A Remote Sensing Rapid Estimation Method for Methane Carbon Flux in the Non-Flooded Drawdown Zone of the Reservoir Area

Abstract

The present invention discloses a remote sensing rapid estimation method for methane carbon flux in the non-flooded water-level-fluctuation zone of a reservoir area, which includes: screening hyperspectral remote sensing images; preprocessing the screened hyperspectral remote sensing images to obtain the spectral reflectance of the hyperspectral images; vectorizing and extracting the water body range of the reservoir area, extracting the spectra of ground monitoring points and the measured methane flux data; extracting spectral features from the spectral information of all bands of the extracted ground monitoring points, and screening the spectral band ranges and corresponding spectral reflectances that are sensitive to changes in methane flux; combining the screened spectra and geographically weighted regression to construct a methane flux estimation model, and inversely obtaining the methane flux in the non-flooded water-level-fluctuation zone of the Three Gorges Reservoir Area. The present invention has few parameters and high efficiency, and can monitor the methane flux in the non-flooded water-level-fluctuation zone of the Three Gorges Reservoir Area, which is greatly affected by hydrology and has strong spatial heterogeneity, with high efficiency and near real-time, providing technical support for scientifically evaluating the methane flux in the water-level-fluctuation zone of the reservoir area and evaluating the benefits of methane emission reduction and sink enhancement.

Description

Technical Field

[0001] The present invention relates to the field of ecological environment monitoring, and specifically to a remote sensing rapid estimation method for methane carbon flux in the non-flooded drawdown zone of a reservoir area. Background Art

[0002] As a super-large annual regulating reservoir, the water level of the Three Gorges Reservoir changes periodically. The water level in front of the dam changes periodically between 145m and 170m, resulting in a water level drop staggered zone (drawdown zone) of about 30m on both banks. Due to environmental changes such as periodic flooding, the soil respiration in the drawdown zone is changed, and thus the methane flux is in dynamic evolution. Estimating the methane flux in the non-flooded drawdown zone is conducive to scientifically understanding the clean energy attribute of the Three Gorges hydropower, and is an important content for the high-quality development and ecological environment protection of the Three Gorges Reservoir area.

[0003] Traditional monitoring methods use the flux chamber measurement method, where instruments are carried manually by researchers to measure methane flux in the research area of the drawdown zone. This method is costly and inefficient, and it is difficult to obtain the methane flux in the non-flooded drawdown zone near real-time. The development of hyperspectral remote sensing technology provides the possibility for high-efficiency large-scale estimation of methane flux. However, the Three Gorges Reservoir area is large in scope and area, and the drawdown zone environment varies greatly. A global regression model that ignores spatial differences is difficult to achieve accurate estimation of methane flux. Therefore, it is very necessary to study a remote sensing estimation method for methane carbon flux in the non-flooded drawdown zone of the Three Gorges Reservoir area that takes into account spatial heterogeneity.

[0004] References for the background art are as follows:

[0005] [1] Hu Chunhong, Fang Chunming, Chen Xujian. Sediment Movement Laws and Simulation Technologies of the Three Gorges Project [M]. 2017, Science Press; Beijing

[0006] [2] Xia Zhenyao, Yan Rubing, Zhang Lun, etc. Response of the Root Tensile Performance of Bermudagrass to the Duration of Flooding [J]. Transactions of the Chinese Society of Agricultural Engineering, 2023, 39(6): 103 - 110.

[0007] [3] Zhao Dengzhong, Tan Debao, Li Chong, etc. Temporal and Spatial Variation of Carbon Dioxide Flux and Its Influencing Factors in Geheyan Reservoir [J]. Environmental Science, 2017, 38(3): 954 - 963.

[0008] [4] Zheng Shouren. Research and Monitoring and Inspection Analysis on the Safety and Long-Term Operation of the Reservoir Dam of the Three Gorges Project [J]. Yangtze River Technology and Economy, 2018, 2(3): 1 - 9.

[0009] [5] Zheng Shouren. The Three Gorges Project Provides Safety Guarantee and Environmental Protection for the Development of the Yangtze River Economic Belt [J]. 2019, 50(1): 1 - 6.

[0010] Barbosa PM, Bodmer P, Stadler M, et al. Ecosystem Metabolism Is the Dominant Source of Carbon Dioxide in Three Young Boreal Cascade‐Reservoirs (LaRomaine Complex, Québec)[J]. Journal of Geophysical Research:Biogeosciences, 2023, 128(4): 1 21.. Summary of the invention

[0011] In view of the shortcomings of the current methane flux estimation method in the unflooded drawdown zone of the Three Gorges Reservoir, the present invention provides a remote sensing rapid estimation method for the methane carbon flux in the unflooded drawdown zone of the reservoir area, which realizes methane flux estimation with both heterogeneity and high efficiency by using a low-manpower and low-parameter monitoring framework, and provides technical support for obtaining the methane flux in a large area of ​​the unflooded drawdown zone in the Three Gorges Reservoir area in near real time.

