Reconstructing Elemental Carbon Long-Term Trend in the Po Valley (Italy) from Fog Water Samples

: Elemental carbon (EC), a ubiquitous component of ﬁne atmospheric aerosol derived from incomplete combustion, is an important player for both climate change and air quality deterioration. Several policy measures have been implemented over the last decades to reduce EC emissions from anthropogenic sources, but still, long-term EC measurements to verify the e ﬃ cacy of such measurements are limited. In this study, we analyze the concentration of EC suspended in fog water samples, collected over the period 1997–2016 in a rural background site of the southern Po Valley. The comparison between EC in fog water and EC atmospheric aerosol concentration measured since 2012 allowed us to reconstruct EC atmospheric concentration from fog water chemical composition dating back to 1997. The results agree with the EC atmospheric observations performed at the European Monitoring and Evaluation Program (EMEP) station of Ispra in the northern part of the Po Valley since 2002, and conﬁrm that the Po Valley is a pollution hotspot, not only in urban areas, but also in rural locations. The reconstructed trend over the period 1997–2016 indicates that EC concentration during the winter season has decreased on average by 4% per year, in agreement with the emission reduction rate, conﬁrming the e ﬀ ectiveness of air quality measures implemented during the past 20 years.


Introduction
The occurrence in the atmosphere of elemental carbon (EC), referred to also as black carbon (BC), is relevant from both climatic and human health standpoints [1,2]. The terms elemental carbon and black carbon are used interchangeably, and refer to the strongly light-absorbing atmospheric

Sampling Sites
Measurements discussed in this paper were performed at two rural sites in the Po Valley, San Pietro Capofiume (SPC) and Ispra (Figure 1a). The Po Valley, Northern Italy, is a highly polluted region [10], due to significant anthropogenic emissions from both the agricultural and industrial sectors and from transportation. In addition, one fourth of the national population lives in this area, with a population density almost double the national average. Finally, the valley is surrounded to the north, west, and south by mountain ranges that prevent pollution dispersion and favor accumulation, especially during the cold season.
SPC is a rural background site, part of the air quality monitoring network of the Regional Environmental Protection Agency (ARPAE). The site is located about 30 km northeast of Bologna, and is surrounded by agricultural fields and small villages. It is characterized by stable meteorological conditions with high relative humidity, low wind speed (generally below 2 m s −1 ), and low temperature during the winter season, leading to frequent fog events [11]. Sampling site and air quality characterization during winter at SPC have been described previously [12,13].
Ispra is a regional background site located in the northwestern part of the Po Valley, about 60 km north of Milan [14]. The nearest urban areas are Novara, 40 km to the south, and Varese, 20 km to the west. The site has been part of the European Monitoring and Evaluation Program (EMEP) since 1985. The site is affected by polluted air masses from the Po Valley in southerly wind conditions, while clean air masses are observed during northerly wind periods [15].
The extensive compilation of PM 2.5 concentrations and carbonaceous aerosol measurements across different sites in Italy show that the two sites are representative of the rural Po Valley [16].

