Method for exploring influence of soil texture on crop growth vigor

Abstract

The invention relates to a method for exploring influence of soil texture on crop growth vigor, and relates to a method for exploring influence of crop growth vigor. The method comprises the following steps: 1, downloading a remote sensing image; 2, acquiring data; 3, constructing a feature variable set; 4, establishing a forest prediction model; 5, obtaining a soil texture spatial distribution map; and 6, extracting NDVI from the remote sensing image as the growth vigor of crops, and performing correlation analysis on the NDVI and the soil texture spatial distribution diagram. The method is based on the soil texture mapping of the random forest, and the soil texture spatial distribution can be quickly and accurately predicted. According to the method, the data is accurate, the model is more efficient, the interior of the farmland can be accurately and individually regulated and controlled, and the soil texture better meets the actual requirements of crop growth. The method can be used in the field of precision planting agriculture.

Description

Technical Field

[0001] The invention relates to a method for exploring the influence of crop growth. Background Art

[0002] Soil texture, as an important indicator of soil physical properties, is an important soil attribute that controls most physical, chemical, and biological processes in the soil. It can affect soil heat capacity, permeability, water holding capacity, and solute movement, and is closely related to applications such as climate, ecology, hydrological models, smart agricultural management, and soil pollution control. Regions and landscapes are often highly spatially heterogeneous in both the horizontal and vertical directions. Currently, there is an increasing demand for detailed soil texture information when dealing with global and national issues such as climate change, soil degradation, water shortage, environmental pollution, and agricultural and ecosystem sustainability.

[0003] Different soil textures and other soil physical properties interact with fertility and environmental conditions to affect crop growth and development. Since soils of different textures contain different amounts of water-absorbing polymers (WAP), they interact with the effective water content, wilting time, and irrigation frequency of crops, which further affects the growth status of crops. In agricultural production, the rational selection of soil texture types is of great significance for improving the yield and quality of farmland. Therefore, understanding the relationship between soil texture and crop growth is important for selecting suitable crops, formulating reasonable fertilization plans, and irrigation management.

[0004] The traditional soil texture prediction method is the geostatistical analysis method. Geostatistics is a science based on regionalized variable theory and variogram model, which studies natural phenomena with spatial correlation and dependence and both randomness and structure in spatial distribution. However, its prediction accuracy is often limited by sample density and sample variation. Only with sufficient number and dense samples can an accurate soil texture spatial distribution map be obtained, and it is difficult to quickly and accurately complete the soil texture prediction. Summary of the invention

[0005] The present invention provides a method for exploring the influence of soil texture on crop growth. The present invention is based on random forest soil texture mapping, which can quickly and accurately predict the spatial distribution of soil texture.

[0006] The method of the present invention for exploring the influence of soil texture on crop growth is carried out according to the following steps:

[0007] 1. Draw the vector boundary of the study area, upload it to the personal database of the Google Earth Engine platform, and download the remote sensing image dataset; further process the remote sensing image dataset and extract the reflectance information of the image band; at the same time, obtain and process the environmental covariate data; construct the feature variable set with the processed remote sensing data and the environmental covariate data;

[0008] Second, a random forest prediction model was established by combining the random forest method. The model was trained using the feature variable set in step one. The model was adjusted and determined to be the best decision tree and then saved. Different feature variable sets were input into the established model to obtain the variable set with the highest soil texture prediction accuracy. The variable set was input into the random forest prediction model to obtain the soil texture spatial distribution map.

[0009] 3. Extract NDVI from remote sensing images as crop growth potential, conduct correlation analysis with the spatial distribution map of soil texture, and obtain the mechanism of the influence of soil texture on crop growth.

[0010] The present invention has the following advantages:

[0011] 1. Fast timeliness: The present invention uses satellite remote sensing images to draw the vector boundaries of the study area, uploads them to the personal database of the Google Earth Engine platform, and downloads remote sensing images to obtain accurate vegetation index and plot information, thereby providing accurate vegetation growth status and soil nutrient demand information. Compared with traditional soil testing formula analysis, the cost is lower and the timeliness is faster.

