Emissions of methane from offshore oil and gas platforms in Southeast Asia

Methane is a substantial contributor to climate change. It also contributes to maintaining the background levels of tropospheric ozone. Among a variety of CH4 sources, current estimates suggest that CH4 emissions from oil and gas processes account for approximately 20% of worldwide anthropogenic emissions. Here, we report on observational evidence of CH4 emissions from offshore oil and gas platforms in Southeast Asia, detected by a highly time-resolved spectroscopic monitoring technique deployed onboard cargo ships of opportunity. We often encountered CH4 plumes originating from operational flaring/venting and fugitive emissions off the coast of the Malay Peninsula and Borneo. Using night-light imagery from satellites, we discovered more offshore platforms in this region than are accounted for in the emission inventory. Our results demonstrate that current knowledge regarding CH4 emissions from offshore platforms in Southeast Asia has considerable uncertainty and therefore, emission inventories used for modeling and assessment need to be re-examined.

Methane is a substantial contributor to climate change. It also contributes to maintaining the background levels of tropospheric ozone. Among a variety of CH 4 sources, current estimates suggest that CH 4 emissions from oil and gas processes account for approximately 20% of worldwide anthropogenic emissions. Here, we report on observational evidence of CH 4 emissions from offshore oil and gas platforms in Southeast Asia, detected by a highly time-resolved spectroscopic monitoring technique deployed onboard cargo ships of opportunity. We often encountered CH 4 plumes originating from operational flaring/venting and fugitive emissions off the coast of the Malay Peninsula and Borneo. Using night-light imagery from satellites, we discovered more offshore platforms in this region than are accounted for in the emission inventory. Our results demonstrate that current knowledge regarding CH 4 emissions from offshore platforms in Southeast Asia has considerable uncertainty and therefore, emission inventories used for modeling and assessment need to be re-examined.
A tmospheric CH 4 is an important component of short-lived climate pollutants that contribute both directly and indirectly to radiative forcing. It is also known that CH 4 contributes to maintaining the background levels of tropospheric ozone 1 . CH 4 is emitted from a variety of natural (e.g., wetlands, oceans, termites, and clathrates) and anthropogenic (e.g., fossil-fuel exploitation, ruminant animals, rice cultivation, waste management, and biomass burning) sources. Because of the shorter atmospheric lifetime (about nine years 2 ) of CH 4 than CO 2 , a reduction of anthropogenic emissions of CH 4 would be an effective means of abating global warming in the near future 3,4 . However, to establish strategies for the mitigation of global warming, a quantitative understanding of the global CH 4 budget is required.
Atmospheric abundance of CH 4 has been increasing from pre-industrial levels of about 700 nmol mol 21 (hereafter referred to as ppb) with large year-to-year fluctuations in its growth rate 5 . Among the many studies that have investigated the distribution and temporal variation of CH 4 , several have reported conflicting results [6][7][8][9] . Some recent studies have suggested the existence of previously unrecognized sources of CH 4 . For example, satellite observations have been combined with inverse modeling techniques using CH 4 retrievals from the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) to provide a global distribution map of CH 4 that suggests there are many emission hot spots in areas where surface observations are scarce 10 . The first airborne in situ measurements of CH 4 over the Amazon region during the BARCA (Balanço Atmosphérico Regional de Carbono na Amazônia) campaign revealed strong CH 4 emissions from the Amazonian wetlands 11 . Regular aircraft observations from the CARIBIC (Civil Aircraft for the Regular Investigation of the atmosphere Based on an Instrument Container) program suggest strong biogenic emissions from India that cannot be attributed solely to rice cultivation 12 . These studies show that our current understanding regarding the sources of CH 4 emissions is inadequate and that greater effort is needed to obtain better knowledge both of the strength of CH 4 emissions and of the distribution of the sources. For this purpose, a more systematic approach is required regarding the acquisition of CH 4 observations in areas where observational data are sporadic or sparse.
