A method and device for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth
The method and device address the challenge of evaluating nitrogen deposition's impact on vegetation growth by preprocessing data, iteratively processing bandwidth intervals, and constructing geographically weighted regression models, enhancing spatial scale appropriateness and accuracy of assessments.
Patent Information
- Application Number
- CN202410159000.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-04
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-02-04
AI Technical Summary
Existing methods struggle to evaluate the impact of atmospheric nitrogen deposition on vegetation growth at an appropriate spatial scale and under the influence of multiple factors, requiring significant manpower and resources while providing limited scalability and accuracy.
A method and device utilizing geographic feature data preprocessing, iterative processing of bandwidth intervals, and geographically weighted regression modeling to assess the regional impact of nitrogen deposition on vegetation growth, incorporating remote sensing, vegetation type, meteorological, terrain, and socio-economic data to determine optimal spatial scales and evaluate the influence of multiple factors.
Enables accurate and efficient evaluation of nitrogen deposition's impact on vegetation growth at appropriate spatial scales, improving the precision and confidence of assessment results.
Smart Images

Figure CN118313541B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of environmental change and its ecological effects, and in particular to a method and device for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth. Background Art
[0002] Since the Industrial Revolution, with the use of chemical fuels and chemical nitrogen fertilizers and the rapid development of animal husbandry, the amount of reactive nitrogen compounds emitted into the atmosphere has increased sharply. Studies have shown that the production of reactive nitrogen caused by global human activities has increased from 165 TgN / a in 2000 to 192 TgN / a in 2008. Its excessive increase will destroy the productivity and stability of the ecosystem, leading to serious ecological problems such as soil acidification, water eutrophication, and reduced biodiversity. Nitrogen in the atmosphere will continue to fall to land and water in the form of NHx (ammonium nitrogen) and NOx (nitrate nitrogen). This process is called nitrogen deposition. Nitrogen deposition is an important part of the global nitrogen biochemical cycle. It affects vegetation growth by changing the absorption and utilization of nutrients by vegetation during metabolism and the strength of vegetation photosynthesis. The effect and degree of its influence depend on the nitrogen saturation of the ecosystem in which the plants are located. Studies have found that in a "nitrogen-limited" ecosystem, an appropriate increase in nitrogen deposition can promote vegetation growth; while in an ecosystem with sufficient nitrogen, an increase in nitrogen deposition will have a negative impact on vegetation growth. In recent years, with the acceleration of industrialization and urbanization, nitrogen deposition in various regions has also increased. And with the development of economy and society, nitrogen deposition in various regions may increase further. Therefore, studying the feedback mechanism of past vegetation growth on nitrogen deposition and other meteorological factors and evaluating the effect of nitrogen deposition on vegetation growth are of great significance to maintaining the sustainable development of regional ecosystems.
[0003] In the prior art, there are two main methods for studying the effect of nitrogen deposition on vegetation growth. The first is a method based on field or forest experiments, which can simulate different levels of nitrogen deposition experiments, observe the effect of nitrogen deposition on the growth of specific vegetation types under different scenarios, and can more accurately quantify the effect of nitrogen deposition on vegetation growth. However, this method requires a lot of manpower, material resources and time. In addition, since most of the experimental devices can only conduct single-factor experiments, and there are limitations in the scale and geographical location of the experiment, the evaluation results obtained are only applicable to single-factor evaluation of the area to be evaluated at a very small spatial scale, and cannot be universally promoted. The second is a method based on a system or device, which collects atmospheric nitrogen deposition information, other influencing factors and growth status information of specific vegetation in a small area by using a system or device, observes and summarizes the effect of nitrogen deposition on the growth status of specific vegetation types, and can achieve long-term, high-precision and accurate observations. However, when it is extended to a large scale, it is necessary to consider multiple variables, and it is necessary to modify the device based on new influencing factors such as geographical location and terrain, and the maintenance of the equipment requires a certain amount of manpower and material resources. Therefore, how to evaluate the impact of nitrogen deposition on vegetation growth at an appropriate spatial scale and under the influence of multiple factors remains a difficult problem in current research. Summary of the invention
[0004] The present invention provides a method and device for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth, so as to solve the technical problem in the prior art that it is impossible to evaluate the impact of nitrogen deposition on vegetation growth conditions at an appropriate spatial scale and under the influence of multiple factors.
[0005] In a first aspect, the present application provides a method for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth, comprising:
[0006] Acquire first geographic feature data of the area to be evaluated, and iterate the first geographic feature data multiple times according to a preset first iteration method to obtain and output second geographic feature data; wherein the first geographic feature data is obtained by preprocessing the original geographic feature data; the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, terrain data and socio-economic data;
[0007] Iteratively process the preset first bandwidth interval based on the second geographic feature data, update the boundary value of the first bandwidth interval according to a preset first calculation method during iteration, until the first bandwidth interval meets a preset condition, and output the first bandwidth corresponding to the first bandwidth interval of the last iteration;
[0008] Based on the first bandwidth and the second geographic feature data, a geographically weighted regression model is constructed and solved, and according to the solution result, the impact of atmospheric nitrogen deposition on vegetation growth in the assessment area is evaluated.
[0009] In this way, the original geographic feature data containing multiple elements are first obtained and preprocessed, and the obtained first geographic feature data are iteratively processed to obtain the second geographic feature data, which can provide geographic feature data of multiple elements for subsequent evaluation. Then, based on the second geographic feature data, the first bandwidth interval is iterated until the first bandwidth interval meets the preset conditions, and the first bandwidth corresponding to the first bandwidth interval of the last iteration is output. In this way, the first bandwidth that meets the actual research needs can be selected, which provides the necessary and appropriate spatial scale for subsequent evaluation. Then, based on the first bandwidth and the second geographic feature data, a geographical weighted regression model is constructed and solved, and an evaluation is performed according to the solution results. The geographical weighted regression model constructed and solved in this way uses geographic feature data containing multiple elements and appropriate bandwidth, which can evaluate the impact of nitrogen deposition on vegetation growth conditions under the appropriate spatial scale and the influence of multiple elements.
