5.1. Active Remote Sensing Observations from Satellites and at Ground
Active remote sensing measurements were performed at three sites of the Poland AOD network in already-introduced Warsaw and Strzyzow, as well as in Raciborz located in the vicinity of the coal industry region of Silesia (Western Carpathians, south-western Poland). In
Figure 2, the PollyXT lidar measurements in Warsaw and the CHM15k ceilometer measurements in Strzyzow and Raciborz are shown for the period of 24–30 August 2016.
In Warsaw, on 26–28 August, untypical for summer conditions, aerosol load and the boundary layer top can be discerned from 24/7 evolution plots of the lidar signal (
Figure 2a). In general, at night-time in central Europe, convection processes are not expected—typical night-time boundary layer height over Warsaw in August is <1.5 ± 0.2 km a.g.l. [
54]. However, during the daytime, the top of the boundary layer can reach 2.9 ± 0.6 km a.g.l. due to the occurrence of strong convection [
60] and heat waves [
5]. For the case of measurements depicted in
Figure 2a, the aerosol was confined to the lowermost 1.0 ± 0.2 km a.g.l. in the morning of 26 August, then the boundary layer top gradually increased to 1.35 ± 0.1 km the following morning, to reach 1.7 ± 0.2 km in the morning of 28 August. The maximum of the aerosol load was recorded in the evening of 28 August at 2.0 ± 0.1 km. The gradual increase of the boundary layer top between 27 and 28 August 2016 was quite dynamic (approximately 1000 m within less than 2 days!), indicating possible accumulation of not only local pollution but also transported into Warsaw boundary layer biomass burning particles (compare with trajectories in
Figure 1). The observed mixing layer height on the days that followed, of 1.1, 1.6 and 2.1 km a.g.l. was rather low for Warsaw summer conditions (the latter value is more typical for summertime, although still relatively low). At night, the residual layer was exceptionally strong; there were almost no differences in night-time-to-daytime conditions in terms of both layer height (as it was persistently occurring up to the same altitude) and aerosol abundance (less evident). The morning transitions were relatively weak, with no distinct structures within (except for the morning of 27 August). For the case discussed here, the gradual increase of both the aerosol load and the boundary layer top certainly was related to large-scale displacement of the high-pressure system, although this could not have been dominating cause, as it did not manifest at the two other sites in Strzyzow and Raciborz. Over Warsaw, for period starting in the evening of 25 August until the evening of 28 August, there was no evidence of clouds, except for a few passing cumulus, forming over the boundary layer in the evening of 28 August. In free troposphere, there was a clear indication of an aerosol layer at 2.5–3 km from approximately 5:00 to 12:00 UTC on 26 August, which was likely long-range transported from the vicinity of the Iberian Peninsula (compare with backward trajectories
Figure 1). The features visible up to 3 km on 29 August, are related to aerosols and clouds, caused by the atmospheric front passing over Poland. As for the other days, some cloud signatures are evident, especially for 24 and 30 August.
In Strzyzow, the 24/7 ceilometer data on 26–28 August (
Figure 2b) also revealed no clouds, the aerosol load on three days showed an accumulation of particles in the lowermost 0.7 ± 0.2 km a.g.l. with slight increase of the boundary layer top at night-time to 0.9 ± 0.2 km. Morning transitions were not discernible but there was some temporal variability of the aerosol load within the boundary layer. In the free troposphere, there was an indication of a very weak, lofted aerosol layer in the evening of 28 August. On other days, similar cloud signatures as in Warsaw are evident, except for 29 August when even precipitation was evident.
As for the 24/7 ceilometer observations at the Raciborz site (
Figure 2c), on 26–28 August, the aerosol load was confined to the lowermost 1.1 ± 0.1 km a.g.l. and there was no gradual increase of the boundary layer. The height of the afternoon’s aerosol mixing layer and the residual layer at night were almost the same, the latter being of only 200 m lower. On each day, a rather weak morning transition was discernible. There was a relatively low variability of structures within the boundary layer, indicating possible local and/or transported aerosol accumulation on 28 August (reddish values) with no increase of boundary layer top. No clouds were observed over the site in this period. The feature observed in free troposphere at 2–2.5 km on the evening of 28 August (15:00 to 21:00 UTC) should be attributed to an aerosol layer, rather than a cloud system. On other days, similar cloud signatures as in Warsaw and Strzyzow are evident, with some precipitation signatures driven by passing front on 29 August.
