The Geographical Tables of Poland [33
] indicate that September 2016 was extremely warm with a monthly anomaly of +2.57 °C, as calculated in relation to the average period of 1961–1990 (+2.36 °C for 1981–2010; +2.13 °C for 1991–2015). During the daytime, the September temperatures reached as high as +31 °C in the shade, and the night-time temperature drop was very low, down to exceptionally high values of 16–19 °C. By comparison, the Polish absolute temperature record was reported as +40.2 °C (in shadow) on 29 July 1921. According to the Synoptic Weather Charts Archive at the IMGW Website (www.pogodynka.pl/polska/mapa_synoptyczna
], one of the biggest September heat waves on record was faced at the end of the first half of the month, spreading mainly over western and southern Poland. It then became cooler, but high temperatures were still observed until the end of the month. Therefore, to analyse the event, we focus on the crucial period of 11–16 September 2016.
The near-surface temperature and relative humidity measurements recorded by the WXT510 Vaisala meteorological station at Warsaw and Strzyzow sites (www.poland.aod.pl
) confirm the occurrence of the heat wave. At both measurement sites, the temperature and relative humidity are inversely proportional. They exhibit a daytime cycle with lower temperatures at night, and higher temperatures during the day, which are accompanied by an opposite cycle of the relative humidity. During the course of this event in Warsaw, the maximum temperature was obtained during the daytime with the highest values of 30–31 °C on 10 and 11 September, gradually dropping down to 25 °C on 16 September. The daytime relative humidity was very low, with minimal values between 29% and 35%. In general, the maximal surface temperature obtained in Strzyzow was lower, being 3 °C, and relative humidity was higher, being up to 12%. The observed temperature ranges indicate a heat wave occurrence as it is in opposition to typical thermal conditions trends for Poland [47
According to the USA National Weather Service—National Oceanic and Atmospheric Administration (NOAA), the so-called heat index (HI) is defined as a measure of how hot it feels when relative humidity is factored with the apparent air temperature. The Heat Index Chart (http://www.nws.noaa.gov/om/heat/heat_index.shtml
) relates the heat index value to the level of safety alert: HI of 26.7 °C to 32.2 °C caution level (fatigue for prolonged sun exposure and/or physical activity, to 38.9 °C extreme caution (possible sun-stroke, muscle cramps and heat exhaustion, to 52.2 °C danger (likely sun-stroke, heat cramps, heat exhaustion and possible heat-stroke) and >52.2 °C extreme danger (heat-stroke likely). The heat index values are defined for shady, light wind conditions and thus an exposure to full sunshine can increase the heat index values by up to 10 °C. On 8–13 September 2017, based on the meteorological measurements obtained in Warsaw, the heat index indicated values reaching 27.2 °C to 30.6 °C, which are labeled as caution level. On 14 September, the atmosphere started to cool down to reach a lower index value of 22.2 °C by 15 September, as related to long-range transport of cool and cleaner Arctic air mass. It then increased to 24.4 °C on 16 September, due to the influence of continental air from Eastern Europe.
The back-trajectories calculated using the HYSPLIT model [48
] at 12:00 of Universal Coordinated Time (UTC) on each day of the event for 4 locations in Poland are given in Figure 1
. Four sites of the PolandAOD network were chosen: Sopot (in the north of Poland, semi-urban site at the Baltic Sea coast), Rzecin (west-central Poland, wetland site, rural conditions), Warsaw (central Poland, urban site) and Strzyzow (in the south-east, a background mountain site). On each day, the backward-trajectories indicated different possible air mass transport. At most of the stations on 11–13 September, the air was advected from over East and South Germany, Czech Republic and Southern Scandinavia, but at Strzyzow, the air mass indicated fast transport from over Ukraine and Belarus. On 14 September, air was transported to all sites from over the Scandinavian and Baltic countries. On 15 September at all sites, the Arctic air masses were uniformly transported from over Northern Europe. On 16 September, air masses had a similar pathway only for the northernmost site in Sopot, for the trajectories which remained over South-Eastern Europe.
