A method for spatializing regional water consumption

Abstract

The application discloses a kind of regional water resource consumption spatialization method, comprising: 1) constructing research area database, integrating the grid data of water consumption data and influence factor in region;2) data is preprocessed, and eliminate dimensional difference;3) screening key influence factor;4) constructing multiple linear regression model;5) based on ArcGIS platform, in combination with the grid data of key factor, generate regional water resource demand quantity spatial distribution diagram;6) according to spatial distribution diagram, the result of rasterization is verified with original observation data, guarantee the accuracy and applicability of model.The application method integrates multiple influence factors, filters out key influence factor, uses linear regression optimization model, realizes the high-resolution spatial distribution expression of regional water resource demand quantity by means of GIS technology, solves the problem that traditional method ignores the coupling relationship of multiple influence factors, and can provide important scientific basis for water resource management, land use planning and ecosystem protection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ecology and geographic information science, and particularly relates to a spatialization method for regional water resource consumption. BACKGROUND

[0002] Ecosystem services refer to the beneficial functions and products obtained from natural ecosystems, such as water supply, climate regulation, and food production. In recent years, as human activities have increasingly impacted ecosystems, scientifically assessing and planning the spatial distribution of regional water resource demand has become a research hotspot and practical need. However, existing spatialization methods have the following problems in terms of accuracy and efficiency:

[0003] 1. Insufficient spatial resolution: Traditional methods often use low resolution, which cannot reflect the micro differences within the region. This limits their application in high-precision resource management.

[0004] 2. Limited model applicability: Existing methods rely on single variables or nonlinear models, lacking comprehensive consideration of multiple influencing factors, resulting in large prediction errors.

[0005] 3. Difficult data integration: The assessment of regional water resource demand requires the integration of meteorological, topographical, land use, and socio-economic multi-source heterogeneous data. However, the data preprocessing and analysis process is complex, restricting the application of the model.

[0006] Therefore, there is an urgent need for a spatialization analysis method that can integrate multi-source data and achieve high-precision gridding, to improve the accuracy and efficiency of regional water resource demand assessment, and support scientific ecological protection and resource planning. SUMMARY

[0007] To overcome the shortcomings of the prior art, the present application provides a spatialization method for regional water resource consumption, which integrates multiple influencing factors, selects key influencing factors, and uses a linear regression optimization model to achieve high-resolution spatial distribution expression of regional water resource demand, solving the problem of ignoring the coupling relationship of multiple influencing factors in traditional methods, and providing important scientific basis for water resource management, land use planning, and ecosystem protection.

[0008] To achieve the above purpose, the present application adopts the following technical solutions.

[0009] A spatialization method for regional water resource consumption, comprising the following steps:

[0010] Step S1, constructing a research area database, integrating water consumption data of different water types and grid data of influencing factors in the research area;

[0011] The impact factor is determined by consulting relevant academic papers, government reports and industry research materials, and the grid data of the impact factor comes from the Resource and Environment Data Center of the Chinese Academy of Sciences, the National Meteorological Data Center and the World Soil Database;

[0012] Step S2, the grid data of the impact factor and the water consumption data of different water types in the research area in step S1 are preprocessed to eliminate the dimensional difference between the data;

[0013] Step S3, based on the RStudio platform, the preprocessed grid data of the impact factor and the water consumption data of different water types in the research area are linearly fitted, and the key impact factors with fitting goodness R²≥0.7 are screened out;

[0014] Step S4, according to the key impact factors screened out in step S3, a multivariate linear regression model is constructed by using the non-negative least squares regression method based on decision tree grouping improvement;

[0015] Step S5, the multivariate linear regression model parameters optimized by the decision tree grouping are input into the ArcGIS platform, combined with the grid data of the key impact factors, and a 30m resolution regional water resource demand spatial distribution map is generated;

[0016] Step S6, the rasterized prediction results are verified with the original observed water consumption data to ensure the accuracy and applicability of the model.

[0017] Specifically, the research area database in step S1 includes precipitation data, temperature data, population data, first industry added value data, second industry added value data, third industry added value data and land use data.

