New particle formation (NPF) events have been observed all around
the world and are known to be a major source of atmospheric aerosol
particles. Here we combine 20 years of observations in a boreal forest at the
SMEAR II station (Station for Measuring Ecosystem–Atmosphere Relations) in
Hyytiälä, Finland, by building on previously accumulated knowledge
and by focusing on clear-sky (noncloudy) conditions. We first investigated
the effect of cloudiness on NPF and then compared the NPF event and nonevent
days during clear-sky conditions. In this comparison we considered, for
example, the effects of calculated particle formation rates, condensation
sink, trace gas concentrations and various meteorological quantities in
discriminating NPF events from nonevents. The formation rate of 1.5 nm
particles was calculated by using proxies for gaseous sulfuric acid and
oxidized products of low volatile organic compounds, together with an
empirical nucleation rate coefficient. As expected, our results indicate an
increase in the frequency of NPF events under clear-sky conditions in
comparison to cloudy ones. Also, focusing on clear-sky conditions enabled us
to find a clear separation of many variables related to NPF. For instance,
oxidized organic vapors showed a higher concentration during the clear-sky
NPF event days, whereas the condensation sink (CS) and some trace gases had
higher concentrations during the nonevent days. The calculated formation rate
of 3 nm particles showed a notable difference between the NPF event and
nonevent days during clear-sky conditions, especially in winter and spring.
For springtime, we are able to find a threshold equation for the combined
values of ambient temperature and CS, (CS (s
The effects of atmospheric aerosol particles on the climate system, human health and environmental interactions have raised interest in various phenomena associated with the formation, growth and loss of these particles (Pöschl, 2005; Seinfeld and Pandis, 2012; Apte et al., 2015). While primary emissions are a very important source of atmospheric aerosol particles, especially in terms of the aerosol mass loading, the particle number concentration is greatly affected by atmospheric new particle formation (NPF). During the last couple of decades, NPF has been observed to take place almost all over the world (Kulmala et al., 2004a; Zhang et al., 2011; Bianchi et al., 2016; Kontkanen et al., 2016a, 2017). Atmospheric NPF is thought to be the dominant source of the total particle number concentration (Kulmala et al., 2016) and a major source of cloud condensation nuclei in the global troposphere (Merikanto et al., 2009; Yu et al., 2010; Kerminen et al., 2012; Salma et al., 2016).
Understanding the NPF phenomenon requires understanding its precursors and pathways involved under different atmospheric conditions. For instance, high concentrations of low-volatility vapors result in a higher probability for NPF (Nieminen et al., 2015), whereas a high relative humidity and condensation sink (CS) tend to suppress NPF (Hyvönen et al., 2005; Nieminen et al., 2014). Recent laboratory experiments have shown the importance of sulfuric acid and low-volatile oxidized organic vapors to NPF (Metzger et al., 2010; Kirkby et al., 2011; Petäjä et al., 2011; Kulmala et al., 2013; Ehn et al., 2014; Riccobono et al., 2014). Additionally, atmospheric observations confirm the importance of these precursor vapors in the initial steps of NPF and in the further growth of newly formed particles (Kulmala et al., 1998; Smith et al., 2005; Kerminen et al., 2010; Paasonen et al., 2010; Ahlm et al., 2012; Bzdek et al., 2014; Nieminen et al., 2014; Vakkari et al., 2015). The Station for Measuring Forest Ecosystem–Atmosphere Relations (SMEAR II), located in Hyytiälä, southern Finland, compiles almost 21 years of particle number size distribution and extensive complementary data, providing the longest size distribution time series in the world, and hence allows for robust NPF analysis which is not readily possible at other sites. The station is located in a homogenous Scots pine forest far from major pollution sources. Hyytiälä is therefore classified as a background site representative of the semi-clean Northern Hemisphere boreal forests.
Many studies have investigated the role of different variables in causing, enhancing or preventing new particle formation (Hyvönen et al., 2005; Baranizadeh et al., 2014; Nieminen et al., 2014). In particular, Baranizadeh et al. (2014) studied the effect of cloudiness on NPF events observed at SMEAR II in Hyytiälä. They concluded, in agreement with some other studies, that clouds tend to attenuate or interrupt NPF events (Sogacheva et al., 2008; Boulon et al., 2010; Baranizadeh et al., 2014; Nieminen et al., 2015). In this study, we eliminated one variable that limits NPF (cloudiness) in order to provide a better insight into the other quantities related to atmospheric NPF. Based on 20 years of observations and data analysis for the SMEAR II station in Hyytiälä, we aim to (i) quantify the effect of cloudiness on new particle formation frequency, (ii) characterize the differences between NPF event and nonevent days during clear-sky conditions, (iii) explore the connections between new particle formation rates calculated from precursor vapor proxies and the occurrence of NPF events, (iv) formulate an equation that predicts whether a clear-sky day with specific temperature and CS is classified as an event, (v) use the clear-sky data set to calculate the NPF probability distribution based on temperature and CS.
