Climate

This study investigates the effect of ozone (O 3) deposition on ground level O 3 concentrations and subsequent human health and ecosystem risk under hot summer “heat wave” type meteorological events. Under such conditions, extended drought can effectively “turn off” the O 3 vegetation sink leading to a substantial increase in ground level O 3 concentrations. Two models that have been used for human health (the CMAQ chemical transport model) and ecosystem (the DO3SE O3 deposition model) risk assessment are combined to provide a powerful policy tool capable of novel integrated assessments of O 3 risk using methods endorsed by the UNECE Convention on Long-Range Transboundary Air Pollution. This study investigates 2006, a particularly hot and dry year during which a heat wave occurred over the summer across much of the UK and Europe. To understand the influence of variable O3 dry deposition three different simulations were investigated during June and July: (i) actual conditions in 2006, (ii) conditions that assume a perfect vegetation sink for O3 deposition and (iii) conditions that assume an extended drought period that reduces the vegetation sink to a minimum. The risks of O 3 to human health, assessed by estimating the number of days during which running 8 h mean O3 concentrations exceeded 100 μg m −3, show that on average across the UK, there is a difference of 16 days exceedance of the threshold between the perfect sink and drought conditions. These average results hide local variation with exceedances between these two scenarios reaching as high as 20 days in the East Midlands and eastern UK. Estimates of acute exposure effects show that O 3 removed from the atmosphere through dry deposition during the June and July period would have been responsible for approximately 460 premature deaths. Conversely, reduced O 3 dry deposition will decrease the amount of O 3 taken up by vegetation and, according to flux-based assess ents of vegetation damage, will lead to a reduction in the impact of O 3 on vegetation across the UK. The new CMAQ-DO 3SE model was evaluated by comparing observation vs. modelled estimates of various health related metrics with data from both urban and rural sites across the UK; although these comparisons showed reasonable agreement ther were some biases in the model predictions with attributable deaths at urban sites being over predicted by a small margin, the converse was true for rural sites. The study emphasises the importance of accurate estimates of O3 deposition both for human health and ecosystem risk assessments. Extended periods of drought and heat wave type conditions are likely to occur with more frequency in coming decades, therefore understanding the importance of these effects will be crucial to inform the development of appropriate national and international policy to mitigate against the worst consequences of this air pollutant.


Introduction
Strong connections exist between ozone (O 3 ) dry deposition and atmospheric O 3 concentrations.Globally the reduction of atmospheric O 3 concentrations through dry deposition processes is equivalent to ∼ 20 % of tropospheric O 3 photochemical production (Royal Society, 2008) though the magnitude of the deposition term will vary with season and land cover.This is because dry deposition is determined by the sink strength of underlying vegetation, which is largely controlled by stomata accounting on average for approximately 40-60 % of total ecosystem O 3 uptake (Cieslik, 2004; Published by Copernicus Publications on behalf of the European Geosciences Union.Fowler et al., 2001Fowler et al., , 2009)).During spring and summer periods, when vegetation is physiologically most active, high stomatal conductance (g sto ) will result in high dry deposition rates thereby increasing the O 3 loss from the lower atmosphere and decreasing atmospheric O 3 concentrations at ground level (Colbeck and Harrison, 1985).Conversely, during periods of extended hot, dry weather conditions the vegetation can become stressed by high temperatures and soil moisture deficits (SMDs).These conditions will see plants reduce g sto in an effort to limit water loss with subsequent reductions in dry deposition which can lead to maintained high atmospheric O 3 concentrations (Pio et al., 2000).This effect has been investigated in a number of studies concerned with understanding the importance of O 3 deposition on atmospheric O 3 concentrations.Vieno et al. (2010) found that "turning-off" the dry deposition increased O 3 concentrations by ∼ 20 to 35 ppb on most days during the August heat wave period of 2003.The Royal Society (2008) also investigated the effects of altered dry deposition on O 3 concentrations using bespoke European model simulations; they found that "turning-off" surface deposition caused a 31 % increase in episodic peak O 3 concentrations and a 19 % increase in annual mean daily maximum 1 h O 3 concentrations.These results were in broad agreement with the regional modelling of Solberg et al. (2008) which determined the key drivers of peak O 3 concentration during the height of the 2003 European heat wave by assessing the sensitivity of the modelled results to various parameter changes.Reducing the O 3 deposition term to zero produced the largest effect with average maximum hourly O 3 concentrations of eight surface measurement sites across Europe increasing from ∼ 75 ppb to over 90 ppb.

