CN109001127A - A kind of heavy metal content in soil space predicting method - Google Patents
A kind of heavy metal content in soil space predicting method Download PDFInfo
- Publication number
- CN109001127A CN109001127A CN201810557158.7A CN201810557158A CN109001127A CN 109001127 A CN109001127 A CN 109001127A CN 201810557158 A CN201810557158 A CN 201810557158A CN 109001127 A CN109001127 A CN 109001127A
- Authority
- CN
- China
- Prior art keywords
- content
- heavy metal
- soil
- auxiliary variable
- sampled point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/3103—Atomic absorption analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
- G01N21/64—Fluorescence; Phosphorescence
- G01N21/6402—Atomic fluorescence; Laser induced fluorescence
- G01N21/6404—Atomic fluorescence
Abstract
The present invention provides a kind of heavy metal content in soil space predicting methods, research area's heavy metal content in soil spatial prediction is carried out by the auxiliary variable including Hyperspectral imaging, in conjunction with the nonlinear prediction feature of artificial neural network and the linear prediction feature of Kriging method, the heavy metal content in soil spatial distribution of research area's entirety is preferably predicted by sampled point, compared with existing prediction technique, prediction result is more close to measured value, it can preferably explain the Spatial Variability of heavy metal-polluted soil and its linear and non-linear relation with auxiliary variable, with good practicability.
Description
Technical field
The present invention relates to land-use study fields, and in particular to a kind of heavy metal content in soil space predicting method.
Background technique
Currently, the spatial distribution drawing of farmland soil heavy metals is typically based on the actual measurement content value of field sampling acquisition, hold
Easily cause agricultural land soil destruction and high cost, it is also difficult to meet space and rapidly and efficiently obtain wanting for heavy metal-polluted soil spatial information
It asks.Therefore, using effective auxiliary variable for farmland soil heavy metals fine space mapping, save field sampling cost and
Ecological agriculture sustainable development is of great significance.
Heavy metal-polluted soil Spatial Variability is often the direct or indirect effect of Various Complex relation factor.Environmental factor is
It is most commonly used to the factors such as the auxiliary variable of soil attribute spatial prediction, including landform, weather and vegetation.Due to their non-demolition
Property and easily obtain, environmental factor is widely used in the prediction soil organism, soil organic matter and heavy metal-polluted soil.Environmental factor is logical
Crossing influences geographical conditions relevant to heavy metal-polluted soil reflects its Spatial Variability indirectly.For example, landform and the gradient usually influence
Soil thickness influences the content of organic matter adhered to by heavy metal in turn；Temperature and precipitation usually influence atmospheric sedimentation, soil acid-base
Then degree changes the migration conversion process of heavy metal-polluted soil further through the absorption of crop root.
In recent years, bloom spectroscopic factor is found and heavy metal-polluted soil close relation.EO-1 hyperion variable and a soil huge sum of money
Relationship mechanism between category is the spectrum change by capturing the soil organism adhered to by heavy metal and ferrimanganic compound, passes through light
Reflectivity changes are composed to reflect the changes of contents of heavy metal.On the one hand target in hyperspectral remotely sensed image has fine spectral resolution,
The faint variation of heavy metal content in soil can be perceived；On the other hand there is spatial continuity, sky preferably can be carried out to it
Between variability simulate.Therefore, it is pre- to carry out farmland soil heavy metals Spatial Variability for the operation environment factor and target in hyperspectral remotely sensed image
It surveys, precision of prediction and efficiency can be effectively provided.However, being used simultaneously during current heavy metal-polluted soil spatial prediction
The multi-source auxiliary variable of environment and Hyperspectral imaging is also less.
In the heavy metal-polluted soil space predicting method of auxiliary variable, multiple linear regression returns Ordinary Kriging Interpolation and artificial
Neural network is using more.However, these methods are linear between heavy metal-polluted soil and auxiliary variable by solely portraying
Or non-linear relation, and then the Spatial Variability of Prediction of Soil Heavy Metal.In fact, past between heavy metal-polluted soil and auxiliary variable
Toward linear and non-linear relation is existed simultaneously, it can reflect that complex relationship between the two is to improve heavy metal-polluted soil space simultaneously
The key of precision of prediction.
Artificial neural network Ordinary Kriging Interpolation model is to combine artificial nerve network model with Ordinary Kriging Interpolation model
Mixedly statistical method, this method on the one hand portrayed between heavy metal-polluted soil and multi-source auxiliary variable with artificial neural network
Non-linear relation, on the other hand the residual error of artificial nerve network model is linearly inserted further through Ordinary Kriging Interpolation model
Value also rarely has research with the heavy metal-polluted soil Spatial Variability prediction of this method combination multi-source auxiliary variable at present.
