Atmospheric Environment 74 (2013) 393e401
Contents lists available at SciVerse ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Investigating the impacts of anthropogenic and biogenic VOC emissions and elevated temperatures during the 2003 ozone episode in the UK Jonathan Strong a, J. Duncan Whyatt a, *, Sarah E. Metcalfe b, Richard G. Derwent c, C. Nicholas Hewitt a a b c
Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YB, UK School of Geography, Nottingham University, Nottingham NG7 2RD, UK rdscientific, Newbury RG14 6LH, UK
h i g h l i g h t s ELMO-v2 models ozone levels during the 2003 heatwave at Writtle, UK. Effects of temperature on chemistry and emissions (AVOCs, BVOCs) considered. Ratio of ozone attributed to AVOC and BVOC emissions is roughly 3:1. Source attribution reveals varied geographic origins of ozone. Future temperature increases may reduce benefits of precursor emission reductions.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 August 2012 Received in revised form 2 April 2013 Accepted 4 April 2013
The Lagrangian tropospheric ozone model ELMO-v2 (Edinburgh Lancaster Model for Ozone) is applied to the intense 2003 ozone episode in SE England. When using model parameters representative of typical episodes, ELMO-v2 was found to underestimate ozone levels substantially during peak ozone days, but, by increasing three parameters (temperature, biogenic and anthropogenic VOC emission rates) to levels close to those observed, good agreement between modelled and observed ozone was achieved. Using attribution techniques possible with a Lagrangian model, the episode was divided into five phases with each exhibiting different geographical origins for ozone precursor emissions. Anthropogenic VOCs, primarily of European (non-UK) origin, made the biggest contribution to modelled ozone levels. European biogenic VOC emissions significantly contributed to ozone levels on some days, whereas the contribution from UK biogenic VOC emissions was comparatively small throughout. The VOC:NOx ratio was also shown to change during the episode, with high ozone days being less VOC-sensitive. The implications of both variable NOx/VOC sensitivity and the possibility of more frequent heatwaves due to climate change need to be taken into account in planning effective future emissions reductions to control ground-level ozone in the UK. Ó 2013 Elsevier Ltd. All rights reserved.
Keywords: ELMO Lagrangian Temperature Source attribution VOC:NOx ratio
1. Introduction During the first two weeks of August 2003 most of Europe, including parts of the UK, experienced an extreme heatwave. This was caused by a large blocking anticyclone that persisted over central Europe leading to cloudless skies (hence uninterrupted sunshine), high temperatures and light winds (Black et al., 2004;
* Corresponding author. Tel.: þ44 (0)1524 510239; fax: þ44 (0)1524 510269. E-mail address:
[email protected] (J.D. Whyatt). 1352-2310/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.atmosenv.2013.04.006
Burt, 2004). The heatwave was accompanied by an intense episode of ground-level ozone pollution with ozone levels peaking on the hottest days and frequently exceeding 90 ppb1 (180 mg m3) across Europe (EEA, 2003). The highest ozone levels (>180 ppb), recorded in southern France, exceeded the EC ‘alert threshold’ (120 ppb) whilst ozone levels in SE England reached the UK Government’s ‘high’ pollution band (90e180 ppb) on 10 days during August 2003
1
1 ppb is the mixing ratio of 1 part in 10þ9.
394
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
(Kent, 2003). This episode was exceptional in its spatial extent, duration and exceedances of the EC ‘information threshold’ (>90 ppb), and occurred despite significant reductions in ozone precursor emissions across Europe since 1990. Such heatwaves and ozone episodes are of concern as they impact upon human health (Fischer et al., 2004; Stedman, 2004) and vegetation (including crops) (Fowler et al., 2008). Given the impacts of ozone and the costs of high resolution measurements, air quality models are commonly used to explore our understanding of ozone formation and to help to develop effective emissions reduction policies. A wide range of models can be used, each with different strengths and weaknesses and with different levels of complexity required to be ‘fit for purpose’ (Derwent et al., 2009). For ozone models, the treatment of emissions, chemical mechanisms, atmospheric dispersion and the use of meteorological data (amongst others) need to be considered (Monks et al., 2007). Tropospheric ozone chemistry is well described elsewhere (e.g. Monks, 2003; Fowler et al., 2008). Net production rates of ozone vary with VOC:NOx concentration ratios giving either VOC- or NOxsensitive conditions (Sillman, 1999). Temperature influences ozone chemistry directly through reaction rates and indirectly through changes in emission rates of anthropogenic VOCs (AVOCs) and biogenic VOCs (BVOCs) and through impacts on dry deposition. Rubin et al. (2006) have shown that AVOC emissions from motor vehicles increase by 6.5% 2.5% per 1 C rise. Plants are also known to react to heat stress and soil moisture deficits by increasing BVOC emissions. For example, emissions of isoprene (which has a high photochemical ozone creation potential (Derwent et al., 1998)) are positively related to temperature, although these become suppressed at temperatures over 35 C (Pacifico et al., 2009). Here we consider the performance of a simple Lagrangian model, ELMO-v2 in relation to peak ozone values during the August 2003 episode as recorded in SE England. In light of the known controls on ozone chemistry, we consider the model’s sensitivity to AVOC and BVOC emissions and to temperature, and use its Lagrangian formulation to explore the source attribution of the ozone produced during different phases of the episode. 1.1. Studies of the August 2003 episode at Writtle, SE England The August 2003 episode was recorded in detail by the NERCfunded TRopospheric Organic CHemistry campaign (TORCH) which ran from July 25 to August 31 2003, at Writtle, in SE England (Lee et al., 2006). Episodic ground-level ozone concentrations (daily maxima >50 ppb) were observed between August 02 and August 13 and coincided with maximum daily temperatures above 25 C. Between August 06 and 11 the daily ozone maxima exceeded 90 ppb and reached 140 ppb on August 06. Maximum daily temperatures exceeded 30 C on four of these days. The August 2003 episode was preceded by a prolonged drought, thus soil moisture content was low (Marsh, 2004). Dry deposition to vegetation, which is normally a major sink for ozone, but is inversely related to temperature, would have been restricted (Emberson et al., 2001). Moreover, by August in the UK, many crops would have been harvested, thus decreasing the vegetation canopy and further reducing the potential for dry deposition. The extreme 2003 heatwave conditions were, therefore, especially favourable for episodic ozone formation (Solberg et al., 2005; Pellegrini et al., 2007). The TORCH campaign also recorded atmospheric conditions and surface level concentrations of many atmospheric species including selected anthropogenic and biogenic VOCs. With exceptions, AVOC concentrations recorded during the heatwave period of the TORCH campaign were typically between 1.5 and 2.0 times higher than during the rest of the campaign (Lee et al., 2006). Peak heatwave day-time mixing ratios (>1 ppb) of isoprene observed at Writtle
