New particle formation (NPF) is a key atmospheric process which may be
responsible for a major fraction of the total aerosol number burden at the
global scale, including in particular cloud condensation nuclei (CCN). NPF
has been observed in various environments around the world, but some
specific conditions, such as those encountered in volcanic plumes, remain
poorly documented in the literature. Yet, understanding such natural
processes is essential to better define pre-industrial conditions and their
variability in climate model simulations. Here we report observations of NPF
performed at the high-altitude observatory of Maïdo (2165 m a.s.l., La
Réunion Island) between 1 January and 31 December 2015.
During this time period, three effusive eruptions of the Piton de la Fournaise,
located
Aerosol particles are a complex component of the atmospheric system, which affects both air quality and climate. They have been the focus of a growing number of studies during the last decades, but our knowledge of their sources and properties, including their ability to interact with other atmospheric components and associated effects on the Earth's climate system, remains nonetheless uncomplete. Specifically, while particles are known to affect the formation of clouds, and in turn their properties (Albrecht, 1989; Rosenfeld et al., 2014), the radiative forcing associated with these effects (usually referred to as “indirect effects”) is known with a still large uncertainty (Myhre et al., 2013). Better understanding and quantification of this indirect effect requires, in particular, more accurate information on secondary aerosol particle sources, and in turn on new particle formation (NPF). Indeed, measurements conducted in various environments suggest that NPF might be an important source of cloud condensation nuclei (CCN) (e.g. Kerminen et al., 2012; Rose et al., 2017), which is further supported by model investigations (Merikanto et al., 2009; Makkonen et al., 2012; Gordon et al., 2017). However, despite significant improvement of instrumental techniques for the characterization of the newly formed particles and their precursors (Junninen et al., 2010; Vanhanen et al., 2011; Jokinen et al., 2012), model predictions are still affected by our limited understanding of NPF. In addition, the scarcity of observations makes it all the more uncertain in hard-to-reach environments, or in specific conditions, such as those encountered in volcanic plumes.
Volcanic eruptions are one of the most important natural sources of some
specific gases and aerosol particles in the atmosphere. A variety of gaseous
species have been identified in volcanic plumes, including halogens
(Aiuppa et al., 2009; Mather, 2015) and sulfur dioxide (
The occurrence of NPF in volcanic plume conditions was suspected to take
place in several earlier studies (e.g. Deshler et al., 1992; Robock, 2000;
Mauldin III et al., 2003), but the first dedicated study was conducted by Boulon
et al. (2011b), during the eruption of the Eyjafjallajökull which happened in
spring 2010. Indeed, using measurements performed at the high-altitude
station of Puy de Dôme (1465 m a.s.l., France), the authors linked the
occurrence of NPF to unusually high levels of
Despite providing new and highly valuable information, the studies by Boulon
et al. (2011b) and Sahyoun et al. (2019), however, had some limitations.
Indeed, they were both based on short datasets, which did not allow for any
statistical approach to evaluate the relevance of the process nor proper
comparison with the occurrence of NPF outside of plume conditions. Also,
airborne measurements conducted in the vicinity of Etna and Stromboli
allowed Sahyoun et al. (2019) to investigate the presence of the newly
formed particles soon after the emission of their precursors from the vent
of the volcanoes up to few tens of kilometres. They were, however,
unfortunately not able to document properly the evolution of the particle
size distribution along the volcanic plumes, but analysed instead the
particle concentration in relatively broad size ranges (2.5–10, 10–250 nm). In addition, as mentioned earlier, this study was focussed on
passive plumes, i.e. in the presence of a limited concurrent emission of
primary particles by the volcanoes. Conducting a similar investigation of
volcanic eruption plumes is also more difficult due to the unexpected
aspect of active eruptions. Taking advantage of ground-based
measurements and broader instrumental set-up, Boulon et al. (2011b) were in
contrast able to study the time variation of the particle size distribution
between 2 nm and 20
In this context, the objectives of the present work were to provide new
observations of NPF in a volcanic eruption plume with detailed analysis of
the event characteristics, including the capacity of the newly formed
particles to reach CCN sizes, and to assess the relevance of the process
with respect to non-plume conditions. For that purpose, we used measurements
of the particle number size distribution and
Measurements were performed at the Maïdo observatory located on La
Réunion Island in the Indian Ocean (21.080
The instrumental set-up used in the present work was previously described in
the companion study by Foucart et al. (2018). The aerosol size distribution
between 10 and 600 nm was measured with a custom-built differential mobility
particle sizer (DMPS), with a time resolution of 5 min. Particles are first
charged to equilibrium using an Ni-63 bipolar charger, after which they enter
the DMPS, which includes a TSI-type differential mobility analyzer (DMA)
operating in a closed loop and a condensation particle counter model TSI
3010. The instrument was operated behind a whole air inlet (higher size
cut-off of 25
Finally, meteorological parameters recorded with a time resolution of 3 s
were used as ancillary data. Global radiation was measured with a sunshine
pyranometer (SPN1, Delta-T Devices Ltd., resolution 0.6 W m
An overview of the data availability for all above-mentioned instruments between 1 January and 31 December 2015 is provided in Foucart et al. (2018, Fig. 2).
