Cross-time and space estimation method for influence degree of fertilization behavior on soil properties and related equipment

Abstract

The application is suitable for the field of soil science and technology, and provides a method for estimating the influence degree of fertilization behavior on soil properties across space and time and related equipment, which comprises the following steps: obtaining initial data of a to-be-predicted area; initializing a farmland grid of the to-be-predicted area, and performing data matching and multi-dimensional clustering partitioning on geographic position data, soil data and meteorological data in the initial data and the farmland grid to obtain a plurality of farmland space-time units corresponding to the to-be-predicted area; constructing a quantitative relationship equation of a planting mode corresponding to each farmland space-time unit for replacing the influence of fertilization behavior on soil properties and performing space-time recursion to determine the influence degree prediction result of the fertilization behavior on the soil properties in each farmland space-time unit of the to-be-predicted area at different time periods, and determining the target global prediction result of the to-be-predicted area according to the influence degree prediction result of all farmland space-time units. The application can realize the estimation of the influence degree of the fertilization behavior on the soil properties in each farmland space-time unit at different target time periods.

Description

Technical Field

[0001] This invention belongs to the field of soil science and technology, and in particular relates to a method and related equipment for cross-temporal estimation of the impact of fertilization behavior on soil properties. Background Technology

[0002] Soil is the material foundation upon which human beings depend for survival and development, and the quality of farmland soil, in particular, is directly related to national food security. In recent decades, land reclamation rates, intensive farmland use, and excessive fertilization have been continuously increasing. Understanding the impact of fertilization practices on farmland soil properties such as organic matter and total nitrogen is of significant decision-making value for refined farmland management, including soil quality and carbon neutrality.

[0003] While current location-based observational experiments under specific soil and climate conditions can reveal the impact of fertilization on farmland soil properties, the soil-crop transmission system is complex, making extensive surveys of farmland fertilization behavior a laborious, time-consuming, and difficult task. Furthermore, due to the limited number of location-based observation points and their location dependence on specific soil and climate conditions, the effects of fertilization behavior on farmland soil properties vary significantly across different geographical, meteorological, and soil environmental zones. In addition, the mismatch between the standardization and stability of observational experiments and the complexity of actual planting management structures leads to considerable uncertainty in the spatiotemporal variability of fertilization behavior on soil properties obtained under location-based observation conditions when extended to other regional scales. This is especially true in areas with high intensive farmland use and diverse planting structures, where farmland fertilization data is scarce, making it impossible to accurately determine how fertilization behavior influences multidimensional spatiotemporal farmland soil properties. Summary of the Invention

[0004] This application addresses the problem in the prior art that it is impossible to effectively analyze the impact of fertilization behavior on soil properties under different spatiotemporal conditions. It provides a spatiotemporal estimation method for the degree of influence of fertilization behavior on soil properties, aiming to provide a methodological basis for realizing the quantitative estimation of the impact of fertilization behavior on the variation of farmland soil properties across spatiotemporal periods.

[0005] In a first aspect, embodiments of this application provide a method for estimating the spatiotemporal impact of fertilization behavior on soil properties, the method comprising the following steps:

[0006] Acquire initial data for the area to be predicted, including fertilization behavior data, geographic location data, soil data, meteorological data, planting patterns, and soil properties;

[0007] Initialize the farmland grid of the area to be predicted, and perform data matching and multi-dimensional clustering and partitioning of the geographical location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted.

[0008] A quantitative equation was constructed to establish the relationship between the planting pattern substitution fertilization behavior and soil properties corresponding to each spatiotemporal unit of farmland in the initial period.

[0009] Based on the quantitative relationship equation, spatiotemporal recursion is performed to determine the predicted results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland in the region to be predicted under different target time periods.

[0010] Based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland under different target time periods, the target global prediction result of the area to be predicted is determined.

[0011] Furthermore, the initial data also includes remote sensing image data, and the acquisition of initial data for the area to be predicted further includes:

[0012] Acquire the remote sensing image data of the region to be predicted, extract multiple data features from the remote sensing image data, perform data fusion based on the multiple data features, and construct a training dataset;

[0013] Based on the training dataset, a spatial distribution prediction model is trained to obtain a target spatial distribution prediction model. The spatial distribution of planting patterns in the area to be predicted is determined through the target spatial distribution prediction model at different time periods.

[0014] Furthermore, the initialization of the farmland grid of the area to be predicted, and the matching and multi-dimensional clustering partitioning of the geographic location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted, includes:

[0015] Generate a corresponding farmland grid based on the area to be predicted;

[0016] The geographic location data, soil data, and meteorological data are analyzed using a preset correlation analysis algorithm, and the results are then distributed to the farmland grid.

[0017] Based on the time and space dimensions, the geographical location data, soil data and meteorological data allocated in the farmland grid are clustered and partitioned using a preset hybrid clustering model to obtain the spatiotemporal units of farmland in the area to be predicted.

[0018] Furthermore, the quantitative relationship equation for the impact of planting pattern substitution fertilization behavior on soil properties corresponding to each of the farmland spatiotemporal units in the initial time period includes:

[0019] Based on the initial data of multiple spatiotemporal units of farmland in different time periods, a random forest model of planting patterns and a random forest model of fertilization behavior are constructed.

[0020] During the initial and target time periods, the first influence value of the planting pattern on soil properties corresponding to the spatiotemporal unit of farmland is calculated using the planting pattern random forest model.

[0021] During the initial time period, the second influence value of fertilization behavior on soil properties corresponding to the spatiotemporal unit of farmland is calculated using the random forest model of fertilization behavior.

[0022] Based on the first influence value and the second influence value within the initial time period, a quantitative relationship equation between fertilization behavior and planting pattern is determined for each of the farmland spatiotemporal units. The output of the quantitative relationship equation includes the influence ratio of fertilization behavior on planting pattern.

[0023] Furthermore, the step of constructing a random forest model for planting patterns and a random forest model for fertilization behavior based on initial data from multiple spatiotemporal units of farmland within different time periods includes:

[0024] Using the geographic location data, soil data, meteorological data, and planting patterns of each spatiotemporal unit of farmland within the initial and target time periods as input variables, and the soil attributes as the dependent variable, a random forest model of the planting patterns is constructed.