[0018] Specifically, the screening process of the key impact factor in step S3 is as follows:

[0019] Step S31, according to the integrated water consumption data of different water types in the research area and the grid data of the impact factor in step S1, a plurality of candidate driving factors affecting water consumption are preset;

[0020] Step S32, using linear regression algorithm, the preset candidate driving factors are respectively linearly fitted with the actual water consumption data of each province and city and each category, and the fitting goodness R² of the model is calculated;

[0021] Step S33, according to the calculation result of the fitting goodness R², the preset candidate driving factors are sorted, and the candidate driving factors with R²≥0.7 are screened out as the key impact factors for analyzing the driving mechanism of different types of water consumption in each province and city.

[0022] Specifically, the multiple linear regression model is constructed by using the non-negative least squares regression method based on the improved decision tree grouping in step S4, and the model construction process is as follows:

[0023] In step S41, according to the key influence factors screened in step S3, the sample data of the research area is divided into several subsets with similar water use characteristics by using the decision tree segmentation method.

[0024] In step S42, for each subset, the improved non-negative least squares regression algorithm is used to perform linear fitting with the key influence factors as independent variables and the actual observed water consumption of various categories in the research area as dependent variables.

[0025] The mathematical expression of the improved non-negative least squares regression algorithm is as follows:

[0026] ;

[0027] In the above formula, x i is the input feature vector; y i is the target value; D l is the set of data points in the l th group; β l is the regression coefficient of the leaf node l , and satisfies β l ≥0, w i is the weight of the data point x i , y i , and the calculation formula is as follows:

[0028] ;

[0029] In the above formula, c l is the center of the leaf node l , and dist is the Euclidean distance function;

[0030] In the regression process, the regression coefficient is forced to be non-negative, and the influence factor with a coefficient of 0 is removed. The water consumption fitting equation of different water types is constructed, and the water consumption fitting expression and the model matrix of the fitting goodness R² corresponding to each subset are generated.

[0031] In step S43, the water consumption fitting expressions and the corresponding R² generated by all subsets are summarized in the order of the subsets to construct a complete split-type water consumption prediction model library.

[0032] Step S44, traverse the fitting results of all subsets in the classification water consumption prediction model library, sort according to R² value from high to low, select the classification model with the maximum R² value as the final water consumption prediction model.

[0033] Further, the mathematical expression of the final water consumption prediction model in step S44 is:

[0034] ;

[0035] In the above formula, WU i is the final water consumption prediction value of the i th sample, i f NN is a neural network residual correction function, (X) is the linear prediction value of the i th sample: i

[0036] ;

[0037] In the above formula, X j is the i th input variable; j j is the regression coefficient of the i th input variable, and the regression coefficient is positive; a is the total number of input variables. j n Specifically, the rasterized results are verified with the original water consumption data in step S6 to ensure the accuracy and applicability of the model, and the specific content is as follows:

[0038] From the regional water resource demand spatial distribution map generated in step S5, the sample grid cells in the research area are extracted, and the original observed water consumption data of the corresponding geographical units are obtained; the final water consumption prediction value of each sample grid cell is matched with the corresponding original observed water consumption data one by one to obtain a prediction-observation pair, and the goodness of fit is calculated according to the obtained prediction-observation pair to judge the performance of the classification water consumption prediction model in different research areas and different water consumption types. If the goodness of fit calculation result is lower than the preset threshold, return to step S4 to adjust the parameters of the classification water consumption prediction model or re-group regression, until the model meets the preset threshold requirement.

[0039] Compared with the prior art, the present application has the following beneficial effects:

[0040] Compared with the prior art, the present application has the following beneficial effects:

[0041] ​​​The method integrates multiple influence factors, screens key influence factors, realizes high-resolution spatial distribution expression of regional water resource demand by means of GIS technology by using a linear regression optimization model, solves the problem that the coupling relationship of multiple influence factors is ignored in the traditional method, and can provide important scientific basis for water resource management, land use planning and ecosystem protection. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 A flowchart of the spatialization method of regional water resource consumption of the present application;

[0043] Figure 2 A linear fitting schematic diagram of domestic water consumption and the third industry added value of 11 provinces in the Yangtze River Economic Belt in an embodiment of the present application;

