7.2.2.1 Multiple Scales are Important

Temporal variability ranges from the daily and weather time scales to annual, interannual, and decadal or longer scales: the amplitudes of shorter time scales change with long-term changes from global warming. The land climate system has controls on amplitudes of variables on all these time scales, varying with season and geography. For example, Trenberth and Shea 2005 [PoC, JoC, SRC] ) evaluate from climatic observations the correlation between surface air temperature and precipitation, and find a strong r > 0.3) positive correlation over most winter land areas (i.e., poleward of 40°N) but a strong (|r| > 0.3) negative correlation over much of summer and tropical land. These differences result from competing feedbacks with the water cycle. On scales large enough that surface temperatures control atmospheric temperatures, the atmosphere will hold more water vapour and may provide more precipitation with warmer temperatures. Low clouds strongly control surface temperatures, especially in cold regions where they make the surface warmer. In warm regions without precipitation, the land surface can become warmer because of lack of evaporation, or lack of clouds. Although a drier surface will become warmer from lack of evaporative cooling, more water can evaporate from a moist surface if the temperature is warmer (see Box 7.1 ).

7.2.2.2 Spatial Dependence

Drivers of the land climate system have larger effects at regional and local scales than on global climate, which is controlled primarily by processes of global radiation balance. Myhre et al. 2005 [JoC, MoS, ARC] ) point out that the albedo of agricultural systems may be only slightly higher than that of forests and estimate that the impact since pre-agricultural times of land use conversion to agriculture on global radiative forcing has been only –0.09 Wm^–2,that is, about 5% of the warming contributed by CO₂ since pre-industrial times (see Chapter2 for a more comprehensive review of recent estimates of land surface albedo change). Land comprises only about 30% of the Earth’s surface, but it can have the largest effects on the reflection of global solar radiation in conjunction with changes in ice and snow cover, and the shading of the latter by vegetation.

At a regional scale and at the surface, additional more localised and shorter time-scale processes besides radiative forcing can affect climate in other ways, and possibly be of comparable importance to the effects of the greenhouse gases. Changes over land that modify its evaporative cooling can cause large changes in surface temperature, both locally and regionally (see Boxes 7.1 , 7.2 ). How this change feeds back to precipitation remains a major research question. Land has a strong control on the vertical distribution of atmospheric heating. It determines how much of the radiation delivered to land goes into warming the near-surface atmosphere compared with how much is released as latent heat fuelling precipitation at higher levels. Low clouds are normally closely coupled to the surface and over land can be significantly changed by modifications of surface temperature or moisture resulting from changes in land properties. For example, Chagnon et al. 2004 [JoC] ) find a large increase in boundary layer clouds in the Amazon in areas of partial deforestation (also, e.g., ( Durieux et al., 2003; ) Ek and Holtslag, 2004 [JoC, ARC] ). Details of surface properties at scales as small as a few kilometres can be important for larger scales. Over some fraction of moist soils, water tables can be high enough to be hydrologically connected to the rooting zone, or reach the surface as in wetlands (e.g., Koster et al., 2000 [JoC, MoS, ARC] ;( Marani et al., 2001; ) Milly and Shmakin, 2002 [JoC, MoS] ; Liang et al., 2003 [JoC, MoS] ; Gedney and Cox, 2003 [JoC, MoS, SRC] ).

The consequences of changes in atmospheric heating from land changes at a regional scale are similar to those from ocean temperature changes such as from El Niño, potentially producing patterns of reduced or increased cloudiness and precipitation elsewhere to maintain global energy balance. Attempts have been made to find remote adjustments (e.g., Avissar and Werth, 2005 [JoC] ). Such adjustments may occur in multiple ways, and are part of the dynamics of climate models. The locally warmer temperatures can lead to more rapid vertical decreases of atmospheric temperature so that at some level overlying temperature is lower and radiates less. The net effect of such compensations is that averages over larger areas or longer time scales commonly will give smaller estimates of change. Thus, such regional changes are better described by local and regional metrics or at larger scales by measures of change in spatial and temporal variability rather than simply in terms of a mean global quantity.

7.2.2.3 Daily and Seasonal Variability

Diurnal and seasonal variability result directly from the temporal variation of the solar radiation driver. Large-scale changes in climate variables are of interest as part of the observational record of climate changes( Chapter3 ). Daytime during the warm season produces a thick layer of mixed air with temperature relatively insensitive to perturbations in daytime radiative forcing. Nighttime and high-latitude winter surface temperatures, on the other hand, are coupled by mixing to only a thin layer of atmosphere, and can be more readily altered by changes in atmospheric downward thermal radiation. Thus, land is more sensitive to changes in radiative drivers under cold stable conditions and weak winds than under warm unstable conditions. Winter or nighttime temperatures (hence diurnal temperature range) are strongly correlated with downward longwave radiation (e.g., Betts, 2006 [JoC] ; Dickinson et al., 2006 [JoC, MoS, SRC] ); consequently, average surface temperatures may change (e.g., Pielke and Matsui, 2005 [JoC] ) with a change in downward longwave radiation.

Modification of downward longwave radiation by changes in clouds can affect land surface temperatures. Qian and Giorgi 2000 [JoC, ARC] ) discussed regional aerosol effects, and noted a reduction in the diurnal temperature range of –0.26°C per decade over Sichuan China. Huang et al. 2006 [JoC, SRC] ) model the growth of sulphate aerosols and their interactions with clouds in the context of a RCM, and find over southern China a decrease in the diurnal temperature range comparable with that observed by Zhou et al. 2004 [JoC, SRC] ) and Qian and Giorgi. They show the nighttime temperature change to be a result of increased nighttime cloudiness and hence downward longwave radiation connected to the increase in aerosols.

In moist warm regions, large changes are possible in the fraction of energy going into water fluxes, for example, by changes in vegetation cover or precipitation, and hence in soil moisture. Bonan 2001 [JoC, SRC] and Oleson et al. 2004 [JoC, MoS, SRC] ) indicate that conversion of mid-latitude forests to agriculture could cause a daytime cooling. This cooling is apparently a result of higher albedo and increased transpiration. Changes in reflected solar radiation due to changing vegetation, hence feedbacks, are most pronounced in areas with vegetation underlain by snow or light-coloured soil. Seasonal and diurnal precipitation cycles can be pronounced. Climate models simulate the diurnal precipitation cycle but apparently not yet very well (e.g., Collier and Bowman, 2004 [JoC, MoS] Betts 2004 [Ambiguous] ) reviews how the diurnal cycle of tropical continental precipitation is linked to land surface fluxes and argues that errors in a model can feed back to model dynamics with global impacts.

7.2.2.4 Coupling of Precipitation Intensities to Leaf Water – An Issue Involving both Temporal and Spatial Scales

The bulk of the water exchanged with the atmosphere is stored in the soil until taken up by plant roots, typically weeks later. However, the rapidity of evaporation of the near-surface stores allows plant uptake and evaporation to be of comparable importance for surface water and energy balances. ( Dickinson et al., 2003 [MoS, SRC] , conclude that feedbacks between surface moisture and precipitation may act differently on different time scales). Evaporation from the fast reservoirs acts primarily as a surface energy removal mechanism. Leaves initially intercept much of the precipitation over vegetation, and a significant fraction of this leaf water re-evaporates in an hour or less. This loss reduces the amount of water stored in the soil for use by plants. Its magnitude depends inversely on the intensity of the precipitation, which can be larger at smaller temporal and spatial scales. Modelling results can be wrong either through neglect of or through exaggeration of the magnitude of the fast time-scale moisture stores.

Leaf water evaporation may have little effect on the determination of monthly evapotranspiration (e.g., as found in the analysis of Desborough, 1999 [JoC, MoS] ) but may still produce important changes in temperature and precipitation. Pitman et al. 2004 [JoC, MoS, SRC] ), in a coupled study with land configurations of different complexity, were unable to find any impacts on atmospheric variability, but Bagnoud et al. 2005 [JoC, SRC] ) found that precipitation and temperature extremes were affected. Some studies that change the intensity of precipitation find a very large impact from leaf water. For example, Wang and Eltahir 2000 [JoC, MoS] ) studied the effect of including more realistic precipitation intensity compared to the uniform intensity of a climate model. Hahmann 2003 [JoC] ) used another model to study this effect. Figure 7.1 compares their tropical results (Wang and Eltahir over equatorial Africa and Hahmann over equatorial Amazon). The model of Wang and Eltahir shows that more realistic precipitation greatly increases runoff whereas Hahmann shows that it reduces runoff. It has not been determined whether these contradictory results are more a consequence of model differences or of differences between the climates of the two continents, as Hahmann suggests.

Figure 7.1. Rainfall, runoff and evapotranspiration derived from climate simulation results of Hahmann (H; 2004 ) and Wang and Eltahir (W; 2000 ). Hahmann’s results are for the Amazon centred on the equator, and Wang and Eltahir’s for Africa at the equator. Both studies examined the differences between ‘uniform’ precipitation over a model grid square and ‘variable’ precipitation (added to about 10% of the grid square). Large differences are seen between the two cases in the two studies: a large reduction in precipitation is seen in the Hahmann variable case relative to the uniform case, whereas an increase is seen for the Wang and Eltahir variable case. The differences are even greater for runoff: Hahmann’s uniform case runoff is three times as large as the variable case, whereas Wang and Eltahir have almost no runoff for their uniform case.

7.2.3.1 Vegetative Controls on Soil Water and its Return Flux to the Atmosphere

Scanlon et al. 2005 [JoC] ) provide an example of how soil moisture can depend on vegetation. They monitored soil moisture in the Nevada desert with lysimeters either including or excluding vegetation and for a multi-year period that included times of anomalously strong precipitation. Without vegetation, much of the moisture penetrated deeply, had a long lifetime and became available for recharge of deep groundwater, whereas for the vegetated plot, the soil moisture was all transpired. In the absence of leaves, forests in early spring also appear as especially dry surfaces with consequent large sensible fluxes that mix the atmosphere to a great depth (e.g., Betts et al., 2001 [JoC] ). Increased water fluxes with spring green-up are observed in terms of a reduction in temperature. Trees in the Amazon can have the largest water fluxes in the dry season by development of deep roots ( (Da Rocha et al., 2004; ) ( Quesada et al., 2004 ) ). Forests can also retard fluxes through control by their leaves. Such control by vegetation of water fluxes is most pronounced for taller or sparser vegetation in cooler or drier climates, and from leaves that are sparse or exert the strongest resistance to water movement. The boreal forest, in particular, has been characterised as a ‘green desert’ because of its small release of water to the atmosphere ( (Gamon et al., 2003 ) ).

7.2.3.2 Land Feedback to Precipitation

Findell and Eltahir 2003 [JoC] ) examine the correlation between early morning near-surface humidity over the USA and an index of the likelihood of precipitation occurrence. They identify different geographical regions with positive, negative or little correlation. Koster et al. 2003 [JoC, ARC] and Koster and Suarez 2004 [JoC, ARC] ) show during summer over the USA, and all land 30°N to 60°N, respectively, a significant correlation of monthly precipitation with that of prior months. They further show that their model only reproduces this correlation if soil moisture feedback is allowed to affect precipitation. Additional observational evidence for such feedback is noted by D’Odorico and Porporato 2004 [JoC] ) in support of a simplified model of precipitation soil moisture coupling (see, e.g., ( Salvucci et al., 2002, ) for support of the null hypothesis of no coupling). Liebmann and Marengo 2001 [JoC] ) point out that the interannual variation of precipitation over the Amazon is largely controlled by the timing of the onset and end of the rainy season. Li and Fu 2004 [JoC] ) provide evidence that onset time of the rainy season is strongly dependent on transpiration by vegetation during the dry season. Previous modelling and observational studies have also suggested that Amazon deforestation should lead to a longer dry season. ( Fu and Li 2004 ) ) further argue from observations that removal of tropical forest reduces surface moisture fluxes, and that such land use changes should contribute to a lengthening of the Amazon dry season. ( Durieux et al. 2003 ) ) find more rainfall in the deforested area in the wet season and a reduction of the dry season precipitation over deforested regions compared with forested areas. Negri et al. 2004 [JoC, ARC] ) obtain an opposite result (although their result is consistent with Durieux during the wet season).