Summary of the invention
In order to overcome the defect of the above method, the present invention provides a kind of heavy metal content in soil space predicting methods, lead to
The auxiliary variable crossed including Hyperspectral imaging carries out research area's heavy metal content in soil spatial prediction, in conjunction with artificial neural network
It is whole preferably to predict research area by sampled point for the nonlinear prediction feature of network and the linear prediction feature of Kriging method
Heavy metal content in soil spatial distribution, compared with existing prediction technique, prediction result, can be more preferable more close to measured value
Explanation heavy metal-polluted soil Spatial Variability and its linear and non-linear relation with auxiliary variable, have good practical
Property.
Correspondingly, the present invention provides a kind of heavy metal content in soil space predicting methods, comprising the following steps:
Determine research area；
Pedotheque is acquired in the research area；
Measure the pedotheque content of beary metal；
It determines auxiliary variable and obtains the auxiliary variable information in the research area；
The auxiliary variable is screened based on Pearson correlation coefficients；
Based on principal component analysis to the auxiliary variable dimensionality reduction；
Based on artificial neural network-Ordinary Kriging Interpolation model and the auxiliary variable after dimensionality reduction, the research area is carried out
Heavy metal content in soil spatial prediction；
Export content of beary metal spatial prediction figure.
Preferred embodiment, it is described in the research area acquire pedotheque the following steps are included:
In the research area in exposed soil, sampled point is determined based on stratified random smapling method；
Five parts of initial soil samples are collected altogether in four corners of each sampled point X-shaped and the sampled point, are taken
Sample depth is 0~20cm；
Five parts of initial soil samples of the sampled point are mixed, the impurity in addition to removing soil, at room temperature certainly
It is screened after so air-drying using 100 meshes；
Take initial soil sample pedotheque as the sampled point of the 100g after screening.
Preferred embodiment, the measurement pedotheque content of beary metal the following steps are included:
Heavy metal arsenic As content is measured based on reduction and gaseous-atomic fluorescence spectrophotometer method；
Heavy metal zinc Zn and chromium Cr content are measured based on atomic absorption spectrophotometry；
Based on graphite furnace atomic absorption spectrometry heavy metal lead Pb and cadmium Cd content.
Preferred embodiment, the auxiliary variable include four classifications, respectively terrain information, vegetation information, weather
Information and high-spectrum remote-sensing information.
Preferred embodiment, the terrain information include elevation, slope aspect, slope aspect variability, the gradient, gradient variability, confluence
Cumulant, topographic relief amplitude, roughness of ground surface, fills out hollow and depression depth totally 11 auxiliary variables at catchment area；
The vegetation information includes enhancing vegetation index and normalized difference vegetation index totally 2 auxiliary variables；
The climatic information includes monthly mean temperature, the monthly highest temperature, monthly mean rainfall and solar radiation quantity totally 4
Auxiliary variable；
High-spectrum remote-sensing information includes Hyperspectral imaging totally 1 auxiliary variable.
Preferred embodiment, it is described based on Pearson correlation coefficients screen the auxiliary variable the following steps are included:
The Pearson correlation coefficients of each auxiliary variable Yu the various content of beary metal are calculated separately, calculation formula is X
For auxiliary variable, Y is content of beary metal, covariance of the cov (X, Y) between X and Y, σXσYFor multiplying for X standard deviation and Y standard deviation
Product；
Retain the auxiliary variable that absolute value is greater than 0.4.
Preferred embodiment, it is described based on principal component analysis to the auxiliary variable dimensionality reduction the following steps are included:
Based on the auxiliary variable and the auxiliary variable information, correlation matrix is calculated；
Based on the correlation matrix, the eigen vector of the correlation matrix is calculated；
Successively calculate principal component contributor rate, contribution rate of accumulative total and principal component load；
Auxiliary variable based on the principal component load, after obtaining dimensionality reduction.
Preferred embodiment carries out soil weight to the research area based on artificial neural network-Ordinary Kriging Interpolation model
Tenor spatial prediction the following steps are included:
The content of beary metal of auxiliary variable and sampled point based on sampled point after dimensionality reduction constructs artificial neural network mould
Type；
Auxiliary variable of the sampled point after dimensionality reduction is imported into the artificial nerve network model, is obtained based on the artificial mind
The prediction content of beary metal of the sampled point through network model；
Calculate the residual values of sampling point prediction content of beary metal and content of beary metal；
Based on normal stabilizing pile, the prediction residual of all positions in research area is obtained by the residual values of the sampled point
Value；
Auxiliary variable of the non-sampled point after dimensionality reduction is imported into the artificial nerve network model, is obtained based on described artificial
The prediction content of beary metal of the non-sampled point of neural network model；
The heavy metal content in soil of the non-sampled point be corresponding position prediction content of beary metal and corresponding position it is pre-
Survey the sum of residual values.
