Multi-scale correlation analysis method for spatial variability analysis of soil organic carbon and prediction method of soil organic carbon
By employing multi-scale correlation analysis, combined with Spearman correlation analysis and the multi-scale geographically weighted regression model MGWR, the problem of identifying the complex relationship between soil organic carbon (SOC) and environmental variables was solved. This enabled accurate analysis and prediction of the spatial distribution of SOC, thereby enhancing the scientific basis for soil management and environmental protection.
Patent Information
- Application Number
- CN202411119854.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-15
AI Technical Summary
Existing technologies struggle to systematically reveal the complex relationships between soil organic carbon (SOC) and environmental variables at different spatial scales, particularly environmental variables that are not significantly correlated globally but are strongly correlated in local regions.
Using multi-scale correlation analysis, combined with Spearman correlation analysis, geographic weighted correlation analysis, and the multi-scale geographic weighted regression model MGWR, we identified and captured multicollinearity of environmental variables through global and local correlation analysis, optimized the selection of environmental variables, and performed fitting and prediction of soil organic carbon (SOC).
This study systematically reveals the complex relationship between soil organic carbon (SOC) and environmental variables at different spatial scales, improves the generalization ability of soil organic carbon (SOC) mapping models, accurately captures the influence of environmental variables in local areas, and enhances the understanding and prediction accuracy of the spatial distribution patterns of soil organic carbon (SOC).
Smart Images

Figure CN119001060A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon and a method for predicting soil organic carbon, and belongs to the technical field of soil science. Background Art
[0002] Soil organic carbon (SOC) is one of the largest carbon pools in terrestrial ecosystems and has an important impact on the global carbon cycle and climate change. Therefore, accurately grasping the spatial distribution characteristics of soil organic carbon (SOC) is of great significance for agricultural management, ecological protection and climate change research. However, the spatial distribution of soil organic carbon (SOC) is complex and heterogeneous, and it shows significant spatial variability at different scales due to the diversity of soil formation processes, geographical conditions and management practices. This variability makes the prediction and mapping of soil organic carbon (SOC) face great challenges, and more advanced analytical methods are urgently needed to reveal the complex relationship between soil organic carbon (SOC) and environmental variables.
[0003] Existing studies have mostly used global analysis methods, such as Pearson correlation analysis and traditional regression models, to analyze the spatial variability of soil organic carbon (SOC) and its relationship with environmental variables. Although these methods reveal the association between soil organic carbon (SOC) and environmental variables to a certain extent, they often ignore spatial heterogeneity. They assume that the effects of environmental variables on soil organic carbon (SOC) are uniform across the study area and fail to capture significant correlations or influencing factors that may exist in local areas, which may mask the complex dynamics of soil organic carbon (SOC) at small scales or in specific areas. For example, factors such as topographic changes, microclimate differences, and land management practices may have significant effects on soil organic carbon (SOC) in local areas, but these effects are often ignored in global scale analyses. Therefore, it is necessary to develop a multiscale analysis framework that combines global and local analyses to better reveal the role of environmental variables in explaining soil organic carbon (SOC) variation and provide an effective variable selection method for soil organic carbon (SOC) mapping.
[0004] Therefore, existing studies have failed to fully reveal the spatial distribution patterns and driving mechanisms of soil organic carbon (SOC), limiting our in-depth understanding of the soil carbon cycle process.
[0005] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgement or any form of suggestion that the information constitutes the prior art already known to ordinary technicians in this field. Summary of the invention
[0006] The technical problem to be solved by the present invention is: how to systematically reveal the complex relationship between soil organic carbon (SOC) and environmental variables at different spatial scales, and at the same time effectively identify and capture environmental variables that may not show significant correlation in the entire domain but have strong correlation in local areas or under specific conditions.
[0007] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions:
[0008] In a first aspect, the present invention provides a multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon, comprising:
[0009] Collect multiple soil samples in the study area and determine the soil organic carbon (SOC) content in the soil samples;
[0010] Acquire multi-source data in the study area and extract environmental variables from the multi-source data; the multi-source data include basic geographic data, topographic data, and remote sensing data; and the environmental variables include topographic factors, location conditions, landscape patterns, and remote sensing spectral indices;
[0011] Determine the correlation coefficient and significance between soil organic carbon (SOC) content and each environmental variable, analyze the global correlation between soil organic carbon (SOC) content and each environmental variable, and obtain globally correlated environmental variables and globally uncorrelated environmental variables;
[0012] The geographically weighted correlation coefficient was determined based on the global relevant environmental variables, the global unrelated environmental variables and the geographical weight of each soil sample, and the local correlation between the soil organic carbon (SOC) content and each environmental variable was analyzed to obtain the local correlation environmental variables.
[0013] According to the variance inflation factor and tolerance of the local correlation environmental variables, the local correlation environmental variables are judged for multicollinearity, and the local correlation environmental variables with multicollinearity are deleted to obtain environmental variables without multicollinearity;
[0014] Based on the spatial weight of each soil sample point and the effective bandwidth of environmental variables without multicollinearity, the multiscale geographically weighted regression model MGWR was used to fit the soil organic carbon (SOC) content with environmental variables without multicollinearity to obtain the local standardized regression coefficients of environmental variables without multicollinearity.
[0015] Environmental variables with significant local effects on soil organic carbon (SOC) were identified based on the local standardized regression coefficients of environmental variables without multicollinearity.
