Soil barrier layer distribution diagram, drawing method thereof and computer equipment

Abstract

The invention belongs to the technical field of digital soil mapping, and particularly relates to a soil barrier layer distribution diagram, a mapping method thereof and computer equipment. According to the method, the soil barrier layer is drawn through the historical remaining soil data in combination with the expert knowledge base, any field investigation work does not need to be carried out, and the investment of manpower and material resources is reduced. Compared with a traditional method, the cost is greatly reduced, and the efficiency is also improved due to the fact that field sampling is not needed. The technical problem that in the prior art, a soil drawing method is high in cost and low in efficiency is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital soil mapping, and in particular relates to a soil barrier layer distribution map, a mapping method thereof, and a computer device. Background Art

[0002] A soil barrier layer refers to a layer of soil with abnormal physical or chemical properties, typically characterized by compaction, hardening, or high salinity. The presence of a barrier layer can impede root growth, soil water infiltration, and nutrient absorption, thereby reducing crop yields, increasing water stress, and impacting the sustainability of agricultural production. Clarifying the spatial distribution of soil barrier layers is crucial for tailoring soil management to local and soil conditions.

[0003] Digital soil mapping technology utilizes geographic information systems (GIS), remote sensing, and statistical methods to create maps of soil distribution and properties. This method can obtain highly accurate and detailed soil information to support decision-making in land management, agricultural planning, and environmental protection, thereby improving resource utilization efficiency and sustainable development. However, traditional digital soil mapping relies heavily on field surveys within the mapping area. The high cost and low efficiency of field surveys make it difficult to meet the needs of modern soil survey research and application. Therefore, exploring effective and rational alternative soil point data acquisition methods is crucial for reducing the investment in human and material resources, lowering the barrier to application of DSM (Digital Surface Model) technology, improving the efficiency of soil classification, and ultimately promoting the production of high-resolution and high-accuracy digital soil maps. Summary of the Invention

[0004] The purpose of the present invention is to provide a soil barrier layer distribution map and its mapping method and computer equipment to solve the technical problem of high cost and low efficiency of soil mapping methods in the prior art.

[0005] To solve the above technical problems, the present invention provides a method for mapping a soil barrier layer distribution map. The method comprises:

[0006] 1) Based on expert knowledge, obtain the probability of the existence of a soil barrier layer in each mapping unit within the mapping area and the environmental factors that affect the formation of the soil barrier layer; and obtain the environmental data corresponding to each mapping unit and the environmental factors within the mapping area;

[0007] 2) Sampling all mapping units to obtain N sample units;

[0008] According to the probability of the presence of a soil barrier layer in each sample unit, labels are labeled on N sample units to obtain a classified data set including the environmental data of each sample unit, wherein the label indicates whether the sample unit has a soil barrier layer;

[0009] A soil barrier layer classification model is constructed based on the classification data set; the environmental data of each mapping unit is input into the soil barrier layer classification model to obtain the result of whether a soil barrier layer exists in each mapping unit;

[0010] 3) Repeat 2) P times, and obtain the probability of the existence of a soil barrier layer in a certain mapping unit based on the number of occurrences of the result that the soil barrier layer exists in the mapping unit.

[0011] The beneficial effects of the above technical solution are as follows: The technical solution of the present invention, a method for mapping the distribution of soil barrier layers, is an improved invention. This invention utilizes historical soil data combined with an expert knowledge base to map soil barrier layers, eliminating the need for any field surveys and reducing the investment in human and material resources. Compared to traditional methods, this method not only significantly reduces costs but also improves efficiency by eliminating the need for field sampling. This solves the technical problem of high cost and low efficiency in existing soil mapping methods.

[0012] Furthermore, the classified data set is obtained in the following manner: N sample units are divided into two categories according to whether there is a soil barrier layer in each sample unit and the environmental data of each sample unit through a clustering algorithm, and the category to which each sample unit belongs is the label of the sample unit.

[0013] Furthermore, the classification data set is M groups of data sets obtained by repeatedly executing the clustering algorithm M times.

