Discrimination between individual dust and bioparticles using aerosol time-of-flight mass spectrometry

Abstract

Ice nucleating particles (INPs) impact cloud properties and precipitation processes through their ability to trigger cloud glaciation. Dust and bioparticles are two important sources of INPs that have markedly different atmospheric loadings and ice nucleating efficiencies. In-situ identification of the sources of INPs in clouds has been accomplished using single particle mass spectrometry (SPMS). However, external mixtures of dust and bioparticles present a unique challenge as they have overlapping mass spectral ion signatures, complicating their unambiguous identification. This study presents a detailed discussion of dust and bioparticle SPMS signatures, uniting data from a broad array of studies. As emphasized, the ion signals from both dust and bioparticles are highly sensitive to ionization conditions. To understand the observed variations, we characterize the mass spectral dependence of distinct dust and bioparticle samples using total positive ion intensity (TPII) as an indicator of the laser pulse energy each particle encountered. Through this analysis, a broad range of characteristic biogenic low intensity ion peaks that may be useful to distinguish bioparticles from dust became apparent and are highlighted. Insights informed by this analysis were utilized to identify bioparticles in ambient SPMS data. Ambient particles exhibiting both dust and characteristic biogenic spectral fingerprints were excluded from the bioparticle classification. Although bioparticles only made up 0.2% of all sampled particles, their abundance was moderately correlated with INP concentrations measured at −15 °C.

Copyright © 2022 American Association for Aerosol Research

EDITOR:

• Nicole Riemer

1. Introduction

Dust and primary bioparticles have been shown to be the most effective ice nucleating particles (INPs) in field and lab studies. INPs trigger the formation of atmospheric ice at temperatures higher than the homogeneous freezing point of water at ∼−38 °C (DeMott et al. 2010). Because only 1 in 10⁶ of all particles in the atmosphere will induce freezing at −20 °C (Rogers et al. 1998), INPs exert an outsized influence upon clouds and precipitation by inducing cloud glaciation (DeMott et al. 2010). Dust has the highest atmospheric loading by mass of any aerosol source (Andreae and Rosenfeld 2008; Engelstaedter, Tegen, and Washington 2006; Zender 2004) and is generally estimated to represent the largest fraction of INP by number at temperatures lower than −20 °C (Kanji et al. 2017). While dust particles can activate at higher temperatures depending on their mineralogy and organic content (Atkinson et al. 2013; Hill et al. 2016; Tobo et al. 2014), bioparticles are generally assumed to be more efficient nucleators at temperatures above −15 °C and may be particularly important due to their influence upon secondary ice processes that occur in these temperature ranges.

Due to a limited number of measurements isolating the nuclei present in cloud droplets, it has yet to be directly demonstrated that bioparticle INPs play a role in cloud ice-processes (Fröhlich-Nowoisky et al. 2016). Bioparticles include whole organisms (e.g., bacteria), viruses, spores, pollen, and cell fragments/excretions with particle diameters ranging from nanometers to hundreds of microns (Fröhlich-Nowoisky et al. 2016). While bioparticles and their fragments can be internally mixed with other particles such as dust (e.g., Yamaguchi et al. 2012), we are primarily interested in cellular particles or particles comprised of undecomposed cellular fragments. The breadth of potential sources coupled with their rarity as INPs increase the difficulty of detection and unambiguous source identification (Kanji et al. 2017).

Many offline techniques have been utilized to identify bioparticles, including cultures, immunological assays, microscopy, and mass spectrometry (Després et al. 2012 and references therein). These techniques require lengthy sampling periods and specialized handling and preservation techniques which can add uncertainty and bias to the measurements (Després et al. 2012). There is a limited understanding of the concentrations of bioparticles at altitudes where clouds exist (Fröhlich-Nowoisky et al. 2016; Twohy et al. 2016), which is due to the fact that techniques that require long sampling times are not a good match for aircraft studies.

Single particle mass spectrometers (SPMS), such as aerosol time-of-flight mass spectrometry (ATOFMS), are uniquely suited for directly identifying INP sources in clouds because they are aircraft-compatible and can distinguish between a wide variety of particle sources. In previous studies, the presence of ion markers ²⁶CN⁻ and ⁴²CNO⁻, have been used to identify the presence of organic nitrogen species, and ⁷⁹PO₃⁻, from phosphates, in single particle dust mass spectra and to classify a distinct internally mixed “biological-dust” particle type, where the biological species have been postulated to be related to the presence of bacteria, fungal spores, or plant detritus (Creamean et al. 2013; Pratt et al., 2009). These ion markers are ubiquitous in cellular (defined as whole cells) single particle mass spectra (e.g., Cahill et al. 2015; Czerwieniec et al. 2005; Fergenson et al. 2004; Steele et al. 2003; Zawadowicz et al. 2017). However, ²⁶CN⁻, ⁴²CNO⁻, and ⁷⁹PO₃⁻ signals have also been frequently reported in dust mass spectra (Creamean et al. 2014; Marsden et al. 2019; Schmidt et al. 2017; Silva, Carlin, and Prather 2000; Zawadowicz et al. 2017), including mass spectra of mineral dust standards, which contain small amounts of organic material, and no cellular material. As summarized by Zawadowicz et al. (2017), the presence of ²⁶CN⁻, ⁴²CNO⁻, or ⁷⁹PO₃⁻ ion markers alone cannot be used as indicators of the presence of biological material.

To better address this ambiguity, Zawadowicz et al. (2017) proposed a new method for identifying single particle mass spectra containing “biological” material. Standards of biological and inorganic particles were analyzed using Particle Analysis by Laser Mass Spectrometry (PALMS), and the data simplified to a two-dimensional space of the ratios of ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ ion signals. A supervised machine learning algorithm was used to distinguish between two particle classes with 97% accuracy within this space. The applicability of this approach to other SPMS datasets, regardless of the specific algorithm, training set, or parameters specified, largely depends on the natural division between biological and inorganic spectra in this two-dimensional space. However, factors in addition to composition can influence individual mass spectra, and may affect partitioning within this 2-D space. For instance, individual aerosol particles analyzed using SPMS have been shown to exhibit a high degree of variability in relative ion signals, that depend in varying degrees on factors such as particle size (Reinard and Johnston 2008), degree of particle desorption (Carson, Johnston, and Wexler 1997), the wavelength of the desorption/ionization laser (Thomson, Middlebrook, and Murphy 1997), and the photon flux experienced by the aerosol within the laser beam (Wenzel and Prather 2004). Mass spectral signatures generated by aerosolized bacterial cells have been shown to be particularly sensitive to the photon flux experienced by each particle (Steele et al. 2003, 2005).