3.1. Optical Properties with Ground-Based Active Remote Sensing
Ground-based, active remote measurements were available at three Poland AOD sites in Warsaw (central Poland), Strzyzow (south-eastern Poland, a vicinity on the Polish-Ukrainian border), and Raciborz (south-western Poland, a vicinity of the coal industry region of Silesia Region). In Warsaw, Lidar measurements with PollyXT Lidar were taken (shown in Figure 2
). Observations with CHM15k ceilometers were performed in Strzyzow (Figure 3
) and Raciborz (Figure 4
). For all sub-plots, in each figure, the same colour scale was kept.
In Warsaw, on 11–13 September, a distinctly high aerosol load can be discerned from 24/7 evolution plots of the 1064 nm Lidar signal. These are days with a well-increased aerosol load and possible occurrence (even accumulation) of transported and local pollution, compared with trajectories in Figure 1
. The boundary layer is very high, reaching up to 3.2 km a.g.l. (even for night-time residual layer), with distinct structures within. During the night in central Europe, a strong convection is not expected. A typical night time boundary layer height over Warsaw in September is <1.5 km and at daytime, the boundary layer top can reach approximately 2.8 km if strong convection and/or heat island effect occurs [37
]. The extraordinarily strong elevation of boundary layer is likely due the heat wave. On 11 and 13 September, there is evidence of passing cumulus forming over the transition zone at around 12–14 UTC. On the next three days, 14–16 September, the boundary layer is much lower at 1.5–2 km a.g.l. (these values are more typical for summertime than autumn [37
]), with still well distinguished morning transition, but not as much layering and structures as on 11–13 September.
At the Strzyzow site, the situation is different. On all days, the aerosol loads discerned from 24/7 evolution plots of the 1064 nm ceilometer signal, show accumulation of the aerosol in the lowermost 2–2.6 km a.g.l. (there is fair cloudiness indication in some profiles). Except for the period between 19:00 UTC on 19 September and 14:00 UTC on 15 September, where clearly, boundary layer height drops down to 1.25 km a.g.l., this drop, as at the Warsaw site, is likely related to the Arctic air mass inflow (compare Figure 1
). However, the remaining aerosol structures seen in Strzyzow are likely to have originated from smoke and biomass burning aerosol transport, as indicated by the HYSPLIT backward trajectories (Figure 1
). Indeed, the MODIS Global Fire Map 10-days composite, available for the period of 7–16 September 2016 (https://lance.modaps.eosdis.nasa.gov/cgi-bin/imagery/firemaps.cgi
), confirmed the likeliness of a wild fire event of forest, peatland and/or grass burning over Ukraine.
As for the Raciborz site, the ceilometer signals (Figure 4
) indicate similarities to the Warsaw site for 11–13 September (although the heat wave driven boundary layer is slightly lower at 2.2–2.9 km a.g.l.). There were even more similarities to the Strzyzow site for 14–16 September, with Arctic air-cleaning intrusion reaching the site at night on 14/15 September. Indeed, backward trajectories (Figure 1
) indicate likely transport and mixing of polluted air mass from Germany and biomass burning particles from Ukraine over the Raciborz site.
3.2. Aerosol Properties from Model Output Simulation
The results of the Navy Aerosol Analysis and Prediction System—NAAPS model [50
] simulations (http://www.nrlmry.navy.mil/aerosol
) providing maps of AOD, sulphates, smoke and dust for Europe are given in Figure 5
. A clear peak-patch of sulphate concentrations (16–32 μg/m3
) in Germany (in the Region of Saxony and on the west side of the borders with the Poland and Czech Republic) is seen for all days but 15 September. This peak-patch was accompanied by enhanced AODs (up to 1.6) only on 11–13 September, whereby AODs ‘drifted’ from Germany towards western Poland. Over Poland, the high AODs (up to 1.6) remained in the north-west and the sulphates were a significant part of the aerosol load, reaching concentrations in a range of 2–16 μg/m3
. In addition, smoke intrusions were indicated over south-eastern Poland (up to 16 μg/m3
). The period of 14–16 September was characterized by cleaner conditions with AOD in the range of 0.1–0.2 and sulphates reaching up to 16 μg/m3
only over south-western Poland. The rest of the country’s AODs were below 0.1 and sulphate concentrations below 8 μg/m3
were forecast. Within the entire period, minimum values of AOD and sulphates are related to the atmosphere’s cleaning due to the Arctic air mass inflow on 15 September. As for the smoke, a tendency to spread over southern and towards central Poland is discernible. Due to the location of the Strzyzow site in the south-east of Poland, there is a clear indication of relatively high smoke concentration (2–8 μg/m3
) during the entire period of 11–16 September 2016. Throughout the entire event, there was no indication of possible dust intrusions. In general, the NAAPS AOD seems to underestimate, with respect to all AOD data obtained for this period with ground-based and satellite instruments.
