Linking Switzerland’s PM 10 and PM 2.5 oxidative potential (OP) with emission sources

. Particulate matter (PM) is the air pollutant which causes the greatest deleterious heath effects across the world and PM is routinely monitored within air quality networks where PM mass according to its size, and sometimes number are reported. However, such measurements do not provide information on the biological toxicity of PM. Oxidative potential (OP) is a complementary metric which aims to classify PM in respect to its oxidising ability in lungs and is being increasingly reported 5 due to its assumed relevance concerning human health. Between June, 2018 and May, 2019, an intensive ﬁlter-based PM sampling campaign was conducted across Switzerland in ﬁve locations which involved the quantiﬁcation of a large number of PM constituents and OP for both PM 10 and PM 2.5 . OP was quantiﬁed by three assays: ascorbic acid (AA), dithiothreitol (DTT), and dichloroﬂuorescein (DCFH). OP v (OP by air volume) was found to be variable in time and space with Bern-Bollwerk, an urban-trafﬁc sampling site having the greatest levels of OP v among the Swiss sites (especially when considering 10 OP AAv ), with more rural locations such as Payerne experiencing lower OP v . However, urban-background and suburban sites did experience signiﬁcant OP v enhancement, as did the rural Magadino-Cadenazzo site during wintertime because of high levels of wood smoke. The mean OP ranges for the sampling period were: 0.4–4.1, 0.6–3.0, and 0.3–0.7 nmolmin − 1 m − 3 for the OP AAv , OP DTTv , and OP DCFHv respectively. A source allocation method using positive matrix factorisation (PMF) models indicated that although all PM 10 and PM 2.5 sources which were identiﬁed contributed to OP v on average, the anthropogenic 15 road trafﬁc and wood combustion sources had the greatest OP m potency (OP per PM mass). A dimensionality reduction procedure coupled to multiple linear regression modelling consistently identiﬁed a handful of metals usually associated with non-exhaust emissions, namely: copper, zinc, iron, tin, antimony and somewhat manganese and cadmium as well as three speciﬁc wood burning-sourced organic tracers – levoglucosan, mannosan, and galactosan (or their metal substitutes: rubidium and potassium) were the most important PM components to explain and predict OP v . The combination of a metal and a wood 20 burning speciﬁc tracer led to the best performing linear models to explain OP v . Interestingly, within the non-exhaust and wood combustion emission groups, the exact choice of component was not critical, the models simply required a variable to be present to represent the emission source or process. The modelling process also showed that OP AAv may be a more speciﬁc metric for OP than the other assays employed in this work. This analysis strongly suggests that the anthropogenic and locally emitted extract. The consumption of DTT ( nmolmin − 1 ) was determined by following the TNB absorbance at 412 nm wavelength at 10 min intervals for a total of 30 min of analysis time. the ﬁrst is that PM coarse is biologically relevant, and the second is the importance continuing PM 10 monitoring 220

what PM emission sources and components are most likely responsible for elevated OP (OP v and OP m ). The implications of 90 Switzerland's OP v patterns and the identification of PM sources and constituents will be discussed in respect to PM and OP v management.

Sampling sites
Daily PM filter samples were taken at five sampling sites across Switzerland (Table 1; Figure 1) between June, 2018 and May, 95 2019. The five monitoring sites used for the PM sampling are included in Switzerland's national air quality monitoring network; NABEL (Federal Office for the Environment, 2021). These established sites are used for compliance or regulatory monitoring and have long-term time series available for most common pollutants (Bundesamt für Umwelt, 2021). The sampling sites are located in different environments, ranging from rural, to urban-traffic surrounds. One site, Magadino-Cadenazzo, is located south of the Alps while the other four are located on the Swiss Plateau.
100 Table 1. Basic information for the five monitoring sites in Switzerland which were used for oxidative potential PM measurements.