were over twice those typical of non-heat-wave conditions (Lee et al., 2006). Corresponding increases in NOx concentrations were not significant; hence increases in VOC emissions during the heatwave will have altered the VOC:NOx ratios and potentially ozone production rates (see Section 3.5). The extreme conditions at Writtle have made it a focus for modelling activity and three studies have focused on ozone specifically. The UK Photochemical Trajectory Model (PTM) reproduced observed levels reasonably well, but underestimated peak levels, especially on high ozone days, and overestimated night-time levels (Utembe et al., 2005). The PTM was able to reproduce the concentrations of most VOCs, but some (including isoprene) were under-predicted, with the shortfall in ozone levels and VOCs attributed to the underestimation of heatwave temperatures. The Eulerian CMAQ model was used to reproduce the episode over SE England, with an emphasis on Writtle (Francis et al., 2011). The UKspecific Eulerian model EMEP4UK, based on the EMEP Unified Model (Simpson et al., 2012), was used to model ozone levels and its sensitivity to increases in temperature, AVOC and BVOC emissions (Vieno et al., 2010). In common with CMAQ, EMEP4UK used meteorology supplied by the Weather Research Forecast (WRF) model (www.wrf-model.org). The differences between the approach to ozone modelling with EMEP4UK and that adopted here are discussed further in Section 4. 2. Methodology ELMO is a simple single-layer Lagrangian trajectory model that was originally developed by Metcalfe et al. (2002) to provide policy makers with insights into the effectiveness of emission reductions across the UK and the rest of Europe on peak ozone levels in the UK through scenario modelling and source attribution. The model provides a simplified description of ozone formation in air parcels as they travel for up to 96 h from origins across Europe to receptors in the UK. Each air parcel is 1200 m in height, representative of the typical boundary layer depth on a hot summer’s day, and its contents are instantaneously mixed within each time step (120 s). The model incorporates a full description of the fast photochemical reactions which drive ozone production in the polluted atmospheric boundary layer over NW Europe (see Metcalfe et al., 2002 for detail) and includes a condensed form of the Master Chemical Mechanism (Jenkin et al., 1997) to degrade 11 emitted hydrocarbon species that are considered representative of the hundreds of hydrocarbons that contribute to ozone formation in the real atmosphere. The chosen hydrocarbons, summarised in Table 1, reflect a mix of high, medium and low reactivity species and their combined emission has been scaled up to reflect the overall emission from the total inventory (Passant, 2002) This approach of
Table 1 Allocation of unspeciated anthropogenic NMVOC emissions into 11 ‘representative’ species used in ELMO-v2. Their names, chemical formulae, allocation by mass within ELMO-v2 and Photochemical Ozone Creation Potential (POCP) are also described. Species
Formula
Proportion by mass
POCP
Methanal Ethane Ethanal n-Butane Ethene Propene o-Xylene Propane Methanol Acetone Toluene
HCHO C2H6 CH3CHO C4H10 C2H4 C3H6 (CH3)2(C6H4) C3H8 CH3OH (CH3)2CO (CH3)(C6H5)
4.1% 8.9% 0.5% 29.3% 21.5% 6.2% 13.1% 1.9% 0.1% 2.4% 12.0%
45 4 55 31 100 117 78 14 13 6 44
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
lumping NMVOC emissions is the same as that adopted in the Unified EMEP model which underpins all EU air quality policy (Simpson et al., 2012). Unlike the EMEP and EMEP4UK models, however, isoprene emissions in ELMO-v2 do not vary with diurnal temperature variations. The version of ELMO used in this study (ELMO-v2) has previously been applied to model peak spring and summer ozone levels in the UK using HYSPLIT back-trajectories derived from NCEP meteorological data (Draxler and Rolph, 2010), with arrival heights of 500 m and a residual boundary layer (Strong et al., 2010). Trajectory temperatures are adjusted to equivalent ground-level temperatures using an adiabatic lapse rate of 1 C 100 m-1, assuming that under summer heatwave conditions, with strong convection, the near-surface temperature is representative of the temperature through the very well-mixed boundary layer. A monthly-resolved pan-European biogenic emissions inventory for 2004 (Karl et al., 2009) was processed to provide UK and European isoprene and undifferentiated monoterpene (treated as a-pinene within ELMO-v2) BVOC inventories for the month of August. The Karl et al. inventory uses contemporary (2004) ECMWF (http:// www.ecmwf.int/) meteorology to calculate emission rates and hence estimates BVOC emissions typical of non-heatwave conditions. Annually-resolved anthropogenic NOx and NMVOC emissions inventories for 2004 were taken from the National Atmospheric Emissions Inventory (NAEI) and EMEP for the UK and continental Europe at 10 10 km and 50 50 km resolution respectively. Boundary conditions for NO, NO2, CO, CH4 and H2 were chosen to represent the 2004 European background concentrations (http:// tarantula.nilu.no/projects/ccc/network/index.html; Derwent et al. (2007); AGAGE, http://agage.eas.gatech.edu/). These inputs form the basis of all runs described in this study. ELMO-v2 was run to Writtle from August 01 to August 12 (inclusive) at three hourly intervals using parameterisations summarised in Table 2. Run 1, or the ‘base’ run, used standard emission rates and HYSPLIT trajectory ground-level adjusted temperatures. Run 2 estimated the potential loss of ozone through dry deposition. Runs 3e8 investigated the model’s sensitivity to increased temperature, BVOC and AVOC emission rates respectively. Run 9, the ‘combined’ run, increased the HYSPLIT temperatures by a uniform 6 C, and the BVOC and AVOC emission rates by factors of 2.0 and 1.5 respectively as these are considered to be more representative of heatwave conditions than the standard model values. Finally, runs 10e13 were used to estimate the attribution of modelled ozone at Writtle to four source sectors: BVOC and AVOC emissions differentiated by the UK and the rest of Europe (Table 3). Source sector contributions were determined by running the model in its ‘combined’ configuration with emissions from each sector suppressed in turn, then calculating the difference from the ‘combined’ run.