[0025] Using the geographic location data, soil data, meteorological data, and fertilization behavior data of each spatiotemporal unit of farmland within the initial time period as input variables and the soil attributes as the dependent variable, a random forest model of fertilization behavior is constructed.

[0026] Furthermore, the step of performing spatiotemporal recursion based on the quantitative relationship equation to determine the predicted impact of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted includes:

[0027] Determine the planting pattern influence index of the global time series, and use the planting pattern influence index to recursively extrapolate based on the global time series, outputting the influence ratio through a quantitative relationship equation;

[0028] Based on the influence ratio output by the quantitative relationship equation corresponding to the spatiotemporal unit of farmland in the initial period, and the first influence value of the spatiotemporal unit of the target farmland in the target period, the initial influence prediction result of the fertilization behavior of the target farmland spatiotemporal unit on soil properties is calculated.

[0029] By recursively extrapolating from the spatial sequence, we obtain each target farmland spatiotemporal unit in the target time period and each initial farmland spatiotemporal unit in the initial time period corresponding to the target time period, and determine the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit.

[0030] The influence degree is weighted according to the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit. Combined with the initial influence degree prediction results, the influence degree prediction results of fertilization behavior on soil properties of each target farmland spatiotemporal unit are calculated.

[0031] Secondly, embodiments of this application provide a spatiotemporal estimation device for the influence of fertilization behavior on soil properties, the device comprising:

[0032] The data acquisition module is used to acquire initial data of the area to be predicted, including fertilization behavior data, geographical location data, soil data, meteorological data, planting patterns and soil properties;

[0033] The modeling module is used to initialize the farmland grid of the area to be predicted, and to perform data matching and multi-dimensional clustering and partitioning of the geographical location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted.

[0034] The relationship construction module is used to construct quantitative relationship equations on the impact of planting pattern substitution fertilization behavior on soil properties for each of the farmland spatiotemporal units in the initial period.

[0035] The influence determination module is used to perform spatiotemporal recursion based on the quantitative relationship equation to determine the predicted influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted.

[0036] The target prediction module is used to determine the global target prediction result of the area to be predicted based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland under different target time periods.

[0037] Furthermore, the initial data also includes remote sensing image data, and the data acquisition module includes:

[0038] The dataset construction subunit is used to acquire the remote sensing image data of the region to be predicted, extract multiple data features from the remote sensing image data, perform data fusion based on the multiple data features, and construct a training dataset.

[0039] The spatial distribution determination subunit is used to train a spatial distribution prediction model based on the training dataset to obtain a target spatial distribution prediction model, and to determine the spatial distribution of planting patterns in the area to be predicted at different time periods through the target spatial distribution prediction model.

[0040] Thirdly, embodiments of this application provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, it implements the steps of the method for estimating the influence of fertilization behavior on soil properties across time and space as described in the first aspect.

[0041] Fourthly, embodiments of this application provide a computer-readable storage medium, comprising: the computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of the method for estimating the influence of fertilization behavior on soil properties across time and space as described in the first aspect.

[0042] Fifthly, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to perform the steps of the method for estimating the spatiotemporal impact of fertilization behavior on soil properties as described in the first aspect.

[0043] The beneficial effects achieved by this invention are as follows: By gridding the area to be predicted, matching geographical location data, soil data, and meteorological data with farmland grids, and performing clustering and partitioning across multiple dimensions, the spatiotemporal unit partitioning of the influence of fertilization behavior on soil properties at different time periods is reasonably determined based on a comprehensive consideration of multiple influencing factors. Furthermore, by associating fertilization behavior with planting patterns and constructing quantitative relationship equations for regression, the applicability of fertilization behavior can be achieved. The influence of fertilization behavior on soil properties can be inferred based on the two influence ratios constructed using random forests, which can isolate the interaction effects of different variables and avoid the problems caused by the binarization of various data and the coupling of spatiotemporal multidimensional variables, thus failing to reflect the interaction relationships between multiple variables. This provides a methodological basis for effectively realizing cross-spatiotemporal quantitative estimation of the impact of fertilization behavior on farmland soil property variation. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic flowchart of a method for estimating the influence of fertilization behavior on soil properties across time and space, as provided in an embodiment of this application.

[0046] Figure 2 This is a schematic flowchart of step S10 provided in an embodiment of this application;

[0047] Figure 3This is a schematic diagram of the specific process of step S20 provided in the embodiments of this application;

[0048] Figure 4 This is a schematic diagram of the specific process of step S30 provided in the embodiments of this application;

[0049] Figure 5 This is another schematic flowchart of the method for estimating the influence of fertilization behavior on soil properties across time and space provided in the embodiments of this application;

[0050] Figure 6 This is a schematic diagram of the specific process of step S40 provided in the embodiments of this application;

[0051] Figure 7 This is a schematic diagram of a spatiotemporal estimation device for the influence of fertilization behavior on soil properties provided in an embodiment of this application;

[0052] Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0054] This application, by gridding the area to be predicted and matching geographic location data, soil data, and meteorological data with farmland grids, performs multi-dimensional clustering and partitioning based on multiple influencing factors to reasonably determine the spatiotemporal unit partitions of the impact of fertilization behavior on soil properties at different times. Furthermore, it correlates fertilization behavior with planting patterns, constructs quantitative relationship equations for regression, and realizes the applicability of fertilization behavior substitution. By inferring the impact of fertilization behavior on soil properties based on the two impact ratios constructed using random forests, it can isolate the interaction effects of different variables, avoiding the problems of binarization of various data and coupling of spatiotemporal multidimensional variables that prevent the reflection of the interaction relationships between multiple variables. This provides a methodological basis for effectively realizing cross-spatiotemporal quantitative estimation of the impact of fertilization behavior on farmland soil property variability.

[0055] Combination Figure 1 As shown, Figure 1 A flowchart of a method for estimating the spatiotemporal impact of fertilization behavior on soil properties, provided in an embodiment of this application, is shown. The method includes steps S10 to S40. The specific implementation principles of each step are as follows:

[0056] S10. Obtain initial data for the area to be predicted. The initial data includes fertilization behavior data, geographic location data, soil data, meteorological data, planting patterns, and soil properties.

