A method for assessing the risk of heavy metal pollution in orchard soils using multi-source data fusion

Abstract

The present invention discloses a method for assessing the risk of heavy metal contamination in orchard soil by fusion of multi-source data, which relates to the technical field of soil heavy metal contamination risk assessment, and comprises the following steps: constructing a two-layer dynamic adaptive grid to divide the orchard, setting sub-sample points based on the drip line of the fruit tree; performing pXRF in situ detection; performing ICP-MS ex situ detection; judging the fusion condition by correlation, and when the fusion condition is met, using ordinary co-kriging and sequential Gaussian random simulation network for co-simulation to draw a heavy metal predicted content distribution map and a prediction error variance map. This application improves sample point coverage, balances low cost and high precision, and improves the reliability of risk assessment by combining two-layer dynamic division, fruit tree drip line, judging fusion conditions, ordinary co-kriging and sequential Gaussian random simulation network co-simulation.

Description

Technical Field

[0001] The present application relates to the technical field of soil heavy metal pollution risk assessment, and in particular to a method for assessing the risk of heavy metal pollution in orchard soil by fusing multi-source data. Background Art

[0002] Heavy metal contamination in orchard soils can enter the food chain through fruit accumulation, posing a direct threat to human health. Therefore, accurately assessing the contamination risk is crucial. However, existing assessment methods have significant limitations: traditional grid divisions are rigid and difficult to adapt to the distribution characteristics of fruit trees, and they can easily miss key areas prone to heavy metal accumulation, such as drip lines. Among single detection methods, pXRF in situ testing is efficient and low-cost but has limited accuracy, while ICP-MS ex situ testing is highly accurate but time-consuming and expensive, making it difficult to strike a balance between efficiency and accuracy. Data fusion lacks scientific conditional judgment, which can easily lead to conflicts in multi-source data and affect prediction reliability. Furthermore, traditional methods often output a single predicted value, unable to quantify uncertainty and making it difficult to support accurate risk decision-making. Therefore, research is needed on methods for assessing the risk of heavy metal contamination in orchard soils using multi-source data fusion.

[0003] In the prior art, Chinese patent CN116561550A discloses a soil sampling optimization method for heavy metal contaminated sites based on information entropy. First, preliminary survey data corresponding to the target heavy metal contaminated site are constructed, and then a target kriging algorithm is executed to obtain the prior mean and prior covariance of the heavy metal concentrations of candidate sampling points in the target heavy metal contaminated site. Then, an optimization design method based on information entropy is used to conduct detailed survey sampling points on the site to obtain detailed survey data corresponding to the target heavy metal contaminated site. Finally, the preliminary survey data and the detailed survey data are coupled, and the target kriging algorithm is executed to perform spatial interpolation on the target heavy metal contaminated site to monitor the heavy metal concentration and spatial distribution of pollution in the target heavy metal contaminated site.

[0004] However, the above-mentioned existing technologies do not take into account the distribution characteristics of fruit trees, and the sample coverage is low; they do not consider the fusion of sampling data, and cannot take into account both low cost and high precision; and they cannot quantify uncertainty, making it difficult to support accurate risk decision-making, and the reliability of risk assessment is low. Summary of the Invention

[0005] This application provides a method for assessing the risk of heavy metal pollution in orchard soil by fusing multi-source data, which is used to solve the problems of low sample coverage of existing soil heavy metal pollution risk assessment technologies, inability to balance low cost and high precision, and low reliability of risk assessment.

[0006] In one aspect, the present application provides a method for assessing the risk of heavy metal pollution in orchard soil by fusion of multi-source data, comprising the following steps:

[0007] Step 1: Construct a two-layer dynamic adaptive grid to divide the orchard twice, obtain the basic unit of the initial division and the grid unit of the secondary division, and set several sub-sample points in each grid unit based on the drip line of the fruit tree.

[0008] Step 2: Use a portable X-ray fluorescence spectrometer to perform pXRF in-situ detection on each sub-sample point, calculate the heavy metal pXRF in-situ detection data of each grid unit, and construct a pXRF in-situ detection dataset.

[0009] Step 3: Select a collection unit in each basic unit, use an ICP-MS inductively coupled plasma mass spectrometer to perform ICP-MS ex situ detection on each collection unit, obtain the heavy metal ICP-MS ex situ detection data of each collection unit, and construct an ICP-MS ex situ detection data set.