Given the variability in dust and bioparticle mass spectra both within single particle data sets and across SPMS designs, it is important to understand the “landscape” that these mass spectral signatures can occupy in order to more accurately distinguish between them. Methods developed to discriminate between the two types utilizing a few select ion markers with specific ratios or thresholds may be useful for a specific instrument and operating parameters, however such methods cannot easily be extended to all single particle datasets.

In this study, the dependence of dust and bioparticle mass spectral fingerprints on laser energy density is explored by examining the mass spectra as a function of total positive ion intensity (TPII). Using this approach, we identify ion markers that can be used to discriminate between the two particle types but that have been rarely mentioned in previous studies of atmospheric aerosols. These ion markers are used to identify single cells in an ambient dataset at a coastal site in California. The temporal evolution of these bioparticles is compared against other real-time bioparticle measurements, as well as measured concentrations of INPs active at −15 °C (INP_-15) in order to quantify the relative contributions of individual cells to INP populations.

2. Materials and methods

2.1. Aerosol standards

Dust and bioparticle standards in this study were analyzed with an aircraft-aerosol time-of-flight mass spectrometer (A-ATOFMS; Pratt et al. 2009b) at the Aerosol Interaction and Dynamics in the Atmosphere (AIDA) facility at the Karlsruhe Institute of Technology (KIT) during the Fifth International Ice Nucleation Workshop (FIN-1) in November 2014 (DeMott et al. 2018). During this portion of the workshop, a total of 10 single particle mass spectrometers from around the world sampled a common set of aerosol standards and participated in cloud expansion experiments. Additional aerosolized unicellular algal and cyanobacteria standards were analyzed at the University of California, San Diego (UC San Diego) using the A-ATOFMS outfitted with an aerodynamic lens optimized for the transmission of large particles (John F. Cahill et al. 2014). Finally, Arizona Test Dust (ATD) was analyzed with a nozzle-ATOFMS (Gard et al. 1997) at UC San Diego. A summary of the samples is provided in Table 1.

Table 1. Summary of dust and biological samples.

Sample	Name	Description	Study	ATOFMS
Ethiopian soil dust	EVSD	Soil from vegetated area	FIN-1	A-ATOFMS
Argentinian soil dust	ASD	Top soil from arable area	FIN-1	A-ATOFMS
Moroccan soil dust	MSD	Saharan desert dust	FIN-1	A-ATOFMS
Bachli soil dust	BSD	Glacial sediment, no vegetation	FIN-1	A-ATOFMS
K-Feldspar	KF	K-rich feldspar sample	FIN-1	A-ATOFMS
Illite NX	INX	Mineral dust standard	FIN-1	A-ATOFMS
Snomax	SX	Irradiated, desiccated, ground Ps	FIN-1	A-ATOFMS
P. syringae	Ps	Gram-negative bacteria	FIN-1	A-ATOFMS
P. fluorescens	Pf	Gram-negative bacteria	FIN-1	A-ATOFMS
C. reinhardtii	Cr	Freshwater microalgae	UC SAN DIEGO	LP-ATOFMS
S. dimorphus	Sd	Freshwater microalgae	UC SAN DIEGO	LP-ATOFMS
D. tertiolecta	Dt	Marine microalgae	UC SAN DIEGO	LP-ATOFMS
C. vulgaris	Cv	Freshwater microalgae	UC SAN DIEGO	LP-ATOFMS
P. purpera	Pp	Marine microalgae	UC SAN DIEGO	LP-ATOFMS
S. elongates	Se	Freshwater cyanobacteria	UC SAN DIEGO	LP-ATOFMS
Arizona test dust	ATD	Mineral dust standard	UC SAN DIEGO	Nozzle-ATOFMS

2.1.1. FIN-1 sampling details

Experiments were conducted in the Aerosol Preparation Chamber (APC) at the AIDA facility. The APC is a 4 m³ stainless-steel cylindrical vessel of controlled pressure used for characterizing aerosols. Samples of pure dust and bacteria were introduced into the APC during separate experiments. A more complete description of these samples is given in Table 1. A rotating brush generator was used to disperse the mineral dust samples into a stream of particle free synthetic air. Larger dust particles were removed using an inertial impactor. For Snomax, a suspension of 1 g in 200 mL of ultrapure water (18.2 MΩ) was created. Bacterial samples were cultured on agar, and suspensions created at concentrations of 10⁹ cell mL⁻¹ of ultrapure water. The Snomax and bacterial suspensions were freshly prepared before dispersion into the APC, and were aerosolized using a two-component jet device, which uses particle free synthetic air to disperse a liquid flow (Wex et al. 2015).

2.1.2. UC San Diego study details

Details of sample preparation and aerosolization for the algal and cyanobacteria standards analyzed at UC San Diego have been previously published (Cahill et al. 2014, 2015). Cell solutions were nebulized by a C-type concentric nebulizer (Precision Glassblowing, CO) using dry nitrogen. The cell solution was stirred to prevent cell agglomeration and settling. The aerosol stream was dried by a heated flow tube and diffusion dryer and the dried cells were then introduced into the A-ATOFMS. The Arizona Test Dust sample was placed in a sealed flask and aerosolized using a flow of filtered air and then sampled into a nozzle-ATOFMS.

2.2. ATOFMS analysis

Single particle mass spectra were acquired using an aerosol time-of-flight mass spectrometer (nozzle-ATOFMS; Gard et al. 1997) and an aircraft aerosol time-of-flight mass spectrometer (A-ATOFMS; Pratt et al. 2009b). While these instruments are conceptually the same, they employ different aerosol focusing methods and use different mass spectrometers. The ATOFMS was configured with a nozzle-skimmer configuration instead of an aerodynamic lens, and thus sampled larger particles than the A-ATOFMS, which uses the aerodynamic lens described in Pratt et al. (2009b). For the algal and cyanobacteria standards analyzed at UC San Diego, a specialized lens that optimizes the transmission of particles between 4 and 10 µm (Cahill et al. 2014) was used with the A-ATOFMS and this configuration is hereafter referred to as LP-ATOFMS. We note that the inlet is the only difference between these configurations. Also, the ATOFMS uses a linear TOF while the A-ATOFMS employs a Z-TOF. While mass spectra cannot be directly compared between these two instruments, mass spectral features and general trends with regard to TPII should still be similar.