[0057] In this embodiment, the electronic device used in the method for estimating the spatiotemporal impact of fertilization on soil properties can be connected to other devices via wired or wireless connections. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wide band) connections, and other currently known or future-developed wireless connection methods.

[0058] Specifically, the area to be predicted can represent the target area for cross-temporal estimation of the impact of fertilization behavior on soil properties. The target area can be a predefined polygonal region. Initial data can include multidimensional historical data of the area to be predicted over the past few years, which can be obtained from historical records. Multidimensional historical data can include fertilization behavior data, geographic location data, soil data, meteorological data, planting patterns, and soil properties.

[0059] More specifically, the aforementioned soil properties include organic matter content, total nitrogen content, etc., which can be obtained by collecting soil samples and analyzing them. It is possible that the latitude and longitude values ​​of the soil sample collection point can be recorded during sample collection, as different latitude and longitude values ​​correspond to different sample data; latitude and longitude can also serve as geographic location data. Fertilization behavior data includes the amount of organic fertilizer applied, the amount of chemical fertilizer used, and the amount of straw returned to the field. Geographical location data includes ground elevation, land use type, and the latitude and longitude of the soil sample location. Soil data includes soil type and topsoil texture. Meteorological data includes annual average temperature, annual rainfall, and accumulated temperature. Planting patterns include single-season rice, double-season rice, dryland crops, and other crops; for example, the soil type is paddy soil, and the planting pattern is single-season rice.

[0060] S20. Initialize the farmland grid of the area to be predicted, and perform data matching and multi-dimensional clustering and partitioning of the geographical location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted.

[0061] Specifically, to more accurately determine the corresponding data in each sub-region of the area to be predicted and improve the accuracy of the estimation, the area to be predicted can be divided into m×n grids to form a farmland grid. Then, correlation analysis can be performed on geographic location data, soil data, and meteorological data. Analysis methods include raster resampling, spatial interpolation, and spatial proximity. Based on the analysis results, the geographic location data, soil data, and meteorological data are assigned to the farmland grid, and the geographic location data, soil data, and meteorological data of each unit in the farmland grid are distributed accordingly based on the correlation analysis results. Furthermore, clustering modeling can be performed on the geographic location data, soil data, and meteorological data in different dimensions based on different time periods to obtain the farmland spatiotemporal units corresponding to different time periods in different dimensions. The farmland spatiotemporal units are clustered and partitioned based on both time and space dimensions.

[0062] S30. Construct a quantitative equation relating the effects of planting patterns to fertilization practices on soil properties in each spatiotemporal unit of farmland during the initial period.

[0063] Specifically, after obtaining the spatiotemporal units of farmland in the area to be predicted, these units can be used as basic units to construct an assessment model based on machine learning algorithms to evaluate the impact of planting patterns and fertilization behaviors on farmland soil properties. Then, based on the assessment models of the impact of planting patterns and fertilization behaviors on soil properties, an analysis of the relationship between planting patterns and fertilization behaviors is conducted to construct a quantitative equation representing the impact of planting patterns on soil properties, replacing fertilization behaviors. This quantitative equation expresses the quantitative relationship between planting patterns and fertilization behaviors on soil properties. In this embodiment, the aforementioned machine learning algorithm can be a Random Forest (RF) model. In this embodiment, Random Forest is a regression prediction ensemble method that includes multiple unrelated decision trees and outputs the class as the average prediction (regression).

[0064] S40. Based on the quantitative relationship equation, perform spatiotemporal recursion to determine the predicted impact of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted.

[0065] Specifically, after obtaining the quantitative relationship equation, it can be used as a basis for spatiotemporal recursion. The quantitative relationship factors of planting patterns and fertilization behaviors in a certain spatiotemporal unit of farmland can be extrapolated to other spatiotemporal units of farmland in other target time periods using a spatially weighted distribution. The predicted impact of fertilization behaviors on soil properties in the extrapolated spatiotemporal units of farmland in the target time periods can then be calculated. Based on the above method, the predicted impact of fertilization behaviors on soil properties in each spatiotemporal unit of farmland under different target time periods in the area to be predicted can be calculated.

[0066] S50. Based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland under different target time periods, determine the target global prediction results for the area to be predicted.

[0067] Specifically, after traversing all farmland spatiotemporal units for all target time periods and obtaining the predicted results of the influence of fertilization behavior on soil properties in each farmland spatiotemporal unit under different target time periods in the area to be predicted, the global prediction result of the target area can be calculated based on the prediction results of multiple farmland spatiotemporal units using the mean, variance, and other methods.

[0068] In this embodiment of the invention, by gridding the area to be predicted, geographic location data, soil data, meteorological data, and farmland grid are matched, and clustering and partitioning are performed across multiple dimensions. Based on a comprehensive consideration of multiple influencing factors, the spatiotemporal unit partitioning of the impact of fertilization behavior on soil properties at different time periods is reasonably determined. Furthermore, fertilization behavior is correlated with planting patterns, and a quantitative relationship equation is constructed for regression analysis to achieve applicability substitution of fertilization behavior. The impact of fertilization behavior on soil properties is inferred based on the two impact ratios constructed using random forests. This method can isolate the interaction effects of different variables, avoiding the problems of binarization of various data and coupling of spatiotemporal multidimensional variables that prevent the reflection of the interaction relationships between multiple variables. This provides a methodological basis for effectively realizing cross-spatiotemporal quantitative estimation of the impact of fertilization behavior on farmland soil property variation.

[0069] refer to Figure 2 As shown, in some implementations, the initial data also includes remote sensing image data, such as... Figure 2 As shown, step S10 above also includes:

[0070] S101. Obtain remote sensing image data of the area to be predicted, extract multiple data features from the remote sensing image data, perform data fusion based on multiple data features, and construct a training dataset.

[0071] S102. Train the spatial distribution prediction model based on the training dataset to obtain the target spatial distribution prediction model. Use the target spatial distribution prediction model to determine the spatial distribution of planting patterns in the area to be predicted at different time periods.

