A block carbon sink distribution optimization method applied to carbon trading

Abstract

The application relates to a block carbon sink distribution optimization method applied to carbon trading, relates to the technical field of carbon sink trading, and aims to solve the problem of low carbon sink result accuracy caused by the fact that existing carbon sink calculation only considers the total carbon fixation amount of green land and ignores the carbon emission amount generated by energy consumption of equipment machinery during operation and maintenance, production, transportation of fertilizers and pesticides, and plant residue burning or landfilling. Unlike the traditional carbon sink calculation method which only focuses on the total carbon fixation amount of green land, the application considers the carbon emission amount generated by the block patch during operation and maintenance, estimates the net carbon fixation amount of the green land by subtracting the carbon emission amount from the total carbon fixation amount, and greatly improves the precision and efficiency of block carbon sink calculation.

Description

Technical Field

[0001] This application relates to the field of carbon sink trading technology, specifically a method for optimizing the distribution of carbon sinks in urban areas for carbon trading. Background Technology

[0002] Against the backdrop of global climate change, green and low-carbon development has become a common goal pursued by all countries. Guided by the "dual carbon" goals (peaking carbon before 2030 and achieving carbon neutrality before 2060), China is actively building a national carbon market system and promoting regional low-carbon development and technological innovation to achieve a high-quality green transformation. Cities, as the main areas of concentrated carbon emissions and also the core areas of carbon reduction potential, have become crucial to the global low-carbon transformation.

[0003] Neighborhood carbon sequestration, as a crucial natural process for achieving urban carbon neutrality, absorbs and fixes atmospheric carbon dioxide through vegetation, soil, and related ecosystems, becoming a key pathway for urban low-carbon transformation. Under the current trend of ecological construction dominated by nature-based solutions, neighborhood carbon sequestration is gradually being systematically incorporated into urban planning and carbon market systems. Carbon trading is an important market mechanism for achieving low-carbon development; its core lies in promoting emission reduction targets through the allocation and trading of emission rights. With increasing global emphasis on carbon emission limits, countries have successively established carbon trading systems to provide emission reduction incentives to enterprises through market economic means.

[0004] Despite significant progress in advancing low-carbon development and achieving its "dual carbon" goals, China still faces numerous challenges in areas such as local climate regulation, enhancing urban carbon sequestration, and establishing a carbon trading system. Current carbon sequestration calculations only consider the total carbon sequestration of green spaces, neglecting the energy consumption of equipment and machinery during operation and maintenance, as well as carbon emissions from fertilizer and pesticide production and transportation, and the incineration or landfilling of plant residues. This leads to low accuracy in carbon sequestration results. Summary of the Invention

[0005] The purpose of this invention is to address the problem that existing carbon sequestration calculations only consider the total carbon sequestration of green spaces, neglecting the energy consumption of equipment and machinery during operation and maintenance, as well as the carbon emissions from fertilizer and pesticide production and transportation, and the incineration or landfilling of plant residues. This leads to low accuracy in carbon sequestration results. Therefore, this invention provides a method for optimizing the distribution of urban carbon sequestration in carbon trading.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] A method for optimizing the distribution of carbon sinks in urban areas for carbon trading, the method comprising the following steps:

[0008] Step 1: Extract street blocks for the study area and classify the street blocks;

[0009] Step 2: For each block unit, generate a carbon sink patch distribution map for each block unit;

[0010] Step 3: Determine the types of carbon sinks, including urban woodlands, community parks, and neighborhood green spaces;

[0011] Step 4: Determine the design parameters based on the factors affecting the carbon sequestration capacity of the carbon sink, and use the net carbon sequestration capacity of the carbon sink as the optimization parameter;

[0012] Step 5: Based on the carbon sink patch distribution map, and according to the type of carbon sink, adjust the design parameters to obtain multiple adjustment schemes, and obtain the optimization parameters corresponding to each adjustment scheme;

[0013] Step 6: Compare the optimization parameters corresponding to each adjustment scheme with the threshold corresponding to the block unit to obtain the scheme with the smallest carbon sink difference;

[0014] Step 7: Repeat the above steps to obtain the scheme with the minimum carbon sink difference for all block units, i.e., the optimal carbon sink distribution map.

