As one part of the NETCARE project (Network on Climate and Aerosols: Addressing Key Uncertainties in Remote Canadian Environments), aircraft-based measurements were deployed from Resolute Bay, Nunavut (Canada), during 4–21 July 2014. In this study, we focus on measurements made between 4 and 12 July 2014. The satellite image from 4 July 2014 shown in Fig.  presents sea ice and open water conditions around Resolute Bay, which can be regarded as typical during 4–12 July. Six research flights (around 20 flight hours) were performed during this time. Flight tracks covered altitudes from 50 to 3000 m above continental as well as marine (partly covered with sea ice) regions (Fig. ). Three flights aimed to sample above two polynyas north of Resolute Bay. Notably, the sea ice south-east of Resolute Bay and close to the ice edge in Lancaster Sound was largely covered with melt ponds.

The instrument platform was the research aircraft Polar 6, a modified Basler BT-67 maintained by Kenn Borek and operated by the Alfred Wegener Institute for Polar and Marine Research . The aircraft was equipped with instruments to measure meteorological state parameters and several trace gases as well as aerosol particle number, size and chemical composition. In general, aerosol instruments were connected to a forward-facing near-isokinetic stainless steel inlet, which was followed by a 1-inch stainless steel manifold inside the cabin. All instruments were connected to the common inlet line system with 1/4-inch stainless steel tubing. Reactive trace gases were measured via a second PTFA inlet line. Further detailed information on the inlet and sampling strategy can be found in , , and .

Number concentrations of particles greater than 5 nm in diameter (Nd>5nm) were measured with a TSI 3787 water-based ultrafine condensation particle counter (UCPC). Particle size distributions of particles greater than 250 nm (Nd>250nm) were measured using an optical particle counter from GRIMM (model 1.129 Sky-OPC). Measurements of carbon monoxide (CO) were conducted with an Aerolaser ultra-fast CO monitor (model AL 5002). Sub-micron bulk aerosol composition was measured with an Aerodyne high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS). Operation of the HR-ToF-AMS aboard Polar 6 during NETCARE is described in . State parameters and meteorological measurements were made using an AIMMS-20 from Aventech Research Inc. Detailed information on measurement principles and instrument calibrations are given in and .

In order to provide information about the chemical composition of single aerosol particles, the ALABAMA (Aircraft-based Laser Ablation Aerosol Mass Spectrometer; ) was deployed on the Polar 6 during NETCARE 2014. The basic measurement principle of the ALABAMA is as follows: first, the particles enter the system through a constant-pressure inlet. While ambient pressure changes, this device (custom-made at the Max Planck Institute for Chemistry) maintains a constant pressure in the following aerodynamic lens by varying the volume flow rate into the instrument. A flexible orifice is either squeezed or relaxed, depending on atmospheric pressure, by bottom and top plates that are connected to a rotor. After passing through the inlet, particles are focused into a narrow beam with the help of a Liu-type aerodynamic lens . The focused particles are detected by two light scattering signals (using 405 nm laser diodes and photo-multipliers) allowing the determination of the size-dependent particle velocity. By comparing these values with the velocity of manufactured monodisperse polystyrene latex particles in five sizes ranging from 190 to 800 nm, we can derive the particle vacuum aerodynamic diameter (dva). Next, the particles enter the ablation and ionization region in the high-vacuum system. The particles are ablated and ionized by a single triggered laser shot (266 nm, frequency-quadrupled Nd:YAG laser). In the final step, cations and anions produced by laser ablation are guided into a bipolar Z-ToF (Z-shaped Time of Flight) mass spectrometer, which provides bipolar mass spectra of individual particles. Due to limitations of the aerodynamic lens transmission efficiency and the lower detection limit of the photo-multipliers, the ALABAMA covers a particle size range from approximately 200 to 1000 nm.

FLEXPART (FLEXible PARTicle dispersion model (here: version 10.0)) is a Lagrangian particle dispersion model e.g.,. For our analysis, we used operational analysis data from the European Centre for Medium-Range Weather Forecast (ECMWF) with 0.125∘ spatial and 3 h time resolution. FLEXPART was operated in backward mode to provide potential emission sensitivity (PES) maps, which are the response functions to tracer releases from a receptor location. The value of the PES function is related to the particles' residence time in the output grid cell (for more details see Sect. 5 in ). We used such PES maps together with sea ice and open ocean coverage derived from the satellite image in Fig.  to determine the total residence time of the measured air mass above open water regions 3 days prior to sampling at altitudes up to 340 m. The model output frequency was set up to 1 h and 0.125∘ spatial resolution.

