Method for analyzing climate change rule based on remote sensing data

Abstract

The invention discloses a method for analyzing a climate change rule based on remote sensing data. Multi-source remote sensing data are obtained through multiple platforms of satellites, aviation and ground, and after preprocessing of geometric correction, radiation correction and the like, features of surface temperature, humidity and the like are extracted through a split window algorithm, a polarization difference index method and the like. Principal component analysis is used for dimensionality reduction, partial least squares regression is used for determining the quantitative relation between features and climate changes, and a correlation model is established by considering human activity influences. The method comprehensively obtains data, improves the processing and analysis precision, establishes the relation between human activities and climate changes, can accurately determine the climate change rule, and provides powerful support for climate change research and response.

Description

Technical Field

[0001] The present invention relates to the technical field of analyzing climate change laws based on remote sensing data, and specifically to a method for analyzing climate change laws based on remote sensing data. Background Art

[0002] At present, when global climate change has an increasingly profound impact on the ecological environment and human life, accurately analyzing climate change laws has become one of the core tasks in the field of environmental science. Climate change research largely relies on data collected by ground meteorological observation stations. However, these stations have significant limitations in geographical distribution and are difficult to achieve comprehensive coverage of the Earth's surface, especially in vast ocean areas, remote and inaccessible areas, and mountainous areas with complex terrains. There are many deficiencies in remote sensing data analysis methods. In the prior art, patent document CN117992757A proposed a method for analyzing remote sensing data of the national territorial ecological environment based on multi-dimensional data. This method improves the discrimination ability of apparent features through the collaborative application of multi-source data, making the inversion of ecological environment parameters more accurate. The specific steps include: obtaining remote sensing data of the national territorial ecological environment from multiple data sources, including various data sources such as satellite remote sensing, aerial remote sensing, and ground monitoring. Upsampling the low-resolution remote sensing images to increase their resolution to a preset target resolution, thereby obtaining high-resolution remote sensing data. Performing radiometric correction and geometric registration on the high-resolution remote sensing data to eliminate radiometric errors and geometric distortions in the data, and obtaining radiometric geometric correction data. The method has made certain progress in multi-source data fusion and hole filling, but there are still the following deficiencies: it proposed methods for radiometric correction and geometric registration, but these steps still require a large amount of manual intervention, the processing process is complex and error-prone. It uses a generative adversarial network (GAN) for hole filling, but in some complex scenarios, the filling accuracy still needs to be improved. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for analyzing climate change laws based on remote sensing data to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: A method for analyzing climate change laws based on remote sensing data, including the following steps: remote sensing data acquisition, data preprocessing, feature extraction, data analysis, and law determination;

[0005] Step 1, remote sensing data acquisition: Collect remote sensing data of different bands with the help of satellite remote sensing platforms, aerial remote sensing platforms, and ground remote sensing platforms; The satellite remote sensing platform selects the Earth Observing System EOS, the aerial remote sensing platform uses an unmanned aerial vehicle equipped with sensors, and the ground remote sensing platform is a fixed observation station; The data obtained covers multi-spectral, hyperspectral, synthetic aperture radar (SAR), lidar, and meteorological satellite data;

[0006] Step 2, data preprocessing: sequentially perform geometric correction, radiometric correction, and data format unification on the acquired remote sensing data; geometric correction corrects the geometric distortion of the image caused by factors such as sensor attitude and earth curvature through a mathematical model; radiometric correction eliminates the influence of the sensor itself and atmospheric factors on radiation transmission; data format unification converts data from different sources into a unified format;

[0007] Step 3, feature extraction: based on the preprocessed remote sensing data, extract surface temperature, vegetation coverage, surface humidity, and topographic and geomorphic features; calculate vegetation coverage through the Normalized Difference Vegetation Index (NDVI), and obtain topographic and geomorphic features using lidar data;

[0008] Step 4, data analysis: perform spatio-temporal analysis on the extracted features to construct an index system related to climate change; analyze the changing trends of features at multi-year and seasonal scales in the time dimension, and study the distribution differences and correlations of features in the global or specific regions in the space dimension, and use statistical methods such as correlation and regression analysis to determine the quantitative relationship between features and climate change;

[0009] Step 5, trend determination: based on the constructed index system, determine the climate change trend, periodic characteristics, and regional differences; judge the trend of global warming or cooling, determine the cold-warm and dry-wet alternation cycles, clarify the climate change differences in different regions, and comprehensively analyze whether there is a globally consistent change trend, local special changes, and multi-period superposition effects.