Albedo (the fraction of reflected solar radiation) and emissivity (the ratio of thermal radiation to that of a black body) are important variables for the radiative balance. Surfaces that have more or taller vegetation are commonly darker than are those with sparse or shorter vegetation. With sparse vegetation, the net surface albedo also depends on the albedo of the underlying surfaces, especially if snow or a light-coloured soil. A large-scale transformation of tundra to shrubs, possibly connected to warmer temperatures over the last few decades, has been observed (e.g., Chapin et al., 2005 [JoC] Sturm et al. 2005 [JoC] ) report on winter and melt season observations of how varying extents of such shrubs can modify surface albedo. New satellite data show the importance of radiation heterogeneities at the plot scale for the determination of albedo and the solar radiation used for photosynthesis, and appropriate modelling concepts to incorporate the new data are being advanced (e.g., Yang and Friedl, 2003 [JoC, MoS] ; Niu and Yang, 2004 [JoC] ; Wang, 2005 [MoS] ; Pinty et al., 2006 [JoC, MoS] ).

7.2.3.3 Properties Affecting Radiation

7.2.3.4 Improved Global and Regional Data

Specification of land surface properties has improved through new, more accurate global satellite observations. In particular, satellite observations have provided albedos of soils in non-vegetated regions (e.g., Tsvetsinskaya et al., 2002 [JoC] ;( Ogawa and Schmugge, 2004; ) Z. Wang, et al., 2004 [Ambiguous] ; Zhou et al., 2005 [JoC, SRC] ) and their emissivities ( Zhou et al., 2003a [JoC, MoS, SRC] ,b). They also constrain model-calculated albedos in the presence of vegetation ( Oleson et al., 2003 [JoC, MoS] ) and vegetation underlain by snow ( Jin et al., 2002 [NPR, JoC] ), and help to define the influence of leaf area on albedo ( Tian et al., 2004 [NPR, JoC, MoS, SRC] ). Precipitation data sets combining rain gauge and satellite observations ( Chen et al., 2002 [JoC] ; Adler et al., 2003 [JoC, ARC] ) are providing diagnostic constraints for climate modelling, as are observations of runoff ( Dai and Trenberth, 2002 [PoC, JoC, MoS, SRC] ; Fekete et al., 2002 [NPR, JoC, MoS] ).

New and improved local site observational constraints collectively describe the land processes that need to be modelled. The largest recent such activity has been the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA) project ( Malhi et al., 2002 [SRC] ; Silva Dias et al., 2002 [JoC] ). Studies within LBA have included physical climate at all scales, carbon and nutrient dynamics and trace gas fluxes. The physical climate aspects are reviewed here. Goncalves et al. 2004 [MoS] ) discuss the importance of incorporating land cover heterogeneity. ( da Rocha et al. 2004 ) ( and Quesada et al. 2004 ) ) quantify water and energy budgets for a forested and a savannah site, respectively. Dry season evapotranspiration for the savannah averaged 1.6 mm day^–1 compared with 4.9 mm day^–1 for the forest. Both ecosystems depend on deep rooting to sustain evapotranspiration during the dry season, which may help control the length of the dry season (see, e.g., Section 7.2.3.2 ( da Rocha et al. 2004 ) ) also observed that hydraulic lift recharged the forest upper soil profiles each night. At Tapajós, the forest showed no signs of drought stress allowing uniformly high carbon uptake throughout the dry season (July through December 2000; ( da Rocha et al., 2004; ) ( Goulden et al., 2004 ) ). Tibet, another key region, continues to be better characterised from observational studies (e.g., Gao et al., 2004 [JoC, MoS] ; Hong et al., 2004 [JoC] ). With its high elevation, hence low air densities, heating of the atmosphere by land mixes air to a much higher altitude than elsewhere, with implications for vertical exchange of energy. However, the daytime water vapour mixing ratio in this region decreases rapidly with increasing altitude ( Yang et al., 2004 [MoS] ), indicating a strong insertion of dry air from above or by lateral transport.

7.2.3.5 Field Observational Programs

7.2.3.6 Connecting Changing Vegetation to Changing Climate

Only large-scale patterns are assessed here. Analysis of satellite-sensed vegetation greenness and meteorological station data suggest an enhanced plant growth and lengthened growing season duration at northern high latitudes since the 1980 s ( Zhou et al., 2001 [JoC, SRC] , 2003c ). This effect is further supported by modelling linked to observed climate data ( Lucht et al., 2002 [JoC] Nemani et al. 2002 [JoC] , 2003 ) suggest that increased rainfall and humidity spurred plant growth in the USA and that climate changes may have eased several critical climatic constraints to plant growth and thus increased terrestrial net primary production.

Box 7.1 provides a general description of water fluxes from surface to atmosphere. The most important factors affected by vegetation are soil water availability, leaf area and surface roughness. Whether water has been intercepted on the surface of the leaves or its loss is only from the leaf interior as controlled by stomata makes a large difference. Shorter vegetation with more leaves has the most latent heat flux and the least sensible flux. Replacement of forests with shorter vegetation together with the normally assumed higher albedo could then cool the surface. However, if the replacement vegetation has much less foliage or cannot access soil water as successfully, a warming may occur. Thus, deforestation can modify surface temperatures by up to several degrees celsius in either direction depending on what type of vegetation replaces the forest and the climate regime. Drier air can increase evapotranspiration, but leaves may decrease their stomatal conductance to counter this effect.

7.2.4.1 How do Models of Vegetation Control Surface Water Fluxes?

7.2.4.2 Feedbacks Demonstrated Through Simple Models

In semi-arid systems, the occurrence and amounts of precipitation can be highly variable from year to year. Are there mechanisms whereby the growth of vegetation in times of adequate precipitation can act to maintain the precipitation? Various analyses with simple models have demonstrated how this might happen ( Zeng et al., 2002 [JoC] ;( Foley et al., 2003; ) G. Wang, et al., 2004 [Ambiguous] ; X. Zeng et al., 2004 [Ambiguous] ). Such models demonstrate how assumed feedbacks between precipitation and surface fluxes generated by dynamic vegetation may lead to the possibility of transitions between multiple equilibria for two soil moisture and precipitation regimes. That is, the extraction of water by roots and shading of soil by plants can increase precipitation and maintain the vegetation, but if the vegetation is removed, it may not be able to be restored for a long period. The Sahel region between the deserts of North Africa and the African equatorial forests appears to most readily generate such an alternating precipitation regime.

experiments and find that changes in roughness, soil depth, vegetation cover, stomatal resistance, albedo and leaf area index all could make significant contributions. Voldoire and Royer 2004 [JoC] ) find that such changes may affect temperature and precipitation extremes more than means, in particular the daytime maximum temperature and the drying and temperature responses associated with El Niño events. Guillevic et al. 2002 [JoC] ) address the importance of interannual leaf area variability as inferred from Advanced Very High Resolution Radiometer (AVHRR) satellite data, and infer a sensitivity of climate to this variation. In contrast, Lawrence and Slingo 2004 [NPR, JoC, ARC] ) find little difference in climate simulations that use annual mean vegetation characteristics compared with those that use a prescribed seasonal cycle. However, they do suggest model modifications that would give a much larger sensitivity. Osborne et al. 2004 [JoC] ) examine effects of changing tropical soils and vegetation: variations in vegetation produce variability in surface fluxes and their coupling to precipitation. Thus, interactive vegetation can promote additional variability of surface temperature and precipitation as analysed by Crucifix et al. 2005 [JoC, MoS, SRC] Marengo and Nobre 2001 [NPR] ) found that removal of vegetation led to a decrease in precipitation and evapotranspiration and a decrease in moisture convergence in central and northern Amazonia. Oyama and Nobre 2004 [JoC, MoS] ) show that removal of vegetation in northeast Brazil would substantially decrease precipitation.

7.2.4.3 Consequences of Changing Moisture Availability and Land Cover

Soil moisture control of the partitioning of energy between sensible and latent heat flux is very important for local and

regional temperatures, and possibly their coupling to precipitation. Oglesby et al. 2002 [JoC, MoS] ) carried out a study starting with dry soil where the dryness of the soil over the US Great Plains for at least the first several summer months of their integration produced a warming of about 10°C to 20°C. Williamson et al. 2005 [JoC, MoS] ), have shown that flaws in model formulation of thunderstorms can cause excessive evapotranspiration that lowers temperatures by more than 1°C. Many modelling studies have demonstrated that changing land cover can have local and regional climate impacts that are comparable in magnitude to temperature and precipitation changes observed over the last several decades as reported in Chapter3 .However, since such regional changes can be of both signs, the global average impact is expected to be small. Current literature has large disparities in conclusions. For example, Snyder et al. 2004 [JoC] ) found that removal of northern temperate forests gave a summer warming of 1.3°C and a reduction in precipitation of 1.5 mm day^-1.Conversely, Oleson et al. 2004 [JoC, MoS, SRC] ) found that removal of temperate forests in the USA would cool summer temperatures by 0.4°C to 1.5°C and probably increase precipitation, depending on the details of the model and prescription of vegetation. The discrepancy between these two studies may be largely an artefact of different assumptions. The first study assumes conversion of forest to desert and the second to crops. Such studies collectively demonstrate a potentially important impact of human activities on climate through land use modification.

Other recent such studies illustrate various aspects of this issue. Maynard and Royer 2004 [NPR, JoC, MoS] ) address the sensitivity to different parameter changes in African deforestation

7.2.4.4 Mechanisms for Modification of Precipitation by Spatial Heterogeneity

Clark et al. 2004 [MoS, SRC] ) show an example of a ‘squall-line’ simulation where soil moisture variation at the scale of the rainfall modifies the rainfall pattern. Pielke 2001 [JoC, MoS] Weaver et al. 2002 [JoC, MoS] ) and S. Roy et al. 2003 [Ambiguous] ) also address various aspects of small-scale precipitation coupling to land surface heterogeneity. If deforestation occurs in patches rather than uniformly, the consequences for precipitation could be different. Avissar et al. 2002 [JoC, MoS] and Silva Dias et al. 2002 [JoC] ) suggest that there may be a small increase in precipitation (of the order of 10%) resulting from partial deforestation as a consequence of the mesoscale circulations triggered by the deforestation.

Prognostic approaches estimate leaf cover based on physiological processes (e.g., Arora and Boer, 2005 [MoS, SRC] Levis and Bonan 2004 [JoC, MoS, SRC] ) discuss how spring leaf emergence in mid-latitude forests provides a negative feedback to rapid increases in temperature. The parametrization of water uptake by roots contributes to the computed soil water profile ( Feddes et al., 2001 [JoC, MoS] ; Barlage and Zeng, 2004 [JoC, MoS] ), and efforts are being made to make the roots interactive (e.g., Arora and Boer, 2003 [MoS, SRC] ). Dynamic vegetation models have advanced and now explicitly simulate competition between plant functional types (e.g., Bonan et al., 2003 [MoS, SRC] ; Sitch et al., 2003 [MoS] ; Arora and Boer, 2006 [MoS, SRC] ). New coupled climate-carbon models ( Betts et al., 2004 [Ambiguous] ; Huntingford et al., 2004 [MoS] ) demonstrate the possibility of large feedbacks between future climate change and vegetation change, discussed further in Section 7.3.5 (i.e., a die back of Amazon vegetation and reductions in Amazon precipitation). They also indicate that the physiological forcing of stomatal closure by rising atmospheric CO₂ levels could contribute 20% to the rainfall reduction. Levis et al. 2004 [Ambiguous] ) demonstrate how African rainfall and dynamic vegetation could change each other.

