CN115906656A - Method for inverting soil water content based on meteorological and gravity satellite data and other data - Google Patents
Method for inverting soil water content based on meteorological and gravity satellite data and other data Download PDFInfo
- Publication number
- CN115906656A CN115906656A CN202211621963.4A CN202211621963A CN115906656A CN 115906656 A CN115906656 A CN 115906656A CN 202211621963 A CN202211621963 A CN 202211621963A CN 115906656 A CN115906656 A CN 115906656A
- Authority
- CN
- China
- Prior art keywords
- data
- water content
- meteorological
- model
- soil
- 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
Images
Classifications
-
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method for inverting soil water content based on meteorological and gravity satellite and other data, which comprises the following steps: s1, preparing data; s2, wavelet transformation; s3, data normalization processing; s4, dividing a model data set; s5, determining super parameters of the long-term and short-term memory network model; s6, building a long-short term memory network model architecture; s7, determining an optimal long-term and short-term memory network model; and S8, testing and evaluating the model. The soil water content inversion method provided by the invention can acquire the time-space change characteristics of large-range soil water without field sample collection and test or arrangement of a large number of monitoring stations; the soil water content inversion method provided by the invention expands the land hydrological application of the gravity satellite data from the directions of groundwater inversion, disaster monitoring and the like to soil water inversion analysis, and further releases the research potential of the gravity satellite data in a water circulation basin.
Description
Technical Field
The invention relates to the field of meteorological and gravity satellites, in particular to a method for inverting soil water content based on data of meteorological and gravity satellites and the like.
Background
Soil water is an important link for connecting the interaction of atmospheric water, surface water and underground water, is a hub in the process of land water circulation, is also a main source of crop and vegetation water, and the content change of the soil water has important influence on regional climate, water resources and an ecological system; the other method is to estimate the soil water content by utilizing the difference of dielectric parameters of signals such as electromagnetic waves between water and soil particles, the first method is large in workload and time-consuming and labor-consuming, and can affect subsequent sampling observation due to the fact that the surrounding soil structure is damaged when a soil sample is drilled.
The GRACE and GRACE-FO gravity satellites measure the earth time-varying gravity field signals to further obtain the global land water reserve change, and have the advantages of high data acquisition efficiency, repeated observation, uniform scale, wide coverage range, uniform distribution, no restriction by landforms and meteorological conditions and the like.
Therefore, it is necessary to provide a method for inverting the water content of soil based on data such as meteorological and gravity satellites to solve the above technical problems.
Disclosure of Invention
The invention provides a method for inverting soil water content based on meteorological and gravity satellite and other data, which solves the defects of time consumption, labor consumption and cost consumption of the traditional method, and simultaneously has the problems of inverting the soil water content in any global range in the historical period and monitoring the current and future soil water changes.
In order to solve the technical problem, the invention provides a method for inverting the water content of soil based on data such as meteorological and gravity satellites, which comprises the following steps:
s1, data preparation: the data required to be prepared in the invention comprises four types, wherein the first three types of data are independent variables, and the last type of data are dependent variables;
s2, wavelet transformation: firstly, continuous wavelet transformation is carried out on GRACE and GRACE-FO gravity satellite data in the step S1 to construct time-frequency information with good time domain and frequency domain localization, morlet wavelets are selected as basic wavelets, the types of the basic wavelets can be adjusted according to the final soil humidity inversion result, the wavelet transformation scale at least ranges from 1 to 24, and high-frequency signals are selected as model input variables.
S3, data normalization processing: normalizing the time-frequency information of the meteorological data, the vegetation data, the soil water data and the gravity satellite data in the step S2 to improve the model construction and verification efficiency, wherein the normalization method can be that the data is zoomed to 0 to 1 (formula 2) through the maximum value and the minimum value in the sequence, or zoomed to-1 to 1 through the maximum absolute value of the data,
X new ＝ (X – X min ) / (X max – X min ) Equation 2
X new (| X |) equation 3;
s4, model data set division: dividing all the normalized time sequence data in the step S3 into a training set, a verification set and a test set, wherein the division ratio is 0.7;
s5, determining super parameters of the long-term and short-term memory network model: the inversion efficiency is improved by utilizing a super parameter automatic optimization method, model super parameters mainly relate to seven items, generally speaking, the more the parameters are selected or the larger the parameter value is, the inversion accuracy of the soil water content is improved, but the higher the configuration requirements of the model on a computer CPU and a GPU are;
s6, building a long-short term memory network model architecture: building a model architecture composed of different parameters by using a Bayesian optimal parameter filter combination and the super parameters determined in the filtering step S5, and inputting the training set data generated in the step S4 for training network models of different architectures;
s7, determining an optimal long-term and short-term memory network model: substituting the verification set data generated in the step S4 into the different architecture models generated in the step S6, selecting the model with the minimum regression loss as the optimal model, and storing corresponding optimal model parameters;
s8, model testing and evaluating: and substituting the test set data generated in the step S4 into the optimal long-short term memory network model generated in the step S7 to obtain soil water content data simulated by the model, and then evaluating the performance of the soil water content obtained by the model simulation by taking the actual soil water content data and the like in the step S1 as references, wherein the evaluation indexes select root mean square error, correlation coefficient and standard error.