Data
High-volume PM 10 and PM 2.5 quartz filter (Pallflex Tissuquartz 2500QAT-UP) samples were collected using Digitel DA-80H samplers with flow rates of 30 m 3 h −1 . Daily sampling ran continuously from midnight and midnight for a 12-month period between June 1, 2018 and May 31, 2019. However, for the quantification of constituents beyond simple mass, punches from every fourth-days' filters were taken and analysed. Because the sites form part of the NABEL network, routine flow checks 105 and various tests were regularly conducted in accordance to standard operating procedures.
In total, 908 filters were analysed with three OP assays. Eight-hundred and ninety-nine valid samples were reported, the missing samples were due to sampling or laboratory issues. Additional filter punches were used for a collection of other laboratory analyses to quantify other PM constituents such as elemental components (with inductively coupled plasma atomic emission spectrometry (ICP-AES) and inductively coupled plasma mass spectroscopy (ICP-MS)), ions (ion chromatography 110 (IC)), elemental and organic carbon (thermal optical transmission (TOT) EN16909 method using the EUSAAR2 temperature protocol (European Committee for Standardization (CEN), 2017)), and a collection of additional organics (high-performance liquid chromatographic method followed by pulsed amperometric detection (HPLC-PAD)). The details of these additional methods have been reported previously by Hüglin and Grange (2021); Grange et al. (2021) and the latter publication can be considered companion to this paper. 115 2.3 Oxidative potential assays OP was analysed with three different assays: ascorbic acid (AA), dithiothreitol (DTT), and dichlorofluorescein (DCFH). These analyses were conducted at the Institute of Environmental Geosciences, University of Grenoble Alpes, Grenoble, France. The three different protocols are described in detail in Kelly and Mudway (2003); Cho et al. (2005); Calas et al. (2018Calas et al. ( , 2019. All extracts were conducted at iso-and low-mass PM concentration (25 µg L −1 ) to prevent a non-linear measurement effect 120 (Charrier and Anastasio, 2012; Calas et al., 2018).
The consumption of DTT in the assay was inferred as a measure of the ability of the PM to transfer electrons from DTT to oxygen, thereby producing reactive oxygen species (ROS). The PM extracts were reacted with DTT, resulting in the consumption of DTT in the solution. The remaining DTT was then titrated with 5,5-dithiobis-(2-nitrobenzoic acid) (DTNB) to produce a yellow chromophore (5-mercapto-2-nitrobenzoic acid or TNB), which was in direct proportion to the amount of reduced 125 DTT remaining in solution after the reaction with the PM extract. The consumption of DTT (nmol min −1 ) was determined by following the TNB absorbance at 412 nm wavelength at 10 min intervals for a total of 30 min of analysis time. For all assays, the mixtures were injected into a 96-well plates and the absorbance was read from the microplate reader (TECAN spectrophotometer Infinite M200 pro). The well plates were shaken for 3 seconds before each measurement and kept at 37 • C. Three laboratory blanks (in Gamble+DPPC) and three positive controls (1,4-napthoquinone at 24.7 µmol L −1 ) were 140 included in each plate. The average values of these blanks were then subtracted from the sample measurements of the given plate. Detection limits (DL) were defined as three times the standard deviation of laboratory blank measurements. Uncertainties were estimated thanks to triplicate measurement of the same well.
The three assays have the same objective of determining the amount of oxidative stress an analyte can elicit, but the three assays have differing sensitivities to various components which form the PM mix and the specific antioxidants within the lung.

145
The three assays have these general characteristics: AA is primarily sensitive to transition metals (Janssen et al., 2014), DTT is the most reported OP assay and is sensitive to organics and to a lesser extent, metals (Janssen et al., 2014;Calas et al., 2019), and DCFH shows a preferential sensitivity to organic compounds. Therefore, the three assays give different perspectives on similar biological processes.

