Methane mole fraction and δ13C above and below the trade wind inversion at Ascension Island in air sampled by aerial robotics

Ascension Island is a remote South Atlantic equatorial site, ideal for monitoring tropical background CH4. In September 2014 and July 2015, octocopters were used to collect air samples in Tedlar bags from different heights above and below the well‐defined Trade Wind Inversion (TWI), sampling a maximum altitude of 2700 m above mean sea level. Sampling captured both remote air in the marine boundary layer below the TWI and also air masses above the TWI that had been lofted by convective systems in the African tropics. Air above the TWI was characterized by higher CH4, but no distinct shift in δ13C was observed compared to the air below. Back trajectories indicate that lofted CH4 emissions from Southern Hemisphere Africa have bulk δ13CCH4 signatures similar to background, suggesting mixed emissions from wetlands, agriculture, and biomass burning. The campaigns illustrate the usefulness of unmanned aerial system sampling and Ascension's value for atmospheric measurement in an understudied region.


Introduction
Global atmospheric methane (CH 4 ) is increasing rapidly. Between 2007 and 2013 CH 4 increased globally by 5.7 AE 1.7 ppb yr À1 with a more rapid increase in the equatorial tropics in 2010-2011. In 2014 the yearly growth increased to 12.5 AE 0.4 ppb . Tropical emissions are thought to be a major contributor to this growth Schaefer et al., 2016]. Natural tropical sources include wetlands and other freshwater systems, lightning-lit fires, geological leaks, termites, and wild ruminants [Dlugokencky et al., 2011]. CH 4 emissions from the tropics produce around~200 Tg/yr which is thought to be about 40% of the global CH 4 budget [Frankenberg et al., 2008;Bousquet et al., 2006]. Variability in CH 4 removal through reaction with OH radicals is also dominated by the tropics as photochemistry is active throughout the year [Bousquet et al., 2006]. δ 13 C CH4 isotope ratios can be used to help apportion emission sources, because the different sources of CH 4 have different ratios of 13 C and 12 C isotopes. Background ambient air at present approximately has a δ 13 C CH4 of À47.4 to À47.2‰ . CH 4 emissions from biological sources are depleted in comparison to background ambient air, for example, swamps give a signature of À55 AE 3‰ [Dlugokencky et al., 2011], because methanogenic archaea preferentially use the lighter 12 C isotope. CH 4 formed at higher temperatures is relatively enriched in the heavier isotope, for example, biomass burning of savannah grassland (C4 plants) gives a signature of À20 to À15‰ [Kirschke et al., 2013;Dlugokencky et al., 2011;Chanton et al., 2005]. See supporting information (S) Table S1 [Teh et al., 2005;Dlugokencky et al., 2011;Chanton et al., 2000] for isotopic CH 4 source values for tropical wetlands and biomass burning.
In situ sampling of carbon isotopes in CH 4 can be used to distinguish sources [Dlugokencky et al., 2011]. Apart from short airborne campaigns, such measurements have, however, usually been confined to within the planetary boundary layer. Unmanned aerial systems (UASs) offer a practical option for regular sampling at higher level. Aircraft measurements and flask sampling [Schuck et al., 2012] Brownlow, R., et al. (2016), Methane mole fraction and δ 13 C above and below the trade wind inversion at Ascension Island in air sampled by aerial robotics, Geophys. Res. Lett., 43, 11,893-11,902, doi:10.1002 abilities [Chang et al., 2016]. Balloon flights lack maneuvrability compared to UASs [Karion et al., 2010;Chang et al., 2016]. AirCore-like systems [Karion et al., 2010] are impractical for routine monitoring of δ 13 C CH4 and mole fraction as it is difficult to achieve sufficient isotopic precision to observe small changes.
In this study, a new approach to sampling air masses at different heights using bespoke octocopters (UAS) was developed during two field campaigns on Ascension Island in September 2014 and July 2015. The target was to sample above the Trade Wind Inversion (TWI), a strong and persistent temperature inversion capping the marine boundary layer at altitudes between 1200 and 1800 m above sea level (asl) [Barry and Chorley, 2009]. There is strong wind shear across the TWI with air trajectories indicating different origins for air above and below the inversion, as discussed in section 1.1 below.
Air masses were sampled at various heights above and below the TWI, using a remotely controlled pump and valve system to fill Tedlar bags, probing air from different origins. Ascension Island's location is ideal for these experiments. The work has proved the ability of using UASs to collect and maintain the integrity of air samples using real time sensors for targeting specific air profile characteristics.