Since 1992, the National Institute for Environmental Studies (NIES) has conducted a voluntary observing ships (VOS) program of long-term atmospheric monitoring of climatically important trace gases over the Pacific Ocean 13,14 . In the Southeast Asian region, atmospheric monitoring as part of the NIES-VOS program commenced in September 2007. Although flask sampling was initially used for CH 4 monitoring, since 2009, the program has been augmented by the use of continuous measurements that capture the highly variable features of CH 4 in the regionally polluted air in Southeast Asia. In this paper, we present the first results of the high-resolution continuous onboard measurements of CH 4 in the marine boundary layer (MBL) in the Southeast Asian region between September 2009 and April 2012. We focus on the CH 4 distribution in the northern equatorial region, where strong CH 4 peaks were observed off the east coast of the Malay Peninsula and the northwest coast of Borneo, and we examine the emission sources responsible for these observed CH 4 peaks.

Results
Although shipping operations were disrupted temporarily by the global financial crisis and the 2011 Great East Japan Earthquake, we were able to undertake observations during eight voyages between September 2009 and April 2012. Many CH 4 peaks were observed in the northern equatorial region along Southeast Asian routes ( Figure 1); see Methods section for further details. The locations at which these peaks occurred were concentrated in two areas: off the east coast of the Malay Peninsula (39 peaks) and off the northwest coast of Borneo (55 peaks); they are referred to hereafter as the Malay and Borneo peaks, respectively. The Malay peaks were observed largely between latitudes 8uN and 5.5uN along both the northbound and southbound routes, while the Borneo peaks were observed between latitudes 6.5uN and 4.5uN along the northbound Borneo route. Although the durations of all observed CH 4 peaks were short, between several minutes to one hour, the increases of the mole fraction of CH 4 were considerable, i.e., up to about 1100 ppb above the baseline levels for the Southeast Asian region. Concurrent with the CH 4 peaks, we observed simultaneous CO 2 peaks, and the positive correlation between these CO 2 and CH 4 mole fractions suggested a common local, non-biogenic emission source for these gases.
To identify sources of CH 4 emission, we examined satelliteobserved night-light data from the US Air Force Defense Meteorological Satellite Project Operational Linescan System (DMSP/OLS), provided by the US National Oceanic and Atmosphere Administration 15 . Using the nighttime lights data ''avg lights x pct'', which are annual composite images of noise-filtered nighttime lights data used to infer gas-flaring volumes 16 , we identified the locations of offshore platforms within the study area. The distribution of these identified offshore platforms remained largely unchanged throughout our study period. Most of these platforms were either off the east coast of the Malay Peninsula or off the northwest coast of Borneo and were near the locations of the CH 4 peaks along the Southeast Asian trade routes ( Figure 2). Generally, CH 4 is a dominant component of emissions from offshore oil platforms, released as a result of gas flaring and venting, equipment leaks, and evaporation losses, with concomitant emissions of CO 2 mainly due to gas flaring 16,17 . These results suggest that the observed CH 4 peaks represent emissions from offshore production platforms.
The CH 4 emissions from offshore platforms are reported in the anthropogenic trace gas emission inventory database EDGAR (Emission Database for Global Atmospheric Research) v.4.2 FT2010 18 . We compared the distribution of offshore platforms identified in this study to that reported in EDGAR. For the comparison, we used the annual composite image for 2010 from the DMSP/OLS data and annual CH 4 emission data for 2010 from EDGAR. This revealed considerable discrepancy in the distribution of offshore platforms, especially off the east coast of the Malay Peninsula, which indicates that the current emission inventories of offshore platforms in Southeast Asia still include considerable uncertainties regarding CH 4 and other co-emitted gas components.
Most of our observations were performed during the boreal fall and winter season when strong northeasterly winds associated with the East Asian monsoon prevail off the east coast of the Malay Peninsula; westerly winds passing over the Malay Peninsula from the Indian Ocean prevailed only during September and October of 2009 during our observations. In contrast, there was no prevailing wind direction off the northwest coastal region of Borneo during our study period. Thus, the Malay peaks observed during the northeasterly wind season should represent emissions from offshore platforms windward of the peak locations, whereas the Borneo peaks represent emissions from both offshore platforms and onshore coastal sources. To characterize the offshore platform emissions, we examined the emissions measured during the CH 4 peaks based on the CH 4 -CO 2 enhancement ratio (DCH 4 /DCO 2 ), which is the linear slope of the correlation of the mole fractions of CH 4 and CO 2 . The observed enhancement ratio can often be used to identify the emission sources because it can be approximated to the emission ratio when observations are performed near the emission sources. For example, past observations at remote sites during wintertime have shown the DCH 4 /DCO 2 ratios are typically less than about 20 ppb/ppm (ppm is defined as mmol mol 21 ) in air masses polluted principally by anthropogenic combustion-related emissions in urban and industrialized areas [19][20][21][22] . The emission factors of CO 2 and CH 4 from biomass-burning sources were determined the typical CH 4 /CO 2 ratios to be less than 20 ppb/ppm 23 . These results provide diagnostic criteria for estimating the contributions from these anthropogenic emissions on land.
