A soil sampling system for detecting soil organic carbon in arid regions

Abstract

The present application relates to the technical field of soil environment monitoring, in particular to a soil sampling system for detecting soil organic carbon in arid regions, which comprises a site planning module, a sampling line and layered sampling points are arranged in the target arid region; a field sampling module, which uses a device with a depth scale to collect undisturbed soil column samples and records the position, depth, vegetation and surface morphology information; a sample processing module, which extracts soil samples at specified depths and performs portable pretreatment; a data transmission module, which transmits the processed soil samples to be tested and related information to an analysis terminal; a content estimation module, which calculates the estimated value of organic carbon content by analyzing the improved spectral estimation algorithm in the analysis terminal and combining the soil sample spectral reflectance; a model generation module, which inputs all estimated contents, corresponding depths, positions, vegetation and surface morphology information into a three-dimensional space interpolation calculation program to generate a three-dimensional space distribution model of soil organic carbon content at different depths in the target arid region.

Description

Technical Field

[0001] This invention relates to the field of soil environmental monitoring technology, and in particular to a soil sampling system for detecting soil organic carbon in arid regions. Background Technology

[0002] Ecosystems in arid regions are fragile, with small and highly spatially heterogeneous soil organic carbon pools. Accurate quantification of their reserves and distribution is crucial for global carbon cycle research. Current methods for soil organic carbon detection primarily rely on traditional chemical analysis, requiring tedious processing and measurement of soil samples collected in the field and then transported to laboratories. This process is time-consuming, costly, and struggles to obtain large-scale, high-density data. In recent years, visible-near-infrared spectroscopy has been introduced for soil property estimation due to its speed and non-destructive nature. However, existing spectral estimation models are mostly developed for surface soils in homogeneous environments. In arid regions with complex and variable environmental factors, the accuracy and stability of soil organic carbon estimation, particularly for middle and deep soils, are generally insufficient. Furthermore, current methods for characterizing the spatial distribution of soil organic carbon commonly employ two-dimensional geographic interpolation to display surface content distribution, or only provide independent two-dimensional displays of content at different depths. This fails to effectively integrate vertical depth and horizontal spatial information, making it difficult to reveal the continuous distribution pattern and transport dynamics of organic carbon in three-dimensional space. Therefore, how to achieve efficient and accurate in-situ detection of soil organic carbon in arid areas, and on this basis, construct a model that can intuitively reflect its distribution characteristics in three-dimensional space, is a key technical bottleneck in accurately assessing the soil carbon pool in arid areas. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a soil sampling system for detecting soil organic carbon in arid areas.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a soil sampling system for detecting soil organic carbon in arid areas, comprising: The sampling point planning module deploys several sampling transects within the target arid area, and sets multiple sampling points at different depths on each sampling transect. The sampling point depths include the surface, middle and deep layers. The on-site sampling module collects undisturbed soil column samples at each sampling point using a soil sampling device equipped with a depth scale, and records the specific location coordinates, sampling depth, vegetation cover type, and surface morphology information of each sampling point. The sample processing module extracts soil samples at a specified depth from the undisturbed soil column sample, and places the soil samples in a portable rapid pretreatment device for preliminary processing to obtain the soil sample to be tested. The data transmission module transmits the soil sample to be tested, location coordinates, sampling depth, vegetation cover type and landform information to the integrated analysis terminal. The content estimation module uses an improved spectral estimation algorithm configured in the analysis terminal, combined with the spectral reflectance of the soil sample to be tested, to calculate the estimated value of the organic carbon content of the soil sample at the corresponding sampling point. The model generation module takes the estimated organic carbon content obtained from all sampling points, combines it with the corresponding sampling depth, location coordinates, vegetation cover type and landform information, and inputs it into the three-dimensional spatial interpolation calculation program to generate a three-dimensional spatial distribution model of soil organic carbon content at different depths in the target arid area.

[0005] As a further aspect of the present invention, the step of collecting undisturbed soil column samples using a soil sampling device with a depth scale and recording the specific location coordinates, sampling depth, vegetation cover type, and surface morphology information of each sampling point includes: At the pre-set location of the sampling line, the sampling tube of the soil sampling device with a depth scale is vertically pressed into the soil until the predetermined sampling depth is reached; Extract the undisturbed soil column sample containing the complete soil layers from the surface to the sampling depth, and immediately load the undisturbed soil column sample into a sealed sample tube with a location identifier; The longitude, latitude, and elevation values ​​of the current sampling point are obtained by using a GPS receiver integrated on the soil sampling device, and the longitude, latitude, and elevation values ​​are combined as the location coordinates. The visual acquisition unit integrated on the soil sampling device is used to acquire images of the vegetation and ground surface around the sampling point. The visual acquisition unit automatically records the acquisition timestamp during image acquisition. The acquired images are transmitted to the image recognition unit, which identifies and outputs vegetation cover type codes and landform descriptors from the images based on a pre-trained model.

[0006] As a further aspect of the present invention, a soil sample at a specified depth is extracted from the undisturbed soil column sample, and the soil sample is placed in a portable rapid pretreatment device for preliminary processing to obtain a soil sample to be tested, comprising: On the temporary sample processing table, the undisturbed soil column sample is taken out from the sealed sample tube and cut at preset depth intervals using a sterile sampler. Soil debris was collected at each depth interval, and the soil debris collected at the same depth was combined to form the soil sample at the corresponding depth. The soil sample is loaded into a dedicated rapid processing crucible, and the rapid processing crucible is placed into the heating chamber of the portable rapid pretreatment device; The portable rapid pretreatment device is activated to remove moisture from the soil sample according to a preset drying program, and then the dried soil sample is ground to the target particle size according to a preset grinding program to obtain homogenized dry soil powder. The rapid processing crucible is removed from the heating chamber, and the homogenized dry soil powder is transferred to a standard sample dish with an optical window. The dish is then sealed, and an electronic tag associated with the location coordinates and the sampling depth is affixed to the outer wall of the standard sample dish.

[0007] As a further aspect of the present invention, the improved spectral estimation algorithm configured within the analysis terminal, combined with the spectral reflectance of the soil sample to be tested, is used to calculate the estimated organic carbon content of the soil sample at the corresponding sampling point, including: The standard sample dish containing the soil sample to be tested is placed in the spectral analysis darkroom of the analysis terminal; The broadband light source in the spectral analysis darkroom is controlled to illuminate the standard sample dish, and the hyperspectral sensor is controlled to collect the spectral reflectance signal of the soil sample to be tested in different bands. After analog-to-digital conversion, a full-band spectral reflectance curve is formed. The reflectance values ​​of multiple characteristic bands corresponding to the characteristic absorption bands of soil organic carbon are extracted from the spectral reflectance curve. The reflectance values ​​of the multiple characteristic bands are input into the improved spectral estimation algorithm, and the improved spectral estimation algorithm outputs the estimated value of organic carbon content corresponding to the soil sample to be tested. The calculated estimated organic carbon content is data-bound with the location coordinates, sampling depth, vegetation cover type, and landform information associated with the current soil sample to be tested, and stored as an organic carbon record.

[0008] As a further aspect of the present invention, the working principle of the improved spectral estimation algorithm is as follows: The improved spectral estimation algorithm internally stores a baseline spectral-organic carbon content database constructed from soil samples from arid regions; When the reflectance values ​​of the multiple characteristic bands of a new soil sample are input, the improved spectral estimation algorithm first performs a spectral morphology similarity search in the benchmark spectral-organic carbon content database. The spectral morphology similarity retrieval is achieved by calculating the Mahalanobis distance between the input spectrum and the database spectrum in the characteristic band, and then filtering out the few database spectral records that are closest in distance. The improved spectral estimation algorithm then calculates the weighted average of the measured organic carbon content values ​​corresponding to several selected database spectral records as a preliminary estimate. The improved spectral estimation algorithm finally introduces a dynamic environmental correction factor, which is obtained based on the vegetation cover type and landform information associated with the current soil sample. The preliminary estimate is multiplied by the dynamic environmental correction factor to obtain the final estimated value of organic carbon content.

