Precise monitoring method for water and salt dynamics of coastal saline-alkali soil based on HYDRUS model

Abstract

This application provides a method for precise monitoring of water and salt dynamics in coastal saline-alkali land based on the HYDRUS model, belonging to the field of coastal saline-alkali land monitoring technology. This application first determines the soil salinity profile type and characteristic ions in different regions; then, it sets the soil profile stratification and initial water and salt transport parameters of the HYDRUS model in conjunction with surface vegetation information. Meteorological and groundwater data are used as boundary conditions to run the model, simulating the conductivity and characteristic ion concentration at different depths of each point over time. This data is input into a two-dimensional convolutional neural network to extract spatiotemporal state features; an adjacency graph is constructed based on geographic location, and adjacent node information is fused through a graph neural network to optimize the spatiotemporal features into global state features reflecting spatial interaction. Finally, by combining the correlation between characteristic ions and total salt, salinization prediction information is generated, achieving precise monitoring. This method integrates water and salt transport simulation with spatiotemporal deep learning to achieve high-precision prediction of water and salt dynamics in coastal saline-alkali land.

Description

Technical Field

[0001] This application relates to the field of coastal saline-alkali land monitoring technology, and in particular to a method for precise dynamic monitoring of water and salt in coastal saline-alkali land based on the HYDRUS model. Background Technology

[0002] The HYDRUS model-based method for precise monitoring of water and salt dynamics in coastal saline-alkali land is a technical means to simulate and predict the spatiotemporal variation of soil salinity by constructing a numerical model of soil water and salt transport. Coastal saline-alkali land is an important reserve of arable land in my country, and precise monitoring of its water and salt dynamics has significant application prospects for guiding the improvement and utilization of saline-alkali land and ensuring regional food security.

[0003] In existing technologies, studies have used the HYDRUS-1D model to simulate water and salt transport processes in coastal saline-alkali land. Basic data were obtained through field surveys, sample collection, and sensor installation. Model parameters and boundary conditions were set to simulate the vertical transport process of soil water and salt under rainfall and evaporation conditions, and the model's simulation effect on water and salt during the secondary rainfall and long-term evaporation stages was verified.

[0004] However, the aforementioned existing technologies primarily focus on one-dimensional vertical water and salt transport simulations at a single point, making it difficult to fully integrate spatial distribution information from different sampling points. This results in limitations in characterizing the spatiotemporal evolution of salinity at the regional scale. Furthermore, traditional methods do not adequately explore the coupling characteristics of soil salinity's temporal evolution and vertical profile distribution, affecting the overall accuracy of salinization trend prediction. Therefore, existing technologies face the technical challenge of achieving high-precision prediction of water and salt dynamics in coastal saline-alkali lands. Summary of the Invention

[0005] The purpose of this application is to provide a method for accurate monitoring of water and salt dynamics in coastal saline-alkali land based on the HYDRUS model, so as to solve the problem that it is difficult to achieve high-precision prediction of water and salt dynamics in coastal saline-alkali land in the existing technology.

[0006] To address the aforementioned technical problems, firstly, this application provides a method for precise dynamic monitoring of water and salinity in coastal saline-alkali land based on the HYDRUS model, comprising: To determine the type of saline-alkali profile and the characteristic ions corresponding to different salinity levels in soil samples within the coastal saline-alkali land; Based on the surface vegetation information, the saline-alkali profile type, and the characteristic ions, the soil profile stratification and the initial water and salt transport parameters of each stratum at each sampling point in the HYDRUS model are set. The collected meteorological data and groundwater level data of the coastal saline-alkali land are used as the boundary conditions of the HYDRUS model. Based on the soil profile stratification, the initial water and salt transport parameters and the boundary conditions, the HYDRUS model is run to simulate and output the conductivity data and characteristic ion concentration data of the soil profile at different depths over time at each sampling point. The conductivity data and characteristic ion concentration data are input into a two-dimensional convolutional neural network to extract spatiotemporal state features that can simultaneously characterize the coupling mode of salt in time evolution and vertical profile distribution. A spatial adjacency graph is constructed based on the geographic location of each sampling point. The feature information of adjacent nodes in the spatial adjacency graph is fused through a graph neural network, and the spatiotemporal state features of each node are iteratively optimized into global state features that can reflect spatial interactions. Based on the global state characteristics and the correlation between the characteristic ions and the total salt content, predictive information on soil salinization in the coastal saline-alkali land is generated.

[0007] Optionally, the step of constructing a spatial adjacency graph based on the geographic location of each sampling point, fusing the feature information of adjacent nodes in the spatial adjacency graph through a graph neural network, and iteratively optimizing the spatiotemporal state features of each node into global state features that can reflect spatial interactions includes: Based on the geographical coordinates of each sampling point, the spatial distance between each sampling point is calculated, and the spatial adjacency relationship between each sampling point is determined based on the spatial distance. A spatial adjacency graph is constructed using each sampling point as a graph node and the spatial adjacency relationship as the edge between nodes; wherein, the initial feature of each node is the spatiotemporal state feature of the corresponding sampling point. The spatial adjacency graph is input into a graph neural network. Through the neighborhood aggregation operation of the graph neural network, each node receives and aggregates the feature information of its neighboring nodes along the edge. Through multiple iterations of the graph neural network, the features of each node are gradually aggregated with spatial context information from multiple-order neighbors to generate global state features corresponding to each node that can comprehensively represent spatial interactions.

[0008] Optionally, the step of inputting the conductivity data and characteristic ion concentration data into a two-dimensional convolutional neural network to extract spatiotemporal state features that can simultaneously characterize the coupling mode of salt in temporal evolution and vertical profile distribution includes: The conductivity data and characteristic ion concentration data corresponding to each sampling point are organized into two-dimensional input data according to the two dimensions of soil vertical profile depth and time. The two-dimensional input data is input into a two-dimensional convolutional neural network. Through the convolution operation of the two-dimensional convolutional neural network, the local coupling features of each sampling point in the depth direction and time direction of the soil vertical profile are extracted. By performing multi-layer convolution and pooling operations on the local coupling features, spatiotemporal state features corresponding to each sampling point are generated that can comprehensively characterize the spatiotemporal evolution law of salinity.

[0009] Optionally, the step of using the collected meteorological data and groundwater level data of the coastal saline-alkali land as boundary conditions of the HYDRUS model, and running the HYDRUS model according to the soil profile stratification, the initial water and salt transport parameters, and the boundary conditions, to simulate and output the conductivity data and characteristic ion concentration data of the soil profile at each sampling point at different depths over time, including: The collected meteorological data is organized into daily meteorological sequences, which include precipitation data and evaporation data. The daily meteorological sequence is set as the surface boundary condition at the top of the soil profile, and the groundwater level data is set as the bottom boundary condition at the bottom of the soil profile. The soil profile is divided into layers, the initial water and salt transport parameters corresponding to each layer, the surface boundary conditions, and the bottom boundary conditions are input into the HYDRUS model. The HYDRUS model performs daily simulations according to a set time step to calculate the conductivity and characteristic ion concentration values ​​of each layer at different times. Record the conductivity and characteristic ion concentration values ​​of each layer at each time point to generate conductivity and characteristic ion concentration data of each layer of the soil profile at each sampling point over time.

[0010] Optionally, the step of generating predictive information on soil salinization in coastal saline-alkali land based on the global state features and the correlation between the feature ions and total salt content includes: Obtain the conversion relationship between each characteristic ion and the total salt content; Based on the characteristic ion concentration data of each sampling point at different depths and times contained in the global state features, the equivalent salt contribution value corresponding to each characteristic ion is calculated using the conversion relationship, and multiple equivalent salt contribution values ​​at the same spatiotemporal location are accumulated to obtain the simulated total salt content. The conductivity data contained in the global state features are matched with a preset conductivity salinity level comparison table to generate a first salinization level sequence, and the simulated total salt content is matched with a preset total salt salinity level comparison table to generate a second salinization level sequence. Based on the first salinization level sequence and the second salinization level sequence, the comprehensive salinization level of each sampling point at different depths and at different times is determined, and a comprehensive salinization level sequence is generated. Based on the concentration ratio of different characteristic ions contained in the global state features, the salinization type of each sampling point at different depths and at different times is determined, and the evolution trend of salinization type is obtained. The comprehensive salinization level sequence and the evolution trend of salinization type are used together as predictive information for soil salinization in coastal saline-alkali land.