[0012] The technical solution adopted by the present invention is: a method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area, comprising the following steps:

[0013] Step 1: According to the complete vector range of the existing reservoir drawdown zone, set the imaging time and cloud coverage, and select the hyperspectral remote sensing image;

[0014] Step 2: Preprocessing the selected hyperspectral remote sensing images to obtain the spectral reflectance of the hyperspectral images, wherein the preprocessing includes orthorectification, radiation correction, and geometric correction;

[0015] Step 3: Vector extraction of the water body range in the reservoir area and spectral extraction of ground monitoring points and measured methane flux data: ArcGIS 10.5 was used to vectorize the extraction of the boundary of the water body, and the vector range of the unflooded drawdown zone was clipped based on the complete vector range data of the drawdown zone; the ROI of the area of ​​interest and the longitude and latitude of the center point of the methane flux ground monitoring point were manually selected, and the spectral information of all bands of the ground monitoring point was extracted based on the longitude and latitude; the methane concentration was measured and converted into methane flux;

[0016] Step 4: Use the competitive adaptive reweighted sampling CARS algorithm to extract spectral features from the spectral information of all bands of the ground monitoring points extracted in step 3, and screen the spectral band range and corresponding spectral reflectance that are sensitive to methane flux changes;

[0017] Step 5: Based on the screened spectral reflectance, the local model geographically weighted regression (GWR) taking into account spatial heterogeneity is used to construct the methane flux inversion model in the drawdown zone;

[0018] Step 6: Model evaluation: Calculate the coefficient of determination R of the measured and estimated methane fluxes in the water-level-fluctuation zone of the reservoir area on the validation set 2 , the residual sum of squares RSS, the corrected Akaike information criterion AICc, and the effective number of parameters ENP to evaluate the reliability of the model, and thus obtain the best model for methane flux inversion;

[0019] Step 7: Estimation of methane flux in the above-water part of the water-level-fluctuation zone of the reservoir area: Extract the corresponding hyperspectral images and the longitude and latitude of all pixels according to the vector range of the non-flooded water-level-fluctuation zone obtained in Step 3. Input the longitude and latitude of all pixels and the spectral reflectance sensitive to methane flux changes obtained in Step 4 into the best model for methane flux inversion to obtain the estimated value of methane flux in the entire non-flooded water-level-fluctuation zone.

[0020] Furthermore, the above-mentioned Step 1 includes:

[0021] Step 1.1, preliminarily screen hyperspectral remote sensing images according to the existing vector range of the complete water-level-fluctuation zone of the reservoir area;

[0022] Step 1.2, add screening conditions, set the cloud cover not to exceed 25%, and set the imaging time to the time of methane flux measurement, and further screen the hyperspectral remote sensing images that meet the requirements.

[0023] Furthermore, the above-mentioned Step 2 includes:

[0024] Step 2.1, open the orthorectification tool in ENVI5.2 to orthorectify the screened hyperspectral images to obtain images without geometric distortion;

[0025] Step 2.2, open the radiometric calibration tool in ENVI5.2 to radiometrically calibrate the screened hyperspectral images to obtain the true spectral reflectance of the surface;

[0026] Step 2.3, open the geometric correction tool in ENVI5.2, call the header file of the hyperspectral remote sensing images, and set parameters to obtain hyperspectral images with accurate positions.

[0027] Furthermore, the above-mentioned Step 3 includes:

[0028] Step 3.1, import the preprocessed hyperspectral images into ArcGIS10.5 to vectorize the vector range of the current water body in the reservoir area;

[0029] Step 3.2, import the existing vector range of the complete water-level-fluctuation zone of the reservoir area, and call the symmetrical difference tool in the analysis tool of ArcGIS10.5 to obtain the vector range of the current non-flooded water-level-fluctuation zone in the reservoir area;

[0030] Step 3.3: ArcGIS calls the masking tool (Extracted by Mask) in "Spatial Analyst Tools" to crop the hyperspectral remote sensing image of the non-flooded area of the water-level-fluctuation zone in the reservoir area;

[0031] Step 3.4: Import the hyperspectral remote sensing image of the non-flooded area of the water-level-fluctuation zone in the reservoir area into ENVI 5.2. Manually select no less than 40 ground monitoring points of methane flux, and manually check one region of interest (ROI) at each measured point. It is required that the number of pixels contained in each ROI is no less than 5. Export the vector range and the longitude and latitude of the center point of the ROI, and obtain the methane concentration data of the water-level-fluctuation zone in the reservoir area on-site and convert it into methane flux;

[0032] Step 3.5: Import the vector ranges of all ROIs and the hyperspectral image of the non-flooded area of the water-level-fluctuation zone in the reservoir area into ArcGIS 10.5, extract the longitude and latitude information of the center point of each ROI, and use the "Extract values to Points" tool to obtain the spectral information of all bands at each ground monitoring point.

