CN103529189A - Soil organic matter space distribution predication method based on qualitative and quantitative auxiliary variables - Google Patents
Soil organic matter space distribution predication method based on qualitative and quantitative auxiliary variables Download PDFInfo
- Publication number
- CN103529189A CN103529189A CN201310268497.0A CN201310268497A CN103529189A CN 103529189 A CN103529189 A CN 103529189A CN 201310268497 A CN201310268497 A CN 201310268497A CN 103529189 A CN103529189 A CN 103529189A
- Authority
- CN
- China
- Prior art keywords
- soil
- organic matter
- value
- distribution
- study area
- 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.)
- Granted
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a soil organic matter space distribution predication method based on qualitative and quantitative auxiliary variables. The soil organic matter space distribution predication method is characterized by comprising the following steps: (1) data sources; (2) auxiliary data sources and processing; (3) a research method; (4) a contrast method. With the adoption of the technical scheme, compared with the prior art, the soil organic matter space distribution predication method applies an artificial neural network model and carries out space distribution predication of the content of soil organic matters on the basis of fusing the qualitative and quantitative auxiliary environment variables including soil types, terrain factors, vegetation indexes and the like, so as to provide method references for the regional high-precision space distribution predication of soil properties.
Description
Technical field
The present invention relates to a kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable.
Background technology
The soil organism is the important indicator of soil characteristic, is also the mineral nutrition of plant and organotrophic source, at aspects such as regional soil fertility, agricultural production and environmental protection, all has important function and significance.Due to the impact of the multiple factors of soil formation and ecological process, soil organic matter content spatially presents non-uniform Distribution.Conventionally the sampling point soil organic matter content information obtaining based on sampling, is difficult to meet the actual demand of opposite region soil content of organic matter distribution characteristics in regional agriculture production and environmental protection.Therefore, accurately obtaining the space distribution information of soil organic matter content, is the actual needs of accurately grasping soil nutrient management and the regional environment protection of regional soil fertility state, science.
The Spatial Distribution Pattern of soil property forms under the acting in conjunction of various ground surface environment factor.Introduce relevant earth's surface envirment factor as the soil property space distribution Forecasting Methodology of auxiliary variable, considering in varying degrees the impact of correlative environmental factors on soil property space distribution, the method that its precision of prediction more only carries out spatial interpolation based on sampling point data is significantly improved.In all kinds of auxiliary variables, Quantitative Factors, as relative easy operatings of the factor such as meteorology, terrain feature parameters, is widely used in the prediction of soil property space distribution.Qualitative factor also has significant impact to soil property spatial variability as Land-Use, soil types and matrix type etc., and such factor even can weaken the effect of Quantitative Factors to soil property spatial variability.Therefore, in Forecasting Methodology, introduce quantitative and qualitative analysis auxiliary variable is further to improve the inevitable requirement of soil property space distribution precision of prediction and one of the most effective approach simultaneously.Yet in the prediction of soil property space distribution, the use of qualitative factor is obviously on the low side.
In introducing the Forecasting Methodology of environmental variance, the methods such as multivariate regression model, Cokriging and Regression-kriging method are widely used in portraying the relation between soil property and envirment factor, and then the space distribution of prediction soil property.Yet, between soil property and environmental factor, be a kind of extremely complicated nonlinear relationship, the complex relationship of accurately expressing between soil property and environmental factor is the key that improves precision of prediction.Artificial nerve network model can disclose the Nonlinear Mapping relation between polynary environmental variance and target variable in complication system more exactly, be successfully applied to the nonlinear relationship of describing between the soil and environment factor, in the space distribution forecasting research of soil property, receive increasing concern.
Summary of the invention
The technical problem to be solved in the present invention is to overcome above-mentioned defect, a kind of utilization artificial nerve network model is provided, merging on the basis of the quantitative and qualitative analysis auxiliary environment variablees such as soil types, terrain factor and vegetation index, carrying out the soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable of the space distribution prediction of soil organic matter content.