[0010] Furthermore, in the above Step 1, specific acquisition frequencies are set for different types of remote sensing data; for meteorological satellite data, considering the rapidity of meteorological changes, it is set to be acquired once every 30 minutes to ensure that the dynamic changes of atmospheric parameters can be captured in a timely manner; for hyperspectral remote sensing data, due to its large data volume and mainly used for analyzing the fine features of surface substances, for key research areas, it is set to be acquired once a week to balance the timeliness of data and processing costs.

[0011] Furthermore, in the above Step 2, polynomial correction models are used for geometric correction. By selecting at least 20 ground control points and using the least squares method to fit the polynomial coefficients, precise geometric correction of the image is achieved, effectively reducing the geometric error of the image to within 0.5 pixels; for radiometric correction, a method based on the radiation transfer equation is used, combined with atmospheric sounding data, to accurately calculate the scattering and absorption parameters of the atmosphere, so as to more accurately eliminate the influence of the atmosphere on radiation transmission.

[0012] Further, in the third step, the split-window algorithm is used for extracting surface temperature features. By utilizing data from two different channels in the thermal infrared band and constructing a specific mathematical model, the influence of the atmosphere on the retrieval of surface temperature is eliminated, and the accuracy of surface temperature retrieval is improved to ±0.5°C. For the extraction of surface humidity features, based on microwave remote sensing data, the polarization difference index method is adopted. By analyzing the differences in microwave signals under different polarization modes, the surface humidity is accurately calculated.

[0013] Further, in the fourth step, the principal component analysis (PCA) method is used to reduce the dimension of the extracted multi-dimensional features, compressing the high-dimensional feature space to 3 - 5 dimensions, removing redundant information in the data while retaining the main feature information, and improving the efficiency of subsequent analysis. When determining the quantitative relationship between features and climate change, the partial least squares regression (PLSR) method is adopted to effectively solve the problem of multicollinearity among independent variables and improve the prediction accuracy of the model.

[0014] Further, in the fifth step, combined with the output data of the global climate model (GCM), the climate change laws obtained from remote sensing data analysis are verified and supplemented. By comparing the GCM simulation results with the actual remote sensing analysis results, the accuracy of the model is evaluated. At the same time, using the long-term prediction ability of GCM, the climate change trends in the next 50 - 100 years are estimated, providing more comprehensive information for climate change research.

[0015] Further, in the entire method process, a data quality control system is established. In the data acquisition stage, the signal-to-noise ratio and radiation calibration accuracy indicators of remote sensing data are monitored in real time. When the indicators are lower than the set thresholds, the data re-acquisition program is automatically started. In the data processing and analysis stage, the quality of each processing result is evaluated, and the accuracy of the model is evaluated through cross-validation methods to ensure that the finally determined climate change laws are highly reliable.

[0016] Further, the method also includes data fusion, which is carried out after data preprocessing and before feature extraction. The weighted average fusion algorithm is adopted. According to the spatial resolution, temporal resolution, and data accuracy of different types of remote sensing data, different weights are assigned to each type of data, and multi-source remote sensing data are fused into a comprehensive data image, improving the integrity and accuracy of the data and providing a better data basis for subsequent feature extraction and analysis.

[0017] Furthermore, in the third step, an adaptive feature extraction strategy is adopted for different land cover types; for water bodies, in addition to extracting conventional surface temperature and surface humidity features, chlorophyll content and suspended sediment concentration features of water bodies are additionally extracted, and an inversion algorithm based on spectral features is used for extraction; for urban areas, the urban heat island intensity and impervious surface ratio features are mainly extracted, and high-resolution remote sensing images and deep learning algorithms are used for extraction.

