Intra-Urban CO₂ Spatiotemporal Patterns and Driving Factors Using Multi-Source Data and AI Methods: A Case Study of Shanghai, China

Abstract

Cities are major sources of anthropogenic carbon dioxide (CO₂) emissions, making the study of intra-urban CO₂ concentration patterns an emerging research priority. However, limited data availability and the complexity of urban environments have impeded detailed spatiotemporal analyses at the city scale. To address these challenges, an analysis supported by multi-source data and GeoAI methods is carried out to examine the spatial distribution, vertical variation, temporal dynamics, and driving factors of CO₂ concentrations in urban areas. We combined OCO-2 satellite-derived XCO₂ data (2014–2024) with ground-based measurements from the Shanghai Tower (August 2024 to March 2025), alongside meteorological and socioeconomic variables. The analysis employed spatial interpolation (inverse distance weighting), nonparametric testing (Mann–Whitney U test), time series decomposition, ordinary least squares (OLS) regression, and machine learning techniques including random forest and SHAP (SHapley Additive exPlanations) analysis. Results reveal that CO₂ concentrations are significantly higher in central urban districts compared to suburban areas, with notable spatial heterogeneity. Elevated levels were detected near ports and ferry routes, with airports and industrial emissions identified as principal contributors. Vertically, CO₂ concentrations decline with increasing altitude but exhibit a peak at mid-level heights. Temporally, a pronounced seasonal pattern was observed, characterized by higher concentrations in winter and lower levels in summer. Both OLS regression and machine learning models highlight proximity to emission sources, wind speed, and temperature as key determinants of spatial CO₂ variability, with these factors collectively explaining 67% of the variance in OLS models. This study demonstrates how multi-source data and advanced methods can capture the spatial, vertical, and seasonal dynamics and driving factors of urban CO₂ concentrations, offering insights for policy, planning, and mitigation.

1. Introduction

With the continuous increase in CO₂ emissions, global warming has become a growing concern [1,2]. Rapid urbanization has exacerbated this issue, posing more severe challenges for cities compared to other areas [3]. Among greenhouse gases, carbon dioxide (CO₂) plays an important role, accounting for over 80% of long-term radiative forcing along with methane [4,5]. This has made carbon neutrality a global priority, with the 2009 Copenhagen Accord establishing “measurable, reportable, and verifiable” standards, emphasizing the importance and uniqueness of CO₂ emissions [2]. Meanwhile, research on CO₂ has become extensive, covering various areas including CO₂ emissions, CO₂ flux, and CO₂ concentrations. Studies on CO₂ emissions indicate that fossil fuel combustion, urbanization, and suburban transportation are key drivers of rising CO₂ levels, while compact urban layouts and vegetation help mitigate emissions and regulate seasonal variations [6,7,8,9,10,11]. Research on CO₂ emissions reveal clear seasonal variations, driven largely by human activities and meteorological conditions [12,13]. Moreover, studies also show that CO₂ concentrations are influenced by local emissions and topography, traffic also has a certain impact on it, exhibiting sharp peaks during traffic rush hours [14]. The studies, which reveal different characteristics of CO₂, provide a basis for reducing carbon emissions, thus making CO₂ a new focal point for research.

However, research on CO₂ concentrations in urban areas, particularly in the city centers of large cities, remains insufficient. Further research in this area is important for addressing this gap. CO₂ emissions and fluxes primarily describe sources and exchange processes in the atmosphere. In contrast, CO₂ concentrations directly indicate the density of CO₂ molecules in the air, accounting for atmospheric diffusion and long-term accumulation. Therefore, focusing on CO₂ concentrations provides a more objective approach for analyzing spatial distribution patterns at the urban scale [14,15]. Extensive studies show that CO₂ concentrations are generally higher in megacities, especially in city centers [5,16,17]. In Essen, Germany, CO₂ concentrations within the urban canopy consistently exceed those in rural areas [18]. Similarly, ground-based observations and handheld measurements in China reveal that urban CO₂ concentrations are typically higher than in suburban areas, with some big cities exhibiting an “urban CO₂ dome” phenomenon [19,20]. These findings consistently indicate that megacities tend to exhibit higher CO₂ concentrations, particularly in urban centers, highlighting the importance of understanding intra-urban CO₂ distribution patterns. Although numerous inter-urban studies exist, research focusing on intra-urban CO₂ concentrations in mega-cities remains limited. Investigating these spatial patterns is therefore important both for informing urban policies aimed at reducing emissions and mitigating climate change, and for addressing the current gap in studies of intra-urban CO₂ dynamics [14,15,19,20].

Research about urban and regional CO₂ concentration encompass multiple dimensions, including temporal variations, spatial distribution, vertical distribution, influencing factors and monitoring technologies, among others. Regarding spatial distribution, urbanization processes and associated anthropogenic activities considerably impact CO₂ patterns, resulting in a typical “CO₂ dome” phenomenon in city centers that is closely linked to emission density, traffic conditions, and industrial activities [1,21,22]. Industrial zones and urban core areas display relatively stable CO₂ levels, while suburban areas experience an upward trend in CO₂ concentration due to increased population density and seasonal human activity fluctuations [23]. In terms of temporal variation, CO₂ concentration exhibits diurnal fluctuations, seasonal variations, and annual trends influenced by meteorological conditions and human activities [22]. The demand for heating in winter and the photosynthetic activity of plants in summer result in a general trend of higher CO₂ levels in winter and lower levels in summer, reflecting a distinct seasonal pattern [24,25]. Furthermore, comprehensive studies of spatiotemporal characteristics reveal a strong coupling between spatial and temporal variations in CO₂ concentration. In certain regions, temporal characteristics of CO₂ levels are influenced by spatial factors, illustrating the combined effects of local climate, topography, and human activities on CO₂ dynamics [23,24,26]. Research on the vertical distribution of CO₂ concentrations is important for developing 3D CO₂ transport models, indicating surface CO₂ flux, estimating the distribution of surface sources and sinks, and validating satellite measurements and ground-based remote sensing instruments [27,28]. However, vertical distribution analysis is more focused on the troposphere and stratosphere, with less attention given to the boundary layer. Urban-scale impacts on CO₂ concentrations are mainly concentrated below the boundary layer, and studies have shown that surface CO₂ concentrations affect high-altitude concentrations only within 300 m, with diurnal variations more pronounced at the surface [29].

