The FAAM aircraft manages a modified BAe-146-300 aircraft which carries core and optional instruments for measuring various components of the atmosphere. Core instruments cover a range of basic atmospheric measurements including thermodynamic properties, wind, turbulence, and some chemical species. These are provided by FAAM as part of the facility. Details of most FAAM instruments can be found on the FAAM web-site: http://www.faam.ac.uk (last access: 22 March 2018). Wind and turbulence are measured using a five-port pressure measurement system in the aircraft radome, combined with two scientific static ports located symmetrically on either side of the aircraft. Wind and thermodynamic profiles from the aircraft down to the surface are also provided by dropsondes which can be released and tracked periodically in flight. Of greatest relevance to the work reported here are the systems for fast methane measurement and for obtaining air samples for laboratory analysis; these are described below.

CO2 and CH4 were measured in situ on the aircraft using a modified Los Gatos Research Inc. Off-Axis Integrated Cavity Output Fast Greenhouse Gas Analyser (FGGA model RMT-200). This was calibrated in-flight against gas standards certified by the Max-Planck Institute for Biogeochemistry (Jena) as part of the Infrastructure for Measurements of the European Carbon Cycle project (EU 13 IMECC; see http://imecc.ipsl.jussieu.fr/, last access: 22 March 2018). The stability of these standards was also cross-checked against Royal Holloway laboratory standards. All reported CH4 mixing ratio data are traceable to the National Oceanographic and Atmospheric Administration NOAA-04 scale (Dlugokencky et al., 2005). A technical summary of the FGGA deployed on-board the FAAM aircraft, the calibration system, data analysis and quality control methods developed by the University of Manchester and FAAM is presented elsewhere (O'Shea et al., 2013), illustrating the airborne performance of the system, chiefly a measurement accuracy of ±1.28 ppb with a 1σ precision at 1 Hz of 2.48 ppb for CH4.

Ambient air was sampled using both the automated whole air sampling (WAS) system fitted to the aircraft and manually into Tedlar bags for post-flight laboratory analysis. The WAS system consists of sixty-four silica passivated stainless steel canisters of 3 L internal volume (Thames Restek, Saunderton UK) fitted in packs of 8, 9, and 15 canisters to the rear lower cargo hold of the aircraft. Each pack of canisters was connected to a 3/8 inch outside diameter stainless steel sample line, in turn connected to an all-stainless steel assembly double-headed three phase 400 Hz metal bellows pump (Senior Aerospace, USA). The pump drew air from the port-side ram air sample pipe and pressurized air into individual canisters to a maximum pressure of 3.25 bar, giving a useable sample volume for analysis of up to 9 L. WAS canisters take approximately 20 s to fill at typical boundary layer pressures, thus they provide an averaged measure of hydrocarbon content. At a typical aircraft science speed of around 100 m s-1, a WAS sample is therefore an average mixing ratio over a spatial extent of ∼ 2 km. The length of sampling manifold within the aircraft creates a delay of around 10 s between air entering the inlet at the front of the aircraft and being available for capture in the hold. This slight delay allowed the real-time CH4 outputs from the FGGA to be used to aid the capture plume samples with canisters. The integrated nature of the WAS means that the concentrations reported do not represent peak plume concentrations, however these can be inferred assuming a constant relationship to CH4. The manual Tedlar bag sampling system employed a Metal Bellows pump (model MB-158) and was more direct, with a few seconds lag time and rapid bag filling (∼ 5 s).

Air samples were analysed for volatile organic compounds (VOCs) within 48 h of collection at the University of York using a dual channel gas chromatograph with two flame ionization detectors (Hopkins et al., 2011). 1 L samples of air were withdrawn from the sample canisters and dried using a glass condensation finger held at -30 ∘C. C2–C7 samples were pre-concentrated onto a multi-bed carbon adsorbent trap, consisting of Carboxen 1000 and Carbotrap B (Supelco), held at -20 ∘C and then heated to 325 ∘C at 16 ∘C s-1 and transferred to the GC columns in a stream of helium. The eluent was split in an approximately 50 : 50 ratio between an aluminium oxide (Al2O3, NaSO4 deactivated) porous layer open tubular PLOT column (50 m, 0.53 µm id) for analysis of NMHCs and two LOWOX columns (10 m, 0.53 um id) in series for analysis of polar VOCs. Both columns were supplied by Varian, Netherlands. Peak identification and calibration was made by reference to a part per billion level certified gas standard (National Physical Laboratory, ozone precursors mixture, cylinder number: D64 1613) for NMHCs. This standard and instrument has in turn been evaluated as part of the WMO GAW programme and was within target operating limits.

