The automatic sun photometer network AERONET was founded by NASA and PHOTONS (Lille University of Science and Technology, National Centre for Space Studies (CNES), and the National Center for Scientific Research of France (CNRS)) and includes several hundreds of sites over the world. The basic principle of the network is to standardize the equipment, measurement techniques, and data processing, which are stored in a freely accessible centralized database. The description of instruments and measurements procedures, calibration of sun photometers, accuracy of measurements, and terminology are described in and in the documentation from the AERONET web page (http://aeronet.gsfc.nasa.gov/). According to the AERONET classification, all observed data are divided into three levels: Level 1.0 is primary unscreened data; Level 1.5 is cloud-screened data; and Level 2.0 is the highest accuracy data, cloud-screened, quality-controlled, and corrected for sun photometers recalibration. We used Level 2.0 data in our research. The spectral AOD in the atmospheric column over the observational site is determined from direct solar irradiance measurements with errors of ±0.01 in the visible and near-infrared regions of the spectrum and with a larger uncertainty (±0.02) in the ultraviolet band . The Ångström exponent (AE) is also determined by the AERONET algorithm for the sun photometer spectral range 340–870 nm from direct sun irradiance measurements. We used AE to interpolate the AOD on the required wavelength. We applied AE determined for 440–870 nm because it is suitable to the aerosol size distribution during wildfires when relative dominance of the fine-mode particles takes place .

To assess the extent of the impact of wildfires in summer 2010 we used data from the following Eastern European AERONET sites (also shown in Fig. ): Minsk (Belarus), Moscow (Russian Federation), Toravere (Estonia), Belsk (Poland), Moldova (the official name of the site is Moldova, although it is located in Chisinau, Moldova), Cluj-Napoca, Bucharest, and Eforie (all Romania). We also used data from the only two Ukrainian sites that measured during summer 2010: Kyiv and Sevastopol.

To analyze aerosol dynamics over Ukraine during the Russian wildfires in summer 2010, we used AERONET data mostly from the Kyiv site. Measurements of aerosol properties over Kyiv by AERONET sun photometers started at the end of March 2008 . The Kyiv PHOTONS/AERONET site is located at the Main Astronomical Observatory of the National Academy of Sciences of Ukraine in the Golosiiv forest located in the southern part of Kyiv, approximately 10 km from the city center. The surrounding landscape around the Kyiv site enables the sun photometer to scan the entire celestial hemisphere. Data are sent to AERONET database twice a day and are accessed and displayed on the Kyiv AERONET website on the same day. Up to 2013 the site was equipped with CIMEL CE-318-2 sun photometer polarized model with filters of 440, 670, 870, 936, and 1020 nm. We obtained AOD 500 nm data for the Kyiv site by data interpolation using Ångström formula in the range of 440–870 nm. Similarly, we obtained AOD 550 nm for the Kyiv site by data interpolation using Ångström formula in the range of 440–675 nm. We calculated AOD values at different wavelengths to be further consistent with measurements, performed by different types of sun photometer and MODIS satellite instrument.

Two MODIS instruments are installed on board the Terra (EOS AM-1) and Aqua (EOS PM-1) satellites. Terra (http://terra.nasa.gov/about/) flies in a sun-synchronous, near-polar, circular orbit with an inclination of 98.5∘ at an altitude of 705 km every 98 min (16 orbits per day). The local equatorial crossing time is approximately 10:30 in a descending node. The Aqua satellite belongs to the A-Train constellation (http://atrain.gsfc.nasa.gov). The local equatorial crossing time is approximately 13:30 in an ascending node of a sun-synchronous, near-polar, circular orbit. The MODIS instruments supply data used to study the Earth's surface and atmosphere from local to global scales. Over land, aerosol properties are retrieved from spectral channels 0.47, 0.66, and 2.12 µm. One of the primary aerosol products of the MODIS algorithms is the AOD at 550 nm in the atmosphere over land and ocean .