[0010] Furthermore, the first geographic feature data is obtained by preprocessing the original geographic feature data, and specifically includes:
[0011] Acquire original geographic feature data of the area to be evaluated, and perform corresponding preprocessing according to the category of the original geographic feature data to obtain preprocessed feature data;
[0012] The preprocessed feature data is input into a preset spatial scale detection model, the optimal spatial scale is determined according to a preset first detection method, and the preprocessed feature data is resampled to the optimal spatial scale to obtain and output first geographic feature data; wherein the first geographic feature data includes normalized vegetation index, vegetation type, atmospheric nitrogen deposition data, total precipitation, wind speed, relative humidity, surface temperature, solar radiation value, elevation, slope, slope aspect, population density, night lights and gross domestic product.
[0013] In this way, the original geographic feature data is obtained and preprocessed accordingly according to the data category, and then the optimal spatial scale is determined according to the first detection method and the preprocessed feature data is resampled to the optimal spatial scale. This can unify the original geographic feature data in data type and spatial scale, which is more in line with research needs.
[0014] Further, the iterative processing of the preset first bandwidth interval based on the second geographic feature data, updating the boundary value of the first bandwidth interval according to the preset first calculation method during iteration, until the first bandwidth interval meets the preset condition, and outputting the first bandwidth corresponding to the first bandwidth interval of the last iteration, specifically includes:
[0015] Based on the second geographic feature data, iteratively process the preset first bandwidth interval, divide the first bandwidth interval into a first interval and a second interval according to the first division method, and update the boundary value of the first bandwidth interval according to the first characteristic value of the first interval and the second characteristic value of the second interval calculated by the first calculation method until the difference between the first characteristic value and the second characteristic value is less than a preset threshold value;
[0016] The first division method specifically includes: calculating the right boundary value of the first interval based on a preset parameter and the left boundary value of the first bandwidth interval, and calculating the left boundary value of the second interval based on the preset parameter and the right boundary value of the first bandwidth interval; in each iteration, dividing the first bandwidth interval into a first interval and a second interval according to the first division method, respectively calculating the first eigenvalue of the right boundary value of the first interval and the second eigenvalue of the left boundary value of the second interval according to the first calculation method, and taking the interval corresponding to the larger one of the first eigenvalue and the second eigenvalue as the first bandwidth interval in the next iteration;
[0017] A first bandwidth interval in the last iteration is selected, a first eigenvalue of the first bandwidth interval is calculated by a first calculation method, and a first bandwidth corresponding to the first eigenvalue is output.
[0018] In this way, the first bandwidth interval is continuously divided during the iteration, and the first bandwidth interval is updated according to the first calculation method until the first bandwidth interval meets the conditions, and then the first bandwidth corresponding to the first bandwidth interval at the last iteration is calculated and output, so that the most appropriate bandwidth in spatial scale that meets the actual research needs can be selected, providing the necessary and appropriate spatial scale for subsequent model construction.
[0019] Furthermore, the first calculation method specifically includes:
[0020] Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample;
[0021] Constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth;
[0022] Based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized difference vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated;
[0023] Based on the spatial weight matrix and the estimated standard deviation, according to a first feature calculation method, an eigenvalue corresponding to the current bandwidth is calculated and output.
[0024] In this way, we first construct multiple corresponding spatial weight matrices based on the spatial distance between each sample and the current bandwidth, and then construct a geographically weighted regression equation and calculate the estimated standard deviation. At the same time, we calculate the eigenvalue corresponding to the current bandwidth based on the first feature calculation method. This allows us to quickly and accurately calculate the estimated standard deviation, and then calculate the eigenvalue corresponding to the current bandwidth, thereby speeding up the iteration of the first bandwidth interval.
[0025] Furthermore, based on the first bandwidth and the second geographic feature data, a geographically weighted regression model is constructed and solved, and according to the solution result, the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated is evaluated, specifically including:
[0026] Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample;
[0027] Constructing a spatial weight matrix corresponding to each sample according to the first bandwidth and the spatial distance;
[0028] Based on the spatial weight matrix, taking the normalized vegetation index as the dependent variable and the variables in the second geographic feature data other than the dependent variable as independent variables, constructing and solving a geographically weighted regression model to obtain a solution result;
[0029] Based on the solution results, the impact of atmospheric nitrogen deposition on vegetation growth in the assessed area is evaluated.
[0030] In this way, we first construct multiple corresponding spatial weight matrices based on the spatial distance and current bandwidth between each sample, and then construct and solve the geographically weighted regression model. At the same time, based on the solution results, we evaluate the regionalized impact of atmospheric nitrogen deposition on vegetation growth in each region. In this way, we can construct a geographically weighted regression model that better meets actual needs and improve the accuracy and confidence of the evaluation results.
[0031] Furthermore, the first geographic feature data is iterated multiple times according to a preset first iteration method to obtain and output second geographic feature data, specifically including:
[0032] Inputting the first geographic feature data into a preset first screening model, performing multiple iterations of training according to a preset first training method, screening the first geographic feature data during iterations, and completing the training when the number of items of the first geographic feature data reaches a preset value;
[0033] The preset first training method includes: according to the category of the first geographic feature data, converting the first geographic feature data into multiple feature variables through a preset first transformation method, and inputting the first screening model, screening and eliminating one or more feature variables during iteration, until the number of the feature variables reaches a preset value to complete the training; in each iteration, obtaining the predicted values of the first screening model corresponding to the multiple feature variables, sorting the predicted values from low to high to obtain a predicted sequence, eliminating one or more feature variables at the low end of the predicted value of the predicted sequence according to a preset threshold, and using the remaining feature variables after elimination as the initial feature variables for the next iteration;
[0034] The first geographic feature data of the last iteration is output as the second geographic feature data.
[0035] In this way, iterating the first geographic feature data and screening out irrelevant feature variables during the iteration can reduce the coupling of the data and select the feature variable combination most closely related to the research objective, which provides assistance for the subsequent selection of the first bandwidth and the construction of a geographically weighted regression model.
[0036] In a second aspect, the present application provides a device for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth, including a first iteration module, a second iteration module and a model solving module;
[0037] The first iteration module is used to obtain first geographic feature data of the area to be evaluated, and iterate the first geographic feature data multiple times according to a preset first iteration method to obtain and output second geographic feature data; wherein the first geographic feature data is obtained by preprocessing the original geographic feature data; the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, terrain data and socio-economic data;
[0038] The second iteration module is used to iteratively process the preset first bandwidth interval based on the second geographic feature data, update the boundary value of the first bandwidth interval according to the preset first calculation method during iteration, until the first bandwidth interval meets the preset condition, and output the first bandwidth corresponding to the first bandwidth interval of the last iteration;
[0039] The model solving module is used to construct and solve a geographically weighted regression model based on the first bandwidth and the second geographic feature data, and to evaluate the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated based on the solution results.