[0033] Further, Step 3.4 includes:

[0034] Step 3.4.1: Import the longitude and latitude file of the selected ground monitoring points into a portable GPS, and manually reach the measured points of methane flux with the assistance of the GPS;

[0035] Step 3.4.2: Use Picarro G2301 to measure the methane concentration of the non-flooded water-level-fluctuation zone at the measured points for 20 - 30 minutes;

[0036] Step 3.4.3: Convert the methane concentration into methane flux according to the conversion formula.

[0037] Further, Step 4 includes:

[0038] Step 4.1: Comprehensively use ENVI 5.2 and Matlab 2018 to perform continuum removal and derivative processing on the hyperspectrum of each measured point to eliminate background noise and obtain the spectral reflectance of each measured point after preprocessing;

[0039] Step 4.2: Import the preprocessed hyperspectral information of each measured point into Matlab 2018, and use the competitive adaptive reweighted sampling (CARS) algorithm to perform preliminary dimensionality reduction on the hyperspectral information of each measured point, and extract the spectral bands sensitive to the change of methane flux from all bands;

[0040] Step 4.3, Discrimination of the number of spectral bands selected by CARS. For the number of spectral bands after CARS dimensionality reduction, if the number is less than 3, all the bands after dimensionality reduction are retained; if the number is greater than 3, the first 3 bands after dimensionality reduction are retained.

[0041] Further, the said Step 4.1 includes:

[0042] Step 4.1.1, Import the hyperspectral data of each measured point into ENVI 5.2, call the tool for removing the envelope line, and obtain the spectra of each measured point after removing the envelope line.

[0043] Step 4.1.2, Import the spectra of each measured point after removing the envelope line into Matlab 2018, call the differential function, perform differential processing on the spectra to further eliminate noise and highlight spectral features, and obtain the preprocessed hyperspectral data of each measured point.

[0044] Further, the said Step 4.2 includes:

[0045] Step 4.2.1, Random division of the dataset. Randomly divide the methane fluxes in the water-level-fluctuation zone of the reservoir area measured in the field in Matlab 2018.

[0046] Step 4.2.2, Selection of exponentially decaying wavelengths. The number of spectral bands in each iteration decreases successively, and the proportion of the number of variables sampled in the i-th sampling is determined through an exponentially decaying function.

[0047] Step 4.2.3, Adaptive reweighted sampling. Resample according to the proportion of the number of variables sampled in the i-th sampling determined in Step 4.2.2, call the partial least squares regression model, and calculate the root mean square error RMSECV of the model on the validation set.

[0048] Step 4.2.4, Loop iteration. Repeat Step 4.2.2 and Step 4.2.3, stop when the set maximum number of iterations is reached, determine the spectral bands sensitive to methane flux changes and the priority order of these bands according to the minimum RMSECV, and extract the spectral reflectances corresponding to these bands for standby.

[0049] Further, the said Step 5 includes:

[0050] Step 5.1, Place the longitude, latitude of the measured points and the spectral reflectances of the spectral bands sensitive to methane flux changes in a specified folder, divide the measured methane fluxes into a modeling set and a validation set according to a ratio of 0.75:0.25, and import them into ArcGIS 10.5.

[0051] Step 5.2: Based on the modeling set and the spectral reflectance corresponding to the spectral bands sensitive to methane flux changes, using the geographically weighted regression (GWR) algorithm that takes into account spatial heterogeneity, and adding GWR successively according to the priority order of the optimal spectral band set, an inversion model for methane flux in the water-level-fluctuation zone that takes into account spatial heterogeneity is established respectively.

[0052] Further, step 7 includes:

[0053] Step 7.1: Crop the hyperspectral image corresponding to the vector range of the non-flooded water-level-fluctuation zone extracted in step 3.3 to obtain the hyperspectral image of the corresponding range, and extract the longitude and latitude of all pixels in the current non-flooded water-level-fluctuation zone in ArcGIS 10.5 as the input data for the location parameters of the inversion model GWR.

[0054] Step 7.2: According to the optimal spectral band set determined in step 4, extract the spectral reflectance corresponding to the optimal spectral bands of the entire non-flooded water-level-fluctuation zone as the input data for the inversion parameters of the inversion model GWR.

