Model construction method for predicting spatial and temporal change of pH of cultivated land soil

Abstract

The invention discloses a model construction method for predicting spatial and temporal change of pH of cultivated land soil, and relates to the technical field of soil environment monitoring, the method comprises the following specific steps: basic data acquisition and preprocessing: acquiring soil related data meeting requirements in at least two periods, and preprocessing the data; screening and quantifying driving factors: screening out soil types, crop harvest salt-based ion carrying amount and chemical fertilizer application intensity as key driving factors, performing quantitative assignment on each driving factor according to a corresponding rule, and converting the driving factors into quantitative data which can be incorporated into a model; according to the method, space-time two-dimensional verification and optimization are carried out when the model is constructed, the space dimension covers the sub-regions of different soil types and agricultural production modes in the research region, and the time dimension selects actually measured data at a certain age limit interval, so that the application range of the model is wide, and high prediction precision is ensured by repeatedly adjusting the weight coefficient of the driving factor.

Description

Technical Field

[0001] This invention relates to the field of soil environmental monitoring technology, specifically to a method for constructing a model to predict the spatiotemporal changes in pH of arable land soil. Background Technology

[0002] Soil pH is one of the core indicators for assessing arable land quality. Its changes directly affect soil microbial activity, nutrient conversion efficiency, element migration patterns, and crop growth status. In recent years, affected by both natural conditions and human agricultural production activities, soil pH in many arable land areas has changed significantly, and some areas have even experienced soil acidification, posing a threat to the stability of agricultural ecosystems. Against this backdrop, scientifically predicting the spatiotemporal changes in arable land soil pH has become an important prerequisite for formulating soil acidification prevention and control strategies and ensuring sustainable agricultural development.

[0003] In existing technologies, research on soil pH mainly focuses on compiling historical data and analyzing trends. While this can reflect changes in soil pH over a period of time, it has significant limitations and cannot effectively predict future spatiotemporal changes in soil pH. Furthermore, traditional prediction methods are often based on limited influencing factors and do not comprehensively consider the combined effects of natural conditions and human activities on soil pH, leading to significant discrepancies between prediction results and actual conditions. Moreover, due to the lack of analysis and modeling of large-scale, long-term series data, it is difficult to meet the needs of long-term, dynamic prediction for large-scale farmland soil acidification control. In the face of complex and ever-changing agricultural ecological environments, traditional technologies cannot provide accurate predictions of spatiotemporal changes in soil pH, making it difficult to provide strong support for scientific and rational soil acidification control decisions. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a model construction method for predicting the spatiotemporal changes of arable soil pH. This method collects multi-period data on arable soil pH, soil type, crop planting and harvesting, and fertilizer application. The soil pH data is preprocessed to ensure reliability and comparability. Regarding the screening and quantification of driving factors, both natural and anthropogenic factors are considered to identify key driving factors such as soil type, crop harvesting base ion removal, and fertilizer application intensity. Each factor is scientifically quantified and assigned values. An initial prediction model is constructed, and it is then spatiotemporally validated and optimized using measured data from different regions and periods to ensure high prediction accuracy in both spatial and temporal dimensions. Finally, independent measured data not involved in model construction are used to verify the optimized model. Only when the relative error rate between the predicted and measured values ​​meets the requirements and the coefficient of determination reaches a high level is the final prediction model output. This provides reliable technical support for the long-term dynamic prediction of arable soil pH over a large area and a scientific basis for the formulation of arable soil acidification control measures.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a model construction method for predicting the spatiotemporal changes of soil pH in arable land, the method comprising the following specific steps: Basic data collection and preprocessing: Collect soil-related data that meet the requirements for at least two periods and preprocess the data; Driver selection and quantification: Soil type, crop harvest base ion removal, and fertilizer application intensity were selected as key driver factors. Each driver factor was quantified and assigned a value according to the corresponding rules, and then converted into quantitative data that can be included in the model. Initial model construction: Using standardized soil pH change as the dependent variable and quantified driving factors as independent variables, an initial prediction model was constructed using a spatiotemporal coupling prediction formula; multiple linear regression analysis was performed on the effective data using statistical analysis software to determine the initial weight coefficients of each driving factor and the model constant term, thereby achieving coupled prediction in time and space dimensions. Model spatiotemporal validation and optimization: The initial model was validated in both spatial and temporal dimensions using measured soil pH data from different sub-regions and at different times. The weight coefficients of the driving factors were adjusted based on the relative error rate and the validation was repeated to obtain the optimized model. Model validation and output: The optimized model is validated using independent measured data that were not involved in the model construction and validation. If the error rate and coefficient of determination requirements are met, the final model for predicting the spatiotemporal variation of farmland soil pH is output.

