CN116341363A - Multilayer soil humidity inversion method - Google Patents
Multilayer soil humidity inversion method Download PDFInfo
- Publication number
- CN116341363A CN116341363A CN202310073979.4A CN202310073979A CN116341363A CN 116341363 A CN116341363 A CN 116341363A CN 202310073979 A CN202310073979 A CN 202310073979A CN 116341363 A CN116341363 A CN 116341363A
- Authority
- CN
- China
- Prior art keywords
- data
- soil
- inversion
- neural network
- soil humidity
- 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
- 239000002689 soil Substances 0.000 title claims abstract description 128
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000003062 neural network model Methods 0.000 claims abstract description 44
- 238000013528 artificial neural network Methods 0.000 claims abstract description 31
- 241000254173 Coleoptera Species 0.000 claims abstract description 30
- 238000010845 search algorithm Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 7
- 238000010606 normalization Methods 0.000 claims abstract description 5
- 230000005855 radiation Effects 0.000 claims description 39
- 230000004907 flux Effects 0.000 claims description 20
- 230000000694 effects Effects 0.000 claims description 13
- 241001481710 Cerambycidae Species 0.000 claims description 12
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 229910002092 carbon dioxide Inorganic materials 0.000 claims description 4
- 239000001569 carbon dioxide Substances 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000001556 precipitation Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 abstract description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 abstract description 6
- 238000010801 machine learning Methods 0.000 abstract description 4
- 238000011160 research Methods 0.000 abstract description 4
- 238000004364 calculation method Methods 0.000 abstract description 3
- 238000004088 simulation Methods 0.000 description 8
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000000919 ceramic Substances 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000008635 plant growth Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 239000013585 weight reducing agent Substances 0.000 description 1
Images
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
- G01N33/246—Earth materials for water content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Chemical & Material Sciences (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Food Science & Technology (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Medicinal Chemistry (AREA)
- Remote Sensing (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a multilayer soil humidity inversion method, which belongs to the field of hydrology water resource subject soil humidity inversion, and comprises the following steps: dividing research data into a training set and a testing set and carrying out normalization treatment; selecting data with higher relativity with soil humidity by using a principal component analysis method; establishing a BP neural network model and initializing the BP neural network model to obtain an initial weight and a threshold; initializing a longhorn beetle whisker search algorithm; calculating an fitness function; iterative calculation; assigning the obtained optimal weight and threshold to a BP neural network, and constructing a BAS_BP neural network model suitable for soil humidity inversion at different depths; the BAS_BP neural network model is based on a machine learning theory, can be used for inversion of soil humidity at different depths, is used for a hydrologic forecasting and flood drought forecasting and early warning system, and improves forecasting and early warning precision.
Description
Technical Field
The invention belongs to the technical field of hydrologic water resource discipline soil humidity inversion, and particularly relates to a multi-layer soil humidity inversion method based on a longhorn beetle whisker search algorithm fused reconstruction BP neural network model.
Background
Soil Moisture (SM) is an important indicator of the dry and wet condition of land Soil, and is usually expressed by the volume Moisture content of the Soil. Soil moisture regulates water circulation and energy balance between the earth's surface and the atmosphere by affecting evaporative emissions, runoff, downflow, etc. links in the water circulation (mccol K a, 2017). Inversion of soil humidity is of great importance for many fields such as drought monitoring (Cao, chen, liu, & Liu, 2022), flood forecasting, plant growth (Annala et al, 2022), runoff simulation (Fidal & Kjeldsen, 2020). Therefore, accurate inversion of soil moisture at different depths is critical.
Machine learning has been widely used in the field of hydrologic water resources. Compared with the traditional statistical regression model, the machine learning model represented by the BP neural network model has remarkable advantages in nonlinear data modeling (Zhao, S nchez, lu, & Li, 2018). Inversion of soil humidity at different depths is a very complex nonlinear problem, and is limited by a plurality of factors, the solving process is similar to a multi-dimensional and multi-peak complex function solving process, and inversion of soil humidity at different depths is performed by using a simple neural network model, so that the requirements of practical application cannot be completely met (Xu Xiuying, 2011). The inversion accuracy of soil humidity at different depths can be effectively improved by fusing and reconstructing BP neural network (Back Propagation Neural Networks) model based on longhorn whisker search algorithm (Beetle Antennae Search Algorithm, BAS). At present, the BAS_BP neural network model (Beetle Antennae Search-Back Propagation Neural Networks) established by fusing the reconstructed BP neural network model based on the longhorn beetle whisker search algorithm is scarce in research and application in the multilayer soil humidity inversion field.
