Domain Editor-in-Chief: Detlev Helmig; Institute of Alpine and Arctic Research, University of Colorado Boulder, US
Associate Editor: Stefan Schwietzke; University of Colorado Boulder, US


Introduction

Minimizing methane (CH4) emissions from the global oil and gas system represents a significant opportunity to reduce the near-term climatic impacts of this short-lived climate pollutant (Bousquet et al., 2006; Alvarez et al., 2012; Shoemaker et al., 2013; Schwietzke et al., 2016). In contrast with other major CH4 anthropogenic sources (e.g. agriculture), the oil and gas system is more physically concentrated and the number of actors is also relatively limited, facilitating the implementation of emission mitigation strategies.

Over the last few years, an extensive body of research has reduced the uncertainty in CH4 emissions from the oil and gas supply chain in the US (Allen et al., 2013; Miller et al., 2013; Brandt et al., 2014; Lamb et al., 2015; Mitchell et al., 2015; Zavala-Araiza et al., 2015; Zimmerle et al., 2015; Conley et al., 2016). With limited empirically-derived emissions data across other global regions, it becomes crucial to study major sources of CH4 emissions within the oil and gas system.

Globally, oil and gas contributes about 22% to anthropogenic methane emissions (Saunois et al., 2016). In Europe, the share of the oil and gas sector is roughly 10% (Saunois et al., 2016), and for the 28 European Union (EU) member states, it is estimated at 7% (EEA, 2017). Zooming in on The Netherlands, the contribution reaches roughly 3% (Coenen et al., 2017).

In the case of Europe, it is important to assess accuracy of inventories and identify mitigation opportunities. The Netherlands is the largest natural gas (NG) producer among the EU-28 and the second largest in Europe after Norway (Eurostat, 2017). However, The Netherlands ranks as number nine in the EU-28 for CH4 emissions from the natural gas sector. Given its small size, relative short distances and well-maintained infrastructure, low emission values for the NG distribution network are not surprising, but its production-related emissions are remarkably low compared to its peers.

Currently available inventory-based data suggest that a large percentage of EU oil and gas methane emissions come from only a handful of few countries (see emissions by country in SM Text S1, Figure S1). Targeted policies in these countries could dramatically reduce EU oil and gas industry methane emissions. For example, Romania could focus on its production related emissions while Germany, Italy, UK and Poland also have an opportunity in reducing emissions from natural gas distribution systems. Field studies on methane emissions would help refine EU understanding of the magnitude of emissions, and their distribution across countries. In this light, it is also critical to understand why the emissions from the two largest producers (i.e., Norway and The Netherlands) are so low. The current paper aims to address this issue for the Netherlands.

The Groningen gas field consists of a limited number of production clusters (~30) with a single operator responsible for all the gas extraction that happens in the field. Each production cluster consists of 8–12 wells with processing and storage of gas happening on site. In addition to NG, water and condensate are also produced (Whaley, 2009; NAM, 2013; NAM, 2016). This clustering is striking in comparison with heavily dispersed NG systems in other regions. For example, US NG fields like the Barnett Shale (North Central Texas) include ~20,000 production sites with ~26,000 wells from multiple operators, with additional separate locations for gathering, processing, and storage facilities.

The goal of this project is to quantify the magnitude of oil and gas methane emissions in and around Groningen, Netherlands – home to one of the major natural gas producing fields in Europe – and compare it to inventory-based estimates. This multi-scale campaign includes inventory analysis (estimate of inventory-based emissions, spatial distributions of methane sources in the region), ground-based measurements (measurements of production clusters and other oil and gas (O&G) facilities) and airborne measurements (regional mass-balance and large facility emissions). The coordinated campaign took place between August 28th and September 1st 2016.

A broad study region is shown in Figure 1, based upon flight paths described later. It includes most production clusters in the Groningen field and traditional wellpads near the city of Drachten. The flight path excludes a few of the Groningen field production clusters on its eastern border. This polygon covers about 6000 km2, 85% of which is land and 15% of which is the Waddensea, a shallow coastal part of the North Sea. The footprint includes portions of three Dutch provinces: Groningen, Friesland, and Drenthe. Note that the city, the O&G and play the province are all called “Groningen”.

Figure 1 

The study region (blue polygon). Several cities are noted (white). Natural gas facility locations (production clusters, compressor stations, production sites) are shown as orange triangles, with production clusters in the Groningen field colored in red. DOI: https://doi.org/10.1525/elementa.308.f1

In this paper, we first outline the methodology involved in the inventory, ground-based and aircraft-based results.

Methods

Inventory

The Netherlands maintains a Pollutant Release & Transfer Register (PRTR) dating back to 1974. Results serve to underpin the national environmental policy and provide data for the many environmental reports to international organizations such as the European Union and the United Nations. The CH4 emissions from the oil and gas industry as well as all other anthropogenic sources in the Netherlands are reported annually in the National Inventory Report (NIR) for the Kyoto Protocol. Emissions are also directly available by source sector or subsector from the PRTR website (Government of the Netherlands, 2017).

