CN117093806A - Remote sensing-based full-space coverage offshore atmosphere CO 2 Column concentration calculation method - Google Patents

Abstract

The invention discloses a remote sensing-based full-space coverage offshore atmosphere CO ₂ The invention relates to a method for calculating column concentration, in particular to CO ₂ The field of column concentration calculation methods includes: step 1: land to sea atmosphere CO ₂ Effect of column concentration the red band reflectance corrected for atmosphere can be used to construct a seawater eutrophication index, step 2: reconstructing missing data by a space-time interpolation method, and step 3: standardization processing, step 4: constructing a predicted random forest model by adopting a random forest method, and step 5: according to the downloaded data and the calculated seawater eutrophication index data, the step 2 space-time interpolation method and the step 3 standardization treatment are carried out, and the random forest model trained in the step 4 is utilized, so that all-weather and full-coverage marine atmosphere CO is predicted ₂ The method provided by the invention can effectively calculate the CO of the offshore atmosphere ₂ Column concentration.

Description

Calculation method of atmospheric CO2 column concentration at sea with full space coverage based on remote sensing

Technical field

The present invention relates to the field of CO ₂ column concentration calculation methods, and specifically relates to a remote sensing-based full space coverage maritime atmospheric CO ₂ column concentration calculation method.

Background technique

The detection of atmospheric _CO2 column concentration can be divided into two methods: ground-based and space-based. Ground-based detection generally refers to the establishment of stations for observation, which has high accuracy. For example, the Global Carbon Column Total Observation Network is an observation network composed of ground-based Fourier transform spectrometers, which can conduct CO ₂ , CH ₄ , N ₂ O, CO, Precise measurement of elements such as H ₂ O, but its disadvantage is that the current observation network has a small number of stations and is sparsely distributed around the world, which cannot reflect the distribution trend of continuous atmospheric CO ₂ in a large area. Space-based detection is generally based on certain satellites dedicated to atmospheric greenhouse gas detection, such as Japan's GOSAT and GOSAT2, the United States' OCO-2 and OCO-3, China's carbon satellite, etc., using the short-wave infrared band carried by them to detect atmospheric CO ₂ for detection, its biggest advantage is that it has global coverage detection capabilities as the satellite moves around the earth. However, its channel characteristics determine that the observation revisit period is long, and it is impossible to achieve daily full coverage detection of large areas.

In order to obtain atmospheric CO ₂ column concentration monitoring data with full space coverage, space-based detection is a low-cost method that can be implemented currently. However, the carbon satellite observation method has the problem of long revisit period. In order to overcome this problem, a common idea is to take advantage of the wide data width of medium-resolution satellites such as Fengyun Satellite and MODIS, which can achieve multiple times of daily coverage monitoring of the earth, and establish its reflection in the sea-land-atmosphere layer. The correlation between the parameters derived and the atmospheric CO ₂ column concentration obtained by carbon satellites is achieved, thereby achieving full coverage monitoring of atmospheric CO ₂ column concentration based on medium-resolution satellite data.

However, there are still two problems in space-based detection:

(1) The carbon cycle process involves multiple layers such as the biosphere, lithosphere, hydrosphere and atmosphere. Although the concentration of CO ₂ in the atmosphere is relatively stable, it is significantly affected by the underlying surface. Between sea-air and land-air Differences in _CO2 partial pressure differences will lead to spatiotemporal differences in _CO2 concentration in the atmosphere. The current focus on _CO2 column concentration is over land, and less attention is paid to over oceans. There are many differences between the ocean and land. The monitoring method of _CO2 column concentration over land is not suitable for the ocean. Therefore, it is necessary to construct a new method for monitoring atmospheric _CO2 column concentration suitable for the ocean.

(2) The monitoring of atmospheric and oceanic parameters by medium-resolution satellite data such as MODIS is usually limited to clear-sky areas. Cloud obstruction in cloudy areas will cause data loss. Therefore, daily monitoring cannot be achieved when there are clouds in the monitoring area. Full spatial coverage of atmospheric _CO2 column concentration monitoring.

To this end, we provide a remote sensing-based full-space coverage maritime atmospheric CO ₂ column concentration calculation method to solve the above problems.

