Influence of semi- and intermediate-volatile organic compounds (S/IVOC) parameterizations, volatility distributions and aging schemes on organic aerosol modelling in winter conditions

Highlights

• Latest experimental studies about emissions of organic matter implemented in CAMx.

• Remarkable improvement on modelled organic aerosol levels.

• S/IVOC emission revisions appear to be the key factor for such improvement.

• Enabling aging processes for biomass burning SOA enhances the performance of the model.

• VBS provides a better reconstruction of POA and SOA relative contribution to the total.

Abstract

This study presents a high–resolution (5 km) set of new simulations performed with CAMx v6.40 over the Po Valley area (Northern Italy), aimed to enhance organic aerosol (OA) levels prediction and to gain insight into the sensitivity of CAMx to different uncertain features of the input setup. In particular, we mainly investigated the role of (i) volatility distributions of organic emissions, (ii) parametrizations of semi– and intermediate–volatile organic compounds (S/IVOC) emissions and (iii) different aging schemes, by exploiting the latest experimental information available in the recent scientific literature. Model results were validated against two OA–specific datasets, available for both an urban site (Bologna, February 2013) and a rural one (Ispra, March 2013).

We found out a remarkable performance enhancement of modelled OA levels when applying revisions in S/IVOC emission parametrizations together with the new volatility distributions, at both the validation sites. This performance enhancement is associated with a very significant improvement in secondary organic aerosol (SOA) prediction, mainly due to revised IVOC emissions. At Bologna urban site, mean fractional bias (MFB) of OA ranged from −80.1% in the worst run to −10.1% in the best one and index of agreement (IOA) from 0.52 to 0.75. Notable improvements but overall poorer metrics were found for Ispra site, where MFB ranges from −84.2% to −35% and IOA from 0.45 to 0.50. These findings indicate that organic matter in the semi– and intermediate–volatile range are most likely underestimated in official emission inventories for each main source category (i.e. biomass burning, diesel and gasoline vehicles exhaust).

Finally, model results did not show a very pronounced sensitivity to aging processes, due to the low photochemical activity typically observed during winter–time. However, we give evidence that enabling aging processes for biomass burning related SOA, which is by default disabled in CAMx v6.40, can help in closing the gap between modelled and observed SOA concentrations.

Introduction

Atmospheric pollution from particulate matter (PM) represents one of the major environmental and social concern for human health and it poses several challenges in terms of management and mitigation of harmful impacts. According to the latest European Environment Agency report (EEA, 2017), approximately 53% of the EU-28 population was exposed to PM concentrations exceeding the WHO Air Quality Guidance value for PM₁₀ (WHO, 2006) in 2015. Premature deaths resulting from such exposure are estimated to be around 400 000 in the EU-28 countries. Nevertheless, the trend of mean PM concentration in Europe is rather flat during the most recent years (Guerreiro et al., 2014; Barmpadimos et al., 2012). Development of cost– effective mitigation policies depends heavily upon reliable air quality models results (Harrison et al., 2008) which can give insights about the impact of a given control strategy on PM concentrations.

A relevant fraction of submicron particulate matter is given by organic aerosol (OA), which accounts for 20–90% of total PM_2.5 (Zhang et al., 2007). However, the large complexity of OA chemical composition, with thousands of organic chemical species found in the ambient aerosol (Goldstein and Galbally, 2007), as well as the complex atmospheric processing of organic compounds strongly limited scientific progress in the OA modelling area (Hallquist et al., 2009; Fuzzi et al., 2015). Within the atmospheric modelling community, there is mounting evidence that – despite an overall good agreement in gaseous pollutants – OA mass is in most applications underestimated mainly because of the not well reproduced secondary (SOA) fraction (Meroni et al., 2017; Ciarelli et al., 2016; Woody et al., 2016; Zhang et al., 2013; Bergström et al., 2012; Hodzic et al., 2010).

The traditional scheme for OA modelling in Chemical Transport Models (CTMs) is based on the so called “Two-product approach” by Odum et al. (1996). This approach considers primary organic aerosol (POA) that is directly emitted from various combustion sources (e.g. vehicles exhaust, biomass burning) as a non–volatile species that does not chemically evolve. SOA is formed from the early generation oxidation of gaseous organic volatile (VOC) precursors, which produces two nonreactive semi-volatile products that are partitioned between gas and aerosol phases depending on temperature and OA mass concentration. However, recent experimental studies highlighted that this approach presents two main limitations. First, Robinson et al. (2007) suggested that POA species should be treated as semi-volatile compounds that can evaporate from the particulate phase, react in the gas-phase and repartition as SOA, as pointed out also in other works (Jimenez et al., 2009; Grieshop et al., 2009). In the conceptual model of Robinson et al. (2007), POA emission is associated with semi–volatile (SVOC) and intermediate–volatile (IVOC) compounds emissions. SVOC compounds are characterized by a relatively low volatility (effective saturation concentration C* between 10⁻¹ and 10³ μg m⁻³) and are in the substantial partitioning with the particulate phase whereas IVOC compounds (C* between 10³ and 10⁶ μg m⁻³) are highly volatile and they partition preferentially to the gas-phase in atmospheric conditions. The second main issue of Odum et al. (1996) approach is related to the further oxidation of SOA in the atmosphere (i.e., the so-called aging process), which is traditionally neglected as the products of VOC oxidation were considered non–reactive. These two limitations led to the development of a new framework for the description of all OA components and their reactions. This new framework – in literature referred to as VBS (Volatility Basis Set) – rethinks the distinction between the traditional primary and secondary OA by grouping organic species into surrogates according to their volatility and degree of oxidation, thus providing a more realistic picture of the behavior of atmospheric organic aerosol. Details about theoretical aspects of VBS framework are provided in Donahue et al. (2006); Donahue et al. (2011); Donahue et al. (2012b).