[0014] Furthermore, the clustering algorithm is a constrained K-means clustering algorithm.

[0015] Furthermore, in 2), a stratified sampling strategy based on natural break points is adopted to obtain sample units.

[0016] Furthermore, sampling is performed in 2) in the following manner: the soil types in the mapping area are divided into multiple levels based on the area of each soil type in the mapping area, and the number of sample units for each level is determined according to the total area of the soil types in each level.

[0017] Furthermore, the soil barrier layer classification model is obtained in the following manner: based on the classification data set, L classification models are constructed for predicting whether there is a soil barrier layer in each sample unit, and the classification model that appears most frequently in the L times is used as the final soil barrier layer classification model.

[0018] Furthermore, the classification model is a random forest model.

[0019] Furthermore, in 3), the probability of a soil barrier layer existing in a certain mapping unit is obtained in the following way: the proportion of a soil barrier layer existing in a certain mapping unit in P sub-classifications is used as the probability of a soil barrier layer existing in the mapping unit.

[0020] Furthermore, the expert knowledge is obtained in the form of a questionnaire survey.

[0021] The present invention also provides a technical solution for a computer device: a computer device comprising a processor, wherein the processor is configured to execute a computer program to implement the steps of the soil barrier layer distribution map making method as described above.

[0022] The present invention also provides a soil barrier layer distribution map: a soil barrier layer distribution map, wherein the soil barrier layer distribution map is obtained by the soil barrier layer distribution map making method as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of an embodiment of a method for mapping a soil barrier layer distribution map according to the present invention;

[0024] Figure 2 A schematic diagram of a study area of an embodiment of a soil barrier layer distribution mapping method of the present invention;

[0025] Figure 3 The sub-class soil type map of the typical black soil area in Northeast China in the embodiment of the method for mapping the distribution of soil barrier layers of the present invention;

[0026] Figure 4 The result labels of four predictions in the embodiment of the soil barrier layer distribution mapping method of the present invention illustrate an example of the integration process of the fuzzy probability of a certain barrier layer;

[0027] Figure 5-1 The white pulp layer mapping result of the soil barrier layer distribution mapping method embodiment of the present invention;

[0028] Figure 5-2 The calcium accumulation layer mapping result of the soil barrier layer distribution mapping method embodiment of the present invention;

[0029] Figure 5-3 The alkali accumulation layer mapping result of the soil barrier layer distribution mapping method embodiment of the present invention;

[0030] Figure 5-4 The result of mapping the eutrophic layer in the embodiment of the method for mapping the distribution of the soil barrier layer according to the present invention;

[0031] Figure 5-5 The salt accumulation layer mapping result of the soil barrier layer distribution mapping method embodiment of the present invention;

[0032] Figure 5-6 The clay layer mapping result of the soil barrier layer distribution mapping method according to the present invention;

[0033] Figure 6 This is a schematic diagram of the verification point set in an embodiment of the soil barrier layer distribution map making method of the present invention. DETAILED DESCRIPTION

[0034] Existing digital soil mapping techniques rely heavily on the need for field surveys to measure points within the mapping area. The high cost and low efficiency of field surveys make it difficult to meet the needs of modern soil survey research and application. This present invention utilizes historical soil data combined with an expert knowledge base to map soil barrier layers, eliminating the need for field surveys and reducing the investment in human and material resources. Compared to traditional methods, this method significantly reduces costs and, because field sampling is not required, improves efficiency. This solves the technical problem of high cost and low efficiency in existing soil mapping methods.

[0035] Example of a method for mapping soil barrier layer distribution map:

[0036] This embodiment takes the Northeast black soil area as an example to introduce the soil barrier layer distribution map making method of the present invention. Figure 2 As shown in Figure 2, the typical black soil area in Northeast China (117°04′9″E~32°48′50″E, 42°57′43″N~50°46′4″N) is located in Northeast China, spanning Heilongjiang Province, Jilin Province, Liaoning Province and Inner Mongolia Autonomous Region, including 138 county-level administrative regions with a total area of 332,954 km 2 According to the second national land survey data and county-level farmland quality survey results, the cultivated land area in the typical black soil region of Northeast China is about 1.85×10 7 hm 2 It is an important grain production area and agricultural base in China.

