Method for spatial mapping of soil organic matter content based on erosion zoning

Abstract

The present application relates to the technical field of soil analysis, and particularly relates to a soil organic matter content spatial mapping method based on erosion zoning, which comprises the following steps: according to the causes of water erosion and wind erosion of the soil in the research area, dividing the multi-dimensional environmental factor data of the research area into a water erosion related factor set and a wind erosion related factor set, and determining the water erosion intensity and the wind erosion intensity respectively; establishing a joint erosion probability model of the water erosion intensity and the wind erosion intensity; taking the joint erosion probability as the zoning basis, spatially zoning the research area, and respectively performing feature screening and prediction model construction in different zones; applying the prediction model to the corresponding spatial region of the research area to obtain the complete soil organic matter spatial distribution result of the research area. The present application can effectively reduce the mixed influence of different erosion processes on the prediction of soil organic matter, enhance the adaptability of the model to regional spatial heterogeneity, and improve the accuracy of soil organic matter spatial mapping and its explanation ability to actual processes.

Description

Technical Field

[0001] This invention belongs to the field of soil analysis technology, and in particular relates to a spatial mapping method for soil organic matter content based on erosion zones. Background Technology

[0002] With the rapid development of remote sensing and geographic information system (GIS) technologies, spatial prediction techniques for soil properties based on multi-source remote sensing data and digital soil mapping have gradually matured. These methods typically use measured soil sample data as a foundation, combining topographic factors, climatic factors, and spectral information extracted from remote sensing images—multi-source environmental variables—to construct quantitative relationship models between soil properties and environmental factors, achieving a continuous spatial expression of soil properties. Compared to traditional empirical or interpolation-based methods, these methods offer advantages such as high data utilization efficiency, wide applicability, and strong scalability, and have become the mainstream technique for regional-scale soil organic matter mapping. In recent years, the application of machine learning methods in spatial prediction of soil organic matter has been increasing. Compared to traditional statistical methods, machine learning models (such as random forests and support vector machines) can effectively handle complex nonlinear relationships and improve prediction accuracy, thus finding widespread application in the field of digital soil mapping.

[0003] Under complex natural environmental conditions, soil erosion is a crucial process influencing the spatial distribution of soil organic matter, primarily encompassing water erosion and wind erosion. Rainfall runoff and wind action cause the migration and redistribution of soil particles and organic matter, thereby altering soil surface structure and nutrient migration processes. While existing spatial mapping techniques for soil organic matter content can utilize multi-source environmental information to represent the spatial distribution of soil organic matter, their reflection of the impact of erosion processes remains insufficient in areas of significant erosion, mainly in the following aspects: (1) Existing spatial mapping methods for soil organic matter mainly rely on environmental variables such as topography, climate and remote sensing to build prediction models. They usually use various factors as independent input variables for modeling, and lack explicit characterization of key soil evolution processes such as soil erosion. In fact, erosion processes such as water erosion and wind erosion are important driving mechanisms affecting the migration and redistribution of soil organic matter. Ignoring such processes will make it difficult for the model to accurately reflect the formation and change of soil organic matter, thereby reducing the reliability of the prediction results. (2) Existing zoning modeling methods are mostly based on classification data such as soil type and land use type to divide the study area. The zoning boundaries are usually set artificially or derived from existing maps, which have obvious discreteness and subjectivity. However, soil organic matter usually shows a continuous and gradual change in space, and there are transition zones between different areas. Zoning with fixed boundaries is prone to artificially dividing areas with similar environmental characteristics or classifying areas with large differences into the same category, making it difficult for the model to accurately characterize spatial heterogeneity, thus affecting the prediction accuracy. Summary of the Invention

[0004] In view of this, the present invention aims to provide a spatial mapping method for soil organic matter content based on erosion zoning. By constructing water erosion and wind erosion intensity models and establishing a joint erosion probability model for both, it achieves coupled characterization of multiple erosion processes. Using the joint erosion probability as the zoning basis, the study area is spatially partitioned, and feature selection and model construction are performed separately within each partition, thereby realizing the transformation from globally unified modeling to partition-differentiated modeling. Through the above technical solution, the present invention can effectively reduce the confounding effects of different erosion processes on soil organic matter prediction, enhance the model's adaptability to regional spatial heterogeneity, and improve the accuracy of soil organic matter spatial mapping and its interpretability of actual processes.