[0010] Step 4: Determine the fusion condition of the pXRF in-situ detection data set and the ICP-MS ex-situ detection data set based on the correlation between the heavy metal pXRF in-situ detection data and the heavy metal ICP-MS ex-situ detection data in the same acquisition unit. When the correlation meets the fusion condition, proceed to step 5.

[0011] Step 5: Ordinary co-kriging and sequential Gaussian random simulation networks are used to co-simulate the pXRF in-situ detection dataset and the ICP-MS ex-situ detection dataset, and a heavy metal predicted content distribution map and a prediction error variance map are drawn to analyze the spatial distribution pattern and hotspot areas and evaluate local uncertainties.

[0012] In one possible implementation, step one includes:

[0013] The orchard was preliminarily divided based on remote sensing images to obtain several basic units.

[0014] According to the distribution of fruit trees in rows and columns, the grid method is used to divide each basic unit into grid units of equal area.

[0015] Based on the drip line of the fruit trees, the plum blossom method was used to set several sub-sample points in each grid unit.

[0016] In one possible implementation, in step 2, the accuracy and stability of pXRF in-situ detection are ensured by combining standard substances with quantitative limit indicators, relative standard deviation indicators, and relative error indicators.

[0017] In one possible implementation, step three includes:

[0018] For regular basic units, the grid units at the four corners and the center are used as collection units. For irregular basic units, randomly sampled grid units are selected as collection units.

[0019] A mixed soil sample was collected from each collection unit, and the mixed soil sample from each collection unit was tested in the laboratory using an ICP-MS inductively coupled plasma mass spectrometer to obtain the heavy metal ICP-MS ex situ detection data of each collection unit and construct an ICP-MS ex situ detection dataset.

[0020] In one possible implementation, in step 4, when the correlation does not meet the fusion condition, soil sampling and pXRF ex situ detection are performed on each sub-sample point, and the correlation is optimized until the fusion condition is met.

[0021] In a possible implementation, the correlation uses an R² indicator. When the R² indicator is greater than or equal to a preset threshold, the correlation meets the fusion condition, otherwise it does not meet the condition.

[0022] In one possible implementation, in step 4, the soil sampling and pXRF ex situ detection at each sub-sample point includes:

[0023] Soil samples were collected at each sub-sample point, dried naturally, sieved through a stainless steel sieve, placed in a polyethylene container, pressed into sheets, and then covered with Mylar film.

[0024] The treated soil samples were subjected to pXRF ex situ detection using a portable X-ray fluorescence spectrometer.

[0025] In one possible implementation, step five includes:

[0026] In the acquisition unit, a linear regression model is constructed between the pXRF in-situ detection dataset and the ICP-MS ex-situ detection dataset, with the pXRF in-situ detection dataset serving as the primary variable and the ICP-MS ex-situ detection dataset serving as the auxiliary variable.

[0027] The variograms of the main variable and the auxiliary variable are calculated respectively, and the cross-variogram is calculated, and a variogram model is obtained by fitting. Based on the fitted variogram model, ordinary cokriging method is used to perform interpolation prediction on each grid cell to obtain the heavy metal prediction value and prediction error variance of each grid cell.

[0028] The pXRF in situ detection dataset is converted into a synergistic variable set through the linear regression model, the synergistic variable set is fused with the ICP-MS ex situ detection dataset to construct a multi-source dataset, the multi-source dataset is input into a sequential Gaussian random simulation network, and multiple independent realizations are generated based on the heavy metal prediction values ​​and prediction error variances. A heavy metal prediction content distribution map and a prediction error variance map are drawn to analyze spatial distribution patterns and hot spots and evaluate local uncertainties.

[0029] In one possible implementation, in step 5, inputting the multi-source dataset into a sequential Gaussian random simulation network, and generating multiple independent realizations based on the heavy metal prediction values ​​and prediction error variances includes:

[0030] The multi-source data sets are input into a sequential Gaussian random simulation network.

[0031] The sequential Gaussian random simulation network defines a random path to visit all grid cells to be simulated.

[0032] In each grid cell to be simulated, the corresponding heavy metal prediction value and prediction error variance are used as the mean and variance of the conditional distribution of the grid cell.

[0033] Assume that the conditional distribution is a Gaussian distribution, randomly extract a value from the Gaussian distribution based on a random seed as the simulated value of the grid cell, and add the simulated value to the conditional data set for subsequent simulation of the grid cell.