150
OP can be represented in two forms: OP per PM mass (OP m ), or OP per volume of air (OP v ). OP per volume of air is a superior unit when representing population exposure and therefore, this unit is mostly used in this analysis. There are three OP assays reported and to differentiate these assays, a superscript notation is used, i.e., OP AA v , OP DTT v , and OP DCFH v using the nmol min −1 m −3 units.

155
Source apportionment for PM 10 and PM 2.5 for the five monitoring sites was conducted with the positive matrix factorisation (PMF) receptor model and the multilinear engine (ME-2) algorithm (Paatero and Tapper, 1994;Paatero, 1999). The PMF approach employed was consistent with the SOURCES programme and is informally known as "extended PMF" (Favez et al., 2017;. The EPA PMF 5.0 software tool was used to apply PMF. The specific settings, constraints, and and output data are also available in a persistent data repository for others' convenience (Grange, 2021a).
The PMF analysis for the particular dataset was challenging because of the existence of fewer than the recommended samples available (91 compared to the recommended at least 100 (Norris et al., 2014)), low signal to noise ratios for many variables because of low ambient concentrations, and the inclusion of extra organic species into the PMF models. Despite the many validation steps conducted, the models had a number of limitations which are discussed fully in the companion paper Grange 165 et al. (2021) and were considered in the current work.

Linking PM sources to OP
The OP measurements were not included in the PMF modelling process, but these observations required linking to the PMFidentified sources. To estimate the source contributions to the three OP assays, weighted robust multiple linear regression (MLR) with an iterative M -estimator was used. Conceptually, the OP observations were explained by the PMF-identified 170 sources, and because linear regression models return coefficients in the dependent variable's response scale, the estimates of the PMF-identified sources for OP are readily interpreted by investigating the models' slope coefficients (β). To allow for evaluation of the models' coefficients uncertainty, the data were bootstrapped 500 times and modelled. Additionally, the analytical uncertainly was included in the models as weights. The MASS R package was used as the interface to the robust linear regression function (Venables and Ripley, 2002). An example of how this process was conducted can be found in a 175 public repository (Grange, 2021c).