Studies also indicate that CO₂ concentrations are influenced by various factors, such as land cover and vegetation, with more green space leading to lower CO₂ levels and impervious surfaces contributing to higher CO₂ levels. Meteorological factors like wind and humidity also regulate CO₂ distribution, especially in non-urban areas [30,31,32]. Future studies need multi-site monitoring and long-term data to improve monitoring technologies and capture urban CO₂ variations [22], which also reflects the importance of CO₂ concentration monitoring technologies in promoting related studies.

CO₂ monitoring technologies have evolved since the 19th century. From early chemical detection methods to the 1950s [33], when Keeling documented the “Keeling Curve” [34]. Since then, CO₂-related issues have received widespread attention [35,36,37]. Advances in technology have led to the development of infrared gas analyzers and NDIR sensors [38,39], the promotion of carbon capture and storage (CCS) and carbon capture and utilization (CCU) technologies [40], and improvements in global climate models (GCMs) and regional climate models (RCMs), enhancing CO₂ concentration predictions [41]. International agreements like the Paris Agreement have further driven CO₂ monitoring and mitigation technologies [42,43]. In the 21st century, satellite remote sensing has become a key tool for global CO₂ monitoring [44,45]. Early satellites, such as ESA’s ENVISAT and NASA’s AQUA and AURA, enabled CO₂ column concentration observations but had low resolution [46,47,48]. The METOP series, equipped with the IASI instrument, improved atmospheric CO₂ data accuracy [49,50,51]. Recent advancements include satellites launched by JAXA (GOSAT), NASA (OCO series), and China (TANSAT), which focus on high-precision CO₂ monitoring, particularly for urban and point source emissions [52,53,54,55]. In addition, the OCO satellites, operating in a sun-synchronous orbit, observe CO₂ in the afternoon with a 16-day revisit cycle, limiting its ability to monitor daily and short-term CO₂ fluctuations [56,57,58,59], but the Snapshot Area Map (SAM) mode of the OCO satellite series has improved the limitations of low resolution and scan cycles for analyzing CO₂ concentrations within individual cities [60,61].

After reviewing studies on CO₂ concentrations and detection methods, several research gaps are identified. There is limited research on CO₂ distribution in large cities at fine spatial scales, due to insufficient ground-based monitoring stations and low accuracy of satellite-derived CO₂ column data. Although these two data types could complement each other, studies integrating both are scarce, highlighting a lack of comprehensive methodology. GeoAI has recently shown promise for integrating multi-source spatial data and modeling complex, non-linear environmental processes at regional scales. However, its application to intra-urban CO₂ concentration analysis remains limited, leaving a gap for approaches that can capture fine-scale spatial variability and the relative importance of influencing factors [62,63,64]. Therefore, this study aims to provide a comprehensive understanding of intra-urban CO₂ concentration dynamics by integrating multi-source datasets to analyze their spatiotemporal patterns, vertical distribution, and influencing factors.

The main research question of this study concerns the spatiotemporal patterns of intra-urban CO₂ concentrations. Using the megacity of Shanghai as a case study area, the research explores the spatial characteristics from four scales of refined grid, district level and urban–rural differences, the vertical distributions from 30 m to 630 m, the temporal patterns of hourly and daily variation in summer and winter, and the influencing factors of the CO₂ concentrations. Based on the findings, the study offers recommendations for policy and planning implementation across five key perspectives. The significance of this study lies in integrating multi-source data for analysis and addressing a novel research question by investigating multidimensional CO₂ concentration patterns at the intra-urban scale.

2. Materials and Methods

To address the research objectives, this study establishes a systematic methodological framework. The analysis begins with data preparation, followed by spatial, temporal, and vertical analyses to characterize distribution patterns, and an assessment of influencing factors to identify key drivers. This integrated approach provides both descriptive and explanatory insights into urban CO₂ dynamics, as outlined in the research flow chart in Figure 1.

2.1. Study Area

This study selects Shanghai as the research area. As a coastal city in eastern China and the core of the Yangtze River Delta region, Shanghai is a major economic center covering 6340.5 square kilometers in China. It experiences a subtropical maritime monsoon climate with distinct seasons, relatively high humidity, and ample sunshine throughout the year. The above climatic conditions, combined with the city’s expansive plains and island terrain, create favorable conditions for the dispersion of carbon emissions. Shanghai is also a major population hub, with a permanent population of 24.87 million in 2024 and an urbanization rate of 89.7% [65]. This high population density contributes to its substantial carbon emissions. According to data from National Bureau of Statistics (2024), although Shanghai occupies only 2% of the total land area of the Yangtze River Delta region, its carbon dioxide emissions reached 131 million tons, accounting for 10.9% of the region’s total emissions [65]. Accordingly, studying the spatiotemporal characteristics of carbon emissions in Shanghai and the differences between urban and suburban areas will help in developing effective emission reduction measures.

2.2. Data Preparation

The measurement took place in nine stations at the different heights in the Shanghai Tower (Figure 2), in the following text, this location will be referred to as Shanghai Center.

Vertical daily average CO₂ concentration data from August 2024 to August 2025 are obtained from the stations at different heights (represented by the daily averages measured at each height). The stations also provided data on the diurnal variation of CO₂ concentrations, expressed as hourly averages throughout the days from August 2024 to August 2025, supporting the analysis of daily variations.

These data cover 8 months, with some dates having missing or abnormal data, which will be addressed in the subsequent analysis. The stations with multiple measurement levels have detection heights at 30 m, 100 m, 180 m, 250 m, 330 m, 410 m, 500 m, 580 m, and the top level at 630 m.

The planetary boundary layer height (PBLH) plays a crucial role in the vertical distribution of CO₂ concentrations. Hourly PBLH data for the year 2024 to 2025 were obtained from the ERA5 reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) via the Copernicus Climate Change Service platform [66] and processed using Python 3.11.9. The hourly PBLH data from August 2024 to August 2025 were averaged across all 365 days for each hour of the day (00:00–23:00) to produce the mean diurnal cycle, representing the typical daily variation over the year. This processing is consistent with that applied to CO₂ concentration data at the SC monitoring station. With a spatial resolution of 0.25° × 0.25°, the PBLH values used correspond to the ERA5 grid cell center nearest to the SC station (31.25° N, 121.50° E), while the SC station itself is located at 31.23° N, 121.48° E.