[0044] Figure 3 A water demand spatial distribution map of 11 provinces in the Yangtze River Economic Belt in an embodiment of the present application;

[0045] Figure 4 A fitting curve diagram of the predicted water consumption and the actual water consumption obtained according to the method of the present application. DETAILED DESCRIPTION

[0046] In order to facilitate those skilled in the art to understand and implement the present application, the steps of the method of the present application are described in detail below, and it should be understood that these embodiments are only used to illustrate the present application and not to limit the scope of the present application. In addition, it should be understood that those skilled in the art can make various modifications or changes to the present application after reading the content taught by the present application, and these equivalent forms also fall within the scope defined by the appended claims of the present application.

[0047] Example 1

[0048] As shown in Figure 1 , the present application discloses a spatialization method of regional water resource consumption, comprising the following steps:

[0049] Step S1, constructing a research region database, integrating water consumption data of different water types and grid data of influence factors in the research region;

[0050] The influence factors are determined by consulting relevant academic papers, government reports and industry research materials, and the grid data of the influence factors come from the data center of the Resource Environment Department of the Chinese Academy of Sciences, the National Meteorological Data Center and the World Soil Database;

[0051] Step S2, preprocessing the water consumption data of different water types and the grid data of influence factors in the research region in step S1 to eliminate the dimensional differences between the data;

[0052] Step S3, based on the RStudio platform, linear fitting is performed on the preprocessed grid data of the influence factors and the water consumption data of different water types in the study area, and key influence factors with a fitting goodness R2≥0.7 are screened out;

[0053] Step S4, according to the key influence factors screened out in step S3, a multivariate linear regression model is constructed by using a non-negative least squares regression method based on decision tree grouping improvement;

[0054] Step S5, the multivariate linear regression model parameters optimized by the decision tree grouping are input into the ArcGIS platform, and combined with the grid data of the key influence factors, a 30m resolution regional water resource demand spatial distribution map is generated;

[0055] Step S6, the rasterized prediction results are verified with the original observed water consumption data to ensure the accuracy and applicability of the model.

[0056] Specifically, the research area database in step S1 includes precipitation data, temperature data, population data, first industry added value data, second industry added value data, third industry added value data and land use data.

[0057] Specifically, the screening process of the key influence factors in step S3 is as follows:

[0058] Step S31, according to the water consumption data of different water types and the grid data of the influence factors integrated in step S1, a plurality of candidate driving factors affecting water consumption are preset;

[0059] Step S32, using linear regression algorithm, the preset candidate driving factors are respectively linearly fitted with the actual water consumption data of each province and city and each category, and the fitting goodness R2 of the model is calculated;

[0060] Step S33, according to the calculation result of the fitting goodness R2, the preset candidate driving factors are sorted, and the candidate driving factors with R2≥0.7 are screened out as the key influence factors for analyzing the driving mechanism of water consumption of different categories in each province and city.

[0061] Specifically, in step S4, the non-negative least squares regression method based on decision tree grouping is used to construct the multivariate linear regression model, and the model construction process is as follows:

[0062] Step S41, according to the key influence factors screened out in step S3, the sample data of the study area is divided into a plurality of subsets with similar water consumption characteristics by using the decision tree segmentation method;

[0063] Step S42, for each subset, a modified non-negative least squares regression algorithm is used to linearly fit the key impact factors as independent variables and the actual observed water consumption of each category in the study area as dependent variables;

[0064] The mathematical expression of the modified non-negative least squares regression algorithm is:

[0065] ;

[0066] In the above formula, x i is the input feature vector; y i is the target value; D l is the set of the first l group of data points; β l is the regression coefficient of the leaf node l , and satisfies β l ≥ 0, w i is the weight of the data point x i , y i , and the calculation formula is:

[0067] ;

[0068] In the above formula, c l is the center of the leaf node l , and dist is the Euclidean distance function;

[0069] In the regression process, the regression coefficient is forced to be non-negative, and the impact factor with a coefficient of 0 is removed, a water consumption fitting equation of different water types is constructed, and a water consumption fitting expression and a model matrix of fitting goodness R² corresponding to each subset are generated;

[0070] Step S43, the water consumption fitting expressions and the corresponding R² generated by all subsets are summarized in the order of subsets to construct a complete split-type water consumption prediction model library;

[0071] Step S44, the fitting results of all subsets in the split-type water consumption prediction model library are traversed, and the classification model with the largest R² value is selected as the final water consumption prediction model according to the R² value from high to low.