[0006] Furthermore, in the basic data collection and preprocessing steps, the soil-related data includes soil pH data, soil type data, crop planting and harvesting data, and fertilizer application data; data preprocessing includes standardizing the soil pH data by referring to verified conversion relationships to uniformly convert pH values ​​measured at different soil-water ratios to pH values ​​under the same soil-water ratio conditions; and using statistical software to identify and remove outliers in the crop harvesting data and fertilizer application data to complete data cleaning and retain valid data for subsequent analysis.

[0007] Furthermore, in the driving factor screening and quantification step, soil type, crop harvest basic ion removal, and fertilizer application intensity are selected as key driving factors. Each driving factor is quantified and assigned a value according to corresponding rules. Soil type is assigned a value based on the acid-base buffering capacity of different soil types and the historical pH change range. A higher value indicates that the soil type is more sensitive to pH changes. Crop harvest basic ion removal is calculated by combining collected crop yield data and basic ion concentration data per unit yield, and is used as a quantitative indicator of this driving factor. Fertilizer application intensity is quantified by the amount of fertilizer applied per unit cultivated land area. At the same time, the amount of nitrogen fertilizer applied per unit cultivated land area is separately calculated and included as a sub-item of the fertilizer application intensity driving factor in subsequent model construction.

[0008] Furthermore, the quantitative formula for the soil type is: ,in, It is the first The pH change sensitivity coefficient for different soil types; the higher the value, the stronger the sensitivity. It is a buffer performance weighting coefficient. It is the first The acid-base buffering capacity of different soil types It is a historical change weighting coefficient. It is the first Historical pH changes for each soil type.

[0009] Furthermore, the quantitative formula for the amount of basic ions carried away by the crop during harvest is as follows: ,in, It is the first Total amount of basic ions carried away by crops in each region It is the first The first region The yield of crops, It is the first The content of basic ions per unit yield of a crop. It is the first The first region The proportion of planting area for each crop It represents the total number of crop species in the region.

[0010] Furthermore, the quantitative formula for the fertilizer application intensity is as follows: ,in, It is the first The comprehensive index of fertilizer application intensity in each region It is the weight of the total amount of fertilizer applied. It is the first Total fertilizer application in each region It is the first Nitrogen fertilizer application rate in each region It is the first The arable land area of ​​each region It is the nitrogen fertilizer weight, reflecting the special impact of nitrogen fertilizer on soil acidification.

[0011] Furthermore, in the initial model construction step, a spatiotemporal coupling prediction formula is used to construct the initial prediction model, the formula of which is: ,in, It is in time Spatial coordinates ( , Soil pH change at location ) , , These are the quantitative values ​​for soil type, crop, and fertilizer driving factors, respectively. , , The weighting coefficients of the corresponding driving factors, , , These are the numbers for soil type, crop type, and regional division. It is the spatiotemporal residual term, reflecting minor influencing factors not included in the model.

