Reasonable plough layer evaluation index system construction method based on dual-objective optimization

Abstract

The invention relates to the technical field of plough layer structure identification and evaluation, in particular to a reasonable plough layer evaluation index system construction method based on dual-objective optimization. Comprising the following steps: acquiring target area cultivated land multi-source soil data, and preprocessing to obtain an original soil attribute variable set; constructing a double-target comprehensive scoring function to obtain a comprehensive scoring value to divide the levels of the plough layer; processing the original soil attribute variable set by adopting a principal component analysis method to obtain a first variable set; constructing an LASSO regression model to screen the first variable set and the original soil attribute variable set to obtain a second variable set and a third variable set; acquiring a fourth variable set by adopting a random forest algorithm; key variable sets of three levels are determined through intersection comparison, and a reasonable plough layer level and index system is constructed in combination with the plough layer level of the cultivated land in the target area. According to the method, collaborative quantification of a plough layer output function and an anti-corrosion function is achieved, and a scientific basis is provided for farmland quality improvement, soil improvement and partition management.

Description

Technical Field

[0001] This invention relates to the field of topsoil structure identification and evaluation technology, specifically to a method for constructing a reasonable topsoil evaluation index system based on dual-objective optimization. Background Technology

[0002] The topsoil layer is a crucial soil layer in arable land, responsible for multiple functions such as crop root growth, nutrient supply, water regulation, and structural support. Its thickness, structure, and functional rationality directly determine the productive potential, stability, and erosion resistance of arable land. Under the combined effects of long-term high-intensity tillage, slope hydraulic erosion, and inappropriate soil management practices, problems such as topsoil structure degradation, reduced thickness, increased compaction, and intensified nutrient differentiation have gradually become prominent, becoming significant factors restricting sustainable agricultural production.

[0003] Current technologies for topsoil evaluation generally equate "topsoil quality" with "high-yield capacity," typically using crop yield or soil fertility indicators as the primary evaluation criteria. However, in areas with severe erosion, sloping farmland, or ecologically fragile regions, yield levels are easily affected by fertilization amounts, farming practices, and short-term control measures. This can lead to "apparent high yields" masking topsoil degradation, making yield-oriented evaluation methods inadequate to reflect the true stability and long-term ecological function of the topsoil. In such areas, even if fertilization or deep tillage can increase yields in the short term, processes such as topsoil shallowing, compaction, and increased erodibility will continuously weaken the soil's stable yield capacity, leading to a decoupling of productive and ecological functions. Therefore, simply using high yield as a criterion for topsoil rationality is insufficient to meet the current needs for improving farmland quality and ensuring long-term sustainable use.

[0004] From the perspective of current technological development, the existing research and application system for topsoil evaluation still has the following main limitations: (1) Lack of functional quantitative standards for “reasonable topsoil”: Most studies define the boundaries of topsoil by fixed depth or profile morphology, ignoring the influence of different soil types, topography and farming systems on the evolution of functional layers, making it difficult to accurately identify the actual functional thickness and structural rationality of topsoil. (2) Single evaluation index orientation: Traditional methods often construct index models with soil fertility or crop yield as a single objective, ignoring the ecological functions of the topsoil in resisting erosion, maintaining structural stability and conserving water and fertilizer, and failing to reveal the synergistic mechanism of "high yield-stable yield". (3) The construction of the indicator system lacks an objective screening mechanism: the existing evaluation system generally relies on experience judgment or expert weighting, the selection of indicators is arbitrary and the weights are subjective, and the collinearity and variable redundancy among multi-source soil data are not considered, resulting in unstable evaluation results and poor repeatability. (4) Insufficient structure-function coupling: Although some studies have tried to use methods such as PCA and random forest to screen variables, they are mainly used for dimensionality reduction or correlation analysis, and still remain at the single objective optimization, lacking a collaborative modeling and trade-off framework for the dual functions of "yield target - corrosion resistance target". (5) Lack of standardized and generalizable methods at the regional scale: Existing models are mostly built at the scale of single plots or small watersheds, relying on high-density sampling points and local empirical weights, making it difficult to apply them uniformly under different landforms, parent materials and climate conditions.