[0072] Further, the mathematical expression of the final water consumption prediction model in step S44 is:

[0073] ;

[0074] In the above formula, WU i is the final water consumption prediction value of the i-th sample, i f NN (X) is the neural network residual correction function, i is the linear prediction value of the i-th sample:

[0075] ;

[0076] In the above formula, X j is the i-th input variable; j a j is the regression coefficient of the i-th input variable, and the regression coefficient is positive; j is the total number of input variables. n

[0077] Specifically, the gridded result is verified with the original water consumption data in step S6 to ensure the accuracy and applicability of the model, and the specific content is as follows:

[0078] From the regional water resource demand spatial distribution map generated in step S5, the sample grid cells in the study area are extracted, and the original observed water consumption data of the corresponding geographical units are obtained; the final water consumption prediction value of each sample grid cell is matched with the corresponding original observed water consumption data one by one to obtain a prediction-observation pair, and the goodness of fit is calculated according to the obtained prediction-observation pair to judge the performance of the type-specific water consumption prediction model on water consumption prediction in different study areas and different water consumption types. If the goodness of fit calculation result is lower than the preset threshold, return to step S4 to adjust the parameters of the type-specific water consumption prediction model or re-group regression, until the model meets the preset threshold requirement.

[0079] The technical effects of the method of the present application will be further described below with the spatialization of water consumption demand in the Yangtze River Economic Belt as an example.

[0080] In this example, the Yangtze River Economic Belt is taken as the target research area, 2020 is taken as the research time, the water consumption demand of the target research area is solved, and spatialization analysis is carried out. A~K in all charts in this example represent 11 provinces in the Yangtze River Economic Belt. All the required data sources are shown in Table 1.

[0081] Table 1, data sources

[0082] ;

[0083] ​​​​Based on the data sources given in Table 1, raster data of water consumption for three water use types and candidate driving factors from 11 provinces in the Yangtze River Economic Belt were used in the RStudio platform to calculate the pairwise fitting relationships between water consumption for each water use type and each candidate driving factor, and the results were output as a document. Based on the output documents, corresponding fitting curves were plotted, and the R-values ​​were selected based on these curves. 2 An index with a value of ≥0.7 was used as a key influencing factor to construct a water consumption prediction model.

[0084] Specifically, such as Figure 2 The figure shown is a curve showing the fitted relationship between domestic water consumption and the added value of the tertiary industry in 11 provinces along the Yangtze River Economic Belt. Figure 2 It can be seen that the goodness of fit R between domestic water consumption and the added value of the tertiary industry in the 11 provinces of the Yangtze River Economic Belt is high. 2 All values ​​were ≥0.7, indicating that the added value of the tertiary industry was a key influencing factor in analyzing the driving mechanism of domestic water consumption in the 11 provinces of the Yangtze River Economic Belt. Further calculation and screening revealed population size, the added value of the tertiary industry, and rainfall as key influencing factors in the driving mechanism of domestic water consumption in the 11 provinces of the Yangtze River Economic Belt (there are some differences among provinces; this only uses domestic water consumption as an example, but the same method applies to agricultural and industrial water consumption). The corresponding fitting formulas for domestic water consumption in the 11 provinces of the Yangtze River Economic Belt are shown in Table 2 below.

[0085] Table 2. Fitting Formula for Domestic Water Consumption

[0086] ;

[0087] Based on the fitting formulas given in Table 2, the raster calculator in ArcGIS Pro software was used for processing and statistics (Table 2 only shows the fitting formulas for domestic water consumption; fitting formulas for agricultural and domestic water consumption can also be constructed using this method, but only domestic water consumption is used as an example here). Figure 3 As shown, the raster data of the dependent variable in the fitting formula (precipitation data, temperature data, population data, added value of primary industry data, added value of secondary industry data, added value of tertiary industry data and land use data; the data in parentheses are not all used, but the corresponding data are used according to the dependent variable in the equation) to complete the quantification of water consumption at the raster scale in the study area.

