Method and apparatus for real-time characterization and prediction of carbon emissions based on land use

Abstract

The application discloses a land use-based carbon emission instant characterization and prediction method and device, which comprises the following steps: preprocessing geographic data to obtain raster data and vector data, finely processing the raster data, establishing a land use GIS database and an expansion driving force factor data set; establishing an energy consumption data set and an economic and social development data set by using annual data; performing carbon accounting according to the energy consumption data set and the land use GIS database, and establishing a first model for representing and predicting the regression relationship between land use and carbon emission; determining various economic and social development indexes of a preset development scenario, and predicting land use demand; obtaining an expansion possibility atlas set of each land use type by using a random forest algorithm and a LEAS model; simulating future land use by using a PLUS model, and calculating the carbon emission of the simulation result by using the first model. Through the above method, the instant characterization and prediction of carbon emission and the fine processing of built-up area land can be realized.

Description

Technical Field

[0001] This application relates to the field of carbon emission analysis and prediction technology, and in particular to a method and apparatus for real-time characterization and prediction of carbon emissions based on land use. Background Technology

[0002] Carbon emissions are the process by which human production and business activities release greenhouse gases (carbon dioxide, methane, nitrous oxide, hydrofluorocarbons, perfluorocarbons, and sulfur hexafluoride, etc.) into the environment. Greenhouse gases are considered the main cause of the greenhouse effect. To reduce carbon emissions and promote technological innovation in green industries, timely, scientific, and accurate carbon emission characterization and prediction are essential.

[0003] In existing technologies, carbon emissions are typically characterized using socio-economic development indicators such as GDP and population, or regional nighttime light data. These socio-economic development indicators are mostly derived from statistical yearbooks or other publications, resulting in data delays and hindering real-time carbon emission representation. This makes it difficult for governments to adjust relevant policies promptly. Carbon emission prediction mainly employs models such as IPAT, STIRPAT, LMDI, and LEAP, or machine learning methods such as grey prediction and neural network models to predict and analyze carbon emissions scenarios. However, these methods primarily rely on socio-economic development indicators for carbon emission prediction, only guiding policymakers in setting development goals. They lack specificity in land development and utilization, making them difficult to apply.

[0004] Current research on land use and carbon emissions by scholars both domestically and internationally has not considered the impact of land use functions and building volume in built-up areas on carbon emission intensity, nor has it conducted refined land use analysis of built-up areas. Furthermore, existing models and methods often treat carbon emission factors in built-up areas as fixed values, failing to account for differences in land use carbon emission intensity caused by land use attributes and climate zone variations. Summary of the Invention

[0005] This application provides a method and apparatus for real-time characterization and prediction of carbon emissions based on land use. It implements a real-time carbon emission characterization module by acquiring satellite remote sensing imagery. The acquired first raster data undergoes refined processing, addressing the problem that existing technologies for land use and carbon emission research have not considered the impact of built-up area land use functions and building volume on land use carbon emission intensity. The model for characterizing and predicting carbon emissions based on land use data solves the problem of significant data delays in existing technologies that mostly use statistical yearbook data for carbon emission characterization. It also addresses the issue that existing technologies mainly rely on various economic and social development indicators for carbon emission prediction, which lacks specificity regarding land development and utilization.

[0006] In a first aspect, embodiments of this application provide a method for real-time characterization and prediction of carbon emissions based on land use, comprising: preprocessing geographic data to obtain first raster data and first vector data; wherein the geographic data includes a first LUCC dataset, remote sensing imagery, and vector data; refining the first raster data and establishing a land use GIS database and a land use expansion driving force factor dataset; wherein the first raster data includes the first LUCC dataset and a second LUCC dataset; establishing an energy consumption dataset and an economic and social development dataset using processed statistical yearbook data; calculating carbon source carbon emission factors and carbon sink carbon sequestration factors based on the energy consumption dataset and the land use GIS database, and establishing a land use data model for characterization and prediction. A first model relating carbon emissions to the following: the first model can characterize the corresponding carbon emissions based on a pre-processed and refined LUCC dataset; it determines various economic and social development indicators for multiple preset development scenarios and uses a trained first neural network to predict the land use demand for each of the economic and social development indicators; it obtains an expansion probability map of each land use type using a random forest algorithm and a LEAS model based on the geographic data and the land use expansion driving factor dataset; it simulates future land use using a PLUS model based on the land use GIS database, the land use demand for multiple development scenarios, and the expansion probability map of the land use types, and uses the first model to calculate the carbon emissions of the simulation results for the development scenarios.

[0007] In conjunction with the first aspect, in a first possible implementation, the vector data includes POI data, administrative boundary data, and road network data.

[0008] In conjunction with the first possible implementation of the first aspect, in the second possible implementation, the second LUCC dataset includes: performing radiometric calibration and atmospheric correction processing on the remote sensing image; performing supervised classification on the processed remote sensing image using a trained second neural network; and performing accuracy verification on the supervised classified remote sensing image to obtain the second LUCC dataset.

