Aerosol particle formation in the upper residual layer

Abstract. According to current estimates, atmospheric new particle formation (NPF) produces a large fraction of aerosol particles and cloud condensation nuclei in the earth’s atmosphere, therefore having implications for health and climate. Despite recent advances, atmospheric NPF is still insufficiently understood in the upper parts of the boundary layer (BL). In addition, it is unclear how NPF in upper BL is related to the processes observed in the near-surface layer. The role of the topmost part of the residual layer (RL) in NPF is to a large extent unexplored. This paper presents new results from co-located airborne and ground-based measurements in a boreal forest environment, showing that many NPF events (∼42 %) appear to start in the upper RL. The freshly formed particles may be entrained into the growing mixed layer (ML) where they continue to grow in size, similar to the aerosol particles formed within the ML. The results suggest that in the boreal forest environment, NPF in the upper RL has an important contribution to the aerosol load in the BL.



Introduction
It has been estimated that atmospheric new particle formation (NPF) is responsible for most of the cloud condensation nuclei (CCN) in the atmosphere (Dunne et al., 2016;Gordon et al., 2017;Pierce and Adams, 2009;Yu and Luo, 2009). Aerosol-cloud interactions, in turn, have important but poorly-understood effects on climate (Boucher et al., 2013). Being a major source of ultrafine aerosol particles in many environments (e.g. Brines et al., 2015;Posner and Pandis, 2015;Salma et al., 2017;Yu et al., 2019), NPF may have implications for human health.
The majority of NPF observations come from ground-based measurements Kulmala et al., 2004), which can be argued to represent NPF within the mixed layer (ML).
Measurements from aircrafts (e.g. Clarke and Kapustin, 2002;Rose et al., 2017) and high-altitude research stations (e.g. Bianchi et al., 2016) demonstrate that NPF frequently takes place in the free troposphere (FT). Entrainment of particles formed in the upper FT was identified as an important source of CCN in the tropical boundary layer (BL) Williamson et al., 2019).
To what extent NPF happens in the lower FT and in the upper parts of the BL is not clear. Freshly formed particles were observed in the inversion capping a ML Platis et al., 2015;Siebert et al., 2004) and in turbulent layers inside the residual layer (RL) (Wehner et al., 2010). NPF was frequently observed in the lower FT over a megacity in a year-long campaign (Quan et al., 2017). Also Qi et al. (2019) reported NPF just above the ML over Yangtze River Delta. In the marine BL, sub-10 nm particles were observed in the entrainment zone above a cloud topped BL (Dadashazar et al., 2018). Layers of sub-10 nm particles, usually less than 500 m in thickness, were often observed in the lower FT over a boreal forest environment (Leino et al., 2019;Schobesberger et al., 2013;Väänänen et al., 2016). On the other hand, Junkermann and Hacker (2018) attributed their observations of ultrafine particle layers to flue gas emissions from stacks with subsequent chemistry taking place during air mass transport over long distances.
In this study we used co-located airborne and ground-based measurements to study NPF in the BL over a boreal forest. We aimed to answer the following questions: (1) where, how often and why does NPF take place in the upper parts of the BL, and (2) how the upper-BL NPF is related to ground-based observations, and what implications this has for data interpretation.

Airborne measurements
We used data from airborne measurement campaigns conducted between 2011 and 2018 around Hyytiälä, Finland. Figure 1 shows the data availability from these measurements. Most of the flights were carried out during spring and early autumn because that is when NPF events are most common in Hyytiälä. Here we focused on the data that was measured within a 40-km radius from Hyytiälä.
The measurement setups changed slightly over the years. Detailed descriptions of the setups on board can be found in our previous studies (Leino et al., 2019;Schobesberger et al., 2013;Väänänen et al., 2016).
The instrumented aircraft was a Cessna 172 operated from the Tampere-Pirkkala airport (ICAO: EFTP). The sample air was collected through an outside inlet into a main sampling line that was inside the aircraft's cabin. The forward movement of the aircraft during flight provided adequate flow rate inside the main sampling line. The flow rate was maintained at 47 lpm by using a manual valve. The instruments drew air from the main sampling line using core sampling inlets. The necessary flow rate to the instruments was provided by pumps. The flow rate in the main sampling line corresponded to roughly isokinetic sampling at the core sampling inlets. The airspeed was kept at 130 km/h during the measurement flights.
The on-board aerosol instruments considered in this study were an ultrafine condensation particle counter (uCPC,TSI,model: 3776), measuring the >3 nm particle number concentration at a 1-s time resolution, a particle size magnifier (PSM, Airmodus, model: A10) operated with a TSI 3010 CPC, measuring the >1.5 nm particle number concentration at a 1-s time resolution, and a custombuilt scanning mobility particle sizer (SMPS) with a short Hauke type DMA and a TSI 3010 CPC, measuring the aerosol number size distribution in the size range of 10-400 nm at a 2-min time resolution. In addition, basic meteorological data (temperature, relative humidity and pressure) and water vapor concentration from Licor Li-840 gas analyzer were used.
Vertically, the measurement profiles extended approximately from 100 m to 3000 m above the ground. This altitude range covered the ML, RL and roughly 1 km of the FT (Figure 2). The measurement flights lasted about 2-3 hours and were flown mostly during the morning (~8:00-12:00 local time) and the afternoon (~13:00-16:00 local time). Horizontally, the profiles were flown perpendicular to mean wind in order to avoid the airplane's exhaust fumes.

