A method for assessing the impact of hydrological regimes on wetland bird food plants

Abstract

The present invention discloses a method for evaluating the impact of hydrological conditions on wetland bird food plants. The method first obtains bird food plant types, normalized vegetation index grid data, plant spatial distribution patterns, and hydrological data in a survey and research area, and constructs a hydrodynamic model based on the above data to obtain the spatial distribution pattern of accurate inundation time and average water depth; then, extracts food plant normalized vegetation index grid data from the data obtained in the survey and research area, and normalizes these data, then uses a random forest model for fitting regression, and combines the SHAP method to calculate the SHAP value of the sample point in the independent variable inundation time and average water depth, and then obtains the average absolute SHAP value of the sample point. By combining the hydrodynamic model and the SHAP method, the present invention can accurately obtain the two hydrological elements of the independent variable inundation time and average water depth, and quantify their impact on food plants.

Description

Technical Field

[0001] The invention relates to the field of eco-hydrology, and in particular to a method for evaluating the impact of hydrological conditions on wetland bird food plants. Background Art

[0002] As a key component of wetland ecosystems, wetland vegetation plays a variety of ecological functions, including reducing water erosion, protecting species diversity, and providing food and nighttime habitats for birds. Hydrological regimes are the main influencing factors of wetland vegetation patterns. Field monitoring and remote sensing methods have been widely used to obtain wetland hydrological elements. However, field monitoring has certain limitations. Generally speaking, it is only suitable for obtaining hydrological elements in a small area because field monitoring often requires a lot of time and manpower, and in some cases, monitoring sites may be difficult to reach. Although remote sensing technology can provide an effective way to obtain large-scale hydrological elements, it still faces many challenges in extracting complete hydrological elements. This is mainly due to the time and conditions of satellite image acquisition intervals, coupled with the influence of factors such as cloud cover. The hydrodynamic model is constructed based on the basic equations of water flow motion. It can accurately and continuously simulate the water flow process of the wetland, thereby meeting the needs of hydrological regime research at different time resolutions and different spatial scales, and has significant advantages in accurately obtaining the changing trends of key hydrological regime elements. However, it cannot directly quantify the impact of hydrological regimes on wetland plants.

[0003] The SHAP method based on random forest is a newly proposed interpretable machine learning method in recent years. It can be used to quantify the driving force (influence) of geographical environmental phenomena. It comes from the Shapley value in cooperative game theory. The Shapley value aims to fairly distribute the contributions of each participant in the cooperative game. In machine learning, the SHAP method can interpret the model's prediction as the contribution of each feature to the prediction. This method is good at dealing with complex nonlinear relationships within the system. Compared with traditional statistical methods such as principal component analysis, regression analysis, and correlation analysis, which can only quantify the global impact of environmental factors on plants, this method can comprehensively quantify the local and global impact of various environmental factors on plants, as well as the positive and negative impact. However, since the most important influencing factor of wetland plants, the hydrological situation factors, are difficult to obtain accurately, the application of the SHAP method in assessing the influence of wetland plants has been limited. Summary of the invention

[0004] In view of the shortcomings of the existing technology, the present invention provides a method for evaluating the impact of hydrological conditions on wetland bird food plants, so as to more scientifically and deeply evaluate the impact of hydrological conditions on wetland bird food plants, and provide more valuable understanding for the management and protection of wetland ecosystems.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for evaluating the impact of hydrological conditions on wetland bird food plants, comprising the following steps:

[0006] Step S1: Obtain bird food source plant types, normalized difference vegetation index raster data, plant spatial distribution pattern, hydrological data, topographic data and meteorological data in the study area;

[0007] Step S2: extracting the normalized vegetation index grid data of food source plants from the normalized vegetation index grid data according to the types of food source plants for birds and the spatial distribution pattern of plants;

[0008] Step S3: construct a hydrodynamic model based on the two-dimensional incompressible Navier-Stokes equations according to the hydrological data, topographic data and meteorological data, obtain the inundation time and average water depth of the study area in the hydrodynamic model, and extract them to obtain the spatial distribution pattern of accurate inundation time and average water depth;