For addressing the above problem, the technical solution adopted in the present invention is:
A soil organism space distribution Forecasting Methodology for quantitative and qualitative analysis auxiliary variable, is characterized in that: step is as follows:
1), Data Source: study area topomap, present landuse map and the soil types distribution plan of choosing of take is supplementary, on the basis of taking into account representative and homogeneity principle, consider the information such as landform, soil types and carry out the laying of soil sampling point, at each soil sampling point place, adopt the method that multiple spot mixes to gather top layer pedotheque; Each sampled point all records its geographic coordinate and sea level elevation with GPS, records in detail the environmental information at sampling point place simultaneously; The sample collecting is taken back laboratory and after natural air drying, was ground 1mm sieve, adopts potassium bichromate titrimetric method to measure each soil sample content of organic matter;
2), auxiliary data source and processing:
In research, Quantitative Factors has mainly been selected terrain factor and vegetation index; Terrain factor is the most frequently used auxiliary environment variable in the prediction of soil property space distribution, comprises elevation (H), the gradient (S), planar curvature (Ct), profile curvature (Cp) and Topographic Wetness Index (TI); Based on study area 1:5 ten thousand topomap, in ArcGIS9.3, generate the data elevation model of 30m resolution, further utilize the spatial analysis of ArcGIS9.3 and the gradient, planar curvature, profile curvature and the Topographic Wetness Index distribution plan that grid computing function is obtained study area 30m resolution;
The vegetation index obtaining based on remote sensing image is growth conditions and the vegetation coverage information of reflecting regional surface vegetation preferably, is the another auxiliary variable that is usually used in the prediction of soil property space distribution; Enhancement mode vegetation index in study area employing MODIS remotely-sensed data product is as the quantitative vegetation factor of forecasting research district soil organic matter content space distribution;
3), research method:
What research method adopted is the radial basis function neural network method that merges quantitative and qualitative analysis auxiliary variable;
Relation table between the measured value of each sampled point soil organism and environmental factor is shown:
Z(x
i,k,y
j,k)=m(k)+r(x
i,k,y
j,k) (1)
r(x
i,k,y
j,k)=f(t(x
i,y
j),v(x
i,y
j),...) (2)
In formula: Z (x
i,k, y
j,k) represent that the soil sample that belongs to k kind soil types is at (x
i, y
j) content of organic matter value located, (x
i, y
j) be sample point coordinate, its ranks number are respectively i and j; M (k) is k kind soil types content of organic matter mean value; r(x
i,k, y
j,k) be sampled point (x
i, y
j) deduct the residual error after soil types content of organic matter mean value under this point.T(x
i, y
j) be point (x
i, y
j) terrain parameter value, v (x
i, y
j) be point (x
i, y
j) vegetation index value.First the variation of supposing the soil organism on specified point position is put residing soil types by this and is determined, its residual values is put the local environmental factors such as residing landform vegetation condition and determined by this; Mean value m (k) and residual error r (x
i,k, y
j,k) separate.
Relation between each point position residual values and local environmental factor adopts radial basis function artificial nerve network model to express; This model is a kind of three layers of feed-forward type neural network model with single hidden layer, and Forecasting Methodology process is as follows:
(1), according to sampled point soil organic matter content value, in statistical research district, the mean value of each soil types soil organic matter content and each soil sampling are put corresponding residual values respectively;
(2), according to each soil types content of organic matter mean value and the study area soil types distribution plan that calculate, the organic value of each soil types of making and research district content mean value distribution plan;
(3), usining landform and the vegetation characteristics parameters value of each soil sampling point inputs as network, the soil organism residual values of usining after the normalization of corresponding point position is exported as network, build neural network model, in MATLAB, simulation obtains the residual distribution figure of the study area soil organism;
(4), finally the organic value mean value of each soil types of study area distribution plan is added to the residual distribution figure that Neural Network model predictive obtains, obtain study area soil organic matter content spatial distribution map;
4), contrast method:
Select method that Ordinary Kriging Interpolation method, Regression-kriging method and neural network model be combined with Ordinary Kriging Interpolation method in contrast; First Regression-kriging method adopts multiple stepwise regression to carry out the regression forecasting between the soil organism and envirment factor, with common Ke Lifa, regression forecasting result residual error is carried out to interpolation again, finally regression forecasting result and Ordinary Kriging Interpolation method are added the estimated value of residual error, obtain the spatial distribution map of the study area soil organism; The method that neural network model is combined with Ordinary Kriging Interpolation is to complete the space distribution prediction to the study area soil organism with the regression model in neural network model alternative regression Kriging method.
Owing to having adopted technique scheme, compared with prior art, being take in the Santai County of hills area, the Sichuan Basin in the present invention is study area, use artificial nerve network model, merging on the basis of the quantitative and qualitative analysis auxiliary environment variablees such as soil types, terrain factor and vegetation index, carry out the space distribution prediction of soil organic matter content, to being region high precision soil Property Spaces forecast of distribution supplying method reference.
Accompanying drawing explanation
Fig. 1 is the histogram of study area soil organic matter content and logarithm conversion value thereof;
Fig. 2 is each data item semi-variance function of study area soil organism figure;
Fig. 3 is the predict the outcome figures of different Forecasting Methodologies to soil organic matter content space distribution.
Embodiment
Embodiment:
A soil organism space distribution Forecasting Methodology for quantitative and qualitative analysis auxiliary variable, step is as follows:
1), Data Source: study area topomap, present landuse map and the soil types distribution plan of choosing of take is supplementary, on the basis of taking into account representative and homogeneity principle, consider the information such as landform, soil types and carry out the laying of soil sampling point, at each soil sampling point place, adopt the method that multiple spot mixes to gather top layer pedotheque.The whole district gathers 2346 sampled points of soil sample altogether, and each sampled point all records its geographic coordinate and sea level elevation with GPS, the environmental information that simultaneously records in detail sampling point place as soil types, cultivate the information such as crop and farmland crop rotation method.The sample collecting is taken back laboratory and after natural air drying, was ground 1mm sieve, adopts potassium bichromate titrimetric method to measure each soil sample content of organic matter.
2), auxiliary data source and processing:
In research, Quantitative Factors has mainly been selected terrain factor and vegetation index; Terrain factor is the most frequently used auxiliary environment variable in the prediction of soil property space distribution, comprises elevation (H), the gradient (S), planar curvature (Ct), profile curvature (Cp) and Topographic Wetness Index (TI); Based on study area 1:5 ten thousand topomap, in ArcGIS9.3, generate the data elevation model of 30m resolution, further utilize the spatial analysis of ArcGIS9.3 and the gradient, planar curvature, profile curvature and the Topographic Wetness Index distribution plan that grid computing function is obtained study area 30m resolution.