[0005] To achieve the above objectives, the technical solution created by this invention is implemented as follows: A spatial mapping method for soil organic matter content based on erosion zones includes: S1: Obtain soil organic matter sample data and corresponding multidimensional environmental factor data for the year in the study area; based on the causes of water erosion and wind erosion in the soil of the study area, divide the multidimensional environmental factor data into a set of water erosion-related factors and a set of wind erosion-related factors. S2: Based on the set of water erosion-related factors and the set of wind erosion-related factors obtained in step S1, the water erosion intensity and wind erosion intensity of the soil in the study area are calculated respectively. S3: Construct the joint distribution relationship between water erosion intensity and wind erosion intensity obtained in step S2, determine the joint erosion probability value of water erosion and wind erosion in each spatial unit in the study area, and obtain the spatial distribution result of joint erosion probability; S4: Cluster the spatial distribution results obtained in step S3 to generate a spatial partition map based on the joint erosion probability; S5: Divide the two types of data obtained in step S1 according to the spatial partition map obtained in step S4, and match the two types of data in each partition to form a partition feature set; filter each partition feature set to obtain the optimal feature set for each partition. S6: Use the optimal feature set obtained in step S5 to train the random forest regression model, and apply the trained random forest regression model of each partition to the corresponding spatial region of the study area to obtain the complete spatial distribution results of soil organic matter in the study area.

[0006] Furthermore, the soil organic matter sample point data in step S1 includes the spatial location information of the sample points and the corresponding measured values ​​of soil organic matter content; step S1 also includes: preprocessing the acquired soil organic matter sample point data, including outlier identification and removal, wherein: based on the mean and standard deviation of the soil organic matter sample point data, sample points that deviate from the mean by more than three times the standard deviation are identified and removed as outlier samples.

[0007] Furthermore, the multidimensional environmental factor data includes topographic factors, climate factors, soil property factors, and remote sensing index factors.

[0008] Furthermore, in step S2, the intensity of water erosion is calculated by modifying the general soil loss equation, and the intensity of wind erosion is calculated by modifying the wind erosion equation.

[0009] Furthermore, the calculation process for water erosion intensity includes: calculating rainfall erosion factors based on multi-year average precipitation data of the study area; constructing a mapping relationship between vegetation cover and vegetation indices obtained from remote sensing inversion in the study area, and then obtaining vegetation cover and management factors based on vegetation cover; assigning values ​​to soil and water conservation measures factors according to the land use type of the study area; calculating soil erodibility factors based on soil particle composition information and soil organic matter content in the soil of the study area; calculating slope and slope length factors in the topographic factors of the study area based on the digital elevation model of the study area; and comprehensively calculating the rainfall erosion factors, vegetation cover and management factors, soil and water conservation measures factors, soil erodibility factors, slope factors, and slope length factors to obtain the spatially continuous water erosion intensity of the study area.

[0010] Furthermore, the calculation process for wind erosion intensity includes: calculating climate factors based on meteorological data of the study area; calculating soil erosibility factors and soil crust factors based on soil particle composition of the soil in the study area; calculating surface roughness factors based on surface micro-topographic features of the study area; calculating vegetation cover factors based on remote sensing vegetation index or land cover data of the study area; comprehensively calculating the climate factors, soil erosibility factors, soil crust factors, surface roughness factors, and vegetation cover factors to obtain the length of key plots in the erosion area of ​​the study area, as well as the maximum amount of soil particles transferred under wind action; and obtaining the spatially continuous wind erosion intensity of the study area using the following formula: ; Among them, S q The value of Q represents the wind erosion intensity, Z represents the distance at which the maximum wind erosion occurs, and Q represents the wind erosion intensity. max and S L These represent the maximum transfer volume and the length of the critical parcel, respectively.

[0011] Furthermore, step S3 includes: S31: Use the Copula function to construct the joint distribution relationship between water erosion intensity and wind erosion intensity; S32: Based on the joint distribution relationship constructed in step S31, perform pixel-by-pixel calculations for each spatial unit in the study area to obtain the joint erosion probability value of each pixel. S33: Based on the combined erosion probability value obtained in step S32, generate the spatial distribution results of the combined erosion probability of water erosion and wind erosion in the study area.

