A city heat-carbon coupling analysis method based on multi-source data and machine learning

Abstract

This invention discloses a method for urban heat-carbon coupling analysis based on multi-source data and machine learning, belonging to the field of urban climate regulation and spatial optimization technology. It includes: acquiring and preprocessing multi-source data of the study area to construct a unified grid database; classifying heat-carbon into four quadrants based on surface temperature and anthropogenic carbon emission data; estimating global heat sensitivity and global carbon sensitivity using ordinary least squares regression; estimating local heat sensitivity and local carbon sensitivity using geographically weighted regression; identifying key driving factors using the XGBoost-SHAP model in conjunction with local climate zoning; and analyzing the interaction pathways between urban morphology, green infrastructure, near-surface atmospheric-radiation conditions, and heat-carbon sensitivity using a partial least squares path model. This invention achieves precise identification and causal analysis of the coupling relationship between urban internal thermal environment and carbon emissions, providing a scientific basis for the overall planning of urban thermal risk management and low-carbon regulation.

Description

Technical Field

[0001] This invention belongs to the field of urban climate regulation and spatial optimization technology, specifically involving an urban heat-carbon coupling analysis method based on multi-source data and machine learning. Background Technology

[0002] Against the backdrop of intensifying global climate change, cities, as areas with high concentrations of population, industry, and infrastructure, are not only key regions for energy consumption and greenhouse gas emissions but also the areas with the highest concentration of high-temperature risks. Urban areas simultaneously face increasing risks of heat waves and carbon emission pressures. In this context, high-temperature risks and carbon emission pressures are not independent but rather form a mutually reinforcing and interdependent heat-carbon coupling relationship within cities. Therefore, revealing the spatial patterns and driving mechanisms of heat-carbon coupling has become an important issue in current urban spatial governance.

[0003] Existing research has explored the relationship between the built environment and carbon emissions, and the built environment and thermal environment, in considerable depth. However, three shortcomings remain: First, most studies are conducted separately, with few systematic analyses of the coupling relationship between urban thermal environment and carbon emissions at a unified spatial scale. Second, existing research often focuses on single factors or linear relationships, with insufficient discussion of the nonlinear mechanisms resulting from the combined effects of urban morphology, green infrastructure, and climate background. Third, there is a lack of detailed empirical analysis on whether changes in the thermal environment will further affect carbon emissions, and how this effect manifests within cities.

[0004] Therefore, there is an urgent need for an urban heat-carbon coupling analysis method that can integrate multi-source data, consider spatial heterogeneity, and identify nonlinear driving mechanisms, so as to provide a scientific basis for the coordinated advancement of urban heat risk management and low-carbon regulation. Summary of the Invention

[0005] To address the aforementioned problems in existing technologies, the present invention aims to provide a method for urban thermal-carbon coupling analysis based on multi-source data and machine learning. By constructing a unified grid database, classifying thermal-carbon coupling types, estimating global and local sensitivities, identifying key driving factors, and analyzing action paths, this method achieves precise identification and mechanism analysis of the coupling relationship between urban thermal environment and carbon emissions.

[0006] This invention provides the following technical solution: a method for urban heat-carbon coupling analysis based on multi-source data and machine learning, comprising the following steps: S1. Acquire multi-source spatial data of the target urban area, perform spatial normalization on the acquired data, and construct a multi-source grid database based on a unified spatial grid; S2. Based on surface temperature and anthropogenic carbon emission data from a multi-source grid database, spatial grid cells are classified into thermal-carbon coupling types. S3. Construct indicators to characterize urbanization intensity, and assess the global sensitivity of thermal environment and carbon emissions to changes in urbanization intensity, as well as the spatially heterogeneous local sensitivity. S4. Group spatial units based on local climate zoning type, and identify the driving factors affecting local sensitivity and their contributions under different groups; S5. Analyze the interaction pathways between urban built morphology, green infrastructure, near-surface climate conditions, and local sensitivity using structural equation modeling; S6. Based on spatial coupling patterns, driving factors, and action pathways, formulate differentiated strategies for thermal environment governance and low-carbon regulation.