[0034] This process continues until all grid cells to be simulated are simulated, completing an independent implementation.

[0035] By varying the random path and random seed, multiple independent realizations are generated.

[0036] The multi-source data fusion method for assessing the risk of heavy metal pollution in orchard soils in this application has the following advantages:

[0037] By combining double-layer dynamic division, fruit tree drip line, judgment fusion conditions, ordinary co-kriging method and sequential Gaussian random simulation network co-simulation, the sample coverage is improved, low cost and high precision are taken into account, and the reliability of risk assessment is improved.

[0038] By using the grid method to divide each basic unit into grid units of equal area according to the distribution of fruit trees in rows and columns, and using the plum blossom method to set several sub-sample points in each grid unit based on the drip line of the fruit trees, a mathematical relationship between grid division and the distribution characteristics and physiological characteristics of fruit trees was established, thereby improving the coverage of sample points.

[0039] When the correlation does not meet the fusion conditions, soil samples are collected at each sub-sample point, and the soil samples are naturally air-dried, passed through a stainless steel sieve, placed in a polyethylene container, pressed into sheets, and then covered with Mylar film. A portable X-ray fluorescence spectrometer is used to perform pXRF ex situ detection on the treated soil samples, eliminating the matrix effect and reducing the error from the data source.

[0040] By constructing a linear regression model, fitting a variation function model, and combining ordinary cokriging and sequential Gaussian random simulation networks for collaborative simulation, the single prediction value is upgraded to a pollution risk probability matrix to support remediation priority decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] Figure 1 A flowchart of a method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion provided in an embodiment of the present application;

[0043] Figure 2 Schematic diagram of basic unit grid division and sub-sample point setting provided in an embodiment of the present application;

[0044] Figure 3 A schematic diagram of the spatial distribution of a multi-source data set constructed by fusion when the research object provided in the embodiment of the present application is one basic unit;

[0045] Figure 4 Schematic diagram of the spatial distribution of a multi-source data set constructed by fusion when the research object provided in the embodiment of the present application is greater than 1 basic unit. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0047] like Figure 1 As shown, the embodiment of the present application provides a method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion, comprising the following steps:

[0048] Step 1: Construct a two-layer dynamic adaptive grid to divide the orchard twice, obtain the basic unit of the initial division and the grid unit of the secondary division, and set several sub-sample points in each grid unit based on the drip line of the fruit tree.

[0049] Step 2: Use a portable X-ray fluorescence spectrometer to perform pXRF in-situ detection on each sub-sample point, calculate the heavy metal pXRF in-situ detection data of each grid unit, and construct a pXRF in-situ detection dataset.

[0050] Step 3: Select a collection unit in each basic unit, use an ICP-MS inductively coupled plasma mass spectrometer to perform ICP-MS ex situ detection on each collection unit, obtain the heavy metal ICP-MS ex situ detection data of each collection unit, and construct an ICP-MS ex situ detection data set.

[0051] Step 4: Determine the fusion condition of the pXRF in-situ detection data set and the ICP-MS ex-situ detection data set based on the correlation between the heavy metal pXRF in-situ detection data and the heavy metal ICP-MS ex-situ detection data in the same acquisition unit. When the correlation meets the fusion condition, proceed to step 5.

[0052] Step 5: Ordinary co-kriging and sequential Gaussian random simulation networks are used to co-simulate the pXRF in-situ detection dataset and the ICP-MS ex-situ detection dataset, and a heavy metal predicted content distribution map and a prediction error variance map are drawn to analyze the spatial distribution pattern and hotspot areas and evaluate local uncertainties.

[0053] The 2014 National Soil Pollution Survey Bulletin shows that the main inorganic pollutants in my country are cadmium (Cd), chromium (Cr), mercury (Hg), arsenic (As), copper (Cu), lead (Pb), zinc (Zn), and nickel (Ni). Due to the limitations of portable X-ray fluorescence spectrometers (LOD > 1 ppm), the detection rates of Cd and Hg are extremely low, and hexavalent chromium cannot be detected. Therefore, the method in this application focuses on Cu, Pb, Zn, Ni, and As in soil.

[0054] Exemplarily, step one includes:

[0055] The orchard was preliminarily divided based on remote sensing images to obtain several basic units.

[0056] According to the distribution of fruit trees in rows and columns, the grid method is used to divide each basic unit into grid units of equal area.