Reference is made to:
Annala,M.J.,Lehosmaa,K.,Ahonen,S.H.K.,Karttunen,K.,Markkola,A.M.,Puumala,I.,&
H.2022.Effect of riparian soil moisture on bacterial,fungal and plant communities and microbial decomposition rates in boreal stream-side forests.Forest Ecology and Management,519:120344.
Cao,M.,Chen,M.,Liu,J.,&Liu,Y.2022.Assessing the performance of satellite soil moisture on agricultural drought monitoring in the North China Plain.Agricultural Water Management,263:107450.
Fidal,J.,&Kjeldsen,T.R.2020.Accounting for soil moisture in rainfall-runoff modelling of urban areas.Journal of Hydrology,589:125122.
McColl K A,A.S.H.,Akbar R,et al..2017.The global distribution and dynamics of surface soil moisture.The global distribution and dynamics of surface soil moisture,10(2).Zhao,W.,Sánchez,N.,Lu,H.,&Li,A.2018.A spatial downscaling approach for the SMAP passive surface soil moisture product using random forest regression.Journal of Hydrology,563:1009-1024.
xu Xiuying, glycerol, ceramic, huang Caojun.2011. Prediction of soil moisture content based on genetic neural network [ A ]. China society of agricultural engineering: 1360-1365.
Disclosure of Invention
The invention aims to: aiming at the problem that an independent BP neural network cannot fully meet the actual requirements of soil humidity inversion application at different depths, the invention provides a multi-layer soil humidity inversion method based on the fusion of a longhorn beetle whisker search algorithm and a reconstructed BP neural network model.
The technical scheme is as follows: in order to achieve the purpose of the invention, the technical scheme adopted by the invention is as follows: a multi-layer soil moisture inversion method comprising the steps of:
step 1, acquiring observation data sets including soil humidity at soil layers with different depths, and dividing a training set and a testing set; normalizing the data;
step 2, screening out characteristic factors with the correlation with soil humidity data within a certain range from the normalized observation data set by using a principal component analysis method;
step 3, establishing and initializing a BP neural network to obtain a network initial weight and a threshold value, and optimizing and reconstructing the network initial weight and the threshold value by using a longhorn beetle whisker search algorithm BAS to obtain an optimal weight and the threshold value;
step 4, the characteristic factors are used as input data of the network, soil humidity at soil layers with different depths is used as output data of the network, and the training set is used for training the network to obtain a fused and reconstructed BAS_BP neural network model;
and 5, inputting the test set data into a BAS_BP neural network model, and outputting the model to obtain the multi-layer soil humidity with different depths by inversion.
Further, the data acquiring and processing in the step 1 specifically includes:
acquiring an observation data set comprising meteorological data, surface radiation data, EC data and soil hydrothermal data; the meteorological data comprise wind speed WS, wind direction WD, air temperature Ta, relative humidity RH, vapor Pressure Vapor, air Pressure and precipitation Prec; the surface radiation data comprises incoming solar radiation Rsd, outgoing solar radiation Rsu, downward long-wave radiation Rld, upward long-wave radiation Rlu and net radiation Rn; the EC data includes sensible heat flux H, latent heat flux LE, carbon dioxide flux Fc; the soil hydrothermal data comprise surface temperature Tg, soil temperature Ts, soil humidity SM and soil heat flux SHF;
interpolation of missing values in data is achieved through a filmsising function in MATLAB, meanwhile abnormal values in the data are monitored through a filutliiers function, interpolation is conducted, and after a data set continuous in time is obtained, the data set is converted into a multi-year daily average data set; the interpolated data set is divided into a training set and a testing set according to a certain proportion.