The vegetation index obtaining based on remote sensing image is growth conditions and the vegetation coverage information of reflecting regional surface vegetation preferably, is the another auxiliary variable that is usually used in the prediction of soil property space distribution; Enhancement mode vegetation index in study area employing MODIS remotely-sensed data product is as the quantitative vegetation factor of forecasting research district soil organic matter content space distribution.Consider the accumulation characteristic of the soil organism, the EVI data of the synthetic 250m resolution of 16d annual July and August in 2001 to 2006 in the present invention, have been selected, after the EVI data to downloaded are carried out quality check and processed, average and be resampled to 30m resolution, the average vegetation index while using biomass maximum in 2001 to 2006 is as one of quantitative auxiliary variable.For convenience of calculating, all Quantitative Factors are all normalized.
Research shows, hills area soil types is large on the impact of soil organic matter content space distribution compared with land use pattern, therefore in research, selects study area soil types as the qualitative auxiliary variable of this research.Study area soil types distribution plan derives from the overall survey of soil for the second time.After to the scanning of study area soil type map, utilize ArcGIS9.3 to carry out vector quantization, obtain study area digital soil Map of Distributions of Types.
3), research method:
What research method adopted is the radial basis function neural network method that merges quantitative and qualitative analysis auxiliary variable;
Relation table between the measured value of each sampled point soil organism and environmental factor is shown:
Z(x
i,k,y
j,k)=m(k)+r(x
i,k,y
j,k) （1）
r(x
i,k,y
j,k)=f(t(x
i,y
j),v(x
i,y
j),...) （2）
In formula: Z (x
i,k, y
j,k) represent that the soil sample that belongs to k kind soil types is at (x
i, y
j) content of organic matter value located, (x
i, y
j) be sample point coordinate, its ranks number are respectively i and j; M (k) is k kind soil types content of organic matter mean value; r(x
i,k, y
j,k) be sampled point (x
i, y
j) deduct the residual error after soil types content of organic matter mean value under this point.T(x
i, y
j) be point (x
i, y
j) terrain parameter value, v (x
i, y
j) be point (x
i, y
j) vegetation index value.First the variation of supposing the soil organism on specified point position is put residing soil types by this and is determined, its residual values is put the local environmental factors such as residing landform vegetation condition and determined by this; Mean value m (k) and residual error r (x
i,k, y
j,k) separate.
Relation between each point position residual values and local environmental factor adopts radial basis function artificial nerve network model to express; This model is a kind of three layers of feed-forward type neural network model with single hidden layer, and Forecasting Methodology process is as follows:
(1), according to sampled point soil organic matter content value, in statistical research district, the mean value of each soil types soil organic matter content and each soil sampling are put corresponding residual values respectively;
(2), according to each soil types content of organic matter mean value and the study area soil types distribution plan that calculate, the organic value of each soil types of making and research district content mean value distribution plan;
(3), usining landform and the vegetation characteristics parameters value of each soil sampling point inputs as network, the soil organism residual values of usining after the normalization of corresponding point position is exported as network, build neural network model, in MATLAB, simulation obtains the residual distribution figure of the study area soil organism;
(4), finally the organic value mean value of each soil types of study area distribution plan is added to the residual distribution figure that Neural Network model predictive obtains, obtain study area soil organic matter content spatial distribution map;
4), contrast method:
Select method that Ordinary Kriging Interpolation method, Regression-kriging method and neural network model be combined with Ordinary Kriging Interpolation method in contrast; First Regression-kriging method adopts multiple stepwise regression to carry out the regression forecasting between the soil organism and envirment factor, with common Ke Lifa, regression forecasting result residual error is carried out to interpolation again, finally regression forecasting result and Ordinary Kriging Interpolation method are added the estimated value of residual error, obtain the spatial distribution map of the study area soil organism; The method that neural network model is combined with Ordinary Kriging Interpolation is to complete the space distribution prediction to the study area soil organism with the regression model in neural network model alternative regression Kriging method.
For the above-mentioned soil organism space distribution Forecasting Methodology predicted exactitude evaluation method based on quantitative and qualitative analysis auxiliary variable, adopting each method of individual authentication sampling point set pair to predict the outcome evaluates, from 2346 soil sampling points, extract 20% (469) as check post at random, all the other 80% (1877) are as modeling point, with mean absolute error (MAE), root-mean-square error (RMSE) and average relative error (MRE), the predicted value of check post and actual observed value are analyzed, draw precision evaluation result.
The above-mentioned soil organism space distribution Forecasting Methodology results and analysis based on quantitative and qualitative analysis auxiliary variable:
1, descriptive statistical analysis.
According to modeling point, statistical study shows, as shown in table 1, and the mean value of study area soil organism massfraction is 17.97g/kg; The coefficient of variation, between 10～100%, is moderate variability.Soil type soil organic matter content differs greatly, and rice soil quality of organic matter mark mean value reaches 22.49g/kg, far above other 3 soil typess; Next is moisture soil and purple soil, and mean value is respectively 15.87 and 14.09g.kg-1; Yellow soil is minimum, and mean value is 13.66g.kg-1.From degree of variation, rice soil and purple soil content of organic matter degree of variation be a little more than another 2 great soil groups, but all belong to moderate variation.
