optimized separation of oc and ec for radiocarbon … · 2018-11-06 · ix summary particulate...

Research Collection

Doctoral Thesis

Organic Aerosol Source Apportionment on Long-Term, Spatially-Dense Observation Networks Using Novel Mass SpectrometryTechniques

Author(s): Dällenbach, Kaspar Rudolf

Publication Date: 2017

Permanent Link: https://doi.org/10.3929/ethz-b-000238925

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-b-000238925

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

DISS. ETH NO. 24200

Organic aerosol source apportionment

on long-term, spatially-dense observation networks

using novel mass spectrometry techniques

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZÜRICH

(Dr. sc. ETH Zürich)

presented by

Kaspar Rudolf Dällenbach

MSc Environmental Sciences, ETH Zürich

born on 12.10.1988

citizen of Oberdiessbach, Switzerland

accepted on the recommendation of

Prof. Dr. Urs Baltensperger (examiner) Prof. Dr. Thomas Peter (co-examiner)

Dr. André Prévôt (co-examiner) Prof. Dr. James Schauer (co-examiner)

2017

v

Content

Summary ................................................................................................................................ ix

Zusammenfassung............................................................................................................... xiii

1 Introduction ........................................................................................................................ 1

1.1 Atmospheric aerosol-definition and properties .......................................................... 1

1.2 Aerosol climate impact .............................................................................................. 3

1.3 Aerosol health impact ................................................................................................ 6

1.4 Chemical composition and sources ............................................................................ 9

2 Motivation and thesis outline .......................................................................................... 13

3 Methodology ..................................................................................................................... 17

3.1 Aerosol Mass Spectrometer ..................................................................................... 17

3.1.1 General ........................................................................................................................... 17

3.1.2 offline AMS ..................................................................................................................... 19

3.2 Laser-Desorption/Ionization mass spectrometer ...................................................... 20

3.3 Source apportionment .............................................................................................. 21

4 Characterization and source apportionment of organic aerosol using offline aerosol

mass spectrometry ........................................................................................................... 25

4.1 Introduction .............................................................................................................. 26

4.2 Methods .................................................................................................................... 28

4.2.1 Aerosol sampling ............................................................................................................ 28

4.2.2 Offline AMS .................................................................................................................... 30

4.3 Results and discussion .............................................................................................. 35

4.3.1 Signal-to-noise, quantification and detection limits ....................................................... 35

4.3.2 OA recovery ................................................................................................................... 37

4.3.3 Mass spectral analysis ................................................................................................... 40

4.3.4 Source apportionment results ......................................................................................... 42

4.4 Conclusions .............................................................................................................. 54

vi

5 Long-term chemical analysis and organic aerosol source apportionment at 9 sites in

Central Europe: source identification and uncertainty assessment ........................... 57

5.1 Introduction ............................................................................................................. 58

5.2 Methods ................................................................................................................... 60

5.2.1 Study area and aerosol sampling ................................................................................... 60

5.2.2 Offline AMS analysis ...................................................................................................... 62

5.2.3 Other chemical analysis ................................................................................................. 63

5.3 Source apportionment .............................................................................................. 63

5.3.1 General principle ........................................................................................................... 63

5.3.2 Preliminary PMF ........................................................................................................... 66

5.3.3 Sensitivity analysis ......................................................................................................... 67

5.3.4 Factor classification ...................................................................................................... 68

5.3.5 Recovery and blank corrections ..................................................................................... 68

5.3.6 Solution selection ........................................................................................................... 69

5.4 Results and discussions ........................................................................................... 72

5.4.1 Interpration of PMF factors ........................................................................................... 72

5.4.2 Uncertainty analysis ...................................................................................................... 80

5.4.3 Factor relative contribution at different sites ................................................................ 86

5.5 Conclusion ............................................................................................................... 88

6 Insights into organic-aerosol sources via a novel laser-desorption/ionization mass

spectrometry technique applied to one year of PM10 samples from nine sites in

central Europe ................................................................................................................. 89

6.1 Introduction ............................................................................................................. 90

6.2 Methods ................................................................................................................... 93

6.2.1 Sample collection and other chemical analysis .............................................................. 93

6.2.2 Laser-desorption/ionization ToF MS analysis ............................................................... 95

6.2.3 Source apportionment / PMF ......................................................................................... 98

6.3 Results ................................................................................................................... 100

6.3.1 Calibration, repeatability, and quantification.............................................................. 100

6.3.2 Combustion source profiles .......................................................................................... 103

6.3.3 Ambient samples .......................................................................................................... 104

6.3.4 Source apportionment results ...................................................................................... 106

6.4 Summary and conclusions ..................................................................................... 119

7 Conclusions and outlook ............................................................................................... 121

8 Bibliography .................................................................................................................. 129

vii

9 List of Figures................................................................................................................. 161

10 List of Tables .................................................................................................................. 167

A Supplementary Material for chapter 5 Long-term chemical analysis and organic

aerosol source apportionment at 9 sites in Central Europe: source identification and

uncertainty assessment .................................................................................................. 169

B Supplementary material for chapter 6 Insights into organic-aerosol sources via a

novel laser-desorption/ionization mass spectrometry technique applied to one year

of PM10 samples from nine sites in central Europe ................................................... 181

11 Acknowledgement .......................................................................................................... 197

12 Curriculum vitae ............................................................................................................ 201

ix

Summary

Particulate matter (PM), liquid or solid particles suspended in the atmosphere,

contributes significantly to air pollution. Besides negative acute and long-term health

effects, PM also affects the Earth’s climate by scattering and absorbing solar radiation

and also by changing cloud properties. A large part of PM consists of organic aerosol

(OA) which is a complex mixture of compounds. OA can be emitted directly as

particleswhich is called primary OA (POA) or can be formed as secondary OA (SOA)

through atmospheric aging of initially emitted volatile organic compounds (VOCs).

Important OA sources are traffic, cooking, combustion of biomass or fossil fuels but

also biogenic VOC emissions from terrestrial and marine ecosystems. For

understanding the many processes involved, measurements with high temporal

resolution are necessary. Similarly, for establishing mitigation goals and strategies also

long-term analyses in spatially dense observation networks are crucial.

In this dissertation, we used two mass spectrometers in a new way for long-term

analyses of OA in order to obtain a deeper understanding of its chemical composition

and the different contributing sources. The Aerodyne aerosol mass spectrometer (AMS)

is designed and used for short-term online measurement campaigns in order to

determine the contribution of the different OA sources. However, the complex

maintenance and the instrument cost hinder long-term and multi-site deployments. In

comparison, collecting PM on filters is easier and is routinely performed for

determining PM concentrations worldwide. In a first study, we developed an analytical

technique, based on analyzing the organic mass spectral signatures of water extracted

filter samples by AMS (termed offline AMS). By comparing offline AMS results to

reference methods like online deployments of AMS or the more robust aerosol chemical

speciation monitor (ACSM), we showed that the analyzed water-soluble OA (WSOA)

covers a large part of OA and also exhibits similar signatures as total OA. While highly

oxygenated organic compound classes seemed to be well represented, hydrocarbons are

strongly underestimated. A statistical source apportionment (SA) analysis pointed out

the value of such offline AMS data. In comparison to SA results from online ACSM

x

analysis, we determined what proportion of the different OA sources was represented in

the offline AMS SA.

In a second study, we determined the contributions of different relevant OA

sources using the offline AMS technique. Thereby, we separated the influence of POA

sources in terms of traffic (HOA), cooking (COA), and biomass burning (BBOA) as

well as secondary OA categories and then validated our results by comparing them to

established state-of-the-art measurements. Moreover also a factor explaining organic

sulfur fragments of yet unknown origin was resolved. This factor is hypothesized to be

related to traffic activity. This dataset of source contributions is unique in its coverage,

both temporal and spatial (covering the entire year 2013 at 9 sites in central Europe).

Our results showed that the influence of BBOA was higher in winter than in summer

and exhibited enhanced concentrations at southern alpine valley sites than the sites in

northern Switzerland (no valley sites). The SOA was separated seasonally in a

component dominant in summer (summer oxygenated organic aerosol, SOOA) and

another component dominant in winter (WOOA). SOOA concentrations increased more

than linearly with rising temperatures, similar to biogenic VOC emissions and biogenic

SOA concentrations in other studies. This suggests that SOOA is strongly influenced by

biogenic emissions. In contrary, WOOA concentrations evolved similarly to

anthropogenically influenced inorganic compounds like ammonium (and also sulfate

and nitrate) suggesting a considerable anthropogenic influence on WOOA.

The benefit of AMS/ACSM data is strongly limited by the loss of chemical

information caused by the fragmentation of the analyzed molecules during the

measurement. This limitation hampers separating sources and is most prominent for

SOA because of the evolution of OA to similar chemical composition during

atmospheric aging. Therefore, we used in a third study in this dissertation a mass

spectrometer equipped with a laser-based ionization (laser-desorption/ionization mass

spectrometer, LDI) inducing less fragmentation. We developed and optimized an

analysis and data analysis framework for LDI analyses of PM collected on quartz-fiber

filters. This framework was applied to samples which were also analyzed by offline

AMS in order to obtain a deeper understanding of OA and its sources. In this study, we

apportioned OA to sources like traffic, wood burning, and also to secondary OA. In

comparison to offline AMS SA results of the same samples (2nd study), LDI SA

suggested higher contributions of primary sources (different response factors for

xi

sources). The influence of traffic is separated into a part interpreted as primary tailpipe

exhaust and another part interpreted as either aged/secondary traffic emissions or

influenced by resuspended road dust. Since also the ratio of fossil organic to fossil

elemental carbon concentration (related to fossil fuel burning, 14C analysis) increased in

summer, the hypothesis of aged/secondary traffic emissions gains further trust.

Furthermore, we separated the influence of biomass burning in different components.

Based on a comparison of the factors’ mass spectral signature to laboratory wood

burning experiments, the factors could be related to the burning conditions. According

to these results, inefficient wood burning POA contributed more to the total wood

burning POA at the southern alpine sites than at the sites north of the alpine crest during

high-pollution winter-days, possibly contributing to the higher OA concentrations

attributed to primary wood burning emissions in the 2nd study.

xiii

Zusammenfassung

Feinstaub, worunter man feste oder flüssige Partikel in Suspension in der

Atmosphäre versteht, trägt substantiell zur Luftverschmutzung in der Atmosphäre bei.

Nebst negativen akuten und Langzeit-Auswirkungen auf die menschliche Gesundheit

beeinflusst der Feinstaub auch das Klima der Erde, durch Streuung und Absorption des

Sonnenlichts und durch die Veränderung von Wolkeneigenschaften. Ein grosser Anteil

des atmosphärischen Feinstaubs (mit aerodynamischem Durchmesser <10µm) besteht

aus organischem Material (OM), einer komplexen Mischung von Molekülen

verschiedener Stoffklassen. Der organische Feinstaub kann direkt als primäres Aerosol

(POA) emittiert werden, oder durch atmosphärische Alterung sekundär aus anfänglich

gasförmigen organischen Stoffen gebildet werden (SOA). Wichtige Quellen sind

Verkehr, Kochen, Verbrennung verschiedener Energieträgern wie Holz und anderer

Biomasse nebst fossilen Brennstoffen, aber auch biogene Emissionen der marinen und

terrestrischen Ökosysteme. Zum Verständnis der Vielzahl involvierter Prozesse sind

Messungen mit einer hohen zeitlichen Auflösung nötig, jedoch sind Langzeitstudien auf

grossen Messnetzwerken ein Kernpunkt bei Bestimmung von

Emissionsreduktionszielen und Strategien zur Verbesserung der Luftqualität.

In dieser Dissertation wurden zwei verschiedene Flugzeitmassenspektrometer in

einer neuartigen Weise für Langzeitmessungen des organischen Feinstaubs zur

Erlangung eines besseren Verständnisses dessen chemischer Zusammensetzung und der

beitragenden Quellen verwendet. Das Aerodyne Aerosolmassenspektrometer (AMS)

wird online für kürzere Messperioden verwendet, um den Anteil verschiedener Quellen

des organischen Feinstaubs zu bestimmen. Jedoch machen es die komplexe Wartung

und der Instrumentpreis schwierig, Langzeitmessungen an verschiedenen Orten

durchzuführen. Vergleichsweise ist die Sammlung von Feinstaub auf Filtern einfacher

und wird weltweit routinemässig zur Bestimmung der totalen Feinstaubkonzentration

(PM) verwendet. Durch den erleichterten Zugang können auf solchen Filtern

Langzeitmessungen durchgeführt werden und zudem sind auch Orte zugänglich, die mit

xiv

dem AMS/ACSM nicht erreichbar sind. Zunächst wurde in einer Studie eine analytische

Methode entwickelt, die auf Messungen der organischen massenspektralen Signatur von

so gesammeltem Feinstaub in wässriger Lösung mithilfe des AMS basiert (offline

AMS). Es wurde anhand von Vergleichen mit Referenzmethoden (AMS, ACSM)

gezeigt, dass der mit dieser Methode analysierte Feinstaubanteil (wasserlöslicher

organischer Feinstaub) einen Grossteil des organischen Feinstaubs abbildet und

grundsätzlich eine ähnliche chemische Zusammensetzung des organischen Anteils

aufweist. Jedoch sind gewisse organische Anteile weniger gut repräsentiert, speziell

unoxidierte Kohlenwasserstoffe im Verhältnis zu hochoxidierten Komponenten. Eine

statistische Quellenzuweisungsanalyse basierend auf offline AMS Daten, zeigte den

Nutzen solcher Daten auf und im Vergleich zu online ACSM Messungen wurde

ermittelt, welcher Massenanteil der verschiedenen aufgelösten Quellen mit der offline

AMS Methode abgebildet werden kann.

In einer Folgestudie wurden basierend auf der offline AMS-Methodik Anteile

verschiedener Feinstaubquellen bestimmt. Dabei wurde der Einfluss primärer

Emissionen des Verkehrs, Kochprozessen und Biomassenverbrennungen nebst

sekundären OA Kategorien bestimmt und mithilfe anderer etablierter Messungen

validiert. Zusätzlich wurde auch ein bis jetzt unbekannter Faktor aufgelöst, der

organische Schwefelfragmente erklärt, und mit der Verkehrsaktivität in Verbindung zu

stehen scheint. Der resultierende Datensatz ist sowohl in der zeitlichen (ganzes Jahr

2013) wie auch geographischen Auflösung (9 Stationen in Zentraleuropa) einzigartig.

Der Einfluss des primären Biomassenverbrennungsfeinstaubs war im Winter höher als

im Sommer und zeigte an den Stationen in den südlichen Alpentälern stark erhöhte

Konzentrationen im Vergleich zu den Stationen nördlich der Alpen (keine Tallagen

vorhanden). Der sekundäre OA-Anteil wurde saisonal aufgetrennt in eine Komponente,

die im Sommer (Sommer-oxidiertes organisches Aerosol, SOOA), und eine andere, die

im Winter (WOOA) dominant war. Die SOOA-Konzentrationen stiegen

überproportional mit steigenden Temperaturen an, ähnlich wie in anderen Studien auch

biogene VOC Emissionen und biogenes SOA. Dies legt den Schluss nahe, dass SOOA

stark von solchen biogenen Emissionen beeinflusst wird. WOOA hingegen zeigte ein

ähnliches zeitliches Verhalten wie Ammonium (und andere anthropogen beeinflusste

xv

inorganische Aerosolkomponenenten wie Sulfat und Nitrat), was wiederum den Schluss

eines starken anthropogenen Einflusses zulässt.

Die Anwendung von AMS/ACSM-Daten ist durch die starke Fragmentierung und

den damit einhergehenden chemischen Informationsverlust limitiert. Aufgrund der

Tendenz, dass während der atmosphärischen Alterung Aerosole zu einer ähnlicheren

chemischen Zusammensetzung konvergieren, ist diese Limitation insbesondere für

SOA, aber auch generell für die Auftrennung chemisch ähnlicher Quellen hinderlich.

Aus diesem Grund wurde in dieser Dissertation auch ein Massenspektrometer mit laser-

basierter Desorption/Ionisation verwendet, welche durch eine sanftere

Ionisierungsmethode zu weniger Fragmentierung führt. Die Anwendung der im Rahmen

dieser Dissertation für die Anwendung auf Feinstaubproben (Quartzfilter) optimierten

Methode wurde auf den gleichen Datensatz wie die offline AMS Messungen

angewendet, um ein tiefergreifendes Verständnis der organischen Feinstaubs und seiner

Quellen zu erlangen. In dieser Studie wurde OA der Einfluss von OA-Quellen wie

Verkehr und Holzfeuerung nebst sekundärem OA aufgelöst. Der Vergleich mit offline

AMS Quellenzuweisungsresultaten legte dar, dass mit LDI-Daten höhere POA-

Konzentrationen als mit offline AMS-Daten abgeschätzt werden (verschiedene

Sensitivitäten zur Messung von Stoffklassen). Der Verkehrseinfluss konnte in einen

Teil, der mit primären Auspuffemissionen in Verbindung steht, und einen Teil, der als

entweder gealterte/sekundäre Auspuffemissionen oder aufgewirbelter Strassenstaub

interpretiert wurde, aufgetrennt werden. Da im Sommer auch das Verhältnis des fossilen

organischen Kohlenstoffs im Vergleich zum fossilen elementaren Kohlenstoff ansteigt

(aus der Verbrennung von fossilen Brennstoffen, 14C Analyse), erhält die Hypothese

gealterter/sekundärer Auspuffemissionen zusätzliches Gewicht. Ausserdem wurde der

mit Biomassverbrennungen einhergehende Feinstaub in Komponenten aufgetrennt, die

aufgrund ihrer massenspektrometrischen Signatur mit der Brenneffizienz in Holzöfen

(Laborexperimente) verbunden werden konnte. Während hochbelasteter Wintertagen

trugen gemäss dieser statistischen Analyse ineffiziente Feuerungen im Süden klar mehr

zum organischen Feinstaub aus Holzfeuerungen bei als an den Stationen nördlich der

Alpen. Diese ineffizienten Feuerungen könnten zu den höheren primären

Biomassenverbrennungs-Konzentrationen in den Alpentälern beitragen.

1

1 Introduction

1.1 Atmospheric aerosol-definition and properties

Aerosols are solid or liquid particles dispersed in a gas phase and are abundant in

the Earth’s atmosphere (summarized under the term particulate matter, PM), which are

emitted by natural and anthropogenic sources. These sources emit aerosol directly as

particles (primary particles) such as dust, pollen, fossil fuel combustion, wood

combustion. However, aerosol can also be formed by gas-phase oxidation of vapors and

subsequent condensation through new particle formation or particle growth from vapors

from e.g. fossil fuel and wood combustion, biogenic emissions (secondary).

According to the particle size PM is divided generally in a coarse fraction (>2.5

µm) and a fine fraction (<2.5 µm) which also have differing formation and removal

processes (Fig.1.1). The coarse fraction is dominated by resuspended dust, sea spray,

emissions from volcanoes, plant particles, and other emissions. A major sink for coarse

particles is dry deposition. The fine fraction can further be divided into accumulation

(0.1 to 2.5 µm), Aitken (10 to 100 nm), and nucleation mode (up to 10 nm). Particles in

the accumulation mode are either formed through coagulation or particle growth

through condensation of vapors on preexisting particles. For particles in the

accumulation range, removal mechanisms are slowest and thus they accumulate in the

atmosphere. Eventually they are deposited on the Earth’s surface by wet deposition (in-

cloud and below-cloud scavenging). Hot vapors produced during combustion processes

Chapter 1: Introduction

2

condense and produce particles in the Aitken mode while the even smaller particles in

the nucleation mode are formed through gas-to-particle processes. These particles are

lost to a considerable extent through coagulation with accumulation mode particles.

Typical ambient particle number concentrations are dominated by particles in the

nucleation and Aitken mode, while for volume and mass concentrations larger particles

have an increased importance (Seinfeld and Pandis, 2006).

Figure 1.1: Idealized ambient particulate size distribution including formation and removal processes (from Finlayson-Pitts and Pitts, 2000).

3

1.2 Aerosol climate impact

The Earth’s climate is governed by the global radiation budget which is in turn

influenced by the incoming solar radiation, as well as the Earth’s atmosphere and

surface. With the task to regularly assess the scientific understanding of parameters

influencing climate change, the Intergovernmental Panel on Climate Change (IPCC)

was created. The radiative forcing (RF) describes the net change of the global energy

balance (W/m2) caused by changes compared to pre-industrial conditions (after the

reference year 1750, Myhre et al., 2013). The RF is used to assess the magnitude of the

different factors contributing to climate change: a positive (negative) RF leads to more

(less) energy per surface and thus to a warming (cooling) effect of the factor in question.

In this comparison, the natural RF (change in solar irradiance) is small compared to the

total anthropogenic RF (Fig. 1.2).

Figure 1.2: Radiative forcing 2011 (1750 as reference for pre-industrial atmosphere) from Myhre et al. (2013).

Greenhouse gases such as e.g. CO2 strongly impact the radiative budget (RF of

1.68 W/m2), but also other atmospheric constituent have a considerable influence

(Myhre et al., 2013). The aerosol’s impact is separated in a direct aerosol-radiation-

interaction (ARI, related to scattering and absorption) and an indirect category (aerosol-


4

cloud-interaction, ACI). Among the different factors, the uncertainty related to aerosol

is highest of which the biggest part stems from ACI.

Parameters important for ARI (Fig. 1.2) are the total number concentration,

optical properties, and also the particle size. On one hand, aerosol particles scatter light,

reducing the incoming solar radiation at the surface. On the other hand, besides

scattering, a fraction of the aerosol, e.g., black carbon, also absorbs radiation and emits

radiation leading to a positive RF in the respective atmospheric layer. Therefore, the

aerosol source contributions determining the chemical composition of the aerosol are

decisive whether the RF is positive or negative and also define the magnitude of the

effect (Fig. 1.3).

Figure 1.3: Global Radiative forcing of different ambient aerosol constituents evolving over time (from Myhre et al., 2013).

ACI refers to the effect of the aerosol on cloud properties. Higher aerosol

concentrations likely lead to a higher abundance of cloud condensation nuclei (CCN),

which leads to a higher number of cloud droplets lowering the albedo of the cloud

because of the limited amount of water in the atmosphere. Since precipitation is

moreover decreased, the lifetime of the cloud is, thereby, also increased making the

5

effect lasting longer (Albrecht, 1989). However, the presence of black carbon reduces

the formation of clouds because of absorption of radiation and heating. The impact of

anthropogenic aerosol emissions is exemplified by the lower global radiation (Gilgen et

al., 1999) between 1960 and 1990 in e.g., Potsdam (Germany) coinciding with increased

sulfur emissions (Stern et al., 2005) which seemingly also impacted temperature

(Brohan et al., 2006) as illustrated in Fig. 1.4.

Figure 1.4: Surface solar radiation in Potsdam, Germany, (GEBA, Gilgen et al., 1999) and sulfur emissions in Europe (Stern et al., 2005) temperature anomalies over Europe from 1930 to 2010, and annual temperature anomaly in Europe over land (Brohan et al., 2006).

In estimating the anthropogenic (industrial) impact on the radiative forcing, both

an accurate representation of the present and past atmosphere is crucial (Carslaw et al.,

2013). While pre-industrial trace gas levels such as CO2 are largely well defined e.g. by

studying climate archives like ice cores , the pre-industrial aerosol burden is uncertain

(Myhre et al., 2013, Masson-Delmotte et al., 2013, Gordon et al., 2016). Recent results

suggest that under pre-industrial conditions pure biogenic new particle formation, which

can act as CCN after growth, is more pronounced than previously thought and,

therefore, the change in RF caused by aerosol has been overestimated by 27% (0.22

W/m2, Gordon et al., 2016).


6

1.3 Aerosol health impact

Figure 1.5: Mortality and SO2 concentrations during the great London smog (adapted from Bell et

al., 2001 and completed with data from the same publication).

With the great London smog in December 1952, the adverse health effects of air

pollution including particulate matter became apparent to a larger public. The great

London smog in December 1952, caused 5000 excess deaths (acute) and as a long-term

effect further 7000 excess deaths between January and March 1953 which could not be

explained by influenza (Bell et al., 2001). Also sensitivity studies on the impact of

influenza underline the fact that only a fraction of excess deaths can be attributed to this

cause (Bell et al., 2004). These findings triggered further research on the health impact

of PM, especially so-called cohort studies assessing the effect of chronic exposure. In

such studies, mortality was found to scale linearly with chronic particle exposure levels

(e.g., a study period of 14-16 years and a PM concentration between 10 and 30 µg/m3 in

Dockery et al., 1993). A reduction of PM levels in the study region led to a lower

mortality (as described in Laden et al., 2006). Such epidemiological studies relate also

7

an increased risk of cardiopulmonary diseases as well as lung cancer to long-term

exposure of PM (Laden et al., 2006, Pope et al., 2006, 2013).

Figure 1.6: Global distribution of mortality attributed to air pollution (from Lelieveld et al., 2015).

On a global scale, ambient air pollution is estimated to have caused 3.7 million

premature deaths in 2012 (WHO, 2013) and in agreement a recent modelling study links

3.15 million deaths to air pollution, mostly to PM2.5 (PM with an aerodynamic

diameter smaller than 2.5 µm) and ozone. These estimates do however not include the

impact of indoor air pollution causing additional deaths of 4.4 million (Lim et al.,

2013). Environmental regulations regarding particulate air pollution aim to limit the

impact of aerosol on human health by establishing limits in the PM10 (smaller than

10µm) and also other size fractions. These limits consider acute and chronic effects by

separately specifying daily and annual limits (Tab. 1.1).

Table 1.1: PM limits established in Switzerland, EU, U.S.A., and WHO. Country/Organization PM10 daily PM10 annual PM2.5 daily

PM2.5 annual *primary/**secondary

Switzerland (FOEN, 2010)

50 20 - -

EU (European Commission, 2010)

50 40 - 25

US EPA, 2012 150 - 35 12*/15** WHO (2006) 50 20 25 10


8

Given the number of fatalities, the link between the toxicity and the chemical

composition and size of the particles is important, but not yet well understood. Exposure

to PM occurs through inhalation, but not all inhaled particles are deposited in the human

body, e.g., more than 80% of the inhaled 300 nm particles are exhaled again. The size of

the particle does not only influence the overall deposition probability but also the

location of the deposition (Fig. 1.7, Maynard and Kuempel, 2005). While coarse

particles are already deposited in the nose region, smaller particles also can travel as far

as to the tracheobronchial region. The even smaller ultrafine particles even penetrate

into the alveolar region and can pass the air-blood barrier (Kelly and Fussel, 2012).

However, even if coarse particles are mostly deposited in the head region, there is

increasing evidence for adverse health effects of such particles (Brunekreef and

Forsberg, 2005). In order to describe the effect of the chemical composition on human

health, the particles’ ability to create oxidative stress and thus to cause inflammations

and cell death is commonly used (Kelly and Fussel, 2012). Therefore, reactive oxygen

species are a focus in current research aiming to understand the health effects of PM

(Seifried et al., 2006). Among such studies, apportioning ROS to its sources is given

special attention (Zhang et al., 2008, Verma et al., 2015). However, the health effects of

single sources remain unclear and especially the impact organic carbon is uncertain due

to its complex composition (Stanek et al., 2011, Kelly and Fussel, 2012).

9

Figure 1.7: Modelled particle deposition probability upon inhalation in the human respiratory

system as a function of particle size and location (from Maynard and Kuempel, 2005).

1.4 Chemical composition and sources

Aerosol sources have distinctly different composition and their contributions to

PM exhibit high temporal and spatial variability which also depends on the assessed

size fraction. This variability influences also the bulk composition of ambient PM (Fig.

1.8, Putaud et al., 2004). PM10 is composed of organic aerosol (OA), black carbon

(BC), and inorganic aerosol (ammonium, nitrate, sulfate, sea salt, mineral dust).


10

Figure 1.8: Aerosol composition for different environments in Europe in PM10 and coarse PM, with unaccounted mass from water and other aerosol constituents which were not analyzed at all sites (adapted from Putaud et al., 2004).

SO2 (emitted by fossil combustion and volcanic eruptions) and dimethylsulfide

(DMS, C2H6S, by marine microorganisms) are oxidized in the atmosphere and form

sulfuric acid. NOx, emitted during combustion processes, similarly undergoes

atmospheric processing leading to the formation of nitric acid. Ammonia, emitted

mainly by agricultural processes (livestock and use of fertilizers, according to

Bouwman et al., 1997), reacts with sulfuric and nitric acid to ammonium sulfate and

nitrate (secondary aerosol formation). In contrast, BC, mineral dust, and sea salt are

emitted as particles (primary aerosol) and OA can be both primary and secondary.

Increasing PM10 and elevated concentrations of combustion related aerosol components

such as BC reflect the anthropogenic influence (through fossil fuel and biomass

combustion) in urban areas (urban and kerbside, Fig. 1.8). A similar behavior can be

observed for the coarse fraction (PM10-PM2.5) at kerbsides, where PMcoarse is

dominated by mineral dust from resuspension caused by traffic while the concentrations

are lower in rural areas or even dominated by sea salt close to the sea (Visser et al.,

2015). However, a large fraction of OM contributing to PMcoarse might stem from

plant abrasion (Andreae et al., 2007).

11

Figure 1.9: NR-PM1 AMS analyses at multiple sites on the northern hemisphere: a) aerosol composition (organic (Org), sulfate, nitrate, ammonium, chloride), b) oxygenated OA concentrations, c) and primary OA factors as HOA and BBOA (from Zhang et al., 2011).

The composition of the non-refractory-PM1 (NR-PM which is smaller than 1 µm)

is commonly analyzed with the aerosol mass spectrometer (AMS, Canagaratna et al.,

2007, described in Chapter 3). The contribution of OA to PM1 is highest (45%), though

also sulfate (32%), ammonium (13%), and nitrate (10%) contribute substantially (Zhang

et al., 2007). The spatial variability of the NR-PM1 composition as well as the

contributing OA sources along a rural-to-urban gradient is displayed in Fig. 1.9. On

average OA contributes 45% to NR-PM1 at the urban sites, 52% downwind from urban

sites, and 43% at rural sites. The organic aerosol is a complex mixture of compounds of

which only a small fraction is identified (Hoffmann et al., 2011). Anthropogenic sources

such as traffic (Schauer et al., 2002a, Platt et al., 2014, Chirico et al., 2011), cooking

(Schauer et al., 2002b, Klein et al., 2016a, 2016b), residential heating using different

fuels such as wood (Schauer et al., 2001, Bruns et al., 2015 and 2016, Heringa et al.,


12

2011) and natural sources such as the terrestrial (Sanchez-Ochoa et al., 2007; Leaitch et

al., 2011, Guenther et al., 1995) and marine environment (Schmale et al., 2013; Zorn et

al., 2008; Crippa et al., 2013c, Seinfeld and Pandis, 2006) contribute to OA by primary

particle emission (POA) as well as the emission of volatile organic compounds (VOC).

VOCs undergo atmospheric aging (radiation, oxidants) leading to the formation of

lower-volatility compounds which subsequently condense on preexisting particles

(SOA). Source apportionment techniques applied to organic mass spectral signatures

from the AMS enable to quantitatively disentangle the contribution of different POA

sources (traffic, cooking, biomass combustion, coal combustion, e.g. in Elser et al.,

2016b) as well as oxygenated OA (OOA) categories typically linked to SOA (Zhang et

al., 2011; Jimenez et al., 2009). POA and SOA levels increase from rural to urban sites

with increasing anthropogenic influence though the relative enhancement for POA is

higher (from POA/OA of 0.1 at rural to 0.42 at urban sites, Fig. 1.9, Zhang et al., 2011).

13

2 Motivation and thesis outline

The atmospheric aerosol impacts climate and human health (IPCC, 2013; Pöschl

et al., 2005). Since the organic aerosol contributes significantly to PM (up to 90% of

PM1, Jimenez et al., 2009), a detailed characterization and a profound understanding of

the contributing sources and formation processes is essential. OA is a complex mixture

of which only a small proportion of all compounds is identified (Hoffmann et al., 2011).

Estimating source contributions has, therefore, been a focus of numerous studies (Zhang

et al., 2011). Such information enables reducing uncertainties in climate modelling and

designing effective mitigation strategies. However, the lack of long-term analyses limits

developing and testing of tropospheric aerosol prediction models. For obtaining a better

understanding of the health effects of OA, among others, long-term source

apportionment analyses in large networks coupled with epidemiological studies will be

of great value.

In earlier studies, different analytical methods were applied on filter samples.

Organic carbon (OC) was separated into a fossil (anthropogenic fossil fuel emissions)

and non-fossil (wood combustion, cooking biogenic, primary biological, etc)

contribution using 14C analyses (Szidat et al., 2006; Cavalli et al., 2010; Zotter et al.,

2014). However, this approach cannot separate the influence of POA and SOA. The

contribution of specific sources is also estimated based on marker concentrations (e.g.

levoglucosan) using scaling factors to obtain the total contribution of the source. This

approach has the drawback that these scaling factors are uncertain. Combining OC/EC

Chapter 2. Motivation and thesis outline

14

analyses with several organic and inorganic markers is also used for statistical

attribution of OC to its sources (e.g. Waked et al., 2014). With this approach only a

small fraction of OA or OC is resolved and the rest apportioned to sources based on

their variability. Thus, sources not represented by the chosen markers are not resolved.

With the invention of the AMS (DeCarlo et al., 2006, Jayne et al., 2000) highly

time-resolved quantitative mass spectral signatures of NR-OA became available for

source apportionment. However, molecular identification is hindered by electron impact

ionization (EI) and the resulting strong fragmentation. With the help of statistical tools

(e.g., positive matrix factorization, PMF, Paatero and Tapper, 1994) and comparison to

emission studies, the influence of different sources could be separated. Numerous

studies (e.g. Fig. 2.1 from Jimenez et al., 2009) provided quantitative estimates of

primary OA source contributions such as traffic related hydrocarbon emissions (HOA),

biomass burning emission (BBOA) as well as of secondary OA (OOA).

Figure 2.1: NR-PM1 mass concentrations (organic, sulfate, nitrate, ammonium, and chloride) measured by AMS at multiple sites. Source apportionment results of OA are displayed: traffic (HOA), other POA as BBOA (other OA), oxygenated OA linked to SOA (OOA), separated into semi-volatile (SV-OOA) and low-volatility (LV-OOA) from Jimenez et al. (2009).

15

However, the AMS is costly and demanding in maintenance and, therefore,

measurement campaigns are typically limited to 1 to 2 months. For this reason, the

aerosol chemical speciation monitor (ACSM, Aerodyne Inc., Ng et al., 2011c, Fröhlich

et al., 2013) was developed for long-term measurements but in order to cover large

networks, still many instruments are required. Both ACSM and AMS are restricted to

the submicron aerosol defined by the particle transmission through the aerodynamic

lens. Recently a new aerodynamic lens was developed accessing particles of up to a size

of 2.5µm (Williams et al., 2013), but PM10 is yet not accessible. This type of

instrument recovers as mentioned quantitative mass spectral signatures of NR-OA but

yet loses a lot of chemical information. For this reason and due to the evolution of the

organic matter to chemically similar composition during atmospheric ageing,

information on the origin of OOA is limited. In contrast to the AMS, laser-

desorption/ionization mass spectrometry fragments the molecules less and, according to

Samburova et al. (2005a), the molecular weight distribution of water extract PM is not

significantly influenced by fragmentation by LDI and thus adding a matrix assisting the

ionization does not lead to a gain in information. The same study also shows the

possibility to analyze PM collected on filters directly without further treatment. On the

other hand, this instrument is not quantitative because of matrix effects, i.e. the

measured intensity of a compound depends not only on its concentration but also on the

concentration of all other present compounds. With this instrument, earlier studies have

shown that it might be possible to distinguish anthropogenic from biogenic SOA

(Baltensperger et al., 2005). Kalberer et al. (2004) discovered the formation of

oligomers to the organic aerosol in a reaction chamber and concluded that such

oligomers are also important in ambient OA. These results indicate that LDI

measurements capture valuable information on the chemical composition of OA.

In this dissertation, we use different approaches to overcome the above

limitations. On one hand, we develop and characterize an offline application of the

AMS based on liquid extracts of PM on filter samples that can cover long-term

measurements in large networks (Chapter 4). This approach is used to quantify the

sources contributing to the organic aerosol in the PM10 size fraction in Switzerland at 9

sites throughout the year 2013 (Chapter 5). On the other hand, we use offline laser-

Chapter 2. Motivation and thesis outline

16

desorption/ionization mass spectrometry for obtaining a better understanding of the

contributing sources and their chemical characteristics. This measurement approach is

described and applied to the same ambient samples as those analyzed by offline AMS

(Chapter 6).

17

3 Methodology

In this dissertation, we characterize the chemical composition of the ambient

organic aerosol using mainly two different mass spectrometers: the Aerodyne aerosol

mass spectrometer and a laser-desorption/ionization mass spectrometer. In this chapter

both instruments are described in more detail along with a detailed explanation of the

statistical technique employed for source apportionment based on the mass spectral

signatures.

3.1 Aerosol Mass Spectrometer

3.1.1 General

The Aerodyne aerosol mass spectrometer (AMS, Fig. 3.1) analyzes the non-

refractory particulate matter (NR-PM) online and provides quantitative measurements

of organics, ammonium, nitrate, sulfate, and chloride, as well as mass spectral

signatures of the organic aerosol and particle size (Canagaratna et al., 2007).

Chapter 3: Methodology

18

Figure 3.1: Scheme of the high-resolution-time-of-flight aerosol mass spectrometer (HR-ToF-AMS) from DeCarlo et al. (2006).

The instrument is composed of three parts: the aerosol inlet, the particle sizing

chamber, and the particle compositon detection section (DeCarlo et al., 2006). PM

suspended in air is sucked through a critical orifice and focused into a particle beam by

an aerodynamic lens and released into the sizing and detection region. The particles are

sized by measuring the particle time-of-flight using a chopper regulating the

transmission of the particle beam with up to 4% transmission rate (chopped). The

background and air signal is separated from the particle signal by either blocking

completely the particle beam (chopper blocked) or transmitting it continuosly (chopper

open). When impacting on the heated tungsten surface (600°C, vaporizer), the particles

are flash-vaporized. Thus refractory material as e.g., black carbon, are not accessible to

the AMS. The vaporized compounds are ionized by electron impact (EI). This hard

ionization pathway causes significant fragmentation. The charged fragments are

transferred by orthogonal extraction to a time-of-flight unit and detected. The AMS has

a mass resolution of 2’500 (at m/z 43, V-mode, Canagaratna et al., 2007) which enables

separating different ions at the same nominal mass (e.g. C2H3O+ and C3H7+ at m/z 43,

Fig. 3.2). With this information the bulk O/C or OM/OC (organic matter to organic

carbon ratio) can be determined.

19

Figure 3.2: Example of mass resolution of 3 AMS types: Quadrupole, C-ToF-MS, HR-ToF-MS from De Carlo et al. (2006).

3.1.2 offline AMS

In several studies, the AMS, developed as an online instrument, is used as an

offline application for analyzing nebulized liquid extracts of filter samples (Lee et al.,

2011; Sun et al., 2011; Mihara and Mochida, 2011). In this study, we further develop

and characterize this method (offline AMS, Chapter 4) and apply it to environmental

filter samples. The technique involves extracting PM in ultrapure water (20 min in

ultrasonic bath at 30°C and 1 min vortexing) and filtering in order to remove suspended

material. The aqueous solution is nebulized, dried and analyzed by an HR-ToF AMS.

The mass spectral signature of the organic fraction represents WSOM which is of

advantage because the signatures can be related to quantitative WSOC analyses which is

not possible using other solvents. Therefore, the non-quantitative organic offline-AMS

signatures can be quantified using the product of the WSOM/WSOC ratio from the

offline-AMS analyses and the water-soluble OC (WSOC) measurements from a total

OC (TOC) analyzer quantitatively oxidizing WSOM to CO2 which is detected by a non-

dispersive infrared (NDIR) detector. In analogy to online AMS OM signatures, the

offline WSOM signatures are used for source apportionment. The characterization of

the method and performance of offline AMS data in source apportionment in


20

comparison to online ACSM data is assessed in Chapter 4. Main advantages and

disadvantages of this method are lined out in Tab. 3.1.

Table 3.1: Advantages and disadvantages of offline AMS technique.

Advantage Disadvantage

• High temporal and spatial

coverage.

• Low time resolution (defined by

filter sampling time).

• Coarse OM also accessible (if water soluble).

• Filter artifacts.

• Analysis of any nebulizable matrix (e.g. lake sediments, snow, rain, soil water, etc.) possible

• External data needed for quantification (WSOC and/or OC).

• Complementary analysis a posteriori with required external data to study the aerosol sources.

• Low recoveries of less water-soluble sources and thus large uncertainties (as e.g. for traffic emissions).

3.2 Laser-Desorption/Ionization mass spectrometer

Laser-desorption/ioniziation mass spectrometry uses a laser for desorbing liquid

or solid material and producing ions (De Hoffmann and Stroobant, 1999). The laser

emits pulses directly focused on the sample surface with a defined frequency. These

laser pulses create first generation ions and neutral molecules by ablating material from

the sample surface. In the dense plume, ions and neutral molecules react and form

second and further generation ions. Even if the ionization pathway is softer than EI (as

mentioned above), the formed ions are typically fragments of the molecules (De

Hoffmann and Stroobant, 1999). The created ions are in a next step analyzed by a time-

of-flight mass spectrometer. In this study, we use an LDI (Shimadzu Axima

Confidence, Shimadzu-Biotech Corp., Kyoto, Japan) equipped with an N2 laser (λ=337

nm) and a time-of-flight mass spectrometer (axial, without orthogonal injection). Laser-

based ionization instrumentation is known to be affected by matrix effects (Ellis et al.,

2014, Borisov et al., 2013). This means that the recorded intensity of a molecule does

not only depend on its concentration but also on the concentration of all other

compounds present. In complex mixtures, such as ambient PM, not all of these

21

dependencies can be understood. Therefore, this type of analysis is not considered

quantitative. However, using online laser-based instrumentation (ATOF-MS), good

correlations with reference analyses analyses could be achieved for OA and different

inorganic components (Healy et al., 2013).

Earlier studies showed that PM collected on filter samples can be analyzed

without matrix addition (Samburova et al., 2005a) and related this observation to the

possible presence of compounds acting as a matrix in PM. In this study, we apply LDI

measurements directly to PM collected on quartz filter samples in order to avoid losses

that may occur during an intermediate extraction step. Prior to the analysis, aqueous

solution of silver nitrate (SN) is applied to fractions of the sample as an internal

standard for m/z calibration. The dry sample on a stainless steel plate is inserted into the

LDI and analyzed. In Chapter 6, the method development and characterization is

described as well as an application to ambient samples. Table 3.2 presents main

advantages and disadvantages of the offline LDI approach.

Table 3.2:Advantages and disadvantages of LDI technique.

Advantage Disadvantage

• Softer ionization than EI. • High temporal and spatial

coverage.

• Not quantitative. • Matrix effects. • Unknown response factors of

aerosol components. • Low time resolution (defined by

filter sampling time).

• Coarse OM also accessible. • Filter artifacts. • Analyzing any material responding

to laser. • External data needed for

quantification (WSOC and/or OC). • Complementary analysis a

posteriori with required external data to study the aerosol sources.

3.3 Source apportionment

OA is a complex mixture with a multitude of contributing primary and secondary

aerosol sources. Source apportionment aims to identify and quantify sources


22

contributing to OA. Bilinear receptor models, e.g. positive matrix factorization (PMF,

Paatero and Tapper, 1994), are widely used to apportion OA to its sources. PMF

explains the variability in a chosen dataset (x) with linear combinations of factor

contributions varying over time (g) and constant factor profiles (f). The index i

represents the measurement time, j the m/z or ion and k the factor number and the model

residuals are described by eij.

(3.1)

The entries of g and f are constrained to be positive, as also the name PMF

specifies. In OA applications based on AMS data, x reflects measured mass spectra, f

the chemical composition of the source, g the time series of the source and eij not

resolved/residual mass. In solving Eq. 3.2, also the measurement uncertainties (sij) of

the mass spectra are considered by minimizing eij/sij, through the object function Q:

(3.2)

PMF can be implemented by the multilinear engine 2 (ME-2, Paatero, 1999)

which allows exploration of the solution space. This is possible through using a priori

information on e.g. chemical composition of sources. In the so called a-value approach,

it is possible also to allow the solver to deviate from the a priori information defined by

the relative deviation, a (Eq. 3.3).