For further analysis of the DCH 4 /DCO 2 ratios observed here, we selected 11 distinct Malay peaks and 16 Borneo peaks that showed substantial CH 4 increases (.50 ppb) for more than 10 min and significant positive correlations between CH 4 and CO 2 (R . 0.4, p-value , 0.05). The DCH 4 /DCO 2 ratios during these peaks, calculated by reduced major-axis regression 24 , ranged from 8 to 1108 ppb/ ppm and from 3 to 880 ppb/ppm for the Malay and Borneo peaks, respectively ( Figure 3). The M1, M3, M4, M5, M6, M8, and M9 peaks show similar DCO 2 -vs.-DCH 4 correlative behavior, suggesting that they originated from the same emission process at the offshore platforms. We examined the approximate flaring efficiency (i.e., DCO 2 / (DCO 2 1 DCH 4 ) in %) using the DCO 2 -vs.-(DCO 2 1 DCH 4 ) regression for these peaks. The mean DCH 4 /DCO 2 ratio for these peaks is 94 ppb/ppm, corresponding to a flaring efficiency of 92%. As gasflaring efficiencies for industrial flares are usually greater than 90%, it is considered that these plumes originated principally from flaring with the contribution from fugitive emissions (inclusive of venting), if any, being relatively small. The M7, M10, and M11 peaks show higher DCH 4 /DCO 2 ratios than those of the gas-flaring plumes, suggesting greater contribution from fugitive emissions. These results www.nature.com/scientificreports indicate that the observed DCH 4 /DCO 2 ratios can vary widely, depending on the contributions from fugitive emissions. In contrast, the contributions from flaring and fugitive emissions to the M2 peak, associated with the DCH 4 /DCO 2 ratio of 8 ppb/ppm, appear negligible. Similarly, the 16 peaks observed in the Borneo area with DCH 4 /DCO 2 ratios higher than 20 ppb/ppm were explained by the mixing of flaring and fugitive emissions. Consequently, we chose those peaks with DCH 4 /DCO 2 ratios higher than 20 ppb/ppm for further analysis.

Discussion
We used these observed CH 4 peaks to estimate the CH 4 emission rates based on a mass balance approach [25][26][27] . Assuming that the CH 4 plume was formed steadily during the observation and that the CH 4 mixing ratio was vertically well mixed in the MBL, the CH 4 emission rate q CH4 can be expressed by: In equation (1), u is the mean horizontal wind speed along the plume axis, a is the angle between the ship transect and the perpendicular to the plume axis, Z MBL is the depth of the MBL, n is the average molar density of air within the MBL, y is the distance from the plume axis, f CH4 (y) is the observed CH 4 mole fraction at y, and C 0 (y) is the background CH 4 mole fraction at y. The start and end points of the integration interval for the individual CH 4 peaks, a and b in equation (1), are determined manually by visual inspection, and the values of C 0 (y) are determined practically by linear interpolation between the CH 4 mole fractions at points a and b. As no meteorological observations were performed onboard, the mean wind speeds and directions and the depths of the MBL were estimated based on the CGER/METEX three-dimensional kinematic trajectory model 28 . This trajectory model was driven by six-hourly meteorological input data from the NCEP/NCAR reanalysis, which has a spatial resolution of 2.5u 3 2.5u. The model calculation was initiated at an altitude of 250 m above sea level at the locations of the CH 4 peaks. For every plume calculation, we adopted the molar density of air (n) of 1.2 kg m 23 , which was the average value of n at 0 and 0.5 km 29 . We applied the mass balance approach to the appropriate 14 peaks that the ship transited straight across the CH 4 peak. Geographical relationships between the observed CH 4 Table 1, together with the values of the parameters used in the calculations. These estimates are comparable with the recent preliminary estimates using the Visible Infrared Imaging Radiometer Suite on board the Suomi National Polar-orbiting Partnership satellite, provided by the National Geophysical Data Center of the National Oceanic and Atmospheric Administration 30 . As conservative error estimates, we evaluated uncertainty ranges for the emission rates by assuming the relative uncertainty of 650% for the two dominant factors (u 3 cosh and Z MBL ), and 60.05 kg m 