Unveiling spatial variations in atmospheric CO₂ sources: a case study of metropolitan area of Naples, Italy

Abstract

In the lower atmosphere, CO₂ emissions impact human health and ecosystems, making data at this level essential for addressing carbon-cycle and public-health questions. The atmospheric concentration of CO₂ is crucial in urban areas due to its connection with air quality, pollution, and climate change, becoming a pivotal parameter for environmental management and public safety. In volcanic zones, geogenic CO₂ profoundly affects the environment, although hydrocarbon combustion is the primary driver of increased atmospheric CO₂ and global warming. Distinguishing geogenic from anthropogenic emissions is challenging, especially through air CO₂ concentration measurements alone. This study presents survey results on the stable isotope composition of carbon and oxygen in CO₂ and airborne CO₂ concentration in Naples’ urban area, including the Campi Flegrei caldera, a widespread hydrothermal/volcanic zone in the metropolitan area. Over the past 50 years, two major volcanic unrests (1969–72 and 1982–84) were monitored using seismic, deformation, and geochemical data. Since 2005, this area has experienced ongoing unrest, involving the pressurization of the underlying hydrothermal system as a causal factor of the current uplift in the Pozzuoli area and the increased CO₂ emissions in the atmosphere. To better understand CO₂ emission dynamics and to quantify its volcanic origin a mobile laboratory was used. Results show that CO₂ levels in Naples’ urban area exceed background atmospheric levels, indicating an anthropogenic origin from fossil fuel combustion. Conversely, in Pozzuoli's urban area, the stable isotope composition reveals a volcanic origin of the airborne CO₂. This study emphasizes the importance of monitoring stable isotopes of atmospheric CO₂, especially in volcanic areas, contributing valuable insights for environmental and public health management.

Introduction

The equilibrium among natural CO₂ emissions, biotic uptake on land, and ocean absorption regulates long-term fluctuations in airborne CO₂, establishing the greenhouse effect essential for the biosphere's existence on Earth. Human activities, particularly fossil fuel combustion, vehicle mobility, house heating, and waste management, disrupt the carbon cycle, leading to an increase in airborne CO₂ levels^1,2,3,4,5. Disruption of this equilibrium worsens the effects of global warming and climate changes.

Global temperature data from Copernicus (https://climate.copernicus.eu/ accessed on 2024, January 10), shows that the mean near-surface temperature in 2023 was ~ 1.4 ± 0.12 °C above the 1850–1900 average. This marked the warmest year in the 174-year observational record, surpassing the joint warmest years of 2016 and 2020. Notably, the last decade (2014–2023) encompasses the nine warmest years on record. Real-time data from specific locations reveals a continued increase in CO₂ levels in 2023, while consolidated concentration datasets of CO₂, methane, and nitrous oxide reached their highest records in 2022.

Several causes contribute to global warming and climate change⁶. Since the eighteenth century the industrialization has led to the gradual abandonment of rural areas and the concentration of people in urbanized zones. Industries, mainly relying on electrical power generated by hydrocarbon combustion, settled in suburban areas contribute significantly to CO₂ emissions^5,7. Urban growth, characterized by skyscrapers and increased vehicle mobility, results in continuous large-scale carbon dioxide release, predominantly concentrated in urban areas, significantly impacting the global atmospheric composition.

Earth degassing, driven by natural sources like soil respiration, volcanic degassing, and photosynthesis, contributes to atmospheric CO₂ concentrations⁸. Regions of active volcanism, responsible for a significant portion of natural gas emissions, release CO₂ of magmatic origin, particularly during eruptions, accounting for ~ 1% of global CO₂ emissions annually^9,10,11. Although this percentage is modest on a global scale, locally, natural emissions may have a more substantial environmental impact, raising hazards for local populations^12,13,14,15. For example, during the recent outgassing crisis at Vulcano, Italy^16,17, gas hazards increased due to either diffuse degassing or crater plume emissions, though human health risk threshold value was not exceeded^18,19,20.

Naples, with around 1 million residents, ranks third in population among Italian cities and is the most densely populated city in Europe. Its strategic location in Mediterranean shipping routes and heavy ship traffic in the harbour make it a potential major source of anthropogenic CO₂. The city is located in a volcanic area with active volcanic and hydrothermal zones, making it an ideal study area to investigate the coexistence of human-related and natural CO₂ emissions.

This study presents the results of a spatial survey on airborne CO₂ in the metropolitan area of Naples. The survey aimed to collect measurements of airborne CO₂ concentration and stable isotopes of CO₂ to differentiate between volcanic and anthropogenic sources, identifying sources that elevate airborne CO₂ concentrations above the background. The study area includes Naples’ downtown and a broad urbanized zone extending from the western edge of Vesuvius volcano to Bacoli and Cuma in the east, and Agnano crater in the north, encompassing the active volcanic/hydrothermal zone of Campi Flegrei (Fig. 1a). The Campi Flegrei area has experienced significant volcanic activity, including supereruptions, the oldest one dating back 40,000 years^21,22. This area exhibits continuous degassing and seismic activity (i.e., Solfatara and Pisciarelli in the municipality of Pozzuoli). Anomalies in CO₂ emissions occur from soils via diffuse degassing and from fumaroles^{23,24,25,26,27}, particularly in the Solfatara area (Zone A in Fig. 1a). The most recent eruption originated from Monte Nuovo in 1538 A.D. Since then, this system has been in a state of persistent degassing and fluctuating seismic activity, leading to ground motion known as bradiseism. The study area has also increased the degassing since 2005 and is currently in unrest^{28,29,30,31,32,33,34,35,36}. Human-related and geologic CO₂ emissions have distinct stable isotopic signatures, allowing differentiation in the air at the district scale through a combination of concentrations and isotopic measurements^{12,13,18,19,20,37,38,39}. The results of the spatial survey enable a comparison between volcanic CO₂ emissions and those of anthropogenic origin in the urbanized area of Naples.

Fig. 1

Study area, the route used during survey and dataset distribution. The survey was conducted in May 2023. (a) The blue line represents the route used during the survey. The selected subsets for Solfatara area (orange zone A), downtown Naples (green zone B), and airport (ice blue zone C) are shown. (b) Probability plot for concentration dataset. Global average value of airborne CO₂ concentration is reported as reference. (blue line indicates 423 ppm vol) for a comparison with the average CO₂ concentration over the target area (50% cumulative probability). (c) Histograms for both the oxygen isotope (δ¹⁸O–CO₂) and carbon isotope (δ¹³C–CO₂) compositions. (d) Three-dimensional view of the study area showing the atmospheric CO₂ concentration measurements at their respective locations. The height of the vertical bars is proportional to the concentration levels. The colour scale and bar height indicate that the highest CO₂ concentration was detected near the port. (e) Placement of the measurements within the study area. The colour scale is identical to that in subplot (d) and indicates the CO₂ concentration measured in the air. The maps (a), (d), and (e) were generated in Qgis 3.34 environment (https://qgis.org/download/).

Results

We developed a measurement program to detect and quantify the spatial variability of CO₂ concentration and its stable isotopes in the near-surface air of the Naples metropolitan area (Fig. 1a). The dataset enables a better determination of the influence of meteorological factors and multiple greenhouse gas sources on the nature of the urban CO₂ dome^40,41,42,43, which is considerably more challenging to identify than its mere presence. For this study, the wind direction was selected as the meteorological factor influencing CO₂ dispersal, while other meteorological factors (e.g., temperature, atmospheric pressure, relative humidity) can be averaged over the survey's completion time (11.4 h of acquisition during daytime over 24 h), as variations at the meteorological station from National Research Counsil (i.e., C.N.R. Long: 432,409; Lat: 4,520,399 UTM) are likely suitable for the entire Naples metropolitan area. Throughout the survey period, the weather remained consistently sunny. Table 1 presents the statistics of both environmental variables and atmospheric measurements.