L. D. Emberson et al.: Scorched Earth
The conditions that are likely to reduce dry deposition (i.e.extended periods of hot dry sunny weather) are the same conditions that are likely to result in the build-up of high O 3 concentrations.The association of poor air quality and extremely warm weather is well established (Lee et al., 2006;Filleul et al., 2006) and is due to a combination of meteorological effects, atmospheric chemical interactions and changes to both the rates and types of terrestrial emissions which occur at elevated temperatures.For example, in the case of UK O 3 concentrations, high summer time concentrations of > 90 ppb are almost always associated with anticyclonic conditions and temperatures in excess of 28-30 • C (Lee et al., 2006) when limited mixing and dilution along with synoptic transport pathways often brings already highly polluted air from mainland Europe to the UK (Jenkin et al., 2002).Increases in biogenic volatile organic carbon (VOC) emissions are also likely to occur with higher temperatures (although the more extreme levels of temperature and drought stress may lead to decreases in biogenic VOC emissions) with such changes being non-linear and species dependant.For example, Lee et al. (2006) found that during the European heat wave of 2003 daytime isoprene concentrations of greater than 1600 ppt were observed in South East England; such concentrations are more typical of high emitting tropical forested regions and were considered likely to, at least in part, have been due to increases in biogenic emissions.
Such conditions were experienced during the late summer heat wave of 2003 which affected much of Western Europe, especially Switzerland, France and southern England.In the UK the heat wave lasted for a 2 week period between 4 to 13 August, during which time temperatures peaked at a new record of 38.5 • C. Stedman (2004) investigated the association of this heat wave with excess deaths caused by air pollution using established dose-response relationships and found that between ∼ 225 and 595 excess deaths were associated with elevated O 3 in the UK during the August 2003 episode as compared with the same period in 2002 (with ∼ 95 % of deaths occurring in England and Wales).These figures represented ∼ 10 to 30 % of the UK Office for National Statistics reported total excess deaths.The ranges given are based on estimates made assuming either a zero threshold or a threshold of 50 ppb for O 3 effects on human health respectively, based on COMEAP (1998).Across Europe, it was estimated that the 2003 heat wave was responsible for 22,000 premature deaths (Schär and Jendritzky, 2004) leading to losses of an estimated £7bn.Of the deaths occurring in European cities, between ∼ 2.5 to 80.0 % could be attributed to O 3 based on data analysis from a study of major French cities (Filleul et al., 2006).The occurrence of such heat wave conditions is likely to increase in the future with climate models suggesting the probability of exceeding 35 • C in the UK will increase from 0.6 % under current conditions to 6 % by 2080 (Schar et al., 2004;Stott et al., 2004).The likelihood that such conditions will co-occur with high O 3 concentrations will depend on how the UK and European air pollution emission reduction policy develops in coming years.
At the same time as increased atmospheric O 3 concentrations may be causing impacts on human health, the reduction of O 3 deposition to vegetation may be viewed as protecting ecosystems from O 3 damage (Fuhrer, 2009;Matyssek et al., 2007).However, the assessment of vegetation risk will be extremely dependent upon the metric used to estimate effects (Ashmore et al., 2004).In Europe, within the United Nations Economic Commission for Europe (UNECE) Convention on Long-Range Transboundary Air Pollution (LRTAP), use of the flux based metric to assess O 3 impacts on ecosystems is now firmly accepted (LRTAP Convention, 2010).This metric is capable of taking into account the influence of environmental conditions on the sensitivity of the vegetation to O 3 and hence is suitable for risk assessment under conditions representative of future, warmer and drier climates (Harmens et al., 2007).
The likely increases in the occurrence of future heat wave events across Europe will lead to a combination of conditions (increases in biogenic O 3 precursor emissions; conditions favouring O 3 formation and conditions likely to reduce O 3 deposition) that are very likely to substantially increase ground level O 3 concentrations.To understand the complex interactions of these factors requires the development and application of coupled models capable of predicting both O 3 concentration as well as O 3 deposition at fine resolution.In the study presented here this is achieved by combining the Community Multiscale Air Quality (CMAQ) chemical transport model (CTM) with the Deposition of Ozone for Stomatal Exchange (DO 3 SE) O 3 dry deposition model.The CMAQ model can be applied using nested grids down to rather fine horizontal resolutions, desirable for human health assessments; in this study we use a 9 km resolution determined by the spatial resolution of the input land cover dataset.The CMAQ model has been found to provide reliable estimates of O 3 concentrations in both rural and urban settings across the UK (Carslaw, 2012;Chemel et al., 2010;Yu et al., 2008;Sokhi et al., 2006).
The DO 3 SE model is the only regionally parameterised O 3 dry deposition model that has been developed specifically to estimate damage to vegetation (Emberson et al., 2001;Simpson et al., 2007).The advantage of this scheme is that the surface resistance component, and particularly the g sto algorithm, have been parameterised, evaluated and used to assess damage for a wide range of species grown across Europe (Emberson et al., 2001;Tuovinen et al., 2004;Simpson et al., 2003;Emberson et al., 2007;LRTAP Convention, 2008).Therefore, as well as providing an estimate of O 3 deposition that is likely to better incorporate the influence of European vegetation on O 3 mass balance, the DO 3 SE dry deposition module can also be used to estimate effects on vegetation across the region.
Due to these advantages, the DO 3 SE deposition modelling scheme was integrated into CMAQ, thus creating the "CMAQ-DO 3 SE" model.The resulting model is used in this study to assess risk and impacts (i) to human health, based on UK air quality objectives and methods to estimate premature mortality due to acute exposure to O 3 (e.g.COMEAP, 1998); and (ii) to ecosystems following methods proposed by the LRTAP Convention (LRTAP Convention, 2008).The study focuses on the heat wave of 2006 which was preceded by a hot dry summer causing extensive drought coupled with high O 3 concentrations across the UK and Europe (Doherty et al., 2009).This study provides a first quantification of the importance of the vegetation sink in determining O 3 dry deposition, atmospheric O 3 concentrations, and associated impacts on human health and ecosystems.