[0057] Based on the drip line of the fruit trees, the plum blossom method was used to set several sub-sample points in each grid unit.

[0058] Specifically, in this embodiment, taking a certain orchard as an example, the geographic information system is used to extract the orchard range from the remote sensing image, and the orchard range is preliminarily divided into several basic units of equal size (each basic unit is about 1 mu); according to the distribution of fruit trees in rows and columns, the grid method is used to divide each basic unit into grid units of the same area. Taking into account the orchard area and planting density, when the orchard area is less than 10 mu, the size of the grid unit should be (3-5) rows × (2-3) columns; when 10 mu ≤ orchard area ≤ 100 mu, the size of the grid unit should be (5-8) rows × (3-5) columns; when the orchard area is greater than 100 mu, the size of the grid unit should be (8-15) rows × (5-10 columns); when the planting density is large (such as the dwarf rootstock dense plantation in Weibei, Shaanxi), the size of the grid unit should be appropriately increased. Figure 2 As shown, Figure 2 The size of the grid cells in is 3 rows × 2 columns.

[0059] In this embodiment, the plum blossom method is used to set 5 sub-sample points in each grid unit, including a sub-sample point S in the center of the grid unit. n5 , the remaining four sub-sample points S n1 、S n2 、S n3 、S n4 (n represents the nth grid unit, 1…n) Located between the two rows of fruit trees in the center of the grid unit, 30-50 cm outside the drip line (can be adjusted according to the crown, age and size of the fruit trees).

[0060] For example, in step 2, the accuracy and stability of pXRF in-situ detection are ensured by combining standard substances with quantitative limit indicators, relative standard deviation indicators, and relative error indicators.

[0061] Specifically, in this embodiment, a portable X-ray fluorescence spectrometer is used to perform pXRF in-situ detection on each sub-sample point. During the pXRF in-situ detection process, by adding standard substances, the quantification limit index, relative standard deviation index, and relative error index of the portable X-ray fluorescence spectrometer are calculated to ensure the accuracy and stability of the pXRF in-situ detection. The heavy metal pXRF in-situ detection data of each grid unit is calculated (in this embodiment, the average value of the heavy metal element i of the five sub-sample points in the grid unit is used). )as follows:

[0062] .

[0063] in, represents the pXRF in-situ detection value of the i-th heavy metal element at the 1st sub-sample point of the n-th grid unit. Construct a pXRF in-situ detection dataset, denoted as ,in, It represents the average value of the pXRF in-situ detection value of the i-th heavy metal element in the n-th grid cell in the q-th basic unit.

[0064] Exemplarily, step three includes:

[0065] For regular basic units, the grid units at the four corners and the center are used as collection units. For irregular basic units, randomly sampled grid units are selected as collection units.

[0066] A mixed soil sample was collected from each collection unit, and the mixed soil sample from each collection unit was tested in the laboratory using an ICP-MS inductively coupled plasma mass spectrometer to obtain the heavy metal ICP-MS ex situ detection data of each collection unit and construct an ICP-MS ex situ detection dataset.

[0067] Specifically, in this embodiment, for the regular basic unit, the grid units at the four corners and the center are used as collection units, such as Figure 2 As shown, Figure 2 The mixed sampling point S in 11 、S 12 、S 13 、S 14 、S 15 That is, the collection unit. Collect 5 soil mixed samples in each collection unit ( ,in, Indicates the A collection unit, Indicates the The first soil mixed sample of each collection unit was collected), and the soil mixed sample of each collection unit was tested in the laboratory using ICP-MS inductively coupled plasma mass spectrometer to obtain the heavy metal ICP-MS ex situ detection data of each collection unit. (Indicates the Heavy metal ICP-MS ex situ detection data for the i-th heavy metal element in each collection unit were collected. The detection methods and related procedures were referenced in the "Soil Environmental Quality Agricultural Land Soil Pollution Risk Control Standard (Trial)" (GB15618-2018). For irregular basic units, five randomly sampled grid cells were selected as collection units. The related sampling methods and procedures were referenced in the "Random Number Generation and Its Application in Product Quality Sampling Inspection" (GB / T 10111-2008).

[0068] In this example, the soil samples from five sub-samples were mixed to represent one collection unit. An ICP-MS ex situ detection dataset was constructed and recorded as ,in, It represents the heavy metal ICP-MS ex situ detection data of the i-th heavy metal element in the 5th collection unit in the q-th basic unit, effectively reducing the error caused by soil spatial heterogeneity.

