Tropospheric and Surface Measurements of Combustion Tracers During the 2021 Mediterranean Wildfire Crisis: Insights from the WMO/GAW Site of Lamezia Terme in Calabria, Southern Italy

Abstract

The central Mediterranean and nearby regions were affected by extreme wildfires during the summer of 2021. During the crisis, Türkiye, Greece, Italy, and other countries faced numerous challenges ranging from the near-complete destruction of landscapes to human losses. The crisis also resulted in reduced air quality levels due to increased emissions of pollutants linked to biomass-burning processes. In the Mediterranean Basin, observation sites perform continuous measurements of chemical and meteorological parameters meant to track and evaluate greenhouse gas and pollutant emissions in the area. In the case of wildfires, CO (carbon monoxide) and formaldehyde (HCHO) are effective tracers of this phenomenon, and the integration of satellite data on tropospheric column densities with surface measurements can provide additional insights on the transport of air masses originating from wildfires. At the Lamezia Terme (code: LMT) World Meteorological Organization–Global Atmosphere Watch (WMO/GAW) observation site in Calabria, Southern Italy, a new multiparameter approach combining different methodologies has been used to further evaluate the effects of the 2021 wildfires on atmospheric measurements. A previous study focused on wildfires that affected the Aspromonte Massif area in Calabria; in this study, the integration of surface data, tropospheric columns, and backtrajectories has allowed pinpointing additional contributions from other southern Italian regions, as well as North Africa and Greece. CO data were available for both surface and column assessments, while continuous HCHO data at the site were only available through satellite. In order to correlate the observed peaks with wildfires, surface BC (black carbon) was also analyzed. The analysis, which focused on July and August 2021, has allowed the definition of three case studies, each highlighting distinct sources of emission in the Mediterranean; the case studies were further evaluated using HYSPLIT backtrajectories and CAMS products. The LMT site and its peculiar local wind patterns have been demonstrated to play a significant role in the detection of wildfire outputs in the context of the Mediterranean Basin. The findings of this study further stress the importance of assessing the effects of wildfire emissions over wide areas.

1. Introduction

Recent assessments on extreme wildfire events and their trends indicate that areas such as the Mediterranean could be exposed to more events of this kind, possibly linked to anthropogenic climate change [1,2,3].

Extreme wildfire events generally have a common factor, which is the phenomenon of pyroconvection [4,5]. This phenomenon results in fire convection columns extending to the middle and upper troposphere [6]. The typical framework of pyroconvection is flammagenitus (or fire) clouds that develop above the smoke plumes linked to a fire event [7]. These clouds, known as pyrocumulonimbus and pyrocumulus—as well as their post-development dynamics—influence the transport of smoke particles and hot gases, which then contribute to the formation of additional clouds following condensation processes [8]. These dynamics can result in several effects that increase the potential of extreme wildfire spreading, such as lightning types of a pyrogenic nature [9,10,11], the injection of aerosols as high as the lower stratosphere [12,13], additional fire growth [14], and significant changes both in surface winds [15,16,17] and synoptic-scale flows [18].

EFFIS, the European Forest Fire Information System, issued a comprehensive report on the effects of the extreme events linked to the 2021 Mediterranean wildfire crisis on European ecosystems [19].

Türkiye has a natural vulnerability to wildfire across multiple regions of the country and the 2021 crisis resulted in unprecedented wildfire occurrences [20] that challenged the country’s capacity to detect them in time and counter the hazard efficiently [21]. The alteration of landscapes resulted in increased exposure to flood hazards [22]. The damage to biodiversity was also notable [23,24,25]. In addition to the main hazards and damage linked to wildfire occurrences, air quality was also affected [26,27]. Satellite data from Sentinel-2 and Sentinel-5P allowed for an assessment of burn severity and CO emissions linked to Turkish fires [28]. The impact of these wildfires on air quality was also assessed via the integration of satellite data with ground measurements [29].

In Greece, heat waves affected the summer of 2021 and triggered a number of wildfires [30]. Fires with similar degrees of intensity had not been seen in Greece for 13 years since the previous summer crisis of 2007, which resulted in the loss of more than 12% of Greece’s forested areas [31,32,33,34,35,36,37]; however, their damage was considerable, and the country’s early warning systems were challenged [38]. Research also evidenced the susceptibility of the country’s wildfire hazard to synoptic-scale drivers [39]. Overall, both Greece and Türkiye were very heavily affected, and surveys attempted to provide a reliable estimate of the combined impact of wildfire-related emissions in the eastern Mediterranean sector [40].

In Italy, fires affected several regions [41] and the atmospheric outputs of these events have also been observed by the WMO/GAW (World Meteorological Organization–Global Atmosphere Watch) observation sites across the country [42,43]. Among the Italian regions affected by the 2021 crisis, Calabria reported extensive wildfire events, especially in the southern area of the region, which coincides with the southernmost area of the entire Italian peninsula in the Aspromonte National Park [43]. The 2021 crisis led to a more detailed analysis of the areas affected by these events in Italy: areas with intense land abandonment have been shown to be vulnerable to such events [44]. The effects on air quality were also assessed [45], as well as the increase in soil exposure to erosion [46].

In this research, three metrics of the main products of wildfires will be analyzed using a combination of ground measurements, satellite data on tropospheric column concentrations, and both methods when applicable: carbon monoxide (CO, ground, and satellite); formaldehyde (HCHO and satellite); and black carbon (BC and ground). Ground measurements have been performed at the WMO/GAW observation site of Lamezia Terme (code: LMT) in Calabria, Southern Italy. The observation site’s location in the central Mediterranean region allowed for the detection of wildfire outputs from various regions across the Mediterranean Basin itself and, therefore, highlighted the diffusion and circulation of these byproducts over a wide area.

Carbon monoxide is a primary output of combustion and is attributed to both anthropogenic and natural sources [47,48]; among the carbon compounds present in the atmosphere, carbon monoxide is now characterized by a declining trend [49], likely linked to more sustainable policies [50] and efficient combustion engines introduced since the 2000s [51]. Prior to these measures, atmospheric carbon dioxide trends and annual emissions were considerable [52]. However, this annual decline rate has lowered in the past few years due to notable wildfire-related outputs [53,54]. CO is used as an effective tracer of wildfire emissions [43].