[0005] In summary, current technologies have not yet established a dual-objective collaborative evaluation method capable of simultaneously characterizing the "yield stability" and "erosion resistance" of the topsoil, nor do they possess a systematic technical solution for objectively selecting core indicators and constructing a rational topsoil evaluation system at a regional scale. Therefore, it is necessary to propose a rational topsoil evaluation method that integrates dual-objective collaborative optimization, multi-algorithm consistency screening, and the construction of an indicator grading system, in order to achieve the scientific quantification of topsoil functions and its scalable application at a regional scale. Summary of the Invention

[0006] To address the problems of existing topsoil evaluation methods, such as single-objective orientation, strong subjectivity in variable selection, insufficient structure-function coupling, and poor regional applicability, this invention provides a method for constructing a reasonable topsoil evaluation index system based on dual-objective optimization. This method obtains an original soil attribute variable set by preprocessing multi-source soil data from cultivated land in the target area; constructs a dual-objective comprehensive scoring function to obtain a comprehensive score value; classifies the topsoil levels; uses principal component analysis to process the original soil attribute variable set to obtain a first variable set; constructs a LASSO regression model to filter the first variable set and the original soil attribute variable set to obtain a second and third variable set; uses a random forest algorithm to obtain a fourth variable set; and determines the key variable sets for the three levels through intersection comparison. Combined with the topsoil levels of cultivated land in the target area, a reasonable topsoil level and index system is constructed. This invention achieves the synergistic quantification of topsoil output function and erosion resistance function, possesses cross-regional applicability and standardized screening capabilities, and can provide a scientific basis for large-scale cultivated land quality improvement, soil improvement, and zoning management.

[0007] This invention adopts the following technical solution: a method for constructing a reasonable topsoil evaluation index system based on dual-objective optimization, comprising: Acquire multi-source soil data for cultivated land in the target area; The multi-source soil data were standardized and subjected to multicollinearity diagnostic processing to obtain the original set of soil attribute variables; A dual-objective comprehensive scoring function is constructed, which includes crop yield target and soil erosion resistance target. The original set of soil attribute variables is input into the dual-objective comprehensive scoring function to obtain the comprehensive score value of cultivated land in the target area. Based on the comprehensive score of cultivated land in the target area, the natural breakpoint method is used to classify the topsoil level of cultivated land in the target area. Principal component analysis was used to extract the main influencing factors from the original soil attribute variable set to obtain the first variable set. A LASSO regression model was constructed based on the comprehensive score. The first variable set and the original variable set were filtered to obtain the corresponding second variable set and third variable set. The random forest algorithm was used to filter the variables in the original soil attribute variable set to obtain a fourth variable set; Based on the intersection comparison of the second, third and fourth variable sets, the first-level key variable set, the second-level key variable set and the third-level key variable set are determined respectively. By combining the topsoil grade of the target area's cultivated land with the identified primary, secondary, and tertiary key variable sets, a reasonable topsoil evaluation index system for the target area is constructed.

[0008] Furthermore, the multi-source soil data includes spatial data of soil properties, meteorological data, and crop yield data; the spatial data of soil properties includes at least: pH value, bulk density, cation exchange capacity, soil volumetric water content, soil available water content, soil temperature, total nitrogen reserves, total phosphorus reserves, and total potassium reserves; the meteorological data includes at least: average rainfall, rainfall erosivity, and average temperature data.

[0009] Furthermore, the multi-source soil data is standardized and subjected to multicollinearity diagnosis to obtain the original set of soil attribute variables. Specifically, the multi-source soil data is standardized, and the Pearson correlation coefficient and variance expansion factor are used to diagnose multicollinearity in the standardized multi-source soil data.

[0010] Furthermore, a dual-objective comprehensive scoring function is constructed, which includes crop yield and soil erosion resistance targets, and is expressed as: ; in, This is the overall score. Assign weights to production targets. Normalized yield value for crops. This represents the normalized value of soil erodibility obtained from multi-source soil data.

