Climatology of New Particle Formation and Corresponding Precursors at Storm Peak Laboratory

Thirteen years of measurements of ultrafine (3–10 nm diameter) aerosols are presented from a remote high elevation (3210 m a.s.l.) site in Colorado, Storm Peak Laboratory. Previous work has shown that frequent new particle formation (NPF) occurs regularly at the site (52% of days). This long-term climatology of ultrafine aerosols clearly shows a seasonal dependence on new particle formation at Storm Peak Laboratory, reaching a maximum during the spring season and a minimum in summer. Recent sulfur dioxide data indicates a strong source region west of Storm Peak Laboratory, and this wind direction corresponds to the predominant wind direction observed during NPF events.


INTRODUCTION
Atmospheric aerosols impact climate, air quality and visibility, and human health (e.g., Akimoto, 2003).Aerosols affect the Earth's radiation balance directly by scattering sunlight and indirectly through their role as cloud condensation nuclei (CCN) (Twomey, 1974;Twomey et al., 1984;Albrecht, 1989;Charlson et al., 1992).Current estimates of aerosol direct and indirect effects remain uncertain because of our inability to accurately estimate the spatial and temporal distributions of aerosol concentrations, size, and composition (Boucher et al., 2013).Radiative effects are primarily related to particles larger than ~50-100 nm, while smaller (< 10 nm) sized particles also have adverse health effects (e.g., Oberdorster et al., 1992;Peters et al., 1997) and are therefore a serious air quality issue.
Aerosol nucleation (formation of aerosol particles directly from gas phase species) is an important atmospheric process that can affect cloud formation.Spracklen et al. (2010) showed that a global aerosol microphysics model was unable to reproduce the observed annual cycles in aerosol number concentrations if nucleation was not included.Merikanto et al. (2009) and Yu and Luo (2009) showed that aerosol nucleation can indeed have a great impact on the number of the global CCN.A global study showed that nucleation in the free troposphere contributes to 35% of CCN that forms low -level clouds at supersaturation ratio of water of 0.2% (Merikanto et al., 2009).In general, the contribution of primary particles to CCN decreases at higher elevations with respect to the contribution from nucleation (Merikanto et al., 2009).Model results by Yu and Luo (2009) showed that secondary particles formed via nucleation in sub-grid SO 2 plumes contribute up to 50-70% of CCN.These results suggest that high altitude NPF can play a critical role in cloud formation and hence climate.