Further, step 2 is to screen out observation data, i.e. characteristic factors, which have a correlation with soil humidity data within a certain range from all observation factors except soil humidity by using a principal component analysis method, and specifically includes:
calculating a covariance matrix of the sample characteristics; the sample characteristics refer to meteorological data, surface radiation data, EC data and soil hydrothermal data in an observation data set;
calculating eigenvalues and eigenvectors of the covariance matrix; the eigenvalues are arranged according to descending order, and the eigenvectors corresponding to the eigenvalues are respectively used as column vectors to form eigenvectors;
calculating the accumulated contribution rate of the characteristic values, sequentially combining the accumulated contribution rates from high to low to perform soil humidity inversion, and selecting a group of data with the best inversion effect as a main component;
the selected principal component is used as a characteristic factor related to soil moisture data.
Further, selecting a group of data with the best inversion effect as a main component includes: downward long wave radiation Rld, incoming solar radiation Rsd, outgoing solar radiation Rsu, upward long wave radiation Rlu, barometric Pressure, surface relative humidity RH, vapor Pressure Vapor, wind direction WD, wind speed WS, sensible heat flux H, latent heat flux LE.
Further, the optimization and reconstruction are carried out on the initial weight and the threshold value of the network by using a longhorn beetle whisker search algorithm BAS, so as to obtain the optimal weight and the threshold value, which comprises the following steps:
setting an initial step length and iteration times, initializing a longhorn beetle whisker search algorithm BAS, creating a random vector of the longhorn beetle whisker orientation, carrying out normalization processing, and creating a longhorn beetle whisker space coordinate;
setting the initial weight and the threshold of the neural network as the direction and the initial position of the longhorn beetle beards respectively; setting the ratio of the step length of the longicorn to the distance between the left antenna and the right antenna;
the seeking smell and advancing operation are carried out: calculating the odor concentration perceived by the left and right tentacles by using the self-adaptive function, if the odor concentration perceived by the left tentacle is stronger than that of the right, advancing the longicorn to the left next, and if the odor concentration perceived by the right tentacle is stronger than that of the left Bian Jiang, advancing the longicorn to the right next;
judging whether an iteration termination condition is reached, namely, the longicorn finds food, namely, the output weight and the threshold value are the global optimal solution, and stopping iteration; otherwise, returning to the operation of searching smell and advancing.
The beneficial effects are that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects.
The invention adopts the principal component analysis technology to select main characteristic variables from a plurality of factors influencing soil humidity and keep original information as much as possible, thereby not only improving model precision, but also reducing dimension, effectively eliminating redundancy and relativity among influencing factors, further realizing weight reduction of the model, and improving model forecasting accuracy. Secondly, the invention provides a BAS fusion reconstruction BP neural network model based on a longhorn beetle whisker search algorithm, and the BAS_BP neural network model suitable for multi-layer soil humidity inversion is constructed.
Inversion of soil humidity at different depths is a complex nonlinear problem, and is limited by various factors, the inversion process is similar to a multi-dimensional and multi-peak complex function solution, and the inversion of soil humidity by using a common BP neural network model cannot completely meet the requirements of practical application. Therefore, the invention provides a technology for introducing a longhorn beetle whisker search algorithm BAS into a BP neural network soil humidity inversion model, optimizing weights and thresholds in the BP neural network, and fusing and reconstructing the BAS_BP neural network model suitable for multi-layer soil humidity inversion. The BAS_BP neural network model can effectively improve the calculation accuracy of the BP neural network model, has the advantages of small operation amount, high convergence rate, global optimizing capability and the like, and greatly improves the inversion accuracy of the BP neural network model to multi-layer soil humidity.
Drawings
FIG. 1 is a schematic flow chart of the present invention;
FIG. 2 is a graph of contrast dispersion of the inversion model of BAS_BP neural network soil moisture to the inversion values and observation values of soil moisture at different depths in a test set;
fig. 3 is a graph of inversion values and observation values of soil moisture at different depths in a test set by using a BP neural network and bas_bp neural network soil moisture inversion model.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples. The following examples are only for more clearly illustrating the technical aspects of the present invention, and are not intended to limit the scope of the present invention.