To estimate atmospheric pollution over Ukraine caused by aerosol from wildfires, we used AOD 550 nm retrieved by the land algorithm and collected in the MODIS Aqua and Terra Level 2 Collection 005 and 051 Optical_Depth_Land_And_Ocean product file (see https://modis-images.gsfc.nasa.gov/_docs/ATBD_MOD04_C005_rev2.pdf, https://modis-images.gsfc.nasa.gov/MOD04_L2/format.html). The documentation of the MODIS Level 2 Collection 051 Optical_Depth_Land_And_Ocean dataset does not recommend the application of these data for quantitative analysis. Nevertheless we applied these data because the alternatively recommended dataset Corrected_Optical_Depth_Land contains only a very small amount of data in the regions for the period we considered in this study. Earlier studies based on the Optical_Depth_Land_And_Ocean product showed good agreement between AOD MODIS (quality assurance confidence (QAC) flags 1, 2, and 3) and AERONET on a global scale . Local biases for special cases of aerosol pollution, such as wildfires, were reported by . At this stage we applied all data over the land with QAC flags 1, 2, and 3 from the dataset without any additional filtering. For comparably clean atmospheric conditions, when AOD is close to zero, AOD values within ±0.05 are practically indistinguishable . Following , we set corresponding data within this range to zero.

We compared MODIS Collection 5 data with AERONET data from the Kyiv site to further examine potential regional peculiarities in the satellite measurements. To define spatial collocations we computed distances and azimuth angles between the centers of each MODIS image pixel and the location of the Kyiv AERONET site. To compare AOD from MODIS and AERONET we averaged the MODIS AOD over the pixels area centered on the Kyiv AERONET site. We did not calculate the spatial biases of the AOD over this area. This simplification of the procedure is acceptable because MODIS images of land do not exceed an area of 50 × 50 km, which is significantly less than the characteristic dimension of inhomogeneities in the spatial distribution of aerosol in the atmosphere . In addition, if several AERONET AOD measurements were available, we chose only one, performed at the closest time to the satellite measurement (not more than 30 min).

The MODIS mission also provides fire mapping on a land surface, which we used to evaluate the activity of fires and their spatial distribution, similarly to . For detecting fires and defining their characteristics, the MODIS data processing algorithm uses two spectral channels 4 and 11 µm;. Combustion temperature (in K) is defined from the measurements of the spectral brightness of the flame, applying the Stefan–Boltzmann law. An algorithm is applied to calculate the number of burning cells and the total area occupied by fires. Data of the spatiotemporal distribution of fires are freely available online from NASA (e.g., https://firms.modaps.eosdis.nasa.gov/map/). To visualize the spatial distribution of fires we used data supplied by the University of Maryland (ftp://fuoco.geog.umd.edu). We applied high-confidence (confidence level of calculated fire pixels is larger than 80 %) brightness temperature of fire pixels (at band 21). More information about the algorithms of fire recognition, principles of archiving, and distributing data can be found in , , and . MODIS sometimes detects volcanic eruptions or flares of gas in addition to vegetation fires. Here, we considered all observed signals as wildfires.

CALIOP is a two-wavelength (532 and 1064 nm) polarization lidar providing high-resolution vertical profiles of aerosol and clouds. It is installed on board the NASA Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite, which belongs to the A-Train constellation. A detailed description is provided on its official website (http://www-calipso.larc.nasa.gov/about/) as well as in and . CALIOP uses three receiver channels: one measures the 1064 nm backscatter intensity and two channels measure orthogonally polarized components of the 532 nm backscattered signal. The laser beam is directed almost at nadir with a slight forward tilt in the direction of motion of the satellite to avoid direct reflection of laser radiation from high-reflectivity objects (surface water, snow, etc.). The divergence of the transmitted laser beam equals 100 µrad. This results in a footprint on the Earth's surface, called in our paper ground track, of about 70 m. A pulse repetition rate of approximately 20 Hz provides vertical resolution of 15 m.