[0016] Based on the globally correlated environmental variables, globally uncorrelated environmental variables, locally correlated environmental variables and environmental variables with significant local standardized regression coefficients, a comprehensive analysis of the impact scales of environmental variables was conducted to analyze the spatial heterogeneity of soil organic carbon (SOC) content.
[0017] Furthermore, the correlation coefficient and significance between the soil organic carbon (SOC) content and each environmental variable are determined, the global correlation between the soil organic carbon (SOC) content and each environmental variable is analyzed, and the method for obtaining the globally correlated environmental variables and the globally uncorrelated environmental variables includes:
[0018] The Spearman correlation analysis method was used to calculate the correlation coefficient between soil organic carbon (SOC) content and each environmental variable;
[0019] According to the correlation coefficient between soil organic carbon (SOC) content and each environmental variable, a statistical significance test was performed to determine the significance between soil organic carbon (SOC) content and each environmental variable;
[0020] According to the correlation coefficient and significance between the soil organic carbon (SOC) content and each environmental variable, the global correlation between the soil organic carbon (SOC) content and each environmental variable was analyzed to obtain globally correlated environmental variables and globally uncorrelated environmental variables.
[0021] Furthermore, a method for determining a geographically weighted correlation coefficient based on the global relevant environmental variables and the global unrelated environmental variables and the geographical weight of each soil sample, analyzing the local correlation between the soil organic carbon (SOC) content and each environmental variable, and obtaining the local correlation environmental variables includes:
[0022] The geographic weight of each soil sample was determined using a Gaussian kernel function;
[0023] According to the geographical weight of each soil sample, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable were calculated;
[0024] The geographically weighted correlation coefficients at different spatial locations were calculated based on the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable;
[0025] According to the globally relevant environmental variables, the globally irrelevant environmental variables and the geographically weighted correlation coefficient, the local correlation between the soil organic carbon (SOC) content and each environmental variable is analyzed to obtain the locally relevant environmental variables.
[0026] Furthermore, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable were calculated;
[0027] Among them, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable is expressed as:
[0028] ;
[0029] In the formula, is the geographically weighted covariance value between the soil organic carbon (SOC) content at different spatial locations and each environmental variable; and are the levels of environmental variables at location i and the levels of soil organic carbon (SOC) in soil samples at location i, respectively; and are the levels of environmental variables at location j and the levels of soil organic carbon (SOC) in soil samples at location j, respectively; and are the geographically weighted average ranks of environmental variables at location i and the geographically weighted average ranks of soil organic carbon (SOC) in soil samples at location i, respectively; For the The sample pair at position The geographic weight of the location samples, is the total number of soil samples.
[0030] The geographically weighted standard deviation of each environmental variable at different spatial locations is expressed as:
[0031] ;
[0032] The geographically weighted standard deviation of soil organic carbon (SOC) content is expressed as:
[0033] ;
[0034] In the formula, and They are the geographically weighted standard deviation values of each environmental variable and the geographically weighted standard deviation values of soil organic carbon (SOC) content at different spatial locations.
[0035] Furthermore, the geographically weighted correlation coefficients at different spatial locations were calculated based on the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable. The geographically weighted correlation coefficients at different spatial locations were expressed as:
[0036] ;
[0037] In the formula, is the geographically weighted correlation coefficient value at different spatial locations.
[0038] Furthermore, according to the variance inflation factor and tolerance of the local correlation environment variables, the local correlation environment variables are judged for multicollinearity, and the local correlation environment variables with multicollinearity are deleted to obtain the method of obtaining the environment variables without multicollinearity, including:
[0039] The variance inflation factor and tolerance of local correlation environmental variables were determined by regression analysis. When the variance inflation factor was greater than 5 and the tolerance was less than 0.2, it indicated that there was multicollinearity in the local correlation environmental variables. The local correlation environmental variables with multicollinearity were deleted to obtain environmental variables without multicollinearity.
[0040] Furthermore, based on the spatial weight of each soil sample and the effective bandwidth of the environmental variables without multicollinearity, the multiscale geographically weighted regression model MGWR is used to fit the soil organic carbon SOC content with the environmental variables without multicollinearity to obtain the local standardized regression coefficient of the environmental variables without multicollinearity, including:
[0041] Use Gaussian kernel function to weight each soil sample spatially to obtain the spatial weight of each soil sample;
[0042] The effective bandwidth of environmental variables without multicollinearity was determined using the Akaike information criterion (AIC) or the Bayesian information criterion (BIC).
[0043] Based on the spatial weight and the effective bandwidth, the multi-scale geographically weighted regression model MGWR is used to fit the soil organic carbon SOC content and the environmental variables without multicollinearity. According to the fitting results of the multi-scale geographically weighted regression model MGWR, the local standardized regression coefficients of the environmental variables without multicollinearity are calculated.
[0044] Furthermore, the multi-scale geographically weighted regression model MGWR is expressed as:
[0045] ;
[0046] In the formula, For soil samples dependent variable; For soil samples No. environment variables; For the Regression equations for environmental variables Differentiation bandwidth of environmental variables; is the total number of variables; For soil samples The spatial location coordinates; The intercept term varies with the spatial position change; For soil samples The spatial position coordinates The The regression coefficients of the variables; is a random error term.
