First Evidences of Methyl Chloride (CH₃Cl) Transport from the Northern Italy Boundary Layer during Summer 2017

Abstract

Methyl Chloride (CH₃Cl) is a chlorine-containing trace gas in the atmosphere contributing significantly to stratospheric ozone depletion. While the atmospheric CH₃Cl emissions are predominantly caused by natural sources on the global budget, significant uncertainties still remain for the anthropogenic CH₃Cl emission strengths. In summer 2007 an intensive field campaign within the ACTRIS-2 Project was hosted at the Mt. Cimone World Meteorological Organization/Global Atmosphere Watch global station (CMN, 44.17° N, 10.68° E, 2165 m a.s.l.). High-frequency and high precision in situ measurements of atmospheric CH₃Cl revealed significant high-frequency variability superimposed on the seasonally varying regional background levels. The high-frequency CH₃Cl variability was characterized by an evident cycle over 24 h with maxima during the afternoon which points towards a systematic role of thermal vertical transport of air-masses from the regional boundary layer. The temporal correlation analysis with specific tracers of anthropogenic activity (traffic, industry, petrochemical industry) together with bivariate analysis as a function of local wind regime suggested that, even if the role of natural marine emissions appears as predominant, the northern Italy boundary layer could potentially represent a non-negligible source of CH₃Cl during summer. Since industrial production and use of CH₃Cl have not been regulated under the Montreal Protocol (MP) or its successor amendments, continuous monitoring of CH₃Cl outflow from the Po Basin is important to properly assess its anthropogenic emissions.

1. Introduction

Methyl chloride, CH₃Cl, is a major carrier of reactive chlorine into the stratosphere and the understanding of its natural and anthropogenic sources is important for the further development of global policies on ozone-depleting chemicals [1]. CH₃Cl has become increasingly important as a source of chlorine to the stratosphere, as the abundances of anthropogenic halogenated trace gases have been declining under the Montreal Protocol (MP) regulation [2]. Our current understanding is that CH₃Cl is largely natural in origin: the largest natural sources are tropical vegetation (e.g., [3]), biomass burning (e.g., [4]), oceans (e.g., [5,6]), and salt marshes (e.g., [7]), while the major sinks include oxidation by the hydroxyl radical (OH), uptake by soils, degradation in oceans, and photolysis in the stratosphere. While recent studies have contributed to a better understanding of emissions from natural sources (e.g., [8,9,10]), the CH₃Cl global budget still remains unbalanced [2]. Such uncertainties are relevant for policy formulation including CH₃Cl when compared with other anthropogenic ozone-depleting chemicals. Furthermore, emission strengths from anthropogenic sources are still affected by important uncertainties at the global scale (e.g., coal combustion was estimated ranging from of 29 to 295 Gg/year, see [2]). Anthropogenic sources include coal combustion for industrial and incineration activities and indoor biofuel use [2,11] as well as chemical feedstock or solvent releases [12]. Even if the large uncertainty of the source and sink estimations could justify the 20% gap in the global budget of CH₃Cl [13], the recent work by [12] suggested that unaccounted emissions from the chemical industry can help in closing this gap [13]. Interestingly, because the use as chemical feedstock is not supposed to release CH₃Cl to the atmosphere, this has not been taken into account in the national inventory reports (NIRs) submitted to the United Nation Framework Convention for Climate Change (UNFCCC).

On the global scale, extensive regular CH₃Cl measurements are conducted by the Advanced Global Atmospheric Gases Experiment (AGAGE, [14]), and by NOAA-ESRL (National Oceanic and Atmospheric Administration, Earth System Research Laboratory). As reported in the latest Stratospheric Ozone Assessment [13], the 2016 CH₃Cl global mean mole fraction was 553 and 559 ppt (parts per trillion, AGAGE and NOAA measurements, respectively). These values are around 2%–3% higher than the 2012 values reported in the previous Assessment [2], although such changes are thought to be consistent with historical variability. As a consequence of such an increase and of the decrease in regulated anthropogenic chlorinated gases, the fractional contribution to total tropospheric chlorine due to CH₃Cl is currently 16.9% [13].

