A Method for Constructing Temporal and Spatial Sequence Data of Farmland Soil Acidification and Trend Diagnosis

Abstract

The present invention discloses a method for constructing spatio-temporal sequence data of farmland soil acidification and trend diagnosis, which relates to the technical field of soil science. The method includes: collecting spatial and temporal observation data of soil pH values, and performing data cleaning and spatio-temporal grid construction; calculating spatio-temporal empirical variograms based on the collected data to quantify the spatio-temporal laws of soil acidification; constructing spatio-temporal variograms through separable models and metric models, and optimizing parameters to achieve high-precision fitting; constructing a unified spatio-temporal grid, and predicting the spatial distribution and temporal changes of soil acidification based on spatio-temporal Kriging interpolation technology; using trend analysis and Hurst index to identify the significance and sustainability of soil acidification. This method can efficiently and accurately evaluate the spatio-temporal changes of soil acidification, providing a scientific basis for agricultural management.

Description

Technical Field

[0001] The present invention relates to the technical field of soil science, and more particularly, to a method for constructing a spatio-temporal sequence data of farmland soil acidification and diagnosing trends. Background Art

[0002] Traditional methods for monitoring soil pH mainly rely on manual sampling and laboratory analysis. Although this method can obtain relatively accurate soil pH value data, there are problems such as limited sampling points, long data acquisition cycle, high cost, and inconsistent spatial coordinates of sampling points over the years. The in-depth application of geostatistical methods has laid a solid foundation for understanding the complex mechanism of the soil acidification process by analyzing the variation law of soil nutrients in time and space. The introduction of Geographic Information System (GIS) technology has further promoted the exploration of the correlation between soil nutrient spatial data and attribute data. In particular, the application of techniques such as Kriging interpolation and geographically weighted regression has effectively improved the accuracy of predicting the spatial distribution of soil attribute values. Due to the spatial heterogeneity of farmland soil, the monitoring data of single or a small number of points often cannot comprehensively reflect the soil acidification status of the entire region. The above methods mostly focus on local or short-term monitoring and are difficult to comprehensively capture the wide distribution characteristics of soil acidification in the spatial dimension and its evolution trajectory over time. At the same time, due to the diversity of data sources, uneven distribution of sample points, and geographical coordinate accuracy problems, it increases the uncertainty of soil acidification trend research. In view of this, integrating soil acidification data through more precise spatial interpolation and other advanced technical means to achieve comprehensive analysis across long time series and wide spatial scales has become a key approach to deeply analyze the internal mechanism of soil acidification and its environmental effects.

[0003] Based on the two major challenges faced by current farmland soil monitoring: spatial discreteness and inconsistency of sampling points. These problems lead to discontinuity and uncertainty of soil data in space and time, seriously affecting the accurate assessment and prediction of soil acidification trends. Farmland soil acidification is a complex and dynamic process, which is affected by various natural and human factors, such as climate, vegetation, fertilization habits, etc. At the technical level, this method first recognizes that soil acidification is a complex spatio-temporal process, which is not only affected by spatial variation but also constantly evolves over time. Therefore, traditional interpolation methods that only consider spatial variation can no longer meet the accuracy requirements of farmland soil acidification trend monitoring. Summary of the Invention

[0004] The purpose of the present invention is to introduce time variation parameters to comprehensively capture the change law of soil acidification in the spatio-temporal dimension, and thus propose a method for constructing a spatio-temporal sequence data of farmland soil acidification and diagnosing trends.

[0005] The technical solution of the present invention is: to provide a method for constructing and trend diagnosing spatiotemporal sequence data of farmland soil acidification, and the method includes:

[0006] S1. Data collection and preprocessing: Collect annual observed data of farmland soil pH values, and the data includes spatial information and time stamps; Clean the data, remove error records and invalid points outside the study area; Construct a spatiotemporal sparse grid data framework, and input the cleaned data into the spatiotemporal sparse grid data framework to uniformly integrate the spatial and temporal dimensions;

[0007] S2. Calculate the spatiotemporal empirical variogram according to the collected data: Calculate the spatiotemporal variogram and empirical estimation of the soil pH value respectively, and list the two to form the spatiotemporal empirical variogram to quantify the variation law of the soil pH value under different spatial and temporal conditions;

[0008] S3. Construct a spatiotemporal variogram model to fit the spatiotemporal variogram: Based on the calculation results of the spatiotemporal variogram, the system constructs a separable model and a metric model to model the spatiotemporal changes of soil acidification, and continuously optimizes the parameters of the two models to fit a more accurate spatiotemporal variogram;

[0009] S4. Conduct spatiotemporal prediction through spatiotemporal Kriging interpolation: Construct a unified spatiotemporal grid, combine the spatial and temporal information in the observed data of farmland soil pH values, and conduct prediction of farmland soil pH values based on spatiotemporal Kriging interpolation; Generate a full spatiotemporal observation grid to predict the spatial distribution of farmland soil acidification and its change trend over time;

[0010] S5. Soil acidification trend analysis: Based on the collected data, establish a soil acidification trend recognition model through time series raster image Theil - Sen Median trend analysis and non - parametric Mann - Kendall test to evaluate the significance of the farmland soil acidification trend;

[0011] S6. Sustainability identification: Use the Hurst index to analyze the sustainability of the soil acidification trend to provide a basis for future soil improvement and management.

