Soil salinity inversion method, apparatus, and electronic device
By combining a multi-level soil inversion sub-model layer and a vertical recursive strategy with a soil water and salt transport model, the problems of time-consuming and labor-intensive traditional soil salinity surveys and insufficient accuracy of deep inversion have been solved, enabling accurate monitoring and prediction of soil salinity and supporting agricultural production and sustainable development.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FARMLAND IRRIGATION RES INST CHINESE ACAD OF AGRI SCI
- Filing Date
- 2026-03-04
- Publication Date
- 2026-06-05
AI Technical Summary
Traditional methods for investigating soil salinity are time-consuming and laborious, and it is difficult to capture the continuous dynamic changes of salinity in time and space. Machine learning-based methods are not accurate enough for deep salinity inversion.
A multi-level soil inversion sub-model layer is adopted. The hydraulic connection between adjacent soil layers is utilized through a vertical recursion strategy. The model is trained in conjunction with a soil water and salt transport model to obtain target feature data and perform inversion. The information relay method is used to reduce the deep prediction error.
It significantly improves the accuracy of salinity inversion, enabling accurate monitoring and prediction of soil salinity changes, and is applicable to agricultural production and sustainable development in arid and semi-arid regions.
Smart Images
Figure CN122157839A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus and electronic device for soil salinity inversion. Background Technology
[0002] Saline-alkali land, also known as saline soil, refers to land types where excessive salt and alkali components accumulate in the topsoil. This includes saline land, alkali land, salinized land, and alkalized land. Soil salinization refers to the phenomenon and process of the continuous increase in surface salt content exceeding a certain limit. Soil salinization is a serious land degradation problem and a significant manifestation of soil environmental disasters. The causes of soil salinization include natural and anthropogenic factors such as seawater erosion, high groundwater levels, and unreasonable land reclamation. Soil salinization is a core issue restricting agricultural productivity and sustainable development in arid and semi-arid regions. Accurate monitoring of soil salinity is a prerequisite for the scientific development and utilization of saline-alkali land, increasing crop yields, and ensuring food security.
[0003] Traditional soil salinity surveys typically employ field sampling and interpolation methods, which are labor-intensive, time-consuming, and destructive, and struggle to capture the continuous dynamic changes in salinity over time and space. Machine learning-based soil salinity retrieval methods suffer from insufficient accuracy because the correlation between surface spectral or environmental information and deep salinity decreases exponentially; therefore, the accuracy of the retrieval results drops significantly with increasing soil depth. Summary of the Invention
[0004] According to one aspect of this application, a soil salinity inversion method is provided, comprising: training an initial soil inversion model to obtain a trained soil inversion model; acquiring target feature data of a target soil region; inputting the target feature data into the soil inversion model to obtain soil salinity inversion data of the target soil region output by the soil inversion model; wherein, the soil inversion model includes multiple levels of soil inversion sub-model layers, and different soil inversion sub-model layers output salinity inversion data of different soil depths; the input of the first-level soil inversion sub-model layer includes the target feature data, the input of the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1.
[0005] According to the soil salinity inversion method provided in this application, the target feature data includes target environmental meteorological data, target soil feature data, and target crop growth data.
[0006] According to the soil salinity inversion method provided in this application, the step of training an initial soil inversion model to obtain a trained soil inversion model includes: a training dataset acquisition step: acquiring a training dataset, the training dataset including historical feature data of the target soil region and corresponding soil salinity data labels; a training step: inputting the historical feature data into the initial soil inversion model to obtain soil salinity prediction data output by the initial soil inversion model; and an optimization step: optimizing the parameters of the initial soil inversion model based on the difference between the soil salinity prediction data and the soil salinity data labels to obtain the soil inversion model.
[0007] According to the soil salinity inversion method provided in this application, the step of obtaining a training dataset includes: obtaining historical feature data of the target soil region; obtaining actual collected salinity data labels corresponding to the historical feature data, wherein the actual collected salinity data labels include salinity data labels of deep soil layers, and the deep soil layers include the portion of soil with a depth above a predetermined depth; obtaining simulated salinity data labels corresponding to the historical feature data, wherein the simulated salinity data labels include salinity data labels of shallow soil layers, and the shallow soil layers include the portion of soil with a depth within a predetermined depth; and constructing a training dataset based on the historical feature data, the actual collected salinity data labels, and the simulated salinity data labels.
[0008] According to the soil salinity inversion method provided in this application, the step of obtaining the simulated salinity data label corresponding to the historical feature data includes: determining the soil water-salt transport model corresponding to the target soil region; inputting the historical feature data of the target soil region into the soil water-salt transport model for numerical simulation to obtain the historical simulated salinity data of the target soil region output by the soil water-salt transport model; and using the shallow soil simulation data in the historical simulated salinity data as the simulated salinity data label.
