Coordinating Lead Authors:

Piers Forster (UK) [SRC:7], Venkatachalam Ramaswamy (USA) [SRC:11],

Concern	Occurrence
SRC >= 5	2
Potentially Biased Authors	2

Lead Authors:

Paulo Artaxo (Brazil), Terje Berntsen (Norway) [SRC:4], Richard Betts (UK) [SRC:6], David W. Fahey (USA), James Haywood (UK) [SRC:13], Judith Lean (USA) [SRC:8], David C. Lowe (New Zealand) [SRC:6], Gunnar Myhre (Norway) [SRC:11], John Nganga (Kenya), Ronald Prinn (USA; New Zealand) [SRC:10], Graciela Raga (Mexico; Argentina), Michael Schulz (France; Germany) [SRC:5], Robert Van Dorland (Netherlands) [SRC:1],

Concern	Occurrence
SRC >= 5	7
SRC 1-4	2
Potentially Biased Authors	9
Impartial Authors	4

Contributing Authors:

G. Bodeker (New Zealand) [SRC:1], O. Boucher (UK; France) [SRC:17], W.D. Collins (USA) [SRC:2], T.J. Conway (USA) [SRC:1], E. Dlugokencky (USA) [SRC:5], J.W. Elkins (USA), D. Etheridge (Australia) [SRC:2], P. Foukal (USA) [SRC:1], P. Fraser (Australia), M. Geller (USA) [SRC:2], F. Joos (Switzerland) [SRC:2][PoC], , C.D. Keeling (USA) [SRC:2], R. Keeling (USA) [SRC:3], S. Kinne (Germany) [SRC:3], K. Lassey (New Zealand), U. Lohmann (Switzerland) [SRC:11], A.C. Manning (UK; New Zealand) [SRC:2], S. Montzka (USA) [SRC:2], D. Oram (UK) [SRC:2], K. O’Shaughnessy (New Zealand), S. Piper (USA) [SRC:1], G.-K. Plattner (Switzerland), M. Ponater (Germany) [SRC:8], N. Ramankutty (USA; India) [SRC:2], G. Reid (USA), D. Rind (USA) [SRC:3], K. Rosenlof (USA), R. Sausen (Germany) [SRC:10], D. Schwarzkopf (USA), S.K. Solanki (Germany; Switzerland) [SRC:2], G. Stenchikov (USA) [SRC:7], N. Stuber (UK; Germany) [SRC:4], T. Takemura (Japan) [SRC:3], C. Textor (France; Germany) [SRC:2], R. Wang (USA), R. Weiss (USA) [SRC:2], T. Whorf (USA) [SRC:2],

Concern	Occurrence
PoC	1
SRC >= 5	6
SRC 1-4	22
Potentially Biased Authors	28
Impartial Authors	9

Review Editors:

Teruyuki Nakajima (Japan) [SRC:5], Veerabhadran Ramanathan (USA) [SRC:4],

Concern	Occurrence
SRC >= 5	1
SRC 1-4	1
Potentially Biased Authors	2

This chapter should be cited as:

Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D.W. Fahey, J. Haywood, J. Lean, D.C. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz and R. Van Dorland, 2007: Changes in Atmospheric Constituents and in Radiative Forcing. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

2.3.1 Atmospheric Carbon Dioxide

This section discusses the instrumental measurements of CO₂,documenting recent changes in atmospheric mixing ratios needed for the RF calculations presented later in the section. In addition, it provides data for the pre-industrial levels of CO₂ required as the accepted reference level for the RF calculations. For dates before about 1950 indirect measurements are relied upon. For these periods, levels of atmospheric CO₂ are usually determined from analyses of air bubbles trapped in polar ice cores. These time periods are primarily considered in Chapter6 .

Awide range of direct and indirect measurements confirm that the atmospheric mixing ratio of CO₂ has increased globally by about 100 ppm (36%) over the last 250 years, from a range of 275 to 285 ppm in the pre-industrial era (AD 1000 – 1750 ) to 379 ppm in 2005 (see FAQ 2.1 ,Figure 1). During this period, the absolute growth rate of CO₂ in the atmosphere increased substantially: the first 50 ppm increase above the pre-industrial value was reached in the 1970 s after more than 200 years, whereas the second 50 ppm was achieved in about 30 years. In the 10 years from 1995 to 2005 atmospheric CO₂ increased by about 19 ppm; the highest average growth rate recorded for any decade since direct atmospheric CO₂ measurements began in the 1950 s. The average rate of increase in CO₂ determined by these direct instrumental measurements over the period 1960 to 2005 is 1.4 ppm yr^-1.

High-precision measurements of atmospheric CO₂ are essential to the understanding of the carbon cycle budgets discussed in Section 7.3 .The first in situ continuous measurements of atmospheric CO₂ made by a high-precision non-dispersive infrared gas analyser were implemented by C.D. Keeling from the Scripps Institution of Oceanography (SIO) (see Section 1.3 ). These began in 1958 at Mauna Loa, Hawaii, located at 19°N ( Keeling et al., 1995 [JoC, SRC] ). The data documented for the first time that not only was CO₂ increasing in the atmosphere, but also that it was modulated by cycles caused by seasonal changes in photosynthesis in the terrestrial biosphere. These measurements were followed by continuous in situ analysis programmes at other sites in both hemispheres ( Conway et al., 1994 [JoC, SRC] ; Nakazawa et al., 1997 [JoC] ; Langenfelds et al., 2002 [JoC] ). In Figure 2.3 ,atmospheric CO₂ mixing ratio data at Mauna Loa in the Northern Hemisphere (NH) are shown with contemporaneous measurements at Baring Head, New Zealand in the Southern Hemisphere (SH; Manning et al., 1997 [NPR, ARC] ; Keeling and Whorf, 2005 [NPR, SRC] ). These two stations provide the longest continuous records of atmospheric CO₂ in the NH and SH, respectively. Remote sites such as Mauna Loa, Baring Head, Cape Grim (Tasmania) and the South Pole were chosen because air sampled at such locations shows little short-term variation caused by local sources and sinks of CO₂ and provided the first data from which the global increase of atmospheric CO₂ was documented. Because CO₂ is a LLGHG and well mixed in the atmosphere, measurements made at such sites provide an integrated picture of large parts of the Earth including continents and city point sources. Note that this also applies to the other LLGHGs reported in Section 2.3 .

In the 1980 s and 1990 s, it was recognised that greater coverage of CO₂ measurements over continental areas was required to provide the basis for estimating sources and sinks of atmospheric CO₂ over land as well as ocean regions. Because continuous CO₂ analysers are relatively expensive to maintain and require meticulous on-site calibration, these records are now widely supplemented by air sample flask programmes, where air is collected in glass and metal containers at a large number of continental and marine sites. After collection, the filled flasks are sent to central well-calibrated laboratories for analysis. The most extensive network of international air sampling sites is operated by the National Oceanic and Atmospheric Administration’s Global Monitoring Division (NOAA/GMD; formerly NOAA/Climate Monitoring and Diagnostics Laboratory (CMDL)) in the USA. This organisation collates measurements of atmospheric CO₂ from six continuous analyser locations as well as weekly flask air samples from a global network of almost 50 surface sites. Many international laboratories make atmospheric CO₂ observations and worldwide databases of their measurements are maintained by the Carbon Dioxide Information Analysis Center (CDIAC) and by the World Data Centre for Greenhouse Gases (WDCGG) in the WMO Global Atmosphere Watch (GAW) programme. ^[6]

The increases in global atmospheric CO₂ since the industrial revolution are mainly due to CO₂ emissions from the combustion of fossil fuels, gas flaring and cement production. Other sources include emissions due to land use changes such as deforestation ( Houghton, 2003 [JoC, MoS] ) and biomass burning ( Andreae and Merlet, 2001 [JoC] ; van der Werf, 2004 [JoC] ). After entering the atmosphere, CO₂ exchanges rapidly with the short-lived components of the terrestrial biosphere and surface ocean, and is then redistributed on time scales of hundreds of years among all active carbon reservoirs including the long-lived terrestrial biosphere and deep ocean. The processes governing the movement of carbon between the active carbon reservoirs, climate carbon cycle feedbacks and their importance in determining the levels of CO₂ remaining in the atmosphere, are presented in Section 7.3 ,where carbon cycle budgets are discussed.

The increase in CO₂ mixing ratios continues to yield the largest sustained RF of any forcing agent. The RF of CO₂ is a function of the change in CO₂ in the atmosphere over the time period under consideration. Hence, a key question is ‘How is the CO₂ released from fossil fuel combustion, cement production and land cover change distributed amongst the atmosphere, oceans and terrestrial biosphere?’. This partitioning has been investigated using a variety of techniques. Among the most powerful of these are measurements of the carbon isotopes in CO₂ as well as high-precision measurements of atmospheric oxygen (O₂)content. The carbon contained in CO₂ has two naturally occurring stable isotopes denoted ¹²Cand ¹³C. The first of these, ¹²C, is the most abundant isotope at about 99%, followed by ¹³Cat about 1%. Emissions of CO₂ from coal, gas and oil combustion and land clearing have ¹³C/¹²Cisotopic ratios that are less than those in atmospheric CO₂,and each carries a signature related to its source. Thus, as shown in Prentice et al. 2001 [NPR] ), when CO₂ from fossil fuel combustion enters the atmosphere, the ¹³C/¹²Cisotopic ratio in atmospheric CO₂ decreases at a predictable rate consistent with emissions of CO₂ from fossil origin. Note that changes in the ¹³C/¹²Cratio of atmospheric CO₂ are also caused by other sources and sinks, but the changing isotopic signal due to CO₂ from fossil fuel combustion can be resolved from the other components ( Francey et al., 1995 [JoC] ). These changes can easily be measured using modern isotope ratio mass spectrometry, which has the capability of measuring ¹³C/¹²Cin atmospheric CO₂ to better than 1 part in 10⁵ ( Ferretti et al., 2000 [JoC, SRC] ). Data presented in Figure 2.3 for the ¹³C/¹²Cratio of atmospheric CO₂ at Mauna Loa show a decreasing ratio, consistent with trends in both fossil fuel CO₂ emissions and atmospheric CO₂ mixing ratios ( Andres et al., 2000 [NPR, MoS] ; Keeling et al., 2005 [NPR, SRC] ).

AtmosphericO₂ measurements provide a powerful and independent method of determining the partitioning of CO₂ between the oceans and land ( Keeling et al., 1996 [JoC, SRC] ). AtmosphericO₂ and CO₂ changes are inversely coupled during plant respiration and photosynthesis. In addition, during the process of combustionO₂ is removed from the atmosphere, producing a signal that decreases as atmospheric CO₂ increases on a molar basis( Figure 2.3 ). Measuring changes in atmosphericO₂ is technically challenging because of the difficulty of resolving changes at the part-per-million level in a background mixing ratio of roughly 209,000 ppm. These difficulties were first overcome by Keeling and Shertz 1992 [JoC, SRC] ), who used an interferometric technique to show that it is possible to track both seasonal cycles and the decline ofO₂ in the atmosphere at the part-per-million level( Figure 2.3 ). Recent work by Manning and Keeling 2006 [JoC, SRC] ) indicates that atmosphericO₂ is decreasing at a faster rate than CO₂ is increasing, which demonstrates the importance of the oceanic carbon sink. Measurements of both the ¹³C/¹²Cratio in atmospheric CO₂ and atmosphericO₂ levels are valuable tools used to determine the distribution of fossil-fuel derived CO₂ among the active carbon reservoirs, as discussed in Section 7.3 .In Figure 2.3 ,recent measurements in both hemispheres are shown to emphasize the strong linkages between atmospheric CO₂ increases,O₂ decreases, fossil fuel consumption and the ¹³C/¹²Cratio of atmospheric CO₂.

From 1990 to 1999, a period reported in Prentice et al. 2001 [NPR] ), the emission rate due to fossil fuel burning and cement production increased irregularly from 6.1 to 6.5 GtC yr^–1 or about 0.7% yr^–1.From 1999 to 2005 however, the emission rate rose systematically from 6.5 to 7.8 GtC yr^–1 ( BP, 2006 [NPR] ; Marland et al., 2006 [NPR] ) or about 3.0% yr^–1,representing a period of higher emissions and growth in emissions than those considered in the TAR (see Figure 2.3 ). Carbon dioxide emissions due to global annual fossil fuel combustion and cement manufacture combined have increased by 70% over the last 30 years ( Marland et al., 2006 [NPR] ). The relationship between increases in atmospheric CO₂ mixing ratios and emissions has been tracked using a scaling factor known as the apparent ‘airborne fraction’, defined as the ratio of the annual increase in atmospheric CO₂ to the CO₂ emissions from annual fossil fuel and cement manufacture combined ( Keeling et al., 1995 [JoC, SRC] ). On decadal scales, this fraction has averaged about 60% since the 1950 s. Assuming emissions of 7 GtC yr^–1 and an airborne fraction remaining at about 60%, Hansen and Sato 2004 [JoC, ARC] ) predicted that the underlying long-term global atmospheric CO₂ growth rate will be about 1.9 ppm yr^–1,a value consistent with observations over the 1995 to 2005 decade.

Carbon dioxide emissions due to land use changes during the 1990 s are estimated as 0.5 to 2.7 GtC yr^–1 ( Section 7.3 , Table 7.2 ), contributing 6% to 39% of the CO₂ growth rate ( Brovkin et al., 2004 [ARC] Prentice et al. 2001 [NPR] ) cited an inventory-based estimate that land use change resulted in net emissions of 121 GtC between 1850 and 1990, after Houghton 1999 [JoC] , 2000 ). The estimate for this period was revised upwards to 134 GtC by Houghton 2003 [JoC, MoS] ), mostly due to an increase in estimated emissions prior to 1960 Houghton 2003 [JoC, MoS] ) also extended the inventory emissions estimate to 2000, giving cumulative emissions of 156 GtC since 1850 . In carbon cycle simulations by Brovkin et al. 2004 [ARC] and Matthews et al. 2004 [JoC, ARC] ), land use change emissions contributed 12 to 35 ppm of the total CO₂ rise from 1850 to 2000 ( Section 2.5.3 , Table 2.8 ). Historical changes in land cover are discussed in Section 2.5.2 ,and the CO₂ budget over the 1980 s and 1990 s is discussed further in Section 7.3 .

In 2005, the global mean average CO₂ mixing ratio for the SIO network of 9 sites was 378.75 ± 0.13 ppm and for the NOAA/GMD network of 40 sites was 378.76 ± 0.05 ppm, yielding a global average of almost 379 ppm. For both networks, only sites in the remote marine boundary layer are used and high-altitude locations are not included. For example, the Mauna Loa site is excluded due to an ‘altitude effect’ of about 0.5 ppm. In addition, the 2005 values are still pending final reference gas calibrations used to measure the samples.

New measurements of CO₂ from Antarctic ice and firn ( MacFarling Meure et al., 2006 [JoC] ) update and extend those from Etheridge et al. 1996 [JoC, SRC] ) to AD 0. The CO₂ mixing ratio in 1750 was 277 ± 1.2 ppm. ^[7] This record shows variations between 272 and 284 ppm before 1800 and also that CO₂ mixing ratios dropped by 5 to 10 ppm between 1600 and 1800 (see Section 6.3 ). The RF calculations usually take 1750 as the pre-industrial index (e.g., the TAR and this report). Therefore, using 1750 may slightly overestimate the RF, as the changes in the mixing ratios of CO₂,CH₄ andN₂Oafter the end of this naturally cooler period may not be solely attributable to anthropogenic emissions. Using 1860 as an alternative start date for the RF calculations would reduce the LLGHG RF by roughly 10%. For the RF calculation, the data from Law Dome ice cap in the Antarctic are used because they show the highest age resolution (approximately 10 years) of any ice core record in existence. In addition, the high-precision data from the cores are connected to direct observational records of atmospheric CO₂ from Cape Grim, Tasmania.

The simple formulae for RF of the LLGHG quoted in Ramaswamy et al. 2001 [NPR, MoS, SRC] ) are still valid. These formulae are based on global RF calculations where clouds, stratospheric adjustment and solar absorption are included, and give an RF of +3.7 Wm^–2 for a doubling in the CO₂ mixing ratio. (The formula used for the CO₂ RF calculation in this chapter is the IPCC 1990 [NPR] ) expression as revised in the TAR. Note that for CO₂,RF increases logarithmically with mixing ratio.) Collins et al. 2006 [JoC, MoS, SRC] ) performed a comparison of five detailed line-by-line models and 20 GCM radiation schemes. The spread of line-by-line model results were consistent with the ±10% uncertainty estimate for the LLGHG RFs adopted in Ramaswamy et al. 2001 [NPR, MoS, SRC] ) and a similar ±10% for the 90% confidence interval is adopted here. However, it is also important to note that these relatively small uncertainties are not always achievable when incorporating the LLGHG forcings into GCMs. For example, both Collins et al. 2006 [JoC, MoS, SRC] and Forster and Taylor 2006 [JoC, MoS, SRC] ) found that GCM radiation schemes could have inaccuracies of around 20% in their total LLGHG RF (see also Sections 2.3.2 and 10.2 ).