[0009] Step S4: The grid data of the normalized vegetation index of food plants and the spatial distribution pattern of the precise flooding time and the average water depth are converted into point data, normalized, and input into the random forest model as independent variables and dependent variables respectively, and the regression equation is fitted by the random forest model to obtain the predicted value of the input sample independent variable;

[0010] Step S5: By comparing the predicted value of the input sample independent variable with the value of the point data of the normalized vegetation index of food source plants, the fitting effect of the random forest model is evaluated, and a point-line graph of the sampling point resolution and the determination coefficient is drawn. The relationship between the sampling point resolution and the determination coefficient is fitted using a Gaussian function, and the mean point of the Gaussian curve is calculated. The mean point is used as the sampling point resolution with the best fitting effect of the random forest model;

[0011] Step S6: Based on the sampling point resolution with the best fitting effect of the random forest model, the independent variables and dependent variables in step S4 are normalized, and a random forest regression is constructed. The random forest regression is calculated using the SHAP method to obtain the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth.

[0012] Furthermore, in step S3, the spatial distribution pattern of the precise submergence time and the average water depth is obtained. The specific process is as follows:

[0013] Step S31, gridding operation is performed on the study area including hydrological data, topographic data and meteorological data, interpolation of the minimum grid resolution is completed according to the topographic data, upper and lower boundaries of the study area are set respectively, roughness parameters of the hydrodynamic model are clarified, and based on the two-dimensional incompressible Navier-Stokes equations, the measured inflow and outflow upper and lower boundaries and meteorological data of the measured study area are used to construct the hydrodynamic model, and the flooding time and average water depth of the study area in the hydrodynamic model are obtained;

[0014] Step S32, extracting the inundation time and average water depth of the grid with the same location as the station in the hydrodynamic model and the study area as the simulated value, and comparing the simulated value with the measured inundation time and average water depth of the corresponding station in the study area as the measured value, and using the Nash efficiency coefficient EF to evaluate the simulation accuracy based on the comparison result, and finally, by extracting the inundation time and average water depth of all grids in the hydrodynamic model after simulation accuracy, the spatial distribution pattern of accurate inundation time and average water depth is obtained;

[0015] The spatial distribution pattern of accurate inundation time and average water depth is obtained, which is expressed as:

[0016] (1);

[0017] (2);

[0018] In the formula, represents the spatial distribution pattern of precise flooding time; WD represents the water depth in the grid; n represents a natural year; D WD Indicates the cumulative time that the water depth in the grid is greater than 0; Represents the spatial distribution pattern of the precise average water depth.

[0019] Further, in step S4, the predicted value of the sample independent variable is input, and the specific steps are:

[0020] Step S41, converting the spatial distribution pattern of precise flooding time and average water depth into raster data;

[0021] Step S42, using the Extract Raster Data to Points tool, converting the raster data of the normalized vegetation index of food plants and the raster data of the spatial distribution pattern of the precise flooding time and the average water depth into point data;

[0022] The point data is normalized and expressed as:

[0023] (3);

[0024] Where F represents the normalized value; x represents the value of the grid data of the food source plant normalized vegetation index and the precise flooding time and average water depth after point conversion; x maxand x min Respectively represent the minimum and maximum values ​​of the normalized vegetation index of food source plants and the precise flooding time and average water depth after the grid data are converted;

[0025] Step S43, during step S42, firstly, the Extract Raster Data to Points tool is set with different resolutions so that the resolution of the point data is the same, and then the point data with the spatial distribution pattern of the precise flooding time and the average water depth with the same resolution after normalization is used as an independent variable, which is recorded as the independent variable S=(S1, S2), where S1 represents the independent variable precise flooding time, and S2 represents the independent variable precise average water depth;

[0026] Step S44, taking the point data of the normalized difference vegetation index of food plants with the same resolution as the point data of the spatial distribution pattern of the precise flooding time and the average water depth as the dependent variable, recorded as the dependent variable Y;

[0027] Step S45, inputting the independent variable S and the dependent variable Y into the random forest model for training, the random forest model will fit the regression equation by learning the relationship between the independent variable S and the dependent variable Y;