In total, 7412 particles were chemically analyzed (mass spectra produced) by the ALABAMA during the study; 94 % of these spectra include size information; 80 % of these spectra have dual polarity. Considering the 20 % single-polarity spectra, potential reasons for the lack of negative ions are discussed in the Supplement, Sect. 1. Briefly, it is likely that single-polarity spectra are produced in high relative humidity (RH) environments , in particular marine environments .

The CRISP software package (Concise Retrieval of Information from Single Particles) was used to perform m/z (mass to ion charge ratio) calibration of particle mass spectra and peak area integration as well as to classify particle mass spectra using ion markers for different species . The marker method requires knowledge of certain ion markers belonging to a certain substance as well as knowledge of a certain marker threshold (ion peak area threshold). The typical fragmentation pattern of a substance due to laser ablation is crucial for defining the distinct ion markers. Fragmentation depends on laser wavelength and energy. Ion markers of many species are already well known from laboratory measurements with the ALABAMA and additionally from the literature of other single particle mass spectrometers (SPMSs) using the same ablation laser wavelength. Table  lists ion markers of substances used in this study to identify the external and internal mixing states of particles. The identification of ion markers m/z +59 and +58 for TMA by was confirmed by additional laboratory measurements with the ALABAMA (Supplement Sect. 2). To decide whether an ion signal is present, we used ion peak area thresholds of 10 and 25 mV for positive and negative mass spectra, respectively. Both thresholds are chosen as a conservative measure on the basis of signal intensities of the non-occupied m/z values. Supplement Sect. 3 presents a detailed explanation of ion peak area threshold determination.

The measurement period during 4–12 July 2014 was characterized by generally clear skies, calm wind speeds (Fig. ) and occasional scattered to broken stratocumulus clouds due to prevailing high-pressure influence in the Resolute Bay region. Based on low CO mixing ratios, low aerosol number concentrations (Fig. ) and backward trajectory analysis, air masses measured in this period experienced a weak mid-latitudinal influence and were mainly affected by local emission sources (also denoted as “Arctic air mass period” in ). As shown in Fig. , our measurements took place largely over remote areas, which are dominated by Arctic vegetation, open water regions (e.g., polynyas, Lancaster Sound) and sea ice coverage. Furthermore, seabird colonies were located close to the ice edge in Lancaster Sound and are likely a source of ammonia . Anthropogenic emissions might have affected our measurements, but are mainly related to the sparse Arctic settlements and can be ruled out by comparison with other tracers (e.g., CO). We can therefore expect that our observations from 4–12 July 2014 were mainly influenced by Arctic marine and terrestrial emissions.

As is evident from vertical profiles of equivalent potential temperature (Θe) (Fig. ), the mean upper boundary layer (BL) height for this measurement period was at around 340 ± 100 m. The vertical resolution of the profile in Fig.  (100 m) justifies the range of the mean BL height. The mean BL height within its range can be confirmed by results from an extensive study on BL height, mixing and stability during the NETCARE 2014 campaign . The capping temperature inversion above 390 m, inferred from values of Θe, represents a transport barrier for air masses between the BL and the free troposphere (FT). The BL, compared to the FT, was characterized by lower wind speeds, higher RH and enhanced Nd>5nm in contrast to Nd>250nm, indicating an enhanced number of ultrafine particles due to nucleation in the Arctic BL. A detailed discussion of this topic is given in .

Applying the marker method (Sect. ), we classified 6676 particle mass spectra (90 % of the mass spectra analyzed by the ALABAMA (Sect. )) into five distinct particle types: TMA-, Na / Cl-, EC-, levoglucosan- and K-containing particles. TMA-, levoglucosan- and K-containing particles, with relative fractions of 23, 18 and 46 %, respectively, appear to be the most prominent particle types. Other alkylamines (other than TMA) and amino acids could not be identified (Supplement Sect. 4). Furthermore, only 2 and 1 % of all analyzed particles are assigned as EC- and Na / Cl-containing particles, respectively. To obtain 100 % as the total particle number, every spectrum is classified into one distinct particle type in the order presented in Table . The mean spectra in Fig.  combined with the additional ion signals listed in Table  provide an overview of the average chemical composition of each particle type; 28 and 9 % of TMA- and K-containing particle spectra lack negative ions, respectively. Potential reasons for the lack of negative ions are discussed in the Supplement, Sect. 1. The mean spectrum of the remaining 736 particles (10 % of mass spectra analyzed by the ALABAMA), which could not be classified into one of the five particle groups outlined above, is shown in Fig. S7 in the Supplement. For further analysis we summarize these remaining particles in “others”.

In order to describe the unique characteristics of TMA-containing particles compared to other particle groups, Figs.  and depict the size and vertical distribution of each particle type, respectively. Both figures show the fractional abundance of each particle type per size and altitude bin, respectively. We show relative numbers of particles in order to eliminate the size-dependent transmission and detection efficiency of the ALABAMA (Fig. ) and the dependence of the number of detected particles on sampling time at different altitudes (Fig. ). The following use of the word fraction always refers to the number fraction measured by the ALABAMA.