[0009] According to a soil salinity inversion method provided in this application, determining the soil water and salt transport model corresponding to the target soil region includes: determining whether the target soil region is covered with a film; determining the one-dimensional unsaturated zone soil water and salt transport model as the soil water and salt transport model when there is no film covering; determining a parameter correction factor when there is film covering, and correcting the one-dimensional unsaturated zone soil water and salt transport model by the parameter correction factor to obtain the soil water and salt transport model; wherein, the parameter correction factor is used to correct the simulation parameters of the one-dimensional unsaturated zone soil water and salt transport model, and the simulation parameters include water infiltration boundary conditions and diffusion parameters.
[0010] According to the soil salinity inversion method provided in this application, the historical characteristic data includes historical environmental meteorological data, historical soil characteristic data, and historical crop growth data.
[0011] According to a second aspect of this application, a soil salinity inversion device is provided, comprising: a training module for training an initial soil inversion model to obtain a trained soil inversion model; an acquisition module for acquiring target feature data of a target soil region; and an inversion module for inputting the target feature data into the soil inversion model to obtain soil salinity inversion data of the target soil region output by the soil inversion model; wherein the soil inversion model includes multiple levels of soil inversion sub-model layers, and different soil inversion sub-model layers output salinity inversion data at different soil depths; the input of the first-level soil inversion sub-model layer includes the target feature data, and the input of the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1.
[0012] According to a third aspect of this application, an electronic device is provided, including a memory and a processor, wherein computer program instructions are stored on the memory, and the processor executes any of the soil salinity inversion methods described above when processing the program instructions.
[0013] According to a fourth aspect of this application, a non-volatile computer storage medium is provided, wherein a computer program is stored therein, and the computer program, when executed by a processor, implements any of the soil salinity inversion methods described above.
[0014] The embodiments described in this application have the following beneficial effects: The soil salinity inversion method in this application requires first training a soil inversion model capable of inverting salinity in a target soil region. After training the soil inversion model, target feature data of the target soil region is acquired, and then this target feature data is input into the trained soil inversion model to obtain the soil salinity inversion data of the target soil region output by the soil inversion model. Since the soil inversion model includes multiple levels of soil inversion sub-model layers, different soil inversion sub-model layers correspond to output salinity inversion data at different soil depths. Furthermore, the input of deeper soil inversion sub-model layers includes the output of the previous soil inversion sub-model layer. Based on this vertical recursive strategy, utilizing the hydraulic connection between adjacent soil layers, through an "information relay" approach, the prediction error generated with increasing soil depth can be significantly reduced, thereby improving the accuracy of salinity inversion. Attached Figure Description
[0015] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The accompanying drawings in the following description are merely exemplary embodiments of this disclosure.
[0016] Figure 1 A schematic diagram of the soil salinity inversion method provided in this application; Figure 2 This is a schematic diagram of the simulation results of soil salinity in the 40-60 cm soil layer without a vertical recursion strategy provided in the embodiments of this application; Figure 3 This is a schematic diagram of the simulation results of soil salinity in the 60-80cm soil layer without a vertical recursion strategy provided in the embodiments of this application; Figure 4 This is a schematic diagram of the simulation results of soil salinity in the 40-60 cm soil layer based on the vertical recursion strategy provided in the embodiments of this application; Figure 5 This is a schematic diagram of the simulation results of soil salinity in the 60-80 cm soil layer based on the vertical recursion strategy provided in the embodiments of this application; Figure 6 This is a schematic diagram of the soil salinity inversion device provided in this application. Detailed Implementation
[0017] The technical solutions in this application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.
[0018] It is understandable that soil salinization is a core issue restricting agricultural productivity and sustainable development in arid and semi-arid regions. Accurate monitoring of soil salinity is a prerequisite for the scientific development and utilization of saline-alkali land, increasing crop yields, and ensuring food security. Therefore, after a certain crop is planted or has already been planted in a target soil area, it is necessary to monitor the soil salinity of that area to determine subsequent maintenance strategies, such as irrigation strategies. Based on the above reasons, the first aspect of this application provides a soil salinity inversion method. This method can monitor or predict the salinity of a target soil area in real time, achieving precise monitoring of the salinity of the target soil area.
[0019] The soil salinity inversion method 100 of this application will be described in detail below with reference to the accompanying drawings.
[0020] Figure 1A schematic flowchart of the soil salinity inversion method provided in this application is shown. Figure 1 As shown, the soil salinity inversion method 100 includes steps 101 to 103. Steps 101 to 103 will be described in detail below with reference to the accompanying drawings.
[0021] Step 101: Train the initial soil inversion model to obtain the trained soil inversion model.
[0022] Specifically, the initial soil inversion model can be an untrained artificial intelligence model. By training the initial soil inversion model, a soil inversion model capable of salinity inversion can be obtained. Since the obtained soil inversion model is used for subsequent salinity inversion of the target soil area, the initial soil inversion model should be trained using training data corresponding to the target soil area to obtain a soil inversion model corresponding to the target soil area.
