ACPAtmospheric Chemistry and PhysicsACPAtmos. Chem. Phys.1680-7324Copernicus PublicationsGöttingen, Germany10.5194/acp-17-11227-2017Comparative measurements of ambient atmospheric concentrations of
ice nucleating particles using multiple immersion freezing methods and a
continuous flow diffusion chamberDeMottPaul J.paul.demott@Colostate.eduhttps://orcid.org/0000-0002-3719-1889HillThomas C. J.https://orcid.org/0000-0002-5293-3959PettersMarkus D.https://orcid.org/0000-0002-4082-1693BertramAllan K.https://orcid.org/0000-0002-5621-2323ToboYutakahttps://orcid.org/0000-0003-0951-3315MasonRyan H.SuskiKaitlyn J.McCluskeyChristina S.LevinEzra J. T.SchillGregory P.https://orcid.org/0000-0002-4084-0317BooseYvonnehttps://orcid.org/0000-0001-9495-2165RaukerAnne MarieMillerAnna J.ZaragozaJakeRocciKatherineRothfussNicholas E.https://orcid.org/0000-0002-1495-1902TaylorHans P.HaderJohn D.ChouCedricHuffmanJ. Alexhttps://orcid.org/0000-0002-5363-9516PöschlUlrichhttps://orcid.org/0000-0003-1412-3557PrenniAnthony J.KreidenweisSonia M.https://orcid.org/0000-0002-2561-2914Department of Atmospheric Science, Colorado State University, Fort
Collins, CO 80523, USADepartment of Marine, Earth and Atmospheric Sciences, North Carolina
State University, Raleigh, NC 27695, USADepartment of Chemistry, University of British Columbia, Vancouver,
BC, V6T1Z1, CanadaNational Institute of Polar Research, Tachikawa, Tokyo 190-8518, JapanDepartment of Polar Science, School of Multidisciplinary Sciences,
SOKENDAI (The Graduate School for Advanced Studies), Tachikawa, Tokyo
190-8518, JapanKarlsruhe Institute of Technology, Institute of Meteorology and
Climate Research (IMK-IFU), 82467 Garmisch-Partenkirchen, GermanyDepartment of Chemistry, Reed College, Portland, OR 97202, USADepartment of Earth Sciences, University of New Hampshire, Durham,
NH 03824, USADepartment of Chemistry & Biochemistry, University of Denver,
Denver, CO 80210, USADepartment of Multiphase Chemistry, Max Planck Institute for
Chemistry, 55128 Mainz, GermanyNational Park Service, Air Resources Division, Lakewood, CO 80228,
USAnow at: Pacific Northwest National Laboratory, Richland, WA 99352, USAnow at: Air Resource Specialists, Fort Collins, CO 80525, USAPaul J. DeMott (paul.demott@Colostate.edu)22September2017171811227112453May20178May201715August201723August2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://acp.copernicus.org/articles/17/11227/2017/acp-17-11227-2017.htmlThe full text article is available as a PDF file from https://acp.copernicus.org/articles/17/11227/2017/acp-17-11227-2017.pdf
A number of new measurement methods for ice nucleating particles (INPs) have
been introduced in recent years, and it is important to address how these
methods compare. Laboratory comparisons of instruments sampling major INP
types are common, but few comparisons have occurred for ambient aerosol
measurements exploring the utility, consistency and complementarity of
different methods to cover the large dynamic range of INP concentrations that
exists in the atmosphere. In this study, we assess the comparability of four
offline immersion freezing measurement methods (Colorado State University ice
spectrometer, IS; North Carolina State University cold stage, CS; National
Institute for Polar Research Cryogenic Refrigerator Applied to Freezing Test,
CRAFT; University of British Columbia micro-orifice uniform deposit
impactor–droplet freezing technique, MOUDI-DFT)
and an online method (continuous flow diffusion chamber, CFDC) used in a manner deemed to promote/maximize
immersion freezing, for the detection of INPs in ambient aerosols at different
locations and in different sampling scenarios. We also investigated the
comparability of different aerosol collection methods used with offline
immersion freezing instruments. Excellent agreement between all methods could
be obtained for several cases of co-sampling with perfect temporal overlap.
Even for sampling periods that were not fully equivalent, the deviations
between atmospheric INP number concentrations measured with different methods
were mostly less than 1 order of magnitude. In some cases, however, the
deviations were larger and not explicable without sampling and measurement
artifacts. Overall, the immersion freezing methods seem to effectively
capture INPs that activate as single particles in the modestly supercooled
temperature regime (>-20∘C), although more comparisons are
needed in this temperature regime that is difficult to access with online
methods. Relative to the CFDC method, three immersion freezing methods that
disperse particles into a bulk liquid (IS, CS, CRAFT) exhibit a positive bias
in measured INP number concentrations below -20∘C, increasing
with decreasing temperature. This bias was present but much less pronounced
for a method that condenses separate water droplets onto limited numbers of
particles prior to cooling and freezing (MOUDI-DFT). Potential reasons for
the observed differences are discussed, and further investigations proposed
to elucidate the role of all factors involved.
Introduction
Heterogeneous ice nucleation by atmospheric aerosols impacts the
microphysical composition, radiative properties and precipitation processes
in clouds colder than 0 ∘C. These interactions are complex, and any
first assessment of the role of different particles on ice formation, cloud
properties and climate requires more observations of ice nucleating particles
(INPs, as defined by Vali et al., 2015) present in ambient air. To quantify
the initial stage of ice nucleation in the atmosphere, multiple sampling
techniques are now being used in field studies (Hader et al., 2014; Mason et
al., 2015; DeMott et al., 2015; Stopelli et al, 2015; Boose et al., 2016;
Schrod et al., 2016, 2017). Since these various measurements are being used
as bases for developing numerical model parameterizations for different
emission sources, their comparability should be assessed. In this study, we
focus on ice nucleation measurements in the mixed-phase cloud temperature
regime (0 to -38∘C), where heterogeneous ice nucleation is the
only trigger for primary ice initiation. Within this regime, INP number
concentration can increase up to 10 orders of magnitude as temperatures cool
from -5 to -35∘C (DeMott et al., 2015, 2016; Hiranuma et al.,
2015; Murray et al., 2012; Petters and Wright, 2015), and there can be up to
2–3 orders of magnitude of temporal and spatial variability at a single
temperature by any given method (DeMott et al., 2010; Petters and Wright,
2015).
This study compares results from an online INP measurement method used over
the last 25 years, the Colorado State University (CSU) continuous flow
diffusion chamber (CFDC), with four offline immersion freezing methods for
INP measurements. These four variants immerse particles into variously sized
liquid volumes/droplets which are cooled to freezing in different ways in
order to measure the immersion freezing INP number per volume of air. In this
study, comparisons are made only for times when the CFDC instrument operated
in a manner which emphasized immersion freezing contributions to ice
nucleation (DeMott et al., 2015). A principal reason to evaluate consistency
between approaches, and in ambient air, is because offline methods collect
large enough sample volumes to estimate INP number concentrations active at
modest supercooling (as warm as -5∘C), a temperature regime where
online instruments are unable to obtain statistically significant data
samples. In contrast, online methods can provide high-time-resolution data at
lower temperatures. Comparability between off- and online methods can be
assessed in temperature regions of overlap. Another reason for such a
comparison is to gauge the magnitude of uncertainties when only a single INP
measurement method is used or when data sets from different instruments are
combined toward addressing a scientific question. This study differs from
previous efforts in that comparisons have been restricted to the ambient
atmosphere, where presumably the compositions of INPs are more diverse and
likely different than for single INP types often examined in laboratory
studies. In one set of laboratory studies (Hiranuma et al., 2015),
discrepancies between online and offline methods were noted for sampling
NX-illite INPs. In particular, bulk, offline freezing methods estimated INP
ice nucleation efficiencies that were 10–1000 times lower than found with
continuous flow chambers and the AIDA (Aerosol Interaction and Dynamics in
the Atmosphere) expansion cloud chamber for temperatures warmer than about
-25∘C. Similar discrepancies were discussed by Emersic et
al. (2016). Impacts of dry dispersion vs. wet immersion on the agglomeration
properties and the exposures of active sites were implicated in varied ways
in both studies for explaining discrepancies. Grawe et al. (2016) also noted
discrepancies occurring in single-particle activation via immersion freezing
in the LACIS (Leipzig Aerosol Cloud Interaction Simulator) instrument for
certain, but not all, combustion ash particles. In contrast, no discrepancies
were reported in processing wet-dispersed ice nucleating bacteria from
Snomax® (Wex et al., 2014). Nevertheless,
many of the reported laboratory results have thus far focused on a specific
INP type that was shared across laboratories and for which individual
investigators were allowed to determine protocols for generation as an
aerosol or production of liquid suspensions for the different methods used.
Here, by contrast, we focus on co-located sampling of ambient aerosol, for
which no more than two methods have hitherto been used in a single published
study using this approach.
The goal of this intercomparison is to assess the status and potential for
using single or combinations of INP measurement methods to access and
measure the dynamic range of atmospheric INP concentrations active for ice
initiation in mixed-phase clouds. The assessment assumes that the
time dependence is subordinate to the temperature dependence of the freezing
nucleation process. The scientific basis of this assumption and its
implications for the assessment are discussed. We address the magnitude of
agreement; how particle collection methods may influence immersion freezing
measurements; and whether obvious biases appear, for example due to the
different size ranges of particles that may be collected in offline and
online measurement systems. This study is intended not as a comprehensive
evaluation but rather as a first assessment using some of the most common
methods likely to be applied for atmospheric sampling in the coming years.
Methods
Several INP measurement methods, most with a legacy of previous atmospheric
measurements, are herein intercompared during sampling of ambient aerosols.
This section describes the instruments, details of sampling protocol and
processing, and sampling sites.
INP measurement systemsColorado State University CFDCs
Online INP measurements were made with two CSU CFDCs, designed for mobile and
aircraft deployments but otherwise identical (Eidhammer et al., 2010; DeMott
et al., 2015). As described in these previous publications, aerosol flows
vertically downward in a central lamina between concentric, cylindrical walls
that are ice coated and thermally controlled at different temperatures.