[0018] Furthermore, in the fifth step, the influence of human activities on the laws of climate change is considered; by collecting data on population density, land use change, and energy consumption, an association model between human activities and climate change is established, and the mechanism of the role of human activities in climate change trends, periodic characteristics, and regional differences is analyzed, so as to determine the laws of climate change more comprehensively and accurately.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0020] Through multiple remote sensing platforms, the present invention can obtain multi-source remote sensing data with different spatio-temporal resolutions, making up for the limitations of single-platform data, covering all regions of the earth's surface, including the ocean, remote areas, etc., and greatly improving the spatial integrity of the data. By setting specific acquisition frequencies for different types of remote sensing data, it is ensured that dynamic changes such as meteorology can be captured in a timely manner, enhancing the timeliness of the data and providing a richer and more timely data basis for climate change research; using a polynomial correction model combined with a large number of ground control points and the least squares method for geometric correction, and radiation correction based on the radiative transfer equation combined with atmospheric sounding data can more accurately eliminate image deformation and radiation distortion, improving the data accuracy. In the feature extraction stage, advanced algorithms such as the split window algorithm and the polarization difference index method are used to extract features such as surface temperature and surface humidity respectively, significantly improving the feature extraction accuracy. When analyzing data, principal component analysis is used for dimensionality reduction to remove redundancy, and partial least squares regression is used to solve the problem of multicollinearity, making the established climate change index system more accurate and the quantitative relationship more reliable, effectively improving the accuracy of climate change law analysis; an association model between human activities and climate change is established, comprehensively considering factors such as population density, land use change, and energy consumption, filling the gap in this aspect of previous research, making the analysis of climate change laws more comprehensive and in-depth, and enabling a better understanding of the complex mechanism of climate change. At the same time, combined with the output data of global climate models for verification and supplementation, and using its long-term prediction ability to estimate future climate change trends, more forward-looking and comprehensive information is provided for climate change research. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. 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.

[0023] Refer to Figure 1 As shown, the present invention provides a method for analyzing climate change laws based on remote sensing data. Satellites with multi-band detection capabilities and a wide coverage range are selected, and they have a long-term observation history. Their data can be used for long-term climate change analysis; the Sentinel series provides high-resolution multi-spectral and radar data, which is suitable for fine observation of surface features.

[0024] According to the research area and research purpose, set the acquisition period of satellite data. For global-scale climate change research, data can be acquired once a month; for key areas of concern, such as ecologically fragile areas or areas with intensive human activities, data can be acquired once every two weeks to capture more subtle changes.

[0025] Establish a dedicated data receiving station, equipped with high-performance antennas and data receiving equipment to ensure stable and efficient reception of satellite-transmitted data. Classify and store the received data according to information such as date, satellite name, and band. The storage medium uses a large-capacity disk array, and data backup is performed regularly to prevent data loss.

[0026] Select a suitable UAV model for different research tasks. For large-area vegetation coverage monitoring, the DJI Phantom series of UAVs equipped with multi-spectral cameras can be selected, which are easy to operate and have a moderate endurance; for high-precision topographic mapping, professional mapping UAVs equipped with lidar, such as the Pegasus D2000, can be used.

[0027] Before flying, use geographic information system (GIS) software to analyze the research area and formulate a detailed flight route. Consider parameters such as flight altitude, speed, and overlap rate to ensure that the acquired data can comprehensively cover the research area and have sufficient accuracy. Generally, the flight altitude is determined according to the required resolution. For example, to obtain an image with a resolution of 10 cm, the flight altitude can be set at 100 - 150 meters; the overlap rate is set at 70% - 80% to ensure the accuracy of image stitching.

[0028] During the flight of the UAV, data is collected according to the preset flight route. The collected data is transmitted to the ground control station in real time through a wireless transmission module, and the ground control personnel monitor the data collection situation in real time to ensure the integrity and quality of the data. After the flight, the data in the UAV memory card is backed up and sorted.

[0029] Within the study area, representative locations are selected to construct ground remote sensing observation stations. The observation stations are equipped with various types of sensors, such as ground hyperspectral radiometers, microwave radiometers, weather stations, etc. The construction of the observation stations should consider the influence of the surrounding environment to avoid the occlusion and interference of tall buildings, trees, etc. on the observation data.

[0030] Various sensors collect data at set time intervals. For example, the ground hyperspectral radiometer collects data every 10 minutes, and the weather station records meteorological parameters such as temperature, humidity, and air pressure every 5 minutes. The collected data is transmitted to the data management system in real time for storage and preliminary processing.

[0031] The sensors of the ground remote sensing observation stations are calibrated regularly to ensure the accuracy of the data. The calibration work is carried out according to the operation manuals of the sensors, and generally a comprehensive calibration is carried out once every quarter. At the same time, the equipment of the observation stations is maintained and inspected regularly, and the aging or damaged components are replaced in time to ensure the normal operation of the observation stations.

[0032] II. Data Preprocessing

[0033] Using high-precision Global Positioning System (GPS) devices, at least 20 ground control points are evenly selected within the study area. The control points should have obvious ground features, such as road intersections, building corners, etc., to ensure accurate identification on remote sensing images.

[0034] According to the terrain complexity of the study area, a suitable polynomial model is selected for geometric correction. For relatively flat areas, a quadratic polynomial model can be used; for mountainous areas with large terrain undulations, a cubic polynomial model is adopted to improve the correction accuracy.