From the distributional pattern of data, raw data is obvious skewness and distributes, by its degree of bias value and kurtosis value and distribution frequency figure after number conversion are obviously more approached to normal distribution, as shown in table 1, Fig. 1.Therefore, the numerical value after number conversion is used for studying the calculating of various Forecasting Methodologies.
Table 1 study area soil type content of organic matter descriptive statistic feature
2, the relationship analysis of the soil organism and environmental variance.
Utilize variance analysis to test to study area soil type content of organic matter difference, result shows, as shown in table 2, there is extremely significant difference (P < 0.001) in 4 kinds of the study areas soil types content of organic matter, the Spatial Distribution Pattern of soil types has important impact to the space distribution of study area soil organic matter content.
The variance analysis of the content of organic matter between table 2 soil type
Correlation analysis result between soil organic matter content and 6 Quantitative Factors shows, study area soil organic matter content is obviously subject to the impact of landform and vegetation factor, as shown in table 3.Wherein, there is extremely significant negative correlativing relation in sea level elevation and the gradient and soil organism massfraction, related coefficient is respectively-0.06 (P=0.0067 < 0.01) and-0.22 (P=4.17E-21 < 0.01), on this some position that explanation physical features is higher, the gradient is larger, soil organic matter content is lower.Related coefficient between Topographic Wetness Index and vegetation index and soil organism massfraction is respectively 0.24 and 0.17, all reach extremely significant positive correlation (P=6.87E-27 and 5.67E-13), this shows that on physical features is more low-lying, vegetation cover degree is larger position, soil organic matter content is higher.In addition, topographic profile curvature and soil organic matter content also have correlativity to a certain degree (related coefficient 0.06, P=0.02).
Relative coefficient (n=1877) between table 3 soil type organic content and each soil types average residual error and envirment factor
The residual error and each Quantitative Factors that under each sampling point content of organic matter is removed, after great soil group mean value, obtain carry out correlation analysis, result shows, as shown in table 3, the related coefficient of average residual error and the gradient is-0.09, reaches extremely significant negative correlativing relation (P=9.35E-05); Be respectively 0.10 and 0.13 with the related coefficient of Topographic Wetness Index and vegetation index, all show as extremely significant positive correlation (P=9.43E-06 and 2.81E-08).This shows, removes after each soil types mean value, and the difference of study area soil organic matter content is main relevant with the difference of surface slope, Topographic Wetness Index and plant cover situation.
3, semivariance analysis
Modeling is put to soil organic matter content value, the residual values of multiple stepwise regression prediction and the residual error of neural network prediction result and carry out number conversion, make it more approach normal distribution, complete on this basis the semivariance analysis to above-mentioned 3 data item.Result shows, as shown in table 4, Fig. 2,3 data item all meet spherical model, and the coefficient of determination of model is all more than 0.85, and the fitting degree of selected theoretical semivariance model is higher, can reflect preferably the spatial structure characteristic of each index.From the parameter of model, the nugget effect value of 3 data item is between 0.742～0.765, and range is between 6.00～7.00km, and this shows, the space correlation degree of the study area soil organism a little less than, the scope of spatial autocorrelation is little; Semivariance model parameter and the former variable change of regression equation and Neural Network model predictive residual error are less, have substantially retained the spatial structure characteristic of former variable.
The semivariance model parameter of the table 4 study area soil organism and recurrence thereof and neural network prediction residual values
Ln SOM: the logarithm conversion value of the soil organism; The logarithm conversion value of MLR residuals soil organism regression residuals; Ln RBFNN residuals: the logarithm conversion value of soil organism neural network prediction residual error.
4, soil organic matter content space distribution predicts the outcome
Fig. 3 is the study area soil organic matter content spatial distribution map that different Forecasting Methodologies obtain.From the figure that predicts the outcome, study area soil organic matter content distribution trend is also not obvious, and this is relevant with the landform of study area fragmentation.The distinct methods difference that predicts the outcome is obvious.Ordinary Kriging Interpolation method (OK) predicts the outcome to walk and is obvious block distribution (Fig. 3 c) compared with Ping Chang Hua，Gao Zhi district and low value district.The method (RBFNN+OK) that Regression-kriging method (RK) combines with Ordinary Kriging Interpolation method with neural network model predicts the outcome similar, can embody to a certain extent the detailed information (Fig. 3 b and d) of soil organic matter content deformation everywhere; But all in northeast, study area and east, there are 2 high value regions that are significantly block distribution.The neural net method that merges quantitative and qualitative analysis auxiliary variable predicts the outcome does not have obviously to become the high value areal distribution of piece in figure, if white portion in Fig. 3 a figure is that ，Gao Zhi district, waters mainly appears in the coombe of hypsography low-lying.
In Sichuan Basin, in the more low-lying coombe of physical features, main distribution soil types is rice soil, and statistics has shown that its content of organic matter is apparently higher than other soil types, as shown in table 1.Correlation analysis shows that the low-lying place of landform soil organic matter content is higher, as shown in table 3.The statistic analysis result of other correlative study shows, under the mima type microrelief of Sichuan Basin, content of organic matter difference is: waist > brae, pin > mound, mound; Under this mima type microrelief, the distribution characteristics of soil organic matter content has determined to become block distribution Gao Zhi district and low value district all not to meet the actual conditions of the study area of breaking topography.Therefore, the neural net method of fusion quantitative and qualitative analysis auxiliary variable is more consistent with study area actual conditions to predicting the outcome of study area soil organism Spatial Distribution Pattern.
