Atmospheric carbon dioxide concentration prediction method, system, equipment and medium

Abstract

The invention discloses an atmospheric carbon dioxide concentration prediction method, system, equipment and medium, and relates to the technical field of concentration prediction, and the method comprises the steps: obtaining multi-source data of a target area; inputting the multi-source data into the convolutional neural network model, and obtaining a nonlinear relationship between the feature and the CO2 concentration under the condition that the temporal-spatial resolution is higher than a threshold value; and obtaining the temporal and spatial change rule of the CO2 concentration through the nonlinear relationship, and predicting the atmospheric carbon dioxide concentration according to the temporal and spatial change rule of the CO2 concentration. According to the method, prediction is carried out through the relation with various kinds of multi-source data influencing the CO2 concentration, the CO2 concentration of all areas can be accurately predicted through the multi-source data of the multi-area range, a CO2 concentration prediction result with higher precision is obtained, and powerful data support is provided for ecological improvement engineering.

Description

Technical Field

[0001] The present invention relates to the technical field of concentration prediction, and in particular to a method, system, equipment and medium for predicting atmospheric carbon dioxide concentration. Background Art

[0002] CO2 is the most important culprit for the global greenhouse effect. With the rapid advancement of global industrialization and the increasing frequency of human activities, the concentration of atmospheric greenhouse gases is increasing, which has a significant impact on the global climate, ecosystem and economy. In recent years, researchers in the field of atmospheric monitoring have made significant progress in deep learning-based methods.

[0003] The distribution of CO2 concentration shows a certain spatial heterogeneity. There are currently two main types of CO2 concentration products. One is the point data directly obtained by remote sensing satellites. Affected by the satellite's orbit, this type of data is mostly strip-type, and there is a lack of data in the blank areas between the strips. The other is the surface data reconstructed based on remote sensing satellite data. However, the current products are for global scales, with a resolution of only 50km, which is insufficient for provincial-scale research.

[0004] However, in the existing technology, CO2 concentration is only measured through atmospheric observation stations, and the atmospheric observation stations are relatively scattered, which makes it impossible to accurately predict the CO2 concentration in most areas and provide strong data support for ecological improvement projects. Summary of the invention

[0005] The purpose of the present invention is to provide a method, system, device and storage medium for predicting atmospheric carbon dioxide concentration in response to the above-mentioned deficiencies in the prior art, so as to solve the problem that the atmospheric observation stations in the prior art are relatively scattered, the CO2 concentration in most areas cannot be accurately predicted, and strong data support cannot be provided for ecological improvement projects.

[0006] The present invention specifically provides the following technical solutions:

[0007] A method for predicting atmospheric carbon dioxide concentration comprises the following steps:

[0008] Acquire multi-source data of the target area within a certain period of time; the multi-source data includes CO2 remote sensing data, road network data, surface meteorological data, normalized difference vegetation index NDVI data and aerosol optical depth AOD data;

[0009] Inputting the multi-source data into a convolutional neural network model to obtain a nonlinear relationship between the feature and the CO2 concentration when the temporal and spatial resolution is higher than a threshold;

[0010] The temporal and spatial variation law of CO2 concentration is obtained through the nonlinear relationship, and the atmospheric carbon dioxide concentration is predicted based on the temporal and spatial variation law of CO2 concentration.

[0011] Preferably, the step of acquiring multi-source data of the target area comprises the following steps:

[0012] Acquire observation spectrum data of a target area through a satellite, acquire CO2 concentration information of the target area through the observation spectrum data, and establish CO2 remote sensing data through the CO2 concentration information;

[0013] Establishing a Fishnet database with a resolution of q km, and extracting road network data from the Fishnet database;

[0014] Download the meteorological data of the target area from the meteorological website, and process the meteorological data of the target area through the Kriging interpolation method to obtain the surface meteorological data with a resolution of q km within the target area;

[0015] Generate NDVI data through satellite sensors and extract NDVI data corresponding to the CO2 concentration sampling points;

[0016] AOD data are obtained through atmospheric aerosol products, and AOD data corresponding to the CO2 concentration sampling point are extracted.

[0017] Preferably, after acquiring the multi-source data of the target area, the acquired multi-source data is preprocessed, specifically:

[0018] De-noising the CO2 remote sensing data to remove the outliers, and construct a CO2 concentration layer by removing the outliers from the CO2 remote sensing data;

[0019] All road network data are fused, and after the road network data is fused, the road network density in each Fishnet database unit is obtained, and the road network density is extracted to the corresponding CO2 concentration layer;

[0020] The NDVI data were resampled to make the resolution consistent with other multi-source data, and the AOD data were format converted and spatially matched.