[0035] The coordinate information of the selected ground control points is input into geometric correction software, such as ENVI, Erdas, etc., and the coefficients of the polynomial model are calculated using the least squares method. According to the calculated coefficients, the remote sensing image is geometrically corrected to correct the ground object coordinates in the image to the accurate geographic coordinate system.

[0036] Atmospheric parameters of the study area, including atmospheric temperature, humidity, air pressure, aerosol concentration, etc., are obtained through means such as ground weather stations and atmospheric sounding satellites. These parameters are used to calculate the scattering and absorption degree of the atmosphere to electromagnetic radiation.

[0037] Using the radiative transfer theory, a radiative transfer equation for remote sensing data is established. According to the obtained atmospheric parameters, the radiative transfer equation is solved to obtain the true radiance value of the ground object. In the solving process, professional radiative transfer simulation software such as MODTRAN can be used to improve the calculation accuracy.

[0038] Verify the radiometrically corrected remote sensing data by comparing it with the radiometric data measured on the ground or by using standard targets with known reflectance. If there is a large deviation between the correction result and the actual situation, recheck the accuracy of the atmospheric parameters and the solution process of the radiative transfer equation, and make adjustments and optimizations.

[0039] Use data format conversion tools such as GDAL to identify the data formats obtained from different remote sensing platforms, such as TIFF, HDF, JPEG, etc. Convert the data in these formats into a format suitable for subsequent processing, such as ENVI standard format or GeoTIFF format.

[0040] During the data format conversion process, preserve and process the metadata information of the data. Metadata contains important information such as the acquisition time of the data, sensor type, geographic coordinates, etc., which is crucial for data analysis and application. Organize and store the metadata according to a unified standard so that it can be accurately read and used in subsequent processing.

[0041] Evaluate the resolution of different types of remote sensing data, including spatial resolution, temporal resolution, and spectral resolution. Satellite optical remote sensing data has a relatively high spatial resolution, up to the meter level or even sub-meter level, but a relatively low temporal resolution; meteorological satellite data has a relatively high temporal resolution, up to the minute level, but a relatively low spatial resolution.

[0042] Analyze the accuracy of different types of remote sensing data by comparing it with the ground-measured data or other high-precision data. For land surface temperature inversion, use the measured temperature data from ground meteorological stations to evaluate the inversion accuracy of different remote sensing data.

[0043] According to the results of resolution evaluation and accuracy analysis, assign weights to each type of remote sensing data. For data with high spatial resolution and high accuracy, assign a higher weight; for data with high temporal resolution but relatively low accuracy, assign an appropriate weight. The weight assignment can be carried out using methods such as the Analytic Hierarchy Process (AHP), and determined through expert scoring and mathematical calculations.

[0044] Adopt a weighted average fusion algorithm for data fusion. This algorithm is simple and effective, and can make full use of the advantages of different remote sensing data. During the fusion process, perform weighted average calculations on the corresponding pixels of different data according to the determined weights.

[0045] Verify the fused remote sensing data by evaluating the fusion effect through methods such as visual interpretation and statistical analysis. Compare the accuracy of the data before and after fusion in aspects such as surface feature recognition and land cover classification, and check whether the fused data retains the important information of the original data while improving the integrity and accuracy of the data.

[0046] Select remote sensing data in the thermal infrared band for land surface temperature inversion, such as the thermal infrared band of Landsat series satellites. Ensure that the data has undergone radiometric correction and geometric correction, with high accuracy.

[0047] Use the split-window algorithm for land surface temperature inversion. This algorithm utilizes data from two different channels in the thermal infrared band and constructs a specific mathematical model to eliminate the influence of the atmosphere on land surface temperature inversion. When applying the algorithm, adjust and optimize the parameters in the algorithm according to information such as the atmospheric parameters and land surface emissivity of the study area.

[0048] Compare and verify the retrieved land surface temperature results with the measured temperature data from ground meteorological stations. If the error between the two is within the allowable range of ±0.5°C, the inversion results are considered reliable; if the error is large, analyze the reasons, which may be inaccurate atmospheric parameters, incorrect estimation of land surface emissivity, etc., and make corresponding adjustments and improvements.

[0049] Calculate the vegetation coverage using the Normalized Difference Vegetation Index (NDVI). The calculation formula for NDVI is: NDVI = (NIR - R) / (NIR + R), where NIR is the reflectance in the near-infrared band and R is the reflectance in the red band. Calculate the NDVI value by computing the data of the near-infrared and red bands of the remote sensing image.