[0012] 2. Predictiveness: Soil texture is one of the important physical and natural properties of soil. Understanding the spatial distribution of cultivated soil texture is crucial for precise agricultural management and farmland protection. The present invention predicts the spatial distribution of global soil texture based on the soil texture data (remote sensing data and environmental data) of the sampling points, and explores the influence mechanism of soil texture on crop growth, so as to make targeted improvements to crop planting types before cultivation.

[0013] The present invention obtains various data through satellite remote sensing, and combines the random forest method to establish a random forest prediction model, uses a characteristic variable set to train the model, and adjusts and determines the best decision tree to obtain the best model; the data described in the present invention is accurate, the model is more efficient, and the present invention can achieve precise and personalized regulation of the interior of the farmland, so that the soil texture is more in line with the actual requirements of crop growth. The method of the present invention can be used in the field of precision planting agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is the spatial distribution map of soil texture-silt in the study area;

[0015] Figure 2It is the spatial distribution map of soil texture-clay in the study area;

[0016] Figure 3 is the soil texture-sand spatial distribution map;

[0017] Figure 4 The correlation between soil texture and crop growth. DETAILED DESCRIPTION

[0018] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0020] Specific implementation method 1: The method of exploring the influence of soil texture on crop growth in this implementation method is carried out according to the following steps:

[0021] 1. Draw the vector boundary of the study area, upload it to the personal database of the Google Earth Engine platform, and download the remote sensing image dataset; further process the remote sensing image dataset and extract the reflectance information of the image band; at the same time, obtain and process the environmental covariate data; construct the feature variable set with the processed remote sensing data and the environmental covariate data;

[0022] Second, a random forest prediction model was established by combining the random forest method. The model was trained using the feature variable set in step one. The model was adjusted and determined to be the best decision tree and then saved. Different feature variable sets were input into the established model to obtain the variable set with the highest soil texture prediction accuracy. The variable set was input into the random forest prediction model to obtain the soil texture spatial distribution map.

[0023] 3. Extract NDVI from remote sensing images as crop growth potential, conduct correlation analysis with the spatial distribution map of soil texture, and obtain the mechanism of the influence of soil texture on crop growth.

[0024] Specific implementation method 2: This implementation method is different from specific implementation method 1 in that: the remote sensing image dataset in step 1 is the Landsat-8 dataset and Sentinel-2 dataset provided by GEE; the Landsat-8 image dataset obtains band information, further performs cloud and shadow mask processing, and synthesizes into a median synthetic image; the Sentinel-2 growing season image dataset further performs NDVI extraction processing. The other steps and parameters are the same as those in specific implementation method 1.

[0025] Specific implementation method three: This implementation method is different from specific implementation method two in that the band information of the Landsat-8 image data set is a blue band, a green band, a red band, a near infrared band and two short-wave infrared bands. Other steps and parameters are the same as those of specific implementation method one.

[0026] Specific implementation method 4: This implementation method is different from specific implementation method 1 in that the environmental covariate data in step 1 are multi-year average precipitation, multi-year average temperature, elevation, slope, drainage density and drainage frequency. The other steps and parameters are the same as those in specific implementation method 1.

[0027] Specific implementation method 5: This implementation method is different from specific implementation method 1 in that: in step 2, ten-fold cross validation is used to evaluate the model performance to determine the best decision tree. Other steps and parameters are the same as those in specific implementation method 1.

[0028] Example 1: Method for exploring the effect of soil texture on crop growth using the method of the present invention

[0029] The specific steps are as follows:

[0030] 1. Draw the vector boundary of the study area, upload it to the personal database of the Google Earth Engine platform, and download the remote sensing image dataset; further process the remote sensing image dataset and extract the reflectance information of the image band; at the same time, obtain and process the environmental covariate data; construct the feature variable set with the processed remote sensing data and the environmental covariate data;

[0031] The details are as follows:

[0032] 1. Obtain vector boundary data of the study area: GEE provides remote sensing datasets, Landsat-8 datasets, and all images from April to June 2022 (bare soil period). The Sentinel-2 dataset has one period of growing season images from 2019 to 2021. The Landsat-8 images are cloud and shadow masked and synthesized into median composite images according to the month, resulting in a total of 3 monthly multi-year composite images from April to June. Image bands (blue band, green band, red band, near-infrared band and two short-wave infrared bands) are mainly used for soil texture prediction research. NDVI is extracted from Sentinel-2 growing season images for crop growth analysis. In order to better predict the soil texture of the study area, six environmental covariates related to climate, topography, and hydrology were used in the study (spatial resolution 30m):

[0033] (1) Average precipitation over many years: Precipitation directly affects soil moisture content, which in turn affects the transformation, migration, leaching and deposition of minerals, organic matter and their products in the soil.