[0012] Furthermore, step S4 includes: S41: The study area is divided into several sub-regions with similar comprehensive erosion characteristics based on the spatial similarity of the joint erosion probability. S42: The elbow rule is used to evaluate the clustering effect under different numbers of clusters, and the optimal number of partitions is determined based on the evaluation results; S43: Generate a spatial partition map based on erosion probability according to the optimal clustering results.

[0013] Furthermore, step S5 includes: assigning the soil organic matter sample point data and multidimensional environmental factor data obtained in step S1 to the corresponding partitions in the spatial partitioning map obtained in step S4, thereby obtaining the partition feature set of each partition; scoring the importance of each partition feature in the partition feature set of each partition, and then using the recursive feature elimination method to filter the partition features according to the importance score, thereby obtaining the optimal feature set of each partition.

[0014] Furthermore, in step S6: using the multidimensional environmental factor data in the optimal feature set as input and the soil organic matter sample point data in the optimal feature set as the true value, the random forest regression model is trained; the random forest regression model for each partition is applied to the corresponding spatial region of the study area to obtain the soil organic matter prediction results for each partition, and all the soil organic matter prediction results are spliced ​​together according to the spatial region to obtain the complete spatial distribution results of soil organic matter in the study area.

[0015] Compared with the prior art, the present invention can achieve the following beneficial effects: (1) In the spatial mapping method of soil organic matter content based on erosion zoning described in this invention, the erosion process is elevated from a traditional input variable to a driving factor of the model structure. Through the integrated framework of "erosion coupling - spatial zoning - zoning modeling", the erosion process is given explicit constraints on the prediction of soil organic matter. This method is not only applicable to the prediction of soil organic matter, but can also be extended to the spatial modeling of other soil properties or ecological variables. (2) In the spatial mapping method of soil organic matter content based on erosion zoning described in this invention, by constructing a water erosion and wind erosion intensity model and using the Copula function to establish a joint erosion probability distribution, the coupled expression of multiple erosion processes in a unified probability space is realized. Compared with the existing methods that only input erosion factors as independent variables into the model, this invention can more realistically reflect the synergistic mechanism of erosion processes. This invention uses the joint erosion probability as the sole zoning basis to realize the spatial zoning of the study area, so that the zoning results directly reflect the differences in the erosion environment. Compared with the existing methods that zonify based on multiple environmental variables or empirical rules, the zoning results of this invention have clearer physical meaning and process consistency. (3) In the spatial mapping method for soil organic matter content based on erosion zoning described in this invention, a zoning modeling strategy based on erosion zoning is adopted. Feature selection and model construction are carried out separately in different erosion environment areas, so that the model is trained only under similar erosion conditions, effectively reducing the confounding effects of different erosion processes on soil organic matter prediction, thereby improving the prediction accuracy and stability of the model. Feature selection is carried out separately in different erosion zoning areas, so that the optimal feature set of each zoning area can reflect the dominant driving factors of soil organic matter formation and evolution in the corresponding area. Compared with the globally unified feature selection method, this invention can more accurately reveal the mechanism of action under different erosion environments. Attached Figure Description

[0016] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic flowchart of the spatial mapping method for soil organic matter content based on erosion zoning as described in the embodiments of the present invention; Figure 2 A flowchart illustrating the spatial mapping method for soil organic matter content based on erosion zoning as described in the embodiments of the present invention; Figure 3 The diagram illustrates the results of an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0018] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] like Figures 1 to 2 As shown in the embodiments of the present invention, the spatial mapping method for soil organic matter content based on erosion zoning includes: S1: Obtain soil organic matter sample data and corresponding multidimensional environmental factor data for the year in the study area; based on the causes of water erosion and wind erosion in the soil of the study area, divide the multidimensional environmental factor data into a set of water erosion-related factors and a set of wind erosion-related factors.

[0020] In some embodiments, the soil organic matter sample point data includes the spatial location information of the sample points and the corresponding measured values ​​of soil organic matter content; step S1 further includes preprocessing the acquired soil organic matter sample point data, including outlier identification and removal, wherein: based on the mean and standard deviation of the soil organic matter sample point data, sample points that deviate from the mean by more than three times the standard deviation are identified and identified as outlier samples and removed.

