following steps:

[0115] Collecting soil samples:

[0116] Sampling was carried out for soils contaminated with different degrees of potential toxic elements in coal mining areas, coal transportation roads and far away from other industrial pollution environments. There were 8-10 sampling points in each area. Mixed samples were taken in a zigzag pattern at each sampling point, and 5 replicates were taken. The sampling points in each area included farmland sampling points for planting wheat, corn and vegetables. Soil was collected from the same sampling points in spring and autumn. When sampling, plant debris and other debris on the surface of the soil were first removed. The soil at the points where plants grew should be loosened first to remove the plants and their roots. Gravel and other foreign matter in the soil samples were removed with a bamboo shovel at the sampling site. Pay attention to cleaning the sampling tools in time to avoid cross contamination. Try to use bamboo shovels and bamboo pieces to take samples directly, and bring the soil samples back to the laboratory for processing within 12 hours after collection. The fresh soil samples were divided into three parts, which were used for the detection of potential toxic elements in the soil, the detection of soil physical and chemical properties, and the analysis of soil nematode community structure.

[0117] Determination of the content of potentially toxic elements Zn, Cu, Cr, Cd, Ni, Mn, Hg, Pb, As, Sb, Se in soil:

[0118] The present invention specifically adopts microwave digestion, inductively coupled plasma mass spectrometry ICP-MS, and atomic absorption spectrometry to detect the content of potentially toxic elements in soil:

[0119] Weigh 0.2g of air-dried soil sample that has passed through a 100-mesh sieve into a polytetrafluoroethylene digestion tube TF-36, add 10mL of hydrochloric acid HCl, and insert it into the digestion hole. Set the temperature of the digestion instrument to 100℃, heat until a small amount of sample remains, then remove and cool. Add 5mL of nitric acid HNO3, 5mL of hydrofluoric acid HF, and 1mL of perchloric acid HClO4, and cover. Set the temperature of the digestion instrument to 150℃, digest for 120min, and open the cover after cooling. Heat to 200℃ until the thick white smoke is gone. Steam until the digestion solution is viscous and repeat the above digestion process. Remove the digestion tube and cool it, add 2mL of hydrochloric acid to dissolve the soluble residue, rinse the inner wall with deionized water, transfer the entire amount to a 50mL volumetric flask to make up the volume, and shake well. Detect using inductively coupled plasma mass spectrometry ICP-MS. Hg is not driven out of acid during digestion to prevent Hg from volatilizing, and is determined by atomic absorption spectroscopy.

[0120] Calculate the soil potential toxic element pollution characteristic index, including:

[0121] (1) The contamination factor (CF) is widely used to evaluate the degree of potential toxic element pollution and is considered the basis for environmental quality assessment. The calculation formula is as follows:

[0122]

[0123] C i represents the concentration of potential toxic element i at the sampling point (mg / kg), and C o is the background value (in this invention, the concentrations of various potential toxic elements in the control area are used as the background value). When CF < 1, it indicates mild soil pollution; when 1 ≤ CF < 3, it indicates moderate soil pollution; when 3 ≤ CF < 6, it indicates high soil pollution; when CF > 6, it indicates extremely high soil pollution.

[0124] (2) The Nemerow synthetic pollution index (NSPI) is used as a comprehensive index for composite potential toxic element pollution. The calculation formula is as follows:

[0125]

[0126] CF imax is the maximum value among all single - pollution indices CF of a certain potential toxic element in a certain area; CF iave is the average value of all single - pollution indices CF of a certain potential toxic element in a certain area. When the Nemerow index NSPI ≤ 0.7, it indicates that the potential toxic element pollution in this area is within the safe range; when 0.7 < NSPI ≤ 1.0, it is within the warning range; when 1.0 < NSPI ≤ 2.0, it is at a slightly polluted level; when 2.0 < NSPI ≤ 3.0, it is at a moderately polluted level; when NSPI > 3.0, it is at a severely polluted level.