[0037] Legacy soil data refers to the results such as books, legacy soil maps and legacy point information obtained and compiled by soil surveyors in the past. It contains rich soil data and expert knowledge and is regarded as an important soil survey resource. Among them, legacy soil maps are accumulated by soil surveyors in the past through a large amount of field survey work. They contain a large amount of expert knowledge about the empirical relationship between soil and environment in the region and have now become the most widely used legacy soil survey results. The present invention proposes a soil barrier layer mapping method that integrates historical data and expert knowledge. The purpose is to achieve rapid mapping of soil barrier layers without the need for new survey samples, thereby meeting the soil information needs of modern agricultural production and ecological protection.

[0038] A method for mapping the distribution of soil barrier layers, such as Figure 1 As shown, the method includes:

[0039] S1. Obtain a vectorized historical soil type map of the target mapping area of the soil barrier layer, use the raster-to-point method to obtain all grid points, and further complete the point division based on the basic mapping units of the historical soil map.

[0040] like Figure 3 As shown in Figure 1, in this example, the selected legacy soil map is a 1:1,000,000 national soil type map based on soil genetic classification during the second soil census, using the WGS 1984 coordinate system. The study area is divided into 8 soil orders, 15 suborders, 19 soil classes, and 59 subclasses. The subclass level, the base unit of the historical soil map, is used for subsequent point selection. The "Raster to Point" function in ArcGIS software was used to convert the subclass soil map into 1 km resolution point data.

[0041] S2. Based on historical data and soil professionals, a probabilistic expert knowledge base containing soil barrier layers in the basic mapping units of historical soil maps is constructed.

[0042] The expert knowledge base should contain two aspects of knowledge: (i) the probability of certain obstacle layers existing in each sub-type polygon, and (ii) the main environmental factor types of each obstacle layer.

[0043] This example is based on soil data from typical sample points of each soil sub-type unit in the soil data left over from the Second National Soil Survey (i.e., the Second National Soil Survey). It comprehensively utilizes soil textbooks such as "China Soils," "Heilongjiang Soils," "Jilin Soils," "Liaoning Soils," and "Inner Mongolia Soils," expert knowledge, and existing soil genetic classification literature. In accordance with the established indicator specifications, probability analysis principles, integer principles, and identifiability principles, a probabilistic matching relationship between soil types and barrier layer information in the historical soil map is formed to construct a soil knowledge base for the study area.

[0044] The expert knowledge was provided by a panel of five national and provincial experts from the Third National Soil Survey of China. All experts were proficient in Chinese soil science and had conducted extensive soil surveys in the study area. At the seminar, the experts were first briefed on the need for expert knowledge. They then discussed and screened out six potential barrier layer types in the study area: the white pulp layer, the calcareous layer, the alkali layer, the tephra layer, the salt layer, and the clay layer. The experts then provided the probability (0-1) of the corresponding barrier layer in each sub-type of soil using a questionnaire. In addition, the questionnaire also included a voting interface for the main controlling environmental factors of each barrier layer.

[0045] S3. Construct a vectorized environmental database for soil barrier layer mapping and use the multi-value extraction to point method to extract the environmental variables corresponding to all grid points in S1.

[0046] As shown in Table 1, a total of 22 environmental factors influencing the development of the soil barrier layer in the study area were extracted, which can be summarized into four categories: topography, climate, vegetation, and land use. Environmental data for these four categories within the study area were obtained from SRTM (Shuttle Radar Topography Mission), WorldClim, MODIS (Moderate Resolution Imaging Spectroradiometer), and CNLUCC (China Multi-period Land Use Remote Sensing Monitoring Dataset), respectively. Image processing was performed in ArcGIS software, including cropping and stitching, resampling to a 1 km resolution, and converting the geographic coordinate system to WGS-84. The "Extract Multi-Values to Points" tool in ArcGIS was used to map the environmental data to each raster cell and present them in its attribute table.