As previously reported in Foucart et al. (2018), the formation rate of 12 nm
particles (
In order to get further insight into the early stages of the NPF process,
and in the absence of direct measurements of sub-3 nm particles,
Four eruptions of the Piton de la Fournaise were observed in 2015: the first
in February, the second in May, the third at the very beginning of August,
and the last from the end of August to late October. More details about
the exact dates and characteristics of the eruptions can be found in Tulet
et al. (2017). Figure 1a presents the time series of the
In total, 30 d were classified as “plume days” following the
above-mentioned criteria, among which 1 was excluded from further analysis
due to DMPS malfunctioning. All these 29 d were previously classified as
plume days by Foucart et al. (2018), who identified in total 44 plume days
with available DMPS measurement. The difference in the classifications
arises from the different time windows investigated in the two studies, i.e.
morning nucleation hours (this study) vs. daytime (Foucart et al., 2018), and
from the criterion on the number of hourly averages of the
In addition to the above-mentioned classification, we further analysed the
characteristics of the plume days in terms of (1) the duration of the plume
conditions detected at Maïdo (from 3 to 5 h between 06:00 and 11:00 LT) and (2) the level of the hourly average
In the absence of direct measurements, we used a proxy to estimate the
concentration of gaseous sulfuric acid. To our knowledge, there is no
specific proxy dedicated to the rather unusual volcanic plume conditions, so
we considered instead the expressions from Petäjä et al. (2009) and
Mikkonen et al. (2011), which have already been widely used in nucleation
studies. The two proxies have the common feature of considering the oxidation
of
As mentioned in Sect. 2.3, 29 plume days were identified at Maïdo as a consequence of the three eruptions of the Piton de la Fournaise which could be documented in 2015. Besides the plume days, 250 d with no influence of the volcanic plume were identified and included in the analysis (days with plume conditions detected in the afternoon or short plume occurrences were excluded, 15 d in total; see Sect. 2.3 for more details); these days will be hereafter referred to as “non-plume days”. Figure 2 shows the monthly NPF frequency separately for plume and non-plume days, with a specific focus on May, August, September and October 2015, when the eruptions were observed. Note that statistics for non-plume days were previously reported for all months in 2015 by Foucart et al. (2018). Our results suggest that volcanic plume conditions favour the occurrence of NPF at Maïdo since all the plume days were classified as NPF event days with the exception of 3 d classified as undefined in September, leading to higher NPF frequencies in plume conditions compared to non-plume days over the months highlighted in Fig. 2 (90 % vs. 71 %). Focussing on the strong plume days, 12 were classified as event days and the remaining 2 d were classified as undefined. At the annual scale, i.e. when including all months in the calculation, the NPF frequency was raised from the already remarkably high value of 67 % when excluding the plume days to 69 % when considering both plume and non-plume days. Such values are among the highest in the literature, similar to those previously reported for the high-altitude station of Chacaltaya (5240 m a.s.l., Bolivia) (64 %, Rose et al., 2015) and the South African savannah (69 %, Vakkari et al., 2011) and slightly lower compared to that reported for the South African plateau, where NPF events are observed on 86 % of the days (Hirsikko et al., 2012).