As these instruments have been described previously, we provide only a short overview here. Particles are introduced into the instrument through either a converging nozzle and series of skimmers (ATOFMS), or an aerodynamic lens (A-ATOFMS) in order to collimate smaller particles into a tight beam. Particle velocity, calculated from the intersection of the particle beam with two continuous wave lasers (532 nm; Nd:YAG), is used with external calibration to compute particle vacuum aerodynamic diameter (D_va). A pulsed 266 nm laser (Nd:YAG; 8 ns pulse width, 1.0–1.3 mJ pulse⁻¹, focused to a 700 µm diameter spot size, peak power ∼10⁸ W cm⁻²) desorbs and ionizes species from individual particles, and positive and negative ions are analyzed using a time-of-flight mass spectrometer. Data were imported into Matlab (The MathWorks, Inc.) and analyzed using the single particle mass spectrometry analysis and visualization software toolkit FATES (Sultana et al. 2017).

Particles analyzed by ATOFMS during the CalWater-2015 field campaign were clustered using a neural net algorithm (Rebotier and Prather 2007; Song et al. 1999) as detailed in Martin et al. (2017). Dust particles were identified based upon their characteristic mass spectra (Schmidt et al. 2017; Silva and Prather 2000). Hereafter Dust-CW refers to particles classified as dust during CalWater-2015. A total of 65,534 of the 1,220,818 particles measured by the ATOFMS were classified as Dust-CW (5.4%). A separate methodology was used to identify bioparticles and is described in Sec. 3.4 and the SI.

2.3. CalWater-2015 field campaign

The CalWater-2015 field campaign was conducted at the Bodega Marine Laboratory in Bodega Bay, California (38.3°N, 123.1°W) from January 15 to March 8, 2015. Details of this study have been published elsewhere (Martin et al. 2017), so only a short overview is provided here. Measurements of aerosol size, number, and composition were made in a UC San Diego mobile laboratory. Section 2.2 contains details on ATOFMS sampling and analysis. Bioparticle concentrations were measured with a wideband integrated bioaerosol sensor (WIBS-4; Droplet Measurement Technologies, Inc.; hereafter WIBS) in a mobile lab operated by Colorado State University (CSU) and North Carolina State University (NCSU) (Sec. 2.3.1). Ambient aerosol samples were collected and analyzed for their INP concentrations (Sec. 2.3.2).

2.3.1. Online bioaerosol concentrations

The wideband integrated bioaerosol sensor (WIBS) measures the size, asymmetry factor, and fluorescence signal of particles with optical diameters between 0.5–20 µm (Hirst et al. 2004). Fluorescence is measured in three different channels: FL1, excitation 280 nm, emission 310–400 nm; FL2, excitation 280 nm, emission 420–650 nm; FL3, excitation 370 nm, emission 420–650 nm. Due to limited sensitivity to signals produced by small particles (Gabey et al. 2010, 2011), data analysis was limited to fluorescent particles above 0.8 µm.

Background fluorescence signal was calculated following Savage et al. (2017). Force-triggered sampling of filtered air was conducted for 5 min and the background signal, E_background,i, calculated using the following equation (1) $$E_{\mathit{background},i}=\acute{E_i}+9{\sigma }_i$$(1) where $\acute{E_i}$ is the average signal in channel i, for a force-triggered sample and ${\sigma }_i$ is the standard deviation of the force triggered sample signal. E_background,i was subtracted from each particle for each channel to yield the final fluorescence signal. Particles were grouped into types based upon their fluorescence signal in each channel using criteria defined in previous studies (Gabey et al. 2010; Perring et al. 2014; Wright et al. 2014) and summarized in Table S1.

2.3.2. Offline sample collection for in analysis

Ambient aerosols were sampled for offline INP measurements using impinger samplers (Willeke, Lin, and Grinshpun 1998, three-piece 5 mL impinger; SKC Inc. BioSampler®). Here, we provide a short overview. A full description of these methods can be found in the online supplementary information (SI). Ambient aerosols were impinged into 20 mL ultrapure water (HPLC Grade Submicron Filtered Water, Fisher Chemical Cat. No. W5-4) at a height of 1 m, and a flow rate of 12.5 L min⁻¹ for ∼4 h. Field blanks, containing only the ultrapure water used to prepare the impinger, were generated for each sampling period. The ambient and field blank samples were immediately frozen after collection and thawed prior to analysis. INP concentrations of samples were measured using the NCSU Cold Stage (Hader, Wright, and Petters 2014; Wright and Petters 2013) with modifications described in (Yadav et al. 2019). Samples were divided into ∼200 1 µL volume drops onto hydrophobic glass cover slides (Siliconized Glass Square Cover Slides, Hampton Research Cat. No. HR3-215). The cover slides were placed on top of the cold stage, and droplets were cooled at a rate of 2 K min⁻¹. Freezing events were detected using an optical change in the droplets’ brightness and recorded every 5 s using a digital camera. Freeze event detection was aided using automated event detection combined with human validation; the source code is freely available (Wright and Petters, 2019). Each sample was tested three times, INP concentrations were binned into 1 K temperature intervals and the mean concentration and 95% confidence interval reported. Points where ambient concentrations are less than the field blank INP concentrations are removed from the dataset.

3. Results and discussion

To facilitate the discrimination between dust and bioparticles, this study provides a detailed exploration of the single particle mass spectral differences. Section 3.1 reviews historical SPMS descriptions of dust and bioparticles. Section 3.2 provides the first description of cellular and dust particle spectra as a function of TPII. Section 3.3 gives a detailed review of ion markers in cellular particle standards analyzed both in this study and previous works. In Sec. 3.4, cellular particles are identified in an ATOFMS dataset from a coastal field study and compared against co-located bioparticle and INP measurements.

3.1. Common “biological” ion markers detected in dust particle spectra

Consistent with previous SPMS studies (Creamean et al. 2014; Schmidt et al. 2017; Silva, Carlin, and Prather 2000; Zawadowicz et al. 2017), ²⁶CN⁻, ⁴²CNO⁻, and ⁷⁹PO₃⁻ ion signals were frequently detected in dust samples analyzed in this work (Figure 1). Note that ATD are not directly comparable to the other dust spectra as ATD was analyzed by a different ATOFMS, but insight can still be gleaned from these samples by studying trends within the dataset. In agreement with previous SPMS dust analyses (Marsden et al. 2019; Schmidt et al. 2017), single particle mass spectra of soil dust samples, especially those collected from noted vegetation areas (Ethiopian soil dust, ESD; and Argentinian soil dust, ASD), generally had higher relative ion signals from organic nitrogen ion markers ²⁶CN⁻ and ⁴²CNO⁻ than other mineral dust samples (Figure 1). Yet due to ion signal variations within each dust sample, there is limited ability to utilize these marker ions alone to definitively identify any single mass spectrum as originating from a relatively “organic-rich” soil dust particle. Furthermore, as ²⁶CN⁻ and ⁴²CNO⁻ are readily produced from mineral dust, as shown here (Figure S2) and in Zawadowicz et al. (2017), there is no way to conclusively link the ion signals to whole or fragmented biological cells.