[0047] Table 1

[0048]

[0049] In other embodiments, environmental data may not be limited to the four categories of topography, climate, vegetation, and land use, but may also include other environmental factors, such as soil parent material; or a soil barrier layer distribution map may be drawn based on only some of the factors among topography, climate, vegetation, and land use.

[0050] S4. Taking the number of grid points in each basic mapping unit of the historical soil map in S1 as the object, the data classification algorithm is used to divide the mapping area, and then the number of virtual sample points to be extracted is determined and sampling is carried out.

[0051] This example uses a stratified sampling strategy based on natural breakpoints to obtain virtual sample points (i.e., sample units). Specifically, the 59 soil subclasses were divided into 10 levels based on the grid area of the soil patches, with each level covering 5 to 7 soil subclasses. The number of random samples was determined based on the total number of soil grid units (i.e., mapping units) in each level, ranging from 5 to 500. A total of 9,570 sample units belonging to different soil subclasses (i.e., N sample units, N = 9,570) were obtained.

[0052] In other embodiments, sampling can be based on other soil types, such as soil classes, soil subclasses, etc. The soil type classification system can be either a soil generation classification or a systematic classification.

[0053] S5. Based on the expert knowledge base in S2 and the sample units extracted in S4, a classification dataset of each soil barrier layer is obtained through constrained K-means clustering.

[0054] Based on the presence probabilities and environmental data for six typical barrier layers corresponding to each soil subtype in the soil knowledge database, constrained K-means clustering was used to assign barrier layers to all extracted sample units. Clustering was performed once for each barrier layer, resulting in two clusters: sample units that did not belong to that barrier layer and sample units that did.

[0055] To ensure the adequacy and reliability of the data, the entire clustering process was repeated 100 times (i.e., M times) to provide a data basis for constructing the probability surface of the barrier layer information. Each soil subtype obtained 100 sample unit sets, and each sample point set contained the latitude and longitude and main controlling environmental factor information of 9570 sample units. After processing, each barrier layer obtained 100 classification data sets containing the above-mentioned 59 types of soil virtual sample point allocation results. Each classification data set includes the environmental data, longitude and latitude of each sample unit, and the label of the sample unit, which is whether the sample unit has a corresponding soil barrier layer.

[0056] In this embodiment, a label is assigned to each sample unit through clustering, and the clustering result is the label of the sample unit. In other embodiments, labels can also be assigned in other ways. For example, if the probability of a certain soil barrier layer (such as a white pulp layer) in a sample unit exceeds a certain value, the sample unit is considered to have the soil barrier layer, that is, the label of the sample unit in the dataset corresponding to the soil barrier layer is directly set to "present".

[0057] S6. Based on the data set obtained in S5, a soil barrier layer classification model is constructed based on the machine learning model.

[0058] Specifically, this embodiment builds 6 obstacle layer prediction models based on the Random Forest Classifier model (RandomForest Classifier) under the Python language sklearn library. The model requires normalization of the input data and manual setting of two main parameters, namely the number of decision trees (n) and the maximum number of features (m). Grid Search is used to traverse the set parameter range to find the optimal parameter combination. This process is based on the GridSearchCV function integrated in the sklearn library. At the same time, combined with 5-fold cross-validation, the model performance evaluation under different parameter combinations is output, and the model accuracy index (Accuracy) is visualized to facilitate further debugging and optimization of the model.

[0059] A random forest model was constructed 100 times (i.e., L times) for each of the six barrier layers. Because the characteristics influencing the generation and importance of each barrier layer differ, the optimal parameter values for the model also differ. For each barrier layer, the combination with the highest frequency of occurrence among the 100 optimal parameter combinations output by the 100 random forest models using the above method was the optimal parameter for single-shot modeling of that barrier layer. The average accuracy of these parameter combinations was used as the model accuracy. The environmental data for each mapping unit was input into the random forest classification model corresponding to the optimal parameters to determine whether a soil barrier layer existed in each mapping unit.