Frequency of occurrence of NPF at Maïdo. Statistics are shown separately for plume and non-plume days, and total frequencies are also reported. Numbers on the plot indicate, for plume and non-plume conditions, the total number of days included in the statistics.
In addition, a quick analysis was also performed on the 8 d for which the
volcanic plume was detected after the morning hours during which nucleation
is usually initiated (see Sect. 3.1.2 for more details about the timing of
NPF). With the exception of one day classified as undefined in October, all
other days were classified as NPF event days, but there was no clear
evidence of an effect of the “late” plume conditions on the ongoing
events, triggered earlier during the day. High
As a first investigation of the specificities of NPF in plume conditions, we performed a simple analysis of the starting time of the NPF events on plume and non-plume days. The starting time of an event was defined by a visual inspection at the time when the 1.5–2.5 nm ions concentration measured with the AIS significantly increased. Only the events simultaneously detected with the AIS and the DMPS were included in this analysis, and the dataset was not limited to May–August–September–October, but included instead all available AIS data between mid-May and the end of October. In total, 36 events observed on non-plume days and 10 events detected in plume conditions were documented.
The median starting time of NPF in non-plume conditions was found at 08:36 LT (25th percentile: 08:15 LT; 75th percentile: 09:06 LT). An earlier rising time of the cluster concentration was observed on plume days, around 07:41 LT (07:18; 08:16 LT). In addition, we also calculated the median time laps between sunrise and the beginning of NPF, since the starting time of NPF was most likely affected by the change in sunrise time over the course of the investigated period. The median time lap between sunrise and rising time of the cluster concentration was 2 h 11 min (1 h 55 min; 2 h 26 min) on non-plume days, with a minimum of 54 min observed on 18 May, and was about 45 min shorter in plume conditions, being 1 h 26 min (57 min; 1 h 38 min), with a minimum of 29 min obtained on 28 August. These observations suggest that on plume days, when precursors related to volcanic plume conditions were available prior to sunrise in a sufficient amount, photochemistry was the limiting factor for NPF to be triggered. In contrast, on non-plume days, NPF was certainly limited by the availability of condensable species, which were most likely transported from lower altitudes by means of convective processes taking place after sunrise. Further discussion on the precursors involved in the process is reported in Sect. 3.2.2.
Figure 3 shows the formation rate of 2 and 12 nm particles (
Monthly medians and percentiles of the NPF event characteristics
observed on plume and non-plume days at Maïdo.
In contrast, the effect of plume conditions was more pronounced on the
particle formation rates, both for
These observations suggest that, despite a limited effect on particle growth, plume conditions do affect NPF, both in terms of frequency of occurrence and particle formation rate. However, assessing the real effect of these specific conditions on the particle formation and growth is challenging. Specifically, as previously highlighted in the companion study by Foucart et al. (2018), the particle growth rates calculated from high-altitude stations such as Maïdo are “apparent” due to the complex atmospheric dynamics around these sites and may in particular be overestimated due to the concurrent transport of growing particles to the site. The influence of the volcanic plume on larger particles, including CCN-relevant sizes, is further investigated in Sect. 3.3., while the next section is focussed on the analysis of key atmospheric components previously reported to influence NPF, both in terms of frequency of occurrence and characteristics.
NPF has been previously reported to be influenced by various atmospheric parameters, including solar radiation, temperature (Dada et al., 2017), and RH, the effect of which on the process is certainly the less evident to predict and understand (e.g. Birmili et al., 2003; Duplissy et al., 2016). In the frame of the present analysis, the median diurnal variations of the above-mentioned parameters reported in Fig. S1 (in the Supplement) did not highlight any specificity for the events observed on plume days and displayed similar behaviour under in-plume and off-plume conditions.
In addition to the aforementioned meteorological variables, which are
thought to directly influence the production of the nucleating and growing
vapours as well as the survival of the newly formed clusters, other factors
were shown to affect NPF, such as for instance the loss rate of the
condensing compounds on pre-existing particles. The effect of the volcanic
plume on this last parameter is discussed below, while the role of
As recalled in Sect. 2.1, the condensation sink represents the loss rate of precursor vapours on pre-existing larger particles and is thus expected to directly affect the amount of precursors available for NPF. In order to further investigate the effect of this parameter on the occurrence of NPF and avoid any interference with the CS increase caused by the newly formed particles themselves, we focus here on the CS observed prior to usual nucleation hours, between 05:00 and 07:00 LT.