Using the global average value of 379 ppm for atmospheric CO₂ in 2005 gives an RF of 1.66 ± 0.17 Wm^–2;a contribution that dominates that of all other forcing agents considered in this chapter. This is an increase of 13 to 14% over the value reported for 1998 in Ramaswamy et al. 2001 [NPR, MoS, SRC] ). This change is solely due to increases in atmospheric CO₂ and is also much larger than the RF changes due to other agents. In the decade 1995 to 2005, the RF due to CO₂ increased by about 0.28 Wm^–2 (20%), an increase greater than that calculated for any decade since at least 1800 (see Section 6.6 and FAQ 2.1 ,Figure 1).

Table 2.1 summarises the present-day mixing ratios and RF for the LLGHGs, and indicates changes since 1998 . The RF from CO₂ and that from the other LLGHGs have a high level of scientific understanding( Section 2.9 , Table 2.11 ). Note that the uncertainty in RF is almost entirely due to radiative transfer assumptions and not mixing ratio estimates, therefore trends in RF can be more accurately determined than the absolute RF. From Section 2.5.3 , Table 2.8 ,the contribution from land use change to the present CO₂ RF is likely to be about 0.4 Wm^–2 (since 1850 ). This implies that fossil fuel and cement production have likely contributed about three-quarters of the current RF.

Table 2.1 .Present-day concentrations and RF for the measured LLGHGs. The changes since 1998 (the time of the TAR estimates) are also shown.

Notes: ^a See Table 2.14 for common names of gases and the radiative efficiencies used to calculate RF. ^b Mixing ratio errors are 90% confidence ranges of combined 2005 data, including intra-annual standard deviation, measurement and global averaging uncertainty. Standard deviations were multiplied by 1.645 to obtain estimates of the 90% confidence range; this assumes normal distributions. Data for CO₂ are combined measurements from the NOAA Earth System Research Laboratory (ESRL) and SIO networks (see Section 2.3.1 ); CH₄ measurements are combined data from the ESRL and Advanced Global Atmospheric Gases Experiment (AGAGE) networks (see Section 2.3.2 ); halocarbon measurements are the average of ESRL and AGAGE networks. University of East Anglia (UEA) and Pennsylvania State University (PSU) measurements were also used (see Section 2.3.3 ). ^c Pre-industrial values are zero except for CO₂ (278 ppm), CH₄ (715 ppb; 700 ppb was used in the TAR),N₂O(270 ppb) and CF₄ (40 ppt). ^d 90% confidence ranges for RF are not shown but are approximately 10%. This confidence range is almost entirely due to radiative transfer assumptions, therefore trends remain valid when quoted to higher accuracies. Higher precision data are used for totals and affect rounding of the values. Percent changes are calculated relative to 1998 . ^e Data available from AGAGE network only. ^f Data for 1998 not available; values from 1999 are used. ^g Data from UEA only. ^h Data from 2003 are used due to lack of available data for 2004 and 2005 . ⁱ Data from ESRL only. ^j 1997 data from PSU ( (Khalil et al., 2003, ) not updated) are used. ^k CFC total includes a small 0.009 Wm^–2 RF from CFC-13, CFC-114, CFC-115 and the halons, as measurements of these were not updated.

2.3.2 Atmospheric Methane

This section describes the current global measurement programmes for atmospheric CH₄,which provide the data required for the understanding of its budget and for the calculation of its RF. In addition, this section provides data for the pre-industrial levels of CH₄ required as the accepted reference level for these calculations. Detailed analyses of CH₄ budgets and its biogeochemistry are presented in Section 7.4 .

Methane has the second-largest RF of the LLGHGs after CO₂ ( Ramaswamy et al., 2001 [NPR, MoS, SRC] ). Over the last 650 kyr, ice core records indicate that the abundance of CH₄ in the Earth’s atmosphere has varied from lows of about 400 ppb during glacial periods to highs of about 700 ppb during interglacials ( Spahni et al., 2005 [JoC, ARC] ) with a single measurement from the Vostok core reaching about 770 ppb (see Figure 6.3 ).

In 2005, the global average abundance of CH₄ measured at the network of 40 surface air flask sampling sites operated by NOAA/GMD in both hemispheres was 1,774.62 ± 1.22 ppb. ^[8] This is the most geographically extensive network of sites operated by any laboratory and it is important to note that the calibration scale it uses has changed since the TAR ( Dlugokencky et al., 2005 [JoC, SRC] ). The new scale (known as NOAA04) increases all previously reported CH₄ mixing ratios from NOAA/GMD by about 1%, bringing them into much closer agreement with the Advanced Global Atmospheric Gases Experiment (AGAGE) network. This scale will be used by laboratories participating in the WMO’s GAW programme as a ‘common reference’. Atmospheric CH₄ is also monitored at five sites in the NH and SH by the AGAGE network. This group uses automated systems to make 36 CH₄ measurements per day at each site, and the mean for 2005 was 1,774.03 ± 1.68 ppb with calibration and methods described by Cunnold et al. 2002 [JoC] ). For the NOAA/GMD network, the 90% confidence interval is calculated with a Monte Carlo technique, which only accounts for the uncertainty due to the distribution of sampling sites. For both networks, only sites in the remote marine boundary layer are used and continental sites are not included. Global databases of atmospheric CH₄ measurements for these and other CH₄ measurement programmes (e.g., Japanese, European and Australian) are maintained by the CDIAC and by the WDCGG in the GAW programme.

Present atmospheric levels of CH₄ are unprecedented in at least the last 650 kyr ( Spahni et al., 2005 [JoC, ARC] ). Direct atmospheric measurements of the gas made at a wide variety of sites in both hemispheres over the last 25 years show that, although the abundance of CH₄ has increased by about 30% during that time, its growth rate has decreased substantially from highs of greater than 1% yr^–1 in the late 1 70s and early 1980 s ( Blake and Rowland, 1988 [JoC] ) to lows of close to zero towards the end of the 1990 s ( Dlugokencky et al., 1998 [JoC, SRC] ; Simpson et al., 2002 [JoC] ). The slowdown in the growth rate began in the 1980 s, decreasing from 14 ppb yr^–1 (about 1% yr^–1)in 1984 to close to zero during 1999 to 2005, for the network of surface sites maintained by NOAA/GMD ( Dlugokencky et al., 2003 [JoC, SRC] ). Measurements by Lowe et al. 2004 [JoC, SRC] ) for sites in the SH and Cunnold et al. 2002 [JoC] ) for the network of GAGE/AGAGE sites show similar features. A key feature of the global growth rate of CH₄ is its current interannual variability, with growth rates ranging from a high of 14 ppb yr^–1 in 1998 to less than zero in 2001, 2004 and 2005 .( Figure 2.4 )

The reasons for the decrease in the atmospheric CH₄ growth rate and the implications for future changes in its atmospheric burden are not understood ( Prather et al., 2001 [NPR, ARC] ) but are clearly related to changes in the imbalance between CH₄ sources and sinks. Most CH₄ is removed from the atmosphere by reaction with the hydroxyl free radical (OH), which is produced photochemically in the atmosphere. The role of OH in controlling atmospheric CH₄ levels is discussed in Section 2.3.5 .Other minor sinks include reaction with free chlorine ( Platt et al., 2004 [SRC] ; Allan et al., 2005 [JoC] ), destruction in the stratosphere and soil sinks ( Born et al., 1990 [JoC] ).

The total global CH₄ source is relatively well known but the strength of each source component and their trends are not. As detailed in Section 7.4 ,the sources are mostly biogenic and include wetlands, rice agriculture, biomass burning and ruminant animals. Methane is also emitted by various industrial sources including fossil fuel mining and distribution. Prather et al. 2001 [NPR, ARC] ) documented a large range in ‘bottom-up’ estimates for the global source of CH₄.New source estimates published since then are documented in Table 7.6 .However, as reported by Bergamaschi et al. 2005 [MoS] ), national inventories based on ‘bottom-up’ studies can grossly underestimate emissions and ‘top-down’ measurement-based assessments of reported emissions will be required for verification. Keppler et al. 2006 [JoC] ) reported the discovery of emissions of CH₄ from living vegetation and estimated that this contributed 10 to 30% of the global CH₄ source. This work extrapolates limited measurements to a global source and has not yet been confirmed by other laboratories, but lends some support to space-borne observations of CH₄ plumes above tropical rainforests reported by Frankenberg et al. 2005 [JoC] ). That such a potentially large source of CH₄ could have been missed highlights the large uncertainties involved in current ‘bottom-up’ estimates of components of the global source (see Section 7.4 ).

Several wide-ranging hypotheses have been put forward to explain the reduction in the CH₄ growth rate and its variability. For example, Hansen et al. 2000 [JoC, ARC] ) considered that economic incentives have led to a reduction in anthropogenic CH₄ emissions. The negligible long-term change in its main sink (OH; see Section 2.3.5 and Figure 2.8 )implies that CH₄ emissions are not increasing. Similarly, Dlugokencky et al. 1998 [JoC, SRC] and Francey et al. 1999 [JoC] ) suggested that the slowdown in the growth rate reflects a stabilisation of CH₄ emissions, given that the observations are consistent with stable emissions and lifetime since 1982 .

Relatively large anomalies occurred in the growth rate during 1991 and 1998, with peak values reaching 15 and 14 ppb yr^–1,respectively (about 1% yr^–1). The anomaly in 1991 was followed by a dramatic drop in the growth rate in 1992 and has been linked with the Mt. Pinatubo volcanic eruption in June 1991, which injected large amounts of ash and (sulphur dioxide) SO₂ into the lower stratosphere of the tropics with subsequent impacts on tropical photochemistry and the removal of CH₄ by atmospheric OH ( Bekki et al., 1994 [JoC] ; Dlugokencky et al., 1996 [JoC, SRC] Lelieveld et al. 1998 [JoC, MoS, ARC] and Walter et al. 2001a [JoC, MoS, ARC] ,b) proposed that lower temperatures and lower precipitation in the aftermath of the Mt. Pinatubo eruption could have suppressed CH₄ emissions from wetlands. At this time, and in parallel with the growth rate anomaly in the CH₄ mixing ratio, an anomaly was observed in the ¹³C/¹²Cratio of CH₄ at surface sites in the SH. This was attributed to a decrease in emissions from an isotopically heavy source such as biomass burning ( Lowe et al., 1997 [JoC, SRC] ; Mak et al., 2000 [JoC, SRC] ), although these data were not confirmed by lower frequency measurements from the same period made by Francey et al. 1999 [JoC] ).

For the relatively large increase in the CH₄ growth rate reported for 1998, Dlugokencky et al. 2001 [JoC, SRC] ) suggested that wetland and boreal biomass burning sources might have contributed to the anomaly, noting that 1998 was the warmest year globally since surface instrumental temperature records began. Using an inverse method, Chen and Prinn 2006 [JoC, MoS, SRC] ) attributed the same event primarily to increased wetland and rice region emissions and secondarily to biomass burning. The same conclusion was reached by Morimoto et al. 2006 [JoC] ), who used carbon isotopic measurements of CH₄ to constrain the relative contributions of biomass burning (one-third) and wetlands (two-thirds) to the increase.

Based on ice core measurements of CH₄ ( Etheridge et al., 1998 [JoC, SRC] ), the pre-industrial global value for CH₄ from 1700 to 1800 was 715 ± 4 ppb (it was also 715 ± 4 ppb in 1750 ), thus providing the reference level for the RF calculation. This takes into account the inter-polar difference in CH₄ as measured from Greenland and Antarctic ice cores.

The RF due to changes in CH₄ mixing ratio is calculated with the simplified yet still valid expression for CH₄ given in Ramaswamy et al. 2001 [NPR, MoS, SRC] ). The change in the CH₄ mixing ratio from 715 ppb in 1750 to 1,774 ppb (the average mixing ratio from the AGAGE and GMD networks) in 2005 gives an RF of +0.48 ± 0.05 Wm^–2,ranking CH₄ as the second highest RF of the LLGHGs after CO₂ ( Table 2.1 ). The uncertainty range in mixing ratios for the present day represents intra-annual variability, which is not included in the pre-industrial uncertainty estimate derived solely from ice core sampling precision. The estimate for the RF due to CH₄ is the same as in Ramaswamy et al. 2001 [NPR, MoS, SRC] ) despite the small increase in its mixing ratio. The spectral absorption by CH₄ is overlapped to some extent byN₂Olines (taken into account in the simplified expression). Taking the overlapping lines into account using currentN₂Omixing ratios instead of pre-industrial mixing ratios (as in Ramaswamy et al., 2001 [NPR, MoS, SRC] ) reduces the current RF due to CH₄ by 1%.

Collins et al. 2006 [JoC, MoS, SRC] ) confirmed that line-by-line models agree extremely well for the calculation of clear-sky instantaneous RF from CH₄ andN₂Owhen the same atmospheric background profile is used. However, GCM radiation schemes were found to be in poor agreement with the line-by-line models, and errors of over 50% were possible for CH₄,N₂Oand the CFCs. In addition, a small effect from the absorption of solar radiation was found with the line-by-line models, which the GCMs did not include( Section 10.2 ).

2.3.3 Other Kyoto Protocol Gases

At the time of the TAR,N₂Ohad the fourth largest RF among the LLGHGs behind CO₂,CH₄ and CFC-12. The TAR quoted an atmosphericN₂Oabundance of 314 ppb in 1998, an increase of 44 ppb from its pre-industrial level of around 270 ± 7 ppb, which gave an RF of +0.15 ± 0.02 Wm^–2.This RF is affected by atmospheric CH₄ levels due to overlapping absorptions. AsN₂Ois also the major source of ozone-depleting nitric oxide (NO) and nitrogen dioxide (NO₂)in the stratosphere, it is routinely reviewed in the ozone assessments; the most recent assessment ( Montzka et al., 2003 [NPR, SRC] ) recommended an atmospheric lifetime of 114 years forN₂O. The TAR pointed out large uncertainties in the major soil, agricultural, combustion and oceanic sources ofN₂O. Given these emission uncertainties, its observed rate of increase of 0.2 to 0.3% yr^–1 was not inconsistent with its better-quantified major sinks (principally stratospheric destruction). The primary driver for the industrial era increase ofN₂Owas concluded to be enhanced microbial production in expanding and fertilized agricultural lands.

Ice core data forN₂Ohave been reported extending back 2,000 years and more before present ( MacFarling Meure et al., 2006 [JoC] ; Section 6.6 ). These data, as for CO₂ and CH₄,show relatively little changes in mixing ratios over the first 1,800 years of this record, and then exhibit a relatively rapid rise (see FAQ 2.1 ,Figure 1). Since 1998, atmosphericN₂Olevels have steadily risen to 319 ± 0.12 ppb in 2005, and levels have been increasing approximately linearly (at around 0.26% yr^–1)for the past few decades( Figure 2.5 ). A change in theN₂Omixing ratio from 270 ppb in 1750 to 319 ppb in 2005 results in an RF of +0.16 ± 0.02 Wm^–2,calculated using the simplified expression given in Ramaswamy et al. 2001 [NPR, MoS, SRC] ). The RF has increased by 11% since the time of the TAR( Table 2.1 ). As CFC-12 levels slowly decline (see Section 2.3.4 ),N₂Oshould, with its current trend, take over third place in the LLGHG RF ranking.

Since the TAR, understanding of regionalN₂Ofluxes has improved. The results of various studies that quantified the globalN₂Oemissions from coastal upwelling areas, continental shelves, estuaries and rivers suggest that these coastal areas contribute 0.3 to 6.6 TgN yr^–1 ofN₂Oor 7 to 61% of the total oceanic emissions ( Bange et al., 1996 [JoC] ; Nevison et al., 2004b [JoC, SRC] ; Kroeze et al., 2005 [MoS] ; see also Section 7.4 ). Using inverse methods and AGAGE Ireland measurements, Manning et al. 2003 [JoC, MoS] ) estimated EUN₂Oemissions of 0.9 ± 0.1 TgN yr^–1 that agree well with the United Nations Framework Convention on Climate Change (UNFCCC)N₂Oinventory (0.8 ± 0.1 TgN yr^–1 Melillo et al. 2001 [JoC] ) provided evidence from Brazilian land use sequences that the conversion of tropical forest to pasture leads to an initial increase but a later decline in emissions ofN₂Orelative to the original forest. They also deduced that Brazilian forest soils alone contribute about 10% of total globalN₂Oproduction. Estimates ofN₂Osources and sinks using observations and inverse methods had earlier implied that a large fraction of globalN₂Oemissions in 1978 to 1988 were tropical: specifically 20 to 29% from 0° to 30°S and 32 to 39% from 0° to 30°N compared to 11 to 15% from 30°S to 90°S and 22 to 34% from 30°N to 90°N ( Prinn et al., 1990 [JoC, SRC] ). These estimates were uncertain due to their significant sensitivity to assumed troposphere-stratosphere exchange rates that strongly influence inter-hemispheric gradients. Hirsch et al. 2006 [JoC, MoS] ) used inverse modelling to estimate significantly lower emissions from 30°S to 90°S (0 to 4%) and higher emissions from 0° to 30°N (50 to 64%) than Prinn et al. 1990 [JoC, SRC] ) during 1998 to 2001, with 26 to 36% from the oceans. The stratosphere is also proposed to play an important role in the seasonal cycles ofN₂O(Nevison et al., 2004a ). For example, its well-defined seasonal cycle in the SH has been interpreted as resulting from the net effect of seasonal oceanic outgassing of microbially producedN₂O, stratospheric intrusion of low-N₂Oair and other processes ( Nevison et al., 2005 [JoC] ). Nevison et al. also estimated a Southern Ocean (30°S–90°S)N₂Osource of 0.9 TgN yr^–1,or about 5% of the global total. The complex seasonal cycle in the NH is more difficult to reconcile with seasonal variations in the northern latitude soil sources and stratospheric intrusions ( Prinn et al., 2000 [JoC, SRC] ; T. Liao et al., 2004 [Ambiguous] ). The destruction ofN₂Oin the stratosphere causes enrichment of its heavier isotopomers and isotopologues, providing a potential method to differentiate stratospheric and surface flux influences on troposphericN₂O(Morgan et al., 2004 ).