(3.3)

In Eq. 3.3 n and m represent any two entries (m/z, ion) in the constrained factor

profile. Different representations of x by PMF, rotations of the linear combinations of f

and g, can have a similar mathematical quality (Q) which is referred to as rotational

ambiguity. Therefore, it is crucial to explore the solution space in order to estimate the

model uncertainty. Uncertainty in PMF results also arise from random fluctuations in

the input data (measurement), which is termed statistical uncertainty. Resampling the

23

input matrices in order to assess the influence of the choice of the model input on the

PMF solution, called bootstrapping, is an efficient approach to estimate the statistical

uncertainty (Paatero et al., 2014, general atmospheric description in Wilks, 2006).

Other statistical tools like principal component analysis (PCA, described in detail

for example in Wilks, 2006), chemical mass balance (CMB, Watson et al., 1998),

cluster analysis (detailed in Wilks, 2006) have similar goals. CMB requires a priori

information on the source profiles (chemical signature) and their uncertainty (Watson et

al., 1998). In general, whenever a priori information is needed, knowledge on the related

uncertainty is crucial. In contrast, PMF does not require a priori information on the

chemical source signature although such information can be used optionally. Moreover,

PMF uses also the measurement uncertainty (see Eq. 3.2). Thus datapoints are given

weight related to their respective uncertainty in the partial least square method. Thereby,

especially outliers with a high uncertainty are given less weight. In comparison to PMF,

CMB requires smaller datasets since it can be theoretically solved for any single

datapoint. CMB, PMF, and PCA separate the influence of different sources rather than

grouping similar spectra as for example cluster analysis (Wilks, 2006). Unlike CMB,

PMF (without a priori information) is limited in separating highly correlated time

sources. However, by using a priori information (as possible using ME-2 as a PMF

implementation), separating such sources is facilitated (Paatero, 1999). As stated earlier,

PMF outputs are constrained to be positive which makes the factorizations (g) directly

interpretable as concentration. This is not the case for other related techniques such as

PCA. Also PCA assumes orthogonal components/factors which is hindering the

separation of highly correlated components, and negative values also hinder the

interpretation in an environmental sense as concentrations. However, PCA is of great

value to assess the influence of environmental parameters as e.g., temperature on PM

levels. CMB is traditionally used to separate different POA and SOA was treated as a

residual. Nevertheless, there were some attempts to use similar approaches using

markers of secondary products of defined precursors to estimate their SOA contribution,

e.g., aromatics, alpha-pinene, isoprene (Kleindienst et al., 2007). Moreover, when

performing PMF using AMS data, SOA is separated by its variability and not modelled

as a residual because the bulk chemical composition and not only selected markers are

used. However, the use of markers can also be advantageous for resolving correlating

POA due to of their high specificity.


24

25

4 Characterization and source

apportionment of organic aerosol using offline aerosol mass spectrometry

Kaspar R. Daellenbach1, Carlo Bozzetti1, Adela Křepelová1, Francesco Canonaco1, Robert Wolf1, Peter Zotter1,a, Paola Fermo4, M. Crippa1,b, Jay G. Slowik1, Yuliya Sosedova1, Yanlin Zhang1,2,3,c, Ru-Jin Huang1, Laurent Poulain5, Sönke Szidat2, Urs Baltensperger1, Imad El Haddad1, and André S. H. Prévôt1 1Laboratory of Atmospheric Chemistry, Paul Scherrer Institute, 5232 Villigen, Switzerland 2Department of Chemistry and Biochemistry & Oeschger Centre for Climate Change Research, University of Bern, 3012 Bern, Switzerland 3Laboratory of Radiochemistry and Environmental Chemistry, Paul Scherrer Institute, 5232 Villigen PSI, Switzerland 4Department of Chemistry, University of Milan, 20133 Milan, Italy 5Leibniz Institute for Troposphärenforschung, Leipzig, Germany anow at: Lucerne School of Engineering and Architecture, Bioenergy Research, Lucerne University of Applied Sciences and Arts, 6048 Horw, Switzerland bnow at: EC Joint Research Centre, Institute for Environment and Sustainability, 21027 Ispra, Italy cnow at: Yale-NUIST Center on Atmospheric Environment, Nanjing University of Information Science and Technology, Nanjing 10044, China

Pulished in Atmospheric Measurement and Technique DOI: 10.5194/amt-9-23-2016

Chapter 4: Characterization and source apportionment of OA using offline AMS

26

Abstract

Field deployments of the Aerodyne Aerosol Mass Spectrometer (AMS) have

significantly advanced real-time measurements and source apportionment of non-

refractory particulate matter. However, the cost and complex maintenance requirements

of the AMS make its deployment at sufficient sites to determine regional characteristics

impractical. Furthermore, the negligible transmission efficiency of the AMS inlet for

supermicron particles significantly limits the characterization of their chemical nature

and contributing sources. In this study, we utilize the AMS to characterize the water-

soluble organic fingerprint of ambient particles collected onto conventional quartz

filters, which are routinely sampled at many air quality sites. The method was applied to

256 particulate matter (PM) filter samples (PM1, PM2.5, and PM10, i.e., PM with

aerodynamic diameters smaller than 1, 2.5, and 10 µm, respectively), collected at 16

urban and rural sites during summer and winter. We show that the results obtained by

the present technique compare well with those from co-located online measurements,

e.g., AMS or Aerosol Chemical Speciation Monitor (ACSM). The bulk recoveries of

organic aerosol (60-91 %) achieved using this technique, together with low detection

limits (0.8 µg of organic aerosol on the analyzed filter fraction) allow its application to

environmental samples. We will discuss the recovery variability of individual

hydrocarbon ions, ions containing oxygen, and other ions. The performance of such data

in source apportionment is assessed in comparison to ACSM data. Recoveries of

organic components related to different sources as traffic, wood burning, and secondary

organic aerosol are presented. This technique, while subjected to the limitations inherent

to filter-based measurements (e.g., filter artifacts and limited time resolution) may be

used to enhance the AMS capabilities in measuring size-fractionated, spatially resolved

long-term data sets.

4.1 Introduction

Aerosols affect climate, air quality, ecosystems, and human health (Kelly et al.,

2012; Griggs and Noguer, 2002). Organic aerosol (OA), a significant fraction of the dry

particle mass (Jimenez et al. (2009) and references therein), is either directly emitted

(primary organic aerosol, POA), or formed in the atmosphere through gas-phase

oxidation of anthropogenic and biogenic volatile organic compounds and subsequent

27

condensation or nucleation (secondary organic aerosol, SOA). Characterization of OA

chemical composition and sources is necessary for understanding the corresponding

atmospheric processes and mitigating the adverse effects of aerosols. Previous studies

have shown that OA contains a variety of organic species, including hydrocarbons,

alcohols, aldehydes, and carboxylic acids. However, only about 10- 30 % of OA has

been chemically speciated so far (Hoffmann et al., 2011).

The High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS,

Aerodyne Research, Inc.) has been widely used for characterizing OA in both field and

laboratory studies. This instrument couples thermal vaporization with electron

ionization (EI, 70 eV) and provides quantitative mass spectra of non-refractory aerosol

components including OA, NH4+, NO3

- , SO42-, and Cl-. Application of advanced factor-

based receptor models such as positive matrix factorization (PMF, Paatero and Tapper,

1994) to these spectra has been proven effective in apportioning OA into different

factors (e.g. Lanz et al., 2007,2010; Jimenez et al., 2009; Zhang et al., 2011; Ng et al.,

2010b; Crippa et al., 2014). These factors are subsequently related to primary sources

like biomass burning (BBOA), traffic (HOA), and cooking (COA), as well as

oxygenated organic aerosol (OOA), which is often attributed to SOA.

The cost and complex operation required by the AMS makes its simultaneous

long-term deployment at many sites impractical. Consequently, current data sets are

typically limited to few weeks and specific sites or measurements from mobile

platforms (Mohr et al., 2011). Recently, a robust, less expensive, Aerosol Chemical

Speciation Monitor (ACSM, Ng et al., 2011c) and a time-of-flight aerosol chemical

speciation monitor (ToF-ACSM, Fröhlich et al., 2013, 2015) were developed to

overcome some of these shortcomings. However, the low mass resolution of these

instruments reduces their utility. Meanwhile, other studies have proposed the use of the

HR-ToF-AMS for the analysis of aqueous or organic solvent extracts of filter samples,

which are already routinely collected at many sites worldwide, offering a greater

coverage than with ACSMs (Mihara and Mochida, 2011; Lee et al., 2011; Sun et al.,

2011). While such methodologies may greatly extend the ability of the AMS to measure

spatially re-solved long-term data sets, the results obtained only pertain to a sub-fraction

of the total organic aerosol and are subject to inherent artifacts of filter-based


28

measurements. It is not clear whether this fraction adequately reflects the chemical

nature of the entire bulk OA and whether these results may be used for OA source

apportionment. Here, we have adopted such an approach based on measurements of the

water-soluble organic fraction. We present a methodology to generalize the results to

bulk OA, based on the analysis of 256 filter samples from 16 urban to rural sites during

different seasons and its comparison to online measurements. These results are expected

to significantly broaden the spatial, temporal, and particle size ranges accessible to

AMS measurements of organic aerosol.

Table 4.1: Filter samples and available supporting measurements used in this study. Location Campaing period Sampling

period (h)

Samples Size Supporting measurements

Zürich

(urban

background)

Apr 2011 12 11 PM1 PM2.5 fingerprints and OA,

SO42-a, gas-phase

measurements (CO)

Aug 2008-

Jul 2009

24 42 PM10 OC/EC, ions

Feb 2011-

Feb 2012

24 41 PM10 PM1 fingerprints and OA,

SO42-b, eBC, WSOC

Paris (urban core) Jul 2009

Jan-Feb 2010

12 12 PM1 OC/EC, PM1 fingerprints

and OA, SO42-a

15 NABEL

stations

in Switzerlandc

Dec 2007-

Feb 2008

Dec 2008-

Feb 2009

24 150 PM10 OC/EC, ions

aHR-ToF-AMS; bQuadrupole ACSM; cNABEL (Swiss National Air Pollution Monitoring Network): the stations represented in the study are Basel, Bern, Chiasso, St. Gallen, Magadino, Massongex, Moleno, Payerne, Reiden, Roveredo, Sissach, Solothurn, San Vittore, Vaduz, and Zürich Zotter et al. (2014).

4.2 Methods

4.2.1 Aerosol sampling

Particulate matter (PM) of different sizes (PM1, PM2.5, and PM10, representing

aerodynamic particle sizes smaller than 1, 2.5, and 10 µm, respectively) were collected

onto pre-heated (800 °C, 12 h) Pall quartz filters (diameter 14.7 cm) using HiVol

samplers (500 Lmin-1). Field blanks were collected using the same method as for the

29

exposed filters. The filters were stored in sealed bags at 18 °C and only transported

cooled. Before handling the filters, they were left at room temperature in the sealed bags

in order to avoid condensation of volatile compounds on the cold surface (> 15 min).

The offline AMS analysis of the filters collected in April 2011 in Zürich were

conducted in October 2011. The other filters were analyzed between April and October

2012. While samples were collected at different seasons at 16 sites including urban,

suburban, and rural sites (Table 4.1), we will mainly focus on the Zürich data sets

(filters evenly distributed in the years 2011-2012) because of the extensive supporting

measurements performed there. Measurements on the remaining samples are used for

the assessment of the bulk OA water solubility.

The urban background site Kaserne in Zürich is located in a park in the middle of

the urban core of a densely populated area (1.2 million inhabitants, including

surrounding communities). In addition to filter sampling, an Aerosol Chemical

Speciation Monitor (ACSM) was operated with a PM1 (standard) aerodynamic lens in

Zürich from February 2011 to February 2012 (Canonaco et al., 2013, 2015, 2018). The

ACSM provides quantitative unit mass resolution (UMR) mass spectra with a time

resolution of 30 min. These mass spectra can be used to determine the concentration of

species such as OA and SO42- , while the OA mass spectra are suitable for source

apportionment (Ng et al., 2011b). At the same site, equivalent black carbon (eBC) was

monitored using an aethalometer, AE 31 (Magee Scientific Inc.) (Hansen et al., 1984;

Herich et al., 2011), and CO by non-dispersive Fourier-transform infrared spectroscopy

(APNA 360, Horiba, Kyoto, Japan). During spring 2011, PM2.5 filter samples were

also collected and a HR-ToF-AMS equipped with a (PM2.5) lens (Williams et al., 2013)

was operated at the same site.

During the winters of 2007/08 and 2008/09, offline AMS measurements (PM10)

were conducted for 15 sites spread over Switzerland including flatland and alpine sites

with varying population density and local emissions. A yearly cycle (August 2008-July

2009) from the urban background station in Zürich described above completes this data

set. For this campaign, measurements of organic carbon (OC), elemental carbon (EC)

(Zotter et al., 2014) and the most com-mon ions are available. Finally, we have also

analyzed 12 filter samples collected in Paris during summer 2009 and winter 2010,


30

where concomitant online HR-ToF-AMS PM1 measurements are available (Crippa et

al., 2013a,b,c; Freutel et al., 2013).

4.2.2 Offline AMS

4.2.2.1 Sample extraction

Sample fractions (2 cm2 or 1.2 % of the entire filter sample, may be increased for

low filter loadings) are collected from each filter sample and extracted in 10 mL

ultrapure water (18.2 MΩ cm, total organic carbon .TOC/ < 5 ppb, 25 °C) by means of

an ultrasonic generator for 20 min at 30 °C. Samples are then briefly vortexed (1 min),

to ensure their homogeneity. The extracts are subsequently filtered with 0.45 µm nylon

membrane syringe filters, prior to AMS analysis.

4.2.2.2 Offline AMS analysis

The water extracts are aerosolized using a custom-built nebulizer designed to

work with small liquid volumes (5-15 mL). When passing through the nebulizer nozzle,

an air stream is accelerated. Simultaneously, liquid is sucked into the nebulizer. The

high velocity air stream breaks up the solution and forms particles. The resulting

particles are dried by a silica gel diffusion dryer, and subsequently analyzed by the HR-

ToF-AMS (V-mode). For each sample, spectra are recorded in the range of 12-300 amu,

with a collection time for each spectrum of 30-60 s. To reduce memory effects,

ultrapure water is nebulized before every sample measurement. This information is used

as a system blank. Raw data depicting the measurement procedure are presented in Fig.

4.1 Field blanks are analyzed using the same procedure as the sample filters, and the

retrieved signals are statistically equal to those obtained from the direct nebulization of

ultrapure water. During each experiment, the nebulizer air is also filtered and measured

with the AMS to remove gas-phase contributions from the mass spectra (Allan et al.,

2004).

The HR-ToF-AMS operating principles, calibration procedures, and analysis

protocols are described in detail else-where (DeCarlo et al., 2006). The instrument

provides quantitative mass spectra of non-refractory PM1 (vacuum aero-dynamic

diameter (Dva) 60-600 nm) components, at 600 °C and 107 Torr (1.3×105 Pa). These

include organic aerosol and ammonium nitrate and sulfate. Data are analyzed using

31

high-resolution analysis fitting procedures, Squirrel v1.52L (SeQUential Igor data

RetRiEvaL) and Pika v1.10C (Peak Integration by Key Analysis, D. Sueper), in the

IGOR Pro software package (Wavemetrics, Inc., Portland, OR, USA).

4.2.2.3 Other chemical analysis

Cations (e.g., K+, Na+, Mg2+, Ca2+, NH4+) and anions (e.g. SO4

2-, NO3-, Cl-) were

analyzed using an ion chromatographic system (850 Professional, Metrohm, Switzer-

land) equipped with a Metrosep C4 cation column and a Metrosep A anion column,

respectively. For this analysis, 1 cm2 filter fractions were extracted in 15 mL ultrapure

water (18.2 M • cm). Filters (1.5 cm2) were also analyzed for EC and OC content by a

thermo-optical transmission method on a Sunset OC/EC analyzer (Birch and Cary,

1996), following the EUSAAR-2 thermal-optical transmission protocol (Cavalli et al.,

2010). Replicate analysis shows a good analytical precision with relative standard

deviations of 7.7, 14.8, and 8.1 % for OC, EC, and TC (total carbon), respectively

(Zotter et al., 2014). The water-soluble organic carbon (WSOC) estimates from the

offline AMS analyses are compared to WSOC measured using a standard method.

Following this method, filter samples are extracted in ultrapure water, they are gently

shaken for 24 h, and the extracts are subsequently analyzed with a TOC analyzer. The

bulk of these offline measurements are used as reference methods to assess the offline

AMS approach.

4.2.2.4 PMF using ME-2

The ability of the offline AMS analysis to characterize the organic aerosol sources

compared with other online techniques (i.e., ACSM) is evaluated by analyzing the

obtained mass spectra from online and offline measurements using positive matrix

factorization (PMF, Paatero and Tapper, 1994) for the case of the yearly cycle from

Zürich (2011-2012). PMF is a bilinear unmixing receptor model used to describe

measurements (in this case AMS or ACSM organic mass spectra time series) as a linear

combination of static factor profiles and their time-dependent source contributions, as

expressed in Eq. (4.1):

(4.1)


32

Here xij, fkj, gik, and eij are matrix elements of the measurement, factor profile,

factor time series, and residual matrices, respectively. The subscript j corresponds to a

measured ion or m/z, i corresponds to a measured time stamp, and k to a discrete factor.

The user determines the number of factors, p, returned by the PMF algorithm. PMF

requires non-negative entries for fkj and gik, suitable for environmental measurements

such as OA mass concentrations. The PMF algorithm solves Eq. (4.1) by iteratively

minimizing the object function Q, defined as

(4.2)

where σij are the elements of the error matrix (measurement uncertainties), which

together with xij and p are provided as model inputs. Measurement uncertainties

considered in the error matrix include electronic noise, ion-to-ion variability at the

detector, and ion counting statistics (Allan et al., 2003). For offline AMS analyses, both

sample and blank uncertainties are incorporated. Following the recommendation of

Paatero and Hopke (2009) variables with low signal-to-noise (SNR < 0.2) are removed

(no variables affected), whereas “weak” variables (0.2 < SNR < 2) are downweighted

by a factor of 3 (26 variables in the PMF input affected). Further, 19 variables were not

considered in PMF because they were not present in the reference spectra used.

In this study, PMF is solved using the multi-linear engine (ME-2) (algorithm

Paatero (1999) and references therein), with the toolkit Source Finder (SoFi version 4.7,

Canonaco et al., 2013) for IGOR Pro (Wavemetrics, Inc., Portland, OR, USA)

employed as a front end for the model. PMF was operated using the robust outlier

treatment mode, in which outliers were dynamically downweighted. Most published

analyses using PMF are limited in their ability to explore rotational ambiguity in the

solution space, which is typically accessible only in a single, random dimension (Zhang

et al., 2011). As a consequence, these analyses do not guarantee the access to

environmentally optimal solutions. In contrast, the ME-2 implementation of PMF

allows efficient exploration of the entire solution space, including approximate matrix

rotations. In the present study, the solutions were directed towards environmentally

meaningful rotations by constraining the elements of one or more profiles in the factor

33

profiles matrix (fkj) to a predetermined range defined by a centervalue (fkj’) and a scalar

defining the width of range (a), such that the returned profile satisfies fkj=fkj’±a×fkj’.

This approach has previously been utilized for AMS data sets to separate distinct

sources with correlated mass spectra profiles or time series (Lanz et al., 2008; Crippa et

al., 2014) and shown to provide improved factor separation compared to conventional

PMF (Canonaco et al., 2013).

In the case of the offline AMS, the HR data matrices were arranged as follows: in

the measurement matrix, each filter sample is represented by on average eight high-

resolution mass spectra (see description above and Fig. 4.1), corrected for the

corresponding average blank measured before the sample. Each mass spectrum is

composed of 154 HR ions (m/z 12-96). 41 samples were considered in this analysis

(total of 41 time points and matrix total dimension of 334×154=51 436). The

corresponding error matrix has the same dimensions. The elements of the error matrix,

σij, include the uncertainties related to the AMS measurements as discussed above

(computed according to Allan et al., 2003; Ulbrich et al., 2009), denoted by δij, added in

quadrature to the variability of the preceding blank βij, which includes the AMS

measurement precision but also accounts for possible drifts in the nebulization:

(4.3)

In order to allow comparisons with external data, the offline AMS data and error

matrices are converted to ambient concentrations. The contribution of δij and βij to σij

depends on the ion in question, but in general δij dominates (98 %, first and third

quartiles of 89 and 100 %). Since the measured data points are not averaged prior to the

ME-2 analysis, but rather used individually, their variability is not included in the error

matrix, but instead directly reflected in the results. This also provides a metric for the

mathematical stability of the ME-2 solution and thus a part of the uncertainties of the

source apportionment results.

In order to assess the performance of offline AMS data in source apportionment

we compare the obtained results to source apportionment results using online ACSM


34

data. For an ideal offline/online comparison, the online data set should resemble the

offline data set as closely as possible. However, the low mass resolution of the ACSM

spectra prevents satisfactory factor resolution, when using 24 h averages of ACSM data

for the selected days. Moreover, running ME-2 on the selected days for the entire year

(retaining 30 min time resolution) results in biases between winter and summer residual

distributions, which was not the case for the offline data. That is, the model tends to

explain the diurnal variation of the online ACSM data, rather than seasonal differences.

For these reasons, a rolling window ME-2 approach was developed to perform source

apportionment analysis on the yearly online UMR ACSM data (Zürich 2011-2012)

(Canonaco et al., 2018). The approach can be described as a controlled bootstrap

technique applied to sorted data, which would help represent summer and winter data

and provide an estimate of the uncertainties (Paatero et al., 2014). In this approach, a

rolling window is capable of capturing seasonal variations in the aerosol factors and/or

variations driven by meteorology. Within a window, which is considerably shorter than

the yearly data set, the ME-2 model is applied, allowing the factors to adapt to the

measured data. A rolling window corresponds to 4 weeks of measurements and rolls

over the whole set of data with a 1 day time step. The PMF window was rolled over the

temperature-sorted Zürich data (by daily average temperature). By sorting the data with

respect to temperature, days with similar conditions in terms of SOA formation and

dominant primary sources (e.g., BBOA at lower temperature) are grouped together. For

every window the solution was optimized using criteria based on correlations between

the time series and the diurnal cycles of the factors and those of the markers. This novel

approach was compared to classical source apportionment results for the winter part of

this data set presented in Canonaco et al. (2013). The rolling window solution presents

an improved representation of OOA (R2 with NH4+ 0.69 vs. on average 0.53 for the

PMF solution in Canonaco et al., 2013) for the overlapping period, which is consistent

with the variable character of OOA. The correlation of HOA and BBOA with their

respective markers is comparable to Canonaco et al. (2013). For the reasons described

above and with the lack of standard techniques to apply PMF to yearly organic mass

spectral data, the rolling window source apportionment results are chosen as reference.

35

4.3 Results and discussion

4.3.1 Signal-to-noise, quantification and detection limits

Figure 4.1: Data recorded with HR-ToF-AMS of filter samples collected in Zürich (2011-2012).

Data from a typical measurement cycle are underlaid in gray. (a) Raw signals obtained for organic aerosol (OA, green), nitrate (NO3

- , blue), sulfate (SO42-, red), and ammonium (NH4

+, orange), where AMS filter air as well as blank and sample measurements are indicated. (b) OA average signal for samples and blanks (logarithmic scale), blank correction curve and the noise (smoothed standard deviation of the blank) associated with the signal of different species used for the calculation of errors. On the y axes, a.u. denotes arbitrary units.

Figure 4.1a shows a typical time pattern of OA, NO3-, SO4

2-, and NH4+ from

offline AMS measurements. The signal intensity of offline AMS measurements can be

expressed in μgm-3 (of nebulized aerosol), but for simplicity we denote this as arbitrary

units (a. u.) to avoid confusion with concentrations in ambient air (μgm-3). This

conversion between AMS signals and real concentrations is explained below. The

intensity of OA is typically 1-2 orders of magnitude higher than that of the

measurement blanks (see Fig. 4.1b). The blank offline AMS signal is typically below

2.1 a. u., with interday and intraday variation (standard deviation) of 0.3 and 1.5 a. u.,


36

respectively. The nebulization efficiency assessedbased on the SO42- signal is

3.8mLsolution mair-3 (first and third quartile of 1.2 and 7.3 mL solution mair

-3). The

particles generated with this nebulizer have a mode diameter of ~200 nm (dV /dlogDp).

The SO42- detected by the offline AMS is related to SO4

2- loadings on the filter

area (calculated from ACSM measurements) analyzed by a power relationship (Fig.

4.2). However, the offline AMS measurements described herein cannot be directly

quantified, without external measurements of e.g., OC, due to variability in the

nebulization process. Another significant source of uncertainty is the ACSM cutoff

(Dva 600 nm).

Figure 4.2: Offline AMS SO4

2- blank corrected concentrations compared to theoretical SO42-

loadings of the filter fractions (μg). The theoretical SO42- loadings are calculated based on ambient

SO42- concentrations measured by the ACSM for the Zürich yearly cycle and the volume of air sampled

through the analyzed filter fraction. Results are fitted using a power function (ln(y) = 2.3 × ln(x) - 5.2).

37

The detection limit (dlj) of species j (e.g., SO42-), in μg on the analyzed filter

fraction, is evaluated based on the blank variability of j in comparison to the signal in

the sample, σblank,j . We define dlj as the mass of j in μg required to produce a signal

equal to 3 σblank,j. dlj is inferred using the existing relationship between blank corrected

offline AMS signals, Ij in a. u., and the mass concentration of j , Mj in μg, in the

analyzed filter fractions (e.g., Fig. 4.2). We estimated the detection limits for OA and

SO42- as 0.80 and 0.25 μg on the analyzed filter area, corresponding to 80 and 25 μgL-1,

respectively.

4.3.2 OA recovery

The loss of hydrophobic or volatile organic species during sample collection,

handling, extraction, and nebulization may significantly hinder the applicability of the

offline AMS technique. In the following, the organic aerosol signals are normalized to

the sulfate mass, in order to evaluate OA recoveries. This is based on the assumption

that sulfate is quantitatively extracted and measured by the AMS, which is expected

since sulfate is mostly bonded to ammonium (watersoluble and non-refractory). This is

not given at all sites, e.g., in the strong presence of potassium. We also assume that the

fractional composition in the size range sampled by both ACSM and filter samples is

the same. Accordingly, the comparison of Ij and Mj both normalized to SO42- yields the

recovery Rj:

(4.4)

The extraction time does not have a statistically significant effect on OA/SO42-

ratios and fingerprints when increasing the extraction time from 20 to 60 min. Likewise,

multiple extractions did not significantly enhance the recovery of the particulate

compounds as the OA signal from the second extraction was below 8% of that from the

first extraction and only 3 times higher than the blank signal. Therefore, we have

concluded that a single extraction step was sufficient in our case to obtain the water-

extractable material.


38

Figure 4.3: Estimated recoveries of organic compounds based on the comparison of OA/SO4

2- ratios using the offline AMS method to reference measurements for different days. The error bars represent the variability of the offline OA/SO4

2- ratio within a sample and were obtained from different runs during the same measurement of the same sample. (a) The reference OA/SO4

2- ratio is obtained by OC filter measurements (Sunset OC/EC analyzer) scaled to OA using OM/OC values from the HR offline AMS data and SO4

2- from IC. (b) OA/SO42- ratios from online measurements were used as

reference values. For both Paris campaigns and the Zürich spring campaign, the online measurements were conducted using HR-ToF-AMS and for the yearly cycle in Zürich by a quadrupole ACSM. (c) For Zürich (2011-2012), probability density functions of Rbulk are presented both using the offline AMS measurements as well as using WSOC from the Sunset OC=EC Analyzer (in combination with OM=OC ratios from offline AMS).

39

We have evaluated the recovery of the bulk OA (Mj here representing OA), by

comparing the offline AMS OA/SO42- ratios with OA/SO4

2- from reference

measurements using the Sunset OC/EC analyzer and ion chromatography (IC)

(described in Sect. 2.3). The recovery of complex mixtures such as ambient OA

depends on the water solubility of its numerous compounds. Figure 4.3a compares

offline AMS and reference measurements for 15 stations in Switzerland where the

reference measurements were performed on the same filters (150 PM10 samples, Table

4.1), using IC and the Sunset OC/EC analyzer for SO42- and OA measurements,

respectively. The latter were calculated by multiplying the Sunset OC/EC analyzer OC

with the OM/OC ratios from the HR analysis of the AMS spectra. While we

acknowledge that also OM/OC from offline AMS is subjected to errors caused by

compound-dependent extraction efficiencies and filter sampling artifacts, such errors do

not significantly affect the results and the OM/OC range found here (median of 1.84,

first quartile of 1.80 and third of 1.87) compare well with previously measured online

ratios (e.g., 1.80 provided by Favez et al. (2010) for Grenoble, January 2009, 1.66 by

Crippa et al. (2013a) for Paris, and 1.6 and 2.0 by Minguillón et al. (2011) for Barcelona

and Montseny, respectively). From this, we estimate a median Rbulk of 0.60 (first and

third quartiles of 0.49 and 0.80), which suggests that the technique can capture a large

part of the organic fraction.

Similar comparisons between offline AMS results and reference measurements

were also performed for other data sets where online AMS data were available (Zürich

spring and Paris campaigns). Offline AMS measurements of filter samples collected in

Paris (summer 2009 and winter 2010) and Zürich (spring 2011) were compared with

online HR-ToF AMS with the same size cutoff (PM1) (Fig. 4.3b). For these data sets,

OA recoveries range between 64 and 76 %. For Zürich, it should be noted that PM1

filter samples are not available and therefore offline PM10 HR-ToF-AMS

measurements are compared to online data from ACSM (yearly cycle) and HR-ToF-

AMS equipped with a PM2.5 lens (spring). We show that for both campaigns the

overall Rbulk is in the same range as values obtained for the other data sets inspected

here, despite the potential contribution of coarse mode OA (median = 0.91; first and

third quartiles of 0.66 and 1.32, respectively for the yearly cycle and median = 1:05;

first and third quartiles of 0.99 and 1.26, respectively for the spring campaign). This


40

implies that the contribution of the latter is not dominant, consistent with previous

measurements at this site, suggesting that the fine particle mass constitutes on average

75 % of the PM10 mass (Putaud et al., 2010). Note that outliers in Rbulk higher than 1

are associated with very low, and therefore highly uncertain sulfate concentrations. For

the Zürich yearly cycle campaign (2011-2012), we validated the Rbulk calculation

approach adopted here to a more conventional approach for the determination of WSOC

(Fig. 4.3c; water-soluble organic aerosol, WSOA = WSOC × (OM/OC)offline AMS). We

show that both approaches give similar estimates (based on the WSOC median Rbulk =

0.74 compared to Rbulk = 0.91 if the calculation is based on the ACSM), suggesting that

offline AMS measurements are related to WSOA and that a great part of the organic

mass is accessible by the analysis procedure followed here.

4.3.3 Mass spectral analysis

Results above raise the question as to whether the offline AMS analysis maintains

the mass spectral signature of the ambient OA. We have addressed this question by

comparing online and offline OA mass spectra in Fig. 4.4, illustrating an example of the

results obtained from Zürich winter and spring campaigns. Such a comparison

implicitly assumes that the mean organic composition across the entire size range

collected by the filter (up to PM10) is identical to that of the approximately 60-600 nm

particles measured by the online ACSM. Although this assumption will not hold for all

conditions, the comparison is nonetheless useful for characterization of the offline AMS

technique. The comparison of offline and online spectra shows a high correlation (R2 >

0.97) irrespective of the seasonal variation in aerosol composition. More importantly, it

can be observed that this method is also able to capture variations in the aerosol

fingerprints between the two seasons. For instance, both online and offline methods

show higher contributions from BBOA and HOA-related fragments (e.g., m/z 60, 73,

and m/z 55, 57, 69, respectively) for the winter samples and higher contributions from

OOA-related m/z values (e.g., m/z 43, 44) for the spring sample.

41

Figure 4.4: Comparison between 24 h average online and offline AMS (both PM2:5) spectra for

winter (a) and spring (b) samples, collected in Zürich. Fragments (m/z) commonly considered as source-specific markers are explicitly labeled with their nominal mass.

HR-ToF-AMS data enable the analysis of individual ions at the same integer m/z,

which in turn provides better assessment of the recovery of the initial parent organic

compounds or ion families. For this analysis, we use Eq. (4.4) to describe the recovery

of individual ions Rfrag with Mj and Ij defined as the concentration of an individual ion

(cfrag). We have grouped the fragments into five different families, based on their

heteroatom content and degree of unsaturation, including N-containing hydrocarbon

ions (CHN), monooxygenated (CHOz=1) and poly-oxygenated (CHOz>1) ions and pure

hydrocarbons (CH) divided into saturated (CHsat) and unsaturated hydrocarbons

(CHunsat)

Figure 4.5 presents Rfrag for the Zürich spring (2011) campaign (see Fig. 4.3b,

green points for Rbulk = 0.65 (first and third quartiles of 0.62 and 0.70). Results show

that highly oxygenated fragments (CHOz>1, mainly organic acids) are well recovered,

RCHOz>1 67% (first and third quartiles of 65 and 72 %). This proportion slightly

decreases to 64% (first and third quartiles of 63 and 71 %) for the CHOz=1 (RCHOz=1)

family, which could mainly be composed of alcohols, aldehydes, and ketones. In

contrast, Rfrag for non-oxygenated species are in general lower, i.e., 55% (first and third

quartiles of 51 and 60 %) for the CHN family (RCHN), and 61% (first and third quartiles


42

of 55 and 64 %) for the CH family (RCH). Within the CH family, the saturated

hydrocarbon fragments (CnH2n+1+), which stem at least in part from the fragmentation of

hydrophobic normal and branched alkanes (Alfarra et al., 2004), are especially strongly

underestimated (RC2H2n+1+=44%, first and third quartiles of 42 and 48%). Note that a

higher variability in the Rfrag value is also observed for the CH fragments, probably due

to the variability in the water solubility of their parent molecules. This may lead to

higher uncertainties in the source apportionment of hydrocarbon-like OA and even to an

underestimation of such sources, using the offline AMS technique, as will be shown

below.

Figure 4.5: Median recovery of single organic fragments, and chemical families for the Zürich

spring campaign (offline vs. online PM2.5 AMS). The first and third quartiles of the inter-sample variability are shown as error bars. A ratio of 1 indicates a recovery of 100 %. The fragments are color-coded with the family (CH (hydrocarbon fragments, split into saturated and unsaturated), CHOz=1 and CHOz>1 (oxygenated fragments), and CHN, nitrogen-containing hydrocarbon fragments). Numbers across the top of the plot indicate the fragments’ nominal mass. Families include all respective fragments weighted by their mass contribution.

4.3.4 Source apportionment results

Differences between offline and online HR-ToF-AMS spectra caused by e.g.,

compound-dependent recoveries may also influence source apportionment results.

Therefore, we assess the ability of the offline AMS in the apportionment of OA sources,

by analyzing the offline Zürich yearly data set using ME-2 and comparing the source

apportionment results to those obtained by applying ME-2 to online ACSM data.

43

4.3.4.1 ME-2 output evaluation

A key consideration for PMF analysis is the number of factors selected by the

user. As mathematical criteria alone are insufficient for choosing the right number of

factors, this selection must be evaluated through comparisons of factor and tracer time

series, analysis of the factor mass spectra, and the evolution of the residual time series

as a function of the number of resolved factors. As described below, a five-factor

solution was selected as the best representation of the offline AMS data. To improve the

resolution of the POA sources by the model, literature profiles were used to define the

range of acceptable profiles (using the a value approach - Sect. 2.4). SOA factors are

not constrained because of the complex dependence of SOA composition on source,

atmospheric age, processing mechanisms, and meteorological conditions. This is

consistent with the approach of Crippa et al. (2014). After determining the optimal

number of factors (and their identity) required for explaining the variability in the data

set, we thoroughly assessed the sensitivity of the PMF results to the selection of the a

values.

Figure 4.6: Change in the time-dependent contribution of Q/Qexp as a function of the number of

factors Δ(Qi,cont/Qexp,i,cont) for a chosen offline solution (for aHOA = 0.0 and aCOA = 0.0).

Previous studies at this site have shown the influence of traffic, cooking, biomass

burning, and secondary organic aerosol (Lanz et al., 2008; Canonaco et al., 2013). Here,

we have constrained HOA and COA (profiles adapted from Mohr et al., 2012) and

optimized the solution by investigating different combinations of a values for the


44

constrained factors. In the selected five-factor solution, the non-constrained factors

extracted by ME-2 were related to BBOA, a highly oxygenated (OOA1) and moderately

oxygenated (OOA2) organic aerosol; the sum of OOA1 and OOA2 will be henceforth

considered as a proxy for secondary organic aerosol (referred to as OOA) which can,

though, be mixed with aged primary organic aerosol. These designations are based on

the correlation between OOA time series and that of secondary inorganic species (i.e.,

SO42- and NH4

+) and the comparison of OOA profile mass spectra with those extracted

from previous AMS data sets.

Number of factors

Figure 4.6 shows the change in the time-dependent Q/Qexp when increasing the

number of factors for the offline data set Δ(Qi,cont/Qexp,i,cont): contribution to Qi,cont for the

(p)- factor solution minus that of the (p+1)-factor solution. A significant decrease in

Δ(Qi,cont/Qexp,i,cont) signifies that structure in the residuals disappeared with the additional

factor. Removed structure is evident up to five factors. This behavior indicates that

while the ME-2 solution is clearly enhanced when increasing the number of factors to

five, addition of further factors does not improve the model description of the input

data. For this solution, there is no statistically significant difference in the residual

distributions of most variables between winter and summer (Fig. 4.7), indicating that the

modeled profiles represent the sources over the entire year well. Lower order solutions

(three and four factors) show one or two OOA factors besides the constrained HOA and

COA. Higher order solutions were explored but yielded additional OOA factors, which

could not be clearly attributed to a distinct source or process. Given this lack of

improvement in Δ(Qi,cont/Qexp,i,cont) and in the understanding of aerosol sources and

formation processes, and the absence of external tracers supporting the additional OOAs

in the high order solutions, the five-factor solution was considered as being optimal.

Furthermore, we consider only the sum of OOAs to facilitate the inter-method

comparison (as explained below). Note that PMF model uncertainties, i.e., imperfect

mathematical unmixing of sources, propagate into this comparison. This setting allows a

direct comparison between the offline and online methodologies, as the same set of

factors are obtained.

45

Figure 4.7: Residuals weighted with the uncertainty (residuals/uncertainty) of the offline solutions

for the periods April-September and October-March (example shown for one chosen solution, aHOA = 0.0; aCOA =0.0). Panels (a, b) show residuals as a function of m/z averaged over the whole periods color-coded with the probability that the residuals for April-September are the same as for October-March (Wilcoxon-Mann-Whitney test). Panels (c, d) shown the probability distribution function (pdf) of R/U during the same periods.

a-value optimization

The a values are independently varied for all constrained factors within a wide

range (a values from 0 to 1 with a step size of 0.1) for offline data in order to find an

optimal solution. Amongst the different solutions obtained, we selected those with

factor time series with the strongest correlation with those of the corresponding tracers.

The a value combinations of the chosen solution are specific for the data set used herein

and the selected reference profiles used, i.e., they may not be directly applicable to other

cases.

For this selection, the approach described above is adopted for the offline data

(illustrated in Fig. 4.8). For each set of a values selected as ME-2 input parameters (two

a values to constrain HOA and COA) five-factor time series are first generated by ME-


46

2. The ratios between factor and marker time series are then displayed as probability

density functions (pdf). The width of this distribution is used as a quality criterion, since

the narrower it is, the closer the linear relation of the factor is to the marker. Here, eBC,

CO, and NH4+ are used as markers for HOA, BBOA, and OOA, respectively. For COA,

no specific marker has yet been identified and studies presenting online data validate

this factor using the daily pattern of its concentration, which typically peaks at lunch-

and dinner-time (e.g. Crippa et al., 2014). For 24 h integrated filter data, this diagnostic

cannot be used and therefore the optimization of COA separation by ME-2 is not used

as a quality criterion. In practice, the solution with the narrowest factor-to-marker

distributions is defined as the best solution with respect to its interpretability in the

environment. For the other factors, we have examined the variability in the ratio, x,

between factor and corresponding marker: . CO0 is the

background concentration, which is estimated to be 100 ppb. This is both in agreement

with measurements at this site and also literature presenting measurements of

background air masses (e.g. Griffiths et al., 2014, for Jungfraujoch). In practice, the best

solution is obtained by minimizing the sum of the ratios of the logarithmic geometric

standard deviations (σg) to the logarithmic geometric averages (µg) of

. Besides using eBC as an HOA marker, the quality of the solution

was also checked using eBCtr. For both markers, the same a value combination was

considered best according the overall criterion. All solutions, for which none of the

single distributions showed a different relative variance than the best solution, were also

accepted (this comparison was performed using an F test). Note that the determination

of the a value ranges resulting in the most satisfactory solutions for the offline data set

is performed independently from the online measurements. The comparison between

source apportionment from offline and online data sets provides, therefore, a direct

measure of the ability of the offline AMS technique to resolve aerosol sources and

formation processes. Systematically stepping through the multi-dimensional a value

space, as opposed to most published analyses using one-dimensional a-value-based ME-

2 or PMF, offers a more effective and objective exploration of the solution space.

Environmentally reasonable factors are obtained by selecting the subset of solutions that

optimizes factor-to-marker relationships.

47

Figure 4.8: Relative width of the distributions cj/cmarker displayed as a function of aHOA and aCOA.

Panel (a) shows the sum of the criteria for HOA, BBOA, and OOA being the sum of OOA1 and OOA2. (The chosen solutions are pointed out in the white area.) Panels (b-d) show the individual criteria as a function of the a values of HOA and COA aHOA, aCOA.

The chosen offline solutions lie in general in a part of the solution space with low

a values for HOA and COA (the single chosen a value combinations are shown as white

boxes in Fig. 4.8a). The relative variability of the signatures of the constrained factors

among the different accepted solutions for the offline source apportionment is below 6

and 3% for HOA and COA, respectively. Note that this is an incomplete exploration of

the rotational ambiguity and thus does not describe the complete model uncertainty.

Online ACSM solution

Like the offline solution, the online ACSM solution yielded a five-factor solution

representing HOA (constrained using the profile reported by Crippa et al., 2013c), COA

(constrained using the profile reported by Crippa et al., 2013c), BBOA (constrained

using the profile reported by Ng et al., 2011b), SV-OOA, and LV-OOA. In contrast to

the offline solution, ME-2 could not extract BBOA independently; thus this factor was

constrained as suggested by Crippa et al. (2014).


48

Comparing the offline to online PMF source apportionment results obtained with

the approaches described earlier has the obvious drawback that we compare OOA

factors extracted in winter-only or summer-only (online) vs. combined winter and

summer (offline). However, this is mitigated by two factors. First, the discrimination

between OOAs for the offline solution is largely driven by seasonal differences (average

relative contributions: in winter OOA2 9 %, OOA1 91% and in summer 87, 13 %,

respectively), indicating only small biases in the composition. Second, the residuals for

both winter and summer are normally distributed and centered around zero (Fig. 4.7),

indicating negligible seasonally dependent bias in the apportioned mass. Therefore,

while this comparison method may contribute somewhat to the overall uncertainties, it

is unlikely to significantly affect the conclusions or values reported below.

Figure 4.9: Comparison of overall factor profiles obtained for the chosen solutions both from the

offline (left, for HOA and COA, spectra from Mohr et al. (2012) were used as reference) and the retrieved factor profiles from the online source apportionment (right).

49

Factor profiles

The averages of factor profiles of the selected ME-2 online and offline solutions

are presented in Fig. 4.9. Apart from the good correlations between factors and external

markers time series used as an acceptance criteria, our results show that the factors

retrieved by ME-2 exhibit spectral profiles are consistent with previous studies. The

BBOA profile extracted from the offline data set closely resembles those reported in the

literature for other locations (Crippa et al., 2013c), characterized by the contribution of

oxygenated fragments at m/z 29 (CHO+), 60 (C2H4O2+), and 73 (C3H5O2

+), from

fragmentation of anhydrous sugars (Ng et al., 2011b). The OOA mass spectra retrieved

by ME-2 for both online and offline data sets is characterized by a typical fingerprint,

dominated by oxygenated fragments at m/z 43 (C2H3O+) and 44 (CO2+) characteristic of

secondary compounds. The consistency of these spectral profiles with previously

reported profiles from online measurements provides additional support to the source

apportionment results presented here.

4.3.4.2 Recoveries of different OA categories (HOA, COA, BBOA, OOA)

While results above show that the bulk OA recovery lies between 60 and 91%

(Sect. 3.2), for the current analysis we assess the recovery of the different factors as

representative of ambient compound classes from various sources/processes determined

by ME-2. This is based on the comparison between online and offline source

apportionment results. However, for this comparison, the approach presented in Eq.