2.5 predicted exactitude evaluation
Table 5 is error statistics results that different Forecasting Methodologies predict the outcome to 469 checking sampling points.From error analysis, can find out, neural net method (ST+RBFNN) predicated error that merges quantitative and qualitative analysis auxiliary variable is significantly less than other 3 kinds of methods.Wherein, the method that the mean absolute error that ST+RBFNN predicts the outcome to 469 checking sampling points is combined with Ordinary Kriging Interpolation compared with Ordinary Kriging Interpolation method, Regression-kriging method and neural network has reduced respectively 30.78%, 27.43% and 25.70%; Average relative error has reduced respectively 35.27%, 31.86% and 30.06%, and root-mean-square error has reduced respectively 22.15%, 19.35 and 17.94%, and it is obvious that error reduces amplitude.
From soil type, the neural net method that merges quantitative and qualitative analysis auxiliary variable is maximum to purple soil check post precision of prediction increase rate, and every error between 25.93～49.27%, on average reduces by 38.24% compared with the reduction amplitude of other 3 kinds of methods; Next is yellow soil, and the reduction amplitude of every error, between 17.92～46.19%, on average reduces by 30.04%; Be moisture soil again, the reduction amplitude of every error, between 12.50～25.70%, on average reduces by 19.42%; And reduction amplitude minimum is rice soil, the reduction amplitude of every error, between 0.59～18.55%, on average reduces by 10.08%.This is mainly because selected orographic factor can disclose the development characteristics of front 3 kinds of soil typess comparatively speaking better, and each landform factor difference of region that rice soil distributes is relatively little.
The different Forecasting Methodology predicated errors of table 5 are analyzed
Note: OK: Ordinary Kriging Interpolation method; RK: Regression-kriging method; RBFNN+OK: the method that radial basis function neural network combines with Ordinary Kriging Interpolation; ST+RBFNN: the neural net method that merges qualitative, quantitative auxiliary variable; MAE: mean absolute error; MRE: average relative error; RMSE: root-mean-square error.
According to embodiments of the invention, can draw the following conclusions:
(1) soil organism massfraction in study area is between 4.20～47.60g.kg-1, average out to 17.97g.kg-1; Value for coefficient of variation 36.89% is moderate variability.In soil type, the rice soil content of organic matter is maximum, and its average quality mark is 22.49g.kg-1; Next is moisture soil and purple soil, and mean value is respectively 15.87 and 14.09g.kg-1; Yellow soil is minimum, and mean value is 13.66g.kg-1.The results of analysis of variance shows, between soil type, the content difference of the soil organism reaches the utmost point level of signifiance (F=411.75, P < 0.001), and soil types factor has significant impact to the space distribution of this regional soil content of organic matter.
(2) correlation analysis shows, there is extremely significant negative correlativing relation (P < 0.01) in soil organic matter content and study area sea level elevation, the gradient, there is extremely significant positive correlation (P < 0.01) with Topographic Wetness Index and vegetation index, also have certain correlativity (P < 0.05) with profile curvature.After removing soil types factor, the gradient, landform humidity and the vegetation factor are the Main Factors that causes study area Spatial Variability of Soil Organic.
(3) semivariance analysis shows, the piece gold number of the study area soil organism and the ratio of base station value between 0.742～0.765, spatial autocorrelation a little less than; Range is between 6.00～7.00km, and the scope of spatial autocorrelation is little.
(4) the method precision that the neural network model of fusion soil types factor and landform vegetation factor combines with Ordinary Kriging Interpolation compared with Ordinary Kriging Interpolation method, Regression-kriging method and neural network is significantly improved; The mean absolute error that the method predicts the outcome to 469 check posts has reduced respectively 30.78%, 27.43% and 25.70% compared with other 3 kinds of methods; Average relative error has reduced respectively 35.27%, 31.86% and 30.06%, and root-mean-square error has reduced respectively 22.15%, 19.35 and 17.94%, and it is obvious that error reduces amplitude.The spatial variability research that the method can be complex environment regional soil character provides Research Thinking.
The present invention is not limited to above-mentioned preferred implementation, and anyone should learn the structural change of making under enlightenment of the present invention, and every have identical or akin technical scheme with the present invention, all belongs to protection scope of the present invention.