[0009] In conjunction with the second possible implementation of the first aspect, in the third possible implementation, the refinement of the first raster data includes: projecting the first raster data and the first vector data according to the geographical location of the study area to obtain planar geographic data; reclassifying the planar geographic data, extracting the raster corresponding to the built-up area, and converting the raster corresponding to the built-up area into area features; completing the road network data, and dividing the area features of the built-up area into street blocks according to the completed road network data; calculating the kernel density value of each type of POI data, and corresponding the kernel density value of each type of POI data to the street block to obtain the kernel density value of each type of POI data for each street block; dividing the street block into single-function land use type units and mixed-function land use type units according to the kernel density value; converting the single-function land use type units and the mixed-function land use type units into raster data types and embedding them back into the planar geographic data.

[0010] In conjunction with the first aspect, in a fourth possible implementation, the use of a trained first neural network to predict land use demand for each of the economic and social development indicators includes: constructing a Pareto optimal solution set for each of the economic and social development indicators using multi-objective programming; and using the trained first neural network to predict the land use demand for each of the economic and social development indicators based on the Pareto optimal solution set.

[0011] In conjunction with the fourth possible implementation of the first aspect, the fifth possible implementation further includes: reducing the autocorrelation between the input samples in the yearbook data; and training the first neural network with the processed input samples as input and the land use GIS database as output.

[0012] In conjunction with the fourth possible implementation of the first aspect, in the sixth possible implementation, the use of the trained first neural network to predict the land use demand for each of the economic and social development indicators further includes: setting a baseline scenario as a control group; wherein the baseline scenario uses a Markov chain to predict the land use demand.

[0013] In conjunction with the first aspect, in the seventh possible implementation, obtaining an expansion probability atlas for each land use type using the random forest algorithm and the LEAS model based on the geographic data and the land use expansion driving factor dataset includes: selecting multiple factors as driving factors from the land use expansion driving factor dataset; wherein, the driving factors include: land succession natural environment driving factors and land succession social development driving factors; performing GIS spatial interpolation and geographic registration on the driving factors, and establishing a driving factor dataset based on the processed driving factors; determining a land use succession dataset based on the geographic data, and calculating the impact of the driving factors on land use succession; and obtaining an expansion probability atlas for each land use type using the LEAS model based on the driving factor dataset and the land use succession dataset.

[0014] Secondly, embodiments of this application provide an apparatus for real-time characterization and prediction of carbon emissions based on land use, comprising: a preprocessing module for preprocessing geographic data to obtain first raster data and first vector data; wherein the geographic data includes a first LUCC dataset, remote sensing imagery, and vector data; a refinement processing module for refining the first raster data and establishing a land use GIS database and a land use expansion driving force factor dataset; wherein the first raster data includes the first LUCC dataset and a second LUCC dataset; a statistical yearbook storage module for establishing an energy consumption dataset and an economic and social development dataset using the processed statistical yearbook data; and a real-time characterization module for calculating carbon source carbon emission factors and carbon sink carbon sequestration factors based on the energy consumption dataset and the land use GIS database, and establishing a land use data model for characterization and prediction. The system utilizes a first model that regresses on carbon emissions; wherein the first model can characterize the corresponding carbon emissions based on a pre-processed and refined LUCC dataset; a land use demand module is used to determine various economic and social development indicators for multiple preset development scenarios, and use a trained first neural network to predict the land use demand for each of the economic and social development indicators; a land use type expansion module is used to obtain an expansion probability map of each land use type using a random forest algorithm and a LEAS model based on the geographic data and the land use expansion driving factor dataset; a simulation and prediction module is used to simulate future land use using a PLUS model based on the land use GIS database, the land use demand for multiple development scenarios, and the expansion probability map of the land use types, and use the first model to calculate the carbon emissions of the simulation results for the development scenarios.

[0015] In conjunction with the second aspect, in the first possible implementation, the vector data includes POI data, administrative boundary data, and road network data.

[0016] In conjunction with the first possible implementation of the second aspect, in the second possible implementation, the second LUCC dataset includes: performing radiometric calibration and atmospheric correction processing on the remote sensing image; performing supervised classification on the processed remote sensing image using a trained second neural network; and performing accuracy verification on the supervised classified remote sensing image to obtain the second LUCC dataset.

[0017] In conjunction with the second possible implementation of the second aspect, in the third possible implementation, the refinement of the first raster data includes: projecting the first raster data and the first vector data according to the geographical location of the study area to obtain planar geographic data; reclassifying the planar geographic data, extracting the raster corresponding to the built-up area, and converting the raster corresponding to the built-up area into area features; completing the road network data, and dividing the area features of the built-up area into street blocks according to the completed road network data; calculating the kernel density value of each type of POI data, and corresponding the kernel density value of each type of POI data to the street block to obtain the kernel density value of each type of POI data for each street block; dividing the street block into single-function land use type units and mixed-function land use type units according to the kernel density value; converting the single-function land use type units and the mixed-function land use type units into raster data types and embedding them back into the planar geographic data.

[0018] In conjunction with the second aspect, in a fourth possible implementation, the use of a trained first neural network to predict land use demand for each of the economic and social development indicators includes: constructing a Pareto optimal solution set for each of the economic and social development indicators using multi-objective programming; and using the trained first neural network to predict the land use demand for each of the economic and social development indicators based on the Pareto optimal solution set.

