4
Emissions Estimated from Atmospheric and Oceanic Measurements

The direct way to measure emissions of greenhouse gases from a source is to collect and analyze exhaust gases as they are emitted. To estimate regional or national emissions from a remote location, one can take advantage of the incomplete mixing of the atmosphere. Because of incomplete mixing, geographical variations in greenhouse gas emissions and removals cause the abundance of a gas to be elevated downwind of a source and reduced downwind of a sink. Measurements of gas abundances around the globe can be used to estimate the locations and sizes of the gas’s sources and sinks through tracer-transport inversion or inverse modeling (Box 4.1). Inverse modeling combines atmospheric and oceanic measurements with a model for atmospheric transport and mixing and models of the natural sources and sinks. Inverse modeling also works for estimates of oceanic sources and sinks, using measurements of greenhouse gases dissolved in water. Atmospheric and oceanic methods can be combined in joint atmosphere-ocean inversions (e.g., Jacobson et al., 2007a,b).

Although, in principle, tracer-transport inversion could provide independent estimates of anthropogenic emissions from individual countries for time scales of several days to a year, uncertainties using state-of-the-art methods are too high for this purpose. Four interacting factors are responsible for the large uncertainty:

  1. Small signals. The emissions of the greenhouse gases covered in this report are small compared to the large atmospheric background. This means that anthropogenic emissions will increase the mole fraction of the gas (Box 1.3) by only a small percentage as air moves across a country (e.g., nitrous oxide [N2O] could be increased by 1 part per billion [ppb] over a global background of 320 ppb).

  2. Incomplete understanding of atmospheric transport. To attribute a local increase in the abundance of a greenhouse gas to emissions by an upwind country, it is necessary to have an accurate reconstruction of air flow and mixing. Errors in the atmospheric transport model will cause errors in the inferred location and magnitude of emissions.

  3. Large natural sources and sinks. Carbon dioxide (CO2), methane (CH4), and N2O all have large and uncertain natural sources and sinks that obscure the signals from anthropogenic emissions. For CO2, the instantaneous flux into and out of the terrestrial biosphere varies with time of day and season and can be an order of magnitude larger than fossil-fuel emissions.

  4. Inadequate observing network. Many surface stations are located too far from intense natural and anthropogenic sources to enable robust determination of global trends and seasonal cycles. Although ground stations and aircraft measurements have been deployed in Europe and North America to study CO2 fluxes from urban and industrial regions, forests, cultivated land, and other terrestrial ecosystems, coverage of the globe is uneven and most countries are not adequately sampled.

Fortunately, technological and methodological remedies for these problems could be implemented within a few years, improving the capability for indepen-



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 53
4 Emissions Estimated from Atmospheric and Oceanic Measurements T he direct way to measure emissions of green- of the gas (Box 1.3) by only a small percentage as air house gases from a source is to collect and moves across a country (e.g., nitrous oxide [N2O] could analyze exhaust gases as they are emitted. To be increased by 1 part per billion [ppb] over a global estimate regional or national emissions from a remote background of 320 ppb). location, one can take advantage of the incomplete 2. Incomplete understanding of atmospheric transport. mixing of the atmosphere. Because of incomplete mix- To attribute a local increase in the abundance of a ing, geographical variations in greenhouse gas emis- greenhouse gas to emissions by an upwind country, it is sions and removals cause the abundance of a gas to be necessary to have an accurate reconstruction of air flow elevated downwind of a source and reduced downwind and mixing. Errors in the atmospheric transport model of a sink. Measurements of gas abundances around the will cause errors in the inferred location and magnitude globe can be used to estimate the locations and sizes of emissions. of the gas’s sources and sinks through tracer-transport 3. Large natural sources and sinks. Carbon dioxide inversion or inverse modeling (Box 4.1). Inverse model- (CO2), methane (CH4), and N2O all have large and ing combines atmospheric and oceanic measurements uncertain natural sources and sinks that obscure the with a model for atmospheric transport and mixing signals from anthropogenic emissions. For CO2, the and models of the natural sources and sinks. Inverse instantaneous flux into and out of the terrestrial bio- modeling also works for estimates of oceanic sources sphere varies with time of day and season and can be an and sinks, using measurements of greenhouse gases order of magnitude larger than fossil-fuel emissions. dissolved in water. Atmospheric and oceanic methods 4. Inadequate observing network. Many surface can be combined in joint atmosphere-ocean inversions stations are located too far from intense natural and (e.g., Jacobson et al., 2007a,b). anthropogenic sources to enable robust determination Although, in principle, tracer-transport inversion of global trends and seasonal cycles. Although ground could provide independent estimates of anthropogenic stations and aircraft measurements have been deployed emissions from individual countries for time scales of in Europe and North America to study CO2 fluxes several days to a year, uncertainties using state-of-the- from urban and industrial regions, forests, cultivated art methods are too high for this purpose. Four interact- land, and other terrestrial ecosystems, coverage of the ing factors are responsible for the large uncertainty: globe is uneven and most countries are not adequately sampled. 1. Small signals. The emissions of the greenhouse gases covered in this report are small compared to Fortunately, technological and methodological the large atmospheric background. This means that remedies for these problems could be implemented anthropogenic emissions will increase the mole fraction within a few years, improving the capability for indepen- 

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS checking self-reported emissions (especially chloro- BOX 4.1 Tracer-Transport Inversion fluorocarbons [CFCs] and HFCs) at national to global scales. A forward model of the atmospheric and/or oceanic abun- dance of a trace gas is based on solution of the continuity Inversions of CH4, N2O, and HFCs equation: Top-down, atmospheric inverse model derivations (X(t + ∆) – X(t)) / ∆ = ∂X(t)/ ∂t = EX(t) + PX(t) – LX(t) – ∇FX(t). (1) of greenhouse gas emissions have been applied exten- The tendency in local abundance of species X at time t sively to CO2, CH4, N2O, and the fluorinated gases. (∂X/∂t), written on the left as a finite difference over the time In general, these approaches use the patterns of vari- interval ∆, is equal to the local emission rate E into the volume ability in the trace gases to infer the geographic pattern being sampled plus in situ chemical production P minus loss rates of emissions, but they require some prior knowledge L minus the divergence of the transport flux ∇F. These models are of the spatial and temporal patterns. For N2O, the often designated chemistry-transport models or tracer-transport Bouwman et al. (1995) work on uncertainties in the models. In a tracer-transport inversion, the measurements of X(t), along with a model for the chemistry and transport (P – L, ∇F), global distribution of emissions forms the core a priori are used to derive emissions (E) by subtracting P – L and ∇F from data that are tested with atmospheric observations and both sides of equation (1). models (Hirsch et al., 2006; Huang et al., 2008). Thus With current chemistry-transport models, this continuity far, these studies, as well as similar ones for HFCs (e.g., equation is solved on a three-dimensional grid of more than a Stohl et al., 2009), have been successful at testing the million points, using a meteorology (including winds, convection, assumed emissions only at the scale of broad latitudinal diffusive mixing, clouds, and precipitation) that varies hourly. Most analyses today employ the Bayesian synthesis method (Ent- bands. Some studies suggest that current observations ing, 2002; Gurney et al., 2003) that includes a priori estimates of and modeling are capable of providing information emissions (the best estimate of the emission patterns, including on individual country emissions (e.g., Manning et al., uncertainties, obtained from independent data and modeling). In 2003), but these claims remain untested. Many such regions where the atmospheric measurements clearly constrain the studies assume that the atmospheric transport model emissions, the resulting a posteriori emissions are independent represents tracer transport perfectly and do not consider of the a priori; but in regions where the available measurements are inadequate for constraining the emissions, the method just the large, uncharacterized errors in the models (e.g., returns what was assumed, the a priori. Thus, Bayesian methods Patra et al., 2003; Rayner, 2004; Prather et al., 2008). result in more stable estimates of emissions, but sometimes add A number of studies have derived top-down emis- no new information. sion patterns for CH4 (e.g., Fung et al., 1991; Hein et al., 1997; Houweling et al., 1999), sometimes tak- ing advantage of additional information contained in dently verifying self-reported emissions by countries. the relative isotopic abundances (e.g., 13CH4 versus This chapter reviews studies that used tracer-transport 12CH ; Fletcher et al., 2004). These studies were able, 4 inversion methods to estimate CO2 emissions and for example, to verify the European Commission’s to check self-reported emissions of other greenhouse Emission Database for Global Atmospheric Research gases. The chapter also identifies four ways to reduce (EDGAR) inventory of CH4 emissions on a scale of uncertainties associated with atmospheric monitoring North America to within an uncertainty of 20 percent, of national CO2 emissions. but they rely strongly on assumed patterns of emissions within the continent. Assimilation of satellite observa- tions of CH4 (Bergamaschi et al., 2007; Meirink et al., INVERSE MODELING STUDIES OF 2008) may be able to constrain large emissions at a sub- GREENHOUSE GAS EMISSIONS national level (e.g., rice paddies in India and Southeast Tracer-transport inversion methods have been Asia), but this has not yet been verified using a multi- used to estimate emissions of CO2, CH4, N2O, and model approach with realistic errors on the satellite data hydrofluorocarbons (HFCs). This section summarizes or by comparing predictions with regional inventories the results of inverse modeling studies for estimating and surface observations. On the other hand, intense greenhouse gas emissions (especially CO2) and for scientific campaigns involving regional measurements