Modelling O 3 photochemistry and dry deposition
The UK O 3 concentration fields used in this study were generated using the USEPA Models-3/CMAQ model, with the central CMAQ model being a third-generation CTM (Byun and Ching, 1999;  CMAQ ordinarily uses one of two different O 3 dry deposition schemes within the MCIP: the surface exchange aerodynamic method (Pleim et al., 2001) or the RADM dry deposition algorithm (Wesely, 1989); both use an electrical resistance approach to estimate dry deposition.Since such an approach is common to most O 3 deposition modelling methods, including the DO 3 SE model, it was relatively straightforward to substitute the DO 3 SE deposition model, described in Emberson et al. (2000Emberson et al. ( , 2001) ) and Simpson et al. (2012), into CMAQ's MCIP.
A particularly unique feature of the DO 3 SE model is the method for estimating surface resistance (R sur ) described in Eq. (1).
where R sur is calculated as a function of stomatal and nonstomatal canopy resistances, the latter including external plant surface (r ext ), within-canopy (R inc ) and ground surface/soil (R gs ) resistances for which empirical methods and constants are employed based on published literature as described in Simpson et al. (2003).Stomatal (r sto ) and external (r ext ) resistances to O 3 deposition are defined at a leaf/needle level (denoted by a lower case r) and scaled according to leaf and surface area indices (LAI and SAI, respectively) to provide canopy scale estimates (denoted by an upper case R).
To estimate g sto (the inverse of r sto ) which represents the stomatal control of O 3 uptake to the sites of O 3 damage within the leaves/needles of plants, the DO 3 SE model employs a multiplicative algorithm, based on that first developed by Jarvis (1976) and modified for O 3 stomatal flux estimates (Emberson et al., 2000(Emberson et al., , 2007) ) according to Eq. (2).
where the species-specific maximum g sto (g max ) is modified within a limit set by a minimum daytime g sto value (f min ) and by functions (scaled from 0 to 1) to account for g sto variation with leaf/needle age (f phen ) or O 3 induced early senescence (f O 3 ) over the course of the growing season and the functions f light , f T , f D and f SW relating g sto to irradiance (PPFD), temperature (T ), vapour pressure deficit (D), and soil water status (SW).The influence of SW on g sto (f SW ) is modelled according to a new method described in Büker et al. (2012) which incorporates the energy balance terms of the Penman-Monteith model (Monteith, 1965) (Büker et al., 2012).Further details of these methods are given in the Supplement.
The DO 3 SE model is parameterised for both broad land cover types (Simpson et al., 2003) and individual forest, crop and grassland species (LRTAP Convention, 2008) such that European parameterisation currently exists for 10 land cover types, 7 forest tree species, 5 crop species and 2 grassland species.Some of the forest species have climate-specific parameterisations that account for the different species ecotypes (LRTAP Convention, 2008).When DO 3 SE is used in combination with regional photochemical models, the broad land cover parameterisations are used to determine O 3 deposition whilst the species-specific parameterisations are used to assess O 3 risk and vegetation damage.The parameterisations of the cover types and species used in this UK based study are given in Table S1.Six broad land cover types: coniferous, deciduous and mixed forests; croplands; productive grasslands and heathlands were used for total deposition estimates since these were considered to represent dominate land cover types in the UK landscape; the ecosystem risk assessment investigated three species: beech, wheat and productive grasslands.

Land cover data
Land cover data for the UK were obtained from the UN-ECE LRTAP Convention harmonised land cover map (Cinderby et al., 2007).These data were compiled specifically for use in assessing the impacts of air pollutants on European ecosystems and agriculture from a mixture of existing digital and paper sources including the European Environment Agency (EEA) Corine Land Cover 2000, the SEI Land European Cover Map (2002 Revision), the FAO Soil Map of the World and the EEA European Biogeographical regions ( 2005).These land cover data were aggregated from 1 km ×1 km resolution to the 9 km ×9 km UK grid used by CMAQ.

Meteorological and emissions data
The meteorological driver for CMAQ-DO 3 SE is the WRF model (Skamarock et al., 2008).The lateral conditions for WRF are provided by the National Centres for Environmental Prediction (NCEP) FNL (Final) Global Tropospheric Analyses at 1 • grid spacing and 6 h temporal resolution (http: //rda.ucar.edu/datasets/ds083.2/).The initial and boundary conditions for CMAQ were derived from the UK Meteorological Office CTM (STOCHEM).Annual anthropogenic emissions (CO, NO x , NH 3 , SO 2 , NMVOC and PM 10 ) data for 2006 were obtained from a number of sources includ-ing the European Monitoring and Evaluation Programme (EMEP; http://www.ceip.at)at a grid resolution of 50 km and the UK National Atmospheric Emissions Inventory (NAEI; http://naei.defra.gov.uk) at a grid resolution of 1 km.The emissions from point sources were derived from the European Pollutant Release and Transfer Register (E-PRTR) and the NAEI databases.The emissions from EMEP were used in CMAQ-DO 3 SE domain 1 (EU 81 km grid) and 2 (EU/UK 27 km) and domain 3 (UK 9 km grid) used the emissions from NAEI.
The annual primary emissions were disaggregated into model chemical species using source specific model species speciation profiles.The profiles for NMVOC were estimated by mapping the UK VOCs emissions (Passant, 2002) with the model chemical species in the USEPA emissions speciation database (http://www.epa.gov/ttn/chief/software/speciate).
These species were then disaggregated into hourly emissions using temporal profiles for 11 CORINAIR/UNECE emission source categories from the City-Delta project (http: //aqm.jrc.ec.europa.eu/citydelta/).The biogenic emissions, isoprene and terpene, were estimated using 100 m grid resolution CORINE land cover data, incoming shortwave radiation and surface temperature, using methods described by Guenther et al. (1995) and Sanderson et al. (2003).