OP modelling
The filter-based measurement campaign resulted in a large number of elements, ions, and organics to be quantified which compose Switzerland's PM. To extract the constituents which were the most important for OP, a multiple step process was conducted to firstly identify the most important constituents which explain OP and secondly, what combination of these con-180 stituents resulted in the best statistical models which explained OP values in Switzerland.
The identification of the most important PM constituents to explain OP was conducted with random forest, an ensemble decision tree machine learning algorithm (Breiman, 2001;Wright and Ziegler, 2017). The entire set of variables available were used to model OP. The random forests' importances for the included variables were extracted and analysed to reduce the feature space (Abdulhammed et al., 2019;Reddy et al., 2020). The variables which were consistently identified as the most 185 important for explanation of OP AA v and OP DTT v by random forest were used in further linear modelling work. Therefore, this dimensionality reduction pre-processing step allowed the dataset to be reduced from over 50 variables to the most important ≈15 for two OP assays.
The most important variables identified by random forest were used to model OP with robust multiple linear regression (Venables and Ripley, 2002). Individual models using all combinations of the ≈15 variables with a maximum of five predictors 190 were created to explain OP v . The intercept term was excluded from the model formulation and over 100 000 models were calculated. An example of how this was achieved is accessible via a public repository (Grange, 2021b). To identify models which were suitable for further use, three filters were applied to the models. Models with a maximum pairwise variance inflation factor (VIF) for independent variables greater than 2.5 were removed because this suggests multicollinearity among the independent variables (Jackson et al., 2009). Models which contained negative term estimates were also dropped, as were 195 models with R 2 values less than 75 %. These filters resulted in 371 models to be kept for further analysis and the majority (77 %) of these models had two independent variables.  (Figure 2; Table 2). For OP AA v , the PM 10 means ranged from 0.7 and 4.1 nmol min −1 m −3 and for PM 2.5 , the corresponding range was 0.4 and 1.6 nmol min −1 m −3 . OP DTT v means ranged from 0.8-3.0 and 0.6-1.1 nmol min −1 m −3 for PM 10 and PM 2.5 respectively. OP DCFH v did not show the same progressive increase across the rural to 205 urban roadside gradient with another rural site, Magadino-Cadenazzo having the highest means (0.7 nmol min −1 m −3 for both PM 10 and PM 2.5 ) while the other four sites were inconsistently ranked for the different PM size fractions and considering the different types of averages ( Table 2). The rural-urban-roadside gradient observed for OP AA v and OP DTT v was also demonstrated by PM mass and most other individual constituents (the exception was secondary components such as nitrate, sulfate, and ammonium) which form the Swiss PM mix, and this has been reported previously in a companion paper (Grange et al., 2021).  Winter and autumn had the highest average OP which is consistent with the common winter situation where primary atmospheric pollutants emissions are higher, and the atmospheric state is less conducive to pollutant transportation and dispersion (Beyrich, 1997;Emeis and Schäfer, 2006). The wintertime OP enhancement was especially clear at Magadino-Cadenazzo, a site known to be heavily burdened by wood smoke during the winter months (Grange et al., 2020). Notably, at Magadino-Cadenazzo, wintertime PM 2.5 OP v was enhanced to nearly the same extent as PM 10 because of wood burning sourced PM 215 being almost all contained in the fine-mode (Kleeman et al., 1999). Bern-Bollwerk was clearly the most polluted site in respect to OP v where the two AA and DTT assays remained elevated for all seasons, but mean OP DTT v was significantly lower during the summer than the other seasons. Another key observation from these aggregations was that PM coarse contained much of the OP v signal and this is only able to be highlighted because the sampling design included both PM 10 and PM 2.5 . This point has two implications, the first is that PM coarse is biologically relevant, and the second is the importance continuing PM 10 monitoring 220 or sampling because of the aforementioned relevance on human health. Figure 2 and Table 2 shows large differences among the three assays used to quantify OP v in this work. The DCFH assay based on a fluorescence method showed much lower levels of structure when compared to the other two, more established AA and DTT assays where the means ranged between 0.3 and 0.7 nmol min −1 m −3 ( Figure A1). The DCFH assay has a lower level of sensitivity when compared to the AA and DTT assays, nevertheless the sensitivity of DCFH to organic-rich PM 225 was observed at the southern Magadino-Cadenazzo sampling location where OP v enhancement was clear during the winter because of high concentrations of wood burning emissions. The AA assay is primarily sensitive to metals (Janssen et al., 2014), and the Bern-Bollwerk site which is known to experience significant non-exhaust emissions from road traffic, observed high levels of OP AA v during the sampling period and the mean was 4.1 nmol min −1 m −3 . Less severe enhancements were also observed for the urban and suburban Zürich-Kaserne and Basel-Binningen sites with the AA assay suggesting some 230 metal contamination of these atmospheres too. These observations are consistent with work exploring the urban and roadside increments in Switzerland, and the importance of non-exhaust emissions to these increments (Grange et al., 2021).

OP comparison with other locations
The comparison of OP metrics among different locations and sampling durations is problematic due to the lack of standardised OP laboratory procedures (Calas et al., 2019). Here however, comparisons can be made with many French sites where OP v 235 has been quantified by the same laboratory with identical analytical approaches. The OP v of PM 2.5 has been rarely reported in Europe, and therefore, only PM 10 's OP v will be discussed here. Bern-Bollwerk also demonstrated high levels of OP DTT v when compared to the other sampling locations, but for this metric, Chamonix was more polluted than Bern-Bollwerk with means of 4.4 and 2.9 nmol min −1 m −3 respectively (