CALIOP measurements allow to derive the vertical distribution of aerosol and clouds. The corresponding AOD is determined by extinction coefficient integration over altitude . For this study we used both parameters, vertical distribution of the extinction coefficient and AOD at 532 nm, defined along the path of the sub-satellite point. We applied Level 2 Cloud and Aerosol Layer and Profile products V 3.01 and 3.02 with a resolution of 5 km on the surface along the sub-satellite point.

A comparison of CALIOP AOD with ground-based AERONET observations can be challenging because of different measurement characteristics of both instruments. The CALIOP lidar provides only fragmentary data on aerosol along CALIPSO satellite's ground track due to the small size of its light beam and cloudy conditions that frequently occur. As long as the instrument's orbital period lasts 98 min, ground tracks of satellite consecutive passages at certain latitudes are shifted 24.5∘ to the west, making its spatiotemporal coverage rather sparse. Consequently, the probability of CALIOP to pass over the atmospheric column observed by the solar photometer AERONET is rather limited e.g.,. During the three summer months of 2010 we found no coincident CALIOP and AERONET measurements over Kyiv, apart from the single collocation, although not exactly matching the selection criteria according to . In this particular case the closest CALIPSO ground track was found 60 km eastwards from the Kyiv AERONET site.

Therefore, in this study we only compared CALIPSO/CALIOP Cloud Aerosol Layer Product AOD 532 nm with Aqua MODIS AOD 550 nm since the orbits of both satellites are in the A-Train constellation (http://atrain.nasa.gov/). The relative positions of CALIPSO and Aqua satellites in the A-Train provide a large number of practically simultaneous measurements with the time span of 2 min, while the spatial difference is only about 10 km. Each granule of MODIS data consists of consecutive scans across the satellite track. The footprint of CALIOP light beam on this granule looks like a sequence of points on the straight line, which are passing close to the center of a granule. Each of these points represent the center of the CALIOP measurement averaged over 5 km, which matches with one of the pixels of a MODIS granule. To find these matches we calculated the distances and azimuth angles between the center of each CALIOP point and the center of each pixel in MODIS granule in the same manner as for the MODIS–AERONET case, described in Sect. 2.2. We averaged MODIS data over areas 50 km × 50 km, while CALIOP data in the CALIPSO Cloud Aerosol Layer Product are averaged on various distances along the satellite ground track, up to 80 km (see CALIPSO Quality Statements: Lidar Level 2 Cloud and Aerosol Layer Products, Version Releases: 3.01, 3.02). We did not apply any correction for potential spectral differences while comparing CALIOP AOD 532 nm and Aqua MODIS AOD 550 nm. It yields to an estimated systematic bias in our AOD comparison of approximately 2–6 % in the AE range between 0.5 and 1.8 (see Fig. 5b) and can be neglected in our cases, following .

To analyze the impact of weather conditions on the distribution of aerosol in the atmosphere during summer 2010, we used weather charts from the Ukrainian Weather Service of isobaric surfaces representing altitudes of 2 m, 1.5, 3, and 5 km (not shown in this paper). We also considered weather charts at 500/1015 hPa of the Global Forecast System (GFS; http://www.wetterzentrale.de) from the National Centers for Environmental Prediction (NCEP) and merged global and European model (GME; http://www.wetter3.de) from the German Weather Service.

To confirm the analysis of weather charts, we also calculated trajectories of air masses using the HYSPLIT model (http://ready.arl.noaa.gov/HYSPLIT.php), developed by the National Oceanic and Atmospheric Administration's (NOAA) Air Resources Laboratory . Furthermore, to study the aerosol inflow towards AERONET sites shown in Fig. 1, we computed back trajectories of air masses at different heights for dates of maximum measured AOD at each of those sites. Back trajectories were simulated for 168 h (7 days) at altitudes of 500 m, 1.5, 3, 4, and 5 km at 12:00 GMT starting time. We chose the lowermost 5 km tropospheric altitudes, taking into account the analysis of the vertical distribution of aerosol according to CALIOP data (shown in Sect. 3.3.2). According to , the uncertainty of calculated HYSPLIT trajectories for a period longer than 24 h is around 20 % in the horizontal direction in the free troposphere; after 120 h the uncertainty increases to about 400 km in the horizontal and about 1300 m in the vertical planes.