[0012] Furthermore, in the model spatiotemporal validation and optimization steps, the initial model is validated in both spatial and temporal dimensions using measured soil pH data from different sub-regions and different periods. For spatial validation: measured soil pH data from different sub-regions within the study area are selected, substituted into the initial model to calculate predicted values, and the predicted values ​​are compared with the measured values ​​to calculate the relative error rate. For temporal validation: measured soil pH data from intermediate periods within the study period that were not involved in model construction are selected, substituted into the initial model to calculate predicted values, and the predicted values ​​are compared with the measured values ​​to calculate the relative error rate. If the relative error rate of the model in spatiotemporal validation is greater than 1%, the influence of each driving factor on the prediction error is analyzed based on the validation results, and the weight coefficients of each driving factor are adjusted. After adjustment, spatiotemporal validation is performed again, and the relative error rate is calculated. The adjustment and validation process is repeated until the relative error rate of the model is ≤1%, resulting in the optimized spatiotemporal change prediction model for cultivated land soil pH.

[0013] Furthermore, in the model spatiotemporal verification and optimization step, the influence of each driving factor on the prediction error is analyzed based on the verification results, and the weight coefficients of each driving factor are adjusted. The calculation formula is as follows: ,in, It is the adjusted number Each driving factor weight coefficient It is the first before the adjustment Each driving factor weight coefficient It is to adjust the step size coefficient. This is the pH value predicted by the model. It is the measured pH value, and sgn(·) is the sign function, which adjusts the weights according to the direction of the prediction deviation.

[0014] Furthermore, in the model verification and output step, independent measured data that did not participate in model construction and spatiotemporal verification are selected and substituted into the optimized model to calculate the predicted value; the predicted value is compared with the measured value of the independent measured data. If the relative error rate between the two is ≤1% and the determination coefficient of the model is ≥0.99, the model verification is passed; the verified spatiotemporal change prediction model of farmland soil pH is output.

[0015] Compared with existing technologies, this method for constructing a model to predict the spatiotemporal changes in arable land soil pH has the following advantages: I. This invention performs spatiotemporal dual-dimensional verification and optimization during model construction. The spatial dimension covers sub-regions with different soil types and agricultural production patterns within the study area, while the temporal dimension selects measured data at certain intervals, making the model widely applicable. Through repeated adjustments to the weight coefficients of driving factors, high prediction accuracy is ensured, which can meet the needs of long-term dynamic prediction of soil pH in large-scale arable land. The final output model results can provide a reliable basis for the formulation of soil acidification prevention and control measures, and strongly contribute to the stable development of agricultural ecosystems.

[0016] Second, this invention screens driving factors that cover key natural and anthropogenic factors such as soil type, crop harvesting of basic ions, and fertilizer application intensity, which conform to the actual impact mechanism of soil pH changes, ensuring the scientific validity and rationality of the model. Furthermore, the model has undergone verification and optimization in both spatial and temporal dimensions, making it widely applicable and highly accurate in prediction. It can meet the needs of long-term dynamic prediction of soil pH in large-scale cultivated land, providing a strong basis for the prevention and control of soil acidification in cultivated land.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 A flowchart illustrating a model construction method for predicting the spatiotemporal variation of soil pH in arable land; Figure 2 This is a flowchart illustrating the steps involved in constructing a model for predicting the spatiotemporal variation of soil pH in arable land. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This invention provides a model construction method for predicting the spatiotemporal changes of arable land soil pH. It collects multi-period data on arable land soil pH, soil type, crop planting and harvesting, and fertilizer application. The soil pH data is preprocessed to ensure reliability and comparability. Regarding the screening and quantification of driving factors, both natural and anthropogenic factors are considered to identify key driving factors such as soil type, crop harvesting base ion removal, and fertilizer application intensity. Each factor is scientifically quantified and assigned values. An initial prediction model is constructed, and it is then spatiotemporally validated and optimized using measured data from different regions and periods to ensure high prediction accuracy in both spatial and temporal dimensions. Finally, independent measured data not involved in model construction are used to verify the optimized model. Only when the relative error rate between the predicted and measured values ​​meets the requirements and the coefficient of determination reaches a high level is the final prediction model output. This provides reliable technical support for the long-term dynamic prediction of arable land soil pH over a large area and provides a scientific basis for the formulation of arable land soil acidification control measures.