23 for the molar density of air (1.2 kg m 23 ). The calculated uncertainty ranges (from lower to upper limits), also listed in Table 1, suggest substantial uncertainty in this approach. The median value of all the CH 4 emission rates is 29.2 g s 21 . Using the DMSP satellite, we identified 112 offshore platforms in the Southeast Asian region (defined as: 15uN-10uS, 90u-140uE); thus, the resultant regional total emission rate is calculated as 3.3 kg s 21 . The total regional annual emission of CH 4 from offshore platforms in the Southeast Asian region is estimated to be about 0.1 Tg y 21 , associated with an uncertainty range of 0.02-0.32 Tg y 21 (the median values of the lower and upper limits). EDGAR reports that annual CH 4 emissions from oil and gas production in 12 Southeast Asian countries (Brunei, Cambodia, Lao, Myanmar, Malaysia, the Philippines, Singapore, Thailand, Timor, Vietnam, Indonesia, and Papua New Guinea) were about 3.7 Tg y 21 for 2010, which corresponds to about 1% of the 335 Tg y 21 global anthropogenic CH 4 emissions 18 . Offshore CH 4 emissions account for about 8% (0.29 Tg y 21 ) of the total emissions from oil and gas production in the Southeast Asian region. Despite the large uncertainty inherent in the mass balance approach, our estimate displays relatively good agreement with that by EDGAR. However, we note substantial differences in the locations of the offshore platforms between the EDGAR inventory and those determined by DMSP satellite observation. The distributions of point sources of CH 4 are an important uncertainty in the existing inventories. The relative contributions of the offshore CH 4 emissions to the regional CH 4 emissions in Southeast Asia are estimated to be about 3% for the oil and gas production sector (both offshore and onshore) (3.7 Tg y 21 , as esti- include fossil fuels, ruminants, landfills/waste, geological sources, freshwater, rice paddies, burning of biomass/biofuels, wild animals, and termites. The emissions from other sources are minor. The emissions (and range) from hydrates, wildfires, and permafrost are estimated to be 6 (2-9), 3 (1-5), and 1 (0-1) Tg y 21 , respectively.
Globally, there are a number of offshore fields for oil and gas production. In addition to Southeast Asia, the North Sea, Persian Gulf, Gulf of Guinea, and Gulf of Mexico are known to be active in oil and gas production. As noted above, our estimate and EDGAR are in relatively good agreement for offshore CH 4 emissions in Southeast Asia. A simple global estimate based on EDGAR implies that CH 4 emissions from worldwide offshore oil and gas platforms are 1-2 Tg y 21 , suggesting that the emissions from offshore sources may be comparable to those from minor natural sources such as wildfires and permafrost.
To our knowledge, this work marks the first top-down constraint on CH 4 emissions from oil and gas platforms in Southeast Asia. On the other hand, we also realize the considerable uncertainty in our estimates, which derive from a combination of features inherent in the mass balance approach and the lack of samples of CH 4 plumes from offshore platforms, due to the sporadic occurrence of gas flaring and fugitive emissions at oil and gas platforms. Hence, our top-down estimates of CH 4 emissions from offshore platforms located in the Southeast Asian region need to be tested and improved. For example, if fugitive plumes were undersampled in our observations, the estimated CH 4 emissions would be much greater, possibly even by one order of magnitude. To better assess the regional total emissions of CH 4 from offshore platforms and thereby improve the current emissions inventory, further top-down constraints by integrated ship, aircraft, and satellite observations are needed. In particular, the detection of fugitive plumes would be useful to reduce uncertainties in estimating the emissions. Current estimates of CH 4 emissions from oil and gas processes were approximately 20% of worldwide anthropogenic emissions in 2010, and they are expected to increase by nearly 35% between 2010 and 2020 32 . The feedback gained from plume observations can help in the reduction of fugitive emissions in Southeast Asian countries and thus, contribute to the mitigation of global warming.