To analyze the impact of wildfires on aerosol loading, we plotted distributions and brightness temperature of the fire pixels over the ETR and Eastern Europe (40–65∘ N and 10–60∘ E) in summer 2010 as observed by MODIS (Fig. 2). Each map accumulates fire pixels over 10-day periods from 1 June to 31 August, except Fig. 2i, which covers a 12-day period during 20–31 August. Black stars and numbers in Fig. 2 indicate the position of the AERONET stations. Figure  shows time series of the total number of fire pixels per day in the same area as in Fig. . Figure 3 indicates that the fire activities increased from mid-July throughout Eastern Europe. The largest number of fires was observed in the ETR, Ukraine, and Moldova. Both the overall number and the brightness temperature of fires reached their maximum between 26 July and 18 August. The largest number of fires, as well as the largest brightness temperature, were observed on 29 July (Figs. f and ). Such large fire areas, high brightness temperature, and long duration were caused by specific weather conditions over the ETR and Eastern Europe during the second half of summer 2010. This favored the accumulation of aerosol in the atmosphere over these regions .

Ground-based AERONET observations showed an increased aerosol content over the Eastern European sites during July and August 2010. We analyzed changes in AOD at 500 nm using all daily averaged measurements from 1 June to 31 August from the AERONET database for Minsk (Belarus), Moscow (Russian Federation), Toravere (Estonia), Belsk (Poland), Moldova (Chisinau/Moldova), Cluj-Napoca, Bucharest, Eforie (all Romania), and Kyiv and Sevastopol (both Ukraine). The analysis revealed significant temporal dynamics of the aerosol content with pronounced peaks during 15–17 August in Kyiv, Sevastopol, Minsk, and Bucharest (see Table ). Also, the highest AOD values were observed in Romania (Cluj-Napoca and Eforie sites) on August 1, in Moscow and Toravere on 7 August, and in Moldova and Belsk in July (observations in Moldova from 25 July until October were not available). The highest air pollution among all analyzed sites caused by aerosol was detected in Moscow in the first half of August due to the proximity of the most active fires e.g.,.

To analyze the aerosol inflow towards AERONET sites, we calculated HYSPLIT back trajectories for dates of maximum AOD from Table  (Fig. ). Our analysis of back trajectories revealed that air movements in the lower 5 km layer of the troposphere corresponded to anticyclonic circulation, which is seen in Fig. a–j at various altitudes as clockwise-shaped curves. The maximum AOD values from ten AERONET sites in the ETR and Eastern Europe were formed under conditions of air stagnation and accumulation of contaminants .

Back trajectories for Moldova and Belsk, where the maximum AOD was observed the earliest in summer 2010 (13 and 16 July, respectively), are shown in Fig. a and b. The trajectories indicate that aerosol was transported to Moldova at altitudes from 0.5 to 1.5 km from the fires in the ETR and southeast of Ukraine (see also Fig. d, e). Into the region of Belsk, aerosol was transported across continental Europe (1.5–5 km) mostly from the Atlantic Ocean, but also from the Baltic across regions of active fires (Fig. ) in the lower atmosphere (500 m).

Transport of aerosol to two Romanian sites (Cluj-Napoca and Eforie) with maximum AOD observed on 1 August (Fig. c, d) also occurred in the lowermost 1.5 km layer, originating from the southeast of Ukraine and Moldova (also the area with active fires). Back trajectories for Moscow and Toravere with a maximum AOD on 7 August are shown in Fig. e and f, respectively. Aerosol to Moscow was transported mostly from the surrounding regions with the most active fires. Air masses over Toravere (Fig. f) originated from Asian regions and crossed areas of active fires in southeastern Ukraine in all analyzed altitudes (Fig. g).