Preferably, the first type in the S1 is meteorological data which comprises monthly rainfall and monthly average air temperature of 2m height above the ground, the rainfall and the air temperature are content which is necessary to be measured by a country or a local meteorological station and are easy to collect, when a meteorological station is absent in a research range, precipitation and air temperature data can be extracted through meteorological grid products issued by a country or an international meteorological organization, the second type is vegetation index monthly homogeneity data, when vegetation in a research area is sparse, an Enhanced Vegetation Index (EVI) can be selected, otherwise, a normalized vegetation index (NDVI) can be selected, the third type is GRACE and GRACE-FO gravity satellite data, the fourth type is soil humidity data used for constructing a model and verifying model precision, soil humidity data measured in the field can be field, and hydrological model product data can also be obtained.
Preferably, Ψ (t) in S2 is a wavelet mother function, a function having continuous properties in both the time domain and the frequency domain, a (— infinity, + ∞) is a translation position, and b (0, + ∞) is a scaling factor.
Preferably, in the S4, random division is preferably selected in general, but considering that there is a certain time interval between the retirement of the GRACE gravity satellite and the transmission of the GRACE-FO gravity satellite, the data of the service period of the GRACE gravity satellite may be divided into a training set and a verification set, and the data of the service period of the GRACE-FO satellite may be divided into a test set.
Preferably, the main super parameters and values thereof in S5 are as follows: (1) selecting relu, sigmoid, tanh, selu, gelu and leak _ relu (alpha = 0.2) as an activation function; (2) the number of output units of the hidden layer is 32-512, and the step length is 32; (3) the data traversal times are at least 200; (4) the rejection ratio of data for preventing overfitting is 0-0.25; (5) taking an absolute error average value as a loss function; (6) the optimization algorithm selects Nadam or adam, and the Nadam or adam is more time-consuming; (7) the number of LSTM layers is 1-10.
Preferably, modeling equipment is required to be used in modeling in S2, the modeling equipment comprises a computer main body, a mounting assembly is arranged on the surface of the computer main body and comprises a mounting frame, and supporting seats are connected to two sides of the surface of the mounting frame.
Preferably, the bottom of mounting bracket is provided with the extension subassembly, the extension subassembly includes two slide bars, two the both sides on slide bar surface all are equipped with the sliding sleeve, two the both sides of sliding sleeve all are connected with the connecting plate.
Preferably, two one side of connecting plate all is connected with the extension board, the surface of extension board is provided with flexible subassembly, flexible subassembly includes a plurality of telescopic links, and is a plurality of be connected with the fixed plate between the telescopic link.
Preferably, a separation block is connected to the center position of the surfaces of the two slide bars.
Preferably, one side of mounting bracket is provided with places the subassembly, it includes two intercommunication pieces to place the subassembly, two the inside of intercommunication piece all is provided with the spliced pole, two be connected with between the spliced pole and place the board, the surface bonding of placing the board has the foam-rubber cushion, the surface of spliced pole is provided with the adapter sleeve.
Compared with the related art, the method for inverting the water content of the soil based on the data such as meteorological data, gravity satellites and the like has the following beneficial effects:
the invention provides a method for inverting soil water content based on data such as meteorological and gravity satellites, and the soil water content inversion method provided by the invention can acquire the time-space change characteristics of large-range soil water without field sample collection and test or the arrangement of a large number of monitoring stations;
the soil water content inversion method provided by the invention expands the land hydrological application of the gravity satellite data from the directions of groundwater inversion, disaster monitoring and the like to soil water inversion analysis, and further releases the research potential of the gravity satellite data in a water circulation basin;
all input data in the soil water content inversion method can be freely obtained from domestic and foreign meteorological or hydrological organizations without spending any cost;
the soil water content inversion method provided by the invention is suitable for various landforms such as mountains, plateaus, plains, deserts and the like, and is not limited by conditions such as sparse vegetation or dense vegetation on the earth surface in space;
the soil water content inversion method provided by the invention is not limited by any meteorological conditions;
except for individual months with missing gravity satellite data such as GRACE, the method provided by the invention can obtain month-by-month data of soil water content at any place since 4 months in 2002, and the time sequence of the soil water content is longer than that of soil water content obtained by utilizing radar image inversion;
the soil water content inversion method provided by the invention introduces a Bayesian parameter screener. Through comparison tests in the same software and hardware computing environment, the method can save the running time by more than 1/5 compared with the manual parameter adjusting method;
the soil water content inversion method provided by the invention is applied to a fixed area, the framework and the optimal parameters of the inversion model are determined for the first time, and the optimal model is called for the subsequent inversion;
through tests, the correlation coefficient of the soil water content obtained by the method and the tested soil water content can reach more than 0.7 (the p value is less than 0.05), and the root mean square error is about 3.5.