Contents of the invention

In view of the problems existing in the above-mentioned existing technologies, the present invention provides a remote sensing-based full-space coverage of the ocean atmospheric CO ₂ column concentration calculation method, which solves the lack of full-space coverage of atmospheric CO ₂ column concentration monitoring due to lack of data such as atmospheric and ocean parameters. The problem.

In order to achieve the above goals, the present invention adopts a remote sensing-based calculation method for the full-space atmospheric CO ₂ column concentration covering the sea, including:

Taking MODIS data as an example, the downloaded data includes three types of MODIS data. The first type is maritime parameters: sea surface temperature: SST, chlorophyll a concentration: Chl-a, photosynthetically active radiation: PAR; the second type is atmospheric parameters: aerosols Optical thickness: AOT; the third category is band reflectivity: atmospherically corrected red light band reflectance, and XCO ₂ data from the OCO-2 satellite. XCO ₂ data is the atmospheric carbon dioxide column concentration;

Step 1: The influence of land on the atmospheric _CO2 column concentration at sea can be used to construct the seawater eutrophication index using the atmospherically corrected red light band reflectance;

Step 2: Reconstruct the missing data through spatiotemporal interpolation method;

Step 3: Standardized processing, sea surface temperature, chlorophyll a concentration, photosynthetically active radiation, aerosol optical depth, seawater eutrophication index parameters and XCO ₂ data of OCO-2 are unified to a spatial resolution of 0.05°, where sea surface temperature, Chlorophyll a concentration, photosynthetically active radiation, and aerosol optical thickness were interpolated using the nearest neighbor method, and seawater eutrophication index and XCO ₂ data were interpolated using the bilinear method and unified into equal longitude and latitude projections;

Step 4: Use the random forest method to build a predictive random forest model, with sea surface temperature, chlorophyll a concentration, photosynthetically active radiation, aerosol optical depth, and seawater eutrophication index as independent variables, and XCO ₂ data as the dependent variable:

Step 5: Based on the downloaded sea surface temperature, chlorophyll a concentration, photosynthetically active radiation, aerosol optical thickness data, and calculated seawater eutrophication index data, go through the spatiotemporal interpolation method in step 2 and the standardization process in step 3, and then use The random forest model trained in step 4 predicts all-weather, full-coverage maritime atmospheric CO ₂ column concentration data.

As a further optimization of the above scheme, the chlorophyll a concentration, photosynthetically active radiation, aerosol optical thickness, and seawater eutrophication index were all reconstructed through spatiotemporal interpolation method to reconstruct the missing data.

As a further optimization of the above scheme, the seawater eutrophication index in step 1 is expressed by SEI:

(1)

Among them, SEI is the estimated seawater eutrophication index, is the atmospherically corrected red light band reflectance.

As a further optimization of the above solution, in step 2, the missing data is reconstructed through the spatiotemporal interpolation method, taking sea surface temperature as an example for illustration:

First, set the sea surface temperature data that needs to be reconstructed for multiple consecutive days in the study area as a matrix , matrix/> The rows are all time series values of a certain spatial location point, the columns are the values of all spatial points at a certain moment, I is the set of missing points to be reconstructed, and the missing values are represented by NaN. The workflow is as follows:

Subtract the effective mean of its time dimension/> , the effective average value of the time dimension is the average value of the time dimension under the sea surface temperature data without missing values, and X is obtained. 1% of the total effective data is randomly taken from ,right The data at the corresponding position is assigned NaN, replace all NaN points in X with 0, define P as the modal retention number, at this time P=1;

pair matrix Perform singular value decomposition, we have

(2)

Use equation (3) to fill in missing point data (3);

Then calculate according to equation (4) with cross-validation set/> The root mean square error R:

(4);

In order to minimize the root mean square error R, let

(5);

Reuse equation (1) for Perform singular value decomposition and repeat step 3 until the root mean square error R converges;

Let the modal retention number P=1,2,…, , repeat formulas (2)-(5), and record the root mean square error under the corresponding P value , at this time there is always a P value that can make/> Minimum, take this P value as the optimal mode retention number k;

Take the optimal mode retention number k to reconstruct the missing data, and the resulting matrix is recorded as ,/> with matrix/> The difference is,/> The missing value NaN has been reconstructed,/> Add/> to each column of , get the final reconstruction matrix;

The sea surface temperature data in areas with missing data are reconstructed through equations (2)-(5), which together with the existing clear sky data form all-weather, full-coverage sea surface temperature data.