The data used for the analysis in this study are from the University of
Helsinki SMEAR II station (Hari and Kulmala, 2005). The station provides long-term
continuous comprehensive measurements of quantities describing
atmospheric–forest–ecosystem interactions. The SMEAR II station is located
in the boreal forest in Hyytiälä, southern Finland (61
In this study, the data analysis is based on four types of measurements: (i) aerosol particle number size distributions, (ii) concentration of the trace
gases (CO, NO, NO
The aerosol number size distributions were measured with a twin DMPS (Differential Mobility Particle Sizer) system (Aalto et al., 2001) for the size ranges 3–500 nm until year 2004 and 3–1000 nm from 2005 onwards. These data were used to classify days as NPF events and nonevents following the method proposed by Dal Maso et al. (2005). The size distributions obtained from the DMPS measurements were used to calculate the CS, which is equal to the rate at which nonvolatile vapors condense onto a pre-existing aerosol particle population (Kulmala et al., 2012).
The CO concentration is measured with one infrared light absorption analyzer
(API 300EU, Teledyne Monitor Labs, Englewood, CO, USA). The NO and NO
Solar radiation in the wavelengths of UV-B (280–320 nm) and global
radiation (0.30–4.8
The formation of new aerosol particles in Hyytiälä is typically observed in the time window of several hours around noon, while this phenomenon seems to be rare during nighttime (Junninen et al., 2008; Buenrostro Mazon et al., 2016). Accordingly, aerosol number size distribution data from the DMPS measurements at around this time window are used for classifying individual days as new particle formation event or nonevent days. The classification follows the guidelines presented by Kulmala et al. (2012) and the procedure presented in Dal Maso et al. (2005). The latter uses a decision criterion based on the presence of particles < 25 nm in diameter and their consequent growth to Aitken mode. Event days are days on which sub-25 nm particle formation and growth are observed. Nonevent days are days on which neither modes are present. Undefined days are the days which do not fit either criterion.
The cloudiness parameter (
The gaseous sulfuric acid concentration is estimated from a
pseudo-steady-state-approximation proxy developed by Petäjä et al. (2009). This proxy takes into consideration the sulfuric acid source and
sink terms as
The concentration of monoterpene oxidation products, called oxidized organic
compounds (OxOrg) here, is estimated using a proxy developed by
Kontkanen et al. (2016b). This proxy is calculated by
using the concentrations of different oxidants (the measured ozone
concentration [O
The formation rate of nucleation mode particles (
The particle growth rate over the particle diameter range of 1.5–3 nm was
calculated by taking into account the size of the condensing vapor molecule
size and the thermal speed of the particle (Nieminen et al.,
2010). The growth rates (1.5–3 nm) were calculated as 30 min averages
and as the sum of the growth rates due to the sulfuric acid (SA) vapor and
OxOrg vapor condensation. The density of the particle was assumed to be
constant (1440 kg m
Air mass trajectories were calculated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT_4) model at 96 h backward trajectories at 100, 250 and 500 m arrival heights once per hour.
We studied NPF events as a function of cloudiness. Figure 1a shows the
fraction of event, nonevent and undefined days as a function of cloudiness
parameter. We can see that clear-sky conditions favor the occurrence of NPF:
the fewer clouds there were, the higher was the fraction of NPF event days.
For instance, for days with the cloudiness parameter of 0.3 or less, the
fraction of event days was less than 0.1 of the total classified days.
However, the fraction of NPF event days reached a maximum of around 0.55
during complete clear-sky conditions (
Monthly variation of cloudiness daily (09:00–12:00) medians and
percentiles recorded during NPF events (E; white) and nonevents (N; shaded).