To Kyiv, where the AOD maximum was observed on 15 August, aerosol was transported in the lower 4 km layer from the most active fires in the ETR, Ukraine, and Moldova (Figs. g, h). On 16 August, the maximum was recorded in Sevastopol on the Black Sea coast, where air masses traveled in almost the entire range of analyzed heights (500 m–5 km) from the ETR and Kazakhstan through the territory of active fires in the southwest of Russia (Fig. h). In Minsk and Bucharest the maximum AOD was observed 1 day later on 17 August (Table , Fig. i, j). Towards Minsk aerosol was transported at 3 km from Kazakhstan across the Caspian Sea, southern Russia, and Ukraine, where the active fires were observed; at 1.5 km from Ukraine; and at 500 m from the western regions of the ETR through Ukraine. Aerosol from fires was transported to Bucharest at 500 m from the northeast, specifically through the ETR and the southeast of Ukraine.

According to the monthly weather reports of the Ukrainian Weather Center, a change in weather was observed on 18–21 August. Atmospheric fronts of an active cyclone which moved from the southern Baltic region to Samara led to a significant change of weather pattern in Eastern Europe. This change caused a distinct decrease in fire activities and a wet deposition of aerosol, lowering its content in the atmosphere above all investigated regions in the second half of August.

Between 2003 and 2014 ground-based and satellite observations showed the highest aerosol content over Kyiv every year in spring (April–May) and late summer July–August;. According to both studies, the observed spring peak in aerosol content is associated with transport of the Saharan dust across Eastern Europe, transport of sea salt aerosol from the Black Sea and the Sea of Azov, and occasionally occurring agricultural fires. The summer peak results from wildfires, soil dust aerosol due to harvesting activity, and transport of Saharan dust. The lowest AOD was observed in June and the middle of autumn. In Table  we show a prolongation of the data record of 2008–2013 for the Kyiv AERONET site by 3 more years up to the end of 2016. Even in this extended record the most significant aerosol pollution was observed in August 2010. This event is related, in particular, to wildfires in the ETR and Eastern Europe.

The impact of the wildfires on the AOD over Kyiv during summer 2010 was less pronounced as over Moscow (Fig. a). However, the aerosol pollution over Kyiv was also exceptional in comparison with the multiannual average. In August 2010, the average AOD of 500 nm exceeded the mean value for the same month over the entire observation period (excluding 2010) by a factor of 2.2 (see Table ). The mean AOD at 500 nm for June 2010 over Kyiv was 0.20, while its daily average on 15 August 2010 was more than 5 times higher, 1.09.

During the period from 1 June to 18 August 2010 the aerosol content over Kyiv gradually increased by more than an order of magnitude (Fig. a). By computing back trajectories to Kyiv at altitudes of 500 m, 1.5, 3, 4, and 5 km, we analyzed possible sources of aerosol from various regions. We provided detailed description of AOD variations and the impact of air transport on those changes in the Fig. S1–S34 in the Supplement.

According to our analysis of air mass transport, spatiotemporal distribution of the wildfires and their brightness temperature, the wildfires were not the main source of increased AOD over Kyiv in June. Both local sources (city transport, heavy industry, etc.) and air transport from Western Europe determined the aerosol content and properties; hence continental and marine aerosol was dominating. In the following, we used corresponding AODs as reference values for the further estimation of the aerosol content and its properties during July and August 2010, when the wildfires took place.

From the end of June 2010 the transport of air masses to Kyiv changed significantly as shown in back trajectory simulations (Supplement Fig. S15–S16). The air masses reached Kyiv at different altitudes after crossing the regions with the wildfires. Additionally, the number and the brightness temperature of wildfires were gradually increasing during first half of July (Figs. 2 and 3). Consequently, the increase of AOD over Kyiv was caused by the accumulation of aerosol transported from the wildfire regions or was formed under the influence of combustion products. In addition, sharp increase of number of wildfires after around 20 July (Fig. 3) and their brightness temperature (Fig. 2f–h) contributed significantly to AOD increase over Kyiv. Observed intermediate decreases of AOD, e.g., during 29–30 July, were mostly caused by deposition from occasional rains over Kyiv. From the beginning of August until 18 August the weather conditions were stable and corresponded to anticyclonic circulation (see Sect. 3.1, Supplement Fig. S29–S32). Those conditions caused an accumulation of aerosol from wildfires over Kyiv and high AOD values, in particular on 15–17 August.

