DETERMINATION OF METHANE SOURCEX GLOBALLY BY SCIAMACHY

Since the beginning of the industrial revolution, the atmospheric concentration of carbon dioxide (CO2) has increased by nearly 30%, and the Methane (CH4) concentration has more than doubled. CH4 is the second most important greenhouse gas, after CO2. Emissions, extrapolated from measurements of actual gas flux from wetlands, vary from place to place, even within the same wetland. This high variability makes large-scale estimates difficult and means that average emissions levels include a large degree of estimated uncertainty. The SCIAMACHY instrument on the European Space Agency satellite ENVISAT measured greenhouse gases in the troposphere and stratosphere. In this study, the CH4 source area is extracted by estimating the concentrations of methane emissions from time-series satellite data. Contamination of the data by cloud is interpolated both spatially and temporally. It is assumed that methane emission is negligible over ocean and that the concentration in the ocean area is due to advection from land. Background CH4 concentration on land was defined as the ocean CH4 concentration at the same latitude. Land CH4 emission concentrations show that areas of concentrated high CH4 emission are not in paddy fields only but also in broadleaf evergreen areas in South America and Central Africa.


INTRODUCTION
Looking at the long-term trend, the global average temperature increased by 0.74 ± 0.18 °C over the 100-year period ending in 2005 (IPCC, 2007).If global warming continues in the same way, changes in temperature and precipitation patterns are expected over the long term of 50 to 100 years, while changes in the intensity and frequency of extreme weather conditions are expected over the short term.Extreme weather conditions, in particular, often lead to massive damage and analysis of them is an important topic in the context of analyzing global warming (Trenberth and Shea, 2005).Much research has found that the main cause of global warming is increased greenhouse gases in the atmosphere due to the use of fossil fuels by humans and is not a natural phenomenon (IPCC, 2007;Houweling et al., 2000).Greenhouse gases are materials in the atmosphere that absorb and re-emit some of the infrared light radiated from the surface of the earth.Carbon dioxide (CO2), methane (CH4), dinitrogen monoxide (N2O) and chlorofluorocarbon gases are all greenhouse gases.The global warming potential (GWP) is a numeric indicator that represents how much a particular greenhouse gas contributes to global warming.GWP is defined as the integral of the amount of radiative energy emitted into the atmosphere per unit mass of greenhouse gas over some fixed period (such as 20 years, 100 years, or 500 years), with the values taken relative to CO2.Across 100 years, CH4 has a global warming effect of 25 (that is, 25-fold the effect of CO2), N2O has an effect of 310, and chlorofluorocarbons have an effect of several thousand to tens of thousands (IPCC, 2007;Forester et al., 2007).Comparing the concentration levels present before the industrial revolution with the mean concentrations in 2012, CO2 concentrations have increased from 278 ppm to 393.1 ± 0.1 ppm (a 41% increase), CH4 has increased from 700 ppb to 1819 ± 1 ppb (160%), and N2O has increased from 270 ppb to 325 ± 0.1 ppb (20%).The increases in the greenhouse gases since the industrial revolution as characterized by the GWP are 63% for CO2, 18% for CH4, 6% for N2O, and 13% for other (IPCC, 2007).This is found by multiplying the proportions of gases present with the GWP for each gas.
Observation of CO2 as a greenhouse gas began in 1957 at the South Pole and in 1958 at Mauna Loa, Hawaii.An increase in CO2 was clearly seen from the observations at Mauna Loa, and this was one of the catalysts for the theory of international global warming.Measurements of greenhouse gases have been conducted by the World Data Center for Greenhouse Gases (WDCGG).The WDCGG was established under the Global Atmosphere Watch (GAW) process of the World Meteorological Organization (WMO), and the WDCGG collects, manages, and provides data on gases (CO, NOX, SO2, etc.) related to greenhouse gases (CO2, CH4, CFCs, N2O) measured in the atmosphere and ocean.Satellite measurement data are available from the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) sensor mounted on the Environmental Satellite (ENVISAT), which was launched on 1 March 2002 by the European Space Agency (ESA) (Burrows et al., 1995) and from the Greenhouse Gases Observing Satellite (GOSAT), which was launched on 23 January 2009 through cooperation between the Japan Aerospace Exploration Agency (JAXA), the National Institute for Environmental Studies (NIES) of Japan, and the Ministry of the Environment (MOE) of Japan.Although the greenhouse gas data measured by the WDCGG cover only a narrow range of geographical locations relative to the globe, satellite collection allows measurement over two-dimensional areas and is often used in research covering wide areas (Berggamaschi et al., 2007;Bracher et al., 2005;Buchwitz et al., 2006;Frankenberg et al., 2005).As global warming intensifies, CH4 that has accumulated in permafrost and CH4 that is trapped in ice is released into the atmosphere, and this drives further warming.From 1974 to 2000, the area of permafrost lakes increased by 14.7% and emissions increased by 58% (Walter et al., 2006).Furthermore, it was recently found that CH4 is also emitted from tropical rainforest and grassland (Keppler et al., 2006;Wang et al., 2008).Sources of CH4 are broadly divided into biological sources and non-biological sources.Non-biological sources of CH4 emissions include mining and burning of fossil fuels, burning of biomass, and garbage disposal.Biological sources include swamps, wet rice paddies, livestock, landfill, termites, and forests.In the complex processes of biological ecosystems, CO2 is produced when organic matter is decomposed by anaerobic bacteria (via fermentation and putrescence), and CH4 is produced by an abundance of CO2 and methane bacteria (Conrad, 1996).Biological sources account for over 70% of total CH4 emissions (IPCC, 2007).Emitters of CH4 include natural and artificial sources.Among natural sources, the largest sources of emissions are bogs and swamps, with emissions also occurring from termites and vegetation (Lelieveld et al., 1998;IPCC, 2007).Among artificial sources, the main sources are wet rice paddies, livestock, waste disposal processes, and mining and use of fossil fuels (Houweling et al., 2000;Ruddiman, 2003).Although there are also sinks for CH4, including absorption by soil (5%) (IPCC, 2007 andInagaki et al., 2004), decomposition by chemical reaction in the atmosphere accounts for over 87% of CH4 elimination.Decomposition by chemical reaction occurs via reaction with hydroxide radicals (OH radicals) by means of chemical reactions involving sunlight, oxygen, ozone, water vapor, etc. (Brasseur et al., 1999).The amount of CH4 that is emitted from sources such as swamps and wet rice paddies has been clarified by many field observations (Walter and Heimann, 2001a,b;Christemsen et al., 2003;Wickland et al., 2006).However, the amount of CH4 emitted varies according to differences in the observation targets and environment.Furthermore, it is difficult to apply data from a narrow observation region to a more global region.Research has therefore also been conducted by using satellite remote sensing technology, which offers wide geographic ranges and temporal periodicity.Frankenberg (2005) shows that CH4 concentrations are high in tropical rainforest zones.However, since the background CH4 concentration was not excluded, accurately determining the source of CH4 was difficult.There are many details of emissions that are insufficiently understood to accurately identify emission sources with the emission quantities in the CH4 balance.For example, the amount of CH4 emitted differs with the model and the field survey method.Estimates of the amount of anthropogenic methane emitted vary greatly between research methods, with a low value of 264 Tg(CH4)yr-1 given by by Scheele et al. (2002), a value of 358 Tg(CH4)yr-1 by Wuebbles and Hayhoe (2002), and a high value of 428 Tg(CH4)yr-1 by Chen and Prinn (2006) using the global inverse model.Furthermore, although the concentration of methane in the atmosphere remained virtually constant from 1999 to 2006, it began increasing again in 2007 (Rigby et al., 2008).It has been noted (WMO Greenhouse Gas Report, 2014) that no variations in the amount of emissions from the Arctic Circle have been observed, despite CH4 emissions from the tropical zones and lower latitudes in the northern hemisphere, and the exact reason for the increase is currently unknown.
In this research, we therefore estimated the surface concentration of methane emissions and extracted the CH4 sources by using time-series data from SCIAMACHY covering the period from 2003 to 2011.We treat the surface background CH4 concentration, which has been problematic in other research, as the CH4 concentration over ocean at the same latitude, and then calculate the surface concentration of CH4 emissions under this assumption.Furthermore, we compare the topographical features of CH4 sources with vegetation maps to investigate the sources of CH4 emissions..