Figure 1. Distribution of relative ion signals for (a) ²⁶CN⁻, (b) ⁴²CNO⁻, (c) ⁷⁹PO₃⁻ from a variety of soil and mineral dust samples. Warmer colors indicate higher relative intensities. The ATD sample was analyzed with a different ATOFMS than the other samples, and thus the relative ion distributions are not directly comparable.

Figure 2. (a) Individual mass spectra for all cellular standards (red) and dust standards (black) plotted in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ space. The blue box in the inset highlights a region of high overlap between the two particle types. Individual (b) ASD and (c) Ps mass spectra in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ space. (d) Individual mass spectra for all Cell-type (red) and Dust-CW (black) particles plotted in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ space. The red box in the inset highlights a region of high overlap between the two particle types. Individual (e) Dust-CW and (f) Cell-type mass spectra in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ space. Warmer shades indicate higher total positive ion intensities.

Zawadowicz et al. (2017) made first use of analyzing spectra in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ coordinate space to separate populations of inorganic and biological particles. In this article, the FIN-1 ATOFMS bacteria and dust spectra plotted in the ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ coordinate space exhibit similar trends as shown in Zawadowicz et al. (2017; Figures 2 and S1). Generally, bacteria spectra exhibit a wide range of ⁷⁹PO₃⁻/⁶³PO₂⁻ (∼10–100) ratios with ²⁶CN⁻/⁴²CNO⁻ values limited to relatively low values (∼<1). For dust spectra, the ⁷⁹PO₃⁻/⁶³PO₂⁻ ratio is relatively narrow and limited to low values (∼<10) with a broader ²⁶CN⁻/⁴²CNO⁻ distribution when compared to the bacteria spectra. However, within the FIN-1 dataset there is also a region of high overlap between the bacteria and dust particle types (Figure 2a). Over 20% of all spectra for the BSD, MSD, KF, and INX dust samples and 14% of the Ps cell sample fall within the highlighted overlap region (see Table 1 for a complete description of all samples). The lack of clear separation of ATOFMS-generated dust and bacteria spectra in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO space indicates utilization of this classification technique with any SPMS instrument first requires a careful study of an array of control samples.

3.2. Variations in total positive ion intensity for cellular and dust mass spectra

For particles of similar composition, the TPII of the mass spectra has been shown to generally increase with increasing laser pulse energy (Sultana, Collins, and Prather 2017; Thomson, Middlebrook, and Murphy 1997; Wade et al. 2008; Woods et al. 2002). Therefore, TPII may be utilized as a proxy for the photon flux experienced by the particle within the much larger laser desorption/ionization beam. For most SPMS, even when laser pulse energy is consistent, photon flux varies across the laser beam profile (Wenzel and Prather 2004). Within a dataset, examining TPII in tandem with ion signals helps disentangle the impact of ionization conditions on mass spectra compared to natural variances in particle composition. For the individual dust and biological particles in this study, their location in ²⁶CN⁻/⁴²CNO⁻ versus ⁷⁹PO₃⁻/⁶³PO₂⁻ space shifts toward lower ⁷⁹PO₃⁻/⁶³PO₂⁻ and higher ²⁶CN⁻/⁴²CNO⁻ values with increasing TPII (Figures 2b, 2c, and S1). This motivated a comprehensive investigation of dust and biological mass spectral trends in the dataset with respect to TPII. The following sections present a holistic discussion of this study’s results and trends contextualized by an array of previous dust and biological SPMS analyses.

3.2.1. Dust mass spectra and total positive ion intensity

Consistent with previous observations, most positive ion dust spectra are dominated by a combination of potassium, calcium, sodium, iron, and aluminum ion markers (Figures 3 and S2). For more general characterizations of mineral and soil dust samples, readers are referred to Marsden et al. (2019), Schmidt et al. (2017), Silva, Carlin, and Prather (2000), and Zawadowicz et al. (2017). Here we focus on TPII-dependent systematic variations within the dust mass spectra.

Figure 3. Representative mass spectra for ASD (dust) with relatively low (a) and high (b) TPII. Representative mass spectra for Ps (cellular) with relatively low (c) and high (d) TPII.

The relative intensity of potassium ion signal has historically been utilized as a key marker for both dust and biological spectra. The most consistent trend in the positive ion mass spectra across all dust samples was the decrease in ³⁹K⁺ relative signal and the increase in ²⁷Al⁺, ⁴⁸Ti⁺, ⁶⁴TiO⁺ and with increasing TPII (Figures 3 and S2). Similar to ³⁹K⁺, the relative ²³Na⁺ signal also decreased with increasing TPII for the KF and BSD samples, the only samples with reported contributions from albite. As detailed below, these observed trends in relative positive ion signals with TPII are consistent with common paradigms applied to LDI.

It has been suggested that both the lattice energy of a compound and the ionization energy of an atomic species influence the generation of metallic cations (Gross et al. 2000; Thomson, Middlebrook, and Murphy 1997). Within soils, aluminum content is typically organized as sheets or in 3D crystal structures, though some soil particles can be contain disordered clay-like colloids (Engelbrecht et al. 2016). However, the aluminum in these colloids is still either tetrahedrally or octahedrally coordinated. Titanium can sometimes substitute for aluminum within these lattices. Unlike aluminum and titanium, interstitial ions, such as potassium in both K-feldspar and illite and sodium in albite, are not tightly bound in the lattice structures. Marsden et al. (2019) suggested that weakly bound interstitial cations between clay sheets desorb and ionize before disintegration of the alumina lattices. Our results are consistent with this paradigm; the higher aluminum and titanium relative ion signals with increasing TPII is consistent with a higher expenditure of energy required to release them from lattice structures compared to potassium. However, potassium and sodium have lower first ionization energies than aluminum and titanium which could also contribute to their preferential detection in lower TPII mass spectra (Sansonetti and Martin 2005).