[0011] Furthermore, the natural breakpoint method is used to classify the topsoil grades of the cultivated land in the target area, specifically as follows: The classification interval for the comprehensive score is determined using the natural breakpoint method, and the specific classification interval is as follows: When the comprehensive score is greater than or equal to the first set threshold, the topsoil grade of the cultivated land in the target area is excellent. When the comprehensive score is greater than or equal to the second set threshold and less than the first set threshold, the topsoil grade of the cultivated land in the target area is good. When the comprehensive score is greater than or equal to the third set threshold and less than the second set threshold, the topsoil grade of the cultivated land in the target area is medium. When the comprehensive score is less than the third set threshold, the topsoil grade of the cultivated land in the target area is poor.

[0012] Furthermore, the method for obtaining the second set of variables is as follows: Using the comprehensive score as the response variable and the first set of variables as the feature variables, a LASSO regression model is constructed. The optimal penalty coefficient for the LASSO regression model was determined using cross-validation. The first variable set is regularized according to the LASSO regression model with the optimal penalty parameters, and the variables with non-zero regression coefficients in the first variable set are used to construct the second variable set.

[0013] Furthermore, based on the intersection comparison of the second, third, and fourth variable sets, the first-level key variable sets, second-level key variable sets, and third-level key variable sets are determined, including: Obtain the variables that exist simultaneously in the second, third, and fourth variable sets to obtain the first-level key variable set; Obtain the variables that exist simultaneously in any two of the second, third, and fourth variable sets to obtain the secondary key variable set; Obtain all variables from the second, third, and fourth variable sets to obtain the three-level key variable sets.

[0014] The beneficial effects of this invention are as follows: By constructing a dual-objective comprehensive scoring function, this invention overcomes the limitations of the single-output orientation in traditional tillage evaluation, achieving a scientific quantification of the comprehensive function of the tillage layer and ensuring the scientific nature and objectivity of the evaluation indicators. It employs principal component analysis, LASSO regression, and random forest algorithms to collaboratively screen the variable set, combining the advantages of dimensionality reduction, linear sparsity, and nonlinear importance ranking methods. This objectively and robustly identifies the core driving factors that truly affect tillage quality from the original data, avoiding subjective assumptions and indicator redundancy. Through multi-algorithm intersection comparison, a three-level variable indicator system is formed. Based on the comprehensive score value, the natural breakpoint method is used to divide the tillage quality into four levels: excellent, good, medium, and poor. This ensures that the evaluation results not only reflect the macroscopic spatial distribution of tillage quality but also provide specific diagnostic criteria and differentiated, precise governance directions for different grade areas, forming a complete closed loop from evaluation to diagnosis to governance. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of a method for constructing a reasonable topsoil evaluation index system based on dual-objective optimization according to an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the results of Pearson correlation coefficient analysis on multi-source soil data according to an embodiment of the present invention. Figure 3 This is a schematic diagram showing the result of multicollinearity diagnosis processing of a variance inflation factor according to an embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the result of processing the original soil attribute variable set using principal component analysis according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the feature importance analysis results according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the result of sparse regression screening of the first variable set using the LASSO regression model according to an embodiment of the present invention. Figure 7 This is a schematic diagram illustrating the evaluation results of an embodiment of the present invention, which uses a random forest algorithm to evaluate the nonlinear explanatory contribution of each variable in the original soil attribute variable set to the comprehensive score value. Figure 8This is a schematic diagram showing the result of an intersection comparison of a second set of variables, a third set of variables, and a fourth set of variables according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The present invention proposes a method for constructing a rational tillage layer evaluation index system based on dual-objective collaborative optimization. This method aims to quantitatively identify the rationality of the tillage layer structure (i.e., the synergistic ability between high and stable yields) through a "data-driven + dual-objective optimization + multi-algorithm collaborative screening" approach. A flowchart illustrating the method for constructing a rational tillage layer evaluation index system based on dual-objective optimization in an embodiment of the present invention is shown below. Figure 1 As shown, it includes: Multi-source soil data of cultivated land in the target area are acquired and preprocessed to obtain the original set of soil attribute variables; In this embodiment of the invention, the target area is a typical sloping farmland in the red soil region of southern China. The terrain of the region is mainly low mountains and hills, and the parent material of the soil is mainly granite and sandstone and shale. Red soil, brick red soil and yellow red soil are developed. The climate type is subtropical humid monsoon climate. The combination of water and heat and the topographic conditions jointly induce strong water erosion and shallow topsoil problems.