7.2.4.5 Interactive Vegetation Response Variables

7.3.1.1 The Natural Carbon Cycle

Over millions of years, CO₂ is removed from the atmosphere through weathering by silicate rocks and through burial in marine sediments of carbon fixed by marine plants (e.g., ( Berner, 1998 ) ). Burning fossil fuels returns carbon captured by plants in Earth’s geological history to the atmosphere. New ice core records show that the Earth system has not experienced current atmospheric concentrations of CO₂,or indeed of CH₄,for at least 650 kyr – six glacial-interglacial cycles. During that period the atmospheric CO₂ concentration remained between 180 ppm (glacial maxima) and 300 ppm (warm interglacial periods) ( Siegenthaler et al., 2005 [JoC] ). It is generally accepted that during glacial maxima, the CO₂ removed from the atmosphere was stored in the ocean. Several causal mechanisms have been identified that connect astronomical changes, climate, CO₂ and other greenhouse gases, ocean circulation and temperature, biological productivity and nutrient supply, and interaction with ocean sediments (see Box 6.2 ).

Prior to 1750, the atmospheric concentration of CO₂ had been relatively stable between 260 and 280 ppm for 10 kyr( Box 6.2 ). Perturbations of the carbon cycle from human activities were insignificant relative to natural variability. Since 1750, the concentration of CO₂ in the atmosphere has risen, at an increasing rate, from around 280 ppm to nearly 380 ppm in 2005 (see Figure 2.3 and FAQ 2.1 ,Figure 1). The increase in atmospheric CO₂ concentration results from human activities: primarily burning of fossil fuels and deforestation, but also cement production and other changes in land use and management such as biomass burning, crop production and conversion of grasslands to croplands (see FAQ 7.1 ). While human activities contribute to climate change in many direct and indirect ways, CO₂ emissions from human activities are considered the single largest anthropogenic factor contributing to climate change (see FAQ 2.1 ,Figure 2). Atmospheric CH₄ concentrations have similarly experienced a rapid rise from about 700 ppb in 1750 ( Flückiger et al., 2002 [NPR, JoC] ) to about 1,775 ppb in 2005 (see Section 2.3.2 ): sources include fossil fuels, landfills and waste treatment, peatlands/wetlands, ruminant animals and rice paddies. The increase in CH₄ radiative forcing is slightly less than one-third that of CO₂,making it the second most important greenhouse gas (see Chapter2 ). The CH₄ cycle is presented in Section 7.4.1 .

Both CO₂ and CH₄ play roles in the natural cycle of carbon, involving continuous flows of large amounts of carbon among the ocean, the terrestrial biosphere and the atmosphere, that maintained stable atmospheric concentrations of these gases for 10 kyr prior to 1750 . Carbon is converted to plant biomass by photosynthesis. Terrestrial plants capture CO₂ from the atmosphere; plant, soil and animal respiration (including decomposition of dead biomass) returns carbon to the atmosphere as CO₂,or as CH₄ under anaerobic conditions. Vegetation fires can be a significant source of CO₂ and CH₄ to the atmosphere on annual time scales, but much of the CO₂ is recaptured by the terrestrial biosphere on decadal time scales if the vegetation regrows.

Carbon dioxide is continuously exchanged between the atmosphere and the ocean. Carbon dioxide entering the surface ocean immediately reacts with water to form bicarbonate (HCO₃^–)and carbonate (CO₃^2–)ions. Carbon dioxide, HCO₃^– and CO₃^2– are collectively known as dissolved inorganic carbon (DIC). The residence time of CO₂ (as DIC) in the surface ocean, relative to exchange with the atmosphere and physical exchange with the intermediate layers of the ocean below, is less than a decade. In winter, cold waters at high latitudes, heavy and enriched with CO₂ (as DIC) because of their high solubility, sink from the surface layer to the depths of the ocean. This localised sinking, associated with the Meridional Overturning Circulation (MOC; Box 5.1 )is termed the ‘solubility pump’. Over time, it is roughly balanced by a distributed diffuse upward transport of DIC primarily into warm surface waters.

Phytoplankton take up carbon through photosynthesis. Some of that sinks from the surface layer as dead organisms and particles (the ‘biological pump’), or is transformed into dissolved organic carbon (DOC). Most of the carbon in sinking particles is respired (through the action of bacteria) in the surface and intermediate layers and is eventually recirculated to the surface as DIC. The remaining particle flux reaches abyssal depths and a small fraction reaches the deep ocean sediments, some of which is re-suspended and some of which is buried. Intermediate waters mix on a time scale of decades to centuries, while deep waters mix on millennial time scales. Several mixing times are required to bring the full buffering capacity of the ocean into effect (see Section 5.4 for long-term observations of the ocean carbon cycle and their consistency with ocean physics).

Together the solubility and biological pumps maintain a vertical gradient in CO₂ (as DIC) between the surface ocean (low) and the deeper ocean layers (high), and hence regulate exchange of CO₂ between the atmosphere and the ocean. The strength of the solubility pump depends globally on the strength of the MOC, surface ocean temperature, salinity, stratification and ice cover. The efficiency of the biological pump depends on the fraction of photosynthesis exported from the surface ocean as sinking particles, which can be affected by changes in ocean circulation, nutrient supply and plankton community composition and physiology.

Figure 7.3. The global carbon cycle for the 1990 s, showing the main annual fluxes in GtC yr^–1:pre-industrial ‘natural’ fluxes in black and ‘anthropogenic’ fluxes in red (modified from Sarmiento and Gruber, 2006 [NPR, SRC] , with changes in pool sizes from Sabine et al., 2004a [JoC, SRC] ). The net terrestrial loss of –39 GtC is inferred from cumulative fossil fuel emissions minus atmospheric increase minus ocean storage. The loss of –140 GtC from the ‘vegetation, soil and detritus’ compartment represents the cumulative emissions from land use change ( Houghton, 2003 [JoC, MoS] ), and requires a terrestrial biosphere sink of 101 GtC (in Sabine et al., given only as ranges of –140 to –80 GtC and 61 to 141 GtC, respectively; other uncertainties given in their Table 1). Net anthropogenic exchanges with the atmosphere are from Column 5 ‘AR4’ in Table 7.1 .Gross fluxes generally have uncertainties of more than ±20% but fractional amounts have been retained to achieve overall balance when including estimates in fractions of GtC yr^–1 for riverine transport, weathering, deep ocean burial, etc. ‘GPP’ is annual gross (terrestrial) primary production. Atmospheric carbon content and all cumulative fluxes since 1750 are as of end 1994 .

In Figure 7.3 the natural or unperturbed exchanges (estimated to be those prior to 1750 ) among oceans, atmosphere and land are shown by the black arrows. The gross natural fluxes between the terrestrial biosphere and the atmosphere and between the oceans and the atmosphere are (circa 1995 ) about 120 and 90 GtC yr^–1,respectively. Just under 1 GtC yr^–1 of carbon is transported from the land to the oceans via rivers either dissolved or as suspended particles (e.g., Richey, 2004 [NPR] ). While these fluxes vary from year to year, they are approximately in balance when averaged over longer time periods. Additional small natural fluxes that are important on longer geological time scales include conversion of labile organic matter from terrestrial plants into inert organic carbon in soils, rock weathering and sediment accumulation (‘reverse weathering’), and release from volcanic activity. The net fluxes in the 10 kyr prior to 1750, when averaged over decades or longer, are assumed to have been less than about 0.1 GtC yr^–1.For more background on the carbon cycle, see Prentice et al. 2001 [NPR] Field and Raupach 2004 [NPR, SRC] and Sarmiento and Gruber 2006 [NPR, SRC] ).

7.3.1.2 Perturbations of the Natural Carbon Cycle from Human Activities

The additional burden of CO₂ added to the atmosphere by human activities, often referred to as ‘anthropogenic CO₂’ leads to the current ‘perturbed’ global carbon cycle. Figure 7.3 shows that these ‘anthropogenic emissions’ consist of two fractions: (i) CO₂ from fossil fuel burning and cement production, newly released from hundreds of millions of years of geological storage (see Section 2.3 )and (ii) CO₂ from deforestation and agricultural development, which has been stored for decades to centuries. Mass balance estimates and studies with other gases indicate that the net land-atmosphere and ocean-atmosphere fluxes have become significantly different from zero, as indicated by the red arrows in Figure 7.3 (see also Section 7.3.2 ). Although the anthropogenic fluxes of CO₂ between the atmosphere and both the land and ocean are just a few percent of the gross natural fluxes, they have resulted in measurable changes in the carbon content of the reservoirs since pre-industrial times as shown in red. These perturbations to the natural carbon cycle are the dominant driver of climate change because of their persistent effect on the atmosphere. Consistent with the response function to a CO₂ pulse from the Bern Carbon Cycle Model (see footnote (a) of Table 2.14 ), about 50% of an increase in atmospheric CO₂ will be removed within 30 years, a further 30% will be removed within a few centuries and the remaining 20% may remain in the atmosphere for many thousands of years ( Prentice et al., 2001 [NPR] ; Archer, 2005 [JoC, SRC] ; see also Sections 7.3.4.2 and 10.4 )

About 80% of anthropogenic CO₂ emissions during the 1990 s resulted from fossil fuel burning, with about 20% from land use change (primarily deforestation)( Table 7.1 ). Almost 45% of combined anthropogenic CO₂ emissions (fossil fuel plus land use) have remained in the atmosphere. Oceans are estimated to have taken up approximately 30% (about 118 ± 19 GtC: Sabine et al., 2004a [JoC, SRC] ; Figure 7.3 ), an amount that can be accounted for by increased atmospheric concentration of CO₂ without any change in ocean circulation or biology. Terrestrial ecosystems have taken up the rest through growth of replacement vegetation on cleared land, land management practices and the fertilizing effects of elevated CO₂ and N deposition (see Section 7.3.3 ).

Because CO₂ does not limit photosynthesis significantly in the ocean, the biological pump does not take up and store anthropogenic carbon directly. Rather, marine biological cycling of carbon may undergo changes due to high CO₂ concentrations, via feedbacks in response to a changing climate. The speed with which anthropogenic CO₂ is taken up effectively by the ocean, however, depends on how quickly surface waters are transported and mixed into the intermediate and deep layers of the ocean. A considerable amount of anthropogenic CO₂ can be buffered or neutralized by dissolution of CaCO₃ from surface sediments in the deep sea, but this process requires many thousands of years.

The increase in the atmospheric CO₂ concentration relative to the emissions from fossil fuels and cement production only is defined here as the ‘airborne fraction’. ^[2] Land emissions, although significant, are not included in this definition due to the difficulty of quantifying their contribution, and to the complication that much land emission from logging and clearing of forests may be compensated a few years later by uptake associated with regrowth. The ‘airborne fraction of total emissions’ is thus defined as the atmospheric CO₂ increase as a fraction of total anthropogenic CO₂ emissions, including the net land use fluxes. The airborne fraction varies from year to year mainly due to the effect of interannual variability in land uptake (see Section 7.3.2 ).

7.3.1.3 New Developments in Knowledge of the Carbon Cycle Since the Third Assessment Report

Sections 7.3.2 to 7.3.5 describe where knowledge and understanding have advanced significantly since the Third Assessment Report (TAR). In particular, the budget of anthropogenic CO₂ (shown by the red fluxes in Figure 7.3 )can be calculated with improved accuracy. In the ocean, newly available high-quality data on the ocean carbon system have been used to construct robust estimates of the cumulative ocean burden of anthropogenic carbon ( Sabine et al., 2004a [JoC, SRC] ) and associated changes in the carbonate system ( Feely et al., 2004 [JoC, ARC] ). The pH in the surface ocean is decreasing, indicating the need to understand both its interaction with a changing climate and the potential impact on organisms in the ocean (e.g., Orr et al., 2005 [JoC, ARC] ; Royal Society, 2005 [NPR] ). On land, there is a better understanding of the contribution to the buildup of CO₂ in the atmosphere since 1750 associated with land use and of how the land surface and the terrestrial biosphere interact with a changing climate. Globally, inverse techniques used to infer the magnitude and location of major fluxes in the global carbon cycle have continued to mature, reflecting both refinement of the techniques and the availability of new observations. During preparation of the TAR, inclusion of the carbon cycle in climate models was new. Now, results from the firstC⁴MIP are available: when the carbon cycle is included, the models consistently simulate climate feedbacks to land and ocean carbon cycles that tend to reduce uptake of CO₂ by land and ocean from 1850 to 2100 (see Section 7.3.5 ).