When focusing on a regional scale, hereinafter referred to as the 20–200 km (meso-beta) scale, modelling studies indicate that the number of measurement sites is still insufficient to constrain the distribution and magnitude of CH₃Cl sources at that specific scales [8,15]. In particular, observations in the outflow of the important source regions should be used to effectively constrain estimates of this source and thus contribute to improving the global CH₃Cl budget.

In this study, for the first time, it is proposed that a highly populated and industrialized area at mid-latitude, such as the Po Basin in northern Italy, may be a non-negligible source area of CH₃Cl. Due to its widespread industrial activities and the high level of household emissions [16], it is of special interest to converge towards a consistent assessment of northern Italy as a CH₃Cl emitter. This assessment would take advantage from the high frequency measurements at the World Meteorological Organization/Global Atmosphere Watch (WMO/GAW) global station of Mt. Cimone (CMN, 2165 m a.s.l.), also part of the global network AGAGE. In summer 2017, the international experimental ACTRIS (Aerosol, Clouds and Trace Gases Research Infrastructure) 2017 campaign, devoted to the investigation of vertical transport of air-masses from northern Italy to the free troposphere, was hosted at the site [17]. Because it overlooks the Po basin, CMN represents a perfect site to investigate the impact of anthropogenic emissions on the regional atmospheric background and on the free troposphere. Here, we discuss the variability of CH₃Cl during June–August 2017 at CMN by analysing the correlations over different time scales (from the 4–5 days typical of the synoptic scale to the hourly scale) with several anthropogenic pollution tracers (e.g., Non-Methane Volatile Compounds (NM-VOCs), carbon monoxide, nitrogen dioxide, sulphur dioxide) and meteorological parameters. A further interesting possibility is related to the availability of carbonyl sulphide (COS) observations at CMN. This atmospheric tracer shares, on the global budget, both natural (mainly wetlands and oceans) and anthropogenic sources [12,18,19].

The aim of this work is to provide a first preliminary understanding about the possible sources of CH₃Cl over northern Italy. The use of several atmospheric trace gases used to trace different anthropogenic sources allowed us to derive first hints about those sectors contributing to the observed variabilities. We primarily focus on interpreting observations at CMN in terms of anthropogenic sources. In particular, we used CH₂Cl₂ as an emission tracer for industrial solvents usage, and HCFC-22 to trace fugitive emissions in hydrofluorocarbons (HFCs) and hydrochlorofluorocarbons (HCFCs) production processes [12]. Indeed, HCFC-22 is supposed to be co-emitted with CH₃Cl and it was used as a robust tracer in a previous CH₃Cl assessment study [2]. Further important hints come from the analysis of 12 anthropogenic NM-VOCs that can trace both combustion and evaporative emissions due to their presence in vehicular emissions or their production in petrochemical processes.

2. Experiments

2.1. Measurement Site

CH₃Cl and other atmospheric tracers are measured since 2002 at Mt. Cimone (2165 m a.s.l.; see Figure 1), which is the highest peak of the northern Apennines ridge and overlooks the Po basin (towards NW–SE) and northern Tuscany (towards S–NW). This particular location, together with the typical meteorological conditions affecting the summer season in this region of the Mediterranean basin (i.e., hot summer with anticyclonic conditions) allows the observation of the upward transport of air-masses from the planetary boundary layer (PBL). Indeed, the vertical transport related to the daytime growth of the mixed layer and the onset of valley and upslope winds during daytime (see e.g., [20]), allow air-masses originating from northern Italy and affected by surface emissions to be caught at this measurement site. Conversely during nighttime, when the measurement site is well above the nocturnal boundary layer, CMN observations can be considered more representative of the background conditions or of aged emissions related both to long-range transport and to residual layers reminiscent of the daytime upward transport of air-masses from the regional boundary layer.