The invention relates to a multilayer soil humidity inversion method, which is shown in a figure 1, and comprises the following steps:
and step 1, obtaining observation data of a Qinghai-Tibet plateau field appearance measuring station MAWORS (Muztagh Ata Westerly Observation and Research Station, mu Shida grid western wind belt environment comprehensive observation research station) and dividing a training set and a testing set.
(1) Acquiring an observation data set comprising meteorological data, surface radiation data, EC data and soil hydrothermal data; the meteorological data comprises Wind Speed (WS), wind Direction (WD), air temperature (Ta), relative humidity (relative humidity, RH), steam Pressure (water Vapor Pressure, vapor), air Pressure (air Pressure), precipitation (Pressure); the surface radiation data includes incoming solar radiation (incoming solar radiation, rsd), outgoing solar radiation (outgoing solar radiation, rsu), downward long wave radiation (Rld), upward long wave radiation (Rlu), net radiation (Rn); the EC data includes sensible heat flux (sensible heat flux, H), latent heat flux (LE), carbon dioxide flux (carbon dioxide flux, fc); the soil hydrothermal data include surface temperature (ground temperature, tg), soil temperature (Ts), soil humidity (SM), soil Heat Flux (SHF).
(2) And interpolating the missing values in the data by using a filmsising function in MATLAB, monitoring the abnormal values in the data by using a filutliiers function, interpolating to obtain a data set which is continuous in time, and converting the data set into a multi-year daily average data set for 365 days.
(3) Dividing the interpolated data set into a training set and a test set according to a certain proportion, taking 1-250 days of data as the training set for establishing a model and training, and 251-365 days of data as the test set for checking the inversion accuracy of the model. All data in the dataset are normalized and mapped between 0-1. Normalization is a way of simplifying computation, namely, an expression with dimension is converted into an expression without dimension through transformation to become a scalar, so that the precision and convergence rate of a model can be improved.
And 2, screening out characteristic factors which are related to soil humidity data within a certain range from the normalized observation data set by using a principal component analysis method, wherein the method specifically comprises the following steps of:
(1) Calculating a covariance matrix of the sample characteristics; the sample characteristics refer to meteorological data, surface radiation data, EC data and soil hydrothermal data in an observation data set;
(2) Calculating eigenvalues and eigenvectors of the covariance matrix;
(3) The eigenvalues are arranged according to descending order, and the eigenvectors corresponding to the eigenvalues are respectively used as column vectors to form eigenvectors;
(4) Calculating the accumulated contribution rate of the characteristic values, sequentially combining the accumulated contribution rates from high to low to perform soil humidity inversion, and selecting a group of data with the best inversion effect as a main component, wherein the method comprises the following steps: downward long wave radiation Rld, incoming solar radiation Rsd, outgoing solar radiation Rsu, upward long wave radiation Rlu, barometric Pressure, surface relative humidity RH, vapor Pressure Vapor, wind direction WD, wind speed WS, sensible heat flux H, latent heat flux LE. The selected principal component is used as a characteristic factor related to soil moisture data.
Step 3, establishing and initializing a BP neural network to obtain a network initial weight and a threshold value, and performing optimization reconstruction on the network initial weight and the threshold value by using a longhorn beetle whisker search algorithm BAS to obtain an optimal weight and the threshold value, wherein the step comprises the following steps:
setting an initial step length and iteration times, initializing a longhorn beetle whisker search algorithm BAS, creating a random vector of the longhorn beetle whisker orientation, carrying out normalization processing, and creating a longhorn beetle whisker space coordinate;
setting the initial weight and the threshold of the neural network as the direction and the initial position of the longhorn beetle beards respectively; setting the ratio of the step length of the longicorn to the distance between the left antenna and the right antenna;
the seeking smell and advancing operation are carried out: calculating the odor concentration of the left and right tentacles by using an adaptive function (calculated by using a fitness function in MATLAB), if the odor concentration of the left tentacle is stronger than that of the right tentacle, the longicorn proceeds to the left next, and if the odor concentration of the right tentacle is stronger than that of the left Bian Jiang, the longicorn proceeds to the right next;
judging whether an iteration termination condition is reached, namely, the longicorn finds food, namely, the output weight and the threshold value are the global optimal solution, and stopping iteration; otherwise, returning to the operation of searching smell and advancing.