395
Table 3 ELMO-v2 modelled contribution to ozone levels from EU and UK BVOC and AVOC emissions. Comparisons are made using data from the whole daily cycle at 3-hourly intervals and for afternoon periods (12:00e18:00 h) only. Run Contribution Parameters used Temp BVOC 10
UK BVOC
11
UK AVOC
12
EU BVOC
13
EU AVOC
þ6 C UK EU þ6 C UK EU þ6 C UK EU þ6 C UK EU
Whole day AVOC
0.0 2.0 2.0 2.0 2.0 0.0 2.0 2.0
UK EU UK EU UK EU UK EU
Afternoon only
Max Avg Max (ppb) (ppb) (ppb) 1.5 31 1.5 0.0 133 1.5 1.5 58 1.5 1.5 94 0.0
Avg (ppb)
5
31
6
24
133
29
9
58
9
19
84
19
3. Results The performance of each run is summarised in Table 2 where modelled ozone levels for all trajectories arriving at Writtle are assessed using average, mean bias error (MBE), Pearson’s correlation (r) and Index of Agreement (IoA) (Legates and McCabe, 1999) metrics. Model output is also compared against observations made at 12:00, 15:00 and 18:00 since ozone levels are highest in the afternoons. Fig. 1 compares the modelled ozone levels from the base run (1) with the corresponding observations. The model does not provide a satisfactory simulation of the episode as it underestimates most afternoon-observed levels (MBE ¼ 30.2 ppb) and over-estimates most night-time observed levels. The underestimation of the diurnal cycle is a consequence of the simplicity of our model and is a limitation of our simple model type. The overestimation of night-time levels is however common with many other more complex models (e.g. CMAQ (Sokhi et al., 2006; Yu et al., 2008), Chimere (Vautard et al., 2005), the PTM (Utembe et al., 2005) and the Oslo CTM 2 (Solberg et al., 2008)). Although ELMO-v2 has previously been used to reproduce ozone levels under typical summer conditions using the equivalent of the ‘base’ run (Strong et al., 2010) it was evident that model parameters needed to be altered to reflect the very unusual meteorological and chemical conditions experienced in August 2003. Run 2 (implementing dry deposition using a simple deposition velocity approach into the ‘base’ run) predictably resulted in greater model under-estimation (MBE ¼ 42.0 ppb); ozone levels are reduced by an average of 11 ppb throughout the episode and by >20 ppb on high ozone days. Francis et al. (2011) estimate that dry deposition could reduce hourly ozone levels by up to 25 ppb on peak ozone days at Writtle, whilst Solberg et al. (2008) showed that dry deposition has the potential to reduce ozone levels by up to 50 ppb during peak episode conditions. Miao et al. (2006) have
Table 2 Comparison of observed and ELMO-v2 modelled ozone levels for the August 2003 episode at Writtle under different parameterisations of temperature, BVOC and AVOC emissions. Comparisons are made using data from the whole daily cycle at 3-hourly intervals and for afternoon periods (12:00e18:00 h) only. Run
Description
Parameters used Temp
1 2 3 4 5 6 7 8 9
Observed Base With DD Temperature BVOC 2.0 BVOC 3.0 BVOC 4.0 AVOC 1.5 AVOC 2.0 Combined
þ0 þ0 þ6 þ0 þ0 þ0 þ0 þ0 þ6
C C C C C C C C C
Whole day
Afternoon only
BVOC
AVOC
Max (ppb)
Avg (ppb)
MBE (ppb)
Correl (r)
IoA
Max (ppb)
Avg (ppb)
MBE (ppb)
Correl (r)
IoA
1.0 1.0 1.0 2.0 3.0 4.0 1.0 1.0 2.0
1.0 1.0 1.0 1.0 1.0 1.0 1.5 2.0 1.5
142 103 86 102 108 113 118 117 145 127
48 37 25 43 41 46 51 54 69 63
11.8 23.2 5.8 6.9 2.1 2.5 5.4 20.2 14.7
0.52 0.50 0.57 0.53 0.53 0.54 0.63 0.67 0.65
0.68 0.60 0.74 0.71 0.73 0.74 0.79 0.75 0.77
142 103 86 102 108 113 118 117 145 127
76 46 34 52 51 56 61 65 82 75
30.2 42.0 23.6 25.3 20.3 15.4 11.0 5.9 0.90
0.56 0.45 0.63 0.60 0.63 0.66 0.78 0.85 0.82
0.60 0.49 0.67 0.66 0.70 0.75 0.85 0.91 0.90
396
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
Fig. 1. Comparison of observed ozone levels at Writtle, UK (01e12 August 2003) and ELMO-v2 modelled values for the ‘base’ run. Data shown at 3-hourly intervals.
recorded low dry deposition rates to vegetation during anticyclonic conditions as plants reduce stomatal conductance in response to heat stress and soil moisture deficits (Emberson et al., 2000). As the temperature, soil moisture and vegetation conditions at Writtle were extreme during the 2003 episode, it was assumed that stomatal conductance was severely or totally limited in the vicinity of the Writtle site during the 2003 episode and hence the dry deposition velocity of ozone was set to zero in subsequent ELMO-v2 runs. 3.1. Sensitivity of ELMO-v2’s chemistry to temperature Temperature data used by ELMO-v2 (see above) are generally lower than observed daily maxima, especially on the hottest days of the episode (August 06, 09 and 10). In run 3 temperatures along each trajectory (see Section 2) were increased by 6 C since this was the maximum difference between temperatures modelled by HYSPLIT and observed at Writtle (on August 10). Modelled ozone levels increased by an average of 6 ppb (using both full-day and afternoon-only evaluation periods). The greatest increase (26 ppb) occurred on August 11 and accounted for 41% of modelled ozone. However, on the second hottest day (August 06), modelled ozone levels showed negligible response to the increase in temperature. The reasons for this discrepancy are not explored further here; however, the back-trajectories associated with these days are quite different (Fig. 2).