[0021] In this embodiment of the invention, the spatial location information is preferably geographic coordinate information obtained through a global satellite navigation system, and the soil organic matter sample point data can be derived from multiple sources such as existing soil survey data, long-term positioning monitoring data, or historical observation data. Furthermore, the preprocessing operation in this embodiment of the invention also includes standardizing or normalizing the soil organic matter sample point data after removing outliers to improve the stability and accuracy of subsequent model training.

[0022] In some embodiments, the multidimensional environmental factor data includes topographic factors, climate factors, soil property factors, and remote sensing index factors. In this embodiment, the remote sensing index factors are preferably constructed based on remote sensing images of bare soil during April and May, and are calculated by statistically synthesizing (preferably median synthesis) images within the target time period. In this embodiment, the multidimensional environmental factor data also undergoes unified preprocessing, including spatial registration, coordinate system unification, and resolution resampling. Preferably, all multidimensional environmental factor data are unified to the same spatial resolution to ensure the consistency of multi-source data at spatial scale, thereby meeting the needs of subsequent modeling and spatial analysis.

[0023] S2: Based on the set of water erosion-related factors and the set of wind erosion-related factors obtained in step S1, calculate the water erosion intensity and wind erosion intensity of the soil in the study area, respectively.

[0024] In some embodiments, water erosion intensity is calculated using the Revised Universal Soil Loss Equation (RUSLE), and wind erosion intensity is calculated using the Revised Wind Erosion Equation (RWEQ).

[0025] Water erosion intensity is used to quantitatively characterize the potential erosion of soil per unit area under the influence of rainfall and surface runoff. It is calculated comprehensively using rainfall erosion factors, soil erodibility factors, topographic factors, vegetation cover and management factors, and soil and water conservation measures factors. These factors synergistically reflect the combined impact of rainfall, topography, soil properties, and human activities on the water erosion process. Specifically, in some embodiments, the calculation process for water erosion intensity includes: The rainfall erosion factor, reflecting rainfall intensity and its erosive capacity, is calculated based on multi-year average precipitation data of the study area. In this embodiment of the invention, an empirical model based on monthly precipitation is preferably used to estimate the rainfall erosion factor using the following formula: ; Where R represents the rainfall erosion factor, r m Let M represent the average monthly precipitation in month m, and M represent the total number of months for which the average monthly precipitation is calculated. In this embodiment of the invention, the average monthly precipitation for a year is calculated, i.e., M=12. This corresponds to the average annual precipitation over many years; A mapping relationship between vegetation cover and vegetation indices obtained from remote sensing inversion in the study area is constructed. Then, vegetation cover and management factors are obtained based on vegetation cover to quantify the inhibitory effect of vegetation on soil erosion. In this embodiment of the invention, the Normalized Difference Vegetation Index (NDVI) is preferably used as the vegetation index, and the following mapping relationship is constructed: ; Where f represents vegetation cover, NDVI veg and NDVI soil These represent the 5th and 95th percentiles of the Normalized Difference Vegetation Index (NDVI), respectively. The vegetation cover and management factors are obtained using the following formula: ; Where C represents vegetation cover and management factors; Since different land use types correspond to different soil and water conservation capacities, soil and water conservation measures factors are assigned values ​​according to the land use type of the study area to reflect the impact of human intervention on the erosion process. In this embodiment of the invention, the soil and water conservation measures factor P values ​​for paddy fields and dry land are 0.15 and 0.352, respectively. Based on the soil particle composition and organic matter content in the study area, a soil erodibility factor is calculated to characterize the soil's resistance to erosion. In this embodiment of the invention, the soil erodibility factor is calculated using the following formula: ; Where K represents the soil erodibility factor, sand represents the percentage of sand content, silt represents the percentage of silt content, clay represents the percentage of clay content, and SOC represents the percentage of soil organic carbon content. Based on the digital elevation model of the study area, the slope factor and slope length factor in the topographic factors of the study area are calculated. In this embodiment of the invention, the slope factor is calculated based on DEM data, and the slope length factor is preferably characterized by spatial resolution or water flow path length, thereby reflecting the amplification effect of topography on the water erosion process, as shown in the following formula: ; ; Where L and S represent the slope length factor and slope gradient factor, respectively, θ represents the slope, and λ represents the slope length. In this embodiment of the invention, 30m is used, and the intermediate parameter β is: .