The spatial distribution of methane emissions is also available from the PRTR (Government of the Netherlands, 2017). These spatially-explicit emissions were used in this study to calculate estimated emissions in the study area. The methodology for the source categories of landfills, livestock, manure, waste water treatment plants and oil and gas production are described in further detail in SM Text S1, Section 1.2.

Comparisons between inventory emissions and measured emissions were also done on a site-by-site basis. In several cases, facility-specific data was not available, and an aggregated emission factor was downscaled in order to estimate a single facility’s emissions. A description of this procedure for each type of measured facility appears in SM Text S1, Section 1.3.

Aircraft measurements

Local and regional-scale flights were carried out in order to quantify total regional and individual source emissions of methane. The Metair-DIMO aircraft (Neininger et al., 2001; Neininger, 2004; Hiller et al., 2014; Metair, 2017b) supported a suite of instruments including a Los Gatos methane (CH4) monitor (Hiller et al., 2014) and additional monitors and sensors for carbon dioxide (CO2, modified LiCOR LI-7500), carbon monoxide (CO, AeroLaser) (Schmitgen et al., 2004) and aerosols. Nitrogen dioxide (NO2) and five other nitrogen-containing species were measured with a custom system based on converters and chemiluminescense of Luminol (Dommen et al., 2000; Metair, 2017a). Meteorological parameters were also measured, including 3-dimensional wind (Neininger et al., 2001; Neininger, 2004; Krings et al., 2016).

Four flights were completed between August 28th and September 1st, 2016. Individual source quantification was performed on August 29th and August 30th. Measurements on August 31st focused on additional sources near the northern coast as well regional measurements that were used for the attribution of methane emissions (oil and gas vs biogenic influence). A mass-balance experiment was conducted on the final September 1st flight, when winds were consistent and from the north-west. Point sources were also targeted on this day and an offshore plume was encountered and quantified. The dual-purpose nature of this flight is the reason that the eastern border excludes a few of the Groningen field production clusters. The methodologies used here to determine emissions from airborne concentration and wind measurements draw upon previously reported methodologies.

Individual source quantification from the aircraft was accomplished using downwind transects of facilities and a previously developed methodology (Bovensmann et al., 2014; Krings et al., 2016). A simplified method was also used (SM Text S1, Section 2.1). These methods are similar to work by Cambaliza et al. (2017), Gordon et al. (2015), Lavoie et al. (2015) and Caulton et al. (2014).

The September 1st flight was used to calculate a regional mass-balance for the study region (Figure 1). Numerous previous studies have performed regional mass-balance experiments using analogous methodology (Karion et al., 2015; Peischl et al., 2015; Heimburger et al., 2017; Johnson et al., 2017; Smith et al., 2017). Methane fluxes were calculated by multiplying background-subtracted CH4 mole fraction (nmol mol–1), wind speed (m s–1) and dry air density (mol m–3 kg mol–1). The result is a flux along the wind per square meter perpendicular to it (kg hr–1 m–2). This was then multiplied by a given cross-sectional area A to yield the total flux through a cross-section (kg hr–1). Here, A = L · z, where L is the projection of a given cross-section perpendicular to the mean wind direction, and z is the height of the boundary layer. Vertical mixing was assumed. Vertical profiles for the September 1st data are shown in SM Text S1, Section 2.2.1.

Four different flux estimates were made by varying how averaging was done, using either fluxes calculated from the 5 Hz data points, to include turbulent contributions, or fluxes averaged for entire transects and proportionately scaled to the cross-wind projection (L). These are described in further detail in SM Text S1, Section 2.2.1. These methods should yield consistent results when data coverage over each cross-section is good, but vary when data coverage is poor.

Consistent winds are crucial for mass-balance experiments. The wind during the September 1st flight was persistently from the north-west (300 degrees). This was further verified by examining ground-level weather station data for the entire day, including the hours preceding takeoff, at six separate stations inland, along the coast and at an offshore helipad (SM Text S1, Section 2.2).

The total regional mass-balance was computed using the difference between upwind and downwind cross-sections. Fluxes for individual portions of these upwind and downwind legs were calculated separately, with summary data shown in SM Text S1, 2.2.3. This study’s one mass balance flight was used to determine a regional flux, with accuracy limited by the available data.

Ground-based measurements

The miniature Aerodyne Mobile Laboratory (minAML) (Yacovitch et al., 2017) supported a suite of meteorological and analytical equipment. The minAML was fully equipped with instruments in the US, and the entire vehicle was put in a shipping container and transported to Europe for measurements.

Ground-based measurements were conducted by the minAML for 7 measurement days (August 4–5 and August 28–September 1st, 2016) in and around the Groningen field, and at facilities to the West, near the town of Drachten.