[0127] (3) The metalion load index (PLI) is used to evaluate the composite potential toxic element pollution load in the study area according to the toxicity of each potential toxic element. It is expressed by the formula:

[0128]

[0129] CF im is the maximum value among all single - pollution indices CF of a certain potential toxic element, PLI i is the pollution load index of a certain sampling point, n is the number of types of potential toxic elements to be evaluated, PLI zoneis the pollution load index of a certain sampling area. The grading standard of the pollution load index is: PLI ≤ 1, no pollution; 1 < PLI ≤ 2, slight pollution; 2 < PLI ≤ 3, moderate pollution; PLI > 3, high pollution.

[0130] (4) The potential ecological risk index (RI) can be used to evaluate the synergistic effect among multiple potentially toxic elements and the degree of potential ecological hazard. The calculation formula is:

[0131] E i = T i × CF i ;

[0132]

[0133] E i is the ecological hazard coefficient of a certain potentially toxic element at a certain sampling point, and T i is the toxicity response coefficient of a certain potentially toxic element. The present invention adopts the toxicity response coefficients in the research of Hakanson: Ti = 1; V = 2; Cr = 2; Mn = 1; Ni = 5; Cu = 5; Zn = 1; As = 10; Cd = 30; Hg = 40; Tl = 40; Pb = 5. The grading of the potential ecological risk index is: E i <40 or RI < 150, slight risk; 40 ≤ E i <80 or 150 ≤ RI < 300, medium risk; 80 ≤ E i <160 or 300 ≤ RI < 600, relatively high risk; 160 ≤ E i <320 or RI ≥ 600, high risk; E i ≥ 320, extremely high risk.

[0134] Analysis is carried out on the soil nematode community structure:

[0135] Firstly, the isolation, collection and counting of soil nematodes are carried out. Specifically, the nematodes in 200 g of soil are isolated and collected by the sucrose centrifugation method, and the total number of nematodes is counted through an Olympus ZX10 stereomicroscope.

[0136] Then, the evaluation indexes of the nematode community structure are calculated, including:

[0137] (1) The nematode abundance is expressed by the total number of nematodes per 100 g of dry soil.

[0138] (2) The relative abundances of each trophic group are respectively the percentages of plant parasitic nematodes, bacterivorous nematodes, fungivorous nematodes, and predatory omnivorous nematodes in the total number of nematodes per 100 g of dry soil.

[0139] (3) Shannon-Wiener diversity index H:

[0140] H=-∑P i (InP i );

[0141] Where P i is the proportion of individuals in the i-th taxonomic unit to the total number of nematodes.

[0142] (4) Soil nematode evenness can reflect the composition structure and stability of the nematode community and explain the spatial characteristics of nematode distribution in the soil (Pielou evenness index). The calculation formula is:

[0143] E = H / InS;

[0144] (5) Maturity index (MI) of free-living nematodes, calculated as:

[0145] MI=∑v i ×f i ;

[0146] Among them, v i is the cp value assigned according to the different life strategies of free-living nematodes in ecological succession, f i is the ratio of the number of individuals of the i-th nematode species to the total number of individuals in the community.

[0147] (6) Plant parasite nematode maturity index PPI (plant parasite index), the calculation formula is:

[0148] PPI=∑v i ×f i ;

[0149] Among them, v i is the cp value of plant parasitic nematodes, f i Same as above.

[0150] (7) Nematode channel ratio (NCR) is the ratio of the number of bacteria-feeding nematodes to the number of micro-feeding nematodes (bacteria-feeding nematodes + fungi-feeding nematodes), and its calculation formula is:

[0151] NCR=Ba / (Ba+Fu):

[0152] (8) The enrichment index (EI) can assess the response of a food web to available resources. Its calculation formula is:

[0153] EI = 100 × e / (e + b);

[0154] According to the cp value of free-living nematodes, b (basal) represents the basic components in the food web, usually referring to the two major groups Ba2 and Fu2; e (enchment) represents the enriched components in the food web, referring to the two groups Ba1 and Fu2. The calculation methods of b and e are ∑k b n b ,∑k e n e , where k b , k e is the weighted value corresponding to each group (its value is between 0.8 and 5.0), n b 、n e is the abundance of each group.