7.3.2.1 Atmospheric Increase

The atmospheric CO₂ increase is measured with great accuracy at various monitoring stations (see Chapter2 ;and Keeling and Whorf, 2005 [NPR, ARC] updated by S. Piper through 2006 ). The mean yearly increase in atmospheric CO₂ (the CO₂ ‘growth rate’) is reported in Table 7.1 .Atmospheric CO₂ has continued to increase since the TAR( Figure 7.4 ), and the rate of increase appears to be higher, with the average annual increment rising from 3.2 ± 0.1 GtC yr^–1 in the 1990 s to 4.1 ± 0.1 GtC yr^–1 in the period 2000 to 2005 . The annual increase represents the net effect of several processes that regulate global land-atmosphere and ocean-atmosphere fluxes, examined below. The ‘airborne fraction’ (atmospheric increase in CO₂ concentration/fossil fuel emissions) provides a basic benchmark for assessing short- and long-term changes in these processes. From 1959 to the present, the airborne fraction has averaged 0.55, with remarkably little variation when block-averaged into five-year bins( Figure 7.4 ). Thus, the terrestrial biosphere and the oceans together have consistently removed 45% of fossil CO₂ for the last 45 years, and the recent higher rate of atmospheric CO₂ increase largely reflects increased fossil fuel emissions. Year-to-year fluctuations in the airborne fraction are associated with major climatic events (see Section 7.3.2.4 ). The annual increase in 1998, 2.5 ppm, was the highest ever observed, but the airborne fraction (0.82) was no higher than values observed several times in prior decades. The airborne fraction dropped significantly below the average in the early 1990 s, and preliminary data suggest it may have risen above the average in 2000 to 2005 .

Figure 7.4. Changes in global atmospheric CO₂ concentrations. (a) Annual (bars) and five-year mean (lower black line) changes in global CO₂ concentrations, from Scripps Institution of Oceanography observations (mean of South Pole and Mauna Loa; Keeling and Whorf, 2005 [NPR, ARC] , updated). The upper stepped line shows annual increases that would occur if 100% of fossil fuel emissions ( Marland et al., 2006 [NPR] , updated as described in Chapter2 )remained in the atmosphere, and the red line shows five-year mean annual increases from National Oceanic and Atmospheric Administration (NOAA) data (mean of Samoa and Mauna Loa; Tans and Conway, 2005 [NPR, ARC] , updated). (b) Fraction of fossil fuel emissions remaining in the atmosphere (‘airborne fraction’) each year (bars), and five-year means (solid black line) (Scripps data) (mean since 1958 is 0.55). Note the anomalously low airborne fraction in the early 1990 s.

The inter-hemispheric gradient of CO₂ provides additional evidence that the increase in atmospheric CO₂ is caused primarily by NH sources. The excess atmospheric CO₂ in the NH compared with the Southern Hemisphere (SH), ΔCO₂^N-S,has increased in proportion to fossil fuel emission rates (which are predominantly in the NH) at about 0.5 ppm per (GtC yr^–1)( Figure 7.5 ). The intercept of the best-fit line indicates that, without anthropogenic emissions, atmospheric CO₂ would be 0.8 ppm higher in the SH than in the NH, presumably due to transport of CO₂ by the ocean circulation. The consistency of the airborne fraction and the relationship between

ΔCO₂^N-S and fossil fuel emissions suggest broad consistency in the functioning of the carbon cycle over the period. There are interannual fluctuations in ΔCO₂^N-S as large as ±0.4 ppm, at least some of which may be attributed to changes in atmospheric circulation ( Dargaville et al., 2000 [JoC, MoS] ), while others may be due to shifts in sources and sinks, such as large forest fires.

Figure 7.5. The difference between CO₂ concentration in the NH and SH (y axis), computed as the difference between annual mean concentrations (ppm) at Mauna Loa and the South Pole ( Keeling and Whorf, 2005 [NPR, ARC] , updated), compared with annual fossil fuel emissions (x axis; GtC; Marland, et al., 2006 [NPR] ), with a line showing the best fit. The observations show that the north-south difference in CO₂ increases proportionally with fossil fuel use, verifying the global impact of human-caused emissions.

7.3.2.2 Uptake of CO₂ by Natural Reservoirs and Global Carbon Budget

7.3.2.2.1 Ocean-atmosphere flux

To assess the mean ocean sink, seven methods have been used. The methods are based on: (1) observations of the partial pressure of CO₂ at the ocean surface and gas-exchange estimates ( (Takahashi et al., 2002 ) ); (2) atmospheric inversions based upon diverse observations of atmospheric CO₂ and atmospheric transport modelling (see Section 7.2.3.4 ); (3) observations of carbon, oxygen, nutrients and chlorofluorocarbons (CFCs) in seawater, from which the concentration of anthropogenic CO₂ is estimated ( Sabine et al., 2004a [JoC, SRC] ) combined with estimates of oceanic transport ( Gloor et al., 2003 [JoC, MoS] ; Mikaloff Fletcher et al., 2006 [NPR, JoC, MoS] ); (4) estimates of the distribution of water age based on CFC observations combined with the atmospheric CO₂ history ( McNeil et al., 2003 [JoC] ); (5) the simultaneous observations of the increase in atmospheric CO₂ and decrease in atmosphericO₂ ( Manning and Keeling, 2006 [JoC, ARC] ); (6) various methods using observations of change in ¹³Cin the atmosphere ( Ciais et al., 1995 [JoC, SRC] ) or the oceans ( Gruber and Keeling, 2001 [JoC, MoS, SRC] ; Quay et al., 2003 [JoC] ); and (7) ocean General Circulation Models ( Orr et al., 2001 [JoC, MoS, ARC] ). The ocean uptake estimates obtained with methods (1) and (2) include in part a flux component due to the outgassing of river-supplied inorganic and organic carbon ( Sarmiento and Sundquist, 1992 [JoC] ). The magnitude of this necessary correction to obtain the oceanic uptake flux of anthropogenic CO₂ is not well known, as these estimates pertain to the open ocean, whereas a substantial fraction of the river-induced outgassing likely occurs in coastal regions. These estimates of the net oceanic sink are shown in Figure 7.3 .

With these corrections, estimates from all methods are consistent, resulting in a well-constrained global oceanic sink for anthropogenic CO₂ (see Table 7.1 ). The uncertainty around the different estimates is more difficult to judge and varies considerably with the method. Four estimates appear better constrained than the others. The estimate for the ocean uptake of atmospheric CO₂ of –2.2 ± 0.5 GtC yr^–1 centred around 1998 based on the atmosphericO₂/N₂ ratio needs to be corrected for the oceanicO₂ changes ( Manning and Keeling, 2006 [JoC, ARC] ). The estimate of –2.0 ± 0.4 GtC yr^–1 centred around 1995 based on CFC observations provides a constraint from observed physical transport in the ocean. These estimates of the ocean sink are shown in Figure 7.6 .The mean estimates of –2.2 ± 0.25 and –2.2 ± 0.2 GtC yr^–1 centred around 1995 and 1994 provide constraints based on a large number of ocean carbon observations. These well-constrained estimates all point to a decadal mean ocean CO₂ sink of –2.2 ± 0.4 GtC yr^–1 centred around 1996, where the uncertainty is the root mean square of all errors. See Section 5.4 for a discussion of changes in the ocean CO₂ sink.

Figure 7.6. Individual estimates of the ocean-atmosphere flux reported in Chapter5 and of the related land-atmosphere flux required to close the global carbon budget. The dark thick lines are the revised budget estimates in the AR4 for the 1980 s, the 1990 s and the early 2000 s, respectively.

7.3.2.3 Regional Fluxes

Quantifying present-day regional carbon sources and sinks and understanding the underlying carbon mechanisms are needed to inform policy decisions. Furthermore, by analysing spatial and temporal detail, mechanisms can be isolated.

7.3.2.3.1 The top-down view: atmospheric inversions

The atmosphere mixes and integrates surface fluxes that vary spatially and temporally. The distribution of regional fluxes over land and oceans can be retrieved using observations of atmospheric CO₂ and related tracers within models of atmospheric transport. This is called the ‘top-down’ approach to estimating fluxes. Atmospheric inversions belong to that approach, and determine an optimal set of fluxes that minimise the mismatch between modelled and observed concentrations, accounting for measurement and model errors. Fossil fuel emissions have small uncertainties that are often ignored and, when considered (e.g., Enting et al., 1995 [JoC] ; Rodenbeck et al., 2003a [JoC, SRC] ), are found to have little influence on the inversion. Fossil fuel emissions are generally considered perfectly known in inversions, so that their effect can be easily modelled and subtracted from atmospheric CO₂ data to solve for regional land-atmosphere and ocean-atmosphere fluxes, although making such an assumption biases the results ( Gurney et al., 2005 [JoC, SRC] ). Input data for inversions come from a global network of about 100 CO₂ concentration measurement sites, ^[3] with mostly discrete flask sampling, and a smaller number of in situ continuous measurement sites. Generally, regional fluxes derived from inverse models have smaller uncertainties upwind of regions with denser data coverage. Measurement and modelling errors and uneven and sparse coverage of the network generate random errors in inversion results. In addition, inverse methodological details, such as the choice of transport model, can introduce systematic errors. A number of new inversion ensembles, with different methodological details, have been produced since the TAR ( Gurney et al., 2003 [JoC, SRC] ; Rödenbeck et al., 2003a [JoC, SRC] ,b; Peylin et al., 2005 [JoC, SRC] ; Baker et al., 2006 [JoC, MoS, SRC] ). Generally, confidence in the long-term mean inverted regional fluxes is lower than confidence in the year-to-year anomalies (see Section 7.3.2.4 ). For individual regions, continents or ocean basins, the errors of inversions increase and the significance can be lost. Because of this, Figure 7.7 reports the oceans and land fluxes aggregated into large latitude bands, as well as a breakdown of five land and ocean regions in the NH, which is constrained by denser atmospheric stations. Both random and systematic errors are reported in Figure 7.7 .

Figure 7.7. Regional ocean-atmosphere and land-atmosphere CO₂ fluxes for the NH (top) and the globe (bottom) from inversion ensembles and bottom-up studies. Fluxes to the atmosphere are positive and uptake has a negative sign. Inversion results all correspond to the post-Pinatubo period 1992 to 1996 . Orange: Bottom-up terrestrial fluxes from Pacala et al. 2001 [JoC, MoS] ( and Kurz and Apps 1999 ) ) for North America, from Janssens et al. 2003 [JoC] ) for Europe and from Shvidenko and Nilsson 2003 [JoC] ) plus Fang et al. 2001 [JoC] ) for North Asia (Asian Russia and China). Cyan (filled circles): Bottom-up ocean flux estimates from ( Takahashi, et al. 2002 ) ). Blue: ocean fluxes from atmospheric inversions. Green: terrestrial fluxes from inversion models. Magenta: total inversion fluxes. Red: fossil fuel emissions. The mean flux of different inversion ensembles is reported. Inversion errors for regional fluxes are not reported here; their values usually range between 0.5 and 1 GtC yr–1. Error bar: range of atmospheric inversion fluxes from the TAR. Squares: Gurney et al. 2002 [JoC, MoS, SRC] ) inversions using annual mean CO₂ observations and 16 transport models. Circles: Gurney et al. 2003 [JoC, SRC] ) inversions using monthly CO₂ observations and 13 transport models. Triangles: Peylin et al. 2005 [JoC, SRC] ) inversions with three transport models, three regional breakdowns and three inversion settings. Inverted triangles: Rödenbeck et al. 2003a [JoC, SRC] ) inversions where the fluxes are solved on the model grid using monthly flask data.

7.3.2.3.4 Robust findings of regional ocean-atmosphere flux

• The regional air-sea CO₂ fluxes consist of a superposition of natural and anthropogenic CO₂ fluxes, with the former being globally nearly balanced (except for a small net outgassing associated with the input of carbon by rivers), and the latter having a global integral uptake of 2.2 ± 0.5 GtC yr^–1 (see Table 7.1 ).

