Soil salinity distribution prediction method based on environmental factors and remote sensing images

Abstract

The invention discloses a soil salinity distribution prediction method based on an environmental factor and a remote sensing image, and relates to the technical field of soil salinity distribution prediction, and the method comprises the following steps: employing a random forest model to carry out the total salinity fitting of environmental factor data, and obtaining a fitting salinity and alkalinity range; performing saline-alkali range extraction on the remote sensing image data by adopting a threshold division mode to obtain a remote sensing saline-alkali range; and superposing the fitting saline-alkali range and the remote sensing saline-alkali range to obtain a soil salinity distribution prediction result. According to the method, the environmental factor data is introduced, the random forest model is adopted to carry out total salt amount fitting, the obtained fitting saline-alkaline range and the remote sensing saline-alkaline range obtained based on the remote sensing image data are superposed, the relationship between the environmental factors and the remote sensing image and the soil salinity distribution is comprehensively considered, and the prediction accuracy of the soil salinity distribution is improved.

Description

Technical Field

[0001] The present application relates to the technical field of soil salt distribution prediction, and particularly relates to a method for predicting soil salt distribution based on environmental factors and remote sensing images. Background Art

[0002] Soil salinity is one of the key factors affecting crop growth and soil health. Excessive or too low salt content will have a negative impact on crop yield and soil quality. With the increasing impact of global climate change and human activities on water resources, the problem of soil salinization has become more prominent. Therefore, it is necessary to study the method for predicting soil salt distribution, so as to provide a basis and support for agricultural production and soil management.

[0003] In the prior art, Chinese Patent CN110243773A discloses a method for calculating the total salt content of soil by using the hyperspectral reflectance of soil. The hyperspectral information of saline soil is used to calculate the total salt content in the soil, which can be used for rapid monitoring and diagnosis of regional soil salinity. Users only need to monitor the total salt content, conductivity and corresponding hyperspectral reflectance information of a limited number of sample soils, and then they can use this method to establish the relationship between the total salt content and the hyperspectral absorption index of the same type of soil in this region. And based on the regression relationship between the total salt content of the soil and the hyperspectral absorption index, users can quickly obtain the hyperspectral absorption index of the same type of soil in this region through a portable ground object hyperspectral spectrometer, and then calculate the total salt content of the soil.

[0004] However, the above prior art only calculates the total salt content of the soil through the hyperspectral reflectance of the soil, without considering the relationship between environmental factors and soil salt distribution, and the accuracy of soil salt distribution prediction is poor. Summary of the Invention

[0005] The present application provides a method for predicting soil salt distribution based on environmental factors and remote sensing images, so as to solve the problem that the prior art does not consider the relationship between environmental factors and soil salt distribution, and the accuracy of soil salt distribution prediction is poor.

[0006] On the one hand, the present application provides a method for predicting soil salt distribution based on environmental factors and remote sensing images, including the following steps:

[0007] Step 1: Obtain the fitted saline-alkali range based on environmental factor data, and obtain the remote sensing saline-alkali range based on remote sensing image data.

[0008] Step 2: Overlay the fitted saline-alkali range and the remote sensing saline-alkali range to obtain the prediction result of soil salt distribution.

[0009] In Step 1, the obtaining of the fitted saline-alkali range based on environmental factor data includes: using a random forest model to fit the total salt content of the environmental factor data to obtain the fitted saline-alkali range.

[0010] In step one, the remote sensing saline-alkali range obtained from the remote sensing image data includes: extracting the saline-alkali range from the remote sensing image data by using the threshold division method to obtain the remote sensing saline-alkali range.

[0011] In a possible implementation manner, in step one, the environmental factor data includes: pH data, evapotranspiration data, and precipitation data.

[0012] In a possible implementation manner, in step one, the total salt content fitting of the environmental factor data by using the random forest model includes:

[0013] Obtaining the environmental factor data and the corresponding total salt content data of the thematic point sampling in the first time period as the training set.

[0014] Constructing a random forest model, training the random forest model by using the training set to obtain a random forest inversion model.

[0015] Obtaining the environmental factor data of the overall sampling in the time period to be fitted, and fitting the total salt content of the environmental factor data in the time period to be fitted by using the random forest inversion model.

[0016] In a possible implementation manner, in the process of training the random forest model by using the training set, the ten-fold cross-validation method is used for parameter tuning.

[0017] In a possible implementation manner, in step one, the extraction of the saline-alkali range from the remote sensing image data by using the threshold division method includes:

[0018] Obtaining the remote sensing image data.

[0019] Using the spectral index as the basis for identifying the saline-alkali range, setting the identification threshold, and extracting the saline-alkali range from the remote sensing image data by using the identification threshold.