The model was then rerun, with emission rates of BVOCs across the UK and European domains increased by factors of 2, 3 and 4 (for both isoprene and monoterpene) in runs 4, 5 and 6, respectively. When BVOC emission rates were quadrupled, modelled isoprene mixing ratios exceeded 1.1 ppb on three days. Although this was similar to the observed maxima, our model was unable to reproduce the observed diurnal pattern of isoprene, presumably due to its treatment of isoprene emissions (see Section 2). Doubling BVOC emission rates increased the modelled average isoprene mixing ratio to 0.3 ppb which is close to the observed episode average recorded between 06:00 and 21:00 inclusive. UK and European BVOC emissions were therefore scaled by a factor of 2.0 in the combined run (see Section 3.4). Each step increase in BVOC emissions resulted in a w5 ppb increase in modelled ozone levels (Fig. 3b). This constant and linear increase in ozone production indicates that ozone production is limited by the availability of VOCs within the model domain during this episode.
3.2. Modelled ozone sensitivity to BVOC emissions Isoprene was the only BVOC species measured during the TORCH campaign and observed mixing ratios follow a clear diurnal cycle (Fig. 3a). The average mixing ratio over the episode was 0.2 ppb, but daily maxima of 0.8 and 1.3 ppb were recorded on the hottest days which coincided with high ozone levels. However, the reverse is not seen: on August 08, ozone levels exceeded 90 ppb but isoprene mixing ratios were low. Modelled isoprene mixing ratios in the ‘base’ run (also shown in Fig. 3a) average 0.1 ppb and peak at 0.3 ppb, but do not follow the diurnal cycle because the BVOC inventories are only resolved on a monthly basis, with no diurnal temperature response (see Section 2). Nevertheless, modelled mixing ratios are comparable with those observed during the summer in the UK, typically w0.2 ppb at 20 C, which is indicative of good model performance in relation to this short-lived species.
Fig. 2. 96-h back-trajectories to Writtle arriving at 15:00 h on 03, 06, 07, 09 and 11 August 2003. Each dot represents one hour travelled.
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
397
Fig. 3. Sensitivity of (a) ELMO-v2 modelled isoprene mixing ratios to increasing BVOC emissions (b) modelled ozone levels to increasing BVOC emissions and (c) modelled ozone levels to increasing AVOC emissions. Data shown at 3-hourly intervals.
3.3. Sensitivity of modelled ozone to AVOC emissions To account for the potential higher AVOC emission rates caused by the heatwave temperatures, AVOC inventory emission rates were scaled uniformly (source sectors, species and both UK and European AVOC inventory domains) by factors of 1.5 and 2.0 in runs 7 and 8 respectively. Fig. 3c shows the sensitivity of modelled ozone to increasing AVOC emissions. The maximum amount of ozone attributable to unscaled AVOC emissions (‘base’ run) varies from 11 ppb on August 02 to >40 ppb on August 06 and 09. When doubled (run 8), the model overestimated observed daily ozone maxima and the MBE value for the afternoon-only evaluation period became positive (MBE ¼ 5.9 ppb). AVOC emission rates were subsequently scaled by a factor of 1.5, compatible with the increase in evaporative AVOCs emission rates reported by Rubin et al. (2006) and the observed increased AVOC concentrations recorded at Writtle (Lee et al., 2006). Further justification comes from comparing
observed and modelled concentrations of a marker species, formaldehyde (methanal; HCHO), although caution should be applied as the differentiation of AVOC emissions within ELMO-v2 is to representative species only. When scaled by a factor of 1.5, the model produced an average HCHO concentration of 2.5 ppb which is comparable to the corresponding observed averaged concentration of 3.0 ppb (Lee et al., 2006). This scaling was therefore applied elsewhere (see Section 3.4). 3.4. Combination of AVOC, BVOC and temperature increases The metrics presented in Table 2 show that the separate increases to temperature and emission rates all improved ELMO-v2 performance. A ‘combined’ run (9) was therefore conducted with temperature increased by 6 C (see Section 3.1) and BVOC and AVOC emission rates increased by factors of 2.0 and 1.5 respectively (see Sections 3.2 and 3.3). These scalars were applied uniformly throughout the episode.