(4.4) cannot be adopted because of the noisy data and/or low sulfate content during

periods critical for recovery determination of a specific factor. We therefore perform a

selfconsistent calculation of factor-dependent recoveries, which, when applied to the

offline data, yield (1) fractional composition consistent with online measurements, and

(2) calculated bulk recovery consistent with the measured bulk recovery using WSOA

measurements. The implementation is described below (Eq. 4.5). Let i denote the time

index and k the factor index. We define the time-dependent recovery of a factor k (Ri,k)

as the time-dependent ratio of the contribution of this factor in the offline (offi,k) and in

the online (oni,k) solution multiplied by the time-dependent bulk recovery of WSOA,

Ri,WSOA (Eq. 4.5):


50

(4.5)

Figure 4.10: Recoveries Rk for HOA, COA, BBOA, and OOA (OOA1COOA2) obtained from the

intercomparison of source apportionment results of offline AMS to online ACSM data (Zürich 2011-2012). 100 000 random combinations of offline and online solutions and randomly chosen offline repeats result in the same amount of time-independent Rk, which are expressed as probability density functions (pdf).

Finally, Rk, the median of Ri,k over time, is computed. Rk reflects not only the bias

caused by the water extraction but also filter sampling/storage effects and differences

between the individual ME-2 solutions. The uncertainty of Rk depends both on the

uncertainty related to the single offline solution point in time as well as on their spread

in comparison to the online solution. The first can be quantified by assessing the model

error for the offline and online using the variability of the solution for different model

runs. The offline approach adopted here, including several measurements of the same

sample in ME-2 (in general eight spectra, called repeats, per sample), enables assessing

the performance of the ME-2 solution for different samples and different factors. For

this reason, we have repeatedly calculated Rk using randomly chosen combinations of

(1) different ME-2 offline solutions (selected in Sect. 3.4.1 and reference online

solutions (due to the rolling window approach providing individually optimized

periods) and (2) different repeats of the offline AMS measurements for the same

51

samples. The result is an ensemble of Rk (for each factor 100 000 Rk values are

calculated) displayed in Fig. 4.10 as probability density functions. The range of these

distributions reflects both model and measurement uncertainties. Note that this range

does not reflect the variability in time of Ri,k. The retrieved factor recoveries are

consistent with our understanding of the chemical nature of the different OA

components, with primary hydrophobic species less efficiently extracted than secondary

oxygenated species. As expected, the most hydrophobic component, HOA, has the

lowest recovery with a median RHOA of 11% (first and third quartiles of 10 and 12 %).

We note that RHOA is lower than the RCH,sat (see Sect. 3.3), which seems to indicate that

these ions can also originate from more hydrophilic molecules than those in traffic

emissions. COA appears to be moderately soluble, with RCOA = 54% (first and third

quartiles of 48 and 60 %). BBOA and OOA species were largely recovered with RBBOA

= 65% (first and third quartiles of 62 and 68 %) and ROOA = 89% (first and third

quartiles of 87 and 91 %). Uncertainties in RWSOA are not included in the calculation.

Further, online measurements have a lower size cutoff than the offline data (600 nm vs.

10 μm), and large accumulation mode particles are expected to preferentially contain

OOA, due to their extended aging. This might provide a positive bias to the OOA Rk

and a negative bias to Rk of the other factors.

4.3.4.3 Quantitative comparison of offline and online OA factors

We assume that the Rk values calculated in the previous section are characteristic

properties of the retrieved OA components, i.e., that they can be applied throughout the

analyzed offline data sets. This allows us to quantitatively compare the mass

concentrations of offline and online OA factors retrieved throughout the year. By

applying the Rk obtained to the offline data set, the source apportionment results

(relative composition) can be corrected. In a second step the results can be scaled to

ambient concentrations. Here OA concentrations from the ACSM measurements are

used. The Rk corrected offline source apportionment results are compared to results

from the ACSM analysis and respective marker concentrations (Fig. 4.11). The

displayed error bars reflect the variability of a factor’s contribution for one offline

sample assessed by the repeats, and also using the different chosen ME-2 solutions

provides an estimate of the measurement and (partial) model uncertainties. Overall the

variability of the offline factor contributions for an individual sample is 0.1 μgm-3. The


52

factors with a lower recovery Rk (HOA and COA), reveal also bigger differences

between the time series for the offline and online data and thus more scattering. In

general, the variability of the offline factor contributions for an individual sample

increases when moving away from the 1 : 1 line. This is especially apparent for BBOA,

where outliers with low offline concentrations are much more uncertain than the points

matching the contribution in the online solution (Fig. 4.11c, f). All factors but COA (for

which no marker is known) show similar relationships with their marker for both offline

and online data (Fig. 4.11e-g).

Figure 4.12 presents the ratio of factor contributions and their respective marker

concentrations for the online and corrected offline solutions. The medians and spreads

of the distributions are comparable between offline and online solutions. Only for

BBOA is the distribution wider (also seen in Fig. 4.11). Chirico et al. (2011) and El

Haddad et al. (2013a) report HOA/EC ratios of 0.4, which is close to the median found

in this study (HOAoff/eBC = 0.57 (first and third quartiles of 0.42 and 0.74),

HOAon/eBC = 0.64; first and third quartiles of 0.42 and 0.79). The ratio BBOA/(CO-

CO0) is 6.1 (first and third quartiles of 2.2 and 7.8) and 5.7 μgm-3 ppm-1 (first and third

quartiles of 4.5 and 8.4) for offline and online, respectively. However, this ratio has to

be considered as a lower limit, as CO may also be emitted by non-biomass-burning

sources (e.g., traffic). While this ratio is significantly lower than values reported for

prescribed/open burns (De Gouw and Jimenez, 2009), values found here are within the

same range as those measured for modern stoves used in Switzerland (Heringa et al.,

2011). Crippa et al. (2014) reported OOA/NH4+ ratios for 25 sites, with an average of

2.0 (0.3 for the site with the lowest ratio and 7.3 for the one with the highest). Lanz et

al. (2010) reported values of 5.6 and 1.5 for Zürich in July 2005 and January 2006,

respectively. The values for Zürich during the period analyzed here are 5.1 (offline, first

and third quartiles of 3.0 and 11.50) and 5.1 (online, first and third quartiles of 3.0 and

10.5). The examination of these ratios and their comparison with previously reported

values provide additional support to the offline AMS methodology and resulting source

apportionments. Indeed, the application of this methodology to additional filters from

other locations where accompanying online AMS measurements are available may aid

the further constraint of the Rk estimates presented here.

53

Figure 4.11: Comparison of factor contributions from separate offline (PM10 AMS, two constrained factors: HOA, COA) and online (PM1 ACSM) source apportionment using ME-2 (traffic (HOA), cooking (COA), biomass burning (BBOA), and oxygenated organic aerosol; OOADOOA1COOA2). Factor-specific recoveries (Rk) are applied to the offline contributions. Error bars (in gray) denote the variability between the different ME-2 solutions and for different recorded spectra per sample for offline and for online only the first of the two. Panels (a-d) show scatter plots comparing the absolute contribution of the respective source/OA category for offline AMS and online ACSM measurements. The color code distinguishes all factor contributions (bullets, saturated colors) from winter points (open circle, light colors). The gray dashed line indicates the 1 V 1 line. Panels (e-g) show the correlation with the respective markers: black symbols represent the absolute contribution of the respective source for the online ACSM measurements and the colored symbols represent the absolute contribution of the same source for the offline AMS measurements.


54

Figure 4.12: Ranges of ratios of the contribution of different factors to their markers for the offline

(corrected with Rk) and online ACSM source apportionment results. Note that OOA is the sum of OOA1 and OOA2.

4.4 Conclusions

In this study, we developed and evaluated an offline method using an HR-ToF-

AMS for the characterization of the chemical fingerprints of aerosol collected on filters

(Pall quartz filters for the current study). Particulate matter on filters is extracted in

water and introduced into the HR-ToF-AMS using a nebulizer. The method was applied

to more than 250 filters from different seasons in different environments in Europe. The

detection limits depend on the nebulizer and species; for the current setup for OA and

55

SO42- they are 80 and 25 μgL-1, respectively. External data are needed for quantification.

We recommend the use of OC analysis from a Sunset OC/EC analyzer and ion

chromatography data for the determination of the inorganic fraction. Estimates of the

recovery of bulk OA using different reference measurements show that OA is largely

captured (60-91% depending on the data set). The obtained organic mass spectra are

comparable to online HR-ToF-AMS spectra, although hydrocarbons are

underestimated. Rbulk also shows a good agreement with the WSOA fraction.

Source apportionment on offline AMS data is conducted with positive matrix

factorization, implemented using the ME-2 algorithm. We investigate a set of PMF

solutions, for which the different OA components show tight relationships with their

respective markers. Thereby, we demonstrate that organic mass spectral data generated

using this method are suitable for identifying different OA sources as HOA, COA,

BBOA, and OOA. By comparing the results for offline AMS and ACSM data, we

retrieved recoveries of the different OA components (Rk): traffic (RHOA = 0.11), cooking

(RCOA = 0.54), biomass burning (RBBOA = 0.65), and secondary OA (ROOA = 0.89).

Qualitatively, Rk also relates to the water solubility of the respective source, e.g.,

primary OA related to hydrocarbons (e.g., HOA) shows a low Rk caused by its low

water solubility. Such Rk should be determined at other sites where also additional

sources might be important, providing an assessment of site-to-site variability.

Nevertheless, these best estimates of Rk may be used to correct source apportionment

results from offline AMS measurements (as in Huang et al., 2014). When combined

with WSOC measurements, one might also be able to assess the applicability of these

values at the site in question by comparing overall modelled Rbulk to RWSOA. Even

though the offline AMS approach might poorly capture sources exhibiting fast changes,

this method broadens the applicability of the AMS to longterm size-segregated (PM1,

PM2.5, PM10) measurements (in contrast to online campaigns of typically 1 month) for

extended monitoring networks.


56

Acknowledgements

This work was supported by the Swiss Federal Office of Environment, inNet

Monitoring AG, Ostluft, the country Liechtenstein, the Swiss cantons Basel-Stadt,

Basel-Landschaft, Graubünden, Solothurn, Ticino, Thurgau, Valais, the Lithuanian-

Swiss Cooperation Programme “Research and Development” project AEROLIT (Nr.

CH-3-ŠMM-01/08), and the IPR-SHOP SNSF starting grant.

57

5 Long-term chemical analysis and organic

aerosol source apportionment at 9 sites in Central Europe: source identification

and uncertainty assessment

K. R. Daellenbach1, G. Stefenelli1, C. Bozzetti1, A. Vlachou1, P. Fermo2, R. Gonzalez2, A. Piazzalunga3,a, C. Colombi4, F. Canonaco1, C. Hueglin5, A. Kasper-Giebl6, J.-L. Jaffrezo7, F. Bianchi1,b, J. G. Slowik1, U. Baltensperger1, I. El-Haddad1, and A. S. H. Prévôt1 1Paul Scherrer Institute (PSI), 5232 Villigen-PSI, Switzerland. 2Università degli Studi di Milano, 20133 Milano, Italy. 3Università degli Studi di Milano-Bicocca, 20126 Milano, Italy. 4ARPA Lombardia, Regional Centre for Air Quality Monitoring, 20122 Milan, Italy. 5Swiss Federal Laboratories for Material Sciences and Technology, 8600, Dübendorf, Switzerland 6Institute of Chemical Technologies and Analytics, Vienna University of Technology, 1060 Wien, Austria. 7Université Grenoble Alpes, CNRS, IGE, 38000 Grenoble, France. anow at: water and soil lab, 24060 Entratico, Italy. bnow at: Department of Physics, University of Helsinki, 00014 Helsinki, Finland.

Pulished in Atmospheric Chemistry and Physics DOI: 10.5194/acp-17-13265-2017

Chapter 5. Long-term organic source apportionment in Central Europe using offline AMS

58

Abstract Long-term monitoring of organic aerosol is important for epidemiological studies,

validation of atmospheric models, and air quality management. In this study, we apply a

recently developed filter-based offline methodology using an aerosol mass spectrometer

(AMS) to investigate the regional and seasonal differences of contributing organic

aerosol sources. We present offline AMS measurements for particulate matter smaller

than 10 μm at nine stations in central Europe with different exposure characteristics for

the entire year of 2013 (819 samples). The focus of this study is a detailed source

apportionment analysis (using positive matrix factorization, PMF) including in-depth

assessment of the related uncertainties. Primary organic aerosol (POA) is separated in

three components: hydrocarbon-like OA related to traffic emissions (HOA), cooking

OA (COA), and biomass burning OA (BBOA). We observe enhanced production of

secondary organic aerosol (SOA) in summer, following the increase in biogenic

emissions with temperature (summer oxygenated OA, SOOA). In addition, a SOA

component was extracted that correlated with an anthropogenic secondary inorganic

species that is dominant in winter (winter oxygenated OA, WOOA). A factor (sulfur-

containing organic, SC-OA) explaining sulfur-containing fragments (CH3SO2+, which

has an event-driven temporal behaviour, was also identified. The relative yearly average

factor contributions range from 4 to 14% for HOA, from 3 to 11% for COA, from 11 to

59% for BBOA, from 5 to 23% for SC-OA, from 14 to 27% forWOOA, and from 15 to

38% for SOOA. The uncertainty of the relative average factor contribution lies between

2 and 12% of OA. At the sites north of the alpine crest, the sum of HOA, COA, and

BBOA (POA) contributes less to OA (POA/OA=0.3) than at the southern alpine valley

sites (0.6). BBOA is the main contributor to POA with 87% in alpine valleys and 42%

north of the alpine crest. Furthermore, the influence of primary biological particles

(PBOAs), not resolved by PMF, is estimated and could contribute significantly to OA in

PM10.

5.1 Introduction

The development and field deployment of the Aerodyne aerosol mass

spectrometer (AMS; Canagaratna et al., 2007) have greatly improved air quality

monitoring by providing real-time measurements of the non-refractory (NR) submicron

59

aerosol (PM1) components. The application of factor analysis on the collected organic

aerosol (OA) mass spectra enabled the efficient disentanglement of aerosol factors,

which could be subsequently related to specific aerosol sources and processes (Lanz et

al., 2007, 2008; Jimenez et al., 2009; Ulbrich et al., 2009, Zhang et al., 2011; Ng et al.,

2011b; Crippa et al., 2014). Factors typically extracted include directly emitted primary

OA (POA) from biomass burning (BBOA) or traffic (HOA), and oxygenated OA

(OOA) that is typically associated with secondary OA (SOA), formed through the

oxidation of organic vapour precursors or heterogeneous processes. The model is not

capable of identifying the main SOA precursors, but often differentiates OOA based on

its volatility and degree of oxygenation (semivolatile fraction and low-volatility

fraction) due to the available highly time-resolved data.

However, the cost and operational requirements of the AMS make its deployment

impractical throughout a dense monitoring network and over longer time periods. As a

result, most available datasets are often limited to a few weeks of measurements, and

factors are extracted mainly based on diurnal variations in POA emission strength and

SOA oxygen content (Zhang et al., 2011; El Haddad et al., 2013b). Highly mobile

measurements on platforms as aircrafts (e.g. DeCarlo et al., 2008) or vehicles (e.g.

Mohr et al., 2011) are designed for regional studies, but are even more limited by cost,

availability, and time than stationary studies. This hinders the determination of the

aerosol regional and seasonal characteristics and evaluation of long-term emission

trends, limiting the information required for model validation and development of

efficient mitigation strategies. Furthermore, the negligible transmission efficiency of the

AMS inlet for coarse particles prevents the characterization of their chemical nature and

contributing sources.

The recent development of the aerosol chemical speciation monitor (ACSM; Ng

et al., 2011c, Fröhlich et al., 2013) has enabled the establishment of dense networks of

long-term AMS-type measurements and source apportionment of the organic aerosol

(e.g. Crippa et al., 2014, using AMSs for shorter campaigns within the EUCAARI

project or EMEP/ACTRIS projects for longer multi-season campaigns using ACSMs).

However, the mass spectrometers used by the ACSMs have far lower mass resolution

than the AMS, reducing their performance for OA characterization and source


60

apportionment. An alternate monitoring strategy involves extending AMS spatial and

temporal coverage by measuring the nebulized water extracts of filter samples

(Daellenbach et al., 2016; Mihara and Mochida, 2011). This approach allows the

retroactive investigation of specific events, e.g. haze events in China (Huang et al.,

2014), as well as AMS measurements of coarse-mode aerosol (Bozzetti et al., 2016) and

long-term source apportionment studies (Bozzetti, 2017a, b). Such an approach was also

used in recent studies for identifying the different types of water-soluble chromophores

(Chen et al., 2016). Additionally, such filters are routinely collected and are already

available over multi-year periods at many air quality monitoring stations around the

world for years and/or decades. Unlike single-season online AMS studies, the offline

AMS analysis of filter samples may reveal seasonal and long-term variations in the

emissions of POA and SOA precursors required for model validations and the

establishment of efficient mitigation strategies.

Here, we present offline AMS measurements of PM10 (particulate matter with an

aerodynamic diameter smaller than 10 μm) at nine stations in central Europe with

different exposure characteristics for the entire year of 2013 (819 samples). The sites

cover rural and urban locations, including urban background and traffic and wood-

burning-influenced stations. Such long-term multi-site analyses allow the quantitative

description of the temporal and spatial variability in the main OA sources and may

provide further insights into SOA precursors and formation pathways. This paper

focuses on the identification of the main factors influencing the OA concentrations at

the different sites and the assessment of the associated uncertainties. In a second paper,

we will investigate the site-to-site differences and general trends in the factor time series

and their relationship with external parameters.

5.2 Methods

5.2.1 Study area and aerosol sampling

PM10 samples were collected at nine sites in Switzerland and Liechtenstein

(Table 5.1 and Fig. 5.1). Seven of the sites (Basel, Bern, Payerne, Zürich, Frauenfeld,

St. Gallen, Vaduz) are located in northern Switzerland and Liechtenstein and two

(Magadino and San Vittore) in southern Switzerland. Aerosol was sampled at the

61

selected sites every fourth day for 24 h throughout the year 2013 on quartz fibre filters

(14.7 cm diameter) using high-volume samplers (500 L min-1). Filters were then

wrapped in aluminium foil or lint-free paper and stored at -20 °C. Field blanks were

collected following the same approach.

Figure 5.1: Map of study area with locations of sites indicating their characteristics. The topography is displayed as meters above sea level.

Table 5.1: Study sites with geographical location and classification Site (station code) Classification General location altitude Basel, St. Johann (bas) Urban/background North of Alps/Swiss plateau 308 m. Bern, Bollwerk (ber) Urban/traffic North of Alps/Swiss plateau 506 m. Frauenfeld, (fra) Suburban/background North of Alps/Swiss plateau 403 m. Payerne (pay) Rural/background North of Alps/Swiss plateau 539 m. St. Gallen,(gal) Urban/traffic North of Alps/Swiss plateau 457 m. Zürich, Kaserne (zue) Urban/background North of Alps/Swiss plateau 457 m. Vaduz, Austrasse (vad) Urban/traffic North of Alps/alpine valley 706 m. Magadino, Cadenazzo (mag) Rural/background South of Alps/alpine valley 254 m. San Vittore, (vi) Rural/traffic South of Alps/alpine valley 330 m.


62

5.2.2 Offline AMS analysis

The offline AMS analysis summarized below was carried out following the

methodology developed by Daellenbach et al. (2016). For each analysed filter sample,

four 16mm diameter filter punches were sonicated together in 10mL ultrapure water

(18.2MΩcm, total organic carbon TOC < 5 ppb, 25 °C) for 20 min at 30 °C. Liquid

extracts were then filtered (0.45 μm) and nebulized in synthetic air (80 % volume N2,

20% volume O2; Carbagas, Gümligen CH-3073 Switzerland) using a customized Apex

Q nebulizer (Elemental Scientific Inc., Omaha, USA) operating at 60°C. The resulting

droplets were dried using a Nafion® dryer and then injected and analysed using the

high-resolution time-of-flight AMS (HR-ToF-AMS). Three types of measurements

were performed: (i) filter samples, (ii) field blanks (collected and treated in the same

way as the exposed filters), and (iii) measurement blanks (nebulized ultrapure water

without filter extract). The measurement blank was determined before and after every

filter sample. Each sample was recorded for 480 s (AMS V-mode, m/z 12-447), with a

collection time for each spectrum of 30 s. Ultrapure water was measured for 720 s.

Once per day, ultrapure Milli-Q water was nebulized with a particle filter interposed

between the nebulizer and the AMS for the determination of the gas-phase contribution

to the measured mass spectrum, which was then subtracted during analysis from both

blanks and filter samples. The filters from Zürich were analysed twice with a time

difference of approximately 5 months to assess the measurement repeatability. High-

resolution mass spectral analysis was performed for each m/z (mass-to-charge) in the

range of 12-115. The measurement blank was subtracted from the sample spectra. In a

previous study, it has been shown that the measurement blank is comparable to the

organic blanks obtained from the nebulization of NH4NO3 (Bozzetti et al., 2017b). The

interference of NH4NO3 in the CO2+ signal described by Pieber et al. (2016) was

corrected as follows (Eq. 5.1):

(5.1)

63

The correction factor was determined based on

measurements of aqueous NH4NO3 conducted regularly during the entire measurement

period and varied between <1% and up to approximately ~5% (Pieber et al., 2016).

5.2.3 Other chemical analysis

Organic and elemental carbon (OC, EC) content were measured using a thermo-

optical transmission method with a Sunset OC/EC analyser (Birch and Cary, 1996),

following the EUSAAR-2 thermal-optical transmission protocol (Cavalli et al., 2010).

Water-soluble carbon was measured with water extraction followed by catalytic

oxidation, nondispersive infrared detection of CO2 using a total organic carbon analyser,

only for the samples from Magadino and Zürich. Water-soluble ions (K+, Na+, Mg2+,

Ca2+, and NH4+ and SO4

2-, NO3-, and Cl- and methane sulfonic acid were analysed using

ion chromatography (Piazzalunga et al., 2013 and Jaffrezo et al., 1998). Levoglucosan

measurements were performed with a high-performance anion exchange

chromatographer (HPAEC) with pulsed amperometric detection (PAD) using an ion

chromatograph (Dionex ICS-1000) following Piazzalunga et al. (2010 and 2013). Free

cellulose was determined using an enzymatic conversion to D-glucose (Kunit and

Puxbaum, 1996) and subsequent determination of glucose with an HPAEC (Iinuma et

al., 2009). Online measurements of gas-phase compounds and meteorology were also

performed at selected sites.

5.3 Source apportionment

5.3.1 General principle

Source apportionment of the organic aerosol is performed using positive matrix

factorization (PMF; Paatero and Tapper, 1994). PMF is a statistical un-mixing model

explaining the variability in the organic mass spectral data (xi,j as linear combinations of

static factor profiles (fj,k and their timedependent contributions (gi,k); see Eq. (5.2)

(where p is the number of factors). The index i represents a specific point in time, j an

ion, and k a factor. The elements of the model residual matrix are termed ei,j.


64

(5.2)

In the input data matrix, each filter sample was represented on average by 11 mass

spectral repetitions to examine the influence of the AMS measurement repeatability on

the PMF outputs. A preceding blank from nebulized ultrapure water was subtracted

from each mass spectrum. The input errors si,j required for the weighted least-squares

minimization by the model consist of the blank variability (σi,j and the uncertainty

related to ion counting statistics and ion-to-ion signal variability at the detector (δi,j

Allan et al., 2003; Ulbrich et al., 2009). We applied a minimum error according to

Ulbrich et al. (2009) and a down-weighting factor of 3 to all fragments with an average

signal-to-noise ratio lower than 2 (Ulbrich et al., 2009). Input data and the

corresponding error matrices consisted of 202 organic ions. The organic fragments, x’i,j,

obtained from offline AMS analyses do not directly represent ambient concentrations.

Therefore, the signal of each fragment was converted to such an ambient concentration

(xi,j in µgm-3) by multiplying the fraction of this signal with the estimated organic matter

(OM) concentration. The latter was calculated as the product of the OC concentrations

measured by the Sunset OC/EC analyser and the OM/OC from the offline AMS

measurements (OM/OC)oAMS (Eq. 5.3). Note that such scaling does not change the

outcome of Eq. (5.2) since both data and error matrices are scaled in the same manner

and the fingerprints (fk,j) are not changed.

(5.3)

The Source Finder toolkit (SoFi v.4.9; Canonaco et al., 2013) for the Igor Pro

software package (Wavemetrics, Inc., Portland, OR, USA) was used to configure the

PMF model and for post-analysis. The PMF algorithm was solved using the multilinear

engine-2 (ME-2; Paatero, 1999). Normalization of the PMF solution during the iterative

minimization process is disabled as implemented in SoFi (Canonaco et al., 2013). ME-2

enables an efficient exploration of the solution space by constraining the fk,j elements a

65

priori within a certain range defined by the scalar a (0 ≤ a ≤ 1) from a starting value f’k,j,

such that the modelled fk,j in the solution satisfies Eq. (5.4):

(5.4)

f’k,j is the starting value used as a priori knowledge from previous studies and fk,j is

the resulting value in the solution. In all PMF runs (unless mentioned otherwise), we

used the high-resolution mass spectra for HOA and COA (cooking OA) from Crippa et

al. (2013c) as constraints, i.e. two rows of f’k,j were set equal to the mass spectra of

HOA and COA. Ions that were present in our datasets but not in the reference profiles

for HOA and COA were inferred from published unit mass resolution (UMR) profiles

(Ng et al., 2011b and Crippa et al., 2013b). For this purpose, the fraction of signal at a

specific m/z in the UMR reference spectrum (fUMR,m/z was compared to the fraction of

signal of all ions at this m/z in the HR reference spectrum (fHR,m/z). The difference

fUMR,m/z - fHR,m/z was used as an entry in f’k,j for such missing ions. For these ions, an a

value of unity was set. For the other factors, the factor elements were fitted using ME-2.

Alternatively, such missing ions can be also treated as ordinary factor elements, to be

fitted using ME-2 with all other ordinary factor elements.

Source apportionment analysis was performed following the scheme shown in

Fig. 5.2 and discussed below. Unconstrained and constrained exploratory PMF runs

provided information on the number of interpretable factors (Sect. 5.3.2). Multiple

constrained PMF runs were then performed to assess the model sensitivity to the chosen

a value, the model starting point and input matrix (entire dataset: PMFblock; only Zürich:

PMFzue,isol; one filter per site and month: PMF1filter/month; repeated measurements for

Zürich: PMFzue,reps and repeated measurements (Sect. 5.3.3). The factors obtained were

then classified and corrected for their recovery (Sect. 5.3.4 and 5.3.5). Finally, the

different solutions were evaluated and only the solutions that satisfied a set of

predefined criteria (Sect. 5.3.6 and Supplement A) were considered.


66

5.3.2 Preliminary PMF

Figure 5.2: Step-by-step outline of adopted source apportionment approach (factor recoveries Rk).

aHOA and aHOA represent the a value applied for HOA and COA, respectively.

We explored constrained PMF solutions, ranging from 1 to 10 factors. This

investigation is performed on the entire dataset, including all stations and seasons

(details in the Supplement A). The impact of the number of factors on the residuals is

examined in the Supplement A. The introduction of two factors, in addition to HOA and

COA, resulted in a significant reduction in the residuals and the separation of BBOA

and OOA contribution. BBOA exhibited a prominent seasonal variation with a

significant increase during winter and contributed most to the explained variation in the

fragment C2H4O2+, originating from the decomposition of anhydrous sugars, i.e. from

cellulose pyrolysis. OOA was identified based on its mass spectral fingerprint, with

high contribution from oxygenated ions at m/z 43 and 44. A further increase in the

number of factors did not significantly contribute to the reduction in the residuals.

However, the introduction of a fifth factor allowed the separation of the OOA into two

different factors, with distinct seasonal variability and different relative contributions

from oxygenated fragments at m/z 43 and 44. The two OOA factors will be referred to

as winter and summer OOA (WOOA and SOOA) according to their seasonality. The

introduction of a sixth factor allowed the resolving of a factor with a distinct time series

explaining the variability of sulfur-containing fragments (e.g. CH3SO2+. This factor will

be referred to as sulfur-containing organic aerosol (SC-OA). We explored higher-order

solutions, but could not interpret the resulting factor separations. Therefore, we further

consider a six-factor solution below.

67

5.3.3 Sensitivity analysis

We assessed the model sensitivity to the chosen a value for HOA and COA and

the model starting point (independently for all four PMF inputs, as described below).

The a values were independently varied for HOA and COA (a value from 0 to 1 with

increments of 0.1, giving 121 a-value combinations). For every a -value combination,

the model was initiated from five different pseudo-random starting points (seeds),

yielding 605 total runs. As the selection of the a value combination was randomized, the

process was repeated four times in order to ensure that every a-value combination was

represented at least once (2420 runs), which in turn provided an assessment of the seed

effect on the results.

While this approach has been proven very effective in selecting a range of

environmentally relevant solutions (Elser et al., 2016a, b, and Daellenbach et al., 2016),

the resulting uncertainties may be underestimated. Paatero et al. (2014) compared the

effectiveness in estimating uncertainties of factor elements using two different

approaches: the displacement (DISP) and bootstrap analysis (BS). BS involves applying

the model to input matrices consisting of a subset of the entire dataset. DISP involves

running PMF several times using systematically perturbed factor profile elements of a

reference solution, but allowing a defined difference in Q from the reference solution.

Both approaches are computationally intensive, especially DISP. Because of such

computational limitations, the combination of BS and DISP was not feasible for the

dataset presented here, especially in combination with a-value sensitivity tests.

Therefore, we chose to perform four sensitivity tests performing PMF runs using four

different input datasets, presented in the following. These sensitivity tests allow

conclusions on the stability of PMF analysis when reducing the temporal or spatial

resolution as well as the influence of the measurement repeatability.

1. PMFblock: PMF was performed on data from all seasons and all sites

combined (all measured in October 2014). The corresponding data and error

matrices involved 819 samples from nine sites with 202 ions and, on average,

11 spectra per sample. This represents the base case.


68

2. PMFzue,isol: PMF was performed on data from Zürich alone (isolated from

PMFblock input). The corresponding data and error matrices involved 91

samples with 202 ions and on average 11 spectra per sample.

3. PMF1filter/month: PMF was performed on data from all sites but only

considering the 1st filter collected for every month (12 filters per site), as for

these samples levoglucosan and cellulose data was available. The

corresponding data and error matrices involved 108 samples with 202 ions

and on average 11 spectra per sample.

4. PMFzue,reps: PMF was performed on data from the repeated measurements of

Zürich samples. The corresponding data and error matrices comprised 91

samples with 196 ions and on average 14 spectra per sample.

For each of the four PMF datasets, 2420 PMF runs were performed for evaluating

the sensitivity of the model to the chosen a value and the seed. The quality of each of

the 2420 PMF runs was individually assessed using criteria lined out in Sect. 5.3.6.

5.3.4 Factor classification

From the sensitivity analysis, a large number of solutions were generated.

Systematic analysis of these solutions required automatic identification and

classification of the retrieved factors within each solution. We applied a sequential

classification algorithm as follows. Since HOA and COA were initially constrained on

preselected rows of fk,j, they did not need to be identified. In a second step, the factor

showing the highest explained variation for C2H4O2+ among the four remaining factors

was identified as BBOA. In a third step, the factor with the highest explained variation

for CH3SO2+ among the three remaining factors was identified as SC-OA. From the last

two factors, the one with the highest explained variation in CO2+ was identified as

WOOA and the other as SOOA.

5.3.5 Recovery and blank corrections

After factor identification, factor time series are corrected using factor-specific

recoveries (Eq. 5.5, resulting in OAi,k) determined in Daellenbach et al. (2016) for

HOA, COA, BBOA, and OOA.

69

(5.5)

where gi,k values are the concentrations of factor k at the time point i, Rk the

recoveries of the respective factor, and OAi the OA concentration. For a limited number

of PMF runs (PMFblock, the field blank analyses were also included in the PMF input

data. This provides the contributions of different factors to the field blanks, which were

used to correct the output factor time series. Uncertainties induced by the blank

subtraction were propagated.

5.3.6 Solution selection

Each of the 2420 PMF solutions per PMF dataset (PMFblock, PMFzue,isol,

PMFzue,reps, PMF1filter/month was evaluated based on their factor profiles, time series, and

the OC mass closure. Solutions were selected if they satisfied the following set of

criteria:

1. fCO2+ < 0.04 in HOA and COA factor profiles (HOA based on Aiken et al.,

2009, Mohr et al., 2012, Crippa et al., 2013c, 2014 and COA based on Crippa

et al., 2013b, 2013c, Mohr et al., 2012);

2. fC2H4O2+ < 0.004 and 0.01 in HOA and COA, respectively (HOA based on

Aiken et al., 2009, Mohr et al., 2012, Crippa et al., 2013b, 2014 and COA

based on Crippa et al., 2013b, 2013c, Mohr et al., 2012);

3. HOA correlates significantly with NOx being the sum of NO and NO2

(defined below);

4. HOA correlates significantly better with NOx than COA; BBOA correlates

significantly with levoglucosan (defined below);

5. SC-OA correlates significantly with CH3SO2+ (defined below);

6. for samples from Zürich and Magadino, where watersoluble organic carbon

(WSOC) data are available, modelled and measured OC mass are comparable

for a set of different conditions (see below and in Supplement A).

The first two criteria (1-2) ensure an appropriate separation of HOA and COA

from OOA and BBOA, respectively. Criteria 3-5 relate to the evaluation of the


70

correlation between factor and marker time series. This was achieved by computing the

Fisher-transformed correlation coefficient z at different stations (Eq. 5.6):

(5.6)

where r is the correlation coefficient between the factor and marker at a given

station. The z values obtained at the different stations are subsequently averaged and

transformed back to ravg before further analysis. A t test is then used to verify the

significance (α = 0.5) of the average correlation coefficient between factor and marker

time series, ravg (Eq. 5.7):

(5.7)

Here, ravg is the correlation coefficient averaged over the different stations,

derived from the average z value, tavg is the corresponding t value, and N is the average

number of samples at the different stations. Results with a significance level of α = 0.05

are summarized in Fig. SI.A.8 in the Supplement A.

To evaluate whether HOA correlated significantly better with NOx than COA did,

the average z values obtained between HOA and NOx and between COA and NOx (Eq.

5.6) were compared using a standard error on the z distribution of (Zar,

1999). The last criterion (6) relates to OC mass closure. A Monte Carlo approach was

applied to evaluate whether a combination of water-soluble factor time series and

recovery parameters would achieve OC mass closure, as described in the following. For

the samples from Zürich and Magadino, for which WSOC concentrations were available

(in contrast to the other samples), offline AMS measurements were scaled to the water-

soluble organic matter (WSOM) calculated using the WSOC measurements and

OM/OC from the HR AMS analysis. The water-soluble contributions from an identified

aerosol source in a sample i were rescaled to their total organic matter concentrations

(OAi,k, where k represents a given factor, using combinations of factor recoveries as

71

determined by Daellenbach et al. (2016, medians of the combinations used being RHOA:

0.11, RCOA: 0.54, RBBOA: 0.65, and ROOA: 0.89, used for WOOA and SOOA). For SC-

OA, whose recovery was not previously determined, a recovery value was stochastically

generated between 0 and 1. The OAi,k concentrations obtained were then converted to

OC concentrations OCi,k, using factor-specific OM/OC determined from the factor

profiles. The sum of OCi,k from all factors k (mod-OCi was then evaluated against the

measured OC (meas-OCi. For this, the residual OC mass (res-OCi for each sample was

calculated (meas-OCi - mod-OCi), and the residual distributions were examined for

different conditions that are specified in the Supplement A. In summary, a solution was

only accepted if res-OCi values were normally distributed around 0 considering all

points and subsets of points: (a) summer, (b) winter, (c) Magadino, (d) Zürich, and (e)

low and high concentrations of the single factors (see Table SI.A.1 in the Supplement

A).

For each of the Monte Carlo simulations, criteria 1-6, which satisfy the water-

soluble factor time series, were used together with a combination of factor recoveries

from Daellenbach et al. (2016) as input data. The WSOC used for scaling the Gi,k matrix

and the meas-OCi used for residual calculation were varied within their uncertainties

(5%) and biases (5%) assuming a normal distribution of the errors. Likewise, constant

biases were also introduced into the initial recovery distributions from Daellenbach et

al. (2016). Monte Carlo simulations were performed and simulations for which res-OCi

distributions were significantly different from 0 (Q25<0<Q75, details in the Supplement

A) were discarded until 500 acceptable simulations were found. Thereby, 331 PMF runs

were selected for PMFblock (230 for PMFzue,isol, 99 for PMFzue,reps, and 269 for

PMF1filter/month. Median factor time series and recovery parameters from all retained

simulations were then determined and the interquartile range (IQR) represents our best

estimate of the uncertainties for the single PMF datasets. The Monte Carlo process was

repeated for the four different PMF datasets described above and the resulting median

time series of their estimated uncertainties were compared. The resulting uncertainty

estimates and the method are described in Sect. 5.4.2.1 and in the Supplement A.


72

5.4 Results and discussions

In this section, the final source apportionment results are presented and validated.

The source signatures are presented in Fig. 5.3 for PMFblock colour-coded with the ion

family. Figure 5.4 shows the time series for Zürich obtained from all PMF approaches

and Table 5.2 summarizes the correlation coefficients between factor and marker time

series for Zürich (all PMF runs) and the other sites in the study area (PMFblock, while the

relation between factor and marker time series is displayed in Figs. 5.4 and 5.5.

Presented are median (and quartile) results for all PMF runs accepted following the

criteria described above.

5.4.1 Interpration of PMF factors

HOA: HOA profile elements were constrained using the reference profile from

Crippa et al. (2013c). The final factor profile (Fig. 5.3) maintains the same features,

characterized by high contributions from hydrocarbon fragments. The fraction of

oxygenated organic fragments that were missing in the initial reference profile, which

were added based on UMR spectra, show an increased contribution to the ions above

m/z=100 (see Sect. 5.3.1). While this indicates a possible overestimation of the

contribution of these fragments, using this methodology, this increase does not

substantially affect the results: e.g. the HOA OM/OC remains low (1.32, IQR 1.30-

1.33). The HOA time series follows an expected pattern that matches the NOx yearly

cycle (Fig. 5.4a), except for San Vittore, which is very likely due to the extremely high

contribution of biomass burning at this site during winter, which may result in

additional NOx inputs and/or may affect the separation of HOA by PMF. The

HOA/NOx (Fig. 5.5a) at the different sites (0.015 - 0.011 µgm-3ppb-1) lies within the

range of literature values (0.001 to 0.028 µgm-3ppb-1, Lanz et al., 2007 and Kirchstetter

et al., 1999). A similar average ratio was obtained for Zürich from the different

sensitivity tests, but with high variability (0.013 - 0.009 µgm-3ppb-1) similar to that

obtained between the different sites. This implies that the observed site-to-site

differences are not statistically significant given our uncertainty in extracting HOA

contributions.

73

Figure 5.3: PMF factor profiles of HOA, COA, BBOA, SOOA, WOOA, SC-OA, color-coded with

ion family of PMFblock (average). fm/z is the relative intensity at a specific mass-to-charge ratio (m/z).

COA: COA profile elements were constrained using the COA profile from

Crippa et al. (2013c) and the obtained factor profile maintains the same features

(OM/OC of 1.32, IQR 1.30-1.33, Fig. 5.3). For COA, no molecular marker is available

for validation purposes. Daellenbach et al. (2016) demonstrated that COA

concentrations can be estimated with offline AMS (in Zürich at the same site) by

constraining its signatures, but only with a high uncertainty. This was performed by

comparing offline AMS results to those from a collocated ACSM, which, owing to its

higher time resolution, enabled the identification of cooking emissions based on their

diurnal cycles (Canonaco et al., 2013). Here, while no ACSM data were available, we

followed the same methodology used in Daellenbach et al. (2016) to estimate the

contribution of COA. The average COA contributions estimated here and their yearly

variability are similar to those from previous studies at the same sites, but as expected

have high uncertainties (Fig. 5.4b).


74

Figure 5.4: HOA, COA, BBOA, SC-OA, SOOA, and WOOA and their respective marker concentrations as a function of time for Zürich in 2013. Depicted are the median factor time series results for the different PMF datasets (median) including the uncertainties for PMFblock (first and third quartile) (green: PMFblock, black: PMFzue,isol, red: PMFzue,reps, pink bullets: PMF1filter/month).

BBOA: BBOA is identified based on its spectral fingerprint (OM/OC of 1.74,

IQR 1.74-1.75; Fig. 5.3), which, similar to previously extracted BBOA factors at other

locations (Daellenbach et al., 2016; Lanz et al., 2007; Crippa et al., 2014), exhibits high

contributions from oxygenated fragments (CHO+, C2H4O2+, C3H5O2

+ from anhydrous

sugar fragmentation (see comparison to nebulized levoglucosan in Supplement A Fig.

SI.A.6). Similar to levoglucosan, the BBOA time series shows an expected seasonal

variation with high concentrations in winter, supporting the identification of this factor

(Fig. 5.4c). Except for Bern and Magadino (7.5 and 11.2), a similar ratio of BBOA to

levoglucosan is found at all other sites (3.9 to 5.7), despite apparent site-to-site

differences in the model residuals during winter due to significantly higher

contributions of BBOA at the southern stations (Fig. 5.5c). The ratios obtained are

within the range of values reported in literature (between 4 and 18 assuming OM/OC)

between 1.6 and 1.8 for the non-AMS analyses; Zotter et al., 2014; Herich et al., 2014;

Minguillón et al., 2011; Crippa et al., 2013a; and Favez et al., 2010). We note that a

similar ratio is also found for the different PMF datasets performed for the case of

Zürich (BBOA/levoglucosan between 3.9 and 12.1). Taken together, the high (for most

75

sites) correlation (R2 = 0.78 for all sites, single sites in Table 5.2) between levoglucosan

and BBOA and their consistent ratios at different sites and between the different PMF

datasets indicates that BBOA is well resolved by PMF at all sites, despite potential site-

to-site differences in BBOA composition.

Figure 5.5: Scatter-plots for the different extreme sensitivity tests for Zürich and for all sites for PMFblock median concentrations): a) HOA vs NOx, b) BBOA vs levoglucosan, c) SOOA vs temperature, d) WOOA vs NH4

+.

SC-OA: Sulfur-containing fragments (e.g. CH3SO2+ are predominantly

apportioned to this factor, which also has a high OM/OC (1.82, IQR 1.80-1.93; Fig.

5.3). As mentioned in Sect. 5.3.6, the recovery of SC-OA was unknown and had to be

determined by mass closure, while the recoveries of the other factors were determined

by comparison to their online counterparts (albeit for a different dataset; Daellenbach et

al., 2016). In the lack of specific constraints (like an online counterpart), the recovery of

SC-OA is highly uncertain and thus the factor time series is also highly uncertain. A

similar factor profile had been extracted from previous online AMS datasets and was

related to the fragmentation of methane sulfonic acid (MSA) present in PM1 particles, a

secondary product of marine origin (Crippa et al., 2013c; Zorn et al., 2008). However,

the SC-OA factor extracted here did not seem to be related to marine emissions because

neither its variability nor its levels matched those of MSA (Fig. 5.4d). First we

compared the MSA levels measured in Zürich using ion chromatography to those


76

estimated based on the concentration of sulfur-containing fragments from offline AMS

measurements in SC-OA (Eq. 5.8), based on Crippa et al. (2013c):

(5.8)

Here, MSAi,est is the estimated MSA concentration, SC-OAi the factor concentration

of the sulfur-containing factor, fSC-OA(CH2SO2+) and the following summands the

fractional contributions of the respective organic fragment to SC-OA, and 0.147 is a

scaling factor from Crippa et al. (2013c). The estimated MSA levels are 6 times higher

than the measured MSA, indicating the presence of another source of sulfur-containing

species. Second, unlike marine OA factors from previous online datasets (lower size

cut-off, typically PM1), the SC-OA time series does not correlate with MSA (R2=0.02).

While MSA concentrations show a clear enhancement during summer, the SC-OA time

series exhibit a very weak seasonal variability with slightly higher concentrations in

winter. SC-OA instead exhibits low background levels episodically intercepted by

remarkable 10-fold enhancements, especially at urban sites affected by traffic emissions

(e.g. the SC-OA contribution is significantly higher at sites with higher yearly NOx

average levels). The hypothesis of an influence of traffic activity on SC-OA is provided

by the correlation of the yearly average concentrations with NOx (Rs,SC-OA,NOx=0.65,

n=9, p<0.06), which is, however, comparable to the correlation of HOA and COA (e.g.

Rs,HOA,NOx=0.68, n=9, p<0.05; Rs,COA,NOx=0.68, n=9, p<0.05). In addition, the SC-OA

time series also correlates with that of NOx (overall R2=0.32, for sites in Table 5.2).