And 4, taking the characteristic factors as input data of the network, taking soil humidity at soil layers with different depths as output data of the network, and training the network by utilizing a training set to obtain a fused and reconstructed BAS_BP neural network model.
And 5, inputting the test set data into a BAS_BP neural network model, and outputting the model to obtain the multi-layer soil humidity with different depths by inversion.
FIG. 2 shows the contrast dispersion of the inversion model of the soil humidity of the BAS_BP neural network to the inversion values and the observation values of the soil humidity of different depths in the test set, wherein the difference between the simulation value and the observation value obtained by using the BAS_BP model at the soil layer of 10cm of the earth surface is between-0.0384 and 0.0082; at the soil layer of 20cm, the difference between the simulation value and the observed value obtained by using the BAS_BP model is between-0.0499 and 0.0188; at the position of 40cm of the soil layer, the difference value between the simulation value and the observed value obtained by using the BAS_BP model is between-0.0628 and 0.0245; at the 80cm position of the soil layer, the difference value between the simulation value and the observed value obtained by using the BAS_BP model is between-0.0167 and 0.0822; at 160cm of soil layer, the difference between the simulation value and the observed value obtained by using the BAS_BP model is between-0.0166 and 0.2407. With the increase of the soil depth, the difference between the simulation value and the observation value obtained by inversion is increased, and the inversion effect is weakened.
This example uses meteorological data, surface radiation, EC data, soil hydro-thermal data from 1 month 1 in 2005 to 31 months 12 in 2016. And respectively using the BP neural network model and the longhorn beetle whisker search algorithm to fuse the reconstructed BAS_BP neural network model to invert soil humidity at different depths.
And respectively constructing BP neural network and BAS_BP neural network multi-layer soil humidity inversion models by using the characteristic factors obtained by the principal component analysis method as input data and the soil humidity at soil layers of 10cm, 20cm, 40cm, 80cm and 160cm as output data, and training and inverting the soil humidity data at different depths.
And (3) adopting three statistical indexes of a correlation coefficient (R), an average absolute error (MAE) and a Root Mean Square Error (RMSE) to test inversion effects of the BP neural network and the BAS_BP neural network model on soil humidity at different depths. The calculation formula of each statistical index is as follows:
mean absolute error (Mean Absolute Error, MAE): and calculating the average value of absolute errors between the observed value and the calculated value, wherein the smaller the average absolute error is, the closer the inversion value is to the observed value, and the more accurate the inversion result is.
Root mean square error (Root Mean Square Error, RMSE): the method is used for measuring the deviation between the observed value and the calculated value, and the smaller the root mean square error is, the closer the inversion value is to the observed value, and the more accurate the inversion result is.
Correlation coefficient (Correlation coefficient, R): according to the variation of the independent variable, the variation of the dependent variable is explained, and the closer the value of the correlation coefficient is to 1, the better the inversion effect is; the closer the correlation coefficient is to 0, the worse the inversion effect is explained.
Wherein n is the total number of samples; y is i Inversion for the ith sampleA value; x is x i An observation value for the i-th sample;
is the average value of inversion data; />
Mean of the observed data.
Table 1 shows the Root Mean Square Error (RMSE), mean Absolute Error (MAE) and correlation coefficient (R) of the simulated values of the BP, bas_bp two models for the soil moisture test sets at different depths, respectively.