Several applications of the VBS scheme to CTMs in different case studies can be found in the recent scientific literature (Fountoukis et al., 2011; Tsimpidi et al., 2011; Bergström et al., 2012; Zhang et al., 2013; Fountoukis et al., 2014; Koo et al., 2014; Ciarelli et al., 2016; Woody et al., 2016; Fountoukis et al., 2016; Meroni et al., 2017). The general conclusion stemming from these works is that the VBS scheme enhances the prediction of both OA levels and degree of oxidation, although the high number of parameters to be constrained in the VBS scheme causes a large uncertainty in the models results. For instance, all the studies cited above scaled the IVOC emissions, which are traditionally neglected in official emission inventories (Ots et al., 2016; Hodzic et al., 2010), on POA emissions using a factor between 1.5$\times$ and 3$\times$, as suggested by Robinson et al. (2007). However, this assumption historically derives from chassis dynamometer tailpipe measurements performed two decades ago on two diesel vehicles (Schauer et al., 1999) and – whilst it might hold true for vehicles exhaust emissions (Kim et al., 2016) – it is likely to be incorrect for other emissions sources (e.g. biomass burning, Ciarelli et al., 2017b). Very recent experimental works presented more detailed and source-specific parametrizations for IVOC emissions, which might be implemented in CTMs to provide more accurate results. As an example, Jathar et al. (2014) performed smog chamber experiments to investigate SOA formation from gasoline vehicles, diesel vehicles and biomass burning, and they reported that unspeciated organics – which are not appropriately included in current emission inventories and, in turn, chemical transport models – account for 10–20% of total non-methane organic gases (NMOG). Zhao et al. (2015) and Zhao et al. (2016) characterized emissions of IVOC from on-road and off-road diesel and gasoline vehicles during dynamometer testing, respectively, reporting both new volatility distributions of the organics emissions and new parametrizations for IVOC emissions calculation. Ciarelli et al. (2017b) performed novel smog chamber experiments for wood combustion emissions, and their result suggest an average ratio of non–traditional VOCs (i.e. IVOC) to POA emissions of 4.75, much higher compared to the widely adopted 1.5, which however was based on diesel vehicles measurements.

Recent European modelling studies attempted to integrate these new parametrizations into CTMs. Ciarelli et al. (2017a) constrained a modified VBS scheme to treat biomass burning OA and evaluated the implementation of this scheme in CAMx. Ots et al. (2016) and Sartelet et al. (2018) investigated different parametrizations for traffic-related S/IVOC emissions for the UK and the greater Paris area, respectively. Chrit et al. (2018) addressed both biomass burning and traffic-related S/IVOC emission parametrizations, volatility distributions and aging by performing a set of sensitivity simulations over western Mediterranean region during winter time. The overall outcome of these works is that updating S/IVOC emission parametrizations and volatility distributions helps in closing the gap between observed and predicted OA concentrations.

Here, we present a new set of sensitivity simulations with CAMx that, differently from previous studies, aims to evaluate model performances in conditions where high OA levels are measured. Following the most recent European studies, we investigate the impact of volatility distributions of organics emissions, S/IVOC emission parametrizations, SOA yields from gaseous precursors and different aging schemes, by implementing the latest experimental information available in the scientific literature. The study area is the Po Valley (Northern Italy) during wintertime (February–March 2013), which is a well-known hotspot where PM levels remain problematic despite the air quality remediation plans intended to get in compliance with current EU air quality standards, mainly because of adverse meteorological conditions (Caserini et al., 2017; Perrino et al., 2014; Pernigotti et al., 2012; Ferrero et al., 2011). We evaluate our model results against two OA–specific datasets, available for both an urban site (Bologna, February 2013) and a rural one (Ispra, March 2013). These two datasets are derived from Positive Matrix Factorization (PMF) analysis of Aerosol Mass Spectrometer (AMS) and Aerosol Chemical Speciation Monitor (ACSM) measurements (DeCarlo et al., 2006; Ng et al., 2011), which allow a thorough comparison of each fraction of organic aerosol (i.e. primary and secondary). We also point out how the development of different meteorological condition can influence the overall model performance as well as, more specifically, the reconstruction of the organic fraction.