[0011] Optionally, determining the salinity profile type of soil samples within coastal saline-alkali land and the characteristic ions corresponding to different salinity levels includes: In the coastal saline-alkali land, soil samples at different depths were obtained through stratified sampling, and the corresponding sampling point locations and surface vegetation information were recorded. The total salt content, electrical conductivity, and composition of water-soluble salt ions in the soil sample were determined. Based on the vertical variation characteristics of the total salt content in the soil profile, the soil profiles at the sampling points are divided into corresponding saline-alkali profile types. Based on the composition information of the water-soluble salt ions, characteristic ions in different salinity areas within the coastal saline-alkali land are determined.

[0012] Optionally, determining the characteristic ions of different salinity levels within the coastal saline-alkali land based on the composition information of the water-soluble salt ions includes: Based on the total salt content of soil samples at different depths at each sampling point, all sampling points were divided into a slightly salinized group and a moderately to severely salinized group. Based on the composition information of the water-soluble salt ions, the correlation coefficient between the ion content of each water-soluble salt ion at different depths and the total salt content is calculated for the mildly salinized group and the moderately severely salinized group, respectively. For the moderately to severely salinized group, the first anion with the highest correlation coefficient is selected as the characteristic ion of the moderately to severely salinized region across all depths. For the mildly salinized group, the soil profile is divided into a shallow zone and a deep zone according to the depth: in the shallow zone, the second anion with the largest correlation coefficient is selected as the shallow characteristic ion of the mildly salinized area; in the deep zone, the third anion with the largest correlation coefficient is selected as the deep characteristic ion of the mildly salinized area. The first anion, the second anion, and the third anion are collectively identified as characteristic ions in different salinity regions within the coastal saline-alkali land.

[0013] Optionally, the saline-alkali profile types include surface-agglomeration type, bottom-agglomeration type, and oscillating type; The process of classifying the soil profiles at the sampling points into corresponding saline-alkali profile types based on the vertical variation characteristics of the total salt content in the soil profile includes: Based on the total salt content of soil samples at different depths of each sampling point, a sequence of salt content depth variation for each sampling point is formed. Based on the salinity depth change sequence, the salinity vertical feature value of each sampling point is calculated, and the salinity vertical feature value includes shallow cumulative value, bottom cumulative value and fluctuation intensity value; When the salinity depth change sequence of the sampling point shows a continuous decreasing trend from the surface to the depth, and the shallow cumulative value is higher than the bottom cumulative value, the soil profile of the sampling point is determined to be surface-aggregate type. When the salinity depth change sequence of the sampling point shows a continuous increasing trend from the surface to the depth, and the cumulative value of the bottom layer is higher than the cumulative value of the shallow layer, the soil profile of the sampling point is determined to be bottom-aggregate type. When the salinity depth change sequence of the sampling point shows an alternating trend of increasing and decreasing values ​​along the depth direction, and the fluctuation intensity value is higher than a preset threshold, the soil profile of the sampling point is determined to be oscillating.

[0014] Optionally, the step of setting soil profile stratification and initial water and salt transport parameters for each sampling point in the HYDRUS model based on surface vegetation information, the saline-alkali profile type, and the characteristic ions includes: The vegetation action coefficient corresponding to each sampling point is determined based on the surface vegetation information, and an initial soil profile stratification is generated based on the vegetation action coefficient and the saline-alkali profile type. Based on the content distribution of the characteristic ions at different depths, the layer boundary depths of the initial soil profile are adjusted to form the final soil profile stratification. Obtain soil physical property data corresponding to each layer in the soil profile, including soil density, particle composition and water holding capacity; Based on the soil physical property data, the water movement parameters corresponding to each layer are determined, including water retention capacity parameters and water conduction capacity parameters; at the same time, based on the type of characteristic ions, the solute movement parameters corresponding to each layer are determined. The water movement parameters and solute movement parameters corresponding to each layer are used as the initial water-salt transport parameters.

[0015] Optionally, determining the vegetation action coefficient corresponding to each sampling point based on the surface vegetation information includes: The surface vegetation type at each sampling point is identified based on surface vegetation information, and the surface vegetation type includes trees, shrubs, herbs and bare ground; Obtain the preset range of action coefficients corresponding to each of the aforementioned surface vegetation types; Based on the surface vegetation cover at the sampling point, the vegetation action coefficient corresponding to the sampling point is determined within the preset action coefficient range; the vegetation action coefficient is used to characterize the comprehensive influence of vegetation on soil moisture absorption capacity and salt accumulation inhibition capacity.

[0016] The method for precise monitoring of water and salt dynamics in coastal saline-alkali land based on the HYDRUS model provided in this application provides a basis for setting model parameters by determining the type of saline-alkali profile and characteristic ions; it sets soil stratification and initial parameters by combining surface vegetation information to make the model more closely match actual conditions; it runs the HYDRUS model with meteorological and groundwater data as boundaries to obtain time-series data of conductivity and ion concentration at different depths of each sampling point; it inputs the data into a two-dimensional convolutional neural network to extract the spatiotemporal features coupled with salinity temporal evolution and vertical distribution; it constructs a spatial adjacency graph based on geographical location, and optimizes the spatiotemporal features into global spatial interaction features by fusing adjacent node information through a graph neural network; and it generates prediction information by combining the correlation between characteristic ions and total salt to achieve high-precision prediction of salinization status.

[0017] Furthermore, spatial distances are calculated and adjacency relationships are determined based on the geographical coordinates of the sampling points. A spatial adjacency graph is constructed with each sampling point as a node and adjacency relationships as edges. The initial features of the nodes are spatiotemporal state features. The graph is then input into a graph neural network, where neighborhood aggregation operations allow nodes to receive and aggregate features from neighboring nodes along the edges. After multiple iterations, a global state feature that integrates multi-level neighbor spatial context information is generated. This step effectively integrates the salinity features of adjacent sampling points through the spatial information transmission of the graph neural network, enhancing the model's ability to model regional spatial correlations. This allows the generated global state feature to more comprehensively reflect spatial interactions, thereby improving the overall accuracy of regional salinization prediction. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, 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.

[0019] Figure 1A flowchart illustrating a method for precise dynamic monitoring of water and salt in coastal saline-alkali land based on a HYDRUS model, provided for an embodiment of this application; Figure 2 A flowchart illustrating another method for precise monitoring of water and salt dynamics in coastal saline-alkali land based on the HYDRUS model, provided for an embodiment of this application; Figure 3 This is a schematic diagram of a dynamic and precise monitoring system for water and salt in coastal saline-alkali land based on the HYDRUS model, provided as an embodiment of this application. Detailed Implementation

[0020] Current methods for monitoring water and salt content in coastal saline-alkali lands primarily rely on the HYDRUS-1D model for single-point vertical simulations. Data is acquired through sampling and sensors to simulate water and salt transport processes under rainfall and evaporation conditions. However, this method struggles to integrate spatial distribution information from different sampling points and fails to adequately explore the coupling characteristics of salinity over time and vertical distribution, thus limiting the accuracy of regional-scale predictions.

[0021] To address the aforementioned issues, this application proposes a method for precise dynamic monitoring of water and salinity in coastal saline-alkali land based on the HYDRUS model. This method first simulates and outputs time-series data of conductivity and ion concentration at different depths of each sampling point using the HYDRUS model. Then, it extracts spatiotemporal coupling features using a two-dimensional convolutional neural network and fuses information from neighboring points based on geographic coordinates using a graph neural network to generate global state features that include spatial interactions, ultimately achieving precise prediction of salinization. This method effectively integrates temporal evolution and spatial correlation features, fundamentally solving the problem of insufficient spatiotemporal information mining in existing technologies.

[0022] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] The core of this application is to provide a method for precise dynamic monitoring of water and salt in coastal saline-alkali land based on the HYDRUS model. A flowchart of one specific implementation method is shown below. Figure 1 As shown, the method includes: S101. Determine the type of saline-alkali profile corresponding to soil samples in coastal saline-alkali land and the characteristic ions corresponding to different salinity levels.

[0024] S101 specifically includes: S1011. In coastal saline-alkali land, soil samples at different depths are obtained through stratified sampling, and the corresponding sampling point locations and surface vegetation information are recorded.

[0025] Stratified sampling refers to collecting samples from different soil layers based on their vertical depth to obtain the true condition of the soil at different depths. The sampling point location is the geographic coordinates of each sampling location. Surface vegetation information includes basic details such as the type and coverage of vegetation at the sampling point.

[0026] In this embodiment of the application, soil samples are first collected layer by layer at preset depth intervals within the target area of ​​coastal saline-alkali land to ensure that each sampling point obtains a continuous layered sample from the ground surface to the specified depth; at the same time, a positioning tool is used to record the precise location of each sampling point, and the vegetation type and coverage of the ground surface at the sampling point are observed and recorded simultaneously.