[0009] As a further aspect of the present invention, the step of inputting the estimated organic carbon content obtained from all sampling points, combined with the corresponding sampling depth, location coordinates, vegetation cover type, and landform information, into a three-dimensional spatial interpolation calculation program includes: Construct an input dataset containing three-dimensional spatial information of all sampling points and the estimated organic carbon value, wherein the three-dimensional spatial information consists of the location coordinates and the sampling depth; Spatial autocorrelation analysis is performed on the input dataset, and the optimal variogram model for spatial interpolation in the horizontal and vertical directions is determined based on the analysis results. Based on the determined optimal variogram model, calculate the theoretical semivariogram of any two sampling points in the input dataset in three-dimensional space; A three-dimensional interpolation grid is defined with the three-dimensional geographic spatial range of the target arid area as the boundary, and each node of the three-dimensional interpolation grid represents a location to be estimated; For each location to be estimated in the three-dimensional interpolation grid, the optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated based on the Kriging interpolation equations, using the estimated organic carbon content of a predetermined number of sampling points around it, the three-dimensional spatial distance from the sampling points, and the theoretical semivariance.

[0010] As a further aspect of the present invention, for each location to be estimated in the three-dimensional interpolation grid, using the estimated organic carbon content of a predetermined number of surrounding sampling points, the three-dimensional spatial distance from the sampling points, and the theoretical semivariance value, the optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated based on the Kriging interpolation equations, including: For the current position to be estimated, a predetermined number of valid sampling points that are closest to it are searched in three-dimensional space to serve as the interpolation neighborhood sample point set; Calculate the three-dimensional spatial distance from the location to be estimated to each sample point in the interpolation neighborhood sample point set; Based on the optimal variogram model and the calculated three-dimensional spatial distance, the theoretical semivariogram between the estimated location and each sample point, as well as the theoretical semivariogram between any two sample points in the interpolation neighborhood sample point set, are calculated. The coefficient vector on the right side of the Kriging equation system is constructed using the theoretical semi-variance between the location to be estimated and each sample point, and the coefficient matrix on the left side of the Kriging equation system is constructed using the theoretical semi-variance between any two sample points in the interpolation neighborhood sample point set. Solving the Kriging equations yields a set of weight coefficients corresponding to each sample point in the interpolation neighborhood sample point set; The optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated by multiplying the weight coefficient of each sample point with its corresponding estimated organic carbon content and summing the results.

[0011] As a further aspect of the present invention, the system further includes: An uncertainty visualization module is used to quantify the uncertainty of the three-dimensional spatial distribution model, specifically including: After calculating the optimal unbiased estimate of the organic carbon content of each node in the three-dimensional interpolation grid, the kriging estimation variance corresponding to each node is calculated simultaneously. The optimal unbiased estimate of the organic carbon content of all nodes in the three-dimensional interpolation grid is used as a three-dimensional scalar field for rendering to generate spatial distribution data of soil organic carbon content. The Kriging estimate variance of all nodes in the three-dimensional interpolation grid is rendered as another three-dimensional scalar field to generate spatial distribution data of model prediction uncertainty. Spatial alignment is performed between the spatial distribution data of soil organic carbon content and the spatial distribution data of model prediction uncertainty in the same coordinate system. The two types of volume data after spatial alignment are fused and visualized. The distribution map of organic carbon content and the uncertainty map of the corresponding position are displayed synchronously on the same cross-section in the display unit. The uncertainty map is represented by the transparency or color depth to indicate the magnitude of the Kriging estimation variance.

[0012] As a further aspect of the present invention, the system further includes: The sampling scheme optimization module, after generating a three-dimensional spatial distribution model of soil organic carbon content at different depths in the target arid area, analyzes the spatial distribution characteristics of high organic carbon content areas, low organic carbon content areas, and areas with drastic changes in content gradient in the three-dimensional spatial distribution model. Based on the spatial distribution characteristics, identify representative insufficient regions where sampling points fail to fully cover the area. These representative insufficient regions include spatial locations with high model uncertainty. In the region of insufficient representativeness, supplementary sampling transects are planned, and sampling depths are recommended for supplementary sampling points on the newly planned supplementary sampling transects; Based on the recommended supplementary sampling transects and supplementary sampling points, formulate supplementary sampling task instructions; The supplementary sampling task instruction is sent to the on-site mobile sampling terminal to guide the on-site supplementary sampling, and the new data obtained from the supplementary sampling is used to update the three-dimensional spatial distribution model.

[0013] As a further aspect of the present invention, in the region of insufficient representativeness, supplementary sampling transects are planned, and sampling depths are recommended for supplementary sampling points on the newly planned supplementary sampling transects, including: Read the spatial distribution data of the model prediction uncertainty in the three-dimensional spatial distribution model, and extract the set of spatial voxel units in which the Kriging estimation variance is higher than a preset variance threshold. In three-dimensional space, the set of spatial voxel units is projected onto the Earth's surface to form a polygon with high uncertainty on the Earth's surface. Within the polygonal region of high uncertainty, a series of parallel straight line segments are generated according to the principle of uniform spatial coverage, serving as supplementary sampling lines; Along each of the supplementary sampling transects, based on the vertical profile of the spatial distribution data of the model prediction uncertainty of the vertical soil column below the supplementary sampling transect, the depth interval with the greatest uncertainty in the vertical direction is determined. The depth of the center point of the depth range is recommended as the target sampling depth of the supplementary sampling point at the corresponding planar position on the supplementary sampling line.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: An improved spectral estimation algorithm, combined with a portable rapid preprocessing device and analysis terminal, enabled rapid on-site prediction of organic carbon content in soil samples. This algorithm was optimized and calibrated for the spectral characteristics of soils in arid regions and for environmental interference factors, enabling more effective extraction of organic carbon content-related feature information from sample spectral reflectance and reducing the impact of background noise. This approach directly utilizes the processed sample for spectral scanning and calculation, avoiding the time delays of long-distance sample transportation and laboratory handling, significantly shortening the cycle from sampling to obtaining a quantitative prediction. Embedding the algorithm in the analysis terminal allows for automated completion of complex calculations in the field, simplifying operation. Its effectiveness lies in significantly improving the accuracy and stability of organic carbon content estimation in complex arid environments, particularly for mid- and deep soil samples. It overcomes the inefficiencies of traditional laboratory methods and the poor applicability of general spectral models in arid regions, providing a large amount of reliable and timely basic data points for subsequent spatial analysis.