Setting a temperature difference between the colder (inner) and warmer
(outer) ice walls in the upper “growth” region establishes a nearly
steady-state relative humidity (RH) where ice nucleation and ice crystal growth
can occur over a few seconds. The temperatures of the inner and outer walls
are set to the same value in the lower “evaporation” region of the chamber,
which promotes evaporation of water droplets and wet aerosols but retains
activated ice particles at larger sizes that can be detected as
optically distinct for counting as INPs with an optical particle counter. For
this study, the aerosol lamina was 15 % of the total volumetric flow of
10 L min-1. Filtered and dried air was recirculated as sheath flow
(8.5 L min-1). Also for this study, a nominal water-supersaturated
condition of 105 % RH was chosen for operation at all temperatures. This
selection was made to force activation of cloud droplets on aerosols at
temperatures where some proportion could freeze during the transit time in
the instruments, allowing for the most direct comparison possible to the
offline immersion freezing methods. Previous studies have explored the need
to set the RH in CFDC style instruments to values far above those expected in
natural clouds (100–101 % RH) in order to mimic this freezing process
(Petters et al., 2009; DeMott et al., 2010, 2015). Although DeMott et
al. (2015) showed in laboratory studies that operational RH up to 109 %
might be required for full expression of freezing in the CFDC; 105 % is
the value that has been consistently used in field studies so that liquid
droplets do not survive through the evaporation region and are not counted as
false-positive INPs. For mineral dusts, at least, operation at 105 %
could miss by up to a factor of 3 (DeMott et al., 2015) INP number
concentrations that ultimately activate via immersion freezing or some
combination of nucleation mechanisms. It is unknown if this factor exists for
all INP types. Hence, no correction factor was applied to the CFDC data here,
but the implications of the factor of 3 will be discussed. CFDC measurement
uncertainties vary with processing conditions, and they are typically
±0.5 ∘C and 2.4 % water relative humidity at
-30∘C (DeMott et al., 2015).
Aerosol particles at sizes that might confound optical detection of (i.e., be
mistakenly counted as) ice crystals were removed upstream of the CFDC using
dual single-jet impactors set to a cut-point aerodynamic diameter of
2.4 µm. This creates a sampling bias for the CFDC vs. other systems
that capture larger particles for immersion freezing experiments but is
required to assure detection of activated ice crystals that typically exit
the CFDC at optical diameters approximately > 4 µm.
Interval periods of sampling filtered air within the overall sampling period
were used to correct for any background frost influences on INP counts. We
follow Schill et al. (2016) for correcting sample concentrations for
background and for defining confidence intervals for CFDC data, which are
represented by error bars in presented plots. Specifically, corrected INP
concentrations are the sample concentrations with the interpolated background
concentrations subtracted. The standard deviation of INP concentrations derived from the Poisson
counting error during both the sample and the interpolated background
period was added in quadrature to obtain the INP concentration
error. Concentrations are considered significant if they are 1.64 times
larger than the INP concentration error, which corresponds to the
Z statistic at 95 % confidence for a one-tailed distribution.
Consequently, although the lowest limit of detection for 10 min sampling
intervals is ∼ 0.2 L-1, significant data often require in excess
of 1 L-1 INP concentrations. As a special sampling aide in these
studies, an aerosol concentrator (Model 4240, MSP Corporation) was used
upstream of the CFDC in some cases to enhance INP number concentrations and
facilitate statistically significant quantification of INP number
concentrations. The enhancement of aerosol concentrations using this dual
virtual impactor method affects only particles of diameter
> 0.5 µm and varies from a factor of 10 at this diameter up to a
factor of about 140 at sizes above 1 µm (Tobo et al., 2013). The
concentration factor achieved for ambient INPs then depends on the INP size
distribution, which is difficult to know a priori. The methods outlined in
Tobo et al. (2013) were followed to define the concentration factor, using
the ratio of CFDC INP number concentrations with and without the concentrator
under conditions where statistical significance of measurement was achieved
without the concentrator. This was assessed over the term of measurements for
each site in the study and applied to all CFDC data when using the aerosol
concentrator. An example of measurements on and off of the concentrator for
one of the sampling periods used in this study is shown in the Supplement,
Fig. S1. Use of the aerosol concentrator is indicated in individual cases in
the data tables, also included in the Supplement.
Particle losses in upstream tubing, the aerosol impactor and the inlet
manifold of the CFDC have previously been estimated as 10 % for particles
with diameter 0.1 to 0.8 µm (Prenni et al., 2009), and we apply
this correction to data for this paper.
North Carolina State University cold stage (CS)
The design of the North Carolina State University (NC State) CS-supported
droplet freezing assay and data reduction methods are
described in detail in Wright and Petters (2013) and Hader et al. (2014).
Droplet populations of three distinct droplet size ranges may be investigated
in the CS; these are termed pico-, nano-, and microdrops. Picodrops are
generated by mixing a 15 µL aliquot of bulk suspension (particles
placed into liquid by methods outlined below) with squalene and emulsifying
the hydrocarbon–water mixture using a vortex mixer. The emulsion is poured
into the CS sample tray, consisting of an aluminum dish holding a hydrophobic
glass slide. Approximately 400–800 droplets with a typical diameter of
∼ 85 µm are analyzed in this manner for each collected
sample. Nanodrops are generated by manually placing drops with a syringe
needle tip on a squalene covered glass slide and letting the drops settle to
the squalene–glass interface. Approximately 80 droplets are typically
analyzed per experiment with a typical diameter of ∼ 660 µm.
Microdrops are placed directly on the hydrophobic glass slide using an
electronic micropipette. In contrast to the pico- and nanodrops, these drops
are exposed to a dry N2 gas phase. Up to 256 drops of
diameter ∼ 1240 µm (1 µL) can be investigated in a
single experiment. For all experiments, the CS was cooled at a constant rate
of 1 ∘C min-1 (2 ∘C min-1 at Bodega Marine
Laboratory), and the number of unfrozen drops was recorded using a microscope
in increments of dT= 0.17 ∘C resolution. Temperature uncertainty
is based on the manufacturer's (Model TR141-170, Oven Industries) stated
tolerance of the cold-plate thermistor (±1 ∘C). To account for
slightly higher temperatures of the squalene relative to the glass slide, a
temperature calibration was applied to the drop-freezing data (Hader et al.,
2014). The resulting data were inverted to find the cumulative concentration
of INPs (CINPs(T)) per volume of liquid at temperature, T, using
the method of Vali (1971):
CINPs(T)=-1VlnNu(T)N,
where Nu is the unfrozen number of an initial N of liquid entities
(droplets in this case) of volume V. Conversion to number concentration of
INPs per volume of air (nINPs(T)) is determined by
nINPs(T)=CINPs(T)VwVs,
where Vw is the volume of liquid suspension (same units as used to
compute CINPs(T)) and Vs is the sample volume (L) of air
collected.
To minimize sample heterogeneity, only droplets with 78 µm <Dp<102µm were included in the calculation of nINPs(T) for picodrops. No restriction was applied to the nanodrops or
microdrops. Furthermore, the warmest 2 % percent of data were removed after
the calculation of CINPs(T) but before plotting for the pico- and
nanodrops due to large uncertainty stemming from poor counting statistics
(Hader et al., 2014). The INP content of the ultrapure water (see Sect. 2.2)
was measured in the above manner between -20 and -35∘C. The
effective INP content was determined by subtracting the background INP
numbers from the ultrapure water from observed nINPs(T). No
impurities were detected at T>-20∘C. Analysis of CS repeat trial
data involved binning data into 1 ∘C intervals. Confidence intervals
were calculated using 2 standard deviations of the geometric mean for each
bin where multiple data points were available.
University of British Columbia MOUDI-DFT
The second immersion freezing method involved freezing of droplets grown on
substrate-collected particles in a temperature- and humidity-controlled flow
cell (Mason et al., 2015) and is referred to as the droplet freezing
technique (DFT). A micro-orifice uniform deposit impactor (MOUDI; MSP Corp.)
was used to size-select particles from known volumes of air onto a substrate
for direct DFT analysis in a number of cases (MOUDI-DFT, Mason et al., 2015).
The MOUDI collected size-selected particles onto multiple hydrophobic glass
cover slips (HR3-215; Hampton Research). For the measurements performed in
Kansas, United States, stages 2–9 of the MOUDI were used corresponding to
particle size bins of 10–5.6, 5.6–3.2, 3.2–1.8, 1.8–1.0, 1.0–0.56,
0.56–0.32, 0.32–0.18 and 0.18–0.10 µm (50 % cutoff
aerodynamic diameter; Marple et al., 1991), respectively. For the
measurements at CSU, stages 2–8 were used; for the measurements at
Manitou (Colorado) Experimental Forest Observatory (MEFO), stages 2–7 were
used.
For DFT analysis, droplets were grown in the flow cell by decreasing
temperature to 0 ∘C and passing a humidified flow of He gas over the
slides. Water was allowed to condense until approximately 100 µm
diameter water droplets formed on the collected particles, typically covering
several to some tens of particles, depending on loading. Droplets were then
monitored for freezing using a coupled optical microscope (Axiolab; Zeiss,
Oberkochen, Germany) with a 5× magnification objective, as temperature
was lowered at a constant rate. A charge-coupled device connected to the optical
microscope recorded a digital video, while a resistance temperature detector
recorded the temperature. A cooling rate of 10 ∘C min-1 (from
0 to -40∘C) was used in these studies to minimize freezing
of droplets due to contact of a growing crystal and to minimize evaporation of
unfrozen droplets due to the Bergeron–Findeisen process, i.e., growth of the
existing ice crystals at the expense of the surrounding liquid droplets
(Mason et al., 2015). The liquid droplet may evaporate, or the frozen droplet
will grow towards and eventually contact a liquid droplet, causing it to
freeze. If a droplet is lost to evaporation or to non-immersion freezing, two
assumptions are made:
That the droplet contained an INP and would have frozen by immersion (on
its own) at the same temperature as the non-immersion/evaporation event. This
gives an upper limit to the calculated INP concentration
That the droplet contained no INPs and would not have frozen until homogeneous
temperatures, which are around -36∘C in the flow cell used. This
assumption provides a lower limit to the calculated INP concentration at a
given T.
The method to obtain the INP number concentrations in air follows a similar
basis as for the CS but with modest differences:
nINPs(T)=-lnNu(T)NNAdepositADFTVsfnufne,
where N is the total number of droplets condensed onto the sample in this
case, Adeposit is the total area of the sample deposit on the
hydrophobic glass cover slip, ADFT is the area of the sample monitored in the
digital video during the droplet freezing experiment and Vs is the volume of
air sampled by the MOUDI. fne is a correction factor to account for
the statistical uncertainty that results when only a limited number of
nucleation events are observed. fne was calculated following
the approach given in Koop et al. (1997) using a 95 % confidence
interval. fnu is a correction factor to account for non-uniformity
in particle concentration across each MOUDI sample (Mason et al., 2015,
2016). This later correction factor consists of two multiplicative terms:
fnu, 1 mm and fnu, 0.25–0.10 mm, with these terms
correcting for non-uniformity in the particle deposits at the 1 mm and
0.25–0.1 mm scale, respectively. Since only a small area (1.2 mm2) of
the particle deposits are analyzed and the concentration of particles are not
uniform across the entire substrate, fnu, 1 mm needs to be applied.