NPF AT MOUNTAIN SITES
There are several potential underlying principles that allow nucleation to occur more frequently at high elevations.First, increased levels of ultra-violet (UV) radiation have been associated with NPF.Previous studies have shown that increased UV irradiance is associated with increased atmospheric OH radical concentrations, which drive the formation of sulfuric acid, a key chemical species known to be involved in aerosol nucleation (e.g., Berndt et al., 2005).Most recently, NPF data from Storm Peak Laboratory (SPL) was used within a global chemical transport model to illustrate the critical role that sulfuric acid plays in the initial nucleation process (Yu and Hallar, 2014).Increased radiation may also allow for increased plant production and emissions of biogenic volatile organic compounds (BVOC) (e.g., Guenther et al., 2006), which potentially impact nucleation and provide condensable species needed for particle growth (Bonn et al., 2008).Previous measurements at SPL, showed higher level of UV irradiance on days with identified NPF, compared with days without these events (Hallar et al., 2011).NPF events were also correlated with UV radiation during a one-year study at Jungfraujoch in the Swiss Alps (3580 m a.s.l.) (Boulon et al., 2010).Secondly, high elevations tend to have lower temperatures and lower surface area of pre-existing aerosol particles than at lower elevations.Nucleation is a non-linear process and temperature is a key thermodynamic parameter (Yu, 2010).At sufficiently low temperatures, it is even possible that the nucleation barrier can disappear and nucleation can take place spontaneously (Yu, 2010).On the other hand, pre-existing aerosol surface area is a sink for nucleation; since nucleation precursors preferentially condense on larger aerosols and newly formed small particles will quickly coagulate with large particles (Kulmala, 2003).Thus, lower temperature and lower surface area of aerosol particles, together with higher precursor concentrations due to convection and enhanced irradiance, can create excellent conditions for nucleation at high elevations.
Additionally, high elevations may provide a favorable thermodynamic condition for NPF due to turbulence, atmospheric waves, storms, and other air mixing processes (e.g., Nilsson and Kulmala, 1998).Turbulence on various scales including the presence of boundary layer turbulence (Nilsson et al., 2001), atmospheric waves (Nilsson et al., 2000), and mixing across temperature and humidity gradients (Kulmala et al., 1998) may promote nucleation.These mixing conditions can create higher supersaturation ratios of nucleation precursors.Recent evidence also supports the role of turbulence in NPF (Wehner et al., 2010).NPF may be also promoted by smaller-scale turbulence associated with surface friction in orographic flow at high elevation ground locations (Boulon et al., 2011).
Measurements above the boundary layer are needed to evaluate the global effects of aerosols on climate.The only high-altitude measurement approach that allows for continuous, long-term observations is the use of mountaintop laboratories (Weber et al., 1995).Using two different high altitude sites, the Puy de Dôme research station (1465 m a.s.l.) and the Opme Station (660 m a.s.l.) in central France, a recent study illustrated the importance of vertical extent for nucleation events (Boulon et al., 2011).At this station Boulon et al. (2011) found a NPF frequency of 36% during a 157-day study, and the NPF frequency was greater (36%) at the higher elevation site compared to the lower elevation site (21%).Using LIDAR measurements, this work highlighted that NPF occurs frequently above the boundary layer and supports the conclusion that nucleation is promoted at high elevations (Venzac et al., 2008).At Jungfraujoch, aerosol nucleation was found on 18% of the 226 measurement days from 2008 to 2009 (Boulon et al., 2010).Venzac et al. (2008) reported a nucleation frequency of 35% during a 511-day study at the Nepal Climate Observatory (5,079 m a.s.l.).Neitola et al. (2011) reported four years of measurements from Mukteshwar in the Himalaya foothills (2180m a.s.l.), where the highest daily NPF frequency was found during the spring (with over 80% of all events), and overall NPF occurred 14.5% of days.García et al. (2014) presented a four-year data set from Izaña Mountain (2373 m a.s.l.) in the North Atlantic demonstrating NPF on 30% of days.There was a clearly marked NPF season at Izaña in the summer (May to August), where events accounted for 50-60% of the days.Most recently, Rose et al. (2015) presented one-year of data from highest in-situ aerosol measurement site in the world, Chacaltaya (5240 m a.s.l.), Bolivia.This site showed remarkably high NPF frequency of 64%, with a clear seasonal dependence.NPF was found primarily in the dry season (May-October), and commonly multiple events were identified per day.Overall, the seasonality of NPF at high elevation sites varies significantly with local conditions.
Frequent NPF has been observed regularly at Storm Peak Laboratory.These events occurred during 52% of the 474 measurement days from 2001 to 2009 (Hallar et al., 2011).This work uses a long-term record of ultrafine aerosol concentration to determine the seasonality of NPF at Storm Peak Laboratory, and consider local precursors of aerosol nucleation.

Location
Storm Peak Laboratory (SPL, 3210 m a.s.l.; 40.455°N, 106.745°W), operated by the Desert Research Institute (DRI), is located on the west summit of Mt.Werner in the Park Range near Steamboat Springs in northwestern Colorado.This site has been used in cloud and aerosol studies for more than 25 years (e.g., Borys and Wetzel, 1997;Lowenthal et al., 2002).SPL is situated at treeline on a 70 km ridge oriented perpendicular to the prevailing westerly winds.SPL is 1120 meters above the nearest population center; at this mountain-top remote location studies of the long-term trends aerosol properties have been made (e.g., Asmi et al., 2013).

Meteorological Data
Meteorological instruments used in this study include research grade temperature, pressure, wind speed and direction, and relative humidity sensors (Campbell Scientific, Inc., Met One and Vaisala).These data are uploaded to the Western Regional Climate Center (http://www.wrcc.dri.edu/) database every 5 min.