Ground-based measurements
Comprehensive atmospheric measurements have been done at the SMEAR II station in Hyytiälä (61°50'40'' N, 24°17'13'' E, 180 m above sea level) since 1996 (Hari and Kulmala, 2005).The landscape around the site is flat and dominated by Scots pine forests, with small farms and lakes scattered nearby. The station represents typical rural background conditions. We used data from the BAECC (Biogenic Aerosols-Effects on Clouds and Climate) campaign, which took place in Hyytiälä during Feb-Sep 2014 , to study the relationship between BL evolution and NPF observed at the station. High spectral resolution lidar (HSRL) measurements and meteorological balloon soundings released every 4 hours by the U.S. Department of Energy ARM mobile facility allowed us to monitor the evolution of the BL (Nikandrova et al., 2018).
From the HSRL data we looked at the values of backscatter cross section in order to see the development of the ML during the day. The data were averaged into 30-m altitude bins and 10-min temporal bins. The ground-based measurements during the BAECC campaign were also supplemented by aircraft measurements using the instrumented Cessna. In case of missing soundings, we also looked at the balloon soundings released from Jokioinen ~120 km south-west from Hyytiälä (WMO: 02963).
The number size distribution of aerosol particles between 3 and 1000 nm was measured at the station using a differential mobility particle sizer (DMPS, Aalto et al., 2001). A neutral cluster and air ion spectrometer (NAIS, Airel Ltd., Mirme and Mirme, 2013) measured the number size distribution of air ions and particles in the size ranges of 0.8-42 nm and 2-42 nm, respectively (Manninen et al., 2009). The time resolutions of the DMPS and NAIS were 10 min and 4 min, respectively. The vertical flux of particles >10 nm was measured by the eddy covariance method from 23 m above ground, which is a couple of meters above the canopy (Buzorius et al., 2000). low aerosol concentration at Hyytiälä, but TKE dissipation rate can be retrieved on most days up to the top of the BL.

Results and discussion
In the aircraft data we frequently observed a layer of nucleation mode (sub-25 nm) particles above the ML. First we introduce how the phenomenon was observed in the airborne and ground-based measurements using a case study. Then we show that the particle layers occurred in the topmost part of the RL, by studying the average vertical profile of particle number-size distribution and temperature as well as the BAECC data. Finally, by using the BAECC data, we associate the nucleation mode particles in the upper RL to a specific signal in the ground-based measurements and use the observations at the SMEAR II station to gather long-term statistics.

Case study: May 2, 2017
On May 2 during the measurement airplane's ascend over Hyytiälä, we observed a layer of freshly formed aerosol particles approximately between 1200 and 2000 m above the ground, in the top parts of the ML (Figure 3). The layer had increased number concentrations of sub-20 nm and sub-3 nm particles. The small size of the particles suggests that they were recently formed in the atmosphere.
The lower edge of the aerosol particle layer was observed at 12:24 UTC. The airplane entered back into the ML at 12:56 UTC and at this point there were no signs of the particle layer, but the particle number concentration had increased inside the ML. On the same day, an early morning flight before the sunrise was also performed. During this flight no elevated particle layer was observed below 3000 m, suggesting that this particle layer had been formed after the sunrise. The air masses came from a non-polluted sector over the Arctic Ocean and northern Scandinavia.
After the aerosol layer was observed from the airplane during the ascend, a new particle mode with a geometric mean diameter of about 10 nm suddenly appeared at the ground-level at 12:36 ( Figure   4). The appearance of this new particle mode was characterized by a negative peak in the vertical particle flux, suggesting that the particles had been mixed down from aloft.
We then studied the vertical profiles of meteorological quantities measured on board the Cessna on TKE dissipation rate clearly reveals the development of the ML on both days. On May 1 the ML reached roughly 1700 m above the ground, while on May 2 the first potential temperature profile measured on board the Cessna revealed the presence of a stable layer (upper RL) at roughly the same altitude. This matches with the height of the aerosol particle layer in Figure 3. The Doppler lidar measurements further show that on May 2 the ML reached this height around the noon UTC, which is when the particle layer was observed to be mixing down. This leads us to hypothesize that NPF was taking place in the upper RL.

