Here, we report the first results of model sensitivity simulations to assess the potential impacts of emissions related to future activities linked to unconventional hydrocarbon extraction (fracking) in the UK on air pollution and human health. These simulations were performed with the Met Office Air Quality in the Unified Model, a new air quality-forecasting model, and included a wide range of extra emissions of volatile organic compounds (VOCs) and nitrogen oxides (NOx) to reflect emissions from the full life cycle of fracking-related activities and simulate the impacts of these compounds on levels of nitrogen dioxide (NO2) and ozone (O3). These model simulations highlight that increases in NOx and VOC emissions associated with unconventional hydrocarbon extraction could lead to large local increases in the monthly means of daily 1-h maximum NO2 of up to + 30 ppb and decreases in the maximum daily 8-h mean O3 up to − 6 ppb in the summertime.
A. T. Archibald1,2, C. Ordóñez3,4, E. Brent1, M. L. Williams5
Department of Chemistry, University of Cambridge, Cambridge CB2 1EW, UK1. NCAS-Climate, University of Cambridge, Cambridge, UK2. Met Office, FitzRoy Road, Exeter EX1 3PB, UK3. Dpto. de Física de la Tierra y Astrofísica, Facultad de Ciencias Físicas, Universidad Complutense de Madrid, 28040 Madrid, Spain4. Environmental Research Group, Kings College London, London, UK5
Received: 11 January 2017 / Accepted: 26 March 2018
©The Author(s) 2018
Broadly speaking, our simulations indicate increases in both of these compounds across the UK air shed throughout the year. Changes in the 1-h maximum of NO2 and 8-h mean of O3 are particularly important for their human health impacts. These respective changes in NO2 and O3 would contribute to approximately 110 (range 50–530) extra premature-deaths a year across the UK based on the use of recently reported concentration response functions for changes in annual average NO2 and O3 exposure. As such, we conclude that the release of emissions of VOCs and NOx be highly controlled to prevent deleterious health impacts.
Hydraulic fracturing or ‘fracking’ describes the industrial process used to extract hydrocarbons by pervasively fracturing a shale unit (termed shale plays) with high-pressure water, mixed with sand and chemical additives (Bickle 2012).
This source of unconventional hydrocarbons has created a large amount of interest from investors in the petroleum industry as well as environmentalists and the general public concerned with the impacts associated with it. Whilst fracking as a technique has a long history, its wide-scale uptake is much more recent. This increased uptake is driven largely by technological breakthroughs such as horizontal drilling, which will ultimately allow those involved to extract more of the oil and gas from geological reservoirs than previous methods would have permitted. In the USA, natural gas (an industry term for gaseous mixtures of hydrocarbons rich in methane) from shale accounted for only ~ 2% of total production in 2004.
Owing to proliferation in the fracking of shale plays, by 2035, shale gas is predicted to account for over 63% of total production (BP 2014). This exploitation of shale oil and gas coincides with the USA becoming the world’s largest producer of natural gas. In the USA, shale fracking sites are generally found in large geological basins, for example in the Rocky and Appalachian Mountains and Great Plains. Generally speaking, the most productive sites are in very rural areas well away from large population settlements (US Energy Information Administration 2015).
Whilst methane is one of the most important hydrocarbons extracted from fracking, there are a large number of non-methane volatile organic compounds (NMVOCs) held within the shale plays, including alkanes, cycloalkanes, aromatic compounds and bicyclic hydrocarbons (Sommariva et al. 2014).
The potential environmental and health impacts associated with fracking are varied in terms of their sources and mechanisms of action (Adgate et al. 2014). Of particular interest is the influence of fracking on air quality. The compositions of the emissions associated with fracking are dependent on the particular phase in the life cycle of the fracking well (Moore et al. 2014). However, significant emissions of nitrogen oxides (NOx) and NMVOCs (also referred to as VOCs) are common throughout this life cycle.
Both NOx and NMVOCs represent important species in the production of ozone (O3), a secondary pollutant associated with adverse health effects (Brunekreef and Holgate 2002). Elevated levels of O3, especially in urban areas, are of particular concern in the USA, where fracking has been adopted on a large scale. As such, much of the literature concerned with air pollution resulting from fracking focuses on the relationship between NOx and NMVOC emissions and O3 concentrations (e.g. Edwards et al. 2014 and references therein) and these interactions provide an interesting point of comparison for an assessment of the potential atmospheric impacts of fracking in the UK.