Since the concentration of particles are not uniform within the small area of
the particle deposits analyzed for freezing, fnu, 0.25–0.10 mm
needs to be applied. Listed in Tables S3 and S4 are the fnu, 1 mm
and fnu, 0.25–0.10 mm values applied to the MOUDI-DFT samples
collected at CSU and Kansas, respectively. Different correction factors were
used for the CSU and Kansas samples since different substrate holders were
used to position the glass slides within the MOUDI at the two sites.
Substrate holders were not yet employed during the earlier MEFO studies
(Huffman et al., 2013). However, by using saved slides from the MEFO
experiments, estimates could be made of the slide offset positions that are needed for
defining the non-uniformity correction at the 1 mm scale in Mason et
al. (2015). Listed in Table S5 are the fnu, 1 mm correction factors
applied to the MEFO samples based on the slide offset positions. Data were
not taken on the non-uniformity within the field of view during the freezing
experiments (fnu, 0.10–0.25 mm) for the MEFO collections, and
hence no correction was applied to the MEFO samples for non-uniformity at
the 0.25–0.1 mm scale. On the basis of Mason et al. (2015; cf. Fig. 8 of
that paper) and calculations using the factors found for CSU and Kansas
sampling, the inability to quantify fnu, 0.10–0.25 mm will lead to
an underprediction of nINPs(T) by a factor that depends on the
frozen fraction of droplets at any temperature, perhaps as high as 1.7 for
the first drops freezing (1 of ∼ 50–100, or 1–2 % frozen
fraction) but less than 1.1 once 25 % of droplets have frozen.
Confidence intervals (95 %) were calculated based on the Poisson
distribution, following Koop et al. (1997). These intervals are nearly
equivalent to Binomial confidence intervals for the data in this study.
Colorado State University IS
The CSU ice spectrometer (IS) (Hill et al., 2014, 2016; Hiranuma et al.,
2015) measures freezing in an array of liquid aliquots held in a
temperature-controlled block. For IS processing, aerosol particles in
suspensions are distributed into 24 to 48 aliquots of 40–80 µL
held in sterile 96-well PCR trays (µCycler, Life Science Products).
The numbers of wells frozen are counted at 0.5 or 1 ∘C intervals
during cooling at a rate of 0.33 ∘C min-1. Temperature was
measured with 0.1 ∘C resolution and 0.4 ∘C accuracy (Hill
et al., 2016). Calculation of nINPs(T) was made using Eqs. (1)
and (2), where V was the aliquot volume. Control wells of ultrapure water
(see Sect. 2.2) were also cooled, and correction for any frozen aliquots in
the pure water control vs. temperature was made in all cases, similar to the
CS method. Binomial sampling confidence intervals (95 %) were determined
for IS data, as described in Hill et al. (2016).
National Institute of Polar Research CRAFT
The Cryogenic Refrigerator Applied to Freezing Test (CRAFT) device has been
described in detail by Tobo (2016). CRAFT is a classical cold-plate device
akin to the DFT and the CS instruments, but it involves procedures to assure
sample isolation, primarily from the cold-plate surface using a layer of
Vaseline®. Droplets containing collected
aerosols are pipetted in a clean hood onto the coated aluminum plate that is
then set on the stage of a portable Stirling-engine-based refrigeration
device (CRYO PORTER, Model CS-80CP, Scinics Corporation). The freezing device
is also operated in a booth that is aspirated with clean air. The temperature
of the plate was measured using a single temperature sensor, and the
uncertainty of temperature is 0.2 ∘C.
For each CRAFT measurement, 49 droplets with a volume of 5 µL were
used, and the temperature was lowered at a rate of 1 ∘C min-1
until all the droplets froze. Results of control experiments with pure water
droplets were used to correct for any contamination introduced by water. Each
freezing experiment was monitored by a conventional video camera. Video image
analysis was used to establish the number fractions of droplets frozen and
unfrozen at 0.5 ∘C intervals. Analyses of nINPs(T)
followed the same scheme as used for the CS and IS measurements. Binomial
confidence intervals (95 %) were determined, as for the IS data.
Aerosol collection methods and processing for immersion freezing
studies
At different times, ambient aerosol samples were collected directly into
liquid or onto filters, for subsequent resuspension into liquid. Collection
directly into liquid was done using a glass bioaerosol sampler (SKC Inc.),
hereafter termed the BioSampler. This unit was typically placed on a table at
1.2 m above ground level. The BioSampler directs particles into a sample cup
filled with 20 mL of ultrapure water (18.2 MΩ cm resistivity and
0.02 µm filtered using an Anotop syringe filter (Whatman, GE
Healthcare Life Sciences)), where they impinge to form an aqueous suspension.
Particle collection efficiencies for this technique exceed 80 % for
particles larger than 200 nm and approach 100 % for particles larger
than 1 µm (Willeke et al., 1998). Particles with diameter
Dp>10µm are expected to impact the inlet wall (Hader et
al., 2014). Sample flow rate was 12.5 L min-1, and impaction liquid
was replenished every 20–30 min by adding ultrapure water into the
collection cup.
For IS-only and some shared samples, particles were also collected onto
pre-sterilized 47 mm diameter Nuclepore™
track-etched polycarbonate membranes (Whatman, GE Healthcare Life Sciences).
Filters were pre-cleaned by soaking in 10 % H2O2 for 10 min,
followed by three rinses in ultrapure water, and were dried on foil in a
particle-free, laminar flow cabinet. Filters were held open-faced in sterile
Nalgene filter units (Thermo Scientific, Rochester, NY). Flow rates varied
from about 8 to 13 L min-1 for ambient temperature and pressure
conditions in different studies. Collection onto 0.2 µm
pore-diameter filters was typical, although comparison vs. 3 µm
pore-diameter filters was also done in some initial experiments. Both filter
types were of ∼ 10 µm average thickness and 15 %
porosity. On the basis of theoretical collection efficiencies (Spurny and
Lodge, 1972), the 0.2 µm pore filters should have collected
particles of all sizes with very high efficiency, the lowest efficiency being
at about 0.1 µm (∼ 80 %). In contrast, the filters with
3 µm pores are expected to collect 15 and 55 % of all particles
at sizes of 0.4 and 1 µm, respectively, increasing to > 75 %
collection at sizes above 1.5 µm. In this manner, the larger pore
size emphasizes the contributions of supermicron aerosols to immersion
freezing INPs.
After particle collection, filters were stored frozen at -25 or
-80∘C in sealed, sterile Petri dishes until they could be
processed (few hours to few months). BioSampler samples were similarly stored
frozen and processed over similar time frames. MOUDI collections for the DFT
method were vacuum-sealed after collection and stored at 4 ∘C in a
refrigerator; shipping was done with cold packs prior to cold-stage flow cell
measurements at the University of British Columbia. We therefore assume
similar impacts, if any, of storage on INPs following thawing for processing.
This study was not initially conceived as one to test storage impacts on
INPs, which should be addressed in future research. We do not expect storage
methods to impact results on the basis of existing documentation in the
literature. For example, in their study of INPs in rainwater, Petters and
Wright (2015) noted that the argument that INP activity remains unaltered by
the freezing of samples and subsequent storage for some time is at the core
of the general application of immersion freezing methods. They noted, with
reference to other literature, the generally better than 1 ∘C
repeatability of freezing temperatures for droplets that undergo repeated
freeze–thaw cycles.
Samples taken during periods when the CFDC was operated at a single
temperature on each date and when immersion freezing methods were aligned in
time, sharing samples in some cases. Data from Waverly, CO, are from Garcia et
al. (2012). Sample volumes ranged from 1600 to 5500 L.
LocationLat, long.DateElevationStandard sampleInstruments(MM/DD/YYYY)(m)typeWaverly, CO40.761, -105.0769/29/101585BioSamplerCFDC, IS10/4/10BioSamplerCFDC, IS10/8/10BioSamplerCFDC, IS11/3/10BioSamplerCFDC, ISCSU Atmos Chem,40.587, -105.1509/6/131591Ultrapure waterCFDC, IS, CSFort Collins, CO9/6/13BioSampler blankCFDC, IS, CS9/6/13BioSamplerCFDC, IS, CS9/6/133 µm filterCFDC, IS, CS9/6/130.2 µm filterCFDC, IS, CS9/12/13BioSamplerCFDC, IS9/12/133 µm filterCFDC, IS9/12/130.2 µm filterCFDC, IS11/12/13BioSamplerCFDC, IS, CS, MOUDI-DFT11/12/133 µm filterCFDC, IS, CS, MOUDI-DFT11/12/130.2 µm filterCFDC, IS, CS, MOUDI-DFT11/13/13BioSamplerCFDC, IS, CS, MOUDI-DFT11/13/133 µm filterCFDC, IS, CS, MOUDI-DFT11/13/130.2 µm filterCFDC, IS, CS, MOUDI-DFT11/14/13BioSamplerCFDC, IS, CS, MOUDI-DFT
For processing of INP freezing spectra, filters were transferred to sterile,
50 mL Falcon polypropylene tubes (Corning Life Sciences), immersed in
7.0–10.0 mL of ultrapure water and tumbled for 30 min in a rotator
(Roto-Torque, Cole-Palmer) to suspend particles in liquid. Common liquid
suspensions were shared amongst methods in some cases (see Sect. 2.3),
following freezing and shipping to different investigators. We detected no
measurable impact of processing rinsed suspensions immediately vs. after
freezing of the bulk water, mostly supported by other recent studies (Beall
et al., 2017). We will note that while all immersion freezing methods
performed tests comparing freezing of the liquid samples and the purified
water used in their setups, and corrected for pure water freezing events, no
correction is made for any INPs that might be released from the filters used
for collection. We have found that filters release a modest number of INPs
active at lower temperatures, even after the pre-cleaning with H2O2
and purified water. A detailed analysis of this will be presented in a future
publication. The percentages of undiluted INPs due to such contamination is
∼ 3 % in the -25 to -30∘C range, and since immersion
freezing measurements at these temperatures require dilution of liquid
samples by 100 to 3000 times, we neglected any corrections.