Estimating O 3 effects on human health
The human health risk from O 3 pollution was assessed according to (i) exceedances of the UKs national air quality objective (http://uk-air.defra.gov.uk/air-pollution/uk-eu-limits).This states that the 8 h mean O 3 concentration should not exceed 100 µg m −3 (DM 100 ) on any more than 10 days per year; and (ii) estimates of the attributable mortalities (or deaths brought forward) due to short-term exposure to O 3 .This was estimated using a concentration-response function derived from the WHO (2004) meta-analysis of timeseries studies of 15 cities in France, Italy, the Netherlands, Spain and the UK.The time-series study reported a risk estimate of 0.3 % increase in daily all natural cause mortality (95 % confidence interval (CI) 0.1 to 0.4 %) per 10 µg m −3 daily maximum 8 h O 3 concentration.This figure is without a threshold and can be translated into the relative risk (RR) of 1.003 (95 % CI 1.001 to 1.004).The concentration response coefficient (β) is a slope of the log-linear relationship between RR and concentrations (RR = exp β X ) and is estimated as ln (RR) /10.
To calculate the acute premature deaths, the fraction of the disease burden attributable to the risk factor (AF) when the daily maximum 8 h mean was greater than a threshold (a level below which O 3 has no effect on mortality) was calculated as shown in Eq. ( 3).
where X is the change in daily maximum 8 h O 3 concentration above a threshold.
The attributable deaths for the days that concentrations were above the threshold were then estimated using the following expression (Eq.4): where y o is the baseline mortality rate (i.e. annual natural cause mortalities per million population), Pop is the annual size of the exposed population.Due to the lack of daily population and mortality rate, these annual values were divided by 365.The annual attributable deaths were then estimated as an accumulation of Mort for the entire year.The attributable deaths during the June to July period were also calculated to quantify the contribution of the peak O 3 heat wave period.
The uncertainty in these estimates associated with the doseresponse relationship is given in Table 2 that shows estimated values within the 95 % confidence interval of the relationship.However, it is acknowledged that there will be other uncertainties that are not quantified such as those associated with O 3 precursor emissions, meteorology, atmospheric O 3 modelling and population estimates, all of which will add to the uncertainty of modelled values.
The 2006 baseline natural cause mortality rate, derived from ONS (2008), was 11 581 per million population.Population data at a spatial resolution of 100 m was taken from the European Environment Agency web site (EEA, http:// dataservice.eea.europa.eu/)and based upon the Eurostat census 2001.As a consequence we have assumed no significant changes in UK population between 2001 and 2006.These population data were aggregated to the 9 km ×9 km CMAQ grid.
It is recognised that the confidence in the existence of associations between O 3 exposures and the health outcomes decrease as concentrations decrease (IGBP, 2007).The estimates of effects were made at concentrations greater than 35 ppb daily maximum 8 h mean based on recommendations by WHO (2004) and UNECE (2004).However, using this cut-off is recognised to underestimate the O 3 effects; therefore, estimates were also made without a threshold to indicate an upper estimate of the attributable effects of O 3 on mortality, although it is recognised that such high mortalities would be extremely unlikely.Further discussion of the uncertainties related to the use of this index, with and without threshold, is provided in Sect. 4.

Estimating O 3 effects on ecosystems
The flux-based method recommended by the UNECE LR-TAP Convention is used in this study to assess O 3 risk to ecosystems during 2006 (LRTAP Convention, 2008).This method requires an estimate of the flux metric POD y (Phytotoxic Ozone Dose over a threshold y) for the three different representative species: beech, wheat and grasslands.The respective critical levels for the POD y metric (CL) represent levels below which damage would not be expected to occur and are provided in Table 1 and based on data described in LRTAP Convention (2008).
The calculation of POD y requires estimates of stomatal O 3 flux (F st ) which is performed according to the methods provided by LRTAP Convention (2008).F st (nmol O 3 m −2 projected leaf area (PLA) s −1 ) is calculated according to Eq. ( 5) which accounts for deposition to the cuticle through incorporation of the leaf surface resistance (r c ) and boundary layer resistance (r b ) terms: where c(z 1 ) is the concentration of O 3 at the top of the canopy (nmol m −3 ) at height z 1 (m), g sto is in m s −1 , r b is the leaf quasi-laminar resistance and r c the leaf surface resistance, both given in s m −1 .For further details on the resistance scheme see LRTAP Convention (2008).The accumulated F st above an O 3 stomatal flux threshold of y, given in nmol m −2 s −1 provides the POD y index and is calculated according to Eq. ( 6).
and n is the number of hours within the accumulation period.The values for the threshold y vary by species as described in Table 1.
The F st accumulation period aims to define that part of the growth period when the species is most sensitive to O 3 and varies for different species.For all species except wheat this is assumed to be equivalent to the period between the start (SGS) and end (EGS) of the growing season, given in the Supplement.For wheat, the accumulation period is considered to be shorter than the entire growth period since evidence shows that the period around wheat grain filling is more sensitive to O 3 (Younglove et al., 1994;Soja et al., 2000).Full methods are again provided in the Supplement.

CMAQ-DO 3 SE model simulations
The CMAQ-DO 3 SE model was applied to simulate O 3 concentrations and deposition across the UK for the year 2006; this year was chosen as during the last couple of weeks of July much of the UK experienced a heat wave that resulted in extreme weather conditions and high O 3 concentrations that came at the end of an extended period of drought.
The CMAQ-DO 3 SE model was run for three alternative model scenarios for the whole of 2006; a "reference" scenario for which O 3 deposition was modelled using the DO 3 SE model method described above; and two scenarios based upon forced alterations to the dry deposition model during the months of June and July 2006.These were achieved by altering Eq. ( 2); the first of the two scenarios assumed a minimum limitation to g sto calculated by g max • f phen • f light (referred to as the "no stress" scenario; for these calculations the irradiance used to estimate f light was assumed to be that for clear sky to avoid limitations to g sto from overcast conditions.Since the stomatal response to light saturates at rather low light levels (∼ 25 % of clear sky conditions), overcast conditions would not be expected to cause much limitation to g sto ).The second scenario assumed full limitation to g sto calculated by g max • f min (referred to as the "stress" scenario).

O 3 concentrations and metrics
The predictive performance of the CMAQ-DO 3 SE model was assessed by comparing the modelled and measured values of a number of O 3 metrics associated with assessing risk to human health.This was done using both scatter plots (Fig. 1) and statistical measures described in Chang and Hanna (2004) and Builtjes (2005), including factor of 2 of the observations (FAC2), mean bias (MB), normalised mean bias (NMB), root mean square error (RMSE) and correlation coefficient (R) (Table 2).Comparisons of these metrics were made both between rural and urban locations as well as dif-ferent time periods (annual as well as winter, spring, summer and autumn).These metrics include daily 8 h mean O 3 concentrations, number of days of exceedance of the DM 100 and the attributable premature mortality that would be associated with the prevailing O 3 concentrations.Measured values were derived from hourly O 3 concentrations collected from 73 urban sites and 19 rural sites across the UK during 2006; these data were collected from sites that the UK Department of environment, food and rural affairs (Defra) consider suitable for evaluation of UK CTMs (Derwent et al., 2010).Figure 1 shows both the location of these sites as well as a modelled vs. observed comparison of the daily 8 h mean O 3 concentrations by season for urban and rural locations.Table 2 and Fig. 1 show that the CMAQ-DO 3 SE model is able to describe these human health risk metrics reasonably well.Overall, the FAC2 values indicate > 75 % of modelled data are in within a factor of 2 of the measurements.Although the model has tendency to over predict the DM 100 by approximately 5 µg m −3 at urban sites and 2 µg m −3 at rural sites, the NMB values show both annual and seasonal model data are well within the ±0.2 range considered acceptable (Derwent et al., 2010).Despite positive biases of the DM 100 , the model slightly under predicts the number of days that DM 100 is exceeded, i.e. approximately 2 (−7 %) and 4 (−13 %) days at urban and rural sites respectively for the period May to July.The attributable deaths at urban sites are over predicted by a small margin due to the tendency of over predicting DM 100 .The attributable deaths at rural sites are marginally under predicted which is likely to be driven by the negative bias of the DM 100 in spring.