[0019] In conjunction with the fourth possible implementation of the second aspect, the fifth possible implementation further includes: reducing the autocorrelation between the input samples in the yearbook data; and training the first neural network with the processed input samples as input and the land use GIS database as output.

[0020] In conjunction with the fourth possible implementation of the second aspect, in the sixth possible implementation, the use of the trained first neural network to predict the land use demand for each of the economic and social development indicators further includes: setting a baseline scenario as a control group; wherein the baseline scenario uses a Markov chain to predict the land use demand.

[0021] In conjunction with the second aspect, in the seventh possible implementation, the step of obtaining an expansion probability atlas for each land use type using the random forest algorithm and the LEAS model based on the geographic data and the land use expansion driving factor dataset includes: selecting multiple factors as driving factors from the land use expansion driving factor dataset; wherein, the driving factors include: land succession natural environment driving factors and land succession social development driving factors; performing GIS spatial interpolation and geographic registration on the driving factors, and establishing a driving factor dataset based on the processed driving factors; determining a land use succession dataset based on the geographic data, and calculating the impact of the driving factors on land use succession; and obtaining an expansion probability atlas for each land use type using the LEAS model based on the driving factor dataset and the land use succession dataset.

[0022] Thirdly, embodiments of this application provide a non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium including storage for storing a computer program or instructions that, when executed, cause the method described in the first aspect or any possible implementation of the first aspect to be implemented.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] This application employs a method for real-time characterization and prediction of carbon emissions based on land use. By acquiring satellite remote sensing images to characterize carbon emissions, it effectively solves the problem of data delays in statistical yearbooks and bulletins, thus realizing a real-time carbon emission characterization method. The acquired geographic data is refined to address the issue in existing technologies that do not consider the impact of built-up area land use functions and building volume on carbon emission intensity. The LEAS and PLUS models are used to predict and simulate future land use demand, and a characterization and prediction model of the relationship between carbon emissions and land use is established, which can predict the carbon emissions of different types of land use. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A research roadmap for the method and apparatus for real-time characterization and prediction of carbon emissions based on land use, as provided in the embodiments of this application;

[0027] Figure 2 A flowchart illustrating the method for real-time characterization and prediction of carbon emissions based on land use, provided for embodiments of this application;

[0028] Figure 3 A flowchart for processing remote sensing images into LUCC data type is provided for embodiments of this application;

[0029] Figure 4 A flowchart for refining geographic data as provided in this application embodiment;

[0030] Figure 5 A flowchart for obtaining an expansion possibility atlas for each land use type, provided as an embodiment of this application;

[0031] Figure 6 A schematic diagram of a device for real-time characterization and prediction of carbon emissions based on land use, provided for an embodiment of this application;

[0032] Figure 7 This is a structural diagram of land refinement processing provided in an embodiment of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] Figure 2 This is a flowchart of a method for real-time characterization and prediction of carbon emissions based on land use, provided in an embodiment of this application, including steps 201 to 207. Figure 2 This is merely one execution order shown in the embodiments of this application and does not represent the only execution order of the method for real-time characterization and prediction of carbon emissions based on land use. Where the final result can be achieved, Figure 2 The steps shown can be performed in parallel or in reverse order.

[0035] Step 201: Preprocess the geographic data to obtain the first raster data and the first vector data. The geographic data includes the first LUCC (Land-Use and Land-Cover Change) dataset, remote sensing imagery, and vector data. Specifically, the first LUCC dataset is land use data from 2000 to 2019 released by the team of Professor Yang Jie and Professor Huang Xin of the State Key Laboratory of Surveying and Mapping Remote Sensing Information Engineering at Wuhan University. The remote sensing imagery is land use data remote sensing imagery from 2020 to 2022. The acquired remote sensing imagery is processed into the second LUCC dataset, obtaining the second LUCC dataset from 2020 to 2022. Specific steps are as follows... Figure 3 As shown, steps 301 to 303 are included.

[0036] Step 301: Perform radiometric calibration and atmospheric correction on the remote sensing image. Specifically, radiometric calibration converts the image's brightness grayscale values ​​into absolute radiance when calculating the spectral reflectance or spectral radiance of ground objects. In this embodiment, the ENVI software's Radiometric Calibration tool is used to perform radiometric calibration on the acquired remote sensing image. The total radiance of the ground target ultimately measured by the sensor does not reflect the true surface reflectance, as it includes radiation errors caused by atmospheric absorption and scattering. Atmospheric correction can eliminate this radiation error and reflect the true surface reflectance of ground objects; specifically, the ENVI software's FLAASHAtmospheric Correction tool is used to perform atmospheric correction on the remote sensing image.

[0037] Step 302: Supervised classification of the processed remote sensing images is performed using the trained second neural network. Specifically, the images are fused. In this embodiment, the vector data is cropped based on the administrative boundaries of the 2019 map of China to obtain the preprocessed satellite remote sensing images. A training element dataset is established. In this embodiment, the training element dataset is divided into six categories: cultivated land, forest land, grassland, water bodies, built-up areas, and unused land. The second neural network is trained using samples, and the trained second neural network is used to perform supervised classification of the remote sensing images processed in step 301.