[0019] In this embodiment of the invention, multi-source soil data refers to spatial data of soil properties, meteorological data, and crop yield data. This data can be obtained by combining data from the National Soil Survey, the National Earth System Science Data Center, and geospatial data platforms, covering the topography, climate, and topsoil characteristics of the target area. Crop yield data can be obtained from local agricultural statistical yearbooks of the target area. Specifically, as shown in Table 1, the crop data in this embodiment uses maize yield data as an example. Spatial data of soil properties includes at least: pH value, bulk density, cation exchange capacity, soil volumetric water content, soil available water content, soil temperature, total nitrogen reserves, total phosphorus reserves, and total potassium reserves. Meteorological data includes at least: average rainfall, rainfall erosivity, and average temperature data. The data on clay, sand, and silt given in Table 1 are used for subsequent calculations of the normalized soil erosibility value when constructing a dual-objective function; therefore, although this data is also obtained here, it is not considered as a variable in the original soil property variables.

[0020] Table 1. Schematic diagram of multi-source soil data ; In this embodiment of the invention, the preprocessing method refers to standardizing the multi-source soil data and using Pearson correlation coefficient and variance expansion factor to perform multicollinearity diagnosis on the standardized multi-source soil data. Specifically, the standardization process adopts the Min-Max standardization method to linearly transform the multi-source soil data to the [0,1] interval to eliminate the influence of dimensional differences on the selection of subsequent indicators. Then, Pearson correlation coefficient and variance expansion factor are used to perform multicollinearity diagnosis on the multi-source soil data.

[0021] In a specific embodiment of the present invention, the results of multicollinearity diagnosis processing of multi-source soil data using Pearson correlation coefficient and variance expansion factor are as follows: Figure 2 and Figure 3 As shown, where, Figure 2 The results of the Pearson correlation coefficient analysis are as follows. Figure 3 Based on the variance expansion factor diagnosis results and Pearson correlation coefficient analysis, it can be seen that the spatial data of soil properties in the multi-source soil data show varying degrees of correlation. Among them, soil temperature (ST) and multi-year average air temperature (T) have a significant positive correlation, while bulk density (BD) and soil volumetric water content (VWC) have a significant negative correlation. Soil volumetric water content (VWC), bulk density (BD), and total phosphorus reserves (PS) are also correlated. The correlation with the target variable was low and not significant. Combined with the diagnostic results of variance expansion factor, the variance expansion factor values ​​of pH, cation exchange capacity (CEC), available soil water content (AWC), total nitrogen stock (NS), total phosphorus stock (PS), total potassium stock (KS), multi-year average rainfall (P), and multi-year average temperature (T) were all less than 5, which provided a statistical basis for entering the subsequent process. Finally, combining the combined analytical and diagnostic results of both, pH, cation exchange capacity (CEC), available soil water content (AWC), soil temperature (ST), total nitrogen stock (NS), total potassium stock (KS), multi-year average rainfall (P), rainfall erosivity (R), and score S were selected as key variables to construct the original set of soil attribute variables.

[0022] A dual-objective comprehensive scoring function is constructed, and the original set of soil attribute variables is input into the dual-objective comprehensive scoring function to obtain the comprehensive score value of cultivated land in the target area; The dual-objective comprehensive scoring function constructed in this embodiment of the invention is expressed as follows: ; in, This is a comprehensive score, which reflects the overall rationality level of the region's topsoil. The production target weight is set to 0.5 in this embodiment of the invention to balance output and corrosion resistance requirements; The normalized yield value is the crop yield value. In this embodiment of the invention, this value is the maize yield value of the target area. To obtain the normalized value of soil erodibility based on multi-source soil data, the soil erodibility value is calculated using the nomograph equation. The parameters selected in the equation are the key variables corresponding to the original soil attribute variable set extracted in this embodiment of the invention.