[0088] like Figure 3 The bar chart presented here visually illustrates the water consumption composition of 11 provinces (A to K) in the Yangtze River Economic Belt, calculated using the method of this invention. The data includes agricultural water consumption, industrial water consumption, domestic water consumption, and total water consumption. All data are expressed in billions of cubic meters. 3 Unit. From Figure 3It is evident that the spatial pattern of water consumption in the Yangtze River Economic Belt is not uniformly distributed along administrative boundaries as assumed by traditional methods. In actual production and daily life, the spatial pattern of water consumption is often related to the region's economic development level, topography, and other factors. Traditional methods often allocate water consumption evenly to different regions within a province based on water resource bulletin data, ignoring the heterogeneity of water consumption across different regions. Therefore, to accurately demonstrate the heterogeneity of water consumption within each province, this invention selects key influencing factors for different types of water consumption within the province and constructs fitting formulas. It adopts a pixel-by-pixel prediction method to finely display the water demand intensity of sub-regions within each province. The water consumption of each unit is calculated based on the selected influencing factor raster data (precipitation data, temperature data, population data, added value of primary industry data, added value of secondary industry data, added value of tertiary industry data, and land use data). Compared with traditional methods, this approach can more accurately and effectively achieve a high-resolution spatial distribution representation of regional water resource demand.

[0089] Through the Figure 3 The analysis of the data clearly identifies the water consumption characteristics and main consumption areas of each region. For example, in the Yangtze River Economic Belt region, the total water consumption of province B is significantly higher than that of other provinces, indicating that its water consumption is relatively large and the implementation of water-saving measures may need to be given special attention. Figure 3 The different colored columns represent agricultural, industrial, and domestic water use, respectively. The proportion of each type of water use in the total water consumption is clear at a glance, which helps to develop and optimize water-saving strategies in a targeted manner. For example, if agricultural water use accounts for too high a proportion in a certain province, efficient water-saving irrigation technology can be promoted in that province first.

[0090] Furthermore, to verify the accuracy and applicability of the predicted water consumption obtained by the method of the present invention, a comparative analysis was conducted between the predicted water consumption obtained by the method of the present invention and the actual water consumption data published in the "Water Resources Bulletin". For example... Figure 4 The figure shown is a fitting curve obtained by linearly fitting the predicted water consumption obtained by the method of the present invention with the actual water consumption. Figure 4 The fit R shown 2 The value of 0.955 indicates that the method of the present invention can accurately and effectively realize the high-resolution spatial distribution expression of regional water resource demand, which further proves the reliability of the method of the present invention in water resource management.

[0091] In summary, this example demonstrates the process of visualizing and interpreting regional water demand data using domestic water consumption as an example. The interpretation results can provide data support and decision-making basis for subsequent water resource allocation optimization, promotion of water-saving technologies, and policy formulation. Furthermore, the water consumption calculated by the method of this invention is at a raster scale, which facilitates further analysis of water consumption in the target study area by combining it with other types of data in a geospatial analysis system.

[0092] The above descriptions are only the preferred embodiments of the present application, not intended to limit the present application in other forms. Any person skilled in the art can make changes or modifications to the equivalent embodiments with the disclosed technical contents. However, any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution of the present application and according to the technical essence of the present application still belong to the protection scope of the present application.