Preferred embodiment, the export content of beary metal spatial prediction figure the following steps are included:
The content of beary metal spatial prediction figure is exported based on ArcGIS software.
Heavy metal content in soil space predicting method provided by the invention is become by the auxiliary including Hyperspectral imaging
Amount carries out research area's heavy metal content in soil spatial prediction, nonlinear prediction feature and Krieger side in conjunction with artificial neural network
The linear prediction feature of method preferably predicts the heavy metal content in soil spatial distribution of research area's entirety by sampled point, with
Existing prediction technique is compared, and prediction result more close to measured value, can preferably explain the spatial variability of heavy metal-polluted soil
Property and its linear and non-linear relation with auxiliary variable, have good practicability.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
Other attached drawings are obtained according to these attached drawings.
Fig. 1 shows the method flow diagram of the heavy metal content in soil space predicting method of the embodiment of the present invention；
Fig. 2 shows the Hyperspectral imagings of the embodiment of the present invention；
Fig. 3 shows SPSS software principal component analysis result figure of the embodiment of the present invention；
Fig. 4 shows the structure chart of ANN model of the embodiment of the present invention；
Fig. 5 shows the heavy metal-polluted soil spatial prediction figure of the embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts all other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1 shows the method flow diagram of the heavy metal content in soil space predicting method of the embodiment of the present invention, the present invention
The heavy metal content in soil space predicting method of embodiment the following steps are included:
S101: research area is determined；
In general, since industrial area human factor is affected, content of beary metal prediction mainly with mankind's activity region
It is associated.Therefore, the heavy metal content in soil space predicting method of the embodiment of the present invention is mainly used in farming region, main to select
It is farming region as research area, the research area geographical coordinate that the embodiment of the present invention selects is 23 ° 5 ' -23 ° 37 ' N and 113 ° 29 ' -
114 ° of 0 ' E, occupied area 1616.47km2, study area's area about 445km2, wherein 35.4% or more is Plain.The research area
Belong to the maritime monsoon climate of south subtropics, 21.8 DEG C of average temperature of the whole year, annual sunshine 1785.5h, annual relative humidity
79%, mean annual precipitation 1994.5mm, for geomorphic type mainly based on Plain and hills, highest height above sea level is 1084.3m.
According to China soil classification, red soil and red earth can be divided by studying area's agricultural land soil type, and research area is the Zhujiang River three
The typical farming region in angle continent, its traditional pattern of farming are a kind of Rice Production systems for not only having produced non-vegetable crop but also produce vegetables
System.
S102: pedotheque is acquired in the research area；
In general, sample acquisition selection is in most dry month in order to accurately obtain the auxiliary variable information of soil
It carries out, to ensure that farmland is exposed soil after crop harvesting；The case where based on research area, the crop harvesting time is generally
September~October, therefore, sampling time are typically chosen in October~December, to ensure that soil is exposed when sampling, to make height
Spectral remote sensing information is more accurate.
The acquisition of sample uses stratified random smapling method, in the gradient, manure pit and the tomb for fully considering ditch area and barrier
After the influence on ground, centered on sampled point, five soil samples are collected altogether in four corners of X-shaped and sampled point, are sampled
Depth is 0~20cm, nearly weighs 1 kilogram.
Specifically, being referred to centered on sampled point in four corners of X-shaped, centered on sampled point with pre-set radius
Circle is done, 4 Along ents are taken on circle.
Five parts of soil samples of each sampled point are thoroughly mixed, remove plant residue and calculus, at room temperature natural wind
Dry doubling is screened using 100 mesh 0.15mm.Finally, the soil for extracting about 100g carries out determining heavy metals, in general, mainly
Common heavy metal arsenic As, cadmium Cd, chromium Cr, lead Pb and zinc Zn are measured.Wherein, arsenic As is glimmering using reduction and gaseous-atom
The measurement of light photometer measuring method, zinc Zn and chromium Cr are measured using atomic absorption spectrophotometry, and lead Pb and cadmium Cd use graphite furnace
Aas determination.
S103: auxiliary variable information is collected；
In order to study the heavy metal content in soil in research area and the relationship between auxiliary variable information, implement in the present invention
In example, it is also necessary to collect auxiliary variable information.
The auxiliary variable information of the embodiment of the present invention includes terrain information, vegetation information, climatic information and high-spectrum remote-sensing
Four classifications of information.
