CN107860889A - The Forecasting Methodology and equipment of the soil organism - Google Patents
The Forecasting Methodology and equipment of the soil organism Download PDFInfo
- Publication number
- CN107860889A CN107860889A CN201710865008.8A CN201710865008A CN107860889A CN 107860889 A CN107860889 A CN 107860889A CN 201710865008 A CN201710865008 A CN 201710865008A CN 107860889 A CN107860889 A CN 107860889A
- Authority
- CN
- China
- Prior art keywords
- soil
- organic matter
- soil organic
- value
- model
- 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
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/24—Earth materials
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Food Science & Technology (AREA)
- Analytical Chemistry (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Medicinal Chemistry (AREA)
- Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a kind of Forecasting Methodology of the soil organism, including：Processor obtains the auxiliary variable data associated with the soil organism of measured position in target area from memory, and wherein memory storage has the auxiliary variable data of each position in target area；The soil organism predicted value of measured position is calculated using the auxiliary variable data of measured position as the input value of the extreme learning machine ELM models pre-established for processor.Processor obtains the corresponding residual error Ordinary Kriging Interpolation interpolation of measured position from memory, and the soil organism predicted value of measured position is obtained into the soil organism forecast value revision value of measured position with the superposition summation of corresponding residual error Ordinary Kriging Interpolation interpolation.Correspondingly, a kind of equipment for predicting the soil organism is also disclosed.By the embodiment of the present invention, the precision of prediction of the soil organism can be improved.
Description
Technical Field
The invention relates to the technical field of computer geographic information, in particular to a method and equipment for predicting soil organic matters.
Background
Soil organic matter is of great interest as an important "indicator" of the agroecological system and is a major natural and artificial driving force for the cause of spatial heterogeneity in environmental carbon cycling and agricultural management, affecting the response process of the crop system in soil productivity. In order to overcome the restriction of regional discreteness of soil organic matters on space surface source expression, soil nutrient space mapping with higher resolution has become an urgent requirement for non-point source space variation estimation, and accurate organic matter content space estimation not only can avoid soil ecological sustainability damage caused by unreasonable sampling, but also further deepens and enriches the space variation theory of soil science.
The traditional soil organic matter space prediction methods are numerous, and geostatistics and kriging interpolation are widely applied due to the good geostatistical interpretation capability, such as the application of a common kriging model. The common kriging model is simple in structure and easy to operate, but is greatly interfered by the density of sampling points in a research area, the spatial autocorrelation and nonlinearity of soil attributes and the like, and has low prediction precision and low generalization on the spatial distribution of soil organic matters.
In recent years, mixed earth statistical models (integrated models of linear or nonlinear algorithms and earth statistical methods) have been introduced into the field of soil property prediction, such as widely applied regression kriging and artificial neural network kriging. However, the regression kriging method still has the problem that the prediction accuracy is not high due to weak resolving power of the nonlinear relation between the soil organic matter and the associated variables, the artificial neural network method effectively improves the fitting effect, but needs complex model parameter adjustment, and the calculation time required by the improvement of the drawing resolution is increased, so that the model execution efficiency is low.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a novel method for predicting soil organic matters, which can improve the accuracy of predicting the soil organic matters.
According to an aspect of the present invention, there is provided a method for predicting soil organic matter, including:
a processor obtains auxiliary variable data associated with soil organic matter at a detected position in a target area from a memory, wherein the memory stores the auxiliary variable data at various positions in the target area;
the processor calculates the predicted value of the soil organic matter of the measured position by taking the auxiliary variable data of the measured position as the input value of the ELM model of the extreme learning machine established in advance.
In alternative embodiments, the auxiliary variables include any combination of the following variables: the system comprises a remote sensing factor, a terrain factor, a climate factor and a soil attribute factor, wherein the auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in the terrain factor, a monthly average temperature and a monthly average precipitation in the climate factor, and a soil pH, a soil total nitrogen, a soil available phosphorus and a soil available potassium in the soil attribute factor.
In an alternative embodiment, the remote sensing factors include band 2 to band 7 of the image and the normalized vegetation index, the terrain factors include elevation, slope, direction of slope, and terrain relief, the climate factors include monthly precipitation and monthly temperature, and the soil property factors include soil pH, soil total nitrogen, soil available phosphorus, and soil available potassium.
In an optional embodiment, the prediction method further comprises the model establishing device establishing an extreme learning machine ELM model, which comprises: acquiring soil organic matter measured values of a plurality of sampling points in the target area; acquiring auxiliary variable data of the plurality of sampling points, which are related to soil organic matters, according to the soil organic matter measured values of the plurality of sampling points; according to the auxiliary variable data of the plurality of sampling points, which are associated with the soil organic matter, and based on the mapping from a high-dimensional space to a low-dimensional space, obtaining the eigenvalue of a matrix formed by the auxiliary variable data and the variance of the eigenvalue, and determining a plurality of principal components when the sum of the variance contribution rates of the eigenvalue is greater than a preset value; and constructing an Extreme Learning Machine (ELM) model by using the data of the plurality of principal components as input values.
In an optional embodiment, the prediction method further comprises: the model establishing device establishes an extreme learning machine common kriging ELMOK model, which comprises the following steps: after an Extreme Learning Machine (ELM) model is constructed, calculating auxiliary variable data, which are associated with soil organic matters, of a plurality of sampling points as input values of the ELM model to obtain soil organic matter predicted values of the plurality of sampling points; calculating a residual value between a soil organic matter actual measurement value and a soil organic matter predicted value of each sampling point in the plurality of sampling points; calculating the residual common kriging interpolation value of each sampling point according to the residual value of each sampling point by the following formula:
wherein, r (x) i ) Is sample point x i The value of the residual error of (a),is a residual ordinary kriging interpolation, lambda i Is the adjustment weight, n is the number of sample points; and determining residual common kriging interpolation of each position in the target area through the residual common kriging interpolation of the plurality of sampling points and storing the residual common kriging interpolation in a memory.
Further, the method comprises the steps that the processor obtains corresponding residual common kriging interpolation values of the measured position from the storage, and the soil organic matter prediction correction value of the measured position is obtained by superposing and summing the soil organic matter prediction value of the measured position and the corresponding residual common kriging interpolation values.
In an alternative embodiment, where the target area is in the range of 60 square kilometers to 1000 square kilometers, the number of sample points ranges from 20 to 600.
According to another aspect of the present invention, there is also provided an apparatus for predicting soil organic matter, including: the storage is used for storing auxiliary variable data which are related to soil organic matters at all positions in the target area; and a processor for retrieving from the memory auxiliary variable data for a located position in the target area; and calculating the auxiliary variable data as an input value of a pre-established Extreme Learning Machine (ELM) model to obtain a predicted value of the soil organic matter of the measured position.