[0038] Step 303: Perform accuracy verification on the supervised classification remote sensing images to obtain the second LUCC dataset. Specifically, after processing the remote sensing images through the above steps, only built-up areas can be distinguished from land use types such as green space, grassland, and cultivated land; further subdivision of specific land use functions within built-up areas is not possible. However, the carbon emission intensity of built-up land use varies significantly across different functions and building sizes, such as industrial, commercial office, and residential areas. Therefore, it is necessary to perform refined classification of land use types within built-up areas and to statistically analyze the area and proportion of each land use type.

[0039] The vector data includes POI (Point of Interest) data, administrative boundary data, and road network data. Specifically, in this embodiment, the POI point feature dataset is obtained by crawling the Gaode Map open-source platform using Python; the administrative boundary dataset is obtained using the China Standard Map; and the road network line feature dataset is obtained using the Open StreetMap (OSM) open-source map platform. The POI data is of point feature type, the administrative boundary data is of polygon feature type, and the road network data is of line feature type. In this embodiment, the vector data consists of the obtained original POI data, administrative boundary data, and road network data; the first vector data is the vector data after missing data has been filled in.

[0040] It should be noted that the LUCC dataset is a raster data type; therefore, in this embodiment, the first LUCC data and the second LUCC data are referred to as the first raster data. ENVI software, the Radiometric Calibration tool, the FLAASHAtmospheric Correction tool, and the 2019 map of China are merely examples of embodiments of this application and are not intended to limit the scope of protection of this application. Figure 3 The step described is to process satellite remote sensing imagery into a LUCC dataset. Those skilled in the art can also implement this step in other ways. Furthermore, this embodiment exemplarily selects remote sensing imagery from 2020 to 2022 to process into a second LUCC dataset. Those skilled in the art can then... Figure 3 The steps shown process remote sensing images from other years. The years 2020 to 2022 are not intended to limit the scope of this application. The first LUCC data from 2000 to 2019 can also be obtained by those skilled in the art through other means.

[0041] Step 202: Refine the first raster data and establish a land use GIS database and a dataset of land use expansion driving factors. The first raster data includes a first LUCC dataset and a second LUCC dataset. Specifically, the first raster data undergoes fine-tuning. For example... Figure 7The land refinement process structure diagram shown below illustrates the specific steps as follows: Figure 4 As shown, steps 401 to 406 are included.

[0042] Step 401: Project the first raster data and the first vector data according to the geographical location of the study area to obtain planar geographic data. Specifically, the acquired surface data is projected according to the geographical location of the study area to obtain the planar geographic data of the study area. LUCC data is a type of raster data that includes geographic coordinates and land use type information.

[0043] Step 402: Reclassify the planar geographic data, extract the rasters corresponding to the built-up areas, and convert the rasters corresponding to the built-up areas into polygon features. Specifically, obtain the raster dataset of the planar geographic data. In this embodiment, ArcGIS is used to reclassify the raster data, setting the land use type outside the built-up areas to NO DATA, extracting all built-up land, and converting the raster data corresponding to the built-up land into polygon features.

[0044] Step 403: Complete the road network data and divide the built-up area's surface features into street block units based on the completed road network data. Specifically, since there are dead ends in the acquired road network data for Class IV roads, it is necessary to complete the acquired road network data. In this embodiment, ArcGIS is used to modify the vector file for completion, and then vector data of the road surface is established and divided according to the road level. In this embodiment, Class III roads and above are uniformly set at 30 meters per direction, and Class IV roads are set at 20 meters per direction. Based on the surface features of the built-up area obtained in Step 402 and the completed road network data, in this embodiment, the ArcGIS feature erasure tool is used to erase the road surface features on the built-up area surface features to obtain the boundary surface feature data of each street block unit.

[0045] Step 404: Calculate the kernel density values ​​of various types of POI data, and map these values ​​to street blocks to obtain the kernel density values ​​of various types of POI data for each street block. Specifically, in this embodiment, Ripley's K function is used to calculate the optimal search radius of the kernel density function. Based on the distance from points within the search radius to the center point, different weights are assigned to each point, which can effectively express the distribution density of different POI data and their impact on urban land use functions. The kernel density calculation formula is as follows: In the formula, f(x) is the kernel density value of the x-th POI data, i is the search radius (i.e., the relative distance from the POI point feature to x), n is the number of feature points included within the search radius of point x, K is the kernel function, and h is the search radius (its value is calculated by Ripley's K function). The kernel density value of each POI data within the block area is calculated using the above kernel density calculation formula.

[0046] Step 405: Divide the block units into single-function land use type units and mixed-function land use type units based on the kernel density values. Specifically, in this embodiment, the kernel density value ratio of each POI data within a certain block area is calculated. When the kernel density value ratio of the POI data corresponding to a certain land use type of a certain block unit reaches 50%, the block unit is marked as a single-function land use type; when the kernel density value ratio of the POI data corresponding to a certain land use type of a certain block unit is less than 50%, the block unit is marked as a mixed-function land use type.

[0047] Step 406: Convert single-function land use type units and mixed-function land use type units into raster data types and embed them back into the planar geographic data. Specifically, in this embodiment, the classified street block unit polygon features are converted into raster data types, and ArcGIS is used to embed the raster data back into the planar geographic data to complete the land refinement process. Embedding back into the raster data in this embodiment refers to converting the built-up area polygon features extracted and converted in step 402 back into raster data types after processing in steps 403 to 406 and embedding them into the corresponding extracted planar geographic data.