[0050] Select a suitable vegetation coverage conversion model according to the vegetation type and growth status of the study area. Commonly used models include the pixel dichotomy model, etc. This model assumes that a pixel consists of two parts: vegetation and non-vegetation, and calculates the vegetation coverage through the NDVI value.

[0051] Compare the calculated vegetation coverage results with the field survey data of vegetation coverage to evaluate the accuracy. The sampling survey method can be used to select multiple sample plots in the study area, measure the vegetation coverage in the field, and compare it with the remote sensing calculation results. If the accuracy does not meet the requirements, optimize the conversion model or consider other influencing factors, such as terrain, soil background, etc.

[0052] Extraction of land surface humidity characteristics

[0053] Select microwave remote sensing data for extracting land surface humidity characteristics, such as the Synthetic Aperture Radar (SAR) data of Sentinel-1 satellite. Preprocess the SAR data, including radiometric calibration, speckle noise removal, etc., to improve the data quality.

[0054] Use the polarization difference index method to extract land surface humidity. This method analyzes the differences in microwave signals under different polarization modes and establishes a relationship with land surface humidity. When applying the algorithm, make adaptive adjustments to the algorithm according to factors such as the soil type and vegetation coverage of the study area.

[0055] Verify and calibrate the extracted surface humidity results using the soil humidity data measured on the ground. If there is a deviation between the two, analyze the reasons, which may be inappropriate algorithm parameters, terrain effects, etc. Improve the accuracy of surface humidity extraction by adjusting algorithm parameters or performing terrain correction, etc.

[0056] Extraction of topographic and geomorphic features

[0057] Obtain the three-dimensional information of the terrain and geomorphology using lidar data. Lidar data has the characteristics of high precision and high resolution, and can accurately reflect the undulation changes of the earth's surface. If there is no lidar data, digital elevation model (DEM) data can also be used for terrain and geomorphology analysis, but the accuracy is relatively low.

[0058] Process the lidar data, including point cloud filtering, ground point extraction, DEM generation, etc. Use GIS software to analyze the generated DEM data and extract topographic and geomorphic features, such as slope, aspect, terrain undulation degree, etc.

[0059] Apply the extracted topographic and geomorphic features to climate change research, and analyze the impact of topography and geomorphology on local climate. Study the blocking effect of mountains on airflows, and the microclimate characteristics in valley areas, etc.

[0060] Arrange the extracted various feature data in chronological order to construct a time series. Use time series analysis methods, such as moving average method, exponential smoothing method, etc., to analyze the change trends of each feature in different time periods. For the surface temperature feature, calculate the average monthly surface temperature over the years, draw the curve of temperature change over time, observe the rising or falling trend of temperature, and whether there are seasonal fluctuations.

[0061] Apply geographic information system (GIS) technology to combine the feature data with geographic spatial information. Through methods such as spatial interpolation and spatial statistics, study the distribution differences and correlations of features at different geographical locations. For the vegetation coverage feature, use Kriging interpolation method to interpolate the discrete vegetation coverage data into a continuous spatial distribution image, analyze the high and low distribution of vegetation coverage in different regions, and the spatial correlation with factors such as terrain and precipitation.

[0062] Before performing principal component analysis (PCA), standardize the extracted multi-dimensional feature data. Adjust the mean of each feature data to 0 and the standard deviation to 1 to eliminate the differences in dimension and order of magnitude between different features and ensure the accuracy of the analysis results.

[0063] Use the PCA algorithm to process the standardized feature data. Calculate the covariance matrix of the data, solve the eigenvalues and eigenvectors of the covariance matrix, sort the eigenvectors according to the magnitudes of the eigenvalues, and select the top 3 - 5 principal components, which can retain most of the information of the original data.

[0064] Visualize the principal component data after dimensionality reduction, such as by plotting scatter plots, 3D plots, etc., to intuitively observe the distribution of the data in the low - dimensional space and facilitate the discovery of potential patterns and outliers in the data.

[0065] Take the feature data after spatio - temporal analysis and dimensionality reduction as independent variables, and take known climate change indicators (such as temperature change, precipitation change, etc.) as dependent variables to construct a dataset for establishing a quantitative relationship model.

[0066] Adopt the partial least squares regression method to establish a quantitative relationship model between features and climate change. This method can effectively handle the problem of multicollinearity among independent variables and improve the prediction accuracy of the model. During the model construction process, determine the optimal parameters of the model, such as the number of principal components, through methods such as cross - validation.