In general, low signal-to-noise ratio at relatively large wavelengths, deployed for the sake of eye-safety in ceilometers, limits the scope of available instrument products [
60]. Moreover, despite efforts at exploiting single-wavelength elastic backscattering lidar/ceilometer data for the aerosol optical properties retrieval, such as a co-location with photometer AOD [
51], a constrain with satellite AODs [
61], a forward retrieval [
62,
63] or a two-stream approach [
64], significant limitations have been reported. Therefore, analyses within the current paper were primarily focused on solely and independently derived, vertically-resolved aerosol optical properties from the complex, multiwavelength, far and near-range PollyXT lidar ground-based observations over the Warsaw EARLINET site. The observations provided a unique set of aerosol wavelength-dependent optical properties: particle extinction (2α) and backscattering (3β) coefficient, linear particle depolarization ratio (2δ).
In total, 33 day-time and night-time profiles were obtained for the investigated period, from 24 to 30 August 2016. The sets of 2α + 3β + 2δ, were inspected visually and divided into 4 time periods in which the similarity of the consecutive profiles was high: the night of 25/26 August, the late afternoon and early evening of 26 August, the night of 26/27 August and the night of 27/28 August. The mean aerosol properties (with their variance indicated by shadows), plotted in
Figure 3, depict the variability of the obtained results. Although according to the synoptic weather charts during this period a quasi-stationary situation was expected, clearly different aerosol properties were observed in the four groups as well as within the boundary layer and in the free troposphere.
In general, on both first panels in
Figure 3 (the night of 25/26 August and the daytime of 26 August), both particle extinction coefficient profiles were very similar, if not the same, within the given uncertainties. However, for the last two panels (the nights of 26/27 and 27/28 August) this holds true only for the free troposphere range, as the extinction profiles were clearly higher at 355 nm within the boundary layer. Moreover, the extinction coefficient peak visible in the free troposphere at the first night seems to lower with altitude and time to appear within the boundary layer at the last night. The particle depolarization above 1.4 km was clearly separated at the two wavelengths (with higher values at 355 nm) at the night of 25/26 August and at the daytime of 26 August. On the following two nights, the particle depolarization at both wavelengths was the same within the given uncertainty range (with a slight tendency for higher values at 532 nm). Within the boundary layer, the depolarization ratio at 532 nm was always higher than its value at 355 nm, both being the highest at the daytime of 26 August. As for the Ångström exponent, it was always significantly lower in the free troposphere than within the boundary layer, except for the daytime of 26 August 2016, when it became higher. The lidar ratio at 355 nm clearly decreased with altitude in all cases depicted in
Figure 3, however, the lidar ratio at 532nm did not revealed such clear regularity, which is visible especially at night of 25/26 and 26/27 August.
In
Figure 3, on the night of 25/26 August (top subfigures), within the boundary layer, below 1.1 km significant wavelength dependence for the mean depolarization ratio (being of roughly 3.4% for 355 nm and higher of 5.2% for 532 nm) with the mean extinction profiles (~ 1 · 10
−4 m
−1 for 355 nm and 532 nm) can be discerned. The mean Ångström exponent was relatively low (~0.9) and the lidar ratio at 355 nm (68 sr) was higher than for 532 nm (60 sr). At the height of 2.2–2.5 km, in the free troposphere, the mean Ångström exponent was low (~0.4), the lidar ratio at 355 nm (46 sr) was lower than for 532 nm (70 sr) and the linear depolarization ratio at 355 nm was slightly increased to a value of about 1.6%.