[0012] In any of the above - mentioned technical solutions, further, the specific process of calculating the spatiotemporal empirical variogram in step S2 includes:

[0013] Assume that the soil pH value is represented by a Gaussian spatiotemporal random field defined on the spatial domain and the temporal domain , and the soil pH value samples PH(s n , t n ) observed at different spatiotemporal positions (s n , t n ) ∈ S × T include repeated measurements at the same location or simultaneous measurements at different spatial positions;

[0014] Introduce the spatio-temporal covariance function: For the spatio-temporal random process {PH(s, t): s ∈ R 2 , t ∈ R}, under the assumption of second-order stationarity, the covariance function of the spatio-temporal random field PH(s, t) depends only on the spatial distance h s between two observation points and the time lag h t , and is independent of the specific spatio-temporal positions of these two observation points. Then the spatio-temporal covariance function is:

[0015] C(h s , h t ) = Cov(PH(s + h s , t + h t ), PH(s, t));

[0016] To more intuitively describe the spatio-temporal variability of soil pH values, the time variation parameter is incorporated into the spatial variation model, and the spatio-temporal variation function γ(h s , h t ) is introduced to measure the degree of difference between random variables:

[0017]

[0018] For the spatio-temporal dataset of soil pH values, the empirical estimate of its spatio-temporal covariance is expressed as:

[0019]

[0020] where N(h s , h t ) is the set of soil pH value observation point pairs that satisfy the spatial offset h s and the time lag h t , is the number of such observation point pairs, PH(s, t) represents the soil pH value measured at position s and time t, μ PH (h s , h t ) and μ PH (0, 0) are the empirical means at the sampling point offset and the origin, respectively;

[0021] Similarly, the empirical estimate of the spatio-temporal variation function of soil pH values is:

[0022]

[0023] To make the covariance function meaningful, the function C(h s , h t ) must satisfy the positive definiteness condition, that is, ensure the non-negativity and invertibility of the covariance matrix. For any real number ai (i = 1, …, n) and for any spatio-temporal point (s i , t i ) there is:

[0024]

[0025] The obtained spatio-temporal variogram γ(h s , h t ) and the empirical estimate constitute the spatio-temporal empirical variogram.

[0026] In any of the above technical solutions, further, the specific process of constructing the spatio-temporal variogram model in step S3 includes: constructing a separable model and a metric model;

[0027] In the separable model, the spatio-temporal covariance function of the soil pH value is expressed as the product of the spatial covariance function and the temporal covariance function, that is:

[0028] C sep (h, u) = C s (h) · C t (u);

[0029] Among them, h represents the spatial distance, u represents the time lag, and the corresponding variogram form is:

[0030]

[0031] are the standardized spatial and temporal variograms, respectively, with a separable nugget effect, and the joint parameter in the sill value is set to 1, and the overall sill value is defined by the sill parameter; in the R language, use the gstat package to construct such a model; first, set the parameters of the spatial and temporal variograms respectively, then specify the separable model through the vgmST function, and use the fit.StVariogram function to fit the empirical variogram to optimize the model parameters;

[0032] The metric model extends the two-dimensional geographical space to three-dimensional spatio-temporal, and matches the spatial and temporal scales by introducing the anisotropy parameter (κ). The spatio-temporal distance is obtained by correcting the spatial distance (h) and the time lag (u) through κ, and the form is as follows:

[0033]

[0034] The corresponding variogram is:

[0035]

[0036] When building a measurement model in R language, first define a variogram model that includes the nugget effect, specify the measurement model through the vgmST function, and use the joint variogram and anisotropy parameters as inputs. Also use the fit.StVariogram function to fit the empirical variogram, optimize the model parameters to minimize the fitting error, and find the optimal variogram parameters; by comparing the MSE values or other statistical indicators of different models, select the optimal model.

[0037] In any of the above technical solutions, further, the process of predicting the soil pH value by spatio-temporal Kriging interpolation in step S4 includes:

[0038] Construct a spatio-temporal observation grid, which includes a spatial grid and a time grid. In terms of the spatial grid, determine a unified spatial reference system, divide the spatial grid according to the size of the research area and the required accuracy, determine the resolution of the grid, and generate a SpatialPixelsDataFrame pixel point. Each point represents a potential farmland soil PH value observation data; in terms of the time grid, determine the time range in the time dimension, and evenly divide this period into several time periods according to the data accuracy. Each time period represents a prediction time point; combine the spatial grid and the time grid to create a spatio-temporal prediction grid, and each grid point corresponds to a specific spatial location and time point;

[0039] Use the optimal spatio-temporal variogram model fitted in step S3, and call the krigeST function on the constructed full spatio-temporal observation grid for spatio-temporal Kriging interpolation to predict the spatial distribution of farmland soil acidification and its change trend over time. For any point (s0, t0) on the STF grid, the spatio-temporal Kriging interpolation prediction value PH(s0, t0) of its soil pH value can be given by the following formula:

[0040]

[0041] In the formula: PH(s i , t i ) is the pH observation value of the soil sampling point, and PH(s0, t0) is the estimated value at the spatio-temporal point (s0, t0); n is the number of observation points participating in the prediction, and λ i is the weight coefficient of the neighboring observation value PH(s i , t i ); μ is the introduced Lagrange coefficient; among them, λ i needs to ensure that the variance of the prediction error is minimized by solving the weight coefficient and satisfy the unbiasedness condition, which is usually transformed into solving the following optimization problem: At the same time, satisfy the unbiasedness condition: Construct the Lagrangian function in the following form:

[0042]

[0043] Let \(L(\lambda i ,\mu)\) take the partial derivatives with respect to \(\lambda i and \(\mu\) and set them to 0, and then solve for the weight coefficient \(\lambda i ;

[0044] After interpolation, it is necessary to evaluate the spatio-temporal Kriging interpolation results based on the parameters. Based on the evaluation results, it is necessary to verify the model and optimize and debug the parameters as needed to improve the prediction performance of the model.

[0045] In any of the above technical solutions, further, step S4 further includes: after interpolation is completed, the following steps are also required:

[0046] Model verification: Check whether the model satisfies the unbiasedness condition, that is, whether the average value of the predicted values is close to the average value of the observed values, analyze the error distribution, check whether there are systematic biases or outliers, evaluate the stability and consistency of the model, and ensure that reasonable prediction results can be obtained for different data subsets or time points;

[0047] Adjust and optimize the spatio-temporal empirical variogram parameters: According to the evaluation results and prior knowledge, adjust the parameters of the spatio-temporal variogram to improve the goodness of fit and prediction accuracy of the model;

[0048] Explore different model types: Try different spatio-temporal variogram models, compare their differences in prediction performance, and select the model most suitable for the current data;

[0049] Consider covariates: If the data contains other covariates that may be related to the soil pH value, incorporate them into the model as part of the fixed effect or random effect to improve the interpretability and prediction accuracy of the model;

[0050] Iterative optimization: Compare and analyze the preliminary evaluation results with the new prediction results after parameter optimization and debugging, judge whether the optimization is effective, and evaluate the accuracy and reliability of the spatio-temporal Kriging interpolation by comparing the prediction results with the data of known observation points;

[0051] Repeat the above steps for multiple iterative optimizations until satisfactory prediction performance is achieved.