For further optimization of the above scheme, in formula (2): ,/> ,/> are respectively the corresponding spatial eigenmodes, singular value matrices and time eigenmodes after SVD decomposition,/> Represents matrix transpose.

As a further optimization of the above scheme, in formula (3): ,/> and/> are the tth column of the spatial and temporal eigenmodes respectively,/> is the corresponding singular value,/> is the missing point data, and P is the number of retained modes.

As a further optimization of the above scheme, in formula (4): N is the cross-validation set The number of data points, R is/> with cross-validation set/> root mean square error,/> is the reconstructed value of missing point data,/> is the original value of the missing point data.

As a further optimization of the above scheme, in formula (5): is the correction value matrix of missing points.

As a further optimization of the above solution, the random forest model formula in step 4 is:

(6).

As a further optimization of the above solution, in step 4, according to the requirements of the random forest model, the number of leaves is set to 5, the number of trees is 70, 90% of the data set is randomly selected as the training set, and the remaining 10% is used as the test set. To evaluate the accuracy of the prediction model, the randomly selected data set refers to 5 independent variables and 1 dependent variable at multiple times after processing in steps 1, 2, and 3.

The present invention's remote sensing-based full-space ocean atmospheric CO ₂ column concentration calculation method has the following beneficial effects:

The method proposed by the present invention can effectively calculate the CO ₂ column concentration in the sea atmosphere.

Fully considering the influence of various sea-land-air elements involved in the carbon cycle on the concentration of _CO2 column in the sea atmosphere, the seawater eutrophication index proposed by the present invention can effectively reflect the impact of offshore suspended sediment and organic matter on the oceanic _CO2 partial pressure. , which can better explain the influence of coastal atmospheric CO ₂ on coastal land.

This invention uses the spatiotemporal interpolation method to first reconstruct the missing data, providing the necessary foundation for all-weather, full-coverage calculation of the CO ₂ column concentration of the maritime atmosphere.

With reference to the following description and drawings, specific embodiments of the present invention are disclosed in detail and the manner in which the principles of the present invention may be adopted is indicated. It should be understood that the scope of the embodiments of the present invention is not thereby limited. Embodiments of the present invention include many alterations, modifications and equivalents within the spirit and scope of the appended claims.

Description of the drawings

Figure 1 is a schematic flow diagram of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below through the drawings and examples. However, it should be understood that the specific embodiments described here are only used to explain the present invention and are not used to limit the scope of the present invention.

It should be noted that when an element is said to be "disposed on, provided with" another element, it can be directly on the other element or there may also be intervening elements. When one element is said to be "connected to, connected to" another element A component, which can be directly connected to another component or there may be an intermediate component at the same time. "Fixed connection" means fixed connection. There are many ways of fixed connection, which are not within the scope of this article. The terms used in this article “Vertical”, “horizontal”, “left”, “right” and similar expressions are for illustrative purposes only and do not represent the only implementation manner.

Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the technical field belonging to the present invention. The terms used in the specification are only for the purpose of describing specific embodiments and not For the purpose of limiting the invention, the term "and/or" as used herein includes any and all combinations of one or more of the associated listed items;

Please refer to Figure 1 of the description. The present invention provides a technical solution: a method for calculating the atmospheric CO ₂ column concentration based on remote sensing covering the whole space covering the sea. In the context of the carbon cycle, the present invention provides a technical solution: the atmospheric CO ₂ column concentration changes in time and space. Therefore, the CO ₂ column concentration in the sea and over land needs to be monitored in order to have an accurate understanding of the total global CO ₂ concentration. The methods in the prior art are all aimed at atmospheric CO ₂ over land. The method proposed by the present invention can effectively calculate the CO 2 concentration over the sea. Atmospheric _CO2 column concentration:

Taking MODIS data as an example, the data used in this invention include three types of MODIS data. The first type is maritime parameters: sea surface temperature (SST), chlorophyll a concentration (Chl-a), and photosynthetically active radiation (PAR); is the atmospheric parameter: aerosol optical thickness (AOT); the third category is the band reflectance: the atmospherically corrected red light band (Band1, 620–670nm) reflectance; and the XCO ₂ data of the OCO-2 satellite, the XCO ₂ data is Atmospheric carbon dioxide column concentration. The above data can be downloaded from the website.