[0069] When the grid unit is the collection unit, the heavy metal ICP-MS ex situ detection data Represents the true value of the i-th heavy metal element in its unit.

[0070] For example, in step 4, when the correlation does not meet the fusion condition, soil sampling and pXRF ex situ detection are performed on each sub-sample point, and the correlation is optimized until the fusion condition is met.

[0071] Exemplarily, the correlation adopts the R² indicator. When the R² indicator is greater than or equal to a preset threshold, the correlation meets the fusion condition, otherwise it does not meet the condition.

[0072] Exemplarily, in step 4, the soil sampling and pXRF ex situ detection at each sub-sample point includes:

[0073] Soil samples were collected at each sub-sample point, dried naturally, sieved through a stainless steel sieve, placed in a polyethylene container, pressed into sheets, and then covered with Mylar film.

[0074] The treated soil samples were subjected to pXRF ex situ detection using a portable X-ray fluorescence spectrometer.

[0075] Specifically, in this embodiment, the heavy metal pXRF in-situ detection data in the same acquisition unit and heavy metal ICP-MS ex situ detection data The correlation of the pXRF in situ detection data set is determined The ICP-MS ex situ detection dataset When the R² index (Pb, Zn, Cu, Ni) ≥ 0.85 and the R² index (As) ≥ 0.70, it indicates that the pXRF in-situ detection data set ICP-MS ex situ detection dataset The consistency is high, and the subsequent steps of building a linear regression model and fusing and constructing a multi-source data set can be carried out; on the contrary, when the R² index does not meet the fusion conditions, soil samples are collected at each sub-sample point, and the soil samples are naturally air-dried, passed through a stainless steel sieve (2mm), placed in a polyethylene container, pressed into a sheet, and covered with Mylar film. After removing the interference of factors such as sample uniformity, surface roughness, and water content, a portable X-ray fluorescence spectrometer is used to perform pXRF ex situ detection on the treated soil samples, and the heavy metal pXRF in situ detection data are analyzed. and heavy metal ICP-MS ex situ detection data The R² indicator is optimized.

[0076] Exemplarily, step five includes:

[0077] In the acquisition unit, a linear regression model is constructed between the pXRF in-situ detection dataset and the ICP-MS ex-situ detection dataset, with the pXRF in-situ detection dataset serving as the primary variable and the ICP-MS ex-situ detection dataset serving as the auxiliary variable.

[0078] The variograms of the main variable and the auxiliary variable are calculated respectively, and the cross-variogram is calculated, and a variogram model is obtained by fitting. Based on the fitted variogram model, ordinary cokriging method is used to perform interpolation prediction on each grid cell to obtain the heavy metal prediction value and prediction error variance of each grid cell.

[0079] The pXRF in situ detection dataset is converted into a synergistic variable set through the linear regression model, the synergistic variable set is fused with the ICP-MS ex situ detection dataset to construct a multi-source dataset, the multi-source dataset is input into a sequential Gaussian random simulation network, and multiple independent realizations are generated based on the heavy metal prediction values ​​and prediction error variances. A heavy metal prediction content distribution map and a prediction error variance map are drawn to analyze spatial distribution patterns and hot spots and evaluate local uncertainties.

[0080] Specifically, in this embodiment, in the acquisition unit (i.e., there is a pXRF in-situ detection data set and ICP-MS ex situ detection datasets ), constructing the pXRF in-situ detection dataset (Primary variable) and ICP-MS ectopic detection data set The linear regression model between (auxiliary variables) is as follows:

[0081] .

[0082] Among them, a represents the intercept, b represents the slope of the line, and e is the error term.

[0083] The variograms of the primary and auxiliary variables were calculated separately, along with the cross-variograms, and a variogram model was fitted to ensure that the model met positive definite conditions (e.g., linear model coregionalization). Based on the fitted variogram model, ordinary cokriging was used to interpolate and predict the heavy metal (Cu, Pb, Zn, Ni, As) content for each grid cell, yielding the predicted heavy metal value and prediction error variance for each grid cell. The prediction error variance quantifies the uncertainty of the prediction.

[0084] The pXRF in situ detection data set was transformed into Convert to a covariate set , the collaborative variable set ICP-MS ex situ detection dataset Multi-source datasets are fused and constructed as input to a sequential Gaussian random simulation network.