[0052] In any of the above technical solutions, further, the soil acidification trend identification model established in step S5 is:

[0053]

[0054] S PHIt is the median slope calculated by the Theil-Sen Median trend analysis method and is used to evaluate the trend of the soil pH value time series. When S PH > 0, it indicates that the soil pH value shows an acidification trend; conversely, it indicates that the soil pH value tends to alkalization;

[0055] Define the statistic S:

[0056]

[0057] Among them, the sign function sgn is defined as:

[0058]

[0059] Among them, pH i and pH j represent the soil pH values in the i-th year and the j-th year respectively, and n represents the length of the time series;

[0060] At a given significance level α, α = 0.05, when |Z| > u 1-α / 2 it indicates that there is a significant change trend in the soil pH value time series at the α level.

[0061] In any of the above technical solutions, further, the specific content of using the Hurst index to analyze the sustainability of the soil acidification trend in step S6 includes:

[0062] The Hurst index calculation method is as follows:

[0063]

[0064] R(τ) = max 1≤t≤τ X(t, τ) - min 1≤t≤t X(t, τ), τ = 1, 2, …, n;

[0065]

[0066] Among them, is the mean time series of the pH value, X(t, τ) is the cumulative deviation of the pH value, R(τ) is the range, S(τ) is the standard deviation, H is the Hurst index. If there is a relationship R(τ) / S(τ) ∝ τ H , it indicates that there is a Hurst phenomenon in the soil pH value time series;

[0067] When 0.5 < H < 1, it indicates that the soil pH value time series has persistence, that is, the soil acidification / alkalization trend may continue in the future, and the closer H is to 1, the stronger this persistence is;

[0068] When H = 0.5, it indicates that the time series of soil pH value is random and there is no long-term correlation, that is, the change of soil pH value is unpredictable.

[0069] When 0 < H < 0.5, it indicates that the time series of soil pH value has anti-persistence, that is, the trend of soil acidification / alkalization may be opposite to the past in the future. The closer H is to 0, the stronger this anti-persistence is.

[0070] The beneficial effects of the present invention are as follows:

[0071] The present invention proposes a method for constructing and trend diagnosing soil acidification data by combining spatial and temporal dimensions. By integrating spatio-temporal data of farmland soil acidification, it realizes a comprehensive understanding of the spatio-temporal variation law of soil acidification. This method makes up for the defects of insufficient information in the spatial and temporal dimensions of traditional monitoring methods, and provides a scientific basis for the sustainable development of regional agriculture.

[0072] Among them, by introducing spatio-temporal variation parameters and using an optimized spatio-temporal variation function model, the accuracy of data analysis and prediction is improved. Through the separable model and the metric model, the dynamic characteristics of soil acidification are deeply explored, and the uncertainty of monitoring data in space and time is significantly reduced.

[0073] By using spatio-temporal Kriging interpolation technology to perform interpolation on the constructed full spatio-temporal prediction grid, it can quickly cover the entire study area, including unmonitored areas, and realize an efficient prediction of the spatial distribution of soil acidification. This method is applicable to the trend analysis of farmland soil acidification in a large range and multiple time periods, reducing the time and cost of manual monitoring.

[0074] The present invention can flexibly adjust the parameters of the spatio-temporal model and the grid resolution according to the actual situation of the study area, and can incorporate other environmental variables (such as rainfall, vegetation cover, etc.) to further enhance the applicability and accuracy of the model. This high adaptability makes this method have wide application value under different regional and environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The above and additional advantages of the present invention will become obvious and easy to understand in the description of the embodiments in conjunction with the following drawings, where:

[0076] Figure 1 is a schematic flowchart of a method for constructing and trend diagnosing spatio-temporal sequence data of farmland soil acidification according to an embodiment of the present invention;

[0077] Figure 2 is a spatio-temporal empirical variation function integrating time variation parameters of a method for constructing and trend diagnosing spatio-temporal sequence data of farmland soil acidification according to an embodiment of the present invention;

[0078] Figure 3It is a separable and measurable spatio-temporal variation function diagram fitted for the construction and trend diagnosis method of farmland soil acidification spatio-temporal sequence data according to an embodiment of the present invention;

[0079] Figure 4 It is a schematic diagram of the full spatio-temporal observation grid prediction based on spatio-temporal Kriging interpolation for the construction and trend diagnosis method of farmland soil acidification spatio-temporal sequence data according to an embodiment of the present invention;

[0080] Figure 5 It is an example diagram for identifying the significance and sustainability trends of soil acidification in the construction and trend diagnosis method of farmland soil acidification spatio-temporal sequence data according to an embodiment of the present invention. Detailed implementation manners

[0081] In order to be able to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the drawings and specific implementation manners. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0082] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0083] As Figure 1 shown, this embodiment provides a method for constructing and trend diagnosing farmland soil acidification spatio-temporal sequence data, and the method includes:

[0084] S1. Collect the annual observed data of the pH value of farmland soil. The data includes spatial information, i.e., longitude and latitude or grid index, and time stamps, i.e., years; clean the above data, remove the error records and invalid points outside the research area, and ensure the integrity and accuracy of the data.

[0085] Construct a spatio-temporal sparse grid data framework, unify and integrate the spatial and temporal dimensions, and mark the missing points to support subsequent interpolation processing.