[0048] This study uses Python to crawl data from the Gaode Map open-source platform, obtaining years of POI data. It then performs refined processing of built-up area land use types and statistically analyzes the area of ​​each refined land use type. For years where POI data is unavailable, spline curve interpolation, with its higher accuracy, can yield the results of refined processing of built-up area land use for each year and statistically analyze the area of ​​each refined land use type.

[0049] Based on the above processed data, a land use GIS database and a dataset of driving factors for land use expansion were established.

[0050] In this embodiment of the application, a single road network of level four or above is used as the basic unit for land fine classification, and the built-up area is finely processed according to the classification method in Table 1 below.

[0051] Table 1

[0052]

[0053]

[0054] Step 203: Establish energy consumption dataset and economic and social development dataset using the processed statistical yearbook data. Specifically, the data obtained from the national economic and social development bulletins of various provinces and prefecture-level cities, as well as data from the China Urban Statistical Yearbook, China Rural Statistical Yearbook, the National Bureau of Statistics' National Statistical Data Center, and prefecture-level city statistical yearbooks, are processed and interpolated to establish energy consumption dataset and economic and social development dataset.

[0055] Step 204: Calculate carbon source carbon emission factors and carbon sink carbon sequestration factors based on the energy consumption dataset and land use GIS database, and establish a first model for characterizing and predicting the regression relationship between land use and carbon emissions. The first model can characterize the corresponding carbon emissions based on the pre-processed and refined LUCC dataset. Specifically, in this embodiment, based on the energy consumption dataset and land use GIS database, and referring to the IPCC calculation method, the carbon dioxide emission factors for urban heating, electricity consumption, total gas supply, total liquefied petroleum gas supply, and various energy sources are calculated. Carbon emissions from 2000 to 2020 are calculated, including carbon source carbon emissions, forest carbon sequestration, and grassland carbon sequestration. The calculation methods mentioned above are not specifically limited or described in this application; those skilled in the art can choose according to the actual situation. Based on the above calculation results, a first model for characterizing and predicting the regression relationship between carbon emissions and land use is constructed. Using the land use data of a certain year in the GIS geographic database, the carbon emissions for that year can be characterized in real time through the first model. The first model can characterize or predict carbon emissions based on land use data. Inputting current land use data into the first model can obtain carbon emissions under current policies, thereby representing the regression relationship between carbon emissions and land use. Inputting land use data under a preset scenario into the first model can obtain carbon emissions under the predicted scenario, thereby realizing carbon emission prediction based on land use data.

[0056] Step 205: Determine various economic and social development indicators for multiple preset development scenarios, and use the trained first neural network to predict the land use demand for each economic and social development indicator. Specifically, the economic and social development factors affecting the land use demand of built-up areas include: economic development factors (total GDP, fixed asset investment growth rate, etc.) and social factors (population, per capita GDP, etc.). Factors affecting the demand for arable land, grassland, forest land, and water areas include: social factors (employed population in various industries, etc.), product output (grain output, etc.), economic factors (total agricultural output value, etc.), and environmental protection factors (area of ​​land closed for afforestation, area of ​​land converted from farmland to forest), etc. Since there is an autocorrelation relationship between the various land use demand influencing factors, in this embodiment, principal component analysis is used to reduce the information overlap between network input samples. Statistical yearbook data for each year is used as input, and land use data for each year is used as output to train the first neural network. Here, the neural network is exemplarily selected as a PCA-BP neural network. This application embodiment presupposes four development scenarios and determines various economic and social development indicators based on each scenario. A Pareto-optimal solution set for each economic and social development indicator is constructed using multi-objective programming. A trained first neural network is then used to predict the land use demand for each of the aforementioned economic and social development indicators based on this Pareto-optimal solution set. In addition, a baseline scenario is set as a control group; the baseline scenario uses Markov chains to predict land use demand. The Markov chain prediction of land use demand is prior art and should be understood by those skilled in the art; therefore, it will not be described in detail here.

[0057] Step 206: Based on the geographic data and the dataset of factors driving land use expansion, use the random forest algorithm and the LEAS model to obtain an expansion probability atlas for each land use type. Specifically, as follows... Figure 5 As shown, steps 501 to 504 are detailed below:

[0058] Step 501: Select multiple factors as driving forces from the land use expansion driving force factor dataset. Specifically, the driving forces of land use expansion mainly include natural environmental factors and human and social factors. Natural environmental factors include altitude, slope, aspect, average temperature, sunshine, precipitation, water bodies, relative humidity, etc.; human and social factors include population, GDP, public service facilities, commercial and service facilities, roads, etc. In this embodiment, 17 factors are selected from the land use expansion driving force factor dataset, including DEM (Digital Elevation Model), distance to roads at all levels, GDP, distance to public service facilities, annual precipitation, slope and aspect, annual average temperature, population, annual sunshine hours, distance to commercial and service facilities, relative humidity, and distance to water systems. These 17 selected factors are merely an example of this embodiment; those skilled in the art can choose the number of factors according to actual circumstances. Theoretically, the more factors, the more accurate the prediction results.