Evidence of NPF in the upper RL based on long-term measurements
We analyzed the airborne data measured during 2011-2018. We plotted the median and 75 th percentile number size distributions measured on board the aircraft as a function of altitude during NPF event days (65 days out of 130 measurement days) between 09:00 and 12:00 ( Figure 6). NPF event days are characterized by a new growing particle mode appearing in the sub-25 nm size range (Dal Maso et al., 2005). If aerosol formation in the upper RL occurs on less than half of the NPF event days, it might not be visible in the median plot, but might still appear in the 75 th percentile plot.
Interestingly, in the 75 th percentile plot a layer of nucleation mode particles is observed at 2500-3000 m above sea level. In the mean temperature profile, an inversion is observed at the same altitude level. The ML and RL are commonly capped by temperature inversions (Stull, 1988).
In this case, the inversion is likely where on average the top of the RL was, since the the top of the ML was well below this altitude. The probable reason for the unusually deep RL is that the NPF event days tend to be sunny spring days and the ML can grow exceptionally high, which also leads to a very deep RL. The vertical profile of particle number size distribution supports the idea that NPF was taking place in the upper RL.

Connection between NPF in the upper RL and ground-based observations
With the BAECC dataset we wanted to investigate whether the sudden appearance of nucleation mode particles with downward particle flux was associated with the ML reaching the upper RL.
This would not only test the hypothesis that NPF happens in the topmost part of the RL, but also provide us with a tool to identify upper RL NPF from the ground-based data alone. We looked for cases where a new particle mode suddenly appeared in the nucleation mode size range during the daytime and the appearance of the particles was associated with a downward particle flux. We noted the times when the particles first appeared, and also estimated a confidence interval of the observation. Then we checked if we could find out the height of the RL from balloon soundings or the Cessna flights. We looked for an elevated temperature inversion that was roughly at the same altitude as the ML of the previous day had reached. We noted the base height of the temperature inversion and took this as the top of the RL. Then we followed the height of the new ML from the HSRL measurements and noted the time when the ML reached the inversion base, also estimating a confidence interval. Figure 7 illustrates an example for this procedure.
We found 8 cases during the campaign where the analysis could be fully carried out. Figure 8 shows a strong positive correlation between the new particle mode appearance time and the time when the ML reached the top of the RL. This suggests that the suddenly appearing nucleation mode particles were entrained into the ML from the upper RL. We found only a weak positive correlation between the new particle mode appearance time and the geometric mean diameter of particles in the new mode at the moment they were first observed. This is probably explained by the NPF starting at different times during the day and variability in growth rates, coupled with the small sample size.

Implications for classifying NPF events
Previous studies that classified NPF events observed in Hyytiälä have collected statistics on the occurrence of suddenly appearing particle modes. For example Buenrostro Mazon et al., (2009) collected statistics on "tail events" where a new particle mode appears at particle diameters greater than 10 nm and grows for several hours. Dada et al., (2018) collected statistics on "transported events" where elevated number concentration of 7-25 nm particles persisted for more than 1.5 hours, but no elevated number concentrations at smaller particle sizes were observed. It was found that ~36% of the NPF events observed for over 10 years in Hyytiälä were transported events. They occurred especially when the conditions inside the ML were less favorable for nucleation.
Here we found cases in the SMEAR II data between 2013 and 2017, in which a new growing particle mode suddenly, without continuous growth from smallest detectable sizes (3 nm), appears in the nucleation mode and is associated with a negative peak in the vertical particle flux. We also noted cases where a new particle mode appears with a continuous growth from the smallest The monthly fractions of the different cases are shown in Figure 9. We found that NPF within the ML occurred on 13% of all the days and NPF in the upper RL on 7% of all the days. During spring (Mar-May) the corresponding percentages were 31% and 17%. On many days NPF took place both in the upper RL and within the ML. According to this analysis, NPF in the upper RL constitutes 42% of the NPF event days in Hyytiälä. Moreover, on 16% of the NPF event days NPF only took place in the upper RL but not in the ML.
The gaseous precursors involved in NPF may end up in the upper RL because of mixing from the surface during the previous day (e.g. organic vapors emitted from the forest or sulfuric acid, ammonia and amines originating from human activities) or because of long-range transport in the FT (e.g. iodine oxides from the ocean).
Many factors favor NPF at higher altitudes, including enhanced photochemistry, reduced sinks and reduced temperature. However, the unique NPF inducing features of the upper RL are probably linked to the mixing that takes place in the interface between RL and FT. For example Nilsson and Kulmala, (1998) found that mixing two air parcels with different initial temperatures and precursor vapor concentrations can lead to a considerable increase in the nucleation rate.
If the new ML reaches the upper RL, particles formed originally in the RL will be mixed into the ML where they continue to grow in size as low-volatility vapors present in the ML are able to condense onto these particles. The processes are illustrated in Figure 10. In case the particles will not be mixed down, they may persist in the FT for a longer time period and possibly have stronger contribution to cloud formation.