Fog Water Collection
Fog water samples were collected at the rural site of SPC. Giulianelli et al. [13] investigated fog occurrence and fog water chemical composition trends during the last twenty years at SPC. They reported a decreasing trend in fog occurrence frequency, as well as a reduction in the amount of sulfate, nitrate, and ammonium dissolved in fog water. Routine fog water collection started in winter 1997, and is performed systematically from November to March.
Fog water is collected with an automated home-made collector, described in detail by Fuzzi et al. [17]. During each fog event, a single fog water sample is collected. The collector is equipped with a fan located on the edge of a tunnel, that creates a flow of about 17 m 3 min −1 . The air stream generated by the fan passes through a grid made of parallel stainless-steel strings, where fog droplets coalesce, to eventually fall into a funnel and then into the sampling bottle. The 50% collection efficiency of each string is about 3 µm.
Atmosphere 2020, 10, x FOR PEER REVIEW 3 of 13 sectors and from transportation. In addition, one fourth of the national population lives in this area, with a population density almost double the national average. Finally, the valley is surrounded to the north, west, and south by mountain ranges that prevent pollution dispersion and favor accumulation, especially during the cold season. SPC is a rural background site, part of the air quality monitoring network of the Regional Environmental Protection Agency (ARPAE). The site is located about 30 km northeast of Bologna, and is surrounded by agricultural fields and small villages. It is characterized by stable meteorological conditions with high relative humidity, low wind speed (generally below 2 m s −1 ), and low temperature during the winter season, leading to frequent fog events [11]. Sampling site and air quality characterization during winter at SPC have been described previously [12,13].
Ispra is a regional background site located in the northwestern part of the Po Valley, about 60 km north of Milan [14]. The nearest urban areas are Novara, 40 km to the south, and Varese, 20 km to the west. The site has been part of the European Monitoring and Evaluation Program (EMEP) since 1985. The site is affected by polluted air masses from the Po Valley in southerly wind conditions, while clean air masses are observed during northerly wind periods [15].
The extensive compilation of PM2.5 concentrations and carbonaceous aerosol measurements across different sites in Italy show that the two sites are representative of the rural Po Valley [16].

Fog Water Collection
Fog water samples were collected at the rural site of SPC. Giulianelli et al. [13] investigated fog occurrence and fog water chemical composition trends during the last twenty years at SPC. They reported a decreasing trend in fog occurrence frequency, as well as a reduction in the amount of sulfate, nitrate, and ammonium dissolved in fog water. Routine fog water collection started in winter 1997, and is performed systematically from November to March.
Fog water is collected with an automated home-made collector, described in detail by Fuzzi et al. [17]. During each fog event, a single fog water sample is collected. The collector is equipped with a fan located on the edge of a tunnel, that creates a flow of about 17 m 3 min −1 . The air stream generated by the fan passes through a grid made of parallel stainless-steel strings, where fog droplets coalesce, to eventually fall into a funnel and then into the sampling bottle. The 50% collection efficiency of each string is about 3 μm.
The fog water collector is driven by a computer program specifically designed to activate the fog sampling when the temperature is higher than 1 °C (to avoid freezing conditions), and liquid water content (LWC) is larger than 0.08 g m −3 (set at arbitrary threshold, and corresponding to a dense fog condition). LWC is continuously measured with a particle volume monitor (PVM-100) at 1-minute time resolution.   The fog water collector is driven by a computer program specifically designed to activate the fog sampling when the temperature is higher than 1 • C (to avoid freezing conditions), and liquid water content (LWC) is larger than 0.08 g m −3 (set at arbitrary threshold, and corresponding to a dense fog condition). LWC is continuously measured with a particle volume monitor (PVM-100) at 1-min time resolution.
The data here reported refer to the winter seasons from 1997 to 2016. The number of fog samples collected during each season varied between 7 and 32, with an average of 20 samples per season (428 samples over 20 years).

Quantification of Carbonaceous Particles Suspended in Fog Water Samples
Immediately after sampling, 10 mL of fog water samples are filtered through a quartz fiber filter (PALL QAO-UP 2500 filters, Ø 47 mm, Port Washington, NY, USA) with a glass vacuum filtration device. Filters are then stored at −4 • C prior to analysis.
Samples collected between 1997 and 2001 were analyzed for total carbon (TC) and EC by double thermal analysis. The analysis of filter punches exposed for several minutes to a stream of pure O 2 at 340 • C led to the determination of EC, and the analysis of filter punches without thermal treatment led to the determination of TC. The temperature of 340 • C was chosen to remove organic aerosol without oxidation of EC [18], and to avoid the potential effects of organic aerosol charring and EC overestimation. No direct comparison of this double thermal analysis procedure with more recently standardized protocols is available for insoluble carbonaceous material.
From 2002 to 2011, we measured TC concentrations in filtered fog water samples using an EA Flash2000 Thermo Fisher Scientific. For each sample, two to four replicate analyses of 6 or 9 mm diameter punches were performed. At the beginning of each sequence of analysis, three capsules of Atropine (IAEA standard) and one of USGS-40 were analyzed. Then, for every 8-10 samples, one capsule of an Adriatic sediment used as laboratory reference was analyzed in order to compensate for potential instrument drift, and as a quality control measure. The average standard deviation of each measurement, determined by replicate analyses of the same sample, was ±0.07% for TC.
From 2012, the EC and OC concentrations collected on the filters were quantified by thermal-optical analysis (Sunset, Laboratory Inc., USA), using the NIOSH protocol [19].