[0023] In one embodiment, training the initial soil inversion model to obtain the trained soil inversion model may include: a training dataset acquisition step: acquiring a training dataset, the training dataset including historical feature data of the target soil region and corresponding soil salinity data labels; a training step: inputting the historical feature data into the initial soil inversion model to obtain soil salinity prediction data output by the initial soil inversion model; and an optimization step: optimizing the parameters of the initial soil inversion model based on the difference between the soil salinity prediction data and the soil salinity data labels to obtain the soil inversion model.
[0024] Specifically, different soil regions may have different environments and soil textures, and their water and salt transport processes may also be different. Therefore, in order to train a soil inversion model suitable for the target soil region, it is necessary to obtain a training dataset related to the target soil region. This training dataset can be pre-stored on the device. For example, the training dataset may include historical feature data of the target soil region and corresponding soil salinity data labels.
[0025] Historical feature data includes feature data of the target soil region at different times within a predetermined historical period. Soil salinity data labels include soil salinity data labels corresponding to the feature data of the target soil region at different times within the predetermined historical period. For example, if the historical feature data includes feature data of the target soil region at M times within a predetermined historical period, then the feature data of the target soil region at each time time corresponds to a soil salinity data label. This soil salinity data label can include soil salinity data at different depths at multiple different locations within the target soil region at that time. For example, it includes shallow salinity data, intermediate salinity data, and deep salinity data at location 1; shallow salinity data, intermediate salinity data, and deep salinity data at location 2; ..., shallow salinity data, intermediate salinity data, and deep salinity data at location n.
[0026] After obtaining the training dataset, historical feature data from the training dataset can be input into the initial soil inversion model for training. The initial soil inversion model will output predicted soil salinity data based on the input historical feature data. The initial soil inversion model has initial parameters (which can be randomly determined or predefined, for example). The initial soil inversion model outputs predicted soil salinity data based on the initial parameters, but these initial parameters are not optimized based on the characteristics of the target soil region, and the accuracy of its output results may be insufficient. Therefore, the parameters of the initial soil inversion model can be optimized based on the difference between the predicted soil salinity data and the soil salinity data labels. For example, during training, the value of the loss function can be calculated based on the difference between the predicted soil salinity data and the soil salinity data labels, and then the model parameters can be updated through the optimizer. This process is repeated iteratively until the model's loss value reaches a predetermined level, thereby optimizing the parameters of the initial soil inversion model and obtaining the soil inversion model. The initial soil inversion model can be, for example, a model built based on Gradient Boosting Tree (GBM) or Deep Neural Network (DNN).
[0027] The historical feature data may include historical environmental meteorological data, historical soil feature data, and historical crop growth data.
[0028] Specifically, historical environmental meteorological data can include environmental meteorological data at different times within a predetermined historical period. This data may include, for example, rainfall, temperature, air humidity, and solar radiation. Historical soil characteristic data can include soil characteristic data at different times within a predetermined historical period. This data may include, for example, soil texture (e.g., sandy or clay soil), soil particle size, soil water holding capacity, and soil moisture content. Historical crop growth data may include, for example, crop growth data at different times within a predetermined historical period. This data may include, for example, crop plant height, crop root density, crop root length, and cumulative leaf area.
[0029] It can be understood that, in order to make the soil inversion model trained to predict the target soil area more accurately, the aforementioned historical crops should be the same as or similar to the crops currently planted or to be planted in the target soil area. This way, the obtained historical crop growth data will better match the current characteristics of the target soil area, helping to improve the inversion accuracy of the soil inversion model. For example, the historical time period can be a portion of the growth period of a historical crop or the entire growth period.
[0030] Further, obtaining the training dataset may include: obtaining historical feature data of the target soil region; obtaining actual collected salinity data labels corresponding to the historical feature data, wherein the actual collected salinity data labels include salinity data labels of deep soil layers, the deep soil layers including the portion of soil at a predetermined depth; obtaining simulated salinity data labels corresponding to the historical feature data, wherein the simulated salinity data labels include salinity data labels of shallow soil layers, the shallow soil layers including the portion of soil at a predetermined depth; and constructing a training dataset based on the historical feature data, the actual collected salinity data labels, and the simulated salinity data labels.
[0031] Specifically, when obtaining the training dataset, it is necessary to obtain historical feature data of the target soil region and the corresponding soil salinity data labels.
[0032] For example, soil water and salt transport models can be used to simulate salt data labels. However, since soil water and salt transport models typically have higher accuracy in simulating shallow soil salt data but lower accuracy in simulating deep soil salt data, shallow soil salt data labels can be obtained through soil water and salt transport models, while deep soil salt data labels can be obtained through actual sampling. It is understood that because simulating salt data labels using soil water and salt transport models is time-saving and labor-saving, more shallow soil salt data labels can be obtained through simulation. Since obtaining salt data labels through actual sampling is time-consuming and labor-intensive, fewer deep soil salt data labels can be obtained through actual sampling. The actually collected salt data labels can be pre-stored in memory or electronic devices for retrieval when training the initial soil inversion model.