Drawings
FIG. 1 is a schematic structural diagram of a first embodiment of a method for inverting water content of soil based on meteorological and gravity satellite data;
FIG. 2 is a flow chart of key links;
FIG. 3 is a schematic diagram of inversion model construction;
FIG. 4 is a schematic structural diagram of a second embodiment of a method for inverting water content of soil based on meteorological and gravity satellite data;
FIG. 5 is a schematic perspective view of the device shown in FIG. 4;
FIG. 6 is an enlarged view of portion A of FIG. 5;
fig. 7 is an enlarged view of the portion B shown in fig. 5.
Reference numbers in the figures: 1. the main body of the computer is provided with a computer,
2. a mounting component 21, a mounting frame 22 and a supporting seat,
3. an extension component 31, a sliding rod 32, a sliding sleeve 33, a connecting plate 34 and an extension plate,
4. a telescopic component 41, a telescopic rod 42 and a fixed plate,
6. a placing component 61, a communicating block 62, a connecting column 63, a placing plate 64, a spongy cushion 65 and a connecting sleeve,
7. a partitioning block.
Detailed Description
The invention is further described with reference to the following figures and embodiments.
First embodiment
Please refer to fig. 1, fig. 2 and fig. 3 in combination, wherein fig. 1 is a schematic structural diagram of a first embodiment of a method for inverting water content of soil based on data of meteorological and gravity satellites according to the present invention; FIG. 2 is a key link flow diagram; fig. 3 is a schematic diagram of inversion model construction. A method for inverting the water content of soil based on data such as meteorological and gravity satellites comprises the following steps:
s1, data preparation: the data required to be prepared in the invention comprises four types, wherein the first three types of data are independent variables, and the last type of data are dependent variables;
s2, wavelet transformation: firstly, performing continuous wavelet transformation on the GRACE and GRACE-FO gravity satellite data in the step S1 to construct a time-frequency information (shown as formula 1) with good time domain and frequency domain localization, wherein the basic wavelet selects morlet wavelet, or the type of the basic wavelet can be adjusted according to the inversion result of the final soil humidity, the wavelet transformation scale at least ranges from 1 to 24, and a high-frequency signal is selected as a model input variable.
S3, data normalization processing: normalizing the time-frequency information of the meteorological data, the vegetation data, the soil water data and the gravity satellite data in the step S2 to improve the model construction and verification efficiency, wherein the normalization method can be that the data is zoomed to 0 to 1 (formula 2) through the maximum value and the minimum value in the sequence, or zoomed to-1 to 1 (formula 3) through the maximum absolute value of the data,
X new ＝ (X – X min ) / (X max – X min ) Equation 2
X new (| X |) equation 3;
s4, model data set division: dividing all the normalized time sequence data in the step S3 into a training set, a verification set and a test set, wherein the division ratio is 0.7;
s5, determining super parameters of the long-term and short-term memory network model: the inversion efficiency is improved by using a super parameter automatic optimization method, the super parameters mainly relate to seven items, generally speaking, the more the parameters are selected or the larger the parameter value is, the inversion accuracy of the soil water content is improved, but the higher the configuration requirements of the model on a computer CPU and a GPU are;
s6, building a long-term and short-term memory network model architecture: building a model architecture composed of different parameters by using a Bayesian optimal parameter filter combination and the super parameters determined in the filtering step S5, and inputting the training set data generated in the step S4 for training LSTM models of different architectures;
s7, determining an optimal long-short term memory network model: substituting the verification set data generated in the step S4 into the network models with different architectures generated in the step S6, selecting the model with the minimum regression loss as the optimal model, and storing corresponding optimal model parameters;
s8, model testing and evaluating: substituting the test set data generated in the step S4 into the optimal long-term and short-term memory network model generated in the step S7 to obtain a model simulation value (Y) of the soil water content LSTM ) Then, based on the measured data of the soil water content in the step S1 and the like (Y), Y is evaluated LSTM The evaluation index is selected from Root Mean Square Error (RMSE), correlation Coefficient (CC) and Standard Error (SE).