[0020] In a possible implementation manner, after obtaining the remote sensing image data, preprocessing the remote sensing image data.

[0021] The preprocessing includes: cropping the region of interest, cloud masking, data filtering, and image synthesis.

[0022] In a possible implementation manner, the spectral index includes: normalized difference saline-alkali index, soil permeability index.

[0023] In a possible implementation manner, in step two, the fitting saline-alkali range and the remote sensing saline-alkali range are superimposed by taking the overlapping region to obtain the final saline-alkali land range, that is, the prediction result of the soil salt content distribution.

[0024] A method for predicting soil salinity distribution based on environmental factors and remote sensing images in this application has the following advantages:

[0025] By introducing environmental factor data and using the random forest model for total salt content fitting, the obtained fitting saline-alkali range is superimposed with the remote sensing saline-alkali range obtained based on remote sensing image data, comprehensively considering the relationship between environmental factors, remote sensing images and soil salinity distribution, and improving the accuracy of soil salinity distribution prediction.

[0026] During the process of training the random forest model with the training set, the ten-fold cross-validation method is used for parameter tuning, improving the accuracy of the random forest inversion model.

[0027] It is proposed that after obtaining remote sensing image data, preprocessing is carried out on the remote sensing image data, and the preprocessing includes: region of interest cropping, cloud masking, data filtering and image synthesis, improving the quality of remote sensing image data, and further improving the accuracy of the remote sensing saline-alkali range.

[0028] By superimposing the fitting saline-alkali range and the remote sensing saline-alkali range in the way of taking the overlapping area, the final range of saline-alkali land, that is, the prediction result of soil salinity distribution, is obtained, avoiding the error of a single method and improving the reliability of the prediction result. Brief Description of the Drawings

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0030] Figure 1 It is a schematic flowchart of a method for predicting soil salinity distribution based on environmental factors and remote sensing images provided by an embodiment of the present application;

[0031] Figure 2 It is a schematic diagram of the final range of saline-alkali land in XX County provided by an embodiment of the present application. Detailed Embodiments

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0033] Such asFigure 1 As shown in Figure 1 , an embodiment of the present application provides a method for predicting soil salinity distribution based on environmental factors and remote sensing images, including the following steps:

[0034] Step 1: Obtain the fitted saline-alkali range based on environmental factor data and the remote sensing saline-alkali range based on remote sensing image data.

[0035] Step 2: Overlay the fitted saline-alkali range and the remote sensing saline-alkali range to obtain the predicted result of soil salinity distribution.

[0036] In Step 1, obtaining the fitted saline-alkali range based on environmental factor data includes: using a random forest model to fit the total salt content of the environmental factor data to obtain the fitted saline-alkali range.

[0037] In Step 1, obtaining the remote sensing saline-alkali range based on remote sensing image data includes: using the method of threshold division to extract the saline-alkali range from the remote sensing image data to obtain the remote sensing saline-alkali range.

[0038] Exemplarily, in Step 1, the environmental factor data includes: pH data, evapotranspiration data, and precipitation data.

[0039] Specifically, in this embodiment, the pH data is the sample-detected pH data after overall and special point sampling. The evapotranspiration data and precipitation data are searched and downloaded through the modis official website and extracted after download processing.

[0040] Exemplarily, in Step 1, using the random forest model to fit the total salt content of the environmental factor data includes:

[0041] Obtain the environmental factor data and the corresponding total salt content data of special point sampling in the first time period as the training set.

[0042] Construct a random forest model, and use the training set to train the random forest model to obtain a random forest inversion model.

[0043] Obtain the environmental factor data of overall sampling in the time period to be fitted, and use the random forest inversion model to fit the total salt content of the environmental factor data in the time period to be fitted.

[0044] Specifically, in this embodiment, the environmental factor data in the first time period is obtained by respectively accumulating the pH data, evapotranspiration data, and precipitation data of 55 special points from November 2023 to April 2024 (6 months) in XX County. The corresponding total salt content data is the sample-detected total salt content data after special point sampling. In this embodiment, the environmental factor data in the time period to be fitted is obtained by respectively accumulating the overall sampling pH data, evapotranspiration data, and precipitation data from May to October 2023 (6 months) in XX County.

[0045] In this embodiment, a soil salt distribution prediction method based on environmental factors and remote sensing images is applied to two situations: the first situation is that partial point sampling is carried out in an area, and the method is used to estimate the soil salinity of the entire area (the overall sampling is estimated by thematic points); the other situation is that sampling was carried out in an area this year, and the method is used to estimate the salt distribution of this area last year (from November 2023 to April 2024 to estimate May to October 2023).