[0033] After acquiring historical feature data, corresponding actual collected salinity data labels, and simulated salinity data labels, a training dataset can be constructed. During the construction of the training dataset, the acquired data can be preprocessed. Data preprocessing may include one or more operations such as data cleaning, data transformation, and data normalization. Data cleaning aims to remove noise and outliers from the data, ensuring its accuracy and consistency. Data transformation converts non-numerical data into digital data to facilitate model computation. Data normalization transforms data from different scales or distributions to the same scale for better comparison and analysis.
[0034] In the above description, the shallow soil layer may be, for example, the part within a predetermined depth, and the deep soil layer may be, for example, the part outside the predetermined depth. The predetermined depth may include, for example, 10cm or 15cm, etc. The specific predetermined depth can be determined according to the depth that the simulation model can accurately simulate. This embodiment does not make a specific limitation here.
[0035] For example, obtaining the simulated salinity data label corresponding to the historical feature data may include: determining the soil water and salt transport model corresponding to the target soil region; inputting the historical feature data of the target soil region into the soil water and salt transport model for numerical simulation to obtain the historical simulated salinity data of the target soil region output by the soil water and salt transport model; and using the shallow soil simulation data in the historical simulated salinity data as the simulated salinity data label.
[0036] Specifically, in the process of obtaining the simulated salinity data labels corresponding to the historical feature data, it is first necessary to determine the soil water and salt transport model corresponding to the target soil area. That is, the soil water and salt transport model used for salinity data label simulation needs to match the characteristics of the target soil area in order to obtain salinity data labels that are more consistent with the target soil area.
[0037] After determining the soil water and salt transport model, historical characteristic data of the target soil area can be input into the model. The model will then perform calculations based on this historical data and output historical simulated salinity data for the target soil area. It is understood that the historical simulated salinity data output by the soil water and salt transport model includes simulated salinity data from both the shallow and deep soil layers of the target soil area. Since the accuracy of the simulated salinity data from the deep soil layer is relatively low, only the simulated data from the shallow soil layer in the historical simulated salinity data is used as the label for the simulated salinity data. The soil water and salt transport model can be, for example, a traditional hydrological model based on the Richards equations.
[0038] Understandably, on the one hand, using a soil water and salt transport model can quickly obtain a large number of salt data labels, reducing the difficulty of acquiring training data and providing sufficient training data for the initial soil inversion model. On the other hand, the soil water and salt transport model can simulate long-term water and salt transport, ensuring the correctness of trends over time. Therefore, this method innovatively uses a physical model as a "data generator," transmitting physical conservation laws to the soil inversion model in the form of data. By combining the soil water and salt transport model with the soil inversion model, this method can maintain stability over long time scales and maintain high accuracy in inversion under complex spatial patterns.
[0039] Further, determining the soil water and salt transport model corresponding to the target soil region includes: determining whether the target soil region is covered with a film; determining the one-dimensional unsaturated zone soil water and salt transport model as the soil water and salt transport model when there is no film covering; determining a parameter correction factor when there is film covering, and correcting the one-dimensional unsaturated zone soil water and salt transport model using the parameter correction factor to obtain the soil water and salt transport model; wherein, the parameter correction factor is used to correct the simulation parameters of the one-dimensional unsaturated zone soil water and salt transport model, and the simulation parameters include water infiltration boundary conditions and diffusion parameters.
[0040] Specifically, since whether or not soil is covered with a film significantly affects soil water and salt transport, it is necessary to first determine whether the target soil area is covered with a film when determining the soil water and salt transport model corresponding to the target soil area.
[0041] If the target soil area is not covered, the one-dimensional unsaturated zone soil water and salt transport model can be directly identified as the soil water and salt transport model. The one-dimensional unsaturated zone soil water and salt transport model is a mathematical model used to simulate the changes of water and salt in soil in a non-saturated state over time and space. The spatial changes are mainly considered in the vertical direction (i.e., from the shallow soil layer to the deep soil layer).
[0042] If the target soil area is covered by a plastic film, directly using a one-dimensional unsaturated zone soil water and salt transport model to simulate salt data labeling will result in inaccuracies. Therefore, it is necessary to consider the impact of plastic film on water and salt transport. A parameter correction factor can be determined based on this impact, and the original parameters of the one-dimensional unsaturated zone soil water and salt transport model can be corrected using this factor. The corrected model yields the soil water and salt transport model for the target soil area under plastic film conditions. The simulation parameters that need correction may include water infiltration boundary conditions and diffusion parameters. Water infiltration boundary conditions refer to the settings of surface water content, flux, or water accumulation during infiltration. Diffusion parameters include, for example, diffusivity and hydraulic conductivity in soil water movement parameters.