Methane isotopic composition (δ13CCH4) was measured at Royal Holloway, University of London (RHUL) in samples collected in WAS canisters during flights on 30 March and 3 April and in Tedlar bag samples collected manually on the 3 April flight. Prior to isotopic analysis, the methane mixing ratio in the samples was measured using a Picarro 1301 cavity ring-down spectrometer, calibrated using NOAA air standards. Repeatability in CH4 mixing ratio measurements was ±0.3 ppb. δ13CCH4 was analysed using a modified gas chromatography isotope ratio mass spectrometry (GC-IRMS) system. The methodology is described in detail by (Fisher et al., 2006). δ13CCH4 repeatability was ∼ 0.05 ‰. All isotope measurements were made in triplicate. Isotope ratios are given in δ notation on the VPDB (Vienna Pee Dee Belemnite) scale. Keeling plot methodology is described by (Pataki et al., 2003) and (Fisher et al., 2017).

The FAAM aircraft is equipped with a system to drop radiosondes (VAISALA, Finland). The sondes (RD94) descend on a parachute with a speed of ∼ 10 m s-1 and measure air pressure, air temperature, relative humidity, and GPS position on their way to the surface. Wind speed and wind direction are calculated from the GPS measurements and the known drag of the dropsonde (Wang, 2005). Data can be received and viewed in real time on the aircraft.

The location of the gas source relative to the sea surface and the mass flux of the emission were initially not well known. A prospective analysis of the gas plume was obtained using HYSPLIT model simulations (Stein et al., 2015), carried out using meteorological fields from the US National Centre for Environmental Prediction Global Forecast System (NCEP-GFS; https://ready.arl.noaa.gov/HYSPLIT.php, last access: 22 March 2018), obtained via the Air Resources Laboratory of the National Oceanic and Atmospheric Administration. NCEP GFS data are high resolution (0.5∘ latitude and longitude and 3 h temporally). Figure 2 shows the modelled CH4 concentration from 0–1000 m above sea level, for 12:00 UTC on 2 April 2012. The modelled start of release was 00:00 UTC and the modelled release rate was 23.5 kg s-1. The model outputs were used for flight planning and to provide a safety case for the flights. Given the explosion risk, and because hydrogen sulfide (H2S) in the Elgin reservoir was reported to be ∼ 45 ppm (Fort and Senequier, 2003), close to the safe human exposure limit, a risk reduction analysis was carried out prior to the first BAe-146 research flight to specify the “turn away” concentrations based on real-time measurements on-board the aircraft using hand held sensors. The flights did not enter a 3 nautical mile radius, 4000 ft altitude exclusion zone imposed by the UK Maritime and Coastguard Agency at the time of the emergency. Outside of this excluded volume, a “turn away” detection value of 40 ppm CH4 was established, which was 20× the background concentration, 10× higher than the forecast of CH4 likely to be present (given an unrealistically high leak rate of 23.5 kg s-1 set in the model) and 100× below any possibly dangerously combustible concentration of the worst case gas mixture.

Method 1 is appropriate when there exists no significant temperature inversions at levels where z≤σz. This requires that measurements are made up to a height of at least σz and that no inversions are encountered up to that level. If an inversion layer does exist, then method 1 may still be used if the measured value of σz is such that σz≪H, where H is the mixing layer height.

When writing Eq. (1) as follows: Cx,y,z=Cz(xz)exp⁡-y-y022σy2, where Cz(xz)=C0(x)exp⁡-z22σz2 and C0(x)=qπσyσzU, then Cz and σy may be obtained from fitting cross-plume data at fixed distance downwind to Eq. (6). Then writing Eq. (5) in the following form: ln⁡Cz=ln⁡C0-z22σz2, where C0 and σz can be obtained by plotting Cz against z2 using data from all transect levels at a fixed downwind distance.

This approach is appropriate when the airborne measurements fully define the vertical extent of plume mixing (e.g., Conley et al., 2016), or the plume is mixed thoroughly in the vertical up to a capping inversion (e.g., Ryerson et al., 2012), such that there exists a clear temperature inversion and elevated stable layer in atmospheric profiles revealed using aircraft measured thermodynamic profiles, dropsondes or radiosondes, and crosswind transects show little decrease of concentration with height (within the uncertainties), up to the inversion level.