2.2. Measurement of CH₃Cl and Other Atmospheric Trace Gases

At CMN, several programs for the continuous monitoring of trace gases are carried out on a systematic basis. CH₃Cl, CH₂Cl₂, HCFC-22, COS and 12 non-Methane Volatile Organic Compounds are measured by an Agilent 6850–5975 GC–MS (Agilent, Santa Clara, CA, USA) equipped with the UNITY2-Air Server2 auto-sampling/pre-concentration device (Markes International, Llantrisant, UK). The ambient air sample is analysed every two hours and two working standard runs bracket every air sample run, in order to detect and correct for short-term instrumental drift. The working standards for CH₃Cl, CH₂Cl₂ and HCFC-22 are regularly calibrated (weekly) against a tertiary standard prepared with the same procedure at the baseline station of Mace Head (Ireland). They are calibrated on the SIO-05, SIO-07, SIO-014 scales developed and maintained at Scripps Institute of Oceanography (San Diego CA, USA) where primary gravimetric standards are prepared to be used in the AGAGE stations. NM-VOC working standards are referenced against a reference standard by NPL (National Physical Laboratory, Teddington, UK). The NPL gas mixture contains 30 ozone precursor hydrocarbon compounds listed in Annex X of the European Directive 2008/50/EC on Ambient Air Quality and Cleaner Air for Europe, at nominal concentrations of 4 nmol/mol, with actual certified concentration ranging from 3.59 to 4.28 ± 0.07 nmol/mol; absolute concentrations are propagated to a lower concentration standard in air that is used to calibrate the working standard. For CH₃Cl and NM-VOCs annual mean relative standard deviations (RSD) are evaluated from the repeated working standard measurements. Measurement repeatability (evaluated from repeated working standards measurements) for CH₃Cl, CH₂Cl₂, HCFC-22 and COS ranged from 0.3% to 5%. Limits of detection (LODs) are in the ranges of 0.06–0.20 ppt. For the considered NM-VOCs these values ranged from 0.7% to 2.9% and around 1.0 ppt, respectively. More details about CH₃Cl and NM-VOC measurements can be found in [21,22].

During summer 2017, CO was measured by non-dispersive infra-red (NDIR) technique. The system is based on a Tei48C-TL analyser (Thermo Fischer Scientific, Waltham, MA, USA) which uses gas filter correlation technology for determining CO ambient concentration. With the aim of minimizing possible influence of water vapor in the NDIR detection, the ambient air passes through a drying system (Nafion© dryer) and is then injected in the measurement cell (more details can be found in [21]). A correction based on the results of a span check is performed once a day, while every 3 months a multipoint calibration is carried out by using a set of 6 NOAA standards. In this way, the measurements are referred to the WMO CO_X2014A calibration scale. NO₂ observations were continuously carried out by using a chemioluminescence analyser (Tei42i-TL, Thermo Fischer Scientific, Waltham, MA, USA) equipped with a photolytic converter (Blue Light Converter by Air Quality Design). Every 48 h, zero and span checks are carried out for NO by using an external zero air source (Thermo 1160 dry air generator equipped with Purafill© and active charcoal scrubbers, Thermo Fischer Scientific, Waltham, MA, USA) and dilution of certified NO mixture in N₂ (5 ppm ± 3.5% by NPL). To determine the efficiency of the NO₂ photolytic converter, gas phase titration (GPT) is used to titrate about 80% of the NO obtained by dilution. SO₂ observations were based on a 43iTLE (Thermo Fischer Scientific, Waltham, MA, USA), which is based on the UV-fluorescence technology. Daily execution of zero/span checks were carried out by an internal span source (permeation tube with emission rate of 44 ng/min ± 10%) and external zero source (activated charcoal filled cartridge). A dilution system (Environics 4000) was used to perform multi-point calibration (range: 0.5–10 ppb) from a certified standard of 100 ppb SO₂ in N₂ (s/n: 12048642) provided by NPL.