Formaldehyde (FA) is a carcinogen and mutagen [55,56] characterized by multiple sources; wildfires contribute to outdoor HCHO concentrations in the atmosphere [57,58,59], in addition to anthropogenic sources such as the combustion of coal, industrial activities, manufacturing, and others [60,61,62]. Formaldehyde also poses a considerable indoor air quality (IAQ) risk due to releases from wooden furniture, tobacco smoking, and other sources, which combine with poor indoor ventilation and further increase exposure-related risks [63,64,65,66,67].

Black carbon, or soot, is a relevant product of combustion processes, like CO [48]. Wildfires are known to result in notable BC emissions into the atmosphere [68]. Black carbon poses health hazards [69,70,71] and is also capable of altering Earth’s climate via perturbations in radiative forcing [72,73,74]. It has a high GWP (global warming potential) [75], which is counterbalanced by its short persistence time in the atmosphere [76,77].

The integrated analysis of these parameters from LMT would, therefore, provide regulators and policy makers alike with new insights into the broad effects and impacts of wildfire emissions on the air quality of the Mediterranean region.

The paper is divided as follows: Section 2 describes the observation site of LMT, its findings, and their implications on the diffusion of greenhouse gases and pollutants; Section 3 and Section 4 report the methods and datasets used for analysis and evaluations, as well as those employed for data gathering; Section 5 and Section 6 focus on the results and discussion, respectively; Section 7 closes the paper with a summary of this study’s findings and their implications.

2. Characteristics of Lamezia Terme WMO/GAW Regional Observation Site

Located 600 m from the Tyrrhenian coast of Calabria, Southern Italy, the Lamezia Terme (LMT) observation site (Lat: 38°52.605′ N; Lon: 16°13.946′ E; Alt: 6 m a.s.l.) (Figure 1) is a regional WMO/GAW (World Meteorological Organization–Global Atmosphere Watch) facility performing continuous measurements of greenhouse gases, aerosols, key meteorological data, and other parameters. Located in the westernmost area of the Catanzaro isthmus—which is the narrowest point in the entire Italian peninsula (≈31 km between the Tyrrhenian and Ionian coasts)—the site is affected by a local wind circulation pattern that is well-oriented on the W/NE axis, as described in two works by Federico et al. (2010a, 2010b) [78,79]. Wind circulation also has an impact on local air traffic, as the runway orientation of the Lamezia Terme International Airport (IATA: SUF; ICAO: LICA) located 3 km north of the WMO station is 10/28 (100/280° N).

The station’s location in the center of the Mediterranean exposes the site to frequent Saharan dust events [81] and summer open fire emissions [43]. During LMT’s observation history, local wind circulation has been characterized using different methodologies, such as the analysis of vertical wind profiles [82] and PBL variability [83,84,85]. An analysis of solar radiation using several methodologies was also performed [86]. With respect to reactive gases, the study by Cristofanelli et al. (2017) [87] provided a milestone characterization of the site with remarks on a number of local emission sources (e.g., the above-mentioned airport, livestock farming, and the A2 highway). Some of these sources could be characterized even further in a study on the first COVID-19 lockdown of 2020 and the exceptional circumstances of reduced anthropic activities that followed [88]. Additional details on specific greenhouse gases became available in cyclic and multi-year studies: in D’Amico et al. (2024a) [89], seven continuous years of methane (CH₄) measurements showed that emission peaks were linked to northeastern-continental winds, while western-seaside winds yielded lower values. A seasonal influence was also observed, with winter and summer yielding the highest and lowest concentrations, respectively. Methane concentrations at the site have also shown a strong correlation with wind speeds in a so-defined HBP (Hyperbola Branch Pattern) as the highest concentrations were linked to low speeds and, vice versa, the lowest mole fractions were correlated with high speeds, further demonstrating the influence of local emission sources. A study on surface ozone (O₃), however, demonstrated that these patterns are not constant among gases as ozone showed a “reversed” pattern, adding additional complexity to the correlations between winds and other parameters on a local scale [90]. The analysis of gases and aerosols at LMT also involved cross-station research with other southern Italian stations [91,92] and the evaluation of weekly patterns in pollutants to discriminate natural and anthropogenic outputs [93]. In D’Amico et al. (2024e) [94], an additional level of complexity was observed in the nature of LMT measurements: the study, which assessed peplospheric influences on a number of parameters at the site, evidenced the presence of four wind regime categories affecting LMT data gathering that need to be taken in consideration.

3. Methodologies for the Analysis of Surface and Satellite Data

Carbon monoxide (CO) mole fractions in ppm (parts per million) were gathered by a Picarro G2401 (Santa Clara, CA, USA) CRDS (Cavity Ring-Down Spectrometry) [95,96] analyzer. The same instrument also gathered data on CO₂ (carbon dioxide), CH₄ (methane), and H₂O (water vapor). The principle of CRD spectroscopy allows measuring, with high degrees of precision, the concentration of trace gases in the atmosphere. The G2401 analyzer at LMT gathered data every 5 s with a precision of 1 ppb, and the resulting outputs were aggregated on an hourly or daily basis, depending on the evaluation. More details concerning G2401 measurements at LMT are available in Malacaria et al. (2024) [43] and D’Amico et al. (2024a) [89].