In an alternative embodiment, the auxiliary variables include any combination of the following variables: the system comprises a remote sensing factor, a terrain factor, a climate factor and a soil attribute factor, wherein the auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in the terrain factor, a monthly average temperature and a monthly average precipitation in the climate factor, and a soil pH, a soil alkaline-hydrolyzable nitrogen, a soil available phosphorus and a soil available potassium in the soil attribute factor.
In an optional embodiment, the memory in the apparatus is further configured to store measured values of soil organic matters at a plurality of sampling points in the target area. The equipment also comprises a model establishing device used for establishing an extreme learning machine ELM model, which comprises: acquiring soil organic matter measured values of a plurality of sampling points in the target area from a memory; acquiring auxiliary variable data of the plurality of sampling points, which are related to soil organic matters, according to the measured values of the soil organic matters of the plurality of sampling points; according to auxiliary variable data of a plurality of sampling points and associated with soil organic matters and based on mapping from a high-dimensional space to a low-dimensional space, solving a characteristic value of a matrix formed by the auxiliary variable data and a variance of the characteristic value, and determining a plurality of principal components when the sum of variance contribution rates of the characteristic value is greater than a preset value; and constructing an Extreme Learning Machine (ELM) model by using the data of the plurality of principal components as input values.
In an alternative embodiment, the model building device is further configured to build an extreme learning machine ordinary kriging ELMOK model, which includes: after an Extreme Learning Machine (ELM) model is constructed, calculating the auxiliary variable data, which are associated with soil organic matters, of the plurality of sampling points according to the input value of the ELM model to obtain the predicted values of the soil organic matters of the plurality of sampling points; calculating a residual value between a soil organic matter measured value and a soil organic matter predicted value of each sampling point in the plurality of sampling points; calculating the residual common kriging interpolation value of each sampling point according to the residual value of each sampling point and the following formula
Wherein, r (x) i ) Is sample point x i The value of the residual error of (a),is a residual ordinary kriging interpolation, lambda i Is the adjustment weight, n is the number of sampling points;
and determining residual common kriging interpolation of each position in the target area through the residual common kriging interpolation of the plurality of sampling points and storing the residual common kriging interpolation in the memory.
In an optional embodiment, the processor is further configured to obtain a corresponding residual common kriging interpolation of the measured position from the memory, and add and sum the predicted soil organic matter value of the measured position and the corresponding residual common kriging interpolation to obtain a corrected soil organic matter prediction value of the measured position.
By the soil organic matter prediction method provided by the embodiment of the invention, the problems of poor nonlinear fitting, complex model parameters and low model performance in the conventional nonlinear algorithm can be solved, and the prediction precision is improved.
Drawings
Fig. 1 is a flowchart of a method for predicting soil organic matter according to an embodiment of the present invention.
Fig. 2 is a flowchart of a method for predicting organic matter of soil according to another embodiment of the present invention.
Fig. 3 is a logical structure diagram of an ELM model according to an embodiment of the present invention.
Fig. 4 is a schematic block diagram of an apparatus for performing soil organic matter prediction according to an embodiment of the present invention.
Detailed Description
The following disclosure provides many different embodiments, or examples, for implementing different features of the invention. Specific examples of components and method steps are described below to simplify the present disclosure. Of course, these are merely examples and are not intended to be limiting. It should be appreciated that the subject matter disclosed herein is capable of being applied in a wide variety of forms and that any specific structure and/or function disclosed herein is merely illustrative. Based on the teachings herein one skilled in the art should appreciate that an aspect disclosed herein may be implemented independently of any other aspects and that two or more of these aspects may be combined in various ways.
The selection of the soil organic matter space prediction auxiliary variable usually influences the model fitting effect and is more important for the feedback interpretation capability of the driving force of the model result. At present, most of soil organic matter space mapping auxiliary variables selected at home and abroad are environmental factors, such as climate, terrain and the like, which are very convenient to obtain [5～7] . However, due to the limited spatial variation of the environmental factors in the regional scale and the weak nonlinear interpretation capability, higher accuracy in the prediction of the organic matter content of the soil is difficult to achieve. In recent years, remote sensing waveband data is also used as an auxiliary factor for soil organic matter prediction and achieves a certain prediction effect, but the remote sensing factor is difficult to better explain the relation mechanism of organic matter, global change and carbon sink [8] . Therefore, it is necessary to select multispectral remote sensing elements, environmental elements and soil attribute elements (such as total nitrogen, available phosphorus, quick-acting potassium and the like) closely related to soil organic matters in the prediction model as multisource auxiliary variables for predicting the soil organic matter space. However, in the prediction of the spatial distribution of the organic matter in the soil, the application of simultaneously considering environmental, remote sensing and soil attribute elements as auxiliary variables is very few.
Fig. 1 is a flowchart of a method for predicting soil organic matter according to an embodiment of the present invention. As shown in fig. 1, the prediction method includes:
s101, the model building device obtains soil organic matter measured values of a plurality of sampling points in a target area from a memory.
And S102, the model establishing device acquires auxiliary variable data associated with the soil organic matters of the plurality of sampling points from the memory according to the soil organic matter measured values of the plurality of sampling points.
In an embodiment of the invention, the auxiliary variables comprise any combination of the following variables: remote sensing factor, terrain factor, climate factor and soil property factor. Preferably, the auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in the terrain factor, a monthly average temperature and a monthly average precipitation in the climate factor, and a soil pH, a soil total nitrogen, a soil available phosphorus and a soil available potassium in the soil attribute factor.
The memory stores the soil organic matter measured value of each sampling point in the target area and auxiliary variable data related to the soil organic matter at each position in the target area.
And S103, determining a plurality of main components for building the ELM by the model building device according to the auxiliary variable data of the plurality of sampling points and associated with the soil organic matter.
Specifically, the model establishing device obtains the eigenvalue of a matrix formed by the auxiliary variable data and the variance of the eigenvalue based on the mapping from the high-dimensional space to the low-dimensional space, and determines a plurality of principal components for establishing the ELM model when the sum of the variance contribution rates of the eigenvalue is greater than a preset value. In alternative embodiments, the number of auxiliary variables associated with soil organic matter may be 5-25, and 2-10 principal components may be determined based on the 5-25 auxiliary variables.
And S104, the model building device takes the data of the main components as input values to build an Extreme Learning Machine (ELM) model.
And S105, the processor acquires auxiliary variable data which is related to soil organic matters and is positioned in the target area from the memory.
And S106, the processor calculates the predicted value of the soil organic matter of the measured position by using the auxiliary variable data of the measured position as the input value of the ELM model.