[0026] By comprehensively calculating the above factors, the spatially continuous distribution of water erosion intensity in the study area is obtained. The specific calculation formula for water erosion intensity in this embodiment of the invention is as follows: A = R × K × L × S × C × P; Where A represents the intensity of water erosion.

[0027] Wind erosion intensity is used to characterize the potential erosion and transport capacity of soil particles under wind action. Specifically, in some embodiments, the calculation process of wind erosion intensity includes: Climate factors are calculated based on meteorological data of the study area. In this embodiment of the invention, the climate factors are calculated using the following formula: ; Where WF represents the climate factor, SW represents the soil moisture in the study area, SD is the snow cover coefficient (dimensionless), representing the ratio of the number of days with snow cover less than 25.4 mm to the total number of days in a year, and N... d In this embodiment of the invention, N represents the period for detecting meteorological data in the study area. d =30 days, meaning that meteorological data of the study area were monitored for 30 days, where N represents the period N. d The frequency of wind speed observations inside the building typically exceeds 500. This represents the threshold wind speed at a set height above the ground during the nth observation. In this embodiment of the invention, the standard observation height of the meteorological station is 10m, and the threshold wind speed at a height of 2m above the ground is set to 5m / s. , The actual wind speed at a set height above the ground (i.e., 2m in this embodiment of the invention) during the nth observation is derived from the wind speed at the standard observation height of the meteorological station through the wind profile equation, where g represents gravitational acceleration and ρ represents air density. The soil moisture SW in the above formula is composed of rainfall (rain), potential evapotranspiration (ETp), irrigation depth (I) (in this embodiment of the invention, I=0), and the number of experimental days (N). d The result is obtained through the following formula: ; Among them, R d Indicates the number of days of rainfall or irrigation per month; Based on the soil particle composition in the study area, the soil erodibility factor and soil crust factor were calculated. The soil erodibility factor characterizes the transportability of soil particles under wind action, while the soil crust factor reflects the influence of soil surface structure on wind erosion. In this embodiment of the invention, the soil erodibility factor is calculated using the following formula: ; Wherein, EF represents the soil erodibility factor, and Ca represents the carbonate content in the soil of the study area; The soil crust factor is calculated using the following formula: ; SCF represents soil crust factor; Based on the surface micro-topographic features of the study area, the surface roughness factor is calculated. The surface roughness factor characterizes the disturbance effect of random roughness and topographic undulation on airflow. In this embodiment of the invention, the surface roughness factor is calculated using the following formula: ; Where K' represents the surface roughness factor, K r C represents the ridge roughness coefficient of the study area. rr Indicates random roughness; Based on remote sensing vegetation index or land cover data of the study area, the vegetation cover factor of the study area is calculated to characterize the effect of vegetation on near-surface wind speed reduction and particle blocking. In this embodiment of the invention, the vegetation cover factor is calculated by the following formula: ; C represents the vegetation cover factor, ai represents the coefficient for different vegetation types, and the value for farmland is -0.0438; SC is the vegetation cover calculated from NDVI. By comprehensively calculating the above factors, the length of the key plot in the erosion area of ​​the study area and the maximum amount of soil particles transferred under wind action are obtained. The specific formula for calculating the maximum transfer amount in this embodiment of the invention is as follows: Q max =109.8×(WF×EF×SCF×K'×C); S L =150.71×(WF×EF×SCF×K'×C) -0.3711 ; Among them, Q max and S L These represent the maximum transfer volume and the length of the key parcel, respectively. The spatially continuous wind erosion intensity of the study area is obtained by the following formula: ; Among them, S q S represents the wind erosion intensity, and Z represents the distance at which maximum wind erosion occurs. Wherein, wind erosion intensity S... q In actual engineering calculations, this refers to the actual amount of wind erosion affecting the soil in the study area.