TABLE 1
Table 1 and fig. 3 are comprehensively compared to analyze the soil moisture inversion effects of BP and bas_bp neural network models on different depths. The result shows that at the place of 10cm on the earth surface, the root mean square error and the average absolute error of the soil humidity and the observed value based on the inversion of the BAS_BP neural network model are respectively 0.0141 and 0.0098, and the correlation coefficient is 0.903, which are superior to the inversion model of the initial BP neural network. At the place of 20cm on the earth surface, the root mean square error and the average absolute error of soil humidity and an observed value based on the inversion of the BAS_BP neural network model are respectively 0.0145 and 0.0107, and the correlation coefficient is 0.854, which is also superior to the inversion model of the BP neural network. And the inversion effect of the BAS_BP neural network model on the deep soil humidity (40 cm to 160 cm) is consistent with that of the surface soil humidity, and is obviously improved compared with that of the BP neural network model. The inversion value of the BAS_BP neural network model on the soil humidity at the place of 40cm on the earth surface has root mean square error and average absolute error of 0.0200 and 0.0170 respectively, and the correlation coefficient is 0.832, which is superior to the inversion value of the BP neural network model on the soil humidity at the same layer (root mean square error is 0.0247, average absolute error is 0.0264, and correlation coefficient is 0.530). At the soil depth of 80cm, the root mean square error and the average absolute error of the inversion value and the observation value of the soil humidity by the BAS_BP neural network model are respectively 0.0457 and 0.0413, the correlation coefficient is 0.800, and the root mean square error and the average absolute error of the inversion value of the soil humidity and the observation value based on the BP neural network model are respectively 0.0532 and 0.0382, and the correlation coefficient is 0.403. At the soil depth of 160cm, the root mean square error and the average absolute error of the inversion value and the observation value of the soil humidity by the BAS_BP neural network model are 0.1434 and 0.1309 respectively, the correlation coefficient is 0.504, and the root mean square error and the average absolute error of the soil humidity and the observation value based on the inversion of the BP neural network model are 0.181 and 0.1654 respectively, and the correlation coefficient is 0.390.
Fig. 3 is a graph of comparison of inversion effects before and after optimization by using a longhorn beetle whisker search algorithm for a BP neural network, wherein at the place of the ground surface of 10cm, the minimum values of root mean square error and average absolute error before and after inversion of soil humidity are respectively 0.0152 and 0.0101, and the maximum correlation coefficient is 0.915, and the results are obtained by inversion of the BAS_BP neural network. At the place of 20cm on the earth surface, the minimum values of root mean square error and average absolute error before and after soil humidity inversion are respectively 0.0152 and 0.0139, and the maximum correlation coefficient is 0.897, and the root mean square error and the average absolute error are obtained through inversion of a BAS_BP neural network. At the place of 40cm on the earth surface, the minimum values of root mean square error and average absolute error before and after soil humidity inversion are respectively 0.0209 and 0.0190, and the maximum correlation coefficient is 0.892, and the soil humidity inversion is obtained by inversion of a BAS_BP neural network. At the place of 80cm on the earth surface, the minimum values of root mean square error and average absolute error before and after soil humidity inversion are respectively 0.0350 and 0.0247, and the maximum correlation coefficient is 0.877, and the soil humidity inversion is obtained by inversion of a BAS_BP neural network. At 160cm of the ground surface, the minimum values of root mean square error and average absolute error before and after soil humidity inversion are 0.0901 and 0.1310 respectively, the maximum correlation coefficient is 0.796, and the root mean square error and the average absolute error are obtained through inversion of a BAS_BP neural network. The comparison analysis shows that the inversion capability of the BP neural network model and the BAS_BP neural network model to soil humidity at different depths has attenuation trend along with the increase of soil layer depth. Compared with a common BP neural network model, the bas_BP neural network model which is fused and reconstructed through the longhorn beetle whisker search algorithm can effectively improve inversion precision and stability of soil humidity at different depths.
In summary, according to the multi-layer soil humidity inversion method for establishing the BAS_BP neural network model based on the longhorn beetle whisker search algorithm fused reconstruction BP neural network model, based on the machine learning theory, weather factors with high correlation with soil humidity are selected as input data by using a principal component analysis method, BP neural network and BAS_BP neural network multi-layer soil humidity inversion models are respectively established, and soil humidity at different depths is trained and inverted. The BASBP neural network model based on the longhorn beetle whisker search algorithm fusion reconstruction constructed by the method is more suitable for inversion of multi-layer soil humidity, and has higher inversion precision and more stable simulation performance.