Claims

1. A method of spatializing regional water consumption, characterized by, The method comprises the following steps: Step S1, constructing a research area database, integrating water consumption data of different water use types and grid data of influence factors in the research area; The influence factors are determined by consulting relevant academic papers, government reports and industry research materials, and the grid data of the influence factors come from the data center of the Resource and Environment Science Research Division of the Chinese Academy of Sciences, the National Meteorological Data Center and the World Soil Database; Step S2, preprocessing the water consumption data of different water use types and the grid data of the influence factors in the research area in step S1 to eliminate the dimensional differences between the data; Step S3, based on the RStudio platform, linearly fitting the preprocessed grid data of the influence factors and the water consumption data of different water use types in the research area, and screening out key influence factors with a goodness of fit R²≥0.7; Step S4, according to the key influence factors screened out in step S3, a multivariate linear regression model is constructed by using a non-negative least squares regression method improved based on decision tree grouping; Step S41, according to the key influence factors screened out in step S3, the sample data of the research area is divided into several subsets with similar water use characteristics by using a decision tree segmentation method; Step S42, for each subset, an improved non-negative least squares regression algorithm is used to linearly fit the key influence factors as independent variables and the actual observed water consumption of all categories in the research area as dependent variables; The mathematical expression of the improved non-negative least squares regression algorithm is: ； In the above formula, x i is the input feature vector; < i>y i is a target value; D l is the first l set of data points; β l is the regression coefficient of the leaf node l , and satisfies β l ≥0, w i is the weight of the data point x i , y i , and the calculation formula is: ； In the above formula, c l is the center of the leaf node l and dist is the Euclidean distance function. During the regression process, the regression coefficients are forced to be non-negative, and the influence factors with coefficients of 0 are removed, and the water consumption fitting equation of different water use types is constructed to generate the water consumption fitting expression and the model matrix of the goodness of fit R² corresponding to each subset; Step S43, the water consumption fitting expressions and the corresponding R² generated by all subsets are summarized in order to construct a complete type-specific water consumption prediction model library; Step S44, the fitting results of all subsets in the type-specific water consumption prediction model library are traversed, and the classification model with the highest R² value is selected as the final water consumption prediction model according to the R² value from high to low; Step S5, the parameters of the multivariate linear regression model optimized by the decision tree grouping are input into the ArcGIS platform, and the grid data of the key influence factors are combined to generate a 30m resolution regional water demand spatial distribution map; Step S6, the rasterized prediction results are verified with the original observed water consumption data to ensure the accuracy and applicability of the model.

2. The method of spatializing regional water consumption according to claim 1, wherein, The research area database in step S1 includes precipitation data, temperature data, population data, first industry added value data, second industry added value data, third industry added value data and land use data.

3. The method of spatializing regional water consumption according to claim 1, wherein, The screening process of the key influence factors in step S3 is as follows: Step S31, according to the integrated water consumption data of different water use types and the grid data of the influence factors in the research area in step S1, a plurality of candidate driving factors affecting water consumption are preset; Step S32, the linear regression algorithm is used to linearly fit the preset candidate driving factors with the actual water consumption data of each province, city and category, and calculate the goodness of fit R² of the model; Step S33, according to the calculation result of goodness of fit R², the candidate driving factors are sorted, and the candidate driving factors with R²≥0.7 are selected as the key influence factors for analyzing the driving mechanism of different types of water use in provinces and cities.

4. The method of claim 1, wherein, The mathematical expression of the final water use prediction model in step S44 is: ； In the above formula, WU i is the final water consumption prediction value for the i th sample, f NN (X) is the neural network residual correction function, is the linear prediction value for the i th sample: ； In the above formula, X j is the first input variable; j is the first input variable; is the regression coefficient for the first input variable, the regression coefficient being positive; j is the regression coefficient for the first input variable, the regression coefficient being positive; n is the total number of input variables.

5. The method of spatializing regional water consumption according to claim 1, wherein, In step S6, the rasterized results are tested with the original water use data to ensure the accuracy and applicability of the model, and the specific contents are as follows: From the spatial distribution map of regional water resource demand generated in step S5, the sample grid cells in the study area are extracted, and the original observed water use data of the corresponding geographical units are obtained; the final water use prediction value of each sample grid cell is matched with its corresponding original observed water use data one by one to obtain the prediction-observation pair, and according to the obtained prediction-observation pair, the goodness of fit is calculated to judge the performance of the type-specific water use prediction model in predicting water use in different study areas and different water use types. If the calculation result of the goodness of fit is lower than the preset threshold, return to step S4 to adjust the parameters of the type-specific water use prediction model or re-group regression, until the model meets the preset threshold requirement.

Images