[0086] Specifically, the first step of the system operation is to integrate the scattered observed data of soil acidity and alkalinity (pH value) into an ordered and easy-to-analyze "skeleton"; to achieve this goal, it is first necessary to collect the annual soil pH observations from various sources (such as monitoring stations or experimental data over the years), and these data are usually stored in tabular form, and each record must carry clear spatial location information (such as longitude and latitude) and time marks (years or dates); however, there are often certain problems with the data: some records may be distorted due to collection errors, and some points are repeated. Therefore, it is necessary to remove the invalid information and ensure the integrity of the data through an automated program or manual cleaning.

[0087] Subsequently, by constructing a spatio-temporal index, each piece of valid data is accurately located in the corresponding geographical location and time dimension, thus forming a sparse but clear three-dimensional spatio-temporal grid. The construction of this grid lays the foundation for subsequent interpolation analysis, just like finding the border for a jigsaw puzzle.

[0088] S2. Calculate the spatio-temporal empirical variogram. By introducing spatio-temporal variation parameters and combining the variation characteristics of space and time, calculate the spatio-temporal empirical variogram of soil pH value. This step is to understand the variation law of soil acidification under different spatial and time conditions.

[0089] Figure 2 Figure for the spatio-temporal empirical variogram, showing the variation law of soil acidification data in the time and space dimensions. The abscissa is the spatial distance h between the observation points in the spatio-temporal empirical variogram s , and the ordinate is the time lag h t , and the color scale represents the empirical estimate of the spatio-temporal variogram of soil pH value The empirical variogram is a mathematical tool used to describe the similarity or difference between observation points. By comparing the variation ranges of soil acidity and alkalinity (pH value) at different locations and times, it reveals the spatial and temporal correlations of the data. In this embodiment, Figure 2 the variogram curve in reflects how to combine the variation characteristics of time (such as the annual acidification trend) with the differences in space (such as the pH value differences between different plots) for modeling, intuitively presenting the spatio-temporal covariance characteristics of the data, and is an important basis for constructing the spatio-temporal variation model.

[0090] When all spatio-temporal data are integrated into the grid, the system needs to reveal the internal law of soil acidification between different regions and time points. Assume that the soil pH value is represented by a Gaussian spatio-temporal random field defined on the spatial domain and the time domain . For the soil pH value samples PH(s n ,t n ) observed at different spatio-temporal positions (s n ,t n ) ∈ S × T, it may include repeated measurements at the same location or simultaneous measurements at different spatial positions.

[0091] To describe the correlation between these spatio-temporal sample data, introduce the spatio-temporal covariance function: For the spatio-temporal random process {PH(s,t): s ∈ R 2 , t ∈ R}, under the assumption of second-order stationarity, the covariance function of the spatio-temporal random field PH(s,t) only depends on the spatial distance h s between two observation points and the time lag h t, which is independent of the specific spatio-temporal positions of these two observation points, the spatio-temporal covariance function is as follows:

[0092] C(h s ,h t ) = Cov(PH(s + h s ,t + h t ), PH(s,t));

[0093] To more intuitively describe the spatio-temporal variability of soil pH values, the time variation parameter is incorporated into the spatial variation model, and the spatio-temporal variation function γ(h s ,h t ) is introduced to measure the degree of difference between random variables:

[0094]

[0095] For the spatio-temporal dataset of soil PH values, the empirical estimate of its spatio-temporal covariance can be expressed as:

[0096]

[0097] where N(h s ,h t ) is the set of soil PH value observation point pairs that satisfy the spatial offset h s and the time lag h t , is the number of such observation point pairs, PH(s,t) represents the soil pH value measured at position s and time t, μ PH (h s ,h t ) and μ PH (0,0) are the empirical means at the sampling point offset and the origin, respectively.

[0098] Similarly, the empirical estimate of the spatio-temporal variation function of soil pH values is:

[0099]

[0100] To make the covariance function meaningful, the function C(h s ,h t ) must satisfy the positive definiteness condition, that is, to ensure the non-negativity and invertibility of the covariance matrix. For any real numbers a i (i = 1,…,n) and any spatio-temporal points (s i ,t i ), there is:

[0101]

[0102] S3. Construct spatio-temporal variogram models. In this embodiment, two main spatio-temporal variogram models are constructed: the separable model and the metric model.

[0103] After step S2, the data pattern is initially clarified, and the system enters the crucial modeling stage. Traditional analysis often only focuses on spatial correlation and ignores the important influence of time on data. Through the constructed separable model and metric model, this system deeply explores the dynamic characteristics of soil acidification. The separable model assumes that the spatial and temporal variations are independent, while the metric model takes into account the correlation between spatial and temporal variations. By optimizing these models, the spatio-temporal variogram can be more accurately fitted.

[0104] In the separable model, the spatio-temporal variation of soil pH value is considered to vary independently in space and time. The model decomposes the spatio-temporal variogram into a linear combination of a pure spatial variogram and a pure temporal variogram; it is assumed that the spatio-temporal covariance function of soil pH value can be expressed as the product of a spatial covariance function and a temporal covariance function, that is:

[0105] C sep (h,u) = C s (h)·C t (u);

[0106] where h represents the spatial distance and u represents the time lag. Then the corresponding variogram form is:

[0107]

[0108] are the standardized spatial and temporal variograms respectively, with a separate nugget effect, and the joint parameter in the sill value is set to 1. The overall sill value is defined by the sill parameter; in the R language (the R language is a free software programming language and operating environment, mainly used for statistical analysis, plotting, and data mining), such models can be constructed using the gstat package; first, the parameters of the spatial and temporal variograms are set respectively (such as part of the sill value, model type, range, and nugget effect), and then the separable model is specified through the vgmST function (stModel = "separable"), and the fit.StVariogram function is used to fit the empirical variogram to optimize the model parameters.