[0015] A three-dimensional spatial interpolation program is used to fuse the estimated organic carbon content obtained from each sampling point with its corresponding three-dimensional geographic coordinates, sampling depth, vegetation cover type, landform, and other multi-dimensional environmental information. This program employs an interpolation algorithm capable of simultaneously processing three-dimensional spatial coordinates and attribute data, transforming the content data from discrete, layered sampling points into a continuous spatial distribution surface. Inputting vegetation and landform information as auxiliary variables constrains and optimizes the interpolation process, making the model more consistent with actual ecological and geomorphological patterns. The result is a three-dimensional model that visually demonstrates the continuous changes in soil organic carbon content in both horizontal and vertical directions. This model overcomes the limitations of traditional two-dimensional planar distribution maps or layered two-dimensional maps, achieving for the first time an integrated and visualized representation of the "length, width, and depth" three-dimensional attributes of the soil organic carbon pool in arid regions. It clearly reveals the spatial heterogeneity of organic carbon distribution in different soil layers, the vertical structure of storage, and possible spatial migration trends, providing an unprecedented spatial analysis tool for a deeper understanding of soil carbon cycling processes in arid regions. Attached Figure Description

[0016] Figure 1 This is a time-series diagram of the soil sampling system for detecting soil organic carbon in arid regions as described in this invention; Figure 2 A flowchart for preparing soil samples for the sample processing module; Figure 3 A flowchart for the improved spectral estimation algorithm. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0019] See Figure 1This invention provides a soil sampling system for detecting soil organic carbon in arid areas. The overall implementation scheme is as follows: The system, through a site planning module, lays out several sampling transects within the target arid area according to preset rules. On each sampling transect, multiple sampling points requiring depth-layered sampling are set, with preset sampling depths covering the surface, middle, and deep layers. The on-site sampling module operates at each designated sampling point. Operators use a soil sampling device with a depth scale to collect undisturbed soil column samples, simultaneously recording the specific location coordinates, actual sampling depth, vegetation cover type, and landform information of the sampling point. The collected samples are processed by the sample processing module, which extracts soil samples from the undisturbed soil column samples at specified depths and places the soil samples in a portable rapid pretreatment device for preliminary processing such as drying and grinding, ultimately obtaining homogenized soil samples for testing. The data transmission module is responsible for transmitting the soil samples to be tested, along with their associated location coordinates, sampling depth, vegetation cover type, and landform information, to an integrated analysis terminal. Within the analysis terminal, the content prediction module is activated, invoking an improved spectral estimation algorithm. Combining this with the measured spectral reflectance of the soil sample, it calculates the predicted organic carbon content at the current sampling point and corresponding depth. The model generation module gathers the predicted organic carbon content from all sampling points and, along with the sampling depth, location coordinates, vegetation cover type, and landform information for each point, imports this data as input into a three-dimensional spatial interpolation calculation program. After running, this program generates a three-dimensional spatial distribution model that reflects the spatial distribution of soil organic carbon content at different depths within the target arid region.

[0020] In one embodiment of the present invention, the on-site sampling module operates as follows: at a pre-defined location on the sampling line, an operator vertically presses the sampling tube of a soil sampling device equipped with a depth scale into the soil until the depth scale indicates a predetermined sampling depth. Subsequently, an undisturbed soil column sample containing complete soil layers from the surface to the sampling depth is extracted and immediately placed into a sealed sample tube with a unique location identifier to maintain its original state. The GPS receiver integrated on the sampling device synchronously acquires the longitude, latitude, and elevation values ​​of the current sampling point, which are then combined and recorded as the location coordinates of the sampling point. Simultaneously, the visual acquisition unit integrated on the soil sampling device acquires images of the vegetation and ground surface surrounding the sampling point and automatically records the timestamp of the images during acquisition. The acquired images are transmitted to an image recognition unit, which, based on a pre-trained deep learning model, automatically identifies and outputs standardized vegetation cover type codes and landform descriptors from the images.

[0021] The execution process of the sample processing module is as follows, please refer to... Figure 2At a temporary sample processing table set up in the field, operators retrieved undisturbed soil column samples from sealed sample tubes. Using a sterile sampler, the undisturbed soil column samples were cut at preset depth intervals. Soil debris generated at each depth interval was collected, and soil debris belonging to the same depth was combined as the soil sample corresponding to that depth. The soil sample was then placed into a dedicated rapid processing crucible, which was placed in the heating chamber of a portable rapid pretreatment device. The portable rapid pretreatment device was activated, and the device heated the soil sample according to a preset drying program to remove moisture. Subsequently, the dried soil clumps were ground to the target particle size according to a preset grinding program, thus obtaining a homogenized soil sample for testing. The data transmission module was responsible for transmitting the soil sample for testing, along with its associated location coordinates, sampling depth, vegetation cover type, and landform information, to the integrated analysis terminal. Within the analysis terminal, the content prediction module was activated, invoking an improved spectral estimation algorithm and, combined with the measured spectral reflectance of the soil sample for testing, calculating the estimated organic carbon content of the soil sample at the current sampling point and corresponding depth. Finally, the model generation module gathers the estimated organic carbon content from all sampling points and combines this with the sampling depth, location coordinates, vegetation cover type, and landform information for each point. This data is then imported into a three-dimensional spatial interpolation program. After running, the program generates a three-dimensional spatial distribution model that reflects the spatial distribution of soil organic carbon content at different depths within the target arid region.

[0022] In practice, the on-site sampling module operates as follows: at a pre-defined location on the sampling transect, the operator vertically presses the sampling tube of the soil sampling device, equipped with a depth gauge, into the soil until the depth gauge indicates the predetermined sampling depth. For example, the surface sampling depth is 0 to 20 cm, the middle layer sampling depth is 20 to 60 cm, and the deep layer sampling depth is 60 to 100 cm. A complete undisturbed soil column sample containing all soil layers from the surface to the predetermined sampling depth is extracted. This undisturbed soil column sample is immediately placed into a sealed sample tube with a unique location identifier to maintain its original state. The location identifier uses the format "SL-2023-001-A," where "SL" represents the transect, "2023" represents the year, "001" represents the transect number, and "A" represents the sampling point number. The GPS receiver integrated on the sampling device works synchronously to acquire the longitude value of the current sampling point. Latitude and elevation values, as well as longitude, latitude, and elevation values, are combined and recorded as the location coordinates of the sampling point. For example, the longitude value is 85.1234 degrees, the latitude value is 42.5678 degrees, and the elevation value is 1500 meters. The visual acquisition unit integrated on the soil sampling device acquires images of the vegetation and ground surface around the sampling point and automatically records the timestamp of the image during acquisition. The timestamp format is "YYYY-MM-DDHH:MM:SS". The acquired images are transmitted to the image recognition unit. Based on a pre-trained deep learning model, the image recognition unit automatically identifies and outputs standardized vegetation cover type codes and landform descriptors from the images. The vegetation cover type codes include "BARE" for bare land, "GRASS" for grassland, and "SHRUB" for shrubs. The landform descriptors include "FLAT" for flat land and "DUNE" for sand dunes.

[0023] In some embodiments, the sample processing module operates as follows: on a temporary sample processing table set up in the field, an operator removes an undisturbed soil column sample from a sealed sample tube, uses a sterile sampler to cut the undisturbed soil column sample at preset depth intervals (every 10 cm), collects soil debris generated at each depth interval, and merges soil debris belonging to the same depth as soil samples corresponding to that depth. For example, soil debris at a depth of 0-10 cm is merged into a topsoil sample, and soil debris at a depth of 10-20 cm is merged into a subsurface soil sample. Then, the soil sample is placed into a dedicated rapid processing crucible, which is then placed in the heating chamber of a portable rapid pretreatment device, and the portable rapid pretreatment device is activated. The portable rapid pretreatment device heats the soil sample according to a preset drying program to remove moisture. The drying program involves continuous heating at 105 degrees Celsius for 2 hours. Subsequently, the dried soil clumps are ground to the target particle size of 0.15 mm according to a preset grinding program, resulting in homogenized dry soil powder. After processing, the rapid processing crucible is removed from the heating chamber, and the homogenized dry soil powder is transferred to a standard sample dish with an optical window and sealed. An electronic tag is attached to the outer wall of the standard sample dish. The information stored in the electronic tag is associated with the location coordinates and sampling depth of the soil sample. The associated information is written using radio frequency identification technology, and the identifier of the electronic tag is consistent with the location identifier of the sealed sample tube.