Sampling sites/periods and objectives
Sampling sites represent a variety of ecosystems, climates and elevations
across the western US, including agricultural regions of the US High Plains,
intermountain desert regions and a coastal site. The majority of data
included in this intercomparison involved periods that did not include all
groups and were not temporally aligned for all instrument systems.
Nevertheless, substantial overlap of sampling periods occurred in all cases.
Very often, the CFDC sampling was conducted to obtain data at multiple
temperatures, while offline collections were made for longer periods to
obtain integrated INP temperature spectra. Times when the sampling periods
were the same for the offline systems and for the CFDC, while it was
operating at a single temperature, are listed in Table 1. Other site
locations, characteristics and instruments participating when there were
overlapping sample periods are listed in Table 2.
Colorado State University, Fort Collins, CO, USA
Sampling was conducted outside of the atmospheric chemistry building at
Colorado State University at different times and including different methods.
The laboratory site is on a small hill on the western edge of the Fort
Collins urban area, residing amongst surrounding grasslands. Initially, a
series of measurement days were conducted in which collections for three
immersion freezing methods were made while the CFDC sampled at a single
temperature for the entire sampling period. While this protocol permitted
only a single comparison point vs. the temperature spectra obtained by
offline measurements, the purpose was to obtain a statistically significant
CFDC nINPs(T) value during the course of time-integrated offline
samples and to assure that any signal variance occurring during sampling was
the same for all measurements. Such aligned sampling was conducted on five
different days (see Table 1). Participating in these temporally aligned
experiments were the IS, CS and MOUDI-DFT instruments. For these periods,
the filter sampling units, BioSampler and (when used) MOUDI sampling units
were set in close proximity and at the same sampling elevation. Filter
suspensions from the two pore-size (0.2 and 3.0 µm) filter
collections and from the BioSampler were shared for IS and CS measurements.
All CS data were analyzed using the pico- and nanodrop technique.
Sampling locations, elevations, dates and instruments involved in
sampling at field sites when the CFDC sampled at varied temperatures during
integral offline collections. All sampling at these sites was by filters
except for the use of a BioSampler for the CS at Bodega Bay Marine
Laboratory, the IS at Waverly, and the MOUDI-DFT at Manitou Experimental Forest (Huffman
et al., 2013) and Colby (Mason et al., 2015). CFDC data from Manitou Experimental Forest
are from Tobo et al. (2013). Data from Waverly, CO, are from Garcia et
al. (2012).
RegionLocationLat, long.DateElevation (m)InstrumentsForestManitou Experimental39.094, -105.1018/17/11,2370CFDC,Forest Observatory, CO8/18/11MOUDI-DFTAgriculturalWaverly, CO40.761, -105.0769/29/10,1585CFDC, IS10/4/10,10/8/10,11/3/10AgriculturalColby, KS39.394, -101.06610/14/14,966CFDC, IS, MOUDI-DFT10/15/14AgriculturalLamont, OK36.607, -97.4884/30/14,315CFDC, IS5/4/14,5/5/14,6/5/14,6/7/14,6/8,14CoastalBodega Bay Marine39.307, -123.0661/26/15,5CFDC, IS, CSLaboratory, CA2/2/15Semi-aridCanyonlands, UT38.071, -109.5635/11/16,1627CFDC, IS, CRAFT5/12/16Semi-ruralCSU Atmos Chem,40.587, -105.1505/18/16,1591CFDC, IS, CRAFTFort Collins, CO5/19/16
Sampling was also conducted at CSU without exact temporal overlap of CFDC,
IS and CRAFT method measurements, as noted in Table 2. CRAFT filters
(0.2 µm pore size) were drawn for 6 h at a flow rate of
10 L min-1 at standard temperature and pressure (STP) conditions (T= 273 K, 1013.5 mb). IS filter (0.2 µm pore size) were drawn
for 4 h at a flow rate of 13 L min-1 at ambient temperature and
pressure. The CFDC sample was temporally aligned with the IS sample, and
single operating temperatures were used.
Northern Colorado, USA, agricultural region
Sampling over previously harvested fields during fall 2010 was conducted at a
rural site approximately 26 km NNE of the CSU atmospheric chemistry
building, at Grant Family Farms, near the village of Waverly, CO. The
sampling field sites on different days, the sampling protocol and the results
used in the present study are discussed in detail by Garcia et al. (2012).
Sampling by CFDC and IS (BioSampler) were temporally overlapped in this
study. This site is referred to as NoCO in the data tables in the
Supplement.
Manitou Experimental Forest, CO, USA
Sampling within an open forest site at MEFO as part of the
Bio-hydro-atmosphere interactions of Energy, Aerosols, Carbon, H2O,
Organics & Nitrogen project (Ortega et al., 2014) during summer 2011 was
conducted as described by Huffman et al. (2013), Prenni et al. (2013) and
Tobo et al. (2013). Only two selected periods from that study for which there
was partial overlap of samples from the CFDC and MOUDI-DFT methods were
available for this study.
Kansas, USA, agricultural region
Sampling periods were conducted in and around the times of different crop
harvesting at Kansas State University Northwest Research Extension Center in
Colby, KS, as part of a larger study.
Sampling periods
used for this study were during mornings before or evenings following
harvesting of various crops and during daytime near fields being harvested
of soy and sorghum crops. CFDC sampling was conducted from the CSU Mobile
Laboratory facility, using gasoline-powered generators, as described
previously by McCluskey et al. (2014). The mobile laboratory was in all cases
well upwind of the generators. Aerosols were sampled through an inlet
comprised of a stainless-steel rain hat with a 1/2 in. OD stainless-steel
tube attached. MOUDI-DFT (Mason et al., 2016) and filter samples were
collected with their inlets at the same approximate elevation as the CFDC
inlet and used separate pumps for drawing samples. The CFDC scanned
different temperatures during the IS filter (0.2 µm) and MOUDI-DFT
sampling periods.
Temperature spectra of INP number concentrations (nINP)
from IS and CS measurements and a CFDC measuring at a single temperature over
a 4 h sampling period. Ambient aerosols were sampled outside of the Colorado
State University atmospheric chemistry building on
(a) 12 September 2013 and (b) 6 September 2013.
Temperature spectra were separately measured for simultaneously collected
filter samples with different pore sizes and liquid samples from a
BioSampler. Uncertainty values (95 % confidence intervals) are shown.
Southern Great Plains (SGP), USA, site
The site at Lamont, OK (Table 2), is the central instrumentation suite
location for the US Department of Energy's Atmospheric Radiation Measurement
Climate Research Facility SGP field site. CFDC and IS instruments both
drew air from a platform at 10 m above ground elevation at this site.
Sampling occurred in a transition from dry to wet conditions in the spring of
2014. The CFDC was operated to scan temperatures during the IS filter
(0.2 µm) sampling period. A selection of representative days of
data were chosen, and full study data will be included in a separate
publication.
Bodega Marine Laboratory, CA, USA
Sampling near Bodega Bay, CA (BBY in subsequent figures), occurred during the
CalWater-2015 study (Ralph et al., 2015; Martin et al., 2016). The sampling
site was at the University of California, Davis Bodega Marine Laboratory,
∼ 100 m ENE of the seashore and ∼ 30 m north of the
northernmost permanent building at the site (Martin et al., 2016). The CFDC
and IS instruments sampled from approximately 4 m above the surface. The
CFDC was operated to scan temperatures during the IS filter
(0.2 µm) sampling period. CS BioSampler samples, overlapping with
IS and CFDC sampling, were drawn from an elevation of 1 m above the
vegetated surface, approximately 20 m west of the other samplers. All BBY CS
data are analyzed using the microdrop technique. A few representative days
are chosen from the data set for comparison of IS and CS data with CFDC data.
Comparison of the complete CS and IS data sets will be included in a
publication in preparation.
Canyonlands Research Center, UT, USA
The Nature Conservancy's Canyonlands Research Center is an intermountain
(Rocky Mountains, US), high-desert site located adjacent to Canyonlands
National Park in SE Utah. Sampling occurred in May of 2016. IS and CRAFT
filters were drawn at 1.2 m above ground, the same elevation as the CFDC
inlet. CRAFT filters were drawn for 6 h at a flow rate of 10 L min-1 at STP conditions (T=273 K,
1013.5 mb), at this site and at CSU. IS filters (0.2 µm pore size)
were drawn for 6 h at a flow rate of 13 L min-1. CFDC sampling
overlapped with the IS filter period, but operating temperature was
varied.
Three additional experimental comparison days, as in Fig. 1 but for
cases where all four methods were operational for consistent sampling
periods. These dates were 12 –14 November in
panels (a)–(c), respectively. The legend is shown in
panel (a). The additional data in green are from the MOUDI-DFT
method (all sizes included), including median (cross) and upper and lower
bounds.
ResultsComparison of cases with perfect temporal overlap of sample data
collections
Figure 1a compares IS and CFDC data for two 4 h study periods at the CSU
site. In the figure CFDC INP concentrations at -16∘C are
integrated and averaged for the entire IS filter sampling period for
comparison to IS data collected both on filters and using the BioSampler.
Considering the capture efficiencies vs. size noted in Sect. 2.2, the lack of
significant difference in IS nINPs(T) measured with the filters of
0.2 and 3 µm pore sizes implies that most INPs were likely large
enough to be captured effectively. This crudely suggests an INP mode size at
about 1 µm or larger. This is also a size that is collected with
high efficiency in the BioSampler, for which similar INP concentrations were
measured. This example also shows the uncertainties in temperature spectra of
INP number concentration from the IS. In this case, one can see a range of
about a factor of 4 in INP number concentration and an equivalent range of
2–4 ∘C using different collection methods, and in consideration of
confidence in measurements made at any particular temperature. The CFDC data
collected using the aerosol concentrator are in agreement within
uncertainties of all particle collection methods in this case.
In Fig. 1b, results are shown from a case where filter rinse suspension and
BioSampler suspension were also shared with the CS instrument for offline
processing of samples collected from the CSU site on 6 September 2013. There
is significant overlap between the IS and CS data in the temperature range
from -6 to -23∘C (the lowest temperature limit of IS processing
for these particular experiments). No significant bias is discernable between
IS and CS data for any of the collection methods. Once again, correspondence
of the CFDC data (using the aerosol concentrator in this case) with other
methods is good at a processing temperature of -18.2∘C. However,
the CFDC data fall a factor of 2–5 lower than the immersion freezing
methods. This is similar to data reported in Garcia et al. (2012) for which
the discrepancy was attributed primarily to the failure of the CFDC
instrument to sample larger aerosols. Nevertheless, results from this
sampling day support the conclusions of general agreement between methods
obtained in Fig. 1a.