[0034] (2) Multi-year average temperature: Air temperature affects soil temperature. Soils at different temperatures will undergo different physical, chemical, and biological processes and changes.

[0035] (3) Elevation: Elevation indirectly affects soil, causing the redistribution of matter and energy in mountainous areas. As temperature, precipitation, and humidity change vertically with altitude, different climate and vegetation zones are formed, resulting in obvious vertical zoning of soil composition and physical and chemical properties.

[0036] (4) Slope: The study used the 30 m elevation data using ArcGIS 10.6 to extract the slope “SL” as the slope input for the study area.

[0037] (5) Drainage density: Drainage density indicates the relationship between landform and river development, and has a strong influence on soil physical and chemical properties. This study used the line density tool in the ArcGIS 10.6 analysis tool with an elevation of 30 m to extract the drainage density “DD” as the drainage density input for the study area.

[0038]

[0039] Where, DD is the drainage density, L is the total length of the river (km), A is the area around the river (km 2 ).

[0040] (6) Drainage frequency: As traditional hydrological data, drainage frequency and drainage density are more important to the model than other hydrological data. This study used drainage density data to extract the drainage frequency “DF” using ArcGIS 10.6 as the drainage frequency input for the study area.

[0041]

[0042] Where, DF: drainage frequency. n: number of channels. A: area around the channel (km 2 ).

[0043] 2. Acquisition of sample data: Soil samples were collected in the field from March 22 to April 5, 2021, and brought back to the laboratory for air-drying, and stones, weed roots and other impurities were removed. The soil was weighed indoors and air-dried, passed through a 2 mm sieve, and the soil particle composition (mass fraction) of the samples was determined using a Malvern MS-2000 laser particle size analyzer (produced by Malvern Instruments Ltd., UK, with a residual error of less than 1%).

[0044] Second, a random forest prediction model was established by combining the random forest method. The model was trained using the feature variable set in step one. The model was adjusted and determined to be the best decision tree and then saved. Different feature variable sets were input into the established model to obtain the variable set with the highest soil texture prediction accuracy. The variable set was input into the random forest prediction model to obtain the soil texture spatial distribution map.

[0045] Specifically:

[0046] 1. Random forest prediction model: prediction is based on the Random Forest algorithm (RF); RF has high prediction accuracy, can handle high-dimensional and multicollinear data, has high tolerance for outliers and missing values, requires little manual intervention, and is not prone to overfitting problems; RF modeling is implemented through the Random Forest package in the R language;

[0047] 2. Model establishment and verification: The random forest prediction model (RF) was used to perform soil texture prediction modeling using the band information of Landsat-8 monthly multi-year synthetic images in different months and six environmental covariates as input; ten-fold cross validation was used to evaluate the model performance to determine the best decision tree, that is, the root mean square error (RMSE) was used to measure the accuracy of the model.

[0048]

[0049] Where n represents the number of samples, y i represents the observed value of the i-th sampling point, Represents the measured value of the i-th sampling point; in general, R 2 The value range is 0 to 1. The closer the absolute value is to 1, the closer the predicted value is to the actual measured value; RMSE is used to test the difference between the actual measured value and the predicted value of soil texture; the smaller the value, the smaller the difference between them, and the better the prediction accuracy of the model;

[0050] 3. Input the remote sensing data into the model by month to obtain the best time window for soil texture prediction (the time window for silt prediction is April, with an RMSE of 7.84%; the time window for clay prediction is May, with an RMSE of 2.14%; the time window for sand prediction is April, with an RMSE of 2.43%). Then, by adding environmental variables, the prediction accuracy is improved, and the prediction data set with the highest soil texture prediction accuracy is obtained (the multi-year monthly synthetic images in April have the highest prediction accuracy for silt after adding climate and terrain data, with an RMSE of 1.71%. The multi-year monthly synthetic images in May have the highest prediction accuracy for clay after adding terrain data, with an RMSE of 2.02%. The multi-year monthly synthetic images in April have the highest prediction accuracy for sand after adding climate and terrain data, with an RMSE of 2.24%.); Use the prediction data set with the highest soil texture prediction accuracy to predict soil texture, input the variable set with the highest accuracy into the random forest prediction model, and obtain the spatial distribution map of soil texture.