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS and modeling (Kort et al., 2008; Zhao et al., 2009) have of our understanding of the carbon cycle, including succeeded in deriving CH4 and N2O emissions at the the link between fossil-fuel combustion and increases subnational scale within the United States, although in CO2 (Revelle and Suess, 1957; Pales and Keeling, these short, intense campaigns cannot be sustained 1965). Measurements of atmospheric CO2 and other year-round over much of the globe. gases are currently made at in situ stations (e.g., sur- One approach to quantifying regional emissions face sites, towers, aircraft profiling) around the world is to use the relative increase in abundances of several (e.g., Figure 4.1, Appendix C) and from satellites. The gases during pollution episodes to derive their relative combined use of CO2 and oxygen (O2) atmospheric emissions. For example, high-resolution time series measurements allows the CO2 removed from the atmo- of gases have been used to identify European pollu- sphere to be partitioned into land and oceanic carbon tion episodes and to quantify the relative emissions of sinks (Keeling et al., 1993; Manning and Keeling, halocarbons and N2O from Western Europe (Prather, 2006; Denman et al., 2007). Observation of isotope analogues of CO2 (including 13C, 14C, 18O), CH4 1985), as well as European emissions of dichlorometh- (13C, 14C, D), and N2O (15N, 18O) as well as O2/N2 ane, trichloroethene, and tetrachloroethene (Simmonds et al., 2006). This method has also been used to try to ratios have increased our understanding of the global separate sources of CO2 by assuming that sulfur hexa- sources and sinks of these greenhouse gases on scales fluoride (SF6) is co-emitted with fossil-fuel combustion of hemispheres or broad latitude bands. All of these (Rivier et al., 2006; Turnbull et al., 2006). The approach measurements would significantly aid in verification obviates the need for accurate tracer-transport model- or falsification of reported national emissions if the ing and can define the relative emissions in a polluted pattern of emissions being tested is provided and the air sample to 10 percent or better. However, it is limited measurements are made at sufficiently high spatial to areas where distinct, single-source pollution plumes resolution and at suitable locations (e.g., within or occur on top of a nearly uniform, clean air background. close to the borders of the country whose emissions S urface-based, diurnal column measurements have are being tested). recently been used to estimate CH4 emissions from The mixing time of tracers in the atmosphere the Los Angeles basin (Wunch et al., 2009). Two key ranges from minutes to days between the ground and criteria need to be met to derive absolute emissions, for the tropopause, about two weeks around a latitude example, of fossil-fuel CO2: (1) the absolute emissions circle at midlatitudes, several months to reach through of the reference gas must be known to equivalent or bet- a hemisphere, about one year between the northern ter accuracy; and (2) the two gases must be co-emitted and southern hemispheres, and several years to mix in the same ratio over the entire region being sampled, through the stratosphere. The density of the observing otherwise the much larger errors in tracer-transport network, together with the atmospheric mixing times, inversion dominate. Both criteria fail for SF6, but for determines the spatial resolution with which sources 14CO , the second criterion allows for the separation and sinks can be inferred using an atmospheric model. 2 of fossil-fuel CO2 from biogenic sources (see “14C In particular, the slow mixing between hemispheres Measurements” below). and the preponderance of emissions in the northern Overall, there are published cases for CH4, N2O, hemisphere mean that the signal from anthropogenic and the synthetic fluorinated gases in which one can emissions is relatively large when hemispheres are distinguish source from sink and other cases in which compared. one cannot. Thus, the committee estimates uncertain- Tans et al. (1990) showed that the interhemispheric ties in national emissions of these gases as 50 percent gradient in CO2 was smaller than it should have been, to greater than 100 percent. given the north-south disparity in fossil-fuel emissions. This implied a “missing sink” of more than one billion tons of carbon per year in the northern hemisphere. CO2 Emission Estimates Subsequent analyses confirmed this qualitative conclu- The long-term, accurate measurements of atmo- sion but have struggled to improve its spatial resolution. spheric CO2 at Mauna Loa by Keeling (1961) and Continued effort has focused on the range of different many others (Figure 1.4) laid the foundation for much atmospheric model results and what model error could

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS FIGURE 4.1  Map of the National Oceanic and Atmospheric Administration Earth System Research Laboratory (NOAA ESRL) global  cooperative air sampling network in 2008. Blue squares are continuous measurements; red circles are weekly or daily flask samples;  green triangles are continuous measurements on tall communications towers; and dark blue stars are weekly to biweekly vertical profiles  by small aircraft. These sites provide a large fraction of the current set of atmospheric measurements of CO2 and related trace gases  that are used in inverse modeling. SOURCE: NOAA ESRL, . do to CO2 inverse modeling (Engelen et al., 2002; errors in changes in emissions than in their abso- Peylin et al., 2002; Gurney et al., 2003; Rayner, 2004; lute magnitude. For example, Denman et al. (2007) Baker et al., 2006a; Prather et al., 2008). The spread compared emissions estimates from different tracer- in north-south transport among the different models, transport models and found that year-to-year differ- while nonnegligible, is not large enough to invalidate ences in estimates for different regions within the same interpretations of annual mean CO2 emissions from latitude zone were surprisingly consistent between broad latitude bands. However, the derived emissions models, despite the large between-model differences in for each continent and ocean basin within each band are absolute magnitude. The implication is that it is easier highly uncertain, even in sign. The addition or loss of to quantify changes in emissions than it is to assign a single site from the observing network in Figure 4.1, absolute magnitudes to those emissions. for example, can in some cases shift derived emissions The first row of Table 4.1 summarizes the current for large regions, such as a half-continent, from positive state of the art in CO2 inversion estimates. It shows to negative (e.g., Rödenbeck et al., 2003; Le Quéré et that although the annual atmospheric increase in CO2 al., 2007; Law et al., 2003, 2008). is known within 7 percent, global annual fossil-fuel Transport uncertainty causes smaller systematic emissions can be estimated from atmospheric and