• The tropical oceans are outgassing CO₂ to the atmosphere (see Figure 7.8 ), with a mean flux of the order of 0.7 GtC yr^–1,estimated from an oceanic inversion ( Gloor et al., 2003 [JoC, MoS] ), in good agreement with atmospheric inversions (0 to 1.5 GtC yr^–1), and estimates based on oceanic pCO₂ observations (0.8 GtC yr^–1( ;Takahashi et al., 2002 ) ).

Figure 7.8. Estimates (4° × 5°) of sea-to-air flux of CO₂,computed using 940,000 measurements of surface water pCO₂ collected since 1956 and averaged monthly, together with NCEP/NCAR 41-year mean monthly wind speeds and a (10-m wind speed)² dependence on the gas transfer rate Wanninkhof, dependence on the gas transfer rate Wanninkhof,). The fluxes were normalised to the year 1995 using techniques described in ( Takahashi et al. 2002 ) ), who used wind speeds taken at the 0.995 standard deviation level (about 40 m above the sea surface). The annual flux of CO₂ for 1995 with 10-m winds is –1.6 GtC yr^–1,with an approximate uncertainty (see Footnote 1) of ±1 GtC yr^–1,mainly due to uncertainty in the gas exchange velocity and limited data coverage. This estimated global flux consists of an uptake of anthropogenic CO₂ of –2.2 GtC yr^–1 (see text) plus an outgassing of 0.6 GtC yr^–1,corresponding primarily to oxidation of organic carbon borne by rivers (Figure 7.3). The monthly flux values with 10-m winds used here are available from T. Takahashi at http://www.ldeo.columbia.edu/res/pi/CO2/carbondioxide/pages/air_sea_flux_rev1.html.

• The extratropical NH ocean is a net sink for anthropogenic and natural CO₂,with a magnitude of the order of 1.2 GtC yr^–1,consistent among various estimates.

• The Southern Ocean is a large sink of atmospheric CO₂( (Takahashi et al., 2002; ) Gurney et al., 2002 [JoC, MoS, SRC] ) and of anthropogenic CO₂ ( Gloor et al., 2003 [JoC, MoS] ; Mikaloff Fletcher et al., 2006 [NPR, JoC, MoS] ). Its magnitude has been estimated to be about 1.5 GtC yr^–1.This estimate is consistent among the different methods at the scale of the entire Southern Ocean. However, differences persist with regard to the Southern Ocean flux distribution between subpolar and polar latitudes (T. Roy et al., 2003 [Ambiguous] ). Atmospheric inversions and oceanic inversions indicate a larger sink in subpolar regions ( Gurney et al., 2002 [JoC, MoS, SRC] ; Gloor et al., 2003 [JoC, MoS] ), consistent with the distribution of CO₂ fluxes based on available ∆pCO₂ observations( Figure 7.8 ( and Takahashi, 2002 ) ).

7.3.2.4 Interannual Changes in the Carbon Cycle

7.3.2.4.2 Interannual variability in regional fluxes, atmospheric inversions and bottom-up models

Year-to-year flux anomalies can be more robustly inferred by atmospheric inversions than mean fluxes. Yet, at the scale of continents or ocean basins, the inversion errors increase and the statistical significance of the inferred regional fluxes decreases. ^[5] This is why Figure 7.9 shows the land-atmosphere and ocean-atmosphere flux anomalies over broad latitude bands only for the inversion ensembles of Baker et al. 2006 [JoC, MoS, SRC] Bousquet et al. 2000 [JoC, SRC] and Rödenbeck et al. 2003b [SRC] ). An important finding of these studies is that differences in transport models have little impact on the interannual variability of fluxes. Interannual variability of global land-atmosphere fluxes (±4 GtC yr^–1 between extremes) is larger than that of air-sea fluxes and dominates the global fluxes. This result is also true over large latitude bands( Figure 7.9 ). Tropical land fluxes exhibit on average a larger variability than temperate and boreal fluxes. Inversions give tropical land flux anomalies of the order of ±1.5 to 2 GtC yr^–1,which compare well in timing and magnitude with terrestrial model results ( Tian et al., 1998 [JoC] ; Peylin et al., 2005 [JoC, SRC] ; Zeng et al., 2005 [JoC] ). In these studies, enhanced sources occur during El Niño episodes and abnormal sinks during la Niña. In addition to the influence of these climate variations on ecosystem processes ( Gérard et al., 1999 [JoC] ; C. Jones et al., 2001 [Ambiguous] ), regional droughts during El Niño events promote large biomass fires, which appear to contribute to high CO₂ growth rates during the El Niño episodes ( Barbosa et al., 1999 [JoC] ; Langenfelds et al., 2002 [JoC] ; Page et al., 2002 [JoC] ;( van der Werf et al., 2003, ) 2004; Patra et al., 2005 [JoC, MoS] ).

Inversions robustly attribute little variability to ocean-atmosphere CO₂ flux (±0.5 GtC yr^–1 between extremes), except for the recent work of Patra et al. 2005 [JoC, MoS] ). This is in agreement with ocean model and ocean observations ( Lee et al., 1998 [JoC] ; le Quéré et al., 2003 [JoC, SRC] ; Obata and Kitamura, 2003 [JoC, MoS] ; McKinley et al., 2004b [JoC, MoS] ). However, inversions and ocean models differ on the dominant geographic contributions to the variability. Inversions estimate similar variability in both hemispheres, whereas ocean models estimate more variability in the Southern Ocean ( Bousquet et al., 2000 [JoC, SRC] ; Rödenbeck et al., 2003b [SRC] ; Baker et al., 2006 [JoC, MoS, SRC] ). Over the North Atlantic, Gruber et al. 2002 [JoC, SRC] ) suggest a regional CO₂ flux variability (extremes of ±0.3 GtC yr^–1)by extrapolating data from a single ocean station, but McKinley et al. 2004a [JoC] ,b) model a small variability (extremes of ±0.1 GtC yr^–1). The equatorial Pacific is the ocean region of the world where the variability is constrained with repeated ∆pCO₂ observations (variations of about ±0.4 GtC yr^–1 ;Feely et al., 2002 [ARC] ), with a reduced source of CO₂ during El Niño associated with decreased upwelling of CO₂-rich waters. Over this region, some inversion results (e.g., Bousquet et al., 2000 [JoC, SRC] ) compare well in magnitude and timing with ocean and coupled model results ( Le Quéré et al., 2000 [JoC, SRC] ; C. Jones et al., 2001 [Ambiguous] ; McKinley et al., 2004a [JoC] ,b) and with ∆pCO₂ observations ( Feely et al., 1999 [JoC, ARC] , 2002 ).

Figure 7.9. Year-to-year anomalies in ocean-atmosphere and land-atmosphere CO₂ fluxes (GtC yr^–1)from interannual inversion ensembles covering the past 20 years or so, grouped into large latitude bands, and over the globe. Three different inversion ensembles from Bousquet et al. 2000 [JoC, SRC] Rödenbeck et al. 2003a [JoC, SRC] and Baker et al. 2006 [JoC, MoS, SRC] ) are shown. For each flux and each region, the anomalies were obtained by subtracting the long-term mean flux and removing the seasonal signal. Grey shaded regions indicate El Niño episodes, and the black bars indicate the cooling period following the Mt. Pinatubo eruption.

7.3.3.1 Processes Driven by Climate, Atmospheric Composition and Land Use Change

7.3.4.1 Overview of the Ocean Carbon Cycle

Oceanic carbon exists in several forms: as DIC, DOC, and particulate organic carbon (POC) (living and dead) in an approximate ratio DIC:DOC:POC = 2000 :38:1 (about 37,000 GtC DIC: Falkowski et al., 2000 [JoC] and Sarmiento and Gruber, 2006 [NPR, SRC] ; 685 GtC DOC: Hansell and Carlson, 1998 [JoC] ; and 13 to 23 GtC POC: Eglinton and Repeta, 2004 [NPR] ). Before the industrial revolution, the ocean contained about 60 times as much carbon as the atmosphere and 20 times as much carbon as the terrestrial biosphere/soil compartment.

Seawater can, through inorganic processes, absorb large amounts of CO₂ from the atmosphere, because CO₂ is a weakly acidic gas and the minerals dissolved in the ocean have over geologic time created a slightly alkaline ocean (surface pH 7.9 to 8.25: Degens et al., 1984 [NPR] ; Royal Society, 2005 [NPR] ). The air-sea exchange of CO₂ is determined largely by the air-sea gradient in pCO₂ between atmosphere and ocean. Equilibration of surface ocean and atmosphere occurs on a time scale of roughly one year. Gas exchange rates increase with wind speed ( Wanninkhof and McGillis, 1999 [JoC, ARC] ; Nightingale et al., 2000 [JoC] ) and depend on other factors such as precipitation, heat flux, sea ice and surfactants. The magnitudes and uncertainties in local gas exchange rates are maximal at high wind speeds. In contrast, the equilibrium values for partitioning of CO₂ between air and seawater and associated seawater pH values are well established ( Zeebe and Wolf-Gladrow, 2001 [NPR] ; see Box 7.3 ).

Box 7.3: Marine Carbon Chemistry and Ocean Acidification

The marine carbonate buffer system allows the ocean to take up CO₂ far in excess of its potential uptake capacity based on solubility alone, and in doing so controls the pH of the ocean. This control is achieved by a series of reactions that transform carbon added as CO₂ into HCO₃^– and CO₃^2–.These three dissolved forms (collectively known as DIC) are found in the approximate ratio CO₂:HCO₃^–:CO₃^2– of 1:100:10 (Equation (7.1)). CO₂ is a weak acid and when it dissolves, it reacts with water to form carbonic acid, which dissociates into a hydrogen ion (H⁺)and a HCO₃^– ion, with some of theH⁺ then reacting with CO₃^2– to form a second HCO₃^– ion (Equation (7.2)).

CO2 +H2O→H+ + HCO3– → 2H+ + CO32–	(7.1)
CO2 +H2O+ CO32– → HCO3– +H+ + CO32– → 2HCO3–	(7.2)

Therefore, the net result of adding CO₂ to seawater is an increase inH⁺ and HCO₃^–,but a reduction in CO₃^2–.The decrease in the CO₃^2– ion reduces the overall buffering capacity as CO₂ increases, with the result that proportionally moreH⁺ ions remain in solution and increase acidity.

This ocean acidification is leading to a decrease in the saturation state of CaCO₃ in the ocean. Two primary effects are expected: (1) the biological production of corals as well as calcifying phytoplankton and zooplankton within the water column may be inhibited or slowed down (Royal Society, 2005 [NPR] ), and (2) the dissolution of CaCO₃ at the ocean floor will be enhanced ( Archer, 2005 [JoC, SRC] ). Aragonite, the meta-stable form of CaCO₃ produced by corals and pteropods (planktonic snails; Lalli and Gilmer, 1989 [NPR] ), will be particularly susceptible to a pH reduction ( Kleypas et al., 1999b [JoC] ; Hughes et al., 2003 [JoC] ; Orr et al., 2005 [JoC, ARC] ). Laboratory experiments under high ambient CO₂ with the coccolithophore species Emiliania huxleyi and Gephyrocapsa oceanica produce a significant reduction in CaCO₃ production and a stimulation of POC production ( Riebesell et al., 2000 [JoC, SRC] ; Zondervan et al., 2001 [JoC, SRC] ). Other species and growth under other conditions may show different responses, so that no conclusive quantification of the CaCO₃ feedback is possible at present ( (Tortell et al., 2002; ) ( Sciandra et al., 2003 ) ).

The sinking speed of marine particle aggregates depends on their composition: CaCO₃ may act as an efficient ballast component, leading to high sinking speeds of aggregates ( Armstrong et al., 2002 [MoS] ; Klaas and Archer, 2002 [JoC] ). The relatively small negative feedback of reduced CaCO₃ production to atmospheric pCO₂ may be compensated for by a change in the ballast for settling biogenic particles and the associated shallowing of re-mineralization depth levels in the water column for organic carbon ( Heinze, 2004 [JoC, MoS, SRC] ). On the other hand, production of extracellular organic carbon could increase under high CO₂ levels and lead to an increase in export ( Engel et al., 2004 [JoC] ).