The tropospheric density of CO and HCHO (or FA) in the vertical column were obtained by Sentinel-5P, an ESA (European Space Agency) satellite aimed at global air pollution monitoring launched as part of the Copernicus mission [97]. Sentinel-5P is in a low-Earth afternoon polar orbit yielding a swath of 2600 km and carries TROPOMI (Tropospheric Monitoring Instrument), a device capable of advanced atmospheric monitoring spectrometry. TROPOMI scans have a spatial resolution of 3.5 × 5.5 km and the signal-to-noise ratio is high. Operational Level 2 (L2) products are publicly available for use via the Copernicus platform [98]. Data on daily observed columns were downloaded in netCDF format, and an algorithm set up at CNR ISAC parses through all parameters required for local data analysis at the LMT observatory. The analysis is divided into several steps: coordinates extraction (latitude and longitude); trace gas data analysis and its division into arrays correlated to a map focused on the region of interest; data processing via a filter applying a Q_a > 0.5 condition to the entire set; generation of a georeferenced map showing data filtered from the array; and a direct comparison between select coordinates and satellite measurements nearby, with the selection of the minimum distance to the selected site with respect to the slightest reported distance in the array. Q_a is a data quality indicator whose assigned values range between 0 and 1: applicable manuals recommend a Q_a greater than 0.5 to exclude influences from a number of conditions such as ice/snow warnings, a Solar Zenith Angle (SZA) ≤ 70°, cloud radiance fraction at 340 nm < 0.5, surface albedo ≤ 0.2, air mass factor > 0.1, and common error flags [99,100]. The implementation of the Q_a filter has reduced the original datasets to coverage rates of 72% (carbon monoxide) and 61% (formaldehyde) for the study period. Specifically, the reduction in terms of coverage was 15.09% for CO and 21.25% for HCHO compared to primary satellite products. Formaldehyde is more sensitive to the Q_a filter due to known difficulties at sites affected by higher HCHO levels from biomass burning and anthropogenic emissions, in addition to effects induced by aerosols and albedo calculated by TROPOMI, which in the case of coastal sites such as LMT can interfere with data gathering.

Ground indoor and outdoor measurements of formaldehyde at the site are limited to a summer 2021 campaign at the local INAIL (Italian National Institute for Insurance against Accidents at Work) department located nearby [101]. Continuous measurements of surface HCHO at LMT are not available on a regular basis. The joint INAIL-ISAC campaign demonstrated good agreement between tropospheric column densities of formaldehyde and ground (indoor/outdoor) concentrations. It also showed well-defined correlations between the orientation of rooms and windows, indoor HCHO concentrations, and the main wind corridors in the area. More details on the campaign are available in Barrese et al. (2024) [101].

Black carbon (BC), and specifically equivalent black carbon (eBC) data in micrograms per cubic meter (μg/m³ or μg PCM) were gathered by a Thermo Scientific 5012 (Franklin, MA, USA) MAAP (Multi-Angle Absorption Photometer) [102,103,104] performing measurements on a per-minute basis. In the literature, the term “equivalent black carbon” (eBC) is used to define BC gathered via optical absorption methodologies such as the MAAP used at LMT [105]. More details concerning eBC data gathering at the site are available in Malacaria et al. (2024) [43] and D’Amico et al. (2024b) [93].

Wind speed (m/s) and direction (°) at a near-ground level were measured by a Vaisala WXT520 (Vantaa, Finland) weather station. The instrument also gathers data on relative humidity, pressure, accumulated rain, and temperature. Additional details on WXT520 measurements are available in D’Amico et al. (2024c) [88].

Hourly and daily averages of surface CO, eBC, and meteorological parameters were aggregated in R 4.4.0. Plots were computed in R (packages/libraries: ggplot2, ggpubr, tidyverse, zoo, dplyr) and MATLAB 2016a.

In addition to the main data processing procedures, statistical analyses were performed using Jamovi v. 2.3. These analyses were aimed at hourly data gathered at LMT, and integrated surface–satellite data evaluated based on satellite flybys above the LMT site to ensure that the daily TVC measured from the satellite was correlated with surface measurements performed at the same time. In order to test possible linear correlations, Spearman’s Rank (SR) and Pearson’s Correlation Coefficient (PCC) were calculated [106,107,108].

Spearman’s Rank in particular is calculated by converting two observed variables—x and y—into ranks (R_x and R_y); the SR, defined as ρ, is calculated for a sample of size n as described in Equation (1):

ρ = (cov(R_x,R_y))/(σR_x σR_y)

(1)

where cov indicates the covariance of x and y; and σR_x and σR_y are the respective standard deviations of the evaluated variables. Both the PCC and SR range in the interval between −1 (negative correlation) and +1 (positive correlation), with 0 indicating the absence of a linear correlation.

4. Data Coverage and Evaluations of Wildfire Sources During the 2021 Crisis

Surface measurements at LMT between July and August 2021 are characterized by high coverage rates, shown in

Table 1. Coverage rate per instrument/dataset compared to the total number of hours (1488) and days (62) elapsed between 1 July and 31 August 2021. Q_a refers to the available dataset after the implementation of the Q_a data quality filter.

Type	G2401	MAAP	WXT520	Sat. CO	Sat. FA	Qa CO	Qa FA
Hours	99.66%	98.18%	100%	-	-	-	-
Days	100%	100%	100%	87.09%	82.25%	72%	61%

as percentages compared to the hours and days elapsed during the study period. All days are covered by surface measurements, while gaps of only a few hours, due to instrument calibration and maintenance, are limited. CO and eBC data were further differentiated by wind sector: northeast (0–90° N); southeast (90–180° N); west (240–300° N). The categorization by wind sectors is due to the characteristics of LMT in terms of local wind circulation and its direct impact on surface measurements of a number of parameters [82,87,88,89,90,94,101].

In

Table 2. Average monthly concentrations of surface carbon monoxide and equivalent black carbon, as well as their respective standard deviations (±), first-quartile (Q1), and third-quartile (Q3) values.

Param.	July			August			Total
	Q1	Mean ± SD	Q3	Q1	Mean ± SD	Q3	Q1	Mean ± SD	Q3
CO (ppb)	76.07	102.532 ± 40.12	117.19	119.01	150.002 ± 50.27	166.4	91.64	126.267 ± 51.21	147.50
eBC (μg/m3)	0.13	0.408 ± 0.36	0.56	0.29	0.607 ± 0.41	0.79	0.19	0.508 ± 0.38	0.7

, the monthly averages of surface CO and eBC are shown, including details on data variability during the study period.