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS TABLE 4.1 Reducing Uncertainties in CO2 Emissions Through New Atmospheric and Ocean Observations Decadal Annual Annual Global Annual Global Annual Regional Continental National Net Flux Net Flux Land- Tropical Mid-latitude to the to the Fossil IPCC ocean Sink Biosphere Biosphere Fossil IPCC Fossil IPCC Method Atmosphere Atmosphere Fuel LUCF Partition Changes Changes Fuel LUCF Fuel LUCF Current uncertainties State of art, including 1a 1a 2b 5b 3b,c,e 3c,f 3c,d,f 5 5 5 5 13CO , O , oceanic, and (1%) (7%) (25%) 2 2 atmospheric data Potential improvements State of art plus 14CO2 1-2 5b 2g 2g 2g 2g 5 2-4 5 All of the above plus ++g ++g intensive aircraft observations All of the above ++g (+)g (+)g (+)g plus intensive ocean observations All of the above plus other ++ ++ tracers of combustion (e.g., CO, NOx, HCN) All of the above plus ++ ++ ++ ++ ++ satellite CO2 observationsh NOTES: IPCC = Intergovernmental Panel on Climate Change; LUCF = land-use change and forestry. Other than %, units are Pg of CO2 per year for the 2000 decade. 1 = 100% (i.e., cannot be certain if it is a source or sink). Potential improvements: ++ = likely and direct; (+) = indirect in the sense that it would directly improve estimates of ocean fluxes which, in a tracer-transport inversion, would reduce errors for land masses, especially for those at similar latitude. a2 σ uncertainty of annual increase, P. Tans, . bDenman et al. (2007). cConway et al. (1994); Ciais et al. (1995). dPeters et al. (2007). eGruber et al. (2009). fTans et al. (1990); Gurney et al. (2002). gImprove transport in models—expert judgment of committee. hIf systematic errors of satellite CO retrievals can be mastered through ongoing comparisons with in situ chemical measurements, frequent resampling can 2 lead to the small errors of annual averages that are required for flux estimates. oceanic data only to within 25 percent. The reason for Because of transport uncertainty and the lack of the difference is the large interannual variation in the measurements over much of the globe, estimates of size of sources and sinks in the terrestrial biosphere and the total net flux of CO2 from broad bands of latitude oceans, which must be separated from the total atmo- on a seasonal time scale are uncertain by at best 25- spheric increase to estimate the contribution from fossil 49 percent, compared to only 7 percent for the globe fuel. Uncertainty in the anthropogenic emissions from (Table 4.1). The situation is far worse for estimates land-use change and forestry is greater than 100 per- of continental or national emissions (>100 percent) cent because both anthropogenic and natural changes because east-west air flow rapidly mixes emissions in the terrestrial biosphere have almost identical effects signals and makes inversions sensitive to model error. on atmospheric CO2, 13C, and O2. For this reason, the In addition, there remains the question of whether the inventory and remote sensing methods described in the amplitude of the greenhouse gas perturbations caused previous chapters are needed for the a priori data on by national emissions is large enough to detect with in emissions from land use and forestry. situ networks or satellites.

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS Appendix B presents a mass-balance analysis for able to separate these terrestrial and ocean fluxes from the 20-largest CO2 emitting nations. Table B.1 shows the anthropogenic emissions. that anthropogenic emissions increase the abundance of CO2 by, on average, a fraction of a part per million Verification of CFCs, HFCs, and Other Synthetic (ppm) in the whole column (ranging from 0.06 ppm Greenhouse Gas Emissions for Australia to 0.76 ppm for the United States). These small signals are measurable, but their detection is con- The best tests of self-reported emissions using founded by the much larger and incompletely under- atmospheric chemistry models and observations have stood signals from terrestrial ecosystems. Figure 4.2 involved the global or hemispheric budgets of long- shows that, even at global scales, and when fluxes are lived, synthetic, fluorinated gases produced solely by averaged over an entire year, the apparent fraction of human activities. Rowland et al. (1982) measured fossil-fuel CO2 that remains in the atmosphere varies CF2Cl2 at several remote sites to determine a global from year to year by as much as a factor of 2. Most of mean abundance. With knowledge of the long atmo- this variation is caused by the response of the terrestrial spheric lifetime (i.e., slow chemical loss), they were able biosphere and oceans to climate anomalies (Francey et to infer annual global emissions to high accuracy (e.g., al., 1995; Keeling et al., 1995; Bacastow, 1976) and to 10 percent) from the annual increase in the atmosphere. increased fire activity during El Niño events (van der The magnitude of CF2Cl2 emissions estimated from Werf et al., 2004, 2008). The magnitude of annual inverse modeling contradicted that claimed by the perturbations (sources or sinks) can be as large as one- chemical industry, which subsequently retracted its quarter of the magnitude of global fossil-fuel emissions reported emissions in favor of Rowland et al.’s derived (e.g., Battle et al., 2000). To monitor anthropogenic emissions. This scenario was replayed in 2008 for the emissions with tracer-transport inversion, one must be long-lived greenhouse gas NF3, which is used in rapidly increasing quantities in the manufacture of large flat- panel displays and photovoltaic cells. Prather and Hsu (2008) reviewed the production and lifetime of NF3, disputing an industry estimate of emissions (Robson et al., 2006) as unrealistically low and argued that this gas should be detectable and increasing in the atmosphere. Within months Weiss et al. (2008) made the measure- ments and confirmed that the reported NF3 emissions were indeed too low. Inverse modeling based on atmospheric measure- ments also indicates much larger emissions of HFC- 134a and SF6 than the sum of emissions in national inventories reported to the United Nations Framework Convention on Climate Change (UNFCCC). Höhne and Harnisch (2002) showed that post-1998 emissions of HFC-134a are 50 percent higher than reported by Annex I countries (see Figure 4.3), which are expected to be the source of nearly all HFC-134a emissions. Similar results have been shown for SF6 (Geller et al., 1997; Höhne and Harnisch, 2002). Although inverse modeling shows that emissions of many HFCs and FIGURE 4.2  Interannual variation in the airborne fraction of  CFCs are underestimated in UNFCCC inventories fossil-fuel  accumulation  in  the  FINAL.eps airborne  frac- Fig 4. 2 atmosphere.  The  (and occasionally overestimated; see discussion of tion is defined as the change in the mass of atmospheric CO2  type has been converted to paths halon-1301 in Clerbaux and Cunnold, 2006), it offers divided  by  the  mass  of  fossil-fuel  CO2  emitted.  The  dark  line  shows successive 5-year averages. SOURCE: Figure 7.4b from  no insight as to the source of error. Global or hemi- IPCC (2007a), Cambridge University Press.