The first type in the S1 is meteorological data, the first type comprises monthly rainfall and monthly average air temperature with the height of 2m on the ground, the rainfall and the air temperature are contents which are necessary to be measured by a national or local meteorological station and are easy to collect, when a meteorological station is lacked in a research range, precipitation and air temperature data can be extracted through meteorological grid products issued by national or international meteorological organizations, the second type is vegetation index monthly homogeneous data, EVI can be selected when vegetation in a research area is sparse, otherwise NDFO can be selected, the third type is GRACE and GRACE-gravity satellite data, the fourth type is soil humidity data used for constructing a model and verifying the precision of the model, the soil humidity data can be measured in the field, and the soil humidity data can also be hydrological model product data.
Ψ (t) in S2 is a wavelet mother function, a function having continuous properties in both the time domain and the frequency domain, a (- ∞, + ∞) is a translation position, and b (0, + ∞) is a scaling factor.
Generally, random division is preferred, but considering that a certain time interval exists between the retirement of the GRACE gravity satellite and the transmission of the GRACE-FO gravity satellite, the data of the service period (2002.03-2017.10) of the GRACE gravity satellite can be divided into a training set and a verification set, and the data of the service period (2018.05-so-far) of the GRACE-FO satellite can be divided into a test set.
The main super parameters and values in the S5 are as follows: (1) selecting relu, sigmoid, tanh, selu, gelu and leak _ relu (alpha = 0.2) as an activation function; (2) the number of output units of the hidden layer is 32-512, and the step length is 32; (3) the data traversal times are at least 200; (4) the rejection ratio of data for preventing overfitting is 0-0.25; (5) taking an absolute error average value from the loss function; (6) the optimization algorithm selects Nadam or adam, and the Nadam or adam is more time-consuming; (7) the number of LSTM layers is 1-10.
Currently, related research mainly focuses on inversion of groundwater changes by gravity satellites (application publication No. CN113887064A, application publication No. 2022.01.04, china; application publication No. CN111752934A, application publication No. 2020.10.09, china; application publication No. CN106529164A, application publication No. 2017.03.22, china) and flood assessment (application publication No. CN114708516A, application publication No. 2022.07.05, china).
Embodiment 1:
the implementation site is as follows: the city of autonomous region of inner Mongolia Darlan unites the Yanghe county.
The implementation time is as follows: 9 months in 2022.
Climate conditions are as follows: in semiarid regions, the average precipitation for many years is 260mm.
Landform characteristics: high plain, agricultural and pastoral interlaced zone.
Preparing data: precipitation and air temperature data come from a bailing temple meteorological station (37 kilometers away from an implementation site); the soil moisture content for modeling and verifying inversion accuracy is obtained through field actual measurement (time domain reflectometry); calculating and obtaining vegetation EVI through Landsat series satellite images; GRACE and GRACE-FO gravity satellite data adopt the latest version land water reserve change grid data respectively issued by the United states air jet propulsion laboratory, the Boettan geographic research center and the spatial research center of the university of Texas.
Data preprocessing: the gravity satellite data of the 2002-2021 times released by three organizations are averaged over time and then the first 12 groups of high frequency signals are extracted as model input variables by morlet wavelet transform. And carrying out maximum and minimum normalization processing on all the data.
Selecting super parameters: the activation function selects relu, sigmoid, tanh, selu, gelu and leak _ relu (alpha = 0.2), and the default function is relu; the number of output units of the hidden layer is 32-512, and the step length is 32; taking 400 data traversal times; the data rejection ratio (dropout) for preventing overfitting is 0-0.25; taking an absolute error average value as a loss function; selecting Nadam by an optimization algorithm; the number of LSTM layers is 1-10.
Model construction: dividing all normalized data into a training set, a verification set and a test set (random screening) according to the proportion of 70%, 15% and 15%; then screening super parameters by using a Bayesian parameter screening device, constructing an LSTM framework, substituting training set data to train an LSTM model, determining an optimal model under the framework through verification set data, and temporarily storing model parameters and simulation precision; repeating the parameter screening, training and verifying process for 20 times, determining the optimal model of the simulation, and storing relevant parameters of the model.
And (3) testing a simulation result: substituting the test set data into the optimal model to obtain a soil water content simulation value; in the simulation, the correlation coefficient digit of the simulation value and the actually measured soil water content is 0.78 (the p value is less than 0.02), the root mean square error is 2.95, the fitting precision is high, and the requirements of related scientific research and production can be met.
Embodiment 2:
the implementation site is as follows: inner Mongolia autonomous region SilingGuo in union, silingHaite.
The implementation time is as follows: 9 months in 2022.
Climate conditions: in semiarid regions, the average precipitation is about 340mm for many years.
Landform characteristics: mongolian plateau, typical grassland area.