[0015] Furthermore, step 1 specifically includes:

[0016] Step 101: Acquire GF-1 satellite panchromatic and multispectral images of the study area, and preprocess the GF-1 satellite panchromatic and multispectral images, including correction and denoising;

[0017] Step 102: The preprocessed image is segmented using the SLIC superpixel algorithm, and the optimal segmentation scale is searched by local variance and edge density to obtain the initial segmented image;

[0018] Step 103: Based on the initial segmented image, the road skeleton extraction network is used to identify the road skeleton, and the road skeleton is buffered to form a back line strip according to the preset buffer distance. Finally, the back line strip is used to segment the image into blocks.

[0019] Step 104: If the area of ​​a single block is not greater than 1 / 2 of the average area of ​​all the divided blocks, the block is considered too small. If the average length-to-width ratio of a single block exceeds 4, the block is considered too long and narrow. Then, blocks that are too small or too long and narrow are merged.

[0020] Step 105: Determine the natural boundary line within each block and perform secondary segmentation on areas where the differences between building clusters exceed a threshold. The natural boundary line includes rivers, water bodies, and green spaces, and the differences between building clusters include abrupt changes in density, layout, and building height.

[0021] Step 106: Classify the block units after segmentation in Step 105 according to their functional attributes.

[0022] Furthermore, in step 106, the classification is performed by a neural network. The neural network extracts the features of the segmented block units and classifies them with functional attributes as labels. The features include spectral and index features, texture features, geometric morphology features, building density and layout features, and building height inference features.

[0023] Furthermore, the functional attributes include residential-dominated areas, production-dominated areas, commercial and business-dominated areas, public service-dominated areas, cultural and educational-dominated areas, green space and water-dominated areas, mixed-function-dominated areas, and special-function-dominated areas.

[0024] Furthermore, the specific steps of step 2 are as follows:

[0025] Step 201: Use the single block unit extracted in step 1 as the base map. Based on this base map, extract the roof discrimination features, and use a binary classification model to identify the greenable roof / non-greenable roof and output a probability map, and then overlay the preset rules to obtain the greenable roof mask; the roof discrimination features include at least one of roof morphology, roof material reflectivity, structural integrity, floor height, and surrounding obstacle shadow duration; the preset rules include setting the minimum continuous greenable area threshold and the roof slope threshold, and cleaning the noise through morphological opening / closing operations;

[0026] Step 202: Based on the greenable roof mask, obtain the building land mask, and obtain the pixel set Mfeasible for arranging carbon sink patches according to the building land mask, which is expressed as:

[0027]

[0028] where, is the block unit space set, is the building land mask, is the road land mask, is the other hardened surface mask, is the convertible roof space set;

[0029] Step 203: Perform connected component analysis and region merging on Mfeasible, remove the patches smaller than the minimum effective ecological patch area threshold, and divide the connected components into linear corridor patches according to the width threshold;

[0030] Step 204: Set the area threshold Sp, the aspect ratio threshold Rl, and the average width threshold Wi. If the area threshold ≥ Sp, it is a planar patch; if the aspect ratio ≥ Rl and the average width ≤ Wi, it is a linear patch; if the area < Sp and it is isolated, it is a point patch. At the same time, calculate and record the morphological indexes of each patch. The morphological indexes include at least one of boundary complexity, aggregation degree, and landscape connectivity.

[0031] Step 205: Convert the patches into vector data and associate them with the block unit ID. Write attribute information into the attribute table. The attribute information includes at least one of the following: the block unit ID, patch type, area S, boundary complexity LSI, proportion of greenable roofs, and candidate range of vegetation types. Thus, a carbon sink patch distribution map of each block unit is generated.

[0032] Furthermore, the specific steps of step 3 are as follows:

[0033] Based on the climate and environment of the city, suitable carbon sink types are selected, and then the initially selected carbon sink types are further classified according to shape and area, as follows:

[0034] Large and medium-sized patches are divided into planar areas, including urban woodlands, community parks, and neighborhood green spaces;

[0035] The corridor patches are divided into linear shapes, including urban corridors, roadside greenways, and leisure greenways;

[0036] Small patches are divided into dots, including roof gardens and bioretention facilities.