Methods
We used two commercial cargo vessels in the VOS program in Southeast Asia: the M/ V Fujitrans World (owned by the Kagoshima Senpaku Kaisya, Ltd., Japan) was the primary vessel with backup provided by the M/V Trans Future 1 (owned by the Toyofuji Shipping Co. Ltd., Japan). These ships regularly sail the trade routes between Japan and Southeast Asia, berthing at Osaka, Yokohama, and Nagoya (Japan); Hong Kong (China); Laem Chabang (Thailand); Singapore; Port Klang, Kuching and Kota Kinabalu (Malaysia); Jakarta (Indonesia); and Muara (Brunei) at four-week intervals ( Figure 4). Two northbound routes are used from Jakarta to Japan: one via Thailand and the Philippines (the northbound Asia route) and the other via Borneo (the northbound Borneo route). Only one southbound route is used from Japan to Indonesia.
Onboard each VOS ship, continuous measurements of CO 2 , CO, and O 3 were performed using a non-dispersive infrared analyzer (NDIR), an NDIR with gas filter correlation, and an ultraviolet absorption analyzer, respectively. The continuous CO 2 data processed for public use are available at our webpage (http://soop.jp/). Flask samples were also collected for laboratory analysis of CO 2 , CO, CH 4 , N 2 O, SF 6 , H 2 ,    13 .
In September 2009, continuous measurements of atmospheric CO 2 and CH 4 were commenced on VOS ships along both of the Southeast Asian routes using wavelength-scanned cavity ring-down spectroscopy (WS-CRDS) instruments (Picarro Inc., Santa Clara, CA, USA, models EnviroSense 3000i and G1301). Air sample was collected from the air intake set at the top deck of the ship (approximately 50 m above sea level) using a diaphragm pump placed in the observation room. The sampled air was dried before analysis to minimize biases due to dilution and pressure-broadening effects of water vapor on the WS-CRDS measurements. The sampled air was dehumidified by passing it through a sample-drying unit consisting of an electric cooler kept at 11uC and a Nafion Perma Pure dryer (Perma Pure LLC, Toms River, NJ, USA). The design and performance of the unit was very similar to that used for the CO measurements 13 . The sampled air was dried to less than ,0.3% water (as measured by the WS-CRDS) before the mole fractions of CO 2 and CH 4 were determined based on the water vapor content according to an instrument-specific water vapor collection function 33 . To our experience, the air samples were rarely contaminated with the ship's exhaust, when the ship sails at approximately 20 knot because the ship's exhaust is located at stern side. When the air samples were contaminated with the exhaust gas, we judged it by the CO 2 and O 3 measurements (CO 2 increase and O 3 decrease), and then rejected the data before the analysis.
For instrument calibration, we prepared a set of three natural or purified airbalanced standard gases with CO 2 and CH 4 (ca. 380, 400, and 420 ppm for CO 2 , and 1800, 2000, and 2200 ppb for CH 4 ) in our laboratory to prevent pressure-broadening effects due to the different air compositions of the samples and standard gases. The standard gases were introduced into the WS-CRDS instrument daily in series, 10 min for each gas. The mole fractions of CO 2 and CH 4 in the standard gases were calibrated against NIES standard gas scales (NIES 09 for CO 2 , NIES 96 for CH 4 ), which are traceable to World Meteorological Organization standard gas scales. The analytical precision for 1-min measurements by the WS-CRDS instruments of CO 2 and CH 4 were typically 0.05 ppm and 0.5 ppb (1 sigma), respectively. In this study, we used only the 1-min temporal mean CH 4 and CO 2 data from the continuous measurements because of the coarse resolution of the CO data, available only as 1-hour means from the gas filter correlation measurements, and because the O 3 data provide little information about CH 4 emission sources.