[0040] Further, the first iteration module includes a first data processing submodule; the first data processing submodule includes a data preprocessing unit and a data resampling unit;
[0041] The data preprocessing unit is used to obtain the original geographic feature data of the area to be evaluated, and perform corresponding preprocessing according to the category of the original geographic feature data to obtain preprocessed feature data;
[0042] The data resampling unit is used to input the preprocessed feature data into a preset spatial scale detection model, determine the optimal spatial scale according to a preset first detection method, and resample the preprocessed feature data to the optimal spatial scale to obtain and output first geographic feature data; wherein the first geographic feature data includes normalized vegetation index, vegetation type, atmospheric nitrogen deposition data, total precipitation, wind speed, relative humidity, surface temperature, solar radiation value, elevation, slope, slope aspect, population density, night lights and gross domestic product.
[0043] Further, the second iteration module includes a second iteration submodule and a bandwidth calculation submodule;
[0044] The second iteration submodule is used to iteratively process the preset first bandwidth interval based on the second geographic feature data, divide the first bandwidth interval into a first interval and a second interval according to the first division method, and update the boundary value of the first bandwidth interval according to the first characteristic value of the first interval and the second characteristic value of the second interval calculated by the first calculation method, until the difference between the first characteristic value and the second characteristic value is less than a preset threshold;
[0045] The first division method specifically includes: calculating the right boundary value of the first interval based on a preset parameter and the left boundary value of the first bandwidth interval, and calculating the left boundary value of the second interval based on the preset parameter and the right boundary value of the first bandwidth interval; in each iteration, dividing the first bandwidth interval into a first interval and a second interval according to the first division method, respectively calculating the first eigenvalue of the right boundary value of the first interval and the second eigenvalue of the left boundary value of the second interval according to the first calculation method, and taking the interval corresponding to the larger one of the first eigenvalue and the second eigenvalue as the first bandwidth interval in the next iteration;
[0046] The bandwidth calculation submodule is used to select a first bandwidth interval in the last iteration, calculate a first eigenvalue of the first bandwidth interval by a first calculation method, and output a first bandwidth corresponding to the first eigenvalue.
[0047] Furthermore, the first calculation method specifically includes:
[0048] Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample;
[0049] Constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth;
[0050] Based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized difference vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated;
[0051] Based on the spatial weight matrix and the estimated standard deviation, according to a first feature calculation method, an eigenvalue corresponding to the current bandwidth is calculated and output. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] Figure 1 : A flow chart of an embodiment of a method for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth provided by the present invention;
[0053] Figure 2 : A module structure diagram of an embodiment of a device for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth provided by the present invention. DETAILED DESCRIPTION
[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0055] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, unless otherwise clearly and specifically defined, "multiple" and "several" mean two or more.
[0056] Embodiment 1
[0057] Please refer to Figure 1 , a method for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth provided by an embodiment of the present invention, includes steps S1 to S3, each of which is specifically as follows:
[0058] Step S1: Acquire first geographic feature data of the area to be evaluated, and iterate the first geographic feature data multiple times according to a preset first iteration method to obtain and output second geographic feature data; wherein the first geographic feature data is obtained by preprocessing the original geographic feature data; the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, terrain data and socio-economic data.
[0059] In an optional embodiment, the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, topographic data and socio-economic data, and the preferred data sets corresponding to each of them are specifically:
[0060] The vegetation growth remote sensing data, the preferred data set is: Normalized Difference Vegetation Index data of MODIS data MOD13A1 product, with a temporal resolution of 16 days and a spatial resolution of 500m;
[0061] The vegetation type data, the preferred data set is: vegetation type data extracted from the International Geo-Biosphere Program Land Cover Classification Scheme (LC_Type1) data of MODIS data MCD12Q1 product;
[0062] The nitrogen deposition data, the preferred data set is: nitrogen deposition data integrated from the spatiotemporal pattern data of nitrogen wet deposition in China from 2001 to 2005, 2006 to 2010, and 2010 to 2015;
[0063] The meteorological data include surface temperature data, 2m dew point temperature data, 2m temperature data, surface net solar radiation data, 10m wind speed latitudinal U component data, 10m wind speed latitudinal V component data and total precipitation data, and the preferred data set is: land reanalysis data set ERA5-Land;
[0064] The terrain data includes elevation data, slope data and aspect data, and the preferred data set is: the Space Shuttle Radar Topography Mission Data Set SRTM3, which has a spatial resolution of 90m;
[0065] The socio-economic data include population density data, night light data and gross domestic product data, and the preferred data sets are: WorldPop, a population information data set of developing countries, with a temporal resolution of 1 year and a spatial resolution of 1 km; DMSP-OLS, a night light data set, with a temporal resolution of 1 year and a spatial resolution of 1 km; and the kilometer grid data set of the spatial distribution of historical GDP from 1990 to 2015 released by the National Tibetan Plateau Science Data Center, with a temporal resolution of 1 year and a spatial resolution of 1 km.
[0066] Furthermore, the first geographic feature data is obtained by preprocessing the original geographic feature data, and specifically includes:
[0067] Acquire original geographic feature data of the area to be evaluated, and perform corresponding preprocessing according to the category of the original geographic feature data to obtain preprocessed feature data;
[0068] The preprocessed feature data is input into a preset spatial scale detection model, the optimal spatial scale is determined according to a preset first detection method, and the preprocessed feature data is resampled to the optimal spatial scale to obtain and output first geographic feature data; wherein the first geographic feature data includes normalized vegetation index, vegetation type, atmospheric nitrogen deposition data, total precipitation, wind speed, relative humidity, surface temperature, solar radiation value, elevation, slope, slope aspect, population density, night lights and gross domestic product.