While HOA and BBOA also correlate with NOx, both of the secondary factors, WOOA

and SOOA, do not, supporting the hypothesis that SC-OA consists of locally emitted

anthropogenic (primary) OA. The site-to-site differences in SC-OA concentrations and

temporal behaviour suggest that this factor, which to the best of our knowledge is

reported here for the first time, is influenced by primary sources.

77

Table 5.2: Comparison of factor time series to reference data for different PMF input datasets runs (by Pearson and Spearman correlation coefficient, Rp

2 and Rs). Displayed are the results for PMFblock unless stated otherwise. R2

(number of points)

HOA vs NOx, Rp

2 BBOA vs levo, Rp

2 WOOA vs NH4

+, Rp2

SOOA vs T, Rs

SC-OA vs NOx, Rp

2

Basel 0.31 (91) 0.91 (11) 0.66 (91) 0.70 (91) 0.17 (91) Bern 0.22* (90) 0.48 (12) 0.53 (90) 0.63 (90) 0.17 (90) Frauenfeld 0.40 (89) 0.73 (12) 0.77 (90) 0.63 (90) 0.28 (89) St. Gallen 0.23 (91) 0.39 (12) 0.78 (91) 0.72 (91) 0.50 (91) Magadino 0.18 (91) 0.55 (12) 0.54 (91) 0.72 (91) 0.63 (91) Payerne 0.48 (91) 0.65 (12) 0.44 (91) 0.68 (91) 0.17 (91) Vaduz 0.38 (91) 0.90 (12) 0.77 (91) 0.68 (91) 0.46 (91) San Vittore 0.02 (90) 0.99 (12) 0.36 (90) 0.76 (68) 0.01 (90) Zürich PMFblock 0.35 (91) 0.43 (12) 0.79 (90) 0.65 (91) 0.40 (91) PMFzue,iso 0.29 (91) 0.59 (12) 0.82 (90) 0.66 (91) 0.27 (91) PMFzue,rep (only 12 points)

0.32 (12) 0.23 (12) 0.84 (12) 0.85 (12) 0.01 (12)

PMF1filter/month 0.30 (91) 0.44 (12) 0.77 (90) 0.59 (91) 0.53 (91) * 1 outlier removed.

Oxygenated OA factors: Unlike oxygenated OA factors from limited-

duration intensive online campaigns characterized by a high temporal resolution in

which factor variability is thought to be primarily driven by volatility and/or local

oxidation reactions, OOA factors are resolved based on differences in their seasonal

behaviour: SOOA (in summer) and WOOA (in winter). The SOOA and WOOA mass

spectral signatures (Fig. 5.3) show similarities with OOA from earlier measurements

(Ng et al., 2011b; Canonaco et al., 2013, 2015), with high contributions of C2H3O+ and

CO2+ and high OM/OC, though SOOA (OM/OC=1.89, IQR 1.88-1.89) is less oxidized

than WOOA (OM/OC=2.12, IQR 2.11-2.14). The mass spectral fingerprints (Fig. 5.3),

the temporal behaviour (Fig. 5.4e and f), and the relation to markers (Fig. 5.5c and d) of

the two factors are in agreement with those from earlier work at other locations,

including Zürich (Daellenbach et al., 2016), Payerne (Bozzetti et al., 2016), and

Lithuania (Bozzetti et al., 2017b). This OOA separation appears to be typical for PMF

analysis of long-term, low-time resolution OA mass spectra of filter samples.


78

SOOA correlates significantly among the different sites (also south and north of

the alpine crest) and with local temperature (Fig. 5.5c). The SOOA exponential increase

with average daily temperatures from 5 to 30°C is consistent with the exponential

increase in terpene emissions, which are dominant biogenic SOA precursors (Guenther

et al., 2006). This is also consistent with the mass spectral fingerprint of this factor,

characterized by an fC2H3O+ of 0.10 and an fCO2+ of 0.13, which are similar to values

reported for chamber SOA from terpenes or at an urban location (Zürich) during

summer (Canonaco et al., 2015). A similar temperature dependence of biogenic SOA

concentrations has been observed for a terpene-dominated Canadian forest (Leaitch et

al., 2011) and for the case in Switzerland, using a similar source apportionment model

(Daellenbach et al., 2016; Bozzetti et al., 2016). Taken together, these observations

suggest that SOOA principally derives from the oxidation of biogenic precursors during

summer. Site-to-site SOOA concentrations were not statistically different within our

model errors, assessed from the different sensitivity tests for the case of Zürich.

Therefore, even though the behaviour of SOOA at the different sites studied here might

be controlled by various parameters, including tree cover, available OA mass, air mass

photochemical age, and oxidation conditions (e.g. NOx concentrations), temperature

seems to be the main driver of the SOOA concentrations. Indeed, the aforementioned

parameters may contribute, together with model and measurement uncertainties, to the

observed scatter in the data. Biogenic volatile organic compound emissions might even

be non-negligible in winter (Oderbolz et al., 2013; Schurgers et al., 2009; Holzke et al.,

2006). Therefore, significant wintertime SOOA concentrations are not in disagreement

with the hypothesized biogenic origin. The lower SOOA concentrations in the

temperature range between 7 and 12°C might be explained by often-occurring

precipitation in this temperature range. We note that relative uncertainties related with

SOOA increase with decreasing concentrations (Fig. 5.7). A small error in modelling

sources with high contributions (BBOA, WOOA) in winter can result in a large error of

SOOA with its small contribution during winter. Furthermore, some other sources like

primary biological OA (PBOA; see Sect. 5.4.2.2) might also mix into SOOA.

Compared to SOOA, the WOOA profile can be distinguished by a higher

contribution from CO2+ and a lower contribution from C2H3O+ (Fig. 5.3), similar to

OOA factors previously extracted in this region during winter based on ACSM

79

measurements. This fingerprint is characteristic of highly oxidized SOA from non-

biogenic precursors with low H/C (e.g. aromatic compounds from wood combustion

emissions; Bruns et al., 2016).WOOA is well correlated with NH4+ (Fig. 5.5d; overall

R2=0.65 for all sites, overall R2=0.81 for all PMF runs for Zürich in Table 5.2), which is

in agreement with earlier studies (e.g. Zürich in Lanz et al., 2008). This is probably

explained by its correlation with other inorganic secondary ions NO3- and SO4

2- (driven

like WOOA by meteorological factors including boundary layer height and

temperature), which govern the NH4+ concentration in the aerosol. Here, we have used

ammonium as a proxy for aged aerosols affected by anthropogenic emissions, as

WOOA correlates better with ammonium than with nitrate sulfate.We note that in

winter, whenWOOA is highest, 56% of ammonium can be attributed to nitrate, whereas

in summer ammonium sulfate dominates (97% of ammonium can be attributed to

sulfate). Therefore, WOOA correlates more with nitrate (R2= 0.64) than sulfate

(R2=0.48). WOOA exhibits a regional behaviour and its concentrations are correlated at

all sites on the Swiss plateau. The WOOA mass spectral fingerprint, its seasonal

variability, and its high correlation with long-range transported anthropogenic inorganic

secondary ions suggest that this factor is characteristic of a highly aged OOA influenced

by wintertime anthropogenic emissions (e.g. biomass burning).


80

5.4.2 Uncertainty analysis

5.4.2.1 Model uncertainties

PMF uncertainties depend on the factor contribution. According to Ulbrich et al.

(2009), reliable interpretation of factors with a low relative contribution is challenging.

However, the specificity of the time series and factor profile (caused by rotational

ambiguity), and in this sense also solution acceptance criteria, influence the uncertainty

as well. In our analysis, we correct our results from WSOM to OM using Rk and thereby

introduce additional uncertainties (caused by the uncertainty of Rk or an unknown Rk.

The more uncertain Rk, the higher is the additional uncertainty in the extrapolation (Eq.

5.5). As mentioned in Sect. 5.3.5, Rk constraints (recovery combinations for different

factors) are available for RHOA, RCOA, RBBOA, and ROOA but not for RSC-OA and not for

individual OOA factors (Daellenbach et al., 2016). With the available constraints of

mass closure (for Magadino and Zürich), RSC-OA can only be determined with a high

uncertainty (Fig. 5.6).

Figure 5.6: Distributions of Rk for HOA, COA, BBOA, OOA (WOOA plus SOOA) and SC-OA (500 pairs). A priori information for HOA, COA, BBOA, and OOA on Rk is used from Daellenbach et al., 2016, with propagated errors and biases, while RSC-OA is determined in this study. Distributions of all factors have a resolution of dRk=0.01 except for dRSC-OA=0.05.

81

Figure 5.7: Relative σa (a) and err‘tot (b) for factor concentrations > 0.1 µg/m3 as a function of factor concentration. err‘tot includes the uncertainties from a-value, seed variability and Rk, and the different PMF datasets.

The variability in the factor time series for the single PMF sensitivity tests

(PMFblock, PMFzue,isol, PMF1filter/month, PMFzue,reps is used as an uncertainty estimate

(shaded area in Fig. 5.4). This estimate (σa) depends on the measurement repeatability

(10 single mass spectra included for each sample) and on the selected PMF solution and

Rk combinations, and therefore also on the a value. However, the variability depending

(1) on the choice of input points (time and site; PMFblock, PMFzue,isol, PMF1filter/month and

(2) on the instrumental reproducibility (PMFzue,reps of the offline AMS measurements is

not accounted for. The contribution of (1) and (2) to the uncertainty is assessed through

the sensitivity tests by examining the variability of the median factor time series (σb). σb


82

is the variability of the median factor concentrations from the PMF sensitivity tests

using PMFblock, PMFzue,isol, PMF1filter/month, PMFzue,reps for the 12 samples common to all

4 PMF datasets. For the 12 filters common in all PMF datasets (PMFblock, PMFzue,isol,

PMF1filter/month, PMFzue,reps), we calculate a best estimate of the overall uncertainty (errtot,

by propagating both error terms: σa and σb. As σb is not available for all datapoints, we

parametrized σb as a function of the factor concentration (details in the Supplement A)

and subsequently used this parameterized σb, σ’b, to calculate an approximated overall

error, err’tot. err’tot is displayed in Fig. 5.7b in comparison with σa (Fig. 5.7a). For all

factors, err’tot are in general high, especially for low factor concentrations (~ a factor of

2). It is worthwhile to note that for major factors exhibiting a similar seasonality, i.e.

WOOA and BBOA, a great part of the uncertainty arises from σb. Thus the variability

between the PMF solutions using PMFblock, PMFzue,isol, PMF1filter/month, PMFzue,reps (σb)

and, therefore, the sensitivity of the factor concentrations on the chosen PMF dataset,

significantly contribute to the uncertainty. By contrast, for moderately soluble fractions

constrained in the PMF, COA and HOA, the major part of err’tot is related to σa.

5.4.2.2 Influence of unresolved primary biological OA.

Unresolved sources in PMF are an inherent uncertainty of source apportionment

analyses. As Bozzetti et al. (2016) show, PBOA can present considerable contributions

to OA in PM10 (constituting a large part of coarse OA). In the present analysis, PBOA

could not be separated by PMF (neither unconstrained nor using the mass spectral

signature from Bozzetti et al., 2016). This inability might be caused by the low water

solubility and the absence of PM2.5 filters in the dataset. Since these coarse particles are

only abundant in PM10 and not in PM2.5 or PM1, the presence of both PM10 and

PM2.5 samples, exhibiting a large gradient in PBOA, might allow an unambiguous

separation of PBOA. The aim of this section is to estimate the influence of PBOA on

the source apportionment results. A quantification of this fraction is, however, beyond

the scope of this paper. In the following, we estimate the influence of PBOA in three

alternative ways.

83

Based on factor profiles: Bozzetti et al. (2016) identified the AMS fragment

C2H5O2+ as a possible tracer ion for PBOA. Based on the seasonality of SOOA (high in

summer and low in winter), one can assume that SOOA in this study is a linear

combination of PBOA and SOOA identified in PM2.5 and PM1. Based on the relative

contribution of the ion C2H5O2+ to the factor profiles of SOOA from this analysis and

literature profiles of PBOA and SOOA from Bozzetti et al. (2016, study site: Payerne),

we estimate that 17% of the watersoluble SOOA is in fact PBOA (between 2 and 23%

for the different sensitivity tests). Using this approach, we estimate that PBOA

contributes 0.30 µgm-3 during the warm months (site-to-site variability computed as the

standard deviation of the average concentration of all sites of 0.03 µgm-3). During the

same period, SOOA concentrations are 1.78 µgm-3 (site-to-site variability of 0.18µgm-3)

and OA concentrations are 4.32 µgm-3 (site-to-site variability of 0.44 µgm-3). This

approach is very uncertain, mainly due to the uncertainty in PBOA and SOOA profiles,

the assumption of a constant PBOA contribution to SOOA, and also the uncertainty of

RPBOA.

Based on coarse OC: Bozzetti et al. (2016) showed that coarse OC

(OCcoarse=OCPM10-OCPM2.5) in summer is dominated by PBOA for samples collected at a

rural site in Switzerland (Payerne). For a subset of the samples used in the present work,

OC in the PM2.5 fraction was also analysed (Basel, Bern, Magadino, Payerne, Zürich

accounting for 149 samples in total). For these samples, the OCcoarse contribution to OC

in the PM10 fraction is 16% higher in summer than in winter (site-to-site variability of

4%). This part of OC might be related to resuspension caused by traffic or emissions of

primary biological particles. The ion C2H5O2+ (indicator for PBOA) shows higher

concentrations with increasing OCcoarse concentrations. Therefore, this increment can

tentatively be ascribed to PBOA, which leads to a contribution of 0.55 µgm-3 to OC in

summer (site-to-site variability 0.16 µgm-3). This results in an average summer PBOA

concentration of 1.21 µgm-3 with a site-to-site variability of 0.39 µgm-3 when assuming

an OM/OC of 2.2 (or 0.66 ± 0.21 µgm-3, for OM/OC=1.2); OM/OC range according to

Bozzetti et al., 2016). ForMagadino (2014, Vlachou et al., 2017), OCcoarse represents 8%

of OC in PM10 in winter while this ratio is 25% in summer. It can be assumed that the

difference of 17% in summer can be attributed to PBOA. Extrapolating this estimate to

the overall dataset from 2013 considered in this study and assuming an OM/OC of 2.2,


84

PBOA contributes on average 0.97 µgm-3 to OA in PM10 in summer, with a site-to-site

variability of 0.13 µgm-3 (or 0.63 ± 0.07 µgm-3 OA with for OM/OC of 1.2).

Figure 5.8: Cellulose concentrations as a function of the season and site. For comparison literature data from other years is added European sites: Payerne (Bozzetti et al., 2016, error bars representingthe standard deviation of the measurements in June and July), Puy de Dôme, Schauinsland, Sonnblick, K-Puszta (Sanchez-Ochoa et al., 2007), Birkenes, Hyytiälä, Lille Valby, and Vavihill (Yttri et al., 2011).

Based on cellulose: It has previously been shown that free cellulose contributes

strongly to PBOA (25% of PBOA mass, for measurements made in Payerne during

summer 2012 and winter 2013; Bozzetti et al., 2016). Therefore, we can attempt to use

cellulose analyses on a subset of samples (the same one as for levoglucosan but Bern;

see Sect. 5.2.3) to estimate PBOA concentrations (Fig. 5.8). As seen for the case of

OCcoarse, cellulose concentrations also increase with higher C2H5O2+ concentrations. For

85

the sites with cellulose measurements available (all sites in the study but Bern),

cellulose average concentrations of 0.17 µgm-3 (site-tosite variability of 0.08 µgm-3, in

the warm season 0.18 ± 0.07 µgm-3) are observed, which corresponds to 0.69 µgm-3

PBOA with a site-to-site variability of 0.34 µgm-3 (in the warm season

0.77 ± 0.29 µgm-3), using the cellulose/PBOA from Bozzetti et al. (2016). In this last

study conducted during summer (15 days in June-July 2012), PBOA concentrations of 3

µgm-3 on average (with cellulose concentrations of 0.8 µgm-3) were estimated, which is

clearly above the observation made here. However, Bozzetti et al. (2016) assessed a

shorter time period with diurnal resolution, instead of one sample per month as in the

present work. Cellulose concentrations from other European sites during other years are

consistent with the results in this study (Sanchez-Ochoa et al., 2007; Yttri et al., 2011).

In general, the background cellulose concentrations at the southern alpine sites are

higher and also the temporal behaviour deviates from that observed at the northern sites:

the maximal concentrations are not reached in July-August but rather in May or

October-November. The different seasonality might be caused by different agricultural

procedures. The higher background concentrations of cellulose for the southern Alpine

sites might be caused by interferences from wood burning, which in the absence of

glucose analyses cannot be excluded.

All these PBOA estimates (between 0.3 and 1.0 µgm-3 during the warm season)

are consistently lower than reported in Bozzetti et al. (2016), with a factor 3 to 10 times

lower depending on the site. One should keep in mind that these estimates are based on

limited datasets in both studies (30 samples in Bozzetti et al. (2016) and 12 samples

from the same site in this study).


86

Figure 5.9: Map of Switzerland with yearly cycles. Negative concentrations were set to 0 prior to normalization for display. The OA mass explained by the source apportionment analysis is termed OAexpl.

5.4.3 Factor relative contribution at different sites

In general, the seasonality of the factor time series is consistent for all nine sites in

the entire study area (Fig. 5.9). In summer, SOOA is the main contributor to OA, while

in winter POA (HOA+COA+BBOA) becomes more important, although WOOA still

contributes significantly. In comparison to the sites in northern Switzerland, OA in the

southern alpine valleys is dominated by BBOA in winter, while in the north WOOA

also plays a role. The different factors contribute 0.47±0.12 (HOA, average and site-to-

site variability), 0.31±0.13 (COA), 1.37±1.77 (BBOA), 0.67±0.31 (SC-OA), 1.11±0.23

(WOOA), and 1.31±0.13 (SOOA) µgm-3 for all sites during the entire year (Table 5.3).

In northern Switzerland, POA contributes less to OA (POA/OA=0.3) than in the

southern alpine valleys, where POA/OA is equal to 0.6. Among POA, BBOA is the

most important, with 87% of POA in the south and 42% in the north. The higher

relative contribution of BBOA to POA in the southern alpine valleys than at the

87

northern sites supports the conclusion that the high BBOA concentrations (e.g. 2.45

µgm-3 in Magadino compared to 0.62 µgm-3 in Vaduz) are not only a consequence of

the meteorological situation in the valleys (strong thermal inversion close to the valley

ground) but mainly reflect the emission strength. SC-OA, which is possibly linked to a

local source of rather primary origin, shows clear site-to-site differences, with high

concentrations at a traffic site in Bern (1.25 µgm-3) and low concentrations at a rural site

in Payerne (0.26 µgm-3), for example. SOOA, believed to have strong influences from

biogenic SOA, shows consistently low concentrations at all sites for low temperatures

(0.76 ± 0.67 µgm-3 at 5-15°C) and clearly increased concentrations under warmer

conditions (4.85 ± 1.51 µgm-3 at 25-35 °C).

Table 5.3: Yearly average contribution and uncertainty of resolved factors for PMFblock run for the different sites and the average for all sites. The uncertainty is calculated based on the variability in the yearly averages from PMFblock and the variability between the 4 sensitivity tests. Factor contribution and uncertainty µg/m3 (%)

HOA COA BBOA SC-OA WOOA SOOA

Basel 0.65±0.23 (14)

0.35±0.19 (8)

0.72±0.15 (16)

0.51±0.24 (11)

1.08±0.24 (24)

1.21±0.30 (27)

Bern 0.61±0.23 (11)

0.59±0.29 (11)

0.64±0.14 (12)

1.25±0.45 (23)

1.21±0.28 (22)

1.11±0.29 (21)

Frauenfeld 0.56±0.22 (12)

0.28±0.19 (6)

0.64±0.14 (14)

0.96±0.35 (20)

0.98±0.22 (21)

1.30±0.32 (27)

St. Gallen 0.40±0.20 (11)

0.15±0.16 (3)

0.42±0.09 (1)

0.71±0.27 (19)

0.83±0.19 (22)

1.22±0.30 (33)

Magadino 0.41±0.20 (6)

0.27±0.21 (4)

2.45±0.50 (37)

0.41±0.20 (6)

1.53±0.32 (23)

1.54±0.35 (24)

Payerne 0.34±0.19 (9)

0.15±0.16 (4)

0.54±0.12 (15)

0.26±0.16 (7)

1.00±0.22 (27)

1.41±0.33 (38)

Vaduz 0.43±0.20 (10)

0.27±0.19 (6)

0.62±0.14 (14)

0.84±0.30 (20)

0.93±0.22 (22)

1.22±0.30 (28)

S. Vittore 0.33±0.18 (4)

0.28±0.22 (3)

5.78±1.16 (59)

0.51±0.23 (5)

1.39±0.30 (14)

1.45±0.33 (15)

Zürich 0.54±0.22 (12)

0.41±0.21 (9)

0.51±0.11 (12)

0.62±0.28 (14)

1.01±0.23 (23)

1.35±0.33 (30)

Average 0.47±0.21 (9)

0.31±0.20 (6)

1.37±0.28 (26)

0.67±0.28 (13)

1.11±0.25 (21)

1.31±0.32 (25)


88

5.5 Conclusion

Sources contributing to OA are quantitatively separated and their uncertainty

estimated statistically at nine sites in central Europe throughout the entire year 2013

(819 samples). Thereby, three primary (HOA, COA, BBOA) OA sources are separated

from two secondary (WOOA, SOOA) categories and a yet unknown source explaining

sulfur-containing fragments (SC-OA). BBOA exhibits clearly higher concentrations at

the alpine valley sites in southern Switzerland than at the sites in northern Switzerland.

SOOA, characterized by high concentrations in summer, shows a more than linear

increase with rising temperatures as is observed from biogenic volatile organic

compound emissions and biogenic SOA concentrations. WOOA, the dominant SOA

category during winter, closely correlates with NH4+. The influence of PBOA, not

resolved by PMF, is estimated using, among others, cellulose analyses and could be an

important contributor. Cellulose’s temporal behaviour suggests maximal PBOA

contributions in northern Switzerland during summer, while at the southern alpine sites

maximal concentrations are reached in spring and autumn.

Acknowledgements

This work was supported by the Swiss Federal Office of Environment;

Liechtenstein; Ostluft; the Swiss cantons Basel, Graubünden, and Thurgau; the

Lithuanian-Swiss Cooperation Programme “Research and Development” project

AEROLIT (no. CH-3-ŠMM-01/08); and the IPR-SHOP SNSF starting grant. The

authors at IGE-Grenoble would like to thank the LABEX OSUG@2020 (ANR-10-

LABX-56) for funding analytical instruments.

6 Insights into organic-aerosol sources via a novel laser-desorption/ionization mass

spectrometry technique applied to one year of PM10 samples from nine sites in

central Europe

K. R. Daellenbach1, I. El-Haddad1, L. Karvonen1, A. Vlachou1, J. C. Corbin1,

J. G. Slowik1, M. F. Heringa1, E. A. Bruns1, S. M. Luedin2,a, J.-L. Jaffrezo3, S.

Szidat4, A. Piazzalunga5,b, R. Gonzalez6, P. Fermo6, V. Pflueger2, G. Vogel2,

U. Baltensperger1, A. S. H. Prévôt 1Laboratory of Atmospheric Chemistry, Paul Scherrer Institute, 5232 Villigen, Switzerland 2MABRITEC AG, Riehen, Switzerland 3Université Grenoble Alpes, CNRS, IGE, 38000 Grenoble, France 4Department of Chemistry and Biochemistry & Oeschger Centre for Climate Change Research,

University of Bern,

3012 Bern, Switzerland 5Università degli Studi di Milano-Bicocca, 20126 Milan, Italy 6Università degli Studi di Milano, 20133 Milan, Italy anow at: University of Geneva, 1211 Geneva, Switzerland bnow at: Water and Soil Lab, 24060 Entratico, Italy

Accepted in Atmospheric Chemistry and Physics

Chapter 6. LDI-MS for understanding of OA sources in Central Europe

90

Abstract We assess the benefits of offline laser-desorption/ionization mass spectrometry

(LDI-MS) in understanding ambient particulate matter (PM) sources. The technique was

optimized for measuring PM collected on quartz-fiber filters using silver nitrate as an

internal standard for m/z calibration. This is the first application of this technique to

samples collected at nine sites in central Europe throughout the entire year 2013 (819

samples). Different PM sources were identified by positive matrix factorization (PMF)

including also concomitant measurements (such as NOx, levoglucosan, and

temperature). By comparison to reference mass spectral signatures from laboratory

wood burning experiments as well as samples from a traffic tunnel, three biomass-

burning factors and two traffic factors were identified. The wood-burning factors could

be linked to the burning conditions; the factors related to inefficient burns had a larger

impact on air quality in southern Alpine valleys than in northern Switzerland. The

traffic factors were identified as primary tailpipe exhaust and most possibly

aged/secondary traffic emissions, respectively. The latter attribution was supported by

radiocarbon analyses of both the organic and elemental carbon. Besides these sources,

also factors related to secondary organic aerosol were separated. The contribution of the

wood burning emissions based on LDI-PMF correlates well with that based on AMS-

PMF analyses, while the comparison between the two techniques for other components

is more complex.

6.1 Introduction

Climate and health are strongly affected by atmospheric aerosols (Kelly et al.,

2007; IPCC, 2013), a substantial fraction of which is organic (Jimenez et al., 2009 and

reference therein). This organic aerosol (OA) is a complex mixture of thousands of

compounds (Goldstein and Galbally, 2007), of which only 10-30% have been speciated

by modern techniques (Hoffmann et al., 2011; Simoneit et al., 2004). Therefore, the

chemical composition of OA, its emission sources and formation processes are under

ongoing investigation. OA can be directly emitted as primary particulate matter

(primary OA, POA) or formed through the oxidation of gas-phase precursors with

subsequent condensation or nucleation (secondary OA, SOA). SOA dominates

91

submicron OA at remote sites (90%) and is also a substantial contributor (30-80%) in

the urban environment (Zhang et al., 2011).

Mass spectrometry has significantly advanced the chemical characterization and

quantification of OA. Instruments equipped with electron ionization (EI), such as the

aerosol mass spectrometer (AMS, Canagaratna et al., 2007) and aerosol chemical

speciation monitor (ACSM, both Aerodyne Research, Inc., Ng et al., 2011c; Fröhlich et

al., 2013) provide quantitative online measurements of OA (Jimenez et al., 2016). The

application of positive matrix factorization (PMF) to AMS and ACSM mass spectra has

allowed the separation of different POA sources such as traffic, cooking, and wood

burning, as well as oxygenated OA factors representing SOA (e.g. Jimenez et al., 2009;

Lanz et al., 2007; Lanz et al., 2008; Lanz et al., 2010; Crippa et al., 2014, Canonaco et

al., 2013). While different SOA factors identified by AMS-PMF have been separated

according to their degrees of oxygenation and volatility, information on the different

origins of this fraction is limited due to significant fragmentation of the molecules by

EI. Moreover, SOA from different sources converges to a chemically similar

composition during oxidation (Kroll et al., 2011, Ng et al., 2011a).

Other strategies for aerosol mass spectrometry have provided complementary

information. Ionization by electrospray (ESI) avoids significant analyte fragmentation.

The coupling of ESI to ultra-high-resolution Fourier-transform mass spectrometers thus

provides detailed information on the chemical composition of a sample. However, using

ESI not all compound classes can be detected efficiently due to ion suppression (Furey

et al., 2013; Trufelli et al., 2011; Kourtchev et al., 2013). In addition, high costs and

labor intensity restrict the application typically to smaller sets of samples. Traditional

techniques such as gas- or liquid-chromatography (Hoffmann et al, 2011) coupled to

MS suffer from similar restrictions.

Laser-based mass spectrometers such as laser-desorption/ionization mass

spectrometers (LDI-MS) and two-step laser mass spectrometers (L2MS) are, similarly

as ESI, less affected by fragmentation than EI. Haefliger et al. (2000a and 2000b) show

that such instrumentation allows for identification of the dominant primary sources

using mass spectral fingerprints (tracer m/z’s identified using emission samples and


92

principle component analysis of ambient measurements). LDI-MS instruments may be

subject to matrix effects, i.e. ion formation from a given compound does not only

depend on its abundance but also on the abundance of all other compounds (De

Hoffmann and Stroobant, 1999). Additional matrix sensitivities may arise if ion-neutral

reactions take place in the desorbed plume, which means that MS design (e.g. time-of-

flight vs. ion trap) also plays a role (Murphy, 2007). Although the importance of matrix

effects is evident, studies with a single-particle LDI-MS (ATOFMS, laser wavelength

266 nm) have achieved good correlations of major aerosol components with well-

established reference measurements such as elemental carbon (EC, Sunset Analyzer,

Sunset Laboratory Inc.), OA (AMS), NH4 (AMS), SO4 (AMS), NO3 (AMS), and K+

(collected with a particle-into-liquid sampler, PILS), suggesting that underlying matrix

effects did not dominate measurement reproducibility (Healy et al., 2013).

Similar to the AMS, online single-particle LDI-MS instruments such as the

ATOFMS also yield extensive fragmentation. However, such fragmentation can be

avoided by measuring offline aerosol samples (filters) using other systems. Samburova

et al. (2005a) showed by comparing measurements with and without matrix addition,

that fragmentation was negligible in their instrument (wavelength 337 nm, LDI-MS,

Shimadzu/Kratos, Axima CFR). Overall, offline LDI-MS may therefore provide quick

access to additional chemical information at near-to-molecular level, potentially

allowing differentiation between several primary organic aerosol sources and even

different precursor-related SOA categories (Kalberer et al., 2004; Samburova et al.,

2005a). Based on offline LDI-MS measurements performed for an urban background

site in Zürich, Switzerland (same site as Haefliger et al., 2000a; 2000b), Baltensperger

et al. (2005) suggested that SOA from biogenic precursors is more important than SOA

from anthropogenic precursors. However, such LDI-MS analyses are rare, typically

focusing on pattern analysis in the mass spectrum since LDI-MS signal quantification is

difficult due to the variability in ionization efficiencies and chemistry for different

compounds. In addition, LDI-MS has not been applied to extensive datasets. In

comparison to online analyses, analyzing offline filter samples allows covering longer

time periods and larger observation networks.

93

In this work, we evaluated the use of offline LDI-MS for the direct measurements

of PM collected on filters, without the addition of an ionization matrix. We assessed the

contributions of different aerosol sources to the total PM. To control for instrumental

differences, we measured all filters on the same instrument. We developed novel

procedures for instrument calibration and uncertainty assessment. We measured three

source reference samples (traffic, wood burning, and cooking) in addition to 819

ambient samples. The ambient samples included filters from the entire year of 2013 at

nine sites in central Europe with different emission conditions (Alpine valleys strongly

influenced by wood burning, as well as urban and rural regions). Based on positive

matrix factorization (PMF) using LDI-MS mass spectral data, the ability to resolve OA

sources in source apportionment was assessed and used for obtaining a deeper

understanding of sources contributing to the organic aerosol in central Europe.

6.2 Methods

6.2.1 Sample collection and other chemical analysis

Ambient samples were collected at nine sites in Switzerland and Liechtenstein

(described in more detail in Daellenbach et al., 2017) covering different atmospheric

conditions (urban/rural and background/curbside). Seven sites were located on a plateau

north of the Alpine crest (Basel, Bern, Payerne, Frauenfeld, St. Gallen, Zürich, Vaduz);

the remaining two sites were located in Alpine valleys south of the Alps (Magadino,

San Vittore). Samples were collected on quartz filters (Pall Corp.) by local air quality

monitoring networks every 4th day during the entire year 2013 (819 filters) and used for

gravimetric PM10 quantification. The samples were stored at -18°C and transported in

cooling boxes. Before any further handling steps, the samples were allowed to warm up

at room temperature for 60 minutes in order to avoid condensation of ambient humidity.

The organic (OC) and elemental (EC) carbon content was determined by a thermo-

optical transmission method using a Sunset OC/EC analyzer (Sunset Laboratory Inc.,

Birch and Cary, 1996), following the EUSAAR-2 thermal-optical transmission protocol

(Cavalli et al., 2010). For 33 samples from Magadino, radiocarbon (14C) analyses were

performed (2013; 10 samples, 2014; 23 samples, Vlachou et al., in prep). The

radiocarbon analyses were conducted at the University of Bern at the Laboratory for the

Analysis of Radiocarbon with the Accelerator MS (LARA; Szidat et al., 2014) to


94

determine the contribution of fossil and non-fossil OC (OCf and OCnf, respectively) and

EC (ECf and ECnf, respectively) content (Zhang et al., 2012). Major ion concentrations

were measured by ion chromatography (IC) by a Dionex ICS1000 instrument

(Piazzalunga et al., 2013; Jaffrezo et al., 1998). Levoglucosan was measured by high-

performance anion exchange chromatography (HPAEC) with pulsed amperometric

detection (PAD) using an ion chromatograph (Dionex ICS1000) following the method

of Piazzalunga et al. (2010; 2013a). Gas-phase NOx was measured online using a

chemiluminescence method, and meteorological parameters (e.g., temperature) were

monitored. Further, equivalent black carbon (eBC) was measured with a multi-

wavelength Aethalometer AE 31 (Magee Scientific Inc.) (Hansen et al., 1984; Herich et

al., 2011) in Magadino, Payerne and Zürich. eBC was separated into a wood burning

influenced (eBCwb) and traffic influenced (eBCtr) fraction based on the enhanced

absorption of eBCwb in the ultraviolet range. For this computation, we used the

Ångström exponents for wood-burning (αwb) of 1.7 and for traffic (αtr) 0.9 (Zotter et al.,

2017; Sandradewi et al., 2008). On all the samples, offline AMS analyses (Daellenbach

et al., 2016) were conducted. This involved the analysis of the water-soluble organic

matter (WSOA) by a high resolution time-of-flight AMS. The offline AMS data were

used for source apportionment (Daellenbach et al., 2017; Bozzetti et al., 2016) and used

in this study for comparison; the source mass concentrations were corrected for the

missing water-insoluble mass fraction using the method described in Daellenbach et al.

(2016).

Reference source samples for traffic (representing PM10) were collected in a

tunnel (Islisberg tunnel on Swiss highway A4, exit Wettswil am Albis). The PM10

samples were collected on Sunday, 2014-05-18, and Tuesday, 2014-05-20, from

midnight to noon with 2 hour intervals (rush hour in the morning). Additional source

samples were collected on quartz filters for beech-wood burning through laboratory

experiments of whole cycle burns and stable flaming phase only. These samples did not

only include the primary aerosol, but also two different levels of aging, i.e., after

simulated aging in a smog chamber for 1 hour (equivalent to an OH dose of 107 cm-3 h )

and 4 hours (OH dose 3·107 cm-3 h) (Bruns et al., 2015; 2016). Additionally, also filter

samples collected during cooking experiments were analyzed (Klein et al., 2016a,

2016b).

95

6.2.2 Laser-desorption/ionization ToF MS analysis

6.2.2.1 Instrument and measurement settings

In laser desorption/ionization mass spectrometry, LDI-MS, liquid or solid material is

simultaneously desorbed and ionized by pulsed laser irradiation. The laser is focused on

the surface of the sample. Ions are produced through two major pathways

(Knochenmuss et al., 2000; Zenobi et al., 1998). In the first pathway, ions are formed by

interaction with the laser beam (primary ions). In the second pathway, in the expanding

primary plume of desorbed molecules and primary ions, ion-molecule reactions produce

secondary ions (Knochenmuss et al., 2000). The detailed ionization mechanisms are still

under investigation (Knochenmuss et al., 2002; 2003; 2006; 2016). Although this type

of ionization is considered softer than electron ionization, the observed ions are

typically still fragments of their parent molecules (De Hoffmann and Stroobant, 1999).

Furthermore, the measured intensity of one compound does not only depend on its

concentration but also on the concentration of all other compounds present (the so-

called matrix effect) making quantification challenging (Ellis et al., 2014; Borisov et al.,

2013).

We recorded the mass spectra of 819 filter samples in an m/z range 65-500 thomsons

(Th; 1 Th =1 Da e-1, where e is the elementary charge, ion gate at m/z 60 Th) using a

laser-desorption/ionization-ToF MS (Shimadzu Axima Confidence, Shimadzu-Biotech

Corp., Kyoto, Japan) equipped with an N2 laser (wavelength 337 nm, frequency 50 Hz,

laser pulse width 3 ns, 130-180 µJ/pulse) in the positive reflectron mode. All the

accessible instrumental parameters were kept constant during the whole period of

measurements taking place from November 2015 to mid-March 2016. Specifically, the

laser intensity was adjustable by means of a rotating wheel of filters with varying

transmissivity (0 being blocked and 180 being completely open). We set the wheel

parameter to 105 of 180 which would result in an estimated laser energy of ~6-9

µJ/pulse or 2.8-4.2*108 W/cm2 with a 3ns pulse and 30µm laser beam diameter. While

this wheel the laser energy was initially set (wheel parameter 105 of 180 equaling

roughly 6-9 µJ/pulse or 2.8-4.2*108 W/cm2 with a 3ns pulse and 30µm laser beam

diameter, 0 being blocked 180 being completely open) and kept constant, the aging of

the laser during the given time period was also expected to reduce its intensity. We


96

monitored and assessed changes in laser power and other instrumental parameters, as

well as possible sources of uncertainty/contamination from sample preparation and intra

and inter-day reproducibility, by repeated measurements of a subset of our samples.

Quartz filter punches of 8 mm were attached to a custom-made stainless steel sample

holder (32 slots). Each of the samples was additionally spiked with a droplet of dilute

aqueous silver nitrate solution (AgNO3, Sigma Aldrich, >99.8%, 500 ppt to 20 ppm), to

provide Ag cations as an internal standard for the m/z calibration. After drying under

ambient conditions, filter punches were inserted into the sampling chamber and

analyzed by the LDI-MS. Blanks were measured according to the same procedure. Both

intra-day and delayed repeated measurements were conducted for the same filters to

assess instrument performance and uncertainty/contamination from sample preparation.

The intra-day repeatability was assessed by measuring 3 filters 10 times on 3 different

occasions as outlined in Section 6.3.1. Overall, 96% of the available ambient filter

samples (785 filters) provided usable data (defined below).

6.2.2.2 Data treatment

While most atmospheric LDI-MS studies present raw mass spectra (Samburova et

al., 2005a; 2005b; Kalberer et al., 2004; Baltensperger et al., 2005), in the present study,

we introduced data treatment techniques in order to perform further mass spectral

analysis on stick-integrated spectra. The techniques we employed are described in the

following.

The m/z calibration parameters were highly variable between different samples.

Therefore, to determine unit mass resolution integration regions it was necessary to

perform an m/z calibration on every single sample. However, unlike in the Aerodyne

AMS (N2+, O2

+, W+), there are no dominant anchor ions present in these spectra that

could be used for an m/z calibration. In absence of such ions, we performed a two-step

calibration procedure as described in the following. Each sample was spiked with silver

nitrate (AgNO3, Sigma Aldrich, >99.8%) as an internal standard (approach illustrated in

Fig. 6.1). In order to avoid the suppression of the sample signal, the internal standard

(aqueous solution 500 ppt to 20 ppm) was only placed (as a droplet) on a small part of

the sample. The 499 spectra from all positions on the measurement grid were separated

97

and defined as (1) silver-spiked (lower panel in Fig. 6.1a), (2) silver-free (upper panel in

Fig. 6.1a), and (3) intermediate-silver. Intermediate-silver cases were defined using the

signal intensity in the regions of the mass spectrum where silver was expected, in

comparison to adjacent silver-free regions of the spectrum, and were discarded. To

calculate the first calibration, we calibrated the average silver-spiked spectrum of each

filter sample, using the peaks of the silver monomer (m/z 107, 109 Th), dimer (m/z 214,

216, 218 Th), and trimer (m/z 321, 323, 325, 327 Th). We found that this calibration of

silver-spiked was not directly applicable to the average silver-free mass spectrum.

Possibly, spiking the filter region with aqueous AgNO3 caused enough of a change to

the surface of the filter sample to influence the ionization physics. This could affect the

m/z calibration, since the delayed-pulse ion extraction in our instrument is not

orthogonal to the ablation plume but nearly parallel. Therefore, a second calibration was

obtained for the averaged silver-free spectra using prominent non-silver peaks present in

both spectra. Such a two-step calibration is necessary to achieve accurate m/z

calibrations of the silver-free spectra.

After calibration, mass spectra were baselined using the following custom algorithm.

A window of width 1 Th (the detector bins were approximately 0.04 Th in width) was

applied to identify the lowest signal intensity of that range, which was defined as the

baseline. The window was moved across the mass spectrum with steps of 0.25 Th to

obtain a baseline spectrum. After linear interpolation, the baseline was subtracted from

the spectrum. Subsequently the spectra were integrated to unit mass resolution sticks

(UMR, 1 stick per unit mass). The UMR integration window was not centered at integer

masses, but rather at integer mass plus the mass defect of an alkyl ion (R-CH2+). The

width of this window was defined as extending to the minimum signal intensity above

and below the center using the first derivative of the signal.

To enable a direct comparison of LDI-MS and HR-AMS results, the intensity of the

calibrated and baseline-subtracted spectra (I’m/z) with arbitrary units were scaled to OA

(µg/m3) using the respective OA contents of the filter samples (see Eq. 6.1, with

rescaled intensities termed Im/z thereafter). OCSunset was determined using the Sunset

OC/EC analyzer and (WSOA/WSOC)oAMS through WSOA measurements (described in

Section 6.2.1, WSOC being water-soluble organic carbon). Note that an additional


98

contribution of non-organic material to the signal could not be fully excluded. However,

we consider the detected signal mainly to be related to organics based on the

measurements of ambient aerosol with offline LDI-MS systems using different laser-

wavelengths (193 nm and 355 nm, Aubriet et al., 2010) and pure components (337 nm,

Goheen et al., 1997) indicating that sulfate and nitrate respond only in the negative

mode. Additionally, molecular nitrate and ammonium ions are too small to be detected

in this study (cutoff at m/z 65 Th). We also do not expect clusters between inorganic

cations and organics because we did not observe significant amounts of silver/organic

clusters in silver-spiked spectra (Fig. 6.1a).

(6.1)

6.2.3 Source apportionment / PMF

6.2.3.1 General and input data

Positive matrix factorization (PMF, Paatero and Tapper, 1994; Paatero, 1997) is a

widely used algorithm for source apportionment. PMF (Eq. 6.2) explains the variability

in a dataset (xi,j, here LDI-MS mass spectra scaled to OA, Im/z) as a linear combination

of constant factor profiles (fk,j, here mass spectral signatures) and their time-dependent

contributions (gi,k, here concentrations of this factor). The index i represents a specific

point in time, j the signal at a specific m/z, and k a factor (up to the number of factors p).

(6.2)

PMF is solved by minimizing the residuals ei,j weighed by the measurement

uncertainty, here σPMF,i,j.. In this study, PMF was solved by the multilinear-engine 2

(Paatero, 1999 and references therein) using the front-end user interface SoFi 4.9

(Canonaco et al., 2013) developed for Igor Pro v6 (Wavemetrics). We performed PMF

without constraints of e.g. reference mass spectra of aerosol sources (unconstrained

PMF). The data matrix consisted of the stick-integrated LDI-MS spectra (m/z 65 to 485

Th scaled to OA) of all 785 filter samples. We excluded all m/z’s for which we expected

99

silver signals (listed in section 6.2.2.2) even though only silver-free spectra (according

to the definition given above) were considered.

We determined the PMF uncertainty matrix (σPMF,i,j) based on the instrument

repeatability (described in Section 6.3.1) according to the two-component model for

measurement uncertainty described by Rocke and Lorenzato (1995) and Wilson et al.

(2004). This approach ultimately means that an absolute-error term (e.g., due to noise)

is combined in quadrature with a relative-error term (e.g., due to scaling or calibration

uncertainties) as shown Fig. 6.2 and in Eq. 6.3 (σ’PMF,m/z, for a single mass spectrum).

The need for and applicability of such an error model has been demonstrated by Corbin

et al. (2015) for mass-spectrometry-PMF. We obtained the two error terms by fitting

Eq. 6.3 to the relative standard deviation of replicate measurements for each filter and

experiment, and applied the averaged fitted parameters to the entire data set (constant

error term, , and an error term proportional to the measured peak intensity,

).

(6.3)

The error was scaled to OA in the same way as the LDI-MS mass spectra (Eq. 6.4).