Human-made PFCs, HFCs and SF₆ are very effective absorbers of infrared radiation, so that even small amounts of these gases contribute significantly to the RF of the climate system. The observations and global cycles of the major HFCs, PFCs and SF₆ were reviewed in Velders et al. 2005 [NPR] ), and this section only provides a brief review and an update for these species. Table 2.1 shows the present mixing ratio and recent trends in the halocarbons and their RFs. Absorption spectra of most halocarbons reviewed here and in the following section are characterised by strongly overlapping spectral lines that are not resolved at tropospheric pressures and temperatures, and there is some uncertainty in cross section measurements. Apart from the uncertainties stemming from the cross sections themselves, differences in the radiative flux calculations can arise from the spectral resolution used, tropopause heights, vertical, spatial and seasonal distributions of the gases, cloud cover and how stratospheric temperature adjustments are performed. IPCC/TEAP ( 2005 ) concluded that the discrepancy in the RF calculation for different halocarbons, associated with uncertainties in the radiative transfer calculation and the cross sections, can reach 40%. Studies reviewed in IPCC/TEAP ( 2005 ) for the more abundant HFCs show that an agreement better than 12% can be reached for these when the calculation conditions are better constrained (see Section 2.10.2 ).

The HFCs of industrial importance have lifetimes in the range 1.4 to 270 years. The HFCs with the largest observed mole fractions in 1998 (as reported in the TAR) were, in descending order, HFC-23, HFC-134a and HFC-152a. In 2005, the observed mixing ratios of the major HFCs in the atmosphere were 35 ppt for HFC-134a, 17.5 ppt for HFC-23 ( 2003 value), 3.7 ppt for HFC-125 and 3.9 ppt for HFC-152a( Table 2.1 ). Within the uncertainties in calibration and emissions estimates, the observed mixing ratios of the HFCs in the atmosphere can be explained by the anthropogenic emissions. Measurements are available from GMD ( Thompson et al., 2004 [NPR] ) and AGAGE ( Prinn et al., 2000 [JoC, SRC] ; O’Doherty et al., 2004 [JoC] ; Prinn et al., 2005b [NPR, SRC] ) networks as well as from University of East Anglia (UEA) studies in Tasmania (updated from Oram et al., 1998 [JoC, SRC] ; Oram, 1999 [NPR, MoS, SRC] ). These data, summarised in Figure 2.6 ,show a continuation of positive HFC trends and increasing latitudinal gradients (larger trends in the NH) due to their predominantly NH sources. The air conditioning refrigerant HFC-134a is increasing at a rapid rate in response to growing emissions arising from its role as a replacement for some CFC refrigerants. With a lifetime of about 14 years, its current trends are determined primarily by its emissions and secondarily by its atmospheric destruction. Emissions of HFC-134a estimated from atmospheric measurements are in approximate agreement with industry estimates ( Huang and Prinn, 2002 [JoC, MoS, SRC] ; O’Doherty et al., 2004 [JoC] ). IPCC/TEAP ( 2005 ) reported that global HFC-134a emissions started rapidly increasing in the early 1990 s, and that in Europe, sharp increases in emissions are noted for HFC-134a from 1995 to 1998 and for HFC-152a from 1996 to 2000, with some levelling off through 2003 . The concentration of the foam blower HFC-152a, with a lifetime of only about 1.5 years, is rising approximately exponentially, with the effects of increasing emissions only partly offset by its rapid atmospheric destruction. Hydrofluorocarbon-23 has a very long atmospheric lifetime (approximately 270 years) and is mainly produced as a by-product of HCFC-22 production. Its concentrations are rising approximately linearly, driven by these emissions, with its destruction being only a minor factor in its budget. There are also smaller but rising concentrations of HFC-125 and HFC-143a, which are both refrigerants.

The PFCs, mainly CF₄ (PFC-14) andC₂F₆ (PFC-116), and SF₆ have very large radiative efficiencies and lifetimes in the range 1,000 to 50,000 years (see Section 2.10 , Table 2.14 ), and make an essentially permanent contribution to RF. The SF₆ andC₂F₆ concentrations and RFs have increased by over 20% since the TAR( Table 2.1 and Figure 2.6 ), but CF₄ concentrations have not been updated since 1997 . Both anthropogenic and natural sources of CF₄ are important to explain its observed atmospheric abundance. These PFCs are produced as by-products of traditional aluminium production, among other activities. The CF₄ concentrations have been increasing linearly since about 1960 and CF₄ has a natural source that accounts for about one-half of its current atmospheric content ( Harnisch et al., 1996 [JoC] ). Sulphur hexafluoride (SF₆)is produced for use as an electrical insulating fluid in power distribution equipment and also deliberately released as an essentially inert tracer to study atmospheric and oceanic transport processes. Its concentration was 4.2 ppt in 1998 (TAR) and has continued to increase linearly over the past decade, implying that emissions are approximately constant. Its very long lifetime ensures that its emissions accumulate essentially unabated in the atmosphere.

2.3.4 Montreal Protocol Gases

The Montreal Protocol on Substances that Deplete the Ozone Layer regulates many radiatively powerful greenhouse gases for the primary purpose of lowering stratospheric chlorine and bromine concentrations. These gases include the CFCs, HCFCs, chlorocarbons, bromocarbons and halons. Observations and global cycles of these gases were reviewed in detail in Chapter1 of the 2002 Scientific Assessment of Ozone Depletion ( WMO, 2003 [NPR] ) and in IPCC/TEAP ( 2005 ). The discussion here focuses on developments since these reviews and on those gases that contribute most to RF rather than to halogen loading. Using observed 2005 concentrations, the Montreal Protocol gases have contributed 12% (0.320 Wm^–2)to the direct RF of all LLGHGs and 95% to the halocarbon RF( Table 2.1 ). This contribution is dominated by the CFCs. The effect of the Montreal Protocol on these gases has been substantial. IPCC/TEAP ( 2005 ) concluded that the combined CO₂-equivalent emissions of CFCs, HCFCs and HFCs decreased from a peak of about 7.5 GtCO₂-eq yr^–1 in the late 1980 s to about 2.5 GtCO₂-eq yr^–1 by the year 2000, corresponding to about 10% of that year’s CO₂ emissions due to global fossil fuel burning.

Measurements of the CFCs and HCFCs, summarised in Figure 2.6 ,are available from the AGAGE network ( Prinn et al., 2000 [JoC, SRC] , 2005b ) and the GMD network ( Montzka et al., 1999 [JoC, MoS, SRC] updated; Thompson et al., 2004 [NPR] ). Certain flask measurements are also available from the University of California at Irvine (UCI; Blake et al., 2001 [JoC] updated) and UEA ( Oram et al., 1998 [JoC, SRC] ; Oram, 1999 [NPR, MoS, SRC] updated). Two of the major CFCs (CFC-11 and CFC-113) have both been decreasing in the atmosphere since the mid- 1990 s. While their emissions have decreased very substantially in response to the Montreal Protocol, their long lifetimes of around 45 and 85 years, respectively, mean that their sinks can reduce their levels by only about 2% and 1% yr^–1,respectively. Nevertheless, the effect of the Montreal Protocol has been to substantially reduce the growth of the halocarbon RF, which increased rapidly from 1950 until about 1990 . The other major CFC (CFC-12), which is the third most important LLGHG, is finally reaching a plateau in its atmospheric levels (emissions equal loss) and may have peaked in 2003 . Its 100-year lifetime means that it can decrease by only about 1% yr^–1 even when emissions are zero. The levelling off for CFC-12 and approximately linear downward trends for CFC-11 and CFC-113 continue. Latitudinal gradients of all three are very small and decreasing as expected. The combined CFC and HCFC RF has been slowly declining since 2003 . Note that the 1998 concentrations of CFC-11 and CFC-12 were overestimated in Table 6.1 of the TAR. This means that the total halocarbon RF quoted for 2005 in Table 2.1 (0.337 Wm^–2)is slightly smaller than the 0.34 Wm^–2 quoted in the TAR, even though the measurements indicate a small 1% rise in the total halocarbon RF since the time of the TAR( Table 2.1 ).

The major solvent, methyl chloroform (CH₃CCl₃), is of special importance regarding RFs, not because of its small RF (see Table 2.1 and Figure 2.6 ), but because this gas is widely used to estimate concentrations of OH, which is the major sink species for CH₄,HFCs, and HCFCs and a major production mechanism for sulphate, nitrate and some organic aerosols as discussed in Section 2.3.5 .The global atmospheric methyl chloroform concentration rose steadily from 1978 to reach a maximum in 1992 ( Prinn et al., 2001 [JoC, SRC] ; Montzka et al., 2003 [NPR, SRC] ). Since then, concentrations have decreased rapidly, driven by a relatively short lifetime of 4.9 years and phase-out under the Montreal Protocol, to levels in 2003 less than 20% of the levels when AGAGE measurements peaked in 1992 ( Prinn et al., 2005a [JoC, SRC] ). Emissions of methyl chloroform determined from industry data ( (McCulloch and Midgley, 2001 ) ) may be too small in recent years. The 2000 to 2003 emissions from Europe estimated using surface observations ( Reimann et al., 2005 [JoC] ) show that 1.2 to 2.3 Gg yr^–1 need to be added over this 4-year period to the above industry estimates for Europe. Estimates of European emissions in 2000 exceeding 20 Gg ( Krol et al., 2003 [JoC] ) are not supported by analyses of the above extensive surface data ( Reimann et al., 2005 [JoC] ). From multi-year measurements, Li et al. 2005 [JoC, MoS] ) estimated 2001 to 2002 emissions from the USA of 2.2 Gg yr^–1 (or about half of those estimated from more temporally but less geographically limited measurements by Millet and Goldstein, 2004 [JoC] ), and suggested that 1996 to 1998 US emissions may be underestimated by an average of about 9.0 Gg yr^–1 over this 3-year period. East Asian emissions deduced from aircraft data in 2001 are about 1.7 Gg above industry data ( Palmer et al., 2003 [JoC] ; see also Yokouchi et al., 2005 ) while recent Australian and Russian emissions are negligible ( Prinn et al., 2001 [JoC, SRC] ; Hurst et al., 2004 [JoC] ).

Carbon tetrachloride (CCl₄)is the second most rapidly decreasing atmospheric chlorocarbon after methyl chloroform. Levels peaked in early 1990 and decreased approximately linearly since then( Figure 2.7 ). Its major use was as a feedstock for CFC manufacturing. Unlike methyl chloroform, a significant inter-hemispheric CCl₄ gradient still exists in 2005 in spite of its moderately long lifetime of 20 to 30 years, resulting from a persistence of significant NH emissions.

HCFCs of industrial importance have lifetimes in the range of 1.3 to 20 years. Global and regional emissions of the CFCs and HCFCs have been derived from observed concentrations and can be used to check emission inventory estimates. Montzka et al. 2003 [NPR, SRC] ) and IPCC/TEAP ( 2005 ) concluded that global emissions of HCFC-22 rose steadily over the period 1975 to 2000, while those of HCFC-141b and HCFC-142b started increasing quickly in the early 1990 s and then began to decrease after 2000 .

To provide a direct comparison of the effects on global warming due to the annual changes in each of the non-CO₂ greenhouse gases (discussed in Sections 2.3.2 , 2.3.3 and 2.3.4 )relative to CO₂, Figure 2.7 shows these annual changes in atmospheric mass multiplied by the GWP (100-year horizon) for each gas (e.g., Prinn, 2004 [NPR, SRC] ). By expressing them in this way, the observed changes in all non-CO₂ gases in GtC equivalents and the significant roles of CH₄,N₂Oand many halocarbons are very evident. This highlights the importance of considering the full suite of greenhouse gases for RF calculations.

2.3.5 Trends in the Hydroxyl Free Radical

The hydroxyl free radical (OH) is the major oxidizing chemical in the atmosphere, destroying about 3.7 Gt of trace gases, including CH₄ and all HFCs and HCFCs, each year ( Ehhalt, 1999 [NPR] ). It therefore has a very significant role in limiting the LLGHG RF. IPCC/TEAP ( 2005 ) concluded that the OH concentration might change in the 21st century by –18 to +5% depending on the emission scenario. The large-scale concentrations and long-term trends in OH can be inferred indirectly using global measurements of trace gases for which emissions are well known and the primary sink is OH. The best trace gas used to date for this purpose is methyl chloroform; long-term measurements of this gas are reviewed in Section 2.3.4 .Other gases that are useful OH indicators include ¹⁴CO, which is produced primarily by cosmic rays ( Lowe and Allan, 2002 [SRC] ). While the accuracy of the ¹⁴CO cosmic ray and other ¹⁴CO source estimates and the frequency and spatial coverage of its measurements do not match those for methyl chloroform, the ¹⁴CO lifetime (2 months) is much shorter than that of methyl chloroform (4.9 years). As a result, ¹⁴CO provides estimates of average concentrations of OH that are more regional, and is capable of resolving shorter time scales than those estimated from methyl chloroform. The ¹⁴CO source variability is better defined than its absolute magnitude so it is better for inferring relative rather than absolute trends. Another useful gas is the industrial chemical HCFC-22. It yields OH concentrations similar to those derived from methyl chloroform, but with less accuracy due to greater uncertainties in emissions and less extensive measurements ( Miller et al., 1998 [JoC] ). The industrial gases HFC-134a, HCFC-141b and HCFC-142b are potentially useful OH estimators, but the accuracy of their emission estimates needs improvement ( Huang and Prinn, 2002 [JoC, MoS, SRC] ; O’Doherty et al., 2004 [JoC] ).

Indirect measurements of OH using methyl chloroform have established that the globally weighted average OH concentration in the troposphere is roughly 10⁶ radicals per cubic centimetre ( Prinn et al., 2001 [JoC, SRC] ; Krol and Lelieveld, 2003 [JoC] ). A similar average concentration is derived using ¹⁴CO ( Quay et al., 2000 [JoC] ), although the spatial weighting here is different. Note that methods to infer global or hemispheric average OH concentrations may be insensitive to compensating regional OH changes such as OH increases over continents and decreases over oceans ( Lelieveld et al., 2002 [JoC, ARC] ). In addition, the quoted absolute OH concentrations (but not their relative trends) depend on the choice of weighting (e.g., ( Lawrence et al., 2001 ) ). While the global average OH concentration appears fairly well defined by these indirect methods, the temporal trends in OH are more difficult to discern since they require long-term measurements, optimal inverse methods and very accurate calibrations, model transports and methyl chloroform emissions data. From AGAGE methyl chloroform measurements, Prinn et al. 2001 [JoC, SRC] ) inferred that global OH levels grew between 1979 and 1989, but then declined between 1989 and 2000, and also exhibited significant interannual variations. They concluded that these decadal global variations were driven principally by NH OH, with SH OH decreasing from 1979 to 1989 and staying essentially constant after that. Using the same AGAGE data and identical methyl chloroform emissions, a three-dimensional model analysis ( Krol and Lelieveld, 2003 [JoC] ) supported qualitatively (but not quantitatively) the earlier result ( Prinn et al., 2001 [JoC, SRC] ) that OH concentrations increased in the 1980 s and declined in the 1990 s. Prinn et al. 2001 [JoC, SRC] ) also estimated the emissions required to provide a zero trend in OH. These required methyl chloroform emissions differed substantially from industry estimates by ( McCulloch and Midgley 2001 ) ) particularly for 1996 to 2000 . However, Krol and Lelieveld 2003 [JoC] ) argued that the combination of possible underestimated recent emissions, especially the >20 Gg European emissions deduced by Krol et al. 2003 [JoC] ), and the recent decreasing effectiveness of the stratosphere as a sink for tropospheric methyl chloroform, may be sufficient to yield a zero deduced OH trend. As discussed in Section 2.3.4 ,estimates of European emissions by Reimann et al. 2005 [JoC] ) are an order of magnitude less than those of Krol et al. 2003 [JoC] ). In addition, Prinn et al. 2005a [JoC, SRC] ) extended the OH estimates through 2004 and showed that the Prinn et al. 2001 [JoC, SRC] ) decadal and interannual OH estimates remain valid even after accounting for the additional recent methyl chloroform emissions discussed in Section 2.3.4 .They also reconfirmed the OH maximum around 1989 and a larger OH minimum around 1998, with OH concentrations then recovering so that in 2003 they were comparable to those in 1979 . They noted that the 1997 to 1999 OH minimum coincides with, and is likely caused by, major global wildfires and an intense El Niño at that time. The 1997 Indonesian fires alone have been estimated to have lowered global late- 1997 OH levels by 6% due to carbon monoxide (CO) enhancements ( Duncan et al., 2003 [JoC] ).