[0028] Step S46, using a grid search method to determine the optimal parameters of the fitted regression equation in the random forest model; the parameters of the random forest model include the number of trees, the depth of the tree, and the minimum sample for splitting a node;

[0029] Step S47, after training, use the random forest model to predict the value of the dependent variable Y for a given independent variable S; specifically, for a new input S new , the random forest model based on the optimal parameters will integrate the prediction results of multiple decision trees to obtain the predicted value of the input sample independent variable, expressed as;

[0030] (4);

[0031] In the formula, Represents the predicted value of the input sample independent variable; represents the number of decision trees; represents the prediction result of the t-th decision tree for input S; S new An independent variable representing a new input of point data of precise inundation time and spatial distribution pattern of mean water depth.

[0032] Further, in step S5, the mean point of the Gaussian curve is calculated, and the specific process is:

[0033] Step S51, by comparing the predicted value of the input sample independent variable with the value of the point data of the normalized vegetation index of food source plants, the fitting effect of the random forest model is evaluated, and the evaluation index is the determination coefficient R 2, and plot the sampling point resolution and determination coefficient R 2 Point and line graph;

[0034] Step S52: Use Gaussian function to fit the sampling point resolution and determination coefficient R 2 The relationship between the two is used to calculate the mean point of the Gaussian curve, and the mean point is used as the sampling point resolution for the best fitting effect of the random forest model.

[0035] Furthermore, the coefficient of determination R 2 The calculation formula is expressed as:

[0036] R 2 = [ ∑ i = 1 m ( p i − p v ) ( y i − y v ) ] 2 ∑ i = 1 m ( p i − p v ) 2 ∑ i = 1 m ( y i − y v ) 2 (5);

[0037] In the formula, represents the coefficient of determination; m represents the number of samples; and represent the actual value and the predicted value respectively; and Expressed as average actual value and predicted value respectively.

[0038] Further, the Gaussian function fitting formula is expressed as:

[0039] h = c exp [ − 1 2 ( u − w ) 2 v 2 ] (6);

[0040] In the formula, h represents the dependent variable of the Gaussian function, that is, the output value of the function; Represents the amplitude, which determines the peak height of the Gaussian function; Represents a number function; represents the mean; represents standard deviation; Represents the independent variable of the Gaussian function.

[0041] Further, in step S6, the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained. The specific process is as follows:

[0042] Step S61, based on the sampling point resolution with the best fitting effect of the random forest model, normalize the independent variable and the dependent variable, and construct a random forest regression using a machine learning library;

[0043] Step S62, inputting the constructed random forest regression into a Shapley value interpretation library, the Shapley value interpretation library including the SHAP method, to obtain the SHAP values ​​of the sample points in the independent variable precise submergence time and the independent variable precise average water depth;

[0044] Step S63, by calculating the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth, the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained.

[0045] Furthermore, the SHAP values ​​of the sample points in the independent variable precise submergence time and the independent variable precise average water depth are obtained. The specific process is as follows:

[0046] The SHAP method is used to calculate the SHAP values ​​of sample points in the independent variable precise submergence time and independent variable precise average water depth in random forest regression;

[0047] The SHAP method explains the output of the random forest model by calculating the contribution of the independent variable precise inundation time and the independent variable precise mean water depth to the prediction of the random forest model;

[0048] Specifically, for a given sample point, assume that the prediction function of the random forest model is f(s), where S=(S1,S2,...,S k ) represents the independent variable; S k represents the value of the kth independent variable; the calculation formula of the SHAP value of the sample point in the independent variable precise submergence time and the independent variable precise average water depth is expressed as:

[0049] ϕ i = ∑ D ⊆ { 1 , 2 ,..., k \ { i } } | D | !( k − | D | − 1 )! k ! [ f ( s D ∪ { s i } ) − f ( s D ) ] (7);

[0050] In the formula, represents the SHAP value of the sample point in the independent variable precise submergence time and the independent variable precise average water depth; D represents the union of the independent variables, that is, a subset of the independent variables; k represents the total number of independent variables; represents factorial; represents the factorial of k; Represents the prediction function of the random forest model; Represents the value corresponding to the independent variable in subset D; Represents the specific value of the independent variable currently considered; among them, Include and , Represents the SHAP value of the sample point in the precise flooding time of the independent variable; Represents the SHAP value of the sample point in the precise average water depth of the independent variable.