To address the limitation in the number of stations for studying spatial characteristics, this study downloaded the OCO-2 Level 2 XCO₂ V11r products (OCO2_L2_Standard) for the winter months (December to February) of 2014–2024 from NASA Earthdata [67]. Valid XCO₂ concentration data (column-averaged CO₂ dry-air mole fraction, ppm) were retrieved and standardized, resulting in a total of 8875 samples. The data effectively covered all major regions of Shanghai (Figure 3) and are suitable for spatial feature analysis. XCO₂ represents the average CO₂ content in the atmospheric column, whereas CO₂ surface concentration is measured at ground-based monitoring stations. The correlation between the two is typically greater than 0.6. On land, although differences between satellite XCO₂ and ground-based CO₂ surface concentrations are influenced by local environmental conditions and human activities around the monitoring stations. However, the two datasets overall exhibit high correlation and consistent spatial patterns [68].

This study incorporated multi-source datasets to investigate the factors influencing CO₂ concentrations in Shanghai. Five categories of data were utilized: meteorological variables, land use, urban surface characteristics, socio-economic indicators, and major emission sources. Meteorological data included wind speed from NOAA’s National Centers for Environmental Information, interpolated using the inverse distance weighting method [69]; temperature and precipitation from the National Tibetan Plateau Data Center, downscaled from CRU and WorldClim datasets using the Delta method and validated by 496 observation stations across China [70,71]; and longitude, used to control for potential zonal climatic gradients. All meteorological variables were aggregated to monthly means and resampled to a 1 km spatial resolution. Land use data were derived from Wuhan University’s 30 m resolution annual land cover maps (1990–2023), reclassified into nine categories and aggregated to 1 km grids based on proportional area [72]. Urban surface data included water body coverage and road network density (expressways, primary and secondary roads) extracted from OpenStreetMap, and NDVI from NASA’s MOD13A3 dataset, representing vegetation cover [73]. Socio-economic variables comprised 2023 population density data provided by the Shanghai Statistics Bureau, derived from the seventh population census and spatially allocated to 1 km grids [74], as well as district-level housing prices from JuHui Data [75]. Locations of major emission sources, including power plants and steel facilities, were obtained from Amap’s POI database. Euclidean distances between each 1 km grid and the nearest major source were calculated using GIS. All datasets were standardized and spatially aligned using ArcGIS 10.8.2 to ensure consistency.

2.3. Model Setting

2.3.1. Spatial Distribution Analysis of CO₂ Concentration in Shanghai

An important area of research at the spatial level, which is also another objective of the research, is the spatial distribution of CO₂ concentration. This study analyzes the spatial distribution of column-averaged CO₂ concentrations (XCO₂) in Shanghai using the OCO-2 Level 2 (OCO2_L2_Standard) product, which provides satellite-retrieved data processed by NASA. The data spans a decade during which global carbon dioxide concentrations have been rapidly increasing. In the Yangtze River Delta region, the growth rate over the past five years has reached as high as 2.37 ppm/year [25]. Thus, differences between data from different scanning tracks are primarily due to temporal variations, with more recent years showing significantly higher carbon dioxide concentrations compared to earlier years. To address this, the data in this study are normalized, all annual data are adjusted to 2023 levels using a standard growth rate of 2.37 ppm/year [25]. Subsequently, spatial interpolation is conducted using the Inverse Distance Weighting (IDW) method in ArcGIS 10.8.2, and the accuracy of the interpolation is assessed by calculating the rRMSE (Relative Root Mean Squared Error).

CO₂ emissions increase its concentration, and understanding the location of emission sources and the surrounding environment is significant for analyzing the spatial relationship of CO₂ levels. This study identifies the primary emission sources of different sectors within the jurisdiction of Shanghai based on the greenhouse gas emission inventory from the Intergovernmental Panel on Climate Change (IPCC). These sources mainly include emission sources of CO₂ from fossil fuel combustion in the energy industry, manufacturing industry, and transportation sector. It then qualitatively compares the impact of key emission sources on the surrounding CO₂ concentrations using interpolated CO₂ values. Finally, we analyze the NDVI values around emission sources and compare CO₂ concentrations in key emission areas under different NDVI conditions to understand the impact of vegetation on CO₂ levels.

2.3.2. Analyzing Urban–Rural and Inter-District CO₂ Concentration Differences in Shanghai

In order to improve the efficiency of planning and governance and to provide decision-makers with more targeted strategies when formulating emission reduction policies, the study calculates the mean CO₂ concentration for each district in Shanghai based on the interpolated CO₂ values, and ranks them accordingly. This analysis aims to help decision-makers identify districts with higher CO₂ concentrations, enabling them to pinpoint the underlying causes and develop more focused countermeasures. Additionally, to better quantify and accurately reflect the urban–rural differences, this study also compares the CO₂ concentrations inside and outside the outer ring road of Shanghai using the Mann–Whitney U test [76,77].

This test is employed to assess the differences between two independent samples when the data does not meet the assumption of normality. These studies offer a quantitative understanding and clearer view of the differences in CO₂ concentrations between urban and suburban areas from a data perspective.

2.3.3. Analyzing the Vertical and Temporal Distribution of CO₂ Concentration in Shanghai

The vertical distribution of CO₂ concentrations at different elevations in Shanghai is a focus of this study. By analyzing the variation of CO₂ concentrations with altitude, we aim to gain deeper insights into atmospheric mixing processes and pollution dispersion patterns. We compared the diurnal variations of CO₂ concentrations at multiple altitudes with planetary boundary layer height (PBLH), and visualized the hourly mean values of both variables from August 2024 to March 2025. This approach enables a clear depiction of inter-altitude differences in CO₂ concentrations, their temporal trends, and the influence of boundary layer dynamics on CO₂ distribution.

Hourly CO₂ concentration and PBLH data were collected continuously over 24 h periods. For each time point, values were averaged across the study period to generate diurnal variation curves for each station (representing different altitudes), along with the corresponding boundary layer height curve. Line plots were used to visualize these trends, with the x-axis representing each hour of the day (0–23), and the primary and secondary y-axes displaying CO₂ concentration (ppm) and boundary layer height (m), respectively. Distinct colors were assigned to each station for clarity, and the boundary layer height was highlighted using a black line with enlarged diamond markers for emphasis. This visualization method effectively identifies typical diurnal patterns of CO₂ concentrations at various altitudes and reveals the influence of boundary layer height fluctuations on vertical CO₂ distribution.