Three Tunable Infrared Laser Direct Absorption Spectroscopy (TILDAS) trace-gas monitors from Aerodyne Research, Inc. (McManus et al., 2015) performed gas phase measurements of the ambient air. These were all single laser mini-TILDAS instruments, measuring 1) nitrous oxide, carbon monoxide and water (N2O, CO, and H2O); 2) CH4, N2O and ethyne (C2H2 also known as acetylene) and 3) C2H6. A LI-COR analyzer measured CO2. An inlet line extended out on a boom mounted to the roof for continuous sampling of ambient air. Every 15 minutes, clean air (ultra-zero air, hydrocarbon-free) was delivered in excess of the intake flow. These gas additions served to spectroscopically background the ethane TILDAS instrument and check zero values for the other instruments. The TILDAS instruments were operated in series at pressures between 30–50 Torr, with an upstream pressure controller managing flow from a downstream scroll pump (Agilent TriScroll TM 600).

Meteorological conditions and GPS-based positioning were measured near the inlet line opening and rooftop. A Hemisphere (V103) GPS Compass was operated in conjunction with an AIRMAR 200WX WeatherStation to determine wind speed and direction and to map vehicle location and bearing during measurements.

Real-time data was logged and displayed on a monitor in the minAML, allowing scientists to rapidly detect and follow plumes of interest. Notes were recorded on the same computer and the observer defined periods of trace-gas data showing enhancements above background (plumes) while in the field.

Instrument calibrations for CH4, CO2 and CO were performed by overblow of the inlet with an ambient air standard, calibrated at the Centre for Isotope Research (CIO), University of Groningen. Ethane instrument calibrations were performed before and after the campaign by precision blending of a ppm-level ethane standard with a diluent, ultra-zero air.

The minAML focused on downwind measurements of individual facilities. Prior to the start of the measurements, the sampling team had a spatially-explicit list of oil and gas facilities in the Groningen field and in nearby areas (e.g., production clusters, compressor stations, production sites outside of the Groningen gas field). A target facility was chosen based on wind direction and road access. The minAML then drove downwind of this facility, looking for enhancements in CH4. The 1-Hz measurement of C2H6 was crucial in distinguishing true NG facility emissions (which contain C2H6) from nearby biogenic emissions (which do not). In this rural region dominated by agriculture, cattle operations and peppered with canals, most NG sites had nearby biogenic CH4 sources.

In Figure 2 below, an example ground-based measurement shows the importance of C2H6 measurements in identifying NG CH4 sources. In this segment of data, three methane enhancements of similar magnitude are apparent (Panel B). Only one of them has co-emitted C2H6, and is attributable to the NG facility (blue box) based on its characteristic C2H6/CH4 ratio of 4.95% (m = 0.0497, Panel C). Other auxiliary tracers on the minAML help further identify interfering signals. Here, a small C2H6 enhancement from vehicle exhaust is indicated by enhancements in CO, CO2. Other species were also measured during the campaign and were useful in identifying agricultural CH4 sources (N2O) or combustion (C2H2).

Figure 2 

Sample ground measurement plume. Panel A shows the path driven by the mobile lab (NE to SW) colored by increasing C2H6 concentration (black to pink). CH4 concentration over this path is also shown (black to yellow), offset for clarity. Wind barbs (white) point into the wind and towards the facility in question (Site 33, blue box). Panel B shows measured concentrations of CH4, C2H6, CO2 and CO over this same path. The data relevant to the NG facility is shown in the blue box. Panel C shows a plot of C2H6 vs. CH4 concentrations and the resulting C2H6/CH4 ratios (m). The data from the NG facility (m = 0.0497, data in Panel B blue box) is clearly distinguishable from the other biogenic signals (m ~ 0). DOI: https://doi.org/10.1525/elementa.308.f2

An estimated CH4 emission magnitude for individual sites was computed using Gaussian plume dispersion methods (Turner, 1994). These simplistic Gaussian dispersion methods are quite uncertain, giving at best a factor of 3 error on emission magnitudes (Yacovitch et al., 2015), and likely more (Abdel-Rahman, 2008), and can present additional systematic errors that are hard to quantify. Other, more robust methods such as tracer-release would be preferred for best accuracy. However, these drive-by dispersion methods can be performed without site-access, on public roads, and pose a significant advantage in terms of the reduced need for equipment, release gases, personnel and time. These were all factors in deciding to limit this campaign to the less accurate Gaussian dispersion methods.

A simple single-point-source technique was used for most production sites, with more sophisticated multiple-point-source models used only when required for large facilities or facilities where measurements were conducted very close by. Dispersion calculations were often done directly on the C2H6 plumes, and a separate proximate measurement of the site’s C2H6/CH4 ratio used to infer the CH4 emission. This method allowed for measurement of natural gas site emissions when nearby sources of biogenic CH4 (e.g., cows, farmland) interfered. Descriptions and examples of these analysis techniques are presented in the SM Text S1, Section 3.1.

At a number of production clusters, no emissions were measured despite an appropriate wind vector. At these sites, the measured emission is set at 0 and the upper limit is estimated by performing a Gaussian dispersion calculation on the measured background. This limit of detection does not depend on instrument performance, which are at sub-ppb levels, but rather on real variability in the real ambient baseline around the facility.