[0024] Optionally, during image acquisition by the visual acquisition unit, the unit is equipped with a wide-angle lens and a polarizing filter to reduce glare caused by direct sunlight. The pre-trained deep learning model used by the image recognition unit is based on a convolutional neural network architecture. During training, the model utilizes an annotated image dataset containing typical vegetation and landforms of arid regions. In the sample processing module, the grinding program of the portable rapid preprocessing device achieves the target particle size by controlling the grinding time and rotation speed. The grinding time is set to 5 minutes, and the rotation speed is set to 200 revolutions per minute. The depth scale has a graduation accuracy of 1 cm to ensure accurate recording of sampling depth. The GPS receiver's positioning accuracy is better than 2 meters horizontally and better than 5 meters vertically. The sealed sample tube is made of polypropylene, which is chemically inert and airtight, preventing soil sample contamination or moisture loss during transportation. The sterile sampler is sterilized by gamma rays to avoid introducing microbial contamination during cutting. The rapid processing crucible is made of high-temperature resistant ceramic, capable of withstanding the high temperatures of the drying process. The optical window of the standard sample dish is made of quartz glass, with a transmittance exceeding 90% in the visible and near-infrared bands.

[0025] In practice, the collaborative operation of the on-site sampling module and the sample processing module is achieved through wireless communication. The location coordinates, vegetation cover type, and landform information recorded by the on-site sampling module are transmitted to the electronic tag reader / writer of the sample processing module via Bluetooth. The electronic tag reader / writer writes the location coordinates, vegetation cover type, and landform information into the electronic tag of the standard sample dish. In some embodiments, the preset sampling depth stratification can be adjusted according to the soil profile characteristics of the target arid area. For example, in areas with thin soil layers, the deep sampling depth is set to 50 cm. Optionally, the visual acquisition unit also records the ambient light intensity and ambient temperature during image acquisition, and the ambient light intensity and ambient temperature data are appended to the image file for subsequent reference. It is understood that the temporary sample processing table is equipped with a windproof cover and a contamination-proof mat to ensure the cleanliness of sample processing in the field environment.

[0026] In one embodiment of the present invention, the content prediction module executes as follows: a standard sample dish containing the soil sample to be tested is placed in the spectral analysis darkroom inside the analysis terminal. A broadband light source in the spectral analysis darkroom illuminates the optical window of the standard sample dish, and a hyperspectral sensor is controlled to collect the spectral signals reflected by the soil sample in different wavelengths. After analog-to-digital conversion, a full-band spectral reflectance curve is generated. From the obtained spectral reflectance curve, reflectance values ​​of multiple characteristic bands corresponding to the characteristic absorption bands of soil organic carbon are extracted. These characteristic band reflectance values ​​are used as input to an improved spectral estimation algorithm, which calculates and outputs a predicted organic carbon content for the soil sample. The calculated predicted organic carbon content is then data-bound with the location coordinates, sampling depth, vegetation cover type, and landform information associated with the soil sample, and stored together as a structured organic carbon record.

[0027] See Figure 3The improved spectral estimation algorithm works by storing a baseline spectral-organic carbon content database constructed from a large number of arid zone soil samples with known organic carbon contents. When the reflectance values ​​of multiple characteristic bands of a new soil sample are input, the algorithm first performs a spectral morphology similarity search in the baseline spectral-organic carbon content database. This search is achieved by calculating the Mahalanobis distance between the input spectrum and each spectral record in the database on the characteristic bands, and then selecting the closest records based on the distance. The algorithm then calculates the weighted average of the measured organic carbon content values ​​corresponding to these selected records, using this weighted average as a preliminary estimate of the organic carbon content of the new soil sample. Finally, the algorithm introduces a dynamic environmental correction factor, which is obtained by querying a pre-defined correction table based on the vegetation cover type and landform information associated with the current soil sample. Multiplying the calculated preliminary estimate by the obtained dynamic environmental correction factor yields the final estimated organic carbon content.

[0028] In practice, the content prediction module operates by placing a standard sample dish containing the soil sample to be tested into the spectral analysis darkroom inside the analysis terminal. A broadband light source within the darkroom illuminates the optical window of the standard sample dish. The broadband light source's spectral range covers 400 nm to 2500 nm. A hyperspectral sensor, with a spectral resolution better than 10 nm, is then used to acquire the spectral signals reflected by the soil sample in different wavelength bands. The acquired analog spectral signals are converted from analog to digital to form a full-band spectral reflectance curve. The horizontal axis of the spectral reflectance curve represents wavelength, and the vertical axis represents reflectance value. From the obtained spectral reflectance curve, reflectance values ​​for several characteristic bands corresponding to the characteristic absorption bands of soil organic carbon are extracted. These characteristic bands include the central bands within the ranges of 500-600 nm, 800-900 nm, 1700-1800 nm, and 2200-2300 nm. The reflectance values ​​of these characteristic bands are used as input to an improved spectral estimation algorithm. This algorithm calculates and outputs a predicted organic carbon content for the soil sample, expressed in grams per kilogram. The calculated predicted organic carbon content is then linked to the soil sample's location coordinates, sampling depth, vegetation cover type, and landform information, and stored together as a structured organic carbon record. This record includes fields such as sample number, longitude, latitude, elevation, depth, vegetation type, landform, and predicted organic carbon content.

[0029] In some embodiments, the improved spectral estimation algorithm operates by internally storing a baseline spectral-organic carbon content database constructed from a large number of arid zone soil samples with known organic carbon content. Each record in the database contains spectral reflectance data, measured organic carbon content, vegetation type, and environmental description of the sampling location for the soil sample. When the reflectance values ​​of multiple characteristic bands of a new soil sample are input, the improved spectral estimation algorithm first performs a spectral morphology similarity search in the baseline spectral-organic carbon content database. This spectral morphology similarity search is achieved by calculating the Mahalanobis distance between the input spectrum and each spectral record in the database on the characteristic bands, and then selecting the closest 20 database spectral records based on this distance. The improved spectral estimation algorithm then calculates the weighted average of the measured organic carbon content values ​​corresponding to these selected database spectral records. The formula for calculating the weighted average is: in: This represents a preliminary estimate. This indicates the number of spectral records selected. This represents the weight of the i-th spectral record. This represents the measured value of organic carbon content corresponding to the i-th spectral record, with weights... The value is obtained by normalizing the reciprocal of the Mahalanobis distance. This weighted average is used as a preliminary estimate of the organic carbon content of the new soil sample. Finally, the improved spectral estimation algorithm introduces a dynamic environmental correction factor. This factor is obtained by querying a pre-defined correction table based on the vegetation cover type and landform information associated with the current soil sample. The correction table stores the corresponding correction coefficients, indexed by vegetation cover type codes and landform descriptors. Multiplying the calculated preliminary estimate by the obtained dynamic environmental correction factor yields the final estimated organic carbon content.

[0030] Optionally, the hyperspectral sensor needs to be calibrated using a standard white board before acquiring spectral reflectance signals to eliminate the influence of ambient light and instrument dark current. The reflectance values ​​of characteristic bands are extracted using the band center method, that is, the average reflectance of three adjacent bands within the characteristic band range is taken as the reflectance value of that characteristic band. It is understood that the benchmark spectrum-organic carbon content database is periodically updated and expanded using new calibration samples to improve the adaptability and accuracy of the improved spectral estimation algorithm. In the spectral morphology similarity retrieval, the Mahalanobis distance calculation considers the covariance relationship between the reflectance of different characteristic bands. The preset calibration table was established through extensive field comparative experiments, statistically analyzing the systematic deviation between spectral estimation values ​​and laboratory measurements under different vegetation and landforms.