Ascension Island
Ascension Island (7°58′S, 14°24′W) (supporting information Text S1 and Figure S1), in the South Atlantic, experiences near constant South East Trade winds below the TWI ( Figure S1a) [Rolph, 2016;Stein et al., 2015] with little diurnal variation, bringing air from central South America and the Southern Ocean. From Ascension the air crosses the Atlantic to become background into Amazonia. Above the TWI, air trajectories are quite different, and the air comes predominantly from tropical and southern Africa ( Figure S1b). Ascension Island is therefore ideally located to measure the tropical Atlantic background air in the boundary layer and to study African sources of CH 4 by sampling the midtroposphere above the TWI.
Cumulative 240 h monthly backtrajectories were simulated for both the September and July campaign period using the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) trajectory tool [Stein et al., 2015]. Air masses at ground level in September and July are from the remote South Atlantic, and above the TWI are from central and West Africa (Figures S1e-S1h) [Rolph, 2016;Stein et al., 2015].
Royal Holloway, University of London (RHUL), hosted by the UK Met Office, has measured CH 4 mole fractions (in situ) and δ 13 C CH4 (in flasks) on Ascension Island from 2010 . NOAA and the University of Colorado's Institute of Arctic and Alpine Research have measured both CH 4 mole fractions and δ 13 C CH4 in flasks collected on the island since 2000 ( Figure S2) Dlugokencky et al., 2016;White et al., 2015].

Methods/Experimental Design
Three 8-rotor multirotor (octocopter) ( Figure S3a) platforms were built from off the shelf components at the University of Bristol. The platforms were custom designed to carry gas sampling equipment and temperature and humidity response sensors to an altitude of 2700 m within 20 min, with 20% battery capacity remaining on landing. The meteorological sensors sent data to the ground control station in real time to allow targeted gas sampling during descent. Flight and avionic specifications of the platform [Greatwood et al., 2016] are detailed in Table S2 and key elements of the operational manual in Text S1 [Thomas et al., 2012].
Octocopters were flown from a site at 340 m asl ( Figure S1c) with a maximum climb rate of 5 m/s. The system retrieved an in situ air sample using an on board diaphragm pump with a flow rate of 4.5 L/min to fill either 3 L or 5 L Tedlar bags (SKC Ltd) as the octocopter hovered for between 45 and 60 s, dependent on atmospheric pressure at the sampling altitude. It was possible to fill two samples per flight at different altitudes. A longer sampling time with a single bag was used at higher altitude ensuring enough sample mass was drawn into the bag for measurement.
High temporal resolution atmospheric profiles of temperature and humidity were telemetered to ground station computers during ascent. The TWI is characterized by temperature increase and relative humidity decrease above the cloud layer. This informed in-flight targeting decisions by the ground observer with remote communication to the UAS for the desired sampling altitude around the unexpectedly tightly defined TWI during the descent leg of the flight. Heights were accurate to within a few meters and allowed samples to be taken above, below, and within the TWI each day, avoiding the use of more uncertain model predictions of Geophysical Research Letters 10.1002/2016GL071155 the TWI boundary ( Figure S3b). Additional 3 L Tedlar bags were filled with air from approximately 1 m above site ground level (340 m asl) and Met Office ground level (75 m asl) each day to compliment the UAS samples. Samples from the flights and ground sampling were subsequently analyzed together, (see Text S2 [Lowry et al., 2014;Fisher et al., 2006;Jones et al., 2007;Cullen, 1993] for methods).

Results
NOAA and RHUL measurements show that the CH 4 mole fraction at Ascension has been increasing sharply since 2007 . A parametric curve fitting program, HPspline, was used to assess the longerterm trends of the RHUL data using the parameters suggested in Pickers and Manning [2015].  . Continuous cavity ring-down spectroscopy (CRDS) monitoring of CH 4 shows regular dip and peak events ( Figure S5). Peak events occur intermittently, with an increase of around 10 ppb and may last for a period of hours or days. A number of dip and peak events in CH 4 mole fraction occur even though the trajectories below the TWI are steady. The field campaigns were in September 2014 and July 2015. July was chosen as a comparison to the September campaign as July is normally the peak biomass burning season in Southern Hemisphere Africa [Roberts et al., 2009], when smoke plumes inject upward into the free troposphere [Chatfield et al., 1998].

Campaign 1-September 2014
The September 2014 campaign period overlaps with one of the numerous dip and peak events with a decline in the average mole fraction measured by the CRDS of 8 ppb on 14 September 2014 then a 6 ppb increase on 15 September 2014 of followed by a further 7 ppb increase on 16 September 2014 (Figures 1b and S9a). There are significant correlations (95% confidence) between the bag sample ground values collected at the UAS site, 340 m asl, and in situ measurements at 75 m asl showing no significant difference between the two sites .
Samples taken above the TWI have a higher CH 4 mole fraction than samples taken below (Figures 1b and 2). The samples retrieved from above the TWI contain higher CH 4 mixing ratios than ground level; however, during periods of lower mixing ratio of CH 4 at ground level there is an increased change in concentration across the TWI. The spread of CH 4 mixing ratios when compared over the whole campaign increases with altitude. No distinct isotope ratio change was seen between samples taken above or below the TWI. Ground values Using the difference between the median and the 95th percentile of the ground, below TWI, and above the TWI samples as a benchmark; ground level values give a difference of 3 ppb, below the TWI gives a difference of 7 ppb, and above the TWI mixing ratios gives a difference of 8 ppb. The difference of the median mixing ratios below and above the TWI is 16 ppb.