Results and Discussion

We begin with an analysis of methane emissions in the Netherlands, based on inventory data, then look at ground-based experimental results for Groningen production clusters, and how these compare to their inventory estimates. Measurements of an offshore plume are shown, and potential sources investigated. A regional aircraft-based flux measurement is then presented, a portion of these methane emissions attributed to oil and gas sources using ethane, and the results compared to inventory estimates for the region.

Inventory CH4 emissions

Total reported anthropogenic methane emissions for 2016 can be derived from the national inventory reports (Coenen et al., 2015; Coenen et al., 2016). The emissions are dominated by agriculture (67%) and waste (18%) followed by energy (12%).

Natural emissions are not included in the detailed emission inventories that support the national inventory reports to UNFCCC (Coenen et al., 2015; Coenen et al., 2016). Though the Netherlands is a relatively wet and flat country, the quantity of natural wetlands is limited because of the intensive land use. Past estimates for natural sources of emissions (van den Born et al., 1991; van Amstel et al., 1993) suggests a share of ~10% in total CH4 emission for wetlands alone, increasing to ~14% for all natural sources, though these estimates are highly uncertain. Dutch emission inventories have been verified using observation data from atmospheric tower measurements over the years (Hollander and Vosbeek, 1996; van der Laan et al., 2009; Röckmann et al., 2016). In all cases, a reasonable agreement with the emission inventories was concluded, suggesting that natural wetland emissions are at least reasonably estimated in Dutch inventories. Some annual variability is likely in both natural and some anthropogenic sources (e.g. agriculture), as has been observed for the greater European continent (Bergamaschi et al., 2017). Such variability has not been explored here.

Inventory data for the oil and gas industry in the Netherlands is managed by a subdivision of the PRTR, The Task force on Energy, Industry and Waste Management (ENINA), with methodology and calculations documented in the report by Dröge et al. (2016). Within the oil and gas industry sector, on-shore emissions make up 16% of the total, with the remainder divided between transport/distribution and off-shore activities. Further breakdown of the oil and gas source types reveals that direct venting of emissions dominates the methane inventory for both on- and off-shore categories. Further breakdown of oil and gas emissions by type and by company is available in SM Text S1, Section 1.6.

Historically, the bulk of Dutch gas production occurs onshore, especially from the Groningen gas field (~70% of total production for the Netherlands in 2013). Recently, the earthquake risks and damages caused by the gas production from the Groningen gas field have received increasing attention. Ultimately, this has led to a reduced production from that field with almost a 50% reduction in 2015 compared to 2013. This resulted in total Dutch gas production dropping by ~38% since 2013. Strikingly, this had little impact on reported CH4 emissions; which were only ~4% lower during the same period. This reflects the fact that 1) reduced on-shore production has no impact on the off-shore venting, which accounts for the bulk (71%) of the reported emissions and 2) on-shore flaring and venting is not proportional to production and may instead be more related to maintenance and safety regulations. Year-to-year trends in oil and gas emissions in The Netherlands are shown in SM Text S1, Section 1.7.

The recent reduction of production from the Groningen gas field has resulted in a leveling out of monthly production for this region, which used to vary seasonally. For example, in August of 2016, the month most representative of the study period, the Groningen gas field produced 7.8% of 2016 total production. This is very close to 1/12th of a year’s production (8.3%). However, production for individual sites varies greatly month-to-month, with some locations having no production in August, and others achieving up to 11–13% of their annual production in August. These site-specific production values were taken into account for comparisons with facility-scale measurements. Production trends by month and by individual Groningen field production cluster are shown in SM Text S1, Sections 1.8 and 1.9, respectively.

Groningen production cluster emissions

In this section, we investigate the ground-based results for individual Groningen gas field production clusters, and compare Gaussian plume estimated emissions to inventory results. These results represent only a subset of potential CH4 sources in the study area shown in Figure 1.

Figure 3, panel A shows the measured CH4 emission estimates (Gaussian plume analysis) plotted against the inventory emissions estimates obtained via the general oil and gas emission factor applied to production (data available in SM Dataset S1). We see only a weak correlation between measurements and inventory (R2 = 0.33). The lack of correlation is not surprising given that the inventory emission factor is industry-wide. Sites in the area surrounding the city of Drachten are also uncorrelated (R2 = 0.17, data in SM Text S1, Figure S16).