[0046] Specifically, random forest is a supervised learning algorithm for solving classification and decision tree problems. "Random forest" refers to a grouping of many decision trees, where each tree relies on the value of a random vector sampled equally from all trees in the forest. Different subsets of the data set are randomly selected with replacement when sampling to train each decision tree, and when splitting each node in the process of building the decision tree, the best split is selected from a random feature subset rather than all features, which ensures the diversity of the created trees and enhances the robustness of the model. At the same time, a specified number of trees will be established according to the parameters, and each tree is established independently, and then the results are summarized and the average prediction value of a single tree is output.

[0047] In this embodiment, the random forest model is constructed using the scikit-learn library in Python 3.9, and the version number of the scikit-learn library is 1.1.1.

[0048] Exemplarily, in the process of training the random forest model using the training set, a ten-fold cross validation method is used to adjust parameters.

[0049] Specifically, in this embodiment, when adjusting parameters using the ten-fold cross-validation method, in the first step, adjust the number of decision trees n_estimator. Starting from 1 and testing sequentially up to 2001 at intervals of 100, select the n_estimator value (n1) with the highest average score; subsequently, test sequentially from n1 - 100 to n1 + 100 at intervals of 10, and select the n_estimator value (n2) with the highest average score; finally, test sequentially from n2 - 10 to n2 + 10 at intervals of 1, and finally determine the number of decision trees with the highest score. In the second step, use the GridSearchCV method in the sklearn library to adjust the maximum depth max_depth of the decision tree. Set the minimum value to 1 and the maximum value to 50, and also select the max_depth value with the highest average cross-validation score. Finally, adjust the number of features max_features, which is determined using the grid search method. The minimum value is the integer part of the square root of the total number of features, and the maximum value is the total number of features. Select the parameter values with the highest average score in 10 validations as the final parameter values of the random forest model.

[0050] Specifically, in this embodiment, use the environmental factor data sampled at the special topic points in the first time period (from November 2023 to April 2024 in XX County) as the input features of the random forest model, and the corresponding total salt content data as the target variable. After training the random forest inversion model, apply the random forest inversion model to the environmental factor data sampled in an overall manner during the period to be fitted, that is, replace the input features with the environmental factor data sampled in an overall manner during the period to be fitted (from May to October 2023 in XX County), obtain the total salt content fitting results for the period to be fitted, and correct the total salt content data sampled in an overall manner using the fitting saline-alkali range. Select cultivated land, orchards, forests, grasslands, and saline-alkali lands for mapping to obtain the fitting saline-alkali range.

[0051] In this embodiment, compare the total salt content fitting results of XX County with the total salt content data sampled in an overall manner, as shown in Table 1:

[0052] Table 1 Comparison of the total salt content fitting results of XX County with the total salt content data sampled in an overall manner

[0053]

[0054] Exemplarily, in step one, the method for extracting the saline-alkali range from the remote sensing image data by using the threshold division method includes:

[0055] Obtain the remote sensing image data.

[0056] Use the spectral index as the basis for identifying the saline-alkali range, set the identification threshold, and use the identification threshold to extract the saline-alkali range from the remote sensing image data.

[0057] Specifically, in this embodiment, the Sentinel-2L2A dataset in the Sentinel-2 high-resolution satellite images is used as the remote sensing image data.

[0058] Exemplarily, after obtaining the remote sensing image data, preprocessing is performed on the remote sensing image data.

[0059] The preprocessing includes: region of interest cropping, cloud masking, data filtering, and image synthesis.

[0060] Specifically, in this embodiment, the boundary data of XX County, Yulin City is selected as the basis for region of interest cropping; cloud masking is used to remove clouds in the remote sensing image data to ensure data quality; common filtering methods (spatial domain filtering or frequency domain filtering) are used to filter the remote sensing image data; and image synthesis is performed by taking the annual median value.

[0061] Exemplarily, the spectral indices include: normalized difference salinity index, soil permeability index.

[0062] Specifically, the normalized difference salinity index is shown as follows:

[0063]

[0064] Where NDSI represents the normalized difference salinity index, R represents the red band value, and NIR represents the near-infrared band value.

[0065] The soil permeability index is shown as follows:

[0066]

[0067] Where SI represents the soil permeability index, and G represents the green band value.

[0068] The recognition threshold can be adaptively set according to actual needs, and this embodiment does not limit it.

[0069] In this embodiment, after extracting the saline-alkali range from the remote sensing image data using the recognition threshold, the extraction result is exported in the GeoTIFF geospatial image format to obtain the remote sensing saline-alkali range.