Methyl chloroform is also destroyed in the stratosphere. Because its stratospheric loss frequency is less than that in the troposphere, the stratosphere becomes a less effective sink for tropospheric methyl chloroform over time ( Krol and Lelieveld, 2003 [JoC] ), and even becomes a small source to the troposphere beginning in 1999 in the reference case in the Prinn et al. 2001 [JoC, SRC] , 2005a ) model. Loss to the ocean has usually been considered irreversible, and its rates and uncertainties have been obtained from observations ( Yvon-Lewis and Butler, 2002 [JoC] ). However, Wennberg et al. 2004 [JoC] ) recently proposed that the polar oceans may have effectively stored methyl chloroform during the pre- 1992 years when its atmospheric levels were rising, but began re-emitting it in subsequent years, thus reducing the overall oceanic sink. Prinn et al. 2005a [JoC, SRC] ) tried both approaches and found that their inferred interannual and decadal OH variations were present using either formulation, but inferred OH was lower in the pre- 1992 years and higher after that using the Wennberg et al. 2004 [JoC] ) formulation.

More recently, Bousquet et al. 2005 [ARC] ) used an inverse method with a three-dimensional model and methyl chloroform measurements and concluded that substantial year-to-year variations occurred in global average OH concentrations between 1980 and 2000 . This conclusion was previously reached by Prinn et al. 2001 [JoC, SRC] ), but subsequently challenged by Krol and Lelieveld 2003 [JoC] ) who argued that these variations were caused by model shortcomings and that models need, in particular, to include observationally-based, interannually varying meteorology to provide accurate annual OH estimates. Neither the two-dimensional Prinn et al. 2001 [JoC, SRC] ) nor the three-dimensional Krol et al. 2003 [JoC] ) inversion models used interannually varying circulation. However, the Bousquet et al. 2005 [ARC] ) analysis, which uses observationally based meteorology and estimates OH on monthly time scales, yields interannual OH variations that agree very well with the Prinn et al. 2001 [JoC, SRC] ) and equivalent Krol and Lelieveld 2003 [JoC] ) estimates (see Figure 2.8 ). However, when Bousquet et al. 2005 [ARC] ) estimated both OH concentrations and methyl chloroform emissions (constrained by their uncertainties as reported by ( McCulloch and Midgley, 2001 ) ), the OH variations were reduced by 65% (dashed line in Figure 2.8 ). The error bars on the Prinn et al. 2001 [JoC, SRC] , 2005a ) OH estimates, which account for these emission uncertainties using Monte Carlo ensembles of inversions, also easily allow such a reduction in OH variability (thin vertical bars in Figure 2.8 ). This implies that these interannual OH variations are real, but only their phasing and not their amplitude, is well defined. Bousquet et al. 2005 [ARC] ) also deduced that OH in the SH shows a zero to small negative trend, in qualitative agreement with Prinn et al. 2001 [JoC, SRC] ). Short-term variations in OH were also recently deduced by Manning et al. 2005 [JoC, ARC] ) using 13 years of ¹⁴CO measurements in New Zealand and Antarctica. They found no significant long-term trend between 1989 and 2003 in SH OH but provided evidence for recurring multi-month OH variations of around 10%. They also deduced even larger (20%) OH decreases in 1991 and 1997, perhaps triggered by the 1991 Mt. Pinatubo eruption and the 1997 Indonesian fires. The similarity of many of these results to those from methyl chloroform discussed above is very important, given the independence of the two approaches.

RF calculations of the LLGHGs are calculated from observed trends in the LLGHG concentrations and therefore OH concentrations do not directly affect them. Nevertheless OH trends are needed to quantify LLGHG budgets( Section 7.4 )and for understanding future trends in the LLGHGs and tropospheric ozone.

2.3.6 Ozone

In the TAR, separate estimates for RF due to changes in tropospheric and stratospheric ozone were given. Stratospheric ozone RF was derived from observations of ozone change from roughly 1979 to 1998 . Tropospheric ozone RF was based on chemical model results employing changes in precursor hydrocarbons, CO and nitrogen oxides (NO_x). Over the satellite era (since approximately 1980 ), stratospheric ozone trends have been primarily caused by the Montreal Protocol gases, and in Ramaswamy et al. 2001 [NPR, MoS, SRC] ) the stratospheric ozone RF was implicitly attributed to these gases. Studies since then have investigated a number of possible causes of ozone change in the stratosphere and troposphere and the attribution of ozone trends to a given precursor is less clear. Nevertheless, stratospheric ozone and tropospheric ozone RFs are still treated separately in this report. However, the RFs are more associated with the vertical location of the ozone change than they are with the agent(s) responsible for the change.

2.3.7 Stratospheric Water Vapour

The TAR noted that several studies had indicated long-term increases in stratospheric water vapour and acknowledged that these trends would contribute a significant radiative impact. However, it only considered the stratospheric water vapour increase expected from CH₄ increases as an RF, and this was estimated to contribute 2 to 5% of the total CH₄ RF (about +0.02 Wm^–2).

Section 3.4 discusses the evidence for stratospheric water vapour trends and presents the current understanding of their possible causes. There are now 14 years of global stratospheric water vapour measurements from Halogen Occultation Experiment (HALOE) and continued balloon-based measurements (since 1980 ) at Boulder, Colorado. There is some evidence of a sustained long-term increase in stratospheric water vapour of around 0.05 ppm yr^–1 from 1980 until roughly 2000, since then water vapour concentrations in the lower stratosphere have been decreasing (see Section 3.4 for details and references). As well as CH₄ increases, several other indirect forcing mechanisms have been proposed, including: a) volcanic eruptions (Considi ne et al., 2001; Joshi and Shine, 2003 [JoC] ); b) biomass burning aerosol ( Sherwood, 2002 [JoC, ARC] ); c) tropospheric (SO₂ ;Notholt et al., 2005 [JoC] ) and d) changes in CH₄ oxidation rates from changes in stratospheric chlorine, ozone and OH ( (Rockmann et al., 2004 ) ). These are mechanisms that can be linked to an external forcing agent. Other proposed mechanisms are more associated with climate feedbacks and are related to changes in tropopause temperatures or circulation ( Stuber et al., 2001a [JoC, SRC] ; Fueglistaler et al., 2004 [JoC] ). From these studies, there is little quantification of the stratospheric water vapour change attributable to different causes. It is also likely that different mechanisms are affecting water vapour trends at different altitudes.

Since the TAR, several further calculations of the radiative balance change due to changes in stratospheric water vapour have been performed ( Forster and Shine, 1999 [JoC, SRC] ; Oinas et al., 2001 [JoC] ; Shindell, 2001 [JoC] ; Smith et al., 2001 [JoC, MoS, ARC] ; Forster and Shine, 2002 [JoC, SRC] Smith et al. 2001 [JoC, MoS, ARC] ) estimated a +0.12 to +0.2 Wm^–2 per decade range for the RF from the change in stratospheric water vapour, using HALOE satellite data. Shindell 2001 [JoC] ) estimated an RF of about +0.2 Wm^–2 in a period of two decades, using a GCM to estimate the increase in water vapour in the stratosphere from oxidation of CH₄ and including climate feedback changes associated with an increase in greenhouse gases. Forster and Shine 2002 [JoC, SRC] ) used a constant 0.05 ppm yr^–1 trend in water vapour at pressures of 100 to 10 hPa and estimated the RF to be +0.29 Wm^–2 for 1980 to 2000 . GCM radiation codes can have a factor of two uncertainty in their modelling of this RF ( Oinas et al., 2001 [JoC] ). For the purposes of this chapter, the above RF estimates are not readily attributable to forcing agent(s) and uncertainty as to the causes of the observed change precludes all but the component due to CH₄ increases being considered a forcing. Two related CTM studies have calculated the RF associated with increases in CH₄ since pre-industrial times ( Hansen and Sato, 2001 [JoC, MoS, ARC] ; Hansen et al., 2005 [JoC, MoS, ARC] ), but no dynamical feedbacks were included in those estimates. Hansen et al. 2005 [JoC, MoS, ARC] ) estimated an RF of +0.07 ± 0.01 Wm^–2 for the stratospheric water vapour changes over 1750 to 2000, which is at least a factor of three larger than the TAR value. The RF from direct injection of water vapour by aircraft is believed to be an order of magnitude smaller than this, at about +0.002 Wm^–2 ( IPCC, 1999 [NPR] ). There has been little trend in CH₄ concentration since 2000 (see Section 2.3.2 ); therefore the best estimate of the stratospheric water vapour RF from CH₄ oxidation (+0.07 Wm^–2)is based on the Hansen et al. 2005 [JoC, MoS, ARC] ) calculation. The 90% confidence range is estimated as ±0.05 Wm^–2,from the range of the RF studies that included other effects. There is a low level of scientific understanding in this estimate, as there is only a partial understanding of the vertical profile of CH₄-induced stratospheric water vapour change( Section 2.9 , Table 2.11 ). Other human causes of stratospheric water vapour change are unquantified and have a very low level of scientific understanding.

2.3.8 Observations of Long-Lived Greenhouse Gas Radiative Effects

Observations of the clear-sky radiation emerging at the top of the atmosphere and at the surface have been conducted. Such observations, by their nature, do not measure RF as defined here. Instead, they yield a perspective on the influence of various species on the transfer of radiation in the atmosphere. Most importantly, the conditions involved with these observations involve varying thermal and moisture profiles in the atmosphere such that they do not conform to the conditions underlying the RF definition (see Section 2.2 ). There is a more comprehensive discussion of observations of the Earth’s radiative balance in Section 3.4 .

Harries et al. 2001 [JoC, MoS] ) analysed spectra of the outgoing longwave radiation as measured by two satellites in 1970 and 1997 over the tropical Pacific Ocean. The reduced brightness temperature observed in the spectral regions of many of the greenhouse gases is experimental evidence for an increase in the Earth’s greenhouse effect. In particular, the spectral signatures were large for CO₂ and CH₄.The halocarbons, with their large change between 1970 and 1997, also had an impact on the brightness temperature. Philipona et al. 2004 [JoC, MoS] ) found an increase in the measured longwave downward radiation at the surface over the period from 1995 to 2002 at eight stations over the central Alps. A significant increase in the clear-sky longwave downward flux was found to be due to an enhanced greenhouse effect after combining the measurements with model calculations to estimate the contribution from increases in temperature and humidity. While both types of observations attest to the radiative influences of the gases, they should not be interpreted as having a direct linkage to the value of RFs in Section 2.3 .

2.4.1 Introduction and Summary of the Third Assessment Report

The TAR categorised aerosol RFs into direct and indirect effects. The direct effect is the mechanism by which aerosols scatter and absorb shortwave and longwave radiation, thereby altering the radiative balance of the Earth-atmosphere system. Sulphate, fossil fuel organic carbon, fossil fuel black carbon, biomass burning and mineral dust aerosols were all identified as having a significant anthropogenic component and exerting a significant direct RF. Key parameters for determining the direct RF are the aerosol optical properties (the single scattering albedo, ω_o,specific extinction coefficient,k_e and the scattering phase function), which vary as a function of wavelength and relative humidity, and the atmospheric loading and geographic distribution of the aerosols in the horizontal and vertical, which vary as a function of time (e.g., Haywood and Boucher, 2000 [JoC, MoS, SRC] ; Penner et al., 2001 [NPR, ARC] ; Ramaswamy et al., 2001 [NPR, MoS, SRC] ). Scattering aerosols exert a net negative direct RF, while partially absorbing aerosols may exert a negative top-of-the-atmosphere (TOA) direct RF over dark surfaces such as oceans or dark forest surfaces, and a positive TOA RF over bright surfaces such as desert, snow and ice, or if the aerosol is above cloud (e.g., Chylek and Wong, 1995 [JoC] ; Haywood and Shine, 1995 [JoC, SRC] ). Both positive and negative TOA direct RF mechanisms reduce the shortwave irradiance at the surface. The longwave direct RF is only substantial if the aerosol particles are large and occur in considerable concentrations at higher altitudes (e.g., Tegen et al., 1996 [JoC, MoS, ARC] ). The direct RF due to tropospheric aerosols is most frequently derived at TOA rather than at the tropopause because shortwave radiative transfer calculations have shown a negligible difference between the two (e.g., Haywood and Shine, 1997 [MoS, SRC] ; Section 2.2 ). The surface forcing will be approximately the same as the direct RF at the TOA for scattering aerosols, but for partially absorbing aerosols the surface forcing may be many times stronger than the TOA direct RF (e.g., Ramanathan et al., 2001b [JoC, MoS, SRC] and references therein).

The indirect effect is the mechanism by which aerosols modify the microphysical and hence the radiative properties, amount and lifetime of clouds( Figure 2.10 ). Key parameters for determining the indirect effect are the effectiveness of an aerosol particle to act as a cloud condensation nucleus, which is a function of the size, chemical composition, mixing state and ambient environment (e.g., Penner et al., 2001 [NPR, ARC] ). The microphysically induced effect on the cloud droplet number concentration and hence the cloud droplet size, with the liquid water content held fixed has been called the ‘first indirect effect’ (e.g., Ramaswamy et al., 2001 [NPR, MoS, SRC] ), the ‘cloud albedo effect’ (e.g., Lohmann and Feichter, 2005 [SRC] ), or the ‘Twomey effect’ (e.g., Twomey, 1977 [JoC] ). The microphysically induced effect on the liquid water content, cloud height, and lifetime of clouds has been called the ‘second indirect effect’ (e.g., Ramaswamy et al., 2001 [NPR, MoS, SRC] ), the ‘cloud lifetime effect’ (e.g., Lohmann and Feichter, 2005 [SRC] ) or the ‘Albrecht effect’ (e.g., Albrecht, 1989 [JoC] ). The TAR split the indirect effect into the first indirect effect, and the second indirect effect. Throughout this report, these effects are denoted as ‘cloud albedo effect’ and ‘cloud lifetime effect’, respectively, as these terms are more descriptive of the microphysical processes that occur. The cloud albedo effect was considered in the TAR to be an RF because global model calculations could be performed to describe the influence of increased aerosol concentration on the cloud optical properties while holding the liquid water content of the cloud fixed (i.e., in an entirely diagnostic manner where feedback mechanisms do not occur). The TAR considered the cloud albedo effect to be a key uncertainty in the RF of climate but did not assign a best estimate of the RF, and showed a range of RF between 0 and –2 Wm^–2 in the context of liquid water clouds. The other indirect effects were not considered to be RFs because, in suppressing drizzle, increasing the cloud height or the cloud lifetime in atmospheric models( Figure 2.10 ), the hydrological cycle is invariably altered (i.e., feedbacks occur; see Section 7.5 ). The TAR also discussed the impact of anthropogenic aerosols on the formation and modification of the physical and radiative properties of ice clouds ( Penner et al., 2001 [NPR, ARC] ), although quantification of an RF from this mechanism was not considered appropriate given the host of uncertainties and unknowns surrounding ice cloud nucleation and physics.

The TAR did not include any assessment of the semi-direct effect (e.g., Hansen et al., 1997 [JoC, MoS, ARC] ; Ackerman et al., 2000a [JoC] ; Jacobson, 2002 [JoC] ; Menon et al., 2003 [JoC, MoS, ARC] ; Cook and Highwood, 2004 [JoC, MoS] ; Johnson et al., 2004 [JoC] ), which is the mechanism by which absorption of shortwave radiation by tropospheric aerosols leads to heating of the troposphere that in turn changes the relative humidity and the stability of the troposphere and thereby influences cloud formation and lifetime. In this report, the semi-direct effect is not strictly considered an RF because of modifications to the hydrological cycle, as discussed in Section 7.5 (see also Sections 2.2 , 2.8 and 2.4.5 ).

Since the TAR, there have been substantial developments in observations and modelling of tropospheric aerosols; these are discussed in turn in the following sections.

2.4.2 Developments Related to Aerosol Observations

Surface-based measurements of aerosol properties such as size distribution, chemical composition, scattering and absorption continue to be performed at a number of sites, either at long-term monitoring sites, or specifically as part of intensive field campaigns. These in situ measurements provide essential validation for global models, for example, by constraining aerosol concentrations at the surface and by providing high-quality information about chemical composition and local trends. In addition, they provide key information about variability on various time scales. Comparisons of in situ measurements against those from global atmospheric models are complicated by differences in meteorological conditions and because in situ measurements are representative of conditions mostly at or near the surface while the direct and indirect RFs depend on the aerosol vertical profile. For example, the spatial resolution of global model grid boxes is typically a few degrees of latitude and longitude and the time steps for the atmospheric dynamics and radiation calculations may be minutes to hours depending on the process to be studied; this poses limitations when comparing with observations conducted over smaller spatial extent and shorter time duration.

Combinations of satellite and surface-based observations provide near-global retrievals of aerosol properties. These are discussed in this subsection; the emissions estimates, trends and in situ measurements of the physical and optical properties are discussed with respect to their influence on RF in Section 2.4.4 .Further detailed discussions of the recent satellite observations of aerosol properties and a satellite-measurement based assessment of the aerosol direct RF are given by Yu et al. 2006 [MoS] ).