[0085] Specifically, the pXRF in situ detection dataset Convert to a covariate set When , firstly transform the above linear regression model into becomes , the second regression equation is as follows:

[0086] .

[0087] The pXRF in situ detection data set was transformed into Convert to a covariate set .

[0088] In the collaborative variable set ICP-MS ex situ detection dataset When fusing and constructing multi-source datasets, there are two situations:

[0089] When the research object is 1 basic unit, 5 acquisition units use ICP-MS ex situ detection data sets , the remaining grid cells use the transformed cooperative variable set , the spatial distribution is as follows Figure 3 As shown (taking the basic unit q=5, n=5 as an example).

[0090] When the research object is greater than 1 basic unit, and if there is no significant difference in the environment, cultivation management, etc. between adjacent grids of different basic units, the adjacent units only need to be measured once to represent the values ​​of the two adjacent collection units and uniformly construct the ICP-MS ex situ detection data set. , the remaining grid cells use the transformed cooperative variable set , the spatial distribution is as follows Figure 4 As shown (taking 2 basic units as an example).

[0091] Exemplarily, in step 5, inputting the multi-source data set into a sequential Gaussian random simulation network, and generating multiple independent realizations based on the heavy metal prediction values ​​and prediction error variances includes:

[0092] The multi-source data sets are input into a sequential Gaussian random simulation network.

[0093] The sequential Gaussian random simulation network defines a random path to visit all grid cells to be simulated.

[0094] In each grid cell to be simulated, the corresponding heavy metal prediction value and prediction error variance are used as the mean and variance of the conditional distribution of the grid cell.

[0095] Assume that the conditional distribution is a Gaussian distribution, randomly extract a value from the Gaussian distribution based on a random seed as the simulated value of the grid cell, and add the simulated value to the conditional data set for subsequent simulation of the grid cell.

[0096] This process continues until all grid cells to be simulated are simulated, completing an independent implementation.

[0097] By varying the random path and random seed, multiple independent realizations are generated.

[0098] Specifically, in this example, a predicted heavy metal content distribution map is plotted based on the means of multiple independent realizations, and a prediction error variance map is plotted based on the variances of multiple independent realizations. The predicted heavy metal content distribution map is used to analyze spatial distribution patterns and hotspots (e.g., areas with high means), while the prediction error variance map is used to assess local uncertainty (e.g., areas with high variance require additional sampling verification to support remediation priority decisions).

[0099] The embodiment of the present application improves the sample point coverage by combining two-layer dynamic partitioning, fruit tree drip line, judgment fusion conditions, ordinary collaborative Kriging method and sequential Gaussian random simulation network collaborative simulation, takes into account low cost and high precision, and improves the reliability of risk assessment.

[0100] By using the grid method to divide each basic unit into grid units of equal area according to the distribution of fruit trees in rows and columns, and using the plum blossom method to set several sub-sample points in each grid unit based on the drip line of the fruit trees, a mathematical relationship between grid division and the distribution characteristics and physiological characteristics of fruit trees was established, thereby improving the coverage of sample points.

[0101] When the correlation does not meet the fusion conditions, soil samples are collected at each sub-sample point, and the soil samples are naturally air-dried, passed through a stainless steel sieve, placed in a polyethylene container, pressed into sheets, and then covered with Mylar film. A portable X-ray fluorescence spectrometer is used to perform pXRF ex situ detection on the treated soil samples, eliminating the matrix effect and reducing the error from the data source.

[0102] By constructing a linear regression model, fitting a variation function model, and combining ordinary cokriging and sequential Gaussian random simulation networks for collaborative simulation, the single prediction value is upgraded to a pollution risk probability matrix to support remediation priority decision-making.