[0043] This method corrects the infiltration boundary conditions by considering the water retention and confluence effect of the membrane through parameter correction factors. It can reflect the flux concentration and local rinsing effect caused by the reduction of the infiltration area in the membrane pore area. It can also consider the nonlinear effect of the under-membrane heating effect on the diffusion rate of solute molecules and make dynamic temperature correction to the diffusion coefficient.
[0044] Next, step 102 is performed, in which target feature data of the target soil area is obtained.
[0045] Specifically, after obtaining the trained soil inversion model, the salinity of a target soil region can be retrieved based on the model. At this point, it is necessary to acquire target feature data for the target soil region. This target feature data can include current real-time feature data of the target soil region, or feature data for a predetermined future period. In other words, the soil inversion model can be used to retrieve the real-time salinity of a target soil region and also to predict its future salinity.
[0046] Of course, the target feature data can also include feature data of a certain historical period of the target soil area, that is, the historical salinity data of the target soil area can also be inverted through the soil inversion model.
[0047] In summary, the target feature data can be reasonably selected according to the inversion needs, and the embodiments of this application do not impose specific limitations here.
[0048] The target feature data includes target environmental meteorological data, target soil feature data, and target crop growth data.
[0049] Specifically, the target environmental meteorological data may include real-time and / or environmental meteorological data at different times within a predetermined target time period. Environmental meteorological data may include, for example, rainfall, temperature, air humidity, and solar radiation. The target soil characteristic data may include real-time and / or soil characteristic data at different times within a predetermined target time period. Soil characteristic data may include, for example, soil texture (e.g., sandy or clay soil), soil particle size, soil water holding capacity, and soil moisture content. The target crop growth data may include, for example, real-time and / or crop growth data at different times within a predetermined target time period. Crop growth data may include, for example, crop plant height, crop root density, crop root length, and cumulative leaf area.
[0050] For example, the predetermined target time period may include a predetermined future time period, wherein the environmental meteorological data of the predetermined future time period can be obtained through weather forecasts, the soil moisture content in the soil characteristic data of the predetermined future time period can be estimated or simulated, and the crop growth data of the predetermined future time period can be obtained through algorithm model simulation, for example, through the one-dimensional unsaturated zone soil water and salt transport model mentioned above.
[0051] It should be noted that the above description of target feature data is merely exemplary, and the target feature data protected by this application is not limited to the contents listed above. Those skilled in the art can set and plan the target feature data according to the actual situation, as long as it can realize the technical principles of this application.
[0052] Finally, step 103 is executed, whereby the target feature data is input into the soil inversion model to obtain soil salinity inversion data of the target soil region output by the soil inversion model. The soil inversion model comprises multiple levels of soil inversion sub-model layers, with different sub-model layers corresponding to output salinity inversion data at different soil depths. The input to the first-level soil inversion sub-model layer includes the target feature data, and the input to the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1.
[0053] Specifically, after obtaining the target feature data of the target soil region, the target feature data can be input into the soil inversion model to obtain the soil salinity inversion data of the target soil region output by the soil inversion model. This soil inversion model includes multiple levels of soil inversion sub-model layers, the number of which can be determined according to the inversion needs. Different soil inversion sub-model layers can output salinity inversion data for different soil depths. Furthermore, the input to the first-level soil inversion sub-model layer includes the target feature data, and the input to the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1. This means that the results output by each soil inversion sub-model layer are vertically recursively applied along the soil depth direction, effectively capturing the nonlinear accumulation characteristics of deep soil salinity.
[0054] For example, the soil inversion model includes four levels of soil inversion sub-model layers. The first level of soil inversion sub-model layer corresponds to outputting salinity inversion data at a soil depth of 0cm-20cm, the second level of soil inversion sub-model layer corresponds to outputting salinity inversion data at a soil depth of 20cm-40cm, the third level of soil inversion sub-model layer corresponds to outputting salinity inversion data at a soil depth of 40cm-60cm, and the fourth level of soil inversion sub-model layer corresponds to outputting salinity inversion data at a soil depth of 60cm-80cm.
[0055] When the soil inversion model has the structure of the example above, the input of the first-level soil inversion sub-model layer includes the target feature data X1, and the output of the first-level soil inversion sub-model layer is Y1. The input of the second-level soil inversion sub-model layer includes the target feature data X1 and Y1, and the output of the second-level soil inversion sub-model layer is Y2. The input of the third-level soil inversion sub-model layer includes the target feature data X1 and Y2, and the output of the third-level soil inversion sub-model layer is Y3. The input of the fourth-level soil inversion sub-model layer includes the target feature data X1 and Y3, and the output of the fourth-level soil inversion sub-model layer is Y4.
[0056] It should be noted that the above description of the structure of the soil inversion model and the soil depth corresponding to each soil inversion sub-model layer are merely exemplary. The structure of the soil inversion model and the soil depth corresponding to each soil inversion sub-model layer protected by this application are not limited to the contents listed above. Those skilled in the art can set and plan the structure of the soil inversion model and the soil depth corresponding to each soil inversion sub-model layer according to the actual situation, as long as the technical principles of this application can be realized.