SCIAMACHY
SCIAMACHY was mounted on ENVISAT and operated from February 2002 to April 2012.The main aim of SCIAMACHY was to measure the various greenhouse gases (including O3, BrO, SO2, HCHO, OCIO, H2O, CH4, CO2, N2O, CO, and NO2) and atmospheric pollutants in the troposphere and stratosphere.SCIAMACHY had 8 spectroscopic channels, ranging from ultraviolet to near infrared range (240 to 2380 nm).The ENVISAT satellite was a polar orbiting satellite that orbited Earth 14 times per day at an altitude of 800 km in space.In the data, the measurement width is 960 km and the surface resolution is 30 km north-south and 60 to 120 km east-west.The vertical column density (VCD) data for CH4 are retrievable by using the iterative maximum a posteriori (IMAP)-DOAS method of Frankenberg et al. (2005).The dataset used in this research was downloaded from the SCIAMACHY website (http://www.sciamachy.org/products/index.php?species=CH4).

GLCMNO
The Global Land Cover by National Mapping Organizations (GLCNMO) was created by the Geospatial Information Authority of Japan and by Chiba University through cooperation with the National Mapping Organization (NMO), in which 181 countries around the world participate.The land cover layer of the GLCNMO is available as a cluster dataset divided into 20 items according to the land cover classification system (LCCS) developed by the Food and Agriculture Organization of the United Nations (FAO).The landcover legend includes evergreen broad-leaf forest, deciduous broadleaf forest, evergreen needle-leaf forest, deciduous needle-leaf forest, mixed forest, grassland, fields, wet rice paddies, mangroves, swamps, bare ground (sand), and bare ground (pebbles and stone) (Tateishi et al., 2014).This research uses the first revision of the 1-km resolution data.The first revision is the result of categorizing by using the MODIS data from 2003.GNCNMO performs sampling at a resolution of 30 × 60 km to perform comparison with the CH4 source distribution.During sampling, the proportion of land cover for each pixel is calculated.

Cloud-free Images
Optical sensor data obtained from satellites are often unclear near the ground surface due to the effects of clouds.We therefore created cloud-free images by determining whether or not an observation point was affected by cloud and using only those data that were not affected by cloud.We used the observation data obtained during a prior pass to make this determination when generating the maps.We used the threshold data (selection_criteria.txt) from the SCIAMACHY product site (ftp://ftp.sron.nl/pub/pub/DataProducts/SCIAMACHY_CH4/)as the threshold values when creating the cloud-free images.

Spatial Interpolation and Time-series Interpolation
When a large proportion of data is missing due to the effects of cloud, analysis becomes difficult.We used spatial and timeseries interpolation to impute missing data.For this, we note that the concentration of CH4 in the atmosphere varies depending on the amounts of CH4 emitted and destroyed.Furthermore, the concentration also varies, depending on advection from the surroundings.In order to accurately estimate the missing data, therefore, it is necessary to reproduce the atmospheric trace components and variations.We use a threedimensional chemistry-transport model for this.However, many of the data items needed in order to accurately employ a threedimensional chemistry-transport model are not easy to obtain, particularly the auxiliary data and initial values required by the model, so it is difficult to process globally.In this research, we therefore assumed that the concentration of CH4 varied gradually with the surrounding concentration environment, and specifically that there are no sudden variations over short periods of time across narrow ranges.In this work, we did not consider the effects of air currents that vary by altitude.Under the above assumptions, we used spatial interpolation and time-series interpolation to simply estimate the missing data.For spatial interpolation of the missing data, we used a weighted spatial filter with a size of 7 × 5 pixels.First, we divided the 34 pixels surrounding the target pixel into 8 regions, and if pixels without cloud existed in at least one pair of opposite directions (0° and 180°, 45° and 225°, 90° and 270°, 135° and 315°), the value for the target pixel was interpolated by applying the 7 × 5 weighted filter.For this, the weighting was applied inversely linearly proportional to the distance from the target pixel.A necessary condition for time-series interpolation of CH4 concentration in the atmosphere was that the CH4 concentration did not change sharply over a short period of time (here, less than 30 days).For example, if a target pixel was missing, the value was found by linear interpolation using the data before (15 days before) and after (15 days after) the target pixel.Where two sets (30 days) or more of consecutive data were missing, interpolation was not performed.Figure 1 shows an example of time-series interpolation by using preceding and following time period data for the case of a missing target pixel.When the fifth data point is missing, linear interpolation is performed using the fourth (1728) and sixth (1732) data points.However, for data points 21 and 22, interpolation is not performed because consecutive data points are missing.