The Copernicus Atmosphere Monitoring Service—CAMS model [44
] forecast of the AOD at 550 nm over Poland for the period of 11 to 16 September 2016 is plotted in Figure 6
. The CAMS AOD maps reflect qualitative similarities with the NAAPS AOD maps. Here too, 11–13 September is indicated as having a higher aerosol load (AOD values of up to 0.5 in the northern part of Poland). The atmosphere begins to clear in the north-east of Poland and shifts the aerosol towards south-west on 14 September (values of 0.2–0.35). Finally, on 15 and 16 September, CAMS output indicates much cleaner conditions for the entire country with AODs of about 0.1.
The Global Environmental Multiscale—Air Quality GEM-AQ model [51
], forecast of the diurnal evolution of the aerosol extinction coefficient profiles for the locations of four sites: Sopot, Rzecin, Warsaw, and Strzyzow for 11–16 September 2016 is shown in Figure 7
. The results for the whole period are in qualitative accordance with the NAAPS and the CAMS simulations; the lowest values are obtained at all sites for 15 September 2016. There are also high similarities of the GEM-AQ model extinction profiles evolution and the ceilometer signal plots in Strzyzow (Figure 3
) and Raciborz (Figure 4
), as well as with the Lidar signal plots in Warsaw (Figure 2
). The similarity here is especially remarkable. Moreover, the values of the daytime AODs obtained in the boundary layer by integrating the model (500 nm) and Lidar (355 nm, 532 nm) aerosol extinction profiles (15 cases, Table 1
) show a similar trend for all days of the event, except for 11 September, (where the GEM-AQ simulated lower values), including AOD obtained by passive sensors, showing strong similarities of trend with AOD of GEM-AQ, can be reported for these four sites.
3.3. Optical Properties from Ground-Based Passive Remote Sensors
In fact, the discussed AOD (respective pixels) of NAAPS and CAMS, as well as GEM-AQ simulations, are in general accordance with the AOD data measured at the ground by the means of photometry and radiometry, as given in Figure 8
. The hourly mean values of AOD were recorded at 5 locations of the PolandAOD network for the period of 11–16 September 2016. The Strzyzow, Rzecin, and Raciborz collected the AOD data with the CE318 photometer and the Sopot and Warsaw stations collected the AOD data with the MFR-7 radiometer. Temporal evolution of the AOD shown in Figure 8
indicates that the aerosol rich atmosphere had already cleared first in Sopot (the northernmost location) by 13 September, in central Poland (Warsaw and Rzecin) only a day later on 14 September and in the south at Strzyzow and Raciborz only on 15 September. On 16 September in Warsaw and Sopot, the conditions were similar; however, at the other three sites, the AOD had increased, which can be attributed to wild fire smoke aerosol transport in the South-East and/or pollution advection in the West.
The spectral dependence of the AOD for period 11–16 September 2016 at the Poland AOD sites in Rzecin, Raciborz, Strzyzow and Warsaw is depicted in Figure 9
. In general, 15 September was distinguishably clearer at all sites (Figure 9
, in green). In Warsaw there is a discernible separation in the spectral plots for 11–13 September (high load) and 14–16 September (lower load). The steep exponential-like shape of the spectra for the higher load period indicates an occurrence of small particles (of an approximate size of 400 nm). For the lower load, a similar abundance of small and larger particles occurred (size range of 400–900 nm). Over other sites, this high aerosol load of small size particles dominated, especially over Rzecin (the west-central location); on 12 and 13 September, values of AOD at 350 nm reached 0.85, clearly indicating anthropogenic pollution. In Rzecin, as in Warsaw, the atmosphere cleared on 14 and 15 September, but on 16 September increased again. In Rzaciborz (the west-south site) the situation is slightly different. The steep AOD spectra are present for all days, except for 15 September. On 11 and 12 September, AOD at 350 nm reaches 0.4–0.5 and on 12, 13 and 16 September, it increases to 0.7, whereby the spectra do not steepen, but are lifted, indicating an increase in aerosol load which may likely be due to smoke aerosol contribution. For Strzyzow again, 15 September AOD spectra are the lowest, and as for the rest, it is not feasible to unambiguously separate them.