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS 100 Global Emissions EDGAR global reported HFC-134a Emissions, Reported and Inverse 96 Inverse 98 80 Inverse CMDL Inverse Model (k t/yr) Inverse AGAGE Inverse AFEAS 60 Annex-IR UNFCCC reporting 40 20 0 1990 1992 1994 1996 1998 2000 2002 Year of Emissions FIGURE 4.3  Global emissions of HFC-134a from several inverse atmospheric models, compared with UNFCCC reported Annex  Fig 4.3 FINAL.eps I emissions and the European Commission’s Emission Database for Global Atmospheric Research (EDGAR), for 1990-2002. Most  production facilities and emissions for HFC-134a are in Annex I countries. SOURCE: Courtesy of Michael Prather, University of Cali- fornia, Irvine. Modified from Prather et al. (2009). Copyright 2009 American Geophysical Union. Reproduced by permission of the  American Geophysical Union. Measurements of Large Emission Sources spheric inverse modeling can provide a strict test of the global sum of national greenhouse gas emissions A large fraction of fossil-fuel emissions emanates inventories on an annual basis, but it has not been able from large local sources, such as cities or power plants, to identify the countries whose emissions are in error. and thus the effect of national mitigation measures should be evident in the “domes” of CO2 that they NEW APPROACHES FOR INCREASING THE produce (Idso et al., 2001; Pataki et al., 2003; Rigby et ACCURACY OF NATIONAL EMISSIONS al., 2008a). For example, more than 57 percent of U.S. ESTIMATES fossil-fuel emissions occur in areas that have a flux rate that exceeds 2 kg C m–2 yr–1, which corresponds to ~1.7 Because of the twin problems of transport error and percent of the total surface area (including power plants, the separation of natural from anthropogenic fluxes, the cities, and other point sources; Table 4.2). Cities also uncertainty in tracer-transport inversion estimates of anthropogenic emissions for continents and nations can be as large as 100 percent (Table 4.1). These errors cur- rently make tracer-transport inversion impractical for TABLE 4.2 U.S. Fossil-Fuel Emissions as a Function of monitoring national emissions. Research is improving Carbon Density at 0.1 Degree Grid Spacing the representation of transport and biogeochemistry in Carbon Density of Emissions Area Percentage of Total models, but at the slow pace at which new observations (kg C m–2 yr–1) (%) U.S. CO2 Emissions are becoming available, advances cannot be expected to ≤2 98.3 42.7 deliver the required accuracy during the coming decade. 2-4 0.9 13.6 4-10 0.6 18.0 The following approaches, which would augment cur- 10-20 0.2 15.1 rent research, would shift the observing paradigm for 20-max 0.1 10.7 the carbon cycle and substantially improve our capabil- SOURCE: VULCAN emissions inventory; .

OCR for page 53
0 VERIFYING GREENHOUSE GAS EMISSIONS provide a broad sample of different emission sectors power plants, and leaks from a geologic sequestra- (Table 4.3). Statistical or systematic sampling of CO2 tion site in Appendix B). Because the increased CO2 emissions from large local sources would provide inde- abundances are largest over the source of emissions pendent data against which to compare trends in emis- and disperse within a few tens of kilometers, they can sions reported by the countries in which those sources usually be attributed unambiguously to their country are located, at least for fossil-fuel emissions. Sampling of origin. This largely eliminates the attribution prob- in cities, however, requires overcoming technical chal- lem created by transport uncertainty in global tracer- lenges, including finding ways to effectively construct transport inversions. For example, Figure 4.4 shows the seasonal averages in the presence of considerable spatial size of the CO2 signature of a hypothetical large power and daily variability and to separate biogenic from fos- plant in the Central Valley of California, as predicted sil-fuel sources (e.g., Pataki et al., 2003). by an atmospheric model. Although the CO2 plume at Working with large localized sources has two any given time varies with wind speed and direction, important advantages. First, their concentrated fossil- CO2 mole fractions would, on average, remain elevated fuel emissions may be large enough to exceed the signal above a continuously emitting power plant compared from local natural sources and sinks. For example, the with the surroundings. In this way, the CO2 within a emissions intensity of the greater Los Angeles metro- radius of the source could be used to infer emissions politan area (20 kg CO2 m–2 yr–1; see Table B.3) is ~20 from power plants. times the annual net sink observed at Harvard Forest The CO2 domes created by large local sources can (0.9 kg CO2 m–2 yr–1; Barford et al., 2001), which is the be mapped using measurements from surface stations difference between two much larger terms of opposite and aircraft in and around cities, measurements of the sign, photosynthesis and respiration. Of course, the radiocarbon content of annual plants found in urban sources and sinks in urban ecosystems are not as large environments, or measurements from high-resolution as in Harvard Forest. satellites. Second, large local sources increase the local CO2 abundance in the atmosphere by a few to more than Surface Network. The variability of CO2 and other 30 ppm, depending on proximity to local sources and greenhouse gases in cities is substantially greater than atmospheric mixing (see Riley et al., 2008; Mays et that measured at clean air marine boundary layer al., 2009; and the analysis of signals from urban areas, stations. Effective sampling of this variability in and TABLE 4.3 CO2 Emissions for Selected Cities Emissions Area Population Total Commerce + industry Residential Utilities Transportation City (km2)a (millions) (Mton CO2 yr–1) (Mton CO2 yr–1) (Mton CO2 yr–1) (Mton CO2 yr–1) (Mton CO2 yr–1) Los Angeles 3,700 17.5 73.2 16.8 8.1 8.8 39.9 Chicago 2,800 9.5 79.1 26.7 14.3 13.9 23.8 Houston 3,300 5.5 101.8 48.8 2.2 32.6 20.1 Indianapolis 900 2 20.1 3.7 2.2 5.5 8.8 Tokyo 1,700 29b 64 29.5 12.5 22 c Seoul 600 13 43 18 13 12 c Beijing 800 15.6 74 57 12 5 c Shanghai 700 18b 112 92 10 10 c NOTE: Mton CO2 is million metric tons of CO2. aArea represents the contiguous area of intense activity, not administrative boundaries. Estimates of such functional boundaries were made from Google maps in “satellite” mode, which shows built up areas by color and road density. bTokyo does not include all of the agglomeration, which has 35 million inhabitants. The current population of Shanghai is 18.9 million, so emissions numbers are underestimates. cCO emissions are included in the commerce + industry and residential sectors. 2 SOURCES: Dhakal et al. (2003) for population (including migrants) and emissions in 1998 for four east Asian cities. U.S. estimates are from the VULCAN emissions inventory for 2002 () and the U.S. Census.

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS around cities is needed to estimate long-term mean values and to detect multiyear changes in emissions. Care must be taken in the sampling to integrate over the daily variations in emissions and meteorological conditions. A sampling network to measure the spatial gradients in greenhouse gases created by large local sources could be established for a sample of sites. The network would require ground-based stations and aircraft sampling. Airborne technologies to support such an effort are described in Appendix D. Enhanced automated meteorological observations in urban areas would help document the dispersion of CO2 and improve the estimation of local source strengths. Urban Plants. Uptake of CO2 by plants integrates the 14C/12C ratio of CO in surface air over a period of 2 several months during the growing season. The plants (by means of photosynthesis) provide long-term inte- grated samples during daytime periods when the plan- etary boundary layer is well developed and variability in CO2 is at a minimum. Decreases in the 14C/12C ratio of plant biomass are directly proportional to the level of excess fossil-fuel CO2 present, thus providing a seasonal measure of the dome of fossil-fuel CO2 over major metropolitan regions (Hsueh et al., 2007; Riley et al., 2008; Pataki et al., 2010). Figure 4.5 shows the distribution of atmospheric radiocarbon anomalies in the Los Angeles basin obtained from annual grasses. By characterizing urban to suburban gradients, it may be possible to detect relative changes in fossil- fuel emissions over time. The 14CO2 measurements of annual plants using accelerator mass spectrometry are FIGURE 4.4  Instantaneous mole fraction of atmospheric CO2  at the surface (top) and averaged for the column using the Or- sufficiently accurate (2.7 permil; Riley et al., 2008) to biting  Carbon  Observatory  (OCO)  averaging  kernal  (bottom)  detect a 6 percent change in fossil-fuel emissions. The 24 hours after emissions from a hypothetical large power plant  actual detection limit will likely be higher because of (emitting 4.16 million tons of carbon per year) at noon in the  year-to-year differences in the growing season of the Central  Valley  of  California.  The  simulation  used  the  Weather  Research  and  Forecast  model  (WRF),  with  meteorology  for  sampled vegetation and variability in the synoptic-scale March 4, 2008, and the model resolution was 2 km. Note the  weather patterns. The preliminary observations shown difference in scales. Wind speed and direction for the surface  in Figure 4.5 suggest that by sampling multiple loca- and for the column are shown as white arrows. Since the hy- tions repeatedly over several years, it would be feasible pothetical  point  source  is  located  in  the  Central  Valley,  high  CO2 at the surface is confined to the valley. Although there is  to detect an emissions change of ±15 percent within vertical mixing, the emitted CO2 is confined to the lower 5 km  a large city. As with other approaches for urban air of the atmosphere and thus would not be seen by AIRS (Atmo- sampling, concurrent meteorological data and high- spheric Infrared Sounder), which senses the upper troposphere.  SOURCE: Courtesy of Zhonghua Yang and Inez Fung, University  resolution tracer-transport modeling would be needed of California, Berkeley. to reduce uncertainties associated with climatic vari- ability of the atmospheric transport.