Wherein, terrain information derives from landform digital complex demodulation, and spatial resolution is 30 meters, in general, domestic
Number from Computer Network Information Center, Chinese Academy of Sciences's International Technology data image station it is believed that obtain；Utilize ArcGIS 10.3
Dem data is handled, obtain elevation, slope aspect, slope aspect variability, the gradient, gradient variability, confluence cumulant, catchment area,
Shape waviness, roughness of ground surface fill out hollow and depression depth totally 11 auxiliary variables.
Vegetation information derives from the MODIS image in same sampling time, therefrom obtains enhancing vegetation index EVI and normalization
Difference vegetation index NDVI totally 2 auxiliary variables.
Climatic select monthly mean temperature, the monthly highest temperature, monthly mean rainfall and solar radiation quantity totally 4 it is auxiliary
Help variable.
Fig. 2 shows the Hyperspectral imagings of the embodiment of the present invention.In embodiments of the present invention, comprehensively consider heavy metal-polluted soil
The factors such as cloud amount, the soil moisture content in month are acquired, high-spectrum remote-sensing information selects Chinese environmental 1A satellite in October, 2008
Hyperspectral imaging carries out radiant correction, atmospheric correction to Hyperspectral imaging using the analysis tool in 5.2 software of ESRI ENVI
FLAASH and mixed pixel decompose.
Finally, original auxiliary variable quantity totally 18 are obtained.For unified resolution, keep each auxiliary variable better
Carry out fusion treatment, using the arest neighbors distribution method in ArcGIS software, by all resamplings of all auxiliary variables be 100 meters ×
100 meters of spatial resolution.
S104: auxiliary variable screening and dimension-reduction treatment；
In different research areas or different time sections, the auxiliary variable information established in not all step S103 is all
There is correlation with heavy metal content in soil, therefore, it is necessary to carry out preliminary screening to auxiliary variable information, exclude certain and soil
Content of beary metal has the auxiliary variable information of weak significance.
Correlation between two variables can be measured with many statistical values, and the most commonly used is Pearson correlation coefficients, Pierres
Gloomy correlation coefficient ρ is defined as the ratio of two covariances and standard deviation product between variable X, Y, and formula isρX,YValue is between -1 to 1.
Pearson correlation coefficients reflect the degree of strength of the linear dependence of two variables, ρX,YThe bigger theory of absolute value
Bright correlation is stronger.Work as ρX,YWhen > 0, show that two variables are positively correlated, i.e. more big then another variate-value of a variate-value also can
It is bigger；Work as ρX,YWhen < 0, show two variable negative correlation, i.e. more big then another variate-value of a variate-value instead can be smaller；When
ρX,YWhen=0, show that two variables are not linearly related (paying attention to being nonlinear correlation), but there may be other modes
Correlation；Work as ρX,YWhen=1 and -1, it is meant that two variable Xs and Y can be very good to be described by linear equation, all samples
Point is all fallen point-blank well.
In general, working as ρX,YAbsolute value less than 0.4 when, the correlation between variable X and Y is weaker, therefore, in general, auxiliary
Help the ρ between variable information and content of beary metalX,YAbsolute value less than 0.4, then give up the auxiliary variable.
In general, there is also more multinomial auxiliary variable, the item number of auxiliary variable will increase too much to be ground after preliminary screening
The complexity studied carefully, therefore, it is necessary to carry out dimensionality reduction to auxiliary variable.
In view of being that there is certain correlativity in many situations, between auxiliary variable, when two or more auxiliary
When helping between variable with certain correlativity, it can be understood as two or more auxiliary variables are to reaction content of beary metal letter
Breath is that have certain plyability, i.e., the content of beary metal information that two or more auxiliary variables are reflected is part or all of
It is consistent；Therefore, the embodiment of the present invention need to carry out dimensionality reduction to auxiliary variable with principal component analytical method.
Principal component analysis is comprehensive at less by a fairly large number of original auxiliary variables premised on least information loss
Auxiliary variable is synthesized, the number for synthesizing auxiliary variable is made to be less than the number of original auxiliary variables；It is general to synthesize auxiliary variable itself
Only there is meaning numerically, for reflecting the most information of original auxiliary variables (single or multiple), and each synthesis is auxiliary
Help it is irrelevant between variable, can effectively solve variable information overlapping, multicollinearity the problems such as.
The basic thought of principal component analysis is to try by the original more index with certain correlation, is reassembled into
The irrelevant overall target of one group of less number replaces original index.