(6.4)

6.2.3.2 Uncertainty estimate of PMF results

A well-established statistical tool for estimating the uncertainty is the bootstrap

technique, which consists of randomly resampling the input data, with replacement, to

create input matrices with the same dimensions as the initial data matrix (Davison and

Hinkley, 1997; Brown et al., 2015). We performed 1’000 bootstrap PMF runs (also

using the replicate measurements) with different initial guesses (“seeds”) using input

matrices of the same size (785 filter samples and 412 ions), resulting in the uncertainty

estimate σbs. We also assessed and parameterized the uncertainty arising from the intra-

day repeatability for three filters analyzed repeatedly by LDI-MS on 3 different days, 10


100

times each day (σintraday, see Supplement B). The reported uncertainty for each PMF

factor is the quadratic sum of σbs and σintraday, as explained in more detail in the

Supplement B. Potential long-term drifts in instrumental response were evaluated

(σinterday, see Supplement B, Fig. SI.B.10) but were not accounted for numerically.

6.3 Results

6.3.1 Calibration, repeatability, and quantification

An example of a sample spectrum after m/z calibration and baseline subtraction,

separated into silver-spiked and silver-free spectra and averaged over the entire filter

punch, is presented in Fig. 6.1a. The silver-spiked spectrum features an intensive signal

at m/z 214 Th related to the 107Ag2 dimer, in contrast to the silver-free spectrum.

The operational mass resolution, m/Δm, based on the measurements conducted on

one sample tray per month and the accuracy of both calibration steps are shown in Fig.

6.1b and 6.1c, respectively. The operational resolution of the instrument was only

determined for the silver mono- (resolution 1100), di- (1700), and trimer (2100) as we

are confident of the absence of interfering ions for these peaks. In comparison to the

LDI-MS, the HR-ToF AMS in V-mode (and W-mode) has a resolution of ca. 2’000

(4’000) at m/z 100 Th and an m/z calibration accuracy < 20 ppm (< 10 ppm, DeCarlo et

al., 2006). With this accuracy and resolution, neither distinguishing different ions at the

same nominal mass nor estimating properties such as O/C and H/C for a nominal mass

following Stark et al. (2015) was possible for the quartz-filter-LDI-MS measurements.

Instead, the spectra were integrated to UMR sticks. The presence of considerable signal

at odd masses might indicate that significant fragmentation occurs during desorption

and ionization of the organic aerosol in the LDI-MS (Fig. 6.1a, 6.3, 6.4). However,

Samburova et al. (2005a) suggested that fragmentation in this instrument is negligible.

The mass spectral signatures acquired in this study are similar to the ones in Samburova

et al. (2005a) and the laser energy applied in this study is lower than in the formed

study. This suggests that also during our measurements fragmentation was not a

prominent process. Overall, the extent of fragmentation remains unclear.

101

The repeatability of the mass spectral signature was assessed in detail for 3 selected

filters (collected in Basel on 2013-06-21, 2013-09-21, and 2013-12-22). Ten punches

from each of these 3 samples were prepared on the sample holder in the same time and

subsequently measured. The given procedure was repeated on 3 occasions (on 2015-11-

25, 2015-12-17, 2016-01-28). The relative error was related to the absolute signal

intensity for each m/z for each sample and experiment (Fig. 6.2), with an asymptotic

value of only ~9 ± 7 % at high signals (average and standard deviation of fitting Eq.

6.3). The average absolute error for small signals was 102 ± 48 a.u. (from Eq. 6.3).

These parameters did not show a temporal trend which suggests that the measurements

are repeatable, despite filter inhomogeneity and laser instability. Analyses of field

blanks (also spiked with AgNO3) exhibited low signal with 99% of peaks below

detection limit (defined as ). The other 1% of the peaks (m/z 197, 249, 251,

322,324 Th) exhibited high variability among the field blank analyses (see details in

Supplement B, Fig. SI.B.1 and SI.B.2). Therefore, no blank subtraction was performed.

The total measured intensity did not show a relation with the filter

loading (OC, OC+EC, PM10, see Supplement B, Fig. SI.B.3), indicating that factors

such as the composition, size, and/or mixing state of the collected aerosol particles had

a larger influence on the measured ion intensities than the mass of PM. Moreover,

replicate measurements indicated that the Itot of a single sample measured on different

instances was affected by an instrumental drift (details in Supplement B, Fig. SI.B.4).


102

Figure 6.1: m/z calibration of LDI-MS analysis of aerosol collected on quartz-fiber filter: a) example of calibrated silver (Ag-spec, blue) and no-silver-containing average mass spectra (noAg-spec, green) of a filter sample. The insert in a) displays a zoom-in of the Ag-spec and noAg-spec. b) operational resolution determined based on silver-mono-, di-, and trimer. c) m/z calibration accuracy for both steps of the m/z calibration.

103

Figure 6.2: Error model parametrization based on 3 samples (Basel, 2013-06-21, 2013-09-21, 2013-12-22) measured on 3 instances with each time 10 repeats on 1 sample holder.

6.3.2 Combustion source profiles

Filters from cooking processes (frying, related largely to oil pyrolysis, Klein et al.

2016a; Allan et al., 2010) did not yield measurable mass spectra in our LDI-MS. This

observation is in agreement with a prior study which performed in-situ single-particle

LDI-MS measurements at 266 nm (using the ATOFMS) and did not observe cooking

particles (Healy et al., 2013). The lack of either graphene-like black carbon or

polycyclic aromatic hydrocarbons (PAHs) to absorb the LDI-MS laser may explain

these observations (Healy et al., 2013), which suggests that our measurements are more

sensitive towards certain PM components. We speculate that it may be possible in some

cases to observe cooking-related ions in LDI-MS of atmospheric samples, if such ions

are secondary ions formed in the ablation plume or cooking particles are coated by

absorbing SOA.

The mass spectra of tunnel filter samples were complex and characterized by distinct

patterns (Fig. 6.3). Some m/z fragments (e.g. 84, 94, 101, 177 Th) were present in the

weekday and weekend tunnel samples, but not in the biomass smoke. Between the two

tunnel samples, large differences could also be identified. The tunnel sample collected

on a Tuesday during rush hour (06:00-08:00) showed many more peaks (among others

163 and 177) which were less prominent during the weekend at the same time and in the

same location (where m/z 101 and 143 Th dominated the weekend spectrum).


104

Figure 6.3: samples representing different combustion sources a) and b) traffic emissions from samples collected in Islisberg tunnel (exit, Wettswil, Switzerland), and c) primary wood burning using whole cycle emissions d) and stable flaming phase emissions d). In absence of a measurable mass spectrum, no spectrum for cooking emissions is displayed.

Previous studies have identified lubrication oil as a major component of diesel- and

gasoline-vehicle exhaust (Gentner et al., 2017; Chirico, et al., 2011), which consists

largely of aliphatic hydrocarbons. Cooking particles similarly consist of such

hydrocarbons (Schauer et al., 2002; Allan et al., 2010; Crippa et al., 2013c). Since we

did not observe a mass spectrum for the cooking sample, but do observe one for the

tunnel sample, our detection of tunnel particles may rely either on the presence of black

carbon or PAHs on the filter samples. To the extent that these species are

heterogeneously distributed within the sampled aerosol particles, our technique may be

specifically sensitive to certain components of traffic emissions.

The mass spectrum of the wood burning sample from the whole burning cycle

shows a complex pattern with a bimodal envelope of relatively-intense ions from m/z

200 Th onwards. By contrast, the sample from the stable-flaming phase shows fewer

peaks (e.g. m/z 85, 140, 213 Th, etc.) with a less significant envelope of ions.

6.3.3 Ambient samples

Typical mass spectra from 4 different locations, known to be heavily influenced by

specific sources at different times of the year are displayed in Figure 6.4. The fragments

are color-coded by their correlation coefficients with 3 different markers or proxies for

aerosol types, i.e., NOx for traffic exhaust (Fig. 6.4a), levoglucosan for biomass burning

emissions (Fig. 6.4c and 4d) and temperature for SOA production (as temperature is

105

expected to exponentially enhance biogenic emissions, Leaitch et al., 2011, Fig. 6.4b).

In general, the mass spectra from winter-time were characterized by higher

contributions from higher-molecular-weight fragments compared to summer. For the

winter sample from San Vittore, a large “hump” was present, on top of which signals

appeared with high intensities and a highly regular pattern of m/z differences of 14 Th.

These fragments, albeit not all detected in smog chamber biomass burning aerosols,

were highly correlated with levoglucosan, a unique marker of biomass smoke.

Figure 6.4: LDI-MS mass spectra from summer and winter samples (dates given in the legend) from a traffic-influenced (Bern), a rural (Payerne), a wood burning-influenced (San Vittore) and an urban background (Zürich) site. Spectra are color-coded with the correlation coefficient (r) between the m/zs and a specific environmental parameter for the whole dataset: for a) the correlation with NOx during summer, for b) the correlation with temperature, for c) and d) the correlation with levoglucosan.

The m/z 84 94, 101, 120, 143, 165, and 177 Th, also detected in tunnel samples,

showed high relative intensities in Bern in summer, of which 94, 120, and 177 Th

correlated with NOx, which suggests a high contribution of traffic emissions at this

location, consistent with previous observations (Zotter et al., 2014). These ions were

also observed, but to a lesser extent, at the rural site of Payerne. At the latter site, the

highest intensities were found for m/z 74 and 104 Th. Only these 2 ions showed a clear

relation with temperature and a clear increase during summer. Overall, this spatial and

temporal variability observed in the spectral fingerprints and the correlation of specific

fragments with environmental parameters suggest that LDI-MS data contain mass


106

spectral information that may be used to separate the contribution from different aerosol

sources. Source apportionment analysis using these data in combination with PMF was

therefore explored in the following.

6.3.4 Source apportionment results

6.3.4.1 PMF setup

Residual analysis of preliminary PMF runs showed that structure in the residuals

was removed when increasing the number of factors up to 5 but not when further

increased. However, when allowing for 7 factors, further environmentally interpretable

separations could be achieved. Increasing to 8 factors led to the separation of a third

traffic related factor, which however contributed little to the overall mass. Therefore, we

opted for 7 factors (see details in Supplement B, Fig. SI.B.6, SI.B.7, and SI.B.8).

6.3.4.2 Interpretation of PMF factors

6.3.4.2.1 Traffic related factors

Two factors could be related to traffic (Fig. 6.5a and 6.5b, termed traffic1 and

traffic2, respectively) based on patterns in the factor profiles similar to patterns in the

samples measured in the tunnel on a weekday (Fig. 6.3a). Since the signatures obtained

from the tunnel filters represent both tailpipe exhaust and resuspended dust, the two

factors identified could be a mixture of both sources. In order to elucidate the reasons

behind the separation of traffic related aerosols into two factors by PMF, we inspected

their relationship with NOx and eBCtr, typical markers of traffic emissions. Both

traffic1 and traffic2 showed increasing concentrations with increasing NOx levels.

However traffic1/NOx and traffic2/NOx were seasonally variable (Rp,traffic1,NOx=-0.17,

Rp,traffic2,NOx=0.24, Fig. 6.6, Fig. SI.B.9). While traffic2 correlated with eBCtr , traffic1

did not show such a dependency (Tab. SI.B.1, Fig. 6.6, Fig. SI.B.9). The correlation

with eBCtr suggests that traffic2 is related to primary emissions from the combustion

process. Based on the lack of correlation between traffic1 and eBCtr, traffic1 might also

be influenced by other processes, such as e.g. secondary production. Additionally,

107

tailpipe exhaust cannot be distinguished from other traffic related emissions in the

tunnel samples and might also contribute to traffic1.

Figure 6.5: PMF factor profiles (colored sticks) and their uncertainty (grey shaded areas, variability among PMF runs): a) traffic1, b) traffic2, c) efficient wood burning (BBeff), d) inefficient wood burning (BBineff1), e) inefficient wood burning 2 (BBineff2), f) lower molecular weight OA (LMW-OA), and g) biogenic OA (bio-OA).

The ratio of both factors to NOx exhibited a clear seasonality, increasing during

summer (Fig. 6.6a and 6.6b). Such a change can either be caused by a change in the

emission patterns, by an enhanced photochemical production of these factors or due to a

change in the lifetime of e.g. NOx. In Fig. 6.6c and d, we compared the contribution of

the traffic related factors with those of eBCtr, another tracer of traffic emissions whose

lifetime is similar to OA (see Fig. SI.B.9 for summer points only). The comparison

shows that while the ratio of traffic2 to eBCtr is not season-dependent, the one of

traffic1 to eBCtr increases during summer. The seasonal variability in the relative

contribution of traffic1 might thus be related to a seasonal change in fleet composition


108

or combustion conditions, or an enhancement of the photochemical production of this

fraction. An enhanced dust resuspension in the warm season (e.g., more dust on the road

due to less precipitation) could also contribute to an increased traffic1 concentration in

summer.

Figure 6.6: Scatterplots between factor-time series and respective markers (traffic1, traffic2, BBeff, BBineff1, BBineff2, bio-OA, LMW-OA, eBCtr, levoglucosan, potassium, ammonium are displayed in µg/m3, NOx in ppm, and temperature in °C).

109

6.3.4.2.2 Wood burning related factors

Three factors could be related to wood burning emissions (Fig. 6.5c, 6.5d, 6.5e). The

first (Fig. 6.5c) showed a similar mass spectral pattern as samples from laboratory

experiments from stable flaming wood burning exhaust from a log wood burner and

thus was termed efficient wood burning (BBeff). The fingerprint of the second (Fig.

6.5d) resembled that of laboratory wood smoke aerosols from the entire burning cycle

including the inefficient starting and burnout phase and, therefore, was related to

inefficient wood burning (BBineff1). BBineff1 showed also a similar signature as

spectra from wood burning haze episodes in San Vittore identified by high levoglucosan

concentrations. The third (Fig. 6.5e) mostly explained the masses above m/z 300 Th.

The signature of the whole cycle wood burning aerosols from a log wood burner (Fig.

6.3c) also showed high relative contributions at high m/z, similar to this factor. For this

reason this factor was termed BBineff2. Both BBeff and BBineff1 correlated with

levoglucosan (Fig. 6.6e, 6.6f, Tab. SI.B.1). Similar to BBeff and BBineff1, BBineff2

correlated with levoglucosan and thus could be related to wood burning emissions (Fig.

6.6g, Tab. SI.B.1). However, BBineff2 also correlated with NH4+, a marker of aged

aerosols, at the northern sites (Fig. 6.6m), which may also indicate that this fraction can

include aged and/or secondary components (Tab. SI.B.1).

Comparing the wood burning related factor time series to potassium (K+), an

inorganic wood burning marker mostly present in ash, provides further insight into the

separation of the 3 wood burning related factors. Among the factors related to wood

burning emissions, BBeff has a lower BB/K+ ratio (2.4, IQR 1.25-4.2) than BBineff1

(6.0, IQR 2.7-12.0) and BBineff2 (11.6, IQR 6.0-21.0). In wood burning experiments, it

was found that OC, EC, PM10, and PAH emissions increase relative to the potassium

output during non-ideal burning conditions (Lamberg et al., 2011). Zotter et al. (2014)

found a north-south gradient in levoglucosan/K+ and OCnf/K+ in Switzerland and

hypothesized that it might be linked to the burning conditions. Thus the higher the LDI-

BB/K+ ratios (north: 16.6, IQR 10.4-30.8 south: 30.8, IQR 22.5-44.2), the less efficient

the burning conditions which in turn supports the hypothesis that BBeff represents the

most efficient burning conditions among the three factors. All wood burning related

factors showed higher median BB/K+ ratios at the southern Alpine valley site (BBeff:

3.0 with IQR 1.1-5.2, BBineff1: 10.9 with IQR 7.5-17.3, BBineff2: 15.8 with IQR 9.9-


110

24.0) than in northern Switzerland (BBeff: 2.1 with IQR 1.2-3.4, BBineff1: 4.3 with

IQR 2.2-8.6, BBineff2: 9.6 with IQR 5.2-18.3) as also visible in Fig. 6.6h/i/j. The north-

south gradient in the BB/K+ ratios might be caused by imperfections of the separation of

the burning conditions by PMF, but other effects such as the age of the stove population

and the used technology could contribute as well. The wood consumption (BFE, 2013)

of automatic burners (>50kW), was higher in northern Switzerland (in the respective

regions, 48m3 wood/km2 area) than in southern Switzerland (8m3 wood/km2 area). Since

the OM, POA, and SOA emissions of pellet burners during the stable phase are

drastically reduced compared to modern logwood burners in a stable flaming phase (e.g.

Heringa et al., 2011), they might contribute over-proportionally to potassium in

northern Switzerland but only little to OA, leading to the higher BB/K+ ratios at the

southern Alpine sites.

6.3.4.2.3 Biogenic-OA, low-molecular-weight OA, and other factor

A factor characterized by high contributions of low-molecular weight ions (Fig. 6.5f,

LMW-OA) correlates with NH4+ suggesting secondary processes as origin (Fig. 6.6l,

Tab. SI.B.1). At the southern Alpine valley sites, the LMW-OA / NH4+ ratio was higher

than in northern Switzerland. As NH4+ is mostly associated with the secondary

inorganic species sulfate and nitrate this variability will mostly be related to differences

in the VOC versus SO2 and NOx emissions, along with temperature differences

influencing the partitioning of nitrate to the particle phase.

The contribution of the last remaining factor showed an exponential increase with

temperature, similar to terpene emissions and biogenic SOA (Leaitch et al., 2011),

suggesting this factor to be strongly influenced by biogenic SOA production. Therefore,

this factor was termed bio-OA. The relationship between bio-OA and temperature was

similar both at the northern and southern sites (Fig. 6.6k, Tab. SI.B.1). Bio-OA had a

highest relative contribution at the most rural site in the dataset (Payerne: yearly average

bio-OA 11%, yearly average NOx concentration: 9 ppb) and the lowest at the most

trafficated site (Zürich: yearly average bio-OA 3%, yearly average NOx concentration:

48 ppb). The chemical signature of bio-OA was dominated by fragments at m/z 74 and

104 Th (Fig. 6.5g). The nature of these fragments remained unidentified, but could not

111

be related to monoterpene or sesquiterpene SOA because of their very low m/z. Further,

the composition of secondary aerosols is expected to be much more complex, showing a

series of fragments distributed over a wide m/z range. We note that the abundance of the

fragments at m/z 74 and 104 Th relative to the total signal should not be directly related

to their absolute concentrations and their ionization efficiency might be superior to that

of other molecules, which may increase their apparent contribution.

6.3.4.3 Comparison to offline AMS and assessment of LDI-MS response factors

The LDI-MS source apportionment results were compared to those based on offline

AMS (oAMS) analyses performed on the same samples (Daellenbach et al., 2017). We

note that this comparison is not straightforward as different sources are separated by the

two methods. For this purpose traffic1 and traffic2 were summed up to LDI-traffic, and

BBeff, BBineff1, and BBineff2 to LDI-BB (Fig. 6.7).

Figure 6.7: Comparison of LDI-MS to reference offline AMS source apportionment results for the sum of traffic related factors (LDI-traffic, R2

LDI,oAMS=0.04), sum of wood burning related factors (LDI-BB, R2

LDI,oAMS =0.83), bio-OA (R2LDI,oAMS =0.62), and LMW-OA (R2

LDI,oAMS =0.45).

As was the case for the comparison to NOx and eBCtr (Section 6.3.4.2), also in the

comparison to the offline AMS traffic (HOA), a higher LDI-traffic/HOA was observed


112

in summer which contributes to the low correlation coefficient (R2=0.04). This might be

related to the fact that HOA is only primary without major contributions of traffic SOA

or dust resuspension while LDI-traffic is potentially also influenced by aged/secondary

traffic aerosol and resuspension. In contrast to summer, we observe a good agreement

between LDI-traffic and HOA in winter. Thus even though a varying relative response

factor (rRF) of the LDI-MS might contribute to these differences, these biases are not

systematic but season dependent. As stated earlier, LDI-traffic is thought to be a

mixture of primary tailpipe exhaust, aged/secondary tailpipe exhaust, and resuspended

dust (as well as tyre break and engine wear). This will be further elucidated in Section

6.3.4.5.

LDI-BB was highly correlated with offline AMS-BBOA (R2=0.83), yet the LDI-BB

concentrations were higher than BBOA from oAMS, especially in northern Switzerland

(with an LDI-BB:AMS-BBOA ratio between 1.4 and 4.1 for the different sites). A

possible reason is also here the mixing of secondary components into LDI-BB when

comparing to primary BBOA from oAMS. However, we cannot exclude that this effect

is due to different rRFs of the LDI-MS for different compound classes.

The identified secondary components, LMW-OA (R2=0.45) and bio-OA (R2=0.62),

also correlate with the corresponding OOA factors from the oAMS analysis, WOOA

and SOOA, yet the correlation coefficients are smaller than for LDI-BB and BBOA. For

these factors, differences between the two methodologies could be related to differences

in the PMF performance or to differences in the response factors for different

components in the LDI-MS (the LMW-OA: WOOA ratio is between 0.7 and 1.1 and the

bio-OA: SOOA ratio between 0.2 and 0.4).

At some of the sites analyzed with LDI-MS in this study, OA was monitored in

previous years with state-of-the-art online aerosol mass spectrometry (AMS or ACSM).

Earlier campaigns with quantitative online AMS analyses show a higher SOA

contribution at those sites: e.g. in Zürich, July 2005 (66% vs 25% for LDI-MS summer)

and December 2005 (55% vs 14% for LDI-MS winter), in Roveredo (close to San

Vittore), in March 2005 (53% vs 46% for LDI-MS summer), and in December 2005

(43% vs 11% for LDI-MS winter), in Payerne, in July 2005 (94% vs 47% for LDI-MS

113

summer) and in December 2005 (71% vs 20% for LDI-MS winter) (Lanz et al., 2010).

Canonaco et al. (2013) presented source apportionment results for the winter of 2011

and 2012 for Zürich (PM1 ACSM), also with higher SOA contributions (71%) than the

LDI-MS in 2013 (20%).

Overall, the source apportionment results based on LDI-MS data provide source

separations with similar temporal behaviors as offline AMS and online AMS and

ACSM analysis, yet seem to overestimate combustion related primary particle sources

and underestimate secondary OA. However, we assume for this comparison that all

factors give an equal response at a given concentration, i.e. the relative response factor

(rRFLDI) of all factors to be 1. The relative contribution of a factor k to the total signal

observed with the LDI-MS for a specific measurement (i) depends on the

factor concentration ( as well as the sum of all factors separated

for the LDI-MS data. Assuming that differences between the LDIr-PMF and the AMS-

PMF arise solely from different rRFLDIs for different factors, is a function of the

AMS factor concentrations and the rRFLDI of this factor (Eq. 6.5):

(6.5)

In order to determine rRFLDI for the LDI-MS several strong assumptions are

required: (1) the sum of traffic1 and traffic2 represents HOA, (2) the sum of WBeff,

WBineff1 and WBineff2 represents BBOA, (3) LMW-OA represents WOOA, (4) bio-

OA represents SOOA and (5) the AMS factors for which there is an LDI-MS

equivalent, (1) - (4), are the only contributors to OA. In scenario 1, all the above

assumption are considered to be true. For scenario 1, we estimate rRFLDI-BB as 1.84,

rRFLDI-LMW-OA as 0.29 and rRFLDI-bio-OA as 0.40 (using LDI-traffic as reference factor,

rRFLDI-traffic = 1.00). The LDI-MS factor concentrations corrected using rRFLDI show a

close relation to the uncorrected LDI factor concentrations (R2LDI-traffic = 0.78,

R2LDI-BB=0.94, R2

LDI-LMW-OA=0.80, R2LDI-bio-OA=0.85, Fig. 6.8).


114

Table 6.1: relative response factors (rRF) for LDI-MS analyses for 3 different scenarios. scenario 1 scenario 2 scenario 3

rRFLDI-traffic = 1.00±0.00 rRFLDI-traffic2 = 1.00±0.00 rRFLDI-traffic = 1.00±0.00

rRFLDI-BB = 1.84±0.06 rRFLDI-BBeff+LDI-BBineff1 = 1.26±0.06 rRFLDI-BB = 5.67±0.13

rRFLDI-LMW-OA = 0.29±0.02 rRFLDI-LMW-OA+LDI-BBineff2 = 1.67 ±0.07 rRFLDI-LMW-OA = 0.30±0.02

rRFLDI-bio-OA = 0.14±0.01 rRFLDI-bio-OA+LDI-traffic1 = 0.39±0.02 rRFLDI-bio-OA = 0.65±0.02

We tested the sensitivity of the rRFLDI estimates (Tab. 6.1) to the assumptions (1)

and (2) being wrong (scenario 2) and only assumption (5) being wrong (scenario 3). In

scenario 2, we alter scenario 1 by comparing the sum of traffic1 and bio-OA to SOOA

and the sum of BBineff2 and LMW-OA to WOOA when computing rRFLDI. For

scenario 2, LDI-MS factor concentrations corrected using rRFLDI show also a close

relation to uncorrected LDI-MS factor concentrations (R2LDI-traffic2=0.96,

R2BBeff+BBineff1=1.00, R2

LDI-BBineff2+LMW-OA=0.99, R2traffic1+bio-OA=0.96). In scenario 3, we

alter scenario 1 by considering also AMS factors without an equivalent in the LDI-MS

PMF. Therefore, we compare LDI-traffic to the sum of HOA, COA, and SC-OA. Under

these conditions LDI-MS factor concentrations corrected using rRFLDI show a close

relation to uncorrected LDI-MS factor concentrations (R2LDI-traffic=0.75, R2

LDI-BB=0.90,

R2LMW-OA=0.74, R2

bio-OA=0.85). The differences in rRFLDI between scenarios 1, 2, and 3

highlight the uncertainties caused by plausible violations of the underlying assumption

when determining rRFLDI. Overall the estimates rRFLDI are highly uncertain and their

accurate determination needs to be the focus of future work. Given the high uncertainty

of rRFLDI and the good correlation between rRFLDI corrected and uncorrected LDI-MS

factor concentrations, we present uncorrected results without considering rRFLDI.

115

Figure 6.8: Comparison of LDI-MS factor concentrations corrected using relative response factors (rRF) to uncorrected LDI factor concentrations (Scenario 1).

6.3.4.4 Uncertainty of PMF results

The uncertainty estimate, σtot, includes both the statistical uncertainty (σbs) and the

uncertainty arising from the intraday variability of the measurements (σintrad, see Fig.

6.9, Section 6.2.3.2 and Supplement B). In order to assess the impact of the intraday

variability consideration (σintrad) in estimating the uncertainty besides the statistical

uncertainty obtained from the bootstrapping approach (σbs), we compared the ratios

σbs/σtot for the different factors: for traffic1 the ratio was 0.83, for traffic2 0.75, for

BBeff 0.86, for BBineff1 0.80, for BBineff2 0.72, for LMW-OA 0.68, and for bio-OA

0.68. Thus, it was important to propagate σintrad. The relative uncertainties for the

median factor concentrations ranged within 0.15 (for traffic1), 0.16 (bio-OA), 0.17

(LMW-OA), 0.18 (BBineff2), 0.20 (traffic2), 0.22 (BBeff, and 0.28 (BBineff1). Unlike

the relative error of σtot of traffic1 (0.63 at the 10th percentile concentration and 0.13 at

the 90th percentile concentration), traffic2 (0.68 and 0.11), BBeff (0.55 and 0.18),

BBineff1 (0.47 and 0.20), BBineff2 (0.37 and 0.14), and LMW-OA (0.27 and 0.16) the


116

relative error of bio-OA (0.20 and 0.15) only depended weakly on the factor

concentration.

Figure 6.9: Relative for the different PMF factors as a function of the factor concentration.

Throughout the measurement campaign, subsets of filters were analyzed

repeatedly in order to assess the repeatability over longer time periods (see details in

Supplement B, Fig. SI.B.10). Most factors did not show significant changes of the

attributed concentration during the measurement campaign. However, BBeff showed

decreasing and LMW-OA increasing concentrations as a function of the measurement

time. This could suggest an uncertain separation of these 2 factors. However, for these

long time delays the variability increased and only few samples were repeated with such

long time delays. Furthermore, the intra-day variance largely explains the total variance

(traffic1 97%, traffic2 94%, BBeff 85%, BBineff1 89%, BBineff2 82%, LMW-OA

79%, bio-OA 97%, details in Supplement B Fig. SI.B.10).

6.3.4.5 Factor variability and contribution

The time series of all factors are shown for the 9 sites in Fig. 6.10 as relative

contributions to OA and summarized as yearly averages in Tab. 6.2. In the yearly

average for all sites, traffic1 contributes 7%, traffic2 12%, BBeff 7%, BBineff1 17%,

BBineff2 32%, LMW-OA 21%, and bio-OA 5% to OA as measured by LDI-MS.

117

Figure 6.10: Relative factor time series of 7 identified factors for all nine sites in study area.

At the southern Alpine Swiss sites, the wood burning influenced categories

(WB=BBeff+BBineff1+ BBineff2) contribute more than at the northern sites (70% vs

50%). Moreover, in winter WB explains 81% and 61% of OA in the south and the

north, respectively. This difference is mostly caused by the higher relative contribution

of BBineff1 (34% in the south vs 14% in the north) since BBineff2 (28% and 30%) and

BBeff (7% and 5%) do not show strong geographical differences. The ratio

BBineff1/BB (Fig. 6.11b) shows enhanced contributions of BBineff1 at the southern

Alpine sites, especially during high pollution episodes in winter. This suggests different

wood burning regimes in the 2 regions, as already discussed above.


118

Figure 6.11: a) time series of traffic1 normalized to eBCtr and ECf in comparison to OCf /ECf in Magadino, b) influence of inefficient wood burning emissions (BBineff1) in comparison to the sum of wood burning influenced factors (BB=BBineff1+BBineff2+BBeff) for the entire datasets.

After scaling of the mass spectra to OA, absolute factor concentrations have still to

be interpreted with caution. One reason for this is that the relative response factors of

the sources/factors are not known, another is that certain species (in analogy to the

cooking-aerosol sample) may not be detected by LDI-MS if externally mixed. In

support of the quantitative interpretation of our results, we note that the sum of the

wood burning related LDI-MS factors correlates well with the offline AMS counterpart

(R2LDI,oAMS =0.82, described in Section 6.3.4.3). On the yearly average 24% of the

measured OA is apportioned to secondary OA. In summer, a bigger fraction is attributed

to secondary OA (35% compared to 16% in winter). 35% (summer) and 1% (winter) of

the secondary OA is attributed to biogenic sources. For some samples from the same

period in Magadino, also the fossil and non-fossil content of OC and EC was

determined (Vlachou et al., in prep) based on the method of Zhang et al. (2012).

Vlachou et al. (in prep.) observed increased OCf/ECf ratios in summer which suggests

119

other fossil POA sources in summer than in winter, or secondary formation of fossil

OA. The higher traffic1 / eBCtr and traffic1/ECf ratios in summer are in agreement with

enhanced OCf /ECf ratios (Fig. 6.11a). However, a part of traffic can also be mixed into

BB leading to an underestimation of the traffic concentrations in winter. Overall this

suggests that traffic1 represents aged or secondary traffic OA.

Table 6.2: yearly averages of the relative factor contributions and NOx concentrations. yearly

average, % traffic1 traffic2 BBeff BBineff1 BBineff2 bio-OA LMW-OA NOx, ppb

Bern 11 24 7 13 30 3 12 48 Zürich 9 27 8 9 27 5 15 24 St. Gallen 8 16 4 17 27 7 21 24 Basel 6 12 7 17 32 5 21 21 Frauenfeld 8 16 6 13 30 7 20 21 Vaduz 7 12 7 20 27 7 20 20 Payerne 5 15 12 9 25 11 23 9 Magadino 3 11 7 27 27 8 17 20 S. Vittore 2 4 3 40 32 5 14 18

6.4 Summary and conclusions

In this study, we advanced a known method for the chemical characterization of

particulate matter collected on quartz-fiber filters by LDI-MS and applied the method to

819 samples. The method included the use of silver nitrate for m/z calibration and the

automated peak integration of the mass spectra at unit-mass resolution. The benefit of

LDI-MS measurements for the chemical characterization and a better understanding of

the sources contributing to the ambient PM10 was assessed at nine sites in central

Europe throughout the entire year 2013.

Wood combustion smog chamber experiments revealed an influence of the

burning conditions on the mass spectral signature. Tunnel samples used as a reference

for traffic related emissions show mass spectral signatures distinctly different from

wood combustion. Key m/z’s identified in the wood burning and traffic signatures

showed links to expected markers as e.g., levoglucosan and NOx, respectively. The

ambient mass spectral information was further used for source apportionment by PMF.

Thereby, the influence of efficient and inefficient wood burning was separated. The


120

extracted wood burning emissions correlated with the results from offline AMS source

apportionment. Other components are more difficult to compare quantitatively because

of different source separations in PMF as well as differences in the relative response

factors (rRF) of OA components. rRF determined in this study are uncertain and,

therefore, not used for correcting the LDI-PMF results. The influence of traffic

emissions was represented by 2 factors. One of these could clearly be linked to BC-

related traffic (eBCtr) and NOx, and thus to primary emissions. The other, when

normalized to eBCtr, showed a similar behavior as OCf /ECf, and was therefore

attributed to aged/secondary traffic OA. A factor was attributed to biogenic SOA based

on its concentration exponentially increasing with temperature. Another OA factor was

characterized by low-molecular-weight ions and was correlated with NH4+ and was

attributed to SOA from an unknown source.

Acknowledgements

This work was supported by the Swiss Federal Office of Environment,

Liechtenstein, Ostluft, the Cantons Basel, Graubünden, and Thurgau. We also thank

AWEL Zürich for providing us with samples collected in Islisbergtunnel. JCC received

financial support from the ERC under grant ERC-CoG-615922-BLACARAT.

121

7 Conclusions and outlook

Linking long-term particulate matter air pollution exposure and the contributing

sources to health effects is important to develop effective mitigation strategies. In this

dissertation, we contribute to this research topic by developing analytical approaches

capable of providing long-term source contributions. In particular, we characterized an

offline application of HR-ToF-AMS based on particulate matter extracted from filter

samples. We applied this technique to create long-term source contribution estimates to

OA at several sites in central Europe. Furthermore, we optimized a method using laser-

desorption/ionization mass spectrometry for measuring PM collected on filter samples.

This type of analysis was applied to the same set of samples as offline AMS and was

used in order to obtain a better understanding of the chemical nature of the contributing

PM sources.

In Chapter 4, we present the method development and characterization of the

offline AMS application using water-extracted PM from filter samples. While

oxygenated fragments are well recovered, hydrocarbon-like fragments are not well

captured. The key element of this study is the comparison of offline AMS source

apportionment results to collocated source apportionment results based on online

ACSM analyses. This comparison yields recovery factors of the different resolved

factors: traffic, cooking, biomass burning, and oxygenated OA. They are essential for

estimating the source contributions to OA based on the offline AMS analysis of WSOA.

Conclusions and outlook

122

For 9 sites in central Europe, sources contributing to OA are estimated using offline

AMS (Chapter 5). For quantification and validation of the source apportionment results,

offline AMS analyses are coupled with state-of-the art measurements of water-soluble

carbon, organic carbon, elemental carbon, water-soluble ions and levoglucosan as an

organic marker for wood burning. In this study, the contribution of primary OA from

traffic, cooking, and biomass burning, besides two seasonally separated OOA factors

could be separated. BBOA shows the highest geographical variability among all factors:

at the alpine valley sites concentrations are clearly higher than in northern Switzerland

and Liechtenstein. This is hypothesized to be related to differences in the burning

conditions between the two geographical regions. Different sensitivity tests also

including the measurement reproducibility allow a detailed examination of the

uncertainty of the factor contributions. Furthermore, cellulose analyses (enzymatic

degradation) are used among other approaches to estimate the contribution of primary

biological OA which could not be resolved based on the offline AMS analyses.

Together with results presented in other studies (Fig. 7.1), the offline AMS approach

yielded results for 13 sites in Europe and for a high pollution episode in China (4 sites).

These wide applications show the potential of the offline AMS approach.

Figure 7.1: Average contributions of OA sources resolved using offline AMS data discussed in this thesis and other studies (this dissertation, and from Bozzetti et al., 2016, 2017a, 2017b, Huang et al., 2014).

123

In a third study (Chapter 6), we optimized the application of laser-

desorpion/ionization mass spectrometry for analyzing PM. Subsequently this approach

was applied to the same samples as analyzed by offline ASM (in Chapter 5) and the

benefit of LDI-MS in understanding OA sources was assessed. In this study, the

influence of traffic (two factors), wood buring (three factors) was separated from two

factors related to secondary OA. Comparison of the cumulative influence of wood

burning resolved by LDI to offline AMS and collocated levoglucosan give trust in the

separation of biomass burning OA. The three wood burning factors could be related to

the burning condition and inefficient burns show a higher contribution at the alpine

valley compared to the site on the northern swiss plateau. This might explain the higher

BBOA concentrations observed with offline AMS. Furthermore, one of the traffic-

related factors showed higher factor concentrations per eBCtr during the warm season

which coincides with higher OCf to ECf ratios at the same site making it possible that

this factor resolves aged traffic emissions.

While routine air quality monitoring encompasses measurements of some gas-

phase constituents, typically PM10 (alternatively PM in other size fractions as PM2.5 or

PM1) concentration is the only parameter assessing PM. Such samples can be accessed

worldwide, also a posteriori. This makes the assessment of past long-term exposure

possible. Moreover, information on extreme pollution episodes, which are only partly

predictable, is accessible through a posteriori analysis. Furthermore, such offline

datasets can be completed with any necessary complementary filter-based chemical

analysis specific to the research question. In contrast to online, offline measurements

also allow testing the measurement repeatability allowing detailed uncertainty

assessments. Centralized analysis of OA sources allows comparing results from

different sites and also limits the influence of instrumental differences by only using one

instrument in the lab, while similar online datasets are much more subjected to such

influences.

Performing measurements online has the major advantage of a better time-

resolution which helps understanding variabilites on a diurnal scale. The low time-

resolution limits analyzing episodes with offline approaches. Increasing the temporal


124

resolution (several filters per day) would result in a better understanding of the

variabilities and would provide further constraints based on diurnal patterns of the

resolved sources. The impossibility to resolve PBOA (Chapter 5) is closely related to

the absence of size resolution (in contrast to Bozzetti et al., 2016). Therefore, for future

studies it is recommended to include PM1 (or PM2.5) filter samples at the cost of

reducing the number of sites which enables the further study of sources contributing to

coarse OA. Including PM1 samples, furthermore, allows a direct comparison to online

AMS/ACSM results and, thus, the determination (confirmation) of the source-specific

recoveries at other sites. Expanding the temporal coverage (to e.g., 4.5 years) using

PM1 and PM10 samples while keeping the temporal resolution constant (1 sample

every 4 days) and only focusing on 1 site, also the number of samples remains constant.

In Dockery et al. (1993), data from 14-16 years is compiled. When reducing the

temporal resolution to 1 sample every 12 days and still keeping the number of samples

constant, the temporal coverage can be further expanded to 14 years. This way it might

be possible to apportion health effects to chronic exposure to single OA sources (given

quantitative source contributions from offline AMS).

The recent development of a time-of-flight unit for the AMS with double the

resolution allows a better separation of ions at the same nominal mass. This is especially

helpful for nitrogen containing species, typically shadowed by other more prominent

peaks. Such ions are possibly important in the coarse PM fraction (Bozzetti et al.,

2016). Yet the AMS OA measurements suffer from an artifact which increases OA

concentrations (CO2+) when injecting nitrate (Pieber et al., 2016). As routinely

performed during offline AMS measurement, regular ammonium nitrate calibrations

should also be performed during online AMS/ACSM campaigns. Offline AMS

measurements also access larger particles than AMS/ACSM. Thus also the interference

with carbonates (decomposing thermally leading to the detection of CO2+) is important

and leading to further inorganic interference at this fragment.

Interference with air (N2+) makes it under most conditions impossible to resolve

the fragment CO+ in AMS analyses (online and offline). As already shown by Bozzetti

(2017b) it is possible to overcome this limitation in offline AMS by changing the carrier

gas to argon making valuable information accessible. As mentioned in Chapter 4,

125

offline AMS analyses are not quantitative alone but need WSOC and/or OC

measurements. By spiking the liquid extracts with isotopically labelled compounds the

main inorganic (15NO3-, 34SO4

2-) can be retrieved quantitatively. Establishing similar

approaches for WSOA might make collocated WSOC and ion chromatography

obsolete.

The sulfur-containing OA factor separated in Chapter 5 is hypothetically linked to

traffic activity which might stem from dust resuspension. For a better understanding of

this influence, separated only in PM10 analyses, further work is required which could

involve the analysis of tunnel samples in different size fractions. Based on the

quantitative source apportionment results obtained from offline AMS, further studies

can be conducted. By performing further measurements of health related parameters

(e.g., ROS, exposing cells to filter extracts) or collecting information on the number of

hospital entries related to respiratory diseases, the implications of the different sources

might be quantified. Completing the dataset with trace metal analyses might also allow

the separation of the impact of organic sources from the one of inorganic species.

Coupled offline AMS and UV-VIS absorption measurements, potentially allow

estimating the direct RF forcing caused by different OA sources. As stated earlier, the

lack of information on the preindustrial aerosol contributes largely to the uncertainty in

estimating RF. Model validation requires datasets like the one presented in Chapter 6

(comparable to the intercomparison in Ciarelli et al., 2016, based on Crippa et al.,

2014). Since any liquid matrix is potentially accessible to offline AMS, it is tempting to

attempt to access information on the bulk OA from preindustrial times using ice and

sediment cores. Environmental matrices in general can be analyzed, such as rain, cloud

and runoff or also ground water.

The LDI based technique does not require an extraction step and thus does not

suffer from losses during such a step, but only matrices desorbing and ionizing upon

interaction with the laser can be detected. Therefore, a better characterization of the

laser (power density) is crucial. The instrument used in Chapter 6 does not extract the

produced ions orthogonallay to the ToF unit which potentially reduces both the

accuracy and precision of the m/z calibration. Another factor potentially playing a role

in the quality of the m/z calibration and reproducibility is the sample surface, i.e.


126

collecting aerosol on a smoother surface like an aluminium plate instead of a quartz-

fiber filter might improve the chemical resolution (attribution of signal at a nominal

mass to different ions/fragments). Similarly also an extraction step and subsequent

deposition of the extract on a steel plate could be applied at the cost of losses during the

extraction. Such improvements would potentially provide further insight into the

precursors of SOA. Moreover, the effect of matrix effects needs to be further studied

explicitly, e.g. by comparing LDI source apportionment results in different

environments (e.g., marine, minimal biogenic emissions) to results from other

techniques. Even if factors like LDI-BB correlate with the corresponding offline AMS

factor, further research is required to understand response factors of different sources

before achieving quantitative relative source contributions. The approach outlined for

LDI analyses of ambient PM samples (Chapter 6) could also be applied to ambient

microorganism samples (where different internal standards are required to account for

the different m/z range) to separate the different species/families/etc.

While ca. 35% of the global city population experiences decreasing PM levels

(mostly in Europe and America), 30% of the city global population suffers from

increasing annual average PM levels between 2008 and 2015 (Eastern Mediterrannean,

Southeast Asia). For Africa similar information is not available. In the Eastern

Mediterrannean, Ryiadh (Saudi Arabia), Ma’amer (Bahrain), Greater Cairo (Egypt),

Doha (Qatar), Abu Dhabi (U.A.E.) are pointed out and in Southeast Asia Delhi (India)

and Dhaka (Bangladesh), but also in other regions there are cities with extreme PM

levels such as Ulaanbaatar (Mongolia), Beijing (China) and Dakar (Senegal) (WHO,

2016, sorted by decreasing PM10 levels within the geographical region). Since not all

these sites are easily accessible to online instrumentation, filter-based offline analysis

approaches can be pointed out as a key tool for analyzing PM. Certain cities are

suspected to be strongly affected by dust from adjacent desertous regions. In such

situations, interference with carbonates are important when using offline AMS, making

it necessairy to resolve (or measure) carbonates or to use alternative measurement

approaches (e.g., acid fumigation of samples). Some sites might also be strongly

affected by several non-water-soluble sources (e.g., in proximity of oil refineries)

making the bulk OA only little water-soluble (as e.g. observed for Baghdad, Hamad et

al., 2015). In such situations, it might be advantageous to use different solvents (offline

127

AMS) or to switch to approaches without an extraction step such as the LDI technique

presented in Chapter 6.

Source apportionment results based on offline AMS and LDI data allowed

separating SOA seasonally, but both approaches could not identify SOA from different

precursors by mass spectral comparison. Offline AMS is limited by the hard ionization

leading to the loss of chemical information required for distinguishing SOA types based

on their precursor and LDI among other factors by the poor chemical resolution and not

well understood ionization pathways. To deepen our understanding of SOA types,

instrumentation equipped with soft ionization and a high chemical resolution is needed.