For the three sites of Warsaw, Strzyzow and Rzciborz, the ground-based active sensors measured the aerosol structures within the boundary layer and radiometers measured the corresponding columnar AOD. An increase of the boundary layer height due to heat trapped near the ground by the heat wave’s stationary conditions can be related to the AOD measured by the radiometers.
For the Warsaw site, NAAPS indicated increased AODs due to a sulphate patch originating in Eastern Germany spreading towards central Poland (Figure 5
). Thus, at this site only anthropogenic pollution is expected. A strong increase in the boundary layer due to the heat wave that manifested on 11–13 September (Figure 2
) is well correlated with an increasing aerosol load and AOD (Figure 9
). There is also a clear daytime AOD variation (Figure 8
), capturing AOD increase and decrease, following the cycle of the Sun’s elevation-angle change; although at night, due to Arctic air inflow on 13/14 September, the boundary layer is abruptly lowered from approximately 3 to 1.2 km a.g.l. and AOD at 675 nm from 0.35 ± 0.025 to 0.12 ± 0.025. The relation of AOD and solar operation is weak but still discernible. Almost no differences in boundary layer height during day and night time for the entire period can be attributed to the heat wave driven lack of typical strong cooling at night, favouring rather strong residual layer formation and hindering nocturnal layer formation. Therefore, no significant differences between day and night AOD are expected.
At the Strzyzow site, NAAPS indicated low sulphates and strong smoke sources over Ukraine advancing towards southern Poland (Figure 5
). Thus, at this site only biomass burning is expected. The increase of the boundary layer due to a heat wave is less intensive and occurs only on 12 and 13 September (Figure 3
). It is anti-correlated with AOD (Figure 9
) and slightly decreases with an increase of the boundary layer. There is no daytime AOD variation (Figure 8
) with the Sun’s operation. At night, due to Arctic-air intrusion on 14/15 September, the boundary layer lowers to circa 1.2 km a.g.l. (as at the Warsaw site), but AOD at 614 nm lowers from 2.5 to 0.05. Over the Raciborz site, NAAPS show an influence of high sulphates advancing from Germany and smoke from Ukraine (Figure 5
), both increasing in intensity with time. The boundary layer increase due to a heat wave is less intensive than for Warsaw (up to 2.2–2.9 km a.g.l.), but persists for a longer time period on 11–14 September (Figure 4
). There is no daytime AOD variation with the Sun’s operation. Finding clear correlations of heat wave conditions with AOD is challenging. As expected from NAAPS, AOD at 614 nm slowly increased on each consecutive day from 0.12 to 0.28, until the Arctic air advection at night, of 14/15 September, when the boundary layer dropped to about 1.2 km a.g.l. and AOD drastically lowered (0.05 ± 0.01), similar to that of Strzyzow. On 16 September the AOD and boundary layer increased, mainly due to biomass burning aerosol advection.
3.4. Satellite-derived Aerosol Optical Depth Maps of Poland
For the derivation of the AOD maps over Poland for 11–16 September 2016, we used the new version of the SEVIRI AOD algorithm. Based on the AOD measurements at the five Poland AOD sites (Sopot, Rzecin, Raciborz, Strzyzow and Warsaw), 15 September 2016 was chosen as a reference day for estimating the surface properties. On each day (except on the reference day), the daytime maps of AOD at 635 nm were calculated with a fine temporal resolution of 15 min from 5:30 till 9:00 UTC (14 maps/day). Within the selected time period, the SEVIRI AOD maps (in total 70; not shown here for brevity) visually revealed a low variability of AOD at such a high resolution. Thus in Figure 10
, only the maps derived for each day at 7:00, 8:00 and 9:00 UTC are depicted. Indeed, for each day in Figure 10
, the aerosol variability obtained at each consecutive hour does not vary much. The data indicated that the highest aerosol load over Poland (AOD > 0.3) was observed on 11–13 September. The strongest spatial variability of AOD was captured on 14 September, where a sharp edge of AOD stretched approximately 100 km off and along the western and southern borders of Poland, which can be attributed to the dynamic air-clearing process (fast transport of the Arctic air-mass had already begun). The lack of data in northern and western areas of the map (white pixels) is due to cloud cover. The SEVIRI AOD maps in Figure 10
well resemble the CAMS AOD maps in Figure 6
, except for the 16 September, where CAMS clearly underestimates the AOD values and spatial variability.