2.4.3 Advances in Modelling the Aerosol Direct Effect

Since the TAR, more complete aerosol modules in a larger number of global atmospheric models now provide estimates of the direct RF. Several models have resolutions better than 2° by 2° in the horizontal and more than 20 to 30 vertical levels; this represents a considerable enhancement over the models used in the TAR. Such models now include the most important anthropogenic and natural species. Tables 2.4 , 2.5 and 2.6 summarise studies published since the TAR. Some of the more complex models now account explicitly for the dynamics of the aerosol size distribution throughout the aerosol atmospheric lifetime and also parametrize the internal/external mixing of the various aerosol components in a more physically realistic way than in the TAR (e.g., Adams and Seinfeld, 2002 [JoC, MoS] ; Easter et al., 2004 [JoC, MoS] ; Stier et al., 2005 [MoS] ). Because the most important aerosol species are now included, a comparison of key model output parameters, such as the total τ_aer,against satellite retrievals and surface-based sun photometer and lidar observations is possible (see Sections 2.4.2 and 2.4.4 ). Progress with respect to modelling the indirect effects due to aerosol-cloud interactions is detailed in Section 2.4.5 and Section 7.5 .Several studies have explored the sensitivity of aerosol direct RF to current parametrization uncertainties. These are assessed in the following sections.

Major progress since the TAR has been made in the documentation of the diversity of current aerosol model simulations. Sixteen groups have participated in the Global Aerosol Model Intercomparison (AeroCom) initiative ( Kinne et al., 2006 [MoS, SRC] ). Extensive model outputs are available via a dedicated website ( Schulz et al., 2004 [NPR, MoS, SRC] ). Three model experiments (named A, B, and PRE) were analysed. Experiment A models simulate the years 1996, 1997, 2000 and 2001, or a five-year mean encompassing these years. The model emissions and parametrizations are those determined by each research group, but the models are driven by observed meteorological fields to allow detailed comparisons with observations, including those from MODIS, MISR and the AERONET sun photometer network. ExperimentB models use prescribed AeroCom aerosol emissions for the year 2000, and experiment PRE models use prescribed aerosol emissions for the year 1750 ( Dentener et al., 2006 [ARC] ; Schulz et al., 2006 [MoS, SRC] ). The model diagnostics included information on emission and deposition fluxes, vertical distribution and sizes, thus enabling a better understanding of the differences in lifetimes of the various aerosol components in the models.

This paragraph discusses AeroCom results from Textor et al. 2006 [MoS, SRC] ). The model comparison study found a wide range in several of the diagnostic parameters; these, in turn, indicate which aerosol parametrizations are poorly constrained and/or understood. For example, coarse aerosol fractions are responsible for a large range in the natural aerosol emission fluxes (dust: ±49% and sea salt: ±200%, where uncertainty is 1 standard deviation of inter-model range), and consequently in the dry deposition fluxes. The complex dependence of the source strength on wind speed adds to the problem of computing natural aerosol emissions. Dust emissions for the same time period can vary by a factor of two or more depending on details of the dust parametrization ( Luo et al., 2003 [JoC] ; Timmreck and Schulz, 2004 [JoC, MoS, SRC] ; Balkanski et al., 2004 [NPR] ; Zender, 2004 [JoC, MoS] ), and even depend on the reanalysis meteorological data set used ( Luo et al., 2003 [JoC] ). With respect to anthropogenic and natural emissions of other aerosol components, modelling groups tended to make use of similar best guess information, for example, recently revised emissions information available via the Global Emissions Inventory Activity (GEIA). The vertical aerosol distribution was shown to vary considerably, which is a consequence of important differences in removal and vertical mixing parametrizations. The inter-model range for the fraction of sulphate mass below 2.5 km to that of total sulphate is 45 ± 23%. Since humidification takes place mainly in the boundary layer, this source of inter-model variability increases the range of modelled direct RF. Additionally, differences in the parametrization of the wet deposition/vertical mixing process become more pronounced above 5 km altitude. Some models have a tendency to accumulate insoluble aerosol mass (dust and carbonaceous aerosols) at higher altitudes, while others have much more efficient wet removal schemes. Tropospheric residence times, defined here as the ratio of burden over sinks established for an equilibrated one-year simulation, vary by 20 to 30% for the fine-mode aerosol species. These variations are of interest, since they express the linearity of modelled emissions to aerosol burden and eventually to RF.

Considerable progress has been made in the systematic evaluation of global model results (see references in Tables 2.4 to 2.6 ). The simulated global τ_aer at a wavelength of 0.55 µm in models ranges from 0.11 to 0.14. The values compare favourably to those obtained by remote sensing from the ground (AERONET, about 0.135) and space (satellite composite, about 0.15) ( Kinne et al., 2003 [JoC, MoS, SRC] , 2006 ), but significant differences exist in regional and temporal distributions. Modelled absorption optical thickness has been suggested to be underestimated by a factor of two to four when compared to observations ( Sato et al., 2003 [JoC] ) and DRE efficiencies have been shown to be lower in models both for the global average and regionally ( Yu et al., 2006 [MoS] ) (see Section 2.4.4.7 ). A merging of modelled and observed fields of aerosol parameters through assimilation methods of different degrees of complexity has also been performed since the TAR (e.g., Yu et al., 2003 [JoC, MoS] ; Chung et al., 2005 [JoC, MoS, SRC] ). Model results are constrained to obtain present-day aerosol fields consistent with observations. Collins et al. 2001 [JoC, MoS, SRC] ) showed that assimilation of satellite-derived fields of τ_aer can reduce the model bias down to 10% with respect to daily mean τ_aer measured with a sun photometer at the Indian Ocean Experiment (INDOEX) station Kaashidhoo. Liu et al. 2005 [JoC, MoS] ) demonstrated similar efficient reduction of errors in τ_aer.The magnitude of the global dust cycle has been suggested to range between 1,500 and 2,600 Tg yr^–1 by minimising the bias between model and multiple dust observations ( Cakmur et al., 2006 [NPR, JoC, MoS] Bates et al. 2006 [MoS] ) focused on three regions downwind of major urban/population centres and performed radiative transfer calculations constrained by intensive and extensive observational parameters to derive 24-hour average clear-sky DRE of –3.3 ± 0.47, –14 ± 2.6 and –6.4 ± 2.1 Wm^–2 for the north Indian Ocean, the northwest Pacific and the northwest Atlantic, respectively. By constraining aerosol models with these observations, the uncertainty associated with the DRE was reduced by approximately a factor of two.

2.4.4 Estimates of Aerosol Direct Radiative Forcing

Unless otherwise stated, this section discusses the TOA direct RF of different aerosol types as a global annual mean quantity inclusive of the effects of clouds. Where possible, statistics from model results are used to assess the uncertainty in the RF. Recently published results and those grouped within AeroCom are assessed. Because the AeroCom results assessed here are based on prescribed emissions, the uncertainty in these results is lowered by having estimates of the uncertainties in the emissions. The quoted uncertainties therefore include the structural uncertainty (i.e., differences associated with the model formulation and structure) associated with the RF, but do not include the full range of parametric uncertainty (i.e., differences associated with the choice of key model parameters), as the model results are essentially best estimates constrained by observations of emissions, wet and dry deposition, size distributions, optical parameters, hygroscopicity, etc. ( Pan et al., 1997 [JoC, MoS, SRC] ). The uncertainties are reported as the 5 to 95% confidence interval to allow the uncertainty in the RF of each species of aerosol to be quantitatively intercompared.

2.4.5 Aerosol Influence on Clouds (Cloud Albedo Effect)

As pointed out in Section 2.4.1 ,aerosol particles affect the formation and properties of clouds. Only a subset of the aerosol population acts as cloud condensation nuclei (CCN) and/or ice nuclei (IN). Increases in ambient concentrations of CCN and IN due to anthropogenic activities can modify the microphysical properties of clouds, thereby affecting the climate system ( Penner et al., 2001 [NPR, ARC] ; Ramanathan et al., 2001a [JoC, SRC] , Jacob et al., 2005 [NPR, MoS, ARC] ). Several mechanisms are involved, as presented schematically in Figure 2.10 .As noted in Ramaswamy et al. 2001 [NPR, MoS, SRC] ), enhanced aerosol concentrations can lead to an increase in the albedo of clouds under the assumption of fixed liquid water content ( Junge, 1975 [NPR, MoS] ; Twomey, 1977 [JoC] ); this mechanism is referred to in this report as the ‘cloud albedo effect’. The aerosol enhancements have also been hypothesised to lead to an increase in the lifetime of clouds ( Albrecht, 1989 [JoC] ); this mechanism is referred to in this report as the ‘cloud lifetime effect’ and discussed in Section 7.5 .

The interactions between aerosol particles (natural and anthropogenic in origin) and clouds are complex and can be nonlinear ( Ramaswamy et al., 2001 [NPR, MoS, SRC] ). The size and chemical composition of the initial nuclei (e.g., anthropogenic sulphates, nitrates, dust, organic carbon and BC) are important in the activation and early growth of the cloud droplets, particularly the water-soluble fraction and presence of compounds that affect surface tension ( McFiggans et al., 2006 [ARC] and references therein). Cloud optical properties are a function of wavelength. They depend on the characteristics of the droplet size distributions and ice crystal concentrations, and on the morphology of the various cloud types.

The interactions of increased concentrations of anthropogenic particles with shallow (stratocumulus and shallow cumulus) and deep convective clouds (with mixed phase) are discussed in this subsection. This section presents new observations and model estimates of the albedo effect. The associated RF in the context of liquid water clouds is assessed. In-depth discussion of the induced changes that are not considered as RFs (e.g., semi-direct and cloud cover and lifetime effects, thermodynamic response and changes in precipitation development) are presented in Section 7.5 .The impacts of contrails and aviation-induced cirrus are discussed in Section 2.6 and the indirect impacts of aerosol on snow albedo are discussed in Section 2.5.4 .

2.5.1 Introduction

Anthropogenic changes to the physical properties of the land surface can perturb the climate, both by exerting an RF and by modifying other processes such as the fluxes of latent and sensible heat and the transfer of momentum from the atmosphere. In addition to contributing to changes in greenhouse gas concentrations and aerosol loading, anthropogenic changes in the large-scale character of the vegetation covering the landscape (‘land cover’) can affect physical properties such as surface albedo. The albedo of agricultural land can be very different from that of a natural landscape, especially if the latter is forest. The albedo of forested land is generally lower than that of open land because the greater leaf area of a forest canopy and multiple reflections within the canopy result in a higher fraction of incident radiation being absorbed. Changes in surface albedo induce an RF by perturbing the shortwave radiation budget ( Ramaswamy et al., 2001 [NPR, MoS, SRC] ). The effect is particularly accentuated when snow is present, because open land can become entirely snow-covered and hence highly reflective, while trees can remain exposed above the snow ( Betts, 2000 [JoC, SRC] ). Even a snow-covered canopy exhibits a relatively low albedo as a result of multiple reflections within the canopy ( Harding and Pomeroy, 1996 [JoC] ). Surface albedo change may therefore provide the dominant influence of mid- and high-latitude land cover change on climate ( Betts, 2001 [MoS, SRC] ; Bounoua et al., 2002 [JoC] ). The TAR cited two estimates of RF due to the change in albedo resulting from anthropogenic land cover change relative to potential natural vegetation (PNV), –0.4 Wm^–2 and –0.2 Wm^–2,and assumed that the RF relative to 1750 was half of that relative to PNV, so gave a central estimate of the RF due to surface albedo change of –0.2 Wm^–2 ± 0.2 Wm^–2.

Surface albedo can also be modified by the settling of anthropogenic aerosols on the ground, especially in the case of BC on snow ( Hansen and Nazarenko, 2004 [JoC, MoS, ARC] ). This mechanism may be considered an RF mechanism because diagnostic calculations may be performed under the strict definition of RF (see Sections 2.2 and 2.8 ). This mechanism was not discussed in the TAR.

Land cover change can also affect other physical properties such as surface emissivity, the fluxes of moisture through evaporation and transpiration, the ratio of latent to sensible heat fluxes (the Bowen ratio) and the aerodynamic roughness, which exerts frictional drag on the atmosphere and also affects turbulent transfer of heat and moisture. All these processes can affect the air temperature near the ground, and also modify humidity, precipitation and wind speed. Direct human perturbations to the water cycle, such as irrigation, can affect surface moisture fluxes and hence the surface energy balance. Changes in vegetation cover can affect the production of dust, which then exerts an RF. Changes in certain gases, particularly CO₂ and ozone, can also exert an additional influence on climate by affecting the Bowen ratio, through plant responses that affect transpiration. These processes are discussed in detail in Section 7.2 .While such processes will act as anthropogenic perturbations to the climate system ( Pielke et al., 2002 [JoC] ) and will fall at least partly within the ‘forcing’ component of the forcing-feedback-response conceptual model, it is difficult to unequivocally quantify the pure forcing component as distinct from feedbacks and responses. The term ‘non-radiative forcing’ has been proposed ( Jacob et al., 2005 [NPR, MoS, ARC] ) and this report adopts the similar term ‘non-initial radiative effect’, but no quantitative metric separating forcing from feedback and response has yet been implemented for climatic perturbation processes that do not act directly on the radiation budget (see Section 2.2 ).

Energy consumption by human activities, such as heating buildings, powering electrical appliances and fuel combustion by vehicles, can directly release heat into the environment. This was not discussed in the TAR. Anthropogenic heat release is not an RF, in that it does not directly perturb the radiation budget; the mechanisms are not well identified and so it is here referred to as a non-initial radiative effect. It can, however, be quantified as a direct input of energy to the system in terms of Wm^–2.

2.5.2 Changes in Land Cover Since 1750

In 1750, 7.9 to 9.2 million km² (6 to 7% of the global land surface) were under cultivation or pasture( Figure 2.15 ), mainly in Europe, the Indo-Gangetic Plain and China ( Ramankutty and Foley, 1999 [JoC, MoS, SRC] ; Klein Goldewijk, 2001 [JoC, MoS] ). Over the next hundred years, croplands and pasture expanded and intensified in these areas, and new agricultural areas emerged in North America. The period 1850 to 1950 saw a more rapid rate of increase in cropland and pasture areas. In the last 50 years, several regions of the world have seen cropland areas stabilise, and even decrease. In the USA, as cultivation shifted from the east to the Midwest, croplands were abandoned along the eastern seaboard around the turn of the century and the eastern forests have regenerated over the last century. Similarly, cropland areas have decreased in China and Europe. Overall, global cropland and pasture expansion was slower after 1950 than before. However, deforestation is occurring more rapidly in the tropics. Latin America, Africa and South and Southeast Asia experienced slow cropland expansion until the 20th century, but have had exponential increases in the last 50 years. By 1990, croplands and pasture covered 45.7 to 51.3 million km² (35% to 39% of global land), and forest cover had decreased by roughly 11 million km² ( Ramankutty and Foley, 1999 [JoC, MoS, SRC] ; Klein Goldewijk, 2001 [JoC, MoS] ; Table 2.8 ).

Overall, until the mid-20th century most deforestation occurred in the temperate regions( Figure 2.15 ). In more recent decades, however, land abandonment in Western Europe and North America has been leading to reforestation while deforestation is now progressing rapidly in the tropics. In the 1990 s compared to the 1980 s, net removal of tropical forest cover had slowed in the Americas but increased in Africa and Asia.

2.5.3 Radiative Forcing by Anthropogenic Surface Albedo Change: Land Use

Since the TAR, a number of estimates of the RF from land use changes over the industrial era have been made( Table 2.8 ). Unlike the main TAR estimate, most of the more recent studies are ‘pure’ RF calculations with the only change being land cover; feedbacks such as changes in snow cover are excluded. Brovkin et al. 2006 [JoC, MoS, ARC] ) estimated the global mean RF relative to 1700 to be –0.15 Wm^–2,considering only cropland changes ( Ramankutty and Foley, 1999 [JoC, MoS, SRC] ) and not pastures. Hansen et al. 2005 [JoC, MoS, ARC] ) also considered only cropland changes ( Ramankutty and Foley, 1999 [JoC, MoS, SRC] ) and simulated the RF relative to 1750 to be –0.15 Wm^–2.Using historical reconstructions of both croplands ( Ramankutty and Foley, 1999 [JoC, MoS, SRC] ) and pasturelands ( Klein Goldewijk, 2001 [JoC, MoS] Betts et al. 2007 [MoS, SRC, 2007] ) simulated an RF of –0.18 Wm^–2 since 1750 . This study also estimated the RF relative to PNV to be –0.24 Wm^–2.Other studies since the TAR have also estimated the RF at the present day relative to PNV( Table 2.8 Govindasamy et al. 2001a [JoC] ) estimated this RF as –0.08 Wm^–2 .Myhre et al. 2005a [JoC, MoS, SRC] ) used land cover and albedo data from MODIS ( (Friedl et al., 2002; ) ( Schaaf et al., 2002 ) ) and estimated this RF as –0.09 Wm^–2.The results of Betts et al. 2007 [MoS, SRC, 2007] and Brovkin et al. 2006 [JoC, MoS, ARC] ) suggest that the RF relative to 1750 is approximately 75% of that relative to PNV. Therefore, by employing this factor published RFs relative to PNV can be used to estimate the RF relative to 1750 ( Table 2.8 ).

In all the published studies, the RF showed a very high degree of spatial variability, with some areas showing no RF in 1990 relative to 1750 while values more negative than –5 Wm^–2 were typically found in the major agricultural areas of North America and Eurasia. The local RF depends on local albedo changes, which depend on the nature of the PNV replaced by agriculture (see top panel of Figure 2.15 ). In historical simulations, the spatial patterns of RF relative to the PNV remain generally similar over time, with the regional RFs in 1750 intensifying and expanding in the area covered. The major new areas of land cover change since 1750 are North America and central and eastern Russia.

Changes in the underlying surface albedo could affect the RF due to aerosols if such changes took place in the same regions. Similarly, surface albedo RF may depend on aerosol concentrations. Estimates of the temporal evolution of aerosol RF and surface albedo RF may need to consider changes in each other ( Betts et al., 2007 [MoS, SRC, 2007] ).