Analysis of EC in Atmospheric Aerosol Samples
PM 2.5 aerosol samples for EC quantification have been collected routinely at SPC, starting in winter 2012, with 24-h aerosol samples collected from midnight. Sampling is performed on quartz fiber filters (PALL QAO-UP 2500 filters, Ø 47 mm), with a low-volume sampler (TCR Tecora) at a flow-rate of 16.7 L min −1 [20]. Filters are pre-baked at 800 • C for 5 h in order to remove any carbonaceous material. EC is then quantified by thermal-optical analysis in transmittance (TOT) using the EUSAAR2 protocol [21,22]. Figure 2 reports the inter-annual variability of insoluble particulate TC measured in fog water over the whole sampling period. Each box corresponds to a fog season starting in November and ending in March of the following year. The year identifying the fog season in the graph corresponds to the beginning of the fog season. Season-averaged TC values ranged from 2.9 µg mL −1 (24 samples) to 16.6 µg mL −1 (7 samples), while single fog event TC concentrations ranged between 0.4 µg mL −1 and 47.0 µg mL −1 . In order to test if a significant time trend could be detected over the investigated period, we calculated the Theil-Sen estimator of the seasonal TC concentration. The resulting slope was −0.08 µg mL −1 year −1 with a median absolute deviation (MAD) equal to 0.23 µg mL −1 year −1 . The large MAD compared to the slope confirms that no time trend can be detected in the concentration of particulate TC suspended in fog water over the last twenty years. A previous analysis of fog water chemical composition trends clearly showed a significant decrease in water-soluble inorganic species dissolved in fog water [13]. In particular, sulfate, nitrate, and ammonium concentration decreased by 76%, 43%, and 55%, respectively, starting from the late 1990s. On the contrary, water-soluble organic carbon did not show any significant trend [13]. The results presented here also indicate that the insoluble carbonaceous particles suspended in fog water follow the same behavior as fog water-soluble carbon.