While the foregoing is directed to the preferred embodiments of the present invention, it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present invention, and such modifications and variations should also be regarded as being within the scope of the invention.
Claims (5)
1. A multi-layer soil moisture inversion method, comprising the steps of:
step 1, acquiring observation data sets including soil humidity at soil layers with different depths, and dividing a training set and a testing set; normalizing the data;
step 2, screening out characteristic factors with the correlation with soil humidity data within a certain range from the normalized observation data set by using a principal component analysis method;
step 3, establishing and initializing a BP neural network to obtain a network initial weight and a threshold value, and optimizing and reconstructing the network initial weight and the threshold value by using a longhorn beetle whisker search algorithm BAS to obtain an optimal weight and the threshold value;
step 4, the characteristic factors are used as input data of the network, soil humidity at soil layers with different depths is used as output data of the network, and the training set is used for training the network to obtain a fused and reconstructed BAS_BP neural network model;
and 5, inputting the test set data into a BAS_BP neural network model, and outputting the model to obtain the multi-layer soil humidity with different depths by inversion.
2. The multi-layer soil moisture inversion method of claim 1 wherein the data acquisition and processing of step 1 specifically comprises:
acquiring an observation data set comprising meteorological data, surface radiation data, EC data and soil hydrothermal data; the meteorological data comprise wind speed WS, wind direction WD, air temperature Ta, relative humidity RH, vapor Pressure Vapor, air Pressure and precipitation Prec; the surface radiation data comprises incoming solar radiation Rsd, outgoing solar radiation Rsu, downward long-wave radiation Rld, upward long-wave radiation Rlu and net radiation Rn; the EC data includes sensible heat flux H, latent heat flux LE, carbon dioxide flux Fc; the soil hydrothermal data comprise surface temperature Tg, soil temperature Ts, soil humidity SM and soil heat flux SHF;
interpolation of missing values in data is achieved through a filmsising function in MATLAB, meanwhile abnormal values in the data are monitored through a filutliiers function, interpolation is conducted, and after a data set continuous in time is obtained, the data set is converted into a multi-year daily average data set; the interpolated data set is divided into a training set and a testing set according to a certain proportion.
3. The multi-layer soil moisture inversion method according to claim 2, wherein step 2 uses principal component analysis to screen out observation data, i.e. characteristic factors, having a correlation with soil moisture data within a certain range from all observation factors except soil moisture, specifically comprising:
calculating a covariance matrix of the sample characteristics; the sample characteristics refer to meteorological data, surface radiation data, EC data and soil hydrothermal data in an observation data set;
calculating eigenvalues and eigenvectors of the covariance matrix; the eigenvalues are arranged according to descending order, and the eigenvectors corresponding to the eigenvalues are respectively used as column vectors to form eigenvectors;
calculating the accumulated contribution rate of the characteristic values, sequentially combining the accumulated contribution rates from high to low to perform soil humidity inversion, and selecting a group of data with the best inversion effect as a main component;
the selected principal component is used as a characteristic factor related to soil moisture data.
4. A multi-layer soil moisture inversion method as claimed in claim 3 wherein selecting a set of data with the best inversion effect as the principal component comprises: downward long wave radiation Rld, incoming solar radiation Rsd, outgoing solar radiation Rsu, upward long wave radiation Rlu, barometric Pressure, surface relative humidity RH, vapor Pressure Vapor, wind direction WD, wind speed WS, sensible heat flux H, latent heat flux LE.