[0055] Step 7.3: Based on the longitude, latitude of the center point of the non-flooded water-level-fluctuation zone and the spectral information corresponding to the optimal spectral bands, call the optimal model for methane flux inversion determined in step 6 to obtain the methane flux of the entire non-flooded water-level-fluctuation zone, and analyze the spatial distribution of methane flux in the non-flooded water-level-fluctuation zone of the reservoir area according to the longitude and latitude position information.

[0056] The present invention proposes a method for rapid remote sensing estimation of methane carbon flux in the non-flooded water-level-fluctuation zone of a reservoir area. This method combines the advantages of large-scale monitoring of hyperspectral remote sensing satellite images and the spatial analysis ability of GWR that takes into account heterogeneity. After envelope removal and differential preprocessing, the improved CARS is used to screen the spectral response information of methane carbon flux in the non-flooded water-level-fluctuation zone of the Three Gorges Reservoir Area. Combining ArcGIS 10.5, ENVI 5.2, and Matlab 2018, an inversion model for methane flux in the non-flooded water-level-fluctuation zone is established and its accuracy is verified based on a small number of measured methane flux data. Based on the model inversion, the methane flux and spatial distribution of the entire non-flooded water-level-fluctuation zone of the Three Gorges Reservoir Area are obtained. This method uses fewer bands and takes into account heterogeneity, and realizes the rapid estimation of methane flux with less spectral information, providing technical support for obtaining the methane carbon flux in the non-flooded water-level-fluctuation zone of the Three Gorges Reservoir by using hyperspectral remote sensing satellites on a large scale and in a wide range. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 Schematic diagram for screening hyperspectral images of the water-level-fluctuation zone of the Three Gorges Reservoir Area in the embodiment of the present invention;

[0058] Figure 2 Schematic diagram for preprocessing hyperspectral images of the water-level-fluctuation zone of the Three Gorges Reservoir Area in the embodiment of the present invention;

[0059] Figure 3Schematic diagram of measured point spectrum extraction and methane flux measurement in the non-flooded area of the water-level-fluctuating zone in the Three Gorges Reservoir Area, which is an embodiment of the present invention;

[0060] Figure 4 Schematic diagram of spectral dimensionality reduction by the competitive adaptive reweighted sampling algorithm, which is an embodiment of the present invention;

[0061] Figure 5 Schematic diagram of the construction and evaluation of the methane flux inversion model in the water-level-fluctuating zone of the Three Gorges Reservoir Area, which is an embodiment of the present invention;

[0062] Figure 6 Schematic diagram of methane flux estimation and spatial pattern revelation in the non-flooded area of the water-level-fluctuating zone of the Three Gorges Reservoir Area, which is an embodiment of the present invention. Detailed implementation manners

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0064] An embodiment of the present invention provides a remote sensing rapid estimation method for methane carbon flux in the non-flooded water-level-fluctuating zone of a reservoir area, including the following steps:

[0065] Step 1: According to the complete vector range of the water-level-fluctuating zone in the Three Gorges Reservoir Area, set the imaging time and cloud cover, and screen hyperspectral remote sensing images as shown in Figure 1 The detailed implementation steps are described as follows:

[0066] Step 1.1, preliminarily screen hyperspectral remote sensing images according to the complete vector range of the water-level-fluctuating zone in the Three Gorges Reservoir Area;

[0067] Step 1.2, add screening conditions, set the cloud cover not to exceed 25%, and set the imaging time to the methane flux measurement time, and further screen the hyperspectral images that meet the requirements.

[0068] Step 2: Pretreat the screened hyperspectral remote sensing images to obtain the spectral reflectance of the hyperspectral images. The pretreatment includes orthorectification, radiometric correction, and geometric correction. As shown in Figure 2 The detailed implementation steps are described as follows:

[0069] Step 2.1, open the orthorectification tool of ENVI5.2, and perform orthorectification on the screened hyperspectral images to obtain images without geometric distortion;

[0070] Step 2.2: Open the radiometric calibration tool in ENVI 5.2 to perform radiometric calibration on the selected hyperspectral images, and obtain the true spectral reflectance of the surface;

[0071] Step 2.3: Open the geometric correction tool in ENVI 5.2, call the header file of the hyperspectral remote sensing images, and set parameters to obtain accurately positioned hyperspectral images.