[0027] S1012. Determine the total salt content, electrical conductivity, and composition of water-soluble salt ions in soil samples.

[0028] Total salt content refers to the total amount of all soluble salts in the soil and is a core indicator for measuring soil salinity. Electrical conductivity is a parameter reflecting the ability of soil solution to conduct electricity and can indirectly characterize the total amount of soil salt. The composition information of water-soluble salt ions refers to the types and contents of various water-soluble anions and cations in the soil, specifically including chloride ions, carbonate ions, bicarbonate ions, sulfate ions, calcium ions, magnesium ions, potassium ions, and sodium ions.

[0029] In this embodiment of the application, the collected soil samples were sent to the laboratory, and the total salt content, electrical conductivity, and specific content of the above eight water-soluble salt ions at different depths of each sample were determined by professional testing instruments and chemical analysis methods.

[0030] S1013. Based on the vertical variation characteristics of total salt content in the soil profile, the soil profiles at the sampling points are divided into corresponding saline-alkali profile types.

[0031] Among them, the saline-alkali profile type is a category classified according to the distribution pattern of salt along the vertical depth of the soil, which is used to visually reflect the accumulation state of salt at different depths; it includes surface accumulation type, bottom accumulation type, and oscillating type; surface accumulation type refers to the type in which salt is mainly concentrated in the shallow soil layer and decreases with increasing depth; bottom accumulation type refers to the type in which salt is mainly concentrated in the deep soil layer and increases with increasing depth; oscillating type refers to the type in which salt alternately increases or decreases along the depth direction and has no obvious accumulation pattern.

[0032] Specifically, S1013 includes: Based on the total salt content of soil samples at different depths at each sampling point, a salt content depth variation sequence is formed for each sampling point. Based on this sequence, the vertical characteristic value of salt content at each sampling point is calculated, including the shallow cumulative value, the bottom cumulative value, and the fluctuation intensity value. When the salt content depth variation sequence at a sampling point shows a continuously decreasing trend from the surface to depth, and the shallow cumulative value is higher than the bottom cumulative value, the soil profile at that sampling point is determined to be surface-aggregated. When the salt content depth variation sequence at a sampling point shows a continuously increasing trend from the surface to depth, and the bottom cumulative value is higher than the shallow cumulative value, the soil profile at that sampling point is determined to be bottom-aggregated. When the salt content depth variation sequence at a sampling point shows an alternating trend of increasing and decreasing values ​​along the depth direction, and the fluctuation intensity value is higher than a preset threshold, the soil profile at that sampling point is determined to be oscillating.

[0033] Among them, the vertical characteristic value of salinity is a key parameter for quantifying the vertical distribution characteristics of salinity. It is calculated based on the sampling depth range of coastal saline-alkali land: the shallow cumulative value is the sum of the total salt content at each depth from the surface to the first set depth, which is used to reflect the degree of shallow salt accumulation; the bottom cumulative value is the sum of the total salt content at each depth from the second set depth to the bottom of the sampling, which is used to reflect the degree of deep salt accumulation; the fluctuation intensity value is an indicator for measuring the degree of drastic change of salinity along the depth. It comprehensively considers the alternation frequency and amplitude of salinity changes between adjacent depths. The more frequent the changes and the greater the amplitude, the higher the fluctuation intensity value.

[0034] In this embodiment, the total salt content of soil samples at different depths of each sampling point is first arranged in order from shallow to deep to form a salt content depth variation sequence for each sampling point. Then, the vertical characteristic value of salt content at each sampling point is calculated based on the salt content depth variation sequence. The vertical characteristic value of salt content includes shallow cumulative value, bottom cumulative value, and fluctuation intensity value. The shallow cumulative value is the sum of the total salt content at each depth from the surface to a first set depth, the bottom cumulative value is the sum of the total salt content at each depth from a second set depth to the bottom of the sampling point, and the fluctuation intensity value is an indicator of the intensity of salt content variation along depth. When the salt content depth variation sequence of a sampling point shows a continuous decreasing trend from the surface to depth and the shallow cumulative value is higher than the bottom cumulative value, the soil profile of that sampling point is determined to be surface-aggregated. When the salt content depth variation sequence shows a continuous increasing trend from the surface to depth and the bottom cumulative value is higher than the shallow cumulative value, it is determined to be bottom-aggregated. When the salt content depth variation sequence shows an alternating increasing and decreasing trend along the depth direction and the fluctuation intensity value is higher than a preset threshold, it is determined to be oscillating.

[0035] In practical applications, taking a sampling point in Dafeng Forest Farm as an example, the measured total salt content data at five depths—0-20 cm, 20-40 cm, 40-60 cm, 60-80 cm, and 80-100 cm—were collected, yielding values ​​of 2.19‰, 2.30‰, 1.94‰, 1.80‰, and 1.84‰, respectively. These values ​​were then arranged from shallowest to deepest to form a salt content depth variation sequence [2.19, 2.30, 1.94, 1.80, 1.84]. The first set depth was set to 40 cm, with the shallow cumulative value being 2.19 + 2.30 = 4.49‰. The second set depth was set to 60 cm, with the bottom cumulative value being 1.80 + 1.84 = 3.64‰. The fluctuation intensity value was obtained by calculating the sum of the absolute values ​​of the salt content differences between adjacent depths. Based on the soil characteristics of the coastal saline-alkali land in Dafeng Forest Farm, a preset threshold of 1.0‰ was set for the fluctuation intensity value. The salinity depth variation sequence at this sampling point showed a continuous decreasing trend from the surface to the depth, with the shallow cumulative value of 4.49‰ higher than the bottom cumulative value of 3.64‰. Simultaneously, the fluctuation intensity value of 0.65‰ was lower than the preset threshold of 1.0‰. Therefore, the soil profile at this sampling point was determined to be surface-accumulating. In practical applications, the first and second preset depths can be adjusted according to the soil characteristics of the study area, and the preset threshold can be determined through statistical analysis of sampling data from a larger area of ​​coastal saline-alkali land. This application does not impose any limitations on this.

[0036] S1014. Based on the composition information of water-soluble salt ions, determine the characteristic ions in areas with different salinity levels within the coastal saline-alkali land.

[0037] It should be noted that the salinity of coastal saline-alkali land mainly originates from seawater tides and groundwater. Soil salts exist primarily in the form of cations and anions, such as sodium chloride and sodium carbonate. However, the core factor influencing soil salinity is the type and content of anions. For example, excessive chloride ions lead to soil salinization, while carbonate and bicarbonate ions cause soil alkali damage. Changes in their content directly drive changes in total salt content and soil properties. Cations, such as calcium and magnesium ions, readily combine with soil colloids or form precipitates, exhibiting strong stability and dynamic changes far less pronounced than anions. Their correlation with total salt content is also weaker, making it difficult to accurately characterize water-salt dynamics. Therefore, this step focuses on calculating anions among the eight water-soluble salt ions in coastal saline-alkali soil. Thus, characteristic ions refer to anions that are most closely correlated with total salt content within a specific salinity region and can accurately reflect the degree of salinization in that area.

[0038] Specifically, S1014 includes: Based on the total salt content of soil samples at different depths at each sampling point, all sampling points were divided into a slightly salinized group and a moderately to severely salinized group. Based on the composition information of water-soluble salt ions, the correlation coefficient between the ion content of each water-soluble salt ion at different depths and the total salt content was calculated for both the slightly and moderately to severely salinized groups. For the moderately to severely salinized group, the first anion with the highest correlation coefficient was selected as the characteristic ion of the moderately to severely salinized area across all depths. For the slightly salinized group, the soil profile was divided into shallow and deep sections based on depth: in the shallow section, the second anion with the highest correlation coefficient was selected as the shallow characteristic ion of the slightly salinized area; in the deep section, the third anion with the highest correlation coefficient was selected as the deep characteristic ion of the slightly salinized area. The first, second, and third anions were collectively identified as the characteristic ions for different salinity levels within the coastal saline-alkali land.

[0039] The mildly salinized group comprises sampling points with low total salt content, while the moderately to severely salinized group comprises sampling points with medium to high total salt content. Grouping is based on the coastal saline-alkali soil salinization classification standard to ensure the grouping reflects the actual regional conditions. The first anion is the anion most closely associated with total salt content in the moderately to severely salinized group, and its content changes most directly reflect the changes in total salt content in the moderately to severely salinized area. The division between shallow and deep layers is based on the vertical distribution characteristics of salt in coastal saline-alkali soils. The second and third anions are the anions most closely associated with total salt content in the shallow and deep layers of the mildly salinized group, respectively, and are used to specifically characterize salinity changes at different depths.