Figure 3 

Comparison of production cluster measurements with inventory. A comparison between measured and inventory emissions estimates is shown in panel A, along with the 1:1 unit ratio line. A binary comparison (yes or no estimated emissions) is shown in panel B. The site count for each type of match is written, and the color scale is propagated to the other graphs. Panel C shows a rough map of Groningen production clusters. Those sites with inventory emissions < 0.1 kg hr–1 for August are outlined in yellow. These results cover the eastern-most portion of the study region shown in Figure 1, with its border drawn in blue here. DOI: https://doi.org/10.1525/elementa.308.f3

During the study period, a number of wells were not in use, or had greatly reduced production (yellow outlined sites, Figure 3, Panel C). Since the inventory emissions estimate is based upon applying an emission factor to reported production values, we investigated whether the presence or lack of production predicts observed emissions or lack of emissions. Sites were noted as having no detected emissions only if repeated downwind transects had clear wind from the facility but showed no enhancement in either CH4 or C2H6 (see SM Text S1, Section 3.1). We found that the experimental observation of emissions downwind of sites is not limited to producing sites. The opposite was also true; a producing site may have no experimentally observed emissions. Non-producing sites may still have equipment under pressure (e.g. natural gas storage equipment, wells), which might explain these observations.

The scale of the emissions measurements is also of interest here. The median experimental emission in the Groningen field is 0.27 kg hr–1, with the highest plotted production cluster measured at an emission rate of 4.6 kg hr–1. By contrast, inventory estimates for these sites give a median emission of 0.48 kg hr–1, with a maximum at 1.4 kg hr–1. Both the measured and inventory values are low when compared to production sector emissions in the United States (SM Text S1, Table S14). The comparatively low emissions found at Groningen production clusters suggest that low emitting production sites are feasible. However, further research is required to explore the reasons behind these emission rates. Several factors are worthy of investigation including the features and infrastructure design of the Groningen field (gas low in C2+, use of production clusters, small number of sites in a tight geographic region), a different culture/approach towards equipment leaks, etc. Future work should also explore the use of production clusters are used in other regions and if they have similar CH4 emissions patterns.

As discussed previously, there may be a few reasons for the relative high proportion of sites with no or low production in August. There are monthly fluctuations in production in response to demand, though overall the region produced close to 1/12th of yearly production during the month of August. However, a number of wells in and around Loppersum (NW portion of the Groningen field) have had their production greatly reduced since 2014 due to concerns about induced seismicity in the region (VanTartwijk and Kent, 2016).

In addition to individual production sites, a number of other larger point sources of CH4 were quantified by ground vehicle, by aircraft, or both. The Grijpskerk natural gas storage facility, for example, shows reasonable agreement between ground-based (17–109 kg hr–1), flight-based (93–100 kg hr–1), and inventory (68 kg hr–1) emissions estimates. This facility was also visited on multiple days and shows evidence of variability in emissions between visits. On the other hand, the Uithuizen Gas Plant along with several landfills show significant discrepancies between experimental and inventory emissions estimates. SM Text S1, Sections 5.2 through 5.4 discuss these larger point sources.

Offshore emissions

One notable source, due to the contribution of offshore emissions to the Netherland’s inventory, is a plume observed just off the northern coast of the Netherlands during the September 1st flight (Figure 4).

Figure 4 

Time series, map and CH4 fluxes for offshore plume (Sept 1st 2016). Panel A shows a map of the plume transects with markers colored by altitude and CH4 concentrations colored by size. The dotted line is the plane through which the flux was calculated. Panel B shows a plot of the cross-wind concentration profiles of each plume transect (rainbow colored red to purple in order of acquired data point). Wind barbs (white) point into the wind. Panel C shows CH4 fluxes through the plane in Panel A, backgrounded and averaged in each grid cell. A point in a grid cell indicates a direct measurement; the other cells contain fluxes linearly extrapolated to the top of the boundary layer. DOI: https://doi.org/10.1525/elementa.308.f4

Seven downwind passes at varying altitude were performed for this plume and a flux calculation was done following methodology in Krings et al. (2016). These seven flight legs yield an average emission of 150 ± 50 kg hr–1. This emission magnitude is much larger than a typical production cluster (0.26 kg hr–1 median emission, see SM Text S1, Section 6) and comparable in size to some of the larger oil and gas facilities or landfills measured in this region (SM Text S1, Section 5.2–5.4). Varying choices in the parameters for the flux algorithm, including background choice, are explored in SM Text S1, Section 5.1.1. No ethane concentrations from flask samples collected during these flight legs were successfully quantified, limiting our ability to do source attribution.

Candidate inventory source locations based on the dominant wind direction, are two offshore wells 30 to 40 km away (SM Text S1, Section 5.1.2). No other reasonable local sources were identified. The observed emission magnitude is of the same order of magnitude as would be expected from a normally operating offshore oil and gas facility like an offshore platform. For example, a component-based estimate of emissions from 15 oil platforms in the Gulf of Mexico yields CH4 emission estimates on the order of 224 kg hr–1 (Bylin et al., 2010). Conversely, measurements from a 2012 uncontrolled gas leak from the North Sea Elgin platform found an emission rate of 4,680 ± 72 kg hr–1 (Lee et al., 2017).