In other embodiments of the invention, after the extreme learning machine ELM model is constructed, the soil organic matter predicted values of the multiple sampling points can be obtained by calculating the auxiliary variable data associated with the soil organic matters of the multiple sampling points as the input values of the ELM model. Then, calculating a residual value between a soil organic matter actual measurement value and a soil organic matter predicted value of each sampling point in the plurality of sampling points, and calculating a residual common kriging interpolation value of each sampling point according to the residual value of each sampling point and the following formula
Wherein, r (x) i ) Is the sampling point x i The residual value of (a) is determined,is a residual ordinary kriging interpolation, lambda i Is the adjustment weight and n is the number of sample points. In this way, the residual common kriging interpolation of each position in the target area can be determined through the residual common kriging interpolation of a plurality of sampling points and stored in the memory.
Fig. 2 is a flowchart of a method for predicting soil organic matter according to another embodiment of the present invention. As shown in fig. 2, the prediction method includes:
s201, the processor acquires auxiliary variable data related to soil organic matters of a measured position in the target area from the memory.
In an embodiment of the invention, the auxiliary variables comprise any combination of the following variables: remote sensing factors, terrain factors, climate factors and soil attribute factors. Preferably, the auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in the terrain factor, a monthly average temperature and a monthly average precipitation in the climate factor, and a soil pH, a soil total nitrogen, a soil available phosphorus and a soil available potassium in the soil attribute factor.
And S202, the processor calculates the auxiliary variable data as the input value of a pre-established extreme learning machine ELM model to obtain the predicted value of the soil organic matter of the measured position.
S203, the processor obtains the corresponding residual error common Kriging interpolation of the measured position from the memory.
And S204, the processor superposes and sums the predicted value of the soil organic matter at the measured position and the corresponding residual common kriging interpolation of the measured position to obtain a predicted corrected value of the soil organic matter at the measured position.
Embodiments of the soil organic matter prediction method of the present disclosure are described in further detail below with reference to fig. 1 to 3.
1. Determining auxiliary variable data required for predicting soil organic matter
1.1 measuring the content of organic matters in soil in a target area
According to an embodiment of the invention, the area is about 450km 2 The method comprises the steps of collecting a plurality of soil surface layer (0-20 cm) samples in a target area, simultaneously recording related environment information such as GPS geographic coordinates, soil types, planting years, crop types and terrains, and then measuring the organic matter content of the soil by adopting an oil bath heating potassium dichromate oxidation-volumetric method.
1.2 determining auxiliary variables associated with soil organic matter prediction of a target area
According to the embodiment of the invention, landsat 7ETM + remote sensing image band data is selected as a remote sensing auxiliary factor. For the purpose of illustration, this embodiment selects Landsat 7ETM + remote sensing image band data of 10 months in 2009 as a remote sensing auxiliary factor, where the remote sensing image data is from the National aerospace administration (NASA) and has a resolution of 30m (where the eighth band is 15 m). After the ETM + wave band data are subjected to radiation correction and atmospheric correction in ENVI5.2, 6 wave bands (wave bands 2-7) are selected, and Normalized Difference Vegetation Indexes (NDVI) are calculated and are jointly used as remote sensing auxiliary variables for soil organic matter prediction. Optionally, the Landsat 8OLI remote sensing image band data can also be selected as a remote sensing auxiliary factor.
According to the embodiment of the invention, the terrain factor can select an ASTER GDEMV2 global digital elevation model provided by geospatial data cloud of Chinese academy as an elevation factor, and the resolution is 30m. In ArcGISI 10.3, a 3D analysis tool is used for respectively calculating the slope, the slope direction and the topographic relief degree which are used as topographic auxiliary factors for predicting soil organic matters.
The climate factor can also be selected from MODLT1M China 1KM earth surface temperature monthly synthesis products provided by Chinese academy geospatial data cloud, and the monthly average temperature and monthly average precipitation are calculated and converted into the spatial resolution consistent with DEM in ArcGISI 10.3 through resampling.
The soil attribute factors include soil pH, soil total nitrogen, soil available phosphorus and soil available potassium. Wherein the pH of the soil is measured by a sleeve type calomel electrode, the total nitrogen of the soil is measured by a semi-micro Karl method, and the available phosphorus and the available potassium of the soil are respectively measured by 0.5 mol.L -1 NaHCO 3 And 1 mol. L -1 NH 4 Flame photometry of OAc. The above point factor data is subjected to Inverse Distance Weighted Interpolation (IDW) to obtain spatial distribution data, and the conversion resolution is 90m.
2. Building an ELM model
2.1 descriptive statistical analysis:
for the area of about 450km 2 The soil organic matter and 17 auxiliary variables of the sampling points in the target area are subjected to descriptive statistics, and after abnormal values are removed by taking plus and minus 3-time standard deviation as a reference, 522 sampling point values are counted in total. The inventor finds that if the number of sampling points in a certain area is too small, the prediction accuracy of the constructed ELM model is not good, namely the prediction value of soil organic matters is not accurate; if the number of the sampling points is too large, the calculation amount for constructing the ELM model is obviously increased, but the prediction precision is not improved much, so that the selection of the proper number of the sampling points can ensure that the prediction precision of the ELM model is better and simultaneously reduce unnecessary calculation amount and calculation complexity. In a preferred embodiment, the area is 100km 2 The number of sampling points is 20-100; in another preferred embodiment, the area is 500km 2 Target area of (1), sampling pointThe number of the active ingredients is 100-600; in a further preferred embodiment, the area is 1000km 2 The number of the sampling points in the target area is 600-1000.
In the embodiment of the invention, as shown in Table 1, the content of the soil organic matters in the target area is 6.00-51.00 g kg through the soil organic matter values of 522 sampling points -1 Mean value of 24.74g kg -1 The Kolmogorov-Smirnov (K-S) test of SPSS22.0 software shows that the P value of the test result (namely the probability of the sample conforming to normal distribution, and the default is that P is the value&And when 0.05, the test result of the original sample conforms to normal distribution) is 0.322, which conforms to normal distribution. As shown in Table 1, among 17 auxiliary variables of 522 sampling points, the variation coefficients of the gradient, the slope direction, the topographic relief degree, the soil total nitrogen, the soil available potassium and the soil available phosphorus are large in difference, and the variation coefficients from the wave band 2 to the wave band 7, NDVI, elevation, monthly mean temperature, soil pH and monthly mean precipitation are low. In general, the variation degree of the terrain factors and the soil attribute factors is higher, the remote sensing auxiliary factors are next to the variation degree of the climate factors, and the analysis reason of the variation degree of the climate factors is probably related to factors such as various land coverage types and integrally higher soil acidity in the research area.
TABLE 1 descriptive statistics of sampling point soil organic matter and auxiliary variables
In the embodiment of the invention, descriptive statistics is performed on the measured values of the soil organic matters of all sampling points in the target area and the auxiliary variable data related to the soil organic matters at each position (including the sampling points) in the target area so as to judge the availability and the correlation of the auxiliary variable and the soil organic matters. Specifically, as shown in table 1, the abnormal value can be found and eliminated by statistical indexes such as the maximum value, the minimum value, the variance, and the coefficient of variation. By preliminarily analyzing the reason of larger variation degree of the auxiliary variables and removing abnormal values, the interference of numerical value abnormality among the variables on subsequent principal component analysis can be avoided.