Table 1 Measurement statistics and features of the survey path.

Figure 1b–c shows statistical distributions of measurements collected during the survey. The dataset collected in the Naples metropolitan area shows airborne CO₂ concentrations higher than 423 ppm vol (Fig. 1b), which is the global reference for airborne CO₂ concentration for May 20, 2023 (https://www.climate.gov/climatedashboard accessed on July 2, 2024). The probability plot⁴⁴ reveals three independent subsets of CO₂ concentrations. The 50% cumulative distribution indicates that the average value for the background CO₂ concentration in the urban area of Naples is 448.1 ± 1.0 ppm vol. The background population comprises more than 98.9% of the cumulative dataset, while the anomalous subset constitutes less than 0.1% of the cumulative dataset, with CO₂ concentrations exceeding 1300 ppm vol (Fig. 1b).

Regarding stable isotopes, the carbon isotope composition of airborne CO₂ (reported in delta notation δ¹³C–CO₂ against the Vienna Pee Dee isotopic ratio-VPDB) shows values more ¹³C-depleted than the theoretical background air (δ¹³C–CO₂ =  − 8‰ vs VPDB). This result indicates that a source of CO₂ forces airborne CO₂ concentration above background values. This gas source has a ¹³C-depleted isotopic signature and establishes an urban CO₂ dome in the Naples metropolitan area. Furthermore, the statistical parameters of the data distribution (skewness =  − 2.27, kurtosis = 15.74) indicate that the dataset has a peak at δ¹³C–CO₂ =  − 10.40‰, which is more ¹³C-depleted than the theoretical atmospheric CO₂ value⁴⁵.

The range of values for δ¹⁸O-CO₂ is wide compared to the spatial and temporal scales of the collected measurements. The oxygen isotope composition of airborne CO₂ depends on both the hydrology of the region and oxygen isotope fractionation in plant leaves during photosynthesis^46,47,48. These factors change over spatial and temporal scales different from those of the measurement acquisition (i.e., ~ 10⁴–10⁵ m and approximately 24 h, respectively). The oxygen isotope values are almost normally distributed (skewness =  − 1.50, kurtosis = 7.3) throughout the study area (Fig. 2b). Gaussian fitting of the oxygen values has a peak at δ¹⁸O–CO₂ =  − 3.16‰ versus VPDB, which is more ¹⁸O-depleted than the expected value for a coastal area of the Mediterranean region^49,50,51,52.

Fig. 2

Spatial variations of the CO₂ measurements collected during survey throughout the target area (cell size 10 m). The maps were generated in Qgis 3.34 environment (https://qgis.org/download/) using Measurement interpolation generated in SAGA GIS environment (https://saga-gis.sourceforge.io/en/index.html). (a) Spatial variation of the airborne CO₂ concentration. (b) Spatial variation of the oxygen isotope composition of the airborne CO₂. (c) Spatial variation of the carbon isotope composition of the airborne CO₂. Traces of the concentration profiles are reported (black lines). See text for description.

The collected dataset was utilized to investigate the spatial variation of airborne CO₂ (Fig. 2). These data allow investigation of whether urban CO₂ sources affect atmospheric chemistry at a district scale or over the urban area (i.e., at the local scale, ~ 10⁴–10⁵ m). The results illustrate a heterogeneous distribution of airborne CO₂ concentration over the Naples metropolitan area, with a concentration gradient from the coast to the inland, likely influenced by local atmospheric circulation. Granieri et al.¹⁴, who conducted detailed micrometeorological studies on atmospheric circulation in the Naples area for gas dispersal simulations, noted a diurnal sea breeze blowing from SW to NE, pushing clean air inland from the seaside during morning hours. A supply of clean air from the sea would dilute CO₂ concentration at relatively low levels. However, measured CO₂ concentrations in the urban area of Naples suggest that atmospheric circulation is insufficient to reduce atmospheric CO₂ concentrations to background levels, at least on days with similar weather conditions to the ones of the day of measurements. Further research should address the issue concerning the critical atmospheric circulation conditions that help to reduce the concentration level of CO₂. The implementation of atmospheric CO₂ monitoring programs in urban areas, particularly when integrated with stable isotope composition analyses, is posited as an effective method for detecting anthropogenic or natural forcings influencing atmospheric CO₂ levels. Elevated atmospheric CO₂ concentrations are frequently correlated with increased levels of other pollutants, suggesting that these monitoring programs can significantly enhance public health management strategies. Additionally, in urbanized regions located within volcanic zones, atmospheric CO₂ monitoring is crucial for mitigating volcanic risks associated with gas emissions (i.e., the gas hazard)¹⁹. Examination of the dataset reveals areas with high airborne CO₂ concentrations, notably near Naples' harbour, where the highest CO₂ concentration was measured (Fig. 1d,e), and the district of Museo square, among others (Zone B in Fig. 1a).

Figure 2 illustrates additional zones with high concentrations of airborne CO₂. The airborne CO₂ concentrations achieve 572 ppm vol in a zone situated in the northeastern sector of the investigated area. While this level does not surpass any established risk threshold for human health⁵³, it exceeds the reference value recorded at NOAA Global Monitoring Laboratory for the investigated time frame (424 ppm by volume) by > 33%. A land use survey in the metropolitan area of Naples reveals the presence of the airport, particularly the runways and aircraft parking areas adjacent to the route used for data collection. Another zone exhibiting elevated concentrations of airborne CO₂ is identified on the western side of downtown Naples, within the municipality of Pozzuoli (i.e., transect B–B′ in Fig. 3b). This area is renowned for its evidence of the underlying volcanic hydrothermal system of Campi Flegrei^26,27,28,29, with airborne CO₂ concentrations reaching 567 ppm vol. The spatial distribution of airborne CO₂ concentrations in this zone appears more heterogeneous compared to other areas, attributable to the presence of several high concentration nuclei near Bagnoli and Baia (Figs. 2a and 3a), eastward and westward of Solfatara, respectively.

Fig. 3

Transects through selected zones of the study area to inspect lateral variations of airborne CO₂ concentration (blue line), δ¹⁸O–CO₂ (red line), and δ¹³C-CO₂ (blue line). (a) A–A′ transect (Bacoli). (b) B–B′ transect (Solfatara). (c) C–C′ transect (Downtown). (d) D–D′ transect (Portici).

The δ¹⁸O–CO₂ has been recognized as a tracer of photosynthesis and the hydrologic cycle's effects on airborne CO₂. These processes play a pivotal role in the fractionation of oxygen in airborne CO₂ at vastly different spatiotemporal scales. While the hydrologic cycle exhibits seasonal effects at the regional scale, notable changes in vegetation (e.g., transition from C3 to C4 or CAM plant dominant types) account for variations in the oxygen isotope composition due to differences in photosynthesis. Since the survey was completed in a few hours, the spatial variations in the oxygen isotope composition resulting from these processes are expected to have negligible effects on the spatial variations of δ¹⁸O–CO₂, which constitutes an ancillary factor for identifying variations in the source of CO₂ at the district scale^{12,13,18,20,51,54}.