The measured CH4 peaks in each plume transect were quite narrow, between 100–300 m wide, and were intercepted at locations that span a width of 2–3 km. These narrow widths cannot be explained by simple Gaussian plume dispersion from sources 30–40 km away, and would instead require filament-like transport of emissions under stable marine boundary layer conditions. An alternate explanation would assign a broad elevated baseline concentration to the offshore source(s) plus an additional unknown local emission event at closer range. This mixed source hypothesis parallels recent studies of a North Sea offshore plume, much bigger in both spatial extent and measured flux, where methane isotope ratios suggested a mixture of oil and gas and other sources (Cain et al., 2017). With the current information currently available no robust conclusions can be drawn. These offshore results highlight the importance of collecting more detailed measurements of offshore oil and gas production facilities, including sampling of upwind and downwind of oil and gas production platforms, as direct venting of CH4 at off-shore facilities dominates (70%) the reported oil and gas CH4 emissions from the Netherlands.

Regional CH4 emissions via mass balance

Measurements of individual facilities are valuable for assessing specific emissions and inventory accuracy for sub-sections of the oil and gas industry. Flight-based mass-balance measurements, however, can provide a broader view of a region’s methane emissions, and assess the accuracy of a regional-scale inventory as a whole.

The September 1st, 2016 flight, where the wind was from the north-west (300 degrees), was used to calculate a regional flux targeting the Groningen gas field. The study area includes the entirety of the polygon region shown in Figure 1. Individual fluxes are calculated through the flight legs defining the north and west border (inflow, 11.5–15.3 Mg hr–1) and the south and east border (outflow, 22.1–22.3 Mg hr–1). The CH4 flux for the entire region defined is taken as the difference between the inflow and outflow transects: 8 ± 2 Mg hr–1.

The inflow and outflow transects were flown in sequence, with the middle portions (Figure 5, cross-sections 5–6 and 7–8) flown earlier in the day. These middle cross-sections were not used in regional flux results. Their original purpose was for vertical profiles and individual source characterization (e.g., Uithuizen gas plant, see SM Text S1, Section 5.3).

Figure 5 

September 1st 2016 mass-balance flight. Panel A shows time traces for measured CH4 and altitude. Panel B shows a map of the mass balance flight, with data points colored and sized by methane concentration. A dotted arrow is drawn along the dominant wind direction (300°), with measured wind (white barbs) also shown along the flight path. Reference points labeled 1 through 16 are shown in both panels, and are used to define cross-sections (blue lines, Panel B). DOI: https://doi.org/10.1525/elementa.308.f5

The top of the boundary layer was clearly visible in the vertical transect done mid-flight along the coast at 700 meters above mean sea level (mAMSL) (SM Text S1, Figure S9) and was used for cross-sections 1–2 and 3–4. Two vertical profiles over land were taken (SM Text S1, Figure S10–S11), with the first done early in the flight (cross-section 5–6) showing a slow transition out of the boundary layer at approximately 700 mAMSL, with possible impact from a residual layer. The other end-of-day profile (cross-section 9–10) shows an approximate 850 mAMSL boundary layer height, with considerably more heterogeneity in the mixed layer. For this reason, an approximate boundary layer height of 800 mAMSL was used for the transects over land (cross-sections 9–10, 11–12 and 15–16).

The indicated range for the total flux and transect fluxes reflect the standard deviation between four separate methods of calculating the flux (see SM Text S1, Section 2.2). This ±2 Mg hr–1 does not reflect other possible expected sources of error or bias, such as effects of sampling any residual layer, or the lack of data at additional altitudes.

The lack of clear boundary layer structure in the first vertical transect of the day suggests that enhanced concentrations inland could also originate from the residual layer from August 31st. Furthermore, for this study, the background was chosen as the methane concentration above the mixing layer. This choice means that emissions are potentially overestimated. For both of these reasons, the mass-balance methane fluxe should be regarded as an upper limit on the emissions from the study region.

Back-trajectories were performed for the September 1st flights (SM Text S1, Figure S13). They show a clear general pattern: the coastal air originates at high altitude, subsiding in the high-pressure region over the North Sea; air in the south, on the other hand, originated closer to the surface, and further south, near London. This again suggests that UK CH4 background emissions could be influencing the outflow transect more than the inflow transect, and that the 8 Mg hr–1 regional estimate is an upper limit.

The observations from this campaign suggests several improvements to the flight plan for future more robust regional measurements. Better flux estimates could be achieved with a flight plan specifically adapted to conducting regional mass-balance observations (as opposed to dividing flight time between regional & point source estimates, as was done here). Additionally, transects ordered to follow the air mass (Lagrangian sampling) could be performed, including measurements at least two different altitudes and a sounding above the mixed layer.

Attribution of emissions

Natural gas contains trace amounts of C2H6, while biogenic CH4 emissions from cattle, agriculture, landfills or wetlands contain no C2H6. This distinction allowed us to separate the total emissions determined from the mass balance flights into two source categories: fossil and biogenic. Whole air flasks taken aboard the aircraft were analyzed for their C2H6/CH4 ratio. Results from flasks collected on an August 31st flight are used here to determine the regional C2H6/CH4 signature (see SM Text S1, Section 4).