Concentration of CH4 Emission
The CH4 concentrations in ocean regions at lower latitudes of the northern hemisphere are higher than in continental regions of the southern hemisphere, despite the fact that virtually no CH4 is emitted over ocean regions.In the lower latitudes of the northern hemisphere, there is a large distribution of wet rice paddies, which are a CH4 source, and the CH4 that is emitted over land flows by advection to neighboring ocean regions.Figure 2 shows a simplified overview diagram of advection of the CH4 emitted over land into the atmosphere.The CH4 concentrations that are observed by satellite are the CH4 concentrations that exist in the atmosphere directly under the satellite, rather than the CH4 concentrations emitted from the land surface directly under the satellite.To determine the CH4 concentration emitted from the land surface (defined here as the CH4 emission concentration), the CH4 concentration that already exists in the atmosphere (i.e., background CH4 concentration) needs to be known.However, the background CH4 concentration cannot be found easily because it varies with location and time.Emission of CH4 in ocean regions is extremely low in comparison with that in continental regions, and the CH4 observed in continental regions is mainly due to advection from land regions.Although air currents vary complexly depending on autorotation, solar heating, terrain, and altitude, in this research we simplified the treatment of air currents.To simplify the air currents, we treat the CH4 concentration in ocean regions at a given latitude as the background CH4 concentration over continental regions at the same latitude.The CH4 emission concentration at a location is then found by determining the difference between the CH4 concentration over continental regions and the background CH4 concentration (ocean-region methane concentration) at the same latitude (Eq.1).
CH4 emission concentration = continental CH4 concentrationocean CH4 concentration (average at same latitude) (1) Figure 2 Overview diagram of advection of CH4 emitted over land into the atmosphere 4. RESULT

3-stage Interpolation for cloud-free image
In this research, we perform 3-stage processing (spatial interpolation followed by time-series interpolation followed by further spatial interpolation) in order to interpolate for the missing data.Figure 3

Time-series Variations in Methane Concentration
The concentration of CH4 in the atmosphere varies with region and time.Figure 4 shows the changes in CH4 concentration between January, April, July, and October 2005.In the diagrams, higher (resp., lower) concentration of CH4 is indicated by redder (resp., bluer) color.In regions of high elevation, such as the Tibetan Plateau and the Andes, the concentration of CH4 is low.When the elevation is high, a lower concentration of CH4 concentration is observed because the distance from the upper edge of the atmosphere to the surface is smaller.
In the image for July, even in desert regions, where CH4 is not emitted, the CH4 concentration was 1730 ppb in the Sahara Desert in Africa and 1670 ppb in the Great Sandy Desert in Australia, with a difference of approximately 60 ppb in CH4 concentration.There is a wide distribution of wet rice paddies in Southeast Asia, in the same latitude range as the Sahara Desert, and the CH4 emitted from the wet rice paddies is thought to be transported via advection to the desert.Since there are few CH4 sources in the same latitude range as the Great Sandy Desert and a large area of ocean, advection from other locations is small.Differences in CH4 concentration in the atmosphere between the two desert regions can be attributed to differences in background concentration.
In the image for October, there is clearly advection of CH4 , which is thought to have been emitted in western Africa and central Africa in July and later appeared over the Atlantic Ocean and Pacific Ocean.The advection is from the point of emission towards the west, and is thought to be due to the effect of westerly winds from the north east near the equator.Most regions in the northern hemisphere exhibit concentrations of 1700 to 1760 ppb, while the southern hemisphere exhibits concentrations 1660 to 1740 ppb, which is lower than in the northern hemisphere.This is because the background concentration that exists in the atmosphere is lower in the southern hemisphere because there are fewer CH4 emission sources than in the northern hemisphere.Regions where the CH4 concentration is high include the tropical rainforests of southern America and central Africa in January (1780 to 1850 ppb) and Southeast Asia from May to October (1850 to 1920 ppb).The regions of high CH4 concentration are thought to arise from differences in land use and cover.In many of the pioneering research studies, it has been reported that CH4 is emitted by wet rice paddies (Inubushi et al. 1989;Inubusuh et al. 1994;Kumagai and Konno 1998;Holzapfel and Seiler 1986) and there have also been reports noting that CH4 is emitted by other vegetation (Keppler et al., 2006;Wang et al., 2008).
It can be judged whether the CH4 concentration is high or low by using atmospheric CH4 concentrations, which include the background CH4 concentration.However, whether the CH4 emission concentration is high or low cannot be seen from data that include background concentration.We therefore need to remove the background concentration to allow comparison of the CH4 emission concentrations over continental regions.

CH4 Emission Concentration
Using the levels of atmospheric CH4 concentration as measured by a satellite-based instrument to judge whether the amounts of CH4 emitted from ground regions are large or small is problematic.For example, the CH4 concentration of 1720 ppb in the Sahara Desert in July 2005 (Figure 4) is higher than the concentration of 1670 ppb over the forest region of Brazil.Despite this, virtually no CH4 is emitted from desert regions with little rain.In contrast, CH4 is emitted from vegetation and from the soil in the forest regions of tropical rainforests, and certainly more CH4 is emitted from these than from desert regions.Although the concentration of CH4 in the atmosphere is affected by the CH4 emitted from the surface, it is also strongly influenced by the background CH4 concentration, which is already present in the atmosphere.In order to investigate the amount of CH4 emitted from the surface, it is essential to know the level of background concentration.We therefore assumed that the mean CH4 concentration over ocean regions is the background concentration for continental regions at the same latitude (Figure 2). Figure 6 shows the CH4 emission concentrations as obtained by subtracting the background CH4 concentrations at the same latitudes from the CH4 concentrations in Figure 4.It is clear that a maximum of 100 ppb of CH4 is emitted from the wet rice paddy region of Southeast Asia and 60 ppb or more is emitted from West Africa and central Africa in July, and a maximum of 80 ppb is emitted from South America from October to April.Although the Sahara Desert exhibits emission concentrations close to virtually 0 ppb, in April the methane emission concentrations are under-evaluated, at -10 ppb.This is because the CH4 concentration on the Pacific Ocean side at the same latitude is 10 ppb higher than in the desert region.This may be due to the effect of OH radical reactions (which are a methane sink), or the removal of cloud from the data for desert regions may not be sufficient; this needs to be investigated in detail in the future.In regions of high elevation, the emission concentrations are under-estimated because the atmospheric CH4 concentrations are low.

Relation between CH4 Sources and Land Cover
To investigate the land cover characteristics of CH4 sources, we performed a comparison with GLCNMO.The coastal region of Cameroon consists of 80% evergreen broad-leaf forest and 8% fields.This shows that the land cover of CH4 sources in Africa consists of a lot of evergreen broadleaf forest.The high-latitude regions in the northern hemisphere, which contain a lot of swamp, are missing from the analysis because sufficient data could not be obtained due to the solar zenith angle and the effect of cloud.

CONCLUSIONS
When investigating CH4 source distribution by using data measured via satellites, it is necessary to know the existing background concentration.In this research, we assumed that since CH4 is not emitted in ocean regions, any CH4 that does exist in the ocean regions is due to the effect of advection from continental regions.Furthermore, we simplify the complex atmospheric cycle model by treating the CH4 over the ocean to be advection from continental regions at the same latitude, as shown in Figure 2, and define the CH4 concentration over ocean regions as the background concentration at the same latitude.
As a result the surface CH4 emissions concentrations are found from the difference between the surface CH4 concentration and the CH4 concentration in ocean regions at the same latitude.
In the IPCC (2007) presentation, the largest anthropogenic source of CH4 emissions was ruminant livestock (23-44%) followed by wet rice paddies (12-26%).From the results for the annual CH4 emissions concentrations (Figure 7), the wet rice paddy regions at the lower latitudes in the northern hemisphere, which are well-known as an existing source, were confirmed to be the largest anthropogenic source of CH4.However, we newly found that tropical evergreen broad-leaf forests in South America and central and western Africa are CH4 sources, which was not previously widely known.Since these regions have a small proportion of fields and similar, it is thought that the CH4 emissions sources are natural sources, rather than artificial sources.At the IPCC meeting (2007), it was presented that swamp and termites are natural sources of CH4 emission.
Although swamps can also be considered as a natural CH4 source, the land cover proportion of swamp in the GLCNMO dataset is extremely small.Termites are also a CH4 emission source in tropical rainforest regions.Termites are widely distributed throughout tropical, subtropical, and temperate zones.Termites mainly consume the fiber (cellulose) of vegetation and emit CH4 through the process of fermentation of cellulose by microorganisms in their digestive organs.However, since there is little research on accurate distributions and quantities of termites, calculating the level of contribution by termites to CH4 emissions is difficult.The land cover of natural CH4 sources was found to be primarily evergreen broad-leaf forest.Although some research has indicated that CH4 is emitted from vegetation, later research found that it is emitted primarily under special environmental conditions, such as strong ultraviolet light or high temperatures.From Figure 6, although seasonal variation occurs in the amount of CH4 emitted, this seasonal variation is not CH4 produced from the action of photosynthesis.Since the timeseries normalized difference vegetation index of evergreen broad-leaf forest has little annual variation, the photosynthesis activity is also thought to have little annual variation.Furthermore, not all of the evergreen broad-leaf forests are CH4 sources, as shown in Figure 7(c).If CH4 is emitted from evergreen broad-leaf forests, then the CH4 source would be expected to extend even further into Brazil.Seasonal variation in CH4 emission concentration in central and western Africa and South America is strongly tied to the wet season.According to the Koppen climate classification, these three regions have climates of tropical rainforest, savannah, and tropical monsoon.The tropical rainforest climate has a lot of rain year-round, and the savannah climate and tropical monsoon climate are divided into a wet season (in central Africa the wet season is from May to July) and a dry season (in Brazil the dry season is from May to September).When the wet season arrives, the ground surface temporarily changes into an anaerobic environment due to the abundance of rain, and CH4 is readily emitted as a result.CH4 that is produced underground is emitted above ground from the leaves of plants by evapotranspiration of the plant (Megonigal and Guenther, 2008).Furthermore, it is thought that CH4 that is produced near the ground surface diffuses directly into the atmosphere.

FUTURE WORK
In this research, we used satellite data to investigate the distribution of CH4 sources.As a result, we found that a lot of CH4 is emitted from evergreen broad-leaf forest regions in tropical rainforest climates, in addition to the known emissions from wet rice paddies.Although the IPCC (2007) report estimated that much more CH4 emissions come from anthropogenic sources than from natural sources, the results of this research showed that natural CH4 sources are more widely distributed than anthropogenic sources.However, changes in the amount of emissions from natural sources are expected to be small unless there are notable changes in the environment.The extraordinary increase in CH4 since the industrial revolution is clearly the result of anthropogenic sources.As problems for the future, we need to investigate the timeseries changes in the CH4 sources that were found in this research and clarify the cause of increases in CH4 since 2007.The emission sources of CH4 can be investigated in more detail by investigating time-series changes in CH4 emissions, timeseries changes in vegetation, and time-series changes in temperature and precipitation.
Figure 3. 3-stage interpolation of data missing due to the effects of cloud

Figure 5
Figure 4. Time-series changes in methane concentration

Figure 5 .
Figure 5. Interannual variation in mean CH4 concentration by latitude zone over continental regions Figure 6.Time-series CH4 emission concentration over continental regions Figure 7. Mean annual CH4 emission concentration for 2009 and land cover coverage of rice paddies and evergreen broadleaf forest