Soil moisture status
Central to this study is the CMAQ-DO 3 SE model's ability to predict SMD and subsequent influence on stomatal O 3 flux and deposition.To assess the former the SMD results of CMAQ-DO 3 SE are compared with equivalent estimates made by the Met Office Rainfall and Evaporation calculation system (MORECS) model described in the National Hydrological Monitoring programmes review for 2006.This review provides estimates of SMDs for the end of July for a grassland cover type over the UK (Marsh et al., 2008).These data showed widespread drought conditions, particularly across almost all of England and the south and east of Scotland (Northern Ireland not shown).Based on an average available water content (AWC) assumed in the MORECS model (Hough and Jones, 1997) these drought areas had less than 25 % of (AWC) remaining.By comparison, the CMAQ-DO 3 SE model estimates a similar pattern of drought across England by the end of July 2006 (shown in Fig. 2) with soils having less than 30 % of AWC remaining.In Scotland CMAQ-DO 3 SE estimates the drier areas to the west of the country rather than the east which seems most likely driven by soil texture (see also Fig. 2).However, the area of dis-crepancy is relatively small and on the whole the CMAQ-DO 3 SE model simulates the same spatial distribution and relative magnitude of reduced plant water availability.Importantly, the CMAQ-DO 3 SE model captured the high July SMDs.These were attributed to the very limited rainfall between November 2005 to February 2006 which allowed significant SMD to be carried through the winter in parts of eastern, central and southern England, this was followed by sharp increases in SMD in April 2006, which were unable to recover in spite of a wet May and which were followed by a steep and more sustained rise in SMD during June (Marsh et al., 2008).Figure 2c shows the seasonal profile of % ASW for a number of grids across the modelling domain.Also shown are the thresholds at which % ASW will cause a 50 % and 75 % reduction in g sto .This figure clearly shows that modelled soil water stress is high for rather extended periods around June and July suggesting the estimated reductions in O 3 deposition are rather insensitive to model uncertainty in the soil water status calculation.

O 3 effects on human health
Table 3 provides estimates of the exceedance of DM 100 for the UK.Results show that exceedance can reach nearly 30 days under the "stress" or drought condition, more than double the number of days of exceedance under the "no stress" condition.For the "reference" and "stress" scenarios, 80 % of the exceedance occurring over the entire year can be attributed to the June and July period.
Table 3. UK exceedance of number of days when 8 h mean O 3 > 100 µg m −3 and attributable mortalities due to short-term O 3 exposure using two different thresholds (0 and 35 ppb), presented for both the entire annual period and the June to July period in 2006 for all three scenarios.The 95 % confidence intervals for estimates of attributable death brought forwards are shown in brackets.The exceedance of this air quality objective varies across the UK as seen in Fig. 3 which describes the DM 100 during the June and July period across the UK for all three scenarios.This variation is driven by the north/south gradient of O 3 concentrations with higher concentrations in the south resulting from stronger photochemistry, higher UK temperatures and increased long-range transport of O 3 from outside the UK in southern England (Lee et al., 2006).Differences in annual mean 24 h O 3 concentrations between the "no stress" and "stress" scenario were on average 2.5 ppb across the whole of the UK.However, for the June-July period these differences increased to 8 ppb (data not shown).These variations in O 3 concentration translate into geographically variable human health risk.The "reference" results in Fig. 3 show the existence of a north-to-south DM 100 gradient, with south east England having the highest number of days (between 30 and 35), exceeding 100 µg m −3 .These results are mirrored by those for the full annual period (data not shown).By contrast the "no stress" scenario is very different, with the southern part of England tending to have only 12 days on average during which the DM 100 is exceeded; for this scenario the DM 100 exceedances for the June-July period are  less important in relation to the total number of days of exceedance for the entire year (Table 3) which is likely due to less extreme conditions during this high summer period.

Scenario
Table 3 shows the estimates of deaths brought forward due to acute O 3 exposure both with and without a threshold of 35 ppb for the "reference", "no stress", and "stress" scenarios (the merits of including estimates made without a threshold are discussed in Sect.4).The upper level of premature annual mortalities for the "reference" case, assuming no threshold, is just above 16 000 (5425-21 420) in which approximately 20 % of excess deaths are attributed to the June-July period.In contrast, the estimate of premature annual mortalities with a 35 ppb threshold is much smaller at approximately 1900 (630-2500) with almost 50 % associated with the June-July period.The estimates with the threshold show greater differences between the annual and June-July periods due to the fact that outside of this high summer period, concentrations are more likely to drop below the threshold.
The influence of altered O 3 deposition on mortality can be inferred from the results presented in Table 3 by comparing the number of attributable deaths between scenarios.The number of UK mortalities in 2006 estimated for the "reference" scenario assuming no threshold increases under the "stress" scenario by ∼ 90 (35-125) premature deaths and is reduced under the "no stress" scenario with ∼ 415 (140-540) premature deaths avoided.Similar results are found using the 35 ppb threshold with the number of premature deaths in 2006 being reduced by ∼ 370 (125-490) under the "no stress" condition and with ∼ 90 (30-115) additional premature deaths estimated under the "stress" scenario.The uncertainties associated with the dose-response relationship suggest that the estimates are tending towards the higher end of the range in premature deaths.
Figure 4 shows the June-July premature mortalities due to O 3 exposure under the "reference", "no stress", and "stress" scenario using the 35 ppb cut-off.The numbers of deaths brought forward are higher in and around urban areas associated with large human populations in polluted areas.This spatial distribution pattern is similar to that found for the an- nual assessment and that estimated without a threshold hence these data are not shown here.The main differences between perfect sink and drought conditions occur over populated areas with the "no stress" scenario reducing premature mortalities in and around these urban centres by approximately 40 % during the June-July period.