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS FIGURE 4.5  Radiocarbon anomalies (in permil) relative to background levels for Southern California. These anomalies were esti- mated relative to radiocarbon levels measured at Point Barrow, Alaska. One part per million of locally added fossil-fuel CO2 causes  a radiocarbon decrease of –2.7 permil or –0.27 percent. The largest negative anomalies were observed near downtown Los Angeles  and exceeded –100 permil, corresponding to over 35 ppm of locally added fossil-fuel CO2. Twenty sampling sites near downtown Los  Angeles had a mean radiocarbon difference relative to Point Barrow, Alaska, of –58 permil, corresponding to a mean of 21 ppm of  locally added fossil-fuel CO2. SOURCE: Courtesy of Wenwen Wang and James Randerson, University of California, Irvine. Modified  from Figure 19.4 from Pataki et al. (2010). Reproduced with kind permission from Springer Science and Business Media. The approaches described above for measuring uncertain for concentration anomalies over a natural or large emission sources avoid some of the problems agricultural region. encountered in studies that have attempted to estimate A research program to measure the atmospheric natural or agricultural fluxes using atmospheric inver- dome of greenhouse gases over a representative sample sion techniques. First, it is possible to directly measure of large local emitters would include three elements: the fossil-fuel emissions that create a CO2 dome over (1) the development of new technologies and analysis a city or power plant. In contrast, accurate independent approaches for detecting long-term trends in urban and estimates of the actual integrated biological fluxes in industrial areas, (2) regional atmospheric modeling to a large region are not available. Second, whereas the help design the local network and to quantify variability concentration anomaly over a city or power plant has an from seasonally and interannually varying winds, and unambiguous origin, transport errors make attribution (3) comparison of different measurement techniques,

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS CO2 yr–1) and that the wind speed is 3 m s–1. These such as flask, isotope, aircraft, sun-viewing Fourier transform spectrometer, mobile platform, and satellite. conditions would produce a perturbation of approxi- Long-term monitoring could begin in approximately mately 0.5 percent (~1.7 ppm) in the average column 10 different U.S. cities with different sizes, population abundance of CO2 within an OCO sample, which is densities, and different mixtures of emissions sources consistent with the instrument’s design uncertainty using currently available sampling technologies. Simul- of 1-2 ppm and significantly larger than the ground- taneous creation of detailed bottom-up inventories of tested value of 1 ppm. In contrast, because a GOSAT emissions for these same representative areas would sample covers a larger area than an OCO sample, the facilitate transport modeling and enable the efficacy of CO2 perturbation within a GOSAT sample would be different atmospheric approaches for trend detection approximately 0.1 percent (~0.4 ppm). This is an order to be assessed. By analogy with research initiatives of of magnitude smaller than GOSAT’s estimation error similar scope, this initiative would require a research of 4 ppm. Note that in target mode, OCO could take budget of about $15 million to $20 million per year up to 7,000 shots at an individual site, under different (e.g., $6 million to $10 million for extending National viewing angles, and could potentially have an uncer- Oceanic and Atmospheric Administration [NOAA] tainty of 0.1 ppm if systematic biases were character- flask and continuous monitoring to 10 cities and the ized and removed. remainder for technology development, modeling, and No other satellite has OCO’s critical combination analysis) in the United States. Ideally, similar programs of high precision, small footprint, readiness, density would be developed in other countries. of cloud-free measurements, and ability to sense CO2 near the Earth’s surface (Table C.1, Appendix C). Its Satellites. By providing repeated global coverage from 1-2 ppm accuracy and 1.29 × 2.25 km sampling area a single instrument, satellites would complement a would have been well matched to the size of a power global network of ground stations and aircraft sam- plant. Yet, the OCO mission as planned would have pling to monitor CO2 emitted from selected cities and had limitations for monitoring CO2 emissions from power plants. They would also largely overcome the large sources because it would have sampled only 7-12 difficulties of national sovereignty and international percent of the land surface (Miller et al., 2007) with cooperation in CO2 monitoring and the verification a revisit period of 16 days and a nominal lifetime of challenges associated with self-reporting. As shown only 2 years (Table C.1, Appendix C). However, many in Table C.1 (Appendix C), Japan’s Greenhouse gases metropolitan areas are large enough to be sampled by O bserving Satellite (GOSAT) is the best available the planned orbit, and OCO would still have provided satellite for spaceborne measurement of CO2 anomalies a sample of a few percentage of the power plants. from emissions. It has lower uncertainty and higher Monitoring urban and power plant emissions from spatial resolution than the Scanning Imaging Absorp- space is challenging and has not been demonstrated. tion Spectrometer for Atmospheric Chartography A replacement OCO could demonstrate these capa- (SCIAMACHY ), the Atmospheric Infrared Sounder bilities. Nevertheless, it would be valuable to explore (AIRS), or the Infrared Atmospheric Sounding Inter- changes in the orbit and other parameters so that a ferometer (IASI), and it senses near the surface where greater fraction of large sources is sampled. For example, emission signals are largest, unlike AIRS and IASI. consider a precessing orbit covering ~100 percent of the However, the CO2 signal produced by the emissions of surface but with only two measurements per year of a large power plant is typically smaller than what can each location. With 100-500 large local sources in high- be measured with GOSAT. In contrast, the Orbiting emitting countries, it might be possible to obtain a sta- Carbon Observatory (OCO), which failed on launch in tistical sample of hundreds of measurements of plumes February 2009, would have had the high spatial resolu- of CO2 being emitted by the large sources in each of tion required to monitor instantaneous CO2 emissions these countries. The trade-offs in optimizing monitor- from such local sources. ing capabilities while meeting scientific objectives (e.g., For example, assume that a 500 MW pulverized observing with high solar zenith angle) would have to coal power plant emits ~0.13 t s–1 of CO2 (e.g., 4 Mt be examined by a technical advisory group. The group