[0007] Furthermore, in step S1, the multi-source spatial data includes surface temperature data, anthropogenic carbon emission data, urban building morphology data, urban green infrastructure data, and near-surface atmospheric-radiation condition data.

[0008] Furthermore, the urban building morphology data includes average building height, building volume, building density, landscape morphology index, sky visibility factor, road network density, and built-up area coverage; the urban green infrastructure data includes average tree height, normalized difference vegetation index, and average distance to the nearest water body; and the near-surface atmospheric-radiation condition data includes shortwave white sky reflectance, near-surface air temperature, relative humidity, and wind speed.

[0009] Furthermore, the specific process of step S2 is as follows: The median surface temperature and median anthropogenic carbon emissions of all grid cells were calculated, and the study area was divided into four quadrants based on the median surface temperature and median anthropogenic carbon emissions. This resulted in four coupling types of regions: high-thermal-high-carbon, high-thermal-low-carbon, low-thermal-high-carbon, and low-thermal-low-carbon, in order to identify the high-thermal-high-carbon superposition zone and the thermal-carbon misalignment zone.

[0010] Further, in step S3, the urbanization intensity index is the built-up area coverage; the global sensitivity is evaluated using an ordinary least squares regression model; and the local sensitivity is evaluated using a geographically weighted regression model to obtain the local thermal sensitivity coefficient and the local carbon sensitivity coefficient for each grid unit.

[0011] Furthermore, in step S4, a machine learning model is used to fit the relationship between local sensitivity and environmental factors, and an interpretability analysis method is used to quantify the impact contribution of each environmental factor.

[0012] Furthermore, in step S4, the process of constructing local climate zones is as follows: Based on the area proportion of each local climate zone type within each grid cell, the type with the largest proportion is assigned as the type of the grid; if the largest proportion does not exceed the preset threshold, the grid is classified as a mixed type.

[0013] Furthermore, in step S5, the structural equation model is a partial least squares path model. The model uses urban building morphology and urban green infrastructure as exogenous latent variables, near-surface atmospheric-radiation conditions as mediating latent variables, and local heat sensitivity coefficient and local carbon sensitivity coefficient as outcome variables.

[0014] Furthermore, based on the analysis of structural equation modeling, the results include at least one of the following: 1) Urban building form has a positive effect on near-surface atmospheric-radiation conditions, while urban green infrastructure has a negative effect on near-surface atmospheric-radiation conditions; 2) The variation in the local thermal sensitivity coefficient depends on the mediating effect of near-surface atmospheric-radiation conditions; 3) The change in the local carbon sensitivity coefficient is simultaneously affected by the direct effect of urban building morphology and the mediating effect of near-surface atmospheric-radiation conditions.

[0015] Furthermore, the aforementioned thermal environment management and low-carbon control strategies include: Strategies for optimizing building openness and increasing blue-green infrastructure should be developed for high-heat and high-carbon areas. Strategies for optimizing energy structure and regulating the intensity of functional activities should be developed for high-carbon, low-heat regions.

[0016] For low-heat, low-carbon areas, formulate strategies to protect the ecological base and restrict the expansion of built-up areas.

[0017] By employing the above-described technology, the beneficial effects of the present invention compared to the prior art are as follows: 1) This invention constructs a unified multi-source grid database, incorporates thermal environment and carbon emissions into the same analytical framework, and identifies the spatial pattern and differences of thermal-carbon coupling within cities through four-quadrant type classification, making up for the shortcomings of existing studies that conduct separate analyses. 2) This invention comprehensively utilizes ordinary least squares regression, geographically weighted regression, XGBoost-SHAP, and partial least squares path models to achieve systematic analysis from global average relationships to local spatial heterogeneity, from linear to nonlinear, and from single-factor to multi-factor transmission paths, significantly improving the scientific rigor and precision of thermal-carbon coupling mechanism identification. 3) This invention, combined with local climate zoning, reveals the differences in the driving mechanisms of heat-carbon sensitivity under different built-up scenarios, which can provide decision support for the coordinated promotion of urban thermal risk management and low-carbon regulation, and has broad prospects for promotion and application. Attached Figure Description