Geophysical Research Letters
10.1002/2016GL071155 ranged between À47.41 and À47.17‰ with samples taken from the UAS ranging between À47.60 and À47.00‰ (see Figure 5). Boundary layer heights (Figure 1) vary between 1350 and 1628 m asl with no correlation between boundary layer height and changes in the CH 4 mole fraction at ground level.
Daily NAME modeled footprint plots ( Figure S6) have been generated from particles released from a height of 0 to 100 m above ground level with time integrated particle density in the boundary layer over the 11 days up to the release time. These plots show air arriving from the remote South Atlantic, and to a lesser extent from southern Africa as indicated in the cumulative trajectories. NAME was run backward for 7 days for individual samples. Particles were released from the sample location (from a depth of 100 m centered on the sample height) in order to model where the measured air had come from. Figure S7 shows all the 7 day integrated particle density plots for each sample, for three different height bands (0.1-1 km, 1-3 km, and 3-9 km). Each plot shows the density of particles that passed through that grid box and height band over the last week. Clear differences can be seen in the air mass histories of different samples with variable influence from the different height layers. Some samples are mainly influenced by the remote South Atlantic, whereas others include air masses coming from the north or central Africa. Figure 2 shows typical NAME plots during each sampling period. Ground samples (Figures 2a to 2d) in general arrive from the remote South Atlantic. Above the TWI samples with influence from the east and northeast with air coming from over Africa (Figures 2e-2g and S7) tend to have higher mole fractions than those with South Atlantic influence (Figure 2h).

Campaign 2-July 2015
Continuous CRDS ground level measurements ( Figure 3b) were on average 8 ppb higher in July 2015 than in September 2014. July 2015 ground samples have a larger range from 1800 to 1820 ppb before the drop of mole fraction on 13 July 2015 after which they range from 1795 to 1810 ppb, whereas the September 2014 campaign mole fractions were between 1790 and 1805 ppb ( Figure S9). In contrast to the September campaign the samples taken above the TWI have a higher CH 4 mole fraction during the period when CH 4 is high at ground level ( Figure 3). There is significant correlation between samples taken from the ground at both the UAS (340 m asl) and Met Office (75 m asl) sites and the Picarro CRDS values.
Boundary layer heights (Figure 3a) range from 1000 to 1500 m asl between 7 July 2015 and 11 July 2015. During the decrease in the ground level mole fraction beginning on 13 July 2015 the boundary layer height increases from~1500 to~2050 m asl.
Ground level mole fraction measurements can be separated into three periods in the July campaign ( Figure 4ii); 6-12 July has the highest ground mole fractions and more enriched δ 13 C CH4 , 13-14 July is an intermediate period, and 15 July has lowest ground mole fraction and depleted δ 13 C CH4 values (Figures 3b and 4). During the 6-12 July back trajectories on the ground show the air masses have a higher possibility of arrival from over Africa (Figures 4a and 4b) or from the remote South Atlantic (Figures 4 and S8).

Campaign Comparisons
Both campaigns show consistently higher CH 4 mole fractions above the TWI with increments up to 31 ppb. NAME modeling indicates these air masses may be influenced by source emissions north of the intertropical a b Figure 3. perhaps due to the sampling height difference. Using the difference between the median and the 95th percentile of the ground, below TWI, and above the TWI samples as a benchmark; ground level values give a difference of 4 ppb, below the TWI gives a difference of 5 ppb, and above the TWI mixing ratios gives a difference of 28 ppb. The difference of the median mixing ratios below and above the TWI is 8 ppb and the 95th percentile mixing ratios is 31 ppb.  Figure 5). This may reflect year-on-year growth and seasonality. There is expected to be more biomass burning (δ 13 C heavy), as shown in the ground level samples (Figure 4ii) and wetland (δ 13 C light) influence from southern Africa in July [Roberts et al., 2009]. Different mixing rates across the TWI are suggested by the increase in mole fraction above the TWI during periods with lower ground values and a decrease in mole fraction above the TWI when higher ground values occurred in the September campaign. Trajectories above the TWI also show a NE component during the dip in ground level mole fraction compared to before and after the dip event. The three ground mole fraction periods shown in Figure 4ii during the July campaign shows that the mole fractions measured above the TWI were higher when higher ground values occurred suggesting the ground mole fraction was influenced by increments along the air mass trajectory.