A time series analysis is conducted in this study to understand the temporal patterns of carbon dioxide (CO₂) concentrations of the monitoring stations at different heights in Shanghai center. Through data processing and integration, daily average CO₂ data are obtained, mitigating the impact of outliers and missing data. Time series plots are then generated for the daily average CO₂ concentration during August 2024 (summer) and January 2025 (winter) at each monitoring station. The data from these two months represent the most typical seasonal CO₂ concentration distribution patterns of the year, corresponding to the lowest (summer) and highest (winter) periods.

Additionally, to further investigate the stationarity and potential seasonal characteristics of carbon dioxide concentrations, we plot the ACF (Autocorrelation Function) graphs for the mean of the daily mean carbon dioxide concentrations across eight stations at different heights in Shanghai center. By analyzing the tapering or cut-off patterns in the results, we evaluate whether carbon dioxide concentrations exhibit periodicity or seasonality, as well as the relationships between carbon dioxide concentrations at different lags.

2.3.4. Diagnostic Screening of Influencing Variables on CO₂ Concentration

To enhance the accuracy and reliability of regression models evaluating the impact of urban factors on CO₂ concentrations, this study implemented a systematic screening process for 13 candidate variables (

Table 1. Overview of PBLH Data.

Parameter	Data Source	Temporal Coverage	Spatial Resolution	Location
PBLH (Planetary Boundary Layer Height)	ERA5 reanalysis dataset, ECMWF	August 2024–August 2025, hourly	0.25° × 0.25°	31.25° N, 121.50° E (SC station at 31.23° N, 121.48° E)

): population density, second-hand housing price, new housing price, water system, impervious surface, distance to major emission sources, longitude, local road network, highway density, NDVI, average wind speed, average precipitation, and average temperature. The selection integrated both theoretical relevance and empirical diagnostics. On the theoretical side, variables were chosen based on prior studies in urban planning and environmental science to ensure the inclusion of well-established drivers of CO₂ variation. Empirically, statistical diagnostics were applied to identify redundant variables and address multicollinearity, thereby improving explanatory power and model stability.

Collinearity was assessed using the Variance Inflation Factor (VIF) and Tolerance (

Table 2. Multicollinearity Diagnostics (VIF and Tolerance) for Candidate Variables.

Variable	VIF	Tolerance
Population density	1.520	0.658
Second-hand housing price	4.796	0.208
New housing price	4.351	0.230
Water system	1.063	0.941
Impervious surface	2.693	0.371
Distance to emissions	3.603	0.278
Longitude	10.022	0.100
Local road network	1.462	0.684
Highway density	1.078	0.927
NDVI	1.111	0.900
Average wind speed	3.200	0.312
Average precipitation	1.999	0.500
Average temperature	3.802	0.263

Note: Bold values indicate VIF > 10.

). Variables with VIF values exceeding 10 were considered to exhibit severe multicollinearity; accordingly, longitude (VIF = 10.022) was excluded. Pearson correlation analysis further identified a strong positive correlation between second-hand housing price and new housing price (r = 0.854, p < 0.01), suggesting redundancy, leading to the removal of the former. As a result, 11 variables were retained for subsequent regression modeling, encompassing demographic pressure (population density), economic activity (new housing price), ecological conditions (NDVI, water system), land cover characteristics (impervious surface), transportation infrastructure (local road network, highway density), proximity to emission sources (distance to major emission sources), and climatic conditions (average wind speed, average precipitation, and average temperature). This screening process minimized statistical distortion, stabilized coefficient estimation, and enhanced model interpretability, laying a solid foundation for the following OLS and spatial analyses.

2.3.5. Assessing the Importance of CO₂ Influencing Factors Using Random Forest

While Ordinary Least Squares (OLS) regression provides interpretable coefficient estimates, it assumes linearity and spatial stationarity, which may limit its ability to capture non-linear relationships and complex interactions between variables. Moreover, OLS results cannot directly quantify the relative contribution of each explanatory variable to CO₂ concentration variations.

To address these limitations and complement the OLS analysis, this study employed the Random Forest (RF) algorithm, a non-parametric ensemble learning method. RF models the relationship between the selected explanatory variables and CO₂ concentrations by aggregating the predictions of multiple decision trees, which enhances model stability and reduces overfitting. In this study, RF was used to derive the relative importance ranking of explanatory variables, providing a quantitative assessment of their contributions to CO₂ concentration patterns.

2.3.6. Explaining the Contributions of Influencing Factors with SHAP

To further elucidate the mechanisms underlying the relationships identified by RF, SHapley Additive exPlanations (SHAP) were applied. SHAP is an interpretable framework grounded in cooperative game theory, which attributes the prediction outcome to individual variables by calculating their marginal contributions across all possible feature combinations.

In this study, SHAP was employed to quantify both the magnitude and direction of each variable’s effect on CO₂ concentrations. It also revealed potential non-linearities, threshold effects, and interactions among variables. By providing both global and local interpretability, SHAP allowed a more nuanced understanding of how socio-economic, meteorological, and built environment factors jointly influence intra-urban CO₂ concentrations.

3. Results

3.1. Spatial Distributions of CO₂ Concentration in Shanghai

The satellite-retrieved XCO₂ data is generally comparable to ground-based CO₂ concentration measurements, with a high correlation observed, particularly over land [68]. Consequently, interpolation studies based on CO₂ satellite data can reflect the spatial distribution of CO₂ concentrations in urban areas.

Results obtained through the IDW interpolation indicate that the spatial distribution of CO₂ concentrations across Shanghai demonstrates urban–rural disparities (Figure 4). The concentration values range from approximately 418.96 ppm in suburban areas to over 427.30 ppm in urban centers. To verify the accuracy and effectiveness of the interpolation methods, this study employed a cross-validation approach to assess the performance of the interpolation results. Specifically, 80% of the monitoring sites were randomly selected for model training, while the remaining 20% were used to validate the model’s predictive accuracy. Evaluation metrics included the Root Mean Square Error (RMSE) and the relative Root Mean Square Error (rRMSE), where RMSE measures the average deviation between predicted and observed values, and rRMSE standardizes RMSE to eliminate the influence of measurement units, allowing for comparison across different scales. All four interpolation methods yielded highly similar rRMSE values (~0.00358), indicating comparable overall accuracy; consequently, RMSE was used as the primary criterion for model selection. Among the four methods, the second IDW (Inverse Distance Weighting) model produced the lowest RMSE (1.363), and was therefore adopted for the spatial interpolation of CO₂ concentrations in this study.