Numbers below the plot correspond to the number of data points included in
each box plot. Number of clear-sky events (E (
Our results emphasize the fact that radiation favors the occurrence of NPF, while clouds tend to decrease the probability of NPF. Undefined days were observed under cloudiness conditions that fell between those for NPF events and nonevents. In general, undefined days can be interrupted NPF events or unclassified plumes of small particles due to pollution (Buenrostro Mazon et al., 2009). The interruption of a NPF event can be due to a change in the measured air mass or to the attenuation of solar radiation caused by the appearance of a cloud during the event. We will not consider undefined days further in our analyses.
The monthly variation of daily median cloudiness parameter within the time window of 09:00–12:00 during the classified days is shown in Fig. 2. Spring showed the best separation between the events and nonevents in terms of the cloudiness parameter, while the separation became weaker during the summer and especially for June and July. Taken together, Figs. 1 and 2 emphasize the observation that the presence of clouds decreases the probability of NPF events.
Upon visualizing the cloudiness conditions during events and nonevents, we
chose a fixed constraint for clear-sky conditions (
The monthly distribution of the event fraction on clear-sky days appeared as double peaks in spring and autumn, with spring having a higher fraction of events (Fig. 3a). The minimum fraction of NPF events was recorded in December. The fraction of nonevent days peaked in winter with another peak in summer. The total number of NPF events varied from year to year between 1996 and 2015. However, this variation did not show any specific trend of frequency (Fig. 3b), which is in agreement with previous statistics reported from studies that did not consider clear-sky classification (Nieminen et al., 2014).
Since NPF is most frequent in spring, we dedicated our focus to this season (Fig. 3a). The springtime medians and percentiles of air mass trajectories arriving at Hyytiälä during clear-sky NPF events and nonevents were calculated 96 h backward in time at the 100, 250 and 500 m arrival heights for the years 1996–2015. The medians and similarly the percentiles were calculated by taking the median compass direction at every point on the trajectory (1 h between every two points), arriving every half an hour at Hyytiälä. The trajectories arriving at Hyytiälä at these three heights were quite similar, and those arriving at the 500 m height are shown in Fig. 4. Medians and percentiles of the routes were calculated by taking the median of the trajectories at every half hour for springtime NPF event days and nonevent days separately. During the NPF event days, the measured air masses were found to originate mainly from the north and passed over Scandinavia before arriving at Hyytiälä. Similarly to previously reported results, air masses arriving from the north and north-west directions result in clean air with low pollutant (particulate matter and trace gas) concentrations (Nieminen et al., 2015). During NPF the nonevent days, air masses originated from more polluted areas in Europe and Russia, resulting in elevated levels of condensation sink and other air pollutants in Hyytiälä, as also seen in previous studies (Sogacheva et al., 2005).
Median and percentiles of 96 h backward air mass trajectories arriving at Hyytiälä during springtime (09:00–12:00).
In Fig. 5a we present the monthly variation of condensation sink during NPF events and nonevents under daytime clear-sky conditions. NPF events tended to be favored by low values of CS throughout the year. In all months except during summer, the 75th percentile of the event day values of CS was lower than the 25th percentile of the nonevent day values of CS. On the NPF event days, CS had its maximum in summer, which might be one of the main reasons for the local minimum in the NPF event frequency during the summer months (Fig. 3a). However, the monthly cycle of CS during nonevent days had two maxima, one in spring and another one in autumn, which might suggest that during these seasons, high values of CS prevented NPF occurring on particular days. The difference in the value of CS between the NPF event and nonevent days was the highest in March and the lowest during the summer months.
Correlation coefficients between different meteorological parameters, gas concentrations and condensation sink (CS) during clear-sky events and nonevents during spring (March–May, 1996–2015) and time window 09:00–12:00. High positive and negative correlations are marked in bold.
Median and percentiles of monthly variation (09:00–12:00) at
Figure 5b shows the monthly temperature conditions (
As with an earlier study (Hamed et al., 2011), our results indicate
that NPF is favored by low values of ambient relative humidity in
Hyytiälä (Fig. 5c). This observation does not conflict with chamber
experiments (e.g., Duplissy et al., 2016) or theory
(Merikanto et al., 2016; Vehkamäki et al., 2002), which suggest
higher nucleation rates at higher values of RH, because binary
H
Springtime (months 3, 4, 5) medians and percentiles of trace gases during clear-sky events (E, white) and nonevents (n, shaded) during daytime (09:00–12:00). See Fig. 1 for explanation of symbols.