[0069] In an optional embodiment, the performing corresponding preprocessing according to the category of the original geographic feature data specifically includes:
[0070] De-clouding the normalized vegetation index data in the vegetation growth remote sensing data in batches to obtain a normalized vegetation index;
[0071] Extract the LC_Type1 data in the vegetation type data according to Table 1 to obtain the vegetation type;
[0072] Integrate the spatiotemporal pattern data of nitrogen wet deposition in China in the nitrogen deposition data to obtain atmospheric nitrogen wet deposition data;
[0073] The 2m dew point temperature data (d2m) and 2m temperature data (t2m) in the meteorological data are first converted from Fahrenheit to Celsius, and then converted according to The relative humidity (RH) is synthesized; the 10m latitudinal U component and the 10m latitudinal V component of the wind speed are calculated according to The total precipitation in the meteorological data is first screened, and the monthly total precipitation is obtained by adding them up by month, and the unit is converted from m to mm to obtain the total precipitation; the surface solar net radiation data in the meteorological data is first screened, and the monthly total solar net radiation is obtained by adding them up by month, and the unit is converted from J / m 2 Convert to MJ / m 2 , get the solar radiation value;
[0074] Solving the elevation data in the terrain data to obtain slope and aspect data;
[0075] The population density data, night light data and gross domestic product data in the socio-economic data are divided according to regions to obtain population density, night light and gross domestic product.
[0076] Table 1
[0077]
[0078]
[0079] In an optional embodiment, the inputting the preprocessed feature data into a preset spatial scale detection model, determining the optimal spatial scale according to a preset first detection method, and resampling the preprocessed feature data to the optimal spatial scale is specifically as follows:
[0080] The preprocessed feature data is input into a spatial scale detection model, and after factor detection and parameter optimization detection, an optimal spatial scale is determined, and the preprocessed feature data is resampled to the optimal spatial scale.
[0081] In an optional embodiment, the optimal spatial scale is specifically: 10 km.
[0082] In this way, the original geographic feature data is obtained and preprocessed accordingly according to the data category, and then the optimal spatial scale is determined according to the first detection method and the preprocessed feature data is resampled to the optimal spatial scale. This can unify the original geographic feature data in data type and spatial scale, which is more in line with research needs.
[0083] Furthermore, the first geographic feature data is iterated multiple times according to a preset first iteration method to obtain and output second geographic feature data, specifically including:
[0084] Inputting the first geographic feature data into a preset first screening model, performing multiple iterations of training according to a preset first training method, screening the first geographic feature data during iterations, and completing the training when the number of items of the first geographic feature data reaches a preset value;
[0085] The preset first training method includes: according to the category of the first geographic feature data, converting the first geographic feature data into multiple feature variables through a preset first transformation method, and inputting the first screening model, screening and eliminating one or more feature variables during iteration, until the number of the feature variables reaches a preset value to complete the training; in each iteration, obtaining the predicted values of the first screening model corresponding to the multiple feature variables, sorting the predicted values from low to high to obtain a predicted sequence, eliminating one or more feature variables at the low end of the predicted value of the predicted sequence according to a preset threshold, and using the remaining feature variables after elimination as the initial feature variables for the next iteration;
[0086] The first geographic feature data of the last iteration is output as the second geographic feature data.
[0087] In the embodiment provided by the present invention, the second geographic feature data includes dependent variables and independent variables. According to the specific problem studied, the dependent variable is specifically: normalized vegetation index; in the second geographic feature data iterated by the first iteration method, except for the dependent variable, the remaining variables are independent variables. In this embodiment, the preferred independent variables are specifically: atmospheric nitrogen deposition data, wind speed, surface temperature, total precipitation, relative humidity and solar radiation value.
[0088] In an optional embodiment, in each iteration, the predicted values of the first screening model corresponding to the multiple feature variables are obtained, the predicted values are sorted from low to high to obtain a prediction sequence, and one or more feature variables at the low end of the predicted value of the prediction sequence are eliminated according to a preset threshold, specifically:
[0089] In each iteration, the predicted values of the first screening model corresponding to the multiple characteristic variables are obtained;
[0090] Calculating the predicted importance of the plurality of characteristic variables according to the predicted values of the plurality of characteristic variables;
[0091] Sort the prediction importance from low to high to obtain a prediction sequence;
[0092] One or more feature variables at the lower end of the predicted value of the predicted sequence are eliminated according to the preset threshold.
[0093] In an optional embodiment, the first screening model is specifically a tree model based on machine learning, including but not limited to a decision tree model, a random forest model, a gradient boosting tree model, an XGBoost model and an AdaBoost model.
[0094] In an optional embodiment, the preferred scheme of the first screening model is: random forest model.
[0095] In this way, iterating the first geographic feature data and screening out irrelevant feature variables during the iteration can reduce the coupling of the data and select the feature variable combination that is most closely related to the research objective, which provides assistance for the subsequent selection of the first bandwidth and the construction of a geographical weighted regression model.
[0096] Step S2: Based on the second geographic feature data, the preset first bandwidth interval is iteratively processed, and the boundary value of the first bandwidth interval is updated according to the preset first calculation method during iteration until the first bandwidth interval meets the preset conditions, and the first bandwidth corresponding to the first bandwidth interval of the last iteration is output.
[0097] Further, the iterative processing of the preset first bandwidth interval based on the second geographic feature data, updating the boundary value of the first bandwidth interval according to the preset first calculation method during iteration, until the first bandwidth interval meets the preset condition, and outputting the first bandwidth corresponding to the first bandwidth interval of the last iteration, specifically includes:
[0098] Based on the second geographic feature data, iteratively process the preset first bandwidth interval, divide the first bandwidth interval into a first interval and a second interval according to the first division method, and update the boundary value of the first bandwidth interval according to the first characteristic value of the first interval and the second characteristic value of the second interval calculated by the first calculation method until the difference between the first characteristic value and the second characteristic value is less than a preset threshold value;
[0099] The first division method specifically includes: calculating the right boundary value of the first interval based on a preset parameter and the left boundary value of the first bandwidth interval, and calculating the left boundary value of the second interval based on the preset parameter and the right boundary value of the first bandwidth interval; in each iteration, dividing the first bandwidth interval into a first interval and a second interval according to the first division method, respectively calculating the first eigenvalue of the right boundary value of the first interval and the second eigenvalue of the left boundary value of the second interval according to the first calculation method, and taking the interval corresponding to the larger one of the first eigenvalue and the second eigenvalue as the first bandwidth interval in the next iteration;
[0100] A first bandwidth interval in the last iteration is selected, a first eigenvalue of the first bandwidth interval is calculated by a first calculation method, and a first bandwidth corresponding to the first eigenvalue is output.