The city center exhibits the highest CO₂ concentrations, exceeding 425 ppm, particularly concentrated in the central business districts and densely populated areas. In contrast, suburban and green areas surrounding the urban core show relatively lower CO₂ concentrations, typically below 420 ppm. These peripheral areas are characterized by lower levels of urbanization and a more dispersed population. Although the difference between urban and suburban areas may appear modest (approximately 5–10 ppm), such a disparity is considered substantial in CO₂ studies, as background concentrations are generally stable, and even small numerical variations can reflect significant differences in local emissions, energy use, and atmospheric mixing processes [1,56].

The analysis highlights localized CO₂ concentration peaks within the urban zone, aligning with known highly populated areas and transportation hubs, mainly near Hongqiao Airport, the city center, and around the inner and outer ring roads. Conversely, pockets of lower concentrations are observed in suburbs with extensive green spaces, such as parks and forested regions. Such results provide clear evidence of a pronounced urban–rural divide in CO₂ concentrations. However, it is worth noting that the CO₂ concentrations near the mouth of the Huangpu River and its surrounding areas are higher than in nearby suburban regions. Considering the numerous ports and maritime routes in this area, this result appears reasonable. Hence, when addressing CO₂ concentrations, port cities, particularly those with large international ports, should not be overlooked as key regions of focus.

Building on the spatial patterns shown in Figure 4, Figure 5 further compares key emission sources with the NDVI distribution to examine how surrounding environments influence CO₂ levels (Figure 5). Areas with concentrated heavy industry generally correspond to higher CO₂ concentrations; however, this relationship is not absolute, as vegetation coverage can mitigate emissions locally. For instance, the southern coastal region, although densely industrialized, exhibits relatively low CO₂ levels due to extensive vegetation, particularly in the southeastern areas with large parks and green belts. Compared with the impervious surfaces of the city center, these vegetated zones help reduce local concentrations. In addition, CO₂ concentrations tend to increase with higher road density, while ports and shipping routes show elevated levels along the coastline. Emissions from ports, ships, and maritime transport are often overlooked, yet they contribute noticeably to CO₂ accumulation near the coast.

3.2. CO₂ Concentration Urban–Rural and Inter-District Differences in Shanghai

The study finds that the mean CO₂ concentration obtained through interpolation across Shanghai (Figure 3) shows evident spatial differences between the districts (Figure 1 &

Table 3. IDW Interpolated CO₂ Concentrations Across Districts in Shanghai.

Districts	Mean CO2 (ppm)	Std
Putuo (Urban)	425.35	0.295
Jingan (Urban)	425.18	0.565
Baoshan (Urban–Suburban)	425.11	0.559
Hongkou (Urban)	424.95	0.490
Xuhui (Urban)	424.74	1.001
Yangpu (Urban)	424.19	0.298
Huangpu (Urban)	424.05	0.375
Changning (Urban)	423.78	0.583
Chongming (Suburban)	423.66	0.533
Minhang (Urban–Suburban)	423.30	0.776
Pudong (Urban–Suburban)	423.24	0.821
Qingpu (Suburban)	422.98	0.388
Jiading (Suburban)	422.85	0.407
Jinshan (Suburban)	422.72	0.409
Fengxian (Suburban)	422.61	0.587
Songjiang (Suburban)	422.58	0.415

). Putuo and Jing’an districts have the highest CO₂ concentrations, highlighting the characteristic trend of higher CO₂ levels in densely populated urban centers. This pattern is also reflected in other districts close to the city center, such as Hongkou, Yangpu, Changning, and Huangpu. In contrast, suburban districts like Jiading, Qingpu, Songjiang, Jinshan, and Fengxian exhibit lower CO₂ concentrations, reinforcing the general pattern of higher CO₂ levels in urban areas and lower concentrations in suburban regions. These characteristics reflect a degree of urban–rural differences.

Notably, although Pudong New Area contains a core urban area represented by Lujiazui, its overall CO₂ concentration remains relatively low. This is likely due to its large geographic area, much of which is located in suburban zones, as well as its proximity to the ocean. On the other hand, Baoshan District, although also located near the ocean and primarily suburban, shows higher CO₂ concentrations. This is due to the influence of shipping and port operations at the Yangtze River estuary, which contribute significantly to CO₂ emissions. A similar pattern is observed in Chongming District, where CO₂ concentrations remain elevated despite its coastal location and distance from the city center. This suggests that factors such as port activities and localized meteorological conditions play a role in maintaining higher CO₂ levels in these areas.

Additionally, the quantitative analysis of urban–rural differences using the Mann–Whitney U test for areas inside and outside the Outer Ring Road also indicate the difference in CO₂ concentrations between the city center (U = 1,234,567.0000, p = 0.0123) and the suburbs, as p < 0.05. Therefore, we reject the null hypothesis and conclude that there is a statistically significant difference in CO₂ concentrations between the two areas. The results further support the idea that CO₂ concentrations tend to be higher in urban areas with dense populations, highlighting the spatial differences between urban and rural regions.

3.3. Vertical and Temporal Distribution Patterns of CO₂ Concentration in Shanghai

Vertical distribution analysis explores the variation of CO₂ concentrations at different heights and the interaction between them and the planetary boundary layer height (PBLH) in Shanghai, providing insights into atmospheric mixing processes and pollutant dispersion.

The vertical distribution of CO₂ concentration generally decreases with increasing altitude but shows distinct patterns influenced by the planetary boundary layer height (PBLH) (Figure 5). At lower altitudes, from 30 m to 180 m, CO₂ concentrations are relatively high, with values rising from approximately 448 ppm at 30 m to around 467 ppm at 180 m, indicating accumulation near the surface under a shallow PBLH during nighttime. Between 180 m and 580 m, concentrations stabilize with moderate fluctuations, and peaks observed at 180 m, 330 m, and 580 m may result from local meteorological or environmental factors. Above 580 m, CO₂ concentration sharply declines, reaching a minimum near 450 ppm at 630 m, showing reduced influence from surface emissions at higher altitudes. These vertical concentration variations are closely linked to the diurnal cycle of PBLH in Shanghai, where the boundary layer remains low and stable at night (around 300 m), trapping emissions near the surface and causing higher CO₂ concentrations at low altitudes such as 30 m. During the day, PBLH rapidly increases, exceeding 1000 m around 14:00, promoting turbulent mixing that dilutes near-surface CO₂ levels but raises concentrations aloft (e.g., at 250 m and 500 m). This process results in a daytime decrease of about 20 ppm at 30 m, contrasted by a 30–50 ppm increase at higher altitudes, reflecting vertical transport and entrainment of CO₂-rich air from above. Consequently, the PBLH modulates the vertical CO₂ gradient by controlling the extent of atmospheric mixing and the height of accumulated emissions, explaining the observed concentration patterns across different elevations.