[0031] In practical implementation, the broadband light source is a halogen tungsten lamp, and the emitted spectrum needs to be filtered to eliminate interference light in the ultraviolet and far-infrared bands. The spectral reflectance signal collected by the hyperspectral sensor is transmitted to the main processor of the analysis terminal in the form of a digital sequence. The improved spectral estimation algorithm is integrated into the operating system of the analysis terminal as a software module. In some embodiments, the selection of characteristic bands can be fine-tuned according to the main mineral composition of the target arid area soil type. For example, in gypsum-rich areas, bands related to gypsum characteristic absorption are added. Optionally, the dynamic environmental correction factor ranges from 0.8 to 1.2, and when the vegetation cover type is "BARE" and the landform is "FLAT", the dynamic environmental correction factor is 1.0. It is understood that organic carbon records are stored in the non-volatile memory of the analysis terminal in the form of database rows or JSON format files, and each record has a unique timestamp. The analysis terminal is equipped with a touch screen to display the spectral reflectance curve, the extracted characteristic band reflectance values, and the calculated estimated organic carbon content.

[0032] In one embodiment of the present invention, the execution process of the three-dimensional spatial interpolation calculation program in the model generation module is as follows: First, an input dataset containing three-dimensional spatial information of all sampling points and estimated organic carbon content is constructed, wherein the three-dimensional spatial information is composed of the location coordinates and sampling depth of the sampling points. Spatial autocorrelation analysis is performed on the input dataset, and the optimal variogram model for spatial interpolation in the horizontal and vertical directions is determined based on the analysis results. Based on the determined optimal variogram model, the theoretical semivariogram value of any two sampling points in the input dataset in three-dimensional space is calculated. A regular three-dimensional interpolation grid is defined with the three-dimensional geographic spatial range of the target arid area as the boundary, and each node of the three-dimensional interpolation grid represents a location to be estimated. For each location to be estimated in the three-dimensional interpolation grid, the optimal unbiased estimate of the organic carbon content of the location is calculated based on the Kriging interpolation equation system, using the estimated organic carbon content of a predetermined number of surrounding sampling points, the three-dimensional spatial distance between the location to be estimated and these sampling points, and the theoretical semivariogram value calculated above.

[0033] The calculation of the optimal unbiased estimate of organic carbon content for each location to be estimated in the 3D interpolation grid includes the following steps: For the location to be estimated, search and determine a predetermined number of effective sampling points that are closest to it in 3D space. These points constitute the neighborhood sample point set used for this interpolation. Calculate the 3D spatial distance from the location to be estimated to each sample point in the neighborhood sample point set. Based on the determined optimal variogram model and each calculated 3D spatial distance, calculate the theoretical semivariogram between the location to be estimated and each sample point, and simultaneously calculate the theoretical semivariogram between any two sample points in the neighborhood sample point set. Construct the coefficient vector on the right side of the Kriging equation system using the theoretical semivariogram between the location to be estimated and each sample point, and construct the coefficient matrix on the left side of the Kriging equation system using the theoretical semivariogram between any two sample points in the neighborhood sample point set. Solve the Kriging equation system to obtain a set of weight coefficients corresponding to each sample point in the neighborhood sample point set. Multiply the weight coefficient of each sample point by its corresponding estimated organic carbon content, then sum all the products to calculate the optimal unbiased estimate of the organic carbon content at the location to be estimated.

[0034] In practical implementation, the execution process of the 3D spatial interpolation calculation program in the model generation module is as follows: First, an input dataset containing the 3D spatial information of all sampling points and the estimated organic carbon content is constructed. The 3D spatial information consists of the position coordinates and sampling depth of the sampling points. An example of the input dataset is shown in the table below, which records information for five sampling points. Spatial autocorrelation analysis is performed on this input dataset. The Moran's index is used to calculate the spatial autocorrelation in the horizontal and vertical directions. Based on the curve characteristics of the Moran's index changing with distance, the optimal variogram model for spatial interpolation in the horizontal and vertical directions is determined. For the horizontal direction, the exponential model is determined as the optimal variogram model, and for the vertical direction, the Gaussian model is determined as the optimal variogram model. Based on the determined optimal variogram model, the theoretical semivariogram value of any two sampling points in the input dataset in 3D space is calculated. When calculating the theoretical semivariogram value, different range and sill value parameters are used for the horizontal and vertical directions. Using the three-dimensional geographic spatial extent of the target arid region as the boundary, defined by minimum longitude, maximum longitude, minimum latitude, maximum latitude, minimum depth, and maximum depth, a regular three-dimensional interpolation grid is defined. Each node of the three-dimensional interpolation grid represents a location to be estimated. The grid spacing in the longitude, latitude, and depth directions is set to 100 meters, 100 meters, and 0.1 meters, respectively. For each location to be estimated in the three-dimensional interpolation grid, using the estimated organic carbon content of a predetermined number of surrounding sampling points, the three-dimensional spatial distance between the location to be estimated and these sampling points, and the theoretical semivariance value calculated above, the optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated based on the Kriging interpolation equations, as shown in Table 1.

[0035] Table 1: Input Dataset Table For each location to be estimated in the 3D interpolation grid, the calculation of the optimal unbiased estimate of its organic carbon content includes the following steps: For the location to be estimated, search and determine a predetermined number of valid sampling points in 3D space that are closest to it. These points constitute the neighborhood sample point set used for this interpolation, and the predetermined number is set to 12. Calculate the 3D spatial distance from the location to be estimated to each sample point in the neighborhood sample point set. The calculation formula is as follows: in: This represents the difference in coordinates between the east and west directions. This represents the difference in coordinates between the north and south directions. Indicates the vertical depth difference. This represents the anisotropy scaling factor, used to reconcile the dimensions and scale of variation between horizontal and vertical distances. Based on the determined optimal variogram model and the calculated 3D spatial distances, the theoretical semivariogram between the estimated location and each sample point is calculated, along with the theoretical semivariogram between any two sample points within the interpolation neighborhood sample point set. The coefficient vector on the right side of the Kriging equations is constructed using the theoretical semivariogram between the estimated location and each sample point, and the coefficient matrix on the left side is constructed using the theoretical semivariogram between any two sample points within the interpolation neighborhood sample point set. Solving the Kriging equations yields a set of weight coefficients corresponding to each sample point in the interpolation neighborhood sample point set. The weight coefficient of each sample point is multiplied by its corresponding estimated organic carbon content, and all products are summed to calculate the optimal unbiased estimate of the organic carbon content at the estimated location.

[0036] In some embodiments, spatial autocorrelation analysis is performed by calculating the experimental variogram at different lag distances and fitting a theoretical model. The optimal variogram model is selected based on minimizing the sum of squared residuals. Optionally, the anisotropy scaling factor α is determined by analyzing the ratio of the vertical to the horizontal range. For example, if the horizontal range is 1000 meters and the vertical range is 2 meters, then α is set to 500. It is understood that the range and spacing of the 3D interpolation grid can be adjusted according to the scale of the study area and computational resources; the smaller the spacing, the higher the resolution of the generated 3D spatial distribution model. The predetermined number of 12 is a configurable parameter that can be appropriately reduced in areas with sparse sampling point distribution.

[0037] In practice, the Kriging equations are solved using the standard matrix inversion method. When the condition number of the coefficient matrix is ​​too large, a small regularization term is added to ensure numerical stability. The calculated optimal unbiased estimate of the organic carbon content at the location to be estimated, along with its three-dimensional coordinates, is stored as a new data point. In some embodiments, the same variogram model can be used for both the horizontal and vertical directions, for example, both being spherical models. Optionally, when calculating the three-dimensional spatial distance, the coordinate difference... and The distance needs to be converted to meters based on latitude. It can be understood that after completing the interpolation calculation for all three-dimensional interpolation grid nodes, the optimal unbiased estimate of the organic carbon content of all nodes together constitutes a three-dimensional scalar field. This three-dimensional scalar field is the three-dimensional spatial distribution model of soil organic carbon content at different depths within the target arid region.