The kriging interpolation of the δ¹⁸O-CO₂ dataset reveals a zone with slightly ¹⁸O-depleted airborne CO₂ westward of downtown Naples, where the δ¹⁸O–CO₂ =  ~  − 2‰. Near Baia, where high concentrations of CO₂ were measured (Fig. 2a), the airborne CO₂ exhibits more ¹⁸O-depleted values, reaching δ¹⁸O–CO₂ =  − 5.38‰ through a steep isotopic gradient (e.g., transect A–A′ in Fig. 3a). The δ¹⁸O–CO₂ abruptly increases to approximately − 2‰ northwestwardly along the transect A–A′ (Fig. 3a). Airborne CO₂ shows less ¹⁸O-depleted values near Solfatara. A concentration profile across the Pozzuoli area (Fig. 3b) depicts the least ¹⁸O-depleted CO₂ in the air, having δ¹⁸O–CO₂ =  − 0.06‰ in the vicinity of Solfatara and toward the northeast (Fig. 3b). The δ¹⁸O-CO₂ values decrease to an average of − 2.5‰ northeast of Astroni. Downtown Naples has been identified as an area where airborne CO₂ exhibits more ¹⁸O-depleted values, although zones with δ¹⁸O–CO₂ <  − 6.5‰ are heterogeneously distributed between Pianura and Capodimonte, where CO₂ exhibits more ¹⁸O-depleted CO₂ (i.e., transect C–C′ in Fig. 3c). In this zone, heavily ¹⁸O-depleted CO₂ (δ¹⁸O–CO₂ <  − 16.0‰) was measured in the harbour district.

Additionally, a wide zone elongated NW–SE exhibits δ¹⁸O–CO₂ <  − 5.3‰, extending from the eastern edge of downtown Naples to the west of Torre del Greco, coinciding with a densely urbanized area and a widespread industrialized area (i.e., Area Est-Centro direzionale). Figure 3d illustrates a step gradient of δ¹⁸O–CO₂ that separates the coastal zone where δ¹⁸O–CO₂ =  ~  − 5.06‰ from the inland area where δ¹⁸O–CO₂ =  ~  − 2.85‰. The ∆¹⁸O–CO₂ = 2.21 represents an order of magnitude greater than the accuracy of the oxygen isotope determination (± 0.25‰). In summary, the spatial variations of the measurements show strong fluctuations of δ¹⁸O–CO₂ in different zones. Kriging interpolation of the δ¹⁸O–CO₂ dataset reveals areas with slightly ¹⁸O-depleted airborne CO₂ westward of Naples' downtown, and more ¹⁸O-depleted values eastwards of downtown Naples. Similarly, wide variations in δ¹³C–CO₂ values correspond to spatial variations in the carbon isotopic signature of airborne CO₂ (Fig. six). Dataset statistics indicate that airborne CO₂ is ¹³C-depleted compared to standard air. Cross-sections show trends indicating potential CO₂ sources with ¹³C-depleted or enriched signatures in different areas, with notable variations near Baia and downtown Naples. These results suggest considerable variability in emission sources at the scale of the urbanized zone, and a dominant source of CO₂ with a ¹³C-depleted signature. This expectation arises because the carbon isotope signature of airborne CO₂ can track the source of the gas^55,56.

The cross-section through the urbanized areas of Bacoli (Figs. 2a and 3a) shows an average value of δ¹³C–CO₂ =  − 10.5‰, indicating airborne CO₂ to be more ¹³C-depleted than theoretical air and global reference values recorded by NOAA (https://www.climate.gov/climatedashboard accessed on July 2, 2024). A significant change in the carbon isotope composition of airborne CO₂ is evident at Baia, where a decrease to a value of δ¹³C–CO₂ <  − 14‰ was measured, coinciding with concentration values higher than those measured at Bacoli (Fig. 2c). Low values of δ¹³C-CO₂ indicates that a heavily ¹³C-depleted source of CO₂ is responsible for forcing airborne CO₂ above background levels and is the main contributor to increased CO₂ concentration. The carbon isotope composition increases to less ¹³C-depleted values north of Cuma and achieves δ¹³C–CO₂ =  ~  − 9‰ in the northern zone of the target area. The B–B′ cross-section shows a different trend compared to the A–A′ profile (Fig. 3a,b respectively). Specifically, δ¹³C–CO₂ decreases from approximately − 9 to − 11‰. Continuing along the Solfatara profile (Fig. 3b), a sudden increase in δ¹³C–CO₂ value is observed, reaching values of approximately − 8‰ at the highest concentration values observed along the same profile.

This trend appears to be clearly opposite to that observed in the A–A′ profile, suggesting the presence of potential CO₂ sources with a less ¹³C-depleted signature compared to those forcing airborne CO₂ concentration in adjacent areas. The alternative hypothesis, suggesting that clean air with δ¹³C–CO₂ =  − 8‰ produces the observed values, can be dismissed based on the evidence that ¹³C-enrichment correlates with an increase in CO₂ concentration. This trend contradicts the expectation of a decrease in CO₂ concentration, which would be consistent with the clean air hypothesis. Furthermore, the A–A′, C–C′, and D–D′ profiles demonstrate that CO₂ concentrations exhibit opposite trends in comparison with δ¹³C–CO₂. Specifically, these transects reveal that increases in CO₂ concentration coincide spatially with decreases in δ¹³C–CO₂, indicating that the effective source of CO₂ in these zones is more ¹³C-depleted. In the surrounding area, δ¹³C–CO₂ values average around − 10‰ regardless of CO₂ concentration in the air. These ¹³C-depleted values reduce evidences of spatial ¹³C-enrichment in airborne CO₂. Therefore, the gas source which causes rise in CO₂ concentration above background levels in the area of Solfatara has a carbon isotope composition only slightly ¹³C-depleted compared to the VPDB standard. Accordingly, to the northeast of the Astroni crater, δ¹³C–CO₂ decrease sharply to values ranging between − 10 and − 11‰. Moreover, zones with high airborne CO₂ concentrations near both Bagnoli and Posillipo also show heavily ¹³C-depleted isotopic composition (i.e., δ¹³C–CO₂ =  − 14.69‰ and δ¹³C–CO₂ =  − 13.85‰, respectively).

The C–C′ profile (Fig. 3c) crosses Naples’ downtown (Fig. 2), which is busiest by vehicle during morning hours. The airborne CO₂ has δ¹³C–CO₂ values from − 17.65 to − 8.54‰ with an average δ¹³C–CO₂ =  ~  − 11‰. High CO₂ concentrations along this profile occur at the harbour district (Fig. 3c and Fig. 2a), which coincides with the zone having the most ¹³C-depleted values of airborne CO₂ (Fig. 2c). A comparison with other profiles reveals that a ¹³C-depleted source of CO₂ forces the airborne CO₂ concentration in downtown Naples more efficiently than in peripheral zones to the west (i.e., Bacoli, Baia, and Posillipo). This source is less effective in forcing CO₂ concentration in the zone near Pozzuoli (B–B′ profile), where the source of CO₂ has a less ¹³C-depleted carbon isotope composition. This northwest-oriented profile shows a zone with less ¹³C-depleted values of airborne CO₂ to northwest (Fig. 2c), consistent with a decrease in airborne CO₂ concentrations (Fig. 2a).