Aerosol Data
Particles were sampled from an insulated, six-inch diameter manifold within four feet of its horizontal entry point through an outside wall.The 4 m high vertical section outside the building is capped with an inverted can to exclude large particles (D > ~8 µm) as well as rain and precipitation.
Aerosol size distributions were measured with a TSI SMPS 3936L22 (scanning mobility particle sizer with a TSI 3022 CPC) for particles with diameters between 8 and 330 nm, and with a nano-SMPS (TSI 3936N76 with a TSI 3776 CPC) for particles with diameters between 3 and 96 nm.For the nano-SMPS and the SMPS, sheath and sample flow rates were 10 L min -1 and 1 L min -1 , respectively.All SMPS and nano-SMPS distributions were corrected for multiple-charge and diffusional losses.Five-minute scans were collected from the SMPS and nano-SMPS.The SMPS dataset started in 2001, although is not continuous.The nano-SMPS data set was collected from December 2010-April 2011.Long-term SMPS data from SPL is primarily available from the spring and winter months, with a smaller dataset representing the summer months.Due to limited access to SPL, combined with measurement complexity, very little long-term SMPS data is currently available representing the fall season.For this study, non-continuous data from April 3, 2013 to July 12, 2014 are presented in detail, as this period overlaps with new measurements of sulfur dioxide (discussed below).This subset of SMPS data primarily represents the winter season (61% of available data) and summer season (28% of available data).The fall season represents only 10% of the available data, and the spring season represents less than 1% of the available data (due to an instrument failure).
Condensation nucleus (CN) concentrations were measured with TSI 3025 and 3010 condensation particle counters (CPC) for particles with diameters greater than 3 and 10 nm, respectively.Concentrations of ultrafine particles (UFCN) were determined by the difference.CN concentrations were averaged and recorded every five minutes, prior to 2011.After 2011, the CN concentrations were averaged and recorded ever minute.SPL has one of the longest records of aerosol number concentration in North America (Asmi et al., 2013).The TSI 3010 CN concentration records started in 1998 and the TSI 3025 CN concentration record started in 2001.The quality assurance protocol for a Global Atmospheric Watch (GAW) station was adapted for this dataset.All data were collected and then meticulously prescreened by SPL staff prior to generation of the final quality controlled data set.Local contamination sources (e.g., snowmobiles and snow cat passing SPL during the ski season) were removed from the final data set.The historical CN data from SPL were incorporated into the aerosol database archive of National Oceanic and Atmospheric Administration Global Monitoring Division's (NOAA/GMD) Aerosol Group when SPL joined NOAA/GMD's collaborative aerosol network.In sum, TSI 3010 data at SPL is available from 1998-2012 with nearly continuous coverage (Asmi et al., 2013;Fig. 2)