[0037] Furthermore, the natural boundary is defined as the river and the green space.

[0038] Furthermore, the design parameters are patch boundary complexity, vegetation cover type, and vegetation cover area.

[0039] Furthermore, the net carbon sequestration of the carbon sink is expressed as:

[0040]

[0041] Where m is the boundary complexity coefficient, n is the number of carbon sink patches in the unit, S is the vegetation coverage area of ​​each patch, x is the carbon sequestration parameter of different types of vegetation, and y is the carbon sink operation and maintenance and its own carbon emission parameter.

[0042] Furthermore, the correction specifically includes:

[0043] First, radiometric calibration and atmospheric correction were performed on the panchromatic and multispectral images of the GF-1 satellite. Then, the elevation control points were fused to perform geometric correction and orthorectification.

[0044] The denoising includes histogram equalization and bilateral filtering / nonlocal mean filtering.

[0045] The beneficial effects of this invention are:

[0046] The technical solution proposed in this application differs from traditional carbon sequestration calculation methods that only focus on the total carbon sequestration of green spaces. This application considers the carbon emissions generated by the blocks during operation and maintenance, estimating the net carbon sequestration of green spaces by subtracting carbon emissions from the total carbon sequestration. This significantly improves the accuracy and efficiency of block carbon sequestration calculations.

[0047] Furthermore, traditional carbon trading has a broad scope of application. This application limits carbon trading to blocks with a single land use, making the carbon trading mechanism more scientific and reasonable. This method reduces the uncertainty caused by diversity in the trading process and provides technical support for promoting a precise carbon trading mechanism. Attached Figure Description

[0048] Figure 1 This is the overall flowchart of this application. Detailed Implementation

[0049] It should be noted that, where there is no conflict, the various embodiments disclosed in this application can be combined with each other.

[0050] Specific Implementation Method 1: This implementation method describes a method for optimizing the distribution of carbon sinks in urban areas used in carbon trading, including:

[0051] Step 1: Extraction and Classification of Street Blocks: Based on GF-1 imagery, a combination of manual visual interpretation and supplementary field surveys was used to identify green space patches and water bodies in the study area. Specifically, urban street block units were initially divided using street-level vehicular road setbacks as boundaries. Subsequently, the unit boundaries were refined based on natural boundaries (rivers, green spaces, etc.) and areas of significant visual difference such as buildings, to reduce the degree of blending between adjacent unit boundaries. Finally, the extracted street block units were classified according to functional attributes (eight dominant functional zones: residential, production, commercial / business, public service, cultural / educational, green space / water area, mixed-use, and special-function).

[0052] Data Acquisition and Preprocessing

[0053] ① Image acquisition: Panchromatic and multispectral images of the study area were acquired from the China Resources Satellite Application Center;

[0054] ②Radiative and atmospheric correction: Radiometric calibration and atmospheric correction are performed on the original images to eliminate the effects of solar elevation angle, sensor gain differences and atmospheric scattering;

[0055] ③ Geometric and orthorectification correction: By integrating DEM / elevation control points, geometric and orthorectification corrections are completed to ensure consistent spatial positioning accuracy of pixels in different scenes and different bands on the ground;

[0056] ④ Image enhancement and denoising: Histogram equalization, bilateral filtering / nonlocal mean filtering and other processing are applied to the image to enhance the distinguishability of texture and structural boundaries.

[0057] Initial segmentation

[0058] ① The super pixel algorithm SLIC is used to divide the image into initial image objects with spectral and textural homogeneity;

[0059] ② Adaptive optimization of scale parameters: The optimal segmentation scale is automatically searched through indicators such as local variance and edge density (constructing a dual-indicator adaptive optimization model based on local variance and edge density. The core is to use the two indicators of local variance and edge density to construct a comprehensive evaluation function, and to automatically search to make the function reach the extreme value, thereby determining the optimal segmentation scale) to avoid oversegmentation and undersegmentation.

[0060] Feature construction (for machine learning to identify street boundaries and types)

[0061] Calculate the following features for the target plot:

[0062] ①Spectral and index-based: NDVI (remote sensing surrogate index for vegetation cover FVC), NDBI (building index), MNDWI (water index), SAVI, etc.