[0022] Example 1 like Figure 1 As shown, soil pH data for a certain region were collected at two different periods. Sampling ensured coverage of topsoil corresponding to different landforms such as plains and hills within the region, and basic information such as sampling depth and time was recorded for each sampling point. Simultaneously, authoritative soil type distribution maps for the region were collected, marking the distribution range of each soil type and its corresponding acid-base buffering capacity parameters. Detailed agricultural production data, including annual planting area and actual yield records of major crops, as well as total fertilizer application, nitrogen fertilizer application, and fertilization time, were also obtained. Soil pH data was standardized by unifying pH values ​​measured using different soil-water ratios at different periods to the same standard through a pre-set water-soil ratio conversion rule. Crop harvest data was logically validated based on crop growth cycles and normal yield ranges. For fertilizer application data, obviously unreasonable extreme values ​​were removed by comparing with the region's agricultural production level, ultimately forming a complete and effective dataset.

[0023] Based on the survey results of soil characteristics in this region and the summary of long-term agricultural production practices, soil type, the amount of basic ions removed by crop harvest, and fertilizer application intensity were identified as the key driving factors affecting soil pH changes. For each soil type, the acid-base buffering capacity was determined by laboratory titration. Combined with the pH change range monitored over the past ten years, a formula was used to... Sensitivity levels were assigned and values ​​were assigned. Higher values ​​indicated that the pH of that soil type was more susceptible to external factors. It is the first The pH change sensitivity coefficient for different soil types; the higher the value, the stronger the sensitivity. It is a buffer performance weighting coefficient. It is the first The acid-base buffering capacity of different soil types It is a historical change weighting coefficient. It is the first Historical pH changes for various soil types were analyzed using a formula that considered crop harvest factors. First, the basic ion content per unit yield of different crops was obtained by testing crop samples. Then, the actual yield and planting area percentage of each crop were combined to determine the historical pH changes. The total amount of basic ions carried away by the region as a whole was obtained, among which, It is the first Total amount of basic ions carried away by crops in each region It is the first The first region The yield of crops, It is the first The content of basic ions per unit yield of a crop. It is the first The first region The proportion of planting area for each crop This refers to the total number of crop species in the region. Taking into account fertilizer application factors, the total fertilizer application per unit area of ​​cultivated land, and the significant impact of nitrogen fertilizer on soil acidification, a formula is used. Construct comprehensive indicators, among which, It is the first The comprehensive index of fertilizer application intensity in each region It is the weight of the total amount of fertilizer applied. It is the first Total fertilizer application in each region It is the first Nitrogen fertilizer application rate in each region It is the first The arable land area of ​​each region It is the nitrogen fertilizer weight, reflecting the special impact of nitrogen fertilizer on soil acidification.

[0024] The model uses the standardized soil pH change (the difference between the pH values ​​of the previous and subsequent periods) as the dependent variable, and the quantified soil type sensitivity coefficient, crop base ion removal, and fertilizer application intensity index as independent variables. Professional statistical analysis software is used to calculate the effective data, and the data undergoes normality testing to ensure compliance with modeling requirements. Finally, a spatiotemporal coupling prediction formula is employed. The initial weighting coefficients and model constants of each driving factor were determined to form an initial prediction model that can preliminarily reflect the trend of soil pH change in the region. It is in time Spatial coordinates ( , Soil pH change at location ) , , These are the quantitative values ​​for soil type, crop, and fertilizer driving factors, respectively. , , The weighting coefficients of the corresponding driving factors, , , These are the numbers for soil type, crop type, and regional division. It is the spatiotemporal residual term, reflecting minor influencing factors not included in the model.