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS 14C Measurements would also have to develop a strategy for integrating the snapshot view of CO2 anomalies from the satellite Systematic radiocarbon (14C) measurements would measurements into annual emissions. largely eliminate the confounding effects of natural Because of its 2-year mission life, OCO would fluxes and thus greatly improve the accuracy of fossil- not have been able to track emission trends. However, fuel emissions estimates. Carbon-14 is produced by it would have provided the first few years of measure- cosmic rays in the higher layers of the atmosphere as ments (a baseline) necessary to verify a decadal trend well as from weapons tests. It is well mixed as 14CO2 in for the large local sources within its footprint and the lower atmosphere and oceans, where it is incorpo- served as a pathfinder for successor satellites designed rated into all living organisms. Its radioactivity decays specifically to support a climate treaty. Moreover, the with a half-life of 5,730 years. Fossil fuels are formed technology used in the OCO instrument has no con- from organic material, but their 14C component has sumables or inherently short-lived components and fully decayed over millions of years. Thus, the CO2 hence could be flown on an extended, decadal mission. emitted from the burning of fossil fuels is characterized A replacement mission is expected to cost about the by its absence of 14C, creating plumes of 14C-depleted same as the original, $278 million.1 Alternate propos- air near source regions (Levin et al., 2003; Levin and als to measure the near-surface CO2 abundance from Rödenbeck, 2008). space with multifrequency laser (e.g., the Active Sens- The addition of 1 ppm of fossil-fuel CO2 to a ing of CO2 Emissions over Nights, Days, and Seasons contemporary air mass with a background CO2 mole [ASCENDS] mission)2 are not yet technologically fraction of 390 ppm and a 14C/12C ratio typical of the ready for long-term operations. free troposphere causes the 14C/12C ratio of CO2 to Deployment of a CO2-sensing satellite with the decrease by 0.27 percent (e.g., Turnbull et al., 2007). capability of OCO, together with a program of surface Accelerator mass spectrometry can measure 14C in and aircraft sampling in and around large local sources CO2 in only 2 liters of ambient air to a precision of would address all four main problems associated with 0.2 percent. Comparison of the 14C depletion in an air tracer-transport inversion. It would enable the network sample to 14C measured in background air determines to support emissions verification by observing CO2 the CO2 component from recent fossil-fuel burning to directly over high-emitting sites, where the signal a precision of 0.7 ppm, which is within the range of from anthropogenic emissions is locally large enough signals produced by fossil-fuel burning in individual to separate from confounding natural fluxes and where nations (Table B.1, Appendix B). the transport problem is far simpler than at regional Global 14CO2 observations would greatly reduce and continental scales because emissions are in most transport errors. The fossil-fuel emissions inventories cases still confined to the boundary layer. Ongoing of most Annex I countries are believed to be fairly sampling by aircraft or from balloons will be essential accurate (Chapter 2). If a baseline of atmospheric to investigate potential systematic biases of column 14CO observations can be established before emission abundances inferred from any satellite and to describe 2 reductions start taking hold, the observations would the vertical distribution necessary to infer the location help constrain atmospheric mixing processes within of emissions. transport models over the continents. A corresponding capability for other greenhouse Box 4.2 presents an analysis of the kind of gains gases does not exist, except for the ozone precursor in estimation accuracy that could be realized from NOx (nitrogen oxide). Satellite remote sensing of NOx systematic 14C measurements if transport errors could emissions is readily able to detect decadal emission be eliminated. The analysis shows that 10,000 14CO2 trends over urban and industrial regions (Martin et al., samples per year could yield fossil-fuel CO2 emission 2006; Stavrakou et al., 2008; Kaynak et al., 2009). uncertainties of 10-25 percent in the United States. W ith a better designed sampling grid, only 5,000 1 S ee . practice, 14CO2 measurements would likely provide 2 See .

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS BOX 4.2 Gains from 14C Sampling A recent Observation System Simulation Experiment (OSSE) tested Note the change in color scale between the panels. Starting with an as- the capacity of 800 atmospheric 14CO2 measurements per month (e.g., sumption of 100 percent uncertainty in the flux from each region, this Turnbull et al., 2007) to constrain U.S. fossil-fuel CO2 emissions. The simulation suggests that 14CO2 measurements can potentially determine results of the experiment are shown in the figure below. The left panel the fossil-fuel flux in the coterminous United States to better than 10 shows the total flux in 60 regions, as represented by the VULCAN emis- percent. This OSSE was a test of the sampling density and assumed sions inventory. The right panel shows the uncertainty in the retrieval that the tracer-transport model was perfect. Errors in the tracer-transport of the VULCAN emissions, which is treated as truth in this simulation. model (as discussed here) would increase these uncertainties. Estimation of CO2 fluxes from fossil-fuel burning using 10,000 14CO2 measurements per year. Left panel: Sources (VULCAN  inventory;  )  aggregated  into  5  ×  5  degree  areas  (colored  rectangles).  The  84  virtual sampling locations, where virtual 14CO2 measurements were made every 3 days at 14:00 local time, are plotted as  “pluses.” Right panel: Uncertainty of the total estimated flux in each 5 × 5 degree area. For both panels, January emissions  and uncertainties are shown, but those for other months are not substantially different. SOURCE: Courtesy of John Miller,  NOAA/ESRL. Measurement Programs to Reduce Model Errors estimates of the global fossil-fuel flux that would be accurate enough to check trends from self-reported To improve estimates of national anthropogenic inventories collectively (Table 4.1). The accuracy of CO2 emissions from tracer-transport inversion, we estimates of fossil-fuel CO2 emissions by continents will ultimately have to reduce transport errors in would be expected to improve from >100 percent atmospheric models (see the plus signs in Table 4.1). uncertainty to 10-25 percent uncertainty for intensely Improvement of atmospheric tracer-transport models sampled regions, but national fossil-fuel CO2 emissions will require not only improved meteorological obser- would still be uncertain at the 50-99 percent level if a vations (especially of vertical mixing) but also new country is not sampled intensely. high-resolution measurements of multiple independent Making 14CO2 measurements in the air samples tracers that will enable the transport characteristics of currently being collected in the global cooperative air the models to be validated. Such data are collected from sampling network (Figure 4.2) would be relatively inex- international networks, supplemented by regional or pensive. It would cost only about $5 million to $10 mil- global campaigns that measure multiple tracers with lion per year to collect and process the samples and to three-dimensional resolution, including vertical pro- make the 14CO2 measurements on 10,000 samples per files. Previous and ongoing studies of this sort include year. A dense deployment in the United States (5,000 the TRACE-P (TRAnsport and Chemical Evolution samples per year) would provide proof of concept and over the Pacific) campaign that examined the emis- identify its limits.