[0051] Furthermore, the calculation formula for the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained, which is expressed as:

[0052] (8);

[0053] In the formula, represents the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth; j represents the number of sample points; Represents the SHAP value of the i-th sample point.

[0054] Compared with the existing technology, the present invention has the following beneficial effects: by combining the hydrodynamic model and the SHAP method, the present invention can accurately obtain the two elements of the independent variable flooding time and the independent variable average water depth and quantify their impact on the plants in the study area, filling the application gap of the SHAP method in this field and providing an effective way to explore the impact of hydrological conditions on plants in the study area. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the present invention.

[0056] Figure 2 It is a schematic diagram for evaluating the simulation results of the hydrodynamic model of the present invention.

[0057] Figure 3 It is a fitting relationship curve diagram of the sampling point resolution and the determination coefficient of the present invention.

[0058] Figure 4 This is a global impact diagram of the flooding time and average water depth on food plants in the present invention. DETAILED DESCRIPTION

[0059] The present invention is further described in detail below with reference to implementation cases and accompanying drawings.

[0060] like Figure 1 As shown in Figure 1, a method for assessing the impact of hydrological conditions on wetland bird food plants includes the following steps:

[0061] Step S1: Obtain bird food source plant types, normalized difference vegetation index raster data, plant spatial distribution pattern, hydrological data, topographic data and meteorological data in the study area;

[0062] Step S2: extracting the normalized vegetation index grid data of food source plants from the normalized vegetation index grid data according to the types of food source plants for birds and the spatial distribution pattern of plants;

[0063] Step S3: construct a hydrodynamic model based on the two-dimensional incompressible Navier-Stokes equations according to the hydrological data, topographic data and meteorological data, obtain the inundation time and average water depth of the study area in the hydrodynamic model, and extract them to obtain the spatial distribution pattern of accurate inundation time and average water depth;

[0064] Step S4: The grid data of the normalized vegetation index of food plants and the spatial distribution pattern of the precise flooding time and the average water depth are converted into point data, normalized, and input into the random forest model as independent variables and dependent variables respectively, and the regression equation is fitted by the random forest model to obtain the predicted value of the input sample independent variable;

[0065] Step S5: By comparing the predicted value of the input sample independent variable with the value of the point data of the normalized vegetation index of food source plants, the fitting effect of the random forest model is evaluated, and a point-line graph of the sampling point resolution and the determination coefficient is drawn. The relationship between the sampling point resolution and the determination coefficient is fitted using a Gaussian function, and the mean point of the Gaussian curve is calculated. The mean point is used as the sampling point resolution with the best fitting effect of the random forest model;

[0066] Step S6: Based on the sampling point resolution with the best fitting effect of the random forest model, the independent variables and dependent variables in step S4 are normalized, and a random forest regression is constructed. The random forest regression is calculated using the SHAP method to obtain the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth.

[0067] Among them, in step S1, the specific process of investigating the research area is as follows:

[0068] Through investigation in the research area, the types of bird food plants were obtained and identified as carex. Carex belongs to the genus Cyperus of the Cyperaceae family. It is a perennial herb that is widely distributed in the Tongjiang wetlands in the middle and lower reaches of the Yangtze River. It is also the main food source for overwintering migratory birds.