[0038] In one embodiment of the present invention, the system further includes an uncertainty visualization module, which is used to quantify and visualize the uncertainty of the generated three-dimensional spatial distribution model. After the model generation module calculates the optimal unbiased estimate of organic carbon content for each node in the three-dimensional interpolation grid, the uncertainty visualization module simultaneously calculates the Kriging estimation variance corresponding to each node. The optimal unbiased estimate of organic carbon content for all nodes in the three-dimensional interpolation grid is rendered as a three-dimensional scalar field to generate spatial distribution data of soil organic carbon content. At the same time, the Kriging estimation variance for all nodes in the three-dimensional interpolation grid is rendered as another three-dimensional scalar field to generate spatial distribution data of model prediction uncertainty. The spatial distribution data of soil organic carbon content and the spatial distribution data of model prediction uncertainty are spatially aligned in the same three-dimensional coordinate system. The two spatially aligned volumetric data are fused and visualized. In the system's display unit, the distribution map of organic carbon content and the uncertainty map at the corresponding location can be displayed synchronously on the same cross-section, wherein the uncertainty map represents the magnitude of the Kriging estimation variance through changes in transparency or color depth.

[0039] In practice, the uncertainty visualization module quantifies the uncertainty of the generated 3D spatial distribution model. After the model generation module calculates the optimal unbiased estimate of the organic carbon content at each node in the 3D interpolation grid, the uncertainty visualization module simultaneously calculates the kriging estimation variance for each node. Calculate using the following formula: in: This indicates the number of neighboring sample points used for interpolation of the current node. This represents the weight coefficient of the j-th neighborhood sample point obtained by solving the Kriging equations. Indicates the current position of the node to be estimated. Location of the j-th neighboring sample point The theoretical semivariance between them This represents the Lagrange multiplier introduced when solving the Kriging equations. The optimal unbiased estimate of organic carbon content at all nodes in the 3D interpolation grid is rendered as a 3D scalar field to generate spatial distribution data of soil organic carbon content. Simultaneously, the Kriging estimate variance at all nodes in the 3D interpolation grid is rendered as another 3D scalar field to generate spatial distribution data of model prediction uncertainty. The spatial distribution data of soil organic carbon content and the spatial distribution data of model prediction uncertainty are spatially aligned in the same 3D coordinate system. Spatial alignment ensures that the two volumes have the same origin, grid spacing, and number of grid nodes in the longitude, latitude, and depth dimensions. The two spatially aligned volumes are then fused and visualized. In the system's display unit, the distribution map of organic carbon content and the corresponding uncertainty map can be simultaneously displayed on the same cross-section. The uncertainty map uses transparency or color depth to represent the magnitude of the Kriging estimate variance; higher transparency or darker color indicates a larger Kriging estimate variance at that location and a higher model prediction uncertainty (see Table 2).

[0040] Table 2: Variance of Kriging Estimation for Some Nodes in the 3D Interpolated Mesh In some embodiments, the rendering of spatial distribution data of soil organic carbon content employs color mapping, mapping the optimal unbiased estimate of organic carbon content to a gradient color scheme from blue to red, where blue represents low content and red represents high content. The rendering of spatial distribution data of model prediction uncertainty employs independent grayscale mapping or transparency mapping, mapping the Kriging estimate variance to transparency or grayscale levels. Optionally, the fusion visualization processing uses overlay rendering technology, first rendering the colors of the spatial distribution data of soil organic carbon content, and then adjusting the transparency of the color at that location based on the spatial distribution data value of the model prediction uncertainty at that location. It is understood that the interactive interface provided by the display unit allows users to freely switch between horizontal cross-sections, vertical profiles, or isosurface display modes of the 3D spatial distribution model; in any display mode, the distribution map and uncertainty map of organic carbon content remain synchronously correlated.

[0041] In practice, the calculation of the Kriging estimate variance and the calculation of the optimal unbiased estimate of organic carbon content are performed in the same interpolation process, using the same neighborhood sample point set and the same Kriging equation coefficient matrix. Spatial alignment is achieved by comparing the metadata (including origin coordinates, grid spacing, and dimension size) of the two volume data and ensuring their consistency. In some embodiments, the spatial distribution volume data of the model prediction uncertainty can be normalized before rendering, linearly scaling the Kriging estimate variance value to the range of 0 to 1 to facilitate uniform control of transparency mapping. Optionally, the display unit can be a standalone computer workstation or a touch screen integrated into the analysis terminal, supporting rotation, scaling, and translation operations of the 3D graphics. It can be understood that in the fused visualization display, users can dynamically change the degree of obscuring of the uncertainty map on the organic carbon content distribution map by adjusting a transparency control slider, thereby more clearly observing the spatial distribution characteristics of organic carbon at different confidence levels.

[0042] In one embodiment of the present invention, the system further includes a sampling scheme optimization module. After generating a three-dimensional spatial distribution model of soil organic carbon content at different depths in the target arid area, the sampling scheme optimization module analyzes the spatial distribution characteristics of high organic carbon content regions, low organic carbon content regions, and regions with drastic changes in organic carbon content gradients within the three-dimensional spatial distribution model. Based on the analyzed spatial distribution characteristics, the module identifies representative insufficient regions where sampling points have not been adequately covered. These representative insufficient regions include spatial locations with high model uncertainty. Within the identified representative insufficient regions, the sampling scheme optimization module plans new supplementary sampling transects and recommends sampling depths for supplementary sampling points on the newly planned supplementary sampling transects. Based on the recommended supplementary sampling transects and supplementary sampling points, the module formulates structured supplementary sampling task instructions. The supplementary sampling task instructions are issued to the field mobile sampling terminal to guide field personnel in conducting supplementary sampling work. Subsequently, the new data obtained from the supplementary sampling is fed back to the system for updating and optimizing the original three-dimensional spatial distribution model.

[0043] The process of planning supplementary sampling transects and recommending sampling depths in the sampling scheme optimization module involves reading the spatial distribution data of the model prediction uncertainty corresponding to the 3D spatial distribution model, extracting all spatial voxel units whose Kriging estimation variance exceeds a preset variance threshold, and forming a set. In 3D space, this set of spatial voxel units is vertically projected onto the Earth's surface, forming one or more high-uncertainty region polygons. Within each high-uncertainty region polygon, a series of parallel straight line segments are generated according to the principle of uniform spatial coverage, and these straight line segments are defined as supplementary sampling transects. Along each supplementary sampling transect, based on the spatial distribution data of the model prediction uncertainty corresponding to the vertical soil column below the transect, its vertical profile is analyzed to determine the depth interval with the greatest uncertainty in the vertical direction. The depth of the center point of this depth interval is recommended as the target sampling depth for the supplementary sampling point at the corresponding planar position on the supplementary sampling transect.

[0044] In practical implementation, the sampling scheme optimization module begins working after generating a three-dimensional spatial distribution model of soil organic carbon content at different depths in the target arid area. The module analyzes the spatial distribution characteristics of high organic carbon content regions, low organic carbon content regions, and regions with drastic changes in organic carbon content gradients within the three-dimensional spatial distribution model. These spatial distribution characteristics are identified by calculating the difference in organic carbon content between adjacent grid nodes in the three-dimensional spatial distribution model and setting a threshold. For example, regions with an optimal unbiased estimate of organic carbon content greater than 8.0 g / kg are defined as high organic carbon content regions, regions with less than 2.0 g / kg are defined as low organic carbon content regions, and regions with a content difference between adjacent nodes greater than 1.0 g / kg / m are defined as regions with drastic changes in content gradients. Based on the analyzed spatial distribution characteristics, the sampling scheme optimization module identifies insufficiently representative regions where sampling points have not adequately covered. These insufficiently representative regions include spatial locations with high model uncertainty. These locations are determined by reading the spatial distribution data of model prediction uncertainty generated by the uncertainty visualization module and selecting grid nodes whose Kriging estimation variance is greater than a preset variance threshold, set to 0.5 (g / kg) squared. Within the identified areas of insufficient representativeness, the sampling scheme optimization module plans new supplementary sampling transects and recommends sampling depths for supplementary sampling points along these transects. Based on the recommended transects and points, the module generates structured supplementary sampling task instructions, presented as a list of numbers and a map overlay. These instructions are then sent to the mobile sampling terminal to guide personnel in conducting supplementary sampling. The new data obtained from the supplementary sampling is subsequently fed back to the system to update and optimize the original 3D spatial distribution model.