[0023] In a specific embodiment of the present invention, in order to determine the optimal weight combination in the dual-objective comprehensive scoring function, a dual-objective collaborative optimization strategy is adopted. That is, Pareto constraint boundaries for yield improvement and corrosion resistance enhancement are set in the sample space, and the optimal solution set is searched using a non-dominated sorting genetic algorithm. The optimal weight combination can be obtained by selecting the weight combination corresponding to the point that minimizes the comprehensive variance.

[0024] In this embodiment of the invention, relevant data on soil erodibility were referenced, and the EPIC model was used to determine soil erodibility factors. Value estimation: ; In the formula: Sand particle content (2~0.05mm); Particle size (0.05mm~0.002mm); The content of clay particles is <0.002 mm. Organic carbon content; =1- / 100, the K values ​​calculated by the EPIC model are all in US units. Multiplying the K value by 0.1317 gives the SI unit K value. K is in SI units (t·hm). 2 ·h / (hm 2 ·MJ·mm)).

[0025] In a specific embodiment of the present invention, the comprehensive score of the cultivated layer of the target area is obtained according to the bi-objective comprehensive scoring function. According to the overall distribution characteristics, the comprehensive score values ​​of all sampling points in the target area are between 0 and 1, with a mean of 0.49, a standard deviation of 0.19, a median of 0.49, and a coefficient of variation of 37.7%. The score distribution is unimodal and skewed to normal, indicating that the cultivated layer of the target area is at a medium level overall, with a few areas having significant advantages or disadvantages. The distribution stability is good, which provides a mathematical basis for establishing a grading standard.

[0026] Based on the comprehensive score of cultivated land in the target area, the natural breakpoint method is used to classify the topsoil level of cultivated land in the target area. In this embodiment of the invention, the classification interval of the comprehensive score value is determined according to the comprehensive score value obtained from each sampling point in the target area, based on the natural breakpoint method. Specifically, when the comprehensive score value is greater than or equal to a first set threshold, the topsoil grade of the cultivated land in the target area is excellent; when the comprehensive score value is greater than or equal to a second set threshold and less than the first set threshold, the topsoil grade of the cultivated land in the target area is good; when the comprehensive score value is greater than or equal to a third set threshold and less than the second set threshold, the topsoil grade of the cultivated land in the target area is medium; and when the comprehensive score value is less than the third set threshold, the topsoil grade of the cultivated land in the target area is poor.

[0027] In a specific embodiment of the present invention, in order to achieve graded identification and regionalized management of the cultivated layer in the target area, the cultivated layer is divided into four grades based on the comprehensive score value. The first set threshold value is 0.59, that is, when the comprehensive score value is greater than or equal to 0.59, the cultivated land in the target area can be considered to be of excellent grade. The second set threshold value is 0.51, and the third set threshold value is 0.43. The above set threshold values ​​are automatically generated by the algorithm and can be adjusted according to the actual area in specific implementation. The proportions of each grade in the sampling points of the target area are 16.0%, 28.2%, 36.2%, and 19.6%, respectively. The overall score distribution pattern shows a median concentration and stretching at both ends, which has a good grade differentiation degree.

[0028] In one specific embodiment of the present invention, a reasonable tillage layer diagnosis-management suggestion table is constructed based on the key variable value ranges of each tillage layer, as shown in Table 2: Table 2 ; By classifying the variables according to their importance as shown in Table 2, differentiated directions for topsoil improvement are provided.

[0029] Principal component analysis was used to process the original soil attribute variable set, and the first variable set was obtained based on the processed original soil attribute variable set. In this embodiment of the invention, principal component analysis was used to process the original soil attribute variable set, and the results are as follows: Figure 4 and Figure 5As shown, the eigenvalues ​​of the first four principal components are all greater than 1, with a cumulative variance contribution rate of 87.24%, which can effectively explain most of the information of the original indicators. Among them, the positive direction of the first principal component (PC1) is mainly driven by soil nutrients (NS, KS) and climate factors (P), while the negative direction is driven by soil physical properties and water retention capacity (AWC). The second principal component (PC2) mainly reflects temperature stress and buffering capacity, dominated by cation exchange capacity (CEC). Total potassium reserves and total nitrogen reserves have moderate projections on both the first and second principal components, reflecting their dual role in structural stability and nutrient supply. NS, KS, P and R show a positive correlation. Principal component analysis shows that the above nine indicators are evenly distributed, directionally dispersed, and have good structural independence. There is no redundant clustering, making them suitable for subsequent machine learning models to rank variable importance. Features with feature importance greater than the 45% threshold are selected to form the first variable set.