Fracking in the UK remains a controversial issue. However, the House of Commons’ rejection of a moratorium in January 2014 means that wide-scale adoption of fracking is a significant possibility. Whilst there has been considerable speculation about the environmental impacts of fracking in the UK (Environmental Audit Committee 2015) mainly based on extrapolation from the USA, there have been no studies investigating the impacts of any potential risks. It is, therefore, important to provide some detail of how the consequences of fracking in the UK may differ from those of the USA. This sort of preparatory investigation may also provide vital information for the development of effective, evidence-based regulation.
The aim of this study is to determine how increased emissions of VOCs and NOx in regions where fracking could take place in the UK may impact air quality and ultimately human health. To achieve this, we make use of a recently developed numerical air quality-forecasting model, Air Quality in the Unified Model (AQUM) (Savage et al. 2013). The ‘The AQUM model’ section provides a description of the AQUM model set-up used in this study. The ‘Emission scenarios’ section provides information on our emission scenarios for increases in VOCs and NOx in the potential fracking regions, whilst in the ‘Results’ section, we present the results obtained from our AQUM simulations, including analyses of comparisons with observations of O3 and NO2 from the UK Automatic Urban and Rural Network (AURN) and the results from our emission scenarios. In the sections ‘Results’ and ‘Discussion’, we discuss and conclude the study and make recommendations for further work.
The AQUM model
AQUM is a limited area configuration of the Met Office Unified Model (UM), used operationally to provide a 5-day air quality forecast for the UK (Savage et al. 2013). The model is operated with a 12-km × 12-km horizontal resolution grid covering part of Western Europe and 38 vertical levels spanning from the ground up to around 40 km altitude. The representation of chemistry in AQUM is given by the regional air quality (RAQ) chemistry scheme of the United Kingdom Chemistry and Aerosols (UKCA) sub-model of the UM, whilst aerosol processes are parameterised within the Coupled Large-Scale Aerosol Simulator for Studies in Climate (CLASSIC) sub-model.
The RAQ chemistry mechanism includes 40 transported species and 18 non-advected species, 116 gas-phase reactions and 23 photolytic reactions. It takes account of emissions for NOx, carbon monoxide (CO) and the following NMVOC species: ethane (C2H6), propane (C3H8), butane (C4H10), ethene (C2H4), propene (C3H6), formaldehyde (HCHO), acetaldehyde (CH3CHO), acetone (CH3COCH3), methanol (CH3OH), isoprene (C5H8), toluene and o-xylene. Removal by wet and dry deposition is considered for 19 and 16 species, respectively. CLASSIC is a bulk aerosol scheme, where the following aerosol species are treated as an external mixture: ammonium sulphate [(NH4)2SO4], ammonium nitrate (NH4NO3), black carbon (BC), fossil fuel organic carbon (FFOC), biomass burning aerosol, mineral dust, sea salt aerosol and a climatology of biogenic secondary organic aerosol (BSOA). The model has two-way coupling of oxidants between the aerosol and gas-phase chemistry schemes.
Within CLASSIC, sulphur dioxide (SO2) and dimethyl sulphide (DMS) are emitted and oxidised into aerosol sulphate (SO42−) by four oxidants whose concentrations are calculated in RAQ: O3, hydroxyl radical (OH), hydroperoxyl radical (HO2) and hydrogen peroxide (H2O2). The reaction of sulphate with ammonia (NH3) produces ammonium sulphate. Any excess ammonia reacts with nitric acid (HNO3), provided by RAQ, to form ammonium nitrate. Finally, the depleted oxidant concentrations are passed back to the RAQ chemistry scheme of UKCA. (For further details on CLASSIC, see Appendix A of Bellouin et al. (2011).) A full description of the RAQ chemistry scheme and the AQUM modelling system is provided by Savage et al. (2013).
Lateral boundary conditions (LBCs) for meteorological fields are provided by the operational forecast of the Met Office global model. LBCs for chemistry and aerosols are derived from the operational forecast of the Monitoring Atmospheric Composition and Climate (MACC) global model (Flemming et al. 2009).