[0018] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a scatter plot showing the changes of global LST and CE with the FBC gradient in an embodiment of the present invention. Figure 3 This is a bar chart comparing α and β under different LCZ types in an embodiment of the present invention; Figure 4 This is a graph showing the fitting effect verification of the GWR model in an embodiment of the present invention; Figure 5 This is a box plot of local_β at different LST levels in an embodiment of the present invention; Figure 6 This is a box plot showing the statistical differences between local_α and local_β under different LCZ values ​​in an embodiment of the present invention. Figure 7 This is a Spearman correlation heatmap of local_α and environmental determinants in an embodiment of the present invention; Figure 8 This is a Spearman correlation heatmap of local_β and environmental determinants in an embodiment of the present invention; Figure 9 This is a diagram illustrating the SHAP driving mechanism of local_α under different LCZ types in this embodiment of the invention; Figure 10 This is a diagram illustrating the SHAP driving mechanism of local_β under different LCZ types in this embodiment of the invention. Figure 11 In this embodiment of the invention, the PLS-PM path diagram (local_α and local_β parallel model) is shown. Figure 12 This is a bar chart showing the decomposition results of the local_α effect in an embodiment of the present invention; Figure 13 This is a bar chart showing the decomposition results of the local_β effect in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment takes the main urban area of ​​City A as the research object and describes the implementation process of the present invention in detail, but the scope of protection of the present invention is not limited to the following embodiment.

[0020] Example: Figure 1 As shown, a method for urban heat-carbon coupling analysis based on multi-source data and machine learning, taking the heat-carbon coupling analysis of the main urban area of ​​City A as an example, includes the following steps: S1. Data Acquisition and Preprocessing

[0021] The study area is the main urban area of ​​City A, and its scope is determined based on the planning boundary of the main urban area. A total of 5494 grids were divided using a 1km × 1km grid as the basic analysis unit.

[0022] The data obtained includes: Anthropogenic carbon emissions data: from ODIAC (Open-source Data Inventory for Anthropogenic) The 2023 monthly-scale grid product with a resolution of 1km will cover major emission sources such as fossil fuel combustion and cement production.

[0023] Thermal environment data: The average land surface temperature (LST) for 2023 was calculated using the MOD11A2 V6.1 land surface temperature product. After removing cloud pollution and low-quality pixels, the LST was reprojected onto the coordinate system of the study area.

[0024] Urban building morphology data: including building height (from GHS-BUILT-S R2023A, 100m resolution) and building outline (from OpenStreetMap), used to calculate average building height (BH_mean), building density (BD), landscape morphology index (LSI_built), sky visibility factor (SVF), and road network density (RD).

[0025] Urban green infrastructure data includes average tree height (TH_mean, from GEDI 30m product), normalized difference vegetation index (NDVI_mean, from multi-source remote sensing data of China region, 250m resolution), and average distance to the nearest water body (Distance_W, calculated based on land use data).

[0026] Near-surface atmospheric-radiative conditions data: including shortwave white sky reflectance (WSA_mean, MCD43A3 V6.1, 500m), near-surface air temperature (Ta_mean), relative humidity (RH_mean), and mean wind speed (WS_mean) (all from the National Meteorological Data Platform rasterized dataset, 1km resolution).

[0027] All data were uniformly resampled to a 1km×1km grid and standardized (Z-score normalization) to eliminate the impact of variable scale differences on subsequent analysis.

[0028] S2, Thermal-Carbon Four-Quadrant Classification

[0029] Using the median LST and CE as thresholds (median LST = 20.8℃, median CE = 12.3 kt / km²), each grid was divided into four types: high-heat-high-carbon (HH), high-heat-low-carbon (HL), low-heat-high-carbon (LH), and low-heat-low-carbon (LL). Among them, the HH region is mainly concentrated in the core built-up area and its extension zone, the LL region is mainly distributed in the peripheral areas with strong natural background, and the HL and LH regions are patchy, showing a complex pattern of overlapping and misalignment.