TC and EC Seasonal Trends
During the first years of fog collection and analysis (from 1997 to 2000), and then more recently during the last five fog seasons analyzed (from 2012 to 2016), the characterization of carbonaceous particles suspended in fog water allowed us to quantify the EC fog water content, together with insoluble particulate TC. Over the whole measurement period, EC averaged 1.2 µg mL −1 , ranging between 0.1 and 7.1 µg mL −1 in single event samples. The trend of EC concentrations measured in fog water shows a large year-to-year variability, with seasonal averages varying from 0.5 µg mL −1 to 2.7 µg mL −1 . Despite the limited dataset, the variability of seasonal averages over the different years is within the single seasonal variability, and no long-term EC trend can be detected.
In order to verify the similarity of the two datasets when EC measurements are available (before 2000 and after 2012), we first tested the normal distribution of EC to TC ratios, and then performed the Welch t-test. Figure 3 shows the Q-Q plot of EC/TC ratios observed during the two periods, reporting on the y-axis the measured ratios, and on the x-axis the distribution of the data around the mean (quantified by z-scores or number of standard deviations). The plots indicate that in both periods, the ratios were distributed around the mean with a variability of about two times its standard deviation. Over the period 1997-2000, the distribution was skewed towards larger values, showing outliers. After removing the outliers observed during the first part of the experiment (4 out of 73 samples), which could possibly come from the different analytical method used at that time, the Welch t-test indicates that the two periods showed identical EC to TC ratios at a confidence interval of 99.95%.
Atmosphere 2020, 10, x FOR PEER REVIEW 5 of 13 During the first years of fog collection and analysis (from 1997 to 2000), and then more recently during the last five fog seasons analyzed (from 2012 to 2016), the characterization of carbonaceous particles suspended in fog water allowed us to quantify the EC fog water content, together with insoluble particulate TC. Over the whole measurement period, EC averaged 1.2 μg ml −1 , ranging between 0.1 and 7.1 μg ml −1 in single event samples. The trend of EC concentrations measured in fog water shows a large year-to-year variability, with seasonal averages varying from 0.5 μg ml −1 to 2.7 μg ml −1 . Despite the limited dataset, the variability of seasonal averages over the different years is within the single seasonal variability, and no long-term EC trend can be detected.
In order to verify the similarity of the two datasets when EC measurements are available (before 2000 and after 2012), we first tested the normal distribution of EC to TC ratios, and then performed the Welch t-test. Figure 3 shows the Q-Q plot of EC/TC ratios observed during the two periods, reporting on the y-axis the measured ratios, and on the x-axis the distribution of the data around the mean (quantified by z-scores or number of standard deviations). The plots indicate that in both periods, the ratios were distributed around the mean with a variability of about two times its standard deviation. Over the period 1997-2000, the distribution was skewed towards larger values, showing outliers. After removing the outliers observed during the first part of the experiment (4 out of 73 samples), which could possibly come from the different analytical method used at that time, the Welch t-test indicates that the two periods showed identical EC to TC ratios at a confidence interval of 99.95%.   During the first years of fog collection and analysis (from 1997 to 2000), and then more recently during the last five fog seasons analyzed (from 2012 to 2016), the characterization of carbonaceous particles suspended in fog water allowed us to quantify the EC fog water content, together with insoluble particulate TC. Over the whole measurement period, EC averaged 1.2 μg ml −1 , ranging between 0.1 and 7.1 μg ml −1 in single event samples. The trend of EC concentrations measured in fog water shows a large year-to-year variability, with seasonal averages varying from 0.5 μg ml −1 to 2.7 μg ml −1 . Despite the limited dataset, the variability of seasonal averages over the different years is within the single seasonal variability, and no long-term EC trend can be detected.
In order to verify the similarity of the two datasets when EC measurements are available (before 2000 and after 2012), we first tested the normal distribution of EC to TC ratios, and then performed the Welch t-test. Figure 3 shows the Q-Q plot of EC/TC ratios observed during the two periods, reporting on the y-axis the measured ratios, and on the x-axis the distribution of the data around the mean (quantified by z-scores or number of standard deviations). The plots indicate that in both periods, the ratios were distributed around the mean with a variability of about two times its standard deviation. Over the period 1997-2000, the distribution was skewed towards larger values, showing outliers. After removing the outliers observed during the first part of the experiment (4 out of 73 samples), which could possibly come from the different analytical method used at that time, the Welch t-test indicates that the two periods showed identical EC to TC ratios at a confidence interval of 99.95%.   The average EC to TC ratio calculated over the entire dataset (0.16 ± 0.01) was then used to estimate the concentration of EC suspended in fog water from the TC concentration between 2000 and 2012 when no direct EC measurements were available. The seasonal averages of measured EC are reported in Figure 4 in red, while the estimated EC seasonal averages are reported in cyan.
Atmosphere 2020, 10, x FOR PEER REVIEW 6 of 13 The average EC to TC ratio calculated over the entire dataset (0.16 ± 0.01) was then used to estimate the concentration of EC suspended in fog water from the TC concentration between 2000 and 2012 when no direct EC measurements were available. The seasonal averages of measured EC are reported in Figure 4 in red, while the estimated EC seasonal averages are reported in cyan.