[0063] ②Texture features: Contrast, homogeneity, entropy, etc. are calculated using the Gray-Level Co-occurrence Matrix (GLCM);

[0064] ③ Geometric morphological characteristics: area, perimeter, aspect ratio, rectangularity, fractal dimension, shape index (such as LSI), etc.

[0065] ④ Spatial structural characteristics: Calculate road density, connectivity, and node degree based on the road skeleton (see 4-①);

[0066] ⑤ Building density and layout: Based on the building extraction results (see 4-②), the building coverage, building spacing distribution, and building orientation consistency within the unit are statistically analyzed (which can be measured by the principal component orientation or the principal orientation of the Hough transform).

[0067] ⑥ Building height estimation: The building height h = L × tan(α) is estimated by using the shadow length L and the solar altitude angle α; or by performing parallax calculation based on multi-period images / stereo pairs; the obtained building height statistics (average height, maximum height, coefficient of variation) are used as auxiliary features to distinguish between high-density / low-density and point-like / panel-like layouts.

[0068] Deep learning segmentation

[0069] ① Road network extraction: Semantic segmentation networks such as U-Net are used to extract road elements (road elements refer to the basic components that need to be considered in the process of road design, construction and maintenance, including geometric characteristics, design standards and traffic flow, etc.).

[0070] Pixel-level segmentation is performed; post-processing includes thinning and morphological opening / closing operations to remove pseudo-noise and fill in broken road segments, constructing a road skeleton network. The road network extraction uses U-Net as the main path and SLIC superpixel results as post-processing constraints.

[0071] ② Building and Impermeable Surface Extraction: U-Net is used for building outline segmentation; NDBI, texture, and shape constraints (compactness, rectangularity) are used to correct model misclassification. A rule-based correction process, based on NDBI / texture / shape rules, corrects the depth model output, reducing misclassification of similar objects and fragmentation.

[0072] Impermeable surfaces refer to artificially covered areas where rainwater cannot naturally infiltrate, including roads, squares, parking lots, and other hard paved surfaces.

[0073] First, multiple indices such as NDBI, UI, and NDVI are calculated to suppress misclassification of non-impermeable categories such as vegetation and water bodies through multi-index combination. Then, using multispectral images and texture features as input, a U-Net or Random Forest model is trained to perform pixel-level classification and output an impermeable surface probability map. Finally, isolated noise is removed based on morphological operations (opening / closing) and connected component analysis to obtain a continuous and complete impermeable surface mask.

[0074] Building outline refers to the external shape and boundaries of a building, and is usually used to describe the building's spatial layout and height information.

[0075] "Land type" refers to land cover / surface type, a commonly used term in remote sensing and geographic information systems. It includes, but is not limited to: vegetation (high-density vegetation, low-density / shrubland), grassland / park, bare land / leaf soil / bare soil, buildings, roads / hard paving (impermeable surfaces), water bodies (rivers, ponds, pools), and other artificial covers (industrial yards, parking lots), etc.

[0076] ③ Identification of natural land cover layers (vegetation, water bodies, bare land, etc.): Random forest (RF) or deep learning multi-classification network is used for land cover identification; manual annotation + field supplementary survey is used as training samples, and K-fold cross-validation and accuracy evaluation (OA, Kappa, mIoU) are performed.

[0077] Automatic segmentation of street blocks based on road framework

[0078] ①Use the setback line of the vehicular road as the initial boundary of the block: buffer the road skeleton at a certain distance (this distance is a fixed value, the setback distance on both sides of the main urban road is 15 meters, the setback distance on the secondary road is 10 meters, and the setback distance on the side road is 5 meters) to form a "setback line" which is used as the block dividing line; perform topological cutting on the building / green space objects within the buffer zone to ensure that the boundaries between blocks are clear and do not overlap.

[0079] ② Boundary Refinement (Natural Boundary Lines + Visual Differences): The initial block is further subdivided based on natural boundaries such as rivers, water bodies, and green spaces; for areas with obvious differences in building clusters (abrupt changes in density, layout, and building height), segmented statistics are used to identify "heterogeneous blocks" and perform secondary subdivision to reduce the degree of mixing between adjacent units.