Figure 4a shows the monthly median of the CS calculated over the above-mentioned time period, separately for plume, strong plume and non-plume days, and event and non-event days. Note that strong plume days were included in the statistics reported for plume days and were also highlighted separately. The comparison of non-plume NPF event and non-event days did not highlight any clear tendency over the months of interest for this study. Indeed, comparable median CS was observed in August regardless of the occurrence of NPF later during the day; higher values were in contrast obtained on event days in May, while the opposite was observed in September and October, most likely related to biomass burning activity in South Africa and Madagascar during austral spring (Clain et al., 2009; Duflot et al., 2010; Vigouroux et al., 2012). The overall number of non-event days included in the statistics was, however, limited, and using a comparable time window Foucart et al. (2018) reported that CS was on average higher on NPF event days compared to non-event days when including all the data from 2015. These contrasting results are representative of the observations from high-altitude observatories at a larger scale, where the location of the sites, their topography and the fast changing conditions related to complex atmospheric dynamics are likely to influence the effect of the CS on the occurrence of NPF. Indeed, Boulon et al. (2010) and Rose et al. (2015) reported that CS was on average positively correlated with the occurrence of NPF at Jungfraujoch (3580 m a.s.l., Switzerland) and Chacaltaya, respectively. These observations suggest that the availability of the precursors was often limiting the process at these sites, which seemed to be fed with vapours transported together with pre-existing particles contributing to the CS. In contrast, the CS was observed to be on average lower on NPF event days compared to non-event days at Puy de Dôme, which could be in a less precursor-limited environment due to its lower altitude (Boulon et al., 2011a).
Despite the important variability of the reported values, the median CSs
obtained on plume days were on average higher than those observed on
non-plume days, with the largest difference in May. The CSs reported for
strong plume days were even higher, with median values on average increased
by 1 order of magnitude compared to non-plume event days (up to 30 times
higher in May). One might have expected those enhanced CSs to inhibit NPF at
Maïdo, which was instead more frequent in plume conditions compared to
non-plume days (Sect. 3.1.1). This non-intuitive result is most likely
explained by the increased mixing ratios of
The origin of the particles responsible for increased CS prior to nucleation
hours on plume event days remains uncertain, but the high
As previously mentioned in Sect. 2.4, sulfuric acid concentrations were
obtained from
Figure 5a shows all
The values calculated for
In order to get further insight into the nucleation mechanism likely to
explain the observed events, we additionally compared the formation rates
derived from DMPS measurements with that predicted by the recent
parameterization developed by Määttänen et al. (2018), which
describes neutral and ion-induced binary nucleation of
A deeper analysis of the ability of the model to represent the observed
events was then performed, and for that purpose the model by
Määttänen et al. (2018) was run for each NPF event, with the
corresponding temperature, relative humidity and CS levels. Figure 5b
shows, for the same dataset as in Fig. 5a, the ratio between
Altogether, these results suggest that higher
The previous section was dedicated to the analysis of NPF occurrence and characteristics in volcanic plume conditions. In this section, we further investigate the effect of such a process on the shape of the particle size distribution, first including all sizes between 10 and 600 nm, and then focussing more specifically on particles large enough to act as CCN. Since the results we have reported so far only revealed a limited signature of strong plume conditions on NPF characteristics, we will no longer put any specific focus on these particular days in this last section. They will, however, still be included in the statistics reported for plume days.
The effect of NPF and/or plume conditions on the particle number size distribution was investigated based on the hourly median particle spectra measured with the DMPS in different conditions between 07:00 and 16:00 LT (Fig. 6). This time period was selected as, besides usual NPF hours, it also includes one hour prior to nucleation hours, which allowed the study of the main features of the particle size distribution in the different conditions (plume and non-plume) without the fresh influence of NPF, as well as several hours to investigate the change of the spectra caused by particle growth processes. Note that in order to increase the statistical relevance of the results (especially for non-event days), the analysis was not restricted to May–August–September–October, and all available data were included in the analysis. All median spectra were in addition fitted with four Gaussian modes, including nucleation, Aitken and two accumulation modes, the parameters of which are shown in Fig. 7 and also reported in Table A1 of Appendix A.