[0067] Use the constructed PLSR model to predict the dataset, and evaluate the performance of the model by calculating indicators such as the root mean square error RMSE and the coefficient of determination R 2 etc. At the same time, apply the model to an independent validation dataset to verify the generalization ability and accuracy of the model. If the model performance is not ideal, analyze the reasons, which may be data quality problems, improper model parameter selection, etc., and make corresponding adjustments and optimizations.

[0068] Establish a human activity association model

[0069] Obtain population density data of the study area from government statistical departments, census agencies, etc., including the population quantity and area information of different years and different administrative regions, and match and organize the population data according to the geographical spatial location.

[0070] Use historical remote sensing image data and land use classification methods to analyze the land use change situation in the study area over the years. Obtain the land use type and its area change data at different times and establish a land use change time series.

[0071] Energy consumption data collation: Collect energy consumption data in the study area, including the total consumption and per capita consumption of various types of energy. The data sources can be energy management departments, statistical yearbooks, etc., and associate the energy consumption data with time and geographical spatial information.

[0072] From the collected data on population, land use, energy consumption, etc., we selected variables closely related to climate change as independent variables, such as population density, rate of change of cultivated land area, energy consumption intensity, etc. We took climate change related indicators as dependent variables and constructed a model for the association between human activities and climate change.

[0073] According to the data characteristics and research purpose, select the appropriate model type, such as multiple linear regression model, neural network model, etc. The multiple linear regression model is simple and intuitive, and can clearly show the linear relationship between independent variables and dependent variables; the neural network model has a strong nonlinear fitting ability and can handle complex nonlinear relationships.

[0074] The selected model is trained using the collected data, and the model parameters are adjusted so that the model can accurately fit the data. During the training process, cross-validation and regularization methods are used to prevent the model from overfitting and improve the generalization ability of the model.

[0075] By analyzing the established human activity association model, the degree and direction of the impact of different human activity factors on climate change are determined. Through regression coefficient analysis, the impact and positive and negative relationship of factors such as increased population density, reduced cultivated land area, and increased energy consumption on climate change indicators such as temperature rise and precipitation changes are determined.

[0076] Human activity correlation model: In the analysis of the correlation between human activities and climate change, multi-dimensional socioeconomic data and remote sensing feature data are integrated, and a mixed effect model is used for cross-scale analysis. The specific implementation steps are as follows:

[0077] The population density data, which was gridded at 1 km, and the energy consumption intensity data at the county level were spatially superimposed with the remote sensing features NDVI and surface temperature of the same period, and unified to the same resolution using the Kriging interpolation method;

[0078] The random forest algorithm was used to evaluate the explanatory power of various human activity factors on climate indicators, and core variables with feature importance > 0.8 were selected;

[0079] Build a spatially explicit regression model:

[0080] ΔT=β 0 +β1*PopDensity+β2*EnergyCons+β3*NDVI+ε, where the spatial autocorrelation term ε is processed using geographically weighted regression (GWR).

[0081] ΔT is the target variable: the surface temperature change rate (unit: °C / year) is obtained through the inversion of remote sensing thermal infrared data, reflecting the quantitative index of regional climate change;

[0082] β 0The spatial heterogeneity intercept term, under the GWR framework, β 0 is a function of geographical coordinates (x, y): β 0 (x, y) = a 0 + a 1 Lon + a 2 Lat, representing the spatial distribution characteristics of the benchmark temperature change:

[0083] β 1 - β 3 The dynamic regression coefficient, calculated using Geographically Weighted Regression (GWR), with each pixel having an independent coefficient: β i (x, y) = f(spatial weight matrix), reflecting the local interaction intensity between human activities and climate;

[0084] PopDensity Population density (persons / km 2 ), a fused product based on NPP-VIIRS night-time light data and LandScan population statistics data, with a spatial resolution of 1 km, and the dimension is eliminated by logarithmic transformation;

[0085] EnergyCons Energy consumption intensity (tons of standard coal / 10,000 yuan of GDP), integrating the DMSP-OLS stable light index and provincial energy statistics data, and downscaled to a 1 km grid through a spatial econometric model.

[0086] NDVI Normalized Difference Vegetation Index (dimensionless), derived from the 30 m resolution time series data of Landsat8, and the inter-annual change rate is calculated after denoising using SG filtering, representing the impact of surface vegetation cover change on temperature

[0087] ε Spatial autocorrelation error term, following a spatial autoregressive process: ε = ρWε + ν, where ρ is the spatial autocorrelation coefficient, W is the Queen adjacency weight matrix, and ν ~ N(0, σ 2 ) white noise.