Atmospheric EC Concentrations
We calculated the atmospheric loading of EC captured by fog droplets, by multiplying the concentration of EC suspended in fog water by the atmospheric liquid water content (LWC). The obtained concentration could be considered as a proxy of EC atmospheric concentrations, under the assumption that EC was completely scavenged by fog droplets. Previous observations showed that the EC scavenging coefficient, e.g., the fraction of EC concentration that is removed from the atmosphere by fog, is actually far from 1. Gilardoni et al. [12] observed that the EC scavenging coefficient in fog can vary significantly from one event to the other, with scavenging coefficients ranging between 0.3 and 0.6. Hallberg et al. [23] reported, instead, a significantly lower fog scavenging coefficient, equal to 0.06. It is likely that, since EC is expected to be found in smaller aerosol particles (generally below 300 nm), the efficiency of scavenging depends strongly on the particles' mixing state [24], which varies over time and space.
An accurate determination of EC scavenging efficiency would require the comparison of atmospheric EC concentration right before and after fog formation [12]. In this study, we are more interested in identifying a conversion parameter to extrapolate atmospheric seasonal variability of atmospheric EC concentrations from long-term fog composition measurements. With this purpose, we calculated the average values of EC concentration in PM2.5 aerosol samples, over each whole fog season (from November to March), and compared them with the seasonal averages of EC fog water concentration, adjusted for LWC. The ratios vary between 0.16 and 0.49, with an average of 0.28 ± 0.05. Based on this conversion factor, we derived the EC atmospheric concentration from fog water composition back to 1997.

Long-Term Trends of Atmospheric EC Concentrations
The fog-season averaged EC concentration estimated for SPC is shown in Figure 5a by the blue bars. The error bars in Figure 5a take into account the seasonal variability, as well as the uncertainty associated with the assumption of a constant EC to TC ratio, to calculate EC fog water concentration, and a constant conversion factor, to calculate atmospheric concentrations. The seasonal averages range between 0.8 and 4.2 μg m −3 . For comparison, Table 1 reports the average EC concentrations observed in PM2.5 and PM10 aerosol samples collected at different European sites. Sandrini et al. [16] reported a survey of EC measurements at different urban and rural sites in the Po Valley, mainly between 2005 and 2008. The seasonal EC average ranged between 1 and 2 µg m −3 at the rural sites, and between 2 and 4 µg m −3 at the urban sites ( [16] and references therein). Over the same period, the

Atmospheric EC Concentrations
We calculated the atmospheric loading of EC captured by fog droplets, by multiplying the concentration of EC suspended in fog water by the atmospheric liquid water content (LWC). The obtained concentration could be considered as a proxy of EC atmospheric concentrations, under the assumption that EC was completely scavenged by fog droplets. Previous observations showed that the EC scavenging coefficient, e.g., the fraction of EC concentration that is removed from the atmosphere by fog, is actually far from 1. Gilardoni et al. [12] observed that the EC scavenging coefficient in fog can vary significantly from one event to the other, with scavenging coefficients ranging between 0.3 and 0.6. Hallberg et al. [23] reported, instead, a significantly lower fog scavenging coefficient, equal to 0.06. It is likely that, since EC is expected to be found in smaller aerosol particles (generally below 300 nm), the efficiency of scavenging depends strongly on the particles' mixing state [24], which varies over time and space.
An accurate determination of EC scavenging efficiency would require the comparison of atmospheric EC concentration right before and after fog formation [12]. In this study, we are more interested in identifying a conversion parameter to extrapolate atmospheric seasonal variability of atmospheric EC concentrations from long-term fog composition measurements. With this purpose, we calculated the average values of EC concentration in PM 2.5 aerosol samples, over each whole fog season (from November to March), and compared them with the seasonal averages of EC fog water concentration, adjusted for LWC. The ratios vary between 0.16 and 0.49, with an average of 0.28 ± 0.05. Based on this conversion factor, we derived the EC atmospheric concentration from fog water composition back to 1997.