And storing the measured value of the soil organic matter of the sampling point obtained after the descriptive statistical analysis and the auxiliary variable data related to the soil organic matter into a memory. In the embodiment of the present invention, after the auxiliary variable data associated with the soil organic matter of the sampling points in the target area is determined, the auxiliary variable data associated with the soil organic matter of other non-sampling position points may be further determined, for example, the research area is divided into a 90m × 90m grid (the area of the research area may be divided into about 55556 grids by 450 square kilometers), 17 pieces of auxiliary variable data (such as planar image data of remote sensing factors, elevations, and the like) are unified into a grid with a size of 90m in the ArcGIS software, and then the auxiliary variable data corresponding to the grid of the non-sampling position points except 522 sampling points (about 55034 grids) is obtained. The grid about 90m and about 90m is selected, so that each sampling point can fall into the corresponding grid as far as possible, the situation that two sampling points exist in one grid can be avoided, and the calculation efficiency can be improved, because the smaller the grid is, the calculation amount is exponentially increased.
2.2 principal component analysis dimensionality reduction treatment:
in the embodiment of the invention, after the abnormal value elimination of the descriptive statistics, the Principal Component Analysis (PCA) can be adopted to reduce the phenomenon of multiple collinearity among variables and simplify the structure of the model. Jolliffe I T. Primary Component Analysis [ M ]. Springer-Verlag,1986. In addition, eight principal components are determined according to the auxiliary variable data and the variance contribution rate, and the information of the original auxiliary variables can be well retained. The main component analysis method is adopted because the inventor finds out in the research that: if high linear correlation exists among the variables, the extracted principal components are excessively concentrated on the variables linearly correlated with the variables, the comprehensive extraction of the original variables is interfered, and the total contribution rate of the extracted principal components is low and the original variable information cannot be well reflected.
In the embodiment of the invention, the principal component analysis is carried out on 17 multisource auxiliary variables, firstly, 17 auxiliary variable original data are arranged in rows (one auxiliary variable data is one row) to form a new matrix X, and the matrix X is standardized to enable the mean value of the matrix X to become zero; then, solving the covariance of the matrix to obtain k eigenvectors, arranging eigenvalues corresponding to the eigenvectors from large to small, and forming a matrix H by the k eigenvectors according to rows; and finally, calculating according to a formula Y = HX to obtain dimension-reduced data Y and obtain a corresponding variance contribution rate. The principal component with variance total contribution rate (the sum of the variance total contribution rates of each eigenvalue, namely the total contribution rate) larger than 90% is usually adopted, and the original information of the multi-source auxiliary variable can be well reserved. In the invention, 17 auxiliary variables are automatically completed in the SPSS22.0 software to perform principal component analysis, as shown in table 2, the principal components are accumulated one by one according to variance contribution rate, 8 principal components are obtained by taking the total variance contribution rate of more than 90%, and the interpretation degree of the original variables reaches 91.46%. Wherein, the first principal component PC1 explains 30.86% of the total variable, and the correlation of the wave band 2, the wave band 3, the wave band 4, the wave band 5 and the wave band 7 is the highest, while the NDVI has higher correlation in the fifth principal component, and the monthly average precipitation and the monthly average temperature, which explains 6.41% of the total variable; the other 6 principal components (PC 2=18.72%, PC3=11.79%, PC4=8.97%, PC6=5.52%, PC7=5.09%, and PC8= 4.10%) reflect mainly the close association of environmental auxiliary variables such as soil properties, climate, terrain, and the like. The resulting 8 principal components can be used as input values to build an ELM model.
TABLE 2 principal Components analysis results
3. ELM model modeling
And constructing an ELM model based on the processed multiple main component data to perform spatial prediction on soil organic matters in the target area.
The Extreme Learning Machine (ELM) algorithm is a Single-Hidden-Layer feed forward neural network (SLFNs) algorithm composed of an input Layer, a Hidden Layer, and an output Layer, which is proposed in 2004 by huang guang of southern university of science and technology, singapore. Huang G B, zhu Q Y, sine C K. Extreme learning machine a new learning scheme of fed forward neural networks [ J ]. Proc. Int. Joint Conf. Neural network, 2004, 2. 985-990. The inventors of the present disclosure found in the study: the ELM algorithm belongs to a neural network algorithm system essentially, but has larger difference compared with the traditional artificial neural network in the algorithm information transmission process, firstly, the ELM algorithm can randomly set the weight and the threshold between an input layer and a hidden layer in the algorithm execution process, and the input weight of the network and the bias of the hidden element do not need to be adjusted and iterated repeatedly, and finally, a unique optimal solution is generated, so that the condition that the artificial neural network algorithm is easy to fall into a local optimal solution is avoided, and the ELM algorithm has ultra-fast learning speed and strong generalization.
The generalized SLFNs output function of ELM, namely formula (1):
where x is the input data of the model, f L (x) Is the output of the model, β = [ β ] 1 ,…,β L ] T Is a set of weight vectors between L nodes and m output nodes (m ≧ 1), h (x) = [ h 1 (x),…,h L (x)]Is a non-linear feature mapping of the ELM hidden layer output row vector, h i (x) Is the output function of the ith hidden layer node, which contains the usual activation function. FIG. 3 shows a logical structure diagram of an ELM model, where the output functions of the hidden layer nodes can be very diverse, with different output functions for different hidden layer neurons. The ELM algorithm finally deduces an optimal weight solution paradigm, namely as shown in formula (2):
wherein beta is the optimal weight value between the hidden layer and the output layer of the ELM model,is Moore-Penrose generalized inverse matrix H, which is shown in formula (3):
t is the expected output matrix of the training data, which is shown in equation (4):
the ELM algorithm derivation process can be known to randomly generate the weight and the threshold between the hidden layer and the input layer according to the continuous probability distribution, and the weight between the hidden layer and the output layer at the second stage is converted into a linear equation set without subsequent adjustment, so that the ELM algorithm derivation process has higher learning efficiency and better generalization performance compared with the traditional artificial neural network.
The single hidden layer structure of the ELM model structure is different from the traditional linear equation expression, belongs to a 'black box' model, and is hidden in the calculation process of model prediction, and the calculation process of the ELM model can be realized in Matlab2013b software through programming.
Through programming in Matlab2013b software, 8 main components in the table 2 are used as input of an ELM model, and a predicted value of the soil organic matter is obtained through calculation in a formula (1) and a black box process in the figure 3. As shown in Table 3, the final structure of the ELM prediction model is 8-32-1 (8 principal component inputs, i.e., x) 1 ＝PC1,x 2 ＝PC2,…,x 8 = PC8, 32 hidden layer nodes, i.e. L =32, and 1 output node, i.e. soil organic mass SOM preMeasured value), the activation function is a sigmoid kernel.