Assuming that there is n sample, each sample shares p variable, constitutes the matrix number X of n × p rank
Remember that former variable index is X1,X2,…,XP, the overall target after dimension-reduction treatment is Z1,Z2,…,Zm(m≤p), then
Coefficient lijIt is calculated based on the following, ZiAnd Zj(i≠j；I, j=1,2 ..., m) it is independent of each other, i.e., finally
Synthesis auxiliary variable between be independent of each other；Secondly, Z1It is X1,X2,…,XpAll linear combinations in variance the maximum, Z2Be with
Z1Incoherent X1,X2,…,XpAll leisurely moods combinations in variance the maximum, ZmIt is and Z1,Z2,…,Zm-1It is all incoherent
X1,X2,…,XpAll linear combinations in variance the maximum；Finally obtain new variables index Z1,Z2,…,ZmIt is referred to as former change
Figureofmerit X1,X2,…,XpFirst, second ..., m principal component.
From above analysis as can be seen that referring to for principal component analysis exactly determines original variable XjIn principal component ZiOn
Load lij, load lijIt is characteristic quantity corresponding to m biggish characteristic values of correlation matrix respectively.
Steps are as follows for the calculating of principal component analysis:
Calculate correlation matrix
Wherein rij(i, j=1,2 ..., p) it is former variable XiAnd XjRelated coefficient, rij=rji, calculation formula is
Eigen vector is calculated, characteristic equation is solved | λ I-R |=0, common Jacobi method finds out characteristic value and by big
It is small to be ordered as λ1≥λ2≥…≥λp≥0；It is found out respectively corresponding to eigenvalue λiFeature vector ei(i=1,2, L, p), it is desirable that | |
ei| |=1, i.e.,Wherein, eijIndicate vector eiJ component；
Calculate principal component contributor rate and contribution rate of accumulative total；
Contribution rate:
Contribution rate of accumulative total:Generally take the characteristic value of contribution rate of accumulative total big 85%~95%, λ1,
λ2,L,λmIt is corresponding be respectively first, second ..., m (m≤p) a principal component；
Calculate principal component loadBack substitution finds out the synthesis after dimension-reduction treatment
Index Z (i.e. after dimensionality reduction).
Fig. 3 shows SPSS software principal component analysis result figure of the embodiment of the present invention.Nowadays, it is calculated to simplify, mainly
Principal component analysis is carried out using SPSS software, the operating procedure of SPSS software is not described in detail in the present embodiment, utilizes SPSS
Derived principal component characteristic value is as shown in Fig. 2, due to color is shown, in the same group of histogram of PC1~PC8, from a left side
It turns right and is followed successively by As, Cd, Cr, Pb, Zn；In the same group of histogram of PC9 and PC10, it is followed successively by Cd, Cr, Pb, Zn from left to right.
Preceding 8 principal component characteristic values of heavy metal Cd, preceding 10 principal component characteristic values of Cr, Pb and Zn and As are greater than 1,
And explain 90% information above of original auxiliary variables；Therefore, by preceding the 8 of preceding 10 principal components of Cd, Cr, Pb and Zn and As
A principal component is used as the input number of ANN (artificial neural network) part of ANNOK (artificial neural network-Ordinary Kriging Interpolation model)
According to.
It should be noted that the principal component its general that principal component analysis obtains is with meaning numerically and and original auxiliary
Helping variable only, there is the transformation relation between formula meaning representated by each principal component is not therefore discussed in detail above.
S105: the spatial prediction of heavy metal-polluted soil is carried out based on artificial neural network-Ordinary Kriging Interpolation model；
In general, being based on auxiliary variable used in the embodiment of the present invention, the relationship between each auxiliary variable and heavy metal can
Expression are as follows: yi,j=f (xi,j,t,xi,j,c,xi,j,v,xi,j,h)；Wherein, yi,jIt is the content value of ith sample point jth heavy metal species,
xi,j,tIt is the corresponding t kind crust deformation magnitude of ith sample point jth heavy metal species, xi,j,cIt is an ith sample point jth kind huge sum of money
Belong to corresponding c kind Climatic value, xi,j,vIt is the corresponding v kind vegetation variate-value of ith sample point jth heavy metal species,
xi,j,hIt is the corresponding h kind EO-1 hyperion variate-value of ith sample point jth heavy metal species, f represents different type auxiliary variable and soil
The prediction process of functional relation between earth heavy metal, i.e. ANNOK (artificial neural network-Ordinary Kriging Interpolation model), hereafter can
It describes in detail.
Fig. 4 shows the structure chart of ANN model of the embodiment of the present invention.Steps are as follows for the calculating of specific ANNOK model:
Firstly, it is necessary to construct ANN model；ANN model belongs to "black box" model, including input layer, hidden layer and output layer,
Input layer is that principal component (i.e. auxiliary variable) and sample point coordinate, output layer of the sampled point after step S104 dimensionality reduction are sampling point
Content of beary metal value hidden layer structure is obtained by input layer and output layer.