The internal mixing state of TMA-containing particles was further classified by applying the marker method introduced in Sect.  and Table  for compounds that are apparent in the mean spectrum (Fig. a and Table ), such as potassium (K), ammonium (NH4), MSA and sulfate (S). The diagram in Fig.  illustrates the classification algorithm as follows: an upper branch always refers to a positive response (“yes”) for whether different ion markers are present in spectra or not; a lower branch shows the opposite answer (“no”). Besides the substances that already appeared in the mean spectrum of TMA-containing particles, here TMA-containing spectra are also viewed based on the concurrent existence of levoglucosan, Na / Cl and EC. We did not consider in detail the concurrent existence of carbon cluster ions (m/z +12, +24,…), different hydrocarbons (m/z +27 and +37) and oxidized organics (m/z +43) since 90 % of all TMA-containing particles contain at least one of these ion signals. The classification of the TMA-containing particle type is further based on an initial differentiation between dual- and single-polarity mass spectra. As can be seen in Fig. , 28 % of TMA-containing particle spectra lack negative ions. Consequently, we cannot state whether species producing anions (such as MSA and sulfate) were present in these particles. Potential reasons for the lack of negative ions are discussed in the Supplement Sect. 1. Particle sub-group notation is based on the existence or non-existence of different species in TMA-containing particles. For reasons of clarity, particle types with less than 7 % fractional abundance (corresponding to a total number of less than 118 particles) are not explicitly considered in this analysis, but are summarized as “others”. Following the categorization in Fig. , five groups of different internal mixing states arise: “K,NH4,MSA,S-”, “K,NH4,S-”, “K-”, “Non-K,NH4-” and “levoglucosan-containing” particles. These five TMA particle sub-types will be divided into those containing biomass burning tracers (such as levoglucosan and potassium) and those not containing these tracers.

As can be seen in Fig. , a large fraction of TMA-containing particles (74 %) are additionally composed of biomass burning tracers such as potassium (67 %) and levoglucosan (7 %). According to , this internal mixture can be explained by potassium-containing particles acting as seeds for the condensation of organic material. Thus, the measured particulate TMA can be considered a secondary component that condensed on pre-existing primary particles. It is also conceivable that TMA particles containing potassium and levoglucosan are a result of biomass burning emissions . The size distribution of the TMA particles containing levoglucosan is shifted towards larger diameters compared to other TMA particle sub-types (Fig. ). Moreover, Fig.  demonstrates that TMA particle sub-types including potassium and levoglucosan were more abundant above the BL, in contrast to “Non-K,NH4-containing” TMA particles. Comparison between CO mixing ratios and TMA sub-types abundance (Fig. ) shows larger fractions of “K,NH4,S-containing” and “levoglucosan-containing” TMA particle sub-types in higher CO environments compared to “Non-K,NH4-containing” TMA particles. Taken together, these results suggest that TMA particles containing levoglucosan and potassium likely originated from remote biomass burning emission sources and were transported to our measurement site.

Another large fraction (25 %, Fig. ) of particulate TMA is neither internally mixed with potassium nor with any other tracer of biomass burning. This result suggests that these TMA-containing particles resulted from SOA formation. This is consistent with results from particle size distributions of TMA sub-types in Fig.  illustrating that the fractional abundance of “Non-K,NH4-containing” TMA particles is highest between 280 and 380 nm compared to other sub-types containing levoglucosan and/or potassium. In particular, positive ion mass spectra of the sub-type “Non-K,NH4-containing” (12 % single polarity (yellow box in Fig. ) and 6 % dual polarity (not colored in Fig. )) show ion signals only for carbon cluster ions and fragments of hydrocarbons (Fig. a, b). Due to a suppression of anion signals, likely in high RH environments (Supplement Sect. 1), we cannot state whether sulfate or MSA was present in these particles. However, the dual-polarity mean spectrum of the 6 % TMA-containing particles not including potassium and ammonium (Fig. b, not colored in Fig. ) indicates the concurrent presence of sulfate or MSA. From the absence of ammonium in these TMA particles containing sulfate or MSA, we can further conclude that aminium salts were present. This result demonstrates that amines, in addition to ammonia, may take part in the neutralization of acidic aerosol. This is of particular interest considering the reduced sources of ammonia in the Arctic and the ocean as a net sink of NH3 in the summertime Canadian Arctic . Furthermore, Fig.  indicates a positive correlation between MSA mass concentrations measured with HR-ToF-AMS and the fraction of “Non-K,NH4-containing” TMA particles. Given that MSA can be used as an indicator of marine influence on sub-micron aerosol, we can conclude that the existence of an inner-Arctic marine-biogenic source of TMA is likely. Moreover, “Non-K,NH4-containing” TMA particles are most abundant at the lowest altitudes (Fig. ) and are coincident with the presence of less aged particulate organic aerosol (Fig. ). Taken together, the characteristics of the “Non-K,NH4-containing” TMA particle sub-type suggest that gaseous TMA emissions from inner-Arctic sources (likely marine-biogenic) act as precursors for the formation of SOA within the summertime Arctic BL.