RESULTS
Previous studies (Hallar et al., 2011;Yu and Hallar, 2014) have used aerosol size distribution data to study NPF at SPL.This work has clearly established that NPF is frequent at the site, during the daytime.As demonstrated in the prior publications (Hallar et al., 2011;Yu and Hallar, 2014) and in Fig. 1 below, these events are clearly distinguishable in both the nano-SMPS and standard SMPS data.Unfortunately, the size distribution data is not continuous, and thus cannot provide a seasonal climatology of NPF.
The seasonal average and one standard deviation of aerosol concentration with diameters greater than 3 and 10 nm are shown in Table 1.Each season is defined as follows: Spring is March, April and May; Summer is June, July, and August; Fall is September, October, and November; Winter is December, January, and February both in Table 1 and for the remainder of this paper.There is no statistically significant seasonal trend in aerosol concentration at SPL, either in the CN concentration or the UFCN concentration, although concentrations of UFCN may be higher in the spring.
Table 1 also shows the meteorological climatology at SPL. Monthly averaged data from September 1992-December 2014 were used to create the seasonal averages, shown in Table 1.The site is consistently impacted by wind from the West, across all seasons.Consistent with a Northern Hemisphere high alpine location, SPL experiences a large temperature span between winter and summer.The fall and spring seasons have a similar temperature profile.
Previous work (Hallar et al., 2011) used 474 measurement days from 2001 to 2009 and calculated an average initiation time for NPF events at SPL.The initiation time of the event on each day was defined if the concentration of particles within the first three SMPS size bins (diameter midpoint between 8-10 nm) was above 47 cm -3 or 3000 in units of dN/dlogDp.The average initiation time for NPF events was 12:11 ± 99 min.No nighttime events were observed at SPL (Hallar et al., 2011).As mentioned in introduction, NPF was found 52% of days at SPL during the timeframe of the Hallar et al. (2011) study.
NPF events can also be identified by considering the difference between the 3010 CPC and 3025 CPC during daytime and nighttime hours.Because these instruments operated more continuously than the SMPS's they can also  The method consists of observing the "difference of differences".First the difference between the 3025 CPC and the 3010 CPC was taken to find the concentration of particles between 3 and 10 nm, for daytime and nighttime, as shown in In general the number of NPF events was highest during the spring (March) followed by the fall (September-November), then winter (Dec-Jan-Feb) and finally a strong decrease was found in the summer (May-June-July).There are limitations to this technique; primarily that it is not a direct observation of particle growth (i.e.banana curves are not observed).
Using the available SO 2 dataset, in conjunction with the meteorology data, source regions or geographic areas of this nucleation precursor were considered.To specifically investigate sources of SO 2 , only time periods with SO 2 events were considered.An event was defined when the SO 2 concentration was above 0.5 ppb, and events were found in 11% of the dataset.Fig. 4(a) demonstrates the wind direction and wind speed during the timeframe where SO 2 data was available (March 2013 to August 2014).The wind direction primarily spans from 200° to 315°, with wind speeds typically below 20 m s -1 .Fig. 4(b) demonstrates the wind direction and SO 2 concentration for events.The SO 2 events are primarily found in a narrower wind direction range, between 265° and 305°, suggesting a specific source region for SO 2 at SPL. Hallar et al. (2011;their Fig. 3) previously demonstrated that NPF events at SPL correspond more closely to a westerly wind direction, specifically the wind direction spanning from 255° to 305°.
Table 2 displays the seasonal averaged SO 2 concentration for the entire dataset, events greater than 0.2 ppb, and events greater than 0.5 ppb.The standard deviation and the 95% confidence interval (calculated from a normal distribution) of the mean are also shown.During the summer months, there is a statistically significant lower SO 2 concentration found using the entire dataset.When considering only events where the SO 2 concentration is greater than 0.2 ppb or 0.5 ppb, the highest concentration is consistently found during the spring months, and the lowest concentration is found in the summer months.These seasonal trends in SO 2 concentration demonstrate strong statistical significance.Again, events with SO 2 concentrations above 0.5 ppb are found in 11% of the overall dataset, but they are only found during 5% of the summertime.Thus the seasonal pattern of the NPF is consistent with the seasonality of SO 2 concentration, i.e., the maximum is found in the spring and the minimum is found in the summer.
In order to identify a potential source region, the spring    2, there is a greater probability of finding high concentrations of SO 2 in the spring.Similar to the climatological patterns displayed in Fig. 4, the springtime SO 2 events clearly show a wind direction representing a source region originating in a narrower region, typically from the wind directions between 265° to 295°.In the summertime the wind direction shows more variation during the SO 2 events.There is also a higher probability of having a Westerly (265° to 295°) wind direction in the spring season than the summer.were measured in all quadrants at SPL, the dominant wind direction is found primarily in the Westerly sector, between 220° to 305°.
Subsequently, the wind direction was considered for events during this timeframe when the SO 2 measurement was greater than 0.5 ppb.This criterion significantly narrows the wind direction sampled.Specifically, SO 2 events were only sampled when the wind direction was within a subset of the Westerly sector, primarily between 265° to 305°, as shown in Fig. 6(b).To evaluate the occurrence of NPF during this timeframe using SMPS data, the first 22 channels of the SMPS data were summed to calculate the concentration of particles smaller than 20 nm (N 20 ).Two thresholds were considered representing all NPF (N 20 > 2000 cm -3 ) and strong NPF events (N 20 > 10000 cm -3 ).It is important to note that this methodology may also include pollution events, which impact particle concentrations below 20 nm.From our visual inspection of the SMPS data, pollution events, which look much different than NPF events (i.e., lack of a distinct diurnal pattern), are a rare occurrence.Fig. 6(c) is a wind rose plot showing SMPS concentration when the concentration of N 20 is greater than 2000 cm -3 .Similar to the SO 2 > 0.5 ppb requirement of an SO 2 event, this N 20 >2000 cm -3 criterion also narrows the wind direction sampled.Specifically, NPF events were only observed when the wind direction was within a subset of the Westerly sector, again primarily between 265° to 305°.When the criterion is further strengthened to N 20 > 10000 cm -3 , representing strong NPF events, the observed wind directions narrow even further (Fig. 6(d)).The probability, indicated by radii on Figs.6(c) and 6(d), of finding a strong NPF event (N 20 > 10000 cm -3 ) increased slightly from 265° to 305°, in comparison to all NPF (N 20 > 2000 cm -3 ).