Preparing data: precipitation and air temperature data come from a great meteorological station (44 kilometers away from an implementation place); the soil moisture content for modeling and verifying inversion accuracy is obtained through field actual measurement (time domain reflectometry); calculating and obtaining vegetation NDVI through Landsat series satellite images; GRACE and GRACE-FO gravity satellite data adopt the latest version land water reserve change grid data respectively issued by the United states air jet propulsion laboratory, the Boettan geographic research center and the spatial research center of the university of Texas.
Data preprocessing: the gravity satellite data of 2002-2021 periods published by three organizations are averaged according to time, and then the first 15 groups of high-frequency signals are extracted as model input variables through Morlet wavelet transformation. And carrying out maximum and minimum normalization processing on all the data.
Selecting super parameters: the activation function selects relu, sigmoid, tanh, selu, gelu and leak _ relu (alpha = 0.2), and the default function is relu; the number of output units of the hidden layer is 32-512, and the step length is 32; the data traversal times are 350; the rejection ratio (dropout) of data for preventing overfitting is 0-0.2; taking an absolute error average value as a loss function; selecting Nadam by an optimization algorithm; the number of LSTM layers is 1-12.
Model construction: dividing all normalized data into a training set, a verification set and a test set (random screening) according to the proportion of 70%, 15% and 15%; then, screening super parameters by using a Bayesian parameter screening device, constructing an LSTM framework, substituting training set data to train an LSTM model, and determining an optimal model under the framework, temporarily storing model parameters and simulation precision through verification set data; repeating the parameter screening, training and verifying processes for 25 times, determining the optimal model of the simulation, and storing relevant parameters of the model.
And (4) testing a simulation result: substituting the test set data into the optimal model to obtain a soil water content simulation value; in the simulation, the correlation coefficient digit of the simulation value and the actually measured soil water content is 0.81 (the p value is less than 0.01), the root mean square error is 2.71, the fitting precision is high, and the requirements of related scientific research and production can be met
Compared with the related technology, the method for inverting the water content of the soil based on the data such as the meteorological satellite and the gravity satellite has the following beneficial effects:
the invention provides a method for inverting soil water content based on data such as meteorological and gravity satellites, and the soil water content inversion method provided by the invention can acquire the time-space change characteristics of large-range soil water without field sample collection and test or the arrangement of a large number of monitoring stations;
the soil water content inversion method provided by the invention expands the land hydrological application of the gravity satellite data from the directions of groundwater inversion, disaster monitoring and the like to soil water inversion analysis, and further releases the research potential of the gravity satellite data in a water circulation basin;
all input data in the soil water content inversion method can be freely obtained from domestic and foreign meteorological or hydrological organizations without spending any cost;
the soil water content inversion method provided by the invention is suitable for various landforms such as mountains, plateaus, plains, deserts and the like, and is not limited by conditions such as sparse vegetation or dense vegetation on the earth surface in space;
the soil water content inversion method provided by the invention is not limited by any meteorological conditions;
except for individual months with missing gravity satellite data such as GRACE, the method provided by the invention can obtain monthly data of the soil water content of any place in 2002 and 4 months, and the time sequence of the soil water content is longer than that of soil water content obtained by utilizing radar image inversion;
the soil water content inversion method provided by the invention introduces a Bayesian parameter screener. Through comparison tests in the same software and hardware computing environment, the method can save the running time by more than 1/5 compared with the manual parameter adjusting method;
by applying the soil water content inversion method provided by the invention to a fixed area, the framework and the optimal parameters of the inversion model are determined for the first time, and the optimal model is called for the subsequent inversion;
through tests, the correlation coefficient of the soil water content obtained by the method and the tested soil water content can reach more than 0.7 (the p value is less than 0.05), and the root mean square error is about 3.5.
Second embodiment
Referring to fig. 4, 5, 6 and 7, a method for inverting the water content of soil based on the data of meteorological and gravity satellites according to a first embodiment of the present disclosure, and a method for inverting the water content of soil based on the data of meteorological and gravity satellites according to a second embodiment of the present disclosure are provided. The second embodiment is only the preferred mode of the first embodiment, and the implementation of the second embodiment does not affect the implementation of the first embodiment alone.
Specifically, the difference of the method for inverting the water content of the soil based on the data such as the weather and the gravity satellite provided by the second embodiment of the application lies in that, the method for inverting the water content of the soil based on the data such as the weather and the gravity satellite is provided, modeling equipment is required to be used in S2 during modeling, the modeling equipment comprises a computer main body 1, an installation component 2 is arranged on the surface of the computer main body 1, the installation component 2 comprises an installation frame 21, and support seats 22 are connected to two sides of the surface of the installation frame 21.
The mounting bracket 21 is mounted on the surface of the computer body 1, and the support base 22 is used to keep a distance between the computer body 1 and an object.