Ecological changes due to expected ocean acidification may be severe for corals in tropical and cold waters ( (Gattuso et al., 1999; ) ( Kleypas et al., 1999a; ) Langdon et al., 2003 [JoC] ; Buddemeier et al., 2004 [NPR] ; Roberts et al., 2006 [JoC] ) and for pelagic ecosystems ( (Tortell et al., 2002; ) Royal Society, 2005 [NPR] ). Acidification can influence the marine food web at higher trophic levels ( (Langenbuch and Pörtner, 2003; ) ( Ishimatsu et al., 2004 ) ).

Since the beginning of the industrial revolution, sea surface pH has dropped by about 0.1 pH units (corresponding to a 30% increase in the H ion concentration). The expected continued decrease may lead within a few centuries to an ocean pH estimated to have occurred most recently a few hundred million years before present ( Caldeira and Wickett, 2003 [JoC] ; Key et al., 2004 [JoC] ; Box 7.3 ,Figure 1).

According to a model experiment based on the IPCC Scenarios 1992a (IS92a) emission scenario, bio-calcification will be reduced by 2100, in particular within the Southern Ocean ( Orr et al., 2005 [JoC, ARC] ), and by 2050 for aragonite-producing organisms (see also Figure 10.24 ). It is important to note that ocean acidification is not a direct consequence of climate change but a consequence of fossil fuel CO₂ emissions, which are the main driver of the anticipated climate change

Box 7.3, Figure 1. (a) Atmospheric CO₂ emissions, historical atmospheric CO₂ levels and predicted CO₂ concentrations from the given emission time series, together with changes in ocean pH based on horizontally averaged chemistry. The emission time series is based on the mid-range IS92a emission scenario (solid line) prior to 2100 and then assumes that emissions continue until fossil fuel reserves decline. (b) Estimated maximum change in surface ocean pH as a function of final atmospheric CO₂ pressure, and the transition time over which this CO₂ pressure is linearly approached from 280 ppm. A: Glacial-interglacial CO₂ changes; B: slow changes over the past 300 Myr; C: historical changes in ocean surface waters; D: unabated fossil fuel burning over the next few centuries. Source: Caldeira and Wickett 2003 [JoC] ). Reprinted with permission from Macmillan Publishers Ltd: Nature, Caldeira and Wickett 2003 [JoC] ), copyright ( 2003 ).

In addition to changes in advection and mixing, the ocean can alter atmospheric CO₂ concentration through three mechanisms ( Volk and Hoffert, 1985 [NPR] ), illustrated in Figure 7.10 :(1) absorption or release of CO₂ due to changes in solubility of gaseous CO₂ (‘solubility pump’); (2) changes in carbon fixation to POC in surface waters by photosynthesis and export of this carbon through sinking of organic particles out of the surface layer (‘organic carbon pump’) – this process is limited to first order by availability of light and nutrients (phosphate, nitrate, silicic acid and micronutrients such as iron); and (3) changes in the release of CO₂ in surface waters during formation of CaCO₃ shell material by plankton (‘CaCO₃ counter pump’).

Figure 7.10. Three main ocean carbon pumps govern the regulation of natural atmospheric CO₂ changes by the ocean ( Heinze et al., 1991 [JoC, MoS, SRC] ): the solubility pump, the organic carbon pump and the CaCO₃ ‘counter pump’. The oceanic uptake of anthropogenic CO₂ is dominated by inorganic carbon uptake at the ocean surface and physical transport of anthropogenic carbon from the surface to deeper layers. For a constant ocean circulation, to first order, the biological carbon pumps remain unaffected because nutrient cycling does not change. If the ocean circulation slows down, anthropogenic carbon uptake is dominated by inorganic buffering and physical transport as before, but the marine particle flux can reach greater depths if its sinking speed does not change, leading to a biologically induced negative feedback that is expected to be smaller than the positive feedback associated with a slower physical downward mixing of anthropogenic carbon. Reprinted with permission, copyright 1991 American Geophysical Union.

Organic particles are re-mineralized (oxidized to DIC and other inorganic compounds through the action of bacteria) primarily in the upper 1,000 m of the oceanic water column, with an accompanying decrease in dissolved O. On the average, CaCO₃ particles sink deeper before they undergo dissolution: deep waters are undersaturated with respect to CaCO₃.The remainder of the particle flux enters marine sediments and is subject to either re-dissolution within the water column or accumulation within the sediments. Although the POC reservoir is small, it plays an important role in keeping DIC concentrations low in surface waters and high in deep waters. The loop is closed through the three-dimensional ocean circulation: upwelling water brings inorganic carbon and nutrients to the surface again, leading to outgassing and biogenic particle production. Dissolved organic carbon enters the ocean water column from rivers and marine metabolic processes. A large fraction of DOC has a long ocean residence time (1–10 kyr), while other fractions are more short-lived (days to hundreds of years; Loh et al., 2004 [JoC] ). The composition of dissolved organic matter is still largely unknown.

In conjunction with the global ocean mixing or overturning time of the order of 1 kyr ( Broecker and Peng, 1982 [NPR, ARC] ), small changes in the large ocean carbon reservoir can induce significant changes in atmospheric CO₂ concentration. Likewise, perturbations in the atmospheric pCO₂ can be buffered by the ocean. Glacial-interglacial changes in the atmospheric CO₂ content can potentially be attributed to a change in functioning of the marine carbon pump (see Chapter6 ). The key role for the timing of the anthropogenic carbon uptake by the ocean is played by the downward transport of surface water, with a high burden of anthropogenic carbon, into the ocean’s interior. The organic carbon cycle and the CaCO₃ counter pump modulate, but do not dominate, the net marine uptake of anthropogenic carbon.

7.3.4.2 Carbon Cycle Feedbacks to Changes in Atmospheric Carbon Dioxide

Chemical buffering of anthropogenic CO₂ is the quantitatively most important oceanic process acting as a carbon sink. Carbon dioxide entering the ocean is buffered due to scavenging by the CO₃^2– ions and conversion to HCO₃^–,that is, the resulting increase in gaseous seawater CO₂ concentration is smaller than the amount of CO₂ added per unit of seawater volume. Carbon dioxide buffering in seawater is quantified by the Revelle factor (‘buffer factor’, Equation (7.3)), relating the fractional change in seawater pCO₂ to the fractional change in total DIC after re-equilibration ( Revelle and Suess, 1957 [JoC] ; Zeebe and Wolf-Gladrow, 2001 [NPR] ):

Revelle factor (or buffer factor) = (Δ[CO₂]/ [CO₂]) / (Δ[DIC] / [DIC]) (7.3)

The lower the Revelle factor, the larger the buffer capacity of seawater. Variability of the buffer factor in the ocean depends mainly on changes in pCO₂ and the ratio of DIC to total alkalinity. In the present-day ocean, the buffer factor varies between 8 and 13 ( Sabine et al., 2004a [JoC, SRC] ; Figure 7.11 ). With respect to atmospheric pCO₂ alone, the inorganic carbon system of the ocean reacts in two ways: (1) seawater re-equilibrates, buffering a significant amount of CO₂ from the atmosphere depending on the water volume exposed to equilibration; and (2) the Revelle factor increases with pCO₂ (positive feedback; Figure 7.11 ). Both processes are quantitatively important. While the first is generally considered as a system response, the latter is a feedback process.

The ocean will become less alkaline (seawater pH will decrease) due to CO₂ uptake from the atmosphere (see Box 7.3 ). The ocean’s capacity to buffer increasing atmospheric CO₂ will decline in the future as ocean surface pCO₂ increases( Figure 7.11 a). This anticipated change is certain, with potentially severe consequences.

Figure 7.11. (a) The Revelle factor (or buffer factor) as a function of CO₂ partial pressure (for temperature 25°C, salinity 35 psu, and total alkalinity 2,300 µmol kg^–1 )(Zeebe and Wolf-Gladrow, 2001 [NPR] , page 73; reprinted with permission, copyright 2001 Elsevier). (b) The geographical distribution of the buffer factor in ocean surface waters in 1994 ( Sabine et al., 2004a [JoC, SRC] ; reprinted with permission, copyright 2004 American Association for the Advancement of Science). High values indicate a low buffer capacity of the surface waters.

Increased carbon storage in the deep ocean leads to the dissolution of calcareous sediments below their saturation depth ( Broecker and Takahashi, 1978 [NPR] ; Feely et al., 2004 [JoC, ARC] ). The feedback of CaCO₃ sediment dissolution to atmospheric pCO₂ increase is negative and quantitatively significant on a 1 to 100 kyr time scale, where CaCO₃ dissolution will account for a 60 to 70% absorption of the anthropogenic CO₂ emissions, while the ocean water column will account for 22 to 33% on a time scale of 0.1 to 1 kyr. In addition, the remaining 7 to 8% may be compensated by long-term terrestrial weathering cycles involving silicate carbonates ( Archer et al., 1998 [JoC, SRC] ). Due to the slow CaCO₃ buffering mechanism (and the slow silicate weathering), atmospheric pCO₂ will approach a new equilibrium asymptotically only after several tens of thousands of years ( Archer, 2005 [JoC, SRC] ; Figure 7.12 ).

Figure 7.12. Model projections of the neutralization of anthropogenic CO₂ for an ocean-only model, a model including dissolution of CaCO₃ sediment and a model including weathering of silicate rocks, (top) for a total of 1,000 GtC of anthropogenic CO₂ emissions and (bottom) for a total of 5,000 GtC of anthropogenic CO₂.Note that the y-axis is different for the two diagrams. Without CaCO₃ dissolution from the seafloor, the buffering of anthropogenic CO₂ is limited. Even after 100 kyr, the remaining pCO₂ is substantially higher than the pre-industrial value. Source: Archer 2005 [JoC, SRC] ).

Elevated ambient CO₂ levels appear to also influence the production rate of POC by marine calcifying planktonic organisms (e.g., Zondervan et al., 2001 [JoC, SRC] ). This increased carbon fixation under higher CO₂ levels was also observed for three diatom (siliceous phytoplankton) species ( Riebesell et al., 1993 [JoC, SRC] ). It is critical to know whether these increased carbon fixation rates translate into increased export production rates (i.e., removal of carbon to greater depths). Studies of the nutrient to carbon ratio in marine phytoplankton have not yet shown any significant changes related to CO₂ concentration of the nutrient utilisation efficiency (expressed through the ‘Redfield ratio’ – carbon:nitrogen:phosphorus:silicon) in organic tissue ( Burkhardt et al., 1999 [SRC] ).

7.3.4.3 Carbon Cycle Feedbacks to Changes in Physical Forcing

Amore sluggish ocean circulation and increased density stratification, both expected in a warmer climate, would slow down the vertical transport of carbon, alkalinity and nutrients, and the replenishment of the ocean surface with water that has not yet been in contact with anthropogenic CO₂.This narrowing of the ‘bottleneck’ for anthropogenic CO₂ invasion into the ocean would provide a significant positive feedback to atmospheric greenhouse gas concentrations ( Bolin and Eriksson, 1959 [NPR] ; see also the carbon cycle climate model simulations by Cox et al., 2000 [JoC, MoS, SRC] ; Friedlingstein et al., 2001 [JoC, MoS, SRC] , 2006 ). As long as the vertical transfer rates for marine biogenic particles remain unchanged, in a more sluggish ocean the biological carbon pump will be more efficient ( Boyle, 1988 [JoC] ; Heinze et al., 1991 [JoC, MoS, SRC] ), thus inducing a negative feedback, which is expected to be smaller than the physical transport feedback ( Broecker, 1991 [JoC] ; Maier-Reimer et al., 1996 [JoC, MoS] ; Plattner et al., 2001 [PoC, JoC, ARC] ; see Figure 7.10 ). However, a modelling study by Bopp et al. 2005 [JoC, MoS] ) predicts a decrease in vertical particle transfer and hence shallower depths of re-mineralization of particulate organic carbon resulting in a positive CO₂ feedback. Further changes in plankton community structure including the role ofN₂-fixing organisms can feed back to the carbon cycle ( Sarmiento et al., 2004 [JoC] ;( Mahaffey et al., 2005 ) ). Changes in ocean circulation can affect the regional circulation of shelf and coastal seas, leading either to increased export of nutrients plus carbon from the shallow seas into the open ocean or to increased upwelling of nutrients plus carbon onto the shelf and towards coastal areas ( Walsh, 1991 [JoC] ; Smith and Hollibaugh, 1993 [JoC] ; Chen et al., 2003 [NPR] ;( Borges, 2005 ) ). A reduction in sea ice cover may increase the uptake area for anthropogenic CO₂ and act as a minor negative carbon feedback ( ACIA, 2005 [NPR] ). The physical ‘bottleneck’ feedback dominates over biological feedbacks induced by circulation change, resulting in an anticipated overall positive feedback to climate change. Both feedbacks depend on details of the future ocean circulation and model projections show a large range.