[0088] Use the established model for scenario simulation to predict the future climate change trends under different human activity scenarios. Set different population growth rates, land use planning, and energy policy scenarios to simulate the possible changes in climate change within the next few decades, providing a scientific basis for policy making.

[0089] Integrate the results obtained from data analysis and human activity correlation model analysis, comprehensively considering various characteristic data and influencing factors. Combine long time series climate change related indicator data, such as temperature, precipitation, sea level rise, etc., and analyze the change trends of these indicators in different time periods.

[0090] Using methods such as linear regression and polynomial fitting, fit the time series data of climate change indicators to obtain the trend line equation. According to the trend line equation, predict the trend of climate change in the future for a period of time, such as predicting the rising amplitude of temperature and the change of precipitation in the next 10 - 20 years. At the same time, considering uncertainty factors, give the confidence interval of the prediction result.

[0091] Adopt methods such as Fourier transform and wavelet analysis to conduct periodic analysis on the time series data of climate change related indicators. These methods can decompose the time series data into components of different frequencies and identify the existing periodic changes. Through Fourier transform, convert the temperature time series data into frequency domain data and find out the main periodic components, such as annual cycle, seasonal cycle, multi - year cycle, etc.

[0092] Describe the identified periodic characteristics in detail, including the length, amplitude, phase, etc. of the cycle. Analyze the mutual relationship between different periodic components and their comprehensive impact on climate change. Study the temperature change characteristics of annual cycle and multi - year cycle and their performance differences in different seasons.

[0093] According to factors such as geographical region, climate type, topography and landform, divide the research area into different sub - regions. Divide the globe into different continents and climate zones, or divide a country into different natural geographical regions.

[0094] Conduct a comparative analysis on the climate change characteristics of different sub - regions, including trend differences, cycle differences, differences in the impact of human activities, etc. Use statistical analysis methods, such as analysis of variance and significance test, to judge whether the differences in climate change characteristics between different regions are significant. Compare the temperature rising trend, precipitation change pattern of different continents, and the degree of influence of human activities on climate change in different regions.

[0095] In the stage of remote sensing data acquisition, set a series of data quality monitoring indicators, such as signal - to - noise ratio, radiometric calibration accuracy, geometric accuracy, etc. For satellite remote sensing data, the signal - to - noise ratio should be greater than a certain threshold (such as 30 dB) to ensure the reliability of the data; the radiometric calibration accuracy should be controlled within a certain range (such as ±5%) to ensure that the data can accurately reflect the radiation characteristics of ground objects.

[0096] Use data receiving equipment and monitoring software to monitor the quality indicators of data in real - time. When it is found that a certain indicator exceeds the set threshold, an alarm is automatically issued and the problem is fed back to the data acquisition operator. The operator checks the operation status of the equipment, adjusts the data acquisition parameters, or re - conducts data collection according to the feedback information.

[0097] Verify the intermediate results at each processing and analysis stage, such as data preprocessing, feature extraction, and data analysis. After geometric correction, verify the correction accuracy by checking the residuals of the control points; after feature extraction, compare the extracted features with known reference data to verify the accuracy of feature extraction.

[0098] For various established models, such as radiative transfer models, feature extraction models, data analysis models, etc., use methods such as cross-validation and leave-one-out method for evaluation. According to the evaluation results, optimize the model, adjust the model parameters, improve the algorithm structure, and improve the accuracy and stability of the model.

Claims

1. A method for analyzing climate change laws based on remote sensing data, characterized in that: The steps include: remote sensing data acquisition, data preprocessing, feature extraction, data analysis and law determination; Step 1: Remote sensing data acquisition: remote sensing data of different bands are collected with the help of satellite remote sensing platforms, aerial remote sensing platforms, and ground remote sensing platforms; the satellite remote sensing platform uses the Earth Observation System EOS, the aerial remote sensing platform uses a drone equipped with sensors, and the ground remote sensing platform is a fixed observation station; the data acquired covers multispectral, hyperspectral, synthetic aperture radar SAR, lidar, and meteorological satellite data; Step 2: Data preprocessing: geometric correction, radiation correction and data format unification are performed on the acquired remote sensing data in turn; geometric correction corrects the geometric deformation of the image caused by the sensor posture and the earth's curvature through mathematical models; radiation correction eliminates the influence of the sensor itself and atmospheric factors on radiation transmission; data format unification converts data from different sources into a unified format; Step 3, feature extraction: extracting surface temperature, vegetation coverage, surface humidity and topographic features based on the preprocessed remote sensing data; The vegetation coverage was calculated by the Normalized Difference Vegetation Index (NDVI), and the topographic features were obtained using LiDAR data; Step 4: Data analysis: Conduct spatiotemporal analysis on the extracted features and construct a climate change-related indicator system; analyze the trend of multi-year and seasonal feature changes in the temporal dimension, and study the differences and correlations in the distribution of global or specific regional features in the spatial dimension; and use correlation and regression analysis statistical methods to determine the quantitative relationship between features and climate change; Step 5: Determine the rules: Based on the constructed indicator system, determine the climate change trends, periodic characteristics and regional differences; judge the global climate warming or cooling trend, determine the cold and warm, dry and wet alternating cycles, clarify the differences in climate change in different regions, and comprehensively analyze whether there are globally consistent change trends, local special changes and multi-cycle superposition effects.

2. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In step one, specific collection frequencies are set for different types of remote sensing data. For meteorological satellite data, considering the rapidity of meteorological changes, collection is set every 30 minutes to ensure that the dynamic changes of atmospheric parameters can be captured in a timely manner. For hyperspectral remote sensing data, due to its large amount of data and its main use for analyzing the fine characteristics of surface materials, it is set to be collected once a week for key research areas to balance the timeliness and processing costs of the data.

3. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In the step 2, the geometric correction adopts a polynomial correction model. By selecting at least 20 ground control points and fitting the polynomial coefficients using the least squares method, accurate geometric correction of the image is achieved, effectively reducing the geometric error of the image to within 0.5 pixels; the radiation correction adopts a method based on the radiation transfer equation, combined with atmospheric detection data, to accurately calculate the scattering and absorption parameters of the atmosphere, so as to more accurately eliminate the influence of the atmosphere on radiation transmission.

4. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In step three, the surface temperature feature extraction adopts a split window algorithm, using two different channel data of the thermal infrared band. By constructing a specific mathematical model, the influence of the atmosphere on the surface temperature inversion is eliminated, and the surface temperature inversion accuracy is improved to ±0.5°C; the surface humidity feature extraction is based on microwave remote sensing data, using the polarization difference index method, and the surface humidity is accurately calculated by analyzing the differences in microwave signals under different polarization modes.

5. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In step four, the principal component analysis (PCA) method is used to reduce the dimensionality of the extracted multidimensional features, compress the high-dimensional feature space to 3-5 dimensions, remove redundant information in the data, and retain the main feature information to improve the efficiency of subsequent analysis; when determining the quantitative relationship between the features and climate change, the partial least squares regression (PLSR) method is used to effectively solve the multicollinearity problem between independent variables and improve the prediction accuracy of the model.

6. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In step five, the output data of the global climate model (GCM) is combined to verify and supplement the climate change laws obtained based on remote sensing data analysis; the accuracy of the model is evaluated by comparing the GCM simulation results with the actual remote sensing analysis results. At the same time, the long-term prediction ability of GCM is used to estimate the climate change trend in the next 50-100 years, providing more comprehensive information for climate change research.

7. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: A data quality control system was established throughout the entire method flow. During the data acquisition phase, the signal-to-noise ratio and radiation calibration accuracy indicators of remote sensing data were monitored in real time. When the indicators were lower than the set threshold, the data re-collection procedure was automatically started. During the data processing and analysis phase, the quality of each processing result was evaluated, and the accuracy of the model was evaluated through cross-validation methods to ensure that the final climate change laws were highly reliable.

8. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: The method also includes data fusion, which is performed after data preprocessing and before feature extraction; a weighted average fusion algorithm is used to assign different weights to each type of data according to the spatial resolution, temporal resolution and data accuracy of different types of remote sensing data, and multi-source remote sensing data are fused into a comprehensive data image, thereby improving the integrity and accuracy of the data and providing a better data foundation for subsequent feature extraction and analysis.

9. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In the step three, an adaptive feature extraction strategy is adopted for different types of land objects; for water areas, in addition to extracting conventional surface temperature and surface humidity features, the chlorophyll content and suspended matter concentration features of the water body are also extracted, and the extraction is performed using an inversion algorithm based on spectral features; for urban areas, the urban heat island intensity and impervious surface ratio features are extracted in particular, and the extraction is performed using high-resolution remote sensing images and deep learning algorithms.

10. The method for analyzing climate change laws based on remote sensing data according to claim 1, characterized in that: In step five, the impact of human activities on the laws of climate change is taken into account; by collecting data on population density, land use changes, and energy consumption, a correlation model between human activities and climate change is established, and the role of human activities in climate change trends, cyclical characteristics, and regional differences is analyzed, so as to determine the laws of climate change more comprehensively and accurately.

Images