[0030] Combining the local climate zoning LCZ types (based on 100m resolution global LCZ products, the type with the largest area proportion within a 1km grid is selected; if the largest proportion is less than 50%, it is classified as mixed LCZ_M), the composition of LCZ in each quadrant is statistically analyzed (see Table 1). It is found that the HH zone is mainly composed of mixed LCZ_M (63.96%) and built-up LCZ_4 (7.91%) and LCZ_5 (7.26%), while the LL zone is composed of the natural-dominated type LCZ_A (74.72%). Table 1. Statistics on the proportion of LCZ in each quadrant

[0031] S3, Global Thermal-Carbon Sensitivity Estimation

[0032] Urbanization intensity is characterized by built-up area coverage (FBC, i.e., the proportion of built-up area to the total grid area), and an ordinary least squares regression model is constructed: LST_i = c1 + α·FBC_i + ε_i; CE_i = c2 + β·FBC_i + ε_i; The regression results show that... Figure 2 As shown, LST and FBC were significantly positively correlated (α=0.032, P<0.001), and CE was also significantly positively correlated with FBC (β=0.78, P<0.001), indicating that the overall thermal environment and carbon emissions in the study area increased with the intensity of urbanization. However, the fitting dispersion of CE and FBC was significantly higher than that of LST, suggesting that carbon emissions are affected by more factors.

[0033] Regression was performed by grouping by LCZ type to obtain α and β within each type, such as... Figure 3 As shown, thermal sensitivity α is significant in most LCZ types, with LCZ_A having the highest α value; carbon sensitivity β shows stronger differentiation among different LCZ types, with LCZ_4 and LCZ_5 having the highest β values, while LCZ_D is not significant.

[0034] S4, Geothermal-Carbon Sensitivity Estimation

[0035] Geographically weighted regression (GWR) was used to estimate the local heat sensitivity coefficient (local_α) and the local carbon sensitivity coefficient (local_β). The model expression is as follows: LST_i = α0(ui,vi) + α(ui,vi)·FBC_i + ε_i; CE_i = β0(ui,vi) + β(ui,vi)·FBC_i + ε_i; An adaptive bandwidth Gaussian kernel function was selected, and the optimal bandwidth was determined through cross-validation. The GWR model showed goodness-of-fit R² values ​​higher than 0.85 for both LST and CE. Figure 4 As shown, this indicates that the model has strong explanatory power.

[0036] The spatial distribution of local_α ranges from approximately -0.229 to 1.150, with most grids showing low to medium thermal sensitivity. However, some areas exhibit higher values, indicating that the impact of enhanced urbanization on the thermal environment varies significantly across different locations. Negative values ​​appear in some areas, suggesting that under specific spatial contexts, enhanced urbanization has not led to significant warming and may even be accompanied by localized cooling. local_β also exhibits significant spatial heterogeneity, ranging from approximately -211.814 to 413.786 Mg / %, with high-value areas showing some clustering, while low-value and negative areas are more dispersed. This indicates that the impact of urbanization on carbon emissions is not uniform across the entire region. In some areas, local_β shows a strong positive response, indicating that urbanization has a stronger promoting effect on carbon emissions; while in other areas, local_β values ​​are low or even negative, reflecting that the emission process is jointly regulated by functional structure and activity type.

[0037] The samples were further grouped according to the LST tertiles, and the distribution of local_β in each group was compared, such as... Figure 5 As shown, the mean local_β value of the high-temperature segment (34.877 Mg / %) was significantly higher than that of the low-temperature segment (14.088 Mg / %), indicating that the thermal environment background amplifies the impact of urbanization on carbon emissions.

[0038] Box plots of local_α and local_β under different LCZ types, such as... Figure 6 As shown, local_α has a high median and large dispersion in the natural and composite classes (LCZA, LCZG); local_β has high dispersion and a longer tail in the built and composite classes (LCZ4, LCZG, LCZM), and the spatial differences in carbon response are more prominent.

[0039] S5, Driver Factor Contribution Identification

[0040] Before identifying the driving mechanism, Spearman rank correlation analysis was first used to explore the monotonic association between various environmental factors and local sensitivity. The results are as follows: Figure 7 and Figure 8 As shown.