[0109] The metric model assumes that the spatio-temporal variability of soil pH values is jointly considered, that is, the spatial and temporal correlations are not completely independent and depend on an anisotropy parameter to adjust the relative contributions of the spatial and temporal scales. The metric covariance model extends the two-dimensional geographical space to three-dimensional spatio-temporal by introducing an anisotropy parameter (κ) to match the spatial and temporal scales. The spatio-temporal distance is obtained from the spatial distance (h) and the time lag (u) after being corrected by κ, in the following form:

[0110]

[0111] The corresponding variogram is:

[0112]

[0113] When constructing the metric model in R language, first define a variogram model including the nugget effect (such as Gaussian, spherical, exponential models, etc.), specify the metric model (stModel = "metric") through the vgmST function, and use the joint variogram and the anisotropy parameter as inputs. Also use the fit.StVariogram function to fit the empirical variogram, optimize the model parameters to minimize the fitting error (such as the mean square error MSE), in order to find the optimal variogram parameters. By comparing the MSE values or other statistical indicators of different models, select the optimal model.

[0114] Figure 3 It shows the complex dynamic characteristics of the soil acidification trend described by the separable model and the metric model, Figure 3 divided into separable on the left half side and metric on the right half side. The separable on the left half side represents the separable model constructed in step S3, and the metric on the right half side represents the metric model constructed in step S3.

[0115] Figure 3 The abscissa on the left half side is the spatial distance h between the observation points in the spatio-temporal empirical variogram fitted based on the separable model in step S3 s , and the ordinate is the time lag h t ; Figure 3 The abscissa on the right half side is the spatial distance h between the observation points in the spatio-temporal empirical variogram fitted based on the metric model in step S3 s , and the ordinate is the time lag h t ; The color scale represents the empirical estimate of the spatio-temporal variogram of the soil PH value after fitting by different models

[0116] S4. Construction and Kriging interpolation of spatio-temporal prediction grid: By constructing a unified spatio-temporal grid, combining spatial and temporal data, and performing spatio-temporal Kriging interpolation to predict soil acidification in unsampled areas. This step combines error analysis and optimization debugging to ensure the accuracy of the interpolation model.

[0117] As Figure 4 shown, through the optimized model, the system extends the prediction tentacles to the entire study area, covering both the monitored plots and the areas with missing data. This step is similar to connecting sparse point data into a surface, mapping each plot and time point to a virtual map. To achieve this goal, the system first divides the spatial grid according to the geographical characteristics of the study area, and sets the prediction time interval according to the time span. Subsequently, these spatial grids and time intervals are integrated into a complete spatio-temporal prediction grid. At each point of the grid, the system performs interpolation calculations based on the influence degree of surrounding data points; finally, the complete spatio-temporal prediction grid shows the soil acidification status of the entire area and its evolution trajectory over time.

[0118] The Spatio-Temporal Full Observation Grid (STF) considers both spatial and temporal dimensions. In terms of the spatial grid, first determine a unified spatial reference system (such as the WGS84 geographic coordinate system / Krassovsky projection coordinate system), divide the spatial grid according to the size of the study area and the required accuracy, and determine the resolution of the grid to generate a SpatialPixelsDataFrame pixel point, where each column corresponds to different longitude and latitude spatial features, so that each pixel point can be accurately located in space, and each point represents a potential farmland soil pH value observation data. In terms of the time grid, determine the time range in the time dimension, and evenly divide this period into several time periods according to the data accuracy, and each time period represents a prediction time point. Then combine the spatial grid and the time grid to create a spatio-temporal prediction grid (STF). Each grid point corresponds to a specific spatial position and time point for subsequent spatio-temporal Kriging interpolation prediction.

[0119] Using the fitted optimal spatio-temporal variogram model, call the krigeST function on the constructed Spatio-Temporal Full Observation Grid (STF) for spatio-temporal Kriging interpolation to predict the spatial distribution of farmland soil acidification and its change trend over time. Specifically, for any point (s0, t0) on the STF grid, the spatio-temporal Kriging interpolation prediction value PH(s0, t0) of its soil pH value can be given by the following formula:

[0120]

[0121] In the formula: PH(s i , t i) is the pH observed value of the soil sampling point, and PH(s0,t0) is the estimated value at the spatio-temporal point (s0,t0); n is the number of observed points participating in the prediction, and λ i is the weight coefficient of the neighboring observed value PH(s i ,t i ), and μ is the introduced Lagrange coefficient. Among them, λ i needs to ensure the minimum variance of the prediction error by solving the weight coefficient and satisfy the unbiasedness condition. This is usually transformed into solving the following optimization problem: Meanwhile, satisfy the unbiasedness condition: The Lagrangian function can be constructed as follows:

[0122]

[0123] Let L(λ i ,μ) take the partial derivatives with respect to λ i and μ and set them to 0, and then solve for the weight coefficient λ i .

[0124] After interpolation, it is necessary to evaluate the spatio-temporal Kriging interpolation results based on the parameters. Based on the evaluation results, we need to verify the model and optimize and debug the parameters as needed to improve the prediction performance of the model. The following are some specific steps: First, model verification: Check whether the model satisfies the unbiasedness condition, that is, whether the average value of the predicted values is close to the average value of the observed values. Analyze the error distribution and check for systematic biases or outliers. Evaluate the stability and consistency of the model to ensure reasonable prediction results can be obtained for different data subsets or time points. Second, adjust and optimize the spatio-temporal empirical variogram parameters: According to the evaluation results and prior knowledge, adjust the parameters such as the partial sill, range, and nugget of the spatio-temporal variogram to improve the goodness of fit and prediction accuracy of the model. Third, explore different model types: Try different spatio-temporal variogram models (such as separable models, metric models, etc.), compare their differences in prediction performance, and select the most suitable model for the current data. Fourth, consider covariates: If the data contains other covariates that may be related to the soil pH value (such as soil type, rainfall, vegetation cover, etc.), they can be incorporated into the model as part of the fixed effect or random effect to improve the explanatory power and prediction accuracy of the model. Fifth, iterative optimization: Compare and analyze the preliminary evaluation results with the new prediction results after parameter optimization and debugging to determine whether the optimization is effective. Evaluate the accuracy and reliability of the spatio-temporal Kriging interpolation by comparing the prediction results with the data of known observed points, including calculating statistical indicators such as the mean squared error (MSE). If necessary, the above steps can be repeated for multiple iterative optimizations until satisfactory prediction performance is achieved.