O 3 effects on ecosystems
The O 3 effects on ecosystems have been assessed using the flux based POD y metric.Under the "reference" scenario described in Fig. 5 the modelling suggests that beech is at risk from O 3 as POD 1.6 values exceed the critical level across most of the country, with more than double the exceedance across approximately half the geographical area.For wheat, the POD 6 index is not exceeded under the "reference" situation (data not shown); a POD 0 value is shown in Fig. 5 to give an indication of the geographical variation in O 3 flux though the index suggests such low fluxes would not be damaging.For grasslands, there is currently no flux based critical level for the LRTAP Convention ( 2008) parameterisation used in this study although recently established flux based critical levels of 1 mmol O 3 m −2 (LRTAP Convention, 2010) would suggest likelihood of a threat from O 3 across all of the UK.The results in Table 2 suggested a slight overestimate of modelled daily 8 h mean O 3 concentrations in rural areas during the summer with a MB value of 4.88 µm −3 which suggests these POD values might be overestimated, especially since the accumulated nature of this index will enhance any model bias.
Under the different scenarios the values recorded for each of these vegetation indices changes quite substantially.The magnitude of risk between the "reference" and "scenario" runs associated with the flux metric can be most easily compared by showing relative exceedances for each, i.e. scaling exceedance by the respective critical level for each index as suggested in Simpson et al. (2007).Figure 6 shows these relative exceedances (RCLe) for beech; a RCLe > 1 denotes an exceedance of the POD y critical levels.The POD y RCLe shows most risk under the "no stress" condition with values Fig. 6.The relative critical level exceedance (RCLe) using POD y metrics for beech under the (i) "reference", (ii) "no stress" and (iii) "stress" scenario for 2006.
between 5 and 8 times the CL across the UK, with the highest values in the south of England.By comparison, the "reference" scenario shows very low exceedances, predominantly between 0 and 3 extending to 4 times the CL in southern England; under the "stress" scenario the CL is not exceeded (i.e.values are less than 1) for the whole of the UK.
The variation in O 3 flux to vegetation (in part described by POD y ) is related to total O 3 deposition and hence O 3 loss from the atmosphere; higher fluxes will equate to higher O 3 deposition rates and reductions in atmospheric O 3 concentration.Deposition rates vary with land cover; the influence of vegetation type on deposition during the growing season can be seen in Fig. 7 which investigates the influence of the three scenarios on the monthly trends of daily maximum 1 h O 3 deposition for four different sites dominated by particular land cover types common in the UK: Easter Bush (grasslands), NE Cambridge (wheat), Alice Holt (forest) and London North Kensington (urban).The are presented as monthly mean box and whisker plots showing the maximum and minimum O 3 deposition rate as well as the inter-quartile range.
At all locations for the "no stress" scenario O 3 deposition increases from a minimum in winter (∼ 0.01 kgO 3 ha −1 for all land cover types) to a maximum during the summer; the influence of land cover is apparent for these maximum deposition rates with the urban locations showing the least O 3 deposition (∼ 0.025 kgO 3 ha −1 ) compared with the agricultural land cover type of NE Cambridge which sees deposition of up to 0.06 kgO 3 ha −1 .The effect of reduced sink strength is apparent during June and July where there is a notable difference in the O 3 deposition for both the "reference" and "stress" scenarios for the three vegetated land cover types.Forests show a reduced O 3 deposition already by June for the "reference" scenario which may be due to a more rapid drying of the soil for this vegetation type.By July, the period of hot dry sunny weather has continued for long enough that all land cover types have dried out the soil so that the "reference" and "stress" scenarios record the same level of reduced O 3 deposition.As such, the land cover type and the degree of stress to which the land cover is exposed will both influ-