[0101] Specifically, in order to illustrate the specific process of the above-mentioned iterative method, in a preferred solution of an optional embodiment, the preset parameters are specifically: The specific process of its iterative processing is:
[0102] Calculate the minimum spatial distance between all samples in the second geographic feature data With the maximum value And set the initial iteration interval to [a,b];
[0103] Calculate the right boundary value of the first interval, specifically: Calculate the left boundary value of the second interval, specifically:
[0104] Based on the right boundary value of the first interval, the corresponding first eigenvalue is calculated as: AICc(h2); based on the left boundary value of the second interval, the corresponding second eigenvalue is calculated as: aICc(h1);
[0105] The interval corresponding to the larger of the first eigenvalue and the second eigenvalue is used as the first bandwidth interval in the next iteration;
[0106] The iteration is completed until the difference between the first eigenvalue and the second eigenvalue is less than a preset threshold, and the first eigenvalue of the first bandwidth interval at the last iteration is calculated, and the first bandwidth corresponding to the first eigenvalue is output.
[0107] In an optional embodiment, the preferred value of the preset threshold is: 0.001.
[0108] In this way, the first bandwidth interval is continuously divided during the iteration, and the first bandwidth interval is updated according to the first calculation method until the first bandwidth interval meets the conditions, and then the first bandwidth corresponding to the first bandwidth interval at the last iteration is calculated and output, so that the most appropriate bandwidth in spatial scale that meets the actual research needs can be selected, providing the necessary and appropriate spatial scale for subsequent model construction.
[0109] Furthermore, the first calculation method specifically includes:
[0110] Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample;
[0111] Constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth;
[0112] Based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized difference vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated;
[0113] Based on the spatial weight matrix and the estimated standard deviation, according to a first feature calculation method, an eigenvalue corresponding to the current bandwidth is calculated and output.
[0114] In an optional embodiment, the spatial distance is specifically:
[0115]
[0116] Among them, d ij is the spatial distance between sample i and sample j, λ is the scaling factor, μ i 、v i is the longitude and latitude of sample i, μ j 、v j are the longitude and latitude of sample j.
[0117] In an optional embodiment, the preferred value of the proportional factor λ is 1.
[0118] In an optional embodiment, constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth is specifically:
[0119] According to the spatial distance, construct a spatial distance attenuation kernel function corresponding to each sample;
[0120] According to the spatial distance attenuation kernel function, a spatial weight matrix corresponding to each sample is constructed.
[0121] In an optional embodiment, constructing a spatial distance attenuation kernel function corresponding to each sample according to the spatial distance is specifically:
[0122]
[0123] Among them, W ij is the weight of sample i observing sample j, and b is the current bandwidth.
[0124] In an optional embodiment, the spatial weight matrix corresponding to each sample is constructed according to the spatial distance attenuation kernel function, specifically:
[0125]
[0126] Among them, W i is the spatial regression weight matrix corresponding to sample i.
[0127] Furthermore, the first feature calculation method is specifically:
[0128]
[0129] Among them, AICc is the first eigenvalue, n is the number of samples of the second geographic feature data, To estimate the standard deviation, S is the hat matrix corresponding to the current spatial bandwidth.
[0130] In an optional embodiment, based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated, specifically:
[0131] Based on the spatial weight matrix, a geographically weighted regression model corresponding to each sample is constructed with the normalized difference vegetation index as the dependent variable, and then the intercept of the geographically weighted regression model is obtained;
[0132] Based on the hat matrix and the intercept of the geographically weighted regression model, a geographically weighted regression equation is constructed to obtain the estimated value of the dependent variable;
[0133] Calculates the estimated standard deviation of the estimated value of the dependent variable from the true value.
[0134] In an optional embodiment, the geographically weighted regression model is specifically:
[0135]
[0136] Among them, y i is the true value of the dependent variable, (μ i , v i ) is the spatial coordinate of sample i, μ i 、v i denote the longitude and latitude of sample i, respectively, β0(μ i , v i ) is the intercept, x ik is the kth independent variable of sample i, β k (μ i , v i ) is the regression coefficient of the kth independent variable of sample i.
[0137] In an optional embodiment, the hat matrix S corresponding to the current spatial bandwidth is specifically:
[0138] S=X(X T WX) -1 X T W;
[0139] in, is the independent variable matrix, n represents the number of samples, p represents the number of independent variables, m1 to m n represents the intercept of the geographically weighted regression model corresponding to each sample, and W is the spatial regression weight matrix.
[0140] In an optional embodiment, the geographically weighted regression equation is specifically:
[0141]
[0142] The regression coefficient is specifically:
[0143]
[0144] Among them, X is the independent variable matrix, W is the spatial regression weight matrix, and Y = [y1, y2, ..., y n ] T is the dependent variable vector, is the estimated value of the dependent variable.
[0145] In this way, we first construct multiple corresponding spatial weight matrices based on the spatial distance between each sample and the current bandwidth, and then construct a geographically weighted regression equation and calculate the estimated standard deviation. At the same time, we calculate the eigenvalue corresponding to the current bandwidth based on the first feature calculation method. This allows us to quickly and accurately calculate the estimated standard deviation, and then calculate the eigenvalue corresponding to the current bandwidth, thereby speeding up the iteration of the first bandwidth interval.
[0146] Step S3: Based on the first bandwidth and the second geographic feature data, a geographically weighted regression model is constructed and solved, and according to the solution result, the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated is evaluated.
[0147] Furthermore, based on the first bandwidth and the second geographic feature data, a geographically weighted regression model is constructed and solved, and according to the solution result, the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated is evaluated, specifically including:
[0148] Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample;
[0149] Constructing a spatial weight matrix corresponding to each sample according to the first bandwidth and the spatial distance;
[0150] Based on the spatial weight matrix, taking the normalized vegetation index as the dependent variable and the variables in the second geographic feature data other than the dependent variable as independent variables, constructing and solving a geographically weighted regression model to obtain a solution result;
[0151] Based on the solution results, the impact of atmospheric nitrogen deposition on vegetation growth in the assessed area is evaluated.
[0152] Furthermore, the geographically weighted regression model is specifically:
[0153]
[0154] Among them, y i is the true value of the dependent variable, (μ i , v i ) is the spatial coordinate of sample i, μ i , v i denote the longitude and latitude of sample i, respectively, β0(μ i , v i ) is the intercept, x ik is the kth independent variable of sample i, β k (μ i , v i ) is the regression coefficient of the kth independent variable of sample i.