398
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
Modelled ozone levels in the ‘combined’ run are shown in Fig. 4b and Table 2 presents performance statistics. Visually, there is good agreement between observed and modelled daily maxima although ozone levels are overestimated by 36% on August 04. Night-time levels are generally overestimated, and this contributes to the high MBE of 14.7 ppb. For the entire episode, however, the IoA value of 0.77 represents improved performance over the ‘base’ run and most alternate parameterisations. The IoA is comparable to values found by Strong et al. (2010) when modelling spring and summer episodes in 1995 (0.71 and 0.85), and those reported by Sokhi et al. (2006) and Yu et al. (2008) when modelling ozone in SE England using CMAQ with annual emissions from EMEP and the NAEI processed into speciated NMVOCs with daily and hourly temporal profiles. When comparison between observed and modelled ozone is restricted to the afternoon period only, performance improves significantly (IoA ¼ 0.90 and MBE ¼ 0.90 ppb) suggesting afternoon levels are particularly well simulated by ELMO-v2. This confirms our previous interpretation of ELMO e that it is particularly suited to describing, analysing and predicting peak ozone levels during the summertime in the UK (Strong et al., 2010). The contributions made by each modification to the default parameterisation are displayed as additive bands above the ozone attributable to the ‘base’ model run in Fig. 4c (for a description of the phases, see Section 3.6). These are expressed as daily averages for clarity, so finer temporal detail is inevitably masked. The sum of the ozone contributions from each scalar is generally higher than the ‘combined’ run (not shown) and reflects the non-linearities of
tropospheric chemistry. The largest contribution comes from the scaling of AVOC emissions which accounts for an average increase of 17.3 ppb over the episode, whereas the average contributions from increasing temperatures and BVOC emissions are 6.2 and 5.0 ppb respectively. Parameter scaling makes the least impact on the lowest ozone days of the episode (August 01, 02 and 03). 3.5. Modelled VOC:NOx ratios during the Writtle episode The sensitivity of modelled ozone to scaled VOC emissions (see Sections 3.2 and 3.3.) suggests that most of the episode was VOCsensitive (i.e. as VOC emission rates were increased, model ozone increased and, by implication, ozone production rates increased). Episode VOC-sensitivity is supported by Vieno et al. (2010) and Solberg et al. (2005) whilst Metcalfe et al. (2002) and Strong et al. (2010) have previously demonstrated wider VOC-sensitivity across southern England. Fig. 4a shows VOC:NOx ratios for the episode based on concentrations associated with air parcels arriving at the Writtle receptor (as opposed to evolving along trajectories) for the ‘combined’ run. The ratios range from 0.7 to 5.3 and show how ozone production rates varied throughout the episode (assuming consistency between the trajectory-end VOC:NOx ratios and ozone production rates) and they correlate well against modelled ozone (r ¼ 0.68). Although the ‘base’ run conditions indicate VOC-sensitivity, there is evidence that the modelled VOC:NOx ratio (calculated from the ‘combined’ run) was changing from VOC-sensitivity (NOx-
Fig. 4. (a) VOC:NOx ratios and (b) ELMO-v2 modelled ozone levels for the ‘combined’ run. Data shown at 3-hourly intervals. The increase in modelled ozone resulting from the application of each scalar (temperature, BVOC, AVOC) is shown in (c). Data shown in 24-h intervals.
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
saturation) to NOx-sensitivity (low to high VOC:NOx ratios). On August 06, when the VOC:NOx ratio was at the episode maximum (5.3) (low VOC-sensitivity), the additional amount of ozone generated when AVOC emission rates were increased by 150% was 19 ppb above ‘base’ conditions, but when doubled (200%) the increase was only 4 ppb. Similar patterns of decreasing ozone production rates are seen on other high ozone days and all have high VOC:NOx ratios. This suggests that the VOC:NOx ratio had moved towards or even reached NOx-sensitivity. In contrast, on a low ozone day (August 03), the VOC:NOx ratio was low (0.7), and the amount of ozone produced when AVOC emissions were increased remained constant, suggesting high VOC-sensitivity. This episode was more severe in continental Europe and Vautard et al. (2005) have demonstrated that it would have been more effective to reduce NOx emissions than VOC emissions when ozone levels were high. This is consistent with our finding that high ozone days have reduced VOC-sensitivity (and increasing NOx-sensitivity). 3.6. Attribution of modelled ozone to VOC source sectors Four source sectors were assessed (UK and Europe differentiated by AVOC and BVOC emissions) to estimate their contributions to modelled ozone at Writtle (runs 10e13), relative to the contribution from background sources. Table 3 presents the maximum and average ozone levels attributable to each sector. Similar amounts of ozone were attributed to UK and European origins throughout the duration of the episode, although their relative contributions varied on different days. The ratio of ozone attributable to AVOC and BVOC emissions is about 3:1 (for both full-day and afternoon-only metrics). This attribution enables us to break the episode down into 5 distinct phases, each of which will be described in more detail below. Fig. 5 shows the daily maximum and percentage contribution to ozone formation from each source sector for five days considered representative of the five different phases of the episode.
399
Corresponding HYSPLIT back-trajectories arriving at Writtle at 15:00 h are shown in Fig. 2. Mean wind speeds along the individual trajectories were generally light (<6 m s1), particularly on days of arrival (w1e4 m s1). Stunder (1996) has previously noted that positional errors associated with trajectories tend to be greatest under slow-moving anti-cyclonic conditions; however, the impacts of positional errors on ozone formation are likely to be small in relation to larger spatial uncertainties associated with the underlying emissions inventories. Phase 1 of the episode (August 01e03), is characterised by most ozone being attributed to UK AVOC and BVOC emissions. Air parcels originating over the Atlantic Ocean were advected east over the UK to Writtle. The low VOC:NOx ratio (<2.0) associated with this trajectory on arrival at Writtle suggests it is highly VOC-sensitive and reflects the short time ozone precursors were active. Phase 2 (August 04e06) marks the transition from clean Atlantic Ocean air to mainly continental anticyclonic air-flows. The air parcel that arrived at Writtle on August 06 (when the episode’s maximum ozone level was observed) was slow moving (w3 m s1) and passed over industrial areas of the near Continent prior to arrival at Writtle, hence most of the modelled ozone is of European origin (Fig. 5). The VOC:NOx ratio on August 06 was high (2.0e5.0) suggesting low VOC-sensitivity with most of the excess NOx in the plume exhausted. Low VOC-sensitivity is typical of ‘ageing’ ozone precursor emissions within an air parcel. Phase 3 of the episode (August 07e08) is characterised by trajectories originating in SW Europe before crossing Wales and England. Solberg et al. (2005) suggested wild fires along the Iberian Peninsula may have contributed to elevated ozone levels across Northern Europe and modelled trajectories from the Lagrangian model FLEXPART (Stohl et al., 1998) and the representative HYSPLIT back-trajectory of August 07 both support this conjecture (Fig. 2). There is a significant contribution from background sources; however, most of the modelled ozone during this phase is attributed to UK sources of precursors (Fig. 5) with only a small
Fig. 5. Percentage contribution of different AVOC and BVOC source sectors to ELMO-v2 modelled ozone levels on selected days of the August 2003 episode at Writtle, UK. Pies are proportional to peak daily ozone levels (56e127 ppb).