5. The multi-layer soil moisture inversion method according to any one of claims 1 to 4, wherein optimizing and reconstructing the initial weights and the thresholds of the network by using a longhorn beetle whisker search algorithm BAS to obtain the optimal weights and the thresholds comprises:
setting an initial step length and iteration times, initializing a longhorn beetle whisker search algorithm BAS, creating a random vector of the longhorn beetle whisker orientation, carrying out normalization processing, and creating a longhorn beetle whisker space coordinate;
setting the initial weight and the threshold of the neural network as the direction and the initial position of the longhorn beetle beards respectively; setting the ratio of the step length of the longicorn to the distance between the left antenna and the right antenna;
the seeking smell and advancing operation are carried out: calculating the odor concentration perceived by the left and right tentacles by using the self-adaptive function, if the odor concentration perceived by the left tentacle is stronger than that of the right, advancing the longicorn to the left next, and if the odor concentration perceived by the right tentacle is stronger than that of the left Bian Jiang, advancing the longicorn to the right next;
judging whether an iteration termination condition is reached, namely, the longicorn finds food, namely, the output weight and the threshold value are the global optimal solution, and stopping iteration; otherwise, returning to the operation of searching smell and advancing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310073979.4A CN116341363A (en) | 2023-02-07 | 2023-02-07 | Multilayer soil humidity inversion method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310073979.4A CN116341363A (en) | 2023-02-07 | 2023-02-07 | Multilayer soil humidity inversion method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116341363A true CN116341363A (en) | 2023-06-27 |
Family
ID=86881261
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310073979.4A Pending CN116341363A (en) | 2023-02-07 | 2023-02-07 | Multilayer soil humidity inversion method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116341363A (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110610054A (en) * | 2019-09-23 | 2019-12-24 | 北京师范大学 | Method and system for constructing cuboid inversion model of soil humidity |
-
2023
- 2023-02-07 CN CN202310073979.4A patent/CN116341363A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110610054A (en) * | 2019-09-23 | 2019-12-24 | 北京师范大学 | Method and system for constructing cuboid inversion model of soil humidity |
Non-Patent Citations (1)
| Title |
|---|
| 李旭强等: "基于BAS－BPNN模型的季节性冻融期土壤含水率预测", 《节水灌溉》, no. 10, pages 66 - 70 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ibrahim et al. | A novel hybrid model for hourly global solar radiation prediction using random forests technique and firefly algorithm | |
| Hao et al. | Risk assessment to China’s agricultural drought disaster in county unit | |
| CN111027175B (en) | Method for evaluating social and economic influences of flood based on coupling model integrated simulation | |
| Hasni et al. | Optimization of greenhouse climate model parameters using particle swarm optimization and genetic algorithms | |
| CN107563565A (en) | A kind of short-term photovoltaic for considering Meteorology Factor Change decomposes Forecasting Methodology | |
| CN109214579B (en) | BP neural network-based saline-alkali soil stability prediction method and system | |
| Li et al. | A novel combined prediction model for monthly mean precipitation with error correction strategy | |
| CN105184400A (en) | Tobacco field soil moisture prediction method | |
| Li et al. | GANs-LSTM model for soil temperature estimation from meteorological: a new approach | |
| Xu et al. | Estimation of potential ecological carrying capacity in China | |
| CN115640956A (en) | Future water resource supply and demand balance analysis method | |
| Bensoussan et al. | Cox–Ingersoll–Ross model for wind speed modeling and forecasting | |
| CN111914488B (en) | Data area hydrologic parameter calibration method based on antagonistic neural network | |
| Li et al. | Comparison study on ways of ecological vulnerability assessment-----A case study in the Hengyang Basin | |
| CN116227692B (en) | Crop heavy metal enrichment risk quantification method, system and storable medium | |
| CN112560633A (en) | Plant key phenological period time point prediction method and system based on deep learning | |
| Mi et al. | Testing the generalization of artificial neural networks with cross-validation and independent-validation in modelling rice tillering dynamics | |
| CN116341363A (en) | Multilayer soil humidity inversion method | |
| Wang et al. | Statistical and machine learning methods applied to the prediction of different tropical rainfall types | |
| Duan et al. | Prediction model of soil heavy metal content based on particle swarm algorithm optimized neural network | |
| Popović et al. | Neural network based system in evapotranspiration time series prediction | |
| Mohan et al. | Crop production rate estimation using parallel layer regression with deep belief network | |
| Jiang et al. | Application and evaluation of an improved LSTM model in the soil moisture prediction of southeast chinese tobacco-producing areas | |
| CN113052379A (en) | Tomato water demand prediction method based on BAS-RBF | |
| Kim et al. | A decision-centric impact assessment of operational performance of the Yongdam Dam, South Korea |
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 |