[0080] Among these, the differences are: (in terms of density) building density by more than double; (in terms of layout and building height) or average floor area of ​​individual buildings by more than double.

[0081] The heterogeneous block is a local heterogeneous region within a block that differs significantly from the surrounding land features in terms of spectral, texture, morphology, or spatial structure.

[0082] ③ Minimum area and shape rule constraints: Automatic merging based on graph cut / region fusion is performed on units with excessively small area (if the area of ​​a single block is not greater than 1 / 2 of the average area of ​​all divided blocks, it is considered too small) or excessively long and narrow shape (if the average length-to-width ratio of a single block exceeds 4, it is considered too long and narrow) to ensure that the block units meet the planning scale and geometric rationality.

[0083] Street block unit functional classification (8 types of dominant functional areas)

[0084] ① Feature system: Spatial and morphological features: building density, road density, average building height, floor area ratio (proxy, elevation × building coverage), cohesion, boundary complexity (LSI); Ecological features: FVC, vegetation cover type (VCT), water surface ratio (WR): POI density, nighttime light intensity, population density (PD), etc. (if data is available);

[0085] ② Multi-classification model: Machine learning models or graph neural networks (GNNs) are used to identify eight dominant functional zones in the blocks (residential dominant zone, production dominant zone, commercial and business dominant zone, public service dominant zone, cultural and educational dominant zone, green space and water area dominant zone, mixed function dominant zone, and special function dominant zone); the classification results are evaluated using confusion matrices and F1-scores; for blocks with confidence scores below the threshold, an active learning mechanism of manual review and retraining iteration is triggered;

[0086] ③ Output: A spatial database of “block unit boundary - functional type - key structural parameters” is formed, providing base map and element fields for subsequent carbon sink patch distribution mapping and parametric optimization.

[0087] The key structural parameters refer to quantifiable elements used to characterize the functional and ecological design of a neighborhood, which may include basic geometry, morphological indices, ecological parameters, building-related parameters, spatial structure, social parameters, operability parameters, etc. These parameters are used for classification and identification, as well as as input fields for subsequent optimization.

[0088] First, the urban blocks requiring optimized design were identified. GF-1 imagery of the corresponding blocks was downloaded from the website of the China Resources Satellite Application Center, and block units were extracted. These blocks were then categorized based on function (eight dominant functional zones, including residential, production, and commercial / business zones). Using these blocks as a base map, building cover, roads, and hard paving were subtracted, and potentially convertible rooftop greening was added to create a carbon sink distribution map for each block. Based on shape and area, carbon sinks were classified into three categories: areal – large and medium-sized patches; linear – corridor patches; and dot – small patches.

[0089] Roads and the boundaries of each block usually have a setback of 2 to 9 meters. The setback distance depends on the road grade. There are usually green spaces within the setback area, but these green spaces do not belong to the block unit, but to the road, and are not within the scope of this study.

[0090] Step 2: Draw a map of carbon sink patch distribution: Using the individual block units extracted in Step 1 as a base map, abstraction is performed using graphical representation methods. Based on this base map, available / unavailable spatial masks are identified:

[0091] ① Automatic identification of unusable spaces: Building coverage: The building surface is directly obtained from the segmentation results in step 4 of step 1; Road traffic and hard paving: Obtained from the road network and impermeable surface segmentation results in step 4 of step 1; Other restricted areas: such as high-voltage corridors, buffer zones for cultural heritage buildings, and the coverage area of ​​important above-ground and underground municipal facilities (if there is auxiliary data, it can be superimposed and excluded).

[0092] ② Roof Greenability Identification: Features for determining roof greenability: roof shape (flat roof / sloping roof, average slope), roof material reflectivity, structural integrity (texture uniformity), floor height (load-bearing proxy), and shadow duration from surrounding obstacles (light accessibility proxy); Discrimination Model: A binary classification model is used to identify "greenable roofs / non-greenable roofs" and output a probability map; Rule Overlay: A minimum continuous greenable area threshold (e.g., ≥Xm²) and a roof slope threshold (e.g., ≤Y°) are set, and noise is cleared through morphological operations (open / closed) to obtain a roof greenability mask.