The bottom of mounting bracket 21 is provided with extension subassembly 3, extension subassembly 3 includes two slide bars 31, two the both sides on slide bar 31 surface are all overlapped and are equipped with sliding sleeve 32, two sliding sleeve 32's both sides all are connected with connecting plate 33.
Both ends at two slide bars 31 all are connected with the fixed block, and the one end of fixed block is connected with the bottom of mounting bracket 21, uses slide bar 31 and sliding sleeve 32 to be convenient for drive connecting plate 33 that has telescopic component 4 and removes to the bottom of computer main body 1.
Two one side of connecting plate 33 all is connected with extension board 34, the surface of extension board 34 is provided with telescopic component 4, telescopic component 4 includes a plurality of telescopic links 41, and is a plurality of be connected with fixed plate 42 between the telescopic link 41.
The use of the telescopic rod 41 facilitates the movement of the fixing plate 42 to be flush with the computer main body 1.
The center position of the surfaces of the two slide bars 31 is connected with a separating block 7.
One side of mounting bracket 21 is provided with places subassembly 6, it includes two intercommunication pieces 61 to place subassembly 6, two the inside of intercommunication piece 61 all is provided with spliced pole 62, two be connected with between the spliced pole 62 and place board 63, the surface bonding of placing board 63 has foam-rubber cushion 64, the surface of spliced pole 62 is provided with adapter sleeve 65.
During the use, will drive placing board 63 and the coincidence of communicating block 61 of spliced pole 62 and foam-rubber cushion 64 and be connected, be connected with communicating block 61 when spliced pole 62, reuse adapter sleeve 65 and spliced pole 62 after spliced pole 62 and the contact of communicating block 61 are connected and can be used.
The working principle of the method for inverting the water content of the soil based on the data of the meteorological satellite, the gravity satellite and the like is as follows:
during the use, the extension board 34 that drives flexible subassembly 4 through the pulling moves to the outside of computer body 1, and when extension board 34 moved to the outside of computer body 1, it moved on the surface of slide bar 31 to drive sliding sleeve 32 through connecting plate 33, after extension board 34 moved to the outside of computer body 1, rethread pulling fixed plate 42 upwards moved under the effect of telescopic link 41, can use after fixed plate 42 moved to and flushly with computer body 1.
Compared with the related technology, the method for inverting the water content of the soil based on the data such as the meteorological satellite and the gravity satellite has the following beneficial effects:
the invention provides a method for inverting the water content of soil based on data such as meteorological and gravity satellites, wherein an extension component 3 with a fixing plate 42 is arranged at the bottom of an installation frame 21 on the surface of a computer main body 1, so that an operator can place data conveniently when using the computer main body 1, and a placing component 6 is arranged on one side of the installation frame 21, so that the operator can lift the hands conveniently when using the computer main body 1.
The above description is only an embodiment of the present invention, and is not intended to limit the scope of the present invention, and all equivalent structures or equivalent processes performed by the present invention or directly or indirectly applied to other related technical fields are included in the scope of the present invention.
Claims (10)
1. A method for inverting the water content of soil based on data such as meteorological and gravity satellites is characterized by comprising the following steps:
s1, data preparation: the data required to be prepared in the invention comprises four types, wherein the first three types of data are independent variables, and the last type of data are dependent variables;
s2, wavelet transformation: firstly, continuous wavelet transformation is carried out on GRACE and GRACE-FO gravity satellite data in the step S1 to construct time-frequency information with good time domain and frequency domain localization, morlet wavelets are selected as basic wavelets, the types of the basic wavelets can also be adjusted according to the final soil humidity inversion result, the wavelet transformation scale at least ranges from 1 to 24, and high-frequency signals are selected as model input variables.
S3, data normalization processing: normalizing the time-frequency information of the meteorological data, the vegetation data, the soil water data and the gravity satellite data in the step S2 to improve the model construction and verification efficiency, wherein the normalization method can be that the data is zoomed to 0 to 1 (formula 2) through the maximum value and the minimum value in the sequence, or zoomed to-1 to 1 through the maximum absolute value of the data,
X new ＝ (X – X min ) / (X max – X min ) Equation 2
X new (| X |) equation 3;
s4, model data set division: dividing all the normalized time sequence data in the step S3 into a training set, a verification set and a test set, wherein the division ratio is 0.7;
s5, determining super parameters of a long-short term memory network model (LSTM): the inversion efficiency is improved by using a super parameter automatic optimization method, model super parameters mainly relate to seven items, generally speaking, the more the parameters are selected or the larger the parameter value is, the inversion accuracy of the soil water content is improved, but the higher the configuration requirements of the model on a computer CPU and a GPU are;
s6, building a long-term and short-term memory network model architecture: building a model architecture composed of different parameters by using a Bayesian optimal parameter filter combination and the super parameters determined in the filtering step S5, and inputting the training set data generated in the step S4 for training network models of different architectures;
s7, determining an optimal long-term and short-term memory network model: substituting the verification set data generated in the step S4 into the different architecture models generated in the step S6, selecting the model with the minimum regression loss as the optimal model, and storing corresponding optimal model parameters;
s8, model testing and evaluating: and substituting the test set data generated in the step S4 into the optimal long-short term memory network model generated in the step S7 to obtain soil water content data simulated by the model, and then evaluating the performance of the soil water content obtained by the model simulation by taking the actual soil water content data and the like in the step S1 as references, wherein the evaluation indexes select root mean square error, correlation coefficient and standard error.