During the day of 26 August, within the boundary layer, similar profiles were derived, whereby significant wavelength dependence in depolarization ratio (roughly 3.1% for 355 nm and higher of 6.5% for 532 nm) was captured. The mean Ångström exponent (~1) and the lidar ratio at 355 nm (70 sr) and 532 nm (60 sr) were slightly higher than at night. In the free troposphere, the mean Ångström exponent was high (1.5), the lidar ratio at 355 nm (40–50 sr) was lower than for 532 nm (50–60 sr) and the linear depolarization ratio at 355 nm, was still slightly increased (up to 2.5%).
On the night of 26/27 August, separation of the properties in free troposphere and within the boundary layer was still visible. Within the boundary layer the Ångström exponent increased to 1.6 and the depolarization ratio and lidar ratio at both wavelengths decreased (2.4%, 56 sr and 4.8%, 50 sr at 355 and 532 nm, respectively). In the upper range, above 1.8 km depolarization at 532 nm increased and lidar ratios at both wavelengths as well as Ångström exponent slightly decreased.
During the following night, 27/28 August, within the boundary layer, a similar relation of the lidar ratio as in the free troposphere is striking (reversed wavelength dependence). High lidar ratios of 80 sr and 83 sr for 355 and 532 were derived for a significantly high Ångström exponent (1.2) and with depolarization ratios of 2.3% for 355 nm and higher, of 5.9% for 532 nm, the latter being higher as on the previous night. In the free troposphere, Ångström exponent was relatively low (~0.3) with significantly high lidar ratios at 355 nm (60 sr) being higher than for 532 nm (72 sr).
The optical properties derived within the boundary layer and the free troposphere are listed in
Table 1. Within the boundary layer, local aerosol properties were likely affected by 1–2 day-old biomass burning transport from over Ukraine (compare
Figure 1, left subfigure). The obtained Ångström exponent (1–1.5) indicated moderately fresh aerosol. The values of ~1.9 for very fresh and <1 for aged biomass burning are reported [
13,
65]. The significant wavelength dependence of the mean depolarization ratio was similar in the current study to those reported for biomass burning by [
7,
25]. The ratio of lidar ratios (LR
355/LR
532) was close to 1; for moderately fresh biomass burning values of 1 or higher can be expected [
13].
At the height of 2–2.5 km, the free troposphere was likely influenced by the long-range transported intrusion of 3–5 day-old biomass burning aerosol from over Portugal (compare
Figure 1, right subfigure). The reversed wavelength dependence for the mean lidar ratio (~45 sr at 355 nm and ~65 sr at 532 nm) characteristic for aged biomass burning was discerned. The reversed dependence in the lidar ratio reported by [
7,
13] was of about 30 sr at 355 nm and 55 sr at 532 nm. In [
24,
25] higher values of about 60 sr and 90 sr were reported, respectively. The depolarization ratio values are similar to those reported by [
25], for the case of the biomass burning layers with low particle abundance. The obtained Ångström exponent (<1) was typical for aged biomass burning aerosol, as in [
13,
65].
As for detailed interpretation of the tropospheric layers, they seem to attribute on the following 4 time periods to: (a), (c), (d) the low in abundance, aged, 3–5 day-old forest fire biomass burning from Portugal and (b) the likely mixture of the aged, 3–4 day-old Portuguese biomass burning with the moderately-fresh, <1 day-old peatland fire biomass burning form Ukraine.
In the layers defined within the boundary layer, the detailed interpretation is as follows: (a), (b) and (c) a slight local, background urban pollution affected with an intruding Ukrainian biomass burning, being at first moderately-fresh <1 day old, then 1–2 day old and (d) mixture of the 2 day-old Ukrainian with the aged 3–4 day-old Portuguese biomass burning.
The yearly mean and the summer mean (JJA) particle extinction profiles were calculated at 355 nm and 532 nm based on four years (2013–2016) of observations at the Warsaw site. The mean profiles were obtained by averaging profiles available within the category
climatology of the EARLINET/ACTRIS Data Base (accessed on 26 January 2018 via
www.earlinet.org). In this category, only data files derived for the Monday evening and Thursday midday and evening measurements are stored. In total 68 and 24 profiles were averaged to obtain the yearly mean and the summer mean, respectively. The climatology profiles with standard deviation are shown in
Figure 4.