[0057] The soil salinity inversion method in this application requires first training a soil inversion model capable of inverting salinity in a target soil region. After training the soil inversion model, target feature data of the target soil region is acquired, and then this target feature data is input into the trained soil inversion model to obtain the soil salinity inversion data of the target soil region output by the soil inversion model. Since the soil inversion model includes multiple levels of soil inversion sub-model layers, different soil inversion sub-model layers correspond to output salinity inversion data at different soil depths. Furthermore, the input of deeper soil inversion sub-model layers includes the output of the previous soil inversion sub-model layer. Based on this vertical recursive strategy, utilizing the hydraulic connection between adjacent soil layers, through an "information relay" approach, the prediction error generated with increasing soil depth can be significantly reduced, thereby improving the accuracy of salinity inversion.
[0058] The following experimental simulation will provide a more intuitive explanation of the effectiveness of the method in this application. Figure 2 This is a schematic diagram of the simulation results of soil salinity in the 40-60cm soil layer without a vertical recursion strategy provided in the embodiments of this application; Figure 3 This is a schematic diagram of the simulation results of soil salinity in the 60-80cm soil layer without a vertical recursion strategy provided in the embodiments of this application; Figure 4 This is a schematic diagram illustrating the simulation results of soil salinity in the 40-60cm soil layer based on a vertical recursion strategy, provided in an embodiment of this application. Figure 5 This is a schematic diagram of the simulation results of soil salinity in the 60-80cm soil layer based on the vertical recursion strategy provided in the embodiments of this application.
[0059] like Figure 2 and Figure 4 As shown, in the simulation results for a 40-60cm soil layer without a vertical recursive strategy: the coefficient of determination (R²) 2 The R² value was 0.79, the root mean square error (RMSE) was 1.64 g / kg, and the normalized root mean square error (NRMSE) was 11.73%. However, in the simulation results based on the vertical recursion strategy (i.e., using the upper-level inversion results as input for the lower-level inversion): R² = 0.79. 2 The value was increased to 0.82, the RMSE decreased to 1.52 g / kg, and the NRMSE decreased to 10.87%.
[0060] like Figure 3 and Figure 5 As shown, in the simulation results for a 60-80cm soil layer without a vertical recursive strategy: R 2 The R value was 0.73, the RMSE was 1.66 g / kg, and the NRMSE was 13.99%. However, in the simulation results based on the vertical recursion strategy: R... 2 The value jumped significantly to 0.83, the RMSE dropped significantly to 1.30 g / kg, and the NRMSE decreased to 10.95%. This shows that for deeper soil layers, the vertical recursive strategy can obtain more accurate inversion results.
[0061] Furthermore, from Figures 2 to 5 Based on the trend of the fitted line and the distribution of sample points, Figure 2 Compared to the fitted lines (straight lines in the figure) Figure 1 The fitted line (straight line in the figure) is closer to the 1:1 standard line (dashed line in the figure), especially in the 60-80cm deep soil layer range. Figure 1 The simulated values show significant dispersion in areas with high salinity, while Figure 2 The simulated values and observed values are in high agreement, indicating that the vertical recursion strategy can effectively capture the variation pattern of deep soil salinity. Figure 2 The confidence intervals (shaded areas around the line) for each sample point are compared to Figure 1 The narrower width indicates that the soil inversion model output is more stable and the uncertainty is significantly reduced after adopting the vertical recursion strategy.
[0062] In summary, this application introduces a vertical recursive strategy, utilizing the vertical continuity of soil water and salt movement, to input the simulation results of the upper soil layer as prior information into the lower model. Experimental results demonstrate that this method exhibits strong robustness and accuracy in simulating deep soil layers (especially at depths of 60-80 cm), and can capture the continuous dynamic changes of salinity in time and space, providing reliable technical support for the accurate inversion of deep soil salinity.
[0063] The soil salinity inversion device provided in this application is described below. The soil salinity inversion device described below can be referred to in correspondence with the soil salinity inversion method described above.
[0064] Figure 6 This is a schematic diagram of the soil salinity inversion device provided in the embodiments of this application, as shown below. Figure 6 As shown, the planned soil salinity inversion device 600 includes: a training module 601, an acquisition module 602, and an inversion module 603.
[0065] The system includes a training module 601 for training an initial soil inversion model to obtain a trained soil inversion model; an acquisition module 602 for acquiring target feature data of a target soil region; and an inversion module 603 for inputting the target feature data into the soil inversion model to obtain soil salinity inversion data of the target soil region output by the soil inversion model. The soil inversion model comprises multiple levels of soil inversion sub-model layers, with different sub-model layers corresponding to different soil depths of salinity inversion data. The input to the first-level soil inversion sub-model layer includes the target feature data, and the input to the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1.