With respect to wildfire analysis on a Mediterranean scale, this research study relied on FIRMS productions to assess the locations of major wildfires prior to the evaluation of backtrajectories from the LMT observation site in Calabria. FIRMS (Fire Information for Resource Management System) provides NRT (Near Real-Time) data on active fires from the MODIS (Moderate Resolution Imaging Spectroradiometer) instrument aboard the Terra Aqua platform, and the VIIRS (Visible Infrared Imaging Radiometer Suite) aboard the NOAA 20, NOAA 21, and S-NPP. Generally, gathered data on a global scale are available within three hours from satellite scanning. Thermal anomalies or active fires represent the center of a 1-km pixel that is flagged by the MODIS MOD14/MYD14 Fire and Thermal Anomalies algorithm as containing one or more fires within the pixel [109]. This is the most basic fire product in which active fires and other thermal anomalies, such as volcanoes, are identified. MCD14DL outputs are available in a number of formats (KML, TXT, WMS, and SHP), meant to be used by a variety of programs for data processing. Only the TXT and SHP formats, however, contain all the attributes. Collection 61 (C61) data effectively replaced Collection 6 (C6) in April 2021 [110]. It is worth noting that C61 did not feature substantial upgrades in terms of the algorithm as the updates mostly focused on improving the calibration in Terra and Aqua MODIS Level 1B data products. The analysis of fires and the consequent detection of daily active fires was obtained by FIRMS via sensors mounted on satellites; these sensors generally have swath widths in the 2300–3000 km range, which provide two daily observations for most of the Earth’s surface.

Backtrajectories aimed specifically at three case studies, each representative of a characteristic source of emission during the 2021 crisis, were computed in HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) by NOAA’s (National Oceanic and Atmospheric Administration) Air Resources Laboratory [111,112]. The use of backtrajectories in case studies, combined with additional evaluations, can provide relevant information on specific emission sources [113]. Although HYSPLIT products are generally used for a variety of air pollutants, there are examples in the literature of evaluations aimed specifically at wildfire emissions [45,114,115]. In this study, 72-h backtrajectories at three distinct altitude thresholds (500, 1000, and 3000 m above ground level) were used. The meteorological input used for HYSPLIT models shown in this study was the Global Forecast System (GFS) with a horizontal resolution of 0.25 degrees.

Specific case studies selected for this paper have relied on CAMS (Copernicus Atmosphere Monitoring Service) [116] products of CO to highlight air mass transport and wildfire source attribution. Multi-layer CAMS forecast products are issued multiple times on a daily basis and provide information on fifty species: forecasts are computed via the combination of a previously issued forecast with currently observed satellite data through a process defined as “data assimilation” [117]; the model also relies on estimates to assess the diffusion of chemical/aerosol species in areas with limited coverage of direct measurements [118,119]. CAMS products are frequently used for assessments of atmospheric compositions [120,121,122], including the diffusion of wildfire outputs [123,124]. CAMS products are also used in conjunction with HYSPLIT backtrajectories to enhance the level of detail on the diffusion of particles in the atmosphere [125]. The 2021 wildfire crisis was well-documented by CAMS as the model reported an 8% increase in boreal CO total column [126].

5. Results

5.1. Tropospheric and Surface Observations of CO, HCHO, and eBC

At the LMT observatory in Calabria, Southern Italy, Picarro G2401 and Thermo Scientific 5012 instruments gathered continuous data on carbon monoxide and equivalent black carbon, respectively. Data on surface meteorological parameters (wind direction and speed) have been gathered by a Vaisala WXT520. As shown in Table 1, the coverage rates of all instruments between July and August are very high. Figure 2 shows the hourly averages of carbon monoxide and equivalent black carbon during the observation period.

In addition to hourly averages, daily averages have been evaluated by integrating CO and eBC concentrations with wind data, thus differentiating these averages on a wind sector basis. As described in Section 2, LMT measurements are heavily affected by local wind circulation patterns, and daily averages accounting for all sectors may not be representative of the variability in oscillations from specific wind corridors [88,94]. Daily tropospheric column densities of CO and HCHO, obtained via satellite data, have also been computed. The results are shown in Figure 3.

Figure 4 shows the daily cycles, differentiated by wind sector, of CO and eBC. The graphs show diurnal hours (11:00–16:00) gaps where the northeastern and southeastern winds are absent during the study period.

Considering that satellite flybys above LMT occur at 14:00 UTC and data obtained during the passage are considered a daily total column density of both CO and HCHO, these values have been compared with the surface data of CO gathered at 14:00 UTC by the Picarro G2401 CRDS analyzer at LMT. Restricting daily surface measurements of CO to a time window coinciding with satellite passages is more representative of the correlation between satellite and surface data. The results are shown in Figure 5.

5.2. Statistical Analysis

The findings of this research have been tested using the statistical parameters and methods described in Section 3.

Table 3. Main statistical parameters of surface wind speed (WS), carbon monoxide (CO), and equivalent black carbon (eBC).

	WS (m/s)	CO (ppb)	eBC (μg/m3)
N	1488	1483	1461
Missing	0	5	27
Mean	3.02	127	0.517
Median	2.69	119	0.386
SD	1.70	51.2	0.455
Variance	2.89	2623	0.207
Min	0.500	61.2	<0.003
Max	9.71	515	4.13
1st Q.	1.51	91.6	0.194
3rd Q.	4.31	148	0.705

shows the main statistics of hourly surface observations at LMT.

In

Table 4. Correlation matrix of surface hourly wind speed (WS, m/s), carbon monoxide (CO, ppb), and equivalent black carbon (eBC, μg/m³). Pearson’s Correlation Coefficient (PCC) and Spearman’s Rank (SR) are reported based on their respective level of significance.

		WS	CO	eBC
WS	PCC	–	–	–
	p-value	–	–	–
	SR	–	–	–
	N	–	–	–
CO	PCC	−0.432 ***	–	–
	p-value	<0.001	–	–
	SR	−0.462 ***	–	–
	N	1481	–	–
eBC	PCC	−0.402 ***	0.840 ***	–
	p-value	<0.001	<0.001	–
	SR	−0.468 ***	0.847 ***	–
	N	1459	1455	–

*** p < 0.001.

, a matrix displays the possible correlations between hourly surface data using the PCC and SR methods for linear correlations.