Claims (1)
1. the soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable, is characterized in that: step is as follows:
1), Data Source: study area topomap, present landuse map and the soil types distribution plan of choosing of take is supplementary, on the basis of taking into account representative and homogeneity principle, consider the information such as landform, soil types and carry out the laying of soil sampling point, at each soil sampling point place, adopt the method that multiple spot mixes to gather top layer pedotheque; Each sampled point all records its geographic coordinate and sea level elevation with GPS, records in detail the environmental information at sampling point place simultaneously; The sample collecting is taken back laboratory and after natural air drying, was ground 1mm sieve, adopts potassium bichromate titrimetric method to measure each soil sample content of organic matter;
2), auxiliary data source and processing:
In research, Quantitative Factors has mainly been selected terrain factor and vegetation index; Terrain factor is the most frequently used auxiliary environment variable in the prediction of soil property space distribution, comprises elevation (H), the gradient (S), planar curvature (Ct), profile curvature (Cp) and Topographic Wetness Index (TI); Based on study area 1:5 ten thousand topomap, in ArcGIS9.3, generate the data elevation model of 30m resolution, further utilize the spatial analysis of ArcGIS9.3 and the gradient, planar curvature, profile curvature and the Topographic Wetness Index distribution plan that grid computing function is obtained study area 30m resolution;
The vegetation index obtaining based on remote sensing image is growth conditions and the vegetation coverage information of reflecting regional surface vegetation preferably, is the another auxiliary variable that is usually used in the prediction of soil property space distribution; Enhancement mode vegetation index in study area employing MODIS remotely-sensed data product is as the quantitative vegetation factor of forecasting research district soil organic matter content space distribution;
3), research method:
What research method adopted is the radial basis function neural network method that merges quantitative and qualitative analysis auxiliary variable;
Relation table between the measured value of each sampled point soil organism and environmental factor is shown:
Z(x
i,k,y
j,k)=m(k)+r(x
i,k,y
j,k) (1)
r(x
i,k,y
j,k)=f(t(
xi,y
j),v(x
i,y
j),...) (2)
In formula: Z (x
i,k, y
j,k) represent that the soil sample that belongs to k kind soil types is at (x
i, y
j) content of organic matter value located, (x
i, y
j) be sample point coordinate, its ranks number are respectively i and j; M (k) is k kind soil types content of organic matter mean value; r(x
i,k, y
j,k) be sampled point (x
i, y
j) deduct the residual error after soil types content of organic matter mean value under this point.T(x
i, y
j) be point (x
i, y
j) terrain parameter value, v (x
i, y
j) be point (x
i, y
j) vegetation index value.First the variation of supposing the soil organism on specified point position is put residing soil types by this and is determined, its residual values is put the local environmental factors such as residing landform vegetation condition and determined by this; Mean value m (k) and residual error r (x
i,k, y
j,k) separate;
Relation between each point position residual values and local environmental factor adopts radial basis function artificial nerve network model to express; This model is a kind of three layers of feed-forward type neural network model with single hidden layer, and Forecasting Methodology process is as follows:
(1), according to sampled point soil organic matter content value, in statistical research district, the mean value of each soil types soil organic matter content and each soil sampling are put corresponding residual values respectively;
(2), according to each soil types content of organic matter mean value and the study area soil types distribution plan that calculate, the organic value of each soil types of making and research district content mean value distribution plan;
(3), usining landform and the vegetation characteristics parameters value of each soil sampling point inputs as network, the soil organism residual values of usining after the normalization of corresponding point position is exported as network, build neural network model, in MATLAB, simulation obtains the residual distribution figure of the study area soil organism;
(4), finally the organic value mean value of each soil types of study area distribution plan is added to the residual distribution figure that Neural Network model predictive obtains, obtain study area soil organic matter content spatial distribution map;
4), contrast method:
Select method that Ordinary Kriging Interpolation method, Regression-kriging method and neural network model be combined with Ordinary Kriging Interpolation method in contrast; First Regression-kriging method adopts multiple stepwise regression to carry out the regression forecasting between the soil organism and envirment factor, with common Ke Lifa, regression forecasting result residual error is carried out to interpolation again, finally regression forecasting result and Ordinary Kriging Interpolation method are added the estimated value of residual error, obtain the spatial distribution map of the study area soil organism; The method that neural network model is combined with Ordinary Kriging Interpolation is to complete the space distribution prediction to the study area soil organism with the regression model in neural network model alternative regression Kriging method;
5), predicted exactitude evaluation method