[0045] The process of planning supplementary sampling transects and recommending sampling depths in the sampling scheme optimization module involves reading the spatial distribution data of the model prediction uncertainty corresponding to the 3D spatial distribution model, extracting all spatial voxel units whose Kriging estimation variance exceeds a preset variance threshold, and forming a set. In 3D space, this set of spatial voxel units is vertically projected onto the Earth's surface, forming one or more high-uncertainty region polygons. These high-uncertainty region polygons are formed by connecting the outer contours of the voxel units projected onto the Earth's surface. Within each high-uncertainty region polygon, a series of parallel straight line segments are generated according to the principle of uniform spatial coverage. These straight line segments are defined as supplementary sampling transects, with a spacing of 200 meters between them and a direction perpendicular to the prevailing wind direction. Along each supplementary sampling transect, based on the spatial distribution data of the model prediction uncertainty corresponding to the vertical soil column below the supplementary sampling transect, its vertical profile is analyzed to determine the depth interval with the greatest uncertainty in the vertical direction. The depth interval is determined by comparing the average Kriging estimation variance corresponding to each depth layer on the vertical profile. The depth of the center point of the depth interval is recommended as the target sampling depth for the supplementary sampling point at the corresponding planar position on that supplementary sampling transect.

[0046] In some embodiments, the planar locations of the supplementary sampling points are arranged at equal intervals along the supplementary sampling transect, with an interval of 100 meters. The direction of generating parallel straight line segments can also be set based on the principal axis direction of the polygon in the high-uncertainty region. Optionally, when planning the supplementary sampling transect, the sampling scheme optimization module calculates a comprehensive uncertainty index. To assist in decision-making, among which This represents a comprehensive uncertainty index. This represents the average variance of the horizontal Kriging estimate within the polygon of the high-uncertainty region. This represents the average variance of the Kriging estimate in the vertical direction for that region. and These are the weighting coefficients for the horizontal and vertical directions, respectively. It can be understood that the supplementary sampling task instruction includes the recommended geographic coordinates, recommended sampling depth, and recommended sampling priority for each supplementary sampling point.

[0047] In practice, the depth of the center point of the depth interval is obtained by calculating the arithmetic mean of the depths of the upper and lower boundaries of the interval. For example, if the identified maximum uncertainty depth interval is 0.4 meters to 0.7 meters, then the recommended target sampling depth is 0.55 meters. After receiving the supplementary sampling task instruction, the mobile sampling terminal will display the high uncertainty region polygon, the supplementary sampling transect, and the location and recommended depth markers of the supplementary sampling points on its electronic map interface. In some embodiments, when the area of ​​the high uncertainty region is small, the supplementary sampling transect may be planned as a single straight line passing through the center of the region. Optionally, the supplementary sampling task instruction can be sent to the mobile sampling terminal via a wireless network or offline file synchronization. It can be understood that the new data obtained through supplementary sampling will be added to the input dataset of the model generation module, and the three-dimensional spatial interpolation calculation program will be rerun to generate an updated, less uncertain three-dimensional spatial distribution model of soil organic carbon content.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A soil sampling system for detecting soil organic carbon in arid regions, characterized in that, The system includes: The sampling point planning module deploys several sampling transects within the target arid area, and sets multiple sampling points at different depths on each sampling transect. The sampling point depths include the surface, middle and deep layers. The on-site sampling module collects undisturbed soil column samples at each sampling point using a soil sampling device equipped with a depth scale, and records the specific location coordinates, sampling depth, vegetation cover type, and surface morphology information of each sampling point. The sample processing module extracts soil samples at a specified depth from the undisturbed soil column sample, and places the soil samples in a portable rapid pretreatment device for preliminary processing to obtain the soil sample to be tested. The data transmission module transmits the soil sample to be tested, location coordinates, sampling depth, vegetation cover type and landform information to the integrated analysis terminal. The content prediction module uses an improved spectral estimation algorithm configured in the analysis terminal, combined with the spectral reflectance of the soil sample to be tested, to calculate the predicted value of the organic carbon content of the soil sample at the corresponding sampling point. The model generation module takes the estimated organic carbon content obtained from all sampling points, combines it with the corresponding sampling depth, location coordinates, vegetation cover type and landform information, and inputs it into the three-dimensional spatial interpolation calculation program to generate a three-dimensional spatial distribution model of soil organic carbon content at different depths in the target arid area.

2. The soil sampling system for detecting soil organic carbon in arid areas according to claim 1, characterized in that, The process involves collecting undisturbed soil column samples using a soil sampling device equipped with a depth scale, and recording the specific location coordinates, sampling depth, vegetation cover type, and landform information of each sampling point, including: At the pre-set location of the sampling line, the sampling tube of the soil sampling device with a depth scale is vertically pressed into the soil until the predetermined sampling depth is reached; Extract the undisturbed soil column sample containing the complete soil layers from the surface to the sampling depth, and immediately load the undisturbed soil column sample into a sealed sample tube with a location identifier; The longitude, latitude, and elevation values ​​of the current sampling point are obtained by using a GPS receiver integrated on the soil sampling device, and the longitude, latitude, and elevation values ​​are combined as the location coordinates. The visual acquisition unit integrated on the soil sampling device is used to acquire images of the vegetation and ground surface around the sampling point. The visual acquisition unit automatically records the acquisition timestamp during image acquisition. The acquired images are transmitted to the image recognition unit, which identifies and outputs vegetation cover type codes and landform descriptors from the images based on a pre-trained model.

3. A soil sampling system for detecting soil organic carbon in arid regions according to claim 2, characterized in that, Soil samples at a specified depth are extracted from the undisturbed soil column sample. These soil samples are then pre-processed using a portable rapid pretreatment device to obtain the soil sample to be tested, comprising: On the temporary sample processing table, the undisturbed soil column sample is taken out from the sealed sample tube and cut at preset depth intervals using a sterile sampler. Soil debris was collected at each depth interval, and the soil debris collected at the same depth was combined to form the soil sample at the corresponding depth. The soil sample is loaded into a dedicated rapid processing crucible, and the rapid processing crucible is placed into the heating chamber of the portable rapid pretreatment device; The portable rapid pretreatment device is activated, and the moisture in the soil sample is removed according to the preset drying program. Then, the dried soil sample is ground to the target particle size according to the preset grinding program to obtain homogenized dry soil powder. The rapid processing crucible is removed from the heating chamber, and the homogenized dry soil powder is transferred to a standard sample dish with an optical window. The dish is then sealed, and an electronic tag associated with the location coordinates and the sampling depth is affixed to the outer wall of the standard sample dish.