[0047] Furthermore, based on the fitting results of the multi-scale geographically weighted regression model MGWR, a method for calculating the local standardized regression coefficients of environmental variables without multicollinearity includes:
[0048] Local standardized regression coefficients of environmental variables without multicollinearity were estimated by weighted least squares ;
[0049] The local standardized regression coefficients of environmental variables without multicollinearity were tested for statistical significance and spatially visualized to identify environmental variables with significant local effects on soil organic carbon (SOC).
[0050] In a second aspect, the present invention further provides a method for predicting soil organic carbon, which selects environmental variables based on the analysis results of the multi-scale association analysis method for analyzing the spatial variability of soil organic carbon described in the first aspect to predict soil organic carbon.
[0051] Compared with the prior art, the present invention has the following beneficial effects:
[0052] The present invention proposes an innovative multi-scale association analysis method for analyzing the spatial variability of soil organic carbon (SOC). This method integrates Spearman correlation analysis, geographically weighted correlation analysis and the multi-scale geographically weighted regression model MGWR, aiming to capture the distribution characteristics of soil organic carbon (SOC) at different spatial scales, and optimize the selection of environmental variables to enhance the generalization ability of the soil organic carbon (SOC) mapping model. Through Spearman correlation analysis, the present invention first reveals the global relationship between environmental variables and SOC. Secondly, through geographically weighted correlation analysis, the local correlation between soil organic carbon (SOC) and environmental variables at different spatial locations is deeply explored, revealing the unique influence of environmental variables in different regions. Subsequently, the introduction of the multi-scale geographically weighted regression model MGWR fully considers the spatial heterogeneity and the differences in the influence of different environmental variables at multiple scales, effectively solving the problem of insufficient explanation of local spatial relationships by traditional methods.
[0053] The present invention systematically reveals the complex relationship between soil organic carbon (SOC) and environmental variables at different spatial scales, while effectively identifying and capturing environmental variables that may not show significant correlation on a global scale but have strong correlation in local areas or under specific conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 It is a schematic diagram of the overall process of a multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon provided by an embodiment of the present invention;
[0055] Figure 2 is a schematic diagram of soil sample point distribution in a study area provided by an embodiment of the present invention;
[0056] Figure 3 is a schematic diagram of some environmental variables in the study area provided by an embodiment of the present invention;
[0057] Figure 4 is a schematic diagram of geographically weighted local correlation formed during the analysis process provided by an embodiment of the present invention;
[0058] Figure 5 It is a schematic diagram of the significant local standardized regression coefficients of the multi-scale geographically weighted regression model MGWR formed in the analysis process provided by the embodiment of the present invention. DETAILED DESCRIPTION
[0059] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.
[0060] According to the correlation coefficient and significance between the soil organic carbon SOC content and each environmental variable, the global correlation between the soil organic carbon SOC content and each environmental variable is analyzed to obtain the globally correlated environmental variables and the globally uncorrelated environmental variables; and according to the globally correlated environmental variables and the globally uncorrelated environmental variables and the geographically weighted correlation coefficient, the local correlation between the soil organic carbon SOC content and each environmental variable is analyzed to obtain the locally correlated environmental variables, etc., which are all prior arts, and the undisclosed parts of this specification are all known arts available to persons skilled in the art.
[0061] Example 1
[0062] like Figure 1 As shown, this embodiment introduces a multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon, which is characterized by comprising:
[0063] Multiple soil samples were collected in the study area, and the soil organic carbon (SOC) content in the soil samples was determined through laboratory analysis, providing a data basis for the spatial distribution and variability of soil organic carbon (SOC) in the study area.
[0064] Acquire multi-source data in the study area and extract environmental variables from the multi-source data; the multi-source data include basic geographic data, topographic data, and remote sensing data; the environmental variables include topographic factors, location conditions, landscape pattern, and remote sensing spectral index; collect data on various environmental factors that may affect the soil organic carbon (SOC) content to comprehensively consider the factors affecting the distribution of soil organic carbon (SOC), which will help reveal the causes of the spatial heterogeneity of soil organic carbon (SOC).
[0065] The correlation coefficient and significance between soil organic carbon (SOC) content and each environmental variable were determined, and the global correlation between soil organic carbon (SOC) content and each environmental variable was analyzed to obtain globally correlated environmental variables and globally uncorrelated environmental variables.
[0066] The global correlation between soil organic carbon (SOC) content and environmental variables was preliminarily analyzed to identify environmental factors that may have significant effects.
[0067] The geographically weighted correlation coefficient was determined based on the globally correlated environmental variables, the globally uncorrelated environmental variables and the geographical weight of each soil sample. The local correlation between the soil organic carbon (SOC) content and each environmental variable was analyzed to obtain the local correlation environmental variables, which can reveal the spatial variability of the relationship between the soil organic carbon (SOC) content and environmental variables and help to more accurately understand the causes of the spatial heterogeneity of soil organic carbon (SOC).
[0068] According to the variance inflation factor and tolerance of the local correlation environmental variables, the local correlation environmental variables are judged for multicollinearity, and the local correlation environmental variables with multicollinearity are deleted to obtain environmental variables without multicollinearity. The environmental variables with multicollinearity are deleted, which can avoid the influence of multicollinearity on the results and make the analysis results more reliable.
[0069] Based on the spatial weight of each soil sample point and the effective bandwidth of environmental variables without multicollinearity, the multi-scale geographically weighted regression model MGWR was used to fit the soil organic carbon SOC content with environmental variables without multicollinearity to obtain the local standardized regression coefficients of environmental variables without multicollinearity; based on the local standardized regression coefficients of environmental variables without multicollinearity, the environmental variables with significant local effects on soil organic carbon SOC were determined.