[0155] (9) Structure index (SI) can reflect the changes in soil food web structure during human disturbance or ecological restoration. Its calculation formula is:

[0156] SI = 100 × s / (s + b);

[0157] s(structure) represents the structural components in the food web, including Ba3-Ba5, Fu3-Fu5, Om3-Om5 and Ca3-Ca5 groups. The s value is calculated as ∑k s n s , where k s is the weighted value corresponding to each group, n s is the abundance of each group.

[0158] The present invention specifically uses Bayesian kernel machine regression (BKMR) to illustrate the dose-response relationship between soil potential toxic elements, nematode community structure and pollution characteristic index. BKMR reveals the different effects of a single potential toxic element on nematode populations and the potential synergistic or antagonistic interactions between potential toxic elements. The contribution of pollutants to ecological impacts was evaluated by five strategies of BKMR: (1) posterior inclusion probability (PIPs) values ​​were calculated to represent the relative importance of each potential toxic element relative to the soil nematode index; (2) mixed exposure effect plots were used to analyze the overall association between potential toxic elements and the soil nematode index; (3) univariate exposure-response curves were generated to illustrate the potential nonlinear association between potential toxic elements and the soil nematode index when all other potential toxic elements were fixed at the 50th percentile; (4) the association between a single potential toxic element and the soil nematode index was evaluated when other potential toxic elements were fixed at the 25th, 50th, or 75th percentile; (5) bivariate exposure-response curves of potential toxic elements and the soil nematode index were plotted to explore the potential interactions between potential toxic elements; in addition, the cross-validated kernel ensemble (CVEK) demonstrated the interaction effect value by utilizing cross-validation to determine the most appropriate kernel function. The above five strategies of BKMR were also used to analyze the dose-response relationship between the soil nematode index and the soil potential toxic element pollution index. The 'bkmr' package was used for rigorous 10,000 iterations of simulation to ensure the robustness of the statistical results. Hypothesis testing of potential toxic element interactions was performed by the CVEK package in R 4.1.0 software. The BKMR calculation formula is:

[0159] Y i =h(z i1 , ..., z im )+βx i +e i ;

[0160] Where: Y i is the soil nematode ecological index of the i-th sample; z i1 , ..., z im are the m potentially toxic elements in soil corresponding to the i-th sample; x i is a group of potential confounding factors; h (·) is the exposure-response function; β is the covariate x i The effect of i is the residual.

[0161] The present invention constructs an ecological risk assessment model through six calculation models, specifically including:

[0162] (1) Constructing a soil nematode ecological index to predict the ecological risk prediction model of potential toxic elements through LR:

[0163] Linear regression (LR) is a statistical method used to establish a linear relationship model between one or more independent variables and a dependent variable. The linear regression model finds the best linear relationship of the data by minimizing the sum of squares of the errors. In R software, the lm() function can be used to fit the linear regression model, using the nematode ecological index as the predictor variable and the ecological risk as the outcome variable to construct a linear regression model. The calculation formula is:

[0164]

[0165] n is the dimension of the sampling point, m is the number of the sampling point, θ0 is the intercept of the regression equation, θ j is the coefficient corresponding to the feature in the regression equation, X ij is the feature vector of the sampling point, y i is the actual value, is the predicted value and L is the value of the loss function.

[0166] (2) Using Ridge to construct a soil nematode ecological index to predict the ecological risk prediction model of potential toxic elements:

[0167] Ridge regression is based on linear regression and introduces L2 regularization terms to deal with collinearity problems and improve the generalization ability of the model. Ridge regression reduces the coefficients and prevents overfitting by adding a penalty term proportional to the square of the coefficient to the loss function. In R software, the ridge() function can be used to implement the ridge regression model, select a suitable regularization parameter lambda, use the nematode ecological index as the predictor variable, and model the ecological risk as the outcome variable. The calculation formula is:

[0168]

[0169] n is the dimension of the sampling point; m is the number of the sampling point; θ0 is the intercept of the regression equation; θ j is the coefficient corresponding to the feature in the regression equation; X ij is the feature vector of the sampling point; y i is the actual value; is the predicted value; α is the structural feature coefficient; L is the value of the loss function.