[0155] In an optional embodiment, the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated is evaluated based on the solution result, specifically:
[0156] According to the solution results, the regression coefficient is obtained, and the fitting accuracy of the geographically weighted regression model is calculated; the fitting accuracy of the geographically weighted regression model includes the overall accuracy and the accuracy of each sample point;
[0157] The overall accuracy is specifically:
[0158]
[0159] The precision of each sample point is specifically:
[0160]
[0161] Among them, y i is the true value of the dependent variable at observation point i, y ij is the true value of the dependent variable of the jth neighboring point of observation point i, is the estimated value of the dependent variable at observation point i, is the estimated value of the dependent variable of the jth neighboring point of observation point i, is the mean of the true values of the dependent variable at all observation points, is the mean of the true values of the dependent variable of all neighboring points of observation point i;
[0162] Based on the regression coefficient, the impact of atmospheric nitrogen deposition on vegetation growth in the area to be assessed is evaluated, and a visualization map of the spatial regression parameters of nitrogen deposition in the area to be assessed is compiled.
[0163] In this way, we first construct multiple corresponding spatial weight matrices based on the spatial distance and current bandwidth between each sample, and then construct and solve the geographically weighted regression model. At the same time, based on the solution results, we evaluate the regionalized impact of atmospheric nitrogen deposition on vegetation growth in each region. In this way, we can construct a geographically weighted regression model that better meets actual needs and improve the accuracy and confidence of the evaluation results.
[0164] In this way, the original geographic feature data containing multiple elements are first obtained and preprocessed, and the obtained first geographic feature data are iteratively processed to obtain the second geographic feature data, which can provide geographic feature data of multiple elements for subsequent evaluation. Then, based on the second geographic feature data, the first bandwidth interval is iterated until the first bandwidth interval meets the preset conditions, and the first bandwidth corresponding to the first bandwidth interval of the last iteration is output. In this way, the first bandwidth that meets the actual research needs can be selected, which provides the necessary and appropriate spatial scale for subsequent evaluation. Then, based on the first bandwidth and the second geographic feature data, a geographical weighted regression model is constructed and solved, and an evaluation is performed according to the solution results. The geographical weighted regression model constructed and solved in this way uses geographic feature data containing multiple elements and appropriate bandwidth, which can evaluate the impact of nitrogen deposition on vegetation growth conditions under the appropriate spatial scale and the influence of multiple elements.
[0165] Embodiment 2
[0166] Please refer to Figure 2 , a regionalized impact assessment device of atmospheric nitrogen deposition on vegetation growth provided by an embodiment of the present invention, comprising a first iteration module 210, a second iteration module 220 and a model solving module 230;
[0167] The first iteration module 210 is used to obtain first geographic feature data of the area to be evaluated, and iterate the first geographic feature data multiple times according to a preset first iteration method to obtain and output second geographic feature data; wherein the first geographic feature data is obtained by preprocessing the original geographic feature data; the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, terrain data and socio-economic data;
[0168] The second iteration module 220 is used to iteratively process the preset first bandwidth interval based on the second geographic feature data, update the boundary value of the first bandwidth interval according to the preset first calculation method during iteration, until the first bandwidth interval meets the preset condition, and output the first bandwidth corresponding to the first bandwidth interval of the last iteration;
[0169] The model solving module 230 is used to construct and solve a geographically weighted regression model based on the first bandwidth and the second geographic feature data, and evaluate the impact of atmospheric nitrogen deposition on vegetation growth in the evaluation area according to the solution result.
[0170] Further, the first iteration module 210 includes a first data processing submodule 211; the first data processing submodule 211 includes a data preprocessing unit 2111 and a data resampling unit 2112;
[0171] The data preprocessing unit 2111 is used to obtain the original geographic feature data of the area to be evaluated, and perform corresponding preprocessing according to the category of the original geographic feature data to obtain preprocessed feature data;
[0172] The data resampling unit 2112 is used to input the preprocessed feature data into a preset spatial scale detection model, determine the optimal spatial scale according to the preset first detection method, and resample the preprocessed feature data to the optimal spatial scale to obtain and output the first geographic feature data; wherein the first geographic feature data includes normalized vegetation index, vegetation type, atmospheric nitrogen deposition data, total precipitation, wind speed, relative humidity, surface temperature, solar radiation value, elevation, slope, slope aspect, population density, night lights and gross domestic product.
[0173] Furthermore, the first iteration module 210 further includes a first iteration submodule 212; the first iteration submodule 212 includes an iteration screening subunit 2121 and a data output subunit 2122;
[0174] The iterative screening subunit 2121 is used to input the first geographic feature data into a preset first screening model, perform multiple iterative training according to a preset first training method, screen the first geographic feature data during iteration, and complete the training when the number of items of the first geographic feature data reaches a preset value;
[0175] The preset first training method includes: according to the category of the first geographic feature data, converting the first geographic feature data into multiple feature variables through a preset first transformation method, and inputting the first screening model, screening and eliminating one or more feature variables during iteration, until the number of the feature variables reaches a preset value to complete the training; in each iteration, obtaining the predicted values of the first screening model corresponding to the multiple feature variables, sorting the predicted values from low to high to obtain a predicted sequence, eliminating one or more feature variables at the low end of the predicted value of the predicted sequence according to a preset threshold, and using the remaining feature variables after elimination as the initial feature variables for the next iteration;
[0176] The data output subunit 2122 is used to output the first geographic feature data of the last iteration as the second geographic feature data.
[0177] Further, the second iteration module 220 includes a second iteration submodule 221 and a bandwidth calculation submodule 222;
[0178] The second iteration submodule 221 is used to iteratively process the preset first bandwidth interval based on the second geographic feature data, divide the first bandwidth interval into a first interval and a second interval according to the first division method, and update the boundary value of the first bandwidth interval according to the first characteristic value of the first interval and the second characteristic value of the second interval calculated by the first calculation method, until the difference between the first characteristic value and the second characteristic value is less than a preset threshold;
[0179] The first division method specifically includes: calculating the right boundary value of the first interval based on a preset parameter and the left boundary value of the first bandwidth interval, and calculating the left boundary value of the second interval based on the preset parameter and the right boundary value of the first bandwidth interval; in each iteration, dividing the first bandwidth interval into a first interval and a second interval according to the first division method, respectively calculating the first eigenvalue of the right boundary value of the first interval and the second eigenvalue of the left boundary value of the second interval according to the first calculation method, and taking the interval corresponding to the larger one of the first eigenvalue and the second eigenvalue as the first bandwidth interval in the next iteration;
[0180] The bandwidth calculation submodule 222 is used to select a first bandwidth interval in the last iteration, calculate a first eigenvalue of the first bandwidth interval by a first calculation method, and output a first bandwidth corresponding to the first eigenvalue.