Assuming conditions are suitable for method 2, then writing Eq. (3) as follows: Cx,y=C0(x)exp⁡-y-y022σy2, where C0(x)=q2πσyUH. Best fitting of the concentration measurements to Eq. (8) is used to determine C0, y0 and σy and then the leak rate q is determined from Eq. (9), using estimates of the inversion height H from the atmospheric soundings. Note that the C0 here is different to the C0 for Solution method 1.

Either of these methods allow for calculation of the mean emission flow rate in a relatively short time period after measurements are taken (potentially within 24 h). This makes airborne sampling useful for emergencies where fast quantification of flow rate can be critical for informed decision making.

Measurements of CH4 were taken at different heights above sea level and different distances from the platform on each of the flights. Figure 4 shows CH4 mixing ratios taken on each flight, plotted as a function of distance along the flight track perpendicular to the plume for all flights at 5 and 15 NM from the Elgin platform. To aid the analysis, profiles of potential temperature and wind speed and direction from dropsonde data taken early and late in the flights (where available) are shown in Fig. 5. Thermodynamic profiles measured by on-board instrumentation during plume sampling were found to be consistent with the dropsonde data.

Non-methane hydrocarbons (NMHC) and other volatile organic compounds in the plumes were determined from whole-air flask samples by offline analysis (Hopkins et al., 2011). NMHC content was dominated by light alkanes ranging from > 20 ethane to < 1 ppb benzene and < 0.1 ppb higher monoaromatics. NMHCs up to C5 all showed enhancements corresponding to enhanced CH4. However, it is noteworthy that the mixing ratios of the heavier hydrocarbons (propane, butanes and pentanes) all fall systematically more rapidly than those of excess methane and ethane, over the entire range of methane mixing ratio (as shown in Fig. 7). We believe this is likely caused by the heavier weight compounds condensing more readily to the cold water surface along the length of the plume due to their increased solubility. The plume was dominated by short chain (< C5) linear and branched chain alkanes and some larger monoaromatic compounds, with up to five alkyl groups substituents attached to the benzene ring. No polycyclic aromatic compounds or oxygenated species were observed in any of the samples.

The spatial mixing of higher condensate species with background air was highly correlated to CH4 and C2–C5 NMHCs, as expected from emissions from a single point source. Atmospheric measurements showed a lower proportion of > C6 species than in Fort and Sénéquier (2003) for Elgin reservoir fluids (∼ 3 % vs 13 %). We speculate that these larger species condensed as liquids to the relatively cold (7 to 8 ∘C) sea surface rather than being transported in the gas phase into the air plume. The corollary is that the NMHC data show no evidence for widespread higher condensate evaporation into air from the seawater sheen, despite reports of significant pollution risks, including condensates from underwater release. This would suggest condensate removal was by biological processes in the water, or simply due to cold surface water, decreasing the evaporation rate to undetectable levels.

NMHC analyses reported here demonstrate that potential fractionation may have occurred as the gas and liquid mix was emitted from the leak, and also that there was likely disproportionation by selective fractionation of volatiles during uptake in the water. Quantification of the gas flux to the atmosphere by taking the ratio to the mass of the condensate sheen, although a useful “first-guess” method, is thus very imprecise. Eventual estimates of condensate mass ranged from approximately 0.04 to 20 t, over an affected area estimated from approximately 15 to 600 km2. This wide range of estimates can potentially hamper a well designed response effort (Ryerson et al., 2012). We emphasize the ability of the airborne chemical data to provide significantly more precise flow rate information than that provided by visual observations alone.

The evidence in the air plume for release of CH4 and C2–C5 alkanes confirmed that the gas leak was not released from a significant depth. Initially it was not clear whether the gas leak was on the wellhead platform, or below sea level, or both. After a Total press statement on 29 March 2012 and updated imagery on 30 March 2012, it became clear that there was indeed a gas leak at the wellhead on the platform. The airborne NMHC evidence supported the inference that release was indeed above sea level.

The height of the release was approximately the same as the aircraft sampling altitude in the lowest sampling crosswind transects. Thus, the aircraft was able to fly through the core of the plume. This contrasts with the early situation in the BP Deepwater Horizon event, where release took place 1.5 km subsurface and CH4 (Camilli et al., 2010; Yvon-Lewis et al., 2011), light alkanes, and light aromatics were essentially completely taken up in the water column (Reddy et al., 2012; Ryerson et al., 2012).