400
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
proportion of ozone apportioned to the European BVOC sector. Neither ELMO-v2 nor the BVOC inventory provide facilities to incorporate episodic emissions events, so direct investigation into the contribution of wild fires is not possible. High ozone levels occurred during the fourth phase of the episode (August 09e10) and the model suggests that most of the ozone may be attributed to AVOCs from the UK. The air parcel (August 09) spent most of its time over England (Fig. 2). The representative trajectory (August 11) associated with the fifth phase of the episode (August 11e12) originated over the North Sea before moving south close to the Rhine Estuary, and then approached Writtle from the Thames Estuary (Fig. 2). European BVOC and AVOC emissions consequently accounted for the majority of the ozone modelled at Writtle (Fig. 5). Although the trajectory itself did not cross much of mainland Europe, the horizontal sampling of emissions by ELMO-v2 captured European AVOCs from industrial areas of the Netherlands, Belgium and France. The sum of ozone generated from each attribution scenario did not match modelled ozone generated from the ‘combined’ run because of the non-linearity of tropospheric photochemistry. Generally, emissions into the air parcel shortly before arrival at Writtle (mainly from the UK) lead to an overestimation of ozone production whilst emissions towards the start of each trajectory in Europe lead to an underestimation of ozone production. As a result, our attribution technique probably underestimates the contribution from European sources; this contribution may also have been enhanced by reduced dry deposition due to the very dry conditions. Nevertheless, the apportionment of ozone to the four VOC emission sectors corresponds with the back-trajectories and demonstrates the diverse origins of ozone at Writtle. 4. Discussion and conclusions ELMO-v2 has been used to model ozone levels during the intense ozone episode recorded at Writtle (which extended across the whole of SE England) in August 2003 (Lee et al., 2006). In common with other modelling studies of this episode, ELMO-v2 underestimated ozone observations using its ‘base’ configuration. Model inputs were then adjusted to incorporate increases in temperature, AVOC and BVOC emissions considered realistic of the heatwave conditions. A ‘combined’ run incorporating all scaled parameters resulted in good performance. We have identified that VOC:NOx ratios varied significantly during the Writtle episode, and this has implications for ozone control in extreme heatwave conditions. The episode was mainly VOC-sensitive especially on low ozone days, but on high ozone days VOC-sensitivity decreased and there is evidence that NOx-sensitivity was approached. Thus, most benefit would be obtained from controlling VOC emissions throughout the episode, although NOx emission controls would have to be considered in the most intense conditions. From this, we infer that different emission control strategies may be applicable to different parts of Europe (including the UK) and at different times during ozone episodes. Vieno et al. (2010, see Section 1.1) using the Eulerian EMEP4UK model, took a rather different approach to modelling the 2003 episode, in that they conducted a baseline run and then carried out a series of sensitivity experiments by changing a number of individual meteorological or chemical parameters in turn, although these were not put together in any combined runs. Corresponding outputs from both ELMO-v2 and EMEP4UK are reasonably consistent and this is encouraging as both models are of different types, use different chemistry schemes, meteorologies and BVOC emission inventories (also, EMEP4UK used isoprene only). Both models indicated that meteorological data often fail to reproduce the highest temperatures observed during major episodes which will
have implications for ozone modelling. The results from both models indicate that high temperatures may lead to high ozone levels, but that this is not always the case. Vieno et al. suggest that when high ozone levels are recorded in the UK, but temperatures are not unusually high, this might reflect ozone or its precursors being imported. Vieno et al. were limited in their ability to explore the effect of specific source regions and sectors on ozone, partly because of the Eulerian format of EMEP4UK, but also because they did not consider the effects of changes in non-UK AVOCs. One of the strengths of ELMO-v2 is that, within the limitations of the HYSPLIT data, individual trajectories can be traced and the importance of different precursor source regions identified. Hence, we have been able to model the 2003 episode on a day to day basis tracing the differences between them. Assessing the performance of ELMO-v2 and comparing this with other ozone models has highlighted both strengths and weaknesses in ELMO in particular and ozone models in general. The reliability of emissions inventories of both AVOCs and BVOCs is obviously central to all models. In addition to overall amounts, there is also the desirability of obtaining AVOC and BVOC speciation profiles that vary spatially and temporally to reflect changes in human activity (AVOCs) and temperature (both). The detailed impact of temperature on isoprene emissions is not currently captured in our model. Here we have applied uniform scalars to the available inventories to begin to explore the effects of temperature, but this could be undertaken in a much more rigorous way (see Simpson et al., 2012 for example). This study and that of Vieno et al. (2010) have demonstrated the need for meteorological inputs which are truly representative of heatwave conditions as these will have impacts on atmospheric chemistry, AVOC and BVOC emissions and the performance of plants as sinks of ozone. The significance of realistic meteorological inputs for heatwave conditions is particularly important given climate change projections (Beniston, 2004; IPCC, 2007). By the end of the 21st century, ‘2003-type’ heatwaves, possibly longer and more intense, could occur about every second summer across Europe (Schär et al., 2004; Fischer and Schär, 2010) and the UK (Hulme et al., 2002) resulting in unacceptably high ozone levels. This paper has shown that ELMO-v2 can replicate peak ozone levels recorded by the TORCH campaign during the intense August 2003 episode, especially when ground-level temperatures are adjusted to match observed values more closely and the general effects of temperature on emissions are taken into account. Variations in photochemistry are driven by diurnal changes in temperature and the amount of incoming solar radiation derived from the HYSPLIT trajectories, but do not account for diurnal changes in emissions. The ability of a relatively simple, but flexible model, which can be used to explore the impact of changes in parameters, emissions and attribution by source region, has been illustrated. The exercise has, however, highlighted knowledge deficiencies which need addressing, including the diurnal and spatial variations in AVOC emission speciation and the necessity of accounting for possible climate change into account when attempting to model future ozone. These measures will help give better understanding of ozone production and levels likely to be encountered in the UK later in the 21st century, and the control strategies which will be required to avoid the exceedance of air quality thresholds. Acknowledgements The authors would like to thank Gemma Davies (Lancaster Environment Centre) for assistance in the production of the figures and the anonymous reviewers for their constructive comments on the original manuscript.