[0072] Specifically, the aforementioned remote sensing image data can be acquired using remote sensing equipment, such as drones. These drones are equipped with image acquisition devices, such as cameras or other similar devices. The drones can be controlled by handheld remote control devices or intelligent control terminals to acquire the remote sensing image data. Remote sensing satellite acquisition offers advantages such as comprehensive coverage across multiple time periods, high accuracy, and fast and convenient data acquisition. The acquired remote sensing image data facilitates a more accurate and rapid understanding of the distribution of planting patterns in different spatiotemporal ranges. This helps address the problem of insufficient spatiotemporal matching of the impact of medium-sized, long-term farmland fertilization practices.

[0073] More specifically, after acquiring remote sensing image data of the area to be predicted over a past period, some data features can be extracted from the remote sensing image data. These extracted features include farmland vegetation index features, color features, and texture features. The farmland vegetation index feature can be the ratio of two-channel reflectance. This index can be constructed using data from different satellite bands and reflects plant growth status. Based on the strong absorption characteristics of plant leaves in the visible red light band and their strong reflectance characteristics in the near-infrared band, different combinations of measurements from these two bands can yield different vegetation indices. After extracting the aforementioned farmland vegetation index features, color features, and texture features, these three types of features can be fused to construct a training dataset. The constructed training dataset can be used in machine learning algorithms such as Support Vector Machines (SVMs) and Random Forests. The training dataset can be input into SVM or Random Forest algorithms for model training, using planting patterns as the dependent variable. The optimal model after training outputs the spatial distribution of planting patterns in the area to be predicted at different time periods.

[0074] In this embodiment, by combining multi-dimensional features of remote sensing image data to predict the spatial distribution of planting patterns, the accuracy of judging the spatial distribution of planting patterns can be improved. This provides a comprehensive and more reliable data foundation for estimating the impact of planting patterns on soil properties, and helps to solve the problem of insufficient spatiotemporal matching of the impact of farmland fertilization behavior with large spatial scale and long time series, thereby improving the effectiveness of the estimation.

[0075] Combination Figure 3 As shown, in some embodiments, step S20 specifically includes the following steps:

[0076] S201. Generate the corresponding farmland grid based on the area to be predicted;

[0077] S202. Correlation analysis is performed on geographic location data, soil data, and meteorological data using a preset correlation analysis algorithm, and the data is allocated to farmland grids based on the analysis results.

[0078] S203. Based on the time and space dimensions, the geographical location data, soil data and meteorological data allocated in the farmland grid are clustered and partitioned using a preset hybrid clustering model to obtain the spatiotemporal units of farmland in the area to be predicted.

[0079] Specifically, considering the imbalance and heterogeneity of the spatiotemporal distribution of multi-source data such as geography, climate, and soil, this embodiment uses soil data, climate data, and geographic location data as the basis for dividing farmland into spatiotemporal units to specifically indicate the impact of planting patterns on farmland soil property variations within these units. Gridding the area to be predicted can be achieved by dividing it into m×n grids based on its area and shape, thereby generating a farmland grid for the area to be predicted. In this embodiment, the aforementioned preset correlation analysis algorithm can be a hybrid clustering model combining Gaussian and classification methods. This hybrid clustering algorithm, based on the conditional independence assumption of maximizing mathematical expectation, uses land use type, ground elevation, annual average temperature, annual rainfall, accumulated temperature, soil type, and topsoil texture as input variables. It then determines the number of cluster partitions based on the Bayesian information criterion, which is equivalent to determining the farmland spatiotemporal unit data, ultimately identifying the farmland spatiotemporal units for the area to be predicted.

[0080] Combination Figure 4 , Figure 5 As shown, in some embodiments, step S30 specifically includes the following steps:

[0081] S301. Based on the initial data of multiple farmland spatiotemporal units in different time periods, construct a random forest model of planting patterns and a random forest model of fertilization behavior.

[0082] S302. Calculate the first influence value of the planting pattern on soil properties for the spatiotemporal unit of farmland using a planting pattern random forest model during the initial and target time periods.

[0083] S303. Calculate the second influence value of fertilization behavior on soil properties in the spatiotemporal unit of farmland using a random forest model of fertilization behavior during the initial time period.

[0084] S304. Based on the first and second influence values ​​within the initial time period, determine the quantitative relationship equation between fertilization behavior and planting pattern for each farmland spatiotemporal unit. The output of the quantitative relationship equation includes the ratio of the influence of fertilization behavior on the planting pattern.

[0085] Specifically, each farmland spatiotemporal unit (S) within the initial time period can be... j T iThe latitude and longitude, annual average temperature, annual precipitation, accumulated temperature, land use type, ground elevation, soil type, topsoil texture, and planting pattern of the soil property sample collection points were used as input variables, and soil properties were used as dependent variables to independently construct planting pattern random forest (RF) models for each spatiotemporal unit of farmland. SjTi_crop Simultaneously, the latitude and longitude, annual average temperature, annual precipitation, accumulated temperature, land use type, ground elevation, soil type, topsoil texture, and fertilization behavior of soil attribute sample collection points for each farmland spatiotemporal unit were used as input variables, with soil attributes as the dependent variable. Random forest (RF) models of fertilization behavior were then independently constructed for each corresponding farmland spatiotemporal unit. SjTi_ofs ).

[0086] More specifically, by performing model verification using a random forest model of planting patterns for each spatiotemporal unit of farmland within the corresponding initial time period, the first influence value (VI) of the corresponding planting pattern on soil properties in the spatiotemporal unit of farmland can be calculated. SjTi_crop Similarly, by validating the model using a random forest model of fertilization behavior for each farmland spatiotemporal unit, the second influence value (VI) of the corresponding fertilization behavior on soil properties in the farmland spatiotemporal unit can be calculated. SjTi_ofs Therefore, based on the second influence value VI of fertilization behavior on soil properties in each farmland spatiotemporal unit... SjTi_ofs And the first influence value VI of planting pattern on soil properties SjTi_crop A quantitative relationship equation between fertilization behavior and planting pattern can be constructed for each spatiotemporal unit of farmland. Based on this equation, the influence ratio of fertilization behavior on planting pattern (the conversion factor of fertilization behavior on planting pattern) can be obtained for each spatiotemporal unit of farmland. SjTi The specific formula is shown in equation (1) below:

[0087] (1)

[0088] In this embodiment, a random forest model of fertilization behavior and a random forest model of planting pattern are constructed for each spatiotemporal unit of farmland based on initial data. This associates fertilization behavior with planting pattern, constructs a quantitative relationship equation for regression, and infers the impact of fertilization behavior on soil properties based on the two influence ratios constructed by the random forest. This approach can isolate the interaction effects of different variables, avoid the problem of binarization of various data and coupling of spatiotemporal multidimensional variables that cannot reflect the interaction relationship between multiple variables, and provides a methodological basis for realizing cross-spatiotemporal quantitative estimation of the impact of fertilization behavior on farmland soil property variation.