Hourly medians of the particle size distribution derived from DMPS measurements conducted in the different conditions (non-plume NPF event and non-event days, plume NPF event days) between 07:00 and 16:00 LT.
Variations of the parameters of the Gaussian modes used to fit the hourly median DMPS size distributions shown in Fig. 6. All the displayed values are also reported in Table A1 of Appendix A.
As previously mentioned, the median spectra recorded at 07:00 LT, i.e. prior
to nucleation hours, gave a unique opportunity to compare the main features
of the particle number size distributions recorded in plume and non-plume
conditions in the absence of freshly nucleated particles. The spectra measured
on non-plume days (both event and non-event days) displayed comparable
shapes as well as similar concentrations, while higher concentrations were
in contrast measured in plume conditions. As discussed in Sect. 3.2.1,
these differences, which were the most pronounced for the two accumulation
modes, were most likely explained by the presence of particles originating
from heterogeneous processes occurring at high temperatures at the vent
during the eruptive periods, and assimilated to volcanic primary particles.
Indeed, in plume conditions the population of the first accumulation mode
(modal diameter
Despite some variations of the particle concentration in the different
modes, the shape of the spectrum observed at 07:00 LT remained the same
throughout the investigated time window on non-event days. The
concentrations of the Aitken and first accumulation modes were increased by
a factor of
Consistent with previous observations reported in Sect. 3.1.2, the beginning of NPF was seen at 08:00 LT on event days (regardless of the occurrence of
plume conditions) from a visual analysis of the spectra and was further
confirmed by the increase in the particle concentration in the nucleation
mode, which lasted until 11:00 LT. The most significant change was observed
in plume conditions, with a 1000-fold increase in the nucleation mode
particle concentration in 4 h, from 5 to 5700 cm
Still focussing on NPF event days, the changes observed in the parameters of
the Aitken and two accumulation modes were less pronounced than for the
nucleation mode. With the exception of the slight difference observed at
07:00 LT, the Aitken mode displayed similar diameters on plume and non-plume
days, with only limited variations over the investigated time window,
especially on non-plume days (32–45 and 38–46.5 nm, on plume and
non-plume days, respectively). The initial concentration of the Aitken mode
measured in plume conditions was slightly higher than on non-plume days (270
vs. 180 cm
Altogether, one can infer from these measurements a distribution of the particles of volcanic origin, including the contributions of both primary and secondary aerosols. Following the above analysis, the concentration of the so-called volcanic primary particles was calculated for each mode as the difference between the concentrations measured at 07:00 LT on plume and non-plume days, and the values obtained at 07:00 LT in plume conditions were used for the other characteristics of the modes (i.e. sigma and modal diameter). In addition, the concentration of secondary aerosol particles of volcanic origin was calculated for each mode as the difference between the maximum concentrations observed on plume event days and that observed on non-plume event days, which were found between 11:00 and 14:00 LT depending on the modes and conditions; the values obtained at 12:00 LT in plume conditions, when the effect of secondary aerosol formation on the spectrum was on average the most pronounced, were used for the other characteristics of the modes. The resulting aerosol spectrum is reported in Fig. 8, with the detailed contributions of volcanic primary and secondary particles for each mode. Secondary aerosol particles formed due to the presence of the plume contributed 93 % of the total concentration observed on plume event days, clearly dominating all the modes but the first accumulation mode, for which the contribution of volcanic primary particles was more significant. The presence of a secondary contribution to the accumulation modes is likely the result of the growth of particles from the Aitken mode, due to the presence of more condensable gases.
Size distribution of the particles of volcanic origin reconstructed using the spectra measured on plume and non-plume event days. The contributions of primary (i.e. formed or transformed via heterogeneous processes in the very close proximity of the vent) and secondary aerosols are shown separately for each mode on the spectrum and are further highlighted on the pie charts. The contribution of primary aerosols was evaluated based on the spectra measured at 07:00 LT under in-plume and off-plume conditions, while the contribution of secondary aerosols was deduced from the maximum concentrations measured for each mode under in-plume and off-plume conditions, between 11:00 and 14:00 LT.