According to our analysis of AE values (Fig. 5b) and results received with the application of a spectral deconvolution algorithm , fine-mode aerosol predominately contributed to the observed AOD increase over Kyiv in July–August 2010. Occasionally, coarse-mode aerosol had a significant impact on the observed AOD increase and the corresponding AE decrease during the time when more fires burned (e.g., 23 July and 2 August; Figs. 2f–g, b). This decrease could occur under the influence of local pollution in combination with weather conditions.

We also applied AERONET observations from Kyiv to analyze the impact of the wildfires on the aerosol spectral SSA and microphysical properties during summer 2010. Effective radius, size distribution (bimodal log-normal particle volume distribution), and RI (real and imaginary parts) were retrieved by and by inverse solution from AERONET sun photometer measurements along the almucantar of the sun. The influence of the wildfires on the aerosol properties over Moscow and surrounding regions in summer 2010 was estimated by , , and by comparing them with multiannual average. In our study we estimated the impact of wildfires over Kyiv by comparing the aerosol properties for the dates when aerosol from wildfires was observed to those dates when aerosol from wildfires was absent. We identified respective dates from the analysis of air masses transport to Kyiv, as described above. We compared aerosol properties as averaged values over different time periods: (1) when the number of fires and their brightness temperature were low, 1–26 June (40–65∘ N and 10–60∘ E, see Fig. 2); (2) when the numbers of fires significantly increased, 18 July–14 August; and (3) when the number of fires and their brightness temperature remained high and the highest AOD values were observed, 15–17 August.

We estimated the impact of fires on the aerosol size from AERONET sun photometer observations by calculating correlation coefficients between the aerosol effective radius and AOD (Fig. a), following the approach of . The correlation coefficient for the fine mode is about 0.65, while for the coarse mode it is only about 0.14. This indicates that the majority of the observed AOD increase was caused by fine-mode aerosol. Changes in the aerosol size distribution in the total atmospheric column are shown in Fig. b. Bimodal volume distributions, inferred according to and averaged over the aforementioned periods, confirm that the main contributor to the aerosol content over Kyiv was the fine mode from wildfires. This corresponds to longer atmospheric lifetime of the fine-mode aerosol in comparison with the coarse-mode aerosol . In the period when most fires burned (18 July–14 August), the volume distribution of both modes increased by about 30–40 % due to the observed AOD increase over Kyiv. In the following 3 days (August 15–17), when maximum AOD was observed (Fig. a, b), only the fine mode increased further by more than a factor of 2. Relative to the earlier period with less fire activities (1–26 June), the overall increase was about a factor of 4. Consistently, after 18 August, when the atmosphere over Kyiv was cleaned again, the aerosol content in both modes decreased and, in turn, was even lower than during earlier periods.

The dependence of visible and near-infrared SSA from the fire activities and aerosol load over Kyiv is evident from Fig. c by reduced values during the more intense fire period between 18 July and 14 August. Most probably, these changes were caused by an increase of the soot content in the air, which was transported from the fires. A corresponding increase of aerosol absorption in the longwave part of spectra is seen in the slope of the SSA spectral dependence, which becomes steeper. It is similar for 15–17 August, although larger SSA values were observed, which were the same level as during the period when less fires burned in June 2010. According to , larger SSA values can be explained by the presence of the coarse-mode particle during those days (Fig. b), which increases the total reflectance of atmospheric column. Also increased real RI and longwave imaginary RI for the days with increased aerosol content (18 July–17 August, Fig. d) refer to aerosol absorption caused by wildfires. We additionally provided the daily averaged SSA as well as real and imaginary RI for 440 and 870 nm, Level 2.0 from AERONET database, for the specific dates during summer 2010 at the Kyiv AERONET site (Table S35).