[0170] (3) Constructing a soil nematode ecological index to predict the ecological risk prediction model of potential toxic elements through RF:

[0171] Random forest (RF) is a machine learning model that uses multiple decision trees to train and predict samples. Random forest can handle classification and regression problems. If the target variable is a categorical variable, it can solve the classification problem; if it is a continuous variable, it can handle the regression problem. The basic principle of random forest regression is: repeatedly and with replacement, randomly select n samples from the original sample to form x regression trees (xtree), and the samples that are not selected each time constitute the out-of-bag data (OOB); then select the m variables (mtry) that can most effectively split the data from the variables, and determine the mtry value according to the principle of minimum out-of-bag error; finally, the multiple regression trees formed are combined into a random forest, and the average value is calculated and output. In R software, the randomForest package can be used to construct the RF model, using the nematode ecological index as the predictor variable and the ecological risk as the outcome variable to construct a random forest prediction model.

[0172] (4) Constructing a soil nematode ecological index to predict the ecological risk prediction model of potential toxic elements through GBRT:

[0173] Gradient boosted regression tree (GBRT) is an ensemble learning algorithm based on decision trees. It iteratively trains decision trees to minimize the gradient of the loss function. Each tree is modified based on the previous tree, thereby gradually improving the performance of the model. In R software, you can use the train() function in the caret package and set the method to "gbrt" to build a GBRT model, using the nematode ecological index as the predictor variable and the ecological risk as the outcome variable.

[0174] (5) Constructing a soil nematode ecological index to predict the ecological risk prediction model of potential toxic elements through DT:

[0175] Decision tree (DT) is an intuitive classification and regression method that makes decisions through the structure of a tree diagram. In a decision tree, each internal node represents a judgment on an attribute, each branch represents a judgment result, and each leaf node represents a decision result. In R software, the rpart package can be used to build a decision tree model by recursively selecting the optimal split point to build the model, using the nematode ecological index as the predictor variable and the ecological risk as the outcome variable.

[0176] (6) Constructing a soil nematode ecological index to predict the ecological risk prediction model of potential toxic elements through ERT:

[0177] Extra Trees is an ensemble learning method that makes predictions by building multiple extremely randomized decision trees. Unlike random forests, Extra Trees randomly selects split points when splitting nodes, rather than selecting the optimal split point. This method can increase the randomness of the model and reduce the risk of overfitting. In R software, the extraTrees package can be used to build an extreme random tree model, using the nematode ecological index as a predictor variable and ecological risk as an outcome variable.

[0178] The above data set is randomly divided into two parts, 70% of which is used for model training and 30% for model testing. The present invention uses the glmnet, rpart, gbm and randomForest functions in R 4.1.0 software, and the parameters in the six models are set by default.

[0179] The present invention uses mean absolute error (MAE), root mean square error (RMSE) and goodness of fit (R 2 ) Evaluate the accuracy of the prediction model, MAE, RMSE and R 2 The calculation formula is as follows:

[0180]

[0181] Where m is the number of sample sites, y i is the real value of the output label, is the predicted value of the output label, is the average value of the output label. The smaller the MAE and RMSE, the higher the prediction accuracy. 2 Used to evaluate the fitting accuracy of modeling, validation set R 2 Used to evaluate prediction accuracy and model generalization ability.

[0182] The embodiments of the present invention include the following verification experiments:

[0183] (1) Experimental verification of the dose-response relationship between soil potential toxic elements and nematode ecological index:

[0184] The soil nematode structure index, nematode channel index, and maturity index were linearly negatively correlated with exposure to composite potential toxic elements in the 25th to 75th percentile range. However, the diversity index, enrichment index, and evenness index of soil nematodes were linearly positively correlated with exposure to composite potential toxic elements. In addition, the total abundance of soil nematodes had a "J-shaped" curve association with exposure to composite potential toxic elements.

[0185] Univariate dose-response results with 95% confidence intervals (CI) showed:

[0186] When the Cu concentration was between 1.30 mg / kg [95% CI: (-51.61, -4.25)] and 1.50 mg / kg [95% CI: (-30.65, -5.62)], the total abundance of soil nematodes showed a negative dose-response relationship with Cu.