Conclusions
We measured aerosol particles, trace gases and meteorological parameters on board an instrumented Cessna 172 over a boreal forest in Hyytiälä, Finland. The airborne data was complemented by the continuous, comprehensive ground-based measurements at the SMEAR II station. We found multiple evidence that NPF frequently takes place in the topmost part of the RL. This is likely related to the unique thermodynamic conditions present in this layer due to mixing between RL and FT air. We estimate that NPF in the upper RL occurs on 42% of the NPF event days in Hyytiälä. Our results provide new information on NPF in the BL and they should be taken into account when interpreting and analyzing ground-based as well as airborne measurements of aerosol particles. Data availability: The particle flux and DMPS data can be accessed from https://avaa.tdata.fi/web/ smart/smear last access: Oct 1, 2020). The BAECC HSRL and radiosonde data is available from https://adc.arm.gov/discovery/ (Bambha et al., 2014;Keeler et al., 2014); last access: Oct 1, 2020). The Jokioinen soundings can be accessed using the Finnish Meteorological Institute's open data service https://en.ilmatieteenlaitos.fi/open-data (last access: Oct 1, 2020). The ERA5 dataset can be accessed from https://cds.climate.copernicus.eu/cdsapp#!/home (last access: May 6, 2020). The rest of the data was gathered into a dataset that can be accessed from https://zenodo.org/record/4063662#.X3cHQnUzY88 (Lampilahti et al., 2020;last access: Oct 2, 2020).       https://doi.org/10.5194/acp-2020-923 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. Figure 6: Panel A shows the median and panel B the 75th percentile vertical profile of particle number-size distribution measured on board the Cessna on NPF event days between 9-12 AM. The number-size distribution was binned into 200 m altitude bins. The data is from the campaigns conducted between 2011 and 2018. It includes only the data that was measured within 40 km radius from Hyytiälä. The dashed line is the mean ML height obtained from the ERA5 reanalysis data. The blue line is the mean temperature profile measured on board the airplane. 20 https://doi.org/10.5194/acp-2020-923 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.
Figure 7: Panel A shows the backscatter cross section measured by the HSRL. The development of the ML is visible from the backscatter cross section signal. Temperature and potential temperature profiles from the 4-hourly balloon soundings are superimposed. The horizontal line rl_h refers to the height of the inversion base observed during the early morning of July 5 th . The bold temperature and potential temperature profiles mark the sounding from which rl_h was determined. The rl_t and crl_t refer to the time when the ML was estimated to reach the rl_h and the confidence interval for this time respectively. Panel B shows the particle number-size distribution measured at the SMEAR II station, the black line is the vertical particle flux. The mode_t and cmode_t respectively refer to the time and the confidence interval, when a nucleation particle mode that is associated with downward particle flux suddenly appears. 21 https://doi.org/10.5194/acp-2020-923 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. Figure 8: The correlation between the times that a new particle mode coupled with downward particle flux is observed at the field site and the times that the ML reaches the top of the RL. Figure 10: Schematic drawing illustrating the proposed mechanism behind NPF in the upper RL. Gaseous precursors released from the surface are mixed throughout the ML. When the mixing stops during the night the gases are stuck in the RL. Also gaseous precursors may be transported in the FT. In the following morning photochemistry and the thermodynamically favorable conditions in the upper RL initiate NPF. The freshly formed particles remain in the elevated layer or get mixed into the a new ML if it reaches the height of the upper RL. The aerosol particles continue to grow larger, contributing to the aerosol load in the BL.