[0089] Combination Figure 5 , Figure 6As shown, in some embodiments, step S40 specifically includes the following steps:

[0090] S401. Determine the planting pattern influence index of the global time series, use the planting pattern influence as the index, perform recursion based on the global time series, and output the influence ratio through a quantitative relationship equation.

[0091] S402. Based on the influence ratio output by the quantitative relationship equation corresponding to the farmland spatiotemporal unit in the initial time period, and the first influence value of the target farmland spatiotemporal unit in the target time period, calculate the initial influence prediction result of the fertilization behavior of the farmland spatiotemporal unit in the target time period on soil properties.

[0092] S403. Recursively deduce from the spatial sequence to obtain each spatiotemporal unit of farmland in the target time period, and each spatiotemporal unit of farmland in the initial time period corresponding to the target time period, and determine the spatial overlap distribution characteristics between each initial spatiotemporal unit of farmland and each target spatiotemporal unit of farmland.

[0093] S404. Based on the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit, the influence degree is weighted, and combined with the initial influence degree prediction results, the influence degree prediction results of the fertilization behavior of each target farmland spatiotemporal unit on soil properties are calculated.

[0094] Specifically, spatiotemporal recursion can be performed along both the time dimension (T) and the spatial dimension (S). When recursing along the time dimension (T), the aforementioned global time series represents a time series constructed from all time periods. Based on quantitative relationship equations, recursion is performed along the time dimension using the influence of planting patterns as an index, judging the impact of planting patterns on soil properties by substituting fertilization behavior for planting patterns. Here, planting patterns include multiple types, and the planting pattern influence index can represent the impact of all planting patterns in each spatiotemporal unit partition of farmland on the variation of soil properties in that farmland's spatiotemporal partition. For example, combining... Figure 5 As shown, a certain time period can be used as the initial time period (T). i0 ) to other time periods (T) i1 T i2 T i3 T i4 ,…,T im ) Perform recursion. If the target time period is T i1 Then the initial time period (T) i0 A certain spatiotemporal unit (S) of farmland in ) j1 The influence ratio (RVI) output by the corresponding quantitative relationship equation SjT0 Multiply by the target time period (T) i1 A specific target farmland spatiotemporal unit (S) k1 The first influence value (VI) of planting patterns in ) SjTi1_cropThus, the target time period T can be calculated. i1 The target farmland spatiotemporal unit S k1 Predicted initial impact of corresponding fertilization practices on soil properties (ofs1VI) Sk1Ti1 The specific calculation formula is shown in equation (2) below:

[0095] (2)

[0096] Here, the target time period can represent the next time period after the initial time period. Similarly, based on the recursion along the time dimension, the initial impact prediction result corresponding to each farmland spatiotemporal unit in each time period can be calculated according to the above method.

[0097] More specifically, in the spatial dimension (S), based on the temporal dimension conversion, the predicted impact of fertilization behavior within the target farmland spatiotemporal unit can be spatially extrapolated, starting from the initial farmland spatiotemporal unit (S) where the initial time period is located. j1 S j2 S j3 S jz (z = 1, 2, 3, ..., Z, where Z is the total number of farmland spatiotemporal units) recursively applied to the target farmland spatiotemporal unit (S) k1 S k2 S k3 S kn When n=1, 2, 3, ..., N, where N is the total number of farmland spatiotemporal units, the target farmland spatiotemporal unit S is used. k1 For the purpose of explanation, based on the target farmland spatiotemporal unit S k1 The spatial overlap distribution characteristics between the initial farmland spatiotemporal units are weighted by influence degree, and combined with the target farmland spatiotemporal unit S k1 The initial impact of fertilization behavior on soil properties can be predicted, and the spatiotemporal unit S of the target farmland can be calculated. k1 The predicted results of the influence of fertilization behavior on soil properties (ofs2VI) Sk1Ti1 The calculation formula is shown in equation (3) below:

[0098] (3)

[0099] Among them, As k1 =As j1 +As j1 +As j1 +, … ,+As jz , where As jz S represents the spatiotemporal unit of farmland in the initial period. jz T i0 and target farmland spatiotemporal unit Sk1 T i1 The spatial overlapping distribution characteristics, that is, the overlapping area, As k1 S represents the spatiotemporal unit of the target farmland. k1 T i1 The total area.

[0100] Similarly, based on the above formula (2), the spatiotemporal units S of all target farmland in the spatial dimension for different target time periods can be calculated. kn T im The influence of fertilization behavior on soil properties is predicted, where n=1, 2, 3, ..., N, m=1, 2, 3, ..., M. Finally, by combining the above formula (3) with the initial influence prediction results, the influence prediction results of each target farmland spatiotemporal unit can be obtained. Finally, the global prediction result of the target area can be calculated based on the influence prediction results of all farmland spatiotemporal units in the area to be predicted. Among them, the weighted average can be calculated based on the influence prediction results of all farmland spatiotemporal units in the target period and the area weight of each farmland spatiotemporal unit in the target period. The calculation result is used as the global prediction result of the target. For example, if the influence prediction results of target farmland spatiotemporal units 1-10 in a certain target period are d1, d2, ..., d10 respectively, and the area distribution of target farmland spatiotemporal units 1-10 in the total area S of 10 target farmland spatiotemporal units are w1, w2, ..., w10 respectively, then the area weighted average is used as the global prediction result of the target period. Among them, the area weighted average ( The formula is shown in equation (4) below:

[0101] (4)

[0102] Optionally, a reliability value for the prediction result can be estimated for each spatiotemporal unit in each time period, and a target reference value can be preset. The target reference value and the estimated reliability value of the prediction result are stored in each spatiotemporal unit of farmland according to different time periods. The target reference value can be used to judge the reliability of the prediction result (impact on soil properties) of the target spatiotemporal unit. There are corresponding results for values ​​greater than or less than the target reference value. For example, when the reliability estimate of the prediction result of the target spatiotemporal unit exceeds a certain target reference value, it indicates that the impact of fertilization on soil properties is more credible.