To evaluate the prediction accuracy and performance of the ELM model, mean Error (ME), equation (5), root Mean Square Error (RMSE), equation (6), and correlation coefficient (R) are selected 2 ) Namely, the formula (7) is used as the criterion.
In the formula, Z (x) i ) Is a measured value of the organic matter of the soil,is a soil organic matter prediction value of the ELM model,is the average value of the measured values of the organic matters in the soil. Where the closer ME and RMSE are to zero, R 2 A closer to 1 indicates a higher prediction accuracy of the model.
Calculating the precision error of the ELM model according to the formula (5), the formula (6) and the formula (7) to obtain that the ME and the RMSE of the prediction of the soil organic matter are respectively 7.92g kg -1 And 2.012g kg -1 While simultaneously interpreting the non-linear interpretive power R between the target content value and the auxiliary variable 2 Reaching 55.1 percent. By contrast, the accuracy error of the artificial neural network ANN prediction model can also be calculated according to the above equations (5), (6) and (7), and as a result, the ELM model soil organic matter prediction has lower ME and RMSE than the ANN prediction model.
In addition, the calculation time of the ELM prediction model is greatly faster than that of the traditional ANN model due to the advantages of random assignment of initial weights, no need of weight adjustment, no need of iteration and other algorithm parameters in the ELM model.
TABLE 3 optimal ELM soil organic matter prediction model parameters and error statistics
4. Analysis of influence of different auxiliary variables on soil organic matter prediction
In order to further explore the influence of 17 auxiliary variables on soil organic matter prediction, an original remote sensing type factor, an original environment factor and an original soil attribute factor are directly used as input of an ELM (element-model) model to predict the content of soil organic matter. In Matlab2013b software, 17 original auxiliary variables in the table 1 are used as the input of an ELM model through programming, and the accuracy of soil organic matter prediction is obtained through calculation in a formula (1) and a black box process in the figure 3. The accuracy sequence of each factor on soil organic matter prediction is obtained through gradual variable addition, and as shown in table 4, the predicted 17 auxiliary variables of the ELM model sequentially have the following influence on the relationship of the soil organic matter from large to small: the effective phosphorus in the soil, the quick-acting potassium in the soil, the monthly temperature, the monthly precipitation, the total nitrogen in the soil, the pH value of the soil, the wave band 5, NDVI, the wave band 4, the slope direction, the gradient, the topographic relief degree, the wave band 2, the wave band 3, the wave band 7 and the wave band 6. Therefore, the 17 auxiliary variables comprise 9 auxiliary variables such as soil available phosphorus, soil available potassium, monthly average temperature, monthly average precipitation, soil total nitrogen, elevation, soil pH, wave band 5, NDVI and the like according to factors with larger influence on the prediction of the soil organic matters, the driving effects of the variables with different sizes jointly restrict the quantitative effect characteristics of the soil organic matters, and the factors are heavily considered in the actual auxiliary factor selection process.
TABLE 4 precision statistics of soil organic matter prediction with different auxiliary variables
In order to further improve the prediction precision of the soil organic matter, the embodiment of the invention also provides an embodiment of a soil organic matter prediction method based on the ordinary kriging ELMOK model of the extreme learning machine.
The common kriging interpolation is a spatial local interpolation method of geostatistics, the solving result cannot be expressed by a clear mathematical equation, and the spatial prediction process of soil organic matters in a research area can be completed by a geostatistical analysis module of ArcGIS10.3 software.
The method for predicting the content of the organic matters in the soil by adopting the ELMOK model comprises the following three steps:
first, an ELM model between soil organic matter and extracted principal components was constructed as described above. For example, the ELM model can be constructed by programming in matlab2013b software. Then, calculating the auxiliary variable data of the sampling point as an input value of the ELM based on the ELM to obtain a predicted value of the soil organic matter of the sampling point;
then, respectively calculating a residual error value and a residual error common kriging value between the soil organic matter measured value and the soil organic matter predicted value based on the ELM model according to a formula (8) and a formula (9);
wherein:
in formula (8), r (x) i ) Is sample point x i The residual value of (d); z (x) i ) Is the sampling point x i Measuring the soil organic matter;is based on the sampling point x of ELM model i The soil organic matter prediction value is obtained.
In formula (9), r (x) i ) Is the sampling point x i The residual value of (a) is determined,is the sampling point x i Residual ordinary kriging interpolation of (a) (-) i Is adjusting the weight, n is sampling point x i The number of known sample points present in a circle with a radius of, for example, 2.04km from the center. Specifically, it is generally assumed that a circle is formed by setting a radius h of, for example, 2.04km (the radius can be determined in a geostatistical analysis module of the arcgis10.3 software according to the actual situation of a research area) with a sampling point x (the actual measured value of the soil organic matter of the sampling point) with an unknown soil organic matter prediction value as a center, n sampling points with known soil organic matter prediction values exist within the range of the circle, and a first-order or multi-order linear equation can be established by using the n known sampling points and the unknown sampling point x. A group of adjusting weights lambda can be obtained through the soil organic matter measured value of the unknown sampling point x, and then the weights are used for predicting the soil organic matter content of other unknown sampling points in the target area. The processes of weight adjustment optimization, function model selection and the like can be automatically completed by a geostatistical analysis module of ArcGIS10.3 software by ordinary technicians in the field.
And then, summing the soil organic matter predicted value based on the ELM model and the residual common Kriging result according to a formula (10) to finally obtain the soil organic matter correction predicted value of the sampling point:
in the formula (10), the reaction mixture is,is a sampling point x based on an ELM mixed ground statistical model i Correcting and predicting the soil organic matter;is the sampling point x i The residual common kriging interpolation;is based on the sampling point x of ELM model i The soil organic matter prediction value is obtained.
In the embodiment of the present invention, the soil organic matter prediction accuracy and performance of the extreme learning machine common kriging (ELMOK) model proposed in the embodiment of the present invention, and the common kriging (OK) model, the multiple linear regression common kriging (MLROK) model, and the artificial neural network common kriging (ANNOK) model as the comparison method can also be evaluated according to the formulas (5), (6), and (7).
As can be seen from Table 5, the ELMOK model has a lower RMSE (1.801 g kg) than the conventional common kriging and other mixed statistical models -2 ) And higher R 2 (0.662), indicating that it has a more accurate prediction level. From the comparison of the predicted value and the actual value range of the soil organic matter of the 4 models, the difference between the predicted result and the actual value range of the common Kriging method is the largest, the critical jump of the organic matter content distribution in the local area is larger, the overall predicted effect and the actual value of the other 3 models are closer, the predicted result range and the actual value of the soil organic matter of the ELMOK model are the closest, the overall predicted effect is better, and the nonlinear analysis capability on the soil organic matter content is deeper.