Changes in relative negative ion intensities as a function of TPII were more subtle compared to the positive ion signatures. Primarily, signals from a wide range of low m/z ion markers, roughly from m/z − 12 to −50, increase with increasing TPII, though most of these relative ion signals remain well below 1% of the total negative ion intensity (Figures 4a, 4b, and S2). Many of the low mass ion peaks, such as ²⁴C₂⁻, ²⁵C₂H⁻, and ²⁶CN⁻, are attributed to fragmentation of organic species. Signal from ¹⁶O⁻ and ¹⁷OH⁻, which could be a result of fragmentation of various mineral oxides or adsorbed water on the particles, also generated relatively intense signals at high TPII. While the growth of low m/z ions with increasing TPII may be unsurprising to SPMS users, to our knowledge this relationship is not documented in the literature. Mixed organic-dust particles can have enhanced IN-activity relative to mineral dust (Conen et al. 2011; O’Sullivan et al. 2014; Tobo et al. 2014), and are thus of great interest to atmospheric scientists. Yet this work highlights that key ion markers typically associated with “organic-rich” dust particles, such as ²⁶CN⁻ and ⁴²CNO⁻, are also tied to the energy experienced by the particle during the LDI process and may be especially important to consider given the variation in photon flux across the LDI beam common to SPMS systems.

Figure 4. For each dust sample analyzed, number fraction of mass spectra with relative ion signals for (a) ³⁹K⁺ > 0.25, (b) ²³Na⁺ > 0.01, (c) ²⁷Al⁺ > 0.05, (d) and ⁴⁸Ti⁺,⁶⁴TiO⁺ > 0.05.

3.2.2. Bioparticle mass spectra and total positive ion intensity

Consistent with Steele et al. (2003, 2005), which examined the influence of laser pulse energy on biological spectra production and signatures, single particle mass spectra for the biological samples analyzed in this study show systematic variations with TPII. A handful of relatively low mass ions, notably 26CN⁻, ⁴²CNO⁻, −66, and −90, increase in relative intensity with increasing TPII (Figures S2–S4). Prior results from SPMS studies of bacterial cells have shown an increase in ²⁶CN⁻, ⁴²CNO⁻, −58, and −66 ion markers with increasing laser pulse energy (Steele et al. 2003, 2005). Though these results were obvious in the figures presented by Steele et al. (2003, 2005), they were not noted at all (Steele et al. 2003) or discussed only briefly as an increase in “negative low-mass peaks” (Steele et al. 2005).

In the positive spectra, fewer coherent trends were discovered across the biological samples studied. With increasing TPII, m/z + 59 and +74 signal consistently decreased (Sec. 3.3.4), and conversely inorganic cation signal, such as ²³Na⁺ and ³⁹K⁺, sometimes increased. Steele et al. (2003, 2005) also showed in figures that m/z + 74 signal decreased markedly with increasing laser pulse energy, though this trend was only noted in the text by Steele et al. (2005). In the Ps, Pf, and Cr samples, signals from potassium-containing cluster ions (Sec. 3.3.5) at m/z 104, 120, 141, 157, 175, 197, and 213 generally increased with increasing TPII, consistent with results from Sultana et al. (2017), which noted an increase in m/z + 104 and +120 with TPII and laser pulse energy. For all cell types sampled in the UC San Diego microalgae study (2–6 μm D_va), an increase in TPII resulted in fragmentation of presumably high mass organic species, as indicated by relatively more intense lower m/z signals at intervals of 14 m/z in the region m/z + 350 to +625 (Figure S5). Also observed with increasing TPII was decreasing relative intensities for m/z − 794 and −820, which have been attributed previously to the [M-H]⁻ ions of sulfoquinovosyldiacylglycerides (Cahill et al. 2015; Plouguerné et al. 2013), a lipid common to chloroplast membranes (Figure S5). These results and discussion highlight that when particles encounter high photon energy densities, organic species specific to bioparticles may be fragmented to nonspecific organic ions at lower m/z, complicating the identification of cellular aerosols within a complex aerosol population.

Figure 5. Temporal profiles of (a) ATOFMS hourly counts for Dust (divided by 10, brown) and Cell-type (green), (b) WIBS concentrations for FP (red, scaled by 1/5), FP3 (blue), and FBAP (black), and (c) INP_-15 (L⁻¹). Correlation coefficients (r²) between INP_-15 and WIBS or ATOFMS particle types are shown in their respective colors.

3.3. Key ion markers for identification of cellular spectra

In addition to the influence of laser energy upon ion signatures, single bioparticle mass spectra have also been shown to vary with biological species, life cycle, and nutrient conditions (Cahill et al. 2015; Czerwieniec et al. 2005; Srivastava et al. 2005; Tobias et al. 2006). We have yet to determine any single ion marker which is both universally generated by aerosolized cells and also can reliably differentiate between cellular and dust mass spectra. Yet as shown by previous studies and supported by the results in this work, bioparticles do produce a range of distinguishing spectral signatures, broadly grouped as nucleobases, amino acids, quaternary amines, negative ion clusters containing phosphate, and positive ion potassium-containing clusters. The following sections briefly describe these ion markers, summarized in Table S2. An extended discussion is presented in Sec. S2. Ion signals from these species are often very low in intensity, causing them to be overlooked in some statistical and machine learning analyses applied to atmospheric datasets. However, by utilizing a combination of these markers, we propose that one should be able to confidently discriminate a fraction of the cellular spectra in a complex dataset. The following discussion could help build a new framework for the identification of bioparticles, divorced from common ion signals shared by bioparticles and dust, such as ²⁶CN⁻, ⁴²CNO⁻, ⁷⁹PO₃⁻, and ³⁹K⁺.

We emphasize that the presence of signal at an m/z associated with a cellular ion marker discussed below is not necessarily indicative of a cellular spectra. Characteristics of the entire spectra must be considered when attempting to discriminate cellular spectra in atmospheric datasets. Researchers should still use historically established signatures and ion markers to identify other atmospheric particle types such as combustion, secondary organic, and sea spray aerosols. The discussion presented herein builds upon this existing foundation and presents supplemental tools to disentangle dust and biological spectra.