Figure 2 shows results from three additional cases for which there was
perfect temporal co-sampling by the CFDC, IS, CS and MOUDI-DFT methods. In
these cases, the IS and CS shared samples of particles collected during the
same time period, while the MOUDI-DFT was operated independently. We note
that the error bars on MOUDI data reflect upper- and lower-bound estimates, as
discussed in Sect. 2.1.3. Figure 2 highlights not only some points already made but
also the occurrence of a range of discrepancies in nINPs(T)
between the MOUDI-DFT and other methods, and for CFDC data collected
simultaneously at temperatures below -20∘C. The CS method
typically measures the highest nINPs(T) overall for the same
collections of aerosols (filter or BioSampler), suggesting a temperature
offset of at least 1 ∘C between these systems that may have as its
source the temperature measurement of the liquid wells or drops. The
MOUDI-DFT results trend with the other immersion freezing methods on all
days but agree quantitatively with them on only one of three days (Fig. 2a)
and fall lower than nINPs(T) determined by the CS and IS on two
other days: by a factor of 2 to 5 (Fig. 2c) in one case and 20 to 50 in the
other (Fig. 2b). These two cases have been discussed previously in Mason et
al. (2015), and we will revisit the largest discrepancies in both cases in
later discussion. Similar to the MOUDI-DFT results, the CFDC data also show a
consistent underestimate of nINPs(T) compared to the CS and IS in
all three cases, with a trend that increases from a factor of 2–4 at
-23∘C up to 10 times at -30∘C (Fig. 2a).
Scatterplot of INP number concentrations obtained with different
immersion freezing methods plotted against CFDC online measurement results
obtained at 105 % RH and temperatures ranging from approximately -15 to
-31∘C: (a) IS, (b) MOUDI-DFT (medians of data
such as shown in Fig. 1), (c) CS and CRAFT, (d) all data
combined from offline immersion freezing tests. The MOUDI-DFT data
in (b) include data for all particles sizes assessed (“all”) and
for the particle size range of 0.3–3.2 µm (“size”) best aligned
with the effective CFDC sampling size range. Error bars represent 95 %
confidence intervals, as defined for each method. Light dashed gray lines are
simple linear relations intended only to guide the eye.
Comparison for cases of imperfect temporal overlap of sample
data collections
The data shown in Figs. 1 and 2, for which there was complete temporal
overlap of observations, provide a limited number of evaluations of
measurement correspondence and uncertainties that may occur due to different
size ranges of collection and natural variations in INP compositions and
concentrations that may occur over varied sampling times as measured across
the mixed-phase cloud temperature regime. This situation will surely be
improved in future studies as many different instrument teams worldwide begin
to compare data collected at common sites. To expand understanding, we
considered all cases in which the CFDC was sampling simultaneously with other
methods but without the restriction of a single CFDC processing temperature
for the full sampling period. There are also cases when the offline sample
periods overlapped but did not perfectly align. Thus, while seeking further
insights by folding in data from additional times and collection sites, we
must acknowledge that such comparisons leave open the possibility that
temporal variability may impact comparison of methods. Nevertheless, this
replicates many field study situations where multiple ice nucleating
instruments may be deployed but may not sample for the same time periods.
In Fig. 3, we combine periods of perfect sampling overlap with these other
cases for which one or more of the immersion freezing methods were performed
for a few-hour period, during which CFDC sampling intervals (typically
10–15 min at a single temperature) occurred. Comparison of the CFDC and IS
measurements is shown in Fig. 3a. These results reinforce those in Fig. 2,
indicating that the IS assessment of nINPs(T) agrees on average
with the CFDC-measured values when the CFDC processed particles at 105 %
RH at the lower end of the dynamic range of nINPs(T). The IS
method, however, measures higher concentrations than the CFDC at higher
nINPs(T), resulting in a non-unity relational slope. The linear
relational slope between IS and CFDC data is shown by the light gray dashed line
in Fig. 3a. The same representation is applied in all panels of Fig. 3. We
provide these fits only to show general trends between the different data
sets and do not provide fit parameters herein because a deeper consideration
of the source of discrepancies requires additional inspection of trends as a
function of temperature, which follows below. Higher nINPs(T)
typically occur at lower temperatures. Results are similar regardless of
measurement site, but with relatively high variability in the relation
between single CFDC and IS measurements even at a single site, and with
greater discrepancy in the data set from Colby, KS, which we suggest is the
result of an abundance of larger INPs not sampled by the CFDC during this
harvesting period.
The MOUDI-DFT data show the best correspondence overall vs. the CFDC
measurements (Fig. 3b), irrespective of whether all aerosol sizes are
considered for the DFT measurement or are limited to a range of particle
sizes similar to those entering the CFDC. There is a slight positive bias for
the MOUDI-DFT method when all sizes are considered, as expected given the
CFDC limitation on particle sizes sampled.
Overlapping comparisons between the CS and CFDC, and CRAFT and CFDC, while
more limited (Fig. 3c), show a relatively high bias of the CS and CRAFT data,
most exaggerated at higher nINPs(T) and correlated with lower
temperatures as discussed shortly.
Overall comparisons by offline methods vs. the CFDC are shown in Fig. 3d.
These demonstrate that, although a consistent linear (but not 1 : 1)
relationship could be inferred between offline immersion freezing and CFDC
measurements, discrepancies for all methods and sampling times taken together
at a CFDC nINP(T) of 1 L-1 can reach nearly 2 orders of
magnitude. Discrepancies appear to reduce to within about 1 order of
magnitude at higher nINPs(T), although the degree to which this is
real or the result of a smaller number of cases is not yet clear. We may note
of course that CFDC measurements have their greatest uncertainties in the
range of concentrations at or below 1 L-1.
The same data sets used in Fig. 3, and compiled in Table S1 in the
Supplement, are used in Fig. 4 to explore the temperature dependence of
immersion freezing measurement results vs. the CFDC when all sampling
scenarios are considered (multiple aerosol scenarios, perfect or imperfect
overlap of sampling times). In examining the IS vs. CFDC comparisons
(Fig. 4a), the scatter in the relation is again the most striking feature,
while the temperature-dependent bias also becomes clear to a greater or
lesser degree at all sampling sites, the least at CSU and the SGP site, and
the most at Bodega Bay and in the harvesting period in Kansas. The strong
positive bias of INP measurements by the IS at lower temperatures in Kansas
is not consistent with the fact that larger INPs (> 2.5 µm),
which are not sampled by the CFDC, are not thought to dominate INPs at lower
temperatures (Mason et al., 2016). A more modest positive temperature bias is
noted in comparing MOUDI and CFDC concentrations vs. temperature at below
-25∘C (Fig. 4b), and the underestimate of INP concentrations due
to the elimination of coarse-mode aerosols in CFDC sampling ranges from about
2 to 4 times (see MOUDI “all” vs. “size” in Fig. 4b), consistent with the
estimates of coarse-mode INP fractions by Mason et al. (2016). We may note
similarly good agreement between INP concentrations measured by the CFDC and
DFT methods across similar temperature ranges for marine aerosols (DeMott et
al., 2016). Strong positive biases of CS- and CRAFT-measured INP
concentrations vs. the CFDC measurements are seen to progressively occur as
temperatures decrease from -20 to -30∘C (Fig. 4c).
Base 10 logarithm of the ratio of INP number concentrations measured by
various immersion freezing methods vs. the CFDC at different sites, denoted
as in previous figures. IS 0.2 µm filter samples are shown
in (a) from five sites. MOUDI-DFT data are compared from three sites
in (b), where “size” and “all” refer to whether INP number
concentrations are from MOUDI size ranges overlapping with sizes permitted
into the CFDC or from all sizes. CS and CRAFT ratios are shown
in (c), where all blue points are for the CS, and “bio” refers to
BioSampler collections.
Discussion
In this section, we summarize observations regarding comparisons of the INP
measurement methods and discuss potential reasons for discrepancies that bear
future investigation. It has been shown that there are times when multiple
measurement techniques give excellent agreement for nINPs(T) in
the immersion freezing mode. Agreement is best at temperatures warmer than
-20∘C and for nINPs(T) less than ∼ 5 L-1.
At lower temperatures and higher nINPs(T), most offline immersion
freezing methods, with the exception of MOUDI-DFT, estimate higher than the
online CFDC method, by ratios ranging from a few to 10 times. We must caution
that the overall range of nINPs(T) assessed and values present at
different temperatures may reflect the aerosol measured at ground level at
the selected sites and times, scenarios that may not represent all locations
and times worldwide. Nor may these results be the same if the comparisons
were made entirely for free-tropospheric aerosols, for example as assessed
from an aircraft or at some mountaintop sites. Nevertheless, the potential
issues in obtaining agreement between methods will be common in any sampling
scenario.
A factor in any series of immersion freezing measurements is the time
dependence of nucleation. In a study of the time-dependent freezing of
kaolinite particles, Welti et al. (2012) demonstrated that the majority of
freezing occurred within about a period of 10 s or less at the temperatures
-30 to -37∘C, with 0.8 µm diameter particles needing
far less time for activation than 0.4 µm particles. Studies of
freezing rates for other natural INP types across broader temperature ranges
indicate that immersion freezing is indeed not a purely stochastic process
and is far more sensitive to temperature, with the consequence that the
increase in nINPs(T) achieved when droplets remain at a single
temperature for periods longer than seconds to minutes is typically overcome
by a few degrees of additional cooling (Vali, 2014; Wright et al., 2013). The
CFDC nINPs(T) attributed here to immersion freezing were obtained
for a total processing time of approximately 7 s, during the last 2 s of which
activated droplets are evaporating (DeMott et al., 2015). This residence time
is constrained by flow rates required for limiting thermally driven reverse
flow circulations in the CFDC. By comparison to the Welti et al. (2012)
study, it seems likely that the CFDC activation times allow for capturing the
majority of immersion freezing activity in most circumstances. Nevertheless,
we expect that the CFDC might underestimate nINPs(T) to a greater
extent than the IS measurements, which are made while ramping at a very slow
cooling rate equivalent to 1 ∘C in 3 min. Since the DFT uses much
faster cooling rates (5–10 ∘C min-1), this might explain the
better correspondence with the CFDC data. However, it cannot explain the
temperature-dependent nature of the bias between other immersion freezing
methods and the CFDC, and so it seems not to be the only source of this
discrepancy.