Remarkable variations in the stable isotope composition of airborne CO₂ can be identified east of the urban area of Naples (Fig. 2c). In concordance with δ¹⁸O–CO₂, δ¹³C–CO₂ shows remarkable variations along the seaside compared to the inland along the D–D′ profile (Fig. 3d). Airborne CO₂ concentrations fluctuate, superimposed on a decrease from the seaside to the inland. According to this trend, the carbon isotope composition shows an opposite trend from the most ¹³C-depleted values in the coastal zone to the less ¹³C-depleted CO₂ inland, revealing that potential sources of CO₂ with heavily ¹³C-depleted signatures force airborne CO₂ concentration in the coastal zone near Portici and Torre del Greco. These sources are less effective in forcing CO₂ concentration inland, near San Giorgio a Cremano.

Discussion

Measurements of CO₂ concentration, combined with stable isotope compositions of airborne CO₂, provide relevant data for distinguishing between natural and anthropogenic CO₂ emissions in the atmosphere, and potentially tracking the gas dispersal from various sources of greenhouse gases at the urban spatial scale (i.e., 10⁴–10⁵ m). This method overcomes the inherent difficulty of studying CO₂ dispersion caused by its high background level and subtle spatial variations of airborne CO₂ concentration. Indeed, various sources of CO₂ have different isotopic signatures for both carbon and oxygen.

There are several methods for tracking the dispersion of gases emitted from a source into the atmosphere. The methods commonly used to track gas dispersion are based on models that require a priori knowledge of the source, the amount of gas emitted, and the geometry of the dispersion area. Isotopic studies combined with atmospheric chemistry follow a different paradigm. The data collected from field measurements underwent analysis utilizing the Keeling plot method mass balance models for oxygen and carbon isotopes^49,50,57. The Keeling plot method facilitates the determination of the primary CO₂ source at the local level using observational data.

At the same time, the mass balance model for oxygen and carbon isotopes allows an assessment of the influences of the individual CO₂ sources on the local air composition. The mathematical expressions governing this model were developed within the framework of previous studies²⁰ and are expounded concisely upon in the method section dedicated to assessing additional CO₂ in the atmosphere. This method allows for detecting the forcing effects introduced by the gas sources on the composition of the atmosphere. The measurements utilized in the theoretical model results (see Eq. (11) in this study) furnish point-by-point estimates of additional CO₂ concentration (i.e., the C_fs) along the trajectory.

Subsequently, the interpolation of C_fs values employing the Kriging algorithm model facilitates the simulation of CO₂ dispersion. This algorithm generates a predictive layer for δ¹³C–CO₂, δ¹⁸O–CO₂, CO₂ concentration, and C_fs. This method has been successfully applied to detect chemical and isotopic effects on the air in the La Fossa caldera on the island of Vulcano, both during periods of quiescent outgassing and during the recent period of increased volcanic outgassing in 2021²⁰.

The Keeling plot illustrates a correlation between the carbon isotope composition of CO₂ and the inverse of airborne CO₂ concentration. Figure 4 shows the concentration dataset normalized by the global reference for airborne CO₂ concentration (i.e., 423 ppm vol). Each straight line on this plot represents binary mixing between the atmospheric background and an additional CO₂ source. The intercept on the isotopic axis provides the carbon isotopic signatures, facilitating the identification of the CO₂ emission source.

Fig. 4

The correlation between δ¹³C–CO₂ and the inverse of airborne CO₂ concentration (i.e., Keeling plot). Data were normalized against the Global reference values recorded by NOAA (a https://www.climate.gov/climatedashboard accessed on July 2, 2024. (a) Dataset collected over the target area. (b) Urbanized areas of Naples. Green circles distinguish the subset of measurement collected near the airport (zone C in Fig. 1a) from those collected in downtown Naples (zone B in Fig. 1a). (c) Pozzuoli–Solfatara–Agnano area (zone A in Fig. 1a) Yellow circles distinguish the subset of magmatic origin from that of anthropogenic origin in the area (blue circles).

Figure 4a displays several mixing lines between background air and various potential sources of CO₂, including natural (e.g., soil and plant respiration or volcanic degassing) and anthropogenic origins (e.g., combustion of fossil fuels or natural gas and landfill CO₂ emissions), whose isotopic signatures were retrieved from previous studies³⁷. A geometric mean regression is recommended for the analysis of a scattered dataset (i.e., R² < 0.980) in the Keeling plot due to the inherent bias associated with determining the carbon isotopic signature through the utilization of a linear regression model⁵⁸. The line representing the isotopic signature of the forcing source can be derived by applying a standard regression and subsequently dividing by the r-coefficient. This corrective approach aims to approximate the geometric mean regression through the utilization of a standard estimate obtained from a linear regression model.

The dataset collected over the target area reveals a variety of mixing lines, highlighting the inherent complexity of identifying a single CO₂ source. The alignments of δ¹³C–CO₂ in the Keeling plot suggests that fossil fuel combustion is a significant source of greenhouse gases, resulting in airborne CO₂ concentrations ranging from > 600 to ~ 1410 ppm vol (i.e., normalized values are from 0.7 to 0.3, respectively). However, multiple CO₂ sources can influence airborne CO₂ concentrations in the target area, especially at low to intermediate values (i.e., from 423 to 600 ppm vol, corresponding to normalized values ranging 0.7–1). These results support findings that human-related activities, such as urban mobility by vehicles and household heating, predominantly based on the combustion of fossil fuels, contribute significantly to rise the airborne CO₂ concentration. Nonetheless, natural CO₂ emissions, such as those from volcanic outgassing, which is estimated on the synoptic scale to account for approximately 1% of total annual emissions, can locally play a pivotal role in the amount of CO₂ injected into the atmosphere.

A sector in Naples’ downtown (i.e., Zone B in Fig. 1a), distinct from Zone A, which includes the Campi Flegrei volcanic/hydrothermal zone and the western suburbs of Naples (i.e., Bagnoli and Posillipo), can serve as a test site to quantify the specific contribution to increasing airborne CO₂ concentrations caused by human-related emissions. Figure 4b illustrates δ¹³C–CO₂ against CO₂ concentrations, showing good agreement with the mixing line between background air and CO₂ produced by fossil fuel combustion, characterized by a heavily ¹³C-depleted signature (i.e., δ¹³C–CO₂ =  − 29.94‰). Furthermore, data collected in the airport zone (i.e., Zone C in Fig. 1a), where high levels of airborne CO₂ concentrations have been measured, indicate that the CO₂ source affecting both concentration and isotope composition of airborne CO₂ is of anthropogenic origin (i.e., δ¹³C–CO₂ =  − 29.31 ‰).

Figure 4c illustrates the complex distribution of concentration and carbon isotope composition values detected in the study area, predominantly located in the urban area of Pozzuoli, in the western suburbs of Naples. Results of cluster analysis applied to a subset of measurements collected in the Zone A (Fig. 1a) reveal that multiple CO₂ sources play an almost equivalent role in elevating the concentration of airborne CO₂ above background levels. One subset of measurements, with CO₂ concentrations in the range 423–700 ppm vol, exhibits an isotopic signature in good agreement with the mixing line between background air and CO₂ produced by the combustion of fossil fuels (i.e., δ¹³C–CO₂ =  − 32.93‰). Another subset of measurements indicates that δ¹³C–CO₂ of the air increases as CO₂ concentrations rise due to the influence of a less ¹³C-depleted CO₂ source, with δ¹³C–CO₂ ≈ − 1.97‰. Although slightly lower, this value aligns with the carbon isotopic signature of CO₂ emitted from Pisciarelli and Bocca Grande fumaroles.