[0070] Exemplarily, in step two, the fitting saline-alkali range and the remote sensing saline-alkali range are superimposed by taking the overlapping area to obtain the final saline-alkali land range, that is, the prediction result of the soil salt distribution.

[0071] Specifically, in this embodiment, the final saline-alkali land range after taking the overlapping area of the fitted saline-alkali range and the remote sensing saline-alkali range of XX County is obtained. Based on the total salt content data obtained by fitting, a salt spatial prediction model is simulated, and salt grading is carried out to finally form a saline-alkali land degree map for the corresponding district or county and extract the corresponding area of the map patches, as Figure 2 shown, where the yellow map patches represent saline-alkali land.

[0072] In the embodiment of the present application, by introducing environmental factor data and using a random forest model for total salt content fitting, the obtained fitted saline-alkali range is superimposed on the remote sensing saline-alkali range obtained based on remote sensing image data, comprehensively considering the relationship between environmental factors and remote sensing images and soil salt distribution, and improving the prediction accuracy of soil salt distribution.

[0073] During the process of training the random forest model with the training set, the ten-fold cross-validation method is used for parameter tuning, improving the accuracy of the random forest inversion model.

[0074] It is proposed that after obtaining the remote sensing image data, preprocessing is carried out on the remote sensing image data, and the preprocessing includes: region of interest cropping, cloud masking, data filtering, and image synthesis, improving the quality of the remote sensing image data, and thus improving the accuracy of the remote sensing saline-alkali range.

[0075] By superimposing the fitted saline-alkali range and the remote sensing saline-alkali range in the way of taking the overlapping area, the final saline-alkali land range, that is, the prediction result of soil salt distribution, is obtained, avoiding the error of a single method and improving the reliability of the prediction result.

[0076] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concept. Therefore, the appended claims are intended to be interpreted to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0077] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these changes and modifications.

Claims

1. A method for predicting soil salinity distribution based on environmental factors and remote sensing images, characterized in that It includes the following steps: Step 1: Obtain the fitted saline-alkali range based on environmental factor data and the remote sensing saline-alkali range based on remote sensing image data; Step 2: Overlay the fitted saline-alkali range and the remote sensing saline-alkali range to obtain the prediction result of soil salt distribution; In Step 1, the obtaining of the fitted saline-alkali range based on environmental factor data includes: using a random forest model to perform total salt content fitting on the environmental factor data to obtain the fitted saline-alkali range; In Step 1, the obtaining of the remote sensing saline-alkali range based on remote sensing image data includes: using the method of threshold division to extract the saline-alkali range from the remote sensing image data to obtain the remote sensing saline-alkali range.

2. The soil salinity distribution prediction method based on environmental factors and remote sensing images according to claim 1, wherein In Step 1, the environmental factor data includes: pH data, evapotranspiration data, and precipitation data.

3. A method for predicting soil salinity distribution based on environmental factors and remote sensing images according to claim 1, characterized in that, In Step 1, the using of the random forest model to perform total salt content fitting on the environmental factor data includes: Obtain the environmental factor data and the corresponding total salt content data of the thematic point sampling in the first time period as the training set; Construct a random forest model, and use the training set to train the random forest model to obtain a random forest inversion model; Obtain the environmental factor data of the overall sampling in the time period to be fitted, and use the random forest inversion model to perform total salt content fitting on the environmental factor data in the time period to be fitted.

4. A method for predicting soil salinity distribution based on environmental factors and remote sensing images according to claim 3, characterized in that, In the process of using the training set to train the random forest model, the ten-fold cross-validation method is used for parameter tuning.

5. A method for predicting soil salinity distribution based on environmental factors and remote sensing images according to claim 1, characterized in that, In Step 1, the using of the method of threshold division to extract the saline-alkali range from the remote sensing image data includes: Obtain the remote sensing image data; Use the spectral index as the basis for identifying the saline-alkali range, set the identification threshold, and use the identification threshold to extract the saline-alkali range from the remote sensing image data.

6. The soil salinity distribution prediction method based on environmental factors and remote sensing images according to claim 5, wherein, After obtaining the remote sensing image data, preprocess the remote sensing image data; The preprocessing includes: region of interest cropping, cloud masking, data filtering, and image synthesis.

7. A method for predicting soil salinity distribution based on environmental factors and remote sensing images according to claim 5, characterized in that, The spectral index includes: normalized difference saline-alkali index, soil permeability index.

8. A method for predicting soil salinity distribution based on environmental factors and remote sensing images according to claim 1, characterized in that In Step 2, use the method of taking the overlapping area to overlay the fitted saline-alkali range and the remote sensing saline-alkali range to obtain the final range of saline-alkali land, that is, the prediction result of soil salt distribution.

Images