The best characterization of the downwind plume for the Groningen field is given by the average of flasks taken downwind and within the gas field. These seven flasks had C2H6/CH4 enhancement ratios (background-subtracted) between 0.21% and 1.19% (mol/mol%), with an average of 0.630.63+0.98%. Error bars are defined at 95% confidence, and are calculated from the standard deviation and the Student’s T, with the lower limit clipped to reflect the fact that C2H6/CH4 ratios cannot be negative. Ethane/methane ratios of the oil and gas production clusters in the region were typically between 2 and 4%, suggesting that the emissions in this region are dominated by biogenic sources.

The production clusters sampled by the ground-based team had C2H6/CH4 enhancement ratios of 3.23% (peak of the Gaussian fit to the histogram, SM Text S1, Section 4.1). The 95% confidence limits of the fit Gaussian distribution occur at ± 1.96σ, yielding a ratio of 3.23 ± 0.85%, or, equivalently, C2H6/CH4 ratios limits of 2.38% and 4.08%.

Given an C2H6/CH4 fossil source signature of 3.23 ± 0.83%, a biogenic signature of 0%, and a regional signature of 0.630.63+0.98%, the apportionment of CH4 emissions for the Groningen region would be 2020+31%oil and gas;8031+20%agricultural/biogenic. Limits to this apportionment are calculated to first order by considering the uncertainty in the oil and gas source distribution propagated with uncertainty in the flask signature measurement, with 95% confidence, and clipped to prevent negative contributions. The variation in the measured flask ethane/methane ratios dominates this uncertainty. Not included in this uncertainty is the impact of background choice and flask choice in the average. If higher background concentrations were chosen, for example using some low-concentration boundary-layer flasks measured in the study region, the oil and gas contribution would climb. Conversely, if only a limited number of the measured flasks were used in calculating the average, the oil and gas contribution could fall or climb. All of these alternate scenarios fall within the error bounds of the above determination. SM Text S1, Section 4.2 outlines the procedure used to choose flasks and background values.

The impact of combustion sources within the city of Groningen was assessed based on measurements of NO2 and NOx from the aircraft. No significant emissions were noted during mass-balance flights, reflecting the dominating impact of rural emissions in these flights.

We consider leaks in the distribution sector (e.g. leaks from residential natural gas piping) to be included in the estimate for fossil sources. The reason for this is that natural gas in the distribution network will mirror the C2H6/CH4 composition measured at the production clusters. In fact, since production clusters in the Netherlands are used to store distribution-ready natural gas, no significant difference in the C2H6/CH4 ratio is expected between production cluster storage and distribution.

Indeed, the natural gas composition and heat content for the City of Groningen is routinely tracked, with gas analyses performed every 15 minutes. For the mass-balance flight dates of August 31st and Sept 1st 2016, the city gas composition averaged 82% CH4 and 3.0% C2H6 (remainder mostly N2, see SM Text S1, Section 6.2 for full gas composition). The C2H6 and CH4 contents can be directly divided, and their errors summed in quadrature to yield the expected C2H6/CH4 ratio of city gas of 3.7 ± 0.04%. This falls within the measured Groningen field C2H6/CH4 ratio of 3.2 ± 0.82%.

Future work on apportionment could include estimating which well sites fall within the regional footprint for the Groningen field methane flux, and obtaining a larger sample of C2H6/CH4 source signatures for apportionment. Data from 3 wellpad sites to the west near Drachten (outside of the Groningen gas field) show comparable C2H6/CH4 signatures (2.0–3.7%) and do not change the result (see SM Text S1, Section 4.3). However, there are a number of non-production cluster sites (e.g., compressor stations, processing plants, condensate storage) that have been measured with varying C2H6 content. These sites could also be included into the C2H6/CH4 source signatures.

Regional comparison with inventory

To compare the flight data with the inventory, CH4 emissions were aggregated from the high-resolution emission grids (SM Text S1, Section 1.2) over the study region (Figure 1, blue polygon). Table 1 shows the breakdown of emissions for the Netherlands by general inventory sector, including agriculture, oil and gas extraction, and “other”, which is dominated by landfills and waste water treatment plants. Gridded inventory data for natural sources of CH4 emissions, such as wetlands, are not available for regional estimates; previous studies (van den Born et al., 1991; van Amstel et al., 1993) have found ~14 Mg hr–1 nationally (SM Text S1, Table S2). Agriculture emissions overwhelmingly dominated the emissions according to the inventory data. Oil and gas is a relatively small component of the inventory, with only 2% of the total national oil and gas emissions thought to be emitted in this region. This is in line with the assumption that the bulk of the oil and gas extraction emissions is due to offshore venting. Overall, our study region would be responsible for about 16% of total national methane emissions excluding natural emissions from wetlands and water bodies.