The increase in potential CCN concentration during NPF was investigated
using DMPS measurements, in a similar way to how it was done earlier by Rose et al. (2017) for the high-altitude station of Chacaltaya, following the approach
originally developed by Lihavainen et al. (2003). It is based on the
hypothesis that the lower cloud droplet activation diameter
This last analysis was not restricted to the months when the volcanic
activity was detected, and 193 NPF event days identified in 2015 were
included in the analysis (167 non-plume days and 26 plume days). As reported
earlier by Foucart et al. (2018), the growth of particles
The increase in
Increase in potential
In a similar way to how it was done previously by Rose et al. (2017), we made an
attempt to decouple the contributions of the above-mentioned CCN sources on
event days. For that purpose, the transport of pre-existing large particles
from the boundary layer was first assumed to have similar magnitude on event
and non-event days, regardless of the occurrence of plume conditions. We also
made the assumption that the contribution of volcanic primary particles did
not vary significantly throughout the day in plume conditions and was thus
systematically removed when calculating the difference between
We investigated the occurrence of NPF in volcanic plume conditions at the
Maïdo observatory based on measurements conducted between 1 January and 31 December 2015. During this time period, four effusive
eruptions of the Piton de la Fournaise, located
Focussing on the months during which the volcanic plume was detected at
Maïdo (May, August, September and October), NPF was observed on 90 %
of the plume days vs. 71 % of the non-plume days. On plume days, when
higher amounts of precursors (such as
The signature of the volcanic plume on the aerosol spectra up to 600 nm was
further investigated based on the analysis and fitting of the particle size
distributions recorded in the different conditions. The spectra measured
prior to nucleation hours (07:00 LT) gave a unique opportunity to compare
the main features of the particle number size distributions recorded in
plume and non-plume conditions in the absence of freshly nucleated particles.
The main differences were observed for the two accumulation modes, which were
more densely populated in plume conditions compared to non-plume days, most
likely because of the contribution of particles formed via heterogeneous
processes at the vent of the volcano during eruptive periods, and
assimilated to primary volcanic particles in the present work. The particle
size distribution only experienced limited changes on non-event days, but
significant variations of the particle concentration were in contrast
observed for the nucleation and Aitken modes on NPF event days between 08:00
and
Specific attention was further paid to the concentration of particles
In order to investigate deeper the influence of volcanic plume conditions on
the occurrence of NPF and related effects on particle concentration, we
first investigated the variations of several atmospheric parameters
(temperature, relative humidity and global radiation). Similar patterns were
observed under in-plume and off-plume conditions, without any specificity for the events
observed on plume days. Attention was then paid to the variations of the
condensation sink (CS) and [
Altogether, our observations show that, based on 1 year of data,
volcanic plume conditions favour the formation of particles that frequently
grow to CCN sizes within the first 40 km of the volcano's vent. The
quantification of the contribution of primary vs. secondary aerosol formation
within a volcanic eruption plume on a statistical basis contributes to
better understanding of this natural process, which might have contributed
significantly to NPF and CCN formation in the pristine pre-industrial era.
Nonetheless, our study should be complemented in the future with a direct
analysis of (1) the cluster formation rates, both charged and neutral, and (2) the precursor vapours involved in their formation. Indeed, our approach to
assess the role of
DMPS data are accessible from the EBAS website (
Parameters of the Gaussians used to fit the hourly median DMPS size distributions shown in Fig. 6.
The supplement related to this article is available online at:
KS and PT organized the measurements. JMM conducted the measurements. DP contributed to the design of the instrumental set-up. CR, BF, KS and AC analysed the data. CR and KS wrote the paper, and all authors commented on the paper.
The authors declare that they have no conflict of interest.
We would like to thank
ATMO Réunion for providing
This research has been supported by the European Commission, H2020 Research Infrastructures (ACTRIS-2 (grant no. 654109)), and by the ANR (grant no. ANR-14-CE03-0004-04). This work has received financial support from the French programme SNO-CLAP; it has also been funded by the OMNCG/OSUR program from La Réunion University.
This paper was edited by Veli-Matti Kerminen and reviewed by two anonymous referees.