[0093] Among them, hard paving refers to the ground surface covered with hard materials such as concrete / asphalt / stone that cannot permeate water, including: roads, squares, parking lots, sidewalks, industrial yards, roofs (non-permeable roofs), etc.

[0094] ③ Identification of potentially carbon sink-increasing bare land / inefficient green spaces: Conduct object-oriented clustering (k-means / DBSCAN) on bare land and areas with low vegetation cover (low NDVI), and combine land use control rules to screen carbon sink-increasing patches.

[0095] Generate the "Carbon Sink Patch Feasibility Map"

[0096] ① Mask operation:

[0097]

[0098] Among them, Mfeasible is the set of pixels available for laying carbon sink patches, is the set of block unit spaces, is the building land mask, is the road land mask, is the mask of other hardened ground surfaces, is the set of convertible roof spaces;

[0099] ② Patch aggregation and minimum scale control:

[0100] Conduct connected component analysis and regional merging on Mfeasible, and remove patches smaller than the threshold of the minimum effective ecological patch area; divide long and narrow connected components into linear corridor patches according to the width threshold;

[0101] ③ Patch morphological indicators and category determination:

[0102] Planar patches (medium and large): area threshold ≥ Sp;

[0103] Linear patches (corridors): length-width ratio ≥ Rl, average width ≤ Wi;

[0104] Point patches (small): convertible roof gardens, bioretention facilities, etc. with an area < Sp and isolated distribution; calculate and record the LSI, Cohesion, TGR, etc. of each patch, providing input for the subsequent calculation of "design parameters - optimization parameters";

[0105] ④ Result vectorization and attribute annotation:

[0106] Convert the patch raster to vector and associate it with the block unit ID; write in the attribute table: the ID of the所属 block, patch type (planar / linear / point), area S, boundary complexity LSI, proportion of greenable roofs, candidate range of vegetation type (VCT), etc.;

[0107] ⑤ Quality control and manual verification closed loop: For locations with low confidence or those misjudged multiple times by the model (based on the inconsistency heatmap), output to the "manual verification checklist"; the manual correction results are fed back into the training set to update the model parameters until the patch extraction accuracy meets the threshold (e.g., mIoU≥T).

[0108] Connection with subsequent optimization stages

[0109] The generated "carbon sink patch distribution map" will serve as the input base map for the Rhino / Grasshopper parametric platform:

[0110] ① Each patch object has initial values ​​for design parameters (such as LSI, VCT, Area);

[0111] ② In the optimization platform, the net carbon sequestration per unit area is used as the optimization target, and iterative searches are conducted on patch type selection, area adjustment, boundary complexity correction, etc.;

[0112] ③ Output the set of optimized parameter values ​​for each iteration, and stop when the maximum balance or minimum difference is reached to obtain the optimal carbon sink distribution scheme.

[0113] Step 3: Carbon sink classification: Based on the climate and environment of the city, select suitable carbon sink types, and then classify the initially screened carbon sinks into three categories according to their shape and area: areal - large and medium-sized patches (urban woodlands, community parks, street green spaces, etc.); linear - corridor patches (urban corridors, roadside greenways, leisure greenways, etc.); point - small patches (roof gardens, bioretention facilities, etc.).

[0114] Step 4: Determine Design and Optimization Parameters: Based on existing research, the following factors influence carbon sequestration: vegetation cover (FVC), vegetation cover area (Area), vegetation cover type (VCT), cohesion, tree green ratio (TGR), patch boundary complexity (LSI), population density (PD), and water surface ratio (WR). Factors related to carbon sequestration distribution at the block scale are selected as design parameters. After initial selection, some parameters are eliminated based on their degree of influence. The final design parameters are: patch boundary complexity (LSI) (influencing m), vegetation cover type (VCT) (influencing x and y), and vegetation cover area (Area) (influencing S). The concept of net carbon sequestration is proposed: Net carbon sequestration = Carbon sequestration amount - Carbon emissions required for maintaining this portion of the carbon sequestration (fertilizer, irrigation, transportation, etc.) - Carbon emissions from the carbon sequestration itself. Step 5: Process Simulation and Result Output: Based on the carbon sink patch distribution map obtained in Step 2, the design parameters of the carbon sink types selected in Step 3 are dynamically adjusted using the Rhino parametric platform. The optimized parameters for each adjustment scheme are output and presented in the form of point sets. The net carbon sequestration of the carbon sink is calculated using the formula... The calculation yields (m is the boundary complexity coefficient, n is the number of carbon sink patches within the unit, S is the vegetation coverage area of ​​each patch, x is the carbon sequestration parameter of different types of vegetation, and y is the carbon sink operation and maintenance and its own carbon emission parameter); the net carbon sequestration per unit area M=F / S is compared with the corresponding functional area standard value M1. When M>M1, the excess carbon sink can be sold to the same land use unit; when M<M1, the carbon sink deficit needs to be purchased from the same land use unit.