Geophysical Research Letters
During the July campaign there was a greater TWI height range (1030 to 2057 m asl). When the inversion is higher, the inversion strength is weaker with more mixing across the boundary layer [Cao et al., 2007]. The 7-11 July campaign showed a larger CH 4 contrast above and below the TWI, indicating a stronger capping of the TWI with less mixing with the free troposphere. On the 15 July, the TWI height was higher, and there was less CH 4 mole fraction contrast across the TWI. The September campaign had more constant TWI height, suggesting mixing above and below the TWI occurred before the air reached Ascension during the periods where there is a lower mole fraction separation above and below the TWI.
Isotopic data give an average of δ 13 C À47.25 AE 0.09‰ during the September campaign and an average of À47.33 AE 0.09‰ during the July campaign ( Figure 5) The shift to more depleted values in July corresponds to the long-term trend both on Ascension Island and globally. Overall, no distinct isotopic ratio change was seen above the TWI during either of the campaigns (Figures 5 and S10) [Pataki et al., 2003]. Figures 5i, 5f, and 5h show two samples at a similar height above the TWI with air masses arriving from different source regions. Sample f has a trajectory with a higher possibility of inputs from African emissions and has a CH 4 mole fraction of 1853 ppb and δ 13 C À47.36‰. Sample h has a trajectory predominately from over the South Atlantic and a CH 4 mole fraction of 1817 ppb and δ 13 C À47.38‰. There is an input of 36.1 ppb CH 4 from African sources but an isotopic shift of only 0.02‰ suggesting the sources (Table S1) have a combined isotopic signature close to the bulk atmospheric value. A mixture of tropical wetland emissions (À55 AE 3‰), agriculture (À62 AE 3‰), and biomass burning of savannah grassland (À20 to À15‰) [Dlugokencky et al., 2011] would fit. Destruction by OH is the main CH 4 sink; this has a kinetic isotope shift of 4 to 6‰ [Allan et al., 2001[Allan et al., , 2007Nisbet et al., 2016]. However, the lifetime of CH 4 is~9 years [Dlugokencky et al., 2011] so even in the intense OH of the tropical midtroposphere the effect of OH destruction between the African sources and Ascension Island is small. A more local influence may be from the marine Cl sink in the marine boundary layer [Allan et al., 2001[Allan et al., , 2007.

Conclusions
Ground values, taken more frequently during the campaigns, show a wider spread of isotopic signatures than the long-term bimonthly RHUL samples , making it more difficult to identify a distinct isotopic change above the TWI. The September ground values have a range of 10 ppb and 0.33‰, and the July ground values have a range of 17 ppb and 0.24‰. This may suggest an influence from possible local factors such as the little studied Cl sink or varied OH loss. Sink processes preferentially remove 12 C CH4 [Schaefer et al., 2016]. The air mass footprints have varying influences from Africa as well as the remote South Atlantic likely contributing to the isotopic changes. NAME modeling has shown that the air mass origin above the TWI can vary daily. Both the origin of air masses and mixing events above and below the TWI influence the mole fractions of the ground samples. The isotopes have no consistent signal above or below the TWI, despite the significant CH 4 increment measured indicating the input from Africa is close to background δ 13 C. This input is likely to be a mixture of emissions from tropical wetlands, agriculture, and biomass burning. The samples taken higher up may be more influenced by these diverse sources over Africa.
Longer-term regular monitoring above the TWI would be useful for identifying isotopic signatures of the mixed tropical sources from Africa, determining seasonality and long-term trends. During such monitoring, replacing Tedlar bags with aluminum flask samples would allow CO/CO 2 to be monitored, which perhaps along with absorbing aerosol measurements would help characterize air masses according to biomass burning history. Other measurements such as water vapor mixing ratios and 18 O/ 16 O [Bailey et al., 2013], or O 3 /CO 2 ratios [Berkes et al., 2016], could allow quantification of vertical mixing over the TWI and aid interpretation of CH 4 isotopic composition. If a UAS were to be used, an increase of approximately 30% in the mass should allow these changes to be realized.
Relatively lightweight sensors (1-2 kg) may be placed on a UAS to measure the CH 4 concentration although the precision of~1% is not sufficient to detect small changes in ambient air [Kahn et al., 2012]. A precision of at least 5 ppb would be needed to distinguish differences across the TWI and a higher precision for smaller mole fraction changes associated with meteorological factors. It is likely that such instrumentation will be developed in the next few years.
Overall, the campaigns have shown that it is possible to use inexpensive UASs to access the midtroposphere above Ascension Island and retrieve air samples. Sampling at Ascension is able to measure both remote South Atlantic air from below the trade wind inversion and also air from above that has been lofted by convective systems in a wide region of the equatorial and southern savannah tropics. Thus, the technique extends Ascension's access from sea level up to 2700 m asl, making it a superb location for long-term global monitoring.