3. Results

As shown by [23], the seasonal variability of CH₃Cl in the northern hemisphere is characterized by a late summer minimum (August–September), which is attributed primarily to the seasonal change in OH [24]. During summer (June–August) 2017, the CH₃Cl mean value was 556.9 ± 2.3 ppt (± 95% confidence level), compared with the concomitant mean value of 540 ppt (R: 0.62, p < 0.01) derived by AGAGE observations at the Jungfraujoch station (3580 m a.s.l., Swiss Alps). The deviation between CMN and Jungfraujoch observations is reduced when only nighttime data are considered for CMN (550.6 ± 2.1 ppt): the higher values observed at CMN can be attributed to the stricter proximity to emission sources.

Figure 2 reports the time series of CH₃Cl (red) for the ACTRIS experiment at CMN, together with CO (black), air-temperature (grey) and atmospheric pressure (blue). For CH₃Cl, variabilities occurring at different time scales are evident: a 24 hour-scale variability is superimposing a multiday component. To filter out this variability at longer time scales, we applied a Loess interpolation [25] to the original time series (thick lines). The smoothed time series showed a certain degree of co-variability between CH₃Cl and CO, a well-known tracer of combustion processes, over time scale of a few days which mimic the behaviour of meteorological parameters. To compare the relative importance of the day-to-day against the diel variability for CH₃Cl, we calculated the standard deviation of the hourly and smoothed datasets [26,27]: sd_hourly (28.6 ppt) and sd_smoothed (23.2 ppt), respectively. Since with the smoothed dataset the diel component of CH₃Cl variability are filtered out, the comparison between the dataset standard deviation allows us to discriminate the hourly variability from other sources of variability. We obtained a ratio sd_hourly/sd_smoothed equal to 0.81, indicating that the diel variability represented about one fifth of the total CH₃Cl variability observed at CMN during summer 2017.

To provide first hints about the source of the diel variability, we compared (Figure 3) the 24 h variability of CH₃Cl with tracers of combustion processes (i.e., CO, NO₂ and SO₂). In particular, CO is a valuable tracer of direct emissions from fossil fuels and biomass combustion, SO₂ is primarily emitted by fossil fuel combustion in industrial processes and energy production, while anthropogenic emissions of NO₂ encompassed high temperature combustion of fossil fuels in electricity generation and petroleum based motors and by lower temperature combustion of biomass in wildfires, deforestation and domestic burning [28]. With the purpose of better characterising the 24 h evolution of CH₃Cl, the variability of O₃/NO_x was also considered (Figure 4); according to [29,30] this ratio can be considered as a proxy to qualitatively evaluate the proximity to major emission sources and photochemical processing. The O₃/NO_x variability already provided useful information to disentangle the role of “fresh” versus “aged” anthropogenic pollution in affecting summer atmospheric composition variability at CMN [20]. As shown in Figure 3 and Figure 4, the afternoon increase in CH₃Cl values was concomitant with an increase in combustion tracers (SO₂ and CO) and a decrease in the O₃/NO_x ratio, which supports the advection of PBL air-masses towards the measurement site due to the thermal transport.