TABLE 5 statistical results of the accuracy of soil organic matter prediction under different models
In the embodiment of the present invention, 522 sampling points may be divided into 380 training samples and 142 testing samples, wherein the soil organic matter measured value and the soil organic matter associated auxiliary variable data of the 380 training samples are used to construct the ELM model and the ELMOK model, and the soil organic matter measured value and the soil organic matter associated auxiliary variable data of the 142 testing samples are used to verify the prediction accuracy of the ELM model or the ELMOK model.
The existing soil organic matter space prediction method mainly comprises three types of classical statistical models such as a multiple linear regression model, geostatistical models such as a simple kriging method and a common kriging method, and mixed geostatistical models such as a regression kriging model and a neural network kriging model. The regression kriging model, the neural network kriging model and the mixed geostatistics model are a more advanced technical scheme in the field of soil organic matter prediction at present by combining high fitting precision of a nonlinear machine learning algorithm and high efficiency of kriging interpolation.
However, the conventional geostatistics method has limited prediction accuracy on the spatial distribution of soil organic matters, for example, although a common kriging model has a simple structure and is easy to operate, the problem of low fitting degree caused by a complex nonlinear relationship between soil organic matters and auxiliary variables is difficult to solve. In addition, the common mixed-land statistical method such as the regression kriging method still has the problem of low prediction precision due to weak resolving power of the nonlinear relation between soil organic matters and associated variables, although the artificial neural network kriging method effectively improves the fitting effect, complex parameter adjustment is required, and the required operation time is increased when the drawing resolution is improved, so that the model execution efficiency is low.
Compared with the prior art, the novel prediction method for excavating regional scale soil organic matter space distribution by adopting the common kriging combination model of the extreme learning machine and combining multi-source auxiliary variables such as the environmental factor, the remote sensing factor, the soil attribute and the like has the following advantages and beneficial effects:
(1) Compared with the conventional Kriging (OK) model, the Multiple Linear Regression conventional Kriging (MLROK) model and the Artificial Neural Network conventional Kriging (ANNOK) model for predicting the content of organic matters in soil, the Extreme Learning Machine conventional Kriging (ELMOK) model has lower root mean square error(RMSE＝1.801g kg -2 ) And higher correlation coefficient (R) 2 = 0.662), has stronger capturing and resolving power on the nonlinear relation between the organic matter content of the soil in the research area and the multisource auxiliary variables, and has higher model fitting precision and better stability;
(2) Compared with an artificial neural network common kriging model, the extreme learning machine common kriging model has the advantages of simple parameter adjustment, difficulty in falling into local optimal solution, simple model structure, higher execution efficiency and great convenience for soil organic matter prediction and soil fertility evaluation.
The device or apparatus of the addressing method for updating and transforming the city according to the embodiment of the invention may be a computer, or an entity related to the computer, etc. The apparatus may further comprise a processor configured to perform actions in a method of predicting organic matter in soil as illustrated in fig. 1 or fig. 2. As shown in fig. 4, an apparatus 40 for predicting organic matter of soil according to one embodiment of the present invention includes: a memory 42 for storing auxiliary variable data associated with soil organic matter at each location in the target area; and a processor 44 for retrieving from the memory 42 auxiliary variable data for the located position in the target area; and calculating the predicted value of the soil organic matter of the measured position by using the auxiliary variable data as the input value of a pre-established extreme learning machine ELM model.
In embodiments of the invention, the auxiliary variables comprise any combination of the following variables: remote sensing factor, terrain factor, climate factor and soil property factor. The auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in a terrain factor, a monthly average temperature and a monthly average precipitation in a climate factor, and soil pH, soil alkaline hydrolysis nitrogen, soil available phosphorus and soil available potassium in a soil attribute factor.
In an optional embodiment, the memory 42 is further configured to store measured soil organic matter values of a plurality of sampling points in the target area. The apparatus 40 further comprises model building means 46 for building an extreme learning machine ELM model. The model building device 46 includes obtaining measured values of soil organic matters of a plurality of sampling points in the target area from the memory 42; acquiring auxiliary variable data, associated with the soil organic matter, of the multiple sampling points from a memory according to the soil organic matter measured values of the multiple sampling points; and according to auxiliary variable data of the plurality of sampling points, which are associated with the soil organic matter, and based on the mapping from the high-dimensional space to the low-dimensional space, solving a characteristic value of a matrix formed by the auxiliary variable data and the variance of the characteristic value, and determining a plurality of principal components when the sum of variance contribution rates of the characteristic value is greater than a preset value. Further, the model building means 46 builds the extreme learning machine ELM model using the data of the plurality of principal components as input values.
In another embodiment, the model building means 46 is also used to build an extreme learning machine common kriging ELMOK model. Specifically, after the extreme learning machine ELM model is constructed, the model establishing device 46 calculates the predicted values of the soil organic matter of the multiple sampling points by using the auxiliary variable data of the multiple sampling points, which are associated with the soil organic matter, as the input values of the ELM model, and then calculates the residual values between the actual measured values of the soil organic matter and the predicted values of the soil organic matter of the multiple sampling points; and calculating the residual common kriging interpolation of each sampling point according to the residual value of each sampling point and the following formula
Wherein, r (x) i ) Is the sampling point x i The residual value of (a) is determined,is residual ordinary kriging interpolation, lambda i Is the adjustment weight and n is the number of sample points.
Finally, the processor 44 determines the residual common kriging interpolation value of each position in the target area through the residual common kriging interpolation values of the plurality of sampling points and stores the residual common kriging interpolation value in the memory 42.
In this embodiment of the present invention, the processor 44 is further configured to obtain a corresponding residual common kriging interpolation of the measured location from the memory 42, and add and sum the predicted soil organic matter value of the measured location and the corresponding residual common kriging interpolation to obtain a predicted soil organic matter correction value of the measured location.
The method for predicting the soil organic matter in each position of the target area by the equipment of the embodiment of the invention can further refer to the description of the corresponding steps in the soil organic matter prediction method in the foregoing disclosure, and details are not repeated here. It is to be understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For a hardware implementation, the processor may be implemented in one or more of the following units: an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a processor, a controller, a microcontroller, a microprocessor, other electronic units designed to perform the functions described herein, or a combination thereof.
The foregoing has outlined features of several embodiments so that those skilled in the art may better understand the aspects of the present disclosure. Those skilled in the art should appreciate that they may readily use the present disclosure as a basis for designing or modifying other processes and structures for carrying out the same purposes and/or achieving the same advantages of the embodiments introduced herein. Those skilled in the art should also realize that such equivalent constructions do not depart from the spirit and scope of the present disclosure, and that they may make various changes, substitutions and alterations herein without departing from the spirit and scope of the present disclosure.