[0115] By repeatedly adjusting the design parameters, the optimal solution with the maximum balance or minimum difference is found, and the optimal carbon sink distribution map is output graphically.

[0116] Design and optimization parameters were determined, and corresponding optimization constraints were set. Design parameters included patch boundary complexity, vegetation cover type, and vegetation cover area; the optimization parameter was net carbon sequestration per unit area. Based on these design parameters, a parametric model of the neighborhood was established using the Rhino and Grasshopper platforms. Using the mathematical components in Grasshopper, the net carbon sequestration per unit area was calculated and output according to the formula. Both design and optimization parameters were simultaneously input into the Galapagos optimization platform, aiming to maximize the optimization parameters. The optimizer iteratively searches for better solutions based on the initial parameters, displaying the optimization process curves and solution distribution in real time. After selecting the optimal solution in Galapagos, its variable values ​​were applied to the Grasshopper model. The net carbon sequestration per unit area M=F / S in the obtained optimization scheme is compared with the standard value M1 of the corresponding functional area. When M>M1, the block can sell excess carbon sequestration to units with the same land use nature. When M<M1, it needs to purchase carbon sequestration deficit from units with the same land use nature.

[0117] It should be noted that the specific embodiments are merely explanations and illustrations of the technical solution of the present invention and should not be used to limit the scope of protection. Any modifications made in accordance with the claims and specification of the present invention that are only partial should still fall within the protection scope of the present invention.

Claims

1. A method for optimizing the distribution of carbon sinks in urban areas for carbon trading, characterized in that... The method includes the following steps: Step 1: Extract street blocks for the study area and classify the street blocks; Step 2: For each block unit, generate a carbon sink patch distribution map for each block unit; Step 3: Determine the types of carbon sinks, including urban woodlands, community parks, and neighborhood green spaces; Step 4: Determine the design parameters based on the factors affecting the carbon sequestration capacity of the carbon sink, and use the net carbon sequestration capacity of the carbon sink as the optimization parameter; Step 5: Based on the carbon sink patch distribution map, and according to the type of carbon sink, adjust the design parameters to obtain multiple adjustment schemes, and obtain the optimization parameters corresponding to each adjustment scheme; Step 6: Compare the optimization parameters corresponding to each adjustment scheme with the threshold corresponding to the block unit to obtain the scheme with the smallest carbon sink difference; Step 7: Repeat the above steps to obtain the scheme with the minimum carbon sink difference for all block units, i.e., the optimal carbon sink distribution map.

2. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 1, characterized in that... Step 1 specifically involves: Step 101: Acquire GF-1 satellite panchromatic and multispectral images of the study area, and preprocess the GF-1 satellite panchromatic and multispectral images, including correction and denoising; Step 102: The preprocessed image is segmented using the SLIC superpixel algorithm, and the optimal segmentation scale is searched by local variance and edge density to obtain the initial segmented image; Step 103: Based on the initial segmented image, the road skeleton extraction network is used to identify the road skeleton, and the road skeleton is buffered to form a back line strip according to the preset buffer distance. Finally, the back line strip is used to segment the image into blocks. Step 104: If the area of ​​a single block is not greater than 1 / 2 of the average area of ​​all the divided blocks, the block is considered too small. If the average length-to-width ratio of a single block exceeds 4, the block is considered too long and narrow. Then, blocks that are too small or too long and narrow are merged. Step 105: Determine the natural boundary line within each block and perform secondary segmentation on areas where the differences between building clusters exceed a threshold. The natural boundary line includes rivers, water bodies, and green spaces, and the differences between building clusters include abrupt changes in density, layout, and building height. Step 106: Classify the block units after segmentation in Step 105 according to their functional attributes.

3. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 2, characterized in that... In step 106, the classification is performed by a neural network. The neural network extracts the features of the segmented block units and classifies them by functional attributes. The features include spectral and exponential features, texture features, geometric features, building density and layout features, and building height inference features.

4. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 3, characterized in that... The functional attributes include residential-dominated areas, production-dominated areas, commercial and business-dominated areas, public service-dominated areas, cultural and educational-dominated areas, green space and water area-dominated areas, mixed-use-dominated areas, and special-function-dominated areas.

5. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 4, characterized in that... The specific steps of step 2 are as follows: Step 201: Use the single block unit extracted in Step 1 as the base map. Based on this base map, extract roof discrimination features, and use a binary classification model to identify greenable roofs / non-greenable roofs and output a probability map. Then, overlay preset rules to obtain a greenable roof mask; the roof discrimination features include at least one of roof shape, roof material reflectivity, structural integrity, floor height, and surrounding obstacle shadow duration; the preset rules include setting a minimum continuous greenable area threshold and a roof slope threshold, and cleaning noise through morphological opening / closing operations. Step 202: Based on the greenable roof mask, obtain a building land mask, and according to the building land mask, obtain a pixel set Mfeasible for arranging carbon sink patches, expressed as: in, A collection of street block unit spaces, For building site cover, For road land cover, For other hardened surface masks, A collection of convertible rooftop spaces; Step 203: Conduct connected component analysis and region merging on Mfeasible, remove patches smaller than the minimum effective ecological patch area threshold, and divide the connected components into linear corridor patches according to the width threshold. Step 204: Set an area threshold Sp, an aspect ratio threshold Rl, and an average width threshold Wi. If the area threshold ≥ Sp, it is a planar patch; if the aspect ratio ≥ Rl and the average width ≤ Wi, it is a linear patch; if the area < Sp and it is isolated, it is a点状 patch. At the same time, calculate and record the morphological indexes of each patch, and the morphological indexes include at least one of boundary complexity, aggregation degree, and landscape connectivity. Step 205: Convert the patches into vector data, associate them with the block unit ID, and write attribute information in the attribute table. The attribute information includes at least one of the block unit ID to which it belongs, patch type, area S, boundary complexity LSI, greenable roof ratio, and candidate range of vegetation types.至此,生成各街区单元的碳汇斑块可分布图.

6. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 5, characterized in that... The specific steps of Step 3 are as follows: According to the climate and environment of the所在城市, screen suitable carbon sink types, and further divide the initially screened carbon sink types according to shape and area, specifically: Divide large and medium-sized patches into planar patches, and the large and medium-sized patches include urban forests, community parks, and block green spaces; Divide corridor patches into linear patches, and the corridor patches include urban corridors, roadside greenways, and leisure greenways; Divide small patches into点状 patches, and the small patches include roof gardens and bioretention facilities.

7. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 6, characterized in that... The natural dividing line is rivers and green spaces.

8. The method for optimizing the distribution of carbon sinks in urban areas for carbon trading according to claim 7, characterized in that... The design parameters are patch boundary complexity, vegetation cover type, and vegetation cover area.

9. A method for optimizing the distribution of carbon sinks in urban areas for carbon trading, as described in claim 8, characterized in that... The net carbon sequestration of the carbon sink is expressed as: where m is the boundary complexity coefficient, n is the number of carbon sink patches in the unit, S is the vegetation cover area of each patch, x is the carbon sequestration parameter of different types of vegetation, and y is the carbon sink operation and maintenance and its own carbon emission parameter.

10. A method for optimizing the distribution of carbon sinks in urban areas for carbon trading, as described in claim 9, characterized in that... The correction is specifically: First, perform radiometric calibration and atmospheric correction on GF-1 satellite panchromatic and multispectral images, and then fuse elevation control points to perform geometric correction and orthorectification; The denoising includes histogram equalization, bilateral filtering / non-local means filtering.

Images