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS sions of CO from combustion in Asia (Kiley et al., calibrated high-precision measurements can be made 2003), the INTEX/NA (Intercontinental Chemical routinely from regular flights at low cost. Although the Transport Experiment—North America; Choi et al., transects are in the upper troposphere and lower strato- 2008) that suggested stronger than expected vertical sphere, a vertical profile is obtained during each take- mixing, the regional campaign in the north central off and landing. Routine measurements of air pollutants United States anchored by the WLEF tall tower in have been made from European commercial airliners Wisconsin (Denning et al., 2003; Wang et al., 2007), as part of the MOZAIC (since 1994) and CARIBIC and European efforts (e.g., see Ahmadov et al., 2007; (since 1997) research programs, and a successor pro- Levin and Karstens, 2007; Vetter et al., 2008). To test gram that will include CO2 and CH4 measurements (IAGOS) is being planned.4 Japanese scientists have and improve tracer-transport models, data on the time evolution of spatially resolved tracer abundances need obtained flask samples on JAL flights between Tokyo to be accompanied by high-resolution gridded data on and Melbourne since 1993 and continuous measure- emissions (e.g., Gurney et al., 2009). ments and flask samples onboard five aircraft traveling Although OCO would be capable of measuring between Asia, Australia, Europe, and North America large urban regions or power plants with precisions as since 2005 (Machida et al., 2008). great as 1-2 ppm (see Appendix C, Table C.1), much of the atmospheric signal from smaller or dispersed Improvements to Tracer-Transport Inversion with sources is much less than 1 ppm. Such signals can be Carbon Data Assimilation measured accurately with in situ networks (e.g., surface sites, towers, aircraft profiling), such as the NOAA The accuracy of sources and sinks inferred from ESRL network (Figure 4.2) and the World Meteo- variations of the atmospheric mole fraction is intimately rological Organization (WMO) Global Atmosphere tied to the accuracy of winds and mixing characteristics Watch (GAW) stations.3 However, many areas, includ- in tracer-transport models. Typically, wind and mixing ing those with high greenhouse gas emissions, are not data are obtained from a meteorological data assimila- adequately sampled. For example, CO2 observations tion system, wherein diverse streams of meteorological are sparse in the tropical latitudes, Africa, and South observations are merged with a numerical weather America. A number of recent studies have examined prediction (NWP) model to yield, every 3 or 6 hours, how to expand the current CO2 measurement sites to the best estimate of the atmospheric state and the optimize the ability to retrieve CO2 emissions from uncertainty in the estimate. A reanalysis is undertaken transport modeling (Gloor et al., 2000; Patra et al., occasionally to assimilate decades of meteorological 2003; Rayner, 2004; Gurney et al., 2008). Expansion observations into a single NWP model (e.g., Kalnay of the GAW network to observe the variations in et al., 1996; Uppala et al., 2005) and, thus, represents greenhouse gas abundances in countries with the larg- the best source of information about climate variability est emissions would greatly improve the independent on diurnal, seasonal, and interannual to decadal time v erification of emissions through tracer-transport scales. Most tracer-transport models or carbon data modeling. assimilation systems employ 3-hourly or 6-hourly Expansion of the GAW network in the vertical atmospheric circulation statistics from a reanalysis. dimension would allow more meaningful comparisons Carbon data assimilation combines, in principle, with satellite retrievals of column-averaged CO2 than emissions inventories and data on land use, greenhouse ground-based measurements alone and would also pro- gas abundances, and meteorology with models of the vide a routine experimental check on how models trans- atmosphere, ecosystems, and oceans into a coherent port emissions from the boundary layer to the free tro- posphere. With the cooperation of commercial airlines, 4 S ee for a description of MOZAIC (Measurement of Ozone and Water Vapour on Airbus in-service Aircraft), CARIBIC (Civil Aircraft for the Regular Investigation of 3 S ee . (In-service Aircraft for a Global Observing System).

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS framework so that emission estimates can be con - model that links the estimated fluxes to atmospheric strained by their consistency with the other observa- CO2 observations. This modeling component ingests tions. Sources and sinks of CO2 and other greenhouse three-dimensional meteorology, various remote sensing gases at the surface and in the atmosphere column can products (e.g., fraction of absorbed photosynthetically be derived, as has been demonstrated by the Observa- active radiation), fossil-fuel emission inventories, and tion System Simulation Experiment using CO2 from time series of atmospheric CO2 mixing ratio data to OCO (Baker et al., 2008), the two-step assimilation produce regional estimates of surface fluxes that are and inversion of AIRS CO2 (Chevallier et al., 2009; used to evaluate regional representativeness of the prior Engelen et al., 2009), and assimilation studies of satel- estimated fluxes. Bayesian inversion is then applied to lite CO, CH4, and O3 using three-dimensional chemi- assign scaling factors that align the surface fluxes with cal transport models (Ménard et al., 2000; Clerbaux the CO2 observation time series (Gökede et al., 2010). et al., 2001; Stajner et al., 2001; Meirink et al., 2008; In this way, the model generates a record of its skill at Tangborn et al., 2009). predicting fluxes from natural ecosystems, agriculture, Most existing carbon data assimilation efforts are forestry, and land use, which should improve through “univariate” in that they are carried out separately from time. The intent is to reduce uncertainties in fluxes over the meteorological assimilation and neglect the spread space and time, which should improve with increased in the reanalysis. As a result, consistency between the density of atmospheric CO2 observations. meteorology and trace gas mole fraction and fluxes A quasi-operational global system is Carbon - is not guaranteed, and the quantifiable uncertainty Tracker, which inverts for net CO2 fluxes for land and in the circulation is not propagated into a spread in ocean regions, with a high-resolution (1 × 1 degree lati- the trace gas mole fractions and, ultimately, to the tude and longitude) zoom region over North America, sources and sinks. Examples of univariate assimilation using the boundary layer CO2 observing network projects include efforts to assimilate anticipated CO2 (Peters et al., 2007). Net fluxes are adjusted each week data from the OCO satellite into atmospheric models by linear scaling factors for 209 land regions and 30 (Stajner et al., 2001; Baker et al., 2006b; Miller et al., ocean regions. Emissions from fossil-fuel burning are 2007; Chevallier et al., 2007; Feng et al., 2009), the held fixed. Initial estimates (before adjustment) are Oregon and California (ORCA) project, and NOAA’s provided by the CASA-GFED2 model for terrestrial CarbonTracker. The Observation System Simulation fluxes (van der Werf et al., 2006) and by an ocean Experiments showed that the vast volume of OCO inverse model for the ocean regions. Vertical profile data (8,000,000 observations over 16 days) could yield data of CO2 from aircraft are used to validate the improvements in CO2 source-sink estimates over con- flux results and the modeled atmospheric transport. tinents with an error reduction of greater than a factor Uncertainties in the net CO2 emissions are estimated of 2 over land, although the degree of improvement by comparing different configurations of the setup. depends on assumptions about errors and biases in the Important limitations are (1) the sparseness of CO2 observations and models (Baker et al., 2008). observations, which forces the assumption of coherent The ORCA project is a regional example of car- behavior of ecosystems over large distances; (2) the bon data assimilation (Law et al., 2004; Quaife et al., presumed pattern of fluxes, which dominate the high- 2008; Gökede et al., 2010). A terrestrial carbon flux resolution results; and (3) the accuracy of atmospheric model uses inputs of Landsat remote sensing data on transport and mixing in the model. In addition, some disturbance history and land cover, and is calibrated of the resulting emission patterns reflect primarily the a with inventory biomass data and seasonal carbon and priori CASA-GFED2 model. However, on the scale of water fluxes from tower sites (Law et al., 2004). The North America, CarbonTracker has consistently found net terrestrial uptake of 0.2 to 0.8 Pg C yr–1, varying improved model computes prior estimates of carbon fluxes for the land surface, including fluxes for land use year to year, which is not a feature of the CASA- and natural ecosystems. The output of the improved GFED2 model. model is then used as input to an atmospheric transport The most sophisticated carbon data assimilation