Here we must also reiterate that the processing of submicron-mode mineral
dust particles at 105 % RH in the CFDC has been shown to underestimate
nINPs(T) by an average temperature-independent factor of 3,
as confirmed by laboratory cloud chamber simulations. This factor was related
to the fact that higher RH is typically required to fully activate all
particles (hygroscopic or hydrophobic) as droplets to subsequently be
available for freezing in the CFDC residence time (DeMott et al., 2015;
Garimella et al., 2017). However, practical operation of the CFDC at higher
RH (109 % may be required for full activation) is prohibited in sampling
of natural aerosol distributions because the largest aerosols could persist
as droplets through the evaporation section of the instrument under these
conditions, thus contaminating INP determination using optical sizing. Hence,
it is unknown if natural INP populations are being underestimated for similar
reasons. Based on the recent study of Garimella et al. (2017), it seems
possible that underestimation of INP concentrations occurs for CFDC-style
instruments independent of the aerosol type. Consequently, lines indicating a
factor of 3 higher than the 1 : 1 relation have been placed on plots in the
panels of Fig. 3. While it is noted that increasing the CFDC nINPs(T) by a factor of 3 leads to better overall agreement of CFDC data with the CS
and CRAFT data especially, this constant expected offset does not explain the
progressive underestimate of the CFDC in comparison to most immersion
freezing methods (the IS, CS and CRAFT being most like other methods used
worldwide) at higher nINPs(T) and lower temperatures.
Immersion freezing methods comparison, shown as the base 10
logarithm of the ratio of the CS, CRAFT and MOUDI-DFT method INP
concentrations for perfect or imperfect overlap of co-sampling periods with
the IS INP number concentrations. Samples collected outside the CSU
atmospheric chemistry facility are shown as filled symbols, while samples
collected at other sites on different days (CS: Bodega Bay; CRAFT:
Canyonlands Research Center; DFT: Colby, KS) are shown as open symbols.
A factor that could artificially increase nINPs(T) at lower
temperatures in methods that immerse the entire aerosol population first into
liquid (IS, CS, CRAFT) is the potential breakup of aggregates containing
multiple INPs (e.g., via the deflocculation of small aggregates as a result
of the strong reduction in di- and trivalent cation concentrations in the
deionized water used for making dilution series, or by the fragmentation of
mucigels (Hill et al., 2016)) and the possible dissolution release of
surface-active INP materials present on single particles when suspended in
deionized water (O'Sullivan et al., 2016). It seems possible that such action
would have the greatest impact on INPs active at lower temperatures (rather
than the most active INPs), since these may be small clay/organic matter
aggregates or flocs that fragment when exposed to deionized water. Since the
MOUDI-DFT method immerses a relatively small number of particles directly and
without agitation in small drops prior to freezing, it is interesting to note
that the least temperature-dependent bias occurs for these measurements in
comparison to the CFDC. This point is shown more clearly by comparing only
the offline immersion freezing methods in Fig. 5. In this figure, the
different measured INP concentrations are taken as a ratio vs. the IS, which
sampled the most times and scenarios. Data at 1 ∘ temperature
resolution are included in this comparison, as compiled in Table S2. Again,
relatively high variability of at least 1 order of magnitude at any
temperature is noted for the relations between methods. Among methods that
involve immersion of all particles in a single volume of water prior to
setting up arrays (CS, CRAFT, IS), the IS falls to the low side of the other
measurements by an average factor of about 2.5. This is not a significant
difference, in consideration of the likely temperature uncertainties
discussed in relation to Figs. 1 and 2. The MOUDI-DFT method that immerses
relatively small populations of particles shows relative equivalence to the
full immersion methods at modest to moderately supercooled temperatures but
measured consistently lower INP concentrations at below about
-20∘C in the few cases when co-sampling was conducted with the IS
(CSU and Kansas). This is consistent with the discrepancy seen also vs. CFDC
data. Interestingly, a lower-temperature enhancement of INPs appearing in
full immersion methods vs. continuous flow methods was not observed in recent
laboratory tests comparing many measurement methods while sampling mineral,
soil dust and biological particle samples that had been purposely limited to
sizes smaller than 2 µm (DeMott et al., 2017). While this points to
coarse-mode particles and their dissolution/fragmentation into multiple INP
units as the source of differences, future experiments will be needed to
confirm or deny that this is either an artifact or a behavior in natural
aerosols that the CFDC cannot effectively capture.
Particle size limitations lead to CFDC underestimates of nINPs(T)
in comparison to some immersion freezing methods. This is because of the need
to remove particles larger than 2.4 µm. This removal of larger
aerosols is necessary when differentiating grown ice crystals from aerosols
by size alone. Even absent the use of impactors, it would be difficult for
most online systems to effectively sample larger particles due to the design
of inlet systems. With reference to the study of Mason et al. (2016), which
entailed sampling with the MOUDI-DFT method at various sites, one might
estimate that on average about 50 % of INPs are at sizes larger than
2.4 µm in the surface boundary layer. Comparison of MOUDI-DFT with
CFDC data in this study is consistent with this same estimate (Fig. 3b).
Again, this would not apparently explain a progressive slight increase in
CFDC underestimation vs. the MOUDI-DFT at lower temperatures unless larger
INPs specially dominate ice nucleation at lower temperatures, a result not
consistent with Mason et al. (2016).
In evaluating the low-temperature discrepancies by noting the better
correspondence of MOUDI-DFT and CFDC methods, it is also necessary to note
the potential issue of particle bounce in the MOUDI in some cases (Mason et
al., 2016). While the conditions for this to occur are not well quantified,
since both INP size and phase state (as this may be influenced by low
relative humidity) may affect bounce, very dry conditions have been indicated
as times when this may become an issue for MOUDI impaction onto the
substrates used in the DFT instrument. Interestingly, average RH during the
sampling period on 13 November 2013 (Fig. 2b) was between 15 and 20 % vs. 40–45 % for days on either side
(Fig. 2a, c), potentially impacting and explaining the MOUDI-DFT results on
this day. For this reason, this day was excluded in Figs. 3–5. Sample
humidification of the system could mitigate this factor as a potential issue
in future sampling.
Conclusions
This study has inspected the correspondence of ice nucleation measurement
systems for co-sampling ambient ice nucleating particles. In this case, we
considered systems for measuring immersion freezing nucleation with a common
online method used in a manner to induce activation of cloud droplets prior
to ice nucleation. Very good agreement within uncertainty limits was obtained
under many conditions for samples that had perfect temporal overlap. In other
cases, discrepancies can approach 2 orders of magnitude and are not
explicable without inferring systematic artifacts inherent to one or more
techniques. The results summarized in Fig. 3d show the uncertainties that can
be expected when employing one or more of these instrument systems for
measuring atmospheric INPs. Within these uncertainties, the data suggest that
the low bias of immersion freezing methods reported by Hiranuma et al. (2015)
for sampling of individual surrogate dusts in the laboratory was not evident
in these ambient data sets.
With regard to particle sampling methods for immersion freezing measurements,
use of a BioSampler or a filter was interchangeable, at least for the
continental boundary layer sampling for which these methods were compared.
This was demonstrated for individual and for cross-technique methods (IS vs.
CS) for assessing immersion freezing from the same samples. Since Nuclepore
filters seem to efficiently capture and release INPs, these provide
ease-of-use benefits in many field scenarios, although the role of retention of
particles on some filter types has not been assessed here. Potential effects
of sample storage protocol also remain to be investigated.
The strongest discrepancies between methods appear at both warmer and colder
ends of the scale of mixed-phase cloud freezing temperatures. At the warmer
end (T>-20∘C), sampling statistics and uncertainties can
dominate comparisons of online and offline methods. Full explanations for the
maximum 2-orders-of-magnitude range of variation in this temperature regime
remain unresolved. In contrast, at lower temperatures the IS, CS and CRAFT
methods measured more INPs than detected by the CFDC and MOUDI-DFT. Potential
artifacts or biases are present in these comparisons and have been discussed
here, including varied assessment of time dependence of ice nucleation;
necessary exclusion of larger INPs by online instruments such as the CFDC;
and immersion of all particles into relatively large volumes of deionized
water in most, but not all, immersion freezing methods vs. activation of
single particles in CFDCs. In addition, it is expected that all CFDC-type
instrument data may require correction for not being able to access full
immersion of particles until higher RH than can commonly be used when
sampling ambient particles, or else this issue requires future mitigation
(e.g., insertion of particles into the instrument lamina could be improved).
Hence, no assured conclusions regarding the sources of discrepancies can be
stated at this time except that size biases in sampling need to be
acknowledged. Effort thus remains to make INP measurements fully quantitative
and comparative across methods if correspondence within less than 1 order
of magnitude is desired. Even amongst standard immersion freezing methods,
uncertainties of a factor of a few nINPs(T) and 2–4 ∘C
are likely common on the basis of this study and may be the best that can be
achieved. Application of size selection to immersion freezing collections for
comparison to MOUDI-DFT data (especially at lower temperatures) and CFDC
data; information on INP compositions inferred under all sampling scenarios
to help constrain influences of various types (e.g., methods of Hill et al.,
2016); and an intercomparison of all methods vs. a cloud parcel simulation
chamber, considered as a de facto standard, would all assist resolution and
improvement of understanding of measurement discrepancies.
Specific data sets are available through referenced studies,
and all data used in figures in this manuscript are tabulated in the
Supplement.
Acronyms and symbols (in italics) used
AdepositTotal area of the sample deposit on the hydrophobic glass cover slip for the MOUDI-DFT methodADFTArea of the sample monitored in the digital video during MOUDI-DFT freezing experimentsAIDAKarlsruhe Institute of Technology Aerosol Interactions and Dynamics of the Atmosphere cloud chamberBBYReference to Bodega Bay, CA, USA, field site located at the University of California,Davis Bodega Marine LaboratoryBioSamplerShorthand for impinger device, the bioaerosol sampler, SKC Inc.CFDCColorado State University continuous flow diffusion chamberCINPs(T)Number concentration of INPs per volume of liquidCRAFTNational Institute of Polar Research (Japan) Cryogenic Refrigerator Applied to Freezing TestCRCThe Nature Conservancy's Canyonlands Research Center, adjacent to Canyonlands National Park, UT, USACSNorth Carolina State University cold-stage freezing assayCSUColorado State University, also used to denote the sampling site outside of theDepartment of Atmospheric Science's atmospheric chemistry (Atmos Chem) buildingDpAerosol particle diameterfneCorrection factor to account for the uncertainty associated with the number ofnucleation events in each experimentfnuCorrection factor to account for non-uniformity in particle concentration across each MOUDI sampleINP(s)Ice nucleating particle(s)ISColorado State University ice spectrometerKansasRefers to state of Kansas sampling, at Colby, KS, USALACISLeibniz Institute for Tropospheric Research's Leipzig Aerosol Cloud Interaction SimulatorMEFOManitou Experimental Forest ObservatoryMOUDI-DFTUniversity of British Columbia's micro-orifice uniform deposit impactor–droplet freezing techniquenINPs(T)Number concentration of INPs per volume of air at a given temperaturenuNumber of drops unfrozen in immersion freezing arraysNTotal number of droplets or liquid entities (in arrays or condensed) in immersion freezing devicesNoCONorthern Colorado, referring to agricultural sampling region in Waverly, COSGPUS Department of Energy's Atmospheric Radiation Measurements Climate Research FacilitySouthern Great Plains site, located near Lamont, OK, USASTPStandard temperature (273 K) and pressure (1013.5 mb) conditions, typically to refer to volumesconverted to these conditionsTTemperature (∘C)VVolume of individual droplets or aliquots in immersion freezing arrayVsSample volume of air collected (L-1)VwTotal liquid volume into which particles are placed (mL)
The Supplement related to this article is available online at https://doi.org/10.5194/acp-17-11227-2017-supplement.