[0069] Among them, the spatial distribution pattern of accurate submergence time and average water depth is obtained in step S3, and the specific process is as follows:

[0070] Step S31, gridding operation is performed on the study area including hydrological data, topographic data and meteorological data, and the minimum grid resolution is interpolated according to the topographic data. The minimum grid resolution is set to 100m, and the upper and lower boundaries of the study area are set respectively. The roughness parameter of the hydrodynamic model is clarified. Based on the two-dimensional incompressible Navier-Stokes equations, the hydrodynamic model is constructed using the measured inflow and outflow boundaries and meteorological data (upper and lower boundaries, located outside the area) of the measured study area from 2009 to 2014, and the flooding time and average water depth of the study area in the hydrodynamic model from 2009 to 2014 are obtained;

[0071] Step S32, extract the inundation time and average water depth of the grid with the same location as the station in the hydrodynamic model in the study area as the simulated value, and compare the simulated value with the measured inundation time and average water depth of the corresponding station in the study area as the measured value, and use the Nash efficiency coefficient EF to evaluate the simulation accuracy based on the comparison result. Finally, by extracting the inundation time and average water depth of all grids in the hydrodynamic model after simulation accuracy, the spatial distribution pattern of accurate inundation time and average water depth is obtained, such as Figure 2 As shown;

[0072] The spatial distribution pattern of accurate inundation time and average water depth is obtained, which is expressed as:

[0073] (1);

[0074] (2);

[0075] In the formula, represents the spatial distribution pattern of precise flooding time; WD represents the water depth in the grid; n represents a natural year; D WD Indicates the cumulative time that the water depth in the grid is greater than 0; Represents the spatial distribution pattern of the precise average water depth.

[0076] Among them, the predicted value of the input sample independent variable is obtained in step S4, and the specific steps are:

[0077] Step S41, converting the spatial distribution pattern of the precise flooding time and the average water depth into raster data, with a resolution consistent with the minimum grid divided in S31;

[0078] Step S42, using the Extract Raster Data to Point tool in ArcMap software, converting the 2010 food source plant normalized difference vegetation index (resolution of 1000m) and the raster data of the spatial distribution pattern of the precise flooding time and average water depth (resolution of 100m) into point data;

[0079] The point data is normalized and expressed as:

[0080] (3);

[0081] Where F represents the normalized value; x represents the value of the grid data of the food source plant normalized vegetation index and the precise flooding time and average water depth after point conversion; x max and x min Respectively represent the minimum and maximum values ​​of the normalized vegetation index of food source plants and the precise flooding time and average water depth after the grid data are converted;

[0082] Step S43, during step S42, firstly, the Extract Raster Data to Points tool needs to be set with different resolutions (the setting range is 100 to 1000m, and the range interval is 100m) so that the resolution of the point data is the same, and then the point data with the spatial distribution pattern of the precise flooding time and average water depth with the same resolution after normalization is used as the independent variable, which is recorded as the independent variable S=(S1,S2), where S1 represents the independent variable precise flooding time, and S2 represents the independent variable precise average water depth;

[0083] Step S44, taking the point data of the normalized difference vegetation index of food plants with the same resolution as the point data of the spatial distribution pattern of the precise flooding time and the average water depth as the dependent variable, recorded as the dependent variable Y;

[0084] Step S45, inputting the independent variable S and the dependent variable Y into the random forest model for training, the random forest model will fit the regression equation by learning the relationship between the independent variable S and the dependent variable Y;

[0085] Step S46, using the grid search method of the scikit-learn library in the Spyder software to determine the optimal parameters for fitting the regression equation in the random forest model; the parameters of the random forest model include the number of trees, the depth of the tree, and the minimum sample for splitting a node;

[0086] Step S47, after training is completed, the optimal parameters of the random forest model can be used to predict the value of the dependent variable Y for a given independent variable S; specifically, for a new input S new , the random forest model based on the optimal parameters will integrate the prediction results of multiple decision trees to obtain the predicted value of the input sample independent variable, expressed as;

[0087] (4);

[0088] In the formula, Represents the predicted value of the input sample independent variable; represents the number of decision trees; represents the prediction result of the t-th decision tree for input S; S new An independent variable representing a new input of point data of precise inundation time and spatial distribution pattern of mean water depth.