[0040] In this embodiment, firstly, based on the total salt content of soil samples at different depths at each sampling point, and referring to the coastal salinization classification standard, all sampling points are divided into a slightly salinized group and a moderately to severely salinized group. Then, for both the slightly salinized and moderately to severely salinized groups, the correlation coefficient between the ion content of each water-soluble salt ion at different depths and the total salt content is calculated. For the moderately to severely salinized group, the first anion with the highest correlation coefficient across all depths is selected as the characteristic ion of the moderately to severely salinized area. For the slightly salinized group, the soil profile is divided into shallow and deep sections based on depth. In the shallow section, the second anion with the highest correlation coefficient is selected as the shallow characteristic ion of the slightly salinized area, and in the deep section, the third anion with the highest correlation coefficient is selected as the deep characteristic ion of the slightly salinized area. Finally, the first, second, and third anions are collectively determined as the characteristic ions of different salinity levels within the coastal salinized area.

[0041] In practical application, taking Dafeng Forest Farm as an example, the total salt content was set at 1.0 to 2.0‰ as the threshold for mild salinization and above 2.0‰ as the threshold for moderate to severe salinization, referencing the coastal salinization classification standard. Total salt content data were collected from 46 sampling points at various depths. Of these, 30 sampling points had total salt content within the range of 1.0 to 2.0‰ at all depths, classifying them as mildly salinized; 16 sampling points had at least one depth with a total salt content greater than or equal to 2.0‰, classifying them as moderate to severe salinization. The shallow layer was defined as 0 to 60 cm, and the deep layer as 60 to 100 cm. For the mildly salinized group, the correlation coefficients between chloride ions, carbonate ions, bicarbonate ions, sulfate ions, and total salt content were calculated in both the shallow and deep layers. For the moderate to severe salinization group, the correlation coefficients between these four anions and total salt content were calculated at all depths. The shallow layer of the mildly salinized group... The correlation coefficients for different depths were calculated as follows: carbonate ion 0.82, chloride ion 0.65, bicarbonate ion 0.58, and sulfate ion 0.49. For deeper depths, the correlation coefficients were 0.76, 0.61, 0.53, and 0.47. For moderately to severely salinized areas, the correlation coefficients for all depths were 0.89, 0.73, 0.62, and 0.51. The chloride ion with the highest correlation coefficient in the moderately to severely salinized area was selected as the first anion; the carbonate ion with the highest correlation coefficient in the shallow depths of the mildly salinized area was selected as the second anion; and the sulfate ion with the highest correlation coefficient in the deep depths of the mildly salinized area was selected as the third anion. Chloride ions, carbonate ions, and sulfate ions were collectively identified as the characteristic ions for different salinity levels in Dafeng Forest Farm.

[0042] Through the above steps, this application obtained the saline-alkali profile type of the sampling points, the characteristic ions of different salinity areas, and the corresponding basic soil data, providing accurate field basis for the subsequent parameter setting of the HYDRUS model.

[0043] S102. Based on surface vegetation information, saline-alkali profile type and characteristic ions, set the soil profile stratification of each sampling point in the HYDRUS model and the initial water and salt transport parameters of each stratum.

[0044] The HYDRUS model is a specialized model for simulating soil moisture and solute transport. Its simulation accuracy is highly dependent on the rationality of soil profile stratification and initial parameters. Soil profile stratification, based on the characteristics of coastal saline-alkali soils, divides the soil into several continuous intervals according to depth, enabling the model to accurately adapt to water and salt transport patterns at different depths. Initial water and salt transport parameters, including water movement and solute movement parameters, are the fundamental input data for model operation and directly affect the accuracy of the simulation results.

[0045] S102 specifically includes: S1021. Determine the vegetation action coefficient corresponding to each sampling point based on the surface vegetation information, and generate the initial soil profile stratification based on the vegetation action coefficient and the saline-alkali profile type.

[0046] In S102, the vegetation action coefficient corresponding to each sampling point is determined based on surface vegetation information, specifically including: The surface vegetation type at each sampling point is identified based on surface vegetation information. The surface vegetation types include trees, shrubs, herbs, and bare ground. A preset range of action coefficients corresponding to each surface vegetation type is obtained. Based on the surface vegetation coverage at the sampling point, the vegetation action coefficient corresponding to that sampling point is determined within the preset range of action coefficients.

[0047] Among them, the vegetation action coefficient is a parameter that quantifies the degree of influence of vegetation on soil water and salt, and is used to characterize the comprehensive influence of vegetation on soil water absorption capacity and salt accumulation inhibition capacity.

[0048] In this embodiment, the surface vegetation type at each sampling point is first identified based on surface vegetation information. The surface vegetation type includes trees, shrubs, herbs, and bare ground. Then, a preset range of action coefficients corresponding to each surface vegetation type is obtained. Next, the vegetation action coefficient corresponding to the sampling point is determined within the preset range of action coefficients based on the surface vegetation coverage at the sampling point. Finally, an initial soil profile stratification is generated based on the vegetation action coefficients and the saline-alkali profile type.

[0049] In practical applications, taking a sampling point in Dafeng Forest Farm as an example, the surface vegetation type of the sampling point is identified as trees, and the preset action coefficient range for trees is 0.7 to 0.9; the vegetation at the sampling point is fully covered, so the vegetation action coefficient is determined to be 0.85 within the preset range; the saline-alkali profile type at the sampling point is bottom-aggregate type with high deep salt content, and combined with the vegetation action coefficient of 0.85, the initial soil profile is divided into layers of 0 to 20 cm, 20 to 40 cm, 40 to 60 cm, 60 to 80 cm, and 80 to 100 cm.

[0050] S1022. Based on the content distribution of characteristic ions at different depths, the layer boundary depth of the initial soil profile is adjusted to form the final soil profile stratification.

[0051] In this embodiment of the application, based on the initial soil profile stratification generated in step S1021, the layer boundary depth of the initial stratification is adjusted according to the content distribution of characteristic ions at different depths, so that the stratification boundary is aligned with the peak content range or change node of characteristic ions, thus forming the final soil profile stratification.

[0052] In practical applications, taking the same sampling point in Dafeng Forest Farm as an example, this sampling point is slightly salinized. Its characteristic ions are carbonate ions in the shallow layer and sulfate ions in the deep layer. The content of carbonate ions is relatively high at a depth of 0 to 60 cm, with the peak value located at 30 cm. The content of sulfate ions is relatively high at a depth of 60 to 100 cm, with the peak value located at 80 cm. Therefore, the initial 60 cm layer boundary was adjusted to 50 cm, resulting in the final soil profile layers of 0 to 30 cm, 30 to 50 cm, 50 to 80 cm, and 80 to 100 cm.

[0053] S1023. Obtain soil physical property data corresponding to each layer in the soil profile.

[0054] The soil physical properties data include soil density, particle composition, and water holding capacity.

[0055] In this embodiment of the application, for each layer in the final soil profile stratification, the soil density, particle composition and water holding capacity data corresponding to each layer are obtained by laboratory measurement or querying the regional soil database.

[0056] In practical application, the final stratification of the sampling point in Dafeng Forest Farm was measured, and the soil density of the 0 to 30 cm layer was 1.35 g / cm³, with sand accounting for 35%, silt accounting for 45%, clay accounting for 20%, and water holding capacity of 28%; the soil density of the 30 to 50 cm layer was 1.40 g / cm³, with sand accounting for 30%, silt accounting for 48%, clay accounting for 22%, and water holding capacity of 30%.

[0057] S1024. Based on soil physical property data, determine the water movement parameters corresponding to each layer; at the same time, based on the type of characteristic ions, determine the solute movement parameters corresponding to each layer.

[0058] Among them, water movement parameters include water retention capacity parameters and water conduction capacity parameters; solute movement parameters are used to characterize the diffusion and migration capacity of characteristic ions in the soil.

[0059] In this embodiment, the water retention capacity parameter is first obtained by fitting the soil hydraulic parameter model built into the HYDRUS model based on the soil physical property data corresponding to each layer, and the water conduction capacity parameter is calculated by empirical formula based on particle composition and soil density. At the same time, the solute movement parameter corresponding to each layer is set according to the type of characteristic ions and the interaction strength between characteristic ions and soil colloids.