The authors declare that they have no conflict of interest.
Acknowledgements
Funding for this work was provided by National Science Foundation grants
AGS1358495 (Paul J. DeMott, Thomas C. J. Hill, Kaitlyn J. Suski and
Ezra J. T. Levin), AGS1010851 (Markus D. Petters and John D. Hader),
AGS1450690 (Markus D. Petters, Nicholas Rothfuss and Hans P. Taylor),
AGS1450760 (Paul J. DeMott, Thomas C. J. Hill, Christina S. McCluskey and
Sonia M. Kreidenweis) and AGS1433517 (Gregory P. Schill). Support for
operations at the US Department of Energy's Southern Great Plains site was
provided by the Atmospheric Radiation Measurement (ARM) Climate Research
Facility, a US Department of Energy Office of Science user facility sponsored
by the Office of Biological and Environmental Research. Allan K. Bertram,
Ryan H. Mason and Cedric Chou acknowledge support of the Natural Sciences
and Engineering Research Council of Canada. Yutaka Tobo acknowledges support
from the Japan Society for the Promotion of Science (JSPS, KAKENHI grant
numbers 15K13570 and 16H06020), Japan's National Institute of Polar Research (NIPR, Project Research
KP-3) and the Arctic Challenge for Sustainability (ArCS) project.
Yvonne Boose thankfully acknowledges support from the Zeno Karl Schindler
foundation. Special thanks to Kim Prather, Andrew Martin and colleagues at the
University of California, San Diego for their logistical support during
studies at Bodega Bay. We also thank the University of California, Davis
Bodega Marine Laboratory for the use of laboratory and office space and
shipping and physical plant support while collecting data. We thank the
National Park Service and Jeffrey Collett for use of their respective mobile
laboratories during various field study deployments. We thank Freddie Lamm,
Dan Foster and Marv Farmer at the Kansas State University Northwest
Research Extension Center for their help with
coordinating the measurements. Finally, we thank the Nature Conservancy's
Canyonlands Research Center and Field Station Manager Philip Adams for
helping arrange and select sites for measurements there. Edited by: Anne Perring Reviewed by: two
anonymous referees
ReferencesBeall, C. M., Stokes, M. D., Hill, T. C., DeMott, P. J., DeWald, J. T., and
Prather, K. A.: Automation and heat transfer characterization of immersion
mode spectroscopy for analysis of ice nucleating particles, Atmos. Meas.
Tech., 10, 2613–2626, 10.5194/amt-10-2613-2017, 2017.Boose, Y., Sierau, B., García, M. I., Rodríguez, S., Alastuey, A.,
Linke, C., Schnaiter, M., Kupiszewski, P., Kanji, Z. A., and Lohmann, U.: Ice
nucleating particles in the Saharan Air Layer, Atmos. Chem. Phys., 16,
9067-9087, 10.5194/acp-16-9067-2016, 2016.
DeMott, P. J., Prenni, A. J., Liu, X., Kreidenweis, S. M., Petters, M. D.,
Twohy, C. H., Richardson, M. S., Eidhammer, T., and Rogers, D. C.: Predicting
global atmospheric ice nuclei distributions and their impacts on climate, P.
Natl. Acad. Sci. USA, 107, 11217–11222, 2010.DeMott, P. J., Prenni, A. J., McMeeking, G. R., Sullivan, R. C., Petters, M.
D., Tobo, Y., Niemand, M., Möhler, O., Snider, J. R., Wang, Z., and
Kreidenweis, S. M.: Integrating laboratory and field data to quantify the
immersion freezing ice nucleation activity of mineral dust particles, Atmos.
Chem. Phys., 15, 393–409, 10.5194/acp-15-393-2015, 2015.DeMott, P. J., Hill, T. C. J., McCluskey, C. S., Prather, K. A., Collins, D.
B., Sullivan, R. C., Ruppel, M. J., Mason, R. H., Irish, V. E., Lee, T.,
Hwang, C. Y., Rhee, T. S., Snider, J. R., McMeeking, G. R., Dhaniyala, S.,
Lewis, E. R., Wentzell, J. J. B., Abbatt, J., Lee, C., Sultana, C. M., Ault,
A. P., Axson, J. L., Diaz Martinez, M., Venero, I., Santos-Figueroa, G.,
Stokes, M. D., Deane, G. B., Mayol-Bracero, O. L., Grassian, V. H., Bertram,
T. H., Bertram, A. K., Moffett, B. F., and Franc, G. D.: Sea spray aerosol as
a unique source of ice nucleating particles, P. Natl. Acad. Sci. USA, 113,
5797–5803, 10.1073/pnas.1514034112, 2016.
DeMott, P. J. et al.: Overview of results from the Fifth International
Workshop on Ice Nucleation – Part 2 (FIN-02): Laboratory intercomparisons of
ice nucleation measurements, Atmos. Chem. Phys., in preparation,
2017.Eidhammer, T., DeMott, P. J., Prenni, A. J., Petters, M. D., Twohy, C. H.,
Rogers, D. C., Stith, J., Heymsfield, A., Wang, Z., Haimov, S., French, J.,
Pratt, K., Prather, K., Murphy, S., Seinfeld, J., Subramanian, R., and
Kreidenweis, S. M.: Ice initiation by aerosol particles: Measured and
predicted ice nuclei concentrations vs. measured ice crystal concentrations
in an orographic wave cloud, J. Atmos. Sci., 67, 2417–2436,
10.1175/2010JAS3266.1, 2010.Emersic, C., Connolly, P. J., Boult, S., Campana, M., and Li, Z.:
Investigating the discrepancy between wet-suspension- and
dry-dispersion-derived ice nucleation efficiency of mineral particles, Atmos.
Chem. Phys., 15, 11311–11326, 10.5194/acp-15-11311-2015,
2015Garcia, E., Hill, T. C. J., Prenni, A. J., DeMott, P. J., Franc, G. D., and
Kreidenweis, S. M.: Biogenic ice nuclei in boundary layer air over two U.S.
High Plains agricultural regions, J. Geophys. Res., 117, D18209,
10.1029/2012JD018343, 2012.Garimella, S., Rothenberg, D. A., Wolf, M. J., David, R. O.,
Kanji, Z. A., Wang, C., Rösch, M., and Cziczo, D. J.: Uncertainty
in counting ice nucleating particles with continuous flow diffusion
chambers, Atmos. Chem. Phys., 17, 10855–10864, 10.5194/acp-17-10855-2017, 2017.Grawe, S., Augustin-Bauditz, S., Hartmann, S., Hellner, L., Pettersson, J. B.
C., Prager, A., Stratmann, F., and Wex, H.: The immersion freezing behavior
of ash particles from wood and brown coal burning, Atmos. Chem. Phys., 16,
13911–13928, 10.5194/acp-16-13911-2016, 2016.Hader, J. D., Wright, T. P., and Petters, M. D.: Contribution of pollen to
atmospheric ice nuclei concentrations, Atmos. Chem. Phys., 14, 5433–5449,
10.5194/acp-14-5433-2014, 2014.Hill, T. C. J., Moffett, B. F., DeMott, P. J., Georgakopoulos, D. G., Stump,
W. L., and Franc, G. D.: Measurement of ice nucleation-active bacteria on
plants and in precipitation by quantitative PCR, Appl. Environ. Microbiol.,
80, 1256–1267, 10.1128/AEM.02967-13, 2014.Hill, T. C. J., DeMott, P. J., Tobo, Y., Fröhlich-Nowoisky, J., Moffett,
B. F., Franc, G. D., and Kreidenweis, S. M.: Sources of organic ice
nucleating particles in soils, Atmos. Chem. Phys., 16, 7195–7211,
10.5194/acp-16-7195-2016, 2016.Hiranuma, N., Augustin-Bauditz, S., Bingemer, H., Budke, C., Curtius, J.,
Danielczok, A., Diehl, K., Dreischmeier, K., Ebert, M., Frank, F., Hoffmann,
N., Kandler, K., Kiselev, A., Koop, T., Leisner, T., Möhler, O., Nillius,
B., Peckhaus, A., Rose, D., Weinbruch, S., Wex, H., Boose, Y., DeMott, P. J.,
Hader, J. D., Hill, T. C. J., Kanji, Z. A., Kulkarni, G., Levin, E. J. T.,
McCluskey, C. S., Murakami, M., Murray, B. J., Niedermeier, D., Petters, M.