[0103] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0104] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion, characterized in that: The following steps are involved: Step 1: Construct a two-layer dynamic adaptive grid to divide the orchard twice, obtaining the basic unit of the initial division and the grid unit of the secondary division. Set several sub-sample points in each grid unit based on the drip line of the fruit tree; Step 2: Use a portable X-ray fluorescence spectrometer to perform pXRF in-situ detection on each sub-sample point, calculate the heavy metal pXRF in-situ detection data of each grid unit, and construct a pXRF in-situ detection data set; Step 3: Select a collection unit within each basic unit, perform ICP-MS ex situ detection on each collection unit using an ICP-MS inductively coupled plasma mass spectrometer, obtain heavy metal ICP-MS ex situ detection data for each collection unit, and construct an ICP-MS ex situ detection data set; Step 4: Determine the fusion condition of the pXRF in-situ detection data set and the ICP-MS ex-situ detection data set based on the correlation between the heavy metal pXRF in-situ detection data and the heavy metal ICP-MS ex-situ detection data set in the same acquisition unit. When the correlation meets the fusion condition, proceed to step 5. Step 5: Co-simulate the pXRF in-situ detection dataset and the ICP-MS ex-situ detection dataset using ordinary co-kriging and sequential Gaussian random simulation network to draw a heavy metal predicted content distribution map and a prediction error variance map to analyze the spatial distribution pattern and hotspot areas and assess local uncertainty; Step five includes: In the acquisition unit, a linear regression model is constructed between the pXRF in-situ detection dataset and the ICP-MS ex-situ detection dataset, with the pXRF in-situ detection dataset serving as the primary variable and the ICP-MS ex-situ detection dataset serving as the auxiliary variable; Calculating the variograms of the primary variable and the auxiliary variable respectively, and calculating the cross-variogram, fitting a variogram model, and using ordinary cokriging to perform interpolation prediction on each grid cell based on the fitted variogram model to obtain the predicted value of heavy metals and the prediction error variance of each grid cell; The pXRF in situ detection dataset is converted into a synergistic variable set through the linear regression model, the synergistic variable set is fused with the ICP-MS ex situ detection dataset to construct a multi-source dataset, the multi-source dataset is input into a sequential Gaussian random simulation network, and multiple independent realizations are generated based on the heavy metal prediction values ​​and prediction error variances. A heavy metal prediction content distribution map and a prediction error variance map are drawn to analyze spatial distribution patterns and hot spots and evaluate local uncertainties.

2. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 1 is characterized in that: Step one includes: The orchard was preliminarily divided based on remote sensing images to obtain several basic units; According to the distribution of fruit trees in rows and columns, each basic unit is divided into grid units of equal area using the grid method; Based on the drip line of the fruit trees, the plum blossom method was used to set several sub-sample points in each grid unit.

3. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 1 is characterized in that: In step 2, the accuracy and stability of pXRF in-situ detection are ensured by combining standard substances with quantitative limit indicators, relative standard deviation indicators, and relative error indicators.

4. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 1, characterized in that: Step three includes: For regular basic units, the grid cells at the four corners and the center are used as collection units. For irregular basic units, randomly sampled grid cells are selected as collection units. A mixed soil sample was collected from each collection unit, and the mixed soil sample from each collection unit was tested in the laboratory using an ICP-MS inductively coupled plasma mass spectrometer to obtain the heavy metal ICP-MS ex situ detection data for each collection unit and construct an ICP-MS ex situ detection dataset.

5. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 1 is characterized in that: In step 4, when the correlation does not meet the fusion conditions, soil sampling and pXRF ex situ detection are performed on each sub-sample point, and the correlation is optimized until the fusion conditions are met.

6. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 5 is characterized in that: The correlation adopts the R² indicator. When the R² indicator is greater than or equal to a preset threshold, the correlation meets the fusion condition, otherwise it does not meet the condition.

7. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 5, characterized in that: In step 4, the soil sampling and pXRF ex situ testing of each sub-sample point includes: Soil samples were collected at each sub-sampling point, dried naturally, passed through a stainless steel sieve, placed in a polyethylene container, pressed into sheets, and then covered with Mylar film. The treated soil samples were subjected to pXRF ex situ detection using a portable X-ray fluorescence spectrometer.

8. The method for assessing the risk of heavy metal pollution in orchard soil by multi-source data fusion according to claim 1, characterized in that: In step 5, the multi-source data set is input into a sequential Gaussian random simulation network, and multiple independent realizations are generated based on the heavy metal prediction values ​​and prediction error variances, including: inputting the multi-source data set into a sequential Gaussian random simulation network; The sequential Gaussian random simulation network defines a random path to visit all grid cells to be simulated; In each grid cell to be simulated, the corresponding heavy metal prediction value and prediction error variance are used as the mean and variance of the conditional distribution of the grid cell; Assume that the conditional distribution is a Gaussian distribution, randomly extract a value from the Gaussian distribution based on a random seed as the simulated value of the grid cell, add the simulated value to the conditional data set, and use it for subsequent simulations of the grid cells; Until all grid cells to be simulated are simulated, an independent realization is completed; By varying the random path and random seed, multiple independent realizations are generated.

Images