3.1. Correlation Analysis of CH₃Cl with Other Tracers: Smoothed Datasets

To identify possible contributions driving the variability of CH₃Cl over a low frequency scale, typical of the synoptic time scales, we analysed the multiple correlations between the smoothed datasets of CH₃Cl and other anthropogenic trace gases including COS, CH₂Cl₂, HCFC-22, CO, SO₂ and those NM-VOCs (ethyne, propane, benzene) which, according to [22], have a typical summer atmospheric lifetime exceeding 7 days and can thus be used to trace long-range transport (see Figure 5). These trace gases have been selected based on their relatively long atmospheric lifetimes, with the aim of minimising possible interference from local sources characterised by the presence of highly reactive NM-VOCs. NO₂ was not considered to avoid introducing inconsistencies related to its non-linear participation in the photochemical reactions. According to [12], COS was included as tracer for biomass/biofuel/coal combustions (included industrial aluminium production), CH₂Cl₂ is suggested to be an emission tracer for industrial solvent usage, while HCFC-22 is used to trace possible co-located industrial emissions. Indeed, HCFC-22 is used as feedstock for Poly-Tetra-Fluoro-Ethylene (PTFE), for which production plants are located in the Po Basin. However, it should be taken into account that at the global scale COS is also affected by significant natural emission [19]. Besides analysing the cross linear correlation between the considered trace gases, we also carried out a hierarchical cluster analysis with the purpose of grouping together the different atmospheric trace gases which are related to each other. Indeed, the application of the cluster analysis would provide a rough identification of those compounds sharing the same source processes [31]. To this aim, we used the Pearson linear correlation as a metric of the distance among the single elements (i.e., the time series of each single trace gas). With the aim of assessing the impact of different agglomeration methods in the clustering results, three different agglomeration methods have been used: “single linkage”, “complete linkage”, and “Ward”. In the following, the clustering results for the “complete” linkage are shown. Similarities and differences between the clustering runs with different agglomeration methods are discussed in Section 4.

The correlation profile for the smoothed datasets showed that CH₃Cl had the highest positive correlation (R = 0.70, p < 0.01) with COS. A relatively high correlation was also found with CO (R = 0.48, p < 0.01). This would support a possible role of emissions related to or co-located with combustion processes. Taking into account the relatively long tropospheric lifetime of CO (~3 months, see [28]) it can be argued that this would trace aged combustion emissions. A relatively low positive correlation (R = 0.37, p < 0.01) with CH₂Cl₂ would suggest a secondary role of evaporative or fugitive emissions related to industrial production processes. Further hints for attributing the variability of CH₃Cl on a synoptic scale can be obtained by the hierarchical cluster analysis reported by the dendrogram in Figure 6, which provides additional information by visualising how groups of variables are related to one another.

From the cluster aggregation, it is clear that CH₃Cl and COS share a high degree of similarity in their variability and behave differently than the other trace gases. Indeed, other compounds sharing variability are ethyne-benzene and propane-CO (tracing emissions from vehicular emission exhausts). These two clusters (not related to CH₃Cl) are also grouped with SO₂, further supporting the role of combustion emissions (fossil fuels). CH₂Cl₂ and HCFC-22, combined together, clearly trace emissions from fugitive sources related to PTFE production process (HCFC-22) and industrial solvents usage (CH₂Cl₂). According to [7], anthropogenic sources of COS appear to be mostly concentrated in Asia and related to Rayon production, but important emissions are still associated to aluminium production, coal use and biomass burning (biofuels, open burning, and agriculture fires). [19] suggested that, even if anthropogenic (+400 ± 140 Gg S yr⁻¹) and biomass burning (+116 ± 52 Gg S yr⁻¹) emissions are dominant on the global scale, natural emissions (ocean: +265 ± 210 Gg S yr⁻¹; wetlands: −150 to + 290) are significant. In our case, the strong similarity of CH₃Cl with COS and the high cluster distance with unambiguous anthropogenic tracers (i.e., CH₂Cl₂, and HCFC-22) and CO, suggest that the synoptic-scale variability of CH₃Cl can be primarily related to natural emissions, most likely related to oceans.