D., O'Sullivan, D., Saito, A., Schill, G. P., Tajiri, T., Tolbert, M. A.,
Welti, A., Whale, T. F., Wright, T. P., and Yamashita, K.: A comprehensive
laboratory study on the immersion freezing behavior of illite NX particles: a
comparison of 17 ice nucleation measurement techniques, Atmos. Chem. Phys.,
15, 2489–2518, 10.5194/acp-15-2489-2015, 2015.Huffman, J. A., Prenni, A. J., DeMott, P. J., Pöhlker, C., Mason, R. H.,
Robinson, N. H., Fröhlich-Nowoisky, J., Tobo, Y., Després, V. R.,
Garcia, E., Gochis, D. J., Harris, E., Müller-Germann, I., Ruzene, C.,
Schmer, B., Sinha, B., Day, D. A., Andreae, M. O., Jimenez, J. L., Gallagher,
M., Kreidenweis, S. M., Bertram, A. K., and Pöschl, U.: High
concentrations of biological aerosol particles and ice nuclei during and
after rain, Atmos. Chem. Phys., 13, 6151-6164,
10.5194/acp-13-6151-2013, 2013.Koop, T., Luo, B., Biermann, U. M., Crutzen, P. J., and Peter, T.: Freezing
of HNO3/H2SO4/H2O solutions at stratospheric temperatures:
Nucleation statistics and experiments, J. Phys. Chem. A, 101, 1117–1133,
10.1021/jp9626531, 1997.Martin, A. C., Cornwell, G. C., Atwood, S. A., Moore, K. A., Rothfuss, N. E.,
Taylor, H., DeMott, P. J., Kreidenweis, S. M., Petters, M. D., and Prather,
K. A.: Transport of pollution to a remote coastal site during gap flow from
California's interior: impacts on aerosol composition, clouds, and radiative
balance, Atmos. Chem. Phys., 17, 1491-1509,
10.5194/acp-17-1491-2017, 2017.Mason, R. H., Chou, C., McCluskey, C. S., Levin, E. J. T., Schiller, C. L.,
Hill, T. C. J., Huffman, J. A., DeMott, P. J., and Bertram, A. K.: The
micro-orifice uniform deposit impactor–droplet freezing technique
(MOUDI–DFT) for measuring concentrations of ice nucleating particles as a
function of size: improvements and initial validation, Atmos. Meas. Tech., 8,
2449–2462, 10.5194/amt-8-2449-2015, 2015.Mason, R. H., Si, M., Chou, C., Irish, V. E., Dickie, R., Elizondo, P., Wong,
R., Brintnell, M., Elsasser, M., Lassar, W. M., Pierce, K. M., Leaitch, W.
R., MacDonald, A. M., Platt, A., Toom-Sauntry, D., Sarda-Estève, R.,
Schiller, C. L., Suski, K. J., Hill, T. C. J., Abbatt, J. P. D., Huffman, J.
A., DeMott, P. J., and Bertram, A. K.: Size-resolved measurements of
ice-nucleating particles at six locations in North America and one in Europe,
Atmos. Chem. Phys., 16, 1637–1651, 10.5194/acp-16-1637-2016,
2016.
McCluskey, C. S., DeMott, P. J., Prenni, A. J., Levin, E. J., McMeeking, G.
R., Sullivan, A. P., Hill, T. C., Nakao, S., Carrico, C. M., and Kreidenweis,
S. M.: Characteristics of atmospheric ice nucleating particles associated
with biomass burning in the US: Prescribed burns and wildfires, J. Geophys.
Res., 119, 10458–10470, 2014.
Murray, B. J, O'Sullivan, D., Atkinson, J. D., and Webb, M. E.: Ice
nucleation by particles immersed in supercooled cloud droplets, Chem. Soc.
Rev., 41, 6519–6554, 2012.Ortega, J., Turnipseed, A., Guenther, A. B., Karl, T. G., Day, D. A., Gochis,
D., Huffman, J. A., Prenni, A. J., Levin, E. J. T., Kreidenweis, S. M.,
DeMott, P. J., Tobo, Y., Patton, E. G., Hodzic, A., Cui, Y. Y., Harley, P.
C., Hornbrook, R. S., Apel, E. C., Monson, R. K., Eller, A. S. D., Greenberg,
J. P., Barth, M. C., Campuzano-Jost, P., Palm, B. B., Jimenez, J. L., Aiken,
A. C., Dubey, M. K., Geron, C., Offenberg, J., Ryan, M. G., Fornwalt, P. J.,
Pryor, S. C., Keutsch, F. N., DiGangi, J. P., Chan, A. W. H., Goldstein, A.
H., Wolfe, G. M., Kim, S., Kaser, L., Schnitzhofer, R., Hansel, A., Cantrell,
C. A., Mauldin, R. L., and Smith, J. N.: Overview of the Manitou Experimental
Forest Observatory: site description and selected science results from 2008
to 2013, Atmos. Chem. Phys., 14, 6345–6367,
10.5194/acp-14-6345-2014, 2014.O'Sullivan, D., Murray, B. J., Ross, J. F., and Webb, M. E.: The adsorption
of fungal ice-nucleating proteins on mineral dusts: a terrestrial reservoir
of atmospheric ice-nucleating particles, Atmos. Chem. Phys., 16, 7879–7887,
10.5194/acp-16-7879-2016, 2016.
Petters, M. D. and Wright, T. P.: Revisiting ice nucleation from
precipitation samples, Geophys. Res. Lett., 42, 8758–8766, 2015.Petters, M. D., Parsons, M. T., Prenni, A. J., DeMott, P. J., Kreidenweis, S.
M., Carrico, C. M., Sullivan, A. P., McMeeking, G. R., Levin, E., Wold, C.
E., Collett, J. L. J., and Moosmüller, H.: Ice nuclei emissions from
biomass burning, J. Geophys. Res., 114, D07209, 10.1029/2008JD011532,
2009.Prenni, A. J., Tobo, Y., Garcia, E., DeMott, P. J., McCluskey, C.,
Kreidenweis, S. M., Prenni, J., Huffman, A., Pöschl, U., and Pöhlker,
C.: The impact of rain on ice nuclei populations, Geophys. Res. Lett.,
40, 227–231, 10.1029/2012GL053953, 2013.
Prenni A. J., DeMott, P. J., Rogers, D. C., Kreidenweis, S. M., McFarquhar,
G. M., Zhang, G., and Poellot, M. R.: Ice nuclei characteristics from M-PACE
and their relation to ice formation in clouds, Tellus B, 61, 436–448, 2009.Ralph, F. M., Prather, K. A., Cayan, D., Spackman, J. R., DeMott, P. J.,
Dettinger, M., Fairall, C., Leung, R., Rosenfeld, D., Rutledge, S., Waliser,
D., White, A. B., Cordeira, J., Martin, A., Helly, J., and Intrieri, J.:
CalWater Field Studies Designed to Quantify the Roles of Atmospheric Rivers
and Aerosols in Modulating U.S. West Coast Precipitation in a Changing
Climate, B. Am. Meteorol. Soc., 97, 1209–1228,
10.1175/BAMS-D-14-00043.1, 2016.Schill, G. P., Jathar, S. H., Krodos, J. K., Levin, E. J. T., Galang, A. M.,
Friedman, B., Farmer, D. K., Pierce, J. R., Kreidenweis, S. M., and DeMott,
P. J.: Ice nucleating particle emissions from photo-chemically-aged diesel
and biodiesel exhaust, Geophys. Res. Lett., 43, 5524–5531,
10.1002/2016GL069529, 2016.Schrod, J., Danielczok, A., Weber, D., Ebert, M., Thomson, E. S., and
Bingemer, H. G.: Re-evaluating the Frankfurt isothermal static diffusion
chamber for ice nucleation, Atmos. Meas. Tech., 9, 1313–1324,
10.5194/amt-9-1313-2016, 2016.Schrod, J., Weber, D., Drücke, J., Keleshis, C., Pikridas, M., Ebert, M.,
Cvetkovic, B., Nickovic, S., Marinou, E., Baars, H., Ansmann, A., Vrekoussis,
M., Mihalopoulos, N., Sciare, J., Curtius, J., and Bingemer, H. G.: Ice
nucleating particles over the Eastern Mediterranean measured by unmanned
aircraft systems, Atmos. Chem. Phys., 17, 4817–4835,
10.5194/acp-17-4817-2017, 2017.
Spurny K. R. and Lodge Jr., J. P.: Collection efficiency tables for membrane
filters usedin the sampling and analysis of aerosols and hydrosols, CAR
Technical Note, National Center for Atmospheric Research, Boulder, CO,
NCAR-TN/STR-77, Vol. I, 1972.Stopelli, E., Conen, F., Morris, C. E., Herrmann, E., Bukowiecki, N., and
Alewell, C.: Ice nucleation active particles are efficiently removed by
precipitating clouds, Sci. Rep.-UK, 5, 16433, 10.1038/srep16433, 2015.Tobo, Y.: An improved approach for measuring immersion freezing in large
droplets over a wide temperature range, Sci. Rep.-UK, 6, 32930,
10.1038/srep32930, 2016.Tobo, Y., Prenni, A. J., DeMott, P. J., Huffman, J. A., McCluskey, C. S.,
Tian, G., Pöhlker, C., Pöschl, U., and Kreidenweis, S. M.: Biological
aerosol particles as a key determinant of ice nuclei populations in a forest
ecosystem. J. Geophys. Res.-Atmos., 118, 10100–10110,
10.1002/jgrd.50801, 2013.
Vali, G.: Quantitative evaluation of experimental results on the
heterogeneous freezing nucleation of supercooled liquids, J. Atmos. Sci., 28,
402–409, 1971.Vali, G.: Interpretation of freezing nucleation experiments: singular and
stochastic; sites and surfaces, Atmos. Chem. Phys., 14, 5271–5294,
10.5194/acp-14-5271-2014, 2014.Vali, G., DeMott, P. J., Möhler, O., and Whale, T. F.: Technical Note: A
proposal for ice nucleation terminology, Atmos. Chem. Phys., 15,
10263–10270, 10.5194/acp-15-10263-2015, 2015.Welti, A., Lüönd, F., Kanji, Z. A., Stetzer, O., and Lohmann, U.:
Time dependence of immersion freezing: an experimental study on size selected
kaolinite particles, Atmos. Chem. Phys., 12, 9893–9907,
10.5194/acp-12-9893-2012, 2012.Wex, H., Augustin-Bauditz, S., Boose, Y., Budke, C., Curtius, J., Diehl, K.,
Dreyer, A., Frank, F., Hartmann, S., Hiranuma, N., Jantsch, E., Kanji, Z. A.,
Kiselev, A., Koop, T., Möhler, O., Niedermeier, D., Nillius, B.,
Rösch, M., Rose, D., Schmidt, C., Steinke, I., and Stratmann, F.:
Intercomparing different devices for the investigation of ice nucleating
particles using Snomax® as test substance,
Atmos. Chem. Phys., 15, 1463–1485, 10.5194/acp-15-1463-2015,
2015.Willeke, K., Lin, X., and Grinshpun, S. A.: Improved Aerosol Collection by
Combined Impaction and Centrifugal Motion, Aerosol Sci. Techno., 28,
439–456, 10.1080/02786829808965536, 1998.Wright, T. P. and Petters, M. D.: The role of time in heterogeneous freezing
nucleation, J. Geophys. Res.-Atmos., 118, 3731–3743, 10.1002/jgrd.50365,
2013.Wright, T. P., Petters, M. D., Hader, J. D., Morton, T., and Holder, A. L.:
Minimal cooling rate dependence of ice nuclei activity in the immersion mode,
J. Geophys. Res.-Atmos., 118, 10535–10
543, 10.1002/jgrd.50810, 2013.