[0059] Step 502: Perform GIS spatial interpolation and georeferencing on the driving factors, and establish a driving factor dataset based on the processed driving factors. Specifically, some driving factors lack geographic coordinates. Georeferencing is used to match the feature points of the driving factors with corresponding control points with known coordinates, thus establishing correct coordinate information for the driving factors. In this embodiment, ArcGIS's Georeferencing tool is used to perform georeferencing on the driving factors.

[0060] Step 503: Determine the land use succession dataset based on geographic data and calculate the impact of driving factors on land use succession. Specifically, in this embodiment, a land use succession raster dataset is calculated, and an optimized random forest algorithm is used to calculate the impact of 17 driving factors on land use succession. Since water bodies and unused land account for a small proportion of the total area of ​​megacities, the number of pixels involved in land use succession is extremely small. Therefore, sampling random forest training samples according to land use expansion type would result in low accuracy due to the extremely small sample size. Therefore, by optimizing the sampling method of random forest training samples, the sampling ratio for water body and unused land succession categories reaches 0.3, and the sampling ratio for other land use types is 0.01, to increase the scientific validity of the results.

[0061] Step 504: Obtain the expansion probability map of each land use type using the LEAS model based on the driving factor dataset and the land use succession dataset. Specifically, construct the LEAS model based on the training results of the optimized random forest algorithm in Step 503, calculate the probability of succession of each land use type at each location in the study area, and obtain the expansion probability map of each land use type in turn.

[0062] Step 207: Based on the land use GIS database and the expansion probability atlas of land use demand and land use types under various development scenarios, the PLUS model is used to simulate future land use, and the carbon emissions of the simulation results under each development scenario are calculated using the first model. Specifically, based on baseline land use data, expansion probability land for each land use type, and land use demand, the PLUS model is used to simulate future land use, and the carbon emissions of each simulation result are calculated using the first model. The PLUS model is based on multiple random patch seeds for simulation, and its predictions have a certain degree of randomness. Therefore, multiple simulations are performed for each year's prediction, and the median is selected as the prediction result. Those skilled in the art can also use the CLUE-s model or the IDRISE model instead of the PLUS model in this embodiment.

[0063] Figure 2 The flowchart shown is consistent with Figure 1 The research roadmap provided in China for methods and devices for real-time characterization and prediction of carbon emissions based on land use. Figure 1 To.

[0064] While this application provides the method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-inventive labor. The order of steps listed in this embodiment is merely one possible execution order among many and does not represent the only possible execution order. In actual device or client product execution, the methods shown in this embodiment or the accompanying drawings can be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0065] like Figure 6 As shown in the illustration, this application also provides an apparatus 600 for real-time characterization and prediction of carbon emissions based on land use. The apparatus includes: a preprocessing module 601, a fine processing module 602, a statistical yearbook data module 603, a real-time characterization and prediction module 604, a land use demand module 605, a land use type expansion module 606, and a simulation module 607. Details are as follows:

[0066] The preprocessing module 601 is used to preprocess geographic data to obtain first raster data and first vector data; wherein, the geographic data includes a first LUCC dataset, remote sensing imagery, and vector data. Specifically, the preprocessing module 601 is used to perform radiometric calibration and atmospheric correction on the remote sensing imagery, use a trained second neural network to perform supervised classification on the processed remote sensing imagery, and perform accuracy verification on the supervised-classified remote sensing imagery to obtain a second LUCC dataset. The vector data includes POI (Point of Interest) data, administrative boundary data, and road network data. Specifically, in this embodiment, the POI point feature dataset is obtained by crawling the Gaode Map open-source platform using Python; the administrative boundary dataset is obtained using the China Standard Map; and the road network line feature dataset is obtained using the Open Street Map (OSM) platform. The POI data is of point feature type, the administrative boundary data is of area feature type, and the road network data is of line feature type.

[0067] The refinement processing module 602 is used to refine the first raster data and establish a land use GIS database and a land use expansion driving factor dataset; wherein, the first raster data includes the first LUCC dataset and the second LUCC dataset. Specifically, the refinement processing module 601 is used to project the first raster data and the first vector data according to the geographical location of the study area to obtain planar geographic data, reclassify the planar geographic data, extract the raster corresponding to the built-up area, convert the raster corresponding to the built-up area into area features, complete the road network data, divide the area features of the built-up area into street blocks based on the completed road network data, calculate the kernel density values ​​of various POI data, and correlate the kernel density values ​​of various POI data with street blocks to obtain the kernel density values ​​of various POI data for each street block. Based on the kernel density values, the street blocks are divided into single-function land use type units and mixed-function land use type units, and the single-function land use type units and mixed-function land use type units are converted into raster data types and embedded back into the planar geographic data.

[0068] The statistical yearbook storage module 603 is used to establish energy consumption datasets and economic and social development datasets using processed statistical yearbook data. Specifically, the statistical yearbook storage module 603 is used to organize and interpolate data from provincial and prefecture-level city national economic and social development bulletins, as well as data from the China Urban Statistical Yearbook, China Rural Statistical Yearbook, the National Bureau of Statistics National Statistical Data Center, and prefecture-level city statistical yearbooks, and to establish energy consumption datasets and economic and social development datasets.