3.2. Correlation Analysis of CH₃Cl with Other Tracers: Residual Analysis

To obtain information about possible local anthropogenic sources of CH₃Cl at CMN, we analysed the time series of trace gas residuals, i.e., the difference between the punctual measurements with the smoothed dataset obtained by the application of the Loess regression. This analysis is intended to provide complementary information with respect to Section 3.1, where the variability on longer time scales is considered. For this purpose, two different temporal windows of data have been considered: residuals for daytime (12–18 UTC) and for night-time (0–4 UTC). To specifically diagnose fresh emissions occurring from the local to the regional scale, selected reactive compounds have been introduced in the analysed dataset. Among the NM-VOCs measured at CMN, we have considered those which are characterised by relatively low summer reactivity with respect to OH, i.e., toluene, pentanes, butanes, propane, benzene and ethyne (see [22]). For pentanes and butanes, only one isomer has been considered (i-butane and i-pentane, namely) because of their very high cross correlation. The correlation and the cluster analysis were carried out for the original time resolution, with the purpose of retaining the highest degree of information from the datasets. This choice would disentangle the different conditions characterised by daytime upward thermal transport (i.e., direct influence of air-masses from the regional PBL, as suggested in Section 3.1) and more aged (background) air-masses during nighttime [20]. With respect to the previous analysis a larger number of trace gases has been considered.

For the daytime dataset of residuals (Figure 7), CH₃Cl mostly correlates with benzene (R = 0.53, p < 0.01), ethyne (0.48, p < 0.01), CO (R = 0.50, p < 0.01) and n-octane (0.47 p < 0.01). In general, the highest (positive) correlation was found for CO and the anthropogenic NM-VOCs, thus supporting a role of anthropogenic emission (possibly co-located with combustion emission) in affecting the variability of CH₃Cl during daytime at CMN. The degree of correlation appeared to be slightly higher with NM-VOCs characterized by higher lifetime with respect to OH mixing ratios (i.e., ethyne, benzene) and lower for the more reactive toluene (R = 0.40, p < 0.01). In contrast to the smoothed dataset, a weak negative correlation was found between CH₃Cl and COS for the residuals. This can be explained by considering the inverse diel variability which characterizes COS during summer 2017 at CMN (i.e., high values during nighttime versus lower values during daytime). On the other hand, this can further stress the relative decoupling of CH₃Cl and COS variability during daytime at CMN (this latter being also influenced by photosynthetic activity of vegetation [19]).

According to the cluster analysis (Figure 8), the strongest similarity is found between CH₃Cl and those anthropogenic NM-VOCs that were identified by [22] as tracers of fuel evaporation (i.e., pentane, octane and butane) and industrial solvent (i.e., toluene). In particular, the lowest distance was found for toluene and octane. Other clusters (but with lower similarity to CH₃Cl) appeared to be related to vehicular fuel combustion (i.e., CO, benzene, ethyne) and fugitive industrial emissions (CH₂Cl₂ and HCFC-22). This clustering analysis confirms the “independent” behaviour of COS during summer daytime at CMN with respect to other anthropogenic tracers.

For the nighttime dataset of residuals (Figure 9), CH₃Cl showed a high correlation with CO (R = 0.57, p < 0.01), benzene, propane, ethyne (R from 0.50 to 0.45, p < 0.01), which are the least reactive compounds, suggesting that they can describe contributions of air masses that underwent more aged transport of automobile exhausts, fuel evaporation and generation of Liquified Petroleum Gas-LPG (ethyne, benzene and propane). A not negligible positive correlation was found also between CH₃Cl and butane (R = 0.44, p < 0.01), a more reactive trace gas (estimated lifetime with respect to OH lower than 1 day in summer, [22]) associated to fuel evaporation.

Looking at the cluster analysis reported in Figure 10, similarities between CH₃Cl and the other trace gases are less pronounced. This is probably related to the fact that during nighttime CMN is well decoupled from the regional PBL and the higher degree of dynamic mixing of air-masses reaching the measurement site during nighttime reduces the typical emission fingerprints. However, in general, CH₃Cl appeared to be decorrelated also from COS (due to the negative linear correlation between them). The mixing occurring in the free troposphere probably has the larger influence on nighttime concentrations.