[0030] A LASSO regression model is constructed using the comprehensive score of cultivated land in the target area as the response variable. The first set of variables is input into the LASSO regression model, and the first set of variables is filtered using a regularization constraint method to obtain the second set of variables. In a specific embodiment of the present invention, the LASSO regression model can be expressed as: ; in, This is the overall score, and here it represents the response variable. This is the first set of variables, which here refers to the feature variables. The penalty coefficient is... This is the coefficient vector.

[0031] In this embodiment of the invention, a LASSO regression model is constructed using the comprehensive score as the response variable and the first variable set as the feature variable. At the same time, the optimal penalty coefficient in the LASSO regression model is determined by the ten-fold cross-validation method. The first variable set is then regularized according to the LASSO regression model under the optimal penalty parameter, and the variables with non-zero regression coefficients in the first variable set are used to construct the second variable set.

[0032] Input the original set of soil property variables into the LASSO regression model to obtain the third set of variables; In one specific embodiment of the present invention, the optimal penalty coefficient value of 0.001 for the LASSO regression model was determined using cross-validation. Sparse regression screening was then performed on the original set of soil attribute variables, and the results are as follows: Figure 6As shown, the regression coefficients of pH, available soil moisture content (AWC), soil temperature (ST), and multi-year average rainfall (P) under this penalty coefficient are non-zero, indicating that they have a significant linear explanatory power for the changes in the comprehensive evaluation value. Thus, a third variable set was constructed using these four variables.

[0033] Based on the comprehensive score of cultivated land in the target area and the original set of soil attribute variables, a fourth set of variables is obtained using the random forest algorithm.

[0034] In this embodiment of the invention, the random forest algorithm uses a regression tree ensemble to evaluate the nonlinear explanatory contribution of each variable in the original soil attribute variable set to the comprehensive score value, setting the variable importance threshold to 45%. The evaluation results are as follows: Figure 7 As shown, two variables were selected: pH and available soil water content (AWC). Their average reduction in Gini index was significantly higher than other variables, thus allowing the construction of a fourth variable set. pH and AWC were identified in the PCA-LASSO method, LASSO regression model, and random forest tree, demonstrating their stabilizing role in both structural and erosion-resistant functions. The results from the three methods were generally consistent, with some core indicators showing high weights in multiple algorithms, validating their representativeness under the dual objectives of "high and stable yield - erosion resistance".

[0035] By comparing the intersections of the second, third, and fourth variable sets, the key variable sets at three levels are determined. In embodiments of the present invention, such as Figure 8 As shown, the variables obtained by combining the second, third, and fourth variable sets are used to obtain the first-level key variable set by acquiring variables that exist simultaneously in the second, third, and fourth variable sets; variables that exist simultaneously in any two of the second, third, and fourth variable sets are acquired to obtain the second-level key variable set; and all variables in the second, third, and fourth variable sets are acquired to obtain the third-level key variable set.

[0036] In one specific embodiment of the present invention, a three-level key variable set classification framework based on methodological consistency is constructed by integrating the second, third, and fourth variable sets. Among them, pH and available soil water content (AWC) are the only variables that are stably identified in all variable sets, exhibiting high consistency and key functionality, and are thus classified as key variable sets. Soil temperature (ST) is selected in any two variable sets, possessing moderate consensus, and is therefore classified as a secondary key variable set. In addition, although cation exchange capacity (CEC) and multi-year average rainfall (P) are significant only in a single method, they have certain functional value, and together with the above indicators, constitute a tertiary key variable set. This classification system helps to clarify the hierarchical role of different variables in the rational tillage layer evaluation system, providing an indicator basis and technical support for the subsequent construction of grading standards and regional suitability recommendations.

[0037] Based on the topsoil level of the target area and the key variable set of the three levels, a reasonable topsoil evaluation index system for the target area is constructed.