Long-Term Trends of Atmospheric EC Concentrations
The fog-season averaged EC concentration estimated for SPC is shown in Figure 5a by the blue bars. The error bars in Figure 5a take into account the seasonal variability, as well as the uncertainty associated with the assumption of a constant EC to TC ratio, to calculate EC fog water concentration, and a constant conversion factor, to calculate atmospheric concentrations. The seasonal averages range between 0.8 and 4.2 µg m −3 . For comparison, Table 1 reports the average EC concentrations observed in PM 2.5 and PM 10 aerosol samples collected at different European sites. Sandrini et al. [16] reported a survey of EC measurements at different urban and rural sites in the Po Valley, mainly between 2005 and 2008. The seasonal EC average ranged between 1 and 2 µg m −3 at the rural sites, and between 2 and 4 µg m −3 at the urban sites ( [16] and references therein). Over the same period, the seasonal EC average estimated at SPC varied between 1 and 3.4 µg m −3 , in agreement with previously-reported EC atmospheric concentrations. Atmosphere 2020, 11, 580 8 of 12 A multi-site analysis performed across European rural sites at different latitudes in fall 2008 and winter 2009 reports EC concentrations ranging from 0.1, in Northern Europe, to 1.5 µg m −3 , in Southern Europe [25]. Cavalli et al. [26] and Zanatta et al. [27] investigated the seasonal average values of EC for several rural background sites across Europe, from multi-year observations. The EC winter averages, excluding those reported for Ispra, ranged between 0.1 and 0.8 µg m −3 . Higher EC concentrations are typically observed in Northeastern Europe during the heating season (typically from September to March). Single season experiments report average EC values varying between 0.8 and 3.9 µg m −3 [28][29][30]. Ispra is a regional background site strongly affected by anthropogenic emissions, located in the northern part of the Po Valley, as previously observed based on back trajectory statistical analysis [14,31]. The average atmospheric EC concentration measured at Ispra during the period 2008-2011 was 1.6 µg m −3 in PM 2.5 samples and 2 µg m −3 in PM 10 samples [26], thus higher than the other rural Western Europe sites, and comparable to concentrations in Eastern Europe sites. Similarly, during the winter seasons of 2008-2011, the EC concentration at SPC, estimated from fog water composition, was 1.4 µg m −3 . Altogether, these results confirm that the Po Valley is a pollution hotspot, not only in its highly urbanized environment, but also in rural locations. Figure 5a reports the EC atmospheric concentration from long-term measurements performed at the EMEP site of Ispra, together with the SPC data. The long-term trend of EC concentration in SPC derived from fog water composition is compared with the atmospheric EC observations in Ispra, to show the consistency of the two series. Ispra monitoring site, like SPC, is located in the Po Valley; it is a regional background site representative of the northern part of the valley, and is characterized by one of the longest time records of atmospheric EC measurements in Europe. The comparability of Ispra data from 2002 to 2004 with the following seasons is limited, since EC was measured with a flash heating procedure, which did not allow for charring correction, and data coverage was also lower. The EC/TC ratio determined with the flash heating procedure (without charring correction) used from 2002 to 2004 was compared with the EC/TC ratio determined with a thermal-optical method implementing the EUSAAR2 thermal protocol against PM 2.5 samples in 2005, and EC data reported in Figure 5 are corrected accordingly. The seasonal EC averages estimated at SPC are comparable or slightly lower than the seasonal EC averages measured at Ispra. The higher concentrations reported in Ispra can be attributed to the strong influence of anthropogenic emissions observed at this site, relative to other rural sites in the Po Valley [31]. During the overlapping period, the highest EC concentrations at SPC in the years 2006 and 2011 coincide with the high EC concentrations at Ispra. These results support the robustness of the procedure employed here to derive atmospheric concentrations from fog water composition data.
We have analyzed the trend of atmospheric EC concentration over the 20-year period from 1997 to 2016 at SPC (Table 2). First, we calculated the Kendall's tau, which defines the strength of the rank correlation between the variables "fog season year" and "EC concentration". The tau value indicates that the null hypothesis, i.e., EC varies independently from the year, has to be rejected at the 95% confidence level. The Theil-Sen's slope quantifies the negative trend, which corresponds to a 0.06 µg m concentrations at Ispra. These results support the robustness of the procedure employed here to derive atmospheric concentrations from fog water composition data. We have analyzed the trend of atmospheric EC concentration over the 20-year period from 1997 to 2016 at SPC (Table 2). First, we calculated the Kendall's tau, which defines the strength of the rank correlation between the variables "fog season year" and "EC concentration". The tau value indicates that the null hypothesis, i.e., EC varies independently from the year, has to be rejected at the 95% confidence level. The Theil-Sen's slope quantifies the negative trend, which corresponds to a 0.06 μg m −3 decrease in EC per year, with a relatively small MAD value (0.03). The same analysis, applied to   Consistent long-term records of atmospheric EC concentration are quite limited worldwide. Murphy et al. [32] investigated the trends of EC at rural and remote sites across the United States from 1990 to 2004 and reported, for low-altitude sites, a decreasing trend ranging between zero and 6% per year. The trend was more prominent during the cold season. Higher reduction rates are reported for urban locations. Yamagami et al. [33] observed a decreasing trend of 13% per year over the period 2003-2016 in the urban area of Nagoya (Japan). Shorter time series, although less representative, still confirm a decreasing tendency over the more recent years. For example, atmospheric EC concentration decreased on average by 10%-20% from 2012 to 2017 in the Los Angeles area, with the highest reduction during the cold season (up to 28%) [34]. Singh et al. [35] reported a statistically significant reduction in EC at the urban and curbside stations across the United Kingdom, ranging between 6% and 8% per year over the period 2009 to 2016. Although representative exclusively of the cold season (when fog is present), in the present study, we estimate a decrease in atmospheric EC of 4% per year, on average, in agreement with observations at US rural sites [32]. Such a rate is also in agreement with the reduction in BC emissions at national level, reported in Figure 5b [36]. Emissions are calculated at the national level every year by the Italian Institute for Environmental Protection and Research (ISPRA). According to the most recent inventory, emission data are characterized by a clear decreasing trend over the period 1997-2016, with steeper reductions after 2008. Sectors for which really high decreases are observed are: (i) combustion in energy and transformation industries; (ii) production processes and road transport; and (iii) other mobile sources; while an increase was observed from the non-industrial combustion plants. Long-term atmospheric EC data confirm the effectiveness of some air quality policy measures implemented during the last 20 years, especially in reducing emissions from road transport [37]. Additional measures will be implemented in this direction, according to the results of air emission and air quality scenarios.