3.3.1. Nucleobases

Negative mass spectra of nucleobases are characterized by the [M-H]⁻ ion (e.g., ¹⁵⁰[Guanine-H]⁻, ¹³⁴[Adenine-H]⁻, with further fragmentation of these species potentially yielding additional ion markers at m/z − 58, −71, −64 to −67, −90, −95, −106, −107, −108, −117, −131 to −134, −149, −150 (Smith, Schram, and Mccloskey 1983; Srivastava et al. 2005; Varmuza et al. 1999). These ion markers have been commonly observed in previous SPMS studies of bacterial cells (Czerwieniec et al. 2005; Srivastava et al. 2005; Steele et al. 2003; Tobias et al. 2006) and likely cell-containing sea spray aerosol (SSA; Sultana, Al-Mashat, and Prather 2017). In this work in the UC San Diego microalgae studies, nucleobase signatures were frequently observed at m/z − 90, −106 to −107, −117, −131 to −134, −149, and −150 (Figures S3 and S6a). With strong negative nucleobase ion signatures, positive ion signals were also commonly detected at m/z + 110, +135, and +152 (Figure S6b), likely a result of guanine ionization (Varmuza et al. 1999), and +185, not assigned at this time. In the FIN-1 bacteria studies weak signal at −90 was sometimes observed, however nucleobase ion markers above −90 were rare. Ion markers, potentially a result of extensive nucleoside fragmentation, at m/z − 58, −64 to −66, and −71, were observed in all the cellular samples examined here as well. Approximately 70% of all bacterial FIN-1 and microalgae negative spectra had relative signal at m/z − 66 above 0.005. Here and in the literature a signature from −64 to −67 m/z range is pervasive. Signal at m/z − 66 is almost always greater than m/z − 65 and −64, and signal at −67 is very low relative to all these three (Figure S6c). With increasing laser pulse energy (Steele et al. 2003, 2005) or TPII, as examined here, the relative intensities of these markers generally increase, with signal at m/z − 65 and −64 approaching the intensity of m/z − 66 (Figure S6d).

3.3.2. Amino acids

Ion markers which have been assigned to an array of amino acids (Czerwieniec et al. 2005; Schneider et al. 2011; Srivastava et al. 2005) have been widely noted in cellular single particle mass spectra (Fergenson et al. 2004; Srivastava et al. 2005; Tobias et al. 2006, 2005). Consistent with these previous studies, peaks which can be attributed to amino acid [M-CO₂]⁺ ions (e.g., ⁷⁰[Proline-CO₂]⁺, ⁸⁶[Leucine-CO₂]⁺) frequently co-occurred in spectra for all biological species analyzed in this work. As has been previously observed (Czerwieniec et al. 2005), in this study signal at m/z + 84 was almost always detected in spectra when a relatively intense ⁸⁶[Leucine-CO₂]⁺ peak was also present. The amino acid markers ¹⁷³[Arginine-H]⁻, ¹⁷⁵[Arginine-H]⁺, and ¹⁴⁶[Glutamic acid-H]⁻ have also been noted in previous studies (Czerwieniec et al. 2005; Srivastava et al. 2005; Sultana et al. 2017). In this work signal at m/z 175 was detected in 32% of bacterial FIN-1 mass spectra, but m/z 173 and 146 were not noted in the FIN-1 or microalgae studies.

3.3.3. Negative ion clusters with phosphate

Signals from characteristic ion fragments of organophosphate species (Strzelecka et al. 2017) have been frequently noted in cellular single particle mass spectra (Fergenson et al. 2004; Srivastava et al. 2005; Tobias et al. 2006, 2005). Unfortunately, the two most common organophosphate ion markers in cellular spectra, ⁷⁹PO₃⁻ and ⁶³PO₂⁻, are also generated by inorganic phosphates (Silva 2000; Zawadowicz et al. 2017). Signal at m/z −97 (⁹⁷H₂PO₄⁻) has been noted in nearly every published SPMS cell analysis study and was also ubiquitous in the cellular particle types sampled here. Utilizing m/z − 97 as an identifier for ambient cellular particles is not possible because sulfates generated during atmospheric processing commonly result in the detection of ⁹⁷HSO₄⁻. However, a series of higher m/z phosphate-derived signals at m/z − 159, −199, −215, −261, −277, −381 characteristic of organophosphate species such as DNA (Strzelecka et al. 2017), has been reported for Bacillus species (Czerwieniec et al. 2005; Tobias et al. 2005). For both the bacterial FIN-1 and UC San Diego microalgae mass spectra, a similar series of phosphate-derived ions are repeatedly observed, with signal at some combination of the markers listed above in addition to m/z − 239 and −301 (Figure S6f). Out of these markers, a relatively intense signal at m/z − 159 is most frequently detected; 44% of bacterial FIN-1 and 59% of microalgae negative spectra had a relative signal above 1% of the total spectrum.

3.3.4. Quaternary amines

A number of SPMS cell studies have noted signal at m/z + 74, often accompanied by signal at m/z + 59 (Czerwieniec et al. 2005; Fergenson et al. 2004; Srivastava et al. 2005; Tobias et al. 2006, 2005). These marker ions have been attributed to the fragmentation of organic species with quaternary amines, such as trimethylglycine, (Srivastava et al. 2005). Mass spectra with relatively intense signals at m/z + 59 and +74 (Figure S6g) were common in the Ps, Pf, and Cr data sets, but were relatively rare in the UC San Diego microalgae study. Using m/z + 59 to identify cellular spectra in ambient datasets should be performed with care as isobaric contributions from ⁵⁹Co⁺, in dust, as well as ⁵⁹Na(H₂O)₂⁺, common to wet sea spray aerosol spectra, are possible. Also single particle measurements of amines (Angelino, Suess, and Prather 2001; Pratt, Hatch, and Prather 2009) and organic carbon (Cahill et al. 2012; Spencer and Prather 2006) have been shown to readily generate ion signal at +59.

3.3.5. Potassium-containing positive ion clusters

A series of relatively low intensity signals above m/z + 100, including +104, +120, +157, +141, +175, +197, and +213, have been noted in some SPMS studies of cells (Tobias et al. 2005), vegetative detritus (Silva 2000), and likely cell-containing SSA (Sultana et al. 2017). Silva (2000) attributed these signals to potassium-cluster ions (Table S2) though this assignment is not treated as definitive. For instance, the ionization of amino acids arginine and phenylalanine can generate signals at +175 (Srivastava et al. 2005) and +120 (Czerwieniec et al. 2005). All of the potassium-cluster ions described by Silva (2000) have distinct isotopic signatures with a X:X + 2 ratio (e.g., ¹⁵⁷K₂PO₃⁺:¹⁵⁹K₂PO₃⁺) within the range of 7:1 to 4:1. Isotopic signal at m/z X + 2 is apparent in the provided mass spectra within Sultana et al. (2017) and Silva (2000), indicating contributions from potassium-cluster ions, though many other cellular spectra with signal reported at m/z + 175 show no significant signal at +177. In this study, only Ps, Pf, and Cr generated spectra with at least some obvious contributions from potassium-cluster ions, as evidenced by ion signal at m/z X + 2 (Figure S6h). Using potassium-cluster ions to identify cellular spectra in ambient datasets should be utilized with care as other particle types are known to generate similar clusters. Silva (2000) readily detected ¹⁴¹K₂PO₂⁺ and ¹⁵⁷K₂PO₃⁺ ions from potassium phosphate standards, and biomass burning particles can also generate ¹⁰⁴K₂CN⁺, ¹²⁰K₂CNO⁺, ¹⁹⁷K₃SO₃⁺, and ²¹³K₃SO₄⁺ ions (Silva et al. 1999).