[0060] In practical applications, taking the 0-30 cm layer of the sampling point in Dafeng Forest Farm as an example, the measured data of soil density 1.35 g / cm³, sand content 35%, silt content 45%, and clay content 20% were input. The van Genuchten hydraulic parameter model built into the HYDRUS model was used for fitting, and the water retention capacity parameters were obtained as follows: field capacity 25% and saturated water content 32%. Based on the particle composition and soil density, the water conduction capacity parameter was calculated using the Mualem-van Genuchten empirical formula, and the saturated hydraulic conductivity was found to be 0.02 cm / s. The characteristic ion of this layer is carbonate ion. Carbonate ions easily combine with calcium and magnesium ions in the soil colloids of coastal saline-alkali land to form precipitates. The interaction with soil colloids is strong, and the hindrance effect is obvious. Therefore, referring to the solute parameter range of similar soils, the dispersion coefficient was set to 0.01 cm / s² and the hindrance factor was set to 1.2.

[0061] S1025. Use the water movement parameters and solute movement parameters corresponding to each layer as the initial water-salt transport parameters.

[0062] In this embodiment of the application, the water movement parameters and solute movement parameters corresponding to each layer determined in step S1024 are integrated as the initial water-salt transport parameters of the corresponding sampling point in the HYDRUS model.

[0063] In practical applications, the parameters of each layer at the sampling point in Dafeng Forest Farm are organized into an input format that the model can recognize. The 0 to 30 cm layer includes water retention capacity parameters such as field water holding capacity of 25% and saturated water content of 32%, water conduction capacity parameters such as saturated hydraulic conductivity of 0.02 cm / s, and solute movement parameters such as diffusion coefficient of 0.01 cm / s and retardation factor of 1.2. The other layers are integrated with their corresponding parameter sets according to the same logic.

[0064] Through the above steps, this application achieves personalized and precise setting of HYDRUS model parameters, enabling the model to adapt to the vegetation, profile, ion and soil characteristics of different sampling points.

[0065] S103. The collected meteorological data and groundwater level data of the coastal saline-alkali land are used as the boundary conditions of the HYDRUS model. Based on the soil profile stratification, initial water and salt transport parameters and boundary conditions, the HYDRUS model is run to simulate and output the conductivity data and characteristic ion concentration data of the soil profile at different depths over time at each sampling point.

[0066] S103 specifically includes: S1031. Organize the collected meteorological data into daily meteorological sequences.

[0067] Among them, the daily meteorological sequence is a continuous meteorological data set organized according to the natural calendar days, which includes precipitation data and evaporation data.

[0068] In this embodiment of the application, meteorological observation data of the target area are collected, and the original records are organized into a daily meteorological sequence in chronological order to ensure that the sequence is continuous and without gaps.

[0069] In practical applications, daily meteorological observation data of the Dafeng Forest Farm area over the past year are collected and compiled into a sequence containing daily precipitation and evaporation. For example, on June 1, 202X, precipitation was 15 mm and evaporation was 8 mm; on June 2, precipitation was 0 mm and evaporation was 10 mm; and on June 3, precipitation was 20 mm and evaporation was 7 mm.

[0070] S1032. Set the daily meteorological sequence as the surface boundary condition at the top of the soil profile, and set the groundwater level data as the bottom boundary condition at the bottom of the soil profile.

[0071] Among them, the surface boundary conditions are the interface constraints for the interaction between soil and atmosphere in the model, and the daily meteorological sequence directly controls the water input and output of the soil surface. The bottom boundary conditions are the interface constraints for the interaction between soil and groundwater in the model. Groundwater level data determines the amount of water and salt replenishment from groundwater to the deeper soil layers. The higher the water level, the more significant the impact on the deeper soil layers. This is highly consistent with the characteristic of groundwater in coastal saline-alkali lands that easily participates in the surface salt cycle.

[0072] In this embodiment of the application, the daily meteorological sequence obtained in step S1031 is set as the surface boundary condition at the top of the soil profile to control the moisture exchange between the soil surface and the atmosphere; the collected groundwater level data is set as the bottom boundary condition at the bottom of the soil profile to control the replenishment of water and salt to the deep soil by groundwater.

[0073] In practical applications, the daily meteorological sequence of Dafeng Forest Farm is used as the surface boundary condition; the groundwater level is recorded once a day by burying a water level gauge, and the collected groundwater level monitoring data ranges from 1.2 to 1.8 meters. This data is set as the bottom boundary condition.

[0074] S1033. Input the soil profile into the HYDRUS model, including the initial water and salt transport parameters for each layer, the surface boundary conditions, and the bottom boundary conditions.

[0075] In this embodiment of the application, the soil profile stratification determined in step S1025, the initial water and salt transport parameters corresponding to each stratum, and the surface boundary conditions and bottom boundary conditions set in step S1032 are jointly input into the HYDRUS model.

[0076] In practical applications, the final soil profile of a sampling point in Dafeng Forest Farm was divided into layers of 0 to 30 cm, 30 to 50 cm, 50 to 80 cm, and 80 to 100 cm. The initial water and salt transport parameter sets corresponding to each layer, as well as the daily meteorological sequence and groundwater level data, were all imported into the HYDRUS model input interface to complete the data configuration.

[0077] S1034. The HYDRUS model performs daily simulations according to the set time step to calculate the conductivity and characteristic ion concentration values ​​of each layer at different times.

[0078] In this embodiment of the application, the HYDRUS model performs daily simulation calculations according to a set time step, and calculates the conductivity and characteristic ion concentration values ​​of each layer at different times by solving the soil moisture movement equation and solute transport equation.

[0079] In practical applications, the time step is set to 1 day and the simulation period is 365 days. The model performs calculations on a daily basis. For example, on June 1, there is 15 mm of precipitation, and the top 0 to 30 cm of soil is saturated with water. The water leaches downward, causing carbonate ions to migrate to the 30 to 50 cm layer. The model calculates the carbonate ion concentration and conductivity of each layer on June 1 by solving the solute transport equation. On June 2, there is no precipitation and 10 mm of evaporation. The surface soil moisture decreases, and the deep salts return upward. The model updates the parameters of each layer simultaneously.

[0080] S1035. Record the conductivity and characteristic ion concentration values ​​of each layer at each time point, and generate conductivity data and characteristic ion concentration data of each layer of the soil profile at each sampling point over time.

[0081] Among them, electrical conductivity data are time-series data on the electrical conductivity of soil at different times and depths, output by the model, which can indirectly reflect changes in total salinity. Characteristic ion concentration data are the content data of target anions at different times and spaces, output by the model, which directly reflect the migration patterns of core salinity.

[0082] In this embodiment of the application, the conductivity and characteristic ion concentration values ​​of each layer at each time are recorded to generate conductivity data and characteristic ion concentration data of each layer of the soil profile at each sampling point as a function of time.

[0083] In practical applications, the conductivity and characteristic ion concentration values ​​of each layer are automatically recorded daily during the model calculation process. For example, the conductivity of the 0 to 30 cm layer on June 1, 202X is recorded as 2.1 millisiemens per centimeter and the carbonate ion concentration as 0.8 g per kilogram. After being sorted by date and depth, a time series dataset of conductivity and a time series dataset of characteristic ion concentration are formed.

[0084] Through the above steps, this application obtained continuous time-series data on conductivity and characteristic ion concentration at each sampling point, providing core data support for subsequent spatiotemporal feature extraction.

[0085] S104. Input the conductivity data and characteristic ion concentration data into a two-dimensional convolutional neural network to extract spatiotemporal state features that can simultaneously characterize the coupling mode of salt in time evolution and vertical profile distribution.

[0086] S104 specifically includes: S1041. Organize the conductivity data and characteristic ion concentration data corresponding to each sampling point into two-dimensional input data according to the two dimensions of soil vertical profile depth and time.

[0087] In this embodiment of the application, the conductivity data and characteristic ion concentration data corresponding to each sampling point are organized into two-dimensional input data according to the two dimensions of soil vertical profile depth and time, wherein the depth direction is arranged according to the soil layer order and the time direction is arranged according to the date sequence.

[0088] In practical applications, taking a sampling point in Dafeng Forest Farm as an example, four layers—0 to 30 cm, 30 to 50 cm, 50 to 80 cm, and 80 to 100 cm—are used as depth dimensions, and 365 days from January 1 to December 31, 202X, are used as time dimensions. The conductivity and characteristic ion concentration values ​​of each layer for each day are filled into the corresponding positions, forming two two-dimensional matrices with a size of 4 rows and 365 columns, which serve as the conductivity input data and the characteristic ion concentration input data, respectively.

[0089] S1042. Input the two-dimensional input data into the two-dimensional convolutional neural network. Through the convolution operation of the two-dimensional convolutional neural network, extract the local coupling features of each sampling point in the depth direction and time direction of the soil vertical profile.