Table 2.8 .Estimates of forest area, contribution to CO₂ increase from anthropogenic land cover change, and RF due to the land use change-induced CO₂ increase and surface albedo change, relative to pre-industrial vegetation and PNV. The CO₂ RFs are for 2000 relative to 1850, calculated from the land use change contribution to the total increase in CO₂ from 1850 to 2000 simulated with both land use and fossil fuel emissions by the carbon cycle models. Carbon emissions from land cover change for the 1980 s and 1990 s are discussed in Section 7.3 and Table 7.2 .

Notes: ^a The available literature simulates CO₂ rises with and without land use relative to 1850 . ^b 1750 forest area reported as 51.85 x 106 km². ^c 1992 forest area. ^d Land use contribution CO₂ rise from Brovkin et al. 2004 [ARC] ). ^e Albedo RF from Betts et al. 2007 [MoS, SRC, 2007] ). Land cover data combined from Ramankutty and Foley 1999 [JoC, MoS, SRC] ), Klein Goldelwijk ( 2001 and Wilson and Henderson-Sellers 1985 [MoS, ARC] ). ^f Albedo RF from Myhre and Myhre 2003 [JoC, MoS, SRC] ). Range of estimate for each land cover data set arises from use of different albedo values. ^g Albedo RF from model of Goosse et al. 2005 [PoC, JoC, MoS, ARC] ) in Brovkin et al. 2006 [JoC, MoS, ARC] ). ^h RF relative to 1750 estimated here as 0.75 of RF relative to PNV following Betts et al. 2007 [MoS, SRC, 2007] and Brovkin et al. 2006 [JoC, MoS, ARC] ). ⁱ Estimate relative to 1700 . ^j Albedo RF from Matthews et al. 2003 [JoC, MoS, ARC] ). ^k Albedo RF from Hansen et al. 2005 [JoC, MoS, ARC] ). lAlbedo RF from Matthews et al. 2004 [JoC, ARC] ). mForest areas aggregated by Richards 1990 [NPR] ). n 1980 forest area. oLand use contribution to CO2 rise from Matthews et al. 2004 [JoC, ARC] ). Estimate only available relative to 1850 not 1750 . pAlbedo RF from Myhre et al. 2005a [JoC, MoS, SRC] ). qAlbedo RF from Betts 2001 [MoS, SRC] ). rSurface and Atmosphere Radiation Budget; http://www-surf.larc.nasa.gov/surf/. sAlbedo RF from Hansen et al. 1997 [JoC, MoS, ARC] ). tAlbedo RF from Govindasamy et al. 2001a [JoC] ).

2.5.4 Radiative Forcing by Anthropogenic Surface Albedo Change: Black Carbon in Snow and Ice

The presence of soot particles in snow could cause a decrease in the albedo of snow and affect snowmelt. Initial estimates by Hansen et al. 2000 [JoC, ARC] ) suggested that BC could thereby exert a positive RF of +0.2 Wm^–2.This estimate was refined by Hansen and Nazarenko 2004 [JoC, MoS, ARC] ), who used measured BC concentrations within snow and ice at a wide range of geographic locations to deduce the perturbation to the surface and planetary albedo, deriving an RF of +0.15 Wm^–2.The uncertainty in this estimate is substantial due to uncertainties in whether BC and snow particles are internally or externally mixed, in BC and snow particle shapes and sizes, in voids within BC particles, and in the BC imaginary refractive index. Jacobson 2004 [JoC] ) developed a global model that allows the BC aerosol to enter snow via precipitation and dry deposition, thereby modifying the snow albedo and emissivity. They found modelled concentrations of BC within snow that were in reasonable agreement with those from many observations. The model study found that BC on snow and sea ice caused a decrease in the surface albedo of 0.4% globally and 1% in the NH, although RFs were not reported. Hansen et al. 2005 [JoC, MoS, ARC] ) allowed the albedo change to be proportional to local BC deposition according to Koch 2001 [JoC, MoS] ) and presented a further revised estimate of 0.08 Wm^–2.They also suggested that this RF mechanism produces a greater temperature response by a factor of 1.7 than an equivalent CO₂ RF, that is, the ‘efficacy’ may be higher for this RF mechanism (see Section 2.8.5.7 ). This report adopts a best estimate for the BC on snow RF of +0.10 ± 0.10 Wm^–2,with a low level of scientific understanding( Section 2.9 , Table 2.11 ).

Since the dominant aspect of land cover change since 1750 has been deforestation in temperate regions, the overall effect of anthropogenic land cover change on global temperature will depend largely on the relative importance of increased surface albedo in winter and spring (exerting a cooling) and reduced evaporation in summer and in the tropics (exerting a warming) ( Bounoua et al., 2002 [JoC] ). Estimates of global temperature responses from past deforestation vary from 0.01°C ( Zhao et al., 2001 [JoC, ARC] ) to –0.25°C ( Govindasamy et al., 2001a [JoC] ; Brovkin et al., 2006 [JoC, MoS, ARC] ). If cooling by increased surface albedo dominates, then the historical effect of land cover change may still be adequately represented by RF. With tropical deforestation becoming more significant in recent decades, warming due to reduced evaporation may become more significant globally than increased surface albedo. Radiative forcing would then be less useful as a metric of climate change induced by land cover change recently and in the future.

2.5.5 Other Effects of Anthropogenic Changes in Land Cover

Anthropogenic land use and land cover change can also modify climate through other mechanisms, some directly perturbing the Earth radiation budget and some perturbing other processes. Impacts of land cover change on emissions of CO₂,CH₄,biomass burning aerosols and dust aerosols are discussed in Sections 2.3 and 2.4 .Land cover change itself can also modify the surface energy and moisture budgets through changes in evaporation and the fluxes of latent and sensible heat, directly affecting precipitation and atmospheric circulation as well as temperature. Model results suggest that the combined effects of past tropical deforestation may have exerted regional warmings of approximately 0.2°C relative to PNV, and may have perturbed the global atmospheric circulation affecting regional climates remote from the land cover change ( Chase et al., 2000 [JoC, MoS] ; Zhao et al., 2001 [JoC, ARC] ; Pielke et al., 2002 [JoC] ; Chapters 7 , 9 and 11 ).

2.5.6 Tropospheric Water Vapour from Anthropogenic Sources

Anthropogenic use of water is less than 1% of natural sources of water vapour and about 70% of the use of water for human activity is from irrigation ( Döll, 2002 [JoC] ). Several regional studies have indicated an impact of irrigation on temperature, humidity and precipitation ( Barnston and Schickedanz, 1984 [JoC] ; Lohar and Pal, 1995 [JoC] ; de Ridder and Gallée, 1998 [JoC] ; Moore and Rojstaczer, 2001 [JoC] ; Zhang et al., 2002 [ARC] Boucher et al. 2004 [JoC, SRC] ) used a GCM to show that irrigation has a global impact on temperature and humidity. Over Asia where most of the irrigation takes place, the simulations showed a change in the water vapour content in the lower troposphere of up to 1%, resulting in an RF of +0.03 Wm^–2.However, the effect of irrigation on surface temperature was dominated by evaporative cooling rather than by the excess greenhouse effect and thus a decrease in surface temperature was found. Irrigation affects the temperature, humidity, clouds and precipitation as well as the natural evaporation through changes in the surface temperature, raising questions about the strict use of RF in this case. Uncertainties in the water vapour flow to the atmosphere from irrigation are significant and Gordon et al. 2005 [JoC] ) gave a substantially higher estimate compared to that of Boucher et al. 2004 [JoC, SRC] ). Most of this uncertainty is likely to be linked to differences between the total withdrawal for irrigation and the amount actually used ( Boucher et al., 2004 [JoC, SRC] ). Furthermore, Gordon et al. 2005 [JoC] ) also estimated a reduced water vapour flow to the atmosphere from deforestation, most importantly in tropical areas. This reduced water vapour flow is a factor of three larger than the water vapour increase due to irrigation in Boucher et al. 2004 [JoC, SRC] ), but so far there are no estimates of the effect of this on the water vapour content of the atmosphere and its RF. Water vapour changes from deforestation will, like irrigation, affect the surface evaporation and temperature and the water cycle in the atmosphere. Radiative forcing from anthropogenic sources of tropospheric water vapour is not evaluated here, since these sources affect surface temperature more significantly through these non-radiative processes, and a strict use of the RF is problematic. The emission of water vapour from fossil fuel combustion is significantly lower than the emission from changes in land use ( Boucher et al., 2004 [JoC, SRC] ).

Urban heat islands result partly from the physical properties of the urban landscape and partly from the release of heat into the environment by the use of energy for human activities such as heating buildings and powering appliances and vehicles (‘human energy production’). The global total heat flux from this is estimated as 0.03 Wm^–2 ( Nakicenovic, 1998 [NPR] ). If this energy release were concentrated in cities, which are estimated to cover 0.046% of the Earth’s surface ( (Loveland et al., 2000 ) ) the mean local heat flux in a city would be 65 Wm^–2.Daytime values in central Tokyo typically exceed 400 Wm^–2 with a maximum of 1,590 Wm^–2 in winter ( (Ichinose et al., 1999 ) ). Although human energy production is a small influence at the global scale, it may be very important for climate changes in cities ( Betts and Best, 2004 [NPR, MoS, SRC] ;( Crutzen, 2004 ) ).

2.5.7 Anthropogenic Heat Release

2.5.8 Effects of Carbon Dioxide Changes on Climate via Plant Physiology: ‘Physiological Forcing’

As well as exerting an RF on the climate system, increasing concentrations of atmospheric CO₂ can perturb the climate system through direct effects on plant physiology. Plant stomatal apertures open less under higher CO₂ concentrations ( (Field et al., 1995 ) ), which directly reduces the flux of moisture from the surface to the atmosphere through transpiration ( Sellers et al., 1996 [JoC] ). A decrease in moisture flux modifies the surface energy balance, increasing the ratio of sensible heat flux to latent heat flux and therefore warming the air near the surface ( Sellers et al., 1996 [JoC] ; Betts et al., 1997 [JoC, MoS, SRC] ; Cox et al., 1999 [JoC, MoS, ARC] Betts et al. 2004 [MoS, SRC] ) proposed the term ‘physiological forcing’ for this mechanism. Although no studies have yet explicitly quantified the present-day temperature response to physiological forcing, the presence of this forcing has been detected in global hydrological budgets ( Gedney et al., 2006 [JoC] ; Section 9.5 ). This process can be considered a non-initial radiative effect, as distinct from a feedback, since the mechanism involves a direct response to increasing atmospheric CO₂ and not a response to climate change. It is not possible to quantify this with RF. Reduced global transpiration would also be expected to reduce atmospheric water vapour causing a negative forcing, but no estimates of this have been made.

Increased CO₂ concentrations can also ‘fertilize’ plants by stimulating photosynthesis, which models suggest has contributed to increased vegetation cover and leaf area over the 20th century ( Cramer et al., 2001 [MoS] ). Increases in the Normalized Difference Vegetation Index, a remote sensing product indicative of leaf area, biomass and potential photosynthesis, have been observed ( Zhou et al., 2001 [JoC, ARC] ), although other causes including climate change itself are also likely to have contributed. Increased vegetation cover and leaf area would decrease surface albedo, which would act to oppose the increase in albedo due to deforestation. The RF due to this process has not been evaluated and there isa very low scientific understanding of these effects.

2.6.1 Introduction

The IPCC separately evaluated the RF from subsonic and supersonic aircraft operations in the Special Report on Aviation and the Global Atmosphere ( IPCC, 1999 [NPR] ), hereinafter designated as IPCC- 1999 . Like many other sectors, subsonic aircraft operations around the globe contribute directly and indirectly to the RF of climate change. This section only assesses the aspects that are unique to the aviation sector, namely the formation of persistent condensation trails (contrails), their impact on cirrus cloudiness, and the effects of aviation aerosols. Persistent contrail formation and induced cloudiness are indirect effects from aircraft operations because they depend on variable humidity and temperature conditions along aircraft flight tracks. Thus, future changes in atmospheric humidity and temperature distributions in the upper troposphere will have consequences for aviation-induced cloudiness. Also noted here is the potential role of aviation aerosols in altering the properties of clouds that form later in air containing aircraft emissions.

2.6.2 Radiative Forcing Estimates for Persistent Line-Shaped Contrails

Aircraft produce persistent contrails in the upper troposphere in ice-supersaturated air masses ( IPCC, 1999 [NPR] ). Contrails are thin cirrus clouds, which reflect solar radiation and trap outgoing longwave radiation. The latter effect is expected to dominate for thin cirrus ( Hartmann et al., 1992 [JoC] ; Meerkötter et al., 1999 [MoS] ), thereby resulting in a net positive RF value for contrails. Persistent contrail cover has been calculated globally from meteorological data (e.g., Sausen et al., 1998 [SRC] ) or by using a modified cirrus cloud parametrization in a GCM ( Ponater et al., 2002 [JoC, MoS, SRC] ). Contrail cover calculations are uncertain because the extent of supersaturated regions in the atmosphere is poorly known. The associated contrail RF follows from determining an optical depth for the computed contrail cover. The global RF values for contrail and induced cloudiness are assumed to vary linearly with distances flown by the global fleet if flight ambient conditions remain unchanged. The current best estimate for the RF of persistent linear contrails for aircraft operations in 2000 is +0.010 Wm^–2 ( Table 2.9 ;Sausen et al., 2005 [MoS, SRC] ). The value is based on independent estimates derived from Myhre and Stordal 2001b [JoC, SRC] and Marquart et al. 2003 [JoC, MoS, SRC] ) that were updated for increased aircraft traffic in Sausen et al. 2005 [MoS, SRC] ) to give RF estimates of +0.015 Wm^–2 and +0.006 Wm^–2,respectively. The uncertainty range is conservatively estimated to be a factor of three. The +0.010 Wm^–2 value is also considered to be the best estimate for 2005 because of the slow overall growth in aviation fuel use in the 2000 to 2005 period. The decrease in the best estimate from the TAR by a factor of two results from reassessments of persistent contrail cover and lower optical depth estimates ( Marquart and Mayer, 2002 [JoC, MoS] ; Meyer et al., 2002 [JoC, MoS] ; Ponater et al., 2002 [JoC, MoS, SRC] ; Marquart et al., 2003 [JoC, MoS, SRC] ). The new estimates include diurnal changes in the solar RF, which decreases the net RF for a given contrail cover by about 20% ( Myhre and Stordal, 2001b [JoC, SRC] ). The level of scientific understanding of contrail RF is considered low, since important uncertainties remain in the determination of global values( Section 2.9 , Table 2.11 ). For example, unexplained regional differences are found in contrail optical depths between Europe and the USA that have not been fully accounted for in model calculations ( Meyer et al., 2002 [JoC, MoS] ; Ponater et al., 2002 [JoC, MoS, SRC] ;( Palikonda et al., 2005 ) ).

Table 2.9 .Radiative forcing terms for contrail and cirrus effects caused by global subsonic aircraft operations.

Notes: ^a Values for contrails are best estimates. Values in parentheses give the uncertainty range. ^b Values from IPCC- 1999 ( IPCC, 1999 [NPR] ). ^c Values interpolated from 1992 and 2015 estimates in IPCC- 1999 ( Sausen et al., 2005 [MoS, SRC] ). ^d Sausen et al. 2005 [MoS, SRC] ). Values are considered valid (within 10%) for 2005 because of slow growth in aviation fuel use between 2000 and 2005 .

2.6.3 Radiative Forcing Estimates for Aviation- Induced Cloudiness

Individual persistent contrails are routinely observed to shear and spread, covering large additional areas with cirrus cloud ( Minnis et al., 1998 [JoC] ). Aviation aerosol could also lead to changes in cirrus cloud (see Section 2.6.4 ). Aviation-induced cloudiness (AIC) is defined to be the sumx of all changes in cloudiness associated with aviation operations. Thus, an AIC estimate includes persistent contrail cover. Because spreading contrails lose their characteristic linear shape, a component of AIC is indistinguishable from background cirrus. This basic ambiguity, which prevented the formulation of a best estimate of AIC amounts and the associated RF in IPCC- 1999, still exists for this assessment. Estimates of the ratio of induced cloudiness cover to that of persistent linear contrails range from 1.8 to 10 ( Minnis et al., 2004 [JoC] ;( Mannstein and Schumann, 2005 ) ^[10] ), indicating the uncertainty in estimating AIC amounts. Initial attempts to quantify AIC used trend differences in cirrus cloudiness between regions of high and low aviation fuel consumption ( Boucher, 1999 [JoC, SRC] ). Since IPCC- 1999, two studies have also found significant positive trends in cirrus cloudiness in some regions of high air traffic and found lower to negative trends outside air traffic regions ( (Zerefos et al., 2003; ) ( Stordal et al., 2005 ) ). Using the International Satellite Cloud Climatology Project (ISCCP) database, these studies derived cirrus cover trends for Europe of 1 to 2% per decade over the last one to two decades. A study with the Television Infrared Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) provides further support for these trends ( Stubenrauch and Schumann, 2005 [JoC] ). However, cirrus trends that occurred due to natural variability, climate change or other anthropogenic effects could not be accounted for in these studies. Cirrus trends over the USA (but not over Europe) were found to be consistent with changes in contrail cover and frequency ( Minnis et al., 2004 [JoC] ). Thus, significant uncertainty remains in attributing observed cirrus trends to aviation.