[0070] By considering the spatial heterogeneity of different environmental variables, the multi-scale geographically weighted regression model MGWR is used to fit the relationship between soil organic carbon (SOC) content and environmental variables, which can more accurately reflect the multi-scale impact of environmental variables on soil organic carbon (SOC) and improve the explanatory power of the spatial heterogeneity of soil organic carbon (SOC).
[0071] Based on the globally correlated environmental variables, globally uncorrelated environmental variables, locally correlated environmental variables and environmental variables with significant local standardized regression coefficients, a comprehensive analysis of the impact scales of environmental variables was conducted to analyze the spatial heterogeneity of soil organic carbon (SOC) content.
[0072] Identifying environmental factors that have significant local effects on soil organic carbon (SOC) content provides key information for understanding the causes of spatial heterogeneity in soil organic carbon (SOC), which helps to formulate targeted soil management measures and improve soil quality. Combining the results of global and local correlation analysis and the multi-scale geographically weighted regression model MGWR, the spatial heterogeneity of soil organic carbon (SOC) content is fully analyzed. This provides a scientific basis for the rational use of soil resources and environmental protection, and helps to achieve sustainable land use and management.
[0073] Example 2
[0074] Based on the same inventive concept as Example 1, this embodiment provides a multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon, including the following steps:
[0075] Step 1: Collect multiple soil samples in the study area and determine the soil organic carbon (SOC) content in the soil samples.
[0076] The expansion of human activities in plateau areas has led to the loss of ecological land and aggravated habitat fragmentation. Qilu Lake is one of the nine plateau lakes in Yunnan Province, China, with a basin area of 354 square kilometers. The basin is located in the Yunnan-Guizhou Plateau. Due to market-oriented and industrialized agricultural management, Qilu Lake is surrounded by highly fragmented farmland and has a large landscape heterogeneity, making it an interesting case study area.
[0077] The soil complexity characterized by land management heterogeneity makes the soil organic carbon (SOC) and its relationship with soil formation factors in the region present a complex distribution pattern.
[0078] Therefore, a total of 216 surface soil samples were collected in the study area from November 16 to 20, 2018, and the GPS location and surrounding environmental information of each soil sample were recorded. In the above process, the local landscape pattern was taken into consideration, and the density of sampling points was relatively high in areas with more fragmented plots. The collected soil samples were placed in sealed bags, labeled, and then quickly transported to the laboratory for analysis. The specific distribution of multiple soil sampling points in the study area is shown in Figure 2. Figure 2 shown.
[0079] Step 2: Obtain multi-source data in the study area and extract environmental variables from the multi-source data. The multi-source data includes basic geographic data, topographic data, and remote sensing data. The environmental variables include topographic factors, location conditions, landscape patterns, and remote sensing spectral indices.
[0080] After analysis and demonstration, it is believed that the factors affecting the soil organic carbon (SOC) in the study area should be considered from the aspects of terrain conditions, location conditions, landscape pattern, vegetation, and water conditions. Based on this, the present invention collected and sorted out different types of environmental variables in the Qilu Lake Basin as shown in Table 1. The spatial distribution of some environmental variables is shown in Table 1. Figure 3 shown.
[0081]
[0082] Step 3: Determine the correlation coefficient and significance between the soil organic carbon (SOC) content and each environmental variable, analyze the global correlation between the soil organic carbon (SOC) content and each environmental variable, and obtain globally correlated environmental variables and globally uncorrelated environmental variables, including:
[0083] The Spearman correlation analysis method was used to calculate the correlation coefficient between soil organic carbon (SOC) content and each environmental variable. , with a value range of -1 to 1; through Spearman correlation analysis, we can fully understand the global relationship between soil organic carbon (SOC) and each environmental variable, laying the foundation for in-depth research on the spatial variability of SOC and its influencing factors.
[0084] Among them, the Spearman correlation analysis method is expressed as:
[0085] ;
[0086] In the formula, is the difference between the ranking of the environmental variables and soil organic carbon SOC of the 𝑖th sample, is the total number of soil samples.
[0087] According to the correlation coefficient between soil organic carbon (SOC) content and each environmental variable, a statistical significance test was performed to determine the significance between soil organic carbon (SOC) content and each environmental variable;
[0088] The p value is usually used for statistical significance test. When the p value is less than the set significance level, such as 0.05, the correlation coefficient is considered to be statistically significant, as shown in Table 2.
[0089]
[0090] According to the Spearman correlation coefficient and significance between the soil organic carbon (SOC) content and each environmental variable, the global correlation between the soil organic carbon (SOC) content and each environmental variable was analyzed to obtain globally correlated environmental variables and globally uncorrelated environmental variables.
[0091] Step 4: Determine the geographically weighted correlation coefficient based on the global relevant environmental variables and the global unrelated environmental variables and the geographical weight of each soil sample, analyze the local correlation between the soil organic carbon (SOC) content and each environmental variable, and obtain the local correlation environmental variables, including:
[0092] The geographic weight of each soil sample was determined using a Gaussian kernel function;
[0093] According to the geographical weight of each soil sample, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable were calculated;
[0094] The geographically weighted correlation coefficients at different spatial locations were calculated based on the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable;
[0095] According to the global relevant environmental variables, the global unrelated environmental variables and the geographical weighted correlation coefficient, the local correlation between the soil organic carbon SOC content and each environmental variable was analyzed to obtain the local correlation environmental variables, and the local correlation environmental variables were visualized in space to reveal the local correlation between the soil organic carbon SOC content and each environmental variable at different spatial locations. The geographical weighted correlation results of some environmental variables and soil organic carbon SOC are shown in the figure. Figure 4 shown.