[0090] Two-dimensional convolutional neural networks are a type of deep learning network capable of processing information in two dimensions simultaneously. The convolution operation involves sliding a convolutional kernel across the input data to extract feature patterns within local regions.

[0091] In this embodiment of the application, the two-dimensional input data obtained in step S1041 is input into a two-dimensional convolutional neural network. By sliding the convolutional kernel in the depth and time directions, the local coupling features of each sampling point in the depth and time directions of the soil vertical profile are extracted, so that the network can simultaneously perceive the correlation changes of salinity between adjacent depths and adjacent times.

[0092] In practical applications, a two-dimensional convolutional neural network with multiple convolutional layers is constructed, and the kernel size is set to 3x3, that is, each time covering 3 depth points and 3 time points; the conductivity matrix and feature ion concentration matrix are respectively input into the network, and local feature maps are extracted through convolution operations. For example, a certain convolution kernel may capture the pattern of continuous increase in conductivity of the 0 to 30 cm layer over three consecutive days.

[0093] S1043. By performing multi-layer convolution and pooling operations on the local coupling features, a spatiotemporal state feature corresponding to each sampling point is generated that can comprehensively characterize the spatiotemporal evolution law of salinity.

[0094] In this embodiment of the application, by performing multi-layer convolution and pooling operations on the local coupling features obtained in step S1042, the receptive field is gradually expanded and information at different scales is fused, and finally, spatiotemporal state features corresponding to each sampling point that can comprehensively characterize the spatiotemporal evolution law of salinity are generated. The spatiotemporal state features are represented in the form of feature vectors.

[0095] In practical applications, multiple convolutional and pooling layers are stacked in a two-dimensional convolutional neural network. After layer-by-layer processing, the final feature map is flattened into a one-dimensional vector. For example, a spatiotemporal state feature vector with a length of 128 is generated. This vector encodes the evolution of salinity at various depths throughout the year at the sampling point and the correlation information between depths.

[0096] Through the above steps, this application transforms the original time-series data into spatiotemporal state features that can simultaneously reflect the time and depth coupling relationship, laying the foundation for subsequent spatial information fusion.

[0097] S105. Construct a spatial adjacency graph based on the geographic location of each sampling point. Use a graph neural network to fuse the feature information of adjacent nodes in the spatial adjacency graph and iteratively optimize the spatiotemporal state features of each node into global state features that can reflect spatial interactions.

[0098] like Figure 2 As shown, S105 specifically includes: S1051. Calculate the spatial distance between each sampling point based on the geographical coordinates of each sampling point, and determine the spatial adjacency relationship between each sampling point based on the spatial distance.

[0099] Spatial adjacency refers to the adjacency relationship determined based on the geographical distance between sampling points.

[0100] In this embodiment of the application, the geographic coordinates of each sampling point are first obtained, then the spatial distance between each sampling point is calculated, and the spatial adjacency relationship between each sampling point is determined based on the spatial distance. Sampling points with a distance less than a preset threshold are usually regarded as adjacent.

[0101] In practical applications, taking 46 sampling points in Dafeng Forest Farm as an example, the latitude and longitude coordinates of each sampling point are obtained, the Euclidean distance between any two points is calculated, the adjacency distance threshold is set to 500 meters, and sampling points with a distance of less than 500 meters are identified as having a spatial adjacency relationship.

[0102] S1052. Construct a spatial adjacency graph using each sampling point as a graph node and spatial adjacency relationships as edges between nodes.

[0103] Among them, the spatial adjacency graph is a graph structure data constructed with sampling points as nodes and adjacency relationships as edges. The initial features of each node are the spatiotemporal state features of the corresponding sampling point.

[0104] In this embodiment of the application, each sampling point is used as a graph node, and the spatial adjacency relationship determined in step S1051 is used as the edge between nodes to construct a spatial adjacency graph; wherein the initial feature of each node is set as the spatiotemporal state feature vector of the sampling point obtained in step S1043.

[0105] In practical applications, a spatial adjacency graph containing 46 nodes is constructed. Node 1 corresponds to sampling point A in Dafeng Forest Farm, and node 2 corresponds to sampling point B. If the distance between sampling point A and sampling point B is less than 500 meters, an edge is connected between them. Each node is appended with an initial feature vector of length 128.

[0106] S1053. Input the spatial adjacency graph into the graph neural network. Through the neighborhood aggregation operation of the graph neural network, each node receives and aggregates the feature information of its neighboring nodes along the edges.

[0107] Graph neural networks are deep learning networks specifically designed for processing graph-structured data. Neighborhood aggregation refers to each node receiving and aggregating feature information from neighboring nodes along the edges to fuse spatial context.

[0108] In this embodiment of the application, the spatial adjacency graph constructed in step S1052 is input into the graph neural network. Through the neighborhood aggregation operation of the graph neural network, each node receives and aggregates the feature information of adjacent nodes along the edge, thereby realizing information interaction between spatially adjacent sampling points.

[0109] In practical applications, a graph convolutional network is used as the graph neural network, and the aggregation method is set to summation aggregation. Node 1 aggregates the feature vectors of all its neighboring nodes, such as the features of neighboring nodes 2, 3 and 5, and then merges and updates the aggregation result with its own features.

[0110] S1054. Through multiple iterations of the graph neural network, the features of each node are gradually aggregated with spatial context information from multi-level neighbors to generate global state features corresponding to each node that can comprehensively represent spatial interactions.

[0111] In this embodiment of the application, through multiple iterations of the graph neural network, the features of each node are gradually aggregated with spatial context information from multiple-order neighbors. Sampling points that are far apart can also influence each other through multiple transmissions, and finally generate global state features corresponding to each node that can comprehensively represent spatial interactions.

[0112] In practical applications, the graph neural network is set to iterate and update 3 times. In the first iteration, each node aggregates the information of its directly adjacent nodes. In the second iteration, it aggregates the information of neighbors within two steps. In the third iteration, it aggregates the information of neighbors within three steps. After 3 iterations, each node generates a new global state feature vector. For example, the global state feature vector of node 1 has a length of 128. This vector not only contains the spatiotemporal evolution information of node 1 itself, but also integrates the salinity features of surrounding multi-level neighborhood sampling points, which can comprehensively reflect the regional spatial interaction.

[0113] Through the above steps, this application integrates the spatiotemporal features of isolated sampling points into global state features containing spatial correlation information, providing a more comprehensive feature representation for regional-scale salinization prediction.

[0114] S106. Based on global state characteristics and the correlation between characteristic ions and total salt content, predictive information on soil salinization in coastal saline-alkali land is generated.

[0115] S106 specifically includes: S1061. Obtain the conversion relationship between each characteristic ion and the total salt content.

[0116] In this embodiment of the application, the conversion relationship between each characteristic ion and the total salt content is obtained. The conversion relationship is usually obtained based on correlation analysis or experimental calibration and is used to convert the characteristic ion concentration into the corresponding salt contribution value.

[0117] In practical applications, based on the correlation calculation results of step 1014, the conversion factor for chloride ions is set to 0.6, the conversion factor for carbonate ions is set to 0.5, and the conversion factor for sulfate ions is set to 0.4. The conversion formula is that the equivalent salt contribution value is equal to the characteristic ion concentration multiplied by the conversion factor.

[0118] S1062. Based on the characteristic ion concentration data of each sampling point at different depths and times contained in the global state features, calculate the equivalent salt contribution value corresponding to each characteristic ion using the conversion relationship, and accumulate multiple equivalent salt contribution values ​​at the same spatiotemporal location to obtain the simulated total salt content.

[0119] In this embodiment of the application, based on the characteristic ion concentration data of each sampling point at different depths and times contained in the global state features generated in step S1054, the equivalent salt contribution value corresponding to each characteristic ion is calculated using the conversion relationship obtained in step S1061, and multiple equivalent salt contribution values ​​at the same spatiotemporal location are accumulated to obtain the simulated total salt content.

[0120] In practical applications, taking the 0-30 cm layer at a sampling point in Dafeng Forest Farm on July 1, 202X as an example, the global state features extracted the carbonate ion concentration at this spatiotemporal location as 0.8 g / kg, sulfate ion concentration as 0.5 g / kg, and chloride ion concentration as 0.3 g / kg. The equivalent salt contribution value of each ion was calculated: carbonate ion = 0.8 x 0.5 = 0.4 g / kg, sulfate ion = 0.5 x 0.4 = 0.2 g / kg, and chloride ion = 0.3 x 0.6 = 0.18 g / kg. The total simulated salt content was obtained by summing the values: 0.4 + 0.2 + 0.18 = 0.78 g / kg.