Adopting each method of individual authentication sampling point set pair to predict the outcome evaluates, from 2346 soil sampling points, extract 20% (469) as check post at random, all the other 80% (1877) are as modeling point (Fig. 1 b), with mean absolute error (MAE), root-mean-square error (RMSE) and average relative error (MRE), the predicted value of check post and actual observed value are analyzed, draw precision evaluation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310268497.0A CN103529189B (en) | 2013-06-28 | 2013-06-28 | A kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310268497.0A CN103529189B (en) | 2013-06-28 | 2013-06-28 | A kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103529189A true CN103529189A (en) | 2014-01-22 |
| CN103529189B CN103529189B (en) | 2016-01-06 |
Family
ID=49931368
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310268497.0A Expired - Fee Related CN103529189B (en) | 2013-06-28 | 2013-06-28 | A kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103529189B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104239446A (en) * | 2014-09-01 | 2014-12-24 | 北京农业信息技术研究中心 | Method for sampling point layout by utilizing auxiliary variable data |
| CN104408258A (en) * | 2014-12-01 | 2015-03-11 | 四川农业大学 | Large-scale soil organic carbon spatial distribution simulation method involving environmental factors |
| CN104933699A (en) * | 2015-04-22 | 2015-09-23 | 吉林大学 | Method for automatically extracting phenology information of earth surface vegetation based on fitting variance of Gaussian function |
| CN105243435A (en) * | 2015-09-15 | 2016-01-13 | 中国科学院南京土壤研究所 | Deep learning cellular automaton model-based soil moisture content prediction method |
| CN106198434A (en) * | 2016-07-26 | 2016-12-07 | 新疆农垦科学院 | A kind of remote-sensing monitoring method of the North SinKiang desert grey soil content of organic matter |
| CN107657633A (en) * | 2017-09-28 | 2018-02-02 | 哈尔滨工业大学 | A kind of soil improving straw mulching rate measuring method based on BP neural network and sensor data acquisition |
| CN107860889A (en) * | 2017-09-22 | 2018-03-30 | 华南农业大学 | The Forecasting Methodology and equipment of the soil organism |
| CN105528523B (en) * | 2015-12-18 | 2018-09-21 | 中国科学院遥感与数字地球研究所 | A kind of thickness of soil inversion method based on remotely-sensed data |
| CN109374860A (en) * | 2018-11-13 | 2019-02-22 | 西北大学 | A kind of soil nutrient prediction and integrated evaluating method based on machine learning algorithm |
| CN111413484A (en) * | 2020-03-02 | 2020-07-14 | 南京信息职业技术学院 | Analysis method for spatial correlation between soil chromium content and land utilization type |
| CN112816658A (en) * | 2021-01-05 | 2021-05-18 | 四川大学 | Large-area-scale surface soil nutrient space characteristic monitoring method based on low-altitude remote sensing |
| CN113379188A (en) * | 2021-05-06 | 2021-09-10 | 贵州省烟草公司贵阳市公司 | Tobacco crop rotation planting method and system based on Internet of things |
| CN114814167A (en) * | 2022-04-08 | 2022-07-29 | 中科吉安生态环境研究院 | Soil heavy metal content inversion method fusing multi-source environment variables and spectral information |
| CN115932154A (en) * | 2022-09-21 | 2023-04-07 | 中国科学院东北地理与农业生态研究所 | Soil organic carbon remote sensing mapping method considering soil type |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109001127A (en) * | 2018-06-01 | 2018-12-14 | 华南农业大学 | A kind of heavy metal content in soil space predicting method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH01311248A (en) * | 1988-06-09 | 1989-12-15 | Soiru & Lock Eng Kk | Method of extracting homogeneous sample soil body |
| JPH11316221A (en) * | 1998-04-30 | 1999-11-16 | Yanmar Agricult Equip Co Ltd | Soil nutrient distribution analyzing method |
| CN101788553A (en) * | 2010-03-02 | 2010-07-28 | 中国农业大学 | Multiscale analysis method for vegetation indexes of refuse dump and soil nutrient space |
| CN102680413A (en) * | 2012-05-25 | 2012-09-19 | 浙江大学 | Device and method for hyperspectral rapid detection on content of field soil organic matters by using panoramic annular belt |
-
2013
- 2013-06-28 CN CN201310268497.0A patent/CN103529189B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH01311248A (en) * | 1988-06-09 | 1989-12-15 | Soiru & Lock Eng Kk | Method of extracting homogeneous sample soil body |
| JPH11316221A (en) * | 1998-04-30 | 1999-11-16 | Yanmar Agricult Equip Co Ltd | Soil nutrient distribution analyzing method |
| CN101788553A (en) * | 2010-03-02 | 2010-07-28 | 中国农业大学 | Multiscale analysis method for vegetation indexes of refuse dump and soil nutrient space |
| CN102680413A (en) * | 2012-05-25 | 2012-09-19 | 浙江大学 | Device and method for hyperspectral rapid detection on content of field soil organic matters by using panoramic annular belt |
Non-Patent Citations (2)
| Title |
|---|
| 李启权等: "丘陵区土壤有机质空间分布预测的神经网络方法", 《农业环境科学学报》, vol. 31, no. 12, 31 December 2012 (2012-12-31), pages 2451 - 2458 * |
| 李启权等: "基于神经网络模型和地统计学方法的土壤养分空间分布预测", 《应用生态学报》, vol. 24, no. 2, 28 February 2013 (2013-02-28), pages 459 - 466 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104239446B (en) * | 2014-09-01 | 2017-06-30 | 北京农业信息技术研究中心 | A kind of method that sampled point layout is carried out using auxiliary variable data |
| CN104239446A (en) * | 2014-09-01 | 2014-12-24 | 北京农业信息技术研究中心 | Method for sampling point layout by utilizing auxiliary variable data |
| CN104408258A (en) * | 2014-12-01 | 2015-03-11 | 四川农业大学 | Large-scale soil organic carbon spatial distribution simulation method involving environmental factors |
| CN104408258B (en) * | 2014-12-01 | 2017-12-15 | 四川农业大学 | The large scale soil organic matter spatial distribution analogy method of integrated environment factor |
| CN104933699A (en) * | 2015-04-22 | 2015-09-23 | 吉林大学 | Method for automatically extracting phenology information of earth surface vegetation based on fitting variance of Gaussian function |