Step 1: Affected by differences in the distribution of nutrients contained in seawater, the distribution of phytoplankton on the surface of seawater also has spatial and temporal differences. Phytoplankton can absorb and release CO ₂ through photosynthesis and respiration, affecting the CO ₂ partial pressure of seawater, and then through the sea- Air CO ₂ exchange affects the atmospheric CO ₂ column concentration. In offshore areas, due to human activities and the large amounts of sediment and nutrients carried by rivers entering the sea, red tides often break out due to eutrophication. At this time, the partial pressure of CO ₂ on the surface of the seawater There will be a greater impact. The impact of land on the sea atmospheric _CO2 column concentration can be expressed by using the atmospherically corrected red light band reflectance to construct the seawater eutrophication index (SEI):

(1)

Among them, SEI is the estimated seawater eutrophication index; is the atmospherically corrected red light band reflectance.

This invention fully considers the influence of various sea-land-air elements involved in the carbon cycle on the concentration of _CO2 column in the sea atmosphere. The seawater eutrophication index proposed by the invention can effectively reflect the impact of offshore suspended sediment and organic matter on the ocean _CO2 partial pressure. The influence can better explain the influence of coastal atmospheric CO ₂ on coastal land.

Step 2: Through the spatiotemporal interpolation method, sea surface temperature (SST), chlorophyll a concentration (Chl-a), photosynthetically active radiation (PAR), aerosol optical depth (AOT), and the calculated seawater eutrophication index (SEI) , can only be obtained under clear sky conditions. When the sea surface is obscured by clouds, the relevant parameters of the sea area and the atmosphere above the sea area will be missing. This step uses the spatiotemporal interpolation method to reconstruct the missing data to prepare for the next step of all-weather , laying the foundation for full coverage of atmospheric CO ₂ column concentration monitoring.

The following strategy is applicable to the above five parameters. Taking sea surface temperature as an example, other parameters are the same:

First, set the sea surface temperature data that need to be reconstructed for multiple consecutive days in the study area (assuming m spatial locations, n days) as a matrix , matrix/> The rows are all time series values of a certain spatial location point, the columns are the values of all spatial points at a certain moment, I is the set of missing points to be reconstructed, and the missing values are represented by NaN. The workflow is as follows:

1: Subtract the valid (non-NaN) mean of its time dimension/> , the effective average value of the time dimension is the average value of the time dimension under the sea surface temperature data without missing values, and we get/> , from/> Randomly take 1% of the total valid data as the cross-validation set/> , right/> The data at the corresponding position is assigned NaN, and // All NaN points in are replaced with 0, and P is defined as the modal retention number. At this time, P=1.

2: Pair matrix Perform singular value decomposition, we have

(2)

in, ,/> ,/> are respectively the corresponding spatial eigenmodes, singular value matrices and time eigenmodes after SVD decomposition,/> Represents matrix transpose.

3: Use equation (3) to fill in the missing point data

3)

in, ,/> and/> are the tth column of the spatial and temporal eigenmodes respectively,/> is the corresponding singular value, is the missing point data, P is the number of retained modes, and then it is calculated according to Equation (4)/> with cross-validation set/> The root mean square error R:

4)

Among them, N is the cross-validation set The number of data points, R is/> with cross-validation set/> The root mean square error of is the reconstructed value of missing point data,/> is the original value of the missing point data.

4: In order to minimize the root mean square error R, let

(5)

in, is the correction value matrix of the missing points, and then use equation (1) to calculate/> Perform singular value decomposition and repeat step 3 until the root mean square error R converges.

5: Let the modal retention number P=1,2,…, , repeat formulas (2)-(5), and record the root mean square error under the corresponding P value/> . At this time, there is always a P value that can make/> small, take this P value as the optimal mode retention number k.

6: Select the optimal mode retention number k to reconstruct the missing data, and the resulting matrix is recorded as ,/> with matrix/> The difference is,/> The missing values (NaN) have been reconstructed,/> Add/> to each column of , get the final reconstruction matrix.

The sea surface temperature data in areas with missing data are reconstructed through equations (2)-(5), which together with the existing clear sky data form all-weather, full-coverage sea surface temperature data.