2. The method for inverting the water content of the soil based on the data of the meteorological satellite, the gravity satellite and the like in the S1 is characterized in that the first type of the S1 is meteorological data which comprises monthly precipitation and monthly average air temperature with the height of 2m on the ground, the precipitation and the air temperature are contents which are necessary to be measured by a national or local meteorological station and are easy to collect, precipitation and air temperature data can be extracted through meteorological grid products issued by national or international meteorological organizations when a meteorological site is absent in a research range, the second type of the S1 is vegetation index monthly homogeneous data, an Enhanced Vegetation Index (EVI) can be selected when vegetation in a research area is sparse, otherwise, a normalized vegetation index (NDVI) can be selected, the third type of the S1 is GRACE and GRACE-FO gravity satellite data, the fourth type of the S is soil humidity data used for constructing a model and verifying the precision of the model, the soil humidity data actually measured in the field can be soil humidity data, and the data can also be hydrological model product data.
3. The method for inversion of water content in soil based on meteorological and gravitational satellite data as claimed in claim 1, wherein Ψ (t) in S2 is a wavelet mother function, a function having continuous properties in both time domain and frequency domain, a (— infinity, + ∞) is a translation position, and b (0, + ∞) is a scaling factor.
4. The method for inversion of soil water content based on meteorological and gravity satellite data as claimed in claim 1, wherein in S4, random division is preferably selected in general, but considering that there is a certain time interval between the decommissioning of the GRACE gravity satellite and the launching of the GRACE-FO gravity satellite, the data of the service period of the GRACE gravity satellite can be divided into a training set and a verification set, and the data of the service period of the GRACE-FO satellite can be divided into a test set.
5. The method for inverting the water content of the soil based on the data of the meteorological satellite, the gravity satellite and the like as claimed in claim 1, wherein the main super parameters and values thereof in the S5 are as follows: (1) selecting relu, sigmoid, tanh, selu, gelu and leak _ relu (alpha = 0.2) as an activation function; (2) the number of output units of the hidden layer is 32-512, and the step length is 32; (3) the data traversal times are at least 200; (4) the rejection ratio of data for preventing overfitting is 0-0.25; (5) taking an absolute error average value as a loss function; (6) the optimization algorithm selects Nadam or adam, and the former is more time-consuming; (7) the number of LSTM layers is 1-10.
6. The method for inverting the water content of the soil based on the data of the meteorological satellite, the gravity satellite and the like as claimed in claim 1, wherein modeling equipment is required to be used for modeling in the S2, the modeling equipment comprises a computer main body, a mounting assembly is arranged on the surface of the computer main body, the mounting assembly comprises a mounting frame, and supporting seats are connected to two sides of the surface of the mounting frame.
7. The method for inverting the water content of the soil based on the data of the meteorological satellite, the gravity satellite and the like as claimed in claim 6, wherein an extension assembly is arranged at the bottom of the mounting frame, the extension assembly comprises two sliding rods, sliding sleeves are sleeved on two sides of the surfaces of the two sliding rods, and connecting plates are connected to two sides of the two sliding sleeves.
8. The method for inverting the water content of the soil based on the data of the meteorological satellite, the gravity satellite and the like as claimed in claim 7, wherein an extension plate is connected to one side of each of the two connecting plates, a telescopic assembly is arranged on the surface of each extension plate, each telescopic assembly comprises a plurality of telescopic rods, and a fixing plate is connected between the plurality of telescopic rods.
9. The method for inverting the water content of the soil based on the data of the meteorological and gravitational satellites and the like as claimed in claim 7, wherein a separation block is connected to the center position of the surfaces of the two sliding rods.