[0121] S1063. Match the conductivity data contained in the global state features with the preset conductivity salinity level comparison table to generate the first salinization level sequence, and match the simulated total salt content with the preset total salt salinity level comparison table to generate the second salinization level sequence.

[0122] The electrical conductivity salinity grade comparison table and the total salt salinity grade comparison table are corresponding rules pre-set with reference to the coastal salinity classification standard for salinization of soil.

[0123] In this embodiment of the application, the conductivity data contained in the global state features is matched with a preset conductivity salinity level comparison table to generate a first salinization level sequence; the simulated total salt content obtained in step S1062 is matched with a preset total salt salinity level comparison table to generate a second salinization level sequence.

[0124] In practical applications, the preset conductivity-salinity classification table is as follows: 1 to 2 millisiemens per centimeter corresponds to mild saltiness, and 2 to 4 millisiemens per centimeter corresponds to moderate saltiness. The preset total salt content classification table is as follows: 0.5 to 1.0 g per kilogram corresponds to mild saltiness, and 1.0 to 2.0 g per kilogram corresponds to moderate saltiness. The conductivity of the 0 to 30 cm layer at this sampling point on July 1, 202X was 1.5 millisiemens per centimeter, matching the first level as mild saltiness. The simulated total salt content was 0.78 g per kilogram, matching the second level as mild saltiness. Two classification sequences are generated by traversing all spatiotemporal locations according to this rule.

[0125] S1064. Based on the first and second salinization level sequences, determine the comprehensive salinization level of each sampling point at different depths and times, and generate a comprehensive salinization level sequence.

[0126] In this embodiment of the application, based on the first salinization level sequence and the second salinization level sequence, the comprehensive salinization level of each sampling point at different depths and at different times is determined by the consensus priority principle. If the two levels are consistent, they are directly adopted; if they are inconsistent, the level that matches the characteristic ion type better is used to generate the comprehensive salinization level sequence.

[0127] In practical applications, the first and second levels of the 0 to 30 cm layer at this sampling point on July 1, 202X were both mild, so the overall salinization level was determined to be mild; other spatiotemporal locations were processed according to the same rules to generate a comprehensive salinization level sequence covering all depths and times throughout the year.

[0128] S1065. Based on the concentration ratio of different characteristic ions contained in the global state characteristics, determine the salinization type of each sampling point at different depths and at different times, and obtain the evolution trend of salinization type.

[0129] Among them, salting-alkalization type is a category classified according to the dominant characteristic ion; salting-alkalization type evolution trend refers to the trajectory of different types over time.

[0130] In this embodiment of the application, based on the concentration ratio of different feature ions contained in the global state features, the feature ion with the highest concentration ratio at the same spatiotemporal location is used as the core to determine the salinization type, and the changes in the type at different time points are tracked to obtain the evolution trend of salinization type.

[0131] In practical application, on July 1, 202X, the concentration of carbonate ions in the 0-30 cm layer at this sampling point was 0.8 g / kg, sulfate ions 0.5 g / kg, chloride ions 0.3 g / kg, and the total ion concentration was 1.6 g / kg. The proportion of carbonate ions was 0.8 divided by 1.6, which equals 50%, making it the dominant ion. The salinization type was determined to be carbonate type. Tracking the data from July to August, it was found that the proportion of carbonate ions gradually decreased to 30%, while the proportion of chloride ions increased to 45%. Therefore, it was concluded that the 0-30 cm layer was evolving from carbonate type to chloride type.

[0132] S1066. The comprehensive salinization level sequence and the evolution trend of salinization type will be used together as predictive information for soil salinization in coastal saline-alkali land.

[0133] In this embodiment of the application, the comprehensive salinization level sequence generated in step S1064 and the salinization type evolution trend obtained in step 1065 are used together as prediction information for soil salinization in coastal saline-alkali land.

[0134] In practical applications, taking a sampling point in Dafeng Forest Farm as an example, the predicted information includes the comprehensive salinization level sequence of the sampling point at various depths and times throughout the year, as well as the trend of the evolution from carbonate type to chloride type.

[0135] This application, through the above steps, fuses model simulation data with deep learning features to generate predictive information that combines salinization level and type evolution trend.

[0136] This application combines HYDRUS model simulation with deep learning technology through the above steps, realizing a complete processing flow from single-point vertical simulation to regional spatial correlation of water and salt dynamics in coastal saline-alkali land. This effectively improves the continuity and spatial adaptability of water and salt dynamics prediction, and provides more accurate decision support for saline-alkali land improvement.

[0137] Figure 3 This application provides a schematic diagram of a specific implementation of a precise monitoring system for water and salt dynamics in coastal saline-alkali land based on the HYDRUS model, as shown in the following embodiment. Figure 3 The system may include: Module 31 is used to determine the type of saline-alkali profile and the characteristic ions corresponding to different salinity levels in soil samples within coastal saline-alkali land. The setting module 32 is used to set the soil profile stratification and the initial water and salt transport parameters of each stratification at each sampling point in the HYDRUS model based on the surface vegetation information, the saline-alkali profile type and the characteristic ions. The simulation module 33 is used to take the collected meteorological data and groundwater level data of the coastal saline-alkali land as the boundary conditions of the HYDRUS model, and run the HYDRUS model according to the soil profile stratification, the initial water and salt transport parameters and the boundary conditions to simulate and output the conductivity data and characteristic ion concentration data of the soil profile at different depths over time at each sampling point. Extraction module 34 is used to input the conductivity data and characteristic ion concentration data into a two-dimensional convolutional neural network to extract spatiotemporal state features that can simultaneously characterize the coupling mode of salt in time evolution and vertical profile distribution. The optimization module 35 is used to construct a spatial adjacency graph based on the geographic location of each sampling point, and to fuse the feature information of adjacent nodes in the spatial adjacency graph through a graph neural network, and to iteratively optimize the spatiotemporal state features of each node into global state features that can reflect spatial interactions. The generation module 36 is used to generate prediction information on soil salinization in the coastal saline-alkali land based on the global state features and the correlation between the feature ions and the total salt content.

[0138] The dynamic and precise monitoring system for water and salt in coastal saline-alkali land based on the HYDRUS model in this application embodiment is used to implement the aforementioned dynamic and precise monitoring method for water and salt in coastal saline-alkali land based on the HYDRUS model. Therefore, the specific implementation of the dynamic and precise monitoring system for water and salt in coastal saline-alkali land based on the HYDRUS model can be found in the embodiment section of the dynamic and precise monitoring method for water and salt in coastal saline-alkali land based on the HYDRUS model above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0139] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described method for dynamic and precise monitoring of water and salt in coastal saline-alkali land based on the HYDRUS model.

[0140] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for dynamic and precise monitoring of water and salt in coastal saline-alkali land based on the HYDRUS model.

[0141] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0142] The embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the method for dynamic and precise monitoring of water and salt in coastal saline-alkali land based on the HYDRUS model.

[0143] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0144] The above provides a detailed description of a method for precise monitoring of water and salt dynamics in coastal saline-alkali land based on the HYDRUS model, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for accurately monitoring water-salt dynamics of coastal saline-alkali soil based on a HYDRUS model, characterized in that, include: To determine the type of saline-alkali profile and the characteristic ions corresponding to different salinity levels in soil samples within the coastal saline-alkali land; Based on the surface vegetation information, the saline-alkali profile type, and the characteristic ions, the soil profile stratification and the initial water and salt transport parameters of each stratum at each sampling point in the HYDRUS model are set. The collected meteorological data and groundwater level data of the coastal saline-alkali land are used as the boundary conditions of the HYDRUS model. Based on the soil profile stratification, the initial water and salt transport parameters and the boundary conditions, the HYDRUS model is run to simulate and output the conductivity data and characteristic ion concentration data of the soil profile at different depths over time at each sampling point. The conductivity data and characteristic ion concentration data are input into a two-dimensional convolutional neural network to extract spatiotemporal state features that can simultaneously characterize the coupling mode of salt in time evolution and vertical profile distribution. A spatial adjacency graph is constructed based on the geographic location of each sampling point. The feature information of adjacent nodes in the spatial adjacency graph is fused through a graph neural network, and the spatiotemporal state features of each node are iteratively optimized into global state features that can reflect spatial interactions. Based on the global state characteristics and the correlation between the characteristic ions and the total salt content, predictive information on soil salinization in the coastal saline-alkali land is generated.