[0066] The soil salinity inversion device in this application can train a soil inversion model capable of inverting salinity in a target soil region. After training the soil inversion model, it can acquire target feature data of the target soil region and input this target feature data into the trained soil inversion model to obtain the soil salinity inversion data of the target soil region output by the soil inversion model. Since the soil inversion model includes multiple levels of soil inversion sub-model layers, different soil inversion sub-model layers correspond to output salinity inversion data at different soil depths. Furthermore, the input of the deep soil inversion sub-model layer includes the output of the previous soil inversion sub-model layer. Based on this vertical recursive strategy, utilizing the hydraulic connection between adjacent soil layers, through an "information relay" method, the prediction error generated with increasing soil depth can be significantly reduced, thereby improving the accuracy of salinity inversion.
[0067] In one embodiment, the target feature data includes target environmental meteorological data, target soil feature data, and target crop growth data.
[0068] In one embodiment, the training module 601 includes: an acquisition submodule, a training submodule, and an optimization submodule.
[0069] The acquisition submodule is used to perform the training dataset acquisition step: acquire the training dataset, which includes historical feature data of the target soil region and corresponding soil salinity data labels.
[0070] The training submodule is used to perform the training steps: inputting the historical feature data into the initial soil inversion model to obtain the soil salinity prediction data output by the initial soil inversion model.
[0071] The optimization submodule is used to perform the optimization steps: based on the difference between the predicted soil salinity data and the soil salinity data labels, the parameters of the initial soil inversion model are optimized to obtain the soil inversion model.
[0072] In one embodiment, the acquisition submodule includes: a first acquisition submodule, configured to acquire historical feature data of the target soil area; a second acquisition submodule, configured to acquire actual collected salinity data labels corresponding to the historical feature data, wherein the actual collected salinity data labels include salinity data labels of deep soil layers, wherein deep soil layers include the portion of soil at a depth above a predetermined depth; a third acquisition submodule, configured to acquire simulated salinity data labels corresponding to the historical feature data, wherein simulated salinity data labels include salinity data labels of shallow soil layers, wherein shallow soil layers include the portion of soil at a depth within a predetermined depth; and a construction submodule, configured to construct a training dataset based on the historical feature data, the actual collected salinity data labels, and the simulated salinity data labels.
[0073] In one embodiment, the third acquisition submodule includes: a determination submodule, used to determine the soil water and salt transport model corresponding to the target soil region; a simulation submodule, used to input historical characteristic data of the target soil region into the soil water and salt transport model for numerical simulation to obtain historical simulated salinity data of the target soil region output by the soil water and salt transport model; and a tagging submodule, used to use shallow soil simulation data in the historical simulated salinity data as tags for the simulated salinity data.
[0074] In one embodiment, the determining submodule includes: a first determining submodule, configured to determine whether the target soil area is covered with a film; a second determining submodule, configured to determine the one-dimensional unsaturated zone soil water and salt transport model as the soil water and salt transport model when there is no film covering; and a third determining submodule, configured to determine a parameter correction factor when there is film covering, and to correct the one-dimensional unsaturated zone soil water and salt transport model using the parameter correction factor to obtain the soil water and salt transport model; wherein the parameter correction factor is used to correct the simulation parameters of the one-dimensional unsaturated zone soil water and salt transport model, and the simulation parameters include water infiltration boundary conditions and diffusion parameters.
[0075] In one embodiment, the historical feature data includes historical environmental meteorological data, historical soil feature data, and historical crop growth data.
[0076] According to a third aspect of this application, an electronic device is also provided, including a memory and a processor. The memory stores computer program instructions, and the processor executes the soil salinity inversion method provided in the above embodiments when processing the program instructions. The principle and scheme of the soil salinity inversion method are described above in conjunction with the various embodiments and accompanying drawings, and will not be repeated here.
[0077] According to a fourth aspect of this application, a non-volatile computer storage medium is also provided, on which a computer program is stored. When executed by a processor, the computer program is implemented to perform the soil salinity inversion method provided in the above embodiments. The soil salinity inversion method includes: training an initial soil inversion model to obtain a trained soil inversion model; acquiring target feature data of a target soil region; inputting the target feature data into the soil inversion model to obtain soil salinity inversion data of the target soil region output by the soil inversion model; wherein, the soil inversion model includes multiple levels of soil inversion sub-model layers, and different soil inversion sub-model layers output salinity inversion data of different soil depths; the input of the first-level soil inversion sub-model layer includes the target feature data, the input of the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1. The principle and scheme of the soil salinity inversion method are described above in conjunction with the various embodiments and accompanying drawings, and will not be repeated here.
[0078] Fifthly, this application also provides a computer program product, which includes a computer program that can be stored on a non-volatile computer storage medium. When the computer program is executed by a processor, the computer can execute the soil salinity inversion method provided by the above methods. The soil salinity inversion method includes: training an initial soil inversion model to obtain a trained soil inversion model; acquiring target feature data of a target soil region; inputting the target feature data into the soil inversion model to obtain soil salinity inversion data of the target soil region output by the soil inversion model; wherein, the soil inversion model includes multiple levels of soil inversion sub-model layers, and different soil inversion sub-model layers output salinity inversion data of different soil depths; the input of the first-level soil inversion sub-model layer includes the target feature data, the input of the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1. The principle and scheme of the soil salinity inversion method are described above in conjunction with various embodiments and accompanying drawings, and will not be repeated here.