Linking OP to PM sources
The PMF source apportionment analysis identified eight PM sources in Switzerland: sulfate-rich, nitrate-rich, road traffic, wood 255 combustion, primary biogenic, secondary biogenic, mineral dust, and aged sea salt. All the eight sources were detected for PM 10 while the primary biogenic, mineral dust, and aged sea salt sources were not identified in the PM 2.5 fraction indicating that these sources were mostly in the coarse-mode. Full discussion of the PMF results, the limitations, and the sources' characteristics can be found in the companion paper, Grange et al. (2021), however, an outline of the PMF results is briefly given below.
The PMF results indicated that about 50 % of the PM 10 and PM 2.5 load in Switzerland was from the three secondary nitrate-260 rich, sulfate-rich, and aged sea salt sources. Based on the models' factor/source profiles, the former two sources contained a significant amount of organic mass. Generally, the primary and secondary biogenic sources were rather low contributors to average mass concentrations, but they were highly seasonal sources and the secondary biogenic source was more important for PM 2.5 than PM 10 . The wood burning, mineral dust, and road traffic sources were more enhanced in urban areas, but their enhancement was highly dependent on the sites' immediate environmental surrounds. Bern-Bollwerk's road traffic source 265 contributed more than a third to both PM 10 and PM 2.5 , while the wood burning source contributed over 20 % to both PM fractions at Magadino-Cadenazzo, despite also being a source which was inactive for about half of the sampling period.
To investigate the relationship between the activities of the identified main PM sources in Switzerland on its OP, the PMF sources were used in conjunction with the OP v observations. OP v was explained using MLR models for each of the five sites with the identified PMF source contributions as independent variables (in µg m −3 ). The units of the estimated model 270 coefficients for the PM sources were then in nmol min −1 µg −1 and interpreted as the intrinsic OP m . This process has been called an 'inversion' by others Borlaza et al., 2021) and was conducted 500 times with bootstrapped inputs for each site, assay, and PM size fraction to allow for robust estimates of the models' terms. The lack of structure in the DFCH observations ( Figure A1) resulted in poorly performing models and therefore, this assay was not included in further analyses.
When the explanatory multiple linear regression models were exposed to the PMF-identified sources it was clear that the 275 anthropogenic road traffic and wood combustion sources had the greatest intrinsic OP m (Figure 4). When combining the five sites' results together, the road traffic and wood combustion sources were always the highest ranked OP m sources, with the exception of DTT for PM 2.5 where wood combustion was ranked first, but road traffic fell to fourth place and the nitraterich source was placed second. The metal-sensitive AA assay showed that the coarse-mode road traffic source was the most potent PM source in Switzerland giving additional evidence that coarse, non-exhaust emissions drove this assay's OP m results.

280
The mostly fine and carbonaceous wood combustion source was always important for the two OP m assays and was clearly the most potent source for PM 2.5 . The other remaining six sources had, on average, positive contributions to OP m , but were far less important for OP m when compared to the road traffic and wood combustion sources based on this analysis. Notably, the nitrate-and sulfate-rich sources generally showed low levels of OP m which outlines a disconnect between average PM mass concentrations and OP m potency. This suggests that all PM has the ability to contribute to OP, but road traffic and 285 wood combustion source are the two sources which should be prioritised for control and management to efficiently reduce OP v in Switzerland. Unlike Samake et al. (2017); Weber et al. (2021), these results do not suggest that biogenic-sourced PM is particularly important for OP m in Switzerland, perhaps due to different fungal and plant species found in different environments or the differing intensities of agriculture and cultivation between the two countries (Samaké et al., 2019;Samake et al., 2017).