J. Strong et al. / Atmospheric Environment 74 (2013) 393e401
References Beniston, M., 2004. The 2003 heat wave in Europe: a shape of things to come? An analysis based on Swiss climatological data and model simulations. Geophysical Research Letters 31, L02202. http://dx.doi.org/10.1029/2003GL018857. Black, E., Blackburn, M., Harrison, G., Hoskins, B., Methven, J., 2004. Factors contributing to the summer 2003 European heat wave. Weather 59 (8), 217e 223. http://dx.doi.org/10.1256/wea.74.04. Burt, S., 2004. The August 2003 heat wave in the United Kingdom: part 1 e maximum temperatures and historic precedents. Weather 59 (8), 199e208. http://dx.doi.org/10.1256/wea.10.04A. Derwent, R.G., Jenkin, M.E., Saunders, S.M., Pilling, M.J., 1998. Photochemical ozone creation potential for organic compounds in Northwest Europe calculated with a master chemical mechanism. Atmospheric Environment 32, 2429e2441. Derwent, R.G., Simmonds, P.G., Manning, A.J., Spain, T.G., 2007. Trends over a 20 year period from 1987 to 2007 in surface ozone at the atmospheric research station, Mace Head, Ireland. Atmospheric Environment 41, 9091e9098. http:// dx.doi.org/10.1016/j.atmosenv.2007.08.008. Derwent, R.G., Fraser, A., Abbott, J., Jenkin, M., Willis, P., Murrells, T., 2009. Evaluating the Performance of Air Quality Models. Report to DEFRA. Welsh Assembly Government, the Scottish Executive and the Department of the Environment for Northern Ireland. AEAT/ENV/R/2873/Issue 2. Draxler, R.R., Rolph, G.D., 2010. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) Model. http://www.ready.arl.noaa.gov/HYSPLIT.php. EEA (European Environment Agency), 2003. Air pollution by ozone in Europe in summer 2003. Available on-line at: http://www.eea.eu.int. Emberson, L.D., Wieser, G., Ashmore, M.R., 2000. Modelling of stomatal conductance and ozone flux of Norway spruce: comparison with field data. Environmental Pollution 109 (3), 393e402. Emberson, L.D., Ashmore, M.R., Simpson, D., Tuovinen, J.-P., Cambridge, H.M., 2001. Modelling and mapping ozone deposition in Europe. Water, Air and Soil Pollution 130, 577e582. Fischer, P.H., Brunekreff, B., Lebret, E., 2004. Air pollution deaths during the 2003 heatwave in the Netherlands. Atmospheric Environment 38, 1083e1085. http:// dx.doi.org/10.1016/j.atmosenv.2003.11.010. Fischer, E.M., Schär, C., 2010. Consistent geographical patterns of changes in highimpact European heatwaves. Nature Geoscience 3, 398e403. http:// dx.doi.org/10.1038/NGEO866. Fowler, D., et al., 2008. Ground-level Ozone in the 21st Century: Future Trends, Impacts and Policy Implications: Science Policy Report 15/08. The Royal Society, London. Available on-line at: http://royalsociety.org/policy/publications/2008/ ground-level-ozone. Francis, X.V., Chemel, C., Sokhi, R.S., Norton, E.G., Ricketts, H.M.A., Fisher, B.E.A., 2011. Mechanisms responsible for the build-up of ozone over South East England during the August 2003 heatwave. Atmospheric Environment 45, 6880e 6890. http://dx.doi.org/10.1016/j.atmosenv.2011.04.035. Hulme, M., Turnpenny, J., Jenkins, G., 2002. Climate Change Scenarios for the United Kingdom: the UKCIP02 Briefing Report. Tyndall Centre for Climate Change Research. University of East Anglia, Norwich. IPCC, 2007. Intergovernmental Panel on Climate Change: Fourth Assessment Report e Climate Change 2007. Jenkin, M.E., Saunders, S.M., Pilling, M.J., 1997. The tropospheric degradation of volatile organic compounds: a protocol for mechanism development. Atmospheric Environment 31, 81e104. Karl, M., Guenther, A., Koble, R., Leip, Seufort, G., 2009. A new European plantspecific emission inventory of biogenic volatile organic compounds for use in atmospheric transport models. Biogeosciences 6, 1059e1087. http://dx.doi.org/ 10.5194/bg-6-1059-2009. Kent, A., 2003. Air Pollution Forecasting: Ozone Pollution Episode Report (August 2003). Department for Environment, Food & Rural Affairs (DEFRA). Available on-line at: http://www.airquality.co.uk/archive/reports/list.php. Lee, J.D., Lewis, A.C., Monks, P.S., Jacob, M., Hamilton, J.F., Hopkins, J.R., Watson, N.M., Saxton, J.E., Ennis, C., Carpenter, L.J., Carslaw, N., Fleming, Z., Bandy, B.J., Oram, D.E., Penkett, S.A., Slemr, J., Norton, E., Rickard, A.R., Whalley, L.K., Heard, D.E., 2006. Ozone photochemistry and elevated isoprene during the UK heatwave of August 2003. Atmospheric Environment 40, 7598e 7613. http://dx.doi.org/10.1016/j.atmosenv.2006.06.057. Legates, D.R., McCabe, G.J., 1999. Evaluating the use of “goodness of fit” measures in hydrologic and hydroclimatic model validation. Water Resources Research 35 (1), 233e241. http://dx.doi.org/10.1029/1998WR900018. Marsh, T.J., 2004. The UK drought of 2003: a hydrological review. Weather 59 (8), 224e230. http://dx.doi.org/10.1256/wea.79.04.