The recent application of extractive electrospray ionization (EESI) provides more

chemical information (given a high enough chemical resolution defined by the ToF unit

or addition of other instrumentation). In a similar way as the AMS, mass spectrometry

coupled with EESI can be used offline for nebulized filter extracts. Solvent impurities

can be thought only to affect single ions because of the soft ionization pathway which

makes it possible to use, other than in the current offline AMS framework, also organic

solvents like methanol. The use of alternative solvents makes more compound classes

accessible than for the current offline AMS setup (e.g., traffic emissions, compare

Chapter 4). Analyses with ultra-high-resolution mass spectrometry (e.g., Orbitrap)

provide even more chemical information than EESI MS but require extensive lab work

and are, therefore, probably not applicable for studies with a similar amount of samples

as in the presented dissertation.


128

129

8 Bibliography

Aiken, A. C., Salcedo, D., Cubison, M. J., Huffman, J. A., DeCarlo, P. F., Ulbrich, I.

M., Docherty, K. S., Sueper, D., Kimmel, J. R., Worsnop, D. R., Trimborn, A.,

Northway, M., Stone, E. A., Schauer, J. J., Volkamer, R. M., Fortner, E., de Foy, B.,

Wang, J., Laskin, A., Shutthanandan, V., Zheng, J., Zhang, R., Gaffney, J., Marley, N.

A., Paredes-Miranda, G., Arnott, W. P., Molina, L. T., Sosa, G., and Jimenez, J. L.:

Mexico City aerosol analysis during MILAGRO using high resolution aerosol mass

spectrometry at the urban supersite (T0) - Part 1: Fine particle composition and organic

source apportionment, Atmos. Chem. Phys., 9, 6633-6653, doi:10.5194/acp-9-6633-

2009, 2009.

Albrecht, B. A.: Aerosols cloud microphysics, and fractional cloudiness, Science,

245, 1227-1230, 1989.

Alfarra, M. R., Coe, H., Allan, J. D., Bower, K. N., Boudries, H., Canagaratna, M.

R., Jimenez, J. L., Jayne, J. T., Garforth, A. A., Li, S.-M., and Worsnop, D. R.:

Characterization of urban and rural organic particulate in the Lower Fraser Valley using

two Aerodyne Aerosol Mass Spectrometers, Atmos. Environ., 38, 5745-5758,

doi:10.1016/j.atmosenv.2004.01.054, 2004.

Bibliography

130

Allan, J. D., Jimenez, J. L., Williams, P. I., Alfarra, M. R., Bower, K. N., Jayne, J. T.,

Coe, H., and Worsnop, D. R.: Quantitative sampling using an aerodyne aerosol mass

spectrometer - 1. Techniques of data interpretation and error analysis, J. Geophys. Res.-

Atmos., 108, 4090, doi:10.1029/2002JD002358, 2003.

Allan, J. D., Delia, A. E., Coe, H., Bower, K. N., Alfarra, M. R., Jimenez, J. L.,

Middlebrook, A. M., Drewnick, F., Onasch, T. B., Canagaratna, M. R., Jayne, J. T., and

Worsnop, D. R.: A generalised method for the extraction of chemically resolved mass

spectra from aerodyne aerosol mass spectrometer data, J. Aerosol Sci., 35, 909-922,

doi:10.1016/j.jaerosci.2004.02.007, 2004.

Allan, J. D.,Williams, P. I., Morgan,W. T., Martin, C. L., Flynn, M.J., Lee, J.,

Nemitz, E., Phillips, G. J., Gallagher, M.W., and Coe, H.: Contributions from transport,

solid fuel burning and cooking to primary organic aerosols in two UK cities, Atmos.

Chem. Phys., 10, 647-668, doi:10.5194/acp-10-647-2010, 2010.

Andreae, M. O.: Aerosols before pollution, Science, 315, 50,

doi:10.1126/science.1136529, 2007.

Aubriet, F., Carré, V.: Potential of laser mass spectrometry for the analysis of

environmental dust particles - a review, Anal. Chim. Act., 659, 34-54,

doi:10.1016/j.aca.2009.11.047, 2010.

Baltensperger, U., Kalberer, M., Dommen, J., Paulsen, D., Alfarra, M. R., Coe, H.,

Fisseha, R., Gascho, A., Gysel, M., Nyeki, S., Sax., M., Steinbacher, M., Prevot, A. S.

H., Sjögren, S., Weingartner, E., Zenobi, R.: Secondary organic aerosols from

anthropogenic and biogenic precursors, Faraday, Discuss., 130, 265-278,

doi:10.1039/b417367h ,2005.

Bell, M. L., Davis, D. L.: Reassessment of the lethal London fog of 1952: novel

indicators of acute and chronic consequences of acute exposure to air pollution,

environmental health perspectives, 109, no. Suppl. 3, 389, doi:10.1289/ehp.01109s3389,

2001.

131

Bell, M. L., Davis, D. L., and Fletcher, T.: A retrospective of mortality from the

London smog episode of 1952: the role of influenza and pollution, Environmental

health perspectives, 112, doi:10.1289/ehp.6539, 2004.

BFE (Bundesamt für Energie Schweiz): Schweizerische Holzenergiestatistik:

Erhebung für das Jahr 2013, Bundesamt für Energie Schweiz, Ittigen, 2013.

Birch, M. E. and Cary, R. A.: Elemental carbon-based method for monitoring

occupational exposures to particulate diesel exhaust, Aerosol Sci. and Tech., 25, 221-

241, doi:10.1080/02786829608965393, 1996.

Borisov, R. S., Polokov, N. Y., Zhilyaev, D. I., and Zaikin, V. G.: Matrix effect in

matrix assisted laser desorption/ionization mass spectra of derivatized oligometric

polyols, Rapid. Commun. Mass. Spectrom., 27, 333-338, doi:10.1002/rcm.6449, 2013.

Bouwman, A. F., Lee, D. S., Asman, W. A. H., Dentener, F. J., Van der Hoek, K.

W., and Olivier, J. G. J.: A global high resolution emission inventory for ammonia,

Glob. Biogeochem. Cycles, 11, 561-587, doi:10.1029/97gb02266, 1997.

Bozzetti, C., Daellenbach, K., R., Hueglin, C., Fermo, P., Sciare, J., Kasper-Giebl,

A., Mazar, Y., Abbaszade, G., El Kazzi, M., Gonzalez, R., Shuster Meiseles, T., Flasch,

M., Wolf, R., Křepelová, A., Canonaco, F., Schnelle-Kreis, J., Slowik, J. G.,

Zimmermann, R., Rudich, Y., Baltensperger, U., El Haddad, I., and Prévôt, A. S. H.:

Size-resolved identification, characterization, and quantification of primary biological

organic aerosol at a European rural site, Environ. Sci. Technol., 50, 3425-3434, doi:

10.1021/acs.est.5b05960, 2016.

Bozzetti, C. El Haddad, I., Salameh, D., Daellenbach, K. R., Fermo, P., Gonzalez, R.,

Minguillón, M. C., Iinuma, Y., Poulain, L., Müller, E., Slowik, J. G., Jaffrezo, J.-L.,

Baltensperger, U., Marchand, N., and Prévôt, A. S. H.: Organic aerosol source

apportionment by offline-AMS over a full year in Marseille, Atmos. Chem. Phys., 17,

8247-8268, doi:10.5194/acp-17-8247-2017, 2017a.

Bibliography

132

Bozzetti, C., Sosedova, Y., Xiao, M., Daellenbach, K. R., Ulevicius, V., Dudoitis, V.,

Mordas, G., Byčenkienė, S., Plauškaitė, K., Vlachou, A., Golly, B., Chazeau, B.,

Besombes, J.-L., Baltensperger, U., Jaffrezo, J.-L., Slowik, J. G., El Haddad, I., and

Prévôt, A. S. H.: Argon offline-AMS source apportionment of organic aerosol over

yearly cycles for an urban, rural and marine site in Northern Europe, Atmos. Chem.

Phys., 17, 117-141, doi:10.5194/acp-17-117-2017, 2017b.

Brohan, P., Kennedy, J. J., Harris, I., Tett, S. F. B., and Jones, P. D.: Uncertainty

estimates in regional and global observed temperature changes: A new data set from

1850, J. Geophys. Res., 111, D12106, doi:10.1029/2005JD006548, 2006.

Brown, S. G., Eberly, S., Paatero, P., and Norris, G. A.: Methods for estimating

uncertainty in PMF solutions: Examples with ambient air and water quality data and

guidance on reporting PMF results, Sci. Tot. Environ., 518-519, 626-635,

doi:10.1016/j.scitotenv.2015.01.022, 2015.

Brunekreef, B., and Forsberg, B.: Epidemiological evidence of effects of coarse

airborne particles on health, Eur. Respir. J., 26, 309-318, doi:

10.1183/09031936.05.00001805, 2005

Bruns, E. A., Krapf, M., Orasche, J., Huang, Y., Zimmermann, R., Drinovec, L.,

Močnik, G., El-Haddad, I., Slowik, J. G., Dommen, J., Baltensperger, U. and Prévôt, A.

S. H.: Characterization of primary and secondary wood combustion products generated

under different burner loads, Atmos. Chem. Phys., 15, 2825-2841, doi:10.5194/acp-15-

2825-2015, 2015.

Bruns, E. A., El Haddad, I., Slowik, J. G., Kilic, D., Klein, F., Baltensperger, U., and

Prévôt, A. S. H.: Identification of significant precursor gases of secondary organic

aerosols from residential wood combustion, Sci. Rep., 6, 27781, doi:10.1038/srep27881,

2016.

133

Canagaratna, M. R., Jayne, J. T., Jimenez, J. L., Allan, J. D., Alfarra, M. R., Zhang,

Q., Onasch, T. B., Drewnick, F., Coe, H., Middlebrook, A., Delia, A., Williams, L. R.,

Trimborn, A. M., Northway, M. J., DeCarlo, P. F., Kolb, C. E., Davidovits, P. and

Worsnop, D. R.: Chemical and microphysical characterization of ambient aerosols with

the Aerodyne aerosol mass spectrometer, Mass Spectrom. Rev. 26:185-222,

doi:10.1002/mas.20115, 2007.

Canonaco, F., Crippa, M., Slowik, J. G., Baltensperger, U., and Prévôt, A. S. H.:

SoFi, an IGOR-based interface for the efficient use of the generalized multilinear engine

(ME-2) for the source apportionment: ME-2 application to aerosol mass spectrometer

data, Atmos. Meas. Tech., 6, 3649-3661, doi:10.5194/amt-6-3649-2013, 2013.

Canonaco, F., Slowik, J. G., Baltensperger, U., and Prévôt, A. S. H.: Seasonal

differences in oxygenated organic aerosol composition: implications for emissions

sources and factor analysis, Atmos. Chem. Phys. 15, 6993-7002, doi:10.5194/acp-15-

6993-2015, 2015.

Canonaco, F., Daellenbach, K. R., Crippa, M., ElHaddad, I., Bozzetti, C., Huang, R.-

J., Baltensperger, U., Hüglin, C., Herich, H., and Prévôt, A. S. H.: A novel strategy for

the source apportionment of long-term ACSM data based on ME-2 with SoFi:

Automatic Rolling PMF window (AuRo SoFi), in preparation, 2018.

Carslaw, K. S., Lee, L. A., Reddington, C. L., Pringle, K. J., Rap, A., Forster, P. M.,

Mann, G. W., Spracklen, D. V., Woodhouse, M. T., Regayre, L. A., Pierce, J. R.: Large

contribution of natural aerosols to uncertainty in indirect forcing, Nature, 503,67-71,

doi: 10.1038/nature12674, 2013.

Cavalli, F., Viana, M., Yttri, K. E., Genberg, J., and Putaud, J. P.: Toward a

standardised thermal-optical protocol for measuring atmospheric organic and elemental

carbon: the EUSAAR protocol, Atmos. Meas. Tech., 3, 79-89, doi:10.5194/amt-3-79-

2010, 2010.

Bibliography

134

Chen, Q, Miyazaki, Y., Kawamura, K., Matsumoto, K., Coburn, S. C., Volkamer, R.,

Iwamoto, Y., Kagami, S., Deng, Y., Ogawa, S., Ramasamy, S., Kato, S., Ida, A., Kajii,

Y., and Mochida, M.: Characterization of chromophoric water-soluble organic matter in

urban, forest, and marine aerosols by HR-ToF-AMS analysis and excitation emission

matrix spectroscopy, Environ. Sci. Technol., 50, 10351-10360,

doi:10.1021/acs.est.6b01643, 2016.

Chirico, R., Prevot, A. S. H., DeCarlo, P. F., Heringa, M. F., Richter, R.,

Weingartner, E., and Baltensperger, U.: Aerosol and trace gas vehicle emission factors

measured in a tunnel using an Aerosol Mass Spectrometer and other online

instrumentation, Atmos. Environ., 45, 2182-2192, doi:10.1016/j.atmosenv.2011.01.069,

2011.

Ciarelli, G., Aksoyoglu, S., Crippa, M., Jimenez, J.-L., Nemitz, E., Sellegri, K.,

Äijälä, M., Carbone, S., Mohr, C., O'Dowd, C., Poulain, L., Baltensperger, U., and

Prévôt, A. S. H.: Evaluation of European air quality modelled by CAMx including the

volatility basis set scheme, Atmos. Chem. Phys., 16, 10313-10332, doi:10.5194/acp-16-

10313-2016, 2016.

Corbin, J. C., Othman, A., Allan, J. D., Worsnop, D. R., Haskins, J. D., Sierau, B.,

Lohmann, U., and Mensah, A. A.: Peak-fitting and integration imprecision in the

Aerodyne aerosol mass spectrometer: effects of mass accuracy on location-constrained

fits, Atmos. Meas. Tech., 8, 4615-4636, doi:10.5194/amt-8-4615-2015, 2015.

Crippa, M., Canonaco, F., Slowik, J. G., El Haddad, I., De-Carlo, P. F., Mohr, C.,

Heringa, M. F., Chirico, R., Marchand, N., Temime-Roussel, B., Abidi, E., Poulain, L.,

Wiedensohler, A., Baltensperger, U., and Prévôt, A. S. H.: Primary and secondary

organic aerosol origin by combined gas-particle phase source apportionment, Atmos.

Chem. Phys., 13, 8411-8426, doi:10.5194/acp-13-8411-2013, 2013a.

Crippa, M., DeCarlo, P. F., Slowik, J. G., Mohr, C., Heringa, M. F., Chirico, R.,

Poulain, L., Freutel, F., Sciare, J., Cozic, J., Di Marco, C. F., Elsasser, M., Nicolas, J.

B., Marchand, N., Abidi, E., Wiedensohler, A., Drewnick, F., Schneider, J., Borrmann,

135

S., Nemitz, E., Zimmermann, R., Jaffrezo, J.-L., Prévôt, A. S. H., and Baltensperger, U.:

Wintertime aerosol chemical composition and source apportionment of the organic

fraction in the metropolitan area of Paris, Atmos. Chem. Phys., 13, 961-981,

doi:10.5194/acp-13-961-2013, 2013b.

Crippa, M., El Haddad, I., Slowik, J. G., DeCarlo, P. F., Mohr, C., Heringa, M. F.,

Chirico, R., Marchand, N., Sciare, J., Baltensperger, U., and Prevot, A. S. H.:

Identification of marine and continental aerosol sources in Paris using high resolution

aerosol mass spectrometry, J. Geophys. Res.-Atmos., 118, 1950-1963,

doi:10.1002/jgrd.50151, 2013c.

Crippa, M., Canonaco, F., Lanz, V. A., Äijälä, M., Allan, J. D., Carbone, S., Capes,

G., Ceburnis, D., Dall’Osto, M., Day, D. A., De-Carlo, P. F., Ehn, M., Eriksson, A.,

Freney, E., Hildebrandt Ruiz, L., Hillamo, R., Jimenez, J. L., Junninen, H., Kiendler-

Scharr, A., Kortelainen, A.-M., Kulmala, M., Laaksonen, A., Mensah, A. A., Mohr, C.,

Nemitz, E., O’Dowd, C., Ovadnevaite, J., Pandis, S. N., Petäjä, T., Poulain, L.,

Saarikoski, S., Sellegri, K., Swietlicki, E., Tiitta, P., Worsnop, D. R., Baltensperger, U.,

and Prévôt, A. S. H.: Organic aerosol components derived from 25 AMS data sets

across Europe using a consistent ME-2 based source apportionment approach, Atmos.

Chem. Phys., 14, 6159-6176, doi:10.5194/acp-14-6159-2014, 2014.

Cubison, M. J. and Jimenez, J. L.: Statistical precision of the intensities retrieved

from constrained fitting of overlapping peaks in high-resolution mass spectra, Atmos.

Meas. Tech., 8, 2333-2345, doi:10.5194/amt-8-2333-2015, 2015.

Daellenbach, K. R., Bozzetti, C., Krepelova, A., Canonaco, F., Huang, R.-J., Wolf,

R., Zotter, P., Crippa, M., Slowik, J., Zhang, Y., Szidat, S., Baltensperger, U., Prévôt,

A. S. H., and El Haddad, I.: Characterization and source apportionment of organic

aerosol using offline aerosol mass spectrometry, Atmos. Meas. Tech., 9, 23-39,

doi:10.5194/amt-9-23-2016, 2016.

Daellenbach, K. R., Stefenelli, G., Bozzetti, C., Vlachou, A., Fermo, P., Gonzalez,

R., Piazzalunga, A., Colombi, C., Canonaco, F., Hueglin, C., Kasper-Giebl, A., Jaffrezo,

Bibliography

136

J.-L., Bianchi, F., Slowik, J. G., Baltensperger, U., El-Haddad, I., and Prévôt, A. S. H.:

Long-term chemical analysis and organic aerosol source apportionment at nine sites in

central Europe: source identification and uncertainty assessment, Atmos. Chem. Phys.,

17, 13265-13282, doi:10.5194/acp-17-13265-2017, 2017.

Davison, A. C. and Hinkley, D. V.: Bootstrap methods and their Application,

Cambridge University Press, Cambridge, UK, 582 pp., doi:

10.1017/CBO9780511802843, 1997.

De Gouw, J. and Jimenez, J. L.: organic aerosols in the earth’s atmosphere, Environ.

Sci. Technol., 43, 7614-7618, doi:10.1021/es9006004, 2009.

De Hoffmann and, E., and Stroobant, V.: Mass spectrometry: principles and

applications, second edition, John Wiley and Sons, Inc., New York, 1999.

DeCarlo, P. F., Kimmel, J. R., Trimborn, A., Northway, M. J., Jayne, J. T., Aiken, A.

C., Gonin, M., Fuhrer, K., Horvath, T., Docherty, K. S., Worsnop, D. R., and Jimenez,

J. L.: Field-deployable, high-resolution, time-of-flight aerosol mass spectrometer, Anal.

Chem., 78, 8281-8289, doi:10.1021/ac061249n, 2006.

DeCarlo, P. F., Dunlea, E. J., Kimmel, J. R., Aiken, A. C., Sueper, D., Crounse, J.,

Wennberg, P. O., Emmons, L., Shinozuka, Y., Clarke, A., Zhou, J., Tomlinson, J.

Collins, D. R., Knapp, D., Weinheimer, A. J., Montzka, D. D., Campos, T., and

Jimenez, J. L.: Fast airborne aerosol size and chemistry measurements above Mexico

city and Central Mexico during the MILAGRO campaign, Atmos. Chem. Phys., 8,

4027-4048, doi:10.5194/acp-8-4027-2008, 2008.

Dockery, D. W., Pope, C. A., Xu, X., Spengler, J. D., Ware, J. H., Fay, M. E., Ferris,

B. G., and Speizer, F. E.: An association between air pollution and mortality in six U.S.

cities, New England Journal of Medicine, 329, 1753-1759,

doi:10.1056/NEJM199312093292401, 1993.

137

El Haddad, I., D’Anna, B., Temime-Roussel, B., Nicolas, M., Boreave, A., Favez,

O., Voisin, D., Sciare, J., George, C., Jaffrezo, J.-L., Wortham, H., and Marchand, N.:

Towards a better understanding of the origins, chemical composition and aging of

oxygenated organic aerosols: case study of a Mediterranean industrialized environment,

Marseille, Atmos. Chem. Phys., 13, 7875-7894, doi:10.5194/acp-13-7875-2013, 2013a.

El Haddad, I., Marchand, N., D’Anna, B., Jaffrezo, J.-L., and Wortham, H.:

Functional group composition of organic aerosol from combustion emissions and

secondary processes at two contrasted urban environments, Atmos. Environ., 75, 308-

320, doi: 10.1016/j.atmosenv.2013.04.019, 2013b.

Ellis, S. R., Bruinen, A. L., and Heeren, R. M.: A critical evaluation of the current

state-of-the-art in quantitative imaging mass spectrometry,. Anal. Bioanal. Chem., 406,

1275-1289, doi: 10.1007/s00216-013-7478-9, 2014.

Elser, M., Bozzetti, C., El-Haddad, I., Maasikmets, M., Teinemaa, E., Richter, R.,

Wolf, R., Slowik, J. G., Baltensperger, U., and Prévôt, A. S. H.: Urban increments of

gaseous and aerosol pollutants and their sources using mobile aerosol mass

spectrometry measurements, Atmos. Chem. Phys., 16, 7117-7134, doi:10.5194/acp-16-

7117-2016 , 2016a.

Elser, M., Huang, R.-J., Wolf, R., Slowik, J. G., Wang, Q., Canonaco, F., Li, G.,

Bozzetti, C., Daellenbach, K. R., Huang, Y., Zhang, R., Li, Z., Cao, J., Baltensperger,

U., El-Haddad, I., and Prévôt, A. S. H.: New insights into PM2.5 chemical composition

and sources in two major cities in China during extreme haze events using aerosol mass

spectrometry, Atmos. Chem. Phys., 16, 3207-3225, doi:10.5194/acp-16-3207-2016,

2016b.

European Commission, Air quality standards:

http://ec.europa.eu/environment/air/quality/standards.htm, access: 2016/05/12.

Favez, O., El Haddad, I., Piot, C., Boréave, A., Abidi, E., Marchand, N., Jaffrezo, J.-

L., Besombes, J.-L., Personnaz, M.-B., Sciare, J., Wortham, H., George, C., and

Bibliography

138

D’Anna, B.: Inter-comparison of source apportionment models for the estimation of

wood burning aerosols during wintertime in an Alpine city (Grenoble, France), Atmos.

Chem. Phys., 10, 5295-5314, doi:10.5194/acp-10-5295-2010, 2010.

Federal Office for the Environment (FOEN): Immissionsgrenzwerte der

Luftreinhalte-Verordnung (LRV), 2010.

Finlayson-Pitts, B. J., and Pitts, J. N. J.: Chemistry of the Upper and Lower

Atmosphere, 39, 327-328,2000.

Freutel, F., Schneider, J., Drewnick, F., von derWeiden-Reinmüller, S.-L., Crippa,

M., Prévôt, A. S. H., Baltensperger, U., Poulain, L., Wiedensohler, A., Sciare, J., Sarda-

Estève, R., Burkhart, J. F., Eckhardt, S., Stohl, A., Gros, V., Colomb, A., Michoud, V.,

Doussin, J. F., Borbon, A., Haeffelin, M., Morille, Y., Beekmann, M., and Borrmann,

S.: Aerosol particle measurements at three stationary sites in the megacity of Paris

during summer 2009: meteorology and air mass origin dominate aerosol particle

composition and size distribution, Atmos. Chem. Phys., 13, 933-959, doi:10.5194/acp-

13-933-2013, 2013.

Fröhlich, R., Cubison, M. J., Slowik, J. G., Bukowiecki, N., Prévôt, A. S. H.,

Baltensperger, U., Schneider, J., Kimmel, J. R., Gonin, M., Rohner, U., Worsnop, D. R.,

and Jayne, J. T.: The ToF-ACSM: a portable aerosol chemical speciation monitor with

TOFMS detection, Atmos. Meas. Tech., 6, 3225-3241, doi:10.5194/amt-6-3225-2013,

2013.

Fröhlich, R., Crenn, V., Setyan, A., Belis, C. A., Canonaco, F., Favez, O., Riffault,

V., Slowik, J. G., Aas, W., Aijälä, M., Alastuey, A., Artiñano, B., Bonnaire, N.,

Bozzetti, C., Bressi, M., Carbone, C., Coz, E., Croteau, P. L., Cubison, M. J., Esser-

Gietl, J. K., Green, D. C., Gros, V., Heikkinen, L., Herrmann, H., Jayne, J. T., Lunder,

C. R., Minguillón, M. C., Močnik, G., O’Dowd, C. D., Ovadnevaite, J., Petralia, E.,

Poulain, L., Priestman, M., Ripoll, A., Sarda-Estève, R., Wiedensohler, A.,

Baltensperger, U., Sciare, J., and Prévôt, A. S. H.: ACTRIS ACSM intercomparison -

Part 2: Intercomparison of ME-2 organic source apportionment results from 15

139

individual, co-located aerosol mass spectrometers, Atmos. Meas. Tech., 8, 2555-2576,

doi:10.5194/amt-8-2555-2015, 2015.

Furey, A., Moriarty, M., Bane, V., Kinsella, B., and Lehane, M.: Ion suppression: A

critical review on causes, evaluation, prevention and applications, Talanta, 115, 104-

122, doi: 10.1016/j.talanta.2013.03.048, 2013.

Gentner, D. R., Jathar, S. H., Gordon, T. D., Bahreini, R., Day, D. A., El Haddad, I.,

Hayes, P. L., Pieber, S. M., Platt, S. M., de Gouw, J., Goldstein, A. H., Harley, R. A.,

Jimenez, J. L., Prévôt, A. S. H., and Robinson, A. L.: Review of urban secondary

organic aerosol formation from gasoline and diesel motor vehicle emissions, Environ.

Sci. Tech., 51, 1074-1093, doi: 10.1021/acs.est.6b04509, 2017.

Gilgen, H. and Ohmura, A.:: The Global Energy Balance Archive (GEBA). Bull.

Am. Met. Soc., 80,831-850, 10.1175/1520-0477(1999)080<0831:tgeba>2.0.co;2, 1999.

Goheen, S. C., Wahl, K. L., Campbell, J. A., and Hess, W. P.: Mass spectrometry of

low molecular mass solids by matrix-assisted laser desorption/ionization, J. mass.

Spectr., 32, 820,828, doi:10.1002/(sici)1096-9888(199708)32:8<820::aid-

jms538>3.0.co;2-q, 1997

Goldstein, A. H., and Galbally, I. E.: Known and unexplored organic constituents in

the earth’s atmosphere, Environ. Sci. Technol., 41, 1514-1521, doi: 10.1021/es072476p,

2007.

Gordon, H., Sengupta, K., Rap, A., Duplissy, J., Frege, C., Williamson, C., Heinritzi,

M., Simon, M., Yan, C., Almeida, J., Tröstl, J., Nieminen, N., Ortega, I. K., Wagner, R.,

Dunne, E. M., Adamov, A., Amorim, A., Bernhammer, A.-K., Bianchi, F.,

Breitenlechner, M., Sophia Brilke, S., Chen, X., Craven, J. S., Dias, A., Ehrhart, S.,

Fischer, L., Flagan, R. C., Franchin, A., Fuchs, C., Guida, R., Hakala, J., Hoyle, C. R.,

Jokinen, T., Junninen, H., Kangasluoma, J., Kim, J., Kirkby, J., Krapf, M., Kürten, A.,

Laaksonen, A., Lehtipalo, K., Makhmutov, V., Mathot, S., Molteni, U., Monks, S. A.,

Onnela, A., Peräkylä, O., Piel, F., Petäjä, T., Praplan, A. P., Pringle, K. J., Richards, N.

Bibliography

140

A. D., Rissanen, M. P., Rondo, L., Sarnela, N., Schobesberger, S., Scott, C. E., Seinfeld,

J. H., Sharma, S., Sipilä, M., Steiner, G., Stozhkov, Y., Stratmann, F., Tomé, A.,

Virtanen, A., Vogel, A. L., Wagner, A. C., Wagner, P. E., Weingartner, E., Wimmer,

D., Winkler, P. M., Ye, P., Zhang, X., Hansel, A., Dommen, J., Donahue, N. M.,

Worsnop, D. R., Baltensperger, U., Kulmala, M., Curtius, J., and Carslaw, K. S.:

Reduced anthropogenic aerosol radiative forcing caused by biogenic new particle

formation, Proc. Natl. Acad. SCi U.S., 113, 12053-12058,

doi:10.1073/pnas.1602360113, 2016.

Griffiths, A. D., Conen, F., Weingartner, E., Zimmermann, L., Chambers, S. D.,

Williams, A. G., and Steinbacher, M.: Surfaceto-mountaintop transport characterised by

radon observations at the Jungfraujoch, Atmos. Chem. Phys., 14, 12763-12779,

doi:10.5194/acp-14-12763-2014, 2014.

Griggs, D. J. and Noguer, M.: Climate Change 2001: the Scientific Basis,

Contribution of working group I to the third assessment report of the Intergovernmental

Panel on Climate Change,Weather, 57, 267-269, 2002.

Guenther, A., Karl, T., Harley, P., Wiedinmyer, C., Palmer, P. I., and Geron, C.:

Estimates of global terrestrial isoprene emissions using MEGAN (Model of Emissions

of Gases and Aerosols from Nature), Atmos. Chem. Phys., 6, 3181-3210,

doi:10.5194/acp-6-3181-2006, 2006.

Haefliger, O. P., Bucheli, T. D., Zenobi, R.: Laser mass spectrometry analysis of

organic atmospheric aerosols. 1. Characterization of emission sources, Environ. Sci.

Technol., 34, 2178-2183, doi:10.1021/es991122y, 2000a.

Haefliger, O. P., Bucheli, T. D., Zenobi, R.: Laser mass spectrometry analysis of

organic atmospheric aerosols. 2. Elucidation of urban air pollution processes, Environ

Sci. Technol., 34, 2184-2189, doi:10.1021/es991123q: 2000b.

141

Hamad, S. H., Schauer, J. J., Jongbae, H., and Kadhim, A. K. H.: Source

apportionment of PM2.5 carbonaceous aerosol in Baghdad, Iraq, Atmos. Res., 156, 80-

90, doi: 10.1016/j.atmosres.2014.12.017, 2015.

Hansen, A. D. A., Rosen, H., and Novakov, T.: The Aethalometer - an instrument for

the real-time measurement of opticalabsorption by aerosol-particles, Sci. Total Environ.,

36, 191-196, doi:10.1016/0048-9697(84)90265-1, 1984.

Healy, R. M., Sciare, J., Poulain, L., Crippa, M. Wiedensohler, A., Prevot, A. S. H.,

Baltensperger, U., Sarda-Estève, R., McGuire, M. L., Jeong, C.-H., McGillicuddy, E.,

O’Connor, I. P., Sodeau, J. R., Evans, G. J., Wenger, J. C.: Quantitative determination

of carbonaceous particle mixing state in Paris using single-particle mass spectrometer

ande aerosol mass spectrometer measurements, Atmos. Chem. Phys., 13, 9479-9496,

doi:10.5194/acp-13-9479-2013, 2013.

Herich, H., Hueglin, C., and Buchmann, B.: A 2.5 year’s source apportionment study

of black carbon from wood burning and fossil fuel combustion at urban and rural sites

in Switzerland, Atmos. Meas. Tech., 4, 1409-1420, doi:10.5194/amt-4-1409-2011,

2011.

Herich, H., Gianini, M. F. D., Piot, C., Močnik, G., Jaffrezo, J.-L., Besombes, J.-L.,

Prévôt, A. S. H., Hueglin, C.: Overview of the impact of wood burning emissions on

carbonaceous aerosols and PM in large parts of the Alpine region, Atmos. Env., 89, 64-

75, doi:10.1016/j.atmosenv.2014.02.008, 2014

Heringa, M. F., DeCarlo, P. F., Chirico, R., Tritscher, T., Dommen, J.,Weingartner,

E., Richter, R.,Wehrle, G., Prévôt, A. S. H., and Baltensperger, U.: Investigations of

primary and secondary particulate matter of different wood combustion appliances with

a high-resolution time-of-flight aerosol mass spectrometer, Atmos. Chem. Phys., 11,

5945-5957, doi:10.5194/acp-11-5945-2011, 2011.

Hoffmann, T., Huang, R.-J., and Kalberer, M.: Atmospheric analytical chemistry,

Anal. Chem., 83, 4649-4664, doi:10.1021/ac2010718, 2011.

Bibliography

142

Holzke, C., Hoffmann, T., Jaeger, L., Koppmann, R., and Zimmer, W.: Diurnal and

seasonal variation of monoterpene and sesquiterpene emissions from Scots pine (Pinus

sylvestris L.), Atmos. Environ., 40, 3174-3185,

doi:org/10.1016/j.atmosenv.2006.01.039, 2006.

Huang, R.-J., Zhang, Y., Bozzetti, C., Ho, K.-F., Cao, J.-J., Han, Y., Daellenbach, K.

R., Slowik, J. G., Platt, S. M., Canonaco, F., Zotter, P., Wolf, R., Pieber, S. M., Bruns,

E. A., Crippa, M., Ciarelli, G., Piazzalunga, A., Schwikowski, M., Abbaszade, G.,

Schnelle-Kreis, J., Zimmermann, R., An, Z., Szidat, S., Baltensperger, U., El Haddad,

I., and Prévôt, A. S. H.: High secondary aerosol contribution to particulate pollution

during haze events in China, Nature, 514, 218-222, doi:10.1038/nature13774, 2014.

Iinuma, Y., Engling, G., Puxbaum, H., and Herrmann, H.: A highly resolved anion-

exchange chromatographic method for determination of saccharidic tracers for biomass

combustion and primary bio-particles in atmospheric aerosol, Atmos. Environ., 43,

1367-1371, 2009.

IPCC, 2013: Climate Change 2013:The Physical Science Basis. Contribution of

Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on

Climate Change; Stocker, T. F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S. K.,

Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P. M., Eds., Cambridge University

Press: Cambridge, United Kingdom and New York, NY, USA, 2013.

Jaffrezo, J. L., Calas, T., and Bouchet, M.: Carboxylic acids measurements with ionic

chromatography, Atmos. Environ., 32, 2705-2708, doi:10.1016/S1352-2310(98)00026-

0, 1998.

Jimenez, J. L., Canagaratna, M. R., Donahue, N. M., Prévôt, A. S. H., Zhang, Q.,

Kroll, J. H., DeCarlo, P. F., Allan, J. D., Coe, H., Ng, N. L., Aiken, A. C., Docherty, K.

S., Ulbrich, I. M., Grieshop, A. P., Robinson, A. L., Duplissy, J., Smith, J. D., Wilson,

K. R., Lanz, V. A., Hueglin, C., Sun, Y. L., Tian, J., Laaksonen, A., Raatikainen, T.,

Rautiainen, J., Vaattovaara, P., Ehn, M., Kulmala, M., Tomlinson, J. M., Collins, D. R.,

143

Cubison, M. J., Dunlea, E. J., Huffman, J. A., Onasch, T. B., Alfarra, M. R., Williams,

P. I., Bower, K., Kondo, Y., Schneider, J., Drewnick, F., Borrmann, S., Weimer, S.,

Demerjian, K., Salcedo, D., Cottrell, L., Griffin, R., Takami, A., Miyoshi, T.,

Hatakeyama, S., Shimono, A., Sun, J. Y., Zhang, Y. M., Dzepina, K., Kimmel, J. R.,

Sueper, D., Jayne, J. T., Herndon, S. C., Trimborn, A. M., Williams, L. R., Wood, E. C.,

Middlebrook, A. M., Kolb, C. E., Baltensperger, U., and Worsnop, D. R.: Evolution of

organic aerosols in the atmosphere, Science, 326, 1525-1529,

doi:10.1126/science.1180353, 2009.

Jimenez, J. L., Canagaratna, M. R., Drewnick, F., Allan, J. D., Alfarra, M. R.,

Middlebrook, A. M., Slowik, J. G., Zhang, Q., Coe, H., Jayne, J. T., and Worsnop, D.

R.: Comment on “The effects of molecular weight and thermal decomposition on the

sensitivity of a thermal desorption aerosol mass spectrometer, Aerosol Sci. Technol.,

50:9, i-xv, doi:10.1080/02786826.2016.1205728, 2016.

Kalberer, M., Paulsen, D., Sax, M., Steinbacher, M., Dommen, J., Prevot, A. S. H.,

Fisseha, R., Weingartner, E., Frankevich, Zenobi, R., Baltensperger, U.: Identification

of polymers as major components of atmospheric organic aerosols, Science, 303, 1659-

1662, doi:10.1126/science.1092185, 2004.

Kelly, F. J., and Fussell, J. C.: Size, source and chemical composition as

determinants of toxicity attributable to ambient particulate matter, Atmos. Env., 60,

504-526, doi:10.1016/j.atmosenv.2012.06.039, 2012.

Kirchstetter, T., Harley, R., Kreisberg, N., Stolzenburg, M., and Hering, S.: On-road

measurement of fine particle and nitrogen oxide emissions from light- and heavy-duty

motor vehicles, Atmos. Environ., 33(18), 2955-2968, , 1999.

Klein, F., Platt, S. M., Farren, N. J., Detournay, A., Bruns, E. A., Bozzetti, C.,

Daellenbach, K. R., Kilic, D., Kumar, N. K., Pieber, S. M., Slowik, J. G., Temime-

Roussel, B., Marchand, N., Hamilton, J. F., Baltensperger, U., Prévôt, A. S. H., and El

Haddad, I.: Characterization of gas-phase organics using proton transfer reaction time-

Bibliography

144

of-flight mass spectrometry: cooking emissions, Environ. Sci. Tech., 50, 1243-1250,

doi: 10.1021/acs.est.5b04618, 2016a.

Klein, F., Farren, N. J., Bozzetti, C., Daellenach, K. R., Kilic, D., Kumar, N. K.,

Pieber, S. M., Slowik, J. G., Tuthill, R. N., Hamilton, J. F., Baltensperger, U., Prévôt, A.

S. H., El and Haddad, I.: Indoor terpene emissions from cooking with herbs and pepper

and their secondary organic aerosol production potential, Sci. Rep., 6, 36623,

doi:10.1038/srep36623, 2016b.

Kleindienst, T. E. Jaoui, M., Lewandowski, M., Offenberg, J. H., Lewis, C. W.,

Bhave, P. V., Edney, E. O.: Estimates of the contributions of biogenic and

anthropogenic hydrocarbons to secondary organic aerosol at a southern US location,

Atmos. Env., 37, 8288-8300, doi:10.1016/j.atmosenv.2007.06.045, 2007.

Knochenmuss, R., Stortfelder, A., Breuker, K., Zenobi, R.: Secondary ion-molecule

reactions in matrix-assisted laser desorption/ionization, J. Mass. Spectrom, 35, 1237-

1245, doi: 10.1002/1096-9888(200011)35:11<1237::AID-JMS74>3.0.CO;2-O, 2000.

Knochenmuss, R.: A quantitative model of ultraviolet matrix-assisted laser

desorption/ionization including analyte ion generation, Anal. Chem., 75, 2199-2207,

doi:10.1021/ac034032r, 2003.

Knochenmuss, R.: A quantitative model of ultraviolet matrix-assisted laser

desorption/ionization, J. Mass. Spectrom., 37, 867-877, doi: 10.1002/jms.349, 2002.

Knochenmuss, R.: Ion formation mechanism in UV-MALDI, Analyst, 131(9), 966-

86, doi: 10.1039/b605646f, 2006.

Knochenmuss, R.: The coupled chemical and physical dynamics model of MALDI,

Annu. Rev. Anal. Chem., 9, 365-385, doi:10.1146/annurev-anchem-071015-041750,

2016

145

Kourtchev, I., Fuller, S., Aalto, J., Ruuskanen, T. M., McLeod, M. W., Maenhau, W.,

Jones, R., Kulmala, M., Kalberer, M.: Molecular composition of boreal forest aerosol

from Hyytiälä, Finland, using ultrahigh resolution mass spectrometry, Environ. Sci.

Technol., 2013, 47, 4069-4079, doi:0.1021/es3051636, 2013.

Kroll, J. H., Donahue, N. M., Jimenez, J. L., Kessler, S. H., Canagaratna, M. R.,

Wilson, K. R., Altieri, K. E., Mazzoleni, L. R., Wozniak, A. S., Blum, H., Mysak, E. R.,

Smith, J. D., Kolb, C. E., and Worsnop, D. R.: Carbon oxidation state as a metric for

describing the chemistry of atmospheric organic aerosol, Nat. Chem., 3, 133-139,

doi:10.1038/NCHEM.948, 2011

Kunit, M. and Puxbaum, H.: Enzymatic determination of the cellulose content of

atmospheric aerosols, Atmos. Environ., 30, 1233-1236, doi:10.1016/1352-

2310(95)00429-7, 1996.

Laden, F., Schwartz, J., Spelzer, F. E., and Dockery, D. W.: Reduction in fine

particulate air pollution and mortality, Am. J. Respir. Crit. Care Med., 173, 667-672,

doi:10.1164/rccm.200503-443OC, 2006.

Lamberg, H., Nuutinen, K., Tissari, J., Ruusunen, J., Yli-Pirilä, P., Sippula., O.,

Tapanainen, M., Jalava, P., Makkonen, U., Teinilä, K., Saarnio, K., Hillamo, R.,

Hirvonen, M.-R., Jininiemi, J.: Physicochemical characterization of fine particles from

small-scale wood combustion, Atmos. Environ., 45, 7635-7643,

doi:10.1016/j.atmosenv.2011.02.072, 2011.

Lanz, V. A., Alfarra, M. R., Baltensperger, U., Buchmann, B., Hueglin, C., and

Prévôt, A. S. H.: Source apportionment of submicron organic aerosols at an urban site

by factor analytical modelling of aerosol mass spectra, Atmos. Chem. Phys., 7, 1503-

1522, doi:10.5194/acp-7-1503-2007, 2007.

Lanz, V. A., Alfarra, M. R., Baltensperger, U., Buchmann, B., Hueglin, C., Szidat,

S., Wehrli, M. N., Wacker, L., Weimer, S., Caseiro, A., Puxbaum, H., and Prévôt, A. S.

H.: Source attribution of submicron organic aerosols during wintertime inversions by

Bibliography

146

advanced factor analysis of aerosol mass spectra, Environ. Sci. Technol., 42, 214-220,

doi:10.1021/es0707207, 2008.

Lanz, V. A., Prévôt, A. S. H., Alfarra, M. R., Weimer, S., Mohr, C., DeCarlo, P. F.,

Gianini, M. F. D., Hueglin, C., Schneider, J., Favez, O., D'Anna, B., George, C., and

Baltensperger, U.: Characterization of aerosol chemical composition with aerosol mass

spectrometry in Central Europe: an overview, Atmos. Chem. Phys., 10, 10453-10471,

doi:10.5194/acp-10-10453-2010, 2010.

Leaitch, W. R., Macdonald, A. M., Brickell, P. C., Liggio, J., Sjostedt, S. J.,

Vlasenko, A., Bottenheim, J. W., Huang, L., Li, S.-M., Liu, P. S. K., Toom-Sauntry, D.,

Hayden, K. A., Sharma, S., Shantz, N. C., Wiebe H. A., Zhang, W., Abbatt, J. P. D.,

Slowik, J. G., Chang, R., Y.-W., Russell, L. M., Schwartz, R. E., Takahama, S., Jayne,

J. T., and Ng, N. L.: Temperature response of the submicron organic aerosol from

temperate forests, Atmos. Environ., 45, 6696-6704,

doi:10.1016/j.atmosenv.2011.08.047, 2011.

Lee, A. K. Y., Herckes, P., Leaitch, W. R., Macdonald, A. M., and Abbatt, J. P. D.:

Aqueous OH oxidation of ambient organic aerosol and cloud water organics: formation

of highly oxidized products, Geophys. Res. Lett., 38, L11805,

doi:10.1029/2011GL047439, 2011.

Lelieveld, J., Evans, J. S., Fnais, M., Giannadaki, D., and Pozzer, A.: The

contribution of outdoor air pollution sources to premature mortality on a global scale,

Nature, 525, 367-371, doi:10.1038/nature15371, 2015.