[0072] Step 3: Vectorized extraction of the water body range in the Three Gorges Reservoir Area, spectral extraction of ground monitoring points, and measured methane flux data: Use ArcGIS 10.5 to vectorize and extract the boundary of the water body, and based on the existing complete vector range data of the water-level-fluctuation zone, crop to obtain the vector range of the non-flooded water-level-fluctuation zone; Manually select the region of interest (ROI) of the ground monitoring points of methane flux and the longitude and latitude of the center point, and extract the spectral information of all bands of the ground monitoring points according to the longitude and latitude; Measure the methane concentration in-situ and convert it into methane flux. As Figure 3 shown, the detailed implementation steps are described as follows:

[0073] Step 3.1: Import the preprocessed hyperspectral images into ArcGIS 10.5, and vectorize the vector range of the current water body in the Three Gorges Reservoir;

[0074] Step 3.2: Import the existing complete vector range of the water-level-fluctuation zone in the Three Gorges Reservoir Area, and call the symmetrical difference tool in the analysis tool of ArcGIS 10.5 to obtain the vector range of the non-flooded water-level-fluctuation zone in the Three Gorges Reservoir;

[0075] Step 3.3: ArcGIS calls the masking tool (Extracted by Mask) in "Spatial Analyst Tools" to crop and obtain the hyperspectral remote sensing images of the non-flooded area of the water-level-fluctuation zone in the Three Gorges Reservoir Area;

[0076] Step 3.4: Import the hyperspectral remote sensing images of the non-flooded area of the water-level-fluctuation zone in the Three Gorges Reservoir Area into ENVI 5.2, manually select no less than 40 ground monitoring points of methane flux, and manually check 1 region of interest (ROI) for each in-situ measurement point, requiring that the number of pixels included in each region of interest (ROI) is no less than 5, export the vector range and the longitude and latitude of the center point of the region of interest (ROI), and obtain the methane concentration data of the water-level-fluctuation zone in the Three Gorges Reservoir Area in-situ and convert it into methane flux. Step 3.4 specifically includes:

[0077] Step 3.4.1: Import the longitude and latitude file of the selected ground monitoring points into a portable GPS, and manually reach the in-situ measurement points of methane flux with the assistance of the GPS;

[0078] Step 3.4.2: Use Picarro G2301 to measure the methane concentration of the non-flooded water-level-fluctuation zone at the in-situ measurement points, and the measurement duration is 20 - 30 minutes;

[0079] Step 3.4.3, using the conversion formula, convert the methane concentration into methane flux according to the formula. The conversion formula is as follows:

[0080]

[0081] where, represents the methane flux at the gas interface of the non-flooded drawdown zone grass; represents the rate of change of methane with time ( ); V is the volume of methane in the measuring instrument ( ); S is the range of the drawdown zone covered by the measuring instrument ( ); is the volume fraction of methane at standard temperature and pressure the conversion factor from to mg units , F is the conversion factor from seconds to hours ( ).

[0082] Step 3.5, import the vector ranges of all ROIs and the hyperspectral images of the non-flooded areas of the drawdown zone in the Three Gorges Reservoir Area into ArcGIS 10.5, extract the longitude and latitude information of the center points of each ROI, and use the "Extract values to Points" tool to obtain the spectral information of all bands at the ground monitoring points.

[0083] Step 4: Use the competitive adaptive reweighted sampling CARS algorithm to extract spectral features from the spectral information of all bands at the ground monitoring points extracted in Step 3, and screen the spectral band ranges and corresponding spectral reflectances that are sensitive to changes in methane flux. As Figure 4 shown, the detailed implementation steps are described as follows:

[0084] Step 4.1, in order to highlight the reflection characteristics of different ROIs, comprehensively use ENVI 5.2 and Matlab 2018 to perform continuum removal and differential processing on the hyperspectra of each measured point, eliminate background noise, and obtain the preprocessed hyperspectral information of each measured point. Step 4.1 specifically includes:

[0085] Step 4.1.1, import the hyperspectral data of each measured point into ENVI 5.2, call the continuum removal tool, and obtain the spectra after continuum removal of each measured point;

[0086] Step 4.1.2, import the spectra after continuum removal of each measured point into Matlab 2018, call the differential function, perform differential processing on the spectra, further eliminate noise and highlight spectral features, and obtain the preprocessed hyperspectral data of each measured point.

[0087] Step 4.2: Import the preprocessed hyperspectral information of each measured point into Matlab 2018, and use the Competitive Adaptive Reweighted Sampling (CARS) algorithm to perform preliminary dimensionality reduction on the hyperspectral information of each measured point, and extract the spectral bands sensitive to methane flux changes from all bands. Step 4.2 specifically includes:

[0088] Step 4.2.1: Randomly divide the dataset. Randomly divide the methane fluxes in the water-level-fluctuation zone of the Three Gorges Reservoir area measured in the field in Matlab 2018.