DISCUSSION AND CONCLUSIONS
Northern hemisphere atmospheric observations in general have shown the highest NPF frequency in the spring, similar to the measurements presented here.A high seasonal event frequency in spring in the northern hemisphere is shared by other statistical descriptions of NPF events across the globe (e.g., Birmili and Wiedensohler, 2000;Hõrrak et al., 2000;Mäkelä et al., 2000;Birmili et al., 2003;Dal Maso et al., 2005;Pryor et al., 2010).There are a few exceptions to this seasonal trend found in the literature, including the high elevations studies discussed in the introduction.Additionally, Jaatinen et al. (2009) found a high NPF frequency in the summer at San Pietro Capofiume, in the Po Valley of Italy, and a high NPF in the fall in Melpitz, Germany.Both of these low elevation locations are strongly urban influenced.More similar to the measurements here, Kanawade et al. (2012) and Stanier et al. (2004) found a high frequency of NPF in both the spring and fall, and this was usually associated with elevated SO 2 concentrations.These two studies represented semi-rural continental (Kent, Ohio) and urban (Pittsburgh, Pennsylvania) environments respectively.To our knowledge, this is the first study reporting a high frequency of NPF in both the spring and fall in a remote, high elevation location.
The observed seasonality of NPF at SPL is consistent with chemical composition data observed during short field campaigns at the site.Friedman et al. (2013) reported single particle composition (size range: 0.05-3 µm) measured at SPL in March 2011, and showed that sulfate with a small amount of organics was the dominant particle class.In contrast, Hallar et al. (2013) showed that organic compounds (OC) compose the majority (~64%) of the mass of particles collected with filters at SPL in July 2010.The dominance of SPL particle mass by sulfate in March and by OC in July provides a contrast during different seasons: sulfate-rich spring and organics-rich summer.Previous model simulations of aerosol chemistry at SPL indicate that condensable aerosol precursors were dominated by H 2 SO 4 in March and by low volatile secondary organic gases (LV-SOGs) in July (Yu and Hallar, 2014), and that H 2 SO 4 plays a much more important role in the initial nucleation process.Since NPF formation is strongly dependent on H 2 SO 4 and temperature (Yu, 2010), the high frequency of NPF in both the spring and fall observed at SPL is likely a result of combined effects of seasonal variations in key parameters including SO 2 concentrations.Previous work at SPL also has shown that NPF is associated with increased UV flux and a low surface area of pre-existing particles (Hallar et al., 2011).
The combined results have interesting implications for nucleation processes, i.e., key parameters controlling NPF in the remote atmosphere.The difference between the general wind direction at SPL, and the wind direction with high concentrations of SO 2 and NPF (shown in Figs. 4, 5 and 6) suggest a specific source region.Three coal-fired power plants are located directly to the west of SPL, at distances of approximately 50, 80, and 250 km, proving ample year-round SO 2 .Previous research (e.g., Borys et al., 2000;Hallar et al., 2013;Zhao et al., 2013) has also noted the potential of these power plants to influence aerosol properties and cloud microphysics at SPL.Data from SPL continues to suggest strong implications for the role of H 2 SO 4 , stemming from SO 2 emitted by coal fired power plants, on the aerosol number concentration and size distribution in remote areas.These data can also be used to evaluate various nucleation theories as well as the performance of regional and global aerosol models.