The validation of the new version of SEVIRI AOD algorithm was done by comparing the obtained SEVIRI AOD values, regarded as representative for the Warsaw pixel, with the AOD values obtained from the PollyXT Lidar profiles (Table 1
) for 11–16 September 2016 (no reference day is included). During selected times, no clouds above Warsaw were observed by Lidar, except for 14 and 16 September, where some signatures of high level Cirrus clouds were captured. For further comparison, the Lidar AOD at 532 nm (VIS) and 355 nm (UV) was scaled to 635 nm, using the power law with Ångström exponent value derived at the two wavelengths. In all cases (15 values), the scaled Lidar derived AODs at 635 nm are higher for VIS and slightly lower for UV, than the SEVIRI AOD for Warsaw pixel (Figure 11
). The overestimation of the SEVIRI AOD is unlikely and caused by simple scaling of Lidar 355 nm AOD to 635 nm AOD for a large difference of the wavelengths (280 nm). The obtained good-agreement of the AOD Lidar and SEVIRI data (high correlation r2
of 0.86 for 355 nm and 0.85 for 532 nm) can be explained by the specifics of both retrieval algorithms. Firstly, for the Lidar AOD algorithm, the near-range detection unit allows for the extension of the Lidar aerosol extinction coefficient profiles down to approximately 400 m with respect to the far-range unit, that typically provides such profiles only down to about 1000 m. Secondly, the uncertainty of the Lidar-derived AOD and the SEVIRI-derived AOD decreases with the increasing aerosol load, unlike the MODIS AOD retrieval, where the uncertainty of AOD derivation is directly proportional to the aerosol load sensed [31
]. The high aerosol loads observed during the studied heat wave event resulted in a good-agreement of the Lidar and SEVIRI AOD (Figure 11
). In fact, an agreement of the obtained SEVIRI AOD at 635 nm maps with the MODIS AOD at 550 nm maps is reasonable (MODIS available only for 12, 13 and 15 September via www.polandaod.pl
; not shown for brevity); however, it is rather general, as the MODIS AOD maps are daily composites and are not directly comparable with the high temporal resolution of the derived SEVIRI AOD maps.
The SEVIRI AOD was also validated by using the AOD data sets measured at the PolandAOD sites, which operate the AERONET (at 614 nm) photometers in Strzyzow, Rzecin, and Raciborz, or the MFR-7 (at 675 nm) radiometers in Warsaw, Strzyzow and Sopot. The correlation plots of the AOD measured at four of these sites with the SEVIRI AOD pixel for each site location are plotted for an observation period of 11–16 September 2016 in Figure 12
(no reference day is included). In general, the correlation is highest for Warsaw (r2
of 0.91) despite a large difference of the measured wavelength. It is also high for Raciborz (0.84) and for Rzecin (0.8), which is in agreement with correlations reported in literature [30
]. For Warsaw and Raciborz clear separation of the data points is visible; however, different trends are discerned. In Figure 12
, the upper right sub-figure for Warsaw, a gradual AOD decrease on each consecutive day of the event is visible (on 11 September of 0.35, 12 September 0.28, 13 September 0.22 and 14–16 September 0.14). In the bottom left sub-figure the opposite for Raciborz is observed, an increase from lower AOD values on 11 September (0.18) and 12 September (0.2) to 0.3 on the remaining days. For Rzecin and Strzyzow the separation is not as clear. As expected, a low correlation and an overestimation of the SEVIRI AOD was obtained for the Strzyzow site (r2
of 0.57), which can be attributed to the high altitude elevation of this site, located at on the hill (444 m a.s.l.) and thus, a low representativeness of the large SEVIRI pixel for this site.