[0089] The specific process of calculating the mean point of the Gaussian curve in step S5 is as follows:

[0090] Step S51, by comparing the predicted value of the input sample independent variable with the normalized value of the point data of the food source plant normalized vegetation index, the fitting effect of the random forest model is evaluated. The evaluation index is the determination coefficient R², and the sampling point resolution (independent variable) and R 2 (dependent variable) dot-line plot;

[0091] Step S52: Use Gaussian function to fit the sampling point resolution and determination coefficient R 2 The relationship between the two is to calculate the mean point of the Gaussian curve (fitting line), and the mean point is used as the sampling point resolution for the best fitting effect of the random forest model, such as Figure 3 As shown, the mean point is 455m;

[0092] Among them, the calculation formula of the determination coefficient is expressed as:

[0093] R 2 = [ ∑ i = 1 m ( p i − p v ) ( y i − y v ) ] 2 ∑ i = 1 m ( p i − p v ) 2 ∑ i = 1 m ( y i − y v ) 2 (5);

[0094] In the formula, represents the coefficient of determination; m represents the number of samples; and represent the actual value and the predicted value respectively; and They are expressed as average actual and predicted values, respectively;

[0095] Among them, the Gaussian function fitting formula is expressed as:

[0096] h = c exp [ − 1 2 ( u − w ) 2 v 2 ] (6);

[0097] In the formula, h represents the dependent variable of the Gaussian function, that is, the output value of the function; represents the amplitude, which determines the peak height of the Gaussian function; It represents a number function, which represents an exponential operation with a natural constant as the base; represents the mean; represents standard deviation; Represents the independent variable of the Gaussian function.

[0098] Among them, the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained in step S6, and the specific process is as follows:

[0099] Step S61, based on the obtained sampling point resolution with the best fitting effect of the random forest model, normalize the independent variable and the dependent variable, and use the machine learning library (scikit-learn library of Spyder software) to construct a random forest regression;

[0100] Step S62, input the constructed random forest regression into the Shapley value interpretation library (shap library), which includes the SHAP method, to obtain the SHAP values ​​of the sample points in the independent variable precise flooding time and the independent variable precise average water depth to quantify the local impact, such as Figure 4 As shown;

[0101] Step S63, by calculating the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth, the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained to quantify the global importance of the independent variable, such as Figure 4 As shown;

[0102] Among them, the SHAP values ​​of the sample points in the independent variable precise submergence time and the independent variable precise average water depth are obtained by the following specific process:

[0103] The SHAP method is used to calculate the SHAP values ​​of sample points in the independent variable precise submergence time and independent variable precise average water depth in random forest regression;

[0104] The SHAP method explains the output of the random forest model by calculating the contribution of the independent variable precise inundation time and the independent variable precise mean water depth to the prediction of the random forest model;

[0105] Specifically, for a given sample point, assume that the prediction function of the random forest model is f(s), where S=(S1,S2,...,S k ) represents the independent variable; S k represents the value of the kth independent variable; the calculation formula of the SHAP value of the sample point in the independent variable precise submergence time and the independent variable precise average water depth is expressed as:

[0106] ϕ i = ∑ D ⊆ { 1 , 2 ,..., k \ { i } } | D | !( k − | D | − 1 )! k ! [ f ( s D ∪ { s i } ) − f ( s D ) ] (7);

[0107] In the formula, represents the SHAP value of the sample point in the independent variable precise submergence time and the independent variable precise average water depth; D represents the union of the independent variables, that is, a subset of the independent variables; k represents the total number of independent variables; represents factorial; represents the factorial of k; Represents the prediction function of the random forest model; Represents the value corresponding to the independent variable in subset D; Represents the specific value of the independent variable currently considered; among them, Include and , Represents the SHAP value of the sample point in the precise flooding time of the independent variable; Represents the SHAP value of the sample point in the precise average water depth of the independent variable.

[0108] Among them, the calculation formula for obtaining the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is:

[0109] (8);

[0110] In the formula, represents the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth; j represents the number of sample points; Represents the SHAP value of the i-th sample point.