Lim, S. S., Vos, T., Flaxman, A. D., Danaei, G., Shibuya, K., Adair-Rohani, H.,

AlMazroa, M. A., Amann, M., Anderson, H. R., Andrews, K. G., Aryee, M., Atkinson,

C., Bacchus, L. J., Bahalim, A. N., Balakrishnan, K., Balmes, J., Barker-Collo, S.,

Baxter, A., Bell, M. L., Blore, J. D., Blyth, F., Bonner, C., Borges, G., Bourne, R.,

Boussinesq, M., Brauer, M., Brooks, P., Bruce, N. G., Brunekreef, B., Bryan-Hancock,

C., Bucello, C., Buchbinder, R., Bull, F., Burnett, R. T., Byers, T. E., Calabria, B.,

Carapetis, J., Carnahan, E., Chafe, Z., Charlson, F., Chen, H., Chen, J. S., Cheng, A. T.-

147

A., Child, J. C., Cohen, A., Colson, K. E., Cowie, B. C., Darby, S., Darling, S., Davis,

A., Degenhardt, L., Dentener, F., Jarlais, D. C. D., Devries, K., Dherani, M., Ding, E.

L., Dorsey, E. R., Driscoll, T., Edmond, K., Ali, S. E., Engell, R. E., Erwin, P. J.,

Fahimi, S., Falder, G., Farzadfar, F., Ferrari, A., Finucane, M. M., Flaxman, S., Fowkes,

F. G. R., Freedman, G., Freeman, M. K., Gakidou, E., Ghosh, S., Giovannucci, E.,

Gmel, G., Graham, K., Grainger, R., Grant, B., Gunnell, D., Gutierrez, H. R., Hall, W.,

Hoek, H. W., Hogan, A., III, H. D. H., Hoy, D., Hu, H., Hubbell, B. J., Hutchings, S. J.,

Ibeanusi, S. E., Jacklyn, G. L., Jasrasaria, R., Jonas, J. B., Kan, H., Kanis, J. A.,

Kassebaum, N., Kawakami, N., Khang, Y.-H., Khatibzadeh, S., Khoo, J.-P., Kok, C.,

Laden, F., Lalloo, R., Lan, Q., Lathlean, T., Leasher, J. L., Leigh, J., Li, Y., Lin, J. K.,

Lipshultz, S. E., London, S., Lozano, R., Lu, Y., Mak, J., Malekzadeh, R., Mallinger,

L., Marcenes, W., March, L., Marks, R., Martin, R., McGale, P., McGrath, J., Mehta, S.,

Memish, Z. A., Mensah, G. A., Merriman, T. R., Micha, R., Michaud, C., Mishra, V.,

Hanafiah, K. M., Mokdad, A. A., Morawska, L., Mozaffarian, D., Murphy, T., Naghavi,

M., Neal, B., Nelson, P. K., Nolla, J. M., Norman, R., Olives, C., Omer, S. B., Orchard,

J., Osborne, R., Ostro, B., Page, A., Pandey, K. D., Parry, C. D., Passmore, E., Patra, J.,

Pearce, N., Pelizzari, P. M., Petzold, M., Phillips, M. R., Pope, D., Pope, C. A. 3rd,

Powles, J., Rao, M., Razavi, H., Rehfuess, E. A., Rehm, J. T., Ritz, B., Rivara, F. P.,

Roberts, T., Robinson, C., Rodriguez-Portales, J. A., Romieu, I., Room, R., Rosenfeld,

L. C., Roy, A., Rushton, L., Salomon, J. A., Sampson, U., Sanchez-Riera, L., Sanman,

E., Sapkota, A., Seedat, S., Shi, P., Shield, K., Shivakoti, R., Singh, G. M., Sleet, D. A.,

Smith, E., Smith, K. R., Stapelberg, N. J., Steenland, K., Stöckl, H., Stovner, L. J.,

Straif, K., Straney, L., Thurston, G. D., Tran, J. H., Van Dingenen, R., van Donkelaar,

A., Veerman, J. L., Vijayakumar, L., Weintraub, R., Weissman, M. M., White, R. A.,

Whiteford, H., Wiersma, S. T., Wilkinson, J. D., Williams, H. C., Williams, W.,

Wilson, N., Woolf, A. D., Yip, P., Zielinski, J. M., Lopez, A. D., Murray, C. J., Ezzati,

M., AlMazroa, M. A., and Memish, Z. A.: A comparative risk assessment of burden of

disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions,

1990-2010: a systematic analysis for the Global Burden of Disease Study 2010, The

Lancet, 380, 2224-2260, doi: 10.1016/S0140-6736(12)61766-8, 2012

Masson-Delmotte, V., Schulz, M., Abe-Ouchi, A., Beer, J., Ganopolski, A.,

González Rouco, J.F., Jansen, E., Lambeck, K., Luterbacher, J., Naish, T., Osborn, T.,

Bibliography

148

Otto-Bliesner, B., Quinn, T., Ramesh, R., Rojas, M., Shao, X., and Timmermann, A.:

Information from Paleoclimate Archives. in: Climate Change 2013: The Physical

Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the

Intergovernmental Panel on Climate Change edited by Stocker, T.F., Qin, D., Plattner,

G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex V., and Midgley,

P.M., Cambridge University Press, Cambridge, United Kingdom and New York, NY,

USA, 2013.

Maynard, A., and Kuempel, E.: Airborne nanostructured particles and occupational

health, J. Nanopart. Res., 7, 587-614, doi:10.1007/s11051-005-6770-9, 2005.

Mihara, T. and Mochida, M.: Characterization of solvent extractable organics in

urban aerosols based on mass spectrum analysis and hygroscopic growth measurement,

Environ. Sci. Technol., 45, 9168-9174, doi:10.1021/es201271w, 2011.

Minguillón, M. C., Perron, N., Querol, X., Szidat, S., Fahrni, S. M., Alastuey, A.,

Jimenez, J. L., Mohr, C., Ortega, A. M., Day, D. A., Lanz, V. A., Wacker, L., Reche,

C., Cusack, M., Amato, F., Kiss, G., Hoffer, A., Decesari, S., Moretti, F., Hillamo, R.,

Teinilä, K., Seco, R., Peñuelas, J., Metzger, A., Schallhart, S., Müller, M., Hansel, A.,

Burkhart, J. F., Baltensperger, U., and Prévôt, A. S. H.: Fossil versus contemporary

sources of fine elemental and organic carbonaceous particulate matter during the

DAURE campaign in Northeast Spain, Atmos. Chem. Phys., 11, 12067-12084,

doi:10.5194/acp-11-12067-2011, 2011.

Mohr, C., Richter, R., DeCarlo, P. F., Prévôt, A. S. H., and Baltensperger, U.: Spatial

variation of chemical composition and sources of submicron aerosol in Zürich during

wintertime using mobile aerosol mass spectrometer data, Atmos. Chem. Phys., 11,

7465-7482, doi:10.5194/acp-11-7465-2011, 2011.

Mohr, C., DeCarlo, P. F., Heringa, M. F., Chirico, R., Slowik, J. G., Richter, R.,

Reche, C., Alastuey, A., Querol, X., Seco, R., Penuelas, J., Jimenez, J. L., Crippa, M.,

Zimmermann, R., Baltensperger, U., and Prevot, A. S. H.: Identification and

quantification of organic aerosol from cooking and other sources in Barcelona using

149

aerosol mass spectrometer data, Atmos. Chem. Phys., 12, 1649-1665, doi:10.5194/acp-

12-1649-2012, 2012.

Murphy, D. M.: The design of single particle laser mass spectrometers, Mass Spectr.

Rev., 26, 150-165, doi:10.1002/mas.20113, 2007.

Myhre, G., Shindell, D., Bréon, F.-M., Collins, W., Fuglestvedt, J., Huang, J., Koch,

D., Lamarque, J.-F., Lee, D., Mendoza, B., Nakajima, T., Robock, A., Stephens, G.,

Takemura, T., and Zhang, H.: Anthropogenic and Natural Radiative Forcing. in:

Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to

the Fifth Assessment Report of the Intergovernmental Panel on Climate Change edited

by Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels,

A., Xia, Y., Bex, V., and Midgley, P.M., pp. 659-740, Cambridge University Press,

Cambridge, United Kingdom and New York, NY, USA, 2013

Ng, N. L., Canagaratna, M. R., Zhang, Q., Jimenez, J. L., Tian, J., Ulbrich, I. M.,

Kroll, J. H., Docherty, K. S., Chhabra, P. S., Bahreini, R., Murphy, S. M., Seinfeld, J.

H., Hildebrandt, L., Donahue, N. M., DeCarlo, P. F., Lanz, V. A., Prévôt, A. S. H.,

Dinar, E., Rudich, Y., and Worsnop, D. R.: Organic aerosol components observed in

Northern Hemispheric datasets from Aerosol Mass Spectrometry, Atmos. Chem. Phys.,

10, 4625-4641, doi:10.5194/acp-10-4625-2010, 2010.

Ng, N. L., Canagaratna, M R., Jimenez, J. L., Chhabra, P. S., Seinfeld, J. H., and

Worsnop, D. R.: Changes in organic aerosol composition with aging inferred from

aerosol mass spectra, Atmos. Chem. Phys., 11, 6465-6474, doi: 10.5194/acp-11-6465-

2011, 2011a.

Ng, N. L., Canagaratna, M. R., Jimenez, J. L., Zhang, Q., Ulbrich, I. M., and

Worsnop, D. R.: Real-time methods for estimating organic component mass

concentrations from aerosol mass spectrometer data, Environ. Sci. Technol., 45, 910-

916, doi:10.1021/es102951k, 2011b.

Bibliography

150

Ng, N. L., Herndon, S. C., Trimborn, A., Canagaratna, M. R., Croteau, P. L., Onasch,

T. B., Sueper, D., Worsnop, D. R., Zhang, Q., Sun, Y. L., and Jayne, J. T.: An aerosol

chemical speciation monitor (ACSM) for routine monitoring of the composition and

mass concentrations of ambient aerosol, Aerosol Sci. Tech., 45, 780-794,

doi:10.1080/02786826.2011.560211, 2011c.

Oderbolz, D. C., Aksoyoglu, S., Keller, J., Barmpadimos, I., Steinbrecher, R., Skjøth,

C. A., Plaß-Dülmer, C., and Prévôt, A. S. H.: A comprehensive emission inventory of

biogenic volatile organic compounds in Europe: improved seasonality and land-cover,

Atmos. Chem. Phys., 13, 1689-1712, doi:10.5194/acp-13-1689-2013, 2013.

Paatero, P. and Tapper, U.: Positive matrix factorization - a nonnegative factor model

with optimal utilization of error-estimates of data values, Environmetrics, 5, 111-126,

doi:10.1002/env.3170050203, 1994.

Paatero, P.: Least squares formulation of robust non-negative factor analysis,

Chemometr. Intell. Lab., 37, 23-35, doi: 10.1016/S0169-7439(96)00044-5, 1997.

Paatero, P.: The multilinear engine - a table-driven, least squares program for solving

multilinear problems, including the n-way parallel factor analysis model, J. Comput.

Graph. Stat., 8, 854-888, doi:10.2307/1390831, 1999.

Paatero, P.: User’s guide for positive matrix factorization programs PMF2 and

PMF3, 2004.

Paatero, P. and Hopke, P. K.: Rotational tools for factor analytic models, J.

Chemometr., 23, 91-100, doi:10.1002/cem.1197, 2009.

Paatero, P., Eberly, S., Brown, S. G., and Norris, G. A.: Methods for estimating

uncertainty in factor analytic solutions, Atmos. Meas. Tech., 7, 781-797,

doi:10.5194/amt-7-781-2014, doi:10.5194/amt-7-781-2014, 2014.

151

Piazzalunga, A., Fermo, P., Bernardoni, V., Vecchi, R., Valli, G., and De Gregorio,

M. A.: A simplified method for levoglucosan quantification in wintertime atmospheric

particulate matter by high performance anion-exchange chromatography coupled with

pulsed amperometric detection, Intern. J. Environ. Anal. Chem., 90, 12, 934-947,

doi:10.1080/03067310903023619, 2010.

Piazzalunga, A., Bernardoni, V., Fermo, P., and Vecchi, R.: Optimisation of

analytical procedures for the quantification of ionic and carbonaceous fractions in the

atmospheric aerosol and applications to ambient samples, Anal., Bioanal. Chem., 56,

30-40, doi:10.1007/s00216-012-6433-5, 2013

Pieber, S. M., El Haddad, I., Slowik, J. G., Canagaratna, M. R., Jayne, J. T., Platt, S.

M., Bozzetti, C., Daellenbach, K. R., Fröhlich, R., Vlachou, A., Klein, F., Dommen, J.,

Miljevic, B., Jimenez, J. L., Worsnop, D. R., Baltensperger, U., and Prévôt, A. S. H.:

Inorganic salt interference on CO2+ in aerodyne AMS and ACSM organic aerosol

composition studies, Environ. Sci. Technol, 50, 10494-10503, doi:

10.1021/acs.est.6b01035,2016.

Platt, S. M., El Haddda, I., Pieber, S. M., Huang, R.-J., Zardini, A. A., Clairotte, M.,

Suarez-Bertoa, R., Barmet, P., Pfaffenberger, L., Wolf, R., Slowik, J. G., Fuller, S. J.,

Kalberer, M., Chirico, R., Dommen, J., Astorga, C., Zimmermann, R., Marchand, N.,

Hellebust, S., Temime-Roussel, B., Baltensperger, U., Prévôt, A. S. H.: two-stroke

scooters are a dominant source of air pollution in many cities, Nat. Comm., 5, 3749,

doi:10.1038/ncomms4749, 2014.

Pope, C. A., and Dockery, D. W.: Health effects of fine particulate air pollution:

Lines that connect, J. Air Waist Manage. Assoc., 56, 709-742,

doi:10.1080/10473289.2006.10464485, 2006.

Pope, D., III, C. A. P., Powles, J., Rao, M., Razavi, H., Rehfuess, E. A., Rehm, J. T.,

Ritz, B., Rivara, F. P., Roberts, T., Robinson, C., Rodriguez-Portales, J. A., Romieu, I.,

Room, R., Rosenfeld, L. C., Roy, A., Rushton, L., Salomon, J. A., Sampson, U.,

Sanchez-Riera, L., Sanman, E., Sapkota, A., Seedat, S., Shi, P., Shield, K., Shivakoti,

Bibliography

152

R., Singh, G. M., Sleet, D. A., Smith, E., Smith, K. R., Stapelberg, N. J., Steenland, K.,

Stöckl, H., Stovner, L. J., Straif, K., Straney, L., Thurston, G. D., Tran, J. H., Dingenen,

R. V., van Donkelaar, A., Veerman, J. L., Vijayakumar, L., Weintraub, R., Weissman,

M. M., White, R. A., Whiteford, H., Wiersma, S. T., Wilkinson, J. D., Williams, H. C.,

Williams, W., Wilson, N., Woolf, A. D., Yip, P., Zielinski, J. M., Lopez, A. D., Murray,

C. J., and Ezzati, M.: A comparative risk assessment of burden of disease and injury

attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a

systematic analysis for the Global Burden of Disease Study 2010, The Lancet, 380,

2224 - 2260, doi:10.1016/S0140-6736(12)61766-8 2013.

Putaud, J. P., Van Dingenen, R., Baltensperger, U., Brüggemann, E., Charron, A.,

Facchini, M. C., Decessari, S., Fuzzi, S., Gehrig, R., Hansson, H.-C., Harrison, R. M.,

Jones, A. M., Laj, P., Lorbeer, G., Maenhaut, W., Mihalopoulos, N., Müller, K.,

Palmgren, F., Querol, X., Rodriguez, S., Schneider, J., Spindler, G., ten Brink, H.,

Tunved, P., Törseth, K., Wehner, B., Weingartner, E., Wiedensohler, A., Wahlin, P.,

Raes, F.: A European aerosol phenomology, Joint Research Center, EUR 20411, 2004.

Putaud, J. P., Van Dingenen, R., Alastuey, A., Bauer, H., Birmili, W., Cyrys, J.,

Flentje, H., Fuzzi, S., Gehrig, R., Hansson, H. C., Harrison, R. M., Herrmann, H.,

Hitzenberger, R., Hueglin, C., Jones, A. M., Kasper-Giebl, A., Kiss, G., Kousa, A.,

Kuhlbusch, T. A. J., Loeschau, G., Maenhaut, W., Molnar, A., Moreno, T., Pekkanen,

J., Perrino, C., Pitz, M., Puxbaum, H., Querol, X., Rodriguez, S., Salma, I., Schwarz, J.,

Smolik, J., Schneider, J., Spindler, G., ten Brink, H., Tursic, J., Viana, M.,

Wiedensohler, A., and Raes, F.: A European aerosol phenomenology-3: physical and

chemical characteristics of particulate matter from 60 rural, urban, and kerbside sites

across Europe, Atmos. Environ., 44, 1308-1320, doi:10.1016/j.atmosenv.2009.12.011,

2010.

Putaud, J. P., Van Dingenen, R., Alastuey, A., Bauer, H., Birmili, W., Cyrys, J.,

Flentje, H., Fuzzi, S., Gehrig, R., Hansson, H. C., Harrison, R. M., Herrmann, H.,

Hitzenberger, R., Hueglin, C., Jones, A. M., Kasper-Giebl, A., Kiss, G., Kousa, A.,

Kuhlbusch, T. A. J., Loeschau, G., Maenhaut, W., Molnar, A., Moreno, T., Pekkanen,

J., Perrino, C., Pitz, M., Puxbaum, H., Querol, X., Rodriguez, S., Salma, I., Schwarz, J.,

153

Smolik, J., Schneider, J., Spindler, G., ten Brink, H., Tursic, J., Viana, M.,

Wiedensohler, A., and Raes, F.: A European aerosol phenomenology-3: physical and

chemical characteristics of particulate matter from 60 rural, urban, and kerbside sites

across Europe, Atmos. Environ., 44, 1308-1320, doi:10.1016/j.atmosenv.2009.12.011,

2010.

Rocke, D. M., and Lorenzato, S.: A two-component model for measurement error in

analytical chemistry, Technometrics, 37, 2, 176-184, doi:

10.1080/00401706.1995.10484302, 1995.

Samburova, V., R. Zenobi, Kalberer, M.: Characterization of high molecular weight

compounds in urban atmospheric particles Atmos. Chem. Phys., 5.8 2163-2170,

doi:10.5194/acp-5-2163-2005, 2005a.

Samburova, V.,Szidat, S., Hueglin, C., Fisseha, R., Baltensperger, U., Zenobi, R.,

Kalberer, M.: Seasonal variation of high-molecular-weight compounds in the water-

soluble fraction of organic urban aerosols, J. Geo. Res., 110, D23,

doi:10.1029/2005jd005910, 2005b.

Sanchez-Ochoa, A., Kasper-Giebl, A., Puxbaum, H., Gelencser, A., Legrand, M., and

Pio, C.: Concentration of atmospheric cellulose: A proxy for plant debris across a west-

east transect over Europe, J.Geo. Res., 112, D23, doi:10.1029/2006jd008180, 2007.

Sandradewi, J., Prévôt, A. S. H., Szidat, S., Perron, N., Alfarra, M. R, Lanz, V. A.,

Weingartner, E., and Baltensperger, U.: Using aerosol light absorption measurements

for the quantitative determination of wood burning and traffic emission contributions to

particulate matter, Environ. Sci. Technol., 42, 3316-3323, doi:10.1021/es702253m,

2008.

Schauer, J. J., Kleeman, M. J., Cass, G. R., and Simoneit, B. R. T.: Measurement of

emissions from air pollution sources. 3. C1-C29 organic compounds from fireplace

combustion of wood, Environ. Sci. Technol., 35, 1716-1728, doi: 10.1021/es001331e,

2001.

Bibliography

154


emissions from air pollution sources. 5. C-1-C-32 organic compounds from gasoline-

powered motor vehicles, Environ. Sci. Technol., 36, 1169-1180, doi:

10.1021/es0108077, 2002a.


emissions from air pollution sources. 4. C-1-C-27 organic compounds from cooking

with seed oils, Environ. Sci. Technol., 36, 567-575, doi: 10.1021/es002053m, 2002b.

Schmale, J., Schneider, J., Nemitz, E., Tang, Y. S., Dragosits, U., Blackall, T. D.,

Trathan, P. N., Phillips, G. J., Sutton, M., and Braban, C. F.: Sub-Antarctic marine

aerosol: dominant contributions from biogenic sources, Atmos. Chem. Phys., 13, 8669-

8694, doi:10.5194/acp-13-8669-2013,2013.

Schurgers, G., Arneth, A., Holzinger, R., and Goldstein, A. H.: Process-based

modelling of biogenic monoterpene emissions combining production and release from

storage, Atmos. Chem. Phys., 9, 3409-3423, doi:10.5194/acp-9-3409-2009, 2009.

Seifried, H. E., Anderson, D. E., Fisher, E. I., Milner, J. A.: A review of the

interaction among dietary antioxidants and reactive oxygen species, J. Nutr. Biochem.,

18, 567-579, doi:10.1016/j.jnutbio.2006.10.007, 2006.

Seinfeld, J. H., and Pandis, S. N.: Atmospheric chemistry and physics: from air

pollution to climate change, 2nd ed., John Wiley and Sons, Inc., New York, 2006.

Simoneit, B. R. T., Kobayashi, M., Mochida, M., Kawamura, K., Lee, M., Lim, H.-

J., Turpin, B. J., Komazaki, Y.: Composition and major sources of organic compounds

of aerosol particulate matter sampled during the ACE-Asia campaign, J. Geophys. Res.,

109, D19S10, doi:10.1029/2004jd004598, 2004.

155

Stanek, L. W., Sacks, J. D., Dutton, S. J., Dubois, J.-J. B.: Attributing health effects

to apportioned components and sources of particulate matte: An evaluation of collective

results, Atmos. Environ., 45, 5655-5663, doi:10.1016/j.atmosenv.2011.07.023, 2011.

Stark, H., Yatavelli, R. L. N., Thompson, S. L., Kimmel, J. R., Cubison, M. J.,

Chhabra, P. S., Canagaratna, M. R., Jayne, J. T., Worsnop, D. R., and Jimenez, J. L.:

Methods to extract molecular and bulk chemical information from series of complex

mass spectra with limited mass resolution, Int. J. Mass Spectrom., 389, 26-38,

doi:10.1016/j.ijms.2015.08.011, 2015.

Stern, D. I.: Global sulfur emissions from 1850 to 2000, Chemosphere, 58, 163-175,

doi: 10.1016/j.chemosphere.2004.08.022, 2005.

Sun, Y., Zhang, Q., Zheng, M., Ding, X., Edgerton, E. S., and Wang, X.:

Characterization and source apportionment of watersoluble organic matter in

atmospheric fine particles (PM2:5) with high-resolution aerosol mass spectrometry and

GC-MS, Environ. Sci. Technol., 45, 4854-4861, doi:10.1021/es200162h, 2011.

Szidat, S., Jenk, T. M., Synal, H.-A., Kalberer, M., Wacker, L., Hajdas, I., Kasper-

Giebl, A., Baltensperger, U.: Contributions of fossil fuel, biomass-burning, and biogenic

emissions to carbonaceous aerosols in Zürich as traced by 14C, J. Geophys. Res.-

Atmos., 111, D07206, 10.1029/2005JD006590, 2006.

Szidat, S., Salazar, G. A., Battaglia, M., Wacker, L., Synal, H.-A., Vogel, E., and

Türler, A.: 14C analyses and sample preparation at the new Bern Laboratory for the

Analyses of Radiocarbon with AMS (LARA), Radiocarbon, 56, No. 2, 561-566,

doi:10.2458/56.17457, 2014.

Trufelli, H., Palma, P., Famiglini, G., and Cappiello, A.: An overview of matrix

effects in liquid chromatography-mass spectrometry, Mass. Spectrom, Rev., 30, 491-

509, doi:10.1002/mas.20298, 2011.

Bibliography

156

Ulbrich, I. M., Canagaratna, M. R., Zhang, Q., Worsnop, D. R., and Jimenez, J. L.:

Interpretation of organic components from Positive Matrix Factorization of aerosol

mass spectrometric data, Atmos. Chem. Phys., 9, 2891-2918, doi:10.5194/acp-9-2891-

2009, 2009.

US Environmental Protection Agency, National Ambient Air Quality Standards

(NAAQS): https://www.epa.gov/criteria-air-pollutants/naaqs-table, access: 2016/12/05.

Verma, V., Fang, T., Xu, L., Peltier, R. E., Russell, A. G., Ng, N. L., and Weber, R.

J.: Organic aerosols associated with the generation of reactive oxygen species (ROS) by

water-soluble PM2.5, Environ. Sci. Technol., 49, 4646-4656, doi:10.1021/es505577w,

2015.

Vlachou, A., Daellenbach, K. R., Bozzetti, C., Szidat, S., Salazar, G., Agrios, K.,

Jaffrezo, J. L., Baltensperger, U., El Haddad, I., and Prévôt, A. S. H.: Radiocarbon

based source apportionment of carbonaceous aerosols in Magadino, Switzerland, in

preparation, 2017.

Visser, S., Slowik, J. G., Furger, M., Zotter, P., Bukowiecki, N., Canonaco, F.,

Flechsig, U., Appel, K., Green, D. C., Tremper, A. H., Young, D. E., Williams, P. I.,

Allan, J. D., Coe, H., Williams, L. R., Mohr, C., Xu, L., Ng, N. L., Nemitz, E., Barlow,

J. F., Halios, C. H., Fleming, Z. L., Baltensperger, U., and Prévôt, A. S. H.: Advanced

source apportionment of size-resolved trace elements at multiple sites in London during

winter, Atmos. Chem. Phys., 15, 11291-11309, doi:10.5194/acp-15-11291-2015, 2015.

Vlachou, A., Daellenbach, K. R., Bozzetti, C., Chazeau, B., Salazar, G. A., Szidat,

S., Jaffrezo, J.-L., Hueglin, C., Baltensperger, U., El Haddad, I., and Prévôt, A. S. H.:

Advanced source apportionment of carbonaceous aerosols by coupling offline AMS and

radiocarbon size segregated measurements over a nearly two-year period, Atmos.

Chem. Phys. Discuss., doi:10.5194/acp-2017-1102, in review, 2017.

Waked, A., Favez, O., Alleman, L. Y., Piot, C., Petit, J.-E., Delaunay, T., Verlinden,

E., Golly, B., Besombes, J.-L., Jaffrezo, J.-L., and Leoz-Garziandia, E.: Source

157

apportionment of PM10 in a north-western Europe regional urban background site

(Lens, France) using positive matrix factorization and including primary biogenic

emissions, Atmos. Chem. Phys., 24, 3325-3346, doi:acp-14-3325-2014, 2014.

Watson, J. G., Robinson, N. F. Fujita, E. M., Chow, J. C., Pace, T. G. Lewis, C.,

Coulter, T.: CMB8 applications and validation protocol for PM2.5 and VOCs, US EPA,

USA, 1998.

Wilks, D. S.: Statistical methods in the atmospheric sciences, second edition, edited

by Dmowska, R., Hartmann, D., and Rossby, H. T., Academic Press, 2006.

Williams, L. R., Gonzalez, L. A., Peck, J., Trimborn, D., McInnis, J., Farrar, M. R.,

Moore, K. D., Jayne, J. T., Robinson, W. A., Lewis, D. K., Onasch, T. B., Canagaratna,

M. R., Trimborn, A., Timko, M. T., Magoon, G., Deng, R., Tang, D., de la Rosa

Blanco, E., Prévôt, A. S. H., Smith, K. A., and Worsnop, D. R.: Characterization of an

aerodynamic lens for transmitting particles greater than 1 micrometer in diameter into

the Aerodyne aerosol mass spectrometer, Atmos. Meas. Tech., 6, 3271-3280,

doi:10.5194/amt-6-3271-2013, 2013.

Wilson, M. D., Rocke, D. M., Durbin, B., and Kahn, H. D.: Detection limits and

goodness-of-fit measures for the two-component model of chemical analytical error,

Anal. Chim. Acta., 509, 197-208, doi:10.1016/j.aca.2003.12.047, 2004.

World Health Organization (WHO): WHO air quality guidelines for particulate

matter, ozone , nitrogen dioxide and sulfur dioxide. Global update 2005. Summary of

risk assessment, WHO/SDE/PHE/OEH/06.02, 2006.

World Health Organization (WHO): Review of evidence on health aspects of air

pollution - REVIHAAP Project, 2013.

World Health Organization (WHO): WHO’s urban ambient air pollution database -

Update 2016, version 2.0, 2016.

Bibliography

158

Yttri, K. E., Simpson, D., Nøjgaard, J. K., Kristensen, K., Genberg, J., Stenström, K.,

Swietlicki, E., Hillamo, R., Aurela, M., Bauer, H., Offenberg, J. H., Jaoui, M., Dye, C.,

Eckhardt, S., Burkhart, J. F., Stohl, A., and Glasius, M.: Source apportionment of the

summer time carbonaceous aerosol at Nordic rural background sites, Atmos. Chem.

Phys., 11, 13339-13357, doi:10.5194/acp-11-13339-2011, 2011.

Zar, J.H., Biostatistical Analysis, 4th ed., 929 pp., Prentice Hall, Englewood Cliffs,

New Jersey, 1999.

Zenobi, R and Knochenmuss, R.: Ion formation in MALDI mass spectrometry, Mass

Spectr. Rev., 17, 337-366, doi:10.1002/(sici)1098-2787(1998)17:5<337::aid-

mas2>3.3.co;2-j, 1998.

Zhang, Q., Jimenez, J. L., Canagaratna, M. R., Allan, J. D., Coe, H., Ulbrich, I.,

Alfarra, M. R., Takami, A., Middlebrook, A. M., Sun, Y. L., Dzepina, K., Dunlea, E.,

Docherty, K., DeCarlo, P. F., Salcedo, D., Onasch, T., Jayne, J. T., Miyoshi, T.,

Shimono, A., Hatakeyama, S., Takegawa, N., Kondo, Y., Schneider, J., Drewnick, F.,

Borrmann, S., Weimer, S., Demerjian, K., Williams, P., Bower, K., Bahreini, R.,

Cottrell, L., Griffin, R. J., Rautiainen, J., J. Y., Zhang, Y. M., and Worsnop, D. R.:

Ubiquity and dominance of oxygenated species in organic aerosols in

anthropogenically-influenced Northern Hemisphere midlatitudes, Geophys. Res. Lett.,

34, L13801, 2007.

Zhang, Y., Schauer, J. J., Shafer, M. M., Hannigan, M. P., and Dutton, S. J.: Source

apportionment of in vitro reactive oxygen species bioassay activity from atmospheric

particulate matter, Environ. Sci. Technol., 42, 7502-7509, doi:10.1021/es800126y,

2008.

Zhang, Q., Jimenez, J. L., Canagaratna, M. R., Ulbrich, I. M., Ng, N. L., and

Worsnop, Y. S.: Understanding atmospheric organic aerosols via factor analysis of

aerosol mass spectrometry: a review, Anal. Bioanal. Chem., 401, 3045-3067,

doi:10.1007/s00216-011-5355-y, 2011.

159

Zhang, Y. L., Perron, N., Ciobanu, V. G., Zotter, P., Minguillón, M. C, Wacker, L.,

Prévôt, A. S. H., Baltensperger, U., and Szidat, S: On the isolation of OC and EC and

the optimal strategy of radiocarbon-based source apportionment of carbonaceous

aerosols, Atmos. Chem. Phys., 12, 10841-108556, doi:10.5194/acp-12-10841-2012,

2012.

Zorn, S. R., Drewnick, F., Schott, M., Hoffmann, T., and Borrmann, S.:

Characterization of the South Atlantic marine boundary layer aerosol using an aerodyne

aerosol mass spectrometer, Atmos. Chem. Phys., 8, 4711-4728, doi:10.5194/acp-8-

4711-2008, 2008.

Zotter, P., Ciobanu, V. G., Zhang, Y. L., El-Haddad, I., Macchia, M., Daellenbach,

K. R., Salazar, G. A., Huang, R.-J., Wacker, L., Hueglin, C., Piazzalunga, A., Fermo, P.,

Schwikowski, M., Baltensperger, U., Szidat, S., and Prévôt, A. S. H.: Radiocarbon

analysis of elemental and organic carbon in Switzerland during winter-smog episodes

from 2008 to 2012 - Part 1: Source apportionment and spatial variability, Atmos. Chem.

Phys., 14, 13551-13570, doi:10.5194/acp-14-13551-2014, 2014.

Zotter, P., Herich, H., Gysel, M., El Haddad, I., Zhang, Y., Močnik, Hüglin, C.,

Baltensperger, U., Szidat, S., and Prévôt, A. S. H.: Evaluation of the absorption

Ångström exponents for traffic and wood burning in the Aethalometer based source

apportionment using radiocarbon measurements of ambient aerosols, Atmos. Chem.

Phys., 17, 4229-4249, https://doi.org/10.5194/acp-17-4229-2017, 2017.

Bibliography

160

161

9 List of Figures

Figure 1.1: Idealized ambient particulate size distribution including formation and removal processes (from Finlayson-Pitts and Pitts, 2000). ................................................................................................. 2

Figure 1.2: Radiative forcing 2011 (1750 as reference for pre-industrial atmosphere) from Myhre et al. (2013). ............................................................................................................................................. 3

Figure 1.3: Global Radiative forcing of different ambient aerosol constituents evolving over time (from Myhre et al., 2013). .................................................................................................................... 4

Figure 1.4: Surface solar radiation in Potsdam, Germany, (GEBA, Gilgen et al., 1999) and sulfur emissions in Europe (Stern et al., 2005) temperature anomalies over Europe from 1930 to 2010, and annual temperature anomaly in Europe over land (Brohan et al., 2006). ............................................. 5

Figure 1.5: Mortality and SO2 concentrations during the great London smog (adapted from Bell et al., 2001 and completed with data from the same publication). ........................................................... 6

Figure 1.6: Global distribution of mortality attributed to air pollution (from Lelieveld et al., 2015). . 7

Figure 1.7: Modelled particle deposition probability upon inhalation in the human respiratory system as a function of particle size and location (from Maynard and Kuempel, 2005). ................................. 9

Figure 1.8: Aerosol composition for different environments in Europe in PM10 and coarse PM, with unaccounted mass from water and other aerosol constituents which were not analyzed at all sites (adapted from Putaud et al., 2004). .................................................................................................... 10

Figure 1.9: NR-PM1 AMS analyses at multiple sites on the northern hemisphere: a) aerosol composition (organic (Org), sulfate, nitrate, ammonium, chloride), b) oxygenated OA concentrations, c) and primary OA factors as HOA and BBOA (from Zhang et al., 2011). .............. 11

Figure 2.1: NR-PM1 mass concentrations (organic, sulfate, nitrate, ammonium, and chloride) measured by AMS at multiple sites. Source apportionment results of OA are displayed: traffic (HOA), other POA as BBOA (other OA), oxygenated OA linked to SOA (OOA), separated into semi-volatile (SV-OOA) and low-volatility (LV-OOA) from Jimenez et al. (2009). ........................ 14

List of figures

162

Figure 3.1: Scheme of the high-resolution-time-of-flight aerosol mass spectrometer (HR-ToF-AMS) from DeCarlo et al. (2006). ................................................................................................................ 18

Figure 3.2: Example of mass resolution of 3 AMS types: Quadrupole, C-ToF-MS, HR-ToF-MS from De Carlo et al. (2006). ........................................................................................................................ 19

Figure 4.1: Data recorded with HR-ToF-AMS of filter samples collected in Zürich (2011-2012). Data from a typical measurement cycle are underlaid in gray. (a) Raw signals obtained for organic aerosol (OA, green), nitrate (NO3

- , blue), sulfate (SO42-, red), and ammonium (NH4

+, orange), where AMS filter air as well as blank and sample measurements are indicated. (b) OA average signal for samples and blanks (logarithmic scale), blank correction curve and the noise (smoothed standard deviation of the blank) associated with the signal of different species used for the calculation of errors. On the y axes, a.u. denotes arbitrary units. ............................................................................. 35

Figure 4.2: Offline AMS SO42- blank corrected concentrations compared to theoretical SO4

2- loadings of the filter fractions (μg). The theoretical SO4

2- loadings are calculated based on ambient SO4

2- concentrations measured by the ACSM for the Zürich yearly cycle and the volume of air sampled through the analyzed filter fraction. Results are fitted using a power function (ln(y) = 2.3 × ln(x) - 5.2). ......................................................................................................................................... 36

Figure 4.3: Estimated recoveries of organic compounds based on the comparison of OA/SO42- ratios

using the offline AMS method to reference measurements for different days. The error bars represent the variability of the offline OA/SO4

2- ratio within a sample and were obtained from different runs during the same measurement of the same sample. (a) The reference OA/SO4

2- ratio is obtained by OC filter measurements (Sunset OC/EC analyzer) scaled to OA using OM/OC values from the HR offline AMS data and SO4

2- from IC. (b) OA/SO42- ratios from online measurements were used as

reference values. For both Paris campaigns and the Zürich spring campaign, the online measurements were conducted using HR-ToF-AMS and for the yearly cycle in Zürich by a quadrupole ACSM. (c) For Zürich (2011-2012), probability density functions of Rbulk are presented both using the offline AMS measurements as well as using WSOC from the Sunset OC=EC Analyzer (in combination with OM=OC ratios from offline AMS). ................................................. 38

Figure 4.4: Comparison between 24 h average online and offline AMS (both PM2:5) spectra for winter (a) and spring (b) samples, collected in Zürich. Fragments (m/z) commonly considered as source-specific markers are explicitly labeled with their nominal mass. ........................................... 41

Figure 4.5: Median recovery of single organic fragments, and chemical families for the Zürich spring campaign (offline vs. online PM2.5 AMS). The first and third quartiles of the inter-sample variability are shown as error bars. A ratio of 1 indicates a recovery of 100 %. The fragments are color-coded with the family (CH (hydrocarbon fragments, split into saturated and unsaturated), CHOz=1 and CHOz>1 (oxygenated fragments), and CHN, nitrogen-containing hydrocarbon fragments). Numbers across the top of the plot indicate the fragments’ nominal mass. Families include all respective fragments weighted by their mass contribution. .............................................. 42

Figure 4.6: Change in the time-dependent contribution of Q/Qexp as a function of the number of factors Δ(Qi,cont/Qexp,i,cont) for a chosen offline solution (for aHOA = 0.0 and aCOA = 0.0). .................. 43

Figure 4.7: Residuals weighted with the uncertainty (residuals/uncertainty) of the offline solutions for the periods April-September and October-March (example shown for one chosen solution, aHOA = 0.0; aCOA =0.0). Panels (a, b) show residuals as a function of m/z averaged over the whole periods color-coded with the probability that the residuals for April-September are the same as for October-

163

March (Wilcoxon-Mann-Whitney test). Panels (c, d) shown the probability distribution function (pdf) of R/U during the same periods. ................................................................................................ 45

Figure 4.8: Relative width of the distributions cj/cmarker displayed as a function of aHOA and aCOA. Panel (a) shows the sum of the criteria for HOA, BBOA, and OOA being the sum of OOA1 and OOA2. (The chosen solutions are pointed out in the white area.) Panels (b-d) show the individual criteria as a function of the a values of HOA and COA aHOA, aCOA. .................................................. 47

Figure 4.9: Comparison of overall factor profiles obtained for the chosen solutions both from the offline (left, for HOA and COA, spectra from Mohr et al. (2012) were used as reference) and the retrieved factor profiles from the online source apportionment (right). ............................................. 48

Figure 4.10: Recoveries Rk for HOA, COA, BBOA, and OOA (OOA1COOA2) obtained from the intercomparison of source apportionment results of offline AMS to online ACSM data (Zürich 2011-2012). 100 000 random combinations of offline and online solutions and randomly chosen offline repeats result in the same amount of time-independent Rk, which are expressed as probability density functions (pdf). ................................................................................................................................... 50

Figure 4.11: Comparison of factor contributions from separate offline (PM10 AMS, two constrained factors: HOA, COA) and online (PM1 ACSM) source apportionment using ME-2 (traffic (HOA), cooking (COA), biomass burning (BBOA), and oxygenated organic aerosol; OOADOOA1COOA2). Factor-specific recoveries (Rk) are applied to the offline contributions. Error bars (in gray) denote the variability between the different ME-2 solutions and for different recorded spectra per sample for offline and for online only the first of the two. Panels (a-d) show scatter plots comparing the absolute contribution of the respective source/OA category for offline AMS and online ACSM measurements. The color code distinguishes all factor contributions (bullets, saturated colors) from winter points (open circle, light colors). The gray dashed line indicates the 1 V 1 line. Panels (e-g) show the correlation with the respective markers: black symbols represent the absolute contribution of the respective source for the online ACSM measurements and the colored symbols represent the absolute contribution of the same source for the offline AMS measurements. .................................. 53

Figure 4.12: Ranges of ratios of the contribution of different factors to their markers for the offline (corrected with Rk) and online ACSM source apportionment results. Note that OOA is the sum of OOA1 and OOA2. .............................................................................................................................. 54

Figure 5.1: Map of study area with locations of sites indicating their characteristics. The topography is displayed as meters above sea level. ............................................................................................... 61

Figure 5.2: Step-by-step outline of adopted source apportionment approach (factor recoveries Rk). aHOA and aHOA represent the a value applied for HOA and COA, respectively. ................................. 66

Figure 5.3: PMF factor profiles of HOA, COA, BBOA, SOOA, WOOA, SC-OA, color-coded with ion family of PMFblock (average). fm/z is the relative intensity at a specific mass-to-charge ratio (m/z). ............................................................................................................................................................ 73

Figure 5.4: HOA, COA, BBOA, SC-OA, SOOA, and WOOA and their respective marker concentrations as a function of time for Zürich in 2013. Depicted are the median factor time series results for the different PMF datasets (median) including the uncertainties for PMFblock (first and third quartile) (green: PMFblock, black: PMFzue,isol, red: PMFzue,reps, pink bullets: PMF1filter/month). ..... 74

Figure 5.5: Scatter-plots for the different extreme sensitivity tests for Zürich and for all sites for PMFblock median concentrations): a) HOA vs NOx, b) BBOA vs levoglucosan, c) SOOA vs temperature, d) WOOA vs NH4

+. ....................................................................................................... 75

List of figures

164

Figure 5.6: Distributions of Rk for HOA, COA, BBOA, OOA (WOOA plus SOOA) and SC-OA (500 pairs). A priori information for HOA, COA, BBOA, and OOA on Rk is used from Daellenbach et al., 2016, with propagated errors and biases, while RSC-OA is determined in this study. Distributions of all factors have a resolution of dRk=0.01 except for dRSC-OA=0.05. ........................ 80

Figure 5.7: Relative σa (a) and err‘tot (b) for factor concentrations > 0.1 µg/m3 as a function of factor concentration. err‘tot includes the uncertainties from a-value, seed variability and Rk, and the different PMF datasets. ...................................................................................................................... 81

Figure 5.8: Cellulose concentrations as a function of the season and site. For comparison literature data from other years is added European sites: Payerne (Bozzetti et al., 2016, error bars representingthe standard deviation of the measurements in June and July), Puy de Dôme, Schauinsland, Sonnblick, K-Puszta (Sanchez-Ochoa et al., 2007), Birkenes, Hyytiälä, Lille Valby, and Vavihill (Yttri et al., 2011). ......................................................................................................... 84

Figure 5.9: Map of Switzerland with yearly cycles. Negative concentrations were set to 0 prior to normalization for display. The OA mass explained by the source apportionment analysis is termed OAexpl. ................................................................................................................................................ 86

Figure 6.1: m/z calibration of LDI-MS analysis of aerosol collected on quartz-fiber filter: a) example of calibrated silver (Ag-spec, blue) and no-silver-containing average mass spectra (noAg-spec, green) of a filter sample. The insert in a) displays a zoom-in of the Ag-spec and noAg-spec. b) operational resolution determined based on silver-mono-, di-, and trimer. c) m/z calibration accuracy for both steps of the m/z calibration. ................................................................................................ 102

Figure 6.2: Error model parametrization based on 3 samples (Basel, 2013-06-21, 2013-09-21, 2013-12-22) measured on 3 instances with each time 10 repeats on 1 sample holder. ............................. 103

Figure 6.3: samples representing different combustion sources a) and b) traffic emissions from samples collected in Islisberg tunnel (exit, Wettswil, Switzerland), and c) primary wood burning using whole cycle emissions d) and stable flaming phase emissions d). In absence of a measurable mass spectrum, no spectrum for cooking emissions is displayed. .................................................... 104

Figure 6.4: LDI-MS mass spectra from summer and winter samples (dates given in the legend) from a traffic-influenced (Bern), a rural (Payerne), a wood burning-influenced (San Vittore) and an urban background (Zürich) site. Spectra are color-coded with the correlation coefficient (r) between the m/zs and a specific environmental parameter for the whole dataset: for a) the correlation with NOx during summer, for b) the correlation with temperature, for c) and d) the correlation with levoglucosan. .................................................................................................................................... 105

Figure 6.5: PMF factor profiles (colored sticks) and their uncertainty (grey shaded areas, variability among PMF runs): a) traffic1, b) traffic2, c) efficient wood burning (BBeff), d) inefficient wood burning (BBineff1), e) inefficient wood burning 2 (BBineff2), f) lower molecular weight OA (LMW-OA), and g) biogenic OA (bio-OA). .................................................................................... 107