[0038] Based on the topsoil grade distribution in the target area, the superior topsoil layer exhibits strong yield-erosion resistance coupling advantages. Its typical soil characteristics are: pH value between 5.16 and 6.05, slightly acidic; cation exchange capacity of 14.12–16.00 cmol(+) / kg; available soil moisture content of 11.46–13.12%; soil temperature of 9.90–14.15℃; and average annual rainfall of 1804.62–1945.77 mm, demonstrating good water and fertilizer buffering capacity and root adaptability. The comprehensive score of the good topsoil layer is mainly concentrated in the range of 0.51–0.59. In this region, the pH and cation exchange capacity are slightly lower or higher than those of the superior topsoil layer, while the available soil moisture content is slightly higher. Soil temperature and rainfall are lower than those of the superior topsoil layer, but the overall topsoil structure is still good. It possesses high plasticity and regulation potential; the soil properties of medium-grade topsoil are at the critical level: such as cation exchange capacity mostly between 17.20 and 19.40 cmol(+) / kg, effective soil water content higher than 14.20%, pH value fluctuates greatly, the overall topsoil stability is poor, and the erosion resistance and nutrient supply capacity are both weak; poor-grade topsoil is generally shallow, with significantly insufficient organic matter and water retention capacity, cation exchange capacity mostly higher than 19.40 cmol(+) / kg, effective soil water content in the range of 15.00 to 16.90%, pH value generally strongly acidic or neutral, soil structure is compact, nutrient availability is poor, which seriously restricts crop root development and yield stability.

[0039] For superior-grade topsoil, no-till mulching is used to reduce disturbance, organic matter is used to maintain and enhance the carbon pool, and precise water and fertilizer management is combined to optimize resource utilization. Structural damage such as mechanical compaction is avoided, prioritizing conservation and maintaining the stability of the existing soil structure and function. For good-grade topsoil, straw return to the field enhances organic input, inter-row cultivation improves aeration, and moderate deep plowing breaks up the plow pan for structural optimization, supplemented by chemical improvement methods such as pH adjustment to continuously enrich the soil and improve its function. For medium-grade topsoil, targeted deep improvement and topsoil reconstruction are implemented. Deep loosening combined with deep application breaks down the barrier layer, humus is added to enhance aggregate stability, compound biological amendments are used to regulate the microbial community, and cropping systems are adjusted to meet ecological needs. For poor-grade topsoil, multi-dimensional comprehensive management is used to restore and reconstruct the topsoil, removing obstacles such as salinity / acidification, implementing terrace restoration projects, using organic-mineral synergistic fertilization to quickly replenish nutrients, and supporting slope protection projects to control erosion, forming a systematic restoration plan. The zoning management strategy proposed in this invention reflects a tiered management logic from "preventive maintenance" to "rescue and restoration," providing a hierarchical decision-making framework for sustainable soil utilization.

[0040] This invention further extracts indicators from the key variable set of each level of tillage layer to construct a management-evaluation threshold reference table, as shown in Table 3: Table 3 ; As can be seen, the pH of the "excellent" grade is slightly acidic, suitable for the growth of most crops, while the "poor" grade tends to be overly acidic or neutral. High pH values ​​reduce the availability of micronutrients. The cation exchange capacity of the "excellent" grade is 14.12~16.00 cmol(+) / kg. An appropriate cation exchange capacity means that the soil can effectively retain potassium. + NH4 + Nutrient ions provide a continuous supply of nutrients to plants. Soil effective water content directly affects the plant's water supply, thus determining the level of primary productivity. Excellent and good soils have moderate effective water content, while medium and poor soils have higher levels. Excellent and good soils have slightly higher temperatures than medium and poor soils, but overall, temperature fluctuations are smaller and tend to be more stable, which is conducive to microbial activity and root growth. Excellent soils have higher average annual rainfall, and abundant rainfall is beneficial to plant development and growth.