[0021] Preferably, the multi-source data is input into a convolutional neural network model to obtain a nonlinear relationship between the feature and the CO2 concentration when the temporal and spatial resolution is higher than a threshold, comprising the following steps:

[0022] Extracting features related to CO2 concentration from the multi-source data, the features including road network density value, air temperature and wind speed value, NDVI value and AOD value;

[0023] Performing grid processing on the features to obtain feature data with consistent data format;

[0024] The features after gridding processing are input into the convolutional neural network model, and the relationship between the CO2 concentration value in the current time and space and the road network density value, air temperature and wind speed value, NDVI value and AOD value at each sampling point is compared to obtain the nonlinear relationship between the CO2 concentration value in the current time and space and the road network density value, air temperature and wind speed value, NDVI value and AOD value.

[0025] Preferably, before inputting the features into the convolutional neural network model, the following steps are also included:

[0026] The performance of the convolutional neural network model is evaluated by the average relative error, correlation coefficient and root mean square error indicators; the specific expression is:

[0027]

[0028]

[0029]

[0030] Where Y i is the actual value, Y p is the predicted value, Y mean is the predicted value of the actual value, SSR is the regression sum of squares, SST is the total sum of squares, which reflects the total deviation of the N observed values ​​of the target variable from the sample mean, SSE is the residual sum of squares, which reflects the impact of other factors other than the predicted variable on the target variable value, RMSE and MRE reflect the prediction error. The smaller the value, the smaller the error. 2 Reflects the accuracy of the model, ranging from 0 to 1. The closer it is to 1, the higher the model accuracy;

[0031] Through the optimal RMSE, MRE and R 2 Select the convolutional neural network model with the best prediction effect.

[0032] Preferably, obtaining the temporal and spatial variation law of CO2 concentration through the nonlinear relationship comprises the following steps:

[0033] Generate a CO2 concentration distribution map with a resolution higher than the threshold through a nonlinear relationship;

[0034] The generated CO2 concentration distribution map is subjected to spatiotemporal analysis to obtain the spatiotemporal variation patterns and trends of CO2 concentration.

[0035] The present invention provides an atmospheric carbon dioxide concentration prediction system, comprising:

[0036] A data collection module is used to obtain multi-source data of a target area within a certain period of time; the multi-source data includes CO2 remote sensing data, road network data, surface meteorological data, normalized difference vegetation index NDVI data and aerosol optical depth AOD data;

[0037] A model processing module, used for inputting the multi-source data into a convolutional neural network model to obtain a nonlinear relationship between the feature and the CO2 concentration when the temporal and spatial resolution is higher than a threshold;

[0038] The prediction module is used to obtain the temporal and spatial variation law of CO2 concentration through the nonlinear relationship, and predict the atmospheric carbon dioxide concentration according to the temporal and spatial variation law of CO2 concentration.

[0039] The present invention provides a computer device, including a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of the above-mentioned atmospheric carbon dioxide concentration prediction method.

[0040] The present invention provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned atmospheric carbon dioxide concentration prediction method are implemented.

[0041] Compared with the prior art, the present invention has the following significant advantages:

[0042] The present invention obtains multi-source data in the target area and analyzes the nonlinear relationship between CO2 remote sensing data and road network data, surface meteorological data, NDVI data and AOD data. It does not require intensive data from atmospheric observation stations, but makes predictions based on the relationship with various multi-source data that affect CO2 concentration. It can accurately predict the CO2 concentration of all areas through multi-source data in multiple regions, obtain more accurate CO2 concentration prediction results, and provide strong data support for ecological improvement projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of the remote sensing prediction method for atmospheric carbon dioxide concentration provided by the present invention;

[0044] Figure 2 A scatter plot of the prediction accuracy of the atmospheric carbon dioxide concentration remote sensing prediction method provided by the present invention; wherein Figure 2 (a) is the training set, Figure 2 (b) is the validation set. DETAILED DESCRIPTION

[0045] The following is a clear and complete description of the technical solutions of the embodiments of the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0046] In order to avoid the need for dense atmospheric observation stations, the present invention uses a convolutional neural network model to predict CO2 concentration and generate CO2 concentration prediction data to fill the gap in CO2 concentration data when the resolution is higher than the threshold, and accurately predict the CO2 concentration in most areas. Specifically:

[0047] like Figure 1 As shown, the present invention proposes a method for predicting atmospheric carbon dioxide concentration to solve the problems of the prior art, which specifically includes the following steps:

[0048] Step S1: Acquire multi-source data of the target area within a certain period of time; the multi-source data includes CO2 remote sensing data, road network data, surface meteorological data, normalized difference vegetation index NDVI data and aerosol optical depth AOD data.