Table 1

CH4 emissions and fossil/biogenic ratios for the Groningen study region. Total inventory CH4 emissions in megagrams per hour (Mg hr1, where 1 Mg = 1000 kg = 1 metric ton) are listed, with a breakdown by sector: agricultural (agri), oil & gas and other. Mass balance CH4 emissions from flight measurements are shown (observed). The apportionment of CH4 emissions to fossil sources is also shown, as a % of the total. DOI: https://doi.org/10.1525/elementa.308.t1

CH4 Emissions (Mg hr–1) Fossil CH4 Emissions Apportionment (proportion of total, %)

Inventory, by sector Observed Inventoryb Observedc

agri oil & gas othera total total

10.9 0.26 2.9 14 8 ± 2 1.9 2020+31

a Dominated by landfills and waste water treatment plants.

b Oil & gas sector vs. inventory total.

c Experimental apportionment uses ethane/methane ratios. Uncertainties are expressed at 95% confidence and are asymmetric. See text.

Comparisons of inventory and observational results are shown in Table 1. Total observed methane emissions were lower than estimated in the inventory (8 ± 2 observed vs. 14 Mg hr–1 inventory). However, the proportion of these observed methane emissions from fossil sources is an order of magnitude higher than estimated in the inventory (2020+31% observed vs. 1.9% inventory). The large uncertainty in the observed emissions attributed to fossil sources is driven by the variability in measured ethane/methane ratios in aircraft flasks. The uncertainty of the inventory emissions estimates is unknown.

The observed proportion of CH4 emissions from oil and gas is sensitive to flask choice and background concentration determinations; on the inventory side, a number of factors can greatly influence results. The use of downscaled oil and gas emission factors is one example, where national-scale emission factors are used with production volumes as the independent variable to estimate the regional emissions. Indeed, in this region, the assumption that emissions scale with production is contradicted by ground-based emissions estimates at production clusters. Many assumptions were also required in preparing the biogenic portion of the inventory, and will significantly impact the final result. Livestock and manure methane emissions, for example, were estimated based on methane/ammonia emission ratios.

There is further uncertainty in the footprint of the flight-based data, including the footprint of the flask-based sampling for airborne ethane. As previously discussed, the mass-balance results themselves are subject to additional and unquantified uncertainties related to the presence of a residual layer, boundary layer height, and the lack of data at several altitudes.

The regional mass-balance was done on a single measurement day, making these emissions estimates a snapshot in time. Conversely, the emission inventory is largely constructed from yearly national estimates. These uncertainties and unknowns suggest the need for additional measurements and research, particularly focusing on regional emissions apportionment, and inventory refinement.

Conclusions

This study presents a multi-scale pilot study of methane emissions in the Groningen region of the Netherlands. This region is home to “production cluster” sites, which combine extraction, processing and storage all within the same facility footprint. Ground-based measurements using simple dispersion methodology find low emission rates compared to production facilities in the United States. Further investigation is required to understand the underlying causes of these comparatively low emission rates, for example, investigation of the characteristics of the Groningen field (gas low in C2+, production cluster type infrastructure, size and density of the region), equipment leak detection and repair strategies, etc. Future work should also explore if similar production clusters are used in other global production regions and if they have similar CH4 emissions patterns.

Inventory estimates reveal that production volume and aggregated emission factors do a poor job of predicting emissions at these sites; indeed, we find that sites with no production still had emissions, and vice-versa. Biogenic methane sources in the studied region (cattle, bodies of water, agricultural fields) could easily have been confused for emissions from oil and gas sources without the continuous measurement of ethane.

Biogenic sources dominate the methane emissions from the study area. Aircraft mass balance yields total regional methane emissions of 8 ± 2 Mg hr–1, lower than the 14 Mg hr–1 predicted by inventory. Experimental apportionment of these methane emissions, however, reveals that 20 ± 31% of the observed methane emissions stem from fossil sources, ten times higher than the 1.9% indicated in the inventory. While the central value for this oil and gas proportional contribution is much larger than expected, the large variance in ethane/methane ratios in aircraft flasks means that the confidence limits of the observed result include the predicted value from the inventory. The results reported here suggest the need for additional regional studies of methane emissions in the Netherlands if the uncertainties in emissions are to be reduced and the sources of emissions effectively resolved.

The dominance of biogenic methane emissions (e.g. agriculture, wetlands, cattle) in the Netherlands underscores the importance of understanding and quantifying emissions from a diversity of sources including agriculture, livestock operations and natural wetlands in order to improve inventory source estimates. In oil and gas regions in particular, it is crucial to collect data that can distinguish fossil sources from biogenic sources. Future flight-based and tower-based campaigns should include more ethane measurements in addition to the ongoing isotope studies (Röckmann et al., 2016). Future campaigns should also focus on understanding temporal and spatial variation in emissions, to provide a more integrated assessment of annual methane fluxes.

Data Accessibility Statements

Raw data (mixing ratios, GPS and measured wind for the ground vehicle and aircraft are available from the authors upon request). The following datasets are available.

Text S1, as supplemental material, includes a dataset listing production cluster latitudes, longitudes and site names. Observed methane emission magnitudes along with upper and lower limits are included. These data were collected by the ground vehicle and worked up using Gaussian plume analysis. Inventory-predicted emissions are also included.