The main difference between the model configuration presented by Savage et al. (2013) and the one used in this study is the treatment of anthropogenic gas-phase emissions within UKCA. In the present study, monthly emission fields used in the AQUM control run are created for the year 2012 by merging data from three emission inventories: the National Atmospheric Emissions Inventory (NAEI) at 1-km resolution over the UK (Passant et al. 2014), ENTEC shipping data around the UK at 5 km (Whall et al. 2010) and the MACC-II regional emission inventory at 0.125° longitude × 0.0625° latitude resolution (Kuenen et al. 2014) for the rest of the domain. The ENTEC and MACC emission data used here were compiled for 2007 and 2009, respectively, but they are scaled according to published totals from the European Monitoring and Evaluation Programme (EMEP) for 2012. Data from all three sources are interpolated to the AQUM 12-km grid prior to merging. In the version of AQUM introduced by Savage et al. (2013), anthropogenic monthly emission fields for gas-phase species were spread equally over the first four model layers (20, 80, 180 and 320 m mid-altitude) and hourly emission factors derived from an analysis of traffic cycles were applied.
This simplistic approach has been replaced here by a more comprehensive system that enables the separate treatment of emission fields according the so-called selected nomenclature for air pollutants (SNAP) source sectors. In particular, the vertical disaggregation of emissions consists of separate vertical profiles for each SNAP sector based on calculations performed with the SMOKE-EU plume rise model (Bieser et al. 2011). Emissions in our simulations are initially redistributed in the vertical among the EMEP vertical layers (Simpson et al. 2012) following the Bieser et al. (2011) recommendations, but adding a supplementary 0–20-m layer as done by Mailler et al. (2013) to account for the fact that the first EMEP layer is relatively thick (92 m). The resulting matrix attributing emissions to vertical layers is equivalent to that of Mailler et al. (2013), but interpolated to the AQUM vertical grid. Hourly and daily emission factors based on those created by TNO for the MACC project (Denier van der Gon et al. 2011) are also applied to account for the diurnal and weekly cycles of each SNAP source sector.
The use of a comprehensive treatment of gas-phase emissions has a moderate impact on the modelled near-surface concentrations of pollutants. Basically, the new vertical disaggregation of emissions results in somewhat higher effective emission heights compared to a previous configuration which resembles that of Savage et al. (2013). This yields some decreases in the peak NOx concentrations at the surface and consequently a reduction in the number of events with very low O3 concentrations. The combined effect of the new vertical and temporal profiles of emissions also results in improved diurnal cycles for both NOx and O3 when compared to near-surface observations from the UK Automatic Urban and Rural Network (AURN). However, this does not remove all biases from the model.
A number of model simulations, summarised in Table 1, have been conducted to simulate the year 2013 and perform an annual assessment of the impact of emissions related with fracking on air quality over the UK.
Table 1 Summary of the AQUM simulations conducted for 2013 to assess the impact of emissions from hydraulic fracturing including the modifications to total VOC and NOx emissions
The location/extent of the additional emissions was constrained by the locations of shale prospective areas for fracking in the UK identified by the British Geological Survey (Fig. 1). These include (i) the Jurassic Weald Basin in the South East of England (Andrews 2014), (ii) the Carboniferous Bowland-Hodder Shale in the North of England (Andrews 2013) and (iii) the Carboniferous Midland Valley of Scotland (Monaghan 2014). These areas already contain licenced blocks, granted by the Department for Energy and Climate Change (DECC), for the exploration and extraction of oil and gas. Only emissions in the AQUM grid cells that were within these licenced blocks were modified.
Fig. 1 Locations of fracking licences granted in the UK (as of 1 January 2015) and used to constrain the extent and locations for extra emissions (purple polygons). The key areas to note are (i) the Jurassic Weald Basin in the South East of England, (ii) the Carboniferous Bowland-Hodder Shale in the North of England and (iii) the Carboniferous Midland Valley of Scotland. Overlaid are the spatial locations of measurement stations that form the AURN network (black dots)
It is critical to note that as there are no wide-scale fracking plants in operation in the UK at present, the scenarios we have explored with AQUM test the sensitivity of air pollution to potential changes in emissions. Fracking wells themselves are not likely to be allowed (by law) to become major sources of VOC emissions in the UK. The DECC and other government agencies have acknowledged that the major risk to air pollution is in the associated surface-based activities that rely on heavy-duty internal combustion engines (particularly diesel fuelled engines which are known to have significant impacts on human health (IARC 2012)).