The solubility of CO₂ gas in seawater and the two dissociation constants of carbonic acid in seawater depend on temperature and salinity ( Weiss, 1974 [ARC] ;( Millero et al., 2002 ) ). A 1°C increase in sea surface temperature produces an increase in pCO₂ of 6.9 to 10.2 ppm after 100 to 1,000 years ( Heinze et al., 2003 [JoC, SRC] ; see also Broecker and Peng, 1986 [ARC] ; Plattner et al., 2001 [PoC, JoC, ARC] ). Warming may increase the biological uptake rate of nutrients and carbon from surface waters, but the net effect on export and DIC is uncertain. Laws et al. 2000 [JoC] ) proposed that export efficiency increases with net photosynthesis at low temperatures, which implies a positive feedback to warming. In addition, DOC may be degraded more quickly at higher temperatures.

7.3.4.4 Carbon Cycle Feedbacks Induced by Nutrient Cycling and Land Ocean Coupling

Rivers deliver carbon (DIC, DOC) and nutrients to the ocean. Rising CO₂ levels in the atmosphere and land use may lead to increased chemical and physical weathering, resulting in increased carbon and alkalinity loads in rivers ( Clair et al., 1999 [JoC] ;( Hejzlar et al., 2003; ) Raymond and Cole, 2003 [JoC] ; Freeman et al., 2004 [JoC] ). Depending on the lithology and soil composition of the catchment areas, increased levels of alkalinity, DIC or DOC can lead to local positive or negative feedbacks. Mobilisation of silicate carbonates from soils and transfer to the ocean would lead to a negative feedback to atmospheric CO₂ on long time scales ( (Dupre et al., 2003 ) ). Variations in nutrient supply can lead to species shifts and to deviations from the large-scale average Redfield ratios mainly in coastal waters, but also in the open ocean ( Pahlow and Riebesell, 2000 [JoC, SRC] ). Nutrient supply to the ocean has been changed through increased nitrate release from land due to fertilizer use as well as nitrogen deposition from the atmosphere in highly polluted areas ( (De Leeuw et al., 2001; ) ( Green et al., 2004 ) ).

Dust deposition to the ocean provides an important source of micronutrients (iron, zinc and others, e.g., ( Frew et al., 2001; ) Boyd et al., 2004 [JoC] ) and ballast material to the ocean. Areas where iron is not supplied by aeolian dust transport in sufficient amounts tend to be iron-limited. A warmer climate may result on the average in a decrease of dust mobilisation and transport ( Werner et al., 2002 [JoC] ; Mahowald and Luo, 2003 [JoC, MoS] ) although increased dust loads may result as well due to changes in land use ( Tegen et al., 2004 [JoC, MoS] ) and in vegetation cover ( Woodward et al., 2005 [JoC, MoS, SRC] ). A decrease in dust loads could result in a net positive feedback, further increasing CO₂ through a weakening of marine biological production and export of aggregates due to clay ballast ( (Haake and Ittekkot, 1990; ) Ittekkot, 1993 [JoC] ). Changes in plankton species composition and regional shifts of high production zones due to a changing climate could lead to a series of further feedbacks. Light absorption due to changes in bio-optical heating may change and induce a respective temperature change in ocean surface water ( Sathyendranath et al., 1991 [JoC] ; Wetzel et al., 2006 [JoC, MoS] ). An increase in blooms involving calcifying organisms as indicated for the high northern latitudes ( (Broerse et al., 2003; ) Smyth et al., 2004 [JoC] ) can temporarily increase surface ocean albedo, though the effect on the radiation budget is small (Tyrell et al., 1999 ).

7.3.4.5 Summary of Marine Carbon Cycle Climate Couplings

Couplings between the marine carbon cycle and climate are summarised in Table 7.3 and below.

7.3.5.1 Introduction

Atmospheric CO₂ is increasing at only about half the rate implied by fossil fuel plus land use emissions, with the remainder being taken up by the ocean, and vegetation and soil on land. Therefore, the land and ocean carbon cycles are currently helping to mitigate CO₂-induced climate change. However, these carbon cycle processes are also sensitive to climate. The glacial-interglacial cycles are an example of tight coupling between climate and the carbon cycle over long time scales, but there is also clear evidence of the carbon cycle responding to short-term climatic anomalies such as the El Niño-Southern Oscillation (ENSO) and Arctic Oscillation ( Rayner et al., 1999 [JoC, SRC] ; Bousquet et al., 2000 [JoC, SRC] ; C. Jones et al., 2001 [Ambiguous] ; Lintner, 2002 [JoC] ; Russell and Wallace, 2004 [JoC] ) and the climate perturbation arising from the Mt. Pinatubo volcanic eruption ( Jones and Cox, 2001a [JoC, MoS, SRC] ; Lucht et al., 2002 [JoC] ; Angert et al., 2004 [JoC] ).

Previous IPCC reports have used simplified or ‘reduced-form’ models to estimate the impact of climate change on the carbon cycle. However, detailed climate projections carried out with Atmosphere-Ocean General Circulation Models (AOGCMs) have typically used a prescribed CO₂ concentration scenario, neglecting two-way coupling between climate and the carbon cycle. This section discusses the first generation of coupled climate-carbon cycle AOGCM simulations, using the results to highlight a number of critical issues in the interaction between climate change and the carbon cycle.

7.3.5.2 Coupled Climate-Carbon Cycle Projections

The TAR reported two initial climate projections using AOGCMs with interactive carbon cycles. Both indicated positive feedback due largely to the impacts of climate warming on land carbon storage ( Cox et al., 2000 [JoC, MoS, SRC] ; Friedlingstein et al., 2001 [JoC, MoS, SRC] ), but the magnitude of the feedback varied markedly between the models ( Friedlingstein et al., 2003 [JoC, SRC] ). Since the TAR a number of other climate modelling groups have completed climate-carbon cycle projections ( Brovkin et al., 2004 [ARC] ; Thompson et al., 2004 [JoC, MoS] ; N. Zeng et al., 2004 [Ambiguous] , Fung et al., 2005 [JoC, SRC] ; Kawamiya et al., 2005 [MoS, SRC] ; Matthews et al., 2005 [JoC, MoS, ARC] ; Sitch et al., 2005 [JoC, MoS] ) as part ofC⁴MIP. The 11 models involved inC⁴MIP differ in the complexity of their components ( Friedlingstein et al., 2006 [JoC, MoS, SRC] ), including both Earth System Models of Intermediate Complexity and AOGCMs.

The models were forced by historical and Special Report on Emission Scenarios (SRES; IPCC, 2000 [NPR] ) A2 anthropogenic CO₂ emissions for the 1850 to 2100 time period. Each modelling group carried out at least two simulations: one ‘coupled’ in which climate change affects the carbon cycle, and one ‘uncoupled’ in which atmospheric CO₂ increases do not influence climate (so that the carbon cycle experiences no CO₂-induced climate change). A comparison of the runs defines the climate-carbon cycle feedback, quantified by the feedback factor: F = ∆C_A^c /∆C_A^u ,where ∆C_A^c is the change in CO₂ in the coupled run, and ∆C_A^u is the change in CO₂ in the uncoupled run. All of the elevenC⁴MIP models produce a positive climate-carbon cycle feedback, but with feedback factors varying from 1.04 (Model E) to 1.44 (Model A). This translates into an additional CO₂ concentration of between 20 and 224 ppm by 2100, with a mean of 87 ppm( Table 7.4 ).

AllC⁴MIP models predict that an increasing fraction of total anthropogenic CO₂ emissions will remain airborne through the 21st century. Figure 7.13 shows the simulated partitioning of anthropogenic CO₂ for the entire simulation period to 2100 from each of the coupled models, and compares this with the partitioning simulated by the same models over the historical period to 1999 . The dashed box shows observational constraints on the historical CO₂ partitioning, based on estimates of changes in ocean carbon storage ( Sabine et al., 2004a [JoC, SRC] ) and total anthropogenic CO₂ emissions. The area of this box is largely due to uncertainties in the net land use emissions. The majority of the models sit within or very close to the historical constraints, but they differ in the magnitude of the changes projected for the 21st century. However, all models produce an increase in the fraction of total emissions that remain in the atmosphere, and most also indicate a decline in the fraction of emissions absorbed by the ocean (9 out of 11 models) and the land (10 out of 11 models). In the case of the oceanic uptake, this is largely a consequence of the reduced buffering capacity as CO₂ increases, and therefore also occurs in the uncoupledC⁴MIP models.

Figure 7.13. Predicted increase in the fraction of total emissions that add to atmospheric CO₂.Changes in the mean partitioning of emissions as simulated by theC⁴MIP models up to 2000 (black symbols) and for the entire simulation period to 2100 (red symbols). The letters represent the models as given in Table 7.4 .The box shown by the dotted line is a constraint on the historical carbon balance based on records of atmospheric CO₂ increase, and estimates of total emissions (fossil fuel plus land use emissions) and the oceanic uptake of anthropogenic CO₂ ( Sabine et al., 2004a [JoC, SRC] ). The black and red diamonds show the model-mean carbon partitioning for the historical period and the entire simulation period, respectively. The red line shows the mean tendency towards an increasing airborne fraction through the 21st century, which is common to all models.

7.3.5.3 Sensitivity Analysis

The coupled and uncoupled model experiments can be used to separate the effects of climate change and CO₂ increase on land and ocean carbon storage ( Friedlingstein et al., 2003 [JoC, SRC] ). Table 7.4 also shows the linear sensitivity parameters diagnosed from each of theC⁴MIP models ( Friedlingstein et al., 2006 [JoC, MoS, SRC] ).

7.3.5.4 Summary of Coupling Between the Carbon Cycle and Climate

7.3.5.4.2 Key uncertainties

TheC⁴MIP models also exhibit uncertainties in the evolution of atmospheric CO₂ for a given anthropogenic emissions scenario. Figure 7.14 shows how uncertainties in the sensitivities of ocean and land carbon processes contribute to uncertainties in the fraction of emissions that remain in the atmosphere. The confidence limits were produced by spanning the range of sensitivities diagnosed from the 11C⁴MIP models (Tables 7.4 and 7.5 ). In the absence of climate change effects (lowest three bars), models simulate increased uptake by ocean and land (primarily as a result of CO₂ enhancement of NPP), with a slight offset of the land uptake by enhancement of the specific heterotrophic respiration rate (see Section 7.3.5.3.2 ). However, there is a wide range of response to CO₂,even in the absence of climate change effects on the carbon cycle. Climate change increases the fraction of emissions that remain airborne by suppressing ocean uptake, enhancing soil respiration and reducing plant NPP. The sensitivity of NPP to climate change is especially uncertain because it depends on changing soil water availability, which varies significantly between General Circulation Models (GCMs), with some models suggesting major drying and reduced productivity in tropical ecosystems ( Cox et al., 2004 [MoS, SRC] ). The transient climate sensitivity to CO₂ is also a major contributor to the overall uncertainty in the climate-carbon cycle feedback (top bar).