10. The method for inverting the water content of soil based on the data of meteorological and gravity satellites and the like according to claim 6, wherein a placing component is arranged on one side of the mounting rack and comprises two communicating blocks, connecting columns are arranged inside the two communicating blocks, a placing plate is connected between the two connecting columns, a sponge mat is bonded on the surface of the placing plate, and connecting sleeves are arranged on the surfaces of the connecting columns.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211621963.4A CN115906656A (en) | 2022-12-16 | 2022-12-16 | Method for inverting soil water content based on meteorological and gravity satellite data and other data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211621963.4A CN115906656A (en) | 2022-12-16 | 2022-12-16 | Method for inverting soil water content based on meteorological and gravity satellite data and other data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115906656A true CN115906656A (en) | 2023-04-04 |
Family
ID=86480585
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211621963.4A Pending CN115906656A (en) | 2022-12-16 | 2022-12-16 | Method for inverting soil water content based on meteorological and gravity satellite data and other data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115906656A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229285A (en) * | 2023-05-06 | 2023-06-06 | 深圳大学 | Soil water content monitoring method integrating Internet of things data and space scene |
| CN116451142A (en) * | 2023-06-09 | 2023-07-18 | 山东云泷水务环境科技有限公司 | Water quality sensor fault detection method based on machine learning algorithm |
| CN117152629A (en) * | 2023-08-23 | 2023-12-01 | 武汉大学 | Method and system for filling gravity vacancy data of drainage basin scale time-varying satellite |
-
2022
- 2022-12-16 CN CN202211621963.4A patent/CN115906656A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229285A (en) * | 2023-05-06 | 2023-06-06 | 深圳大学 | Soil water content monitoring method integrating Internet of things data and space scene |
| CN116229285B (en) * | 2023-05-06 | 2023-08-04 | 深圳大学 | Soil water content monitoring method integrating Internet of things data and space scene |
| CN116451142A (en) * | 2023-06-09 | 2023-07-18 | 山东云泷水务环境科技有限公司 | Water quality sensor fault detection method based on machine learning algorithm |
| CN117152629A (en) * | 2023-08-23 | 2023-12-01 | 武汉大学 | Method and system for filling gravity vacancy data of drainage basin scale time-varying satellite |
| CN117152629B (en) * | 2023-08-23 | 2024-03-22 | 武汉大学 | Method and system for filling gravity vacancy data of drainage basin scale time-varying satellite |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115906656A (en) | Method for inverting soil water content based on meteorological and gravity satellite data and other data | |
| Zaitchik et al. | Assimilation of GRACE terrestrial water storage data into a land surface model: Results for the Mississippi River basin | |
| Zhang et al. | Assimilation of precipitation-affected radiances in a cloud-resolving WRF ensemble data assimilation system | |
| Dai et al. | Development of a China dataset of soil hydraulic parameters using pedotransfer functions for land surface modeling | |
| Koch et al. | Inter-comparison of three distributed hydrological models with respect to seasonal variability of soil moisture patterns at a small forested catchment | |
| CN109709577A (en) | A kind of Three-dimensional Variational Data Assimilation method of the aerosol LIDAR inverting PM2.5 based on WRF-Chem mode | |
| Loew et al. | A dynamic approach for evaluating coarse scale satellite soil moisture products | |
| CN105243435B (en) | A kind of soil moisture content prediction technique based on deep learning cellular Automation Model | |
| CN107390298A (en) | A kind of analogy method and device of Complex Mountain underlying surface strong wind | |
| CN110363327A (en) | Short based on ConvLSTM and 3D-CNN faces Prediction of Precipitation method | |
| CN106404620A (en) | Method and system for inverting ground PM 2.5 through combination of geostatistical interpolation and satellite remote sensing | |
| De Lannoy et al. | Soil moisture data assimilation | |
| CN105069295B (en) | Satellite and surface precipitation measured value assimilation method based on Kalman filtering | |
| CN105067534A (en) | Pollutant transport flux measurement and calculation method based on ground-based MAX-DOAS | |
| CN112785024A (en) | Runoff calculation and prediction method based on watershed hydrological model | |
| CN108875905A (en) | A kind of visibility function Direct Inverse Method of Atmosphere and humidity profiles | |
| Gu et al. | Hydrological assessment of TRMM rainfall data over Yangtze River Basin | |
| Liu et al. | Wind dynamics over a highly heterogeneous oasis area: An experimental and numerical study | |
| CN105975702B (en) | Cable-stayed bridge health monitoring sensors optimum placement method and system | |
| Vich et al. | Ensemble prediction of Mediterranean high-impact events using potential vorticity perturbations. Part I: Comparison against the multiphysics approach | |
| CN113723000A (en) | Farmland soil moisture simulation method based on Sentinel data and deep learning model | |
| Goodrich | Basin scale and runoff model complexity | |
| CN115840908B (en) | Method for constructing PM2.5 three-dimensional dynamic monitoring field based on microwave link of LSTM model | |
| CN106154309A (en) | Short term prediction method in a kind of network earthquake December | |
| CN113361114B (en) | Multi-scale non-point source pollutant river entering coefficient measuring and calculating method based on runoff path |
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 |