Fig. 1 .
Fig. 1.Particle size distributions are presented as a concentration matrix, with the x-axis representing time and the y-axis representing the particle size and the colors represents the concentration.Concentrations are reported in units of dN/dlogDp.Fig. 1(a) presents data from the standard SMPS at SPL from one day, March 1, 2011.Fig. 1(b) presents nano-SMPS data for the same time period at SPL.

Fig. 2 .
This figure highlights the seasonality of ultrafine aerosol concentration at SPL. Nighttime values show a maximum in the summer season and a minimum in the winter.Next, the nighttime difference (no NPF) was subtracted from the daytime difference (indicating NPF).This separation into daytime and nighttime was performed for the time intervals 8 am to 4 pm and 11 pm to 4 am, respectively, based on Hallar et al. (2011).The resulting values, shown as a monthly average spanning from 2001 to 2007 and 2009-2014 in Fig. 3, clearly displays seasonality in NPF.The maximum of NPF occurs in March, while the three lowest NPF months are May through July.A second smaller peak of NPF is observed in the fall season, with a maximum in October.

Fig. 2 .
Fig. 2. Monthly averages showing the difference between the 3025 CPC and the 3010 CPC to find the concentration of particles between 3 and 10 nm.Nighttime difference is shown in blue, and daytime difference is shown in red.Standard deviations are shown with the vertical bars.

Fig. 3 .
Fig.3.Monthly averages showing the "difference of the difference".First, the difference between the 3025 CPC and the 3010 CPC was taken to find the concentration of particles between 3 and 10 nm, then nighttime difference (no NPF) was subtracted from the daytime difference (indicating NPF).Standard error bars calculated from the daily measurements within each month are shown.

Finally
Fig. 6.A) SPL Wind rose from 4/3/2013 to 7/12/2014, when both SMPS and SO 2 data were available; B) Wind rose showing SO 2 concentration for events with SO 2 greater than 0.5 ppb.Colors demonstrate SO 2 concentration.C) Wind rose showing SMPS concentration for particles smaller than 20 nm (N20) when concentration of N20 is greater than 2000 cm -3 .Colors demonstrate aerosol concentration for particles smaller than 20 nm.D) Wind rose showing N20 concentration only for events when concentration of N20 is greater than 10000 cm -3 .Colors demonstrate aerosol concentration for particles smaller than 20 nm.For all four plots in this figure the radius indicates probability.
, and the TSI 3010 has continued monitoring CN concentration uninterrupted since 2012.TSI 3025 data is available from 2001-2007 and from 2009-2014, and thus this is the timeframe used in this study.Hourly averaged CN and UFCN data are used in this study.Model 43iTLE).The Air Pollution Control Division of Colorado Department of Public Health and the Environment performs an audit of this instrument annually.As suggested by the Quality Assurance Handbook for Air Pollution Measurement Systems, published by the U.S. EPA, 2006, the SO 2 instruments is calibrated daily with zero air and a NIST-traceable primary standard.The SO 2 data record spans the time range from March 28, 2013 to August 11, 2014.Unlike the SMPS, the sulfur dioxide instrument ran nearly continuously during this timeframe.Thus, all seasons are well represented.The spring and summer season each covering approximately 30% of the data.Both the fall and winter season represented approximately 20% of the total dataset.
provide a means of detecting a seasonal pattern in NPF.

Table 2 .
Seasonality of SO 2 concentration for all data and as a function of event constraints.