[0041] Urban building morphology factors (BH_mean, BD, LSI_built, SVF, RD), green infrastructure factors (TH_mean, NDVI_mean, Distance_W), and near-surface atmospheric-radiation factors (WSA, Ta, RH, WS) were selected as input variables, with local_α and local_β as output variables, respectively, to construct an XGBoost regression model. Grid search was used to optimize hyperparameters (number of trees = 200, maximum depth = 6, learning rate = 0.1). The model was trained on 80% of the data and tested on 20%, with R² values ​​consistently greater than 0.75.

[0042] SHAP values ​​are introduced for interpretability analysis. Figure 9 The contributions of various factors to the SHAP value of local_α under different LCZ types are shown. Global results indicate that SVF, LSI_built, BD_mean, and WSA are the most important explanatory variables, while the blue-green space factors (TH_mean, NDVI_mean, Distance_W) generally have a negative contribution. LCZ stratification shows that SVF, BD, and LSI consistently rank highest in the built-up space (LCZ4, LCZ5, LCZ8); while the hydrothermal background factors increase in the natural space (LCZA, LCZB).

[0043] Figure 10 The SHAP contribution of each factor to local_β is shown. In the global results, RH has the most significant contribution, followed by LSI_built, TH_mean, WS, NDVI_mean, and Ta. LCZ stratification shows that in built-up spaces, building density and humidity jointly affect carbon sensitivity; in mixed-type spaces, humidity and spatial morphological complexity have stronger effects; and in natural-type spaces, the importance of meteorological background factors such as temperature, humidity, and wind speed increases.

[0044] S6, Action Path Resolution

[0045] The structural equation model was constructed using a partial least squares path model (PLS-PM). Latent variables were defined as: urban building morphology (UB, reflected by BD and LSI_built), green infrastructure (UGI, reflected by NDVI_mean and Distance_W), and near-surface atmospheric-radiation conditions (NSA-R, reflected by Ta, RH, WS, and WSA). The outcome variables were local_α and local_β (modeled separately).

[0046] External model tests (Tables 2 and 3) show that the external loadings and external weights of all indicators are significant (P<0.05), indicating high sign stability and valid measurement model. Table 2 External model indexes, external loads, and their stability test results

[0047] Note: External loading reflects the strength of the association between the indicator and the latent variable, and is usually used to test the validity of the measurement. The 95% confidence interval (CI) is calculated based on 2000 bootstrap resampling. The dominant direction refers to the sign (+ or -) that appears most frequently in the bootstrap sample. Sign stability represents the proportion of consistent signs, divided into three categories: high (≥0.95), medium (0.90–0.95), and low (<0.90). Table 3 External weights of external model indicators and their stability test results

[0048] Note: External weights reflect the degree and direction of each indicator's contribution to the latent variable score, and are mainly used to determine whether each indicator has a positive or negative impact on the latent variable.

[0049] like Figure 11 The pathway diagram shows that the total effect of UB on NSA-R is 0.455 (P<0.001), and the total effect of UGI on NSA-R is -0.538 (P<0.001), exhibiting a reinforcing-slow-release inverse effect. For local_α, the direct effect of NSA-R is -0.304 (accounting for 100% of the total effect), indicating that the formation of thermal sensitivity depends entirely on the mediating process of the near-surface environment. For local_β, the direct effect of UB is 0.104 (accounting for 64.2% of the total effect), and the indirect effect is 0.058 (35.8%), showing that mediating and direct effects coexist. Among them, the figure shows... This indicates that the path relationship is highly significant (P<0.001).

[0050] Effect decomposition results (e.g.) Figure 12 , Figure 13 (As shown) This further verifies the path differences between thermal sensitivity and carbon sensitivity: thermal sensitivity is more dependent on the near-surface environment, while carbon sensitivity is simultaneously constrained by morphology and regulated by the near-surface environment.

[0051] This embodiment successfully applied the method of the present invention to conduct a heat-carbon coupling analysis of the main urban area of ​​City A, identifying the spatial pattern of heat-carbon coupling, the spatial heterogeneity of local sensitivity, key driving factors and their mechanisms of action, thus verifying the feasibility and effectiveness of the method. The results show that this method can provide refined scientific support for urban thermal risk management and low-carbon regulation.