[0187] Ni concentrations ranging from 1.45 mg / kg [95% CI: (-37.63, -5.57)] to 1.73 mg / kg [95% CI: (-39.90, -0.10)] were negatively correlated with the total abundance of soil nematodes.

[0188] When the Pb concentration ranged from 1.33 mg / kg [95% CI: (-31.35, -7.00)] to 1.56 mg / kg [95% CI: (-39.04, -6.56)], the total abundance of soil nematodes was negatively correlated with Pb.

[0189] The Ni concentration ranged from 1.42 mg / kg [95% CI: (-8.17, -0.13)] to 1.67 mg / kg [95% CI: (-8.89, -0.58)], and the soil nematode enrichment index was negatively correlated with Ni.

[0190] When the Mn concentration ranged from 2.24 mg / kg [95% CI: (-55.47, -7.88)] to 2.49 mg / kg [95% CI: (-16.62, -0.08)], the soil nematode enrichment index was negatively correlated with Mn.

[0191] The Pb concentration ranged from 1.43 mg / kg [95% CI: (-0.06, -0.00)] to 1.85 mg / kg [95% CI: (-0.21, -0.06)], and Pb was negatively correlated with the soil nematode passage index.

[0192] When the As concentration ranged from 1.31 mg / kg [95% CI: (-11.83, -1.35)] to 1.26 mg / kg [95% CI: (-50.43, -21.34)], As was negatively correlated with the soil nematode structure index.

[0193] (2) Experimental verification of the dose-response relationship between soil nematode ecological index and potential toxic element pollution characteristic index:

[0194] The Nemerow index, pollution load index and mixed soil nematode index showed negative linear correlation, while the potential ecological risk index showed positive linear correlation.

[0195] Univariate dose-response results with 95% CIs showed:

[0196] The soil nematode structure index ranged from 0.09 [95% CI: (-1.05, -0.92)] to 2.16 [95% CI: (-0.13, -0.05)], showing a negative dose-response relationship with the Nemerow index.

[0197] The soil nematode enrichment index was negatively correlated with the Nemerow index when the enrichment index ranged from 0.09 [95% CI: (-0.56, -0.45)] to 2.16 [95% CI: (-0.06, -0.03)].

[0198] When the soil nematode maturity index was between 0.34 [95% CI: (-0.17, -0.14)] and 0.96 [95% CI: (-0.02, -0.01)], the maturity index was negatively correlated with the Nemerow index.

[0199] The soil nematode passage index was negatively correlated with the Nemerow index when the passage index ranged from 0.01 [95% CI: (-0.08, -0.07)] to 0.30 [95% CI: (-0.02, -0.01)].

[0200] The total abundance of soil nematodes ranged from 3.72 [95% CI: (-0.05, -0.04)] to 5.69 [95% CI: (-2.00, -1.97)], which was negatively correlated with the Nemerow index.

[0201] When the soil nematode diversity index was between 0.87 [95% CI: (-0.04, -0.02)] and 1.25 [95% CI: (-0.42, -0.40)], the diversity index was negatively correlated with the Nemerow index.

[0202] The soil nematode evenness index ranged from 0.50 [95% CI: (-0.03, -0.01)] to 0.69 [95% CI: (-0.21, -0.20)], which was negatively correlated with the Nemerow index.

[0203] (3) Result prediction based on the ecological risk model of potential toxic elements of soil nematode index:

[0204] The decision tree model has the highest MAE (NSPI = 3.05, PLI = 0498, RI = 177.217) and RMSE (NSPI = 16.583, PLI = 0.221, RI = 7645.183) values. Among the two linear regression-based algorithms, the ridge regression model has the highest MAE (NSPI = 3.05, PLI = 0498, RI = 177.217) and RMSE (NSPI = 16.583, PLI = 0.221, RI = 7645.183) values. 2 =0.106, MAE=3.009, RMSE=4.53) and RI(R 2=0.124, MAE = 181.734, RMSE = 272.34) had the best prediction effect. However, the random forest model showed the best performance on PLI (R 2 =0.144, MAE=0.427, RMSE=0.515). In the present invention, the extreme random tree model ranks third, second only to the ridge regression and random forest models. The specific prediction data are shown in Table 1 below.