[0111] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the impact of hydrological conditions on wetland bird food plants, characterized in that: The following steps are included: Step S1: Obtain bird food source plant types, normalized difference vegetation index raster data, plant spatial distribution pattern, hydrological data, topographic data and meteorological data in the study area; Step S2: extracting the normalized vegetation index grid data of food source plants from the normalized vegetation index grid data according to the types of food source plants for birds and the spatial distribution pattern of plants; Step S3: construct a hydrodynamic model based on the two-dimensional incompressible Navier-Stokes equations according to the hydrological data, topographic data and meteorological data, obtain the inundation time and average water depth of the study area in the hydrodynamic model, and extract them to obtain the spatial distribution pattern of accurate inundation time and average water depth; Step S4: The grid data of the normalized vegetation index of food plants and the spatial distribution pattern of the precise flooding time and the average water depth are converted into point data, normalized, and input into the random forest model as independent variables and dependent variables respectively, and the regression equation is fitted by the random forest model to obtain the predicted value of the input sample independent variable; Step S5: By comparing the predicted value of the input sample independent variable with the value of the point data of the normalized vegetation index of food source plants, the fitting effect of the random forest model is evaluated, and a point-line graph of the sampling point resolution and the determination coefficient is drawn. The relationship between the sampling point resolution and the determination coefficient is fitted using a Gaussian function, and the mean point of the Gaussian curve is calculated. The mean point is used as the sampling point resolution with the best fitting effect of the random forest model; Step S6: Based on the sampling point resolution with the best fitting effect of the random forest model, the independent variables and dependent variables in step S4 are normalized, and a random forest regression is constructed. The random forest regression is calculated using the SHAP method to obtain the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth.

2. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 1, characterized in that: In step S3, the spatial distribution pattern of accurate submergence time and average water depth is obtained. The specific process is as follows: Step S31, gridding operation is performed on the study area including hydrological data, topographic data and meteorological data, interpolation of the minimum grid resolution is completed according to the topographic data, upper and lower boundaries of the study area are set respectively, roughness parameters of the hydrodynamic model are clarified, and based on the two-dimensional incompressible Navier-Stokes equations, the measured inflow and outflow upper and lower boundaries and meteorological data of the measured study area are used to construct the hydrodynamic model, and the flooding time and average water depth of the study area in the hydrodynamic model are obtained; Step S32, extracting the inundation time and average water depth of the grid with the same location as the station in the hydrodynamic model and the study area as the simulated value, and comparing the simulated value with the measured inundation time and average water depth of the corresponding station in the study area as the measured value, and using the Nash efficiency coefficient EF to evaluate the simulation accuracy based on the comparison result, and finally, by extracting the inundation time and average water depth of all grids in the hydrodynamic model after simulation accuracy, the spatial distribution pattern of accurate inundation time and average water depth is obtained; The spatial distribution pattern of accurate inundation time and average water depth is obtained, which is expressed as: （1）； （2）； In the formula, represents the spatial distribution pattern of precise flooding time; WD represents the water depth in the grid; n represents a natural year; D WD Indicates the cumulative time that the water depth in the grid is greater than 0; Represents the spatial distribution pattern of the precise average water depth.

3. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 2, characterized in that: In step S4, the predicted value of the sample independent variable is input, and the specific steps are: Step S41, converting the spatial distribution pattern of precise flooding time and average water depth into raster data; Step S42, using the Extract Raster Data to Points tool, converting the raster data of the normalized vegetation index of food plants and the raster data of the spatial distribution pattern of the precise flooding time and the average water depth into point data; The point data is normalized and expressed as: （3）； Where F represents the normalized value; x represents the value of the grid data of the food source plant normalized vegetation index and the precise flooding time and average water depth after point conversion; x max and x min Respectively represent the minimum and maximum values ​​of the normalized vegetation index of food source plants and the precise flooding time and average water depth after the grid data are converted; Step S43, during step S42, firstly, the Extract Raster Data to Points tool is set with different resolutions so that the resolution of the point data is the same, and then the point data with the spatial distribution pattern of the precise flooding time and the average water depth with the same resolution after normalization is used as an independent variable, which is recorded as the independent variable S=(S1, S2), where S1 represents the independent variable precise flooding time, and S2 represents the independent variable precise average water depth; Step S44, taking the point data of the normalized difference vegetation index of food plants with the same resolution as the point data of the spatial distribution pattern of the precise flooding time and the average water depth as the dependent variable, recorded as the dependent variable Y; Step S45, inputting the independent variable S and the dependent variable Y into the random forest model for training, the random forest model will fit the regression equation by learning the relationship between the independent variable S and the dependent variable Y; Step S46, using a grid search method to determine the optimal parameters of the fitted regression equation in the random forest model; the parameters of the random forest model include the number of trees, the depth of the tree, and the minimum sample for splitting a node; Step S47, after training, use the random forest model to predict the value of the dependent variable Y for a given independent variable S; specifically, for a new input S new , the random forest model based on the optimal parameters will integrate the prediction results of multiple decision trees to obtain the predicted value of the input sample independent variable, expressed as; （4）； In the formula, Represents the predicted value of the input sample independent variable; represents the number of decision trees; represents the prediction result of the t-th decision tree for input S; S new An independent variable representing a new input of point data of precise inundation time and spatial distribution pattern of mean water depth.

4. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 3, characterized in that: In step S5, the mean point of the Gaussian curve is calculated, and the specific process is as follows: Step S51, by comparing the predicted value of the input sample independent variable with the value of the point data of the normalized vegetation index of food source plants, the fitting effect of the random forest model is evaluated, and the evaluation index is the determination coefficient R 2 , and plot the sampling point resolution and determination coefficient R 2 Point and line graph; Step S52: Use Gaussian function to fit the sampling point resolution and determination coefficient R 2 The relationship between the two is used to calculate the mean point of the Gaussian curve, and the mean point is used as the sampling point resolution for the best fitting effect of the random forest model.

5. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 4, characterized in that: Coefficient of determination R 2 The calculation formula is expressed as: （5）； In the formula, represents the coefficient of determination; m represents the number of samples; and represent the actual value and the predicted value respectively; and Expressed as average actual value and predicted value respectively.

6. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 5, characterized in that: The Gaussian function fitting formula is expressed as: （6）； In the formula, h represents the dependent variable of the Gaussian function, that is, the output value of the function; Represents the amplitude, which determines the peak height of the Gaussian function; Represents a number function; represents the mean; represents standard deviation; Represents the independent variable of the Gaussian function.

7. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 6, characterized in that: In step S6, the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained. The specific process is: Step S61, based on the sampling point resolution with the best fitting effect of the random forest model, normalize the independent variable and the dependent variable, and construct a random forest regression using a machine learning library; Step S62, inputting the constructed random forest regression into a Shapley value interpretation library, the Shapley value interpretation library including the SHAP method, to obtain the SHAP values ​​of the sample points in the independent variable precise submergence time and the independent variable precise average water depth; Step S63, by calculating the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth, the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is obtained.

8. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 7, characterized in that: The SHAP values ​​of the sample points in the independent variable precise submergence time and the independent variable precise average water depth are obtained by: The SHAP method is used to calculate the SHAP values ​​of sample points in the independent variable precise submergence time and independent variable precise average water depth in random forest regression; The SHAP method explains the output of the random forest model by calculating the contribution of the independent variable precise inundation time and the independent variable precise mean water depth to the prediction of the random forest model; Specifically, for a given sample point, assume that the prediction function of the random forest model is f(s), where S=(S1,S2,...,S k ) represents the independent variable; S k represents the value of the kth independent variable; the calculation formula of the SHAP value of the sample point in the independent variable precise submergence time and the independent variable precise average water depth is expressed as: （7）； In the formula, represents the SHAP value of the sample point in the independent variable precise submergence time and the independent variable precise average water depth; D represents the union of the independent variables, that is, a subset of the independent variables; k represents the total number of independent variables; represents factorial; represents the factorial of k; Represents the prediction function of the random forest model; Represents the value corresponding to the independent variable in subset D; Represents the specific value of the independent variable currently considered; among them, Include and , Represents the SHAP value of the sample point in the precise flooding time of the independent variable; Represents the SHAP value of the sample point in the precise average water depth of the independent variable.

9. The method for evaluating the impact of hydrological conditions on wetland bird food plants according to claim 8, characterized in that: The calculation formula for obtaining the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth is expressed as: （8）； In the formula, represents the average absolute SHAP value of the sample points in the independent variable precise submergence time and the independent variable precise average water depth; j represents the number of sample points; Represents the SHAP value of the i-th sample point.

Images