Simulation of the cooking organic aerosol concentration variability in an urban area

Highlights

• Developed COA emissions with temporal, spatial, and volatility distribution.

• Framework for treating cooking OA in CTMs using the VBS.

• COA up to 15 μg m⁻³ during the night in area with high restaurant density.

• During dinner time cooking can be responsible 40–50% of the PM_2.5

• Low secondary COA production for meat charbroiling.

Abstract

Cooking activities are a source of urban air pollution that has been often ignored. The corresponding particulate matter (PM) emissions are not included or are seriously underestimated in urban emission inventories. However, several field studies suggest that cooking organic aerosol (COA) can be an important component of the organic PM in urban areas. In this study we propose and evaluate a methodology for the simulation of the COA concentration and its variability in space and time in an urban area. The city of Patras, the third biggest in Greece, is used for this first application during a typical late summer period.

The spatial distribution of COA emissions was based on the exact location of restaurants and grills, while the emission rates on the per person meat consumption in Greece. We estimated COA emissions of 0.6 g d⁻¹ per person that corresponds to 150 kg d⁻¹ for Patras. The temporal distribution of COA emissions was based on the known cooking times and the results of past field studies in the area. We estimated that half of the daily COA is emitted during dinner time (21:00–0:00 LT), while approximately 25% during lunch time (13:00–16:00 LT). The COA is simulated using the Volatility Basis Set approach and is treated as semivolatile and reactive. Its volatility distribution is based on laboratory measurements of COA from meat grilling.

The chemical transport model PMCAMx predicts that the COA concentration reaches values up to 15 μg m⁻³ during the nights of the simulated summertime period in an area with high restaurant density. The average predicted COA concentration is around 1.2 μg m⁻³ in the center and 0.1–0.2 μg m⁻³ in the suburbs of the city. COA has a distinct daily profile that peaks during lunch (13:00–15:00 LT) and dinner (21:00–23:00 LT) periods. The local production of secondary COA is predicted to be slow and it represents just a few percent of the total COA. The model reproduces well the average PM_2.5 concentrations at the outskirts of Patras and its overall performance for hourly PM_2.5 values is rated as good. A 50% uncertainty of the reported COA emissions is estimated based on the results of the sensitivity tests and the observations in the city core. The emissions can be further constrained and the corresponding uncertainty can be further reduced with measurements focusing on COA and not on total PM_2.5 as was the case in this study. These should take place in different areas of the city, selected based on the model predictions, to also improve our estimates of the spatial distribution of the emissions.

Introduction

Exposure to PM_2.5 (particles with a diameter lower than 2.5 μm) can lead to serious respiratory and cardiovascular problems (Du et al., 2016; Xing et al., 2016). Organic aerosol (OA) often represents more than 50% of PM_2.5 in urban areas (Kanakidou et al., 2005) and is a major component of atmospheric pollution (Zhang et al., 2007; Jimenez et al., 2009). OA can be emitted directly by a source or can be formed in-situ by condensation of low volatility products of the oxidation of organic vapors (Seinfeld and Pandis, 2006).

Cooking operations can be an important fine PM source for urban areas. In Los Angeles in the 1980s, approximately 20% of the primary organic aerosol was due to meat cooking (Hildemann et al., 1991). Cooking emissions contributed 19% to the total OA in Manchester during winter (Allan et al., 2010) and 17% in Barcelona (Mohr et al., 2012). In New York City, COA accounted for 16% of the total OA during summertime (Sun et al., 2011). In London, COA represented 22 and 30% of the total OA for two winter periods with different temperatures (Allan et al., 2010). Similar behavior was observed in Paris with cooking contributing 18 and 20% to the total OA during winter and summer time respectively (Crippa et al., 2013a, 2013b). In Athens, the contribution of COA to the total OA during summer and winter 2012 was 17 and 16% respectively (Kostenidou et al., 2015; Florou et al., 2017). In China, the contribution of cooking emissions is similar to that in European countries. During the 2008 Beijing Olympic Games 24% of the total OA was due to cooking (Huang et al., 2010). In Lanzhou, northwest China, during summertime COA contributes 24% to the total OA (Xu et al., 2014).

Meat charbroiling is one of the most important COA sources in many cities (Hildemann et al., 1991; Kostenidou et al., 2015). Hildemann et al. (1991) measured aerosol emissions up to 40 g kg⁻¹ of meat cooked, during charbroiling of regular meat (with a fat percentage of 21%). During frying of regular meat, the measured aerosol emissions were a lot lower at 1 g kg⁻¹ of meat cooked. Kostenidou et al. (2015) have shown that in Greek cities COA emissions are mainly due to the charbroiling of meat. The average COA contribution to the total OA in Patras was 10–15% depending on the season (Kostenidou et al., 2015; Florou et al., 2017). COA concentrations in Patras peaked between 14:00 and 15:00 LT and during the night (22:00–23:00 LT). Similar daily behavior has been observed in Paris, Athens, London, Manchester and Barcelona (Allan et al., 2010; Mohr et al., 2012; Crippa et al., 2013a; Kostenidou et al., 2015) which indicates the importance of cooking emissions during lunch and dinner periods.

COA can partially evaporate in the atmosphere after its emission and can also react with O₃, OH, etc. Significant changes in COA composition during photooxidation in combination with small changes in mass have been reported by Kaltsonoudis et al. (2017). Heterogeneous reactions seem to be the most probable reason for these changes. Louvaris et al. (2017) determined the volatility distribution of COA during meat charbroiling combining thermodenuder and isothermal dilution measurements.

There have been relatively few modeling efforts of the spatial distribution of COA in urban areas mainly due to the lack of suitable emission inventories. Fountoukis et al. (2016) applied the PMCAMx model over Paris assuming COA emissions of 5.3 tons d⁻¹ for summer and 5.1 tons d⁻¹ for winter. These emissions were estimated based on the measured concentrations. They also assumed that 50% of the COA was emitted during lunch time (12:00–14:00 LT), while 20% during dinner time (20:00–22:00 LT). The emissions were distributed in space based on the population density. Ots et al. (2016) applied the EMEP4UK model (a regional version of the EMEP MSC-W model) over Britain for a year assuming emissions of 7.4 Gg COA. These cooking emissions were based on measurements of COA at two sites in London during the same period. The temporal variation of these emissions was based on measurements at a hot spot and peaked during lunch (12:00–14:00 LT) and dinner times (18:00–21:00 LT). The spatial distribution was also based on population density.

Despite the progress in understanding COA in urban areas a number of issues are yet to be addressed. The estimation of emissions and their distribution in space and time are still quite uncertain. It is not clear if there are COA hot spots in urban areas and how high the corresponding concentration levels are. Also, the evaluation of the few modeling efforts has been limited due to the sparse observations which may be far from the concentration hot spots.

In this work we estimate the COA emissions in a major Greek city (Patras) based on the individual locations of each restaurant and the corresponding emission factors from meat grilling. A high-resolution version of PMCAMx is applied over Patras in an effort to simulate the spatial and temporal distribution of COA in this area. COA is assumed to be semi-volatile and to also react forming secondary COA. The predictions of the model are evaluated against measurements from a low-cost PM_2.5 sensor network in the city including measurements in the area with the highest restaurant density.