[0125] S5. Soil acidification trend analysis: Based on the Theil-Sen Median trend analysis and non-parametric Mann-Kendall test of time series raster images, a soil acidification trend recognition model is established to evaluate the significance of farmland soil acidification trends.

[0126] Merely plotting the spatial distribution map of soil acidification is far from enough. The system further delves into whether these changes are meaningful and whether there are persistent trends. Two tools are introduced here: one is the Theil-Sen trend analysis method, which is used to calculate the change slope of the pH value at each grid point over time; the other is the Mann-Kendall test, which is used to evaluate whether these trends are significant.

[0127] When identifying the acidification trend of soil pH values, the combination of Theil-Sen Median trend analysis and Mann-Kendall test methods provides an effective and robust non-parametric statistical method. Theil-Sen Median evaluates trends by calculating the median slope, while Mann-Kendall is used to verify the significance of this trend. The formulas are as follows:

[0128]

[0129] S PH is the median slope calculated by the Theil-Sen Median trend analysis method and is used to evaluate the trend of the soil pH value time series. When S PH > 0, it indicates that the soil pH value shows an acidification trend; conversely, it indicates that the soil pH value tends to be alkalized.

[0130] Define the statistic S:

[0131]

[0132] Among them, the sign function sgn is defined as:

[0133]

[0134] Among them, pH i and pH j represent the soil pH values in the i-th year and the j-th year respectively, and n represents the length of the time series. At a given significance level α (α = 0.05), when |Z| > u 1-α / 2 , it indicates that there is a significant change trend in the soil pH value time series at the α level.

[0135] S6. Identification of Soil Acidification Sustainability: Construct a model for identifying soil acidification sustainability through the Hurst index, analyze the sustainability of the soil acidification trend, evaluate the persistence or anti-persistence trend of future soil acidification, and provide a basis for soil improvement and management.

[0136] After analyzing the trend, the system can also predict whether these changes will continue or reverse. This assessment is based on a mathematical tool called the Hurst index, whose core idea is to measure the long-term correlation of data. Through this assessment, the system can not only point out which plots require long-term attention for acidification but also provide alerts for potential risk areas.

[0137] The Hurst index is used to evaluate the long-range dependence of time series. The Hurst index can be used to evaluate the persistence trend characteristics of predicting future soil pH value changes. The calculation method is as follows:

[0138]

[0139] R(τ) = max 1≤t≤τ X(t, τ) - min 1≤t≤τ X(t, τ), τ = 1, 2, …, n;

[0140]

[0141] where is the mean time series of pH values, X(t, τ) is the cumulative deviation of pH values, R(τ) is the range, S(τ) is the standard deviation. If there is a relationship R(τ) / S(τ) ∝ τ H (where H is the Hurst index), it indicates that there is a Hurst phenomenon in the soil pH value time series. When 0.5 < H < 1, it shows that the soil pH value time series has persistence, that is, the soil acidification (or alkalization) trend may continue in the future, and the closer H is to 1, the stronger this persistence. When H = 0.5, it means that the soil pH value time series is random and there is no long-term correlation, that is, the change of soil pH value is unpredictable. When 0 < H < 0.5, it indicates that the soil pH value time series has anti-persistence, that is, the soil acidification (or alkalization) trend may be opposite to the past in the future, and the closer H is to 0, the stronger this anti-persistence.

[0142] As Figure 5 shown, the observed data of farmland soil pH values in a certain area and period are used for actual analysis. Finally, the Hurst index of each block is obtained, and the numerical values are segmented and marked in different colors on the map to visually display the diagnostic results.

[0143] Figure 5(a) is the Theil - Sen Median trend result in step S5, that is, based on the Theil - Sen Median trend of the time - series raster images over 12 years from 2009 to 2020. When S PH > 0, it indicates that the soil pH value shows an acidification trend; conversely, it indicates that the soil pH value tends to be alkalized.

[0144] Figure 5 (b) is the result after verifying the significance of the trend by Mann - Kendall in step S5. At a given significance level α (α = 0.05), when |Z| > u 1-α / 2 , it indicates that there is a significant change trend in the soil pH value time series at the α level. (Among them, extremely significant, significant, relatively significant, insignificant, no change, etc. are distinguished).

[0145] Figure 5 (c) is the classification result of the Hurst index value in the continuous trend feature of predicting the future change of soil pH value by the Hurst index evaluation in step S6.

[0146] Figure 5 (d) is the type area divided based on the judgment criteria of the Hurst index value in step S6. (Among them, strong continuous acidification, weak continuous acidification, anti - continuous acidification, etc. are distinguished).

[0147] This system transforms scattered and discrete soil acidification data into a complete and visual trend analysis chart and prediction report through multi - level and multi - dimensional technical means. This integrated and innovative method not only solves the blind spots in traditional monitoring but also provides a reliable basis for agricultural management, policy - making, and ecological protection.

[0148] To sum up, the present invention proposes a method for constructing and trend diagnosing spatio - temporal sequence data of farmland soil acidification, including:

[0149] S1. Data collection and pre - processing: Collect annual observed data of farmland soil pH value, including spatial information and time stamps; clean the data, remove error records and invalid points outside the study area to ensure the integrity and accuracy of the data; construct a spatio - temporal sparse grid data framework to uniformly integrate spatial and temporal dimensions.

[0150] S2. Calculate the spatio - temporal empirical variogram: Introduce spatio - temporal variation parameters, and calculate the spatio - temporal empirical variogram of soil pH value by combining spatial and temporal variation characteristics to reveal the variation law of soil acidification under different spatial and temporal conditions.

[0151] S3. Construct a spatio - temporal variogram model: Construct a separable model and a metric model, and optimize the parameters of the two models to more accurately fit the spatio - temporal variogram.