The diurnal variations in CO₂ concentrations at different vertical heights exhibit distinct stratification characteristics, with an overall trend of first increasing and then decreasing along the vertical profile (Figure 6). Specifically, in the lower layers below 100 m (30 m and 100 m), CO₂ concentrations follow a “morning peak–afternoon trough” pattern: at 30 m, concentrations reach a peak around 09:00, then gradually decline to the lowest point by 16:00. At 100 m, the peak occurs slightly later, around 09:30–10:00, with a weaker fluctuation compared to 30 m.

In contrast, the middle-to-upper layers above 100 m (180–630 m) show the opposite trend. CO₂ concentrations remain at lower levels during nighttime and increase continuously during the day as the boundary layer rises, reaching a peak between 16:00 and 19:00 before declining. The peak time at higher elevations (e.g., 330–630 m) is slightly delayed with increasing height. Notably, the vertical distribution of CO₂ concentrations follows a “rise-then-fall” pattern, with 100 m serving as a dividing line: below this height (≤100 m), CO₂ concentrations are predominantly influenced by surface activities, resulting in significant diurnal fluctuations, whereas above 180 m, variations are closely tied to the diurnal evolution of the boundary layer. Enhanced daytime mixing facilitates the upward transport of CO₂ from lower layers, while nighttime boundary layer descent suppresses vertical exchange. This stratification highlights the role of boundary layer dynamics in shaping the vertical distribution of CO₂, revealing distinct diurnal variation mechanisms at different height levels.

The analysis of the CO₂ time series (Figure 7a,b) reveals remarkable seasonal fluctuations and short-term variability at the monitoring station of Shanghai center in August 2024 (summer) and January 2025 (winter). Overall, CO₂ concentrations are consistently higher in January than in August, with more pronounced daily fluctuations. In August, CO₂ concentrations remain relatively low (approximately 430–480 ppm) and exhibit stable daily variations with minimal differences across heights. In contrast, January shows significantly higher CO₂ concentrations (approximately 440–550 ppm) with a more pronounced vertical gradient. Regardless of the season, CO₂ concentrations followed a pattern of being highest at mid-level heights and lower at both extremes, with the highest values observed at 330 m, while concentrations at 30 m and 630 m remained lower. Additionally, the daily fluctuation in January is more significant than in August, with distinct peaks on specific days (e.g., 12–14 January and 22–23 January), whereas August exhibits a more synchronized trend across different heights. Compared to August, the vertical differences in CO₂ concentrations are more pronounced in January, indicating a stronger influence of altitude on winter CO₂ levels. These results highlight a clear seasonal pattern, with higher and more unevenly distributed CO₂ concentrations in winter, while summer concentrations are lower and more uniform across different heights.

From the ACF and PACF plots (Figure 8), it can be observed that carbon dioxide concentrations in Shanghai center are non-stationary, indicating that concentrations change over time and exhibit periodic characteristics. This is reflected in the ACF plot by a trailing pattern and in the PACF plot by a cutoff pattern, suggesting the potential presence of seasonal variations in carbon dioxide concentrations. The concentrations show strong autocorrelation over specific periods, with minimal variation in concentrations over consecutive days. The PACF plot reveals a strong correlation between carbon dioxide concentrations and those of the preceding day, while the relationship with concentrations beyond the third lag diminishes rapidly.

3.4. OLS Model Results for CO₂ Concentration and Urban Factors

Ordinary Least Squares (OLS) regression was conducted using the 11 selected variables to quantify their influence on spatial CO₂ distribution (

Table 4. OLS Regression Results for CO₂ Concentration in Shanghai.

	Coef.	Std. Coef.	Std. Error	t	p	95% CI
Constant	423.401	-	0.497	852.732	0.000 **	422.428~424.374
Population density	0.021	0.025	0.008	2.680	0.007 **	0.006~0.036
New housing price	0.000	0.029	0.000	2.876	0.004 **	0.000~0.000
Water system	−0.000	−0.024	0.000	−3.086	0.002 **	−0.000~−0.000
Impervious surface	−0.000	−0.107	0.000	−8.683	0.000 **	−0.000~−0.000
Distance to emissions	−0.000	−0.750	0.000	−59.844	0.000 **	−0.000~−0.000
Local road network	−0.060	−0.053	0.010	−5.792	0.000 **	−0.081~−0.040
Highway density	−0.076	−0.023	0.026	−2.947	0.003 **	−0.126~−0.025
NDVI	−0.217	−0.015	0.115	−1.881	0.060	−0.443~0.009
Average wind speed	−0.168	−0.205	0.007	−23.375	0.000 **	−0.182~−0.154
Average precipitation	0.012	0.017	0.007	1.760	0.078	−0.001~0.025
Average temperature	0.751	0.183	0.041	18.128	0.000 **	0.670~0.832
R2	0.670
Adjusted R2	0.669
F	F(11,5615) = 1036.036, p = 0.000
Durbin-Watson	0.152

Dependent variable: CO₂ concentration, n = 5627; ** p < 0.01.

). The model achieved strong explanatory power (R² = 0.670), indicating that the retained variables collectively account for 67% of the variance in CO₂ concentration. Among the predictors, distance to major emission sources emerged as the most influential factor with a substantial negative coefficient (β = −0.750, p < 0.01), confirming that proximity to emissions strongly elevates local CO₂ levels. Population density (β = 0.025, p < 0.01) and new housing price (β = 0.029, p < 0.01) were positively associated with CO₂ concentration, reflecting the influence of urban demand and affluence. Transportation infrastructure variables, including local road network (β = −0.053, p < 0.01) and highway density (β = −0.023, p < 0.01), also showed significant albeit negative coefficients, suggesting complex interactions potentially driven by spatial allocation rather than traffic volume. Climatic variables such as average wind speed (β = −0.205, p < 0.01) and average temperature (β = 0.183, p < 0.01) exerted strong effects as well, underscoring the meteorological modulation of CO₂ dispersion and accumulation. In contrast, NDVI and average precipitation had weaker or non-significant impacts, indicating limited short-term mitigation through green cover and rainfall.