[0041] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a reasonable topsoil evaluation index system based on dual-objective optimization, characterized in that, include: Acquire multi-source soil data for cultivated land in the target area; The multi-source soil data were standardized and subjected to multicollinearity diagnostic processing to obtain the original set of soil attribute variables; A dual-objective comprehensive scoring function is constructed, which includes crop yield target and soil erosion resistance target. The original set of soil attribute variables is input into the dual-objective comprehensive scoring function to obtain the comprehensive score value of cultivated land in the target area. Based on the comprehensive score of cultivated land in the target area, the natural breakpoint method is used to classify the topsoil level of cultivated land in the target area. Principal component analysis was used to extract the main influencing factors from the original soil attribute variable set to obtain the first variable set. A LASSO regression model was constructed based on the comprehensive score. The first variable set and the original variable set were filtered to obtain the corresponding second variable set and third variable set. The random forest algorithm was used to filter the variables in the original soil attribute variable set to obtain a fourth variable set; Based on the intersection comparison of the second, third and fourth variable sets, the first-level key variable set, the second-level key variable set and the third-level key variable set are determined respectively. By combining the topsoil grade of the target area's cultivated land with the identified primary, secondary, and tertiary key variable sets, a reasonable topsoil evaluation index system for the target area is constructed.

2. The method for constructing a reasonable tillage layer evaluation index system based on dual-objective optimization according to claim 1, characterized in that: The multi-source soil data includes spatial data of soil properties, meteorological data, and crop yield data; the spatial data of soil properties includes at least: pH value, bulk density, cation exchange capacity, soil volumetric water content, soil available water content, soil temperature, total nitrogen reserves, total phosphorus reserves, and total potassium reserves; the meteorological data includes at least: average rainfall, rainfall erosivity, and average temperature data.

3. The method for constructing a reasonable tillage layer evaluation index system based on dual-objective optimization according to claim 1, characterized in that: The multi-source soil data is standardized and subjected to multicollinearity diagnosis to obtain the original set of soil attribute variables. Specifically, the multi-source soil data is standardized, and the Pearson correlation coefficient and variance expansion factor are used to diagnose multicollinearity in the standardized multi-source soil data.

4. The method for constructing a reasonable tillage layer evaluation index system based on dual-objective optimization according to claim 1, characterized in that: A dual-objective integrated scoring function, comprising crop yield and soil erosion resistance targets, is constructed and expressed as follows: ； in, This is the overall score. Assign weights to production targets. Normalized yield value of crop This represents the normalized value of soil erodibility obtained from multi-source soil data.

5. The method for constructing a reasonable tillage layer evaluation index system based on dual-objective optimization according to claim 1, characterized in that: The natural breakpoint method was used to classify the topsoil grades of the cultivated land in the target area, specifically as follows: The classification interval for the comprehensive score is determined using the natural breakpoint method, and the specific classification interval is as follows: When the comprehensive score is greater than or equal to the first set threshold, the topsoil grade of the cultivated land in the target area is excellent. When the comprehensive score is greater than or equal to the second set threshold and less than the first set threshold, the topsoil grade of the cultivated land in the target area is good. When the comprehensive score is greater than or equal to the third set threshold and less than the second set threshold, the topsoil grade of the cultivated land in the target area is medium. When the comprehensive score is less than the third set threshold, the topsoil grade of the cultivated land in the target area is poor.

6. The method for constructing a reasonable tillage layer evaluation index system based on dual-objective optimization according to claim 1, characterized in that, The method for obtaining the second set of variables is as follows: Using the comprehensive score as the response variable and the first set of variables as the feature variables, a LASSO regression model is constructed. The optimal penalty coefficient for the LASSO regression model was determined using cross-validation. The first variable set is regularized according to the LASSO regression model with the optimal penalty parameters, and the variables with non-zero regression coefficients in the first variable set are used to construct the second variable set.

7. The method for constructing a reasonable tillage layer evaluation index system based on dual-objective optimization according to claim 1, characterized in that: Based on the intersection comparison of the second, third, and fourth variable sets, the first-level key variable sets, second-level key variable sets, and third-level key variable sets are determined, including: Obtain the variables that exist simultaneously in the second, third, and fourth variable sets to obtain the first-level key variable set; Obtain the variables that exist simultaneously in any two of the second, third, and fourth variable sets to obtain the secondary key variable set; Obtain all variables from the second, third, and fourth variable sets to obtain the three-level key variable sets.

Images