As such, we took the grid square in the UK AQUM emission data sets (see the ‘The AQUM model’ section for details of the sources of emissions data) with the highest emissions of VOC and NOx from SNAP sector 5 (extraction and distribution of fossil fuels and geothermal energy) or SNAP sector 7 (road transport) for 2012 as a starting point under the premise that additional emissions from fracking-related activity could potentially mirror the current UK emission situation. These emissions were then added to each of the grid squares in the model where shale gas extraction licences have been granted (Fig. 1) to create scenario 1 (Table 1). The NOx and VOC emissions were then further increased through scenarios 2–4 to span a wide range of potential increases in VOC and NOx emissions (following Edwards et al. (2014)) to determine more about the role of NOx limitation and VOC limitation on potential O3 increases (Sillman 1999). These increments to the emissions in AQUM have been added to SNAP sector 5 and have the appropriate temporal and vertical profiling applied (see the ‘The AQUM model’ section).
An important question to ask is whether or not these emission scenarios are realistic. Given the uncertainty present, it is impossible to answer this with a yes or a no. Our aim is to span plausible to potential worst-case scenarios (outline in Table 1). Vengosh et al. (2017) analysed over 100 papers published since 2013 on the topic of unconventional hydrocarbon extraction. Despite this large number of papers on the topic, very few published studies exist which quantify NMVOC and NOx emission rates. By far, the majority of emission-based studies have focused on CH4 which is not so important for air quality.
The analysis of these papers highlights large basin-basin variability which causes huge uncertainty in trying to estimate emissions from a new basin. Marrero et al. (2016) provide one of the only alternative studies to those in Uintah cited by Edwards et al. (2014), to evaluate how plausible our emission scenarios could be. Marrero et al. (2016) used whole air samples, collected during October 2013 in the Barnett Shale deposit in northern Texas and found that NMVOC emission rates were greater than 600–1000 kg h−1. Our extra NMVOC emissions range from 88 to 373 kg h−1 (per grid box) and so are of similar magnitude to those observed in the Barnett Shale deposit.
NO2 and O3 are key pollutants included in the calculation of the Daily Air Quality Index (DAQI) for the UK (Connolly et al. 2013; Neal et al. 2014). Since the emission scenarios considered here only include additional sources for NOx and NMVOCs which are implemented in AQUM, the main impact will be on the modelled simulation of surface NO2 and O3. The impact on the surface concentrations of PM2.5 is expected to be minor because the emissions of aerosol precursors in CLASSIC have not been altered. Note, however, that the change in the atmospheric concentrations of oxidants (O3, OH, HO2, H2O2 and HNO3) within the RAQ chemistry scheme of AQUM—resulting from the photochemical processing of the additional sources related to fracking—will have an impact on the formation of ammonium sulphate and ammonium nitrate in CLASSIC. Therefore, only a lower limit of the effect of fracking emissions on PM2.5 can be assessed with the current experimental set-up.
The main focus of our work here is on NO2 and O3 and we omit assessing the impacts on PM2.5, but note that it is likely that increases in PM2.5 emissions (especially from the use of heavy diesel machinery (Hasheminassab et al. 2014)) would have significant health impacts.
Here, we present an analysis of hourly mixing ratios of gas-phase species simulated at the lowest model layer in AQUM. The calculation of the DAQI treats each pollutant with its own average period and concentration threshold levels, derived from health impact studies (Connolly et al. 2013; Neal et al. 2014). Following those indications, the following analysis focuses on the maximum daily 8-h running average ozone (MDA8 O3) and the maximum daily 1-h average NO2. Given the lack of changes in emissions of PM2.5 and consequently small changes in their abundance, we neglect any further analysis of these species.
Figure 2 shows a comparison of the results from the year 2013 control simulation against observations of NO2 and O3 at a selection of rural and suburban locations across the UK (see panels for number of sites included in the analysis). Greater than 80% of all simulated mixing ratios of NO2 and O3 in the control simulation are within a factor of two of the observations considering hourly data over the full year. The average correlation coefficients for the comparison of modelled against observed hourly NO2 and MDA8 O3 are 0.62 and 0.64, respectively, across all sites over the full year.