Then, by after step S104 dimensionality reduction principal component and sample point coordinate back substitution to the ANN model, obtain logical
Cross the predicted value z' of ANN modelANN(xi,j)；Specifically, the pre- of ANN model can be realized by programming in matlab2013 software
Survey process；
Then, the residual values r of the practical content of beary metal of sampled point and ANN model prediction content of beary metal is calculatedANN(xi,j),
Its formula is as follows: rANN(xi,j)=z (xi,j)-z′ANN(xi,j), wherein z (xi,j) it is that ith sample point jth heavy metal species are adopted
Measured value；
Then, then the residual error Ordinary Kriging Interpolation value of ANN model is calculatedIts formula is as follows:
Wherein, λi,jIt is the weight in residual error Ordinary Kriging Interpolation calculating process, passes through the Geostatistics analysis mould of ArcGIS10.3 software
Block automatically obtains；Residual error Ordinary Kriging Interpolation valueIt is according to the residual error of known sampled point, based on ordinary Kriging
Calculate the residual values of each coordinate in survey region；
Then, the principal component by non-sampled point after step S104 dimensionality reduction and sample point coordinate input ANN model, obtain
The prediction content of beary metal value of the ANN model of non-sampled point；
Finally, calculating the predicted value of ANNOK modelIts predictor formula is as follows:
Wherein,It is the best predictor studied i-th of coordinate jth heavy metal species in area and use ANNOK model.
z'ANN(xi,j) it is estimated value of the ANN model to i-th of coordinate jth heavy metal species,It is i-th of coordinate jth kind huge sum of money
The residual error Ordinary Kriging Interpolation value of category.By to research area in all coordinate points traverse, realize each coordinate points based on
The content of beary metal of ANNOK model is predicted.
S105: export heavy metal-polluted soil spatial prediction figure；
Fig. 5 shows the heavy metal-polluted soil spatial prediction figure of the embodiment of the present invention.Based on data are calculated, each soil is exported respectively
Earth heavy metal spatial distribution prognostic chart, since picture color limits, the specific color of partial content can not be shown, only for signal
With reference to.
Heavy metal content in soil space predicting method provided in an embodiment of the present invention, by including Hyperspectral imaging
Auxiliary variable carry out research area's heavy metal content in soil spatial prediction, in conjunction with artificial neural network nonlinear prediction feature with gram
The linear prediction feature of league (unit of length) method preferably predicts the heavy metal content in soil space point of research area's entirety by sampled point
Cloth, compared with existing prediction technique, prediction result more close to measured value, can preferably explain the space of heavy metal-polluted soil
Variability and its linear and non-linear relation with auxiliary variable have good practicability.
It is provided for the embodiments of the invention a kind of heavy metal content in soil space predicting method above and has carried out detailed Jie
It continues, used herein a specific example illustrates the principle and implementation of the invention, and the explanation of above embodiments is only
It is to be used to help understand method and its core concept of the invention；At the same time, for those skilled in the art, according to this hair
Bright thought, there will be changes in the specific implementation manner and application range, in conclusion the content of the present specification should not manage
Solution is limitation of the present invention.
Claims (9)
1. a kind of heavy metal content in soil space predicting method, which comprises the following steps:
Determine research area；
Pedotheque is acquired in the research area；
Measure the pedotheque content of beary metal；
It determines auxiliary variable and obtains the auxiliary variable information in the research area；
The auxiliary variable is screened based on Pearson correlation coefficients；
Based on principal component analysis to the auxiliary variable dimensionality reduction；
Based on artificial neural network-Ordinary Kriging Interpolation model and the auxiliary variable after dimensionality reduction, soil is carried out to the research area
Content of beary metal spatial prediction；
Export content of beary metal spatial prediction figure.
2. heavy metal content in soil space predicting method as described in claim 1, which is characterized in that acquired in the research area
Pedotheque the following steps are included:
In the research area in exposed soil, sampled point is determined based on stratified random smapling method；
Five parts of initial soil samples are collected altogether in four corners of each sampled point X-shaped and the sampled point, and sampling is deep
Degree is 0~20cm；
Five parts of initial soil samples of the sampled point are mixed, the impurity in addition to removing soil, at room temperature natural wind
It is screened after dry using 100 meshes；
Take initial soil sample pedotheque as the sampled point of the 100g after screening.
3. heavy metal content in soil space predicting method as claimed in claim 2, which is characterized in that the measurement soil
Sample content of beary metal the following steps are included:
Heavy metal arsenic As content is measured based on reduction and gaseous-atomic fluorescence spectrophotometer method；
Heavy metal zinc Zn and chromium Cr content are measured based on atomic absorption spectrophotometry；
Based on graphite furnace atomic absorption spectrometry heavy metal lead Pb and cadmium Cd content.