In the present invention, sea surface temperature (SST), chlorophyll a concentration (Chl-a), photosynthetically active radiation (PAR), aerosol optical depth (AOT), and seawater eutrophication index (SEI) can only be obtained under clear sky conditions. If we only rely on these clear-sky data, it is impossible to establish an all-weather, full-coverage maritime atmospheric _CO2 column concentration monitoring algorithm. Therefore, the present invention uses the spatiotemporal interpolation method to first reconstruct the missing data, providing the necessary foundation for all-weather, full-coverage calculation of the CO ₂ column concentration in the maritime atmosphere.

Step 3: Standardization processing, the above five parameters and the XCO ₂ data of OCO-2 are unified to a spatial resolution of 0.05°, including sea surface temperature (SST), chlorophyll a concentration (Chl-a), photosynthetically active radiation (PAR) The aerosol optical thickness (AOT) is interpolated using the nearest neighbor method, and the seawater eutrophication index (SEI) and XCO ₂ data are interpolated using the bilinear method (linear) and unified into equal latitude and longitude projections.

Step 4: Use the random forest method to build a predictive random forest model, sea surface temperature (SST), chlorophyll a concentration (Chl-a), photosynthetically active radiation (PAR), aerosol optical thickness (AOT), seawater eutrophication index (SEI) ) as the independent variable and XCO ₂ as the dependent variable:

6)

Equation (6) is a conceptual model, with the dependent variable on the left and the independent variables in parentheses on the right. RF only shows that the random forest method (RF, Random Forest) is used.

According to the requirements of the random forest model, set the number of leaves to 5 and the number of trees to 70. Randomly select 90% of the data set (referring to 5 independent variables and 1 dependent variable at multiple times after steps 1, 2, and 3) as the training set, and the remaining 10% as the test set for evaluation and prediction. Model accuracy.

Step 5: In actual use, it can be based on the downloaded sea surface temperature (SST), chlorophyll a concentration (Chl-a), photosynthetically active radiation (PAR), aerosol optical depth (AOT) data, and calculated seawater enrichment Trophic index (SEI) data, through the spatiotemporal interpolation method in step 2, the standardization process in step 3, and then using the random forest model trained in step 4, all-weather, full-coverage maritime atmospheric CO ₂ column concentration data can be predicted.

They are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention. .

Claims (10)

1. Remote sensing-based full-space coverage offshore atmosphere CO ₂ A method for calculating column concentration, comprising:

taking MODIS data as an example, the downloaded data comprise three types of MODIS data, one type is offshore parameters: sea surface temperature: SST, chlorophyll a concentration: chl-a, photosynthetically active radiation: PAR; class IIIs an atmospheric parameter: aerosol optical thickness: an AOT; three categories are band reflectivities: atmospheric corrected red band reflectivity, and XCO for OCO-2 satellites ₂ Data, XCO ₂ Data, namely atmospheric carbon dioxide column concentration;

step 1: land to sea atmosphere CO ₂ The effect of column concentration can be used to construct a seawater eutrophication index using the atmospheric corrected red band reflectivity;

step 2: reconstructing missing data by a space-time interpolation method;

step 3: standardized treatments, sea surface temperature, chlorophyll a concentration, photosynthetically active radiation, aerosol optical thickness, sea water eutrophication index parameters, XCO of OCO-2 ₂ The data unify the spatial resolution to 0.05 DEG, wherein the sea surface temperature, chlorophyll a concentration, photosynthetically active radiation, aerosol optical thickness are interpolated using nearest neighbor method, sea water eutrophication index and XCO ₂ The data is interpolated by a bilinear method and is unified into equal longitude and latitude projections;

step 4: constructing a predicted random forest model by adopting a random forest method, wherein sea surface temperature, chlorophyll a concentration, photosynthetic effective radiation, aerosol optical thickness and sea water eutrophication index are taken as independent variables, and XCO is taken as a reference value ₂ Data as dependent variables:

step 5: according to downloaded sea surface temperature, chlorophyll a concentration, photosynthetic effective radiation, aerosol optical thickness data and calculated sea water eutrophication index data, through a space-time interpolation method of step 2, standardized treatment of step 3, and then utilizing a random forest model trained in step 4, all-weather and full-coverage sea atmosphere CO is predicted ₂ Column concentration data.