Satellite TVC values have been evaluated in conjunction with surface measurements performed at the time of satellite passages. The main statistical information concerning dual surface and satellite measurements is shown in

Table 5. Main statistical parameters of surface wind speed (WS), carbon monoxide (CO), equivalent black carbon (eBC), TVC CO, and TVC formaldehyde (HCHO) based on the timing of satellite flybys above LMT’s coordinates.

	WS (m/s)	Surface CO (ppb)	eBC (μg/m3)	TVC CO (mol/cm2)	TVC HCHO (mol/cm2)
N	62	62	62	54	50
Missing	0	0	0	8	12
Mean	4.85	110	0.382	1.61 × 1016	1.30 × 1016
Median	4.72	110	0.275	4.06 × 1015	1.04 × 1016
SD	1.30	36.3	0.341	1.51 × 1016	1.19 × 1016
Variance	1.70	1315	0.116	2.29 × 1032	1.41 × 1032
Min	2.48	65.3	0.0543	3.23 × 1014	−2.35 × 1015
Max	8.62	260	2.06	5.59 × 1016	5.59 × 1016
1st Q.	3.89	84.9	0.150	3.44 × 1015	2.92 × 1015
3rd Q.	5.69	126	0.535	2.96 × 1016	2.03 × 1016

.

These data, limited to a well-defined time span, have been statistically evaluated to test for possible correlations. The results are shown in

Table 6. Correlation matrix of surface hourly wind speed (WS, m/s), carbon monoxide (CO, ppb), equivalent black carbon (eBC, μg/m³), TVC CO (mol/cm²), and TVC formaldehyde (HCHO, mol/cm²). Pearson’s Correlation Coefficient (PCC) and Spearman’s Rank (SR) are reported based on their respective level of significance.

		WS	Surface CO	eBC	TVC CO
WS	PCC	–	–	–	–
	p-value	–	–	–	–
	SR	–	–	–	–
	N	–	–	–	–
Surface CO	PCC	−0.497 ***	–	–	–
	p-value	<0.001	–	–	–
	SR	−0.495 ***	–	–	–
	N	62	–	–	–
eBC	PCC	−0.521 ***	0.887 ***	–	–
	p-value	<0.001	<0.001	–	–
	SR	−0.544 ***	0.828 ***	–	–
	N	62	62	–	–
TVC CO	PCC	0.158	−0.081	−0.071	–
	p-value	0.25	0.561	0.608	–
	SR	0.159	0.027	0.099	–
	N	54	54	54	–
TVC HCHO	PCC	0.080	−0.095	−0.110	−0.218
	p-value	0.582	0.513	0.445	0.165
	SR	0.061	−0.108	−0.113	−0.249
	N	50	50	50	42

*** p < 0.001.

.

5.3. Case Studies: Sardinian, Algerian, and Greek Wildfires

As described in Section 1, the Mediterranean wildfire crisis of summer 2021 affected several countries and caused enormous environmental damage. Analyzing two continuous months and surface and satellite data, three case studies (CS1, CS2, and CS3) have been selected for additional evaluations. These case studies go beyond the characterization of Calabrian open fire emissions described in Malacaria et al. (2024) [43], which focused on the Aspromonte Massif area wildfires that affected the southernmost sector of the Italian peninsula in August 2021.

The 10–11 July peaks (CS1) observed at Lamezia Terme (LMT) could be attributable to the vast wildfires that occurred in Calabria itself, the Italian regions of Sicily, Apulia, and Sardinia, and northern African regions in Tunisia and Algeria, as shown by the FIRMS data (Figure 6) and related estimates on the number of days of prolonged fire activity in these areas. Therefore, the total column densities of CO and HCHO have been analyzed and the results are shown in Figure 7. Via the HYSPLIT model, backtrajectories have been computed using LMT’s coordinates as the starting point (Figure 8), indicating that the sources of CS1 peaks at LMT are in fact attributable to Sardinian wildfires. CAMS products (Figure 9) also show a diffusion pattern compatible with this hypothesis.

The second case study (CS2) is aimed at the peaks observed between 29 July and 1 August. Algeria and the southern Italian regions of Calabria and Sicily were affected by a significant number of prolonged wildfires. FIRMS data on the week elapsed between 28 July and 3 August show several wildfires affecting these areas (Figure 10). The tropospheric column densities computed on the 29–31 July data, shown in Figure 11, and the HYSPLIT backtrajectory, shown in Figure 12, both indicate the northern African country of Algeria as the most probable source of these emissions. CAMS outputs shown in Figure 13 are also consistent with this hypothesis. Just like in CS1, air mass transport in this case was dominated by large-scale westerly flows.

The third case study (CS3) focused on 4 August, which provided evidence of an eastern source such as Greece and Türkiye, which—as described in Section 1—were both heavily affected by the wildfire crisis at the time. FIRMS data on the entire week between 31 July and 6 August show vast areas in both countries being affected by wildfires (Figure 14). The Italian regions of Calabria and Apulia were also affected by prolonged wildfire activity. Tropospheric column data, shown in Figure 15, show relevant Greek peaks in density. One of the HYSPLIT backtrajectories (Figure 16) centered at LMT is compatible with the surge in CO and HCHO emissions reported in Greece at the time. CAMS products related to CS3 are shown in Figure 17.