Those data retrieved from application of laboratory techniques to condensed fumarolic fluids have accuracy ± 0.1‰²⁶. Differences in the range Δ¹³C < 0.4‰ can be neglected because of the accuracy of the measurements with Deltaray (i.e., ± 0.25‰ according to^{12,13,18,37,58}).

Equation (11), included in the method section, facilitates the calculation of additional CO₂ in the air owing to either natural (i.e., volcanic/hydrothermal CO₂) or anthropogenic (i.e., produced by the combustion of fossil fuels) emissions. This calculation is based on input parameters in a theoretical model and measurements of airborne CO₂ concentration, δ¹³C–CO₂, and δ¹⁸O–CO₂ in the field. C_fs provides the concentration of the forcing source of CO₂, exceeding local background levels in the atmosphere. A combination of the positioning of the endogenous sources of CO₂ and results of the Keeling plot helps distinguish the application of the mass balance model to the dispersal of volcanic CO₂ in the zone Solfatara (i.e., Zone A in Fig. 1a) and the dispersal of CO₂ produced by the combustion of fossil fuels downtown Naples (i.e., Zone B in Fig. 1a).

Figure 5 shows dispersions of CO₂ from anthropogenic origin in Naples’ downtown. In particular, the excess CO₂ concentrations in air produced by hydrocarbon combustion, which has a ¹³C-depleted isotope composition compared to standard air (Fig. 4b). For the calculation of the additional amount of CO₂ in the air, an anthropogenic source of CO₂ with the isotopic signature δ¹³C =  − 31.00‰ and δ¹⁸O =  − 16.00‰ has been adopted as the model parameters. Figure 5a shows the fossil fuel-derived CO₂ has a heterogeneous distribution across the target area. A CO₂ dome^40,41,42,43 appears irregular and has numerous lobulations. The dome encloses islands where hydrocarbon combustion forces the CO₂ above the atmospheric background and generates concentration peaks even greater than + 300 ppm above the airborne CO₂ background (Fig. 5b). One such island of high CO₂ concentration is well delineated in the harbour area, which is renowned for being among the Mediterranean's major harbours. In fact, the burning of hydrocarbons sustains the majority of the ship traffic in these areas. Another area with high CO₂ concentrations is located in the western downtown, near one of the most densely populated areas of Naples.

Fig. 5

Dispersal of anthropogenic CO₂ in downtown Naples (cell size 10 m). The maps were generated in Qgis 3.34 environment (https://qgis.org/download/) using Measurement interpolation generated in SAGA GIS environment (https://saga-gis.sourceforge.io/en/index.html). (a) CO₂ concentration map that shows the CO₂ concentration excess above the reference background. The concentration excess value of 83 ppm vol has been set as the threshold for transparency. (b) Vertical profile (black line in subplot a) of the excess CO₂ concentration across downtown Naples. (c) Wind vectors and speed recorded at C.N.R. station. (d) Wind direction frequency during survey.

The results of isotopic investigations prove the anthropogenic origin of atmospheric CO₂. It is reasonable to assume that most of the anthropogenic CO₂ found in downtown Naples is the result of hydrocarbon combustion produced by urban mobility, given that the average air temperature during measurement collection was 22 °C (with an air temperature range of 19–23 °C). Within the one-day measurement acquisition timescale, variations in wind intensity and direction affecting the dispersion of CO₂ cannot be ruled out. This is particularly expected in Zone B, where the wind can influence the dispersion of emitted plumes near Naples' harbour. However, the data on wind direction (Fig. 5c) and speed indicate that during the acquisition time window, the atmospheric circulation brought in SSW air, which is generally less enriched in anthropogenic CO₂. Given the morphology of the study area and the local effects of densely built environments (Fig. 5d), it is reasonable to assume a dilution effect of anthropogenic CO₂ due to the influence of less CO₂—rich air from the sea. Accordingly, the anthropogenic CO₂ concentration along the C–C′ profile (Fig. 5c) shows a notable increase in airborne CO₂ near the harbour and above a pedestrian area, suggesting that proximal sources of greenhouse gas emissions in the nearby areas are responsible for the increase in CO₂ above background levels.

Measurements collected at Pozzuoli (Zone A in Fig. 1a) reveal multiple origins for CO₂ present in the air, namely volcanic and anthropogenic. Although human-related activities cause high concentrations of airborne CO₂, a comparison with downtown can be made concerning the dispersal of geogenic CO₂ in the Pozzuoli area because the Campi Flegrei volcanic/hydrothermal system was in a state of unrest at the time of measurement collection^{28,29,30,31,32,33,34,35,36}. Results of the cluster analysis provide a subset for calculating the amount of geogenic CO₂ that the main degassing zones at Campi Flegrei discharge into the atmosphere. The isotope composition and the airborne CO₂ concentration values of this subset were used in Eq. (11), with the values for the isotopic signatures of both carbon and oxygen for CO₂ emitted at Solfatara and Pisciarelli in May 2023 serving as model parameters (i.e., δ¹³C–CO₂ =  − 1.67‰ vs VPDB and δ¹⁸O–CO₂ =  − 7.85‰ vs VPDB) to obtain point-to-point calculations of volcanic CO₂ dispersal. These values provide insight into the dispersal of volcanic CO₂ from the main degassing vents at Campi Flegrei based on direct measurements and model parameters (Fig. 6a). A comparison with the downtown map (Fig. 5a) shows a more homogeneous dispersal of volcanic CO₂ along an N–S oriented dispersal zone. Furthermore, the volcanic CO₂ concentration is higher than 124 ppm vol above background levels in the area lying between Pozzuoli and Pisciarelli alone. Considering a background air CO₂ concentration of 423 ppm vol and the volcanic input calculated in the present study, this result is in good agreement with dispersal simulations averaged over a whole diurnal cycle obtained using the DISGAS software¹⁴. Measurements of concentration, corroborated by isotopic determinations, reveal volcanic CO₂ dispersion in the area of Bagnoli and eastward towards the urbanized area of Naples. In this area, measurements of airborne CO₂ concentrations alone are not able to track the dispersal of volcanic CO₂ because comparable absolute concentration values are found throughout the urban areas, where the additional CO₂ has anthropogenic origins. According to Granieri et al.¹⁴, inland air circulation prevails during nighttime in the Gulf of Naples, when volcanic CO₂ dispersal occurs towards the sea.

Fig. 6

Dispersal of volcanic CO₂ in Pozzuoli-Solfatara zone (cell size 10 m). The maps were generated in Qgis 3.34 environment (https://qgis.org/download/) using Measurement interpolation generated in SAGA GIS environment (https://saga-gis.sourceforge.io/en/index.html). (a) CO₂ concentration map that shows the CO₂ concentration excess above the reference background. The concentration excess value of 75 ppm vol has been set as the threshold for transparency. (b) Wind vectors and speed recorded at C.N.R. station. (c) Wind direction frequency during survey. (d) Vertical profile (black line in subplot a) of the excess CO₂ concentration across Pozzuoli-Solfatara area.