Identifying important PM constituents with random forest 290
Within the PMF-identified sources shown in Figure 4, there are a large number of constituents which give the sources their characteristics. To better identify the specific components which compose the PM sources identified in Switzerland that were important and potent drivers of OP v , presumably, mostly contained within the road traffic and wood combustion sources, a multi-step modelling process was conducted. The random forest algorithm was used to calculate importance and were ranked for all variables included in the data set (Breiman, 2001). A high importance ranking indicates that the variable is more has gained traction in many fields (Behnamian et al., 2019). The motivation for this process was to simplify and resolve the lower-level linkages between PM components and OP v when the PMF sources were potentially heterogeneous among the sampling sites and PM size fractions.
When the random forest importances were calculated for each site, PM fraction, and OP v assay, it was clear that a collection 300 of organics and metals were commonly identified as being the most important variables for the explanation of OP v . Elements and organic compounds associated with wood combustion: rubidium, potassium, levoglucosan, mannosan and galactosan were constantly ranked highly in terms of importance ( Figure 5). The other group of components which were identified were metals such as copper, zinc, iron, tin, antimony, and to some extent manganese and cadmium. These collection of metals are usually associated with vehicular non-exhaust emissions and are generated by abrasive or wear processes Harrison 305 et al., 2021). EC and OC were also commonly identified and these variables are associated with both wood burning and vehicle exhaust emissions. Despite these two groups of PM constituents being identified, both mass and ions (especially nitrate) were also present in the most important variables identified by random forest. We interpret the presence of these variables as proxies of total PM mass indicating that although for a given PM mass, OP v may vary depending on its make-up, total PM mass is still an important, and related metric. Therefore, the importance analysis was consistent with the PMF inversion process discussed 310 in Section 3.2.
Figure 5 also shows some site specific variation due to the sites' different local emissions. For example, in Bern-Bollwerk, the non-exhaust sourced metals such as copper, iron, and zinc were ranked higher than the mean importance rank across the five sites. This feature was present in both assays and was somewhat clearer in PM 10 due to the tendency of abrasive processes to emit PM larger than 2.5 µm (Harrison et al., 2021). Magadino-Cadenazzo on the other hand, demonstrated a tendency of 315 rubidium, potassium, mannosan, and levoglucosan to be more important than the sites' mean ranking which was consistent with what is known about this site's exposure to local emissions because it experiences a heavy wood smoke load (Sandradewi et al., 2008;Chen et al., 2021). When comparing the two PM size fractions, there was no clear dominating source and the differences between PM 10 and PM 2.5 were overshadowed by site specific differences. This supports the conclusions made in a companion paper (Grange et al., 2021) where non-exhaust PM 2.5 emissions were found to be considerable and are important 320 to consider across the Swiss sampling sites. When exposing the PMF sources (eight for PM 10 and five for PM 2.5 ) to the same random forest importance analysis, the road traffic and wood combustion sources were clearly the most important sources for OP explanation, as shown in Figure 4.
A slightly different representation of the random forest importance rankings are provided in Figure 6, where the presence of variables in the group which was considered highly important were counted for the five sites, two OP v assays, and two PM 325 fractions. It is noticeable that rubidium and copper, two tracers for wood burning and non-exhaust emissions, were ranked as the most important variables for PM 10 at all five sites and for both OP v assays. For PM 2.5 , where concentrations of many metals were lower than in PM 10 , only a wood burning tracer (either potassium or rubidium) together with PM mass were identified across all five sites and both OP v assays. All variables which were identified more than once for each OP v assay and PM size fraction (the variables shown in the y-axes of Figure 6) were used in the next step of linear modelling to identify what variables 330 are best to be used when forming predictive models to explain OP v .   Figure 6. Counts of how many times an independent variable was ranked highly in terms of random forest importance for two OPv assays, two particulate size fractions, and five sampling sites. Variables with counts of five shows that for every site included in the analysis, this variable was identified as important for the explanation of OPv.