[0028] Water erosion intensity and wind erosion intensity are not only used to calculate erosion volume, but also serve as the basic variables for subsequent coupled erosion modeling, thereby realizing the transformation of the erosion process from a single description to joint modeling.

[0029] S3: Construct the joint distribution relationship between water erosion intensity and wind erosion intensity obtained in step S2, determine the joint erosion probability value of water erosion and wind erosion in each spatial unit in the study area, and obtain the spatial distribution result of joint erosion probability in the study area.

[0030] In some embodiments, step S3 includes: S31: The Copula function is used to construct the joint distribution relationship between water erosion intensity and wind erosion intensity. By establishing the correlation structure between the two, the original univariate erosion intensity is transformed into a unified joint probability space. In this embodiment of the invention, the Gaussian Copula function is preferably used to construct the joint distribution relationship. S32: Based on the joint distribution relationship constructed in step S31, each spatial unit in the study area is calculated pixel by pixel to obtain the joint erosion probability value of each pixel, thereby characterizing the degree of synergistic effect of different erosion processes in space. S33: Based on the combined erosion probability value obtained in step S32, generate the spatial distribution result of the combined erosion probability of water erosion and wind erosion in the study area. This result is used to quantitatively describe the probability of the occurrence of comprehensive erosion in the study area and its spatial heterogeneity characteristics.

[0031] The joint erosion probability proposed in this invention can be further used for erosion zoning or as a key driving factor input for subsequent soil organic matter spatial prediction models, thereby realizing coupled modeling between erosion processes and soil organic matter changes. In addition, this invention introduces the Copula function to construct a joint erosion probability model, so that the originally independent water erosion and wind erosion processes can be expressed in a unified probability space, thereby realizing coupled modeling of multiple erosion processes.

[0032] S4: Cluster the spatial distribution results obtained in step S3 to generate a spatial partition map based on the joint erosion probability.

[0033] In some embodiments, step S4 includes: S41: The study area is divided into several sub-regions with similar comprehensive erosion characteristics based on the spatial similarity of the joint erosion probabilities. In this embodiment of the invention, the K-means clustering algorithm is preferred. S42: The elbow rule is used to evaluate the clustering effect under different numbers of clusters, and the optimal number of partitions is determined based on the evaluation results; In this embodiment of the invention, the within-cluster sum of squares (WCSS) is used as the evaluation index of the clustering effect, the trend of the within-cluster sum of squares with the number of clusters is analyzed, and the number of clusters corresponding to the inflection point from a significant decrease to a gradual change is determined as the optimal number of partitions. S43: Generate a spatial partition map based on erosion probability according to the optimal clustering results.

[0034] The zoning results based on the joint erosion probability are used to guide subsequent zoning modeling of soil organic matter, so that the same zoning has a similar erosion background, thereby reducing the spatial heterogeneity of the model input and improving the spatial prediction accuracy of soil organic matter.

[0035] S5: Divide the two types of data obtained in step S1 according to the spatial partition map obtained in step S4, and match the two types of data in each partition to form a partition feature set; filter each partition feature set to obtain the optimal feature set for each partition.

[0036] In some embodiments, step S5 includes: assigning the soil organic matter sample point data and multidimensional environmental factor data obtained in step S1 to the corresponding partitions in the spatial partitioning map obtained in step S4, thereby completing the matching of the two types of data and obtaining the partition feature set of each partition; scoring the importance of each partition feature in the partition feature set of each partition, and then using the recursive feature elimination method to filter the partition features according to the importance score to obtain the optimal feature set of each partition.

[0037] In this embodiment of the invention, it is preferred to use a combination of the Random Forest algorithm (Recursive Feature Elimination based on Random Forest, RFE-RF) and the recursive feature elimination method to complete feature selection. That is, in each iteration, based on the training results of the Random Forest algorithm model, the importance score of each feature in the feature set of each partition is calculated, and the partition features with low importance scores are gradually eliminated. The model training and evaluation process is repeated until the optimal feature set of each partition is obtained.

[0038] The optimal feature sets obtained in different partitions differ, which can reflect the differences in the dominant driving factors of soil organic matter formation and evolution under different erosion environments, and realize the transformation from global unified feature modeling to partition adaptive feature modeling.