The mean particle extinction coefficient profiles derived over Warsaw during stable weather conditions on 25–28 August depicted in (
Figure 3) showed differences in signatures in comparison with the mean profiles derived for the summer season (
Figure 4, in purple). The characteristic extinction peak at about 1.4–2.2 km altitude visible in the mean summer profiles at both wavelengths (up to roughly 2.4 · 10
−4 m
−1 at 355 nm and 1.3 · 10
−4 m
−1 at 532 nm), was also visible in
Figure 3. In both figures, within the boundary layer (up to 1–1.8 km), the extinction at 355 nm was higher than at 532 nm but above there was not much wavelength dependence in
Figure 3. By comparison, in the mean summer profiles, the wavelength dependence is observed regardless of altitude, with the mean summer extinction coefficients being higher at 355 nm than at 532 nm.
The temporal evolution of the lidar-derived aerosol properties within the boundary layer during the analysed period is given in
Figure 5, where along with the boundary layer height, the extinction-related Ångström exponent derived at 355 and 532 nm, as well as wavelength dependent aerosol optical depth, particle lidar ratios and particle depolarization ratios are depicted. An opposite trend in the lidar ratio and Ångström exponent is weakly manifested but still discernible, as expected. The scatter plot of the extinction related Ångström exponent (355/532) and lidar ratios at both wavelengths revealed similar relations, indicating decreasing lidar ratios for high values of Ångström exponent (negative correlation coefficient was low of −0.4 and −0.3 for 532 nm and 355 nm, respectively). Lower lidar ratios were obtained for the evening of 26 till the morning of 27 August (40 sr to 60 sr). In general, lower depolarization ratios are obtained for 355 nm (1.6 to 4%) than for 532 nm wavelength (4 to 7.5%) for the same Ångström exponent, whereby the highest depolarizations were observed in the evening of 26 August (6 to 8%). Similar tendencies as for lidar ratio, seem visible in the relation between the extinction related Ångström exponent (355/532) and linear depolarization ratio at the two wavelengths, i.e., decreasing depolarization ratios were obtained for increasing Ångström exponent values, whereby this being significantly weaker at 355 nm. The increase of the boundary layer height seems to be in general accordance with the increase of the aerosol optical depth and Ångström exponent. Increasing Ångström exponent was observed on the evening of 26 August onward, reaching its highest values in the morning of 27 August. At the same time period, the lidar ratio had lowest variability. The depolarization ratio revealed significant differences at the beginning of the evening of 26 August; before and after, the differences were at a similar level.
As for the satellite active remote sensing, within the analysed time period of 24–30 August 2016, the Cloud-Aerosol Transport System (CATS) products over the Warsaw site, were available only on the first and the last day of the analysed period. They were used to indicate conditions before and after the high-pressure situation. The CATS overpasses at a distance regarded sufficiently close to the Warsaw lidar site were chosen according to the EARLINET inter-comparison rules. The L2-products for 24 and 30 August 2016, available via the CATS Website (
https://cats.gsfc.nasa.gov), include the backscattering coefficient at 1064 nm, the particulate depolarization ratio at 1064 nm and the feature type and aerosol subtype for atmospheric layer classifications. For the Warsaw CATS overpass on 24 August (11:28 UTC; 44.85 km closest orbital distance), a dust-clean continental aerosol layer coexisting with water clouds was detected at an altitude of approximately 1–3 km a.g.l. On 30 August (09:26 UTC; 53.06 km closest orbital distance) at about 2–3 km, the backscattering and particle depolarization ratio showed a polluted dust aerosol subtypes. On both days, the lidar data was available 24 h/day and for 24 August indicated a thin, liquid water cloud signature, building on the boundary layer top, sensed at about 1.8 km; and for 30 August, a strongly depolarizing aerosol, apart from weak and scattered clouds, accompanying the morning transition in the boundary layer.