After looking at the characteristics of clear-sky NPF event and nonevent days
in terms of meteorological parameters and CS, we looked at the variation of
trace gas (CO, SO
In this study, we determined
Monthly variation of medians and percentiles of
Being a function of temperature, the OxOrg proxy concentration was generally found to follow the monthly cycle of the ambient temperature. The median value of [OxOrg] was higher on NPF events days in every month compared with nonevent days (Fig. 7b). The largest difference in [OxOrg] between the NPF events and nonevents, in terms of its median value, was recorded for January and the least difference was recorded for May. It is to be noted that the proxy values represent the measured values less accurately during winter than during the other periods (Kontkanen et al., 2016b).
The calculated new particle formation rate,
Since previous studies have shown that there is a clear difference in
observed
Diurnal cycle of median values of calculated formation rate of 3 nm
particles (
Relationship between temperature and CS during springtime
(11:00–12:00) NPF clear-sky (
In Fig. 10 we present the median diurnal cycles of
On NPF event days, the median-approximated formation rate of 3 nm particles
had its maximum value at about midday and was significantly higher than on
nonevents days (Figs. 9b and 10). A clear negative relation could be seen
between the median seasonal diurnal cycles of CS and
Since quite a visible separation could be observed in the calculated values
of
The separation between the clear-sky NPF events and nonevents in the CS
versus
Relationship between CS and temperature (time window: 11:00–12:00) NPF clear-sky event days and nonevent days. Horizontal line is calculated from spring LDA at 95 % confidence relative to nonevents and is demonstrated by Eq. (6).
NPF probability distribution based on the CS and temperature conditions during clear-sky days (11:00–12:00). Marker size indicates number of days included in the probability calculation within every cell.
Since the biggest difference in the calculated 3 nm particle formation rates
between the NPF events and nonevents was observed around noon (Fig. 9b), and
since CS and temperature showed promising threshold values for predicting the
occurrence of NPF nonevents during spring (up to 95 %) (Fig. 11), Fig. 13
presents the probability of having a NPF event in Hyytiälä at a
specific CS and temperature within the time window 11:00–12:00. The
probability was calculated by taking the fraction of events to the total
events and nonevents in every cell which is 2.5 K on the
In this study we combined 20 years of data collected at the SMEAR II station in order to characterize the conditions affecting the frequency of NPF events in that location. By focusing only on clear-sky conditions, we were able to get a new insight into differences between the NPF events and nonevents. In clear-sky conditions, the meteorological conditions, trace gas concentrations and other studied variables on NPF event days appeared to be similar to those presented in the previous studies which did not consider clear-sky classification. Furthermore, the monthly data refined the analysis so that the differences caused by different quantities became more visible compared the previous studies conducted for this site. Our work confirms the conclusions of Baranizadeh et al. (2014) with a complementary data set: NPF events and nonevents are typically associated with clear-sky and cloudy conditions, respectively.
Our results showed that using SA and OxOrg proxies to calculate the apparent formation rates of 1.5 and 3 nm particles works well in differentiating the clear-sky NPF events from nonevents. Moreover, during clear-sky conditions the effect of CS on attenuating or even preventing NPF was quite visible: CS was, on average, two times higher on the nonevent days compared with the NPF event days. Similarly, many other meteorological variables affected NPF. By using CS and ambient temperature, we were able to find a threshold above which no clear-sky NPF events occurred. This threshold is described with an equation that is able to separate 97.4 % of the NPF events from nonevents during springtime. In clear sky conditions, when there is plenty of radiation available, NPF events take place as long as the CS is low enough and temperature is moderate. Although a weaker separation was observed in the other seasons, considering only clear-sky conditions enabled us to form a map of the probability of having a NPF event within specific CS and temperature conditions. Using clear-sky conditions appears to bring us one step forward towards understanding NPF and predicting their occurrences in Hyytiälä. Our study serves as a basis of future detailed comparisons with observations to formulate even more robust conclusions.
Data measured at the SMEAR II station are available on the
following website:
The authors declare that they have no conflict of interest.
This work was supported by the Academy of Finland Centre of Excellence program (grant no. 272041) and Nordic Top-level Research Initiative (TRI) Cryosphere-Atmosphere Interactions in a Changing Arctic Climate (CRAICC). Lubna Dada acknowledges the doctoral programme in Atmospheric Sciences (ATM-DP, University of Helsinki) for financial support. We also thank Ksenia Tabakova for providing air mass trajectory data. Edited by: D. Spracklen Reviewed by: two anonymous referees