[0096] Among them, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable is expressed as:
[0097] ;
[0098] In the formula, is the geographically weighted covariance value between the soil organic carbon (SOC) content at different spatial locations and each environmental variable; and are the levels of environmental variables at location i and the levels of soil organic carbon (SOC) in soil samples at location i, respectively; and are the levels of environmental variables at location j and the levels of soil organic carbon (SOC) in soil samples at location j, respectively; and are the geographically weighted average ranks of environmental variables at location i and the geographically weighted average ranks of soil organic carbon (SOC) in soil samples at location i, respectively; ; n is the total number of soil samples.
[0099] The geographically weighted standard deviation of each environmental variable at different spatial locations is expressed as:
[0100] ;
[0101] The geographically weighted standard deviation of soil organic carbon (SOC) content is expressed as:
[0102] ;
[0103] In the formula, and They are the geographically weighted standard deviation values of each environmental variable and the geographically weighted standard deviation values of soil organic carbon (SOC) content at different spatial locations.
[0104] Among them, the geographically weighted average grade of the ith environmental variable and the geographically weighted average grade of soil organic carbon (SOC) in the ith soil sample are expressed as:
[0105] ;
[0106] ;
[0107] The geographically weighted correlation coefficient at different spatial locations is expressed as:
[0108] ;
[0109] In the formula, is the geographically weighted correlation coefficient value at different spatial locations.
[0110] Step 5: According to the variance inflation factor and tolerance of the local correlation environment variables, the local correlation environment variables are judged for multicollinearity, and the local correlation environment variables with multicollinearity are deleted to obtain the environment variables without multicollinearity, including:
[0111] Before conducting multiscale geographically weighted regression modeling, it is necessary to conduct multicollinearity diagnosis of variables to ensure the stability and reliability of the multiscale geographically weighted regression model.
[0112] The variance inflation factor and tolerance of the local correlation environmental variables were determined by regression analysis. When the variance inflation factor was greater than 5 and the tolerance was less than 0.2, it indicated that there was multicollinearity in the local correlation environmental variables. The local correlation environmental variables with multicollinearity were deleted, and the variance inflation factor VIF and tolerance Tolerance of each local correlation environmental variable were calculated again. When the variance inflation factor was less than 5 and the tolerance was greater than 0.2, the environmental variables without multicollinearity were obtained, and the multi-scale geographically weighted regression model MGWR was then used for modeling. The multicollinearity diagnosis results are shown in Table 3.
[0113]
[0114] Among them, the tolerance and variance inflation factor of the local correlation environment variable are expressed as:
[0115] ;
[0116] ;
[0117] In the formula, is the variance inflation factor of the local correlation environment variable, Tolerance is the tolerance of the local correlation environment variable, R 2 is the coefficient of determination of the regression analysis method.
[0118] Among them, the coefficient of determination of the regression analysis method It is expressed as:
[0119] ;
[0120] In the formula, is the actual measured value of soil organic carbon (SOC) in the soil sample at the i-th position, is the predicted value of soil organic carbon (SOC) in the soil sample at the i-th position, is the mean of the actual measured values of soil organic carbon SOC, is the total number of soil samples.
[0121] Step 6: Based on the spatial weight of each soil sample and the effective bandwidth of the environmental variables without multicollinearity, the multiscale geographically weighted regression model MGWR is used to fit the soil organic carbon SOC content with the environmental variables without multicollinearity to obtain the local standardized regression coefficients of the environmental variables without multicollinearity, including:
[0122] Use Gaussian kernel function to weight each soil sample spatially to obtain the spatial weight of each soil sample;
[0123] The effective bandwidth of each environmental variable is determined using the Akaike information criterion (AIC) or the Bayesian information criterion (BIC);
[0124] Based on the spatial weight and the effective bandwidth, the multi-scale geographically weighted regression model MGWR was used to fit the soil organic carbon SOC content with each environmental variable without multicollinearity, and the local standardized regression coefficients of the environmental variables without multicollinearity were obtained, revealing the complex relationship between soil organic carbon SOC and environmental variables at different spatial scales. The standardized regression coefficients and bandwidth statistics of each environmental variable are shown in Table 4.
[0125]
[0126] The spatial weight of each soil sample is calculated by using Gaussian kernel function or biquadratic kernel function, reflecting the spatial distribution of the soil samples. Soil samples from locations The degree of influence of the regression coefficient estimate for each location.
[0127] Taking the Gaussian kernel function as an example, the specific formula of the spatial weight of each soil sample is expressed as:
[0128] ;
[0129] In the formula, , It is Location and The distance between the locations, It is The spatial bandwidth of the environmental variables, .
[0130] The multi-scale geographically weighted regression model MGWR is expressed as:
[0131] ;
[0132] In the formula, For soil samples dependent variable; For soil samples No. environment variables; For the Regression equations for environmental variables Differentiation bandwidth of environmental variables; is the total number of variables; For soil samples The spatial location coordinates; The intercept term varies with the spatial position change; For soil samples The spatial position coordinates The The regression coefficients of the variables; is a random error term.