[0103] In this embodiment of the invention, by gridding the area to be predicted, geographic location data, soil data, meteorological data, and farmland grid are matched, and clustering and partitioning are performed across multiple dimensions. Based on a comprehensive consideration of multiple influencing factors, the spatiotemporal unit partitioning of the impact of fertilization behavior on soil properties at different time periods is reasonably determined. Furthermore, fertilization behavior is correlated with planting patterns, and a quantitative relationship equation is constructed for regression analysis to achieve applicability substitution of fertilization behavior. The impact of fertilization behavior on soil properties is inferred based on the two impact ratios constructed using random forests. This method can isolate the interaction effects of different variables, avoiding the problems of binarization of various data and coupling of spatiotemporal multidimensional variables that prevent the reflection of the interaction relationships between multiple variables. This provides a methodological basis for effectively realizing cross-spatiotemporal quantitative estimation of the impact of fertilization behavior on farmland soil property variation.

[0104] Combination Figure 7 As shown, corresponding to the above Figure 1 The method shown is a spatiotemporal estimation method for the impact of fertilization behavior on soil properties. Figure 7 The diagram shown is a schematic diagram of a spatiotemporal estimation device for the influence of fertilization behavior on soil properties provided in an embodiment of this application. The device M70 includes:

[0105] The data acquisition module M701 is used to acquire initial data for the area to be predicted. The initial data includes fertilization behavior data, geographic location data, soil data, meteorological data, planting patterns, and soil properties.

[0106] The modeling module M702 is used to initialize the farmland grid of the area to be predicted, and to perform data matching and multi-dimensional clustering and partitioning of the farmland grid with geographic location data, soil data, and meteorological data to obtain the spatiotemporal units of farmland in the area to be predicted.

[0107] The relationship construction module M703 is used to construct quantitative relationship equations on the impact of planting pattern substitution fertilization behavior on soil properties in each spatiotemporal unit of farmland during the initial period.

[0108] The influence determination module M704 is used to perform spatiotemporal recursion based on quantitative relationship equations to determine the predicted influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted.

[0109] The target prediction module M705 is used to determine the global target prediction result of the area to be predicted based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland at different time periods.

[0110] Optionally, the initial data also includes remote sensing image data, and the data acquisition module M701 includes:

[0111] The dataset construction subunit M7011 is used to acquire remote sensing image data of the area to be predicted, extract multiple data features from the remote sensing image data, and perform data fusion based on multiple data features to construct the training dataset.

[0112] The spatial distribution determination subunit M7012 is used to train a spatial distribution prediction model based on a training dataset to obtain a target spatial distribution prediction model. The target spatial distribution prediction model is then used to determine the spatial distribution of planting patterns in the area to be predicted at different time periods.

[0113] Optionally, the modeling module M702 includes:

[0114] The generation sub-unit M7021 is used to generate the corresponding farmland grid based on the area to be predicted;

[0115] The analysis subunit M7022 is used to perform correlation analysis on geographic location data, soil data and meteorological data using a preset correlation analysis algorithm, and allocate the results to farmland grids.

[0116] The clustering subunit M7023 is used to cluster and partition the geographical location data, soil data and meteorological data allocated in the farmland grid based on the time and space dimensions through a preset hybrid clustering model, so as to obtain the farmland spatiotemporal units of the area to be predicted.

[0117] Optionally, the relationship building module M703 includes:

[0118] The model building subunit M7031 is used to build a random forest model of planting patterns and a random forest model of fertilization behavior based on the initial data of multiple farmland spatiotemporal units in different time periods.

[0119] The first calculation subunit M7032 is used to calculate the first influence value of the planting pattern on soil properties of the spatiotemporal unit of farmland using the planting pattern random forest model during the initial and target time periods.

[0120] The second calculation subunit M7033 is used to calculate the second influence value of fertilization behavior on soil properties corresponding to the spatiotemporal unit of farmland through the random forest model of fertilization behavior in the initial time period.

[0121] The quantitative relationship determination subunit M70314 is used to determine the quantitative relationship equation between fertilization behavior and planting pattern for each farmland spatiotemporal unit based on the first influence value and the second influence value within the initial time period. The output of the quantitative relationship equation includes the influence ratio of fertilization behavior on planting pattern.

[0122] Optionally, the model building subunit M7031 includes:

[0123] The first construction subunit M70311 is used to construct a random forest model of planting patterns by taking the geographical location data, soil data, meteorological data and planting patterns of each farmland spatiotemporal unit in the initial time period and the target time period as input variables and soil properties as dependent variables.

[0124] The second sub-unit, M70312, is used to construct a random forest model of fertilization behavior by taking the geographical location data, soil data, meteorological data, and fertilization behavior data of each farmland spatiotemporal unit in the initial time period as input variables and soil attributes as dependent variables.

[0125] Optionally, the impact determination module M704 includes:

[0126] The index determines the sub-unit M7041, determines the planting pattern influence index of the global time series, uses the planting pattern influence as the index, recursively extrapolates based on the global time series, and outputs the influence ratio through a quantitative relationship equation;

[0127] The third calculation subunit M7042 is used to calculate the initial influence prediction result of fertilization behavior on soil properties of the target farmland spatiotemporal unit based on the influence ratio output by the quantitative relationship equation corresponding to the farmland spatiotemporal unit in the initial time period and the first influence value of the target farmland spatiotemporal unit in the target time period.

[0128] The subunit M7043 is obtained to recursively obtain each target farmland spatiotemporal unit in the target time period and each initial farmland spatiotemporal unit in the corresponding initial time period from the spatial sequence, and to determine the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit.

[0129] The fourth calculation subunit M7044 is used to calculate the influence degree prediction results of fertilization behavior on soil properties of each target farmland spatiotemporal unit based on the spatial overlap distribution characteristics of the initial farmland spatiotemporal unit and the target farmland spatiotemporal unit, combined with the initial influence degree prediction results.