6. Discussion

Data on the summer 2021 wildfire crisis in the Mediterranean Basin have been subject to new analyses aimed at an enhanced characterization of carbon monoxide (CO) and formaldehyde (HCHO) observations, with the addition of equivalent black carbon (eBC) as an additional tracer of combustion processes. At the WMO/GAW (World Meteorological Organization–Global Atmosphere Watch) regional site of Lamezia Terme (code: LMT) in Calabria, Southern Italy, surface observations of CO and eBC have been integrated with tropospheric total column data on CO and HCHO to assess the influence of wildfires over air quality in the central Mediterranean during the peak of that crisis in July and August 2021. This evaluation considers a wider area (i.e., the broader context of the Mediterranean Basin) compared to a previous study aimed at local Calabrian wildfires that heavily affected the region in 2021, especially the Aspromonte Massif National Park (≈90 km S-SW from the observation site), which is the southernmost part of continental Italy [43]. The previous work focused on the peaks attributable to the Aspromonte environmental crisis, while in this research, more remote sources (e.g., Sardinia, Greece) have been considered. Furthermore, the entire month of July 2021 is also evaluated in this study, while the previous work compared the monthly averages of August 2017–2021 to assess indicators of variability through multiple seasons affected by wildfires.

Satellites can provide column-integrated values (e.g., total vertical columns of CO or HCHO), while ground-based measurements can focus on surface concentrations. Comparing these datasets provides information on the interactions between near-surface and tropospheric transport over wide areas. The two methodologies are characterized by distinct measurement frequencies: surface data are generally continuous and can be gathered multiple times per minute; satellite data normally occur once or twice per day for a particular location on the Earth’s surface. In terms of spatial resolution, however, satellite measurements are more efficient than their surface counterparts, which provide punctual information instead.

LMT’s location in the Mediterranean (Figure 1) exposes the observation site to Saharan dust events [81] and wildfire inputs from multiple sources: in the previous study, releases from wildfires affecting the Aspromonte Massif in August 2021 were observed via a multiparameter approach. The findings of the previous study demonstrated how open fires can impact air quality on a regional level [43].

In this study, a longer observation period has been selected between 1 July and 31 August of the same year: at the LMT station, this period is characterized by an excellent coverage rate of surface CO and eBC, with instruments nearing 100% in terms of coverage rates, thus allowing for a detailed analysis of all data gathered during the period (Table 1). The monthly averages of surface CO and eBC both show August peaks (Table 2). Considering both July and August, CO (126.267 ± 51.21 ppb) and eBC (0.508 ± 0.38 μg/m³) have shown different degrees of variability, with eBC’s standard deviation encompassing a greater interval around the average value compared to CO. In the atmosphere, black carbon is characterized by a short lifetime [76,77], and the greater variability at LMT is interpreted by notable changes in the detection of outputs during the study period. The observed CO, which has a longer atmospheric lifetime, is influenced by cumulative outputs from various wildfires in the Mediterranean Basin and, therefore, shows a lower variability. These differences in behavior highlight the spatial and temporal influences of wildfire emissions in terms of air quality over a wide area.

Using hourly averages as a reference (Figure 2), the entire observation period has been assessed with respect to surface CO (ppm) and eBC (μg/m³ or μg PCM), which both provided new insights on possible wildfire outputs—other than those already described in Malacaria et al. (2024) [43], which accounted for the 10–12 August peaks linked to wildfires located in the Aspromonte Massif in Calabria. The evaluation of hourly averages pinpointed a number of circumstances with high peaks.

Using daily averages and integrating the tropospheric total column densities of CO and HCHO, a cross-analysis of surface and column data was possible (Figure 3). The cross-analysis allowed for the determination of the difference between high-altitude air mass transport from remote sources and nearby surface/near-surface outputs. The LMT coastal site is affected by local wind circulation patterns (Section 2), which have a direct impact on atmospheric measurements: without a proper differentiation on a wind sector basis, daily averages may not be considered representative of the actual variability at the LMT regional station [88,94]. In Figure 3B,C, data have been differentiated by wind sector, which could provide more precise information on the source of observed peaks. The differentiation was also applied to daily cycle analyses (Figure 4) and highlighted the presence of gaps, i.e., hours, that were not covered by specific combinations of wind direction and the surface concentration of CO and eBC.

A comparison between daily averages alone, however, cannot be deemed sufficient to compare surface and tropospheric column data. In the case of CO—which is the only evaluated parameter measured continuously from both surface analyzers and satellite sensors—the daily average calculated from surface data is not directly comparable to the total tropospheric column density observed by satellite scans. Considering that these scans occur at 14:00 UTC, the surface hourly averages of CO measured during the same time window have been compared with column data in Figure 5. Although divergences are present—each indicating near-surface peaks that were not distributed on the tropospheric column and, vice versa, high-altitude air mass transport that did not result in surface concentration peaks—the data shown in Figure 5 indicate an increase in CO between July and August 2021. Formaldehyde, which is not subject to continuous surface measurements at LMT, also shows a trend compatible with carbon monoxide variability. It is worth noting that tropospheric column densities are not equally susceptible at all altitudes as the employed instruments are more sensitive to surface and near-surface concentrations: in the case of CO, this feature enhances the correlation between tropospheric column and surface data.

Surface measurements (Table 3) gathered on an hourly basis have been statistically evaluated to test possible linear correlations. The analysis was based on PCC (Pearson’s Correlation Coefficient) and SR (Spearman’s Rank) in addition to the p-values as indicators of statistical significance. The results shown in Table 4 indicate that CO and WS yield a statistically significant (p-value < 0.001) negative correlation (PCC = −0.432; SR = −0.462). WS and eBC also display a similar pattern (PCC = −0.402; SR = −0.468; p-value < 0.001). These findings are consistent with the precipitation of compounds at LMT especially in the occurrence of wind inversion, as described in D’Amico et al. (2024c) [88] and D’Amico et al. (2024e) [94]. As expected, the correlation between CO and eBC is positive (PCC = 0.840; SR = 0.847), a finding that further corroborates the hypothesis by which—as observed in Malacaria et al. (2024) [43]—measured peaks of CO and eBC are attributable to open fire outputs.