Modelling OP
The most important variables at each site, identified by the rank of the random forest importance (Figure 6), were used to build multiple linear regression models to explain OP v . Every combination of the variables were used to calculate linear regression models (with a maximum of five independent variables and the intercept terms omitted) and after training, only the models 335 with positive slope estimates, those which had a maximum pairwise variance inflation factor (VIF) of less-than 2.5, and had an R 2 greater-than 75 % were kept. These three filters ensured the models selected did not suffer from undesirable levels of multicollinearity among their independent variables (Jackson et al., 2009;Barmpadimos et al., 2011) and performed adequately on their training set. The VIF filter removed all models with more than four independent variables due to the increased probability of multicollinearity when including additional independent variables in the same model. A total of 100 342 models were 340 trained and 371 models passed the filters. The number of models trained for each site, PM fraction, and OP assay are shown in Table A2.
When analysing the models with the best performance based on their R 2 values, 77 % had two independent variables while models with one or three independent variables only composed 13 % and 10 % of the total set. Almost without exception, the best models' independent variables included a metal and an organic compound. The metals contained in the models were 345 the same as those identified and discussed previously ( Figure 5; Figure 6) and are generally emitted from abrasive processes related to road vehicles (iron, zinc, copper, antimony, but also cadmium), while the organics were the specific biomass burning markers of levoglucosan, mannosan, and galactosan. Table 3 shows equations of the best performing models based on their R 2 values for each sampling site, the two PM size fractions and the two OP v . However, all models fulfilling the applied filter criteria can be considered as appropriate and considered as suitable models for explaining the observed OP v . The full list of 350 these suitable models are provided in the Supplementary Information (Table S1), the counts of all pairwise combinations of variables in the suitable models with two or more independent variables are shown in Figure 7.
The best performing models demonstrated that the combination of vehicular non-exhaust emission and wood burning tracers were required to generate the best models to explain OP v . Interestingly, the exact tracers or markers used for the modelling was not critical. For example, using antimony, copper, or iron as the representative non-exhaust emission species resulted in 355 models which performed very similarly and showed that these three metals were effectively interchangeable with one-another.
Cadmium, manganese, and zinc could also be added to this group, but the use of these metals resulted in models which performed slight worse on average and such patterns may be related to the differing elements' analytical detection limits or the multiple emission sources these metals have. The same phenomenon was present for the wood burning tracers of levoglucosan, mannosan, and galactosan where the selection of one of these organics over the other was not critical for the explanation of   to note that although rubidium and potassium had higher ranks in the random forest importance, the suitable models for explaining mostly included an organic tracer for wood burning emissions (levoglucosan, mannosan, or galactosan). This could be explained by rubidium and potassium having multiple emission sources and therefore were removed by the multicollinearity filter used for the model selection.
We interpret the presence of levoglucosan, mannosan, and galactosan in this analysis as simply indicators of biomass burning 370 emission sources. This is because these particular organic compounds are not redox-active and therefore, they cannot be the components of PM which drove OP. Quinones, rubidium, and/or other co-emitted products from biomass burning are most likely the responsible components, and this is a clear example of how an observational study can suggest and highlight associations or correlations, but not necessarily causality.
In contrast to OP AA v , PM mass or ammonium and nitrate were present in the better performing models for OP DTT v at times.

375
It is unlikely that ammonium and nitrate are indeed strong drivers of OP v since ammonium sulfate and nitrate ((NH 4 ) 2 SO 4 and NH 4 NO 3 ) and have been shown to have negligible OP , the presence of these inorganic ions might   be acting as a proxy for total ambient PM concentrations or perhaps seasonal emission cycles due to its shift between gas and aerosol phases in the different seasons because of changes in ambient air temperature. For PM 2.5 's OP DTT v , OC as well as pinic acid (a tracer for biogenic secondary organic aerosol) were frequently found in the 371 models which passed the model 380 selection criteria. OC and pinic acid might also be understood as proxies for total PM concentrations or specific conditions leading to elevated PM levels. Such mentioned proxies were in the models for explaining OP DTT v mostly combined with an organic wood burning emission tracer and for PM 2.5 also with copper and tin.
The combinations of pairs of independent variables in suitable models for explaining OP v in PM 10 and PM 2.5 as shown in Figure 7 indicates that the OP AA v assay provided a response that was more specific to the chemical composition of PM than 385 the OP DTT v assay. It is also noticeable that for both OP v assays there are more pairwise combinations of independent variables in the suitable models for PM 2.5 than for PM 10 . The reason for this observation is currently unclear and further research will be required to fully elucidate these features.