3.5. Random Forest Results for CO₂ Driver Importance

The RF analysis results (Figure 9) show clear disparities in the relative importance of the influencing variables. The distance to major emission sources emerged as the dominant factor, with a contribution far exceeding that of other variables. Among meteorological factors, average wind speed and average temperature also showed notable importance, followed by socio-economic indicators such as population density and new housing price, albeit with smaller contributions. Other built environment variables, including impervious surface ratio, road density, and water system density, had only marginal effects.

3.6. SHAP Results for CO₂ Driver Mechanisms

The SHAP summary plot (Figure 10) further revealed the mechanisms behind these relationships. The contribution values of the distance to major emission sources were concentrated in the positive range, indicating that greater distance generally increased the predicted CO₂ concentration. Average wind speed exhibited both positive and negative effects, reflecting its role in dynamically modulating pollutant dispersion. Average temperature had weaker, more variable effects. New housing price and population density displayed contributions clustered around zero, implying more localized and indirect influences, while impervious surface ratio, road density, and water system density exerted minimal impact.

These results highlight pronounced disparities in the relative importance of influencing factors, with CO₂ concentration patterns in Shanghai being shaped primarily by emission source distribution and meteorological dynamics, and secondarily by socio-economic and built environment characteristics. In summary, the analyses identifies a small set of dominant drivers, clarifies their mechanisms of action, and underscores the need to prioritize these key factors in targeted emission reduction strategies.

4. Discussion

We first conduct a general discussion of the spatial, vertical, and temporal distribution characteristics of CO₂ concentrations. Then, we provide an detailed discussion on several key features that are relevant to urban environmental development. Finally, we propose a series of recommendations for urban policy-making and urban planning.

The analysis of the spatial interpolation shows the city center of Shanghai has higher CO₂ concentrations and broader data ranges, attributed to dense populations, heavy traffic, and concentrated CO₂ emissions from buildings [17]. This spatial difference of CO₂ concentration also reflects the spatial heterogeneity in emission sources and the mitigating effects of vegetation. The findings underscore the importance of urban planning and green infrastructure in mitigating urban CO₂ emissions and enhancing air quality in metropolitan areas like Shanghai [78]. It is clear that increasing the area of green spaces or the coverage of green vegetation is an important way to achieve the goal of reducing CO₂ concentrations, especially in city centers and industrial areas, where the impact of green plants on CO₂ concentrations within a certain range is decisive [79]. Additionally, multiple outliers in urban stations’ box plots suggest sudden spikes in CO₂ concentrations likely caused by traffic congestion, industrial surges, or adverse meteorological conditions such as low wind speed or temperature inversions, which hinder pollutant dispersion [78]. These findings are crucial for developing targeted air quality management strategies, especially in urban areas where air quality fluctuations are significant [80].

Vertical analysis indicates that CO₂ concentrations are generally higher at mid-level altitudes (particularly between 180 m and 330 m), showing an overall trend of increasing before decreasing. However, this trend is not absolute, as there is considerable fluctuation within this height range. This stratification suggests that limited atmospheric mixing during certain periods can lead to CO₂ accumulation at these heights. For example, during winter, lower temperatures and stable atmospheric conditions often result in temperature inversions that prevent upward diffusion of CO₂ [81]. CO₂ concentration is closely linked to boundary layer height (PBLH). At night, limited turbulent diffusion leads to CO₂ accumulation near the surface. With the onset of daytime heating, the rising PBLH promotes convective mixing and stimulates photosynthetic uptake, reducing near-surface CO₂. In the afternoon, intensified turbulence facilitates upward transport, increasing concentrations at higher altitudes. As nocturnal cooling sets in, the PBLH becomes shallow again, restricting CO₂ within the surface layer [82,83]. Additionally, the vertical distribution of CO₂ also shows seasonal variation, with higher concentrations at mid-level altitudes in winter and more uniform distribution in summer. Such seasonal variations may be driven by changes in atmospheric stability and temperature gradients, which affect vertical air movement and CO₂ distribution [84].

The daily variation of CO₂ concentration is likely related to plant photosynthesis. Research has shown that the decrease in CO₂ concentration from 8 AM to 11 AM aligns with the enhancement of photosynthesis [85]. Meteorological factors such as temperature and humidity may also indirectly influence the temporal characteristics of CO₂ concentration by affecting plant photosynthesis and respiration [31]. Studies on Seoul also indicate that the distribution of CO₂ concentrations exhibits a similar daily variation pattern, with peaks typically occurring at 7:00 AM and troughs at 4:00 PM [86]. The study also observed that the diurnal variation of CO₂ concentrations differs at different heights, with notable distinctions occurring between 100 m and 180 m. This may be related to the nighttime boundary layer decline. During the day, this height range is typically within the boundary layer, but at night, areas above 100 m may extend beyond the boundary layer, leading to distinctly different diurnal variation patterns of CO₂ concentrations. These phenomenon suggests that while CO₂ concentration variations exhibit similarities across cities, they differ at different heights, though further research is needed to verify this. Additionally, temporal analysis found that CO₂ concentration in Shanghai exhibits seasonal variations, with lower levels in summer and higher levels in winter. Considering that Shanghai is located at the southernmost edge of the mid-latitude region in the Northern Hemisphere, and there is no large-scale fossil fuel emissions during the winter, this variation is more likely to be associated with changes in NDVI. From spring to summer, as the photosynthesis of vegetation gradually strengthens and reaches its peak, CO₂ concentration decreases and reaches its lowest point. In contrast, with the increase in plant respiration during autumn and winter, the ability of NDVI to absorb CO₂ also decreases, leading to a peak in CO₂ concentration [24,25].