[0089] Step 4.2.2: Exponential decay wavelength selection. The number of spectral bands in each iteration decreases successively. The parameters of the exponential decay function are:

[0090]

[0091] Through the constraint condition , finally determine the proportion of the number of variables in the i-th sampling:

[0092]

[0093] Step 4.2.3: Adaptive reweighted sampling. Resample according to the proportion of the number of variables in the i-th sampling determined in Step 4.2.2, call the partial least squares regression model, and calculate the root mean square error of cross-validation (RMSECV) of the model on the validation set.

[0094] Step 4.2.4: Loop iteration. Repeat Step 4.2.2 and Step 4.2.3, stop when the set maximum number of iterations is reached, determine the spectral bands sensitive to methane flux changes and the priority order of these bands according to the minimum RMSECV, and extract the spectral reflectance corresponding to these bands for standby.

[0095] Step 4.3: Discrimination of the number of spectral bands screened by CARS. Discriminate the number of spectral bands after CARS dimensionality reduction. If the number is less than 3, retain all the bands after dimensionality reduction; if the number is greater than 3, retain the first 3 bands after dimensionality reduction.

[0096] Finally, the number and range of bands selected for the methane flux in the water-level-fluctuation zone of the Three Gorges Reservoir area are shown in Table 1.

[0097] Table 1 Number and range of bands screened by Competitive Adaptive Reweighted Sampling

[0098]

[0099] Indicates the spectral reflectance of the band with nm;

[0100] Step 5: Based on the selected spectral reflectance, a methane flux inversion model for the water-level-fluctuation zone is constructed using the local model Geographically Weighted Regression (GWR) that takes into account spatial heterogeneity. As Figure 5 shown, the detailed implementation steps are described as follows:

[0101] Step 5.1: Place the longitude and latitude of the measured points and the spectral reflectance of the spectral bands sensitive to methane flux changes in a specified folder. Divide the measured methane flux into a modeling set and a validation set according to the ratio of 0.75:0.25, and import them into ArcGIS 10.5;

[0102] Step 5.2: Based on the modeling set and the spectral reflectance corresponding to the spectral bands sensitive to methane flux changes, use the Geographically Weighted Regression (GWR) algorithm that takes into account spatial heterogeneity. According to the priority order of the optimal spectral band set, add GWR successively to establish a methane flux inversion model that takes into account spatial heterogeneity.

[0103] Step 6: Calculate the coefficient of determination R 2 , residual sum of squares RSS, corrected Akaike information criterion AICc, and effective number of parameters ENP for the measured and estimated methane fluxes in the validation set of the reservoir area water-level-fluctuation zone to evaluate the reliability of the model, and thus obtain the optimal model for methane flux inversion. As Figure 5 shown, the detailed implementation steps are described as follows:

[0104] The optimal inversion models obtained for each quality index are:

[0105]

[0106] where, is the methane flux, is the j-th dimension-reduced spectral band at the i-th methane observation point (in this study, it is ), is the regression coefficient of the j-th selected band at the position, and 0.315 is the random error term.

[0107] Step 6.1: On the validation set, use multiple inversion models to calculate the coefficient of determination R 2 , residual sum of squares (RSS), corrected Akaike information criterion (AICc), and effective number of parameters (ENP) for methane flux estimation respectively;

[0108] Step 6.2: Compare the R 2 , RSS, AICc, and ENP parameters of each model, and select the model with the highest R2 and lower RSS, AICc, and ENP as the optimal model for methane flux estimation.

[0109] Step 7: Estimation of methane flux in the water part of the water-level-fluctuating zone in the Three Gorges Reservoir Area: Extract the corresponding hyperspectral images and the longitude and latitude of all pixels according to the vector range of the non-flooded water-level-fluctuating zone obtained in Step 3. Input the longitude and latitude of all pixels and the spectral reflectance sensitive to methane flux changes obtained in Step 4 into the best model for methane flux inversion to obtain the estimated value of water methane flux in the entire non-flooded water-level-fluctuating zone. As Figure 6 shown, the detailed implementation steps are described as follows:

[0110] Step 7.1, Crop the corresponding hyperspectral images according to the vector range of the non-flooded water-level-fluctuating zone corresponding to the hyperspectral images extracted in Step 3.3, and extract the longitude and latitude of all pixels in the current non-flooded water-level-fluctuating zone range in ArcGIS 10.5 as the input data for the location parameters of the GWR inversion model.

[0111] Step 7.2, According to the optimal spectral band set determined in Step 4, extract the spectral reflectance corresponding to the optimal spectral bands of the entire non-flooded water-level-fluctuating zone as the input data for the inversion parameters of the GWR inversion model.

[0112] Step 7.3, Based on the longitude and latitude of the center point of the non-flooded water-level-fluctuating zone range and the spectral information corresponding to the optimal spectral bands, call the inversion model determined in Step 6 to obtain the methane flux of the entire non-flooded water-level-fluctuating zone, and analyze the spatial distribution of methane flux in the non-flooded water-level-fluctuating zone of the reservoir area according to the longitude and latitude position information.