[0205]

[0206] Table 1 Performance prediction data of ecological risk model of potential toxic elements based on soil nematode index

[0207] The present invention adopts Bayesian kernel machine regression to analyze the intrinsic dose-response relationship between soil potential toxic elements, nematode ecological index and pollution characteristic index under complex field environment, confirms that soil nematode structure index, nematode channel index and maturity index have negative linear dose-response relationship with soil potential toxic element concentration, and adopts six prediction models: linear regression (LR), ridge regression (Ridge), random forest (RF), gradient boosted regression tree (GBRT), decision tree (DT) and extreme random tree (ERT) to predict the ecological risk of potential toxic element pollution with soil nematode ecological index, and compares the generalization and precision of the models, and experimentally verifies that the Ridge model performs best in Nemerow index and potential ecological risk index, while the RF model performs well in pollution load index, thereby selecting the most suitable evaluation model.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the ecological risk of toxic element pollution based on soil nematode ecological index, characterized by: The evaluation steps include the following: Step 1: Sampling the soil in the region that is contaminated by potentially toxic elements to varying degrees is carried out at multiple time periods and multiple locations, and the sampled soil is used for testing the content of potentially toxic elements in the soil, testing the physical and chemical properties of the soil, and analyzing the structure of the soil nematode community; Step 2: Detect the content of potentially toxic elements in the soil; Step 3: Calculate the characteristic indexes of potential toxic element pollution in the soil, including: pollution factor, Nemerow index, pollution load index, and potential ecological risk index; Step 4: Analyze the soil nematode community structure and calculate the evaluation indicators of the soil nematode community structure, including: nematode abundance, proportion of different soil nematodes, Shannon-Wiener diversity index, soil nematode evenness index, nematode maturity index, nematode channel index, enrichment index, and structure index; Step 5: Use Bayesian kernel machine regression to analyze the dose-response relationship between the content of soil potential toxic elements, soil nematode community structure, and pollution characteristic index; Step 6: Based on Bayesian kernel machine regression analysis, the nematode ecological index and pollution characteristic evaluation index sensitive to potential toxic elements are obtained, and multiple computational models are used to construct ecological risk prediction models of potential toxic element pollution based on soil nematode ecological index. The computational models used include: linear regression model, ridge regression model, random forest model, gradient boosting regression tree model, decision tree model, and extreme random tree model; Step 7: Evaluate the constructed ecological risk prediction models and select the best fitting model.

2. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1 is characterized by: The potentially toxic elements sampled in the contaminated soil in step one specifically include Zn, Cu, Cr, Cd, Ni, Mn, Hg, Pb, As, Sb, and Se.

3. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1, characterized in that: The specific method for detecting the content of potentially toxic elements in the soil in step 2 is: Weigh 0.2 g of air-dried soil sample that has passed through a 100-mesh sieve into a polytetrafluoroethylene digestion tube TF-36, add 10 mL of hydrochloric acid, and insert it into the digestion hole; Set the digestion instrument temperature to 100°C, heat until a small amount of sample remains, then remove and cool; Add 5 mL of nitric acid, 5 mL of hydrofluoric acid, and 1 mL of perchloric acid and cover; Set the digestion instrument temperature to 150°C, digest for 120 min, and open the lid after cooling; Raise the temperature to 200℃ and wait until the thick white smoke disappears; Steam until the digestion solution becomes viscous and repeat the above digestion process; Remove the digestion tube and cool it, add 2mL of hydrochloric acid to dissolve the soluble residue, rinse the inner wall with deionized water, transfer the entire amount to a 50mL volumetric flask, and shake well; Inductively coupled plasma mass spectrometry was used for detection, in which Hg was not driven out of acid during digestion to prevent Hg volatilization, and atomic absorption spectroscopy was used to detect the content of potentially toxic elements in the soil.

4. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1, characterized in that: The specific method for calculating the potential toxic element pollution characteristic index in step 3 is: In order to assess the degree of potential toxic element contamination, the contamination factor CF is used for evaluation. The calculation formula of this factor is: Where: C i represents the concentration of potentially toxic elements at sampling point i, C o is the background value; The classification standard for defining the pollution factor CF is: When CF < 1, it means the soil is slightly polluted; When 1≤CF<3, it means the soil is moderately polluted; When 3≤CF<6, it means the soil is highly polluted; When CF>6, it means the soil is extremely highly polluted; To evaluate the comprehensive index of the pollution of composite potentially toxic elements, the Nemerow index NSPI is used for evaluation. The calculation formula of this index is as follows: Among them: CF imax It is the maximum value among all single pollution indexes CF of a potential toxic element in a certain area; CF iave It is the average value of all single pollution indexes CF of a certain potential toxic element in a certain area; The grading criteria for defining the Nemerow index NSPI are as follows: When NSPI ≤ 0.7, the pollution of potentially toxic elements in this area is within the safe range; When 0.7 < NSPI ≤ 1.0, the pollution of potentially toxic elements in this area is within the warning range; When 1.0 < NSPI ≤ 2.0, the pollution of potentially toxic elements in this area is at a slight pollution level; When 2.0 < NSPI ≤ 3.0, the pollution of potentially toxic elements in this area is at a moderate pollution level; When NSPI > 3.0, the pollution of potentially toxic elements in this area is at a severe pollution level; To evaluate the pollution load of composite potentially toxic elements in the study area according to the toxicity of each potentially toxic element, the pollution load index PLI is used for evaluation. The calculation formula of this index is as follows: Among them, CF im It is the maximum value of all single pollution indexes CF of a potential toxic element, PLI i is the pollution load index of a sampling point, n is the number of potentially toxic elements to be evaluated, and PLI zone is the pollution load index of a certain sampling area; The grading criteria for defining the pollution load index PLI are as follows: When PLI ≤ 1, there is no pollution; When 1 < PLI ≤ 2, it is at a slight pollution level; When 2 < PLI ≤ 3, it is at a moderate pollution level; When PLI > 3, it is at a high pollution level; To evaluate the synergistic effect among multiple potentially toxic elements and the potential degree of ecological harm, the potential ecological risk index RI is used for evaluation. The calculation formula of this index is as follows: E i =T i ×CF i ; Where: E i is the ecological hazard coefficient of a potential toxic element at a certain sampling point, T i is the toxicity response coefficient of a potential toxic element; The potential ecological risk index is graded, and the grading rule is as follows: When E i <40 or RI<150, defined as mild risk; When 40≤E i <80 or 150≤RI<300, defined as moderate risk; When 80≤E i <160 or 300≤RI<600, defined as higher risk; When 160≤E i <320 or RI ≥600, defined as high risk; When E i ≥320, defined as very high risk.

5. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1, characterized in that: The specific method for calculating the evaluation index of the soil nematode community structure in the fourth step is as follows: When calculating the nematode abundance, it is expressed by the total number of nematodes in every 100 g of dry soil; In every 100 g of dry soil for the relative abundances of each trophic group, calculate the percentages of plant-parasitic nematodes, bacterivorous nematodes, fungivorous nematodes, and predatory omnivorous nematodes in the total number of nematodes respectively; Calculate the Shannon-Wiener diversity index H, and the calculation formula is as follows: H=-∑P i (InP i ); Among them, P i is the proportion of individuals in the i-th taxonomic unit to the total number of nematodes; To illustrate the spatial characteristics of the distribution of nematodes in the soil, calculate the soil nematode evenness E, and the calculation formula is as follows: E = H / InS; Calculate the maturity index MI of free-living nematodes, and the calculation formula is as follows: MI=Σv i ×f i ; Among them, v i is the cp value assigned according to the different life strategies of free-living nematodes in ecological succession, f i is the ratio of the number of individuals of the i-th nematode to the total number of individuals in the community; Calculate the maturity index PPI of plant-parasitic nematodes, and the calculation formula is as follows: PPI=∑v i ×f i ; Among them, v i is the cp value of plant parasitic nematodes, f i is the ratio of the number of individuals of the i-th nematode to the total number of individuals in the community; Calculate the ratio of the number of bacterivorous nematodes to microbivorous nematodes. The calculation formula of the nematode channel ratio NCR is as follows: NCR = Bα / (Ba + Fu); To evaluate the response of the food web to available resources, calculate the enrichment index EI, and the calculation formula is as follows: BI = 100×e / (e + b); Among them, according to the c-p value of free-living nematodes, b represents the basic components in the food web, referring to the two major groups of Ba2 and Fu2; e represents the enriched components in the food web, referring to the two groups of Ba1 and Fu2; The calculation formulas for the b and e values are as follows: b=∑k b n b ; e=∑k e n e ; Where: k b , k e is the weighted value corresponding to each group, n b 、n e is the abundance of each group; To reflect the changes in the soil food web structure during human interference or ecological restoration processes, calculate the structure index SI, and the calculation formula is as follows: SI = 100×s / (s + b); In the formula: s represents the structural components in the food web, including the groups of Ba3 - Ba5, Fu3 - Fu5, Om3 - Om5, and Ca3 - Ca5; The calculation formula for the s value is as follows: s=∑k s n s ; Where: k s is the weighted value corresponding to each group, n s is the abundance of each group.

6. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1, characterized in that: The specific method of using Bayesian kernel machine regression in step 5 to analyze the dose-response relationship between the content of soil potential toxic elements, soil nematode community structure, and pollution characteristic index is: By analyzing the dose-response relationship among the content of potentially toxic elements in soil, the structure of soil nematode communities, and the pollution characteristic index, the nematode ecological index sensitive to potentially toxic elements and the pollution characteristic evaluation index were determined. The calculation formula is: Y i =h(z i1 ,...,z im )+βx i +e i ; Where: Y i is the soil nematode ecological index of the i-th sample; z i1 , ..., z im are the m potentially toxic elements in soil corresponding to the i-th sample; x i is a group of potential confounding factors; h (·) is the exposure-response function; β is the covariate x i The effect of i is the residual.

7. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1, characterized in that: The specific method for constructing the ecological risk prediction model of potential toxic element pollution based on soil nematode ecological index in step 6 is: The data set was randomly divided into two parts, 70% for model training and 30% for model testing. The parameters of each model were set to the default parameters. The linear regression model is used to build a prediction model. The expression of the linear regression model is: Among them, n is the dimension of the sampling point, m is the number of the sampling point, θ0 is the intercept of the regression equation, θ j is the coefficient corresponding to the feature in the regression equation, X ij is the feature vector of the sampling point, y i is the actual value, is the predicted value, L is the value of the loss function; The ridge regression model is used to build a prediction model. The expression of the ridge regression model is: Where: n is the dimension of the sampling point; m is the number of the sampling point; θ0 is the intercept of the regression equation; θ j is the coefficient corresponding to the feature in the regression equation; X ij is the feature vector of the sampling point; y i is the actual value; is the predicted value; α is the structural characteristic coefficient; L is the value of the loss function; A random forest model was used to construct a prediction model, with the nematode ecological index as the predictor variable and ecological risk as the outcome variable, implemented using the randomForest package in R software; A gradient boosting regression tree model was used to construct a prediction model, with the nematode ecological index as the predictor variable and ecological risk as the outcome variable, implemented using the caret package in R software; A decision tree model was used to construct a prediction model, with the nematode ecological index as the predictor variable and ecological risk as the outcome variable, which was implemented using the rpart package in R software; The extreme random tree model was used to construct the prediction model, with the nematode ecological index as the predictor variable and the ecological risk as the outcome variable, which was implemented using the extraTrees package in R software.

8. The method for ecological risk assessment of toxic element pollution based on soil nematode ecological index according to claim 1, characterized in that: The specific method for selecting the best fitting model in step seven is: The mean absolute error (MAE), root mean square error (RMSE), and goodness of fit (R) were used. 2 To evaluate the accuracy of each prediction model, the calculation formula is: Where m is the number of sample sites, y i is the real value of the output label, is the predicted value of the output label, is the average of the output labels; According to the calculation results, the smaller the MAE and RMSE, the higher the prediction accuracy. 2 Used to evaluate the fitting accuracy of modeling, validation set R 2 Used to evaluate prediction accuracy and model generalization ability.

Images