[0133] Step 7: Based on the fitting results of the multiscale geographically weighted regression model MGWR, calculate the local standardized regression coefficients of environmental variables without multicollinearity.
[0134] Local standardized regression coefficients of environmental variables without multicollinearity were estimated by weighted least squares ;
[0135] The local standardized regression coefficients of environmental variables without multicollinearity were tested for statistical significance and spatially visualized to identify environmental variables with significant local effects on soil organic carbon (SOC).
[0136] Among them, the weighted sum of squares residuals WSSR is minimized to estimate the The local standardized regression coefficients of the environmental variables at each location are expressed as:
[0137] ;
[0138] In the formula, To minimize the weighted residual sum of squares; For location SOC content of soil samples at It is The sample pair at position The geographic weight of the location samples depends on the spatial bandwidth and the distance decay function, It is The spatial bandwidth of the independent variable; Soil sample point No. independent variables; For the The regression equation for the independent variables has the bandwidth of differentiation of the independent variables; is the total number of variables; Soil sample point The spatial location coordinates; is the intercept term, which varies with the spatial position change; Soil sample point The spatial position coordinates The The regression coefficients of the variables; is the total number of soil samples.
[0139] Among them, the residual sum of squares RSS is expressed as:
[0140] ;
[0141] Among them, the Akaike information criterion AIC is expressed as:
[0142] ;
[0143] In the formula, Represents the number of all regression coefficients and intercept terms in the multi-scale geographically weighted regression model MGWR, Represents the logarithm with constant e as base.
[0144] The significance test and spatial visualization of the local standardized regression coefficients of each environmental variable without multicollinearity were performed to interpret the results of the multiscale geographically weighted regression model MGWR, identify the environmental variables with significant local effects on soil organic carbon (SOC), analyze their spatial heterogeneity, and deeply analyze the multiscale effects of environmental variables on soil organic carbon (SOC), such as Figure 5 shown.
[0145] Step 8: Based on the globally correlated environmental variables, globally uncorrelated environmental variables, locally correlated environmental variables, and environmental variables with significant local standardized regression coefficients, a comprehensive analysis of the impact scale of environmental variables was conducted to analyze the spatial heterogeneity of soil organic carbon (SOC) content.
[0146] First, according to the results of global correlation analysis as shown in Table 2, soil moisture content (SMC), terrain wetness index (TWI), distance to road (Road_D), enhanced vegetation index (EVI), brightness index (BI), normalized shortwave infrared difference bare soil moisture index (NSDSI1), brightness (Brightness), greenness (Greenness), and humidity (Wetness) are positively correlated with soil organic carbon SOC, and elevation (DEM), slope (Slope), distance to river (River_D), landscape abundance density (PRD), and color index (CI) are negatively correlated with soil organic carbon SOC, while the global correlation between soil total porosity (SP), spread index (CONTAG), and maximum patch area ratio (LPI) and soil organic carbon SOC is low and not significant.
[0147] Secondly, in the local correlation analysis, it was found that the geographically weighted correlation coefficients between each environmental variable and soil organic carbon SOC had spatial differences in positive and negative and size, reflecting that the variables had different local effects on soil organic carbon SOC. In addition, the three variables with no significant global correlation with soil organic carbon SOC, soil total porosity (SP), spread index (CONTAG), and maximum patch area ratio (LPI), all had geographically weighted local correlations with soil organic carbon SOC. Figure 4 As shown in the figure, even variables that are not significant in the global analysis can still play an important role in the local analysis, revealing the complex spatial heterogeneity between soil organic carbon (SOC) and environmental variables.
[0148] In addition, the standardized regression coefficients of the multiscale geographically weighted regression model MGWR further reflect the overall influence of each environmental variable on soil organic carbon SOC and the local explanatory power as shown in Tables 4 and Figure 5 As shown. The multiscale geographically weighted regression model MGWR analysis mainly revealed the local scale effects of soil water content (SMC), soil total porosity (SP), distance to river (River_D), distance to road (Road_D), landscape abundance density (PRD), spread index (CONTAG), proportion of largest patch area (LPI) and enhanced vegetation index (EVI) on soil organic carbon (SOC), with geographical differences. Although variables such as soil water content (SMC), distance to river (River_D), distance to road (Road_D), landscape abundance density (PRD) and enhanced vegetation index (EVI) showed significant global correlations with soil organic carbon (SOC), their effects were mainly local, probably due to the aggregation of local effects. However, soil total porosity (SP), contagion index (CONTAG), and maximum patch area ratio (LPI) also have local effects, but they are easily considered to have no effect on soil organic carbon SOC due to their insignificant correlation with soil organic carbon SOC in global correlation analysis, which leads to them being considered invalid variables and being omitted when predicting soil organic carbon SOC. In contrast, variables such as slope, terrain wetness index (TWI), color index (CI), normalized shortwave infrared difference bare soil moisture index (NSDSI1), brightness, and greenness, which are significantly correlated with soil organic carbon SOC in global correlation analysis, do not show significant effects in the MGWR model. Their local scale effects may be masked or balanced by other more locally significant variables.
[0149] On a global scale, global variables can be used to construct a preliminary soil organic carbon SOC prediction model, providing an overview of the soil organic carbon SOC distribution trend in the entire region. However, in order to carry out detailed management and precise prediction applications on a local scale, variables that are important at the local level must be further incorporated to achieve more accurate prediction results and effective management strategies. Therefore, the example results have verified that the present invention can reveal the relationship between environmental variables and soil organic carbon SOC on a global scale, accurately capture the impact of overall environmental variables on soil organic carbon SOC; and reflect more subtle changes in local areas, and can identify and explain the local impact of environmental variables at different spatial locations, thereby enhancing the understanding of the spatial distribution law of soil organic carbon SOC.