[0069] The real-time characterization module 604 is used to calculate carbon source carbon emission factors and carbon sink carbon sequestration factors based on the energy consumption dataset and the land use GIS database, and to establish a first model for characterizing and predicting the regression relationship between land use and carbon emissions. Specifically, the real-time characterization module 604 enables the first model to characterize the corresponding carbon emissions based on the pre-processed and refined LUCC dataset. Using the carbon emission factor method, based on the energy consumption dataset and the land use GIS database, and referring to the IPCC calculation method, the carbon dioxide emission factors for urban heating, electricity consumption, total gas supply, total liquefied petroleum gas supply, and each energy source are calculated. Carbon emissions from 2000 to 2020 are calculated, including carbon source carbon emissions, forest carbon sequestration, and grassland carbon sequestration. The above-mentioned calculation methods are not specifically limited or described in this application; those skilled in the art can choose according to the actual situation. Based on the above calculation results, a first model for characterizing and predicting the regression relationship between carbon emissions and land use is constructed.

[0070] The land use demand module 605 is used to determine various economic and social development indicators for multiple preset development scenarios, and to predict the land use demand for each of these economic and social development indicators using a trained first neural network. Specifically, the land use demand module 605 uses principal component analysis to reduce information overlap between network input samples, using annual statistical yearbook data as input and annual land use data as output to train the first neural network; here, the neural network is exemplarily selected as a PCA-BP neural network. In this embodiment, four development scenarios are preset, and various economic and social development indicators are determined according to each scenario. A Pareto optimal solution set for each economic and social development indicator is constructed using multi-objective programming; the trained first neural network is used to predict the land use demand for each of these indicators based on the Pareto optimal solution set. In addition, a baseline scenario is set as a control group; wherein, the baseline scenario uses a Markov chain to predict land use demand. The Markov chain prediction of land use demand is prior art, which should be understood by those skilled in the art, and therefore will not be described in detail here.

[0071] The land use type expansion module 606 is used to obtain expansion probability atlases for each land use type based on the geographic data and the land use expansion driving factor dataset using a random forest algorithm and a LEAS model. Specifically, the land use type expansion module 606 is used to: select multiple factors as driving factors from the land use expansion driving factor dataset; perform GIS spatial interpolation and georeferencing on the driving factors, and establish a driving factor dataset based on the processed driving factors; determine the land use succession dataset based on the geographic data, and calculate the impact of the driving factors on land use succession; and obtain expansion probability atlases for each land use type using the LEAS model based on the driving factor dataset and the land use succession dataset.

[0072] The simulation and prediction module 607 is used to simulate future land use using the PLUS model based on the land use GIS database, the land use demand of multiple development scenarios, and the expansion probability atlas of the land use types, and to calculate the carbon emissions of the simulation results of the development scenarios using the first model. Specifically, the simulation and prediction module 607 is used to simulate future land use using the PLUS model based on baseline land use data, the expansion probability of each land use type, and land use demand, and to calculate the carbon emissions of each simulation result using the first model. The PLUS model is based on multiple random patch seeds for simulation, and its predictions have a certain degree of randomness. Therefore, multiple simulations are performed for each year's prediction, and the median is selected as the prediction result. Those skilled in the art can also use the CLUE-s model or the IDRISE model instead of the PLUS model in this application embodiment, but the simulation accuracy will vary depending on the model. The simulation accuracy obtained using the PLUS model in this application embodiment is the highest.

[0073] Some modules in the apparatus described in this application can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0074] The apparatus or module described in the above embodiments can be implemented by a computer chip or physical entity, or by a product with a certain function. For ease of description, the above apparatus is described by dividing it into various modules according to their functions. When implementing the embodiments of this application, the functions of each module can be implemented in one or more software and / or hardware. Of course, a module that implements a certain function can also be implemented by combining multiple sub-modules or sub-units.

[0075] This application also provides a non-volatile computer-readable storage medium storing a computer program or instructions thereon, which, when executed, enables the method described in this application embodiment to be implemented.

[0076] Furthermore, in the various embodiments of the present invention, each functional module can be integrated into a processing module, or each module can exist independently, or two or more modules can be integrated into a single module.

[0077] The aforementioned storage media include, but are not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Cache, Hard Disk Drive (HDD), or Memory Card. The memory can be used to store computer program instructions.

[0078] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary hardware. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, or it can be embodied in the process of data migration. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0079] The various embodiments described in this specification are presented in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. All or part of this application can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0080] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of this application.