Regional cirrus trends were used as a basis to compute a global mean RF value for AIC in 2000 of +0.030 Wm^–2 with a range of +0.01 to +0.08 Wm^–2( (Stordal et al., 2005 ) ). This value is not considered a best estimate because of the uncertainty in the optical properties of AIC and in the assumptions used to derive AIC cover. However, this value is in good agreement with the upper limit estimate for AIC RF in 1992 of +0.026 Wm^–2 derived from surface and satellite cloudiness observations ( Minnis et al., 2004 [JoC] ). A value of +0.03 Wm^–2 is close to the upper-limit estimate of +0.04 Wm^–2 derived for non-contrail cloudiness in IPCC- 1999 . Without an AIC best estimate, the best estimate of the total RF value for aviation-induced cloudiness( Section 2.9.2 ,Table 2.12 and Figure 2.20 )includes only that due to persistent linear contrails. Radiative forcing estimates for AIC made using cirrus trend data necessarily cannot distinguish between the components of aviation cloudiness, namely persistent linear contrails, spreading contrails and other aviation aerosol effects. Some aviation effects might be more appropriately considered feedback processes rather than an RF (see Sections 2.2 and 2.4.5 ). However, the low understanding of the processes involved and the lack of quantitative approaches preclude reliably making the forcing/feedback distinction for all aviation effects in this assessment.

Two issues related to the climate response of aviation cloudiness are worth noting here. First, Minnis et al. 2004 [JoC] , 2005 ) used their RF estimate for total AIC over the USA in an empirical model, and concluded that the surface temperature response for the period 1973 to 1994 could be as large as the observed surface warming over the USA (around 0.3°C per decade). In response to the Minnis et al. conclusion, contrail RF was examined in two global climate modelling studies ( Hansen et al., 2005 [JoC, MoS, ARC] ; Ponater et al., 2005 [JoC, SRC] ). Both studies concluded that the surface temperature response calculated by Minnis et al. 2004 [JoC] ) is too large by one to two orders of magnitude. For the Minnis et al. result to be correct, the climate efficacy or climate sensitivity of contrail RF would need to be much greater than that of other larger RF terms, (e.g., CO₂). Instead, contrail RF is found to have a smaller efficacy than an equivalent CO₂ RF ( Hansen et al., 2005 [JoC, MoS, ARC] ; Ponater et al., 2005 [JoC, SRC] ) (see Section 2.8.5.7 ), which is consistent with the general ineffectiveness of high clouds in influencing diurnal surface temperatures ( Hansen et al., 1995 [ARC] , 2005 ). Several substantive explanations for the incorrectness of the enhanced response found in the Minnis et al. study have been presented ( Hansen et al., 2005 [JoC, MoS, ARC] ; Ponater et al., 2005 [JoC, SRC] ; Shine, 2005 [JoC] ).

The second issue is that the absence of AIC has been proposed as the cause of the increased diurnal temperature range (DTR) found in surface observations made during the short period when all USA air traffic was grounded starting on 11 September 2001 ( Travis et al., 2002 [JoC] , 2004 ). The Travis et al. studies show that during this period: (i) DTR was enhanced across the conterminous USA, with increases in the maximum temperatures that were not matched by increases of similar magnitude in the minimum temperatures, and (ii) the largest DTR changes corresponded to regions with the greatest contrail cover. The Travis et al. conclusions are weak because they are based on a correlation rather than a quantitative model and rely (necessarily) on very limited data ( (Schumann, 2005 ) ). Unusually clear weather across the USA during the shutdown period also has been proposed to account for the observed DTR changes ( (Kalkstein and Balling, 2004 ) ). Thus, more evidence and a quantitative physical model are needed before the validity of the proposed relationship between regional contrail cover and DTR can be considered further.

2.6.4 Aviation Aerosols

Global aviation operations emit aerosols and aerosol precursors into the upper troposphere and lower stratosphere ( IPCC, 1999 [NPR] ; Hendricks et al., 2004 [MoS] ). As a result, aerosol number and/or mass are enhanced above background values in these regions. Aviation-induced cloudiness includes the possible influence of aviation aerosol on cirrus cloudiness amounts. The most important aerosols are those composed of sulphate and BC (soot). Sulphate aerosols arise from the emissions of fuel sulphur and BC aerosol results from incomplete combustion of aviation fuel. Aviation operations cause enhancements of sulphate and BC in the background atmosphere ( IPCC, 1999 [NPR] ; Hendricks et al., 2004 [MoS] ). An important concern is that aviation aerosol can act as nuclei in ice cloud formation, thereby altering the microphysical properties of clouds ( Jensen and Toon, 1997 [JoC] ; Kärcher, 1999 [ARC] ; Lohmann et al., 2004 [JoC, SRC] ) and perhaps cloud cover. A modelling study by Hendricks et al. 2005 [JoC, SRC] ) showed the potential for significant cirrus modifications by aviation caused by increased numbers of BC particles. The modifications would occur in flight corridors as well as in regions far away from flight corridors because of aerosol transport. In the study, aviation aerosols either increase or decrease ice nuclei in background cirrus clouds, depending on assumptions about the cloud formation process. Results from a cloud chamber experiment showed that a sulphate coating on soot particles reduced their effectiveness as ice nuclei ( Möhler et al., 2005 [JoC] ). Changes in ice nuclei number or nucleation properties of aerosols can alter the radiative properties of cirrus clouds and, hence, their radiative impact on the climate system, similar to the aerosol-cloud interactions discussed in Sections 2.4.1 , 2.4.5 and 7.5 .No estimates are yet available for the global or regional RF changes caused by the effect of aviation aerosol on background cloudiness, although some of the RF from AIC, determined by correlation studies (see Section 2.6.3 ), may be associated with these aerosol effects.

2.7.1 Solar Variability

The estimates of long-term solar irradiance changes used in the TAR (e.g., Hoyt and Schatten, 1993 [JoC] ; Lean et al., 1995 [PoC, JoC, SRC] ) have been revised downwards, based on new studies indicating that bright solar faculae likely contributed a smaller irradiance increase since the Maunder Minimum than was originally suggested by the range of brightness in Sun-like stars ( (Hall and Lockwood, 2004; ) M. Wang et al., 2005 [Ambiguous] ). However, empirical results since the TAR have strengthened the evidence for solar forcing of climate change by identifying detectable tropospheric changes associated with solar variability, including during the solar cycle( Section 9.2 ;van Loon and Shea, 2000 [JoC, ARC] ; Douglass and Clader, 2002 [JoC] ; Gleisner and Thejll, 2003 [JoC] ; Haigh, 2003 [JoC, ARC] ; Stott et al., 2003 [JoC, MoS] ; White et al., 2003 [JoC, ARC] ; Coughlin and Tung, 2004 [JoC] ;( Labitzke, 2004; ) Crooks and Gray, 2005 [JoC, ARC] ). The most likely mechanism is considered to be some combination of direct forcing by changes in total solar irradiance, and indirect effects of ultraviolet (UV) radiation on the stratosphere. Least certain, and under ongoing debate as discussed in the TAR, are indirect effects induced by galactic cosmic rays (e.g., ( Marsh and Svensmark, 2000a, ) b; Kristjánsson et al., 2002 [NPR, JoC] ; Sun and Bradley, 2002 [PoC, JoC] ).

2.7.2 Explosive Volcanic Activity

2.8.1 Vertical Forcing Patterns and Surface Energy Balance Changes

The vertical structure of a forcing agent is important both for efficacy (see Section 2.8.5 )and for other aspects of climate response, particularly for evaluating regional and vertical patterns of temperature change and also changes in the hydrological cycle. For example, for absorbing aerosol, the surface forcings are arguably a more useful measure of the climate response (particularly for the hydrological cycle) than the RF ( Ramanathan et al., 2001a [JoC, SRC] ; Menon et al., 2002b [JoC, ARC] ). It should be noted that a perturbation to the surface energy budget involves sensible and latent heat fluxes besides solar and longwave irradiance; therefore, it can quantitatively be very different from the RF, which is calculated at the tropopause, and thus is not representative of the energy balance perturbation to the surface-troposphere (climate) system. While the surface forcing adds to the overall description of the total perturbation brought about by an agent, the RF and surface forcing should not be directly compared nor should the surface forcing be considered in isolation for evaluating the climate response (see, e.g., the caveats expressed in Manabe and Wetherald, 1967 [JoC] ; Ramanathan, 1981 [JoC, SRC] ). Therefore, surface forcings are presented as an important and useful diagnostic tool that aids understanding of the climate response (see Sections 2.9.4 and 2.9.5 ).

2.8.2 Spatial Patterns of Radiative Forcing

Each RF agent has a unique spatial pattern (see, e.g., Figure 6.7 in Ramaswamy et al., 2001 [NPR, MoS, SRC] ). When combining RF agents it is not just the global mean RF that needs to be considered. For example, even with a net global mean RF of zero, significant regional RFs can be present and these can affect the global mean temperature response (see Section 2.8.5 ). Spatial patterns of RF also affect the pattern of climate response. However, note that, to first order, very different RF patterns can have similar patterns of surface temperature response and the location of maximum RF is rarely coincident with the location of maximum response ( Boer and Yu, 2003b [JoC] ). Identification of different patterns of response is particularly important for attributing past climate change to particular mechanisms, and is also important for the prediction of regional patterns of future climate change. This chapter employs RF as the method for ranking the effect of a forcing agent on the equilibrium global temperature change, and only this aspect of the forcing-response relationship is discussed. However, patterns of RF are presented as a diagnostic in Section 2.9.5 .

RFs are increasingly being diagnosed from GCM integrations where the calculations are complex ( Stuber et al., 2001b [MoS, SRC] ; Tett et al., 2002 [JoC, MoS, ARC] ; Gregory et al., 2004 [JoC, MoS, ARC] ). This chapter also discusses several mechanisms that include some response in the troposphere, such as cloud changes. These mechanisms are not initially radiative in nature, but will eventually lead to a radiative perturbation of the surface-troposphere system that could conceivably be measured at the TOA. Jacob et al. 2005 [NPR, MoS, ARC] ) refer to these mechanisms as non-radiative forcings (see also Section 2.2 ). Alternatives to the standard stratospherically adjusted RF definition have been proposed that may help account for these processes. Since the TAR, several studies have employed GCMs to diagnose the zero-surface-temperature-change RF (see Figure 2.2 and Section 2.2 ). These studies have used a number of different methodologies. Shine et al. 2003 [JoC, MoS] ) fixed both land and sea surface temperatures globally and calculated a radiative energy imbalance: this technique is only feasible in GCMs with relatively simple land surface parametrizations. Hansen et al. 2005 [JoC, MoS, ARC] ) fixed sea surface temperatures and calculated an RF by adding an extra term to the radiative imbalance that took into account how much the land surface temperatures had responded. Sokolov 2006 [JoC, MoS, ARC] ) diagnosed the zero-surface-temperature-change RF by computing surface-only and atmospheric-only components of climate feedback separately in a slab model and then modifying the stratospherically adjusted RF by the atmospheric-only feedback component. Gregory et al. 2004 [JoC, MoS, ARC] ; see also Hansen et al., 2005 [JoC, MoS, ARC] ; Forster and Taylor, 2006 [JoC, MoS, SRC] ) used a regression method with a globally averaged temperature change ordinate to diagnose the zero-surface-temperature-change RF: this method had the largest uncertainties. Shine et al. 2003 [JoC, MoS] Hansen et al. 2005 [JoC, MoS, ARC] and Sokolov 2006 [JoC, MoS, ARC] ) all found that that the fixed-surface-temperature RF was a better predictor of the equilibrium global mean surface temperature response than the stratospherically adjusted RF. Further, it was a particularly useful diagnostic for changes in absorbing aerosol where the stratospherically adjusted RF could fail as a predictor of the surface temperature response (see Section 2.8.5.5 ). Differences between the zero-surface-temperature-change RF and the stratospherically adjusted RF can be caused by semi-direct and cloud-aerosol interaction effects beyond the cloud albedo RF. For most mechanisms, aside from the case of certain aerosol changes, the difference is likely to be small ( Shine et al., 2003 [JoC, MoS] ; Hansen et al., 2005 [JoC, MoS, ARC] ; Sokolov, 2006 [JoC, MoS, ARC] ). These calculations also remove problems associated with defining the tropopause in the stratospherically adjusted RF definition ( Shine et al., 2003 [JoC, MoS] ; Hansen et al., 2005 [JoC, MoS, ARC] ). However, stratospherically adjusted RF has the advantage that it does not depend on relatively uncertain components of a GCM’s response, such as cloud changes. For the LLGHGs, the stratospherically adjusted RF also has the advantage that it is readily calculated in detailed off-line radiation codes. For these reasons, the stratospherically adjusted RF is retained as the measure of comparison used in this chapter (see Section 2.2 ). However, to first order, all methods are comparable and all prove useful for understanding climate response.

2.8.3 Alternative Methods of Calculating Radiative Forcing

2.8.4 Linearity of the Forcing-Response Relationship

Reporting findings from several studies, the TAR concluded that responses to individual RFs could be linearly added to gauge the global mean response, but not necessarily the regional response ( Ramaswamy et al., 2001 [NPR, MoS, SRC] ). Since then, studies with several equilibrium and/or transient integrations of several different GCMs have found no evidence of any nonlinearity for changes in greenhouse gases and sulphate aerosol ( Boer and Yu, 2003b [JoC] ; Gillett et al., 2004 [JoC, MoS, ARC] ; Matthews et al., 2004 [JoC, ARC] ; Meehl et al., 2004 [JoC, MoS, ARC] ). Two of these studies also examined realistic changes in many other forcing agents without finding evidence of a nonlinear response ( Meehl et al., 2004 [JoC, MoS, ARC] ; Matthews et al., 2004 [JoC, ARC] ). In all four studies, even the regional changes typically added linearly. However, Meehl et al. 2004 [JoC, MoS, ARC] ) observed that neither precipitation changes nor all regional temperature changes were linearly additive. This linear relationship also breaks down for global mean temperatures when aerosol-cloud interactions beyond the cloud albedo RF are included in GCMs ( Feichter et al., 2004 [JoC, MoS, SRC] ; see also Rotstayn and Penner, 2001 [JoC, MoS, ARC] ; Lohmann and Feichter, 2005 [SRC] ). Studies that include these effects modify clouds in their models, producing an additional radiative imbalance. Rotstayn and Penner 2001 [JoC, MoS, ARC] ) found that if these aerosol-cloud effects are accounted for as additional forcing terms, the inference of linearity can be restored (see Sections 2.8.3 and 2.8.5 ). Studies also find nonlinearities for large negative RFs, where static stability changes in the upper troposphere affect the climate feedback (e.g., Hansen et al., 2005 [JoC, MoS, ARC] ). For the magnitude and range of realistic RFs discussed in this chapter, and excluding cloud-aerosol interaction effects, there is high confidence in a linear relationship between global mean RF and global mean surface temperature response.

2.8.5 Efficacy and Effective Radiative Forcing

Efficacy (E) is defined as the ratio of the climate sensitivity parameter for a given forcing agent (λ_i)to the climate sensitivity parameter for CO₂ changes, that is,E_i = λ_i / λ_CO2 ( Joshi et al., 2003 [JoC, MoS] ; Hansen and Nazarenko, 2004 [JoC, MoS, ARC] ). Efficacy can then be used to define an effective RF (=E_i RF_i )(Joshi et al., 2003 [JoC, MoS] ; Hansen et al., 2005 [JoC, MoS, ARC] ). For the effective RF, the climate sensitivity parameter is independent of the mechanism, so comparing this forcing is equivalent to comparing the equilibrium global mean surface temperature change. That is, ΔT_s =λ_CO2 xE_i x RF_i Preliminary studies have found that efficacy values for a number of forcing agents show less model dependency than the climate sensitivity values ( Joshi et al., 2003 [JoC, MoS] ). Effective RFs have been used get one step closer to an estimator of the likely surface temperature response than can be achieved by using RF alone ( Sausen and Schumann, 2000 [JoC, MoS, SRC] ; Hansen et al., 2005 [JoC, MoS, ARC] ; Lohmann and Feichter, 2005 [SRC] ). Adopting the zero-surface-temperature-change RF, which has efficacies closer to unity, may be another way of achieving similar goals (see Section 2.8.3 ). This section assesses the efficacy associated with stratospherically adjusted RF, as this is the definition of RF adopted in this chapter (see Section 2.2 ). Therefore, cloud-aerosol interaction effects beyond the cloud albedo RF are included in the efficacy term. The findings presented in this section are from an assessment of all the studies referenced in the caption of Figure 2.19 ,which presents a synthesis of efficacy results. As space is limited not all these studies are explicitly discussed in the main text.

2.8.6 Efficacy and the Forcing-Response Relationship

Efficacy is a new concept introduced since the TAR and its physical understanding is becoming established (see Section 2.8.5 ). When employing the stratospherically adjusted RF, there is medium confidence that efficacies are within the 0.75 to 1.25 range for most realistic RF mechanisms aside from aerosol and stratospheric ozone changes. There is medium confidence that realistic aerosol and ozone changes have efficacies within the 0.5 to 2.0 range. Further, zero-surface-temperature-change RFs are very likely to have efficacies significantly closer to 1.0 for all mechanisms. It should be noted that efficacies have only been evaluated in GCMs and actual climate efficacies could be different from those quoted in Section 2.8.5 .