Discussion
The results presented in this paper have quantified the likely influence of extreme meteorological events on O 3 dry deposition, subsequent O 3 concentration and ultimately human health and ecosystem effects.This study focussed on a heat wave period that occurred across the UK and much of Europe during June-July of 2006.The study clearly demonstrated the substantial influence that a drought limited vegetation sink had on atmospheric O 3 concentrations, increasing 24 h mean values by approximately 8 ppb as an average over the entire UK over the June-July period.Under such conditions, the human health risk, quantified by the DM 100 index, was increased by almost two thirds (see Table 3).The increased ground level O 3 concentrations resulted in ∼ 460 additional premature deaths for the June to July period assuming a damage threshold under the "stress" drought scenario compared to the "no stress" perfect O 3 sink scenario, with the largest differences in the estimated premature deaths of the different scenarios reaching 40 % in and around the urban centres.In contrast to the risk to human health, the occurrence of drought reduces stomatal O 3 flux to vegetation as plants close stomata in an attempt to conserve water.The resulting reduction in O 3 dose means that under the dry "stress" and "reference" conditions ecosystems were at less risk from O 3 damage with RCLe being halved across much of the UK when compared with the "no stress" conditions.
There are of course associated with these calculations.In relation to the dose-response relationships used to assess premature mortalities under heat wave events there are two main issues: firstly, the question of whether a threshold for O 3 effects on human health exists; and secondly, what is the likely confounding influence of other co-occurring environmental (and socio-economic) factors on the health impact results.
Due to uncertainties in the shape of the dose-response function at very low O 3 concentrations, a threshold for the effects of O 3 on mortality is likely to exist.Although there is insufficient evidence for identifying such a threshold, a range of values have been assumed in previous studies such as 50 ppb (COMEAP, 1998 andStedman et al., 1997) and 35 ppb (WHO, 2004 andUNECE, 2004).The 35 ppb cut-off was recommended by UNECE (2004) and was used in this study as it is more relevant to the European seasonal variation and geographical distribution of background O 3 concentrations and importantly, the range of concentrations for which most CTMs provide reliable data.It is recognised that the human health effects estimated using this threshold are likely to underestimate the real effects of O 3 ; as such, estimates without a threshold were also made to indicate the upper estimate of the attributable effects of O 3 on mortality though this is considered extremely unlikely to occur.The choice of threshold makes a substantial difference to absolute estimates of premature deaths.Over the entire year, assuming a zero threshold gave estimates of ∼ 16 140 premature deaths due to O 3 compared with the 35 ppb threshold value of ∼ 1880.These estimates can be compared with a study conducted for the entire EU which estimated ∼ 22 000 premature mortalities for the year 2000 assuming a 35 ppb threshold (EEA, 2007) suggesting the zero threshold estimates are extremely exaggerated.However, the difference in estimates of premature deaths between the "no stress" and "stress" scenarios was similar irrespective of the use of a zero or 35 ppb threshold with values of ∼ 505 and 460, respectively.Therefore, the importance of the O 3 deposition term on the magnitude of the human health risk, at least in absolute terms, is largely independent of the threshold value chosen.
In terms of confounding effects it is widely recognised that due to the processes governing O 3 formation, it is likely that high O 3 concentrations will often co-occur with high temperatures.The singular effects of O 3 (e.g.Stedman, 2004) and temperature stress (e.g.Johnson et al., 2005) on human health have been investigated frequently with results clearly showing that both stresses can cause substantial impacts when acting individually.However, there have only been a small number of studies investigating both stresses acting together.Those studies that have been conducted have shown the importance of geographical variation (perhaps due to the frequency of air conditioning use, personal activity and pollution exposure levels, and environmental conditions) as determinants of the O 3 -heat mortality effects (Filleul et al., 2006;Ren et al., 2008).Doherty et al. (2009) performed simulations for three years (2003, 2005 and 2006) for 15 UK conurbations and found that overall the number of deaths associated with O 3 appeared to be slightly higher than those attributable to heat and that proportionately more O 3 deaths occurred during periods of very high temperature.The work presented here suggests that similar studies to that by Doherty et al. (2009), which also specifically address issues related to variable O 3 deposition under climate change, should be performed in the future complimenting efforts that have tended to focus only on improving our understanding of how O 3 transport and formation vary under climate change (Royal Society, 2008).
In terms of O 3 impacts on vegetation our results suggest a reduced risk under the extreme meteorological events.The findings presented here for the "reference" scenario provide the first national level POD y assessments that have incorporated the influence of drought (soil water stress) on stomatal O 3 flux; these follow the methods described in Büker et al. (2012) and are consistent with methods recommended by the LRTAP Convention (2008).Previously, regional modelling of O 3 effects on vegetation has assumed non-limiting soil water and hence could only provide a "worst case" assessment of O 3 risk to vegetation (Simpson et al., 2007).The results of Simpson et al. (2007) showed a range of 4 to 12 mmol O 3 m −2 for POD 1.6 values for forests across the UK with a north-south gradient from low to high values.The "no stress" scenario of this study mimics this geographical pattern though the magnitude of O 3 flux is higher (with values frequently reaching the 25 mmol O 3 m −2 across parts of southern England) as would be expected given the higher O 3 levels of this 2006 study year and the near unlimited stomatal conductance to O 3 of the scenario run.
In contrast, the "reference" case has a similar range of POD 1.6 values (4 to 12 mmol O 3 m −2 ) across the UK but the spatial gradient is the opposite with highest values occurring across Scotland and lowest values in southern England.This pattern is driven by soil moisture and highlights the importance of incorporating the influence of soil moisture in estimates of both stomatal O 3 flux for ecosystem risk assessment but also for accurate estimates of O 3 deposition.
However, it should be noted that enhanced SMD will not only affect stomatal O 3 fluxes; increased SMD will also adversely impact vegetation growth reducing net primary productivity (NPP) (Ciais et al., 2005) and increase the likely frequency of outbreaks of forest fires, which in turn can influence air quality through biomass burning emissions (Lyamani et al., 2006).There is also evidence to suggest that SMD itself may influence the formation of extreme meteorological events such as heat waves through alterations to the balance between latent and sensible heat fluxes; this is due to the absence of soil moisture reducing latent heat cooling which amplifies surface temperature anomalies (Fischer et al., 2007).As such, it seems inappropriate in this study to make any attempt to quantify changes in yield or biomass loss that might result from altered O 3 flux due to changes in O 3 deposition; our study can only indicate changes in O 3 risk.
Although this is the first time that the influence of dry deposition has been related to health effects for ground level O 3 , this is not the case for other pollutants.A number of studies have investigated the relationship between particulate matter deposition to vegetated surfaces and human health (Tiwary et al., 2009).Recognition of the important role that vegetation (especially trees in urban centres) plays in improving urban air quality has even led to the development of planning policies to help control exposure of citizens to this toxic pollutant (McDonald et al., 2007).
To truly understand the implications of changes in O 3 deposition for human health and ecosystem risk assessment the tools employed in the assessment must be "fit for purpose".CMAQ has been used in a number of previous modelling studies across the UK and has shown good agreement with observations in both urban and rural areas (Carslaw, 2012;Chemel et al., 2010;Yu et al., 2008;Sokhi et al., 2006).In this study, good agreement between model results and measurements over rural locations is observed.Although there is a tendency for CMAQ to overestimate O 3 concentrations (as is the case for many O 3 CTMs, Carslaw, 2012), the positive biases at most sites are below 10 % and as such the estimates of the effects of O 3 on human health and ecosystems are expected to be overestimated only by a small margin.
The ability of the DO 3 SE model to estimate SMD is an important component of the modelling performed here; comparisons with MORECS modelled estimates of SMD (Marsh et al., 2008) confirmed the DO 3 SE model's estimate of high SMD across most of England and Wales.However, it is the influence of SMD on stomatal O 3 flux that is the direct driver of O 3 deposition.A recent evaluation paper by Büker et al. (2012) provides some evidence that this term is being modelled with reasonable accuracy.However, more testing of the module is required to ensure appropriate capture of the influence of SMD on O 3 deposition to different land cover types over the course of the growing season.This will be important for future O 3 deposition simulations given observations that suggest an increasing frequency of hot summers followed by winter rainfall deficits over southern Europe (Vautard et al., 2007).
In addition to SMD there are other recognised uncertainties in the estimates of O 3 deposition, which include aspects of the non-stomatal deposition with evidence that this can vary with environmental conditions, such as surface wetness and temperature (Fowler et al., 2009).As such, the use of constant deposition terms to these non-stomatal sinks assumed in the DO 3 SE model would benefit from a rigorous review.Additionally, efforts to re-formulate and parameterise the stomatal component of the DO 3 SE model may benefit from a focus on the incorporation of more process orientated algorithms that can account for changes in photosynthetic ca-pacity and hence atmospheric CO 2 concentrations, especially under changing climates; such methods have been developed and trialled (Büker et al., 2007) but are yet to be used to estimate critical levels and hence are not yet available for use in O 3 risk assessment.
Finally, it is also recognised that under extreme meteorological conditions mechanisms other than O 3 deposition will affect O 3 concentrations.For example, the NO x or VOC limitation to O 3 formation was found to change during spring and summer O 3 episodes occurring in 1995 across the UK with implications for NO scavenging and subsequent O 3 assessments (Strong et al., 2010).Vieno et al. (2010) also described the importance of capturing changes in temperature and subsequent biogenic isoprene emission that will occur under heat wave conditions.This is important due to the role of isoprene as an O 3 radical source both due to its short lifetime (∼ 5 h) as well as the temperature dependence of its emissions with isoprene concentrations having been shown to increase rapidly according to a non-linear relationship with temperature during heat wave O 3 episodes (Lee et al., 2006).If these biogenic emissions are underestimated whilst NO x emissions remain high then less local O 3 production would be simulated (Vieno et al., 2010).Similar effects would also be caused by underestimating temperature as this would favour peroxyacetyl nitrate (PAN) formation tying up NO x that would otherwise lead to O 3 formation (Vieno et al., 2010).An additional consideration is the deposition of other pollutants that follow the same pathway as O 3 , for example, NO x dry deposition may also be reduced by stomatal closure, and may lead to increased chemical titration of O 3 within the plant canopy (Fares et al., 2012).Further, capturing the wind direction accurately is extremely important for assessment of background O 3 concentrations upon which local O 3 production relies; for example, in the UK many O 3 episodes can be attributed to long-range transport arising from precursors originating over continental Europe (Vieno et al., 2010).This also means that the trajectory pathway of these plumes will be affected by O 3 chemistry in the atmosphere and O 3 deposition; an understanding of the influence of these factors along trajectories is fundamental to accurate assessments of O 3 concentrations downwind of major sources.