[0049] Acquiring multi-source data of the target area includes the following steps:

[0050] The observed spectral data of the target area are obtained through the satellite, the CO2 concentration information of the target area is obtained through the observed spectral data, and the CO2 remote sensing data is established through the CO2 concentration information. The CO2 concentration remote sensing data comes from the OCO2_L2_Lite_FP.11r dataset of OCO-2. OCO-2 is a sun-synchronous satellite developed by the National Aeronautics and Space Administration (NASA). It carries a three-channel imaging grating hyperspectral CO2 detector with a spatial resolution of 1.29km×2.25km and a scanning width of 10.6km.

[0051] A Fishnet database with a resolution of q km is established, and road network data is extracted from the Fishnet database.

[0052] The meteorological data of the target area is downloaded from the meteorological website, and the meteorological data of the target area is processed by the Kriging interpolation method to obtain the surface meteorological data with a resolution of q km within the target area. The qkm here is 1 km in actual use. The meteorological data comes from the National Center for Environmental Information (NCEI) under the National Oceanic and Atmospheric Administration (NOAA) of the United States. The original csv data downloaded from the NCEI website contains all the daily average temperatures and daily average wind speeds of a specific station in a year, recorded on a daily basis. The CO2 concentration sampling point obtains the temperature and wind speed values ​​of its corresponding position through value extraction.

[0053] NDVI data is generated through satellite sensors, and the normalized vegetation index NDVI data corresponding to the CO2 concentration sampling point is extracted. Specifically: NDVI data comes from the data shared by the National Tibetan Plateau Science Data Center platform, based on the Aqua / Terra-MODIS satellite sensor MOD13Q1 product, with a data resolution of 250m.

[0054] AOD data is obtained through atmospheric aerosol products, and aerosol optical thickness AOD data corresponding to the CO2 concentration sampling point is extracted. Specifically, the AOD data uses the MOD04_3K (MODIS Terra / AquaAerosol 5-Min L2 Swath 3km) product released by NASA, which is used to obtain the optical properties and mass concentration of atmospheric aerosols in the global ocean and land environment, with a spatial resolution of 3km.

[0055] After obtaining multi-source data of the target area, the following steps are also included:

[0056] Preprocess the acquired multi-source data, including format conversion, data cleaning, spatial alignment, and temporal alignment, to ensure that all data are consistent in space and time. Specifically:

[0057] The CO2 remote sensing data is denoised to remove the existing outliers, and a CO2 concentration layer is constructed using the CO2 remote sensing data without outliers.

[0058] All road network data were fused through the intersection and fusion operations in ArcGIS software. After the road network data were fused, the road network density in each Fishnet database unit was obtained, and the road network density was extracted to the corresponding CO2 concentration layer.

[0059] The NDVI data were resampled to make the resolution consistent with other multi-source data, and the AOD data were format converted and spatially matched.

[0060] Step S2: Input the features into the convolutional neural network model to obtain the nonlinear relationship between the features and CO2 concentration when the spatiotemporal resolution is higher than the threshold.

[0061] Features related to CO2 concentration are extracted from multi-source data, including road network density, air temperature and wind speed, NDVI and AOD.

[0062] When extracting features related to CO2 concentration from multi-source data, it specifically includes:

[0063] Perform feature extraction on CO2 remote sensing data to obtain the CO2 concentration value in the current time and space; perform feature extraction on road network data to obtain the road network density value of each sampling point in the current time and space; perform feature extraction on meteorological data to obtain the temperature and wind speed values ​​of each sampling point in the current time and space; perform feature extraction on NDVI data to obtain the NDVI value of each sampling point in the current time and space; perform feature extraction on AOD data to obtain the AOD value of each sampling point in the current time and space.

[0064] S2 also includes:

[0065] The features are gridded to obtain feature data with consistent data format; different types of convolutional neural network models are used to train the extracted features.

[0066] The gridded features are input into the convolutional neural network model. The relationship between the CO2 concentration value in the current time and space and the road network density value, temperature and wind speed value, NDVI value and AOD value at each sampling point is compared to obtain the nonlinear relationship between the CO2 concentration value in the current time and space and the road network density value, temperature and wind speed value, NDVI value and AOD value.

[0067] Among them, training the convolutional neural network model includes:

[0068] The extracted multi-source data are used as the model input data set to train and verify the CO2 concentration prediction model. The model input data set is sampled and divided into training data and verification data.