Mineral dust intrusions often accompanied by cloud systems are not uncommon over Warsaw [
37] and cannot be excluded in this case. This could be positively confirmed with AERONET measurements. Dust-dominated conditions are defined when the AREONET Level 2.0 Ångström exponent (440/870) is less than or equal to 0.75 [
66,
67]. Unfortunately, for the entire investigated period of 24–30 August 2016, there was no Level 2.0 AERONET data available for the Belsk site, being the closest to Warsaw (40 km distance). Therefore, the lidar-derived Ångström exponent (355/532) on 24 and 30 August was used instead. Values in the range of 0.6–0.8 and 0.4–0.8 were obtained, which on both days indicated the existence of relatively large particles that could have been related to dust or other large-sized aerosol (e.g., pollen, reported by [
68]). These values are significantly lower in comparison to the Ångström exponent (355/532) of roughly 0.8 to 1.8 on 25–27 August (compare with
Figure 5) during the intrusion of biomass burning particles into the boundary layer. The lidar-derived depolarization profiles can serve as a good indication on particle shape used for separation of the mineral dust from other aerosol types. An increased depolarization of up to 12% was observed in the lidar data but only within the boundary layer on 29 and 30 August, which supports the interpretation of the observed features such as pollen and not dusts mixtures.
5.2. Passive Remote Sensing from Satellites and at the Ground Level
The SEVIRI AOD maps covering the entire area of Poland were obtained every 15 min, on each day of the analysed period. There is a lack of maps on 26 August, as this day was suitable for a as reference day, exhibiting the lowest cloud coverage and the lowest aerosol optical depths. Air mass transport for this day showed possible contamination with biomass burning but considering the transport on all other days, this day was the best candidate. Moreover, the maps cannot be provided for the night-time or for high and low Sun elevation angles. Very high cloud coverage over Poland on 24 August prevented the retrieval on this day. For the remaining days, the retrievals proved successful. In total, 98 maps were derived (for brevity, only selection of those can be shown). The SEVIRI AOD Maps are depicted in
Figure 6.
The SEVIRI AOD maps calculated on 25, 27–30 August 2016 are shown in
Figure 6, where the morning maps obtained at 5:45 UTC (in the left column) and afternoon maps at 14:45 UTC (in the right column) are inter-compared. On 29 and 30 August retrieval at 14:45 UTC was hindered (due to high cloud coverage) and therefore the closest-in-time maps were plotted instead (i.e., 9:30 UTC and 15:00 UTC, respectively). Relatively high cloudiness is still evident on 29 and 30 August (white pixels over the land); on the other three days, the cloudiness was low. A striking feature in the AOD patterns is their persistent increase between the values observed in the morning with respect to the afternoon values. This is evident in all maps but it unlikely to be an artefact and should not be related to the zenith elevation angle nor (dependent on it) scattering angle, as retrieval in the more challenging autumn conditions of September 2016 did not reveal such behaviour [
5]. It is speculated that on 29 August the clear increase of the AOD between 5:45 UTC and 9:30 UTC must have propagated towards even higher AOD values in the afternoon. For the cloudless 25–28 August in the morning, the AOD values are evenly spread over Poland, in the afternoon, slightly higher AODs are observed over eastern Poland. Interestingly, despite the cloud coverage on each consecutive day, beginning on 25 August until 29 August, there is a clear gradual increase of the AOD practically over the entire territory of Poland, regardless of whether in the mornings (increase from <0.15 to <0.35) or in the afternoons (increase from <0.35 to <0.5). On 30 August, AODs dropped to background values both in the morning (<0.15) and afternoon (<0.35).
The daily mean of the SEVIRI AOD maps at 635 nm (available on 25, 27–30 August), although slightly higher, still show generally-good agreement with the daily MODIS AOD at 550 nm available on 26–29 August (not shown for brevity).