Claims (10)
1. A prediction method of soil organic matter comprises the following steps:
the processor acquires auxiliary variable data associated with soil organic matter at a detected position in a target area from a memory, wherein the memory stores the auxiliary variable data of each position in the target area;
and the processor calculates the predicted value of the soil organic matter of the measured position by taking the auxiliary variable data of the measured position as an input value of a pre-established extreme learning machine ELM model.
2. The prediction method of claim 1, wherein the auxiliary variables comprise any combination of the following variables: the system comprises a remote sensing factor, a terrain factor, a climate factor and a soil attribute factor, wherein the auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in the terrain factor, a monthly average temperature and a monthly average precipitation in the climate factor, and a soil pH, a soil total nitrogen, a soil available phosphorus and a soil available potassium in the soil attribute factor.
3. The prediction method of claim 2, wherein the remote sensing factors comprise band 2 to band 7 of the image and a normalized vegetation index, the terrain factors comprise elevation, slope and relief, the climate factors comprise monthly precipitation and monthly temperature, and the soil property factors comprise soil pH, soil total nitrogen, soil available phosphorus and soil available potassium.
4. The prediction method according to claim 1, further comprising a model building means building the extreme learning machine ELM model, which comprises:
acquiring soil organic matter measured values of a plurality of sampling points in the target area;
acquiring auxiliary variable data of the plurality of sampling points, which are related to soil organic matters, according to the soil organic matter measured values of the plurality of sampling points;
according to the auxiliary variable data of the plurality of sampling points, which are associated with the soil organic matter, and based on the mapping from a high-dimensional space to a low-dimensional space, solving a characteristic value of a matrix formed by the auxiliary variable data and the variance of the characteristic value, and when the sum of the variance contribution rates of the characteristic value is greater than a preset value, determining a plurality of principal components;
and constructing an extreme learning machine ELM model by using the data of the plurality of principal components as input values.
5. The prediction method according to claim 4, further comprising:
the model establishing device establishes an extreme learning machine common kriging ELMOK model which comprises the following steps of;
after an Extreme Learning Machine (ELM) model is built, calculating the auxiliary variable data of the plurality of sampling points, which are associated with the soil organic matter, as the input values of the ELM model to obtain the predicted values of the soil organic matter of the plurality of sampling points;
calculating a residual value between the soil organic matter actual measurement value and the soil organic matter predicted value of each sampling point in the plurality of sampling points;
calculating the residual common kriging interpolation value of each sampling point according to the residual value of each sampling point and the following formula
Wherein, r (x) i ) Is the sampling point x i The value of the residual error of (a),is residual ordinary kriging interpolation, lambda i Is the adjustment weight, n is the number of sample points;
determining residual common kriging interpolation values of all positions in the target area through the residual common kriging interpolation values of the plurality of sampling points and storing the residual common kriging interpolation values in the memory;
and the processor acquires corresponding residual common kriging interpolation of the measured position from the memory, and superposes and sums the soil organic matter predicted value of the measured position and the corresponding residual common kriging interpolation to obtain the soil organic matter prediction correction value of the measured position.
6. The prediction method according to claim 4 or claim, wherein the number of sample points ranges from 20 to 600 when the target area is in the range of 60 to 1000 square kilometers in area.
7. An apparatus for predicting soil organic matter, comprising:
the storage is used for storing auxiliary variable data which are related to soil organic matters at all positions in the target area;
a processor for retrieving from the memory auxiliary variable data for a located position in the target area; and calculating the predicted value of the soil organic matter of the measured position by taking the auxiliary variable data as the input value of a pre-established extreme learning machine ELM model.
8. The apparatus of claim 7, wherein the auxiliary variables comprise any combination of the following variables: the system comprises a remote sensing factor, a terrain factor, a climate factor and a soil attribute factor, wherein the auxiliary variables at least comprise an image waveband 5 and a normalized vegetation index in the remote sensing factor, an elevation in the terrain factor, a monthly average temperature and a monthly average precipitation in the climate factor, and a soil pH, a soil alkaline-hydrolyzable nitrogen, a soil available phosphorus and a soil available potassium in the soil attribute factor.
9. The apparatus of claim 7, wherein: the memory is also used for storing measured values of soil organic matters of a plurality of sampling points in the target area;
the equipment also comprises a model establishing device used for establishing an extreme learning machine ELM model, which comprises the following steps:
acquiring soil organic matter measured values of a plurality of sampling points in the target area from the memory;
acquiring auxiliary variable data of the plurality of sampling points, which are related to soil organic matters, according to the soil organic matter measured values of the plurality of sampling points;
according to the auxiliary variable data of the plurality of sampling points, which are associated with the soil organic matter, and based on the mapping from a high-dimensional space to a low-dimensional space, obtaining the eigenvalue of a matrix formed by the auxiliary variable data and the variance of the eigenvalue, and determining a plurality of principal components when the sum of the variance contribution rates of the eigenvalue is greater than a preset value;
and constructing an extreme learning machine ELM model by using the data of the plurality of principal components as input values.