Claims

1. A method for real-time characterization and prediction of carbon emissions based on land use, characterized in that, include: Geographic data is preprocessed to obtain first raster data and first vector data; wherein, the geographic data includes a first LUCC dataset, remote sensing imagery, and vector data; The first raster data is refined, and a land use GIS database and a land use expansion driving force factor dataset are established; wherein, the first raster data includes the first LUCC dataset and the second LUCC dataset; Use the processed statistical yearbook data to establish an energy consumption dataset and an economic and social development dataset; Carbon source carbon emission factors and carbon sink carbon factors are calculated based on the energy consumption dataset and the land use GIS database, and a first model is established to characterize and predict the regression relationship between land use and carbon emissions; wherein, the first model can characterize the corresponding carbon emissions based on the preprocessed and refined LUCC dataset. Determine various economic and social development indicators for multiple preset development scenarios, and use a trained first neural network to predict the land use demand for each of the economic and social development indicators. Based on the geographic data and the land use expansion driving factor dataset, an expansion probability atlas for each land use type is obtained using the random forest algorithm and the LEAS model. This includes: selecting multiple factors as driving factors from the land use expansion driving factor dataset; wherein the driving factors include: land succession natural environment driving factors and land succession social development driving factors; performing GIS spatial interpolation and georeferencing on the driving factors, and establishing a driving factor dataset based on the processed driving factors; determining the land use succession dataset based on the geographic data, and calculating the impact of the driving factors on land use succession; and obtaining an expansion probability atlas for each land use type using the LEAS model based on the driving factor dataset and the land use succession dataset. Based on the land use GIS database, the land use demand and expansion probability atlas of the land use types under the various development scenarios, the PLUS model is used to simulate future land use, and the carbon emissions of the simulation results under the development scenarios are calculated using the first model.

2. The method according to claim 1, characterized in that, The vector data includes POI data, administrative boundary data, and road network data.

3. The method according to claim 2, characterized in that, The second LUCC dataset includes: The remote sensing images are subjected to radiometric calibration and atmospheric correction. The processed remote sensing images are classified under supervised supervision using a trained second neural network. The accuracy of the supervised classification remote sensing images is verified to obtain the second LUCC dataset.

4. The method according to claim 3, characterized in that, The refinement processing of the first raster data includes: Based on the geographical location of the study area, the first raster data and the first vector data are projected to obtain planar geographic data; The planar geographic data is reclassified, the raster corresponding to the built-up area is extracted, and the raster corresponding to the built-up area is converted into polygon features; Complete the road network data, and divide the surface elements of the built-up area into street blocks based on the completed road network data; Calculate the kernel density value of each type of POI data, and associate the kernel density value of each type of POI data with the block unit to obtain the kernel density value of each type of POI data for each block unit; Based on the kernel density value, the block unit is divided into single-function land use type units and mixed-function land use type units; The single-function land use type unit and the mixed-function land use type unit are converted into raster data types and embedded back into the planar geographic data.

5. The method according to claim 1, characterized in that, The method of using a trained first neural network to predict land use demand for various economic and social development indicators includes: Use multi-objective programming to construct the Pareto optimal solution set for each of the aforementioned economic and social development indicators; The trained first neural network is used to predict the land use demand for each of the economic and social development indicators based on the Pareto optimal solution set.

6. The method according to claim 5, characterized in that, Also includes: Reduce the autocorrelation between the input samples in the yearbook data; The first neural network is trained using the processed input sample as input and the land use GIS database as output.

7. The method according to claim 5, characterized in that, The method of using the trained first neural network to predict land use demand for each of the aforementioned economic and social development indicators also includes: A baseline scenario is set as the control group; wherein, the baseline scenario uses Markov chains to predict the land use demand.

8. A device for real-time characterization and prediction of carbon emissions based on land use, characterized in that, include: A preprocessing module is used to preprocess geographic data to obtain first raster data and first vector data; wherein, the geographic data includes a first LUCC dataset, remote sensing imagery, and vector data; The fine-processing module is used to perform fine-processing on the first raster data and establish a land use GIS database and a land use expansion driving factor dataset; wherein, the first raster data includes the first LUCC dataset and the second LUCC dataset; The statistical yearbook storage module is used to create energy consumption datasets and economic and social development datasets using processed statistical yearbook data. The real-time characterization module is used to calculate carbon source carbon emission factors and carbon sink carbon factors based on the energy consumption dataset and the land use GIS database, and to establish a first model for characterizing and predicting the regression relationship between land use and carbon emissions; wherein, the first model can characterize the corresponding carbon emissions based on the preprocessed and refined LUCC dataset. The land use demand module is used to determine various economic and social development indicators under multiple preset development scenarios, and to predict the land use demand for each of the economic and social development indicators using a trained first neural network. The land use type expansion module is used to obtain an expansion probability atlas for each land use type based on the geographic data and the land use expansion driving factor dataset using a random forest algorithm and a LEAS model. This includes: selecting multiple factors as driving factors from the land use expansion driving factor dataset; wherein the driving factors include: land succession natural environment driving factors and land succession social development driving factors; performing GIS spatial interpolation and georeferencing on the driving factors, and establishing a driving factor dataset based on the processed driving factors; determining a land use succession dataset based on the geographic data, and calculating the impact of the driving factors on land use succession; and obtaining an expansion probability atlas for each land use type using the LEAS model based on the driving factor dataset and the land use succession dataset. The simulation and prediction module is used to simulate future land use using the PLUS model based on the land use GIS database, the land use demand of multiple development scenarios, and the expansion probability atlas of the land use types, and to calculate the carbon emissions of the simulation results of the development scenarios using the first model.

9. A non-volatile computer-readable storage medium, characterized in that, Includes storage of computer programs or instructions that, when executed, cause the method as described in any one of claims 1 to 7 to be implemented.

Images