4. heavy metal content in soil space predicting method as claimed in claim 3, which is characterized in that the auxiliary variable includes
Four classifications, respectively terrain information, vegetation information, climatic information and high-spectrum remote-sensing information.
5. heavy metal content in soil space predicting method as claimed in claim 4, which is characterized in that
The terrain information includes elevation, slope aspect, slope aspect variability, the gradient, gradient variability, confluence cumulant, catchment area, landform
Waviness, roughness of ground surface fill out hollow and depression depth totally 11 auxiliary variables；
The vegetation information includes enhancing vegetation index and normalized difference vegetation index totally 2 auxiliary variables；
The climatic information includes monthly mean temperature, the monthly highest temperature, monthly mean rainfall and solar radiation quantity totally 4 auxiliary
Variable；
High-spectrum remote-sensing information includes Hyperspectral imaging totally 1 auxiliary variable.
6. heavy metal content in soil space predicting method as claimed in claim 5, which is characterized in that described to be based on Pearson's phase
Relationship number sieve select the auxiliary variable the following steps are included:
Calculate separately the Pearson correlation coefficients ρ of each auxiliary variable Yu the various content of beary metalX,Y, calculation formula isX is auxiliary variable, and Y is content of beary metal, covariance of the cov (X, Y) between X and Y, σXσYFor X mark
The product of quasi- difference and Y standard deviation；
Retain ρX,YAbsolute value is greater than 0.4 auxiliary variable.
7. heavy metal content in soil space predicting method as claimed in claim 6, which is characterized in that described based on principal component point
Analysis to the auxiliary variable dimensionality reduction the following steps are included:
Based on the auxiliary variable and the auxiliary variable information, correlation matrix is calculated；
Based on the correlation matrix, the eigen vector of the correlation matrix is calculated；
Successively calculate principal component contributor rate, contribution rate of accumulative total and principal component load；
Auxiliary variable based on the principal component load, after obtaining dimensionality reduction.
8. heavy metal content in soil space predicting method as claimed in claim 7, which is characterized in that be based on artificial neural network
Network-Ordinary Kriging Interpolation model to the research area carry out heavy metal content in soil spatial prediction the following steps are included:
The content of beary metal of auxiliary variable and sampled point based on sampled point after dimensionality reduction constructs artificial nerve network model；
Auxiliary variable of the sampled point after dimensionality reduction is imported into the artificial nerve network model, is obtained based on the artificial neural network
The prediction content of beary metal of the sampled point of network model；
Calculate the residual values of sampling point prediction content of beary metal and content of beary metal；
Based on normal stabilizing pile, the prediction residual value of all positions in research area is obtained by the residual values of the sampled point；
Auxiliary variable of the non-sampled point after dimensionality reduction is imported into the artificial nerve network model, is obtained based on the artificial neuron
The prediction content of beary metal of the non-sampled point of network model；
The heavy metal content in soil spatial prediction of the non-sampled point is the prediction content of beary metal and corresponding position of corresponding position
The sum of prediction residual value.
9. heavy metal content in soil space predicting method as claimed in claim 8, which is characterized in that the export heavy metal contains
Quantity space prognostic chart the following steps are included:
The content of beary metal spatial prediction figure is exported based on ArcGIS software.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810557158.7A CN109001127A (en) | 2018-06-01 | 2018-06-01 | A kind of heavy metal content in soil space predicting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810557158.7A CN109001127A (en) | 2018-06-01 | 2018-06-01 | A kind of heavy metal content in soil space predicting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109001127A true CN109001127A (en) | 2018-12-14 |
Family
ID=64573266
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810557158.7A Pending CN109001127A (en) | 2018-06-01 | 2018-06-01 | A kind of heavy metal content in soil space predicting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109001127A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113008806A (en) * | 2021-03-02 | 2021-06-22 | 农业农村部环境保护科研监测所 | Agricultural product production area heavy metal spatial distribution determination method |
| CN113609940A (en) * | 2021-07-26 | 2021-11-05 | 南通智能感知研究院 | Wide-coverage soil heavy metal remote sensing mixed mapping method |