2.9.1 Uncertainties in Radiative Forcing

The TAR assessed uncertainties in global mean RF by attaching an error bar to each RF term that was ‘guided by the range of published values and physical understanding’. It also quoted a level of scientific understanding (LOSU) for each RF, which was a subjective judgment of the estimate’s reliability.

The concept of LOSU has been slightly modified based on the IPCC Fourth Assessment Report (AR4) uncertainty guidelines. Error bars now represent the 5 to 95% (90%) confidence range (see Box TS.1). Only ‘well-established’ RFs are quantified. ‘Well established’ implies that there is qualitatively both sufficient evidence and sufficient consensus from published results to estimate a central RF estimate and a range. ‘Evidence’ is assessed by an A to C grade, with an A grade implying strong evidence and C insufficient evidence. Strong evidence implies that observations have verified aspects of the RF mechanism and that there is a sound physical model to explain the RF. ‘Consensus’ is assessed by assigning a number between 1 and 3, where 1 implies a good deal of consensus and 3 insufficient consensus. This ranks the number of studies, how well studies agree on quantifying the RF and especially how well observation-based studies agree with models. The product of ‘Evidence’ and ‘Consensus’ factors give the LOSU rank. These ranks are high, medium, medium-low, low or very low. Ranks of very low are not evaluated. The quoted 90% confidence range of RF quantifies the value uncertainty, as derived from the expert assessment of published values and their ranges. For most RFs, many studies have now been published, which generally makes the sampling of parameter space more complete and the value uncertainty more realistic, compared to the TAR. This is particularly true for both the direct and cloud albedo aerosol RF (see Section 2.4 ). Table 2.11 summarises the key certainties and uncertainties and indicates the basis for the 90% confidence range estimate. Note that the aerosol terms will have added uncertainties due to the uncertain semi-direct and cloud lifetime effects. These uncertainties in the response to the RF (efficacy) are discussed in Section 2.8.5 .

Table 2.11 indicates that there is now stronger evidence for most of the RFs discussed in this chapter. Some effects are not quantified, either because they do not have enough evidence or because their quantification lacks consensus. These include certain mechanisms associated with land use, stratospheric water vapour and cosmic rays. Cloud lifetime and the semi-direct effects are also excluded from this analysis as they are deemed to be part of the climate response (see Section 7.5 ). The RFs from the LLGHGs have both a high degree of consensus and a very large amount of evidence and, thereby, place understanding of these effects at a considerably higher level than any other effect.

Table 2.11 .Uncertainty assessment of forcing agents discussed in this chapter. Evidence for the forcing is given a grade (A to C), with A implying strong evidence and C insufficient evidence. The degree of consensus among forcing estimates is given a 1, 2 or 3 grade, where grade 1 implies a good deal of consensus and grade 3 implies an insufficient consensus. From these two factors, a level of scientific understanding is determined (LOSU). Uncertainties are in approximate order of importance with first-order uncertainties listed first.

2.9.2 Global Mean Radiative Forcing

The RFs discussed in this chapter, their uncertainty ranges and their efficacies are summarised in Figure 2.20 and Table 2.12. Radiative forcings from forcing agents have been combined into their main groupings. This is particularly useful for aerosol as its total direct RF is considerably better constrained than the RF from individual aerosol types (see Section 2.4.4 ). Table 2.1 gives a further component breakdown of RF for the LLGHGs. Radiative forcings are the stratospherically adjusted RF and they have not been multiplied by efficacies (see Sections 2.2 and 2.8 ).

In the TAR, no estimate of the total combined RF from all anthropogenic forcing agents was given because: a) some of the forcing agents did not have central or best estimates; b) a degree of subjectivity was included in the error estimates; and c) uncertainties associated with the linear additivity assumption and efficacy had not been evaluated. Some of these limitations still apply. However, methods for objectively adding the RF of individual species have been developed (e.g., Schwartz and Andreae, 1996 [JoC, ARC] ; Boucher and Haywood, 2001 [JoC, MoS, SRC] ). In addition, as efficacies are now better understood and quantified (see Section 2.8.5 ), and as the linear additivity assumption has been more thoroughly tested (see Section 2.8.4 ), it becomes scientifically justifiable for RFs from different mechanisms to be combined, with certain exceptions as noted below. Adding together the anthropogenic RF values shown in panel (A) of Figure 2.20 and combining their individual uncertainties gives the probability density functions (PDFs) of RF that are shown in panel (B). Three PDFs are shown: the combined RF from greenhouse gas changes (LLGHGs and ozone); the combined direct aerosol and cloud albedo RFs and the combination of all anthropogenic RFs. The solar RF is not included in any of these distributions. The PDFs are generated by combining the 90% confidence estimates for the RFs, assuming independence and employing a one-million point Monte Carlo simulation to derive the PDFs (see Boucher and Haywood, 2001 [JoC, MoS, SRC] ; and Figure 2.20 caption for details).

The PDFs show that LLGHGs and ozone contribute a positive RF of +2.9 ± 0.3 Wm^–2.The combined aerosol direct and cloud albedo effect exert an RF that is virtually certain to be negative, with a median RF of –1.3 Wm^–2 and a –2.2 to –0.5 Wm^–2 90% confidence range. The asymmetry in the combined aerosol PDF is caused by the estimates in Tables 2.6 and 2.7 being non-Gaussian. The combined net RF estimate for all anthropogenic drivers has a value of +1.6 Wm^–2 with a 0.6 to 2.4 Wm^–2 90% confidence range. Note that the RFs from surface albedo change, stratospheric water vapour change and persistent contrails are only included in the combined anthropogenic PDF and not the other two.

Statistically, the PDF shown in Figure 2.20 indicates just a 0.2% probability that the total RF from anthropogenic agents is negative, which would suggest that it is virtually certain that the combined RF from anthropogenic agents is positive. Additionally, the PDF presented here suggests that it is extremely likely that the total anthropogenic RF is larger than +0.6 Wm^–2.This combined anthropogenic PDF is better constrained than that shown in Boucher and Haywood 2001 [JoC, MoS, SRC] ) because each of the individual RFs have been quantified to 90% confidence levels, enabling a more definite assessment, and because the uncertainty in some of the RF estimates is considerably reduced. For example, modelling of the total direct RF due to aerosols is better constrained by satellite and surface-based observations( Section 2.4.2 ), and the current estimate of the cloud albedo indirect effect has a best estimate and uncertainty associated with it, rather than just a range. The LLGHG RF has also increased by 0.20 Wm^–2 since 1998, making a positive RF more likely than in Boucher and Haywood 2001 [JoC, MoS, SRC] ).

Nevertheless, there are some structural uncertainties associated with the assumptions used in the construction of the PDF and the assumptions describing the component uncertainties. Normal distributions are assumed for most RF mechanisms (with the exceptions noted in the caption); this may not accurately capture extremes. Additionally, as in Boucher and Haywood 2001 [JoC, MoS, SRC] ), all of the individual RF mechanisms are given equal weighting, even though the level of scientific understanding differs between forcing mechanisms. Note also that variation in efficacy and hence the semi-direct and cloud lifetime effects are not accounted for, as these are not considered to be RFs in this report (see Section 2.2 ). Adding these effects, together with other potential mechanisms that have so far not been defined as RFs and quantified, would introduce further uncertainties but give a fuller picture of the role of anthropogenic drivers. Introducing efficacy would give a broader PDF and a large cloud lifetime effect would reduce the median estimate. Despite these caveats, from the current knowledge of individual forcing mechanisms presented here it remains extremely likely that the combined anthropogenic RF is both positive and substantial (best estimate: +1.6 Wm^–2).

2.9.3 Global Mean Radiative Forcing by Emission Precursor

The RF due to changes in the concentration of a single forcing agent can have contributions from emissions of several compounds ( Shindell et al., 2005 [PoC, JoC, MoS, ARC] ). The RF of CH₄,for example, is affected by CH₄ emissions, as well as NO_x emissions. The CH₄ RF quoted in Table 2.12 and shown in Figure 2.20 is a value that combines the effects of both emissions. As an anthropogenic or natural emission can affect several forcing agents, it is useful to assess the current RF caused by each primary emission. For example, emission of NO_x affects CH₄,tropospheric ozone and tropospheric aerosols. Based on a development carried forward from the TAR, this section assesses the RF terms associated with each principal emission including indirect RFs related to perturbations of other forcing agents, with the results shown in Figure 2.21 .The following indirect forcing mechanisms are considered:

• fossil carbon from non-CO₂ gaseous compounds, which eventually increase CO₂ in the atmosphere (from CO, CH₄,and NMVOC emissions);
• changes in stratospheric ozone (fromN₂Oand halocarbon (CFCs, HCFC, halons, etc.) emissions);
• changes in tropospheric ozone (from CH₄,NO_x,CO, and NMVOC emissions);
• changes in OH affecting the lifetime of CH₄ (from CH₄,CO, NO_x,and NMVOC emissions); and
• changing nitrate and sulphate aerosols through changes in NO_x and SO₂ emissions, respectively.

For some of the principal RFs (e.g., BC, land use and mineral dust) there is not enough quantitative information available to assess their indirect effects, thus their RFs are the same as those presented in Table 2.12. Table 2.5 gives the total (fossil and biomass burning) direct RFs for BC and organic carbon aerosols that are used to obtain the average shown in Figure 2.21 . Table 2.13 summarises the direct and indirect RFs presented in Figure 2.21 ,including the methods used for estimating the RFs and the associated uncertainty. Note that for indirect effects through changes in chemically active gases (e.g., OH or ozone), the emission-based RF is not uniquely defined since the effect of one precursor will be affected by the levels of the other precursors. The RFs of indirect effects on CH₄ and ozone by NO_x,CO and VOC emissions are estimated by removing the anthropogenic emissions of one precursor at a time. A sensitivity analysis by Shindell et al. 2005 [PoC, JoC, MoS, ARC] ) indicates that the nonlinear effect induced by treating the precursors separately is of the order of 10% or less. Very uncertain indirect effects are not included in Table 2.13 and Figure 2.21 .These include ozone changes due to solar effects, changes in secondary organic aerosols through changes in the ozone/OH ratio and apportioning of the cloud albedo changes to each aerosol type ( Hansen et al., 2005 [JoC, MoS, ARC] ).

Table 2.12. Global mean radiative forcings since 1750 and comparison with earlier assessments. Bold rows appear on Figure 2.20. The first row shows the combined anthropogenic RF from the probability density function in panel B of Figure 2.20. The sum of the individual RFs and their estimated errors are not quite the same as the numbers presented in this row due to the statistical construction of the probability density function.

Notes: ^a For the AR4 column, 90% value uncertainties appear in brackets: when adding these numbers to the best estimate the 5 to 95% confidence range is obtained. When two numbers are quoted for the value uncertainty, the distribution is non-normal. Uncertainties in the SAR and the TAR had a similar basis, but their evaluation was more subjective. [15%] indicates 15% relative uncertainty, [2x], etc. refer to a factor of two, etc. uncertainty and a lognormal distribution of RF estimates. ^b The TAR RF for halocarbons and hence the total LLGHG RF was incorrectly evaluated some 0.01 Wm^–2 too high. The actual trends in these RFs are therefore more positive than suggested by numbers in this table( Table 2.1 shows updated trends).

2.9.4 Future Climate Impact of Current Emissions

The changes in concentrations since pre-industrial time of the long-lived components causing the RF shown in Figure 2.20 are strongly influenced by the past history of emissions. A different perspective is obtained by integrating RF over a future time horizon for a one-year ‘pulse’ of global emissions (e.g., Jacobson 2002 [JoC] ) used this approach to compare fossil fuel organic and BC aerosols to CO₂). Comparing the contribution from each forcing agent as shown in Figure 2.22 gives an indication of the future climate impact for current (year 2000 ) emissions of the different forcing agents. For the aerosols, the integrated RF is obtained based on the lifetimes, burdens and RFs from the AeroCom experiments, as summarised in Tables 2.4 and 2.5 .For ozone precursors (CO, NO_x and NMVOCs), data are taken from Derwent et al. 2001 [JoC, ARC] Collins et al. 2002 [JoC, ARC] Stevenson et al. 2004 [JoC, MoS, ARC] and Berntsen et al. 2005 [JoC, SRC] ), while for the long-lived species the radiative efficiencies and lifetimes are used, as well as a response function for CO₂ (see Section 2.10.2 , Table 2.14 ). Uncertainties in the estimates of the integrated RF originate from uncertainties in lifetimes, optical properties and current global emissions.

Figure 2.22 shows the integrated RF for both a 20- and 100-year time horizon. Choosing the longer time horizon of 100 years, as was done in the GWPs for the long-lived species included in the Kyoto Protocol, reduces the apparent importance of the shorter-lived species. It should be noted that the compounds with long lifetimes and short emission histories will tend to contribute more to the total with this ‘forward looking’ perspective than in the standard ‘IPCC RF bar chart diagram’( Figure 2.20 ).

2.9.5 Time Evolution of Radiative Forcing and Surface Forcing

There is a good understanding of the time evolution of the LLGHG concentrations from in situ measurements over the last few decades and extending further back using firn and ice core data (see Section 2.3 , FAQ 2.1 ,Figure 1 and Chapter6 ). Increases in RF are clearly dominated by CO₂.Halocarbon RF has grown rapidly since 1950, but the RF growth has been cut dramatically by the Montreal Protocol (see Section 2.3.4 ). The RF of CFCs is declining; in addition, the combined RF of all ozone-depleting substances (ODS) appears to have peaked at 0.32 Wm^–2 during 2003 . However, substitutes for ODS are growing at a slightly faster rate, so halocarbon RF growth is still positive( Table 2.1 ). Although the trend in halocarbon RF since the time of the TAR has been positive (see Table 2.1 ), the halocarbon RF in this report, as shown in Table 2.12, is the same as in the TAR; this is due to a re-evaluation of the TAR results.

Radiative forcing time series for the natural (solar, volcanic aerosol) forcings are reasonably well known for the past 25 years; estimates further back are prone to uncertainties( Section 2.7 ). Determining the time series for aerosol and ozone RF is far more difficult because of uncertainties in the knowledge of past emissions and chemical-microphysical modelling. Several time series for these and other RFs have been constructed (e.g., Myhre et al., 2001 [MoS, SRC] ; Ramaswamy et al., 2001 [NPR, MoS, SRC] ; Hansen et al., 2002 [JoC, MoS, ARC] ). General Circulation Models develop their own time evolution of many forcings based on the temporal history of the relevant concentrations. As an example, the temporal evolution of the global and annual mean, instantaneous, all-sky RF and surface forcing due to the principal agents simulated by the Model for Interdisciplinary Research on Climate (MIROC) + Spectral Radiation-Transport Model for Aerosol Species (SPRINTARS) GCM ( Nozawa et al., 2005 [JoC, ARC] ; Takemura et al., 2005 [JoC, MoS, SRC] ) is illustrated in Figure 2.23 .Although there are differences between models with regards to the temporal reconstructions and thus present-day forcing estimates, they typically have a qualitatively similar temporal evolution since they often base the temporal histories on similar emissions data.

General Circulation Models compute the climate response based on the knowledge of the forcing agents and their temporal evolution. While most current GCMs incorporate the trace gas RFs, aerosol direct effects, solar and volcanoes, a few have in addition incorporated land use change and cloud albedo effect. While LLGHGs have increased rapidly over the past 20 years and contribute the most to the present RF (refer also to Figure 2.20 and FAQ 2.1 ,Figure 1), Figure 2.23 also indicates that the combined positive RF of the greenhouse gases exceeds the contributions due to all other anthropogenic agents throughout the latter half of the 20th century.

The solar RF has a small positive value. The positive solar irradiance RF is likely to be at least five times smaller than the combined RF due to all anthropogenic agents, and about an order of magnitude less than the total greenhouse gas contribution (Figures 2.20 and 2.23 and Table 2.12; see also the Foukal et al., 2006 [PoC, JoC, SRC] review). The combined natural RF consists of the solar RF plus the large but transitory negative RF from episodic, explosive volcanic eruptions of which there have been several over the past half century (see Figure 2.18 ). Over particularly the 1950 to 2005 period, the combined natural forcing has been either negative or slightly positive (less than approximately 0.2 Wm^–2), reaffirming and extending the conclusions in the TAR. Therefore, it is exceptionally unlikely that natural RFs could have contributed a positive RF of comparable magnitude to the combined anthropogenic RF term over the period 1950 to 2005 ( Figure 2.23 ). Attribution studies with GCMs employ the available knowledge of the evolution of the forcing over the 20th century, and particularly the features distinguishing the anthropogenic from the natural agents (see also Section 9.2 ).

The surface forcing( Figure 2.23 ,top panel), in contrast to RF, is dominated by the strongly negative shortwave effect of the aerosols (tropospheric and the episodic volcanic ones), with the LLGHGs exerting a small positive effect. Quantitative values of the RFs and surface forcings by the agents differ across models in view of the differences in model physics and in the formulation of the forcings due to the short-lived species (see Section 10.2 ,Collins et al. 2006 [JoC, MoS, SRC] and Forster and Taylor 2006 [JoC, MoS, SRC] ) for further discussion on uncertainties in GCMs’ calculation of RF and surface forcing). As for RF, it is difficult to specify uncertainties in the temporal evolution, as emissions and concentrations for all but the LLGHGs are not well constrained.