[0152] S4, Spatiotemporal Kriging Interpolation and Prediction: Construct a unified spatiotemporal grid, combine spatial and temporal data, and make predictions based on spatiotemporal Kriging interpolation; generate a full spatiotemporal observation grid to predict the spatial distribution of farmland soil acidification and its change trend over time.

[0153] S5, Trend Analysis: Based on the Theil-Sen Median trend analysis and non-parametric Mann-Kendall test of time series raster images, establish a soil acidification trend identification model to evaluate the significance of the farmland soil acidification trend.

[0154] S6, Sustainability Identification: Use the Hurst index to analyze the sustainability of the soil acidification trend and provide a basis for future soil improvement and management.

[0155] The steps in the present invention can be adjusted, combined, and deleted in order according to actual needs.

[0156] The units in the device of the present invention can be combined, divided, and deleted according to actual needs.

[0157] Although the present invention has been disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely exemplary and are not used to limit the application of the present invention. The protection scope of the present invention is defined by the appended claims and may include various modifications, improvements, and equivalent solutions made to the invention without departing from the protection scope and spirit of the present invention.

Claims

1. A method for constructing and diagnosing farmland soil acidification spatiotemporal series data, characterized in that: The method comprises: S1. Data collection and preprocessing: Collect annual farmland soil pH observation data, including spatial information and timestamps; clean the data, remove erroneous records and invalid points beyond the study area; construct a spatiotemporal sparse grid data framework, input the cleaned data into the spatiotemporal sparse grid data framework, and unify the spatial and temporal dimensions; S2. Calculate the spatiotemporal empirical variogram based on the collected data: Calculate the spatiotemporal variogram and empirical estimate of soil pH value separately, and put them together to form the spatiotemporal empirical variogram to quantify the variation of soil pH value under different spatial and temporal conditions; S3. Construct a spatiotemporal variogram model to fit the spatiotemporal variogram: Based on the calculation results of the spatiotemporal variogram, a separable model and a measurement model are systematically constructed to model the spatiotemporal changes of soil acidification, and the parameters of the two models are continuously optimized to fit a more accurate spatiotemporal variogram; in the separable model, the spatiotemporal covariance function of soil pH is expressed as the product of the spatial covariance function and the temporal covariance function, and the measurement model expands the two-dimensional geographic space to three-dimensional space-time, and matches the spatial and temporal scales by introducing anisotropy parameters (κ); S4. Spatiotemporal prediction through spatiotemporal kriging interpolation: Construct a unified spatiotemporal grid, combine the spatial and temporal information in the farmland soil pH observation data, and predict the farmland soil pH value based on spatiotemporal kriging interpolation; generate a full spatiotemporal observation grid to predict the spatial distribution of farmland soil acidification and its changing trend over time; S5. Soil acidification trend analysis: Based on the collected data, a soil acidification trend identification model was established through the Theil-Sen Median trend analysis of time series raster images and the non-parametric Mann-Kendall test to evaluate the significance of farmland soil acidification trends; S6. Sustainability identification: Use the Hurst index to analyze the sustainability of soil acidification trends and provide a basis for future soil improvement and management.

2. The method for constructing and diagnosing farmland soil acidification spatiotemporal series data according to claim 1, characterized in that: The specific process of calculating the spatiotemporal empirical variogram in step S2 includes: Assume that soil pH is defined in the spatial domain and time domain The Gaussian space-time random field on the n ,t n )∈S×T The observed soil pH sample PH(s n ,t n ), including repeated measurements at the same location or simultaneous measurements at different spatial locations; Introducing the space-time covariance function: For the space-time random process {PH(s,t):s∈R 2 ,t∈R}, under the assumption of second-order stationarity, the covariance function of the spatiotemporal random field PH(s,t) depends only on the spatial distance h between the two observation points s and time lag h t , and has nothing to do with the specific time and space positions of the two observation points, then the space-time covariance function is: C(h s ,h t )=Cov(PH(s+h s ,t+h t ),PH(s,t)); In order to more intuitively describe the spatiotemporal variability of soil pH, the temporal variation parameters were integrated into the spatial variation model and the spatiotemporal variation function γ(h s ,h t ) measures the degree of difference between random variables: For the spatiotemporal dataset of soil pH, the empirical estimate of its spatiotemporal covariance is expressed as: Among them, N(h s ,h t ) is to satisfy the spatial offset h s and time lag h t The set of soil pH observation point pairs, is the number of such observation point pairs, PH(s,t) represents the soil pH value measured at location s and time t, μ PH (h s ,h t ) and μ PH (0,0) are the empirical means at the sampling point offset and the origin respectively; Similarly, the empirical estimation of the spatiotemporal variogram of soil pH is for: In order to make the covariance function meaningful, the function C(h s ,h t ) must satisfy the positive definiteness condition, that is, to ensure the non-negative definiteness and invertibility of the covariance matrix. For any real number a i (i=1,…,n) and any space-time point (s i ,t i ) have: The above-mentioned spatiotemporal variation function γ(h s ,h t ) and empirical estimates Construct the spatiotemporal empirical variation function.

3. The method for constructing and diagnosing farmland soil acidification spatiotemporal series data according to claim 1, characterized in that: The specific process of constructing the spatiotemporal variogram model in step S3 includes: constructing a separable model and a measurement model; In the separable model, the spatiotemporal covariance function of soil pH is expressed as the product of the spatial covariance function and the temporal covariance function, that is: C sep (h,u)=C s (h)·C t (u); Among them, h represents the spatial distance and u represents the time lag, then the corresponding variation function form is: are standardized spatial and temporal variograms, with separated nugget effects, and the joint parameter in the sill is set to 1, and the overall sill is defined by the sill parameter; in R, the gstat package is used to build such a model; first, the parameters of the spatial and temporal variograms are set separately, then the separable model is specified by the vgmST function, and the empirical variogram is fitted using the fit.StVariogram function to optimize the model parameters; The metric model expands the two-dimensional geographic space to three-dimensional space-time, and introduces an anisotropy parameter (κ) to match the spatial and temporal scales. The space-time distance is obtained by correcting the spatial distance (h) and the time lag (u) by κ, as follows: The corresponding variogram is: When building a metric model in R language, first define a variogram model that includes the nugget effect, specify the metric model through the vgmST function, and use the joint variogram and anisotropy parameters as input. Also use the fit.StVariogram function to fit the empirical variogram, optimize the model parameters to minimize the fitting error, and find the optimal variogram parameters; select the optimal model by comparing the MSE values ​​or other statistical indicators of different models.