Apart from the MODIS AOD product, the MODIS fire maps (available via NASA Worldview Website (
https://worldview.earthdata.nasa.gov) were searched to allocate possible sources of active wild-fires that occurred before and during the analysed time period. Two composite maps are shown in
Figure 7, where high surface-fire activity (red dots) over Portugal (left subfigure) on 23 August and over the north-west of Ukraine on 26 August 2016 (right subfigure) is indicated. According to the backward trajectories given in
Figure 1, the air masses were likely to pick up the biomass burning aerosol from possible sources over Portugal and Ukraine, whereby along the pathway of the air masses no other possible fire activity could have played a role. There is also negligible indication of local fires over Poland and it is unlikely that it could have actively contributed to the measured aerosol properties.
The ground-based passive remote sensing conducted within the PolandAOD network, agrees with the passive satellite SEVIRI and MODIS retrievals. The temporal evolution of the hourly mean values of AOD obtained at four PolandAOD sites for the period of 24 to 30 August 2016, is depicted in
Figure 8. In Strzyzow, Rzecin and Raciborz AOD was detected with the CIMEL 318 photometer and in Warsaw with the MFR-7 radiometer. The cloud-screened AERONET product of Level 2.0 was available for Strzyzow and Raciborz, therefore the cloudiness over sites in Rzecin and Warsaw was assessed from radiation flux measurements (which in fact agree with the clouds measured by the lidar in Warsaw). During the analysed period, the observed AOD values were in general low, for most sites and days remaining below 0.15, with exception of the higher aerosol loads observed during 29 August. Unfortunately, on that day there is a significant number of missing data for many hours and sites, due to extensive cloud cover over most of the Poland territory (white pixels in
Figure 6) with exception for Strzyzow site. However, it is unlikely that the AODs derived with radiometer and lidar over Warsaw, are contaminated by clouds, as the cloud screening was carefully performed. The clearest day in terms of cloud coverage and aerosol load was 26 August, with the hourly mean AODs below 0.1 at all sites (reference day for the SEVIRI AOD retrieval). The comparative analyses of the SEVIRI AOD, derived at pixels representative for the Warsaw, Rzecin, Raciborz and Strzyzow sites with the AOD data measured with CE318 photometers, MFR-7 radiometers and PollyXT lidar were revealing good agreement (correlation coefficients r
2 of 0.91, 0.84, 0.8 and 0.57, respectively) as reported by [
5]. Within the current study, lower correlation coefficients (~0.65) were obtained, which is mainly attributed to the generally lower AOD values.
During the course of events, the lowest AOD values were observed at the mountain site of Strzyzow in the south-east, especially on 26–28 August, characterized by an exceptionally low boundary layer depicted in
Figure 2. The AODs in Strzyzow remained typical for this site, at values between 0.05–0.07 on 24–25 August. On 26–28 August AODs dropped even lower to a background value of 0.03 ± 0.01. On 29 August, within only 9h they increased steeply up to 0.2 and fell back to typical values in the afternoon of 30 August. Somewhat similar behaviour but with roughly 2 times higher AODs and less sharp increase (from about 0.05 to 0.15 within 3 days) was observed at the semi-urban site of Raciborz in the south-west. In central Poland, at the urban site in Warsaw and the wetland site in Rzecin, the aerosol load enlarged persistently from below 0.1 up to 0.22 within 3 days and from below 0.07 to 0.16 within 5 days, respectively. Also at these two mid-country sites, on 30 August AOD decreased significantly to similar values as obtained before the event, on 24 August. Increase of the AODs on 29 August was due the frontal system passing over the country between the night of 28/29 August till the night of 29/30 August. The Ångström exponent (440/870) measured with the CIMEL instrument in Raciborz was rather stable 1.4–1.8 on 24–30 August but in Strzyzow it was strongly oscillating between 1.1 and 2.0. The Ångström exponent (415/870) measured with the MFR-7 radiometer in Strzyzow was alike (with slightly lower values). In Warsaw, the Ångström exponent (415/870) was between 0.4 and 1.6.