At the time of the survey, weather datasets reveal that NW winds blew from Solfatara towards the sea, even during the early morning (i.e., by ~ 8:00 UTC), after which sea breeze dominated air circulation from the SE throughout the daytime hours (Figura 6b,c). Arguably, the dispersal of volcanic CO₂ results from a combination of volcanic CO₂ dispersal and the residual layer that develops during nighttime and has not yet been disrupted by diurnal atmospheric turbulence. These results show that spatial surveys for studying airborne CO₂ helps in identifying multiple sources of greenhouse gases at the district scale of urban areas. Furthermore, stable isotope measurements allow an assessment of the impact of either volcanic degassing or anthropogenic emissions on airborne CO₂ concentrations.

The results of this study illustrate that integrating measurements of carbon and oxygen isotopic composition with those of CO₂ concentration aids in elucidating the genesis and development of CO₂ dome in urbanized areas. This represents a step forward in evaluating the impact of specific carbon dioxide sources, whether anthropogenic or natural, on the progression of climate change, as it facilitates the discernment of the underlying causes of urban domes through direct investigations.

The findings of this study also suggest that surveys conducted in urban areas such as Naples can be utilized to identify the primary regions for continuous monitoring of both natural and anthropogenic CO₂ emissions against global warming. Climate change has reached a global scale and threatens the stability of various vital sectors, including infrastructure, the economy, electricity production, international relations, biodiversity, and freshwater and food resources. Climate change affects all regions of the world, and its macroscopic effects manifest through extreme weather events, producing vast damage in cities and rural areas.

The international community is implementing a series of measures to combat ongoing climate change, which significantly impacts economic and social systems globally. For instance, several ambitious plans aim to reduce greenhouse gas emissions by 2050, mainly CO₂. To achieve such ambitious goals, it is crucial to estimate and monitor CO₂ emissions, especially in urban areas where most CO₂ is produced through hydrocarbon combustion. Currently, no monitoring tools are available to detect near-real-time CO₂ emissions for individual countries. Therefore, efforts to monitor CO₂ in the air on a regional scale (synoptic ~ 10⁶ m) with low latency (through the publication of hourly, daily, weekly, and annual data) via networks of stations installed in densely urbanized areas are becoming increasingly relevant. However, monitoring CO₂ in the atmosphere is not straightforward due to the high background concentration (approximately 400 ppm vol), which limits the potential for spatial variability. Consequently, monitoring the concentration alone may not always provide sufficient data for real-time estimation. Various studies demonstrate that integrating isotopic and concentration data provides information on the origin of CO₂ emissions^{12,13,16,18,20,37,38,39,58,59,60}.

The δ¹⁸O–CO₂ largely depends on the CO₂ partitioning among the atmosphere, hydrosphere, lithosphere, and biosphere and can be deciphered through isotopic fractionation processes. Recent studies ^12,18,19,20 show that it is possible to quantify atmospheric CO₂ emissions from natural and anthropogenic sources, isotopically characterized by δ¹³C–CO₂ and δ¹⁸O–CO₂ values, through integrated monitoring of atmospheric CO₂ concentration, isotopic composition, and meteorological data (direct investigations).

Therefore, the implementation of an active monitoring system is urgent and represents a paradigm shift in quantifying atmospheric CO₂ emissions at the scale of individual urbanized areas, compared to the currently applied methods based on statistical data at the national level for countries that are signatories to the United Nations Framework Convention on Climate Change⁶¹.

Methods

Instrument setup

The instrument employed for data acquisition in this study is a Delta Ray–Thermo Fisher Scientific. It measures the concentration of the isotopologues ¹³COO, ¹²COO, and CO¹⁸O based on the adsorption strength of light in the mid-infrared region (~ 4.3 μm) following the Lambert–Beer law. The ¹³C/¹²C and ¹⁸O/¹⁶O ratios are calculated using different concentration ratios of the isotopologues, while the total CO₂ concentration is determined by summing the concentrations of the three CO₂ isotopologues. Stable isotope ratios are expressed in agreement with the VPDB scale using the δ-notation (i.e., δ¹³C–CO₂ and δ¹⁸O–CO₂, respectively) within the CO₂ concentration range of 200–3500 ppm vol.

The Delta Ray instrument is equipped with the QTegra software. A specially designed template includes protocols for recording δ¹³C–CO₂, δ¹⁸O–CO₂, and CO₂ concentration values, along with information on the sample list, acquisition parameters, referencing, evaluation settings, and sample definition. Instrument calibration and referencing against two working standards ensure an accuracy of ± 0.25‰ for isotope determinations and ± 1 ppm vol for CO₂ concentration measurements.

The instrument records each measurement of δ¹³C–CO₂ and δ¹⁸O–CO₂ at a frequency of 1 Hz. Before data acquisition, the instrument conducts isotope ratio referencing on the working standards at a fixed CO₂ concentration (i.e., CO₂ = 400 ppm vol) approximating background airborne CO₂. After purging the unknown air sample for 60 s, the instrument skips the purge and measures the concentration of CO₂ isotopologues in the air. Once the air has purged the gas inlet, the instrument calculates δ¹³C–CO₂ and δ¹⁸O–CO₂, as well as CO₂ concentration.

Measurement strategies

An off-road vehicle housed the instrument, and the equipment for measuring δ¹³C–CO₂, δ¹⁸O–CO₂, and airborne CO₂ concentrations during the studies across the urbanized zone of Naples. The positioning of the vehicle was recorded by a global positioning system device (GARMIN GPSMAP® 64 s), time-synchronized with the Delta Ray's internal clock. In specific urban environments^{12,37,38,39,55,56,62} and, more recently, in volcanic regions^18,20, investigations have been conducted utilizing mobile laboratories to analyze the spatial variability of CO₂.

An inverter (12 V input–output, pure sine wave) was connected to the car's electrical system, supplying power to the instrument (~ 300 W). A stainless-steel capillary (1/16 in.; Swagelok-typeTM, 3 m long) was connected to the instrument's inlet, with the other end attached to the front of the car roof (~ 2.3 m above the ground) to avoid potential contamination from the gasoline engine exhaust. The air passed through a filter (2 μm, 1/16 in, capillary aperture) to prevent contamination from dust on the roads. Considering the volume of the sampling capillary, the instrument's flow rate (approximately 100–110 ml min⁻¹), and the average speed of the mobile laboratory (approximately 3.5 m s⁻¹), the delay between measurements and their corresponding positions is approximately 25 m. This delay is comparable to the GPS positioning.

A route of approximately 170 km (Table 1) was designed in the laboratory to obtain a continuous, non-overlapping path, covering various environments in the wide urbanized area of Naples (Fig. 1a). The route includes Miseno, Bacoli, Agnano, Campi Flegrei caldera, Pozzuoli, Capodimonte, Bagnoli, Posillipo to the east of Naples' downtown, and Portici, Ercolano, Torre del Greco, and San Giorgio a Cremano to the west, respectively. The route was planned to ensure that segments did not overlap, preventing an increase in the statistical weight of some route segments over others. The route was meticulously followed using a routing application (e.g., Google Maps). The survey was completed in thirteen hours at an average speed of 13 km h⁻¹, with the spatial density of measurements corresponding to the metric order (~ 4 m average distance between measurements). The dataset encompasses ~ 41,000 georeferenced measurements for δ¹³C–CO₂, δ¹⁸O–CO₂, and CO₂ concentration, respectively⁶³. This method was already employed for a simultaneous airborne CO₂ spatial survey at Vulcano and revealed the dispersion of volcanic CO₂ through direct measurements^18,20.