4. A soil sampling system for detecting soil organic carbon in arid areas according to claim 3, characterized in that, Using the improved spectral estimation algorithm configured within the analysis terminal, and combining it with the spectral reflectance of the soil sample, the estimated organic carbon content of the soil sample at the corresponding sampling point is calculated, including: The standard sample dish containing the soil sample to be tested is placed in the spectral analysis darkroom of the analysis terminal; The broadband light source in the spectral analysis darkroom is controlled to illuminate the standard sample dish, and the hyperspectral sensor is controlled to collect the spectral reflectance signal of the soil sample to be tested in different wavelengths. After analog-to-digital conversion, a full-band spectral reflectance curve is formed. The reflectance values ​​of multiple characteristic bands corresponding to the characteristic absorption bands of soil organic carbon are extracted from the spectral reflectance curve. The reflectance values ​​of the multiple characteristic bands are input into the improved spectral estimation algorithm, and the improved spectral estimation algorithm outputs the estimated value of organic carbon content corresponding to the soil sample to be tested. The calculated estimated organic carbon content is data-bound with the location coordinates, sampling depth, vegetation cover type, and landform information associated with the current soil sample to be tested, and stored as an organic carbon record.

5. A soil sampling system for detecting soil organic carbon in arid areas according to claim 4, characterized in that, The working principle of the improved spectral estimation algorithm is as follows: The improved spectral estimation algorithm internally stores a baseline spectral-organic carbon content database constructed from soil samples from arid regions; When the reflectance values ​​of the multiple characteristic bands of a new soil sample are input, the improved spectral estimation algorithm first performs a spectral morphology similarity search in the baseline spectral-organic carbon content database. The spectral morphology similarity retrieval is achieved by calculating the Mahalanobis distance between the input spectrum and the database spectrum in the characteristic band, and then filtering out the few database spectral records that are closest in distance. The improved spectral estimation algorithm then calculates the weighted average of the measured organic carbon content values ​​corresponding to several selected database spectral records as a preliminary estimate. The improved spectral estimation algorithm finally introduces a dynamic environmental correction factor, which is obtained based on the vegetation cover type and landform information associated with the current soil sample. The preliminary estimate is multiplied by the dynamic environmental correction factor to obtain the final estimated value of organic carbon content.

6. A soil sampling system for detecting soil organic carbon in arid regions according to claim 1, characterized in that, The estimated organic carbon content obtained from calculating all sampling points, combined with the corresponding sampling depth, location coordinates, vegetation cover type, and landform information, is input into a three-dimensional spatial interpolation calculation program, including: Construct an input dataset containing three-dimensional spatial information of all sampling points and the estimated organic carbon value, wherein the three-dimensional spatial information consists of the location coordinates and the sampling depth; Spatial autocorrelation analysis is performed on the input dataset, and the optimal variogram model for spatial interpolation in the horizontal and vertical directions is determined based on the analysis results. Based on the determined optimal variogram model, calculate the theoretical semivariogram of any two sampling points in the input dataset in three-dimensional space; A three-dimensional interpolation grid is defined with the three-dimensional geographic spatial range of the target arid area as the boundary, and each node of the three-dimensional interpolation grid represents a location to be estimated; For each location to be estimated in the three-dimensional interpolation grid, the optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated based on the Kriging interpolation equations, using the estimated organic carbon content of a predetermined number of sampling points around it, the three-dimensional spatial distance from the sampling points, and the theoretical semivariance.

7. A soil sampling system for detecting soil organic carbon in arid regions according to claim 6, characterized in that, For each location to be estimated in the three-dimensional interpolation grid, using the estimated organic carbon content of a predetermined number of surrounding sampling points, the three-dimensional spatial distance from the sampling points, and the theoretical semivariance, the optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated based on the Kriging interpolation equations, including: For the current position to be estimated, a predetermined number of valid sampling points that are closest to it are searched in three-dimensional space to serve as the interpolation neighborhood sample point set; Calculate the three-dimensional spatial distance from the location to be estimated to each sample point in the interpolation neighborhood sample point set; Based on the optimal variogram model and the calculated three-dimensional spatial distance, the theoretical semivariogram between the estimated location and each sample point, as well as the theoretical semivariogram between any two sample points in the interpolation neighborhood sample point set, are calculated. The coefficient vector on the right side of the Kriging equation system is constructed using the theoretical semi-variance between the location to be estimated and each sample point, and the coefficient matrix on the left side of the Kriging equation system is constructed using the theoretical semi-variance between any two sample points in the interpolation neighborhood sample point set. Solving the Kriging equations yields a set of weight coefficients corresponding to each sample point in the interpolation neighborhood sample point set; The optimal unbiased estimate of the organic carbon content at the location to be estimated is calculated by multiplying the weight coefficient of each sample point with its corresponding estimated organic carbon content and summing the results.

8. A soil sampling system for detecting soil organic carbon in arid regions according to claim 7, characterized in that, The system also includes: An uncertainty visualization module is used to quantify the uncertainty of the three-dimensional spatial distribution model, specifically including: After calculating the optimal unbiased estimate of the organic carbon content of each node in the three-dimensional interpolation grid, the kriging estimation variance corresponding to each node is calculated simultaneously. The optimal unbiased estimate of the organic carbon content of all nodes in the three-dimensional interpolation grid is used as a three-dimensional scalar field for rendering to generate spatial distribution data of soil organic carbon content. The Kriging estimate variance of all nodes in the three-dimensional interpolation grid is rendered as another three-dimensional scalar field to generate spatial distribution data of model prediction uncertainty. Spatial alignment is performed between the spatial distribution data of soil organic carbon content and the spatial distribution data of model prediction uncertainty in the same coordinate system. The two types of volume data after spatial alignment are fused and visualized. The distribution map of organic carbon content and the uncertainty map of the corresponding position are displayed synchronously on the same cross-section in the display unit. The uncertainty map is represented by the transparency or color depth to indicate the magnitude of the Kriging estimation variance.

9. A soil sampling system for detecting soil organic carbon in arid regions according to claim 8, characterized in that, The system also includes: The sampling scheme optimization module, after generating a three-dimensional spatial distribution model of soil organic carbon content at different depths in the target arid area, analyzes the spatial distribution characteristics of high organic carbon content areas, low organic carbon content areas, and areas with drastic changes in content gradient in the three-dimensional spatial distribution model. Based on the spatial distribution characteristics, identify representative insufficient regions where sampling points fail to fully cover the area. These representative insufficient regions include spatial locations with high model uncertainty. In the region of insufficient representativeness, supplementary sampling transects are planned, and sampling depths are recommended for supplementary sampling points on the newly planned supplementary sampling transects; Based on the recommended supplementary sampling transects and supplementary sampling points, formulate supplementary sampling task instructions; The supplementary sampling task instruction is sent to the on-site mobile sampling terminal to guide the on-site supplementary sampling, and the new data obtained from the supplementary sampling is used to update the three-dimensional spatial distribution model.

10. A soil sampling system for detecting soil organic carbon in arid regions according to claim 9, characterized in that, Within the region of insufficient representativeness, supplementary sampling transects are planned, and sampling depths are recommended for supplementary sampling points on the newly planned supplementary sampling transects, including: Read the spatial distribution data of the model prediction uncertainty in the three-dimensional spatial distribution model, and extract the set of spatial voxel units in which the Kriging estimation variance is higher than a preset variance threshold. In three-dimensional space, the set of spatial voxel units is projected onto the Earth's surface to form a polygon with high uncertainty on the Earth's surface. Within the polygonal region of high uncertainty, a series of parallel straight line segments are generated according to the principle of uniform spatial coverage, serving as supplementary sampling lines; Along each of the supplementary sampling transects, based on the vertical profile of the spatial distribution data of the model prediction uncertainty of the vertical soil column below the supplementary sampling transect, the depth interval with the greatest uncertainty in the vertical direction is determined. The depth of the center point of the depth range is recommended as the target sampling depth of the supplementary sampling point at the corresponding planar position on the supplementary sampling line.

Images