OCR for page 53
 VERIFYING GREENHOUSE GAS EMISSIONS to date is the Global and Regional Earth-System ate atmosphere-land-ocean carbon data assimilation Monitoring Using Satellite and In-situ Data (GEMS) system, it should be feasible to greatly improve several project (Hollingsworth et al., 2008),5 funded by the aspects of current data assimilation systems. A high European Commission. The central atmospheric model spatial resolution nested subgrid could surround each is the European Centre for Medium-Range Weather observation site for a better representation of local Forecasts Integrated Forecast System (ECMWF IFS), meteorology. Multiple transport models and multiple which includes CO2 as a model prognostic variable, source models of the oceans and terrestrial biosphere forced by prescribed surface sources and sinks. Radi- could be employed to “internalize” biases between dif- ance data from AIRS have been assimilated into the ferent models as increased (and more realistic) uncer- ECMWF IFS using the 4DVAR system to constrain tainty estimates. Expanded observation systems would the modeled atmospheric CO2 mole fraction in the increase the constraints, allowing less dependence on u pper troposphere. The assimilation consistently each site. Assumptions of coherent behavior of eco- improves the meteorology, albeit slightly (McNally systems over very large spatial scales could be relaxed et al., 2006), and essentially creates, for the first time, because they are being made to compensate for the gridded synoptic maps of CO2 in the midtroposphere sparseness of observations. (Engelen et al., 2009). The modeled upper-troposphere CO2 concentrations are then used as “observations” RECOMMENDATIONS in a standard tracer-inversion step, thus updating the surface sources and sinks (Chevallier et al., 2009). These The remote sensing programs described in Chap- ter 3 together with initiatives to measure 14CO2 and are promising new techniques, but the AIRS observa- tions are weighted in the midtroposphere, so their emissions from large local sources would significantly ability to differentiate surface emissions from different improve our capability to check self-reported emis- countries is limited. sions of CO2 (see Table 4.4). In particular, they would A new effort in the United States attempts to carry allow monitoring of land-use activities responsible for out a multivariate assimilation of meteorology and CO2 a large share of emissions from agriculture, forestry, and (Kang, 2009) using a carbon-climate model, so that other land use and the significant portion of fossil-fuel uncertainties in the air flow and transport estimates emissions from countries that are produced by large are propagated to the trace gas variables, and the trace local sources. They would also greatly improve our gas data inform the meteorological state, especially in capacity to estimate total fossil-fuel emissions from regions with few wind observations. Early results sug- continents. Implementation of initiatives to improve gest that the spread in the vertical profile of CO2 is on tracer-transport inversions would improve estimates the order of 1 ppm due to the uncertainty in meteoro- for total fossil- and non-fossil-fuel emissions at the logical fields alone. With the simultaneous assimilation national level and for time scales ranging from daily of AIRS CO2 observations, the spread in the CO2 to annual, enabling decadal changes to be detected. profile grows to 1.5 ppm due to uncertainty in the Specific recommendations include: AIRS observations and the propagation of spread in the meteorology to the CO2 forecast (Liu et al., 2009). This • The National Aeronautics and Space Adminis- approach, if combined with accurate CO2 observations, tration (NASA) should build and launch a replacement especially near the surface, could effectively quantify for the Orbiting Carbon Observatory. A technical advi- the error due to imperfect knowledge of the meteorol- sory group should be convened to assess the optimal ogy, but leaves the errors due to numerical precision and orbit and sampling strategies for estimating emissions coding structure of the tracer-transport model. from large local sources. In addition to developing a fully coupled multivari- • Extend the surface-based atmospheric sam- pling network to research the atmospheric “domes” of greenhouse gases over a representative sample (e.g., 5 See also GEMS, . GEMS concluded 5-10) of large local emitters, such as cities and power on May 31, 2009, and was replaced by a new project Modelling plants. Key goals of the research program would be (1) Atmospheric Composition and Climate (MACC).

OCR for page 53
 EMISSIONS ESTIMATED FROM ATMOSPHERIC AND OCEANIC MEASUREMENTS TABLE 4.4 Potential Improvements in National Emissions Estimates from Atmospheric and Oceanic Measurements and Models Current Uncertainty for Annual National Uncertainty of Gas Major Sectors or Activities Emissions Possible Improvements in 3-5 Years Improved Methods CO2 Large local sources (e.g., 5 CO2 satellite program, including an OCO replacement, 2 (annual) cities, power plants) new in situ measurements in cities, and a research program 1 (decadal change) to guide network design and satellite validation CO2 Fossil-fuel combustion 4-5 Improved tracer-transport inversion through new 1-3 (annual) observations (14CO2, additional ground stations, OCO 1-2 (decadal change) replacement) and data assimilation CO2 Agriculture, forestry, and 5 Improved tracer-transport inversion through new satellite 5 other land-use net emissions and in situ observations CH4 Total anthropogenic 3-5 Improved tracer-transport models, new satellite and in 2-3 situ observations, and improved emission models through research N 2O Total anthropogenic 4-5 Improved tracer-transport and emission models, additional 3-5 observations CFCs, PFCs, Industrial processes 4-5 Gridded inventories, improved tracer-transport inversion, 2-5 HFCs, and SF6 and measurement of correlated variations of gases NOTES: 1 = 100% (i.e., cannot be certain if it is a source or sink). Ranges represent emission uncertainties in different countries (e.g., 1-3 means that uncertainties are <10% in some countries, 10-25% in some, and 25-50% in others). Uncertainty levels correspond to 2 standard deviations. to optimize trend detection using intensive sampling tory is needed to handle approximately 10,000 new atmospheric 14CO2 samples a year. approaches, (2) to develop easily deployable and cost- effective sampling approaches for a globally extensive • An interagency group, with broad participa- ground-based network, and (3) to provide a means for tion from the research community, should design a validating satellite measurements in these complex and research program to develop gridded high-resolution understudied environments. data on U.S. fossil-fuel emissions and HFC, CFC, • Extend the international WMO Global Atmo- and PFC emissions. An important component of these spheric Watch network of in situ sampling stations maps should be uncertainty estimates that can be used to fill in underrepresented regions globally, thereby directly in data assimilation programs. To support the improving national sampling of regional greenhouse research, the National Science Foundation (NSF), gas emissions. Expanding the network to increase col- NOAA, NASA, and the Department of Energy should lection of vertical profiles of greenhouse gases would expand campaigns to sample the time evolution of constrain atmospheric transport and facilitate inter- tracer fields at high resolution as well as studies that pretation of satellite data. The vertical expansion could use the data to improve transport modeling of tracers. be done with the cooperation of commercial aircraft This research would feed into an international initiative and with balloons flown to higher altitudes. Ideally, to publish gridded estimates for fossil-fuel emissions, all major emitting nations and groups of neighboring as recommended in Chapter 2. smaller nations would participate in the cooperative • Develop a state-of-the-art carbon data assimi- network. The latter may require financial assistance lation system that is coupled and/or synergistic with and capacity building to aid the poorest nations that meteorological, land, and oceanographic data assimila- dominate the most undersampled regions. tion systems for the United States. This would require • Extend the capability of the existing CO2 sam- new approaches for coupling circulation and biogeo- pling network to measure atmospheric 14C. At least one chemical models and for deriving biogeochemical additional U.S. accelerator mass spectrometry labora- properties (and hence surface fluxes) from the obser-

OCR for page 53
0 VERIFYING GREENHOUSE GAS EMISSIONS vations. It would also require enhanced collaboration must be separated from the total non-fossil-fuel flux among federal agencies with carbon observations, to estimate agriculture, forestry, and other land-use especially between NASA and NOAA, so that the (AFOLU) emissions. This requires sustaining the best estimates and the uncertainties in the meteorology terrestrial flux network of sites (see Chapter 2)—aug- become integral components of emission estimation mented with high-precision, high-accuracy CO 2 from a replacement OCO. sampling for bottom-up and top-down model calibra- • Sustain the infrastructure to measure natural tion—and continued measurements of the oceanic sink sources and sinks on land and in the ocean, which (see Appendix C).