[0181] Furthermore, the first calculation method specifically includes:
[0182] Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample;
[0183] Constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth;
[0184] Based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized difference vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated;
[0185] Based on the spatial weight matrix and the estimated standard deviation, according to a first feature calculation method, an eigenvalue corresponding to the current bandwidth is calculated and output.
[0186] Furthermore, the first feature calculation method is specifically:
[0187]
[0188] Wherein, AI Cc is the first characteristic value, n is the number of samples of the second geographic characteristic data, To estimate the standard deviation, S is the hat matrix corresponding to the current spatial bandwidth.
[0189] Further, the model solving module 230 includes a distance calculation submodule 231, a matrix construction submodule 232, a model solving submodule 233 and an impact assessment submodule 234;
[0190] The distance calculation submodule 231 is used to obtain the spatial coordinates of all samples in the second geographic feature data and calculate the spatial distance between each sample;
[0191] The matrix construction submodule 232 is used to construct a spatial weight matrix corresponding to each sample according to the first bandwidth and the spatial distance;
[0192] The model solving submodule 233 is used to construct and solve a geographically weighted regression model based on the spatial weight matrix, with the normalized vegetation index as the dependent variable and the variables in the second geographic feature data other than the dependent variable as independent variables, to obtain a solution result;
[0193] The impact assessment submodule 234 is used to assess the impact of atmospheric nitrogen deposition on vegetation growth in the area to be assessed based on the solution results.
[0194] Furthermore, the geographically weighted regression model is specifically:
[0195]
[0196] Among them, y i is the true value of the dependent variable, (μ i , v i ) is the spatial coordinate of sample i, μ i 、v i denote the longitude and latitude of sample i, respectively, β0(μ i , v i ) is the intercept, x ik is the kth independent variable of sample i, β k (μ i , v i ) is the regression coefficient of the kth independent variable of sample i.
[0197] In this way, the original geographic feature data containing multiple elements are first obtained and preprocessed, and the obtained first geographic feature data are iteratively processed to obtain the second geographic feature data, which can provide geographic feature data of multiple elements for subsequent evaluation. Then, based on the second geographic feature data, the first bandwidth interval is iterated until the first bandwidth interval meets the preset conditions, and the first bandwidth corresponding to the first bandwidth interval of the last iteration is output. In this way, the first bandwidth that meets the actual research needs can be selected, which provides the necessary and appropriate spatial scale for subsequent evaluation. Then, based on the first bandwidth and the second geographic feature data, a geographical weighted regression model is constructed and solved, and an evaluation is performed according to the solution results. The geographical weighted regression model constructed and solved in this way uses geographic feature data containing multiple elements and appropriate bandwidth, which can evaluate the impact of nitrogen deposition on vegetation growth conditions under the appropriate spatial scale and the influence of multiple elements.
[0198] Accordingly, the embodiments of the present invention also adaptively provide a computer device and a computer-readable storage medium.
[0199] The computer device comprises: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor;
[0200] When the processor executes the computer program, the method for evaluating the regionalized impact of atmospheric nitrogen deposition on vegetation growth as described above is implemented.
[0201] The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the regionalized impact assessment method of atmospheric nitrogen deposition on vegetation growth as described above.
[0202] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for assessing the regional impact of atmospheric nitrogen deposition on vegetation growth, characterized in that: include: Acquire first geographic feature data of the area to be evaluated, and iterate the first geographic feature data multiple times according to a preset first iteration method to obtain and output second geographic feature data; wherein the first geographic feature data is obtained by preprocessing the original geographic feature data; the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, terrain data and socio-economic data; Iteratively process the preset first bandwidth interval based on the second geographic feature data, update the boundary value of the first bandwidth interval according to a preset first calculation method during iteration, until the first bandwidth interval meets a preset condition, and output the first bandwidth corresponding to the first bandwidth interval of the last iteration; Based on the first bandwidth and the second geographic feature data, construct and solve a geographically weighted regression model, and evaluate the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated according to the solution result; The first geographic feature data is obtained by preprocessing the original geographic feature data, and specifically includes: Acquire original geographic feature data of the area to be evaluated, and perform corresponding preprocessing according to the category of the original geographic feature data to obtain preprocessed feature data; Input the preprocessed feature data into a preset spatial scale detection model, determine the optimal spatial scale according to a preset first detection method, and resample the preprocessed feature data to the optimal spatial scale to obtain and output first geographic feature data; wherein the first geographic feature data includes normalized vegetation index, vegetation type, atmospheric nitrogen deposition data, total precipitation, wind speed, relative humidity, surface temperature, solar radiation value, elevation, slope, slope aspect, population density, night lights and gross domestic product; The iterative processing of the preset first bandwidth interval based on the second geographic feature data, updating the boundary value of the first bandwidth interval according to a preset first calculation method during iteration, until the first bandwidth interval meets a preset condition, and outputting the first bandwidth corresponding to the first bandwidth interval of the last iteration specifically includes: Based on the second geographic feature data, iteratively process the preset first bandwidth interval, divide the first bandwidth interval into a first interval and a second interval according to the first division method, and update the boundary value of the first bandwidth interval according to the first characteristic value of the first interval and the second characteristic value of the second interval calculated by the first calculation method until the difference between the first characteristic value and the second characteristic value is less than a preset threshold value; The first division method specifically includes: calculating the right boundary value of the first interval based on a preset parameter and the left boundary value of the first bandwidth interval, and calculating the left boundary value of the second interval based on the preset parameter and the right boundary value of the first bandwidth interval; in each iteration, dividing the first bandwidth interval into a first interval and a second interval according to the first division method, respectively calculating the first eigenvalue of the right boundary value of the first interval and the second eigenvalue of the left boundary value of the second interval according to the first calculation method, and taking the interval corresponding to the larger one of the first eigenvalue and the second eigenvalue as the first bandwidth interval in the next iteration; Selecting a first bandwidth interval in the last iteration, calculating a first eigenvalue of the first bandwidth interval by a first calculation method, and outputting a first bandwidth corresponding to the first eigenvalue; The first calculation method specifically includes: Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample; Constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth; Based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized difference vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated; Based on the spatial weight matrix and the estimated standard deviation, according to a first feature calculation method, calculate and output an eigenvalue corresponding to the current bandwidth; The step of iterating the first geographic feature data multiple times according to the preset first iteration method to obtain and output the second geographic feature data specifically includes: Inputting the first geographic feature data into a preset first screening model, performing multiple iterations of training according to a preset first training method, screening the first geographic feature data during iterations, and completing the training when the number of items of the first geographic feature data reaches a preset value; The preset first training method includes: according to the category of the first geographic feature data, converting the first geographic feature data into multiple feature variables through a preset first transformation method, and inputting the first screening model, screening and eliminating one or more feature variables during iteration, until the number of the feature variables reaches a preset value to complete the training; in each iteration, obtaining the predicted values of the first screening model corresponding to the multiple feature variables, sorting the predicted values from low to high to obtain a predicted sequence, eliminating one or more feature variables at the low end of the predicted value of the predicted sequence according to a preset threshold, and using the remaining feature variables after elimination as the initial feature variables for the next iteration; the first screening model is a tree model based on machine learning; The first geographic feature data of the last iteration is output as the second geographic feature data.