2. Remote sensing-based full space coverage marine atmospheric CO as defined in claim 1 ₂ The column concentration calculation method is characterized in that: the chlorophyll a concentration, photosynthetic effective radiation, aerosol optical thickness and sea water eutrophication index are all reconstructed into missing data by a space-time interpolation method.

3. Root of Chinese characterRemote sensing-based full space coverage marine atmospheric CO as defined in claim 1 ₂ The column concentration calculation method is characterized in that: the seawater eutrophication index in step 1 is represented using SEI:（1）

wherein SEI is the estimated seawater eutrophication index,is the red light band reflectivity after atmospheric correction.

4. A remote sensing based full space coverage marine atmospheric CO as defined in claim 3 ₂ The column concentration calculation method is characterized in that: in the step 2, missing data is reconstructed by a space-time interpolation method, and the sea surface temperature is taken as an example for explanation:

first, sea surface temperature data which need to be reconstructed for a plurality of continuous days in a research area are set as a matrixMatrix->The row of (1) is all time sequence values of a certain spatial position point, the column is the value of all spatial points at a certain moment, and I is a missing point set to be reconstructed, wherein the missing value is expressed by NaN, and the workflow is as follows:

subtracting the effective average of its time dimension +.>The effective average value of the time dimension is the average value of the time dimension under the sea surface temperature data without the missing value, X is obtained, and 1% of the total effective data is randomly taken out from the X to be used as a cross verification set +.>For->Assigning data of the corresponding positions as NaNs, replacing all NaN points in X with 0, and defining P as a mode reserved number, wherein P=1;

singular value decomposition of matrix X is performed, with

（2）；

Using (3) to complement missing point data

（3）；

Then calculate according to equation (4)Cross-validation set->Root mean square error R:

（4）；

to minimize the root mean square error R, let

（5）；

Reuse of the pair of (1)Performing singular value decomposition, and repeating the step 3 until the root mean square error R converges;

let the mode retention number p=1, 2, …,repeating the formulas (2) - (5), and recording the root mean square error +.>At this time, there is always a P value enabling +.>Minimum, taking the P value as an optimal mode retention number k; reconstructing the missing data by taking the optimal mode retention number k, and marking the obtained matrix as +.>And matrix->Is distinguished by->The missing values NaN have been reconstructed, < >>Add +.>Obtaining a final reconstruction matrix;

the sea surface temperature data of the area with missing data are reconstructed through the formulas (2) - (5), and all-weather and full-coverage sea surface temperature data are formed together with the existing clear sky data.

5. The remote sensing-based full space coverage marine atmospheric CO of claim 4 ₂ The column concentration calculation method is characterized in that: in formula (2): u, S, V are the corresponding spatial characteristic mode, singular value matrix and time characteristic mode after SVD decomposition, T represents matrix transposition.

6. The remote sensing-based full space coverage sea of claim 4Upper atmosphere CO ₂ The column concentration calculation method is characterized in that: in formula (3):，/>and->Column t, < > -of spatial and temporal feature modes respectively>For the corresponding singular value +.>For missing point data, P is the mode retention.

7. The remote sensing-based full space coverage marine atmospheric CO of claim 4 ₂ The column concentration calculation method is characterized in that: in formula (4): n is a cross validation setR is +.>Cross-validation set->Root mean square error of>For the reconstruction value of the missing point data, +.>Is the original value of the missing point data.

8. Tele-based according to claim 4Full space coverage of the sea atmosphere CO ₂ The column concentration calculation method is characterized in that: in formula (5):is a correction value matrix of missing points.

9. Remote sensing-based full space coverage marine atmospheric CO as defined in claim 1 ₂ The column concentration calculation method is characterized in that: the random forest model formula in the step 4 is as follows:

（6）。

10. remote sensing-based full space coverage marine atmospheric CO as defined in claim 1 ₂ The column concentration calculation method is characterized in that: in the step 4, according to the requirement of the random forest model, the number of leaves is set to be 5, the number of trees is 70, 90% of the randomly selected data set is used as a training set, the remaining 10% is used as a test set for evaluating the accuracy of the prediction model, and the randomly selected data set refers to 5 independent variables and1 dependent variable of a plurality of times after the processing in the steps 1,2 and 3.