3.3. Analysis of CH₃Cl with Local Wind Regime

In an attempt at providing a first spatial identification of regional emissions affecting hourly CH₃Cl enhancements at CMN, we analysed the positive residuals of CH₃Cl and residuals of selected tracers as a function of wind speed and directions during daytime. The choice of restricting our attention only to the daytime time window for this analysis is related to the following reasons. Firstly, during daytime, the thermal transport is active, making observations at CMN more representative of the PBL air masses. Secondly, it is for daytime that we obtained the more robust relationship between CH₃Cl and other trace gases from the cluster analysis.

To this aim, we calculated bivariate polar plots (Figure 11) of mixing ratios [32] for CO (combustion tracer), benzene (as traffic exhaust tracer) and toluene and n-octane (tracers of evaporative emissions): all these trace gases showed a high degree of temporal correlation (like in the case of CO and benzene) or similarity (like in the case of toluene and n-octane) with CH₃Cl, as deduced by the multiple correlation analysis and the clustering. For the sake of completeness, the equivalent nighttime analysis is presented in the supplementary material (SM). Moreover, while CO and benzene are expected to be influenced also by long-range emissions, toluene is expected to be more representative of local/regional emissions due to its shorter atmospheric lifetime. The bivariate analysis has been performed using the “Openair” package [33] included in the “R” Language and Environment for Statistical Computing. Wind speed, wind direction and mixing ratio data are partitioned into wind speed-direction bins and the mean mixing ratio calculated for each bin. The wind components, u and v, are calculated by using the equation:

u = ū sin (2π/θ), v = ū cos (2π/θ)

(1)

where ū is the mean hourly wind speed and θ is the mean wind direction in degrees, with 90 degrees indicating wind coming from the East. The field of mixing ratio (mr) as a function of (ū, θ) is obtained by using a Generalized Additive Model (GAM), e.g., [34,35] which is useful to model the non-linear relationship between the variables. GAMs can be expressed as follows:

$$\sqrt{\mathit{m}{\mathit{r}}_{\mathit{i} }}={\mathit{\beta }}_{\mathit{o}}+{\displaystyle \sum }_{\mathit{j}=\mathit{1}}^{\mathit{n}}{\mathit{s}}_{\mathit{j}} ({\mathit{x}}_{\mathit{i}\mathit{j}})+{\mathit{e}}_{\mathit{i}}$$

(2)

where mr_i is the i^th pollutant mixing ratio, β₀ is the overall mean of the response, s_j(x_ij) is the smooth function of i^th value of covariate j, n is the total number of covariates, and e_i is the i^th residual. More details can be found on [33].

Noteworthy, independently of wind speed or direction, CH₃Cl residuals showed only positive values: this would indicate the existence of emissions occurring at regional scale over northern Italy. The largest CH₃Cl enhancements occurred in air masses coming from S to NW and wind speeds of 1 to 10 m/s.

In particular, the high CH₃Cl values associated with wind from S–NW are coherent with the presence of important power production plants (La Spezia and Livorno) and a petrochemical industry (Livorno area), which are located along the western coastlines of northern Italy (see Figure 1) in a range of 60–80 km from CMN and the northern Apennines.

The bivariate analysis of other atmospheric tracers can provide further hints towards the attribution of the CH₃Cl enhancements. In good agreement with the clustering analysis, toluene and n-octane showed large enhancements from the southern sector, for which CH₃Cl is also largely enhanced. Since these NM-VOC are well representative of evaporative sources, this would support an impact from the region South to CMN, where petrochemical plants are located (Figure 1). For CO and benzene, a somewhat different bivariate distribution was observed with respect to toluene and n-octane: during daytime the highest values were associated with air-masses from NE to SW and thus spanning a range of spatial origin encompassing the Po basin, the Florence urban basin as well as the region for which toluene and n-octane enhancements are evident (Figure 1). The high CH₃Cl enhancements along the wind sectors W–NW, would suggest a role of marine emissions from the Thyrrenian Sea which might affect daytime CH₃Cl variability at CMN. Even if a previous summer field campaign assessed the impact of marine emissions to CMN as marginal [36], this possibility can be supported in light of the small enhancements of the most reactive NM-VOCs observed for this sector (e.g., see toluene). For the nighttime selections (see Figure S1), the bivariate analysis provided less consistent results than for the daytime selection. This is in agreement with the dissimilarity observed during nighttime with cluster analysis and can support the possibility of contributions related to large-scale ocean emissions when the measurement site is more exposed to large-scale circulation.