3.4. Bioparticles in ambient studies

Accurately distinguishing between externally mixed bioparticles and dust in the ambient atmosphere is important due to their dissimilar ice nucleating potentials (Pratt et al. 2009a; Sun and Ariya 2006 and references within). Dust particles typically have higher atmospheric concentrations and are broadly acknowledged as the most abundant source of INPs below −20 °C (Andreae and Rosenfeld 2008; Kanji et al. 2017). Certain bioparticle types can be effective ice nucleators at temperatures ≥ −5 °C (Vali et al. 1976), and may be the exclusive source of INPs at temperatures warmer than −15 °C, despite their rarity (DeMott and Prenni 2010).

In Secs. 3.1–3.3, mass spectral features were discussed that may be used to distinguish between dust and bioparticles, the possible identities of those mass spectral features, and the trends associated with variations in laser fluence. In the following sections, we utilize that information for the identification of bioparticles in a complex, ambient SPMS dataset collected during a field study at a coastal site in California. Once identified, the temporal trends in bioparticles are compared to bioparticle concentrations measured by the WIBS and INP concentrations measured by droplet freezing assays.

3.4.1. Cell-type classification

Using the information in Secs. 3.2 and 3.3, we developed a decision tree (Figure S7) in order to identify cellular particles (hereafter Cell-type particles) during the CalWater-2015 field campaign. A fuller description of the development of this decision tree is given in the SI. Cell-type particles were identified after other particle classes were assigned and this classification was assumed to take precedence. Our decision tree heavily weights low intensity ion peaks outside of those previously used to identify bioparticles and explicitly rules out internally mixed dust and bioparticles from the Cell-type classification. Thus, the identified Cell-type population does not necessarily contain all aerosol particles with spectra characteristic of cellular material. While we believe our process can be adapted to select for internally mixed dust/bioparticles, we have elected not to do so here. While our decision tree does not employ every potential bioparticle ion marker, it does use each class in the identification process. It should be noted that many of the steps taken in this process are unique to the context of these measurements and instrumental parameters. While we believe that the classification scheme utilized here can serve as a useful template for future endeavors, it should not be applied to other datasets indiscriminately. Rather, each dataset should be investigated independently for the presence of characteristic ion markers (Secs. 3.2 and 3.3), to identify bioparticles.

A total of 2528 Cell-type particles were identified out of 1,220,818 particles measured by the ATOFMS at during the CalWater-2015 study (0.2% of all particles) and is in line with previous SPMS measurements of bioparticle abundance (0.4–2%; (Zawadowicz et al. 2017, 2019). Figure S8 shows the average spectrum of Cell-type particles identified during CalWater-2015. In addition to intense signals from ³⁹K⁺, ⁷⁹PO₃⁻, ²⁶CN⁻, and ⁴²CNO⁻, many of the mass spectral features noted in studies of control bioparticles, detailed in Table S2, were present in Cell-type spectra, including nucleobases (66⁻, 71⁻, 90⁻, 117⁻, 131–134⁻), negative phosphates (¹⁵⁹H(PO₃)₂⁻, ¹⁹⁹NaH₂P₂O₇⁻), quaternary amines (⁵⁹N(CH₃)₃⁺, ⁷⁴N(CH₃)₄⁺), and positive potassium clusters (¹⁰⁴K₂CN⁺, ¹⁵⁷K₂PO₃⁺, ¹⁷⁵K₂HSO₄). These particles also contained ion markers that have been associated with amino acids (86⁺, 175⁺). While many Cell-type particles contained multiple of these bioparticle markers, no single spectrum contained all of them. Cell-type particles also showed similar trends with regards to increasing TPII (Figure S9) as those observed in lab studies of aerosolized controls (Figure S4). For instance, signals at +59 and +74 were relatively more intense at lower TPII, while signals from phosphate clusters (¹⁵⁹H(PO₃)₂⁻, ¹⁹⁹NaH₂P₂O₇⁻) and potassium clusters (¹⁵⁷K₂PO₃⁺, ¹⁷⁵K₂HSO₄) increased with increasing TPII. To our knowledge this latter relationship has not previously been described in the literature. Our confidence that identified Cell-type particles are bioparticles is strengthened by the fact that they have TPII trends similar to our standards and those found in the literature. These results also highlight that understanding the full bioparticle mass spectral space is critical to identifying them as using just one or two of these ion markers to distinguish potential bioparticles would be insufficient as very few are consistently found across the entire TPII range.

A total of 65,534 Dust-CW particles were identified by the ATOFMS during the CalWater-2015 field campaign. When translated into CN⁻/CNO⁻ versus PO₃⁻/PO₂⁻ space, Dust-CW (more detail on the typing and classification of these particles can be found in Sec. 2.2) and Cell-type particles exhibited similar trends to dust and bioparticles analyzed during FIN-1 (Figure 2). The PO₃⁻/PO₂⁻ and CN⁻/CNO⁻ values for Dust were primarily located between ∼0–20. Cell-type PO₃⁻/PO₂⁻ ratios ranged broadly, from ∼2–50, and the CN⁻/CNO ratios were primarily below 2. As with the FIN-1 samples, identified Dust-CW and Cell-type particles are not well separated in CN⁻/CNO⁻ versus PO₃⁻/PO₂⁻ space. A region of particularly high overlap is highlighted by a red box in Figure 2d and contains 17.5% of all Dust-CW particles and 6.8% of Cell-type particles. The asymmetry between the relative number of Dust-CW and Cell-type means that even small numbers of misclassified particles can lead to a gross overestimate of the number of bioparticles. For instance, even just 5% of Dust-CW particles being misclassified as Cell-type would more than double the total number of Cell-type particles. However, the average spectra for the two types in this overlapping space have numerous distinct ion markers (Figure S10) which can be utilized for distinguishing between these two classes.