[0150] Example 3
[0151] Based on the same inventive concept as other embodiments, the present invention also provides a method for predicting soil organic carbon, which selects environmental variables according to the analysis results of the multi-scale association analysis method for analyzing the spatial variability of soil organic carbon described in Example 1 or 2 to predict soil organic carbon.
[0152] In summary, the present invention proposes a multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon (SOC), which is based on Spearman correlation analysis, geographically weighted correlation analysis and a multi-scale geographically weighted regression model. It captures the distribution characteristics of soil organic carbon (SOC) at different spatial scales and optimizes the selection of environmental variables to improve the generalization ability of the soil organic carbon (SOC) mapping model.
[0153] The present invention first reveals the global relationship between environmental variables and SOC through Spearman correlation analysis; secondly, by introducing geographically weighted correlation, the local correlation between soil organic carbon SOC and environmental variables at different spatial locations is explored, revealing the local impact of environmental variables in different regions; then, by introducing the multi-scale geographically weighted regression model MGWR, the heterogeneity of spatial processes and the degree of influence of different environmental variables at different spatial scales are fully considered, thereby solving the problem of insufficient explanation of local spatial relationships in the prior art.
[0154] After adopting the method of the present invention, firstly, the relationship between environmental variables and soil organic carbon SOC is revealed globally, accurately capturing the impact of overall environmental variables on soil organic carbon SOC; secondly, more subtle changes are reflected in local areas, which can identify and explain the local impact of environmental variables at different spatial locations, and enhance the understanding of the spatial distribution law of soil organic carbon SOC. Finally, the results of the soil organic carbon SOC variability analysis based on the method of the present invention can provide a scientific basis and decision-making support for soil management, agricultural planning and ecological environmental protection. By revealing the multi-scale relationship of the impact of environmental variables on soil organic carbon SOC, the present invention can not only improve the accuracy and reliability of spatial prediction of soil organic carbon SOC, but also has certain universality and extensibility, and can provide effective tools and methods for soil organic carbon SOC management and research in different regions.
[0155] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0156] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0157] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0158] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0159] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.
Claims
1. A multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon, characterized in that: include: Collect multiple soil samples in the study area and determine the soil organic carbon (SOC) content in the soil samples; Acquire multi-source data in the study area and extract environmental variables from the multi-source data; the multi-source data include basic geographic data, topographic data, and remote sensing data; and the environmental variables include topographic factors, location conditions, landscape patterns, and remote sensing spectral indices; Determine the correlation coefficient and significance between soil organic carbon (SOC) content and each environmental variable, analyze the global correlation between soil organic carbon (SOC) content and each environmental variable, and obtain globally correlated environmental variables and globally uncorrelated environmental variables; The geographically weighted correlation coefficient was determined based on the global relevant environmental variables, the global unrelated environmental variables and the geographical weight of each soil sample, and the local correlation between the soil organic carbon (SOC) content and each environmental variable was analyzed to obtain the local correlation environmental variables. According to the variance inflation factor and tolerance of the local correlation environmental variables, the local correlation environmental variables are judged for multicollinearity, and the local correlation environmental variables with multicollinearity are deleted to obtain environmental variables without multicollinearity; Based on the spatial weight of each soil sample point and the effective bandwidth of environmental variables without multicollinearity, the multiscale geographically weighted regression model MGWR was used to fit the soil organic carbon (SOC) content with environmental variables without multicollinearity to obtain the local standardized regression coefficients of environmental variables without multicollinearity. Environmental variables with significant local effects on soil organic carbon (SOC) were identified based on the local standardized regression coefficients of environmental variables without multicollinearity. Based on the globally correlated environmental variables, globally uncorrelated environmental variables, locally correlated environmental variables and environmental variables with significant local standardized regression coefficients, a comprehensive analysis of the impact scales of environmental variables was conducted to analyze the spatial heterogeneity of soil organic carbon (SOC) content.
2. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 1 is characterized in that: The method of determining the correlation coefficient and significance between the soil organic carbon (SOC) content and each environmental variable, analyzing the global correlation between the soil organic carbon (SOC) content and each environmental variable, and obtaining the globally correlated environmental variables and the globally uncorrelated environmental variables includes: The Spearman correlation analysis method was used to calculate the correlation coefficient between soil organic carbon (SOC) content and each environmental variable; According to the correlation coefficient between soil organic carbon (SOC) content and each environmental variable, a statistical significance test was performed to determine the significance between soil organic carbon (SOC) content and each environmental variable; According to the correlation coefficient and significance between the soil organic carbon (SOC) content and each environmental variable, the global correlation between the soil organic carbon (SOC) content and each environmental variable was analyzed to obtain globally correlated environmental variables and globally uncorrelated environmental variables.
3. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 1 is characterized in that: The method of determining the geographically weighted correlation coefficient according to the global relevant environmental variables and the global unrelated environmental variables and the geographical weight of each soil sample, analyzing the local correlation between the soil organic carbon (SOC) content and each environmental variable, and obtaining the local correlation environmental variables includes: The geographic weight of each soil sample was determined using a Gaussian kernel function; According to the geographical weight of each soil sample, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable were calculated; The geographically weighted correlation coefficients at different spatial locations were calculated based on the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable; According to the globally relevant environmental variables, the globally irrelevant environmental variables and the geographically weighted correlation coefficient, the local correlation between the soil organic carbon (SOC) content and each environmental variable is analyzed to obtain the locally relevant environmental variables.
4. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 3 is characterized in that: Calculate the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable; Among them, the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable is expressed as: ; In the formula, is the geographically weighted covariance value between the soil organic carbon (SOC) content at different spatial locations and each environmental variable; and are the levels of environmental variables at location i and the levels of soil organic carbon (SOC) in soil samples at location i, respectively; and are the levels of environmental variables at location j and the levels of soil organic carbon (SOC) in soil samples at location j, respectively; and are the geographically weighted average ranks of environmental variables at location i and the geographically weighted average ranks of soil organic carbon (SOC) in soil samples at location i, respectively; For the The sample pair at position The geographic weight of the location samples, is the total number of soil samples; The geographically weighted standard deviation of each environmental variable at different spatial locations is expressed as: ; The geographically weighted standard deviation of soil organic carbon (SOC) content is expressed as: ; In the formula, and They are the geographically weighted standard deviation values of each environmental variable and the geographically weighted standard deviation values of soil organic carbon (SOC) content at different spatial locations.
5. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 4, characterized in that: The geographically weighted correlation coefficients at different spatial locations were calculated based on the geographically weighted covariance between the soil organic carbon (SOC) content at different spatial locations and each environmental variable, as well as the geographically weighted standard deviation of the soil organic carbon (SOC) content at different spatial locations and the geographically weighted standard deviation of each environmental variable. The geographically weighted correlation coefficients at different spatial locations were expressed as: ; In the formula, is the geographically weighted correlation coefficient value at different spatial locations.
6. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 1, characterized in that: The method of performing multicollinearity judgment on the local correlation environment variables according to the variance inflation factor and tolerance of the local correlation environment variables, deleting the local correlation environment variables with multicollinearity, and obtaining the environment variables without multicollinearity includes: The variance inflation factor and tolerance of local correlation environmental variables were determined by regression analysis. When the variance inflation factor was greater than 5 and the tolerance was less than 0.2, it indicated that there was multicollinearity in the local correlation environmental variables. The local correlation environmental variables with multicollinearity were deleted to obtain environmental variables without multicollinearity.
7. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 1, characterized in that: Based on the spatial weight of each soil sample and the effective bandwidth of environmental variables without multicollinearity, the multiscale geographically weighted regression model MGWR is used to fit the soil organic carbon SOC content with environmental variables without multicollinearity to obtain the local standardized regression coefficients of environmental variables without multicollinearity, including: Use Gaussian kernel function to weight each soil sample spatially to obtain the spatial weight of each soil sample; The effective bandwidth of each environmental variable is determined using the Akaike information criterion (AIC) or the Bayesian information criterion (BIC); Based on the spatial weight and the effective bandwidth, the multi-scale geographically weighted regression model MGWR is used to fit the soil organic carbon SOC content and the environmental variables without multicollinearity. According to the fitting results of the multi-scale geographically weighted regression model MGWR, the local standardized regression coefficients of the environmental variables without multicollinearity are calculated.
8. The multi-scale association analysis method for analyzing the spatial variability of soil organic carbon according to claim 7, characterized in that: The multi-scale geographically weighted regression model MGWR is expressed as: ; In the formula, For soil samples dependent variable; For soil samples No. environment variables; For the Regression equations for environmental variables Differentiation bandwidth of environmental variables; is the total number of variables; For soil samples The spatial location coordinates; The intercept term varies with the spatial position change; For soil samples The spatial position coordinates The The regression coefficients of the variables; is a random error term.
9. The multi-scale correlation analysis method for analyzing the spatial variability of soil organic carbon according to claim 7, characterized in that: Based on the fitting results of the multi-scale geographically weighted regression model MGWR, the method of calculating the local standardized regression coefficients of environmental variables without multicollinearity includes: Local standardized regression coefficients of environmental variables without multicollinearity were estimated by weighted least squares ; The local standardized regression coefficients of each environmental variable without multicollinearity were tested for statistical significance and spatially visualized to identify environmental variables with significant local effects on soil organic carbon (SOC).
10. A method for predicting soil organic carbon, characterized in that: According to the analysis results of the multi-scale association analysis method for analyzing the spatial variability of soil organic carbon according to claim 1, environmental variables are selected to predict soil organic carbon.
Citation Information
Patent Citations
Soil organic carbon predication method based on geographically weighted regression
CN104764868A
Soil organic carbon density spatial heterogeneity-considered Kriging interpolation model
CN107194054A
Forest aboveground biomass downscaling method based on multi-scale geographically weighted regression
CN116091939A
Systems and methods for predicting sample properties based on spectral measurements
US20230194424A1
Cited By
Spatial variation multi-scale soil organic matter remote sensing estimation method
CN119377864A
A multi-scale remote sensing estimation method for spatially variable soil organic matter
CN119377864B
Carbon emission fine scale comparison method and system based on geographical weighted correlation analysis
CN120163340A
Regional scale future soil carbon spatial-temporal change prediction method based on geographically weighted regression model
CN120509752A
A method for predicting spatio-temporal variation of future soil carbon at regional scale based on a geographically weighted regression model
CN120509752B