4. The method for constructing and diagnosing farmland soil acidification spatiotemporal series data according to claim 1, characterized in that: The process of predicting soil pH value by spatiotemporal Kriging interpolation in step S4 includes: Construct a spatiotemporal observation grid, which includes a spatial grid and a temporal grid. In terms of the spatial grid, a unified spatial reference system is determined. According to the size of the study area and the required accuracy, the spatial grid is divided, and the resolution of the grid is determined to generate SpatialPixelsDataFrame pixel points. Each point represents a potential farmland soil pH value observation data. In terms of the temporal grid, the time range is determined in the time dimension, and this period is evenly divided into several time periods in combination with the data accuracy. Each time period represents a prediction time point. Combine the spatial grid and the temporal grid to create a spatiotemporal prediction grid. Each grid point corresponds to a specific spatial position and time point. Using the optimal spatiotemporal variogram model fitted in step S3, the krigeST function is called on the constructed full spatiotemporal observation grid to perform spatiotemporal Kriging interpolation to predict the spatial distribution of farmland soil acidification and its changing trend over time. For any point (s0, t0) on the STF grid, the spatiotemporal Kriging interpolation prediction value of the soil pH value PH(s0, t0) can be given by the following formula: Where: PH(s i ,t i ) is the pH observation value at the soil sampling point, PH(s0,t0) is the estimated value at the time and space point (s0,t0); n is the number of observation points involved in the prediction, λ i is the neighboring observation value PH(s i ,t i ) weight coefficient; where λ i It is necessary to solve the weight coefficient to ensure that the variance of the prediction error is minimized and the unbiased condition is met, which is transformed into solving the following optimization problem: At the same time, the unbiasedness condition is satisfied: The Lagrangian function is constructed as follows: μ is the introduced Lagrange coefficient, let L(λ i ,μ) vs.λ i Find the partial derivative of μ and set it to 0, thus solving the weight coefficient λ i ; After the interpolation is completed, the spatiotemporal Kriging interpolation results need to be evaluated based on the parameters. Based on the evaluation results, the model needs to be verified and the parameters need to be optimized and debugged as needed to improve the prediction performance of the model.

5. The method for constructing and diagnosing the spatiotemporal sequence data of farmland soil acidification according to claim 4, characterized in that: The step S4 further comprises: after the interpolation is completed, performing the following steps: Model validation: Check whether the model meets the unbiased condition, that is, whether the average value of the predicted value is close to the average value of the observed value, analyze the error distribution, check whether there are systematic deviations or outliers, evaluate the stability and consistency of the model, and ensure that reasonable prediction results can be obtained on different data subsets or time points; Adjust and optimize the parameters of the spatiotemporal empirical variogram: According to the evaluation results and prior knowledge, adjust the parameters of the spatiotemporal variogram to improve the model's fit and prediction accuracy; Explore different model types: try different spatiotemporal variogram models, compare their differences in predictive performance, and choose the model that best suits the current data; Consider covariates: If the data contain other covariates that may be associated with soil pH, include them in the model as fixed effects or as part of random effects to improve the explanatory power and prediction accuracy of the model; Iterative optimization: Compare and analyze the preliminary evaluation results and the new prediction results after parameter optimization and debugging to determine whether the optimization is effective. By comparing the prediction results with the data of known observation points, the accuracy and reliability of spatiotemporal Kriging interpolation are evaluated. Repeat the above steps for multiple iterative optimizations until satisfactory prediction performance is achieved.

6. The method for constructing and diagnosing farmland soil acidification spatiotemporal series data according to claim 1, characterized in that: The soil acidification trend recognition model established in step S5 is as follows: S PH is the median slope calculated by the Theil-Sen Median trend analysis method, which is used to evaluate the trend of the soil pH time series. PH When it is >0, it indicates that the soil pH value is acidifying; Conversely, it indicates that the soil pH value tends to alkalize; Define the statistic S: Among them, the sign function sgn is defined as: Among them, pH i and pH j represent the soil pH values ​​in the i-th and j-th years, respectively, and n represents the length of the time series; At a given significance level α, α = 0.05, when |Z|>u 1-α / 2 , it indicates that there is a significant change trend in the soil pH time series at the α level.

7. The method for constructing and diagnosing farmland soil acidification spatiotemporal series data according to claim 1, characterized in that: The specific content of using the Hurst exponent to analyze the sustainability of the soil acidification trend in step S6 includes: The calculation method of the Hurst exponent is as follows: R(τ)=max 1≤t≤τ X(t,τ)-min 1≤t≤τ X(t,τ),τ=1,2,…,n; in, is the mean time series of pH value, X(t,τ) is the cumulative deviation of pH value, R(τ) is the range, S(τ) is the standard deviation, H is the Hurst index, if there is a relationship R(τ) / S(τ)∝τ H , it indicates that the soil pH time series has the Hurst phenomenon; When 0.5 < H < 1, it indicates that the soil pH value time series has persistence, that is, the soil acidification / alkalization trend may continue in the future, and the closer H is to 1, the stronger this persistence; When H = 0.5, it indicates that the soil pH value time series is random and there is no long-term correlation, that is, the change of the soil pH value is unpredictable; When 0 < H < 0.5, it indicates that the soil pH value time series has anti-persistence, that is, the soil acidification / alkalization trend may be opposite to the past in the future, and the closer H is to 0, the stronger this anti-persistence.

Images