Fig. 2 Comparison of observed (dark grey) and simulated (orange and red from the control run) monthly averaged O3 and NO2 mixing ratios across rural and remote and urban background and suburban measurement stations in the AURN network. The shaded areas and vertical bars represent the standard deviations of the hourly observations and the hourly simulated values for each month, respectively
There is a slight positive bias in modelled MDA8 O3 (average of the mean normalised bias error (MNBE) over all stations ~ 0.04 ppb) and a negative bias in maximum daily 1-h average NO2 (average MNBE over all stations ~ − 0.28 ppb), but in general, the comparison with the AURN observations gives us good confidence that AQUM performs reasonably well over the period investigated. A more detailed evaluation of AQUM has been documented by Savage et al. (2013) and the inclusion of the temporal and vertical emission scaling (discussed in the ‘The AQUM model’ section) moderately improves the fit to observations over previous versions of the model (not shown). A more comprehensive evaluation of the performance of the control simulation against surface observations across the UK is given in the online supplementary information.
The impacts of including emissions related to the processes associated with fracking, as explored through scenarios 1–4 (Table 1), are shown in Fig. 3 and Table 2 for O3 and Fig. 4 and Table 3 for NO2. Tables 2 and 3 detail the summary statistics from the analysis for the changes in O3 and NO2, respectively, for every month of our simulation for 2013. Table 2 highlights that in general, the largest changes in the area averages for O3 occur in the summer months (June and July).
Fig. 3 Monthly mean MDA8 surface O3 in the control simulation (a) and simulated differences for the inclusion of fracking-related activity emissions in scenarios 1–4 (b–e) for June 2013. The differences are calculated as scenario X − control simulation. a MDA8 O3 levels simulated by the model for June 2013 are lowest in regions of large shipping activity in the English Channel and generally higher over land—in particular over Northwest France. b–d the inclusion of additional VOCs and NOx emissions in the regions where fracking licences have been granted results in local decreases in O3 (owing to titration) but varying levels of downwind increases in O3. The largest increases in O3 are found for scenario 4 (e) where the monthly mean ∆MDA8 O3 increases by 0.3 ppb on average over the domain shown in the figure
In Figs. 3 and 4, panel a displays the monthly mean MDA8 O3 and monthly mean daily 1-h maximum NO2 for June 2013 in the control simulation, respectively. Whilst the introduction of the fracking-related emissions leads to changes in O3 and NO2 year-round (as shown in Tables 2 and 3), photochemical secondary oxidant production is usually greatest in the summer months, hence the focus of our analysis of the June results. We restrict the geographical focus of our work to areas of the UK, as emissions from the UK tend to have negligible impacts on O3/NO2 in mainland Europe.
Table 3 Results for changes in monthly mean 1-h daily mean maximum NO2 mixing ratio (ppb) between control simulation and scenarios for each month in 2013. The results show the area averaged mean change across the study area (avg) (11° W–6° E, 49° N–61° N) as well as the maximum (max) and minimum (min) changes in any grid box. The implementation of large increases in NOx emissions in scenario 2 leads to large maximum increases in NO2 and the largest area wide average changes in NO2
In general, there is a degree of negative correlation in the spatial distribution of O3 and NO2—where NO2 is high, O3 is low and vice versa. This pattern is well explained through the reaction of NO with O3, which is a primary mechanism for the production of NO2 and accounts for a process termed O3 titration in urban areas. Figure 3a highlights that O3 titration takes place in the English Channel whereas MDA8 O3 shows highest levels over suburban areas in Northern France and Southern England. Panels b–e in Fig. 3 show the difference in monthly mean MDA8 (∆MDA8) for June 2013 between the fracking activity scenarios and the control simulation shown in panel a.
Panel b displays ∆MDA8 between scenario 1 and the control simulation. Over the domain, there is an average increase in MDA8 O3 of 0.1 ppb, but larger local decreases in the regions where fracking activity emissions were included (a maximum decrease of − 1.9 ppb). Panel c highlights that the impacts of scenario 2 are large local decreases in O3 in the grid squares where emission perturbations are applied (up to − 6 ppb), relatively small decreases in the surrounding cells and a smaller regional increase in O3 elsewhere (up to 1.5 ppb) which results in effectively no change in the net area averaged ∆MDA8 O3. Scenario 2 included much larger perturbations to the NOx and VOC emissions than scenario 1.