[0052] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for urban heat-carbon coupling analysis based on multi-source data and machine learning, characterized in that, Includes the following steps: S1. Acquire multi-source spatial data of the target urban area, perform spatial normalization on the acquired data, and construct a multi-source grid database based on a unified spatial grid; S2. Based on surface temperature and anthropogenic carbon emission data from a multi-source grid database, spatial grid cells are classified into thermal-carbon coupling types. S3. Construct indicators to characterize urbanization intensity, and assess the global sensitivity of thermal environment and carbon emissions to changes in urbanization intensity, as well as the spatially heterogeneous local sensitivity. S4. Group spatial units based on local climate zoning type, and identify the driving factors affecting local sensitivity and their contributions under different groups; S5. Analyze the interaction pathways between urban built morphology, green infrastructure, near-surface climate conditions, and local sensitivity using structural equation modeling; S6. Based on spatial coupling patterns, driving factors, and action pathways, formulate differentiated strategies for thermal environment governance and low-carbon regulation.

2. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, In step S1, the multi-source spatial data includes surface temperature data, anthropogenic carbon emission data, urban building morphology data, urban green infrastructure data, and near-surface atmospheric-radiation condition data.

3. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 2, characterized in that, The urban building morphology data includes average building height, building volume, building density, landscape morphology index, sky visibility factor, road network density, and built-up area coverage; the urban green infrastructure data includes average tree height, normalized difference vegetation index, and average distance to the nearest water body. The near-surface atmospheric-radiation conditions data include shortwave white sky reflectance, near-surface air temperature, relative humidity, and wind speed.

4. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, The specific process of step S2 is as follows: The median surface temperature and median anthropogenic carbon emissions of all grid cells were calculated, and the study area was divided into four quadrants based on the median surface temperature and median anthropogenic carbon emissions. This resulted in four coupling types of regions: high-thermal-high-carbon, high-thermal-low-carbon, low-thermal-high-carbon, and low-thermal-low-carbon, in order to identify the high-thermal-high-carbon superposition zone and the thermal-carbon misalignment zone.

5. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, In step S3, the urbanization intensity index is the built-up area coverage; the global sensitivity is evaluated using an ordinary least squares regression model; and the local sensitivity is evaluated using a geographically weighted regression model to obtain the local thermal sensitivity coefficient and the local carbon sensitivity coefficient for each grid unit.

6. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, In step S4, a machine learning model is used to fit the relationship between local sensitivity and environmental factors, and interpretability analysis is used to quantify the impact contribution of each environmental factor.

7. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, In step S4, the process of constructing local climate zones is as follows: Based on the area proportion of each local climate zone type within each grid cell, the type with the largest proportion is assigned as the type of the grid; if the largest proportion does not exceed the preset threshold, the grid is classified as a mixed type.

8. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, In step S5, the structural equation model is a partial least squares path model. The model uses urban building morphology and urban green infrastructure as exogenous latent variables, near-surface atmospheric-radiation conditions as mediating latent variables, and local heat sensitivity coefficient and local carbon sensitivity coefficient as outcome variables.

9. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, Based on the analysis of structural equation modeling, the results include at least one of the following: 1) Urban building form has a positive effect on near-surface atmospheric-radiation conditions, while urban green infrastructure has a negative effect on near-surface atmospheric-radiation conditions; 2) The variation in the local thermal sensitivity coefficient depends on the mediating effect of near-surface atmospheric-radiation conditions; 3) The change in the local carbon sensitivity coefficient is simultaneously affected by the direct effect of urban building morphology and the mediating effect of near-surface atmospheric-radiation conditions.

10. The urban heat-carbon coupling analysis method based on multi-source data and machine learning according to claim 1, characterized in that, The aforementioned strategies for thermal environment management and low-carbon regulation include: Strategies for optimizing building openness and increasing blue-green infrastructure should be developed for high-heat and high-carbon areas. Strategies for optimizing energy structure and regulating the intensity of functional activities should be developed for high-carbon, low-heat regions. For low-heat, low-carbon areas, formulate strategies to protect the ecological base and restrict the expansion of built-up areas.

Images