[0113] The parameters of the present invention are few and the efficiency is high. It can monitor the methane flux in the non-flooded water-level-fluctuating zone of the Three Gorges Reservoir Area with large hydrological influence and strong spatial heterogeneity in real time with high efficiency, providing technical support for the scientific evaluation of methane flux and the evaluation of the benefits of reducing sources and increasing sinks in the water-level-fluctuating zone of the reservoir area.

[0114] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area, comprising the following steps: Step 1: According to the complete vector range of the existing reservoir drawdown zone, set the imaging time and cloud coverage, and select the hyperspectral remote sensing image; Step 2: Preprocessing the selected hyperspectral remote sensing images to obtain the spectral reflectance of the hyperspectral images, wherein the preprocessing includes orthorectification, radiation correction, and geometric correction; Step 3: Vector extraction of the water body range in the reservoir area and spectral extraction of ground monitoring points and measured methane flux data: ArcGIS 10.5 was used to vectorize the extraction of the boundary of the water body, and the vector range of the unflooded drawdown zone was clipped based on the complete vector range data of the drawdown zone; the ROI of the area of ​​interest and the longitude and latitude of the center point of the methane flux ground monitoring point were manually selected, and the spectral information of all bands of the ground monitoring point was extracted based on the longitude and latitude; the methane concentration was measured and converted into methane flux; Step 4: Use the competitive adaptive reweighted sampling CARS algorithm to extract spectral features from the spectral information of all bands of the ground monitoring points extracted in step 3, and screen the spectral band range and corresponding spectral reflectance that are sensitive to methane flux changes; Step 5: Based on the screened spectral reflectance, the local model geographically weighted regression (GWR) taking into account spatial heterogeneity is used to construct the methane flux inversion model in the drawdown zone; Step 6: Model evaluation: Calculate the determination coefficient R of the measured and estimated methane flux in the drawdown zone of the reservoir on the validation set 2 , residual sum of squares RSS, modified Akaike information criterion AICc, and effective parameter ENP are used to evaluate the reliability of the model, so as to obtain the best model for methane flux inversion; Step 7: Estimation of methane flux above water in the drawdown zone of the reservoir: Extract the corresponding hyperspectral image and the longitude and latitude of all pixels according to the vector range of the unflooded drawdown zone obtained in step 3, input the longitude and latitude of all pixels and the spectral reflectance sensitive to methane flux changes obtained in step 4 into the optimal model for methane flux inversion to obtain the estimated value of methane flux in the entire unflooded drawdown zone.

2. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 1 is characterized in that: The step 1 comprises: Step 1.1, preliminarily screen the hyperspectral remote sensing images according to the vector range of the complete drawdown zone of the existing reservoir area; Step 1.2, add screening conditions, set the cloud coverage not to exceed 25%, set the imaging time to the actual methane flux measurement time, and further screen the hyperspectral remote sensing images that meet the requirements.

3. According to the method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area in claim 1, the step 2 comprises: Step 2.1, open the ENVI5.2 orthorectification tool and perform orthorectification on the selected hyperspectral image to obtain an image without geometric distortion; Step 2.2, open the ENVI5.2 radiation correction tool, perform radiation calibration on the selected hyperspectral image, and obtain the true spectral reflectance of the surface; Step 2.3, open the ENVI5.2 geometric correction tool, call the header file of the hyperspectral remote sensing image, and set the parameters to obtain a hyperspectral image with accurate location.

4. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 1 is characterized in that: The step 3 comprises: Step 3.1, import the preprocessed hyperspectral image into ArcGIS10.5 to vectorize the vector range of the current water body in the reservoir area; Step 3.2, import the existing complete drawdown zone vector range of the reservoir area, call the symmetric difference tool in the ArcGIS10.5 analysis tool, and obtain the vector range of the drawdown zone of the reservoir area that is not currently flooded; Step 3.3, ArcGIS calls the mask tool in "Spatial Analyst Tools" to crop the hyperspectral remote sensing image of the unflooded area of ​​the reservoir drawdown zone; Step 3.4, import the hyperspectral remote sensing image of the unflooded area of ​​the reservoir drawdown zone into ENVI5.2, manually select no less than 40 ground monitoring points for methane flux, and manually select one region of interest (ROI) at each measured point, requiring that each region of interest (ROI) contain no less than 5 pixels, derive the vector range and longitude and latitude of the center point of the region of interest (ROI), obtain the methane concentration data of the reservoir drawdown zone on site and convert it into methane flux; In step 3.5, the vector ranges of all ROIs and the hyperspectral images of the unflooded range of the reservoir drawdown zone were imported into ArcGIS 10.5, and the longitude and latitude information of the center point of each ROI were extracted. The spectral information of all bands of each ground monitoring point was obtained using the "Multi-value Extraction to Point" tool.

5. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 4 is characterized in that: The step 3.4 comprises: Step 3.4.1, import the latitude and longitude files of the selected ground monitoring points into the portable GPS, and manually reach the actual measurement point of methane flux with the assistance of GPS; Step 3.4.2, use Picarro G2301 to measure the methane concentration in the unflooded drawdown zone at the measurement point, and the measurement time is 20-30 minutes; In step 3.4.3, the methane concentration is converted into methane flux according to the conversion formula.

6. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 1 is characterized in that: The step 4 comprises: Step 4.1, using ENVI5.2 and Matlab 2018 to perform envelope removal and differential processing on the hyperspectral spectrum of each measured point, eliminating background noise, and obtaining the spectral reflectance of each measured point after preprocessing; Step 4.2, import the preprocessed hyperspectral information of each measured point into Matlab 2018, use the competitive adaptive reweighted sampling CARS algorithm to perform preliminary dimensionality reduction on the hyperspectral information of each measured point, and extract the spectral bands sensitive to methane flux changes from all bands; Step 4.3, the number of spectral bands screened by CARS is determined. The number of spectral bands after dimensionality reduction by CARS is determined. If the number is less than 3, all bands after dimensionality reduction are retained; if the number is greater than 3, the first 3 bands after dimensionality reduction are retained.

7. A method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 6, characterized in that: The step 4.1 comprises: Step 4.1.1, import the hyperspectral data of each measured point into ENVI5.2, call the envelope removal tool, and obtain the spectrum of each measured point after the envelope is removed; In step 4.1.2, the spectrum after removing the envelope of each measured point is imported into Matlab 2018, and the differential function is called to perform differential processing on the spectrum to further eliminate the noise and highlight the spectral features, and obtain the high-spectral number of each measured point after preprocessing.

8. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 6 is characterized in that: The step 4.2 comprises: Step 4.2.1, random division of the data set, randomly divide the methane flux in the reservoir drawdown zone measured in the field in Matlab 2018; Step 4.2.2, exponential decay wavelength selection, the number of spectral bands in each iteration decreases gradually, and the ratio of the number of variables sampled in the i-th time is determined by the exponential decreasing function; Step 4.2.3, adaptive reweighted sampling, resampling is performed according to the ratio of the number of variables sampled in the i-th time determined in step 4.2.2, the partial least squares regression model is called, and the root mean square error RMSECV of the model on the validation set is calculated; Step 4.2.4, loop iteration, repeat steps 4.2.2 and 4.2.3, stop when the maximum number of iterations is reached, determine the spectral bands sensitive to methane flux changes and the priority order of these bands based on the minimum RMSECV, and extract the spectral reflectances corresponding to these bands for later use.

9. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 1 is characterized in that: The step 5 comprises: Step 5.1, place the latitude and longitude of the measured points and the spectral reflectance of the spectral bands sensitive to methane flux changes into a specified folder, divide the measured methane flux into a modeling set and a validation set at a ratio of 0.75:0.25, and import them into ArcGIS10.5; Step 5.2: Based on the modeling set and the spectral reflectance corresponding to the spectral bands sensitive to methane flux changes, the geographically weighted regression (GWR) algorithm that takes into account spatial heterogeneity is used to add GWRs one by one according to the priority order of the optimal spectral band set, and the methane flux inversion models of the drawdown zone that take into account spatial heterogeneity are established.

10. The method for rapid remote sensing estimation of methane carbon flux in the unflooded drawdown zone of a reservoir area according to claim 1, characterized in that: The step 7 comprises: Step 7.1, according to the vector range of the hyperspectral image corresponding to the unflooded drawdown zone extracted in step 3.3, the hyperspectral image of the corresponding range is obtained by clipping, and the longitude and latitude of all pixels in the current unflooded drawdown zone are extracted in ArcGIS 10.5 as the input data of the GWR position parameters of the inversion model; Step 7.2, according to the optimal spectral band set determined in step 4, extract the spectral reflectance corresponding to the optimal spectral band of the entire unflooded drawdown zone as the input data of the GWR inversion parameters of the inversion model; Step 7.3, based on the longitude and latitude of the center point of the unflooded drawdown zone and the spectral information corresponding to the optimal spectral band, call the optimal model for methane flux inversion determined in step 6 to obtain the methane flux of the entire unflooded drawdown zone, and analyze the spatial distribution of the methane flux in the unflooded drawdown zone of the reservoir area based on the longitude and latitude location information.

Images