[0079] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0080] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0081] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.
[0082] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.
[0083] In this application, unless otherwise stated, directional terms such as "up" and "down" are generally used in relation to the direction shown in the accompanying drawings, or in relation to the vertical, perpendicular, or gravitational direction; similarly, for ease of understanding and description, "left" and "right" are generally used in relation to the left and right shown in the accompanying drawings; "inner" and "outer" refer to the inner and outer contours of each component itself, but the above directional terms are not intended to limit this application.
[0084] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope described in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for soil salinity inversion, characterized in that, include: The initial soil inversion model is trained to obtain the trained soil inversion model. Acquire target feature data for the target soil region; The target feature data is input into the soil inversion model to obtain the soil salinity inversion data of the target soil area output by the soil inversion model; The soil inversion model includes multiple soil inversion sub-model layers, with different soil inversion sub-model layers corresponding to output salinity inversion data at different soil depths; the input of the first-level soil inversion sub-model layer includes the target feature data, and the input of the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1.
2. The soil salinity inversion method according to claim 1, characterized in that, The target feature data includes target environmental meteorological data, target soil feature data, and target crop growth data.
3. The soil salinity inversion method according to claim 1, characterized in that, The process of training the initial soil inversion model to obtain the trained soil inversion model includes: Training dataset acquisition steps: Acquire the training dataset, which includes historical feature data of the target soil region and corresponding soil salinity data labels; Training steps: Input the historical feature data into the initial soil inversion model to obtain the soil salinity prediction data output by the initial soil inversion model; Optimization steps: Based on the difference between the predicted soil salinity data and the soil salinity data labels, the parameters of the initial soil inversion model are optimized to obtain the soil inversion model.
4. The soil salinity inversion method according to claim 3, characterized in that, The acquisition of the training dataset includes: Obtain historical characteristic data of the target soil area; Obtain the actual collected salinity data tags corresponding to the historical feature data, wherein the actual collected salinity data tags include salinity data tags of deep soil layers, and deep soil layers include the portion of soil at a depth above a predetermined depth; Obtain simulated salinity data labels corresponding to the historical feature data, wherein the simulated salinity data labels include salinity data labels of the shallow soil layer, and the shallow soil layer includes the portion of the soil within a predetermined depth; A training dataset is constructed based on the historical feature data, the actual collected salinity data labels, and the simulated salinity data labels.
5. The soil salinity inversion method according to claim 4, characterized in that, The step of obtaining the simulated salinity data labels corresponding to the historical feature data includes: Determine the soil water and salt transport model corresponding to the target soil region; The historical characteristic data of the target soil area are input into the soil water and salt transport model for numerical simulation to obtain the historical simulated salinity data of the target soil area output by the soil water and salt transport model. The shallow soil simulation data from the historical simulated salinity data is used as the label for the simulated salinity data.
6. The soil salinity inversion method according to claim 5, characterized in that, The step of determining the soil water and salt transport model corresponding to the target soil region includes: Determine whether the target soil area is covered with a film; Without a film covering, a one-dimensional unsaturated zone soil water and salt transport model was determined as the soil water and salt transport model; In the case of mulching, a parameter correction factor is determined, and the one-dimensional unsaturated zone soil water and salt transport model is corrected using the parameter correction factor to obtain the soil water and salt transport model; wherein, the parameter correction factor is used to correct the simulation parameters of the one-dimensional unsaturated zone soil water and salt transport model, and the simulation parameters include water infiltration boundary conditions and diffusion parameters.
7. The soil salinity inversion method according to claim 5 or 6, characterized in that, The historical characteristic data includes historical environmental meteorological data, historical soil characteristic data, and historical crop growth data.
8. A soil salinity inversion device, characterized in that, include: The training module is used to train the initial soil inversion model to obtain the trained soil inversion model. The acquisition module is used to acquire target feature data of the target soil area; The inversion module is used to input the target feature data into the soil inversion model to obtain the soil salinity inversion data of the target soil area output by the soil inversion model; The soil inversion model includes multiple soil inversion sub-model layers, with different soil inversion sub-model layers corresponding to output salinity inversion data at different soil depths; the input of the first-level soil inversion sub-model layer includes the target feature data, and the input of the Nth-level soil inversion sub-model layer includes the target feature data and the output data of the (N-1)th-level soil inversion sub-model layer, where N is a positive integer greater than 1.
9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and the processor executes the soil salinity inversion method according to any one of claims 1-7 when processing the program instructions.
10. A non-volatile computer storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the soil salinity inversion method according to any one of claims 1-7.