Data processing approach

The data acquired from onsite measurements underwent processing utilizing the Keeling plot approach and mass balance models for oxygen and carbon isotopes. The Keeling plot enables the identification of the predominant CO₂ source at the local scale through observational data. The mass balance model for oxygen and carbon isotopes aims to quantify the impact of the CO₂ source on the local air. The algebraic equations for the model were developed as part of a previous study¹⁸ and are detailed in the following paragraph of this paper, addressing the assessment of either volcanic or anthropogenic CO₂ in the air at Naples’ urban area. This methodology integrates measurements of stable CO₂ isotopes in the air with isotopic signatures of both the local CO₂ source, determined through the Keeling plot method^49,50, and CO₂ in the background air. The theoretical outcomes of the model facilitate the partitioning of CO₂ in the air between the local background air and the CO₂ source.

The Keeling plot^49,50, is the method broadly used to identify the isotopic signature of the gas source that increases CO₂ concentrations at the atmospheric background. The Keeling plot method facilitates the examination of the primary origin of atmospheric CO₂ by analyzing the δ¹³C–CO₂ against the reciprocal of CO₂ concentration. This method relies on mass balance principles, wherein a local CO₂ source alters the concentration from the atmospheric baseline. Mathematically, this is expressed by equations:

$$C_{m} = C_{a} + C_{fs}$$

(1)

and

$$\delta^{13} C_{m} \cdot C_{m} = \delta^{13} C_{a} \cdot C_{a} + \delta^{13} C_{fs} \cdot C_{fs}$$

(2)

where C and δ¹³C denote CO₂ concentration and δ¹³C–CO₂, respectively. Subscripts denote measured values (m), atmospheric background (a), and local source (fs). The linear combination of these equations generates a straight line in the δ¹³C versus 1/C plot^{50,51,52,53,54,55,56,57,58}as delineated by equation

$$\delta^{13} C_{m} = \frac{{C_{a} }}{{C_{m} }}\left( {\delta^{13} C_{a} - \delta^{13} C_{fs} } \right) + \delta^{13} C_{fs}$$

(3)

Equation (3) provides insight into the carbon isotope composition of the local CO₂ source under constant background and CO₂ source conditions.

To identify the main CO₂ source downtown Naples, a subset of measurements was selected. This subset encompasses measurements collected in an area of 24.60 km² centred in the Plebiscito square district of Naples (436,676.0 E and 4,520,817.0 N). Another subset, with its centre in the Pozzuoli area (Lat: 428,058.0 E; Long: 4,520,131.0 N, UTM), was selected for comparative purposes with data collected in Naples’ downtown. Specifically focusing on Pozzuoli (Zone A), the assessment focused on volcanic CO₂ as the primary source of CO₂ in the air. The measurements used in the theoretical model results (Eq. (11) reported below in this study) provide the concentration of the isotopically marked CO₂ source (e.g., volcanic or anthropogenic), causing the airborne CO₂ concentration to exceed the background concentration, point by point within the area. In the case of the Solfatara-Pisciarelli degassing area (Zone A in Fig. 1), the circular area is 47.28 km², and the theoretical model provides the concentration of volcanic CO₂ (C_V). Following this calculation, the interpolation of C_V values using the Kriging algorithm generates simulations of CO₂ from the forcing source (volcanic or anthropogenic for Solfatara-Pisciarelli and Naples' downtown, respectively). This algorithm produces the prediction layer for δ¹³C–CO₂, δ¹⁸O–CO₂, CO₂ concentration, and concentration (C_V or C_F, respectively) based on the assumption that each interpolating variable changes linearly with the distance between adjacent measurements. This assumption aligns with the expected homogeneity of spatial variations in atmospheric variables at the local scale⁷. Kriging interpolation is a geostatistical method used to estimate unknown values of each spatial variable based on known measurements of δ¹³C–CO₂, δ¹⁸O–CO₂, and CO₂ concentration at specific measurement points. The spatial correlation of the data is modeled using a Gaussian variogram, a standard variogram model defined by the equation:

$$\gamma \left( h \right) = C\left( {1 - exp\left( { - \left( \frac{h}{a} \right)^{2} } \right)} \right)$$

(4)

where γ(h) is the semivariance at lag distance h, C₀ is the nugget, C is the partial sill, and a is the range. The kriging system of equations is set up using the Gaussian variogram model to determine the weights assigned to each known data point. These weights are calculated to minimize the estimation variance for the unknown points. The Gaussian model ensures smooth interpolation with continuous and differentiable transitions between estimated values, reflecting the assumed autocorrelation structure of the data.

Based on variogram analysis, CO₂ concentration measurements (Supplementary Fig. S1 online) are spatially dependent up to 700 m (i.e., the range), beyond which they become substantially independent. The range for δ¹³C–CO₂, indicating the distance at which spatial correlation between carbon isotope measurements becomes negligible, has also been set to 700 m for kriging interpolation. For δ¹⁸O–CO₂ measurements, the range was determined to be 800 m. The partial sill was calculated as 1780 for CO₂ concentration, 1.62 for δ¹³C–CO₂, and 1.65 for δ¹⁸O–CO₂, indicating the variance attributable to the spatial structure for each variable. Simulations of stable isotope variables, airborne CO₂ concentration, and volcanic CO₂ dispersion were executed using the SAGA GIS software package (https://saga-gis.sourceforge.io/en/index.html).

Quantification of the CO₂ input in the atmosphere

An appropriate mass balance model for airborne CO₂ incorporates both isotopic parameters and concentration. Utilizing literature values for δ¹³C–CO₂ and δ¹⁸O–CO₂ of standard air (e.g., δ¹³C–CO₂ =  − 8‰ and δ¹⁸O–CO₂ =  − 0.1‰⁵⁰) alongside values specific to CO₂ of external sources (e.g., either volcanic/hydrothermal or fossil fuel derived CO₂), an isotopic mass balance model incorporates four unknowns: background air CO₂ concentration, CO₂ concentration in the forcing source of gas, air CO₂ mixing fraction, and volcanic CO₂ mixing fraction.

The model is expressed by Eq. (5), which represents the CO₂ concentration in the air:

$$C_{m} = X_{a} \cdot C_{a} + X_{fs} \cdot C_{fs}$$

(5)

where C represents the CO₂ concentration and X denotes the mixing fraction between forcing source and atmospheric CO₂, with subscripts m, a, and fs referring to measured, background, and local forcing source of CO₂, respectively. This model operates under the assumptions that external source (i.e., volcanic or fossil fuel derived CO₂) significantly elevates CO₂ concentration relative to background levels.