| CN104933699B (en) * | 2015-04-22 | 2017-06-30 | 吉林大学 | The method that surface vegetation phenology information is automatically extracted based on Gaussian function fitting variance |
| CN105243435B (en) * | 2015-09-15 | 2018-10-26 | 中国科学院南京土壤研究所 | A kind of soil moisture content prediction technique based on deep learning cellular Automation Model |
| CN105243435A (en) * | 2015-09-15 | 2016-01-13 | 中国科学院南京土壤研究所 | Deep learning cellular automaton model-based soil moisture content prediction method |
| CN105528523B (en) * | 2015-12-18 | 2018-09-21 | 中国科学院遥感与数字地球研究所 | A kind of thickness of soil inversion method based on remotely-sensed data |
| CN106198434A (en) * | 2016-07-26 | 2016-12-07 | 新疆农垦科学院 | A kind of remote-sensing monitoring method of the North SinKiang desert grey soil content of organic matter |
| CN107860889A (en) * | 2017-09-22 | 2018-03-30 | 华南农业大学 | The Forecasting Methodology and equipment of the soil organism |
| CN107657633A (en) * | 2017-09-28 | 2018-02-02 | 哈尔滨工业大学 | A kind of soil improving straw mulching rate measuring method based on BP neural network and sensor data acquisition |
| CN109374860A (en) * | 2018-11-13 | 2019-02-22 | 西北大学 | A kind of soil nutrient prediction and integrated evaluating method based on machine learning algorithm |
| CN111413484A (en) * | 2020-03-02 | 2020-07-14 | 南京信息职业技术学院 | Analysis method for spatial correlation between soil chromium content and land utilization type |
| CN112816658A (en) * | 2021-01-05 | 2021-05-18 | 四川大学 | Large-area-scale surface soil nutrient space characteristic monitoring method based on low-altitude remote sensing |
| CN113379188A (en) * | 2021-05-06 | 2021-09-10 | 贵州省烟草公司贵阳市公司 | Tobacco crop rotation planting method and system based on Internet of things |
| CN114814167A (en) * | 2022-04-08 | 2022-07-29 | 中科吉安生态环境研究院 | Soil heavy metal content inversion method fusing multi-source environment variables and spectral information |
| CN115932154A (en) * | 2022-09-21 | 2023-04-07 | 中国科学院东北地理与农业生态研究所 | Soil organic carbon remote sensing mapping method considering soil type |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103529189B (en) | 2016-01-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103529189B (en) | A kind of soil organism space distribution Forecasting Methodology based on quantitative and qualitative analysis auxiliary variable | |
| CN109142679B (en) | Forest soil nutrient space prediction method based on artificial neural network kriging interpolation | |
| Jin et al. | Spatio-temporal distribution of NDVI and its influencing factors in China | |
| Huang et al. | Jointly assimilating MODIS LAI and ET products into the SWAP model for winter wheat yield estimation | |
| CN110751094B (en) | Crop yield estimation method based on GEE comprehensive remote sensing image and deep learning method | |
| Zhang et al. | Quantifying methane emissions from rice paddies in Northeast China by integrating remote sensing mapping with a biogeochemical model | |
| CN106372277B (en) | Method for optimizing variation function model in forest land index space-time estimation | |
| CN109657988B (en) | Tobacco leaf quality partitioning method based on HASM and Euclidean distance algorithm | |
| Dan et al. | Delineating the rice crop activities in Northeast China through regional parametric synthesis using satellite remote sensing time-series data from 2000 to 2015 | |
| CN113011372B (en) | Automatic monitoring and identifying method for saline-alkali soil | |
| WO2024152865A1 (en) | Meteorological-station-based ecosystem carbon-water flux calculation method and system | |
| Hu et al. | Integrating CART algorithm and multi-source remote sensing data to estimate sub-pixel impervious surface coverage: a case study from Beijing Municipality, China | |
| CN105528523A (en) | Soil thickness inversion method based on remote sensing data | |
| Jiang et al. | Monitoring the coastal environment using remote sensing and GIS techniques | |
| Xiang et al. | Mapping potential wetlands by a new framework method using random forest algorithm and big Earth data: A case study in China's Yangtze River Basin | |
| Luo et al. | Mapping the soil organic matter content in a typical black-soil area using optical data, radar data and environmental covariates | |
| Cui et al. | Estimating maize yield in the black soil region of Northeast China using land surface data assimilation: integrating a crop model and remote sensing | |
| Liu et al. | Unrevealing past and future vegetation restoration on the Loess Plateau and its impact on terrestrial water storage | |
| Li et al. | Monitoring spatial and temporal patterns of rubber plantation dynamics using time-series landsat images and google earth engine | |
| Jiang et al. | Contrasting multiple deterministic interpolation responses to different spatial scale in prediction soil organic carbon: A case study in Mollisols regions | |
| Yasin et al. | URBAN SPRAWL ASSESSMENT USING TIME-SERIES LULC AND NDVI VARIATION: A CASE STUDY OF SEPANG, MALAYSIA. | |
| Zhang et al. | Enhanced Feature Extraction From Assimilated VTCI and LAI With a Particle Filter for Wheat Yield Estimation Using Cross-Wavelet Transform | |
| Zhuo et al. | Aboveground biomass retrieval of wetland vegetation at the species level using UAV hyperspectral imagery and machine learning | |
| Zhao et al. | The simultaneous prediction of yield and maturity date for wheat–maize by combining satellite images with crop model | |
| Wu et al. | Optimal Sample Size for SOC Content Prediction for Mapping Using the Random Forest in Cropland in Northern Jiangsu, China |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160106 Termination date: 20170628 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |