See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/313269623

Temporal and spatial variability of ground level atmospheric methane

concentrations in the Ebro River Delta

Article in Atmospheric Pollution Research · February 2017

DOI: 10.1016/j.apr.2017.01.009

CITATIONS READS

3 896

12 authors, including:

Alba Àgueda Claudia Grossi

Universitat Politècnica de Catalunya Universitat Politècnica de Catalunya
49 PUBLICATIONS 715 CITATIONS 35 PUBLICATIONS 324 CITATIONS

SEE PROFILE SEE PROFILE

Laura sanchez-garcia Oscar Batet

Centro de Astrobiología IC3 Catalan Climate Sciences Institute
93 PUBLICATIONS 2,003 CITATIONS 16 PUBLICATIONS 157 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Xavier Rodo on 07 February 2018.

The user has requested enhancement of the downloaded file.

 Atmospheric Pollution Research 8 (2017) 741e753

H O S T E D BY Contents lists available at ScienceDirect

Atmospheric Pollution Research

journal homepage: http://www.journals.elsevier.com/locate/apr

Temporal and spatial variability of ground level atmospheric methane

concentrations in the Ebro River Delta


Alba Agueda a, *
, Claudia Grossi a, **, Elsa Pastor b, Eva Rioja a, b, Laura Sa
nchez-García a, 1,


Oscar Batet a, Roger Curcoll a, Marina Ealo a, 2, Manel Nofuentes a, Paola Occhipinti a,
Xavier Rodo a, c, Josep-Anton Morguí a, d
a
Catalan Institute of Climate Sciences (IC3), Barcelona, Spain
b
Centre for Technological Risk Studies (CERTEC), Department of Chemical Engineering, Universitat Polit

ecnica de Catalunya·BarcelonaTech, Barcelona,
Spain
c
Catalan Institution for Research and Advanced Studies (ICREA), Barcelona, Spain
d
Ecology Department, Universitat de Barcelona (UB), Barcelona, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Deltas provide many worthy ecosystem services. Yet, delta basins are quite vulnerable, especially in the
Received 8 October 2016 face of climate change, which can affect the outcome of both agriculture and biodiversity. Moreover, rice
Accepted 23 January 2017 paddy cultivation is well known to contribute with strong emissions of greenhouse gases (GHGs), such as
Available online 1 February 2017
methane (CH4). Thus, knowing the atmospheric variability of CH4 in relation to the different stages of the
rice culture cycle could help to improve GHGs' mitigation strategies in deltas.
Keywords:
The Ebro River Delta, in the northwestern Mediterranean basin, forms part of the largest Spanish river
Methane
basin and is mainly covered by rice fields. In this study, for the first time, ground level (40 cm a.g.l.)
Atmosphere
Mobile survey
atmospheric CH4 concentrations have been monitored in this area, through twenty-seven car mobile
Rice fields transects, over the course of one year. Seasonal, diurnal and spatial variability of CH4 concentrations were
Straw studied to identify its relationship with rice cultivation, meteorological conditions and land-use
Delta distribution.
Ebro River With regard to seasonal variability, autumn transects showed the highest mean values for atmospheric
CH4 (2.466 ppm) when dead rice straw is mixed with the sediment, and weed growth is prevented.
Spring and summer measurements gave the highest mean CH4 values at dawn (1.897e3.544 ppm),
whereas autumn and winter produced the mean values after sunset (2.148e2.930 ppm). Spatial differ-
ences were accounted for by proximity to urban areas, presence of shallow water storage structures, and
distance to seawater.
© 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by
Elsevier B.V. All rights reserved.

1. Introduction concentrations have risen from 722 ± 25 ppb in 1750e1803 ± 2 ppb

by 2011, and this increase has been linked predominantly to human
Methane (CH4) contributes greatly to the natural and anthro- activities (IPCC, 2013).
pogenic greenhouse gas effect. Globally averaged surface CH4 Sources of CH4 have been identified to be of three main types: (1)
thermogenic: seepages, mud volcanoes, geothermal vents, leakages
from fossil fuel extraction and use; (2) pyrogenic: incomplete
* Corresponding author.
combustion of fossil fuel and plant biomass; (3) biogenic: wetlands,
** Corresponding author.


E-mail addresses: alba.agueda@ic3.cat (A. Agueda), claudia.grossi@ic3.cat termites, oceans, rice paddies, ruminant livestock, landfills, etc.
(C. Grossi). (IPCC, 2013). The most important sinks of CH4 include its tropo-
Peer review under responsibility of Turkish National Committee for Air Pollu- spheric and stratospheric oxidation by OH$ radicals and its microbial
tion Research and Control. oxidation in methanogenic soils and upland soils.
1
Present address: Centro de Astrobiología (CSIC-INTA), Madrid, Spain.
2 According to IPCC (2013) estimates, natural wetlands and rice
Present address: Institute of Environmental Assessment and Water Research,
CSIC, Barcelona, Spain. paddies contribute about 37.3% (32% and 5.3%, respectively) of the

http://dx.doi.org/10.1016/j.apr.2017.01.009
1309-1042/© 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved.
742

A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753
CH4 emitted into the atmosphere, while upland soils are respon-
sible for 4.4% of the CH4 consumption. These values change a little if
Spanish inventory data are taken into consideration. Spanish rice
paddies contribute only about 1% of the net CH4 emitted by this
country, while values for wetlands and upland soils are not re-
ported (Ministerio de Agricultura and Alimentacio  n y Medio
Ambiente, 2016). Although main soil related sources and sinks of
CH4 are well identified, their relative contributions to atmospheric
CH4 levels are still highly uncertain (Kirschke et al., 2013).
Flooded rice paddies and wetland environments have a pre-
dominantly oxygen-free (anoxic) soil profile. In these ecosystems,
CH4 is produced by methanogenic bacteria that digest organic
matter under anaerobic conditions (methanogenesis) (Zhang et al.,
2011). CH4 can also be oxidized into CO2 by methanotrophs in
aerobic zones (e.g. rhizosphere, lower part of culms, soil-water
interface and submersion water) (Le Mer and Roger, 2001). Atmo-
spheric CH4 concentrations measured in the lower boundary layer
of these ecosystems are the result of the transfer by diffusion,
ebullition or mostly through aerenchyma of the net CH4 produced
at the soil-water/soil-atmosphere interface, as well as the result of
atmospheric mixing.
In recent years, there have been many studies on the CH4
emitted from rice fields. Several factors that control these emis-
sions have been highlighted, including both environmental and
agricultural factors. To mention only a few: soil and air temperature
(e.g. Pereira et al., 2013; Chanton et al., 1997; Khalil et al., 1998b;
Simpson et al., 1995), soil redox potential (e.g. Oo et al., 2015;
Simpson et al., 1995), water management (e.g. Meijide et al.,
2011; Yan et al., 2005; Miyata et al., 2000), organic amendment
(e.g. Yan et al., 2005; Fitzgerald et al., 2000) and fertilizers man-
agement (e.g. Oo et al., 2015; Linquist et al., 2012; Bodelier et al.,
2000). Most of those studies consider only CH4 emissions during
the crop cycle, and very few attempts have been made to measure
CH4 emissions during the fallow periods of rice soils (e.g. Fitzgerald
et al., 2000; Bronson et al., 1997), although straw management (e.g.
incorporation into the field, removal from the field, or burning) and
flooding practices after harvest can substantially affect CH4 emis-
sions (e.g. Fitzgerald et al., 2000).
Moreover, these studies have been performed in specific plots of
research stations located in the USA (Chanton et al., 1997) and Japan
(Miyata et al., 2000), or directly in individual rice paddies located in
China (Khalil et al., 1998b), Philippines (Simpson et al., 1995), the
USA (Fitzgerald et al., 2000), Italy (Meijide et al., 2011), Portugal
(Pereira et al., 2013) and Myanmar (Oo et al., 2015). Thus, to date
little effort has been devoted to assessing CH4 emissions in flooded
rice field areas in Europe.
Although continuous atmospheric CH4 observations are
currently being performed at several worldwide existing mea-
surement sites to improve our understanding of CH4 concentration
trends at the regional and continental level (e.g. InGOS - http://
www.ingos-infrastructure.eu/ - and ICOS - https://www.icos-ri.
eu/ - European projects), spatial CH4 mobile surveys have only
been recently performed to identify areas of high CH4 concentra-
tions, and link them to specific sources. Research studies where CH4
transects have been executed have usually focused on surveying
long distances covered by different land-uses (Bamberger et al.,
2014; Farrell et al., 2013), cities (Zazzeri et al., 2015; Phillips et al.,
2013; Shorter et al., 1996) or natural gas and oil production sites
(Eapi et al., 2014; Petron et al., 2012), where localized sources can
be identified. In contrast, spatial differences within rice field
dominated areas have rarely been considered (e.g., Oo et al., 2015;
Mukherjee et al., 2012).
In this study the Ebro River Delta (ERD), a rice paddy dominated
region in the northwestern Mediterranean basin, has been sur-
veyed twenty seven times (9 days) by car, over the course of a year
(December 2011eNovember 2012). The ERD is of international
importance due to its abundant flora and fauna, and belongs to the
largest river basin in Spain. The car surveys were executed during
different phases of the diurnal-seasonal cycles to understand the
temporal and spatial variation of ground level CH4 concentrations
in relation to atmospheric conditions and agricultural practices. We
consider that improving knowledge of CH4 concentration levels in
natural areas of the Mediterranean basin is very important given
that it is a vulnerable region in terms of climate change. In addition,
since this area is dominated by rice paddies, the results of this study
could offer useful information to help in the improvement of the
agricultural Best Management Practices for CH4 emissions
reduction.
2. Material and methods
2.1. Study site: Ebro River Delta
The present study was carried out in the ERD, in northeastern
Spain (Fig. 1). The ERD is one of the largest wetland areas (over
300 km2) in the western Mediterranean region and is located on a
relatively small peninsula.
The Ebro Deltaic area has a typical Mediterranean climate with
mild winters and warm summers (Casanova, 1998). Drought is
observed in summer and rainfall is usually concentrated in two
different periods of the year (MarcheMay and SeptembereNo-
vember). Wind blows with high mean annual velocities
(>30 km h
1) during the entire year (Generalitat de Catalunya,
2011), and blows predominantly from the NW in winter (e.g.
Casanova, 1998; Grossi et al., 2016), and from the S-SE in summer
(Generalitat de Catalunya, 2011). The atmospheric relative humid-
ity is high over the entire year (>65%) (e.g. Generalitat de Catalunya,
2011; Grossi et al., 2016).
The ERD has a flat orography. Maximum elevations are found
near the town of Amposta (8 m above sea level (a.s.l.)) and close to
the river, with values of only 4e5 m a.s.l. (Generalitat de Catalunya,
2011). About 60% of the whole area has elevations lower than 1 m
a.s.l. (Generalitat de Catalunya, 2011). Sediment supply to the
deltaic plain has been reduced during recent years, with the result
that the vertical accretion process in the deltaic plain has been
limited (Alvarado-Aguilar et al., 2012).
The ERD has a shallow aquifer (6e8 m deep) formed of low-
permeability fine sand, allowing the appearance of frequent sa-
line and hypersaline zones (Bayo  et al., 1993). A deep aquifer
(40e60 m) also exists and is present in several layers (gravels, sand,
silts), but it has been poorly characterized so far (Bayo et al., 1993).
The ERD is irrigated with surface water taken from the Ebro river
56 km upstream. Two main canals distribute water through a
network of smaller canals at both margins of the river.
The dominant land-use in the ERD is rice agriculture. Rice
paddies are spread over the major part of the Delta (Fig. 1, zoom in
panel) and represent 62% of the cultivated area (Fatoric and
Chelleri, 2012). As there is a saline environment at the upper soil
level, continuous flooding allows rice to be grown in the Delta,
because this type of plant can get oxygen through the aerenchyma
when flooded, and salt levels are lowered.
Shallow coastal waters, beaches, dunes, saline lagoons, salinas,
freshwater marshes, and freshwater pools fed by groundwater
springs are also present in this fluvial Delta. Most of the natural
wetland areas are found parallel to the coastline (Fig. 1, zoom in
panel) and within the limits of the Natural Park, declared a pro-
tected area in 1983. The ERD is included in the List of Wetlands of
International Importance (RAMSAR, 2016) and supports numerous
species of waterbirds, commercially valuable fish, mollusc, shrimp
and endemic fish populations.


A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753 743

Fig. 1. CORINE land cover map 2006 with a zoom in over the Ebro River Delta location (EEA, 2006).

In Fig. 2, a Gantt diagram is presented with the most important the different activities are generally observed during the life cycle
agricultural activities performed in the rice fields during the period of the rice crop present in this area, and they are provided as a
of study (December 2011eNovember 2012). Start and end dates of guideline to understand the condition of the fields during the

Fig. 2. Gantt chart of anthropogenic activities performed in the Ebro River Delta (ERD) during the period of study: agricultural practices (black e fields; orange e straw and weed
management; red e chemicals; blue e water management; green e rice phenology); mitigation measures performed to fight against apple snail pest (magenta and cyan). The
starting day of the nine mobile campaigns performed for this study are indicated as Ci (where i ¼ 1e9) on the upper axis and dashed vertical lines are also included for each
campaign. Dec/11 and Jan/12 refer to December 2011 and January 2012.
744

A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753
different mobile surveys. The starting day of the two-day long
mobile surveys performed within the frame of this work are also
indicated in Fig. 2.
Rice paddies are permanently flooded during most of the
growth cycle, with a water column ranging from 8 to 15 cm
(Alvarado-Aguilar et al., 2012). Before irrigating the fields (normally
mid-April), land is prepared (tilled and leveled) and fertilized. After
the flooding, direct sowing takes place (mid-April e mid-May) and
plants grow until mid-August, when they are mature and har-
vesting starts. After harvest, rice straw is incorporated into the soil
using mechanical methods. There are only two specific intervals
during the growth cycle when water is drained: (1) Mid-May: fields
are drained for approximately 2 days only to spur young seedling
development and prevent damage by mosquito larvae or chirono-
mids; (2) Mid-June: fields are drained for approximately 1 week to
apply herbicides for weed management.
During the period of study, agri-environmental payments
encouraged rice growers to adopt agricultural activities to preserve
the environment. The most relevant activity was to flood the fields
in autumn-winter to provide substituted wetlands for birds com-
munities (Generalitat de Catalunya, 2008). This activity was per-
formed during those periods when rice straw was left on the fields
(orange rectangles in Fig. 2).
In addition, since 2009 a species of snail (Pomacea insularis) has
invaded most of the rice production areas and the river shore in the
left margin of the ERD (1576 ha affected in 2012; Lo pez, 2015;
personal communication). The apple snail has become a serious
threat to rice production and natural ecosystem equilibrium, and a
variety of measures in response to this infestation were imple-
mented in the rice fields during 2012, i.e.: (i) drying out part of the
fields over the winter (10,200 ha; only the whole left margin); (ii)
flooding some fields with salt water (650 ha); (iii) applying saponin
products to the fields (480 ha); (iv) setting up contention systems
to catch and eliminate the snail (Lo  pez, 2015; personal communi-
cation). The implementation of these measures may have had an
unequal effect on CH4 emissions. As a first approximation, the first
two mitigation measures (i and ii) may have had a reducing effect
on CH4 emissions since non-submerged soils are more aerated and
thus more efficient in methanotrophy (e.g. Serrano-Silva et al.,
2014), and salinity has an inhibitory effect on methanogenesis ac-
cording to Le Mer and Roger (2001). In contrast, since the amount of
dissolved oxygen in water decreases when saponins are applied
(FAO, 1978), treatments with this chemical product (measure (iii))
may have had a positive influence on CH4 emissions, because CH4
oxidation at the soil-water interface might have been hindered.
Physical measures, such as the one in point (iv), may have had no
implications in terms of CH4 emissions.
2.2. Campaigns
2.2.1. Experimental design
We were primarily interested in evaluating the spatial vari-
ability of atmospheric CH4 concentrations in the ERD area, at both
seasonal and diurnal scales. For this reason, mobile measurements
were taken over a one-year period to investigate different seasonal
states of vegetation and meteorological conditions. To assess the
diurnal pattern of CH4, the mobile concentration measurements
were made successively, at different times of the day. More spe-
cifically, nine measuring campaigns were carried out from
December 2011 to November 2012 (Table 1) and each campaign
consisted of three transects which were executed: (1) on the eve-
ning of the first day of the campaign (average start time: 16:23 UTC;
transects ID suffix: Ev); (2) in the morning of the second day of the
campaign (average start time: 05:41 UTC; transects ID suffix: Mo);
(3) in the afternoon of the second day of the campaign (average
start time: 12:04 UTC; transects ID suffix: Af).
The average transect duration was 2 h 49 min (see Table 1) and
the average distance covered per transect was 72 km. Sample size
values shown in Table 1 indicate that transects 20111220Ev,
20120418Ev, 20120419Af, 20120613Ev, 20120802Af and
20120907Mo (see Table 1 for transect ID nomenclature) were
incomplete due to technical restrictions (GPS recording frequency
incorrectly set -20111220Ev- or batteries lifetime eall other pre-
viously quoted transects-). The low number of datapoints was
taken into account in the data analysis.
2.2.2. Transect pattern
The transect pattern was designed as a closed loop to cover
different habitats of the ERD, and a typical one is presented in Fig. 3.
At each margin of the river two locations were selected, one sur-
rounded by paddy fields and another one close to a natural wetland
(in the example of Fig. 3, #2 and #1, respectively, for the left margin,
and #4 and #5, respectively, for the right margin). Another spot
(#3) next to the river (in front of the Illa Sapinya) was also included.
On the basis of this 5-spot skeleton, the vehicle was driven to reach
the distinct locations. Several roads were driven both ways. The
start and end point was at the Canal Vell lagoon e Estacio  Biolo

gica
(#1).
2.2.3. Instrumentation
CH4 measurements were performed using a flight-ready CO2/
CH4/H2O PICARRO, INC analyzer (G-2301m). This instrument is
based on the cavity ring down spectroscopy technique (Crosson,
2008) and can measure CO2, CH4 and H2O data at a frequency of
1 Hz with a precision of <200 ppb for CO2, <1.5 ppb for CH4 and
<150 ppm for H2O according to manufacturer's specifications
(PICARRO, INC, 2010). This spectrometer and its associated vacuum
pump were powered by a smart-UPS device (SURT2000XLI, APC)
and 2-battery modules (SURT48XLBP, APC) that gave a runtime of
about 4 h. These devices were placed in the luggage compartment
of a car, which circulated at a mean velocity of 60 km h
1 during the
transects. At this velocity, one measurement could be taken every
17 m, approximately. This implies that broad emission plumes
could be detected, whereas sharp peaks could not. However, we
expected this velocity to be adequate for the area of study because
regions with similar CH4 emission patterns (rice fields, urban areas,
lagoons or grasslands) cover areas much larger than this length
value. The delay-time between air entering the inlet and the
measurement being recorded was calculated to be approximately
5 s (taking the length of the air inlet to be 4 m and an internal
diameter of the tubing to be 4.3 mm). This means that the air
measured was recorded ~83 m after it entered the inlet. This time-
delay was not accounted for, but the spatial resolution used in this
study is coarse enough (see Data analyses section) to consider that
this is not going to limit the results obtained.
The air inlet was placed in the front lower part of the car (at
around 40 cm above ground level) to avoid contamination from the
exhaust gases of the car itself. The inlet consisted of a methacrylate
cubic structure, with three orifices (about 10 mm diameter) on one
of the sides perpendicular to the ground. A 1/400 outside diameter
(OD) sampling line (Synflex, type 1300) was used to draw ambient
air from the inlet to the analyzer at a flow of 0.685 L min
1 passing
through a 0.7 mm filter. In addition, a GPS device (Map 62s, Garmin)
attached to a roof cross-bar of the car was used to register latitude
and longitude coordinates during the transects at a frequency of
1 Hz.
2.2.4. Agronomic state of paddy fields
From the Gantt diagram presented in Fig. 2, an idea of the
agronomic state of the fields could be inferred. Photographs taken


A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753 745

Table 1
Campaign ID, transect ID, date, duration, sample size, season of the year.

Campaign ID Transect ID Start time (dd/mm/yyyy hh:mm UTC) Duration (hh:mm) Sample size Season

C1 20111220Ev* 20/12/2011 14:04 03:39 153 Winter

20111221Mo 21/12/2011 07:08 03:59 5253
20111221Af 21/12/2011 14:03 03:25 5334
C2 20120123Ev 23/01/2012 15:38 03:28 4787 Winter
20120124Mo 24/01/2012 07:13 03:06 4938
20120124Af 24/01/2012 15:10 03:38 4729
C3 20120418Ev* 18/04/2012 17:35 03:46 4792 Spring
20120419Mo 19/04/2012 05:28 03:29 4987
20120419Af* 19/04/2012 12:44 02:21 3792
C4 20120507Ev 07/05/2012 16:46 02:32 4751 Spring
20120508Mo 08/05/2012 05:02 02:18 4830
20120508Af 08/05/2012 10:38 02:15 4921
C5 20120613Ev* 13/06/2012 18:23 02:15 4547 Spring
20120614Mo 14/06/2012 04:45 02:17 4678
20120614Af 14/06/2012 10:23 02:07 4630
C6 20120801Ev 01/08/2012 15:24 05:07 6015 Summer
20120802Mo 02/08/2012 04:31 03:54 4930
20120802Af* 02/08/2012 11:27 01:23 3832
C7 20120906Ev 06/09/2012 16:56 02:48 5061 Summer
20120907Mo* 07/09/2012 04:52 01:27 2358
20120907Af 07/09/2012 11:16 01:19 4207
C8 20120927Ev 27/09/2012 16:38 02:48 5562 Autumn
20120928Mo 28/09/2012 05:27 02:44 5793
20120928Af 28/09/2012 10:59 01:56 5122
C9 20121102Ev 02/11/2012 16:05 02:38 4849 Autumn
20121103Mo 03/11/2012 06:47 03:44 4826
20121103Af 03/11/2012 12:01 01:52 4739

Transect ID: It was named considering year (yyyy), month (mm), day (dd) and time of the day (Ev e evening; Mo e Morning; Af e Afternoon); e.g. 20111220Ev refers to the
evening transect (Ev) performed on 20 December 2011. Incomplete transects are indicated with *.

 Biolo
Fig. 3. Example of a typical routing of the transects (red line). Selected spots: #1 e Canal Vell lagoon e Estacio
gica; #2 e Desaigüe de la Unio ; #3 e Ebro river in front of Illa
Sapinya; #4 e Can Balada/Se
quia del Rampaire, #5 e La Tancada lagoon. Color legend of the selected spots: Blue e Close to natural wetlands; Red e Close to rice fields; Violet: Next
to the river. The location of the meteorological stations of the Meteorological Service of Catalonia (SMC) present in the area are also indicated in green (see Table 2 for details). Image
source: Google Earth; Data SIO, NOAA, US Navy, NGA, GEBCO.

during the campaigns were also useful to support the accompa-
nying descriptions.
During campaigns C1 and C2, fields from the right hemidelta
were in a condition of straw incorporation (partially flooded),
whereas fields from the left margin were dry. During campaign C3,
most of the rice-pads in the Delta were still fallow after the winter
period. During campaign C4, most of the fields were flooded
already, and sowing activities were being performed by then.
Before being sown, some fields of the left hemidelta had already
been, or were being, treated with saponins to fight against apple
snail pest.
Campaign C5 was carried out during the vegetative phase of rice
746

A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753
(before panicles appeared), whereas C6 was executed during the
reproductive stage (panicles were already present but flowering
had not yet started). Campaign C7 was performed late in the
maturation phase and harvesting operations were beginning by
then.
Fields of both hemideltas were already harvested at the time of
campaign C8. Some of them were flooded after harvesting opera-
tions to perform straw incorporation and mechanical weed man-
agement actions. During campaign C9, fields were still in conditions
of straw incorporation.
2.2.5. Meteorological and atmospheric information
Data from four stations of the Meteorological Service of Cata-
lonia (SMC), all located in the ERD region (Fig. 3 and Table 2), were
used to characterize the meteorological conditions during the
campaigns. Differences in meteorological conditions between
transects were obtained by calculating average values only over the
time period of each mobile survey.
Planetary boundary layer height (PBLH) data were extracted
from the operational high resolution atmospheric model of the
European Center for Medium-Range Weather Forecasting (ECMWF
HRES) (ECMWF, 2006) for the specific dates of the campaigns, and
also for the period 2011e2014. The horizontal spatial resolution of
this model is about 16 km. Average values were calculated for the
time period of each mobile survey. Also, a brief climatic study of the
variability of the PBLH at the ERD area was performed to attempt to
better explain the observed CH4 concentration variability.
2.3. Data analyses
2.3.1. GPS and CH4 measurements
Since the PICARRO, INC analyzer is extremely linear (Crosson,
2008), CH4 concentration raw data were corrected according to
the results of a single calibration performed during the campaigns
(2012/03/12). This calibration was carried out using seven cylinders
assigned by the Central Calibration Laboratory operated by the
Earth System Research Laboratory (ESRL) of the National Oceanic
and Atmospheric Administration (NOAA) (US Department of
Commerce). The cylinders used spanned a representative range of
CH4 values over which the analyzer typically operates
(1.795 ppme2.127 ppm) and a linear regression was applied (slope:
0.999; intercept: 0.010806; Residual Standard Error: 0.245 ppb).
Taking into consideration the time variable, latitude and longi-
tude coordinates were linked to calibrated CH4 concentrations.
Since in several campaigns there was a time difference between the
GPS (time automatically set) and the analyzer, a correction of 31 s,
at the most, had to be applied first to synchronize time measure-
ments of the spectrometer with the GPS signal. Time periods when
the car was stopped for several minutes (mostly at the five spots
previously indicated in Fig. 3) were removed from the data sets, to
avoid giving an extra importance to all the data points registered
there.
Table 2
Automatic meteorological stations from the Meteorological Service of Catalonia (SMC) located in the Ebro Delta region (Fig. 3). Abbreviations: a.s.l e above sea level; a.g.l. e
above ground level.
Station Latitude ( ) Longitude ( )
Els Alfacs 40.62725 0.65922
Illa de Buda 40.70719 0.83449
Amposta 40.70776 0.63210
L'Aldea 40.76947 0.61635
a
A conversion factor of 0.748 was applied to express wind velocity at 2 m, as it is used in agrometeorological studies (FAO, 2006).
2.3.2. Rasterization for spatial analysis
Due to the high density of the data points along transects, a
rasterization approach was used to pixel base our spatial analysis,
merge the extent of all transects and reduce the number of data to
work with. The different phases of the rasterization approach are
described here:
(1) CH4 values of the data points of each transect were
normalized to a common scale (0e1), in order to get some
spatial sense of the emission locations, but do not give a
sense of the seasonal or diurnal variability. This normaliza-
tion step was done using the following formula:
½Valuet
Mint =½Maxt
Mint 
where subscript t refers to a given transect; Maxt and Mint refer to
the maximum and minimum values registered during the t tran-
sect; and Valuet refers to each CH4 value associated to each data
point of the t transect.
(2) Afterwards, the three matrices (one per transect) containing
normalized CH4 values, latitude and longitude data were
bound together by rows, so that a unique matrix was ob-
tained per campaign.
(3) Nine hundred cells (30  30), of about 0.17 km2 each, were
selected to cover the spatial extent of all transects and the
value of each cell was determined by calculating the median
of the data points assigned to the cells.
2.3.3. Statistical analyses
Basic descriptive statistics were computed for each transect
(sample size, median, interquartile range, mean and standard de-
viation). We used box plots to show the variability in the data set,
although most of the data discussion throughout the paper was
performed considering conventional mean and standard deviation
values.
Given that available samples were not normally distributed and
the variance was not homogeneous, the Kruskal-Wallis test (non-
parametric test) was used to check if at least one of the groups was
different from at least one of the others. Then, a multiple compar-
ison test (as described in Siegel and Castellan (1988)) was applied
for determining which groups were different, with pairwise com-
parisons adjusted appropriately.
For the spatial analysis, campaigns which were rasterized ac-
cording to the methodology described in the previous section were
analyzed by performing a hierarchical cluster analysis. By applying
this approach we were evaluating whether grid cells could be
clustered considering the “time” variable as independent variable
(represented indirectly by the campaign number). Other variables
likely to influence spatial differences in ground level CH4 concen-
trations, such as soil salinity, type of soil or redox potential, were
not included in this analysis because of the lack of availability for
the ERD region.
Elevation of the station (m a.s.l.) Anemometer elevation (m a.g.l.)
0 2
a
0 10
a
3 10
62 2


A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753 747

The hierarchical cluster analysis was specifically performed by
taking the following steps:
(i) Pairwise dissimilarities between cell observations during the
different campaigns were computed using the Euclidean
distance method;
(ii) The resulting dissimilarity matrix was then used to perform a
hierarchical cluster analysis using the complete linkage
method;
(iii) Dendrograms were used to describe the output of the clus-
tering algorithm and display the relationship among the grid
cells.
3. Results
The scatter plots of the 27 transects analyzed in this study are
included in the supplement (Figs. S1eS3). They are presented using
a small range of CH4 values (from <1.9 ppm to 2.2 ppm) to better
highlight the spatial variability in CH4 concentrations. The low limit
value selected for this range (1.9 ppm) is slightly higher than the
background values (1.834e1.884 ppm), computed as CH4 NOAA
MBL (Marine Boundary Layer) surface reference for the latitude of
the ERD (Dlugokencky et al., 2015) (see Table S1). According to
Figs. S1eS3, regular values above 1.9 ppm were observed in most of
the transects, indicating a very strong CH4 source in the ERD area.
We selected a high limit value of 2.2 ppm because, as shown in the
box plot of each transect (see Fig. 4), most of the values registered
during the study were below 2.2 ppm (about 6 campaigns out of 9;
only C6, C8 and C9 presented much higher values).
3.1. CH4 measurements statistics
The box plots shown in Fig. 4 present median CH4 values of each
transect, together with the interquartile range (IQR ¼ Q3-Q1; see
Fig. 4 for further explanation) and the outliers (i.e. those values
specifically defined as <2nd percentile and >98th percentile).
The interquartile range was low (defined as IQR<0.1 ppm of
CH4) in all transects performed during campaigns C1, C3, C4 and C5.
In contrast, campaigns C2, C8 and C9 had higher variability with
IQR higher than 0.1 ppm for all transects. Campaigns C6 and C7 had
only morning and evening transects with high IQR.
The results of the multiple comparison test performed after the
Kruskal-Wallis test indicated that, pairwise, differences between
Evening (Ev), Morning (Mo) and Afternoon (Af) transects were
significant at p < 0.05 for all campaigns, except for campaign C2,
when the Mo transect was not significantly different from the Af
transect. According to these results and the values obtained for the
observed means (Table 3), we could sort the three transects of each
campaign from the highest to the lowest mean CH4 values, as
follows.
Spring and summer campaigns (C3-C7) had morning transects
with the highest mean values. However, the complete diurnal
pattern was not identical for all 5 campaigns. Late spring and
summer campaigns (C5-C7) presented the following pattern:
Mo>Ev>Af, whereas spring campaigns C3 and C4 showed the
pattern Mo>Af>Ev.
In winter and autumn campaigns (C1, C2, C8, C9) evening
transects had the highest mean values. Afternoon transects of these
campaigns presented the lowest values, except for campaign C1.

Fig. 4. Box plot of each transect. Nomenclature: Af e Afternoon; Ev eEvening; Mo e Morning; Ci - Campaign number (where i ¼ 1e9); W - Winter; A - Autumn; S - Spring; Su -
Summer. Each box goes from the first (Q1; 25%) to the third quartile (Q3; 75%). Within the box, a horizontal line is drawn at the median of the data set (50%). The lower whisker goes
from Q1 to the smallest non-outlier in the data set, and the upper whisker goes from Q3 to the largest non-outlier. Outliers defined as: >98th or <2nd percentile.
748

A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753

Table 3
Mean and standard deviation of each transect and each campaign. Transects classification within each campaign according to pairwise differences in mean CH4 values and the
results of the multiple comparison test.

Campaign ID Transect ID Transects Campaigns

Mean value [ppm] Standard deviation [ppm] Transects classification Mean value [ppm] Standard deviation [ppm]

C1 20111220Ev 1.908 0.053 Ev>Af>Mo 1.897 0.059

20111221Mo 1.889 0.054
20111221Af 1.906 0.063
C2 20120123Ev 2.148 0.358 Ev>Mo ¼ Af 2.103 0.425
20120124Mo 2.122 0.613
20120124Af 2.039 0.152
C3 20120418Ev 1.887 0.014 Mo>Af>Ev 1.893 0.022
20120419Mo 1.897 0.016
20120419Af 1.893 0.033
C4 20120507Ev 1.903 0.012 Mo>Af>Ev 1.950 0.104
20120508Mo 2.030 0.148
20120508Af 1.917 0.024
C5 20120613Ev 1.935 0.078 Mo>Ev>Af 1.931 0.084
20120614Mo 1.980 0.097
20120614Af 1.876 0.013
C6 20120801Ev 2.188 0.338 Mo>Ev>Af 2.142 0.309
20120802Mo 2.233 0.323
20120802Af 1.951 0.056
C7 20120906Ev 2.064 0.118 Mo>Ev>Af 2.333 0.812
20120907Mo 3.544 1.170
20120907Af 1.978 0.044
C8 20120927Ev 2.930 1.082 Ev>Mo>Af 2.459 0.751
20120928Mo 2.372 0.315
20120928Af 2.046 0.095
C9 20121102Ev 2.749 1.237 Ev>Mo>Af 2.466 1.024
20121103Mo 2.502 0.552
20121103Af 2.140 1.060

Thus, C8 and C9 showed this classification: Ev>Mo>Af; whereas C2
presented this one: Ev>Mo ¼ Af; and C1: Ev>Af>Mo.
Results of the multiple comparison test performed, considering
the nine campaigns as different levels, indicated that all campaigns
were significantly different at p < 0.05 except C8 and C9. Campaigns
could be ordered according to the values obtained for the observed
means (Table 3) and the results of the multiple comparison test. The
final ranking was the following one:
C3<C1<C5<C4<C2<C6<C7<C8¼C9.
The lowest absolute mean difference values (limiting value set
at 0.050 ppm) were obtained for C3-C1 (
0.005 ppm), C8-C9
(
0.007 ppm), C5-C4 (
0.019 ppm), C5-C1 (0.033 ppm), C5-C3
(0.038 ppm) and C6-C2 (0.038 ppm).
On the basis of the presented results, the campaigns could not
be grouped according to calendar seasons. In contrast, other pat-
terns were observed. The autumn campaigns (C8 and C9) showed
the highest mean CH4 concentrations and the highest variability,
whereas spring campaigns presented very low mean CH4 concen-
trations. In the winter campaign (C1) very low mean CH4 concen-
trations were measured, which will be analyzed in detail in the
Discussion section.
3.2. CH4 measurements spatial clusterization
The spatial variability of ground level CH4 concentrations was
evaluated on the basis of the values obtained after normalizing
each transect and binding together the three transects corre-
sponding to each campaign. The combined values were then ras-
terized, and afterwards the nine campaigns were clusterized. The
dendrogram obtained is presented in the left panel of Fig. 5. A three
group division was specified considering only the appearance of the
dendrogram; no specific quantitative criterion to set the height of
cut was established.
The spatial distribution of the 3 clusters obtained is presented in
the right panel of Fig. 5. The largest group (green coloured)
includes, not only cells covered by rice fields, but also cells placed
over urban areas. This group also includes the easternmost cells of
the rasterized path. The medium-size group (red coloured) is
distributed both at the right and the left margin, and includes only
cells containing rice fields.
The smallest group (violet coloured) is located at the northern
region of the Delta, over the TV-3451 road, and represents a group
of only 5 cells. In this area, according to the visual inspection of a
Google Earth Image, there is a high density at both sides of the road
of small-scale, shallow water storage structures normally used for
hunting or as water ponds (see Fig. S4). Although reservoirs of this
type are widespread over the left margin of the Delta (Fig. S4) and
some are close or within green and red cells on Fig. 5, the density is
much larger in the violet cells.
The box plots of the three clusters established in Fig. 5 are
shown in Fig. 6. The distribution of the values included in each
cluster shows significant differences, yet median values are not
significantly different. The Green cluster has a lower variability
(IQR ¼ 0.05) than the Red and Violet clusters (IQR ¼ 0.11 and 0.11,
respectively). The Violet cluster contains high outliers values (>0.4),
whereas all the values from the Green and Red clusters are below
0.4.
3.3. Meteorological and atmospheric conditions
Average meteorological conditions observed during each tran-
sect are presented in Table 4. A detailed description of the meteo-
rological conditions is given as supplementary information and
only the main findings are highlighted here.
The lowest mean temperature was registered during C2 and the
highest one during C6. The highest mean wind speed was observed
in C1with all other campaigns being performed with mean wind
velocities <3 m s
1. Wind direction was predominant from W (SW-
NW) from NoveApr (C9, C1, C2, C3), and from NE-SSW from
MayeSep (C4 e C8). The highest PBLH values were registered


A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753 749

Fig. 5. (a) Cluster dendrogram results with rectangles indicating the 3 groups division established. (b) Spatial distribution of the 3 groups identified, where green cells correspond to
those ones included in the green rectangle from the upper panel (idem for red and violet cells, respectively). Image source: Google Earth; TerraMetrics.

Fig. 6. Box plot of each of the three clusters set in Fig. 5. n: number of median values included in each box plot.

during the daytime (afternoon transects) from ApreNov (C3 eC9).
In Fig. 7, the box plots of the composite diurnal pattern of the
PBLH at ERD area divided by calendar seasons are presented. PBLH
reaches its minimum during the nighttime, due to weak vertical
atmospheric mixing, and its maximum during the daytime, due to
the insulation which facilitates the development of a vertical mix-
ing of the atmospheric layers. Winter and spring seasons are
characterized by a high variability of the PBLH compared with
summer and autumn seasons. The PBLH starts to increase during
the summer at 06 UTC, whereas its development in winter starts at
09 UTC. The highest diurnal PBLH values are observed during spring
and summer seasons, whereas the lowest values are observed in
winter.
4. Discussion
4.1. Diurnal variability
The diurnal variability between transects can be qualitatively
explained by considering the local meteorological conditions and
750

A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753

Table 4
Average meteorological conditions during the transects (calculated using data from four automatic stations of the Meteorological Service of Catalonia (SMC) located in the Ebro
River Delta area) and Planetary Boundary Layer Height (PBLH) (according to ECMWF model data).

Campaign ID Transect ID Temperature [ C] Relative humidity [%] Pressure [hPa] Wind speed [m s
1] Wind direction [ ] PBLH [m]

C1 20111220Ev 16.43 50.86 1018.62 3.4 315 1072

20111221Mo 16.37 51.56 1022.45 4.4 321 808
20111221Af 17.29 53.83 1021.09 4.1 308 901
C2 20120123Ev 13.05 72.27 1018.08 1.4 262 385
20120124Mo 12.62 57.22 1022.46 1.4 336 218
20120124Af 11.83 80.32 1021.89 1.5 236 144
C3 20120418Ev 14.67 59.16 1000.99 1.6 235 128
20120419Mo 16.07 41.16 999.61 2.2 306 1306
20120419Af 16.02 54.17 999.60 2.5 254 2526
C4 20120507Ev 17.98 72.60 1017.87 2.8 196 386
20120508Mo 15.49 82.63 1016.91 0.7 182 83
20120508Af 21.05 65.62 1016.54 1.6 143 1036
C5 20120613Ev 20.03 64.08 1016.98 0.6 195 455
20120614Mo 19.85 85.18 1017.05 1.1 73 608
20120614Af 22.87 70.75 1017.09 2.4 147 1124
C6 20120801Ev 26.59 81.01 1010.63 1.5 165 420
20120802Mo 25.39 77.59 1014.60 1.4 152 800
20120802Af 27.89 63.67 1014.39 2.1 115 1178
C7 20120906Ev 23.54 73.64 1019.06 2.4 203 441
20120907Mo 18.99 78.79 1021.29 0.7 332 56
20120907Af 27.38 62.33 1021.07 2.1 144 1800
C8 20120927Ev 19.94 67.88 1015.69 1.5 54 872
20120928Mo 18.21 72.54 1014.85 1.7 245 470
20120928Af 20.44 63.37 1013.54 3.0 42 1169
C9 20121102Ev 15.89 73.56 1012.71 0.9 263 67
20121103Mo 14.49 74.69 1010.28 0.7 201 83
20121103Af 17.93 67.33 1008.02 1.0 150 809

Fig. 7. Diurnal composite box plots of the PBLH calculated according to ECMWF model data (over 2011e2014) at the Ebro Delta area for each season. Each box goes from the first
(Q1; 25%) to the third quartile (Q3; 75%). Within the box, a horizontal line is drawn at the median of the data set (50%). The lower vertical black line goes from Q1 to the smallest
non-outlier in the data set, and the upper vertical black line goes from Q3 to the largest non-outlier. Outliers defined as >1.5 IQR or <1.5 IQR.

CH4 emission patterns. because, according to other authors, CH4 emissions are decreasing
We would expect the lowest mean CH4 concentrations in af- after midday (Chanton et al., 1997), early afternoon (Simpson et al.,
ternoon (Af) transects, mainly for two reasons. The first one is 1995; Miyata et al., 2000) or in the late evening (Meijide et al.,


A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753
2011), which in turn coincides in time with maximum transpiration
rates, air and leaf temperatures (Chanton et al., 1997), soil tem-
peratures (Simpson et al., 1995; Miyata et al., 2000) and water table
temperature (Meijide et al., 2011). Furthermore, stabilization and
destabilization processes of the atmosphere occur mainly in the
evening (Ev) and in the morning (Mo), respectively. Therefore, in
the afternoon there is a high convective mixing layer, as observed in
Fig. 7 that means that accumulation of CH4 at the ground level is
minimal. All campaigns, except C1, C3 and C4, matched this
reasoning. Campaign C1 was affected by strong synoptic winds (see
wind values in Table 4), coming from the NW which could have
overridden any local effects. Spring campaigns C3 and C4 presented
the lowest mean concentrations during the evening transect,
although mean differences between Af and Ev transects were very
low (0.006 ppm and 0.017 ppm, respectively) in comparison with
those from the campaigns that followed the trend (average abso-
lute value: 0.127 ppm). A possible reason for the differences in C3
and C4 could be because of the wind speed. Wind was blowing
strongly during the days before C3, thus fostering CH4 dispersion
and low background concentrations when the evening transect
began. During C4 wind was also blowing strongly during the eve-
ning transect (see wind values in Table 4) and air mixing was
probably still taking place then.
Regarding the highest mean CH4 concentration values, they
should be expected early in the morning if only local scale phe-
nomena occurred during the campaign, giving very stable and very
unstable classes alternating between night and day (Milla n et al.,
1991). Local CH4 emissions accumulated during the night, under
the shallow boundary layer, would be measured as high ground
level atmospheric CH4 concentrations early in the morning, if the
nocturnal boundary layer was still not completely broken up
(Fig. 7).
According to the modeled PBLH values available for the specific
dates, all the morning transects were performed when the
boundary layer was beginning to break up (note that the PBLH for
C4 and C7 was especially shallow (see Table 4)). Thus, according to
the reasoning stated previously Mo transects should present the
highest values if the PBLH dynamics was the only critical driver for
the surface concentration of CH4. All spring and summer campaigns
followed this trend. However, C1, C2, C8 and C9 did not; they pre-
sented higher Ev than Mo mean values.
As already mentioned, diurnal differences between transects
executed during campaign C1 are difficult to explain because strong
synoptic winds were blowing during those days, and also because
the evening transect of this campaign was incomplete. Campaigns
C2, C8 and C9 were executed in winter (campaign C2) and autumn
(campaigns C8 and C9), under low wind conditions, which facilitate
the accumulation of CH4 emitted by local sources, and when fields
were muddy (C2 e right margin, C8 and C9) or dried (C2 e left
margin).
No significant differences which could justify the different re-
sults were observed in the diurnal pattern of the meteorological
variables registered at the meteorological stations for the autumn
and winter transects C2, C8 and C9. Thus, we believe that the dif-
ference observed between winter-autumn and spring-summer
campaigns, regarding Ev and Mo transects, is due to a variation in
CH4 source emission patterns and its accumulation during the
lower winter-autumn PBL (Fig. 7).
As observed by Meijide et al. (2011), CH4 fluxes show a clear
diurnal cycle in CH4 emissions in systems where a water table is
present. During campaigns C2, C8 and C9 the water table was lower
than during spring-summer campaigns (these three campaigns
were performed when straw management activities were executed
and fields were only partially flooded). This could lead to a poorly
defined diurnal cycle that would reduce differences between
751
evening and morning concentrations, and even a reversed pattern.
Unfortunately, to the best of the authors' knowledge, experimental
data is not available in the literature to corroborate this assumption.
4.2. Seasonal variability
As the ERD region is affected by temperate weather conditions,
temperature and insulation variations throughout the year are not
specially marked here. The seasonal variability observed in ground
level CH4 concentrations throughout the 9 mobile surveys is
probably mainly related to crop cycle stages. In areas where rice is
grown, cultural practices and substrate availability for CH4 pro-
duction play a major role in the dynamics of CH4 emissions and,
consequently, in ground level atmospheric CH4 concentrations.
Although no specific measurements were carried out to charac-
terize substrate availability, results from other studies can be used
to explain the differences observed between campaigns.
The lowest CH4 values were observed during campaigns C1 and
C3. As previously mentioned, C3 was executed at the end of the dry
fallow period that affected both hemideltas for at least two and a
half months. At this time, fields were extremely dry and the
methanogenic activity was probably very low. The agronomic state
of the fields in campaign C1 was not the same as in C3. In fact,
higher CH4 concentration values should be expected for C1 because
rice straw decomposition, which favors CH4 emissions (e.g.
Serrano-Silva et al., 2014), was taking place under flooded condi-
tions in the fields from the right margin. The fact that both cam-
paigns present similarly low mean values can be attributed to the
meteorological conditions registered during C1, with very strong
wind velocities that could trigger a large footprint area, including
non-rice-dominated regions and disperse CH4 easily.
Campaigns performed when rice straw was in the fields (C2, C8
and C9) were chronologically ranged, although C6 and C7 fell in
between and C8 and C9 presented non-significant differences. This
subgroup arrangement can be explained by considering that: (i) C2
presented probably the lowest mean values of this subgroup
because only half of the ERD fields were under rice straw decom-
position conditions and the other half was dry; (ii) C8 was per-
formed right after harvest, and mechanical actions to incorporate
rice straw to the fields (which stir mud and favour trapped CH4 to
escape) had not been performed yet; (iii) In C9, fields were already
flooded after executing mechanical actions for straw incorporation,
so soil submersion (which promotes the establishment of anaero-
biosis), together with the fact that organic matter was incorporated
into the soil for a longer period than in C8 (and could be used as
substrate for methanogens) may explain why the highest mean
values were observed during this campaign.
Campaigns performed during the different phases of the rice
growing cycle (C5 e vegetative phase; C6 e reproductive phase; C7
e ripening phase) were also temporally ordered (although other
campaigns fell in between (C4, C2)). Plant density could be one of
the main factors explaining this result, since it increases
throughout the growing season and it has been correlated with CH4
emissions by providing C sources and by favouring CH4 transfer to
the atmosphere through the aerenchyma (Le Mer and Roger, 2001).
Despite this correlation, Khalil et al. (1998a) indicate that this does
not imply that there is a uniform change in CH4 emissions during
the growing season.
Campaign C4 presented higher mean CH4 concentrations than
C5. The difference between both campaigns was low (see Table 3).
The result C4>C5 is probably because especially high concentra-
tions were measured during the morning C4 transect (2.030 ppm
median value). On a vegetation level, C4 was performed during the
sowing period, whereas C5 was executed during the tilling stage of
the plants and when herbicides were applied to control weeds.
752

A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753
Thus, from a rice growth stage perspective, both campaigns did not
have great differences because plant-mediated processes for CH4
emissions were not present yet. Although C5 was executed during
the period when herbicides were applied and several products have
been reported to decrease CH4 emissions (Mohanty et al., 2001), it
is more likely that meteorological conditions trigger the differences
that both campaigns showed because these differences were only
particularly evident during the morning transects.
4.3. Spatial variability
According to the differences observed between clusters from the
box plots in Fig. 6, the Green cluster presents a low variability. This
could be related with the fact that: (i) urban areas, included in this
cluster, have seasonally low stable CH4 emission sources and; (ii)
the cells closer to the sea are more influenced by air with back-
ground CH4 concentration levels. Red and Violet clusters presented
a similar variability, though significantly high values were observed
for the Violet cluster, the one that included cells where water
storage structures are present in a high density. They referred to
median CH4 cell values of 3.340, 3.051, 3.027, 2.978 and 2.884 ppm
registered during the evening transect of campaign C6
(20120801Ev).
These structures might act as CH4 sources causing high CH4
concentration levels. However, since there is no specific inventory
of CH4 emissions in the ERD, these hotspots cannot be compared.
Clusterization groups obtained for the path followed during the
different mobile surveys were justified by considering differences
recognizable through Google Earth images (e.g. water storage
structures or proximity to urban areas). Other spatial differences
strictly related to environmental properties, such as salinity levels
or type of soil, could not be taken into consideration here because
they were not available in detail for the path of interest.
5. Conclusions
This study focuses on the characterization of ground level CH4
concentrations in the Ebro Delta region, located on a relatively
small peninsula. This region was surveyed using a mobile system
based on a PICARRO, INC spectrometer (G-2301m). Nine cam-
paigns, lasting two days each and consisting of three transects at
different times of the day (morning, afternoon and evening), were
executed over the period of a year. Seasonal and diurnal variations
of ground level CH4 concentrations were assessed, and spatial dif-
ferences determined.
Seasonal patterns seem to be related to agricultural practices
executed during the rice crop growing cycle, including rice straw
management activities performed after harvest, and water regime
during the fallow season. The greatest mean concentration
(2.466 ppm; standard deviation (SD) 1.024 ppm) was observed
during the campaign performed in November 2012, when fields
were flooded after incorporating straw.
The diurnal variability showed different patterns in spring and
summer in comparison with winter and autumn. The highest
ground level concentrations appeared early in the morning in
summer and spring (mean morning values: 1.897e3.544 ppm),
whereas in winter and autumn they were observed in the evening
(mean evening values: 2.148e2.930 ppm). This is probably related
to different CH4 emission patterns during the two periods (spring-
summer versus autumn-winter), while the planetary boundary
layer dynamics seems to be a non-determining factor. Further
studies should be carried out to characterize emissions patterns
over a whole year, including the fallow period. It is over this period
when very high CH4 concentrations are measured in the ERD.
According to the spatial variability, three main groups were
determined. The largest one included both rice fields and urban
areas; the smallest one covered an area where shallow water
storage structures are abundant; and the last group included areas
where urban agglomerations were absent. It would be desirable, in
the future, to assess whether other areas of the ERD region, not
surveyed here, could trigger a different spatial clusterization. It
would also be valuable to consider spatial biogeochemical data (e.g.
salinity levels, redox potential, phenological state of the vegetation)
of the area, for a better understanding of the spatial differences
which, in turn, could vary temporally.
The results presented in this study will be further applied using
modelling approaches to properly simulate the patterns obtained,
and to increase the knowledge base in this study area.
Acknowledgments
This work was supported by the ClimaDat project [Obra Social
“La Caixa”, 2010e2015]. Part of the research leading to these results
has received funding from the European Community's Seventh
Framework Programme (FP7/2007e2013) in the InGOS project
under grant agreement n 284274. The authors of this work are
thankful to the Meteorological Service of Catalonia for providing
~ ol
meteorological data. C.G. particularly thanks the Ministerio Espan
de Economía y Competitividad, for funding the project “Methane
interchange between soil and air over the Iberian peninsula”,
reference: CGL2013-46186-R. The authors thank the reviewers for
their valuable comments.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.apr.2017.01.009.
References
Alvarado-Aguilar, D., Jime nez, J.A., Nicholls, R.J., 2012. Flood hazard and damage
assessment in the Ebro Delta (NW Mediterranean) to relative sea level rise. Nat.
Hazards 62, 1301e1321. http://dx.doi.org/10.1007/s11069-012-0149-x.
Bamberger, I., Stieger, J., Buchmann, N., Eugster, W., 2014. Spatial variability of
methane: attributing atmospheric concentrations to emissions. Environ. Pollut.
190, 65e74. http://dx.doi.org/10.1016/j.envpol.2014.03.028.
, A., Loaso, C., Aragone
Bayo s, J.M., Custodio, E., 1993. Marine intrusion and brackish
 (Spain): a brief
water in coastal aquifers of southern Catalonia and Castello
survey of actual problems and circumstances. Study and modeling of saltwater
intrusion into aquifers. In: Proceedings 12th Saltwater Intrusion Meeting, Bar-
celona, November 1992. CIHS. CIMNE, Barcelona, pp. 741e766.
Bodelier, P.E., Roslev, P., Henckel, T., Frenzel, P., 2000. Stimulation by ammonium-
based fertilizers of methane oxidation in soil around rice roots. Nature 403,
421e424. http://dx.doi.org/10.1038/3500019.
Bronson, K.F., Singh, U., Neue, H.U., Abao Jr., E.B., 1997. Automated chamber mea-
surements of methane and nitrous oxide flux in a flooded rice soil: II. Fallow
period emissions. Soil Sci. Soc. Am. J. 61 (3), 988e993. http://dx.doi.org/
10.2136/sssaj1997.03615995006100030039x.
Casanova, D., 1998. Quantifying the Effects of Land Conditions on Rice Growth: a
Case Study in the Ebro Delta (Spain) Using Remote Sensing. PhD Thesis.
Wageningen University. ISBN 90 548 586 72.
Chanton, J.P., Whiting, G.J., Blair, N.E., Lindau, C.W., Bollich, P.K., 1997. Methane
emission from rice: stable isotopes, diurnal variations, and CO2 exchange. Glob.
Biogeochem. Cycles 11 (1), 15e27. http://dx.doi.org/10.1029/96GB03761.
Crosson, E.R., 2008. A cavity ring-down analyzer for measuring atmospheric levels
of methane, carbon dioxide, and water vapor. Appl. Phys. B 92 (3), 403e408.
http://dx.doi.org/10.1007/s00340-008-3135.
Dlugokencky, E.J., Lang, P.M., Crotwell, A.M., Masarie, K.A., Crotwell, M.J., 2015. At-
mospheric Methane Dry Air Mole Fractions from the NOAA ESRL Carbon Cycle
Cooperative Global Air Sampling Network. Data Path. ftp://aftp.cmdl.noaa.gov/
data/trace_gases/ch4/flask/surface/.
Eapi, G.R., Sabnis, M.S., Sattler, M.L., 2014. Mobile measurement of methane and
hydrogen sulfide at natural gas production site fence lines in the texas barnett
shale. J. Air Waste Manag. Assoc. 64 (8), 927e944. http://dx.doi.org/10.1080/
10962247.2014.907098.
European Center for Medium-Range Weather Forecasting (ECMWF), 2006. Diag-
nostic Boundary Layer Height, in IFS Documentation CY31r1, vol. 4. Physical
Processes, Reading, U. K. (Available at. http://www.ecmwf.int/sites/default/files/
elibrary/2007/9221-part-iv-physical-processes.pdf.


A. Agueda et al. / Atmospheric Pollution Research 8 (2017) 741e753
European Environment Agency (EEA), 2006. CORINE Land Cover (CLC) 2006,
v.18.5.1. Available at: http://land.copernicus.eu/pan-european/corine-land-
cover/clc-2006 [2017/01/13].
Farrell, P., Culling, D., Leifer, I., 2013. Transcontinental methane measurements: Part
1. A mobile surface platform for source investigations. Atmos. Environ. 74,
422e431. http://dx.doi.org/10.1016/j.atmosenv.2013.02.014.
Fatoric, S., Chelleri, L., 2012. vulnerability to the effects of climate change and
adaptation: the case of the spanish ebro delta. Ocean Coast. Manag. 60, 1e10.
Fitzgerald, G.J., Scow, K.M., Hill, J.E., 2000. Fallow season straw and water man-
agement effects on methane emissions in California rice. Glob. Biogeochem.
Cycles 14 (3), 767e776. http://dx.doi.org/10.1029/2000GB001259.
Food and Agriculture Organization of the United Nations (FAO), 1978. Manual on
Pond Culture of Penaeid Shrimp. 9. Predators, Competitors and Pests. ASEAN77/
SHR/CUL3. http://www.fao.org/docrep/field/003/ac006e/AC006E09.htm.
Food and Agriculture Organization of the United Nations (FAO), 2006. Evap-
otranspiracio n del cultivo. Guías para la determinacio  n de los requerimientos
de agua de los cultivos. PARTE A. Evapotranspiracio  n de referencia (ETo). ISSN
0254-5293. http://www.fao.org/3/a-x0490s/x0490s01.pdf.
Generalitat de Catalunya, 2008. Informe intermedio anual del programa de desar-
rollo rural de Catalunya para el período de programacio n 2007-2013 an ~ o 2007.
Available at. http://agricultura.gencat.cat/web/.content/09-desenvolupament-
rural/programa-desenvolupament-rural/periode-2007-2013/enllacos-docu-
ments/fitxers-binaris/informeanualpdr2007.pdf.
Generalitat de Catalunya, 2011. Plec de condicions de la denominacio  d’origen
protegida “Arro
s del Delta de l’Ebre” o “Arroz del Delta del Ebro”. Versio n
Decisio  n Favorable 9 de junio de 2011. Revisada per la informacio  com-
plement
aria CE del 24-02-2012. Available at. http://agricultura.gencat.cat/web/
sites/agricultura/.content/al_alimentacio/al02_qualitat_alimentaria/al02_04_
dop/documents/fitxers_estatics/plec_condicions_versio_favorable_catala_dop_
arros_delta_ebre.pdf [10/05/2016].


Grossi, C., Agueda, A., Vogel, F.R., Vargas, A., Zimnoch, M., Wach, P., Martín, J.E.,
Lopez-Coto, I., Bolívar, J.P., Morguí, J.-A., Rodo , X., 2016. Analysis of ground-
222
based Rn measurements over Spain: filling the gap in south-western
Europe. J. Geophys. Res. Atmos. 121 http://dx.doi.org/10.1002/2016JD025196.
Intergovernmental Panel on Climate Change (IPCC), 2013. Climate Change 2013-The
Physical Science Basis. Working Group I Contribution to the Fifth Assessment
Report of the IPCC. ISBN. 9781107661820.
Khalil, M.A.K., Rasmussen, R.A., Shearer, M.J., 1998a. Effects of production and
oxidation processes on methane emissions from rice fields. J. Geophys. Res.
Atmos. 103 (D19), 25233e25239. http://dx.doi.org/10.1029/98JD01116.
Khalil, M.A.K., Rasmussen, R.A., Shearer, M.J., Dalluge, R.W., Ren, L., Duan, C.-L.,
1998b. Factors affecting methane emissions from rice fields. J. Geophys. Res.
Atmos. 103 (D19), 25219e25231. http://dx.doi.org/10.1029/98JD01115.
Kirschke, S., Bousquet, P., Ciais, P., Saunois, M., Canadell, J.G., Dlugokencky, E.J.,
Bergamaschi, P., Bergmann, D., Blake, D.R., Bruhwiler, L., Cameron-Smith, P.,
Castaldi, S., Chevallier, F., Feng, L., Fraser, A., Heimann, M., Hodson, E.L.,
Houweling, S., Josse, B., Fraser, P.J., Krummel, P.B., Lamarque, J.-F.,
Langenfelds, R.L., Le Que r e
, C., Naik, V., O'Doherty, S., Palmer, P.I., Pison, I.,
Plummer, D., Poulter, B., Prinn, R.G., Rigby, M., Ringeval, B., Santini, M.,
Schmidt, M., Shindell, D.T., Simpson, I.J., Spahni, R., Steele, L.P., Strode, S.A.,
Sudo, K., Szopa, S., van der Werf, G.R., Voulgarakis, A., van Weele, M., Weiss, R.F.,
Williams, J.E., Zeng, G., 2013. Three decades of global methane sources and
sinks. Nat. Geosci. 6, 813e823. http://dx.doi.org/10.1038/NGEO1955.
Le Mer, J., Roger, P., 2001. Production, oxidation, emission and consumption of
methane by soils: a review. Eur. J. Soil Biol. 37 (1), 25e50. http://dx.doi.org/
10.1016/S1164-5563(01)01067-6.
Linquist, B.A., Adviento-Borbe, M.A., Pittelkow, C.M., van Kessel, C., van
Groenigen, K.J., 2012. Fertilizer management practices and greenhouse gas
emissions from rice systems: a quantitative review and analysis. Field Crops
Res. 135, 10e21. http://dx.doi.org/10.1016/j.fcr.2012.06.007.
Meijide, A., Manca, G., Goded, I., Magliulo, V., di Tommasi, P., Seufert, G., Cescatti, A.,
2011. Seasonal trends and environmental controls of methane emissions in a
rice paddy field in northern Italy. Biogeosciences 8, 3809e3821. http://
dx.doi.org/10.5194/bg-8-3809-2011.
Mill
an, M.M., Artin ~ ano, B., Alonso, L., Navazo, M., 1991. The effect of meso-scale
View publication stats
753
flows on regional and long-range atmospheric transport in the western Medi-
terranean area. Atmos. Environ. 25 (5e6), 949e963. http://dx.doi.org/10.1016/
0960-1686(91)90137-V.
Ministerio de Agricultura, Alimentacio n y Medio Ambiente, 2016. Sistema espan ~ ol
de inventario. Inventario GEI 2016-Serie 1990-2014. Tablas de datos de reporte
(CFR). ESP_Inventario GEI-2016. Datos de 2014. Summary1.As1. Available at.
http://www.magrama.gob.es/es/calidad-y-evaluacion-ambiental/temas/
sistema-espanol-de-inventario-sei-/.
Miyata, A., Leuning, R., Denmead, O.T., Kim, J., Harazono, Y., 2000. Carbon dioxide
and methane fluxes from an intermittently flooded paddy field. Agric. For.
Meteorol. 102, 287e303. http://dx.doi.org/10.1016/S0168-1923(00)00092-7.
Mohanty, S.R., Bharati, K., Moorthy, B.T.S., Ramakrishnan, B., Rao, V.R.,
Sethunathan, N., Adhya, K., 2001. Effect of the herbicide butachlor on methane
emission and ebullition flux from a direct-seeded flooded rice field. Biol. Fertil.
Soils 33 (3), 175e180. http://dx.doi.org/10.1007/s003740000301.
Mukherjee, R., Sarkar, U., Hobbs, S.E., 2012. Source areas for a passively diffusing
methane gas from South Asian paddy fields using flux footprint model. Atmos.
Res. 104e105, 154e163. http://dx.doi.org/10.1016/j.atmosres.2011.09.011.
Oo, A.Z., Win, K.T., Bellingrath-Kimura, S.D., 2015. Within field spatial variation in
methane emissions from lowland rice in Myanmar. SpringerPlus 4, 145. http://
dx.doi.org/10.1186/s40064-015-0901-2.
Pereira, J., Figueiredo, N., Goufo, P., Carneiro, J., Morais, R., Carranca, C., Coutinho, J.,
Trindade, H., 2013. Effects of elevated temperature and atmospheric carbon
dioxide concentration on the emissions of methane and nitrous oxide from
Portuguese flooded rice fields. Atmos. Environ. 80, 464e471. http://dx.doi.org/
10.1016/j.atmosenv.2013.08.045.
tron, G., Frost, G., Miller, B.R., Hirsch, A.I., Montzka, S.A., Karion, A., Trainer, M.,
Pe
Sweeney, C., Andrews, A.E., Miller, L., Kofler, J., Bar-Ilan, A., Dlugokencky, E.J.,
Patrick, L., Moore Jr., C.T., Ryerson, T.B., Siso, C., Kolodzey, W., Lang, P.M.,
Conway, T., Novelli, P., Masarie, K., Hall, B., Guenther, D., Kitzis, D., Miller, J.,
Welsh, D., Wolfe, D., Neff, W., Tans, P., 2012. Hydrocarbon emissions charac-
terization in the Colorado Front Range: a pilot study. J. Geophys. Res. 117,
D04304. http://dx.doi.org/10.1029/2011JD016360.
Phillips, N.G., Ackley, R., Crosson, E.R., Down, A., Hutyra, L.R., Brondfield, M.,
Karr, J.D., Zhao, K., Jackson, R.B., 2013. Mapping urban pipeline leaks: methane
leaks across Boston. Environ. Pollut. Rapid Commun. 173, 1e4. http://dx.doi.org/
10.1016/j.envpol.2012.11.003.
PICARRO, INC, 2010. CRDS Analyzer CO2 þ CH4 þ H2O for Flight Model G2301-m
(Datasheet).
RAMSAR, 2016. Wetlands of International Importance (RAMSAR Sites). http://www.
ramsar.org/sites-countries/the-ramsar-sites.
Serrano-Silva, N., Sarria-Guzma n, Y., Dendooven, L., Luna-Guido, M., 2014. Meth-
anogenesis and methanotrophy in soil: a review. Pedosphere 24 (3), 291e307.
http://dx.doi.org/10.1016/S1002-0160(14)60016-3.
Shorter, J.H., MacManus, J.B., Kolb, C.E., Allwine, E.J., Lamb, B.K., Mosher, B.W.,
Harriss, R.C., Partchatka, U., Fischer, H., Harris, G.W., Crutzen, P.J., Karbach, H.-J.,
1996. Methane emission measurements in urban areas in Eastern Germany.
J. Atmos. Chem. 24 (2), 121e140. http://dx.doi.org/10.1007/BF00162407.
Siegel, S., Castellan, N.J., 1988. Non Parametric Statistics for the Behavioural Sci-
ences. MacGraw Hill Int., New York, pp. 213e214.
Simpson, I.J., Thurtell, G.W., Kidd, G.E., Lin, M., Demetriades-Shah, T.H., Flitcroft, I.D.,
Kanemasu, E.T., Nie, D., Bronson, K.F., Neue, H.U., 1995. Tunable diode laser
measurements of methane fluxes from an irrigated rice paddy field in the
Philippines. J. Geophys. Res. Atmos. 100 (D4), 7283e7290. http://dx.doi.org/
10.1029/94JD03326.
Yan, X., Yagi, K., Akiyama, H., Akimoto, H., 2005. Statistical analysis of the major
variables controlling methane emission from rice fields. Glob. Change Biol. 11,
1131e1141. http://dx.doi.org/10.1111/j.1365-2486.2005.00976.x.
Zazzeri, G., Lowry, D., Fisher, R.E., France, J.L., Lanoiselle , M., Nisbet, E.G., 2015.
Plume mapping and isotopic characterisation of anthropogenic methane sour-
ces. Atmos. Environ. 110, 151e162. http://dx.doi.org/10.1016/
j.atmosenv.2015.03.029.
Zhang, G., Zhang, X., Ji, Y., Ma, J., Xu, H., Cai, Z., 2011. Carbon isotopic composition,
methanogenic pathway, and fraction of CH4 oxidized in a rice field flooded year-
round. J. Geophys. Res. 116 (G04) http://dx.doi.org/10.1029/2011JG001696.