The binary mixing equation to determine the relative weights of CO₂ from volcanic and atmospheric sources is given by Eq. (6):

$$X_{a} + X_{fs} = 1$$

(6)

Similarly, Eqs. (7)

$$\delta^{13} C_{m} \cdot C_{m} = \delta^{13} C_{a} \cdot C_{a} \cdot X_{a} + \delta^{13} C_{fs} \cdot C_{fs} \cdot X_{fs}$$

(7)

and (8)

$$\delta^{18} O_{m} \cdot C_{m} = \delta^{18} O_{a} \cdot C_{a} \cdot X_{a} + \delta^{18} O_{fs} \cdot C_{fs} \cdot X_{fs}$$

(8)

describe the isotopic mass balance models for carbon and oxygen isotopes of CO₂, respectively. The combination of Eq. (6) and (7) provides Eq. (9), which allows for the calculation of X_a

$$X_{a} = \frac{{C_{m} \cdot \delta^{13} C_{m} - C_{fs} \cdot \delta^{13} C_{fs} \cdot X_{fs} }}{{C_{a} \cdot \delta^{13} C_{a} }}$$

(9)

Using Eq. (9) in Eq. (8) yields Eq. (10), enabling the determination of X_FS

$$X_{fs} = \frac{{C_{m} }}{{C_{fs} }}\frac{{\left( {\delta^{18} O_{m} \cdot \delta^{13} C_{a} - \delta^{13} C_{m} \cdot \delta^{18} O_{a} } \right)}}{{\left( {\delta^{18} O_{fs} \cdot \delta^{13} C_{a} - \delta^{13} C_{fs} \cdot \delta^{18} O_{a} } \right)}}$$

(10)

By employing both Eqs. (9) and (10) and rearranging Eq. (5), we derive Eq. (11)

$$C_{fs} = \frac{{C_{a} \cdot C_{m} \cdot \left( {\delta^{18} O_{m} \cdot \delta^{13} C_{a} - \delta^{13} C_{m} \cdot \delta^{18} O_{a} } \right)}}{{\left( {C_{m} - C_{a} } \right) \cdot \left( {\delta^{18} O_{fs} \cdot \delta^{13} C_{a} - \delta^{13} C_{fs} \cdot \delta^{18} O_{a} } \right) + C_{m} \left( {\delta^{18} O_{m} \cdot \delta^{13} C_{a} - \delta^{13} C_{m} \cdot \delta^{18} O_{a} } \right)}}$$

(11)

which provides the concentration of CO₂ produced by the local effective gas source in the air C_fs.

Airborne CO₂ partitioning of between volcanic and human related components

Cluster analysis was conducted to explore the relationships between airborne CO₂ concentrations and carbon isotope composition. Cluster analysis facilitates the classification of observational datasets into distinct classes based on specified similarity criteria. The objective of this analysis is to discern several groups of data that exhibit internal homogeneity (i.e., similarity criteria) while displaying heterogeneity among themselves concerning both CO₂ concentration and stable isotope compositions (i.e., δ¹³C–CO₂ and δ¹⁸O–CO₂ values). Various clustering methods are available for partitioning datasets (e.g., k-means, hierarchical, and two-way clustering), each differing in the requirement of preselecting the number of clusters, statistical properties of the dataset, or computational complexity.

Hierarchical clustering enables the grouping of objects such that those within a group are similar to each other and distinct from objects in other groups. Hierarchical clustering holds an advantage over alternative methods as it obviates the necessity of specifying the number of clusters a priori. The hierarchical structure of clusters can be formed using partitioning algorithms, initially considering all objects as individual clusters. Subsequently, through an iterative process, objects are assigned to different clusters based on principles maximizing the inter-cluster distances. One variant of hierarchical clustering is agglomerative clustering, where each object begins as its own cluster, and pairs of smaller clusters are successively merged until all data is encompassed within a single cluster. Essentially, hierarchical clustering assesses object similarity (i.e., distance) to form new clusters. Cluster merging is predicated on the Euclidean distance metric, reflecting the sum of squares of object coordinates in Euclidean space. Calculation of Euclidean distances leads to the updating of the distance matrix, with the iterative process culminating in the merging of the last two clusters into a final cluster encompassing the entire dataset.

Multiple approaches exist for computing inter-cluster distances and updating the proximity matrix, with some (e.g., single linkage or complete linkage) assessing minimum or maximum distances between objects from different clusters. In the cluster analysis of our dataset, we employed the Ward approach, which evaluates cluster variance rather than directly measuring distances, aiming to minimize variance among clusters. In Ward's method, the distance between two clusters is contingent upon the increase in the sum of squares when the clusters are combined. Ward's method implementation seeks to minimize the sum of squares distances of points from cluster centroids. In contrast to other distance-based methods, Ward's method exhibits less susceptibility to noise and outliers. Hence, in this paper, the Ward method is preferred over alternative methods for clustering.

Conclusions

This study presents findings from a spatial survey conducted in the metropolitan area of Naples, Italy, aimed at examining potential variations in atmospheric CO₂ sources. The urban zone of Naples was chosen due to its diverse CO₂ sources, including those from both geological (e.g., volcanic/hydrothermal emissions) and anthropogenic (e.g., combustion-related) origins. Situated within the extensive volcanic zone hosting Vesuvius, Campi Flegrei, and active volcano on Ischia, Naples provides a compelling location for such investigation owing to its dense urban population compared to other urban areas in the European continent.

Identification of CO₂ sources was facilitated through a combination of stable isotopic analysis and concentration measurements. Stable isotopic composition (i.e., carbon and oxygen isotopic ratios) and airborne CO₂ concentration were measured using a high-precision laser-based analyzer installed in an SUV vehicle. Measurements, recorded at 1 Hz, were synchronized with GPS data to ascertain spatial positioning, achieving a spatial resolution on a metric scale.

Spatial variations in both isotopic composition and concentration were derived from the dataset using the kriging algorithm with Gaussian autocorrelation. Resulting maps delineated three zones characterized by elevated CO₂ concentrations exhibiting distinct stable isotopic signatures. The zone with the highest CO₂ concentration encompassed Naples’ downtown and harbour district, while intermediate concentrations were observed inland across the urban area. Spatial simulations indicated lower CO₂ concentrations along the seaside to the west of downtown, consistent with local morning atmospheric circulation patterns oriented from SW to NE. Additional zones of heightened CO₂ concentrations were identified near the airport, situated northeast of downtown, and in proximity to inhabited areas such as Pozzuoli and Pisciarelli, near Solfatara to the west. These last areas (Pozzuoli and Pisciarelli) exhibit manifestations of a broad hydrothermal/magmatic system beneath the Campi Flegrei, constituting a geological source of airborne CO₂. Anthropogenic CO₂ emissions, primarily from vehicular engine combustion, were found to elevate CO₂ concentrations above background levels in downtown Naples, near the airport, and in the vicinity of Solfatara.

A mixing model incorporating stable isotope composition and airborne CO₂ concentration allowed quantification of CO₂ contributions from different sources. Geochemical modeling based on this approach revealed spatial dispersal patterns of additional CO₂ near Solfatara and downtown Naples, with volcanic CO₂ dispersing northeastward under prevailing morning winds northeast oriented. This volcanic CO₂ extends beyond the hydrothermal zone, supplementing anthropogenic CO₂ emissions from vehicular traffic.

This study underscores the utility of combining isotopic and CO₂ concentration data for discerning the dispersion of both endogenous greenhouse CO₂ and emissions from anthropogenic activities. Particularly relevant in densely populated volcanic/hydrothermal regions, this methodology effectively distinguishes between natural and anthropogenic gas emissions in the atmosphere, overcoming challenges associated with high background levels and subtle spatial variations of the airborne CO₂. Measurements in Naples were collected within a single day, during the diurnal phase of the planetary boundary layer (PBL) evolution, under turbulent conditions and mixing of the atmospheric layer closest to the ground. Consequently, the CO₂ dispersal maps represent average conditions for the urban area of Naples. Establishing monitoring programs for the concentration and isotopic composition of airborne CO₂ in Naples and other cities is crucial for studying the impact of the daily evolution of the PBL on potential variations in airborne CO₂. This is particularly important in areas where geogenic sources (i.e., volcanic or hydrothermal) coexist with anthropogenic CO₂ emissions (e.g., from fossil fuel combustion) resulting from high population density.