A similar approach was aimed at dual surface–satellite measurements (Table 5) by restricting surface data to measurements compatible with the timing of satellite flybys at 14:00 UTC. When considering 14:00 UTC (the time of satellite flyby), the tested correlations show variations in terms of statistical significance and linear correlation factors (Table 6). Surface CO and WS have a stronger negative correlation than before (PCC = −0.497; SR = −0.495; p-value < 0.001). Similarly, WS and eBC also yield a stronger negative correlation (PCC = −0.521; SR = −0.544; p-value < 0.001). When compared to satellite data, however, both TVC CO (PCC = 0.158; SR = 0.159; p-value = 0.25) and TVC HCHO (PCC = 0.080; SR = 0.061; p-value = 0.582) do not show a correlation with WS, whether positive or negative. In the case of CO, this is important due to the dual surface–satellite nature of its measurement: the precipitation known to occur at LMT under the circumstances described above (e.g., wind inversion patterns) is influenced by surface and near-surface wind speeds, not by TVC density. The influence of local wind patterns on LMT surface observations is further corroborated by the CO-eBC relationship, which is still high at 14:00 UTC (PCC = 0.887; SR = 0.828; p-value < 0.001). Surface CO and eBC do not correlate with TVC CO and HCHO: p-values are in the 0.445–0.608 range, thus, greater than the 0.05 value required to verify statistical significance. This is also in accordance with local wind patterns dominating the near-surface concentrations observed at LMT of air masses carrying combustion products that would otherwise be subject to transport at higher altitudes. TVC CO-HCHO show a slightly negative correlation (PCC = −0.218; SR = 0.249), which is not statistically relevant (p-value = 0.165).

CO and HCHO are both tracers of combustion (Section 1); however, formaldehyde is characterized by a lower persistence time in the atmosphere due to chemical and photochemical processes [127], thus making HCHO a proximity indicator. Air mass transport from remote wildfires would, therefore, be significantly depleted in HCHO.

The variability in CO and HCHO during the study period (July–August 2021) and related peaks have allowed for the identification of three case studies (CSs); via FIRMS—daily tropospheric column data, surface measurements, and HYSPLIT backtrajectories—each CS has been constrained to a probable source of emission in the Mediterranean Basin.

In the process of evaluating CS1, FIRMS data showed a significant number of wildfires active for 2+ days in Calabria itself, and the Italian regions of Sardinia, Sicily, and Apulia (Figure 6). TVC data on CO and HCHO between 9 and 12 July, subject to availability limitations (CO data not available on 12 July and HCHO data not available on the 9 and 11), highlight the presence of Calabrian and Sardinian wildfires in particular (Figure 7). At the LMT observatory, this period was already known to be characterized by CO and eBC concentrations above seasonal and yearly averages, thus indicating wildfires as possible sources [43]. Using the computed HYSPLIT backtrajectory shown in Figure 8, the peaks of CS1 are attributable to Sardinian wildfires: at the site, these trajectories are linked to western wind corridors, which are generally depleted in GHGs and pollutants [89]. However, the perturbation of background atmospheric levels caused by wildfires and the wind inversion processes linked to the daily cycle at LMT [88,93] are both compatible with westerly winds for CS1.

The HYSPLIT backtrajectory in particular shows the transport of air masses toward LMT from the west-northwest, a direction compatible with Sardinian sources of emissions. Due to its nature as a coastal site, and considering specifically the western sector, from LMT there are no obstacles in that direction for hundreds of kilometers: prevailing winds from that direction, and the analysis of satellite data (Figure 7), indicate that the transport from Sardinia to Calabria of wildfire outputs has occurred. Peaks in surface CO and eBC, seen in Figure 2, are linked to this CS; specifically, an hourly peak of 214.69 ppb for CO and 4.12 μg/m³ for eBC have been observed. The CO peak constitutes the highest value reported in the entire month of July, while the eBC peak is very similar to the August peak described in Malacaria et al. (2024) [43], attributable to local wildfires in the Aspromonte Massif area. CAMS maps of CO (Figure 9) show that wildfires that occurred in northern Sardinia resulted in the release of combustion byproducts and the consequent transport toward the east in the direction of continental Italy. In addition to westerly winds driving the transport to the peninsula, a northwestern component of air mass transport leads air masses yielding high CO toward the Tyrrhenian coasts of Calabria, and ultimately LMT. CAMS modeling is, therefore, in accordance with the HYSPLIT backtrajectories shown in Figure 8 and identifies a phenomenon of air mass transport from Sardinia to Calabria.

A similar circumstance is observed in CS2, where FIRMS data indicate wildfire activity in North Africa, specifically in Algeria (Figure 10), which is known to be susceptible to wildfires [128,129]. The peaks in CO and HCHO from Algerian wildfires are confirmed by TVC data (Figure 11), and the computed HYSPLIT backtrajectory also results in a western air mass transport observed at LMT (Figure 12), which culminated with the precipitation to surface levels of pollutants at high altitudes, linked to daily cycle wind circulation inversions at LMT [78,79]. In fact, surface measurements of both CO and eBC linked to CS2 show a number of peaks (Figure 2): CO reached an hourly peak of 468.65 ppb during wind corridor inversion, a well-described circumstance at the LMT site that causes the precipitation of pollutants [88,94]. eBC reached a peak of 2.94 μg/m³, which is not among the highest reported during the observation period. CS2 measurements are characterized by low wind speeds (<3 m/s), which are typical for local wind inversion patterns and contribute to the precipitation of pollutants. The integration of CAMS modeling in CS2 (Figure 13) provides additional information on the transport of air masses enriched in CO due to biomass combustion processes. Plumes carrying higher concentrations of CO move toward the northwest, in the direction of the Italian peninsula. On 29 July at 18:00 UTC, an increase in CO concentrations was reported in northern Algeria, as previously indicated by FIRMS (Figure 10). These air masses move toward the northwest in the following hours. A diffuse area of elevated CO concentrations is visible stretching across the central region of the map. The highest concentrations appear localized in a narrow corridor, especially in the central region, and air masses enriched in CO arrive in central Italy. Surrounding regions show lower concentration levels, indicating relatively clean air in those areas, closer to background CO levels. A shift toward the southeast is observed on July 30th. A general reduction in CO concentrations is noticeable in the northwestern sectors, suggesting possible dispersion or atmospheric processes reducing concentrations in that area. Consequently, the CO concentration field appears to redistribute, with significant increases in the central and southeastern sectors of the area, suggesting the potential for transport and accumulation of CO in the Tyrrhenian sector and—in particular—the LMT site. Red patches on the CAMS maps point to a localized emission source or a stagnation effect enhancing CO in specific regions.