2. The method for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth according to claim 1, characterized in that: The step of constructing and solving a geographically weighted regression model based on the first bandwidth and the second geographic feature data, and evaluating the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated according to the solution result, specifically includes: Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample; Constructing a spatial weight matrix corresponding to each sample according to the first bandwidth and the spatial distance; Based on the spatial weight matrix, taking the normalized vegetation index as the dependent variable and the variables in the second geographic feature data other than the dependent variable as independent variables, constructing and solving a geographically weighted regression model to obtain a solution result; Based on the solution results, the impact of atmospheric nitrogen deposition on vegetation growth in the assessed area is evaluated.
3. A device for evaluating the regional impact of atmospheric nitrogen deposition on vegetation growth, characterized in that: It includes a first iteration module, a second iteration module and a model solving module; The first iteration module is used to obtain first geographic feature data of the area to be evaluated, and iterate the first geographic feature data multiple times according to a preset first iteration method to obtain and output second geographic feature data; wherein the first geographic feature data is obtained by preprocessing the original geographic feature data; the original geographic feature data includes vegetation growth remote sensing data, vegetation type data, nitrogen deposition data, meteorological data, terrain data and socio-economic data; The second iteration module is used to iteratively process the preset first bandwidth interval based on the second geographic feature data, update the boundary value of the first bandwidth interval according to the preset first calculation method during iteration, until the first bandwidth interval meets the preset condition, and output the first bandwidth corresponding to the first bandwidth interval of the last iteration; The model solving module is used to construct and solve a geographically weighted regression model based on the first bandwidth and the second geographic feature data, and evaluate the impact of atmospheric nitrogen deposition on vegetation growth in the area to be evaluated according to the solution result; The first iteration module includes a first data processing submodule; the first data processing submodule includes a data preprocessing unit and a data resampling unit; The data preprocessing unit is used to obtain the original geographic feature data of the area to be evaluated, and perform corresponding preprocessing according to the category of the original geographic feature data to obtain preprocessed feature data; The data resampling unit is used to input the preprocessed feature data into a preset spatial scale detection model, determine the optimal spatial scale according to the preset first detection method, and resample the preprocessed feature data to the optimal spatial scale to obtain and output first geographic feature data; wherein the first geographic feature data includes normalized vegetation index, vegetation type, atmospheric nitrogen deposition data, total precipitation, wind speed, relative humidity, surface temperature, solar radiation value, elevation, slope, slope aspect, population density, night lights and gross domestic product; The second iteration module includes a second iteration submodule and a bandwidth calculation submodule; The second iteration submodule is used to iteratively process the preset first bandwidth interval based on the second geographic feature data, divide the first bandwidth interval into a first interval and a second interval according to the first division method, and update the boundary value of the first bandwidth interval according to the first characteristic value of the first interval and the second characteristic value of the second interval calculated by the first calculation method, until the difference between the first characteristic value and the second characteristic value is less than a preset threshold; The first division method specifically includes: calculating the right boundary value of the first interval based on a preset parameter and the left boundary value of the first bandwidth interval, and calculating the left boundary value of the second interval based on the preset parameter and the right boundary value of the first bandwidth interval; in each iteration, dividing the first bandwidth interval into a first interval and a second interval according to the first division method, respectively calculating the first eigenvalue of the right boundary value of the first interval and the second eigenvalue of the left boundary value of the second interval according to the first calculation method, and taking the interval corresponding to the larger one of the first eigenvalue and the second eigenvalue as the first bandwidth interval in the next iteration; The bandwidth calculation submodule is used to select a first bandwidth interval in the last iteration, calculate a first eigenvalue of the first bandwidth interval by a first calculation method, and output a first bandwidth corresponding to the first eigenvalue; The first calculation method specifically includes: Obtaining the spatial coordinates of all samples in the second geographic feature data, and calculating the spatial distance between each sample; Constructing a spatial weight matrix corresponding to each sample according to the spatial distance and the current bandwidth; Based on the spatial weight matrix, a geographically weighted regression equation is constructed with the normalized difference vegetation index as the dependent variable to obtain an estimated value of the dependent variable, and an estimated standard deviation between the estimated value of the dependent variable and the true value is calculated; Based on the spatial weight matrix and the estimated standard deviation, according to a first feature calculation method, calculate and output an eigenvalue corresponding to the current bandwidth; The first iteration module further comprises a first iteration submodule; the first iteration submodule comprises an iteration screening subunit and a data output subunit; The iterative screening subunit is used to input the first geographic feature data into a preset first screening model, perform multiple iterative training according to a preset first training method, screen the first geographic feature data during iteration, and complete the training when the number of items of the first geographic feature data reaches a preset value; The preset first training method includes: according to the category of the first geographic feature data, converting the first geographic feature data into multiple feature variables through a preset first transformation method, and inputting the first screening model, screening and eliminating one or more feature variables during iteration, until the number of the feature variables reaches a preset value to complete the training; in each iteration, obtaining the predicted values of the first screening model corresponding to the multiple feature variables, sorting the predicted values from low to high to obtain a predicted sequence, eliminating one or more feature variables at the low end of the predicted value of the predicted sequence according to a preset threshold, and using the remaining feature variables after elimination as the initial feature variables for the next iteration; the first screening model is a tree model based on machine learning; The data output subunit is used to output the first geographic feature data of the last iteration as the second geographic feature data.
Citation Information
Patent Citations
Air quality prediction method based on spatio-temporal bandwidth adaptive geographically weighted regression
CN112990609A
Tobacco total nitrogen content remote sensing inversion algorithm based on satellite image
CN115861832A