| WO2021226976A1 (en) * | 2020-05-15 | 2021-11-18 | 安徽中科智能感知产业技术研究院有限责任公司 | Soil available nutrient inversion method based on deep neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529189B (en) * | 2013-06-28 | 2016-01-06 | 四川农业大学 | A kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable |
| CN107860889A (en) * | 2017-09-22 | 2018-03-30 | 华南农业大学 | The Forecasting Methodology and equipment of the soil organism |
| CN108088981A (en) * | 2017-12-13 | 2018-05-29 | 安徽大学 | A kind of soil sulphur element constituent content Forecasting Methodology based on collocating kriging interpolation method |
-
2018
- 2018-06-01 CN CN201810557158.7A patent/CN109001127A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529189B (en) * | 2013-06-28 | 2016-01-06 | 四川农业大学 | A kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable |
| CN107860889A (en) * | 2017-09-22 | 2018-03-30 | 华南农业大学 | The Forecasting Methodology and equipment of the soil organism |
| CN108088981A (en) * | 2017-12-13 | 2018-05-29 | 安徽大学 | A kind of soil sulphur element constituent content Forecasting Methodology based on collocating kriging interpolation method |
Non-Patent Citations (3)
| Title |
|---|
| 张新: "平原区土壤重金属镉空间分布预测方法研究", 《工程科技Ⅰ辑》 * |
| 张明芝 等: "《实用医学统计学与SAS应用》", 31 August 2015, 苏州大学出版社 * |
| 章清 等: "基于主成分分析的协同克里格插值模型对土壤铜含量的空间分布预测", 《华中农业大学学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021226976A1 (en) * | 2020-05-15 | 2021-11-18 | 安徽中科智能感知产业技术研究院有限责任公司 | Soil available nutrient inversion method based on deep neural network |
| CN113008806A (en) * | 2021-03-02 | 2021-06-22 | 农业农村部环境保护科研监测所 | Agricultural product production area heavy metal spatial distribution determination method |
| CN113609940A (en) * | 2021-07-26 | 2021-11-05 | 南通智能感知研究院 | Wide-coverage soil heavy metal remote sensing mixed mapping method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zeng et al. | A review of vegetation phenological metrics extraction using time-series, multispectral satellite data | |
| Gholizadeh et al. | Soil organic carbon and texture retrieving and mapping using proximal, airborne and Sentinel-2 spectral imaging | |
| Malenovský et al. | Influence of woody elements of a Norway spruce canopy on nadir reflectance simulated by the DART model at very high spatial resolution | |
| Garioud et al. | Recurrent-based regression of Sentinel time series for continuous vegetation monitoring | |
| Garcia et al. | Detection of interannual vegetation responses to climatic variability using AVIRIS data in a coastal savanna in California | |
| CN105809148B (en) | Crop drought recognition and risk assessment method based on remote sensing space-time spectrum fusion | |
| Camacho-De Coca et al. | Vegetation cover seasonal changes assessment from TM imagery in a semi-arid landscape | |
| Odebiri et al. | Deep learning-based national scale soil organic carbon mapping with Sentinel-3 data | |
| Yu et al. | A method to avoid spatial overfitting in estimation of grassland above-ground biomass on the Tibetan Plateau | |
| CN109001127A (en) | A kind of heavy metal content in soil space predicting method | |
| CN114926748A (en) | Soybean remote sensing identification method combining Sentinel-1/2 microwave and optical multispectral images | |
| Niculescu et al. | Synergy between lidar, radarsat-2, and spot-5 images for the detection and mapping of wetland vegetation in the danube delta | |
| Pang et al. | Pixel-level rice planting information monitoring in Fujin City based on time-series SAR imagery | |
| Yin et al. | Estimation of grassland height based on the random forest algorithm and remote sensing in the Tibetan Plateau | |
| Castaldi et al. | Assessing the capability of Sentinel-2 time-series to estimate soil organic carbon and clay content at local scale in croplands | |
| Liu et al. | Evaluating how lodging affects maize yield estimation based on UAV observations | |
| Zhao et al. | Cropland abandonment mapping at sub-pixel scales using crop phenological information and MODIS time-series images | |
| Garioud et al. | Challenges in grassland mowing event detection with multimodal sentinel images | |
| Nigam et al. | Modeling temporal growth profile of vegetation index from Indian geostationary satellite for assessing in-season progress of crop area | |
| Guo et al. | Estimating aboveground biomass of alpine grassland during the wilting period using in situ hyperspectral, Sentinel-2 and Sentinel-1 data | |
| Sanam et al. | Multi-sensor Approach for the Estimation of Above-Ground Biomass of Mangroves | |
| Fayad et al. | Vision Transformers, a new approach for high-resolution and large-scale mapping of canopy heights | |
| Gupta et al. | Fraction of vegetation cover and its application in vegetation characterization in the Hazaribagh wildlife sanctuary, Jharkhand, India | |
| Gragn | Mapping Arable Field Fractions with Muultisensor Remote Sensing Data-Driven Gradient Boosted and Classical GAMs | |
| Blasch | Multitemporal soil pattern analysis for organic matter estimation at croplands using multispectral satellite data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181214 |
|
| RJ01 | Rejection of invention patent application after publication |