[0039] S6: Use the optimal feature set obtained in step S5 to train the random forest regression model, and apply the trained random forest regression model of each partition to the corresponding spatial region of the study area to obtain the complete spatial distribution results of soil organic matter in the study area.

[0040] In some embodiments, multidimensional environmental factor data from the optimal feature set are used as input, and soil organic matter sample point data from the optimal feature set are used as the true values ​​to train the random forest regression model. The random forest regression model for each partition is applied to the corresponding spatial region of the study area to obtain the soil organic matter prediction results for each partition. All soil organic matter prediction results are spliced ​​together according to the spatial region to obtain the complete spatial distribution results of soil organic matter in the study area.

[0041] In this embodiment of the invention, the random forest regression model establishes a nonlinear mapping relationship between soil organic matter and environmental factors through multi-decision tree ensemble learning. The process of training the random forest regression model also includes generating multiple feature subsets by bootstrap sampling of the optimal feature set, randomly selecting some features at each bootstrap node for splitting, and using the bootstrap sampling and the multiple split feature subsets to train the random forest regression model, thereby improving the model's generalization ability and robustness.

[0042] To verify that the constructed random forest regression model can effectively and reliably predict soil organic matter spatially, this embodiment of the invention also provides a process for evaluating the accuracy of the model prediction results. Specifically, this includes: acquiring new soil organic matter sample point data and multidimensional environmental factor data for the study area again; using the method provided by this invention to obtain the predicted soil organic matter data corresponding to the multidimensional environmental factor data; using the acquired soil organic matter sample point data as the true value; and comparing and analyzing the predicted value with the corresponding measured value.

[0043] Furthermore, multiple statistical indicators are used to quantitatively evaluate the model accuracy, including the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and interquartile range performance index (RPIQ), to comprehensively characterize the model's fitting ability and prediction accuracy. Further, the model accuracy within different erosion zones is evaluated separately to analyze the applicability and stability of the zoned modeling strategy under different erosion environments. Finally, by comparing the accuracy differences between the zoned modeling results and the global modeling results, the effectiveness of the proposed method in improving the spatial prediction accuracy of soil organic matter is verified. The results obtained from the embodiments of this invention are as follows: Figure 3 As shown. From Figure 3 As can be seen from this, the determination coefficient (R²) of the partitioning modeling scheme provided by this invention is... 2 =0.74) is significantly higher than that of global modeling (R = 0.74) 2 =0.64), while the root mean square error (RMSE) and mean absolute error (MAE) are significantly lower than those of the global model. This means that the prediction results of the partition modeling provided by this invention are closer to the true values ​​and can effectively reduce prediction errors and improve the prediction accuracy of soil organic matter. This confirms that the method of constructing a local random forest model based on erosion partitions in this invention has better prediction performance than traditional global modeling.

[0044] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0045] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A spatial mapping method for soil organic matter content based on erosion zones, characterized in that, include: S1: Obtain soil organic matter sample data and corresponding multidimensional environmental factor data for the year in the study area; based on the causes of water erosion and wind erosion in the soil of the study area, divide the multidimensional environmental factor data into a set of water erosion-related factors and a set of wind erosion-related factors. S2: Based on the set of water erosion-related factors and the set of wind erosion-related factors obtained in step S1, the water erosion intensity and wind erosion intensity of the soil in the study area are calculated respectively. S3: Construct the joint distribution relationship between water erosion intensity and wind erosion intensity obtained in step S2, determine the joint erosion probability value of water erosion and wind erosion in each spatial unit in the study area, and obtain the spatial distribution result of joint erosion probability; S4: Cluster the spatial distribution results obtained in step S3 to generate a spatial partition map based on the joint erosion probability; S5: Divide the two types of data obtained in step S1 according to the spatial partition map obtained in step S4, and match the two types of data in each partition to form a partition feature set; filter each partition feature set to obtain the optimal feature set for each partition. S6: Use the optimal feature set obtained in step S5 to train the random forest regression model, and apply the trained random forest regression model of each partition to the corresponding spatial region of the study area to obtain the complete spatial distribution results of soil organic matter in the study area.

2. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 1, characterized in that, The soil organic matter sample point data in step S1 includes the spatial location information of the sample points and the corresponding measured values ​​of soil organic matter content. Step S1 also includes: preprocessing the acquired soil organic matter sample point data, including outlier identification and removal, wherein: based on the mean and standard deviation of the soil organic matter sample point data, sample points that deviate from the mean by more than three times the standard deviation are identified and removed as outliers.

3. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 1, characterized in that, Multidimensional environmental factor data include topographic factors, climate factors, soil property factors, and remote sensing index factors.

4. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 1, characterized in that, In step S2, the intensity of water erosion is calculated by modifying the general soil loss equation, and the intensity of wind erosion is calculated by modifying the wind erosion equation.

5. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 4, characterized in that, The calculation process for water erosion intensity includes: Rainfall erosion factor was calculated based on multi-year average precipitation data for the study area. A mapping relationship between vegetation cover and vegetation indices obtained by remote sensing inversion was constructed in the study area, and then vegetation cover and management factors were obtained based on vegetation cover. Based on the land use type of the study area, values ​​were assigned to the soil and water conservation measures factors. Based on the soil particle composition information and soil organic matter content in the study area, the soil erodibility factor was calculated. Based on the digital elevation model of the study area, the slope factor and slope length factor in the topographic factors of the study area are calculated. By comprehensively calculating rainfall erosion factors, vegetation cover and management factors, soil and water conservation measures factors, soil erodibility factors, slope factors, and slope length factors, the spatially continuous distribution of water erosion intensity in the study area was obtained.

6. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 4, characterized in that, The calculation process for wind erosion intensity includes: Climate factors were calculated based on meteorological data from the study area. Based on the soil particle composition in the study area, the soil erodibility factor and soil crust factor were calculated. Based on the surface micro-topographic features of the study area, the surface roughness factor was calculated. Based on remote sensing vegetation index or land cover data of the study area, calculate the vegetation cover factor of the study area; By comprehensively calculating climatic factors, soil erodibility factors, soil crust factors, surface roughness factors, and vegetation cover factors, the length of key plots in the erosion area of ​​the study area and the maximum amount of soil particles transferred under wind action were obtained. The spatially continuous wind erosion intensity of the study area is obtained by the following formula: ； Among them, S q Indicates wind erosion intensity, Z represents the distance at which maximum wind erosion occurs, and Q represents the distance at which maximum wind erosion occurs. max and S L These represent the maximum transfer volume and the length of the critical parcel, respectively.

7. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 1, characterized in that, Step S3 includes: S31: Use the Copula function to construct the joint distribution relationship between water erosion intensity and wind erosion intensity; S32: Based on the joint distribution relationship constructed in step S31, perform pixel-by-pixel calculations for each spatial unit in the study area to obtain the joint erosion probability value of each pixel. S33: Based on the combined erosion probability value obtained in step S32, generate the spatial distribution results of the combined erosion probability of water erosion and wind erosion in the study area.

8. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 1, characterized in that, Step S4 includes: S41: The study area is divided into several sub-regions with similar comprehensive erosion characteristics based on the spatial similarity of the joint erosion probability. S42: The elbow rule is used to evaluate the clustering effect under different numbers of clusters, and the optimal number of partitions is determined based on the evaluation results; S43: Generate a spatial partition map based on erosion probability according to the optimal clustering results.

9. The method for spatial mapping of soil organic matter content based on erosion zoning according to claim 1, characterized in that, Step S5 includes: assigning the soil organic matter sample point data and multidimensional environmental factor data obtained in step S1 to the corresponding partitions in the spatial partitioning map obtained in step S4, and obtaining the partition feature set of each partition; scoring the importance of each partition feature in the partition feature set of each partition, and then using the recursive feature elimination method to filter the partition features according to the importance score, and obtaining the optimal feature set of each partition.

10. The spatial mapping method for soil organic matter content based on erosion zoning according to claim 1, characterized in that, In step S6: The random forest regression model is trained using multidimensional environmental factor data from the optimal feature set as input and soil organic matter sample point data from the optimal feature set as the true values. The random forest regression model for each partition is applied to the corresponding spatial region of the study area to obtain the soil organic matter prediction results for each partition. All the soil organic matter prediction results are then spliced ​​together according to the spatial region to obtain the complete spatial distribution results of soil organic matter in the study area.

Images