This section will further explore potential emission sources of TMA in the Arctic BL. Thus, the following analysis was restricted to measurements below 340 m (mean upper BL height, Sect. ). Figure  shows the temporal distribution of non-TMA-containing particles (such as Na / Cl-, EC-, levoglucosan- and K-containing) and TMA-containing sub-types. Figure  depicts the spatially resolved fraction of TMA-containing particles below 340 m (left panel) as well as the measured wind direction (right panel) for measurements on 4, 7 and 8 July. We further used 3-day FLEXPART backward simulations (Sect. ) for air mass history analysis of the three measurement legs (Fig. ) to understand the source regions of TMA-containing particles.

Potential emission sensitivity (PES) maps combined with sea ice coverage (Fig. ) show that air masses measured on 4 and 7 July spent less than 1 h and around 7 h, respectively, in the previous 3 days in regions of open water (polynyas north of Resolute Bay and Nares Strait). On both days the air was mainly advected above sea ice and snow covered regions north of Resolute Bay (Fig.  compared with Fig. ). The prevailing wind direction on 4 and 7 July along the flight tracks (Fig. ) is from the north and east and therefore consistent with FLEXPART backward simulations (Fig. ). From measurements on 4 and 7 July it is not possible to attribute TMA emissions to marine-biogenic or anthropogenic sources (e.g., vehicle exhaust, residential heating and waste incineration emissions in Resolute Bay). A more detailed air mass history analysis was carried out on observations from 8 July.

The case of 8 July provides further evidence for a marine-biogenic influence on TMA-containing particles through secondary processes. The prevailing wind direction along the presented flight leg is from the east (Fig. ), with low wind speeds up to a maximum of 7 m s-1 (Fig. S9 in the Supplement). The fraction of TMA-containing particles decreases with a shift to a more southerly wind direction (yellow to green colors, Fig. ). The highest fractional abundance of particulate TMA was measured close to the ice edge (Fig. ) at low wind speeds close to 0 m s-1 (Fig. S9 in the Supplement). Thus, the ice edge in the western section of Lancaster Sound where the highest surface phytoplankton production rate and chlorophyll a concentration were measured (M. Gosselin, personal communication, 2017) and large bird colonies at Prince Leopold Island (Fig. ) likely contribute to TMA emissions in the area. Consistent with these observations, previous aerosol chemical composition measurements on Bird Island in the South Atlantic (> 50∘ S) have reported the presence of amines and amino acids emitted from local fauna, including seabirds, penguins and fur seals . Further, air mass history predicted by FLEXPART 3-day backward simulations (Fig. ) illustrates that these air masses were advected at low levels above open water regions in Lancaster Sound, Baffin Bay and Nares Strait (compare with Fig. ). Air masses measured during this flight leg on 8 July resided for more than 17 h during the 3 days prior to sampling above regions of open water. Further, anthropogenic influences on amine emissions from nearby Resolute Bay are likely negligible since CO concentrations are very low. Another important finding is that primary sea spray particles (Na / Cl-containing) and TMA-containing particles measured on 8 July are externally mixed (Fig. ), although both substances seem to be released from the ocean. This analysis solidifies the earlier hypothesis (Sect. ) that particulate TMA presents secondary aerosol . The higher abundance of the TMA-containing particle sub-type “Non-K,NH4-containing” on 8 July (Fig. ), compared to other days, further supports the hypothesis of SOA formation. It is further relevant to discuss that on 8 July from 15:50 until 17:20 UTC (respective flight leg in Fig. ) we flew low over sea ice in the vicinity of dissipating low-level clouds. These clouds had formed above the open water regions east of our flight leg . We can therefore assume that cloud processing likely contributed to enhanced gas-to-particle partitioning of TMA as earlier reported in . In addition, high organic-to-sulfate and MSA-to-sulfate ratios measured with the HR-ToF-AMS during this flight leg (see Sect. 4.3 in ) indicate that particle growth was driven by ocean-derived precursor gases (dimethylsulfide and organic species). Taken together, results from 8 July demonstrate secondary organic aerosol formation from marine-biogenic sources of gas-phase precursors, including TMA.