[0025] Spatial dimension validation was performed by selecting representative sub-regions within the region. These sub-regions should cover different soil type distribution areas (e.g., clay soil areas, sandy soil areas) and different crop planting structure areas (e.g., monoculture areas, crop rotation areas). Measured soil pH data from these sub-regions were collected and substituted into the model, and the difference between predicted and measured values ​​was compared. Temporal dimension validation was performed by selecting measured soil pH data from 3-5 intermediate years within the study period. The predictive accuracy of the model at different time points was analyzed, and the relative error rate was calculated based on the validation results. If the error rate exceeded the preset range of 1%, a dynamic weight adjustment formula was used. Based on the direction of the error (predicted value is too high or too low) and the magnitude of the error, the weight coefficients of the corresponding driving factors are adjusted accordingly. It is the adjusted number Each driving factor weight coefficient It is the first before the adjustment Each driving factor weight coefficient It is to adjust the step size coefficient. This is the pH value predicted by the model. The measured pH value is sgn(·), which is the sign function. The weights are adjusted according to the direction of the prediction deviation. After adjustment, the model is re-verified. This process is repeated 3-5 times until the model prediction accuracy meets the requirements, resulting in an optimized prediction model.

[0026] Independent data not involved in the model construction and validation process were selected for validation. The data must cover different geographical zones and different time points within the region, and include measured soil pH values ​​corresponding to various soil types and crop planting patterns. The model performance was comprehensively evaluated by calculating indicators such as the relative error rate and goodness of fit between the predicted and measured values. If the relative error rate is ≤1% and the goodness of fit meets the preset standard, the final model for predicting the spatiotemporal change of arable land soil pH will be output. This model will be used to predict the changes in arable land soil pH in different plots within the region over the next 5-10 years, providing data support for the formulation of targeted acidification control measures.

[0027] Example 2 like Figure 2As shown, soil pH data for a certain region was obtained in two periods: one period from an early soil survey and the other from a recent farmland fertility evaluation. The sampling range for both types of data was ensured to cover the main farmland within the region, and the sampling depth was uniformly set to the standard depth of the topsoil layer. Soil type classification reports for the region were collected to clarify the physicochemical properties of each soil type, such as parent material and organic matter content. Simultaneously, statistical data on the planting varieties, growth cycles, and annual yields of major crops were obtained from the statistics department, along with detailed records of fertilizer application types, amounts, and methods in different years. Soil pH data was standardized, and systematic errors caused by different measuring instruments and personnel were eliminated based on pre-set correction rules for regional soil characteristics. For crop harvest data, outliers were identified by considering local climate conditions and crop variety characteristics. For fertilizer application data, unreasonable data was removed by referring to regional fertilization recommendations, ensuring the validity and consistency of the final dataset.

[0028] Based on the characteristics of agricultural production and soil environment in the region, soil type, crop harvest basic ion removal, and fertilizer application intensity were selected as the core driving factors affecting soil pH changes. When quantifying soil type, the influence of parent material and organic matter content on acid-base buffering capacity was comprehensively considered. At the same time, historical trend was determined by combining pH monitoring data from the past 15 years, and multi-dimensional value assignment standards were formulated. When calculating crop harvest basic ion removal, the proportion of basic ion absorption by different crops at different growth stages was first determined according to the crop nutrient absorption pattern. Then, the basic ion removal was calculated by crop type and summed after combining actual yield and planting area ratio to obtain the regional comprehensive value. When quantifying fertilizer application intensity, in addition to statistically analyzing the total fertilizer application per unit area, the frequency and single application amount of nitrogen fertilizer were analyzed. Appropriate weights were assigned according to the contribution of nitrogen fertilizer to soil acidification, and a fertilizer application intensity index that can accurately reflect the acidification risk was constructed.

[0029] The standardized change in soil pH was used as the dependent variable, while the quantified soil type, crop base ion removal, and fertilizer application intensity were used as independent variables. The processed effective data were modeled and analyzed using professional statistical analysis software. Before modeling, correlation analysis was performed on each variable to eliminate multicollinearity. Then, stepwise regression was used to screen variables that had a significant impact on pH changes. Finally, the initial weights and model constants of each driving factor were determined, and an initial model capable of predicting soil pH changes in the region was established.