401
Metcalfe, S.E., Whyatt, J.D., Derwent, R.G., O’Donoghue, M., 2002. The regional distribution of ozone across the British Isles and its response to control strategies. Atmospheric Environment 36, 4045e4055. Miao, J.-F., Chen, D., Wyser, K., 2006. Modelling subgrid scale dry deposition velocity of O3 over the Swedish west coast with MM5-PX model. Atmospheric Environment 40, 415e429. http://dx.doi.org/10.1016/j.atmosenv.2005.09.057. Monks, P.S., 2003. Tropospheric photochemistry. In: Hewitt, C.N., Jackson, A.V. (Eds.), Handbook of Atmospheric Science: Principles and Applications. Blackwell, Oxford, pp. 156e187. Monks, P.S., Blake, R.S., Borrell, P., 2007. Review of Tools for Modelling Tropospheric Ozone Formation and Assessing Impacts on Human Health and Ecosystems. DEFRA. AQ0706. Pacifico, F., Harrison, S.P., Jones, C.D., Sitch, S., 2009. Isoprene emissions and climate. Atmospheric Environment 43, 6121e6135. http://dx.doi.org/10.1016/ j.atmosenv.2009.09.002. Passant, N., 2002. Speciation of UK Emissions of Non-methane Volatile Organic Compounds. AEA Technology Report ENV-0545, Abingdon, Culham, United Kingdom. Pellegrini, E., Lorenzini, G., Nali, C., 2007. The 2003 European heatwave: which role for ozone? Some data from Tuscany, Italy. Water, Air and Soil Pollution 181 (1e 4), 401e408. http://dx.doi.org/10.1007/s11270-006-9310-z. Rubin, J.I., Kean, A.J., Harley, R.A., Millet, D.B., Goldstein, A.H., 2006. Temperature dependence of volatile organic compound evaporative emissions for motor vehicles. Journal of Geophysical Research 111, D03305. http://dx.doi.org/ 10.1029/2005/JD006458. Schär, P.L.V., Lüthi, D., Frei, C., Häberli, Liniger, M.A., Appenzeller, C., 2004. The role of increasing temperature variability in European summer heatwaves. Nature 427, 332e336. http://dx.doi.org/10.1038/nature02300. Sillman, S., 1999. The relation between ozone, NOx and hydrocarbons in urban and polluted rural environments. Atmospheric Environment 33, 1821e1845. Simpson, D., Benedictow, A., Berge, H., Bergstrom, R., Emberson, L.D., Fagerli, H., Flechard, C.R., Hayman, G.D., Gauss, M., Jonson, J.E., Jenkin, M.E., Nyiri, A., Richeter, C., Semeena, V., Tsyro, S., Tuovinen, J.-P., Veldebenito, A., Wind, P., 2012. The EMEP MSC-W chemical transport model e technical description. Atmospheric Chemistry and Physics 12, 7825e7865. http://dx.doi.org/10.5194/ acp-12-7825-2012. Sokhi, R.S., San José, R., Kitwiroon, N., Fragkou, E., Pérez, J.L., Middleton, D.R., 2006. Prediction of ozone levels in London using the MM5eCMAQ modelling system. Environmental Modelling & Software 21, 566e576. http://dx.doi.org/10.1016/ j.envsoft.2004.07.016. Solberg, S., Coddeville, P., Forster, C., Hov, O., Orsolini, Y., Uhse, K., 2005. European surfaced ozone in the extreme summer of 2003. Atmospheric Chemistry and Physics Discussions 5, 9003e9038. Solberg, S., Hov, Ø., Sovde, A., Isaksen, I.S.A., Coddeville, P., De Backer, H., Forster, C., Orsolini, Y., Uhse, K., 2008. European surface ozone in the extreme summer 2003. Journal of Geophysical Research 113, D07307. http://dx.doi.org/10.1029/ 2007JD009098. Stedman, J.R., 2004. The predicted number of air pollution related deaths in the UK during the August 2003 heatwave. Atmospheric Environment 38, 1087e1090. http://dx.doi.org/10.1016/j.atmosenv.2003.11.01. Stohl, A., Hittenberger, M., Wotawa, G., 1998. Validation of the Lagrangian particle dispersion model FLEXPART against large scale tracer experiment data. Atmospheric Environment 32, 4245e4264. Strong, J., Whyatt, J.D., Hewitt, C.N., Derwent, R.G., 2010. Development and application of a Lagrangian model to determine the origins of ozone episodes in the UK. Atmospheric Environment 44, 631e641. http://dx.doi.org/10.1016/ jatmosenv.2009.11.019. Stunder, B.J.B., 1996. An assessment of the quality of forecast trajectories. Journal of Applied Meteorology 35, 1319e1331. Utembe, S.R., Jenkin, M.E., Derwent, R.G., Lewis, A.C., Hopkins, J.R., Hamilton, J.F., 2005. Modelling the ambient distribution of organic compounds during the August 2003 ozone episode in the southern UK. Faraday Discussions 130, 311e 326. http://dx.doi.org/10.1039/B417403H. Vautard, R., Honore, C., Beekmann, M., Rouil, L., 2005. Simulation of ozone during the August 2003 heat-wave and emission control scenarios. Atmospheric Environment 39, 2957e2967. http://dx.doi.org/10.1016/j.atmosenv.2005.01.039. Vieno, M., Dore, A.J., Stevenson, D.S., Doherty, R., Heal, M.R., Reis, S., Hallsworth, L., Tarrason, P., Wind, P., 2010. Modelling surface ozone during the 2003 heat-wave in the UK. Atmospheric Chemistry and Physics 10, 7963e7978. http:// dx.doi.org/10.5194/acp-10-7963-2010. Yu, Y., Sokhi, R.S., Kitwiroon, N., Middleton, D.R., Fisher, B., 2008. Performance characteristics of MM5eSMOKEeCMAQ for a summer photochemical episode in southeast England, United Kingdom. Atmospheric Environment 42, 4870e 4883. http://dx.doi.org/10.1016/j.atmosenv.2008.02.051.