Figure 6.6: Scatterplots between factor-time series and respective markers (traffic1, traffic2, BBeff, BBineff1, BBineff2, bio-OA, LMW-OA, eBCtr, levoglucosan, potassium, ammonium are displayed in µg/m3, NOx in ppm, and temperature in °C)................................................................................ 108

Figure 6.7: Comparison of LDI-MS to reference offline AMS source apportionment results for the sum of traffic related factors (LDI-traffic, R2

LDI,oAMS=0.04), sum of wood burning related factors (LDI-BB, R2

LDI,oAMS =0.83), bio-OA (R2LDI,oAMS =0.62), and LMW-OA (R2

LDI,oAMS =0.45). ......... 111

165

Figure 6.8: Comparison of LDI-MS factor concentrations corrected using relative response factors (rRF) to uncorrected LDI factor concentrations (Scenario 1). .......................................................... 115

Figure 6.9: Relative for the different PMF factors as a function of the factor concentration. ... 116

Figure 6.10: Relative factor time series of 7 identified factors for all nine sites in study area. ........ 117

Figure 6.11: a) time series of traffic1 normalized to eBCtr and ECf in comparison to OCf /ECf in Magadino, b) influence of inefficient wood burning emissions (BBineff1) in comparison to the sum of wood burning influenced factors (BB=BBineff1+BBineff2+BBeff) for the entire datasets. ....... 118

Figure 7.1: Average contributions of OA sources resolved using offline AMS data discussed in this thesis and other studies (this dissertation, and from Bozzetti et al., 2016, 2017a, 2017b, Huang et al., 2014). ................................................................................................................................................ 122

SI Figure A.1: Qi,j as a function of the number of factors for a reference experiment with all data used in PMF (9 sites, full year 2013, HOA and COA constrained with a=0.0 (b and d). Δ(median(Qi,j))max is evaluated for the different periods during the year 2013 (January-February-March, April-Mai-June, July-August-September, October-November-December) and for all sites (a and c). The grey line depicts the difference between the category (geographical or season) with the highest and the lowest median Qi,j. ................................................................................................... 170

SI Figure A.2:Qi,j as a function of the day of the week. ................................................................... 171

SI Figure A.3: a) Average Qi,j of ions in PMFblock as a function of their mass-to-charge ratio (m/z). The ions are color-coded with their composition (CH: ions consisting only of C and H; CHO1: ions consisting of C, H, and 1 O; CHOgt1: ions consisting of C, H, and more than 1 O; CHN: ions consisting of C, H, N, (and O); CS: ions consisting of C, H, S, (and O)). b) Average Qi,j of the ions in PMFblock as a function of their mass defect (exact mass - nominal mass) as well as a histogram of the number of ions with a certain mass defect. The mean Qi,j of the ion families is displayed separately. ......................................................................................................................................... 172

SI Figure A.4: Cumulative density functions of a-values for HOA and COA for the accepted solutions. ........................................................................................................................................... 173

SI Figure A.5: Histograms of yearly average factor concentrations of all selected PMFblock solutions (after Rk correction). ......................................................................................................................... 174

SI Figure A.6: mass spectral fingerprints of BBOA (PMFblock) and nebulized levoglucosan. fion is the fraction of signal of a respective ion to the sum of the total signal. ................................................. 176

SI Figure A.7: SOOA concentrations compared to temperature, ozone, and Ox (O3+NO2) for Zürich. .......................................................................................................................................................... 177

SI Figure A.8: number density functions of source apportionment results obtained using a significance level of 0.05 normalized to results obtained using a significance level of 0.5: a) Comparison of factor concentrations b) Comparison of uncertainty estimate (σa). .......................... 179

SI Figure B.1: Cumulative density function (CDF) of the peak intensity in field blank analyses (Iavg,m/z) normalized to the minimal absolute error (σabs). ................................................................. 182

List of figures

166

SI Figure B.2: mass spectral signature of field blank analysis: a) colour-coded with the ratio of peak intensity (Iavg,m/z) divided by the minimal absolute error (σabs), b) colour-coded with the ratio of the measurement of the peak intensity (Istdev,m/z) divided by Iavg,m/z. ...................................................... 183

SI Figure B.3: measured total intensity plotted against filter loadings in OC, OC+EC, and PM10 (color-coded with measurement date). ............................................................................................. 184

SI Figure B.4: Difference of total measured signal intensity (Itot) of samples repeated on different occasions to the average of Itot normalized to the average of Itot (lines with dots with varying colors for the different samples). The data is summarized using the median (black diamond) and quartiles (black horizontal lines), binned as intra-day repeats, repeats within 30, 30-60,60-90,90-120 days, respectively. ..................................................................................................................................... 185

SI Figure B.5: Scatterplots of selected m/zs (scaled to OM, µg/m3) and NOx (ppm), levoglucosan (µg/m3), and temperature (°C). ......................................................................................................... 186

SI Figure B.6:Qavg as a function of the number of factors for the single sites (color-coded with geographical region). ........................................................................................................................ 187

SI Figure B.7: change in time-dependent contribution Qavg,i as a function of the number of factors, ΔQavg,i. .............................................................................................................................................. 188

SI Figure B.8: Qavg as a function of m/z when allowing for 7 factors. ............................................. 189

SI Figure B.9: Scatterplots of selected traffic 1 / 2 (µg/m3) and NOx (ppm), and temperature (°C). Correlation coefficients are computed on all summer points together. ............................................ 192

SI Figure B.10: Difference in factor concentration of samples measured on different instances to the average factor concentration normalized to the average concentration as a function of measurement time delay. The data is summarized using the median (diamond) and quartiles (horizontal lines) of the samples, binned as intra-day repeats, repeats within 30, 30-60,60-90,90-120 days, respectively (195, 240, 46, 53, 28 points per bin, respectively). .......................................................................... 194

SI Figure B.11: relative yearly average factor contributions for the different sites (grey: traffic1, black: traffic2, orange: BBeff, brown: BBineff1, violet: BBineff2, light green: bio-OA, dark green: LMW-OA. ........................................................................................................................................ 195

167

10 List of Tables

Table 1.1: PM limits established in Switzerland, EU, U.S.A., and WHO. ........................................... 7

Table 3.1: Advantages and disadvantages of offline AMS technique. ............................................... 20

Table 3.2:Advantages and disadvantages of LDI technique. .............................................................. 21

Table 4.1: Filter samples and available supporting measurements used in this study. ....................... 28

Table 5.1: Study sites with geographical location and classification ................................................. 61

Table 5.2: Comparison of factor time series to reference data for different PMF input datasets runs (by Pearson and Spearman correlation coefficient, Rp

2 and Rs). Displayed are the results for PMFblock unless stated otherwise. ...................................................................................................................... 77

Table 5.3: Yearly average contribution and uncertainty of resolved factors for PMFblock run for the different sites and the average for all sites. The uncertainty is calculated based on the variability in the yearly averages from PMFblock and the variability between the 4 sensitivity tests. ...................... 87

Table 6.1: relative response factors (rRF) for LDI-MS analyses for 3 different scenarios. ............. 114

Table 6.2: yearly averages of the relative factor contributions and NOx concentrations. ................ 119

SI Table A.1:set of acceptance criteria used. r is the correlation coefficient between a factor time series and the respective marker. Q25 is the 1st quartile and Q75 the 3rd quartile. .............................. 175

SI Table A.2: σminimal and k for the different factors including their uncertainty.............................. 179

SI Table B.1: correlation coefficients between factor and marker time series (Rp: pearson correlation coefficient, Rs: Spearman correlation coefficient). ........................................................................... 191

SI Table B.2: error model coefficients for parameterized σintrad. ...................................................... 193

List of tables

168

169

A

Supplementary Material for chapter 5

Long-term chemical analysis and organic aerosol source apportionment at 9 sites in Central Europe:

source identification and uncertainty assessment

A: Supplementary material

• Number of factors:

Qi,j is computed using the PMF residuals (eij) and the PMF input errors (si,j):

(SI.A.1)

SI Figure A.1: Qi,j as a function of the number of factors for a reference experiment with all data used in PMF (9 sites, full year 2013, HOA and COA constrained with a=0.0 (b and d). Δ(median(Qi,j))max is evaluated for the different periods during the year 2013 (January-February-March, April-Mai-June, July-August-September, October-November-December) and for all sites (a and c). The grey line depicts the difference between the category (geographical or season) with the highest and the lowest median Qi,j.

Fig. SI.A.1 shows Qi,j s as a function of the number of factors for different sites (b)

and seasons (d) and the difference between the highest (a) and lowest (c) median to

evaluate the maximal difference in the mathematical quality of the solutions. As

expected, forcing PMF to explain the variability in the dataset only with the 2

constrained factors (p=2), results in very high median Qi,j . Δ(median(Qi,j))max shows the

difference in the median Qi,j between groups of points like sites or season. The smaller

the Δ(median(Qi,j))max, the smaller are the differences in the mathematical quality of the

PMF solution for the different seasons/sites. To explain the temporal and geographical

variability at least 5 factors are required. However, the difference between the site that

is best explained and the site that is least explained is approximately 6 when using 5 or

171

6 factors. When increasing to 6 factors, also a factor explaining the variability of sulfur-

containing organic ions (especially, CH3SO2+) is resolved. Therefore, we opted to

perform PMF using 6 factors. Using 6 factors, there is also no difference between the

average Qi,j on week-days and weekend (Fig. SI.A.2).

SI Figure A.2:Qi,j as a function of the day of the week.

However, for PMFblock also with 6 factors, the average Qi,j is clearly larger (7 only

the Zürich data points) than the ideal value of 1, i.e. the PMF residuals are larger than

the measurement uncertainties. In comparison to PMFblock, the average Qi,j for Zürich is

slightly reduced for the same number of factors when only including 1 site in PMF

(PMFzue,isol, PMFzue,reps, average Qi,j 6). In this study, we analyse yearly cycles and,

thereby, assume constant factor profiles throughout the year which can contribute to

Q>1.

Another possible reason for Q>1 is an underestimation of the measurement

uncertainty. A main contributor in high-resolution AMS data treatment (attribution of

the signal at a nominal mass to several ions) stems from errors in the m/z calibration

which could not be incorporated in the current data analysis. Recent studies demonstrate

that for overlapping peaks (ions) the measurement uncertainties are strongly

underestimated (Cubison et al., 2015; Corbin et al., 2015). For PMFblock using 6 factors,

average Qi,j do not depend on m/z but rather on the ion family (Fig. SI.A.3): ions

consisting of C, H, S, (and O) summarized under the name (CS) and ions consisting of

C, H, N, (and O) summarized under the name CHN have a higher Qi,j than hydrocarbon


ions (CH, only C and H) and oxygenated ions (CHOz=1 with 1 oxygen and CHOz>1 with

more than 1 oxygen). Since the time series of CH3SO2+ is event-driven, the high Qi,j of

this ion hints to the fact that PMF is unable to accurately resolve all of these events.

The average Qi,j for ions with a mass defect (nominal mass - exact ion mass) around

0.03 a.m.u. is higher than for the other ions (Fig SI.A.3). Mass defects in this range are

most common in our dataset. This makes these peaks prone to overlap with other ions

and thus their error prone to an underestimation because this effect is not considered in

the sij calculation (described above).

SI Figure A.3: a) Average Qi,j of ions in PMFblock as a function of their mass-to-charge ratio (m/z). The ions are color-coded with their composition (CH: ions consisting only of C and H; CHO1: ions consisting of C, H, and 1 O; CHOgt1: ions consisting of C, H, and more than 1 O; CHN: ions consisting of C, H, N, (and O); CS: ions consisting of C, H, S, (and O)). b) Average Qi,j of the ions in

173

PMFblock as a function of their mass defect (exact mass - nominal mass) as well as a histogram of the number of ions with a certain mass defect. The mean Qi,j of the ion families is displayed separately.

SI Figure A.4: Cumulative density functions of a-values for HOA and COA for the accepted solutions.

Cumulative density functions for the a-values of HOA and COA are presented for

the accepted solutions in Fig. SI.A.4. We found that 80% of the accepted solutions have

an a-value≤0.3 for HOA and an a-value≤0.5 for COA. The output HOA and COA

factor profiles are therefore not significantly variable and very similar to the input

profiles, indicating that similar solutions were selected. Furthermore, the yearly average

factor concentrations of all selected PMFblock solutions after Rk correction are shown for

the case of Zürich as an illustration in Fig. SI.A.5. The distributions of each of the

different factors do not show more than 1 distinct mode, indicating that we do not have

several populations of solutions.


The yearly average factor concentrations of all selected PMFblock solutions after Rk

correction areshown for the case of Zürich as an illustration (Fig. SI.A.5). The

distributions of each of the different factors do not show more than 1 distinct mode.

SI Figure A.5: Histograms of yearly average factor concentrations of all selected PMFblock solutions (after Rk correction).

175

• Quality assessment of solutions:

Set of criteria used when assessing quality of a single PMF run:

SI Table A.1: set of acceptance criteria used. r is the correlation coefficient between a factor time series and the respective marker. Q25 is the 1st quartile and Q75 the 3rd quartile. criteria on profile f(CO2

+) f(C2H4O2+)

HOA <0.4 <0.004 COA <0.4 <0.01 Criteria on time series HOA r(HOA,NOx)> 0 & r(HOA,NOx)> r(COA,NOx) BBOA r(BBOA,levo)> 0 SC-OA r(SC-OA,CH3SO2

+)>0 Mass closure criteria

OCres total Q25(res-OC)<0 & Q75(res-OC)>0 Magadino winter, Magadino summer Q25(res-OC)<0 & Q 75(res-OC)>0 Zürich winter, Zürich summer Q25(res-OC)<0 & Q 75(res-OC)>0 Magadino, Zürich Q25(res-OC)<0 & Q 75(res-OC)>0 HOC<median, HOC>median Q25(res-OC)<0 & Q 75(res-OC)>0 COC<median, COC>median Q25(res-OC)<0 & Q 75(res-OC)>0 BBOC<median, BBOC>median Q25(res-OC)<0 & Q 75(res-OC)>0 SC-OC<median, SC-OC>median Q25(res-OC)<0 & Q 75(res-OC)>0 WOOC<median, WOOC>median Q25(res-OC)<0 & Q 75(res-OC)>0 SOOC<median, SOOC>median Q25(res-OC)<0 & Q 75(res-OC)>0 for PMF with 12 filters per site summer Magadino and Zürich

Q25(res-OCi)<0 & Q 75(res-OCi)>0

for PMF with 12 filters per site winter Magadino and Zürich

Q25(res-OCi)<0 & Q 75(res-OCi)>0


• Comparison of mass spectral signature of BBOA and nebulized levoglucosan:

Fig. SI.A.6 demonstrated the high similarity between the retrieved BBOA

signature and the mass spectrum of nebulized levoglucosan.

SI Figure A.6: mass spectral fingerprints of BBOA (PMFblock) and nebulized levoglucosan. fion is the fraction of signal of a respective ion to the sum of the total signal.

177

• Comparison of SOOA to ozone and Ox:

In Figure SI.A.7, we compare the SOOA concentrations to ozone and Ox (O3+NO2)

for Zürich. The SOOA concentrations follow best the temperature (Rs,SOOA,temp=0.65,

Fig. SI.A.7.a) but show also some correlation to ozone Rs,SOOA,O3=0.33, Fig. SI.A.7.b)

and Ox (Rs,SOOA,Ox=0.38, Fig. SI.A.7.c).

SI Figure A.7: SOOA concentrations compared to temperature, ozone, and Ox (O3+NO2) for Zürich.


• Uncertainty estimation and propagation:

The uncertainty described by the interquartile range from the a-value sensitivity

assessment (σa) does not fully explain the variability between the 4 sensitivity tests. In

the following, we use the source apportionment results of the 12 filters common to all 4

sensitivity tests for achieving a better estimate of the uncertainty of the factor

concentrations. For these 12 filters the uncertainty is estimated by propagating the

variability between the median concentrations for the 4 sensitivity tests (σb) and half the

interquartile range of PMFblock (σa, Eq. SI.A.2):

(SI.A.2)

In absence of σb for all other points, we parametrize σb. We express σb as a function

of a minimal uncertainty (σminimal) and an uncertainty proportional (k) to the factor

concentration and fit the equation using the 12 points in common to all datasets (Eq.

SI.A.3):

( SI.A.3)

The uncertainty (σa and σb) is propagated for all points using the parameters from Eq.

SI.A.3 in order to obtain the total uncertainty for all points in the dataset (Eq. SI.A.4):

(SI.A.4)

The resulting coefficients of the error model are presented in table SI.A.2:

179

SI Table A.2: σminimal and k for the different factors including their uncertainty.

factor σminimal k

HOA 0.16±0.06 0.39±0.24

COA 0.09±0.01 0.52±0.09

BBOA 0.06±0.01 0.48±0.05

SC-OA 0.30±0.00 0.32±0.27

WOOA 0.28±0.08 0.42±0.27

SOOA 0.05±0.01 0.24±0.05

• Sensitivity to significance level of statistical tests in PMFblock:

For PMFblock, a sensitivity test with significance level of 0.05 instead of 0.5 as in the

base case was performed. The factor concentrations and their corresponding

uncertainties (σa) are compared and displayed as number density functions (Fig. SI.A.8).

Changes in the estimated factor concentrations are within 10% of the factor

concentrations for SCOA and smaller for all other factors. The uncertainty related to

COA is decreased when lowering the significance level to 0.05, while the other factors

remain largely unaffected.

SI Figure A.8: number density functions of source apportionment results obtained using a significance level of 0.05 normalized to results obtained using a significance level of 0.5: a) Comparison of factor concentrations b) Comparison of uncertainty estimate (σa).

181

B

Supplementary material for chapter 6

Insights into organic-aerosol sources via a novel laser-desorption/ionization mass spectrometry

technique applied to one year of PM10 samples from nine sites in central Europe


• Field blank analyses

Analyses of field blanks (also spiked with AgNO3) exhibited low signal with 99% of

peaks below detection limit (defined as 3*σabs, Fig SI.B.1).

SI Figure B.1: Cumulative density function (CDF) of the peak intensity in field blank analyses (Iavg,m/z) normalized to the minimal absolute error (σabs).

Mass spectra from field blanks exhibit low signals throughout the spectrum (Fig

SI.B.1 and SI.B.2). 1% of all peaks (m/z 197, 249, 251, 322,324) show signals which

are higher than the detection limit (Fig SI.B.2a, highlighted in black). These peaks are

also subjected to a high variability between the repeated analyses (Figure SI.B.2b,

highlighted in black).

183

SI Figure B.2: mass spectral signature of field blank analysis: a) colour-coded with the ratio of peak intensity (Iavg,m/z) divided by the minimal absolute error (σabs), b) colour-coded with the ratio of the measurement of the peak intensity (Istdev,m/z) divided by Iavg,m/z.


• Comparison of total LDI-MS signal and external measurements:

The total LDI-MS-intensity measured for the filter samples does not show a

relation to the filter loading in terms of OC, sum of OC and EC, or PM10 (Fig SI.B.3).

SI Figure B.3: measured total intensity plotted against filter loadings in OC, OC+EC, and PM10 (color-coded with measurement date).

185

• Influence of measurement time delays on total LDI-MS signal recorded:

In order to assess the repeatability of the total intensity, all repeats are normalized

to the first measurement of the respective sample (Fig. SI.B.4). The intra-day variability

(without the ones used to parametrize the error model) ranges from -27% (first quartile)

to +33% (third quartile) of the average measurement. The inferred instrumental drift has

an effect on the measured intensity (Kruskal-Wallis-test, p-value<0.05, Kendall-Tau-

Test, p-value<0.05). Therefore, for quantification purposes we use external data like

OM.

SI Figure B.4: Difference of total measured signal intensity (Itot) of samples repeated on different occasions to the average of Itot normalized to the average of Itot (lines with dots with varying colors for the different samples). The data is summarized using the median (black diamond) and quartiles (black horizontal lines), binned as intra-day repeats, repeats within 30, 30-60,60-90,90-120 days, respectively.


• Comparison of selected m/zs and external measurements:

For visual comparison certain m/zs are plotted as a function of external

measurements (Fig. SI.B.5).

SI Figure B.5: Scatterplots of selected m/zs (scaled to OM, µg/m3) and NOx (ppm), levoglucosan (µg/m3), and temperature (°C).

• Preliminary source apportionment analysis and factor identification:

A preliminary source apportionment analysis is performed by unconstrained PMF

(no a priori information on source fingerprints used) in order to determine the number

of factors for the source apportionment analysis. Based on Qavg, defined as

187

(n: number of spectra, m:number of m/zs), the model explains the sites

north and south of the alpine crest equally well when allowing for at least 4 factors (Fig.

SI.B.6). Furthermore, we assess the change in time-dependent Qavg,i, ,

when increasing the number of factors, (Fig. SI.B.7). A significant

signifies that structure in the residuals disappeared when adding an additional factor. Up

to 5 factors, removed structure is evident but not when adding further factors. When

increasing the number of factors up to 7, a factor with a similar signature as stable phase

wood burning and a second traffic factor appear. A further increase leads to the

separation of another possibly traffic related factor which contributes less than 5%.

Therefore, we consider 7 factors to be optimal for this dataset.

SI Figure B.6:Qavg as a function of the number of factors for the single sites (color-coded with geographical region).


SI Figure B.7: change in time-dependent contribution Qavg,i as a function of the number of factors, ΔQavg,i.

For the 7 factor solution, we assess how well the different m/zs are explained by

PMF using the quantity Qavg,j, (Fig. SI.B.8). Qavg,j shows that even with 7

factors not all m/zs are explained within their measurement uncertainty. This might be

linked to an underestimation of the measurement uncertainty itself. Additionally, factor

profiles, which are assumed to be constant, might vary with changing seasons

contributing to increased Qavg,j of certain m/zs. The m/zs 322, 324, 326 might be affected

by the signal from silver-trimers, even though these peaks were removed prior to PMF.

189

SI Figure B.8: Qavg as a function of m/z when allowing for 7 factors.

The explained variation (EV, Paatero et al., 2004, Canonaco et al., 2013) describes

how much of the measured variation (time or variable) is explained by each PMF-factor

(Eq. SI.B.1, for variable j):

for k=1,…,p (SI.B.1)

fk,j are the constant factor profiles and gi,k their time-dependent contributions. The

index i represents a specific point in time (up to the number of points in time n, j the

signal at a specific m/z, and k a factor (up to the number of factors p). ei,j are the PMF-

residuals and σ’PMF,i,j the measurement uncertainties. Using the factor identifications

based on the preliminary source apportionment analysis, the factors of the bootstrap

runs are identified sequentially based on the explained variation as follows:

• traffic1: factor with maximal explained variation of m/z 177. • traffic2: factor with maximal explained variation of m/z 163 of the

remaining factors. • BBeff: factor with maximal average explained variation of the m/zs

85, 124, 140, 197, and 213 of the remaining factors. • BBineff: factor with maximal average explained variation of the m/zs

284 and 298 of the remaining factors.


• BBineff2: factor with maximal average explained variation of ions bigger than 300 a.m.u. of the remaining factors.

• bio-OA: factor with maximal average explained variation of m/z 74 and 104 of the remaining factors.

• LMW-OA: factor with maximal average explained variation of ions smaller than 150 a.m.u. of the remaining factors.

Bootstrap runs that showed mixing between factors described above were not further

considered for the analysis. Criteria used for minimizing the effect of mixing based on

the explained variation of groups of m/zs are lined out in the following:

1. EV(traffic2,m/zs>300) < EV(BBineff 2,m/zs>300) 2. EV(traffic2, m/zs 85, 124, 140, 197, 213) <

EV(BBeff, m/zs 85, 124, 140, 197, 213) 3. EV(traffic2, m/zs 284, 298) <

EV(BBineff1, m/zs 284, 298) 4. EV(LMW-OA, m/zs 284, 298) < EV(BBeff, m/zs 284, 298)

191

• Factor time series vs marker time series

Tab. SI.B.1 summarizes the correlation coefficients of all the factor time series

and respective markers for all sites and as an average.

SI Table B.1: correlation coefficients between factor and marker time series (Rp: pearson correlation coefficient, Rs: Spearman correlation coefficient).

Correlation bas* ber* fra* gal* mag* pay* vad* vi* zue* avg

Rp traffic1 NOx -0.18 -0.31 -0.08 0.13 -0.23 -0.26 -0.1 -0.29 -0.19 -0.18

Rp traffic2 NOx 0.10 -0.06 0.29 0.35 0.54 0.5 0.24 -0.23 0.32 0.24

Rp traffic1 eBCtr

-0.31 0.15

-0.01 -0.08

Rp traffic2 eBCtr

0.73 0.68

0.63 0.69

Rp BBeff levo 0.71 0.65 0.78 0.44 0.83 0.88 0.42 0.18 0.9 0.70

Rp BBineff1 levo 0.88 0.11 0.64 0.91 0.59 0.93 0.93 0.98 0.58 0.83

Rp BBineff2 levo 0.9 0.85 0.77 0.44 0.96 0.81 0.65 0.9 0.79 0.83

Rp BBeff K+ 0.29 0.61 0.78 0.32 0.67 0.1 0.35 -0.02 0.77 0.48

Rp BBineff1 K+ 0.08 0.08 0.85 0.26 0.75 0.11 0.5 0.9 0.4 0.52

Rp BBineff2 K+ 0.19 0.59 0.75 0.4 0.76 0.09 0.48 0.81 0.68 0.57

Rp BBeff BCwb 0.76 0.42 0.75 0.67

Rp BBineff1 BCwb 0.91 0.49 0.55 0.71

Rp BBineff2 BCwb 0.69 0.4 0.82 0.67

Rs bio-OA temp 0.73 0.69 0.74 0.76 0.69 0.78 0.78 0.72 0.75 0.74

Rp LMW-OA NH4+ 0.7 0.46 0.75 0.78 0.78 0.31 0.75 0.76 0.75 0.69

*ber: Bern, bas: Basel, fra: Frauenfeld, gal: St. Gallen, mag: Magadino, Pay: Payerne: vi: S. Vittore, zue, Zürich


Fig. SI.B.9 illustrates the relation between traffic1, traffic2 and NOx and eBCtr

when only using the summer points.

SI Figure B.9: Scatterplots of selected traffic 1 / 2 (µg/m3) and NOx (ppm), and temperature (°C). Correlation coefficients are computed on all summer points together.

• Uncertainty estimation of PMF results:

The variability (standard deviation) of the apportioned factor concentration for the

same measurement in the bootstrap runs is the base for the uncertainty estimate ( ).

Besides , we also account for the effect of the repeatability on the source

apportionment results (assessed by the standard deviation of the apportioned factor

concentration, σintraday). For the latter purpose, 3 filters were measured on 3 instances

throughout the measurement campaign (10 times during 1 day). For these filters, both

types of uncertainty can be readily propagated to (Eq. SI.B.2):

(SI.B.2)

For the other filters, cannot be estimated directly. Therefore, we parameterized

σintraday as a function of two error terms, one absolute and one relative, similarly to the

PMF input uncertainty described in Section 2.3.2: (Eq. SI.B.3):

193

(SI.B.3)

Eq. SI.B.3 was fitted using the repeatability tests for the 3 filters, to obtain the

parameters and for each of the PMF factors (Tab. SI.B.2) and

the factor concentration (conc). These parameters were then extrapolated to the

other filter samples, such that could be applied to all filter samples.

SI Table B.2: error model coefficients for parameterized σintrad.

, µg/m3 , -

traffic1 0.01±0.01 0.08±0.02 traffic2 0.09±0.13 0.00±0.21 BBeff 0.01±0.0 0.10±0.02

BBineff1 0.04±0.04 0.13±0.06 BBineff2 0.11±0.05 0.08±0.07 LMW-OA 0.02±0.01 0.12±0.04

bio-OA 0±0.01 0.12±0.04


• Influence of instrumental drifts on source apportionment results:

Fig SI.B.10 illustrates the difference in the apportioned factor concentrations as a

function of the measurement time. All factors are affected by instrumental changes

occurring during the measurement campaign (Kruskal-Wallis-test, p-value<0.05) and

for all factors but traffic1 (p-value=0.58) and BBineff1 (p-value=0.23) a trend is

identified consistent with an impact of an instrumental drift (ranked Mann-Kendall test,

p-value<0.05). The intra-day variance explains largely the total variance, which is

defined as the sum of intra-day (σintrad2) and inter-day (σinter

2) variance (for traffic1 97%,

traffic2 94%, BBeff 85%, BBineff1 89%, BBineff2 82%, LMW-OA 79%, bio-OA

97%).

SI Figure B.10: Difference in factor concentration of samples measured on different instances to the average factor concentration normalized to the average concentration as a function of measurement time delay. The data is summarized using the median (diamond) and quartiles (horizontal lines) of the samples, binned as intra-day repeats, repeats within 30, 30-60,60-90,90-120 days, respectively (195, 240, 46, 53, 28 points per bin, respectively).

195

• Relative yearly average contributions of PMF factors for all nine sites: Fig. SI.B.11 presents the relative yearly average contributions of the PMF factors for all nine sites as presented in Tab. 6.2 in the manuscript.

SI Figure B.11: relative yearly average factor contributions for the different sites (grey: traffic1, black: traffic2, orange: BBeff, brown: BBineff1, violet: BBineff2, light green: bio-OA, dark green: LMW-OA.

197

11 Acknowledgement

During my time as a Ph.D. student at the Laboratory of Atmospheric Chemistry, I

could pursue my curiosity and met great people. I wish all of you who were part of this

journey all the best:

In particular I would like to thank:

… My doctoral advisor Prof. Dr. Urs Baltensperger for being my doctoral father

and giving me the opportunity to work at LAC.

… My direct supervisor Dr. André Prévôt for all the interesting discussion and

giving me the freedom also to follow side projects.

… Prof. Dr. Thomas Peter and Prof. Dr. James Schauer for accepting to be my co-

referees.

… Prof. Dr. Harald Bugmann for being the chairman at the Ph.D. exam.

… Dr. Guido Vogel and Dr. Valentin Pflüger for introducing my to the world of

LDI, helping me with all the issues, and always warmly welcoming me at MABRITEC.

Acknowledgement

198

… René Richter, Günther Wehrle for helping me set up the autosampler and all the

other times I didn’t know how to approach a problem.

… Thomas Attinger for providing me with a big screen, and especially his help

whenever I had major disagreements with my laptop.

… Lola Schmid for helping me when I was in need for advice or needed something

in the lab and especially for taking such good care of the milliQ system. Without this

water I wouldn’t have been able to do any of my measurements.

… Jay and Maarten for the help and insights related to AMS.

… Dr. Rolf Siegwolf for reminding me that plants are not only aerosol emitters and

all the discussions we had.

… My office was chameleon: in the mornings the smell of past centuries pointed

out the century-long history of PSI. But sometimes, a cinema-like atmosphere was

created by the odors of popcorn. Carlo’s popcorn, well, we all lived on it. Sometimes it

also felt like in a circus with all the madness. I highly enjoyed your company.

… With all that popcorn, how good that Imad “lent” me his squash racket. You lent

me much more than that. Thanks for all your time.

… Dogushan and Manuel who did their best in helping me burn the popcorn when

playing squash. Outstanding.

… Felix for appointing me (besides others) to ensure the edibility of the dishes he

prepared during his cooking emission campaigns.

… Manolis who shared my love for cheese and who never would let me it eat all

alone.

… Stephen helping me eat my fruits despite the danger of sugar.

Acknowledgement

199

… All you offline people: How many filters did we punch? How many liters of

milliQ water did we use? How many hours did we spend besides our dear old manual

nebulizer waiting to switch sample. Thank you, modern world, for the invention of the

auto-sampler. We could create quite some amount of data and had the chance to think in

detail about it.

… Giulia, Nassia, Lassi, Kevin, Emmanuel for all the efforts in the lab.

... Rotational ambiguity, we spoke often about it and accepted it as the reason for

headaches, but Francesco showed us how to enjoy all the craziness related to rotational

ambiguity.

… Joel, one day the people will understand the beauty of playing PMF residuals as

music, we shouldn’t give up on that because it is not crazy and will become a thing at

some point.

… Marco who shared my passion for naming vectors in funny ways.

… the LAC house inhabitants accepting my frequent, self-invited visits

… Andrin, Johanna, Ju, Christoph, Samuel, and all of you, in a lot of ways I grew

up with you, thanks for all the time since 2007.

… I remember I was so annoyed that I had to practice reading, I always thought that

I would just be able to read, thanks Christine for forcing me to learn it, I think it was

worth it. And lastly, yes Hans, I admit it, monkeys can indeed be blue.

Merci öich aune wo sone wichtige Teil vo denä vier Jahr syt gsi.

Acknowledgement

200

12 Curriculum vitae

Kaspar Rudolf Dällenbach

born on October 12, 1988

in Bern, Switzerland

Education

04/2013- 03/2017 PhD, Laboratory of Atmospheric Chemistry, Villigen PSI and

Department of Environmental Systems Science, ETH Zürich.

Dissertation advisor: Prof. Urs Baltensperger (ETH)

2007-2013 Bachelor’s and Master’s Program in Environmental Science

with a subject focus on Atmosphere and Climate at ETH

Zürich, Zürich, Switzerland.

• Master thesis on “Sensitivity Study on the Chlorine

Activation by Volcanic Aerosols” supervised by Prof.

Thomas Peter.

• Internship at Laboratory for Atmospheric Chemistry, Paul

Scherrer Institute, Villigen, Switzerland.

• ERASMUS program: Studies of Biology and Earth

Sciences at the University Uppsala, Sweden

Curriculum vitae

202

2001-2007 Gymnasium at Freies Gymnasium Bern, Bern, Switzerland

1995-2001 Elementary School in Schliern b. Köniz, Switzerland

Platform presentations at conferences

American Association for Aerosol Research 35th Annual Conference, Portland, U.S.A.,

2016 (oral presentation).

17th Annual AMS Users Meeting, Portland, U.S.A., 2016.

European Aerosol Conference, Milan, Italy, 2015 (two oral presentations).

16th Annual AMS Users Meeting, Milan, Italy, 2015.

European Geoscience Union General Assembly, Vienna, Austria, 2014 (poster).

European Aerosol Conference, Prague, Czech Republic, 2013 (oral presentation).

14th Annual AMS Users Meeting, Prague, Czech Republic, 2013.

Invited presentation

Laboratory for environmental chemistry, Paul Scherrer Institute, Villigen PSI.

Publications

Daellenbach, K. R., Bozzetti, C., Krepelova, A., Canonaco, F., Huang, R.-J., Wolf, R.,

Zotter, P., Crippa, M., Slowik, J., Zhang, Y., Szidat, S., Baltensperger, U., Prévôt, A. S.

H., and El Haddad, I.: Characterization and source apportionment of organic aerosol

using offline aerosol mass spectrometry, Atmos. Meas. Tech., 9, 23-39,

doi:10.5194/amt-9-23-2016, 2016.

Daellenbach, K. R., Stefenelli, G., Bozzetti, C., Vlachou, A., Fermo, P., Gonzalez, R.,

Piazzalunga, A., Colombi, C., Canonaco, F., Hueglin, C., Kasper-Giebl, A., Jaffrezo, J.-

L., Bianchi, F., Slowik, J. G., Baltensperger, U., El-Haddad, I., and Prévôt, A. S. H.:

Long-term chemical analysis and organic aerosol source apportionment at nine sites in

central Europe: source identification and uncertainty assessment, Atmos. Chem. Phys.,

17, 13265-13282, doi:10.5194/acp-17-13265-2017, 2017.

203

Daellenbach, K. R., El-Haddad, I., Karvonen, L., Vlachou, A., Corbin, J. C., Slowik, J.

G., Heringa, M. F., Bruns, E. A., Luedin, S. M., Jaffrezo, J.-L., Szidat, S., Piazzalunga,

A., Gonzalez, R., Fermo, P., Pflueger, V., Vogel, G., Baltensperger, U., and Prévôt, A.

S. H.: Insights into organic-aerosol sources via a novel laser-desorption/ionization mass

spectrometry technique applied to one year of PM10 samples from nine sites in central

Europe, Atmos. Chem. Phys., in press, 2018.

Bozzetti, C., Daellenbach, K., R., Hueglin, C., Fermo, P., Sciare, J., Kasper-Giebl, A.,

Mazar, Y., Abbaszade, G., El Kazzi, M., Gonzalez, R., Shuster Meiseles, T., Flasch, M.,

Wolf, R., Křepelová, A., Canonaco, F., Schnelle-Kreis, J., Slowik, J. G., Zimmermann,

R., Rudich, Y., Baltensperger, U., El Haddad, I., and Prévôt, A. S. H.: Size-resolved

identification, characterization, and quantification of primary biological organic aerosol

at a European rural site, Environ. Sci. Technol., 50, 3425-3434,

doi:10.1021/acs.est.5b05960, 2016.

Bozzetti, C. El Haddad, I., Salameh, D., Daellenbach, K. R., Fermo, P., Gonzalez, R.,

Minguillón, M. C., Iinuma, Y., Poulain, L., Müller, E., Slowik, J. G., Jaffrezo, J.-L.,

Baltensperger, U., Marchand, N., and Prévôt, A. S. H.: Organic aerosol source

apportionment by offline-AMS over a full year in Marseille, Atmos. Chem. Phys., 17,

8247-8268, https://doi.org/10.5194/acp-17-8247-2017, 2017.

Bozzetti, C., Sosedova, Y., Xiao, M., Daellenbach, K. R., Ulevicius, V., Dudoitis, V.,

Mordas, G., Byčenkienė, S., Plauškaitė, K., Vlachou, A., Golly, B., Chazeau, B.,

Besombes, J.-L., Baltensperger, U., Jaffrezo, J.-L., Slowik, J. G., El Haddad, I., and

Prévôt, A. S. H.: Argon offline-AMS source apportionment of organic aerosol over

yearly cycles for an urban, rural, and marine site in northern Europe, Atmos. Chem.

Phys., 17, 117-141, https://doi.org/10.5194/acp-17-117-2017, 2017.

Elser, M., Huang, R.-J., Wolf, R., Slowik, J. G., Wang, Q., Canonaco, F., Li, G.,

Bozzetti, C., Daellenbach, K. R., Huang, Y., Zhang, R., Li, Z., Cao, J., Baltensperger,

U., El-Haddad, I., and Prévôt, A. S. H.: New insights into PM2.5 chemical composition

and sources in two major cities in China during extreme haze events using aerosol mass

Curriculum vitae

204

spectrometry, Atmos. Chem. Phys., 16, 3207-3225, doi:10.5194/acp-16-3207-2016,

2016.

Klein, F., Platt, S. M., Farren, N. J., Detournay, A., Bruns, E. A., Bozzetti, C.,

Daellenbach, K. R., Kilic, D., Kumar, N. K., Pieber, S. M., Slowik, J. G., Temime-

Roussel, B., Marchand, N., Hamilton, J. F., Baltensperger, U., Prévôt, A. S. H., and El

Haddad, I.: Characterization of gas-phase organics using proton transfer reaction time-

of-flight mass spectrometry: cooking emissions, Environ. Sci. Tech., 50, 1243-1250,

doi:10.1021/acs.est.5b04618, 2016

Klein, F., Farren, N. J., Bozzetti, C., Daellenach, K. R., Kilic, D., Kumar, N. K., Pieber,

S. M., Slowik, J. G., Tuthill, R. N., Hamilton, J. F., Baltensperger, U., Prévôt, A. S. H.,

El and Haddad, I.: Indoor terpene emissions from cooking with herbs and pepper and

their secondary organic aerosol production potential, Sci. Rep., 6, 36623,

doi:10.1038/srep36623, 2016.

Krapf, M., El Haddad, I., Bruns, E. A., Molteni, U., Daellenbach, K. R., Prévôt, A. S.

H., Baltensperger, U., Dommen, J.: Labile peroxides in secondary organic aerosol,

Chem, 1, 603-616, doi:10.1016/j.chempr.2016.09.007, 2016.

Pieber, S. M., El Haddad, I., Slowik, J. G., Canagaratna, M. R., Jayne, J. T., Platt, S. M.,

Bozzetti, C., Daellenbach, K. R., Fröhlich, R., Vlachou, A., Klein, F., Dommen, J.,

Miljevic, B., Jimenez, J. L., Worsnop, D. R., Baltensperger, U., and Prévôt, A. S. H.:

Inorganic salt interference on CO2+ in Aerodyne AMS and ACSM organic aerosol

composition studies, Environ. Sci. Technol, 50, 10494-10503,

doi:10.1021/acs.est.6b01035, 2016.

Platt, S. M., El Haddad, I., Pieber, S. M., Zardini, A. A., Suarez-Bertoa, R., Clairotte,

M., Daellenbach, K. R., Huang, R.-J., Slowik, J. G., Hellebust, S., Temime-Roussel, B.,

Marchand, N., de Gouw, J., Jimenez, J. L., Hayes, P. L., Robinson, A. L.,

Baltensperger, U., Astorga, C., and Prévôt, A. S. H.: Gasoline cars produces more

carbonaceous particulate matter than modern filter-equipped diesel cars, 7, 4926,

doi:10.1038/s41598-017-03714-9, 2017.

205

Ulevicius, V., Byčenkienė, S., Bozzetti, C., Vlachou, A., Plauškaitė, K., Mordas, G.,

Dudoitis, V., Abbaszade, G., Remeikis, V., Garbaras, A., Masalaite, A., Blees, J.,

Fröhlich, R., Dällenbach, K. R., Canonaco, F., Slowik, J. G., Dommen, J.,

Zimmermann, R., Schnelle-Kreis, J., Salazar, G. A., Agrios, K., Szidat, S., El Haddad,

I., and Prévôt, A. S. H.: Fossil and non-fossil source contributions to atmospheric

carbonaceous aerosols during extreme spring grassland fires in Eastern Europe, Atmos.

Chem. Phys., 16, 5513-5529, doi:10.5194/acp-16-5513-2016, 2016.

Zhang, Y.-L., Huang, R.-J., El Haddad, I., Ho, K.-F., Cao, J.-J., Han, Y., Zotter, P.,

Bozzetti, C., Daellenbach, K. R., Canonaco, F., Slowik, J. G., Salazar, G.,

Schwikowski, M., Schnelle-Kreis, J., Abbaszade, G., Zimmermann, R., Baltensperger,

U., Prévôt, A. S. H., and Szidat, S.: Fossil vs. non-fossil sources of fine carbonaceous

aerosols in four Chinese cities during the extreme winter haze episode of 2013, Atmos.

Chem. Phys., 15, 1299-1312, doi:10.5194/acp-15-1299-2015, 2015.

Huang, R.-J., Zhang, Y., Bozzetti, C., Ho, K.-F., Cao, J., Han, Y., Dällenbach, K. R.,

Slowik, J. G., Platt, S. M., Canonaco, F., Zotter, P., Wolf, R., Pieber, S. M., Bruns, E.

A., Crippa, M., Ciarelli, G., Piazzalunga, A., Schwikowski, M., Abbaszade, G.,

Schnelle-Kreis, J., Zimmermann, R., An, Z., Szidat, S., Baltensperger, U., Haddad, I. E.,

and Prévôt, A. S. H.: High secondary aerosol contribution to particulate pollution during

haze events in China, Nature, 514, 218-222, doi:10.1038/nature13774, 2014.

Wang, Y. C., Huang, R.-J., Ni, H. Y., Chen, Y., Wang, Q. Y., Li, G. H., Tie, X. X.,

Shen, Z. X., Huang, Y., Liu, S. X., Dong, W. M., Xue, Fröhlich, R., Canonaco, F.,

Elser, M., Daellenbach, K. R., Bozzetti, C., El Haddad, I., A. S. H. Prévôt, Canagaratna,

M. R., Worsnop, D. R., and Cao, J. J.: Chemical composition, sources and secondary

processes of aerosols in Baoji city of northwest China, Atmos. Environ. 158, 128-137,

doi: 10.1016/j.atmosenv.2017.03.026, 2017.

Zotter, P., Ciobanu, V. G., Zhang, Y. L., El-Haddad, I., Macchia, M., Daellenbach, K.

R., Salazar, G. A., Huang, R.-J., Wacker, L., Hueglin, C., Piazzalunga, A., Fermo, P.,

Schwikowski, M., Baltensperger, U., Szidat, S., and Prévôt, A. S. H.: Radiocarbon

Curriculum vitae

206

analysis of elemental and organic carbon in Switzerland during winter-smog episodes

from 2008 to 2012 - Part 1: Source apportionment and spatial variability, Atmos. Chem.

Phys., 14, 13551-13570, doi:10.5194/acp-14-13551-2014, 2014.

optimized separation of oc and ec for radiocarbon … · 2018-11-06 · ix summary particulate...

Documents