[0030] Sub-regions with different topographic conditions and agricultural management models were selected within the region. Measured soil pH data from these sub-regions were collected for spatial verification to analyze the model's adaptability under different geographical environments. Soil pH monitoring data from different stages during the study period were selected for temporal verification to assess the model's predictive stability over long-term time series. The relative error rate was calculated based on the verification results. If the error rate did not meet the requirement of ≤1%, the weight coefficients of each driving factor were dynamically adjusted. For example, soil type has a greater impact on pH changes in mountainous sub-regions, so the weight of soil type can be appropriately increased. After adjustment, the data was re-substituted into the verification data for calculation. After 4-6 rounds of iterative optimization, the model's predictive accuracy reached the specified standard, resulting in an optimized model with higher stability and accuracy.

[0031] The model was validated using independent measured data that were not involved in its construction and verification. The data covered the distribution areas of all major soil types in the region and included soil pH values ​​from different monitoring months over the past 3-5 years to ensure the comprehensiveness and representativeness of the validation data. The model performance was evaluated by calculating indicators such as average error and goodness of fit. If all indicators met the preset accuracy standards, the final model for predicting the spatiotemporal variation of farmland soil pH was output. This model can be used for long-term dynamic prediction of farmland soil pH in the region and can also provide a scientific basis for agricultural departments to formulate differentiated acidification control measures and optimize planting structures.

[0032] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A model construction method for predicting the spatiotemporal variation of pH in cultivated soil, characterized by, The method comprises the following specific steps: Basic data collection and preprocessing: Collecting soil-related data meeting the requirements in at least two periods, and preprocessing the data; Driving factor screening and quantification: Screening soil type, crop harvested salt-based ion carrying amount, and fertilizer application intensity as key driving factors, quantitatively assigning each driving factor according to corresponding rules, and converting it into quantitative data that can be included in the model; Initial model construction: Taking the standardized soil pH change as the dependent variable, and the quantified driving factors as the independent variables, an initial prediction model is constructed by using a space-time coupling prediction formula; the effective data is subjected to multiple linear regression analysis by a statistical analysis software to determine the initial weight coefficient of each driving factor and the model constant term, and the coupling prediction of time and space dimensions is realized; Model space-time verification and optimization: The initial model is verified in the spatial and temporal dimensions by using the measured soil pH data of different sub-regions and different periods, the driving factor weight coefficient is adjusted according to the relative error rate and repeated verification, and the optimized model is obtained; Model verification and output: The optimized model is verified by using independent measured data not involved in the model construction and verification, and if the error rate and the determination coefficient requirements are met, the final prediction model of the cultivated land soil pH space-time change is output.

2. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 1, characterized in that, In the basic data collection and preprocessing step, the soil-related data includes soil pH data, soil type data, crop planting and harvesting data, and fertilizer application data; the data preprocessing includes standardizing the soil pH data, uniformly converting the pH values measured under different soil-water ratios to the pH values under the same soil-water ratio according to the verified conversion relationship, and identifying and removing the outliers in the crop harvesting data and the fertilizer application data by using a statistical software to complete data cleaning and retain effective data for subsequent analysis.

3. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 1, characterized in that, In the driving factor screening and quantification step, the soil type, the crop harvested salt-based ion carrying amount, and the fertilizer application intensity are screened as the key driving factors, and each driving factor is quantitatively assigned according to the corresponding rules, wherein the soil type is assigned according to the acid-base buffer performance and the historical pH change amplitude of different soil types, and the higher the assignment, the stronger the pH change sensitivity of the soil type; the crop harvested salt-based ion carrying amount is combined with the collected crop yield data and the salt-based ion concentration data per unit yield to obtain the salt-based ion carrying amount of different crops through statistical calculation as the quantitative index of this driving factor; the fertilizer application intensity takes the fertilizer application amount per unit cultivated land area as the quantitative index of the fertilizer application intensity, and separately counts the nitrogen fertilizer application amount per unit cultivated land area as a sub-item of the fertilizer application intensity driving factor for subsequent model construction.

4. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 3, characterized in that, The quantitative formula for the soil type is: ,in, It is the first The pH change sensitivity coefficient for different soil types; the higher the value, the stronger the sensitivity. It is a buffer performance weighting coefficient. It is the first The acid-base buffering capacity of different soil types It is a historical change weighting coefficient. It is the first Historical pH changes for each soil type.

5. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 3, characterized in that, The quantitative formula of the amount of salt-based ion taken away by the crop is: Wherein, is the total amount of salt-based ion taken away by the crop in the first region, is the yield of the first crop in the first region, is the salt-based ion content per unit yield of the first crop, is the planting area proportion of the first crop in the first region, is the total number of crop types in the region.

6. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 3, characterized in that, The quantitative formula of the chemical fertilizer application intensity is: Wherein, is the comprehensive index of the chemical fertilizer application intensity of the first area, is the total fertilization amount weight, is the total chemical fertilizer application amount of the first area, is the nitrogen fertilizer application amount of the first area, is the cultivated land area of the first area, is the nitrogen fertilizer weight, reflecting the special influence of nitrogen fertilizer on soil acidification.

7. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 1, characterized in that, In the initial model construction step, an initial prediction model is constructed using a space-time coupling prediction formula, which is as follows: wherein, is the change in soil pH at time , spatial coordinates , , , , are the quantified values of soil type, crop, and chemical fertilizer driving factors, , , are the weight coefficients of the driving factors, , , are the numbers of soil types, crop types, and regional divisions, is a space-time residual term, reflecting small influencing factors not included in the model. 8.The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 1, characterized in that, In the model space-time verification and optimization step, the measured soil pH data of different sub-regions and different periods are used to verify the initial model in space and time dimensions. For the space dimension verification, the measured soil pH data of different sub-regions in the study area are selected to calculate the predicted values by substituting into the initial model, and the relative error rate is calculated by comparing the predicted values with the measured values. For the time dimension verification, the measured soil pH data of the intermediate period not involved in the model construction in the study period are selected to calculate the predicted values by substituting into the initial model, and the relative error rate is calculated by comparing the predicted values with the measured values. If the relative error rate of the model in the space-time verification is greater than 1%, the influence degree of each driving factor on the prediction error is analyzed according to the verification result, and the weight coefficient of each driving factor is adjusted. After adjustment, the space-time verification is performed again, and the relative error rate is calculated. The adjustment and verification process is repeated until the relative error rate of the model is ≤1%, and the optimized cultivated land soil pH space-time change prediction model is obtained.

9. The model construction method for predicting the temporal and spatial variation of pH of cultivated soil according to claim 8, characterized in that, In the model spatiotemporal verification and optimization step, the influence of each driving factor on the prediction error is analyzed based on the verification results, and the weight coefficients of each driving factor are adjusted. The calculation formula is as follows: ,in, It is the adjusted number Weight coefficients of each driving factor It is the first before the adjustment Weight coefficients of each driving factor It is to adjust the step size coefficient. This is the pH value predicted by the model. It is the measured pH value, and sgn(·) is the sign function, which adjusts the weights according to the direction of the prediction deviation.

10. The model construction method for predicting the spatiotemporal variation of pH in cultivated soil according to claim 1, characterized in that, In the model verification and output step, independent measured data not involved in the model construction and space-time verification are selected to calculate the predicted values by substituting into the optimized model. The relative error rate of the predicted values and the measured values of the independent measured data is ≤1%, and the determination coefficient of the model is ≥0.99, and the model verification is passed. The cultivated land soil pH space-time change prediction model that passes the verification is output.

Images