[0130] It is understood that the spatiotemporal estimation device for the influence of fertilization behavior on soil properties provided in this embodiment of the invention can realize all the processes implemented by the spatiotemporal estimation method for the influence of fertilization behavior on soil properties in the above-described method embodiments. To avoid repetition, these processes will not be repeated here. Furthermore, it can achieve the same beneficial effects.

[0131] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0132] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0133] Combination Figure 8 As shown, Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 8 As shown, the electronic device D80 of this embodiment includes: at least one processor D800 ( Figure 8 Only one is shown in the diagram. The device includes a memory D801, a network interface D802, and a computer program stored in the memory D801 that can run on at least one processor D800. The processor D800 is used to call the computer program stored in the memory D801 and perform the following steps:

[0134] Acquire initial data for the area to be predicted, including fertilization behavior data, geographic location data, soil data, meteorological data, planting patterns, and soil properties;

[0135] Initialize the farmland grid of the area to be predicted, and perform data matching and multi-dimensional clustering and partitioning of the geographical location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted.

[0136] A quantitative equation was constructed to establish the relationship between the planting pattern substitution fertilization behavior and soil properties corresponding to each spatiotemporal unit of farmland in the initial period.

[0137] Based on the quantitative relationship equation, spatiotemporal recursion is performed to determine the predicted impact of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted.

[0138] Based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland under different target time periods, the target global prediction result of the area to be predicted is determined.

[0139] Optionally, the initial data also includes remote sensing image data. The initial data acquisition process performed by the processor D800 for the area to be predicted also includes:

[0140] Acquire remote sensing image data of the area to be predicted, extract multiple data features from the remote sensing image data, perform data fusion based on multiple data features, and construct a training dataset;

[0141] A spatial distribution prediction model is trained based on the training dataset to obtain a target spatial distribution prediction model. The spatial distribution of planting patterns in the area to be predicted at different time periods is then determined using the target spatial distribution prediction model.

[0142] Optionally, the processor D800 initializes the farmland grid of the area to be predicted, and performs data matching and multi-dimensional clustering partitioning with the geographic location data, soil data, and meteorological data to obtain the spatiotemporal units of farmland in the area to be predicted, including:

[0143] Generate a corresponding farmland grid based on the area to be predicted;

[0144] The correlation analysis algorithm is used to perform correlation analysis on geographic location data, soil data and meteorological data. The planting pattern random forest model calculates the first influence value of the planting pattern on soil properties corresponding to the spatiotemporal unit of farmland.

[0145] During the initial time period, the second influence value of fertilization behavior on soil properties corresponding to the spatiotemporal unit of farmland is calculated using the random forest model of fertilization behavior.

[0146] Based on the first influence value and the second influence value within the initial time period, a quantitative relationship equation between fertilization behavior and planting pattern is determined for each of the farmland spatiotemporal units. The output of the quantitative relationship equation includes the influence ratio of fertilization behavior on planting pattern.

[0147] Optionally, the processor D800 executes the following steps: constructing a random forest model for planting patterns and a random forest model for fertilization behavior based on initial data from multiple spatiotemporal units of farmland within different time periods, including:

[0148] Using the geographic location data, soil data, meteorological data, and planting patterns of each spatiotemporal unit of farmland within the initial and target time periods as input variables, and the soil attributes as the dependent variable, a random forest model of the planting patterns is constructed.

[0149] Using the geographic location data, soil data, meteorological data, and fertilization behavior data of each spatiotemporal unit of farmland within the initial time period as input variables and the soil attributes as the dependent variable, a random forest model of fertilization behavior is constructed.

[0150] Optionally, the processor D800 executes the step of performing spatiotemporal recursion based on the quantitative relationship equation to determine the predicted impact of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods, including:

[0151] Determine the planting pattern influence index of the global time series, and use the planting pattern influence index to recursively extrapolate based on the global time series, outputting the influence ratio through a quantitative relationship equation;

[0152] Based on the influence ratio output by the quantitative relationship equation corresponding to the spatiotemporal unit of farmland in the initial period, and the first influence value of the spatiotemporal unit of the target farmland in the target period, the initial influence prediction result of the fertilization behavior of the target farmland spatiotemporal unit on soil properties is calculated;

[0153] By recursively extrapolating from the spatial sequence, we obtain each target farmland spatiotemporal unit in the target time period and each initial farmland spatiotemporal unit in the initial time period corresponding to the target time period, and determine the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit.

[0154] The influence degree is weighted according to the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit. Combined with the initial influence degree prediction results, the influence degree prediction results of fertilization behavior on soil properties of each target farmland spatiotemporal unit are calculated.

[0155] The electronic device D80 provided in this embodiment of the invention can realize various implementation methods in the method embodiment for estimating the influence of fertilization behavior on soil properties across time and space, as well as the corresponding beneficial effects. To avoid repetition, these will not be described again here.

[0156] It should be noted that only components D800-D802 are shown in the figure; however, it should be understood that implementation of all shown components is not required, and more or fewer components may be implemented alternatively. Those skilled in the art will understand that the electronic device described herein is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions.

[0157] In some embodiments, the processor D800 may be a Central Processing Unit (CPU). However, the processor D800 may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0158] In some embodiments, memory D801 can be an internal storage unit of electronic device D80, such as a hard disk or memory of electronic device D80. In other embodiments, memory D801 can also be an external storage device of electronic device D80, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on electronic device D80. Furthermore, memory D801 can include both internal and external storage units of electronic device D80. Memory D801 is used to store operating system, application programs, boot loader, data, and other programs, such as program code of computer programs. Memory D801 can also be used to temporarily store data that has been output or will be output.

[0159] The network interface D802 may include a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the electronic device D80 and other electronic devices.

[0160] This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by the processor D800, it implements the various processes of the cross-temporal estimation method for the influence of fertilization behavior on soil properties provided in this invention, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0161] This application provides a computer program product that, when run on an electronic device, enables the electronic device to implement the steps described in the various method embodiments above.