The OLS regression analysis further elucidates the multifaceted drivers behind the spatial variability of CO₂ concentrations in Shanghai, complementing the observed distribution patterns. The significant positive associations between population density and new housing price with CO₂ concentration highlight the strong influence of urbanization intensity and economic activity on local carbon emissions. These findings align with previous studies indicating that densely populated and economically developed urban cores tend to experience elevated CO₂ levels due to higher energy consumption, vehicular traffic, and building-related emissions [17]. Conversely, the negative coefficients for NDVI and water systems reaffirm the critical role of green and blue infrastructure in mitigating CO₂ accumulation, supporting the spatial heterogeneity discussed earlier. Notably, the strong negative effect of distance to major emission sources underscores the localized impact of industrial and traffic-related emissions on urban air quality [87]. The observed negative associations for transportation infrastructure variables, such as local road network and highway density, may reflect complex spatial dynamics where higher road densities coincide with areas that have been urbanized with mitigation measures or where traffic patterns differ, warranting further investigation. Climatic factors, particularly average wind speed and temperature, also exhibited significant effects, consistent with their roles in influencing pollutant dispersion and atmospheric chemistry. Overall, the regression results reinforce the intertwined influence of socioeconomic, environmental, and meteorological factors on urban CO₂ distribution, providing actionable insights for integrated urban planning and emission reduction strategies.

Based on the analysis of the research results, our study offers several new directions for policy-making. The first is the monitoring and reduction of CO₂ emissions in activity-dense areas of the city center. The city center is characterized by high foot traffic, especially in areas with concentrations of shopping, dining, and office activities, combined with crowded transportation. These areas experience high levels of emissions, which contribute to the elevated CO₂ concentrations in urban centers. While background concentrations cannot be influenced, a reduction in activity levels can improve urban CO₂ concentrations, thus improving air quality [88,89,90]. Measures such as increasing green spaces, promoting energy-efficient buildings, and optimizing public transportation systems could help reduce CO₂ emissions in certain areas [91].

The second measure focuses on traffic-dense areas, particularly elevated roads, where CO₂ management should be enhanced. Studies have shown that transportation is a major source of urban CO₂ emissions, particularly in large cities like London, Paris and Seoul. CO₂ concentrations are closely linked to vehicle emissions, particularly from diesel vehicles. The increase in traffic volume will lead to increased CO₂ and affects environmental quality [89,92]. Considering that Shanghai is also a megacity with heavy traffic, strict emission controls should be implemented in high-traffic corridors, such as establishing low-emission zones and promoting electric vehicles, which will help reduce CO₂ concentrations.

Next, as an important transportation hub, airports often generate high levels of CO₂ emissions [93], which could be a major factor contributing to the increased CO₂ concentrations at airports. This issue is becoming more pronounced due to the rise in air traffic, the expansion and construction of airports, which are all driven by urban development. With the civil aviation industry in China ranking among the largest in the world, CO₂ emissions from the aviation sector are expected to continue growing [94]. Shanghai, with its two large international airports, faces significant challenges in its efforts to reduce emissions locally. Hence, regardless, stricter CO₂ monitoring and reduction measures at airports are crucial. Specifically, promoting sustainable aviation fuels and improving operational efficiency will help reduce CO₂ emissions at airports.

Moreover, the ports of large cities are often overlooked as high CO₂ concentration areas. Due to their international nature and large-scale transportation activities, ports can result in high CO₂ emissions, which considerably affect the distribution of CO₂. Some cities, such as Mumbai and Kaohsiung, have already implemented measures to limit CO₂ emissions from ports [95,96]. Consequently, policies could include tightening emission standards for port facilities, promoting cleaner fuels, and enhancing cooperation with surrounding cities and countries to reduce CO₂ emissions from ports.

Finally, progress in monitoring technology is also essential. Due to some differences between satellite data and surface monitoring stations, satellite data cannot fully reflect surface CO₂ concentrations in urban scale and, therefore, monitoring efforts, particularly in underrepresented areas, should be strengthened by expanding the monitoring network and increasing the number of stations to provide more detailed CO₂ spatial distribution information [56,61,97]. This would help to better understand local sources and sinks of CO₂, allowing for more targeted and effective mitigation strategies.

Based on the above discussion, to comprehensively address CO₂ concentrations, urban planning is an essential tool for policymakers, integrating land use, transportation, and green infrastructure to create a low-carbon, resilient urban environment. A well-structured spatial layout promotes compact, mixed-use development, reduces urban sprawl, and balances residential, commercial, and industrial areas to minimize emissions. Transportation planning prioritizes public transit, sidewalks, and cycling infrastructure, while limiting high-emission vehicles in key areas, thereby reducing CO₂ emissions from private cars, a major source of emissions. Combining efficient road networks with integrated public transportation systems can significantly reduce transportation-related emissions. At the building and community level, urban design can optimize energy use by improving building orientation, natural ventilation, and material selection, thus reducing heating and cooling demands. Green infrastructure, such as urban forests, parks, and rooftop gardens, provides dual benefits by sequestering carbon and mitigating the urban heat island effect, while also improving air quality and enhancing urban livability. Through efficient land use, sustainable transportation, and green urban design, cities can minimize emissions, enhance energy efficiency, and promote ecological balance, ensuring that CO₂ management is not carried out in isolation but is integrated into the broader vision of sustainable urban development.

5. Conclusions

This study provides a comprehensive summary of CO₂ concentration patterns in Shanghai, highlighting urban–rural differences, seasonal variation, and vertical distribution. The analysis reveals that CO₂ levels are generally higher in urban centers, traffic-dense areas, airports, and ports, whereas suburban and green areas exhibit lower concentrations. Vertically, CO₂ initially increases with altitude and stabilizes above approximately 500 m, with moderate fluctuations at intermediate heights. By integrating satellite XCO₂ data with ground-based measurements, the study offers a multi-dimensional perspective on urban CO₂ dynamics and identifies the main contributors to intra-urban variability. These findings provide a clear basis for prioritizing areas for monitoring and mitigation.

The study makes two main contributions. Academically, it addresses the limited research on intra-urban CO₂ concentrations in mega-cities, filling a gap and advancing our understanding of spatial and vertical distribution patterns. Practically, it informs policymakers by highlighting key sources and regions—such as densely populated urban centers, traffic corridors, airports, and ports—that should be targeted for emission reduction. This evidence-based guidance supports urban planning, sustainable transportation, and green infrastructure development to manage CO₂ emissions effectively.

Despite these achievements, several limitations should be acknowledged. The study relies on data from a limited number of ground monitoring stations, which may not fully represent the diversity of urban and suburban environments. Vertical measurements contain gaps at certain heights, potentially increasing uncertainty in characterizing the true vertical distribution of CO₂. Moreover, satellite data precision within urban scales is limited, constraining detailed analysis. Future research should expand monitoring networks and integrate additional data sources to further improve understanding and support effective CO₂ mitigation strategies.