[0069] The weight and bias parameters of each convolutional neural network model are continuously adjusted, and the gridded feature data is processed to capture the optimal spatial features and nonlinear relationships.

[0070] The trained convolutional neural network model is obtained through the optimal spatial features and nonlinear relationships.

[0071] Before inputting the features into the convolutional neural network model, the trained convolutional neural network model is evaluated using the validation data, which includes the following steps:

[0072] The performance of the convolutional neural network model is evaluated by the average relative error, correlation coefficient and root mean square error indicators, and the convolutional neural network model with the best prediction effect is selected; the specific expression is:

[0073]

[0074]

[0075]

[0076] Where Y iis the actual value, Y p is the predicted value, Y mean is the predicted value of the actual value, SSR is the regression sum of squares, SST is the total sum of squares, which reflects the total deviation of the N observed values ​​of the target variable from the sample mean, and SSE is the residual sum of squares, which reflects the impact of other factors other than the predicted variable on the target variable value. RMSE and MRE reflect the error of the prediction. The smaller the value, the smaller the error. 2 Reflects the accuracy of the model, ranging from 0 to 1. The closer it is to 1, the higher the model accuracy. Figure 2 (a) and Figure 2 as shown in (b).

[0077] Through the optimal RMSE, MRE and R 2 Select the convolutional neural network model with the best prediction effect.

[0078] Among them, it was found in the experiment that the CO2 prediction accuracy of the present invention is R 2 It reached 0.62, indicating that the model has good prediction performance for CO2 concentration. At the same time, its RMSE and MRE were 3.21ppm and 1.8% respectively. 2 It reached 0.69, RMSE was 2.87ppm, and MRE was 1.20%, which performed better than the verification process. Moreover, the difference between the two was small, and the risk of overfitting was relatively low.

[0079] Step S3: Obtain the temporal and spatial variation law of CO2 concentration through nonlinear relationship, and predict the atmospheric carbon dioxide concentration according to the temporal and spatial variation law of CO2 concentration.

[0080] S3 is specifically:

[0081] The temporal and spatial variation law of CO2 concentration is obtained through nonlinear relationship, including the following steps:

[0082] Comparing different algorithms in R 2 , RMSE and MRE, and obtain the optimal algorithm, that is, the convolutional neural network model with the best prediction effect.

[0083] The optimal algorithm is used to predict the multi-source data in the full range of the target area, and a CO2 concentration distribution map with a resolution higher than the threshold is generated through nonlinear relationships.

[0084] The generated CO2 concentration distribution map is subjected to spatiotemporal analysis to obtain the spatiotemporal variation patterns and trends of CO2 concentration.

[0085] Based on the above method, the present invention provides an atmospheric carbon dioxide concentration prediction system, including: a data collection module, a model processing module and a prediction module.

[0086] Among them, the data collection module is used to obtain multi-source data of the target area within a certain period of time; the multi-source data includes CO2 remote sensing data, road network data, surface meteorological data, normalized difference vegetation index NDVI data and aerosol optical depth AOD data; the model processing module is used to input the multi-source data into the convolutional neural network model to obtain the nonlinear relationship between the characteristics and CO2 concentration when the spatiotemporal resolution is higher than the threshold; the prediction module is used to obtain the spatiotemporal variation law of CO2 concentration through nonlinear relationship, and predict the atmospheric carbon dioxide concentration according to the spatiotemporal variation law of CO2 concentration.

[0087] The present invention also provides a computer device, including a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for predicting atmospheric carbon dioxide concentration.

[0088] In accordance with the disclosed embodiments, a computing device may communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth communications, etc.), or with any device (e.g., routers, modems, etc.) that enables a computing device to communicate with one or more other computing devices.

[0089] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, a method for predicting atmospheric carbon dioxide concentration is implemented.

[0090] According to the disclosed embodiments, the storage medium may be a non-volatile computer-readable storage medium, such as but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0091] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0092] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for predicting atmospheric carbon dioxide concentration, characterized in that: The steps include: Acquire multi-source data of the target area within a certain period of time; the multi-source data includes CO2 remote sensing data, road network data, surface meteorological data, normalized difference vegetation index NDVI data and aerosol optical depth AOD data; Inputting the multi-source data into a convolutional neural network model to obtain a nonlinear relationship between the feature and the CO2 concentration when the temporal and spatial resolution is higher than a threshold; The temporal and spatial variation law of CO2 concentration is obtained through the nonlinear relationship, and the atmospheric carbon dioxide concentration is predicted based on the temporal and spatial variation law of CO2 concentration.