[0051] 3. Effect of soil texture on crop growth

[0052] The NDVI extracted from remote sensing images was used as the crop growth potential, and correlation analysis was performed with the spatial distribution map of soil texture to obtain the mechanism of soil texture's influence on crop growth.

[0053] Specifically:

[0054] Based on ArcGIS 10.6, the Normalized Difference Vegetation Index (NDVI) of images in different growing seasons (July or August) was extracted, and the NDVI was clipped according to the crop classification of different years from 2019 to 2021. The growth of soybeans, corn, and rice was divided into 1 to 10 levels using the natural breakpoint method as the growth of different crops in different years. The NDVI of different years and the prediction results of silt, clay, and sand were subjected to Pearson correlation analysis. The Pearson correlation coefficient was recorded as r. When P<0.01, the correlation was significant.

[0055] In this embodiment Figure 1 The soil texture-silt spatial distribution map of the study area, Figure 2 Soil texture-clay spatial distribution map of the study area, Figure 3 Soil texture - spatial distribution map of sand particles, Figure 4 The correlation between soil texture and crop growth. Figure 1 to Figure 4 It can be seen that soil texture is significantly correlated with crop growth, which means that soil texture has a significant impact on crop growth.

[0056] The method of the present invention can be widely used in agricultural production to monitor and analyze soil texture in real time and efficiently, and make cultivated land planting, management and related decisions based on the analysis results; relevant departments can carry out related work of agricultural planting guidance through accurate soil texture distribution, which will promote the development of agriculture to a certain extent; the mapping results can also provide a data basis for regional agricultural production management and ecological protection.

Claims

1. A method for exploring the effect of soil texture on crop growth, characterized in that The method to explore the effect of soil texture on crop growth is as follows:

1. Draw the vector boundary of the study area, upload it to the personal database of the Google Earth Engine platform, and download the remote sensing image dataset; further process the remote sensing image dataset and extract the reflectance information of the image band; at the same time, obtain and process the environmental covariate data; construct the feature variable set with the processed remote sensing data and the environmental covariate data; Second, a random forest prediction model was established by combining the random forest method. The model was trained using the feature variable set in step one. The model was adjusted and determined to be the best decision tree and then saved. Different feature variable sets were input into the established model to obtain the variable set with the highest soil texture prediction accuracy. The variable set was input into the random forest prediction model to obtain the soil texture spatial distribution map.

3. Extract NDVI from remote sensing images as crop growth potential, conduct correlation analysis with the spatial distribution map of soil texture, and obtain the mechanism of the influence of soil texture on crop growth.

2. The method for exploring the effect of soil texture on crop growth according to claim 1, characterized in that The remote sensing image datasets in step 1 are the Landsat-8 dataset and Sentinel-2 dataset provided by GEE. The Landsat-8 image dataset obtains band information, further performs cloud and shadow mask processing, and synthesizes it into a median composite image. The Sentinel-2 growing season image dataset is further processed for NDVI extraction.

3. The method for investigating the effect of soil texture on crop growth according to claim 2, characterized in that The band information of the Landsat-8 image dataset includes blue band, green band, red band, near infrared band and two short-wave infrared bands.

4. The method for investigating the effect of soil texture on crop growth according to claim 1, characterized in that The environmental covariate data in step 1 are multi-year average precipitation, multi-year average temperature, elevation, slope, drainage density and drainage frequency.

5. The method for investigating the effect of soil texture on crop growth according to claim 1, characterized in that In step 2, ten-fold cross validation is used to evaluate the model performance and find the best decision tree.

Images