[0060] In other implementations, the machine learning model may not be limited to the random forest classification model, and may also adopt an XGBoost model or a neural network model.

[0061] S7. Use an iterative method to repeat the process of S4 to S6 P times to obtain P mapping results of each soil barrier layer, count the number of times the soil barrier layer appears in each grid, and obtain the mapping results of the soil barrier layer.

[0062] After completing a single prediction for a particular obstacle layer, the entire process of extracting sample units, assigning obstacle information, and modeling (i.e., S4–S6) was repeated 100 times. By calculating the percentage of "presence" labels relative to the total number of predictions, a fuzzy mapping result was generated for the final probability surfaces of the six obstacle layers. In this probability surface, each grid contains a probability value integrated from the 100 prediction results, which specifies the probability of the obstacle layer corresponding to that grid being present. Figure 4 The result labels of four predictions show an example of the integration process of the fuzzy probability of a certain obstacle layer. Figure 5-1 to Figure 5-6 The mapping results of six types of soil barrier layers are white pulp layer, calcium accumulation layer, alkali accumulation layer, eutrophic layer, salt accumulation layer and clay layer.

[0063] In other implementations, the probability of each grid being in the soil barrier layer may also be obtained in other ways. For example, if the number of times follows a normal distribution, it may be obtained based on a normal distribution probability model.

[0064] like Figure 6 This example uses 157 typical profiles within the study area, extracted from the Jilin Soil Species Directory, Liaoning Soil Species Directory, China Inner Mongolia Soil Species Directory, and the Heilongjiang Volume of the Chinese Soil System Directory, as the validation point set. First, based on expert knowledge of diagnostic layers, diagnostic characteristics, and texture configurations within these data, each profile was systematically evaluated. This was combined with the specific soil science standards for barrier layer demarcation in each province to determine the presence of barrier factors within the 157 typical profiles.

[0065] A mapping accuracy index (MAI) is used to measure the difference between the predicted results and the actual results for the fuzzy mapping scheme. MAI quantifies the prediction accuracy by combining the predicted probability value and its consistency with the validation set results. The index ranges from 0 to 1. The larger the value, the closer the predicted result is to the actual result, that is, the higher the accuracy of the fuzzy mapping of the obstacle layer. The calculation formula of MAI is as follows:

[0066]

[0067] Where i represents the verification point number. i =p i ] is a function used to judge the prediction accuracy. The logic is as follows: Taking the white pulp layer as an example, when the predicted existence probability at the i-th point is p i (p i ≠0), and the actual existence of the white pulp layer indicated in the validation set r i When the prediction is correct, the prediction accuracy of the point is defined as p i ; If the prediction and actual results r i If the prediction is wrong, the prediction accuracy of this point is 1-p i In particular, when the predicted probability at the i-th point is p i , and p i = 0, which means that the prediction result believes that there is no white pulp layer at this point. If the validation set result also indicates that there is no white pulp layer at this point (r i =0), the prediction is completely correct and the prediction accuracy of this point is 1; if the validation set result r i Is existence (r i =1), then the prediction of this point is wrong and the accuracy is equal to 0.

[0068] Table 2

[0069]

[0070] Table 2 shows the accuracy evaluation of the mapping results for each soil barrier layer. Overall, the soil barrier layer distribution mapping method of this embodiment achieved excellent accuracy, ranging from 74.87% to 97.27%, indicating that the prediction results are consistent with actual results. This embodiment achieved the most significant results in predicting the white slurry layer, achieving the highest accuracy. The solution also performed well for the alkali layer, eutrophic layer, salt layer, and viscous layer, with accuracy exceeding 88%.

[0071] Computer device embodiment:

[0072] A computer device includes a processor configured to execute a computer program to implement the steps of the soil barrier layer distribution mapping method described above. The specific soil barrier layer distribution mapping method has been described in sufficient detail in the above-mentioned soil barrier layer distribution mapping method embodiment and will not be repeated here.

[0073] Example of soil barrier layer distribution map:

[0074] A soil barrier layer distribution map is obtained by the soil barrier layer distribution map drawing method described above. The specific soil barrier layer distribution map drawing method has been described in sufficient detail in the above embodiment of the soil barrier layer distribution map drawing method and will not be repeated here.