2. A method for predicting atmospheric carbon dioxide concentration according to claim 1, characterized in that: The step of obtaining multi-source data of the target area includes the following steps: Acquire observation spectrum data of a target area through a satellite, acquire CO2 concentration information of the target area through the observation spectrum data, and establish CO2 remote sensing data through the CO2 concentration information; Establishing a Fishnet database with a resolution of q km, and extracting road network data from the Fishnet database; Download the meteorological data of the target area from the meteorological website, and process the meteorological data of the target area through the Kriging interpolation method to obtain the surface meteorological data with a resolution of q km within the target area; Generate NDVI data through satellite sensors and extract NDVI data corresponding to the CO2 concentration sampling points; AOD data are obtained through atmospheric aerosol products, and AOD data corresponding to the CO2 concentration sampling point are extracted.

3. A method for predicting atmospheric carbon dioxide concentration according to claim 2, characterized in that: After obtaining the multi-source data of the target area, the method further includes preprocessing the obtained multi-source data, specifically: De-noising the CO2 remote sensing data to remove the outliers, and construct a CO2 concentration layer by removing the outliers from the CO2 remote sensing data; All road network data are fused, and after the road network data is fused, the road network density in each Fishnet database unit is obtained, and the road network density is extracted to the corresponding CO2 concentration layer; The NDVI data were resampled to make the resolution consistent with other multi-source data, and the AOD data were format converted and spatially matched.

4. A method for predicting atmospheric carbon dioxide concentration according to claim 1, characterized in that: The multi-source data is input into a convolutional neural network model to obtain a nonlinear relationship between the feature and the CO2 concentration when the temporal and spatial resolution is higher than a threshold, comprising the following steps: Extracting features related to CO2 concentration from the multi-source data, the features including road network density value, air temperature and wind speed value, NDVI value and AOD value; Performing grid processing on the features to obtain feature data with consistent data format; The features after gridding processing are input into the convolutional neural network model, and the relationship between the CO2 concentration value in the current time and space and the road network density value, air temperature and wind speed value, NDVI value and AOD value at each sampling point is compared to obtain the nonlinear relationship between the CO2 concentration value in the current time and space and the road network density value, air temperature and wind speed value, NDVI value and AOD value.

5. The method for predicting atmospheric carbon dioxide concentration according to claim 1, characterized in that: Before inputting the features into the convolutional neural network model, the following steps are also included: The performance of the convolutional neural network model is evaluated by the average relative error, correlation coefficient and root mean square error indicators; the specific expression is: Where Y i is the actual value, Y p is the predicted value, Y mean is the predicted value of the actual value, SSR is the regression sum of squares, SST is the total sum of squares, which reflects the total deviation of the N observed values ​​of the target variable from the sample mean, SSE is the residual sum of squares, which reflects the impact of other factors other than the predicted variable on the target variable value, RMSE and MRE reflect the prediction error. The smaller the value, the smaller the error. 2 Reflects the accuracy of the model, ranging from 0 to 1. The closer it is to 1, the higher the model accuracy; Through the optimal RMSE, MRE and R 2 Select the convolutional neural network model with the best prediction effect.

6. A method for predicting atmospheric carbon dioxide concentration according to claim 1, characterized in that: Obtaining the temporal and spatial variation law of CO2 concentration through the nonlinear relationship includes the following steps: Generate a CO2 concentration distribution map with a resolution higher than the threshold through a nonlinear relationship; The generated CO2 concentration distribution map is subjected to spatiotemporal analysis to obtain the spatiotemporal variation patterns and trends of CO2 concentration.

7. An atmospheric carbon dioxide concentration prediction system, characterized in that: include: A data collection module is used to obtain multi-source data of a target area within a certain period of time; the multi-source data includes CO2 remote sensing data, road network data, surface meteorological data, normalized difference vegetation index NDVI data and aerosol optical depth AOD data; A model processing module, used for inputting the multi-source data into a convolutional neural network model to obtain a nonlinear relationship between the feature and the CO2 concentration when the temporal and spatial resolution is higher than a threshold; The prediction module is used to obtain the temporal and spatial variation law of CO2 concentration through the nonlinear relationship, and predict the atmospheric carbon dioxide concentration according to the temporal and spatial variation law of CO2 concentration.

8. A computer device, characterized in that: It comprises a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a method for predicting atmospheric carbon dioxide concentration as claimed in any one of claims 1 to 6.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for predicting atmospheric carbon dioxide concentration according to any one of claims 1 to 6 are implemented.

Images