Conclusions
An intensive PM and OP sampling campaign conducted across Switzerland between 2018 and 2019 demonstrated that OP v was 390 variable in time and space. OP v patterns followed the familiar pattern of atmospheric pollutants where urban locations were more polluted than their rural counterparts and wintertime saw enhanced OP v . Although the differences between rural and urban locations were important for mass, the OP metrics constantly showed a greater difference indicating OP was more heterogeneous than PM mass across Switzerland. When comparing Switzerland's OP v with 14 sites in France where data exists and were produced by the same sampling and laboratory procedures, Switzerland's OP v was comparable to that observed in France, but

395
Bern-Bollwerk, a semi-canyonised urban-traffic sampling location had the highest mean OP AA v (4.1 nmol min −1 m −3 ) contained in the dataset. The lack of current standardisation for OP measurement, quantification, and calibration is an issue which the air quality community should address and would allow for reliable comparisons among different locations and times in the future. The AA and DTT assays showed much more structure than the third DCFH assay which made the former approaches more useful for data analysis than the latter.

400
An analysis of Switzerland's PM 10 and PM 2.5 sources identified by PMF models suggested that two major anthropogenic emission sources, namely road traffic and wood combustion were the most important drivers of OP v in Switzerland. Contrasting this was the inorganic nitrate-and sulfate-rich sources which generally had low levels of intrinsic OP m across Switzerland, as did the two biogenic sources (primary and secondary). This outlines the potential disconnect between total PM mass concentration and OP m which has been noted by others, for example, Daellenbach et al. (2020) and this observation may update 405 the management priorities of PM sources with a focus on health impacts rather than total mass.
Further investigation into the components of PM using a random forest dimensionality reduction technique and multiple linear regression models demonstrated that a collection of metals associated with non-exhaust emissions such as copper, zinc, antimony, iron, tin, manganese, and cadmium as well as the specific wood combustion tracers of levoglucosan, mannosan, and galactosan (or associated elements such as rubidium and potassium) were consistently important for the explanation of OP v .

410
The combination of a non-exhaust sourced metal and a biomass burning tracer provided very good models which could explain OP v well when considering their training sets. The observations also suggested that OP AA v was a more specific or sensitive OP metric than the other assays employed, in the Swiss locations where sampling took place. To consider OP AA as a future, standard OP metric, additional evidence of associated health outcomes will be required, however.
The results above point towards the need to control wood burning sourced PM and non-exhaust emissions to reduce the 415 OP v of Switzerland's atmospheres. Such conclusions are not out of step with current air quality management practices and priorities, but reinforce the importance of these sources and their respective chemistry in respect to OP v -potentially a health relevant metric for PM. Therefore, a renewed focus on wood burning and non-exhaust emissions is encouraged to reduce the deleterious heath effects of PM. Because non-exhaust emissions and wood burning emissions can be effectively controlled at a local level, it is likely that significant reductions of OP v could be achieved without the need for regional and trans-boundary 420 management collaboration. OP v DTT J u l 2 0 1 8 O c t 2 0 1 8 J a n 2 0 1 9 A p r 2 0 1 9 J u l 2 0 1 8 O c t 2 0 1 8 J a n 2 0 1 9 A p r 2 0 1 9 J u l 2 0 1 8 O c t 2 0 1 8 J a n 2 0 1 9 A p r 2 0 1 9 J u l 2 0 1 8 O c t 2 0 1 8 J a n 2 0 1 9 A p r 2 0 1 9 J u l 2 0 1 8 O c t 2 0 1 8 J a n 2 0 1 9 A p r 2