2. The method according to claim 1, characterized in that, The process involves constructing a spatial adjacency graph based on the geographic location of each sampling point, fusing the feature information of adjacent nodes in the spatial adjacency graph using a graph neural network, and iteratively optimizing the spatiotemporal state features of each node into global state features that can reflect spatial interactions, including: Based on the geographical coordinates of each sampling point, the spatial distance between each sampling point is calculated, and the spatial adjacency relationship between each sampling point is determined based on the spatial distance. A spatial adjacency graph is constructed using each sampling point as a graph node and the spatial adjacency relationship as the edge between nodes; wherein, the initial feature of each node is the spatiotemporal state feature of the corresponding sampling point. The spatial adjacency graph is input into a graph neural network. Through the neighborhood aggregation operation of the graph neural network, each node receives and aggregates the feature information of its neighboring nodes along the edge. Through multiple iterations of the graph neural network, the features of each node are gradually aggregated with spatial context information from multiple-order neighbors to generate global state features corresponding to each node that can comprehensively represent spatial interactions.

3. The method according to claim 1, characterized in that, The step of inputting the conductivity data and characteristic ion concentration data into a two-dimensional convolutional neural network to extract spatiotemporal state features that can simultaneously characterize the coupling patterns of salt in temporal evolution and vertical profile distribution includes: The conductivity data and characteristic ion concentration data corresponding to each sampling point are organized into two-dimensional input data according to the two dimensions of soil vertical profile depth and time. The two-dimensional input data is input into a two-dimensional convolutional neural network. Through the convolution operation of the two-dimensional convolutional neural network, the local coupling features of each sampling point in the depth direction and time direction of the soil vertical profile are extracted. By performing multi-layer convolution and pooling operations on the local coupling features, spatiotemporal state features corresponding to each sampling point are generated that can comprehensively characterize the spatiotemporal evolution law of salinity.

4. The method according to claim 1, characterized in that, The collected meteorological data and groundwater level data of the coastal saline-alkali land are used as boundary conditions for the HYDRUS model. Based on the soil profile stratification, the initial water and salt transport parameters, and the boundary conditions, the HYDRUS model is run to simulate and output the electrical conductivity and characteristic ion concentration data of the soil profile at different depths over time at each sampling point, including: The collected meteorological data is organized into daily meteorological sequences, which include precipitation data and evaporation data. The daily meteorological sequence is set as the surface boundary condition at the top of the soil profile, and the groundwater level data is set as the bottom boundary condition at the bottom of the soil profile. The soil profile is divided into layers, the initial water and salt transport parameters corresponding to each layer, the surface boundary conditions, and the bottom boundary conditions are input into the HYDRUS model. The HYDRUS model performs daily simulations according to a set time step to calculate the conductivity and characteristic ion concentration values ​​of each layer at different times. Record the conductivity and characteristic ion concentration values ​​of each layer at each time point to generate conductivity and characteristic ion concentration data of each layer of the soil profile at each sampling point over time.

5. The method according to claim 1, characterized in that, The method of generating predictive information for soil salinization in coastal saline-alkali land based on the global state characteristics and the correlation between the characteristic ions and the total salt content includes: Obtain the conversion relationship between each characteristic ion and the total salt content; Based on the characteristic ion concentration data of each sampling point at different depths and times contained in the global state features, the equivalent salt contribution value corresponding to each characteristic ion is calculated using the conversion relationship, and multiple equivalent salt contribution values ​​at the same spatiotemporal location are accumulated to obtain the simulated total salt content. The conductivity data contained in the global state features are matched with a preset conductivity salinity level comparison table to generate a first salinization level sequence, and the simulated total salt content is matched with a preset total salt salinity level comparison table to generate a second salinization level sequence. Based on the first salinization level sequence and the second salinization level sequence, the comprehensive salinization level of each sampling point at different depths and at different times is determined, and a comprehensive salinization level sequence is generated. Based on the concentration ratio of different characteristic ions contained in the global state features, the salinization type of each sampling point at different depths and at different times is determined, and the evolution trend of salinization type is obtained. The comprehensive salinization level sequence and the evolution trend of salinization type are used together as predictive information for soil salinization in coastal saline-alkali land.

6. The method according to claim 1, characterized in that, The determination of the salinity profile type and characteristic ions corresponding to different salinity levels in soil samples within coastal saline-alkali land includes: In the coastal saline-alkali land, soil samples at different depths were obtained through stratified sampling, and the corresponding sampling point locations and surface vegetation information were recorded. The total salt content, electrical conductivity, and composition of water-soluble salt ions in the soil sample were determined. Based on the vertical variation characteristics of the total salt content in the soil profile, the soil profiles at the sampling points are divided into corresponding saline-alkali profile types. Based on the composition information of the water-soluble salt ions, characteristic ions in different salinity areas within the coastal saline-alkali land are determined.

7. The method according to claim 6, characterized in that, The step of determining the characteristic ions of different salinity levels in the coastal saline-alkali land based on the composition information of the water-soluble salt ions includes: Based on the total salt content of soil samples at different depths at each sampling point, all sampling points were divided into a slightly salinized group and a moderately to severely salinized group. Based on the composition information of the water-soluble salt ions, the correlation coefficient between the ion content of each water-soluble salt ion at different depths and the total salt content is calculated for the mildly salinized group and the moderately severely salinized group, respectively. For the moderately to severely salinized group, the first anion with the highest correlation coefficient is selected as the characteristic ion of the moderately to severely salinized region across all depths. For the mildly salinized group, the soil profile is divided into a shallow zone and a deep zone according to the depth: in the shallow zone, the second anion with the largest correlation coefficient is selected as the shallow characteristic ion of the mildly salinized area; in the deep zone, the third anion with the largest correlation coefficient is selected as the deep characteristic ion of the mildly salinized area. The first anion, the second anion, and the third anion are collectively identified as characteristic ions in different salinity regions within the coastal saline-alkali land.

8. The method according to claim 6, characterized in that, The saline-alkali profile types include surface-agglomeration type, bottom-agglomeration type, and oscillating type; The process of classifying the soil profiles at the sampling points into corresponding saline-alkali profile types based on the vertical variation characteristics of the total salt content in the soil profile includes: Based on the total salt content of soil samples at different depths of each sampling point, a sequence of salt content depth variation for each sampling point is formed. Based on the salinity depth change sequence, the salinity vertical feature value of each sampling point is calculated, and the salinity vertical feature value includes shallow cumulative value, bottom cumulative value and fluctuation intensity value; When the salinity depth change sequence of the sampling point shows a continuous decreasing trend from the surface to the depth, and the shallow cumulative value is higher than the bottom cumulative value, the soil profile of the sampling point is determined to be surface-aggregate type. When the salinity depth change sequence of the sampling point shows a continuous increasing trend from the surface to the depth, and the cumulative value of the bottom layer is higher than the cumulative value of the shallow layer, the soil profile of the sampling point is determined to be bottom-aggregate type. When the salinity depth change sequence of the sampling point shows an alternating trend of increasing and decreasing values ​​along the depth direction, and the fluctuation intensity value is higher than a preset threshold, the soil profile of the sampling point is determined to be oscillating.

9. The method according to claim 1, characterized in that, The step of setting soil profile stratification and initial water and salt transport parameters for each sampling point in the HYDRUS model based on surface vegetation information, saline-alkali profile type, and characteristic ions includes: The vegetation action coefficient corresponding to each sampling point is determined based on the surface vegetation information, and an initial soil profile stratification is generated based on the vegetation action coefficient and the saline-alkali profile type. Based on the content distribution of the characteristic ions at different depths, the layer boundary depths of the initial soil profile are adjusted to form the final soil profile stratification. Obtain soil physical property data corresponding to each layer in the soil profile, including soil density, particle composition and water holding capacity; Based on the soil physical property data, the water movement parameters corresponding to each layer are determined, including water retention capacity parameters and water conduction capacity parameters; at the same time, based on the type of characteristic ions, the solute movement parameters corresponding to each layer are determined. The water movement parameters and solute movement parameters corresponding to each layer are used as the initial water-salt transport parameters.

10. The method according to claim 9, characterized in that, The step of determining the vegetation action coefficient corresponding to each sampling point based on the surface vegetation information includes: The surface vegetation type at each sampling point is identified based on surface vegetation information, and the surface vegetation type includes trees, shrubs, herbs and bare ground; Obtain the preset range of action coefficients corresponding to each of the aforementioned surface vegetation types; Based on the surface vegetation cover at the sampling point, the vegetation action coefficient corresponding to the sampling point is determined within the preset action coefficient range; the vegetation action coefficient is used to characterize the comprehensive influence of vegetation on soil moisture absorption capacity and salt accumulation inhibition capacity.

Images