4. Conclusions

We analysed a summer dataset of high frequency CH₃Cl observations at the high mountain station Mt. Cimone (northern Italy). The execution of an intensive field experiment (ACTRIS-2 Mt Cimone and Po Valley Field Campaign 2017) allowed us to perform a first detailed investigation of CH₃Cl variability over the synoptic and 24 h scales. At the synoptic scale, the highest correlations were found with COS (R = 0.70), CO (R = 0.48) and CH₂Cl₂ (R = 0.37). A cluster analysis clearly indicated a similarity between CH₃Cl variability and COS suggesting a dominant role of natural emissions from oceans with respect to anthropogenic sources, which nevertheless contributed to the synoptic scale variability of CH₃Cl at CMN, as deduced by the co-variability with CO and CH₂Cl₂. A sensitivity study was carried out to assess the impact of using different agglomeration methods in the clustering: “single linkage”, “multiple linkage”, and “Ward”. The similarity between CH₃Cl and COS was robust for all the agglomeration methods used.

Similarly to other atmospheric tracers emitted in the northern Italy PBL, CH₃Cl showed a diel cycle with a maximum during late afternoon as results of the occurrence of upward transport of PBL air masses by thermal wind circulation and daytime PBL height growth. The analysis of CH₃Cl multiple correlation revealed co-variability with CO, benzene, ethyne and propane, suggesting the possibility of co-location of CH₃Cl sources with anthropogenic combustions (like vehicular traffic). In particular, a specific role of emissions related to evaporative sources of fuel (i.e., pentane, octane and butane) and industrial solvents (i.e., toluene) was supported by the cluster analysis. This would support a role of petro-chemical or industrial processes. In this case, some differences exist when the various agglomeration methods are considered. The similarity with evaporative-released NM-VOCs (toluene, butane, octane, propane, pentane) is confirmed when the “Ward” agglomeration method is used (here not shown), while in the case of the “single linkage” method, the similarity is higher with SO₂ and NO_2, thus supporting a role of fossil fuel/coal emissions in determining CH₃Cl variability.

Even if far from being conclusive, the bivariate analysis of CH₃Cl as a function of wind direction and speed, suggests the possible existence of anthropogenic sources in the western region of northern Italy, where plants for power production and petrochemical industries are located. The high CH₃Cl enhancements along the wind sectors W-NW would suggest a possible role of the near Tyrrhenian Sea in affecting the daytime CH₃Cl variability at CMN. However, it must be considered that this wind sector is much less important (in term of frequency) than the SW–S for which we obtained more robust evidence of co-located emission of anthropogenic NM-VOCs. These results would indicate that, during the summer, the northern Italy PBL would represent a source of CH₃Cl for the lower free troposphere of the Mediterranean basin/South Europe thanks to thermal transport along the mixed layer and up-slope/up-valley winds across the Apennine ridge: once transported at the altitude of CMN, material from PBL can be diffused over large distance. Even if a comparative analysis of hour-to-hour vs. day-to-day variability (mostly affected by natural CH₃Cl emissions, as deduced by the high similarity with COS variability) indicated that the former only contributed to 1/5 of the overall CH₃Cl variability, the possible role of emissions from the regional PBL cannot be neglected. Further work, by exploiting the long-term dataset of observations available at CMN, will be carried out to systematically assess the impact of this process for the whole year and to better constraint the identification of the anthropic CH₃Cl sources and the variability of their strength with time. Since industrial production and use of CH₃Cl have not been regulated under the Montreal Protocol (MP) or its successor amendments, continuous monitoring of CH₃Cl outflow from the Po basin is important to properly evaluate its anthropogenic emissions.