3.4.2. Comparisons of Cell-type particles with WIBS and INP concentrations

Hourly counts of Cell-type particles were compared to the WIBS particle concentrations typed using FP, FBAP, and FP3 particle classification schemes (Figure S11). We note that by using our ambient dataset to develop the classification methods we are unable to quantify the classification uncertainty in our method because we do not know the true identity of measured particles. However, methodologies that use only laboratory measurements to develop their classifiers are unable to account for factors affecting ambient particle composition such as atmospheric processing. As such, it is important to understand how our method compares with other bioparticle measurements. The FP type (a particle with fluorescence signal in any channel) is the broadest class of fluorescent particle and is most likely to contain non-biological aerosol due to the potential interference from weakly fluorescing materials such as polyaromatic hydrocarbons, soot, and mineral dust (Toprak and Schnaiter 2013) and correlated very poorly with Cell-type (r²=0.26). The FBAP class is defined by signal in channels FL1 and FL3 and has found to have less interference from weakly fluorescing, non-biological materials than the FP class (Gabey et al. 2010). We found a similarly poor correlation between FBAP concentrations and Cell-type particle counts (r² = 0.26). Finally, the FP3 class was first described by Wright et al. (2014) and is defined by its extremely high fluorescence in FL1 (broadly thought to correspond to tryptophan) and little signal in FL2 and FL3. These particles were hypothesized to be fungal spores, though to our knowledge their identity has not been confirmed in laboratory studies. This class has also shown a strong relationship with apparent biological INPs in field studies, and thus could be an important proxy for future studies (Suski et al. 2018; Wright et al. 2014). We found that the FP3 class correlated moderately well with Cell-type hourly particle counts (r² = 0.54), suggesting that it may be comprised of a large population of whole or fragmented cells.

Measurements of WIBS particle classes and the Cell-type counts were also compared to INP concentrations measured using offline sampling methods. While the strongest predictor of INP_-15 was FP3 concentrations (r² of 0.63; Figure 5), Cell-type particle counts also correlated well (r² of 0.50) and both of these particle classes may be useful proxies for high temperature INPs. Neither Dust-CW (r² of 0.18), FP (r² of 0.15), nor FBAP (r² = 0.14) were good predictors of INP_-15. This suggests that bioparticles (i.e., whole or fragmented cells) controlled warm INP concentrations during the CalWater-2015 study. Due to the scarcity of Cell-type particles in the ATOFMS dataset (0.2%), special care should be taken when identifying this particle type. Even small errors in classification can lead to large relative changes in number of particles, and the predictive utility of this particle type may be negated.

4. Conclusions

The study builds a foundation for researchers to distinguish biological particles in complex aerosol populations using SPMS techniques. The dependence of cellular and dust mass spectral fingerprints on the amount of laser energy absorbed by each particle, increases the difficulty in accurately identifying bioparticles utilizing only a handful of ion markers. This dependence on ionization conditions likely contributes to the relatively poor separation of cellular and dust spectra in ⁷⁹PO₃⁻/⁶³PO₂⁻ versus ²⁶CN⁻/⁴²CNO⁻ space, despite demonstrated success by researchers utilizing a different SPMS.

We propose common cellular ion markers from the ionization of biological species such as DNA and amino acids, can be carefully utilized to distinguish bioparticles from complex ambient single particle mass spectral datasets. These low-intensity but characteristic signals have been frequently documented in laboratory studies, but have yet to be leveraged in ambient studies which frequently utilize classification algorithms driven by high intensity peaks. Due to the variation in biological mass spectra with ionization conditions, to maximize the number of bioparticles accurately selected out of an ambient SPMS dataset, an array of characteristic biological signals should be utilized.

The characteristic ion markers discussed were utilized to identify Cell-type particles at a coastal site in California. The Cell-type particles showed similar mass spectral trends with TPII as bioparticles in laboratory studies, suggesting that our methods successfully identified cellular particles in an ambient dataset. While Cell-type particles only made up 0.2% of all ambient particle spectra, Cell-type particle counts correlated moderately well with the WIBS-derived FP3 subclass and INP concentrations at T = −15 °C. The correlation between Cell-type and the FP3 class is particularly interesting as the FP3 type has been strongly implicated for representing microbial INPs in field campaigns that measured INPs active at moderate supercooling (Suski et al. 2018; Wright et al. 2014). While INP composition was not directly measured, this evidence suggests that whole or fragmented cells were primary contributors to warm INPs during the CalWater-2015 field campaign. Due to the extreme rarity of ambient biological particles minimizing nonbiological particles from “cell-type” classification was crucial to attaining these results. This work illustrates that special care needs to be taken when identifying biological particles and that careful consideration of low intensity peaks, in tandem with understanding how these features vary with TPII, can improve the mass spectral identification of this important class of particles.

Authors’ contributions

GCC participated in field study measurements, analyzed data, and wrote the article. CMS analyzed data and helped write the article. HA analyzed lab standards. NER participated in field study. MDP analyzed, quality assessed, and compiled cold-stage experiments and WIBS data. KAP, ACM, PJD, OM, and MDP provided over-arching ideas for study, measurements, subsequent sample processing, and were the leads of the key research projects.

Acknowledgments

The authors also thank the UC Davis Bodega Marine Laboratory for the use of laboratory and office space and shipping and physical plant support while collecting data, as well as the California Air Resources Board for the trailer used for sampling. The authors also thank the organizers and other participants in the FIN-1 workshop, including the staff of the Aerosol Interaction and Dynamics in the Atmosphere (AIDA) facility at the Karlsruhe Institute of Technology for their assistance in sampling, and Dan Cziczo for his role in organizing the FIN-01 campaign. Hans Taylor, Kylie Hoffman, Mary Hester, Margaret Scott, Hannah Tekleab, Mark Wu, and Seth Goodnight are thanked for their assistance in performing the cold-stage experiments. Hans Taylor is thanked for his help during sample collection.

Additional information

Funding

This research was supported by the CalWater-2015 field campaign, a multi-agency study with participants from NSF, NOAA, NASA, and DOE-PNNL. GCC was supported in part by the U.S. Department of Energy (BER), through the Early Career Research Program. This research was funded by NSF award number 145147 (GCC, CMS, HA, ACM, KAP), NSF award number 1450690 (NER, MDP) and NSF award number AGS-1450760 (PJD). The FIN-01 campaign was partially supported by the U.S. National Science Foundation grant no. AGS-1339264, and by the U.S. Department of Energy’s Atmospheric System Research, an Office of Science, Office of Biological and Environmental Research program, under grant no. DE-SC0014487.