Conclusions
The study clearly showed the importance of the O 3 absorbing capacity of vegetation in determining human health risk through alterations in ground level O 3 concentrations.For extreme meteorological events characterised by heat wave conditions lasting only a few weeks, the model estimates that the effect of reducing O 3 dry deposition due to drought are estimated to exceed ∼ 460 excess deaths in the UK in a worst case scenario, though there is some uncertainty in the absolute value of these numbers.At the same time, O 3 damage to vegetation will likely be reduced, although it is acknowledged that the NPP is also likely to be decreased by drought stress.As such, not only is it important to improve our of how emissions and meteorology couple to influence O 3 formation, but also how seasonal environmental conditions will affect the physiological activity and hence O 3 sink strength of the underlying vegetation.Understanding how these factors are likely to interact under those conditions, most likely to lead to high O 3 episodes in the future under changing climates, will provide valuable information to help inform policy decisions on emission reductions that can alleviate the worst effects of O 3 pollution both to human health as well as vegetation and subsequent ecosystem services.

4Figure 1 .
Figure 1.(a) Scatter plot of observed vs. modelled daily maximum 8-hr mean O 3 concentration by season (winter, spring, summer and 2 autumn) for 2006 and (b) the UK location of the 73 urban sites (comprising 19 urban centres, 13 suburban and 41 urban background; red dots) 3 and 19 rural sites (green stars) from which the observed data were collected.4 5 (a) 6

Figure 2 .
Figure 2. (a) Soil moisture deficits (SMD, mm) estimated for the end of July for a grassland cover-type using the CMAQ-DO3SE model; (b) 1 the corresponding dominant soil type at the 9 x 9 km grid resolution used within the modelling and (c) seasonal profiles of CMAQ-DO3SE 2 modelled %ASW for grasslands from 6 sites (shown as black dots on Fig 2a) in relation to the threshold for 50% (solid line) and 75% (dashed 3 line) of reduction in gsto.The shaded area highlights the June to July period.N.B.The major soil types in UK include clay loam (9), loam (6) 4 and sandy loam (3). 5 6 7 (a) (b) 8

Figure 4 .Fig. 4 .
Figure 4. Number of deaths brought forward due to short-term exposure to O3 with threshold value of 35 ppb for the June to July period for 1(i) "reference", (ii) "no stress" and (iii) "stress" scenarios.2

Figure 5 . 4 Fig. 5 .
Figure 5.The PODy in mmol m −2 for beech, wheat and grasslands (with associated critical levels of 4 and 1 mmo 1 there is no flux critical level established for grasslands) under the "reference" scenario for 2006. 2 3

Table 2 .
Statistical measures of model performance in predicting O 3 metrics associated with human health at 19 rural and 73 urban sites in the UK for 2006.The reliability of the different metrics are compared for different periods (annual, winter, spring, summer and autumn) and by location (urban and rural).