[0075] The present invention has the following characteristics:

[0076] The method provided by this invention requires no field surveys, simplifies expert knowledge acquisition, and offers a simple, flexible, and highly compatible operation process, making it easily applicable and scalable. It can rapidly analyze the spatial information of soil barrier layers in target areas. The fuzzy probability mapping method employed can visualize the spatial gradients of soil barrier layers and provide users with richer probabilistic statistical results, facilitating subsequent analysis and application. This method can provide technical support for the current large-scale national analysis of soil barrier layer spatial information, with broad scope and promising application prospects.

[0077] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments without inventive effort, or replace some of the technical features therein with equivalents. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for mapping soil barrier layer distribution, characterized in that: include: 1) Based on expert knowledge, the probability of the existence of a soil barrier layer in each mapping unit within the mapping area and the environmental factors that affect the formation of the soil barrier layer are obtained; And obtain the environmental data corresponding to each mapping unit and environmental factors in the mapping area; 2) Sampling all mapping units to obtain N sample units; According to the probability of the presence of a soil barrier layer in each sample unit, labels are labeled on N sample units to obtain a classified data set including the environmental data of each sample unit, wherein the label indicates whether the sample unit has a soil barrier layer; A soil barrier layer classification model is constructed based on the classification data set; the environmental data of each mapping unit is input into the soil barrier layer classification model to obtain the result of whether a soil barrier layer exists in each mapping unit; 3) Repeat 2) P times, and obtain the probability of the existence of a soil barrier layer in a certain mapping unit based on the number of occurrences of the result that the soil barrier layer exists in the mapping unit.

2. The method for mapping the soil barrier layer distribution map according to claim 1, characterized in that: The classification data set is obtained in the following way: N sample units are divided into two categories according to whether there is a soil barrier layer in each sample unit and the environmental data of each sample unit through a clustering algorithm. The category to which each sample unit belongs is the label of the sample unit.

3. The method for mapping the soil barrier layer distribution map according to claim 2, characterized in that: The classification data set is M groups of data sets obtained by repeatedly executing the clustering algorithm M times.

4. The method for mapping the soil barrier layer distribution map according to claim 2 or 3, characterized in that: The clustering algorithm is a constrained K-means clustering algorithm.

5. The method for mapping the soil barrier layer distribution map according to claim 1, wherein: 2) A stratified sampling strategy based on natural break points is used to obtain sample units.

6. The method for mapping the soil barrier layer distribution map according to claim 1 or 2, characterized in that: 2) Sampling was performed in the following manner: soil types in the mapping area were divided into multiple levels based on the area of each soil type in the mapping area, and the number of sample units for each level was determined according to the total area of the soil types in each level.

7. The method for mapping the soil barrier layer distribution map according to claim 1, wherein: The soil barrier layer classification model is obtained in the following manner: based on the classification data set, L classification models are constructed for predicting whether a soil barrier layer exists in each sample unit, and the classification model that appears most frequently in the L times is used as the final soil barrier layer classification model.

8. The method for mapping the soil barrier layer distribution map according to claim 1 or 7, characterized in that: The classification model is a random forest model.

9. The method for mapping the soil barrier layer distribution map according to claim 1, wherein: In 3), the probability of a soil barrier layer existing in a certain mapping unit is obtained by the following method: the proportion of the soil barrier layer existing in a certain mapping unit in P sub-classifications is used as the probability of the soil barrier layer existing in the mapping unit.

10. The method for mapping the soil barrier layer distribution map according to claim 1, characterized in that: The expert knowledge is obtained in the form of a questionnaire survey.

11. A computer device comprising a processor, characterized in that: The processor is used to execute a computer program to implement the steps of the soil barrier layer distribution map making method according to any one of claims 1 to 10.

12. A soil barrier layer distribution map, characterized in that: The soil barrier layer distribution map is obtained by the soil barrier layer distribution map making method according to any one of claims 1 to 10.

Images