Conclusions
The measurement of the EC concentrations in fog water samples collected over a period of ca. 20 years at the station of San Pietro Capofiume allowed the reconstruction of a long-term trend of aerosol EC in the southeastern part of the Po Valley, Italy.
Over the study period, a reduction in EC aerosol concentration of about 4% per year was estimated, indicating the effectiveness of the implemented policy aimed at emission reduction. Similar trends are reported for PM; in fact, Bigi et al. [38] calculated that PM 10 in the Po Valley decreased by 2%-5% per year over the period 2002-2011. PM 2.5 atmospheric concentrations have been routinely monitored only more recently [39]. Trend analysis across the Po Valley identifies a decreasing tendency between 1% and 8% per year since the mid-2000s [15,39]. Longer time series analysis of total suspended particles (TSP) reported even a stronger decreasing trend from the end of the 1990s.
The concentration of aerosol EC at the rural station of San Pietro Capofiume is comparable to that of urban sites across Europe and of Eastern Europe locations. This is also true for the Ispra station that is located in the northern part of the Po Valley. These results confirm that the Po Valley is clearly a pollution hotspot, not only in its highly urbanized environment, but also in rural locations. This results from a combination of the effects of the diffuse sources of pollution, and also by the Po Valley orography that prevents pollution dilution.
Funding: This research received no external funding.