The obtained SEVIRI AOD is slightly underestimated in comparison to the Lidar measurements and the radiometric observations. On the one hand, this can be attributed to the inaccurate estimation of the surface properties during the reference day (15 September). The Lidar measurements confirm that this day was not a clean one (aerosol signatures are not for a typical clean background in Warsaw). Although on that day, the maximum value of the Lidar aerosol extinction coefficient at 532 nm was nearly 10 times lower than the maximum value measured on 11 September. A closer look at the Lidar derived intensive and extensive properties, indicates an existence of small particles with Ångström exponent AE (532/355) strongly oscillating around 2 within the ground and 2.4 km. During 14 September, such low values were also obtained. This, accompanied with a very low linear polarization ratio of 0.018 ± 0.04 at 532 nm, Lidar ratio of 50 ± 12 sr at 532 nm, indicates that 15 September was typical for background conditions contaminated with slight anthropogenic pollution of urban origin.
On the other hand, one has to bear in mind that the SEVIRI AODs are not directly comparable with the AODs obtained from Lidar, nor from radiometric observations. First of all, the difference in the compared AOD wavelengths has to be taken into account (even if AODs are scaled to the SEVIRI wavelength). Secondly, ground-based measurements are representative for a single point, while the SEVIRI data cover an area of around 5.5 × 5.5 km2, and thus the AOD is in fact a mean value for the whole pixel. Regarding inherent limitations of the least squares method itself, it always produces the result with the smallest sum of squares of errors, although there is no guarantee that this result has any physical meaning. In particular, if there are many outliers in the dataset, the results may have nothing to do with the actual trend line or the relationship between the phenomena described by the random variables. The least squares method is adapted to the points furthest from the mean, which can introduce the greatest error. A single outlier that is very distant from the rest will force a trend line. As outliers may be common in real data, prior to the use of the least squares method, careful data checking for outliers (scatter plot) was performed. No outliers were found/removed in the analyzed data sets.
Finally, the SEVIRI AOD was converted to the particulate matter concentrations of a size of less than 2.5 μm (the SEVIRI PM2.5) with a simple approach, by assuming that the daily mean of the SEVIRI AOD obtained by averaging all of the derived daytime AOD values for the Warsaw pixel, normalized to the value of the daytime AOD obtained by the ground-based Lidar in Warsaw (scaled to SEVIRI wavelength), shall be proportional to the daily PM2.5 derived from the satellite data normalized to the daily PM2.5 value at the surface. For normalization, the PollyXT AODs calculated from the ground level to the dynamically derived boundary layer and the surface daily mean PM2.5 data of the Regional Inspectorate of Environmental Protection (WIOS) monitoring site in Warsaw-Ursynow (http://sojp.wios.warszawa.pl/raport-dobowy-i-roczny
) were used. The obtained values of the SEVIRI PM2.5 and the surface WIOS PM2.5 at Warsaw-Ursynow site are given in Table 1
, where as expected, the satellite derived values are higher than the values obtained at the surface. Since for the SEVIRI AOD normalization the Lidar-derived AODs were calculated within the boundary layer, the contribution of the free troposphere to the SEVIRI PM2.5 can be assessed. It was in the order of 4.8 to 14.4% on all days of the event, with the exception of 13 September, when it increased to 22.9%. Hence, most of the aerosol load and particulate matter were confined to the boundary layer during the analyzed period. These results also indicate a good qualitative correspondence between the range-corrected attenuated backscatter Lidar signals (Figure 2
) and the PM2.5 concentration at the ground (Table 1
), regardless of poor (PM2.5 > 30 µg/m3
), moderate or high air-quality (PM2.5 < 10 µg/m3
). The Lidar signals intensity increases with the increasing PM2.5 concentration and/or decreases with an inflow of clean air masses to Warsaw. Moreover, both the daily mean of the SEVIRI PM2.5 data at the Warsaw pixel, as well as the daily mean surface WIOS PM2.5 data are anti-correlated with respect to the Ångström exponent (ÅE) obtained from the Lidar extinction profiles at 532 nm and 355 nm, as depicted in Figure 13
and Table 1
, whereby the poor and moderate air-quality conditions observed on the first three days of the event revealed low and moderate values of ÅE. After the air cleaning related to the inflow of the Arctic air mass on 14 September, the decrease of the PM2.5 values is accompanied with a significant ÅE increase, indicating occurrence of small size particles.