10. The apparatus of claim 9, wherein:
the model establishing device is also used for establishing an extreme learning machine common kriging ELMOK model, which comprises the following steps of;
after an Extreme Learning Machine (ELM) model is constructed, calculating auxiliary variable data, associated with soil organic matters, of the multiple sampling points as input values of the ELM model to obtain predicted values of the soil organic matters of the multiple sampling points;
calculating a residual value between a soil organic matter actual measurement value and a soil organic matter predicted value of each sampling point in the plurality of sampling points;
calculating the residual common kriging interpolation value of each sampling point according to the residual value of each sampling point and the following formula
Wherein, r (x) i ) Is sample point x i The residual value of (a) is determined,is a residual ordinary kriging interpolation, lambda i Is the adjustment weight, n is the number of sampling points;
determining residual common kriging interpolation values of all positions in the target area through the residual common kriging interpolation values of the plurality of sampling points and storing the residual common kriging interpolation values in the memory;
the processor is further configured to obtain a corresponding residual common kriging interpolation of the measured position from the memory, and add and sum the predicted value of the soil organic matter of the measured position and the corresponding residual common kriging interpolation to obtain a predicted correction value of the soil organic matter of the measured position.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710865008.8A CN107860889A (en) | 2017-09-22 | 2017-09-22 | The Forecasting Methodology and equipment of the soil organism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710865008.8A CN107860889A (en) | 2017-09-22 | 2017-09-22 | The Forecasting Methodology and equipment of the soil organism |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107860889A true CN107860889A (en) | 2018-03-30 |
Family
ID=61698320
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710865008.8A Pending CN107860889A (en) | 2017-09-22 | 2017-09-22 | The Forecasting Methodology and equipment of the soil organism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107860889A (en) |
Cited By (9)
| 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 |
| CN109086823A (en) * | 2018-08-01 | 2018-12-25 | 中国科学院合肥物质科学研究院 | A kind of wheat scab disease tassel yield method for automatically counting |
| CN109541172A (en) * | 2018-10-25 | 2019-03-29 | 北京农业信息技术研究中心 | The calculation method and device of soil attribute value |
| CN110824142A (en) * | 2019-11-13 | 2020-02-21 | 杭州鲁尔物联科技有限公司 | Geological disaster prediction method, device and equipment |
| CN112417655A (en) * | 2020-11-05 | 2021-02-26 | 天津市农业质量标准与检测技术研究所 | Method for establishing farmland soil organic matter prediction model |
| CN112630408A (en) * | 2020-12-16 | 2021-04-09 | 天津市农业科学院 | Method for establishing prediction model of effective zinc content in farmland soil |
| CN112734133A (en) * | 2021-01-25 | 2021-04-30 | 南宁师范大学 | Method for predicting soil pH based on hydrothermal conditions |
| 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 |
| CN114140591A (en) * | 2021-11-29 | 2022-03-04 | 中国科学院南京土壤研究所 | Soil organic matter remote sensing mapping method combining machine learning and land statistics |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529189A (en) * | 2013-06-28 | 2014-01-22 | 四川农业大学 | Soil organic matter space distribution predication method based on qualitative and quantitative auxiliary variables |
| CN104537415A (en) * | 2014-12-02 | 2015-04-22 | 北京化工大学 | Non-linear process industrial fault prediction and identification method based on compressed sensing and DROS-ELM |
-
2017
- 2017-09-22 CN CN201710865008.8A patent/CN107860889A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529189A (en) * | 2013-06-28 | 2014-01-22 | 四川农业大学 | Soil organic matter space distribution predication method based on qualitative and quantitative auxiliary variables |
| CN104537415A (en) * | 2014-12-02 | 2015-04-22 | 北京化工大学 | Non-linear process industrial fault prediction and identification method based on compressed sensing and DROS-ELM |
Non-Patent Citations (2)
| Title |
|---|
| 宋英强等: "基于多源辅助变量和极限学习机的蔬菜地土壤有机质预测研究", 《土壤通报》 * |
| 李启权等: "基于定性和定量辅助变量的土壤有机质空间分布预测——以四川三台县为例", 《地理科学进展》 * |
Cited By (13)
| 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 |
| CN109086823A (en) * | 2018-08-01 | 2018-12-25 | 中国科学院合肥物质科学研究院 | A kind of wheat scab disease tassel yield method for automatically counting |
| CN109086823B (en) * | 2018-08-01 | 2022-02-11 | 中国科学院合肥物质科学研究院 | Automatic statistical method for wheat scab ear disease rate |
| CN109541172A (en) * | 2018-10-25 | 2019-03-29 | 北京农业信息技术研究中心 | The calculation method and device of soil attribute value |
| CN110824142B (en) * | 2019-11-13 | 2022-06-24 | 杭州鲁尔物联科技有限公司 | Geological disaster prediction method, device and equipment |
| CN110824142A (en) * | 2019-11-13 | 2020-02-21 | 杭州鲁尔物联科技有限公司 | Geological disaster prediction method, device and equipment |
| CN112417655A (en) * | 2020-11-05 | 2021-02-26 | 天津市农业质量标准与检测技术研究所 | Method for establishing farmland soil organic matter prediction model |
| CN112630408A (en) * | 2020-12-16 | 2021-04-09 | 天津市农业科学院 | Method for establishing prediction model of effective zinc content in farmland soil |
| 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 |
| CN112734133A (en) * | 2021-01-25 | 2021-04-30 | 南宁师范大学 | Method for predicting soil pH based on hydrothermal conditions |
| CN112734133B (en) * | 2021-01-25 | 2024-02-13 | 南宁师范大学 | Method for predicting soil pH based on hydrothermal condition |
| CN114140591A (en) * | 2021-11-29 | 2022-03-04 | 中国科学院南京土壤研究所 | Soil organic matter remote sensing mapping method combining machine learning and land statistics |
| CN114140591B (en) * | 2021-11-29 | 2024-04-26 | 中国科学院南京土壤研究所 | Soil organic matter remote sensing mapping method combining machine learning and ground statistics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107860889A (en) | The Forecasting Methodology and equipment of the soil organism | |
| CN109884664B (en) | Optical microwave collaborative inversion method and system for urban overground biomass | |
| CN110309985B (en) | Crop yield prediction method and system | |
| CN110427995B (en) | Bayesian soil moisture estimation method based on multi-source remote sensing data | |
| Whittaker et al. | Detection of overparameterization and overfitting in an automatic calibration of SWAT | |
| Wang et al. | Integrating remotely sensed leaf area index and leaf nitrogen accumulation with RiceGrow model based on particle swarm optimization algorithm for rice grain yield assessment | |
| CN110689179A (en) | Water bloom prediction method based on space-time sequence mixed model | |
| Ferraciolli et al. | Neglecting spatial autocorrelation causes underestimation of the error of sugarcane yield models | |
| CN116205310B (en) | Soil water content influence factor sensitive interval judging method based on interpretable integrated learning model | |
| CN116307191B (en) | Water resource configuration method, device and equipment based on artificial intelligence algorithm | |
| Hill et al. | The use of CO 2 flux time series for parameter and carbon stock estimation in carbon cycle research | |
| CN114723149A (en) | Soil moisture content prediction method and device, electronic equipment and storage medium | |
| Monteil et al. | Multi-objective calibration by combination of stochastic and gradient-like parameter generation rules–the caRamel algorithm | |
| CN110826764B (en) | Random forest algorithm-based vegetation index prediction method, system and equipment | |
| Ugbaje et al. | Fuzzy measure-based multicriteria land assessment for rainfed maize in West Africa for the current and a range of plausible future climates | |
| CN115758270A (en) | Soil mineral bound state organic carbon prediction method and device based on random forest and environment variables | |
| CN115755221A (en) | Multi-source precipitation data fusion method based on mathematical uncertainty analysis | |
| Nenkam et al. | Using homosoils for quantitative extrapolation of soil mapping models | |
| CN113139717B (en) | Crop seedling condition grading remote sensing monitoring method and device | |
| Tsai et al. | From parameter calibration to parameter learning: Revolutionizing large-scale geoscientific modeling with big data | |
| Mahmoud et al. | Land cover change dynamics mapping and predictions using EO data and a GIS-cellular automata model: the case of Al-Baha region, Kingdom of Saudi Arabia | |
| CN110852472A (en) | Land water reserve prediction method and equipment based on random forest algorithm | |
| CN113222288B (en) | Classified evolution and prediction method of village and town community space development map | |
| Gupta et al. | Markov Chain Model Used in Agricultural Yield Predictions Utilizing on Indian Agriculture | |
| Kumar et al. | Soil properties prediction for agriculture using machine learning techniques |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180330 |