[0162] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0163] The terms "first," "second," etc., used in the specification, claims, or accompanying drawings of this application are used to distinguish different objects and not to describe a specific order. The reference to "embodiment" herein means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0164] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for estimating the spatiotemporal impact of fertilization behavior on soil properties, characterized in that, The method includes the following steps: Acquire initial data for the area to be predicted, including fertilization behavior data, geographic location data, soil data, meteorological data, planting patterns, and soil properties; Initialize the farmland grid of the area to be predicted, and perform data matching and multi-dimensional clustering and partitioning of the geographical location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted. A quantitative equation was constructed to establish the relationship between the planting pattern substitution fertilization behavior and soil properties corresponding to each spatiotemporal unit of farmland in the initial period. Based on the quantitative relationship equation, spatiotemporal recursion is performed to determine the predicted impact of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted. Based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland under different target time periods, the target global prediction result of the area to be predicted is determined. The step of performing spatiotemporal recursion based on the quantitative relationship equation to determine the predicted impact of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted includes: Determine the index of the impact of planting patterns in the global time series, and use the impact of planting patterns as the index to recursively extrapolate based on the global time series, and substitute the impact of fertilization behavior according to the quantitative relationship equation; Based on the influence ratio output by the quantitative relationship equation corresponding to the spatiotemporal unit of farmland in the initial period, and the first influence value of the spatiotemporal unit of the target farmland in the target period, the initial influence prediction result of the fertilization behavior of the target farmland spatiotemporal unit on soil properties is calculated; By recursively extrapolating from the spatial sequence, we obtain each target farmland spatiotemporal unit in the target time period and each initial farmland spatiotemporal unit in the initial time period corresponding to the target time period, and determine the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit. The influence degree is weighted according to the spatial overlap distribution characteristics between each initial farmland spatiotemporal unit and each target farmland spatiotemporal unit. Combined with the initial influence degree prediction results, the influence degree prediction results of fertilization behavior on soil properties of each target farmland spatiotemporal unit are calculated.

2. The method according to claim 1, characterized in that, The initial data also includes remote sensing image data, and the process of obtaining the initial data for the area to be predicted further includes: Acquire the remote sensing image data of the region to be predicted, extract multiple data features from the remote sensing image data, perform data fusion based on the multiple data features, and construct a training dataset; Based on the training dataset, a spatial distribution prediction model is trained to obtain a target spatial distribution prediction model. The spatial distribution of planting patterns in the area to be predicted is determined through the target spatial distribution prediction model at different time periods.

3. The method according to claim 1, characterized in that, The process involves initializing a farmland grid for the area to be predicted, and then matching the geographic location data, soil data, and meteorological data with the farmland grid and performing multi-dimensional clustering and partitioning to obtain spatiotemporal units of farmland for the area to be predicted. This includes: Generate a corresponding farmland grid based on the area to be predicted; The geographic location data, soil data, and meteorological data are analyzed using a preset correlation analysis algorithm, and the results are then distributed to the farmland grid. Based on the time and space dimensions, the geographical location data, soil data and meteorological data allocated in the farmland grid are clustered and partitioned using a preset hybrid clustering model to obtain the spatiotemporal units of farmland in the area to be predicted.

4. The method according to claim 1, characterized in that, The quantitative relationship equations for the impact of planting patterns replacing fertilization behavior on soil properties corresponding to each of the farmland spatiotemporal units in the initial period are as follows: Based on the initial data of multiple spatiotemporal units of farmland in different time periods, a random forest model of planting patterns and a random forest model of fertilization behavior are constructed. During the initial and target time periods, the first influence value of the planting pattern on soil properties corresponding to the spatiotemporal unit of farmland is calculated using the planting pattern random forest model. During the initial time period, the second influence value of fertilization behavior on soil properties corresponding to the spatiotemporal unit of farmland is calculated using the random forest model of fertilization behavior. Based on the first influence value and the second influence value within the initial time period, a quantitative relationship equation between fertilization behavior and planting pattern is determined for each of the farmland spatiotemporal units. The output of the quantitative relationship equation includes the influence ratio of fertilization behavior on planting pattern.

5. The method according to claim 4, characterized in that, The step of constructing a random forest model for planting patterns and a random forest model for fertilization behavior based on initial data from multiple spatiotemporal units of farmland within different time periods includes: Using the geographic location data, soil data, meteorological data, and planting patterns of each spatiotemporal unit of farmland within the initial and target time periods as input variables, and the soil attributes as the dependent variable, a random forest model of the planting patterns is constructed. Using the geographic location data, soil data, meteorological data, and fertilization behavior data of each spatiotemporal unit of farmland within the initial time period as input variables and the soil attributes as the dependent variable, a random forest model of fertilization behavior is constructed.

6. A device for estimating the spatiotemporal influence of fertilization behavior on soil properties, characterized in that, The device includes: The data acquisition module is used to acquire initial data of the area to be predicted, including fertilization behavior data, geographical location data, soil data, meteorological data, planting patterns and soil properties; The modeling module is used to initialize the farmland grid of the area to be predicted, and to perform data matching and multi-dimensional clustering and partitioning of the geographical location data, soil data, and meteorological data with the farmland grid to obtain the spatiotemporal units of the farmland in the area to be predicted. The relationship construction module is used to construct quantitative relationship equations on the impact of planting pattern substitution fertilization behavior on soil properties for each of the farmland spatiotemporal units in the initial period. The influence determination module is used to perform spatiotemporal recursion based on the quantitative relationship equation to determine the predicted influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland in different time periods of the area to be predicted. The target prediction module is used to determine the global target prediction result of the area to be predicted based on the prediction results of the influence of fertilization behavior on soil properties in each spatiotemporal unit of farmland under different target time periods.

7. The apparatus according to claim 6, characterized in that, The initial data also includes remote sensing image data, and the data acquisition module includes: The dataset construction subunit is used to acquire the remote sensing image data of the region to be predicted, extract multiple data features from the remote sensing image data, perform data fusion based on the multiple data features, and construct a training dataset. The spatial distribution determination subunit is used to train a spatial distribution prediction model based on the training dataset to obtain a target spatial distribution prediction model, and to determine the spatial distribution of planting patterns in the area to be predicted at different time periods through the target spatial distribution prediction model.

8. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps in the method for estimating the spatiotemporal impact of fertilization behavior on soil properties as described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps in the method for estimating the impact of fertilization behavior on soil properties across time and space as described in any one of claims 1 to 5.

Images