A further key goal of the airborne survey flights was to identify the geologic source of the gas leak using the CH4 isotopic measurements (δ13CCH4) of the gas plume using the Tedlar bag and flask samples collected during the aircraft transects. This technique requires rapid sampling during the brief fly-through. Figure 8 shows δ13CCH4 versus 1/CH4 in air samples from the first two flights, following the Keeling plot methodology of (Pataki et al., 2003). The source gas has δ13CCH4 of -42.3 ± 0.7 ‰ (±2σ) using geometric mean regression and a Monte Carlo style simulation to determine the propagation of errors into the fitting process where a geometric mean regression defines a line whose intercept on the δ13CCH4 axis gives the endmember source value. The similarity of results from plotting separately the data from the two flights implies the gas source did not change between flights.

Compared to the first flight on 30 March 2012, the second flight on 3 April 2012 found significantly weaker plumes, suggesting that the gas source was depleting. This was significant in that it supported the inference that the source was comparatively small and depressurizing: i.e that the gas leak was indeed from a restricted source such as may be found in the Hod Formation and not from the main production depth (Bergerot, 2011). Information released by Total indicated that the main production depth had been plugged prior to the blowout.

The δ13CCH4 isotopic ratio gives direct insight on the source of the gas. δ13CCH4 is related to the fractionation that has occurred because of the thermal history of the geological source of the gas. In very hot deep gas fields, where early-formed biogenic gas may have escaped and later-formed gases include thermogenic methane, CH4 is typically enriched in 13C (i.e. δ13CCH4 is less negative). In contrast, in shallower strata δ13CCH4 is likely to be dominated by early-formed biogenic gas and lighter (i.e. more negative).

The source rocks below the main gas field would have been at 5.5 km depth and at 200 ∘C or more. In contrast, the over-pressured interval in the overlying Hod Formation is at about 4.2 km depth and 165 ∘C (Isaksen, 2004). The gas in the Hod Formation likely formed in situ, trapped by the rock without early leakage of isotopically lighter gas. Thus gas in the Hod Formation will likely be much more negative in δ13CCH4 than gas in the significantly hotter source regions underlying the Elgin field.

Methane isotopic information on the Elgin gas field and the Hod Formation was not available; instead, we estimate these based on published stable isotopic values for ethane (C2H6) from the Elgin field. Isaksen (2004) shows δ13C around -29 to -30 ‰. The data on the oils and ethane from the Elgin field suggest that hydrocarbons from the producing gas reservoir are in equilibrium with the setting (Isaksen, 2004). Under this assumption, and given the relatively high maturity of the field, in the Elgin production gas we expect δ13C in methane to be similar to the δ13C ratio in ethane (Berner and Faber, 1996), perhaps in the range -25 to -35 ‰. If significant methane loss had occurred, or if methane had been introduced from below, we would expect it to be less negative. A δ13CCH4 of -42 ‰ from whole-air samples collected from the gas plume is thus consistent with a source in the shallower, lower temperature Hod Formation, rather than the deeper main Elgin reservoir. Alternatively, a signature of 42 ‰ could be generated by mixing shallow gas with gas from the main reservoir. For future events, it is clear that the techniques described here combined with detailed isotopic analysis from the production field would considerably aid source identification.

In order to assess any wider regional impact of the Elgin incident, HYSPLIT model simulations were carried for each day between 25 March 2012 at 18:00 UTC and 16 May 2012 at 18:00 UTC. For each day, a 72 h dispersion forecast was produced and the concentration at 72 h after initialization was recorded. Then, a time average of these 72 h concentration distributions was produced. Thus, dispersion predictions were produced valid for the period 28 March 18:00 UTC until 19 May 2012 18:00 UTC. Calculations have only been made for CH4 and assume that the methane is long-lived (a lifetime much greater than the 3 day model runs). The source strength was allowed to vary temporally using an interpolated time series from the measured flow rate described earlier. Figure 9 shows HYSPLIT results broken down by week, integrated over all levels and displayed over a domain containing all of Europe. The majority of the CH4 was distributed mainly to the south of the source. Low concentrations of methane (< 1 ppbV) travelled as far as mainland UK (principally the Humber Estuary, the northern Norfolk Coast and the northern Yorkshire coast) and continental Europe (Netherlands). The highest levels of concentration, however, appear to be confined to a rectangular box that extended from 56 to 57∘ N and from 1 to 3∘ E. This confinement is true at all levels. Above approximately 1 km above sea level the concentrations were negligible. There is some evidence of the plume reaching as far south as Switzerland (at very low concentrations) during periods 28 March to 10 April and 9 to 19 May.