In CS3’s case, FIRMS data confirm very heavy wildfire activity in central Greece and in several regions across Türkiye in early August (Figure 14). Tropospheric column data of CO and HCHO highlight, especially for August 4th (Figure 15), significant peaks in column density in areas that were significantly affected by wildfires in central Greece at the time (Section 1). However, the analysis of backtrajectories via HYSPLIT (Figure 16) has shown once again a westerly wind corridor linked to surface LMT observations via a clockwise pattern instead of a northeastern wind corridor. Surface LMT detections are, in CS3, also correlated with inversions in wind circulation caused by the local daily cycle and the precipitation of pollutants from higher altitudes [88,94]. CO shows a peak of 259.76 ppb, while eBC peaks at 2.41 μg/m³; these values do not constitute absolute peaks within the context of the study period, although they are characterized by a peculiar backtrajectory. CAMS modeling maps of CS3 (Figure 17) allow for highlighting the evolution of CO concentration increases linked to Greek wildfires and the consequent clockwise transport toward LMT. A plume enriched in CO originates from the area affected by intense wildfires; at 18:00 UTC, a significant increase in CO concentrations near Greece is observed, suggesting more intense fire activity or stagnant conditions trapping emissions. CO plumes are more visible and moving toward central Mediterranean sectors, likely influenced by easterly or south-easterly winds. In this case, HYSPLIT backtrajectories (Figure 16) in this case better highlight the occurrence of a clockwise air mass transport that culminates at LMT. As it progresses in a clockwise motion, the plume becomes more elongated, possibly driven by a stronger eastern component of these winds. The progression inferred from combined HYSPLIT and CAMS models suggests that fire emissions in Greece significantly influenced air quality over Lamezia Terme in early August 2021. The exact local impact depends on boundary layer conditions, vertical mixing, and the ground-level diffusion of pollutants. Locally, mechanisms are known to further increase surface and near-surface concentrations of pollutants at LMT [88,94].

Overall, the findings shown during the evaluation of all three case studies demonstrate the complexity of wildfire detections from surface observations in the central Mediterranean as air mass transport frequently combines with local phenomena and influences the impact of wildfire emissions on the surface. Peaks in the concentrations of pollutants have direct consequences on local air quality and contribute to the total environmental damage caused by wildfire emissions. This study shows, in particular, the effects of local wind circulation patterns on the increase in pollutant concentrations close to the surface: without the local wind inversion cycle at LMT, the peaks linked to CS1 through CS3 would have likely lacked a surface counterpart of tropospheric density peaks. These evaluations highlight the importance of punctual analyses of wildfire emission outputs and their consequent impact on air quality and the environment over wide areas.

7. Conclusions

By expanding the findings of a previous study on the summer of 2021 wildfire crisis in the Mediterranean via the implementation of additional methodologies and a longer study period (July–August), the effects of wildfires on atmospheric measurements performed at the WMO/GAW (World Meteorological Organization–Global Atmosphere Watch) regional site of Lamezia Terme (LMT) could be characterized.

Surface measurements at the site of carbon monoxide (CO) are characterized by a very high coverage rate, with nearly 100% of the study period covered by ground-based measurements. Surface equivalent black carbon (eBC) concentrations, although not classified as gaseous, have been used as an effective tracer of wildfire emissions and also have a very high coverage rate in terms of ground-based measurements. Surface findings have been correlated with tropospheric total column densities of CO and formaldehyde (HCHO) to observe the air mass transport of wildfire byproducts in the central Mediterranean during the crisis of 2021. At the observation site of LMT, HCHO is not subject to continuous measurements. Statistical parameters (Pearson’s Correlation Coefficient, PCC, and Spearman’s Rank, SR) have been calculated to test correlations between surface measurements and, where applicable, dual surface–satellite data. Combustion tracers correlate well, further corroborating the hypothesis that observed peaks are attributable to biomass burning; wind speed has shown a negative correlation, which is consistent with wind inversion patterns at the site and the precipitation of combustion products from higher altitudes.

Surface measurements at LMT have pinpointed peaks that resulted in the evaluation of three distinct case studies (CSs). The analysis of these CSs—which integrated a previous research study on Calabrian wildfires that struck the Aspromonte Massif in the southernmost area of the Italian peninsula—has revealed a number of possible sources of the observed surface concentration peaks of CO and eBC. The sources of emissions are heterogeneous and highlight the variability in air mass transport in the region. Specifically, the Italian region of Sardinia, northern Algeria, and central Greece have been pinpointed as the wildfire emission sources of the evaluated case studies via the integration of surface measurements with TVC and HYSPLIT backtrajectories. CAMS products by Copernicus have also shown well-defined diffusion patterns of combustion tracers. Although they were also affected by heavy wildfire activity, the Italian regions of Sicily and Apulia, as well as Türkiye, are apparently not linked to any of the peaks observed by LMT between July and August 2021. Sardinian and Algerian outputs have been subject to air mass transport from the west, as shown by computed backtrajectories. Central Greek outputs have been detected at LMT from the western-seaside wind corridor instead of the eastern-continental sector, thus demonstrating the importance of the role played by wind circulation in the diffusion of wildfire emissions. In this case, as well as in the other two case studies, surface peaks have been influenced by local wind inversion patterns and the consequent precipitation of pollutants, which are known in the literature to be peculiar characteristics of the LMT site due to the configuration of the Catanzaro isthmus where the station is located.

These findings provide a tangible example of several factors that interplay with each other and add more degrees of complexity to the diffusion of wildfire outputs in the Mediterranean Basin. Although the primary focus with respect to wildfires is the effective containment of the damage caused to the environment, enhanced analyses of wind circulation and the broader impacts of wildfires on air quality can also be assessed to issue ad hoc warnings and mitigate the risks associated with them.