Other potentially important climate-carbon cycle interactions were not included in these first generationC⁴MIP experiments. The ocean ecosystem models used inC⁴MIP are at an early stage of development. These models have simple representations of the biological fluxes, which include the fundamental response to changes in internal nutrients, temperature and light availability, but for most models do not include the more complex responses to changes in ecosystem structure. Changes in ecosystem structure can occur when specific organisms respond to surface warming, acidification, changes in nutrient ratios resulting from changes in external sources of nutrients (atmosphere or rivers) and changes in upper trophic levels (fisheries). Shifts in the structure of ocean ecosystems can influence the rate of CO₂ uptake by the ocean ( Bopp et al., 2005 [JoC, MoS] ).

Figure 7.14. Uncertainties in carbon cycle feedbacks estimated from analysis of the results from theC⁴MIP models. Each effect is given in terms of its impact on the mean airborne fraction over the simulation period (typically 1860 to 2100 ), with bars showing the uncertainty range based on the ranges of effective sensitivity parameters given in Tables 7.4 and 7.5 .The lower three bars are the direct response to increasing atmospheric CO₂ (see Section 7.3.5 for details), the middle four bars show the impacts of climate change on the carbon cycle, and the top black bar shows the range of climate-carbon cycle feedbacks given by theC⁴MIP models.

The first-generationC⁴MIP models also currently exclude, by design, the effects of forest fires and prior land use change. Forest regrowth may account for a large part of the land carbon sink in some regions (e.g., Pacala et al., 2001 [JoC, MoS] ; Schimel et al., 2001 [JoC] ; Hurtt et al., 2002 [JoC, MoS] ; Sitch et al., 2005 [JoC, MoS] ), while combustion of vegetation and soil organic matter may be responsible for a significant fraction of the interannual variability in CO₂ ( Cochrane, 2003 [JoC] ;( Nepstad et al., 2004; ) Kasischke et al., 2005 [JoC] ; Randerson et al., 2005 [JoC] ). Other important processes were excluded in part because modelling these processes is even less straightforward. Among these are N cycling on the land (which could enhance or suppress CO₂ uptake by plants) and the impacts of increasing ozone concentrations on plants (which could suppress CO₂ uptake).

7.4.1.1 Biogeochemistry and Budgets of Methane

Atmospheric CH₄ originates from both non-biogenic and biogenic sources. Non-biogenic CH₄ includes emissions from fossil fuel mining and burning (natural gas, petroleum and coal), biomass burning, waste treatment and geological sources (fossil CH₄ from natural gas seepage in sedimentary basins and geothermal/volcanic CH₄). However, emissions from biogenic sources account for more than 70% of the global total. These sources include wetlands, rice agriculture, livestock, landfills, forests, oceans and termites. Emissions of CH₄ from most of these sources involve ecosystem processes that result from complex sequences of events beginning with primary fermentation of organic macromolecules to acetic acid (CH₃COOH), other carboxylic acids, alcohols, CO₂ and hydrogen (H₂), followed by secondary fermentation of the alcohols and carboxylic acids to acetate,H₂ and CO₂,which are finally converted to CH₄ by the so-called methanogenic Archaea: CH₃COOH → CH₄ +CO₂ and CO₂ +4H₂ → CH₄ +2H₂O(Conrad, 1996 ). Alternatively, CH₄ sources can be divided into anthropogenic and natural. The anthropogenic sources include rice agriculture, livestock, landfills and waste treatment, some biomass burning, and fossil fuel combustion. Natural CH₄ is emitted from sources such as wetlands, oceans, forests, fire, termites and geological sources( Table 7.6 ).

The net rate of CH₄ emissions is generally estimated from three approaches: (1) extrapolation from direct flux measurements and observations, (2) process-based modelling (bottom-up approach) and (3) inverse modelling that relies on spatially distributed, temporally continuous observations of concentration, and in some cases isotopic composition in the atmosphere (top-down approach). The top-down method also includes aircraft and satellite observations ( Xiao et al., 2004 [NPR, JoC] ; Frankenberg et al., 2005 [JoC] , 2006 ). When the bottom-up approach is used to extrapolate the emissions to larger scales, uncertainty results from the inherent large temporal and spatial variations of fluxes and the limited range of observational conditions. The top-down approach helps to overcome the weaknesses in bottom-up methods. However, obstacles to extensive application of the top-down approach include inadequate observations, and insufficient capabilities of the models to account for error amplification in the inversion process and to simulate complex topography and meteorology ( Dentener et al., 2003a [JoC, SRC] ; Mikaloff Fletcher et al., 2004a [JoC, MoS] , 2004b; Chen and Prinn, 2005 [JoC, MoS, ARC] , 2006 ). Measurements of isotopes of CH₄ (¹³C, ¹⁴C, and ²H) provide additional constraints on CH₄ budgets and specific sources, but such data are even more limited ( Bergamaschi et al., 2000 [JoC] ; Lassey et al., 2000 [JoC, SRC] ; Mikaloff Fletcher et al., 2004a [JoC, MoS] , 2004b ).

Since the TAR, availability of new data from various measurement networks and from national reporting documents has enabled re-estimates of CH₄ source magnitudes and insights into individual source strengths. Total global pre-industrial emissions of CH₄ are estimated to be 200 to 250 Tg(CH₄)yr^–1 ( Chappellaz et al., 1993 [JoC, SRC] ; Etheridge et al., 1998 [JoC, ARC] ; Houweling et al., 2000 [JoC, MoS, SRC] ; Ferretti et al., 2005 [JoC] ; Valdes et al., 2005 [NPR, JoC] ). Of this, natural CH₄ sources emitted between 190 and 220 Tg(CH₄)yr^–1,and anthropogenic sources (rice agriculture, livestock, biomass burning and waste) accounted for the rest ( Houweling et al., 2000 [JoC, MoS, SRC] ; Ruddiman and Thomson, 2001 [JoC] ). In contrast, anthropogenic emissions dominate present-day CH₄ budgets, accounting for more than 60% of the total global budget( Table 7.6 ).

The single largest CH₄ source is natural wetlands. Recent estimates combine bottom-up and top-down fluxes, and global observations of atmospheric CH₄ concentrations in a three-dimensional Atmospheric Transport and Chemical Model (ATCM) simulation ( Chen and Prinn, 2005 [JoC, MoS, ARC] , 2006 ). In these estimates, southern and tropical regions account for more than 70% of total global wetland emissions. Other top-down studies that include both direct observations and ¹³C/¹²Cratios of CH₄ suggest greater emissions in tropical regions compared with previously estimates ( Mikaloff Fletcher et al., 2004a [JoC, MoS] , 2004b; Xiao et al., 2004 [NPR, JoC] ; Frankenberg et al., 2006 [NPR, JoC] ). However, several bottom-up studies indicate fewer emissions from tropical rice agriculture ( Li et al., 2002 [NPR, JoC] ; Yan et al., 2003 [MoS] ;( Khalil and Shearer 2006 ) Frankenberg et al. 2005 [JoC] , 2006 and Keppler et al. 2006 [JoC] ) suggest that tropical trees emit CH₄ via an unidentified process. The first estimate of this source was 10 to 30% (62–236 Tg(CH₄)yr^–1)of the global total, but ( Kirschbaum et al. 2006 ) ) revise this estimate downwards to 10 to 60 Tg(CH₄)yr^–1.Representative ¹³C/¹²Cratios (δ¹³Cvalues) of CH₄ emitted from individual sources are included in Table 7.6 .Due to isotope fractionation associated with CH₄ production and consumption processes, CH₄ emitted from each source exhibits a measurably different δ¹³Cvalue. Therefore, it is possible, using mixing models, to constrain further the sources of atmospheric CH₄.

Geological sources of CH₄ are not included in Table 7.6 .However, several studies suggest that significant amounts of CH₄,produced within the Earth’s crust (mainly by bacterial and thermogenic processes), are released into the atmosphere through faults and fractured rocks, mud volcanoes on land and the seafloor, submarine gas seepage, microseepage over dry lands and geothermal seeps ( (Etiope and Klusman, 2002; ) ( Etiope, 2004; ) ( Kvenvolden and Rogers, 2005 ) ). Emissions from these sources are estimated to be as large as 40 to 60 Tg(CH₄)yr^–1.

The major CH₄ sinks are oxidation by OH in the troposphere, biological CH₄ oxidation in drier soil, and loss to the stratosphere( Table 7.6 ). Oxidation by chlorine (Cl) atoms in the marine atmospheric boundary layer is suggested as an additional sink for CH₄,possibly constituting an additional loss of about 19 Tg(CH₄)yr^–1 ( Gupta et al., 1997 [JoC] ; Tyler et al., 2000 [JoC] ; Platt et al., 2004 [ARC] ; Allan et al., 2005 [JoC] ). However, the decline in the growth rate of atmospheric CH₄ concentration since the TAR shows no clear correlation with change in sink strengths over the same period ( Prinn et al., 2001 [JoC, ARC] , 2005; Allan et al., 2005 [JoC] ). This trend has continued since 1993, and the reduction in the CH₄ growth rate has been suggested to be a consequence of source stabilisation and the approach of the global CH₄ budget towards steady state ( Dlugokencky et al., 1998 [JoC, SRC] , 2003 ). Thus, total emissions are likely not increasing but partitioning among the different sources may have changed (see Section 2.3 ). Consequently, in the Fourth Assessment Report (AR4) the sink strength is treated as in the TAR (576 Tg(CH₄)yr^–1). However, the AR4 estimate has been increased by 1% (to 581 Tg(CH₄)yr^–1)to take into account the recalibration of the CH₄ scale explained in Chapter2 .The main difference between TAR and AR4 estimates is the source-sink imbalance inferred from the annual increment in concentration. The TAR used 8 ppb yr^–1 for a period centred on 1998 when there was clearly an anomalously high growth rate. The present assessment uses 0.2 ppb yr^–1,the average over 2000 to 2005 (see Section 2.3 and Figure 2.4 ). Thus, using the CH₄ growth rate for a single anomalous year, as in the TAR, gives an anomalously high top-down value relative to the longer-term average source. For a conversion factor of 2.78 Tg(CH₄)per ppb and an atmospheric concentration of 1,774 ppb, the atmospheric burden of CH₄ in 2005 was 4,932 Tg, with an annual average increase ( 2000 – 2005 ) of about 0.6 Tg yr^–1.Total average annual emissions during the period considered here are approximately 582 Tg(CH₄)yr^–1.

Uncertainty in this estimate may arise from several sources. Uncertainty in the atmospheric concentration measurement, given in Chapter2 as 1,774 ± 1.8 ppb in 2005, is small (about 0.1%). Uncertainty ranges for individual sink estimates are ±103 Tg(CH₄)(20%), ±15 Tg(CH₄)(50%), ±8 Tg(CH₄)(20%) for OH, soil and stratospheric loss, respectively (as reported in the Second Assessment Report). The use of a different lifetime for CH₄ (8.7 ± 1.3 years) leads to an uncertainty in overall sink strength of ±15%. Thus, the top-down method used in AR4 is constrained mainly by uncertainty in sink estimates and the choice of lifetime used in the mass balance calculation.

7.4.1.2 Effects of Climate

Effects of climate on CH₄ biogeochemistry are investigated by examining records of the past and from model simulations under various climate change scenarios. Ice core records going back 650 ka ( Petit et al., 1999 [JoC] ; Spahni et al., 2005 [JoC, ARC] ) reveal that the atmospheric concentration of CH₄ is closely tied to atmospheric temperature, falling and rising in phase with temperature at the inception and termination of glacial episodes ( (Wuebbles and Hayhoe, 2002 ) Brook et al. 2000 [JoC] ) show that, following each transition, temperature increased more rapidly than CH₄ concentration. Since biogenic CH₄ production and emission from major sources (wetlands, landfills, rice agriculture and biomass burning) are influenced by climate variables such as temperature and moisture, the effect of climate on emissions from these sources is significant.

Several studies indicate a high sensitivity of wetland CH₄ emissions to temperature and water table. Before the 1990 s, elevated surface temperature and emissions from wetlands were believed to contribute to the increase in global CH₄ emissions ( Walter and Heimann, 2001a [JoC, MoS, SRC] ,b; Christensen et al., 2003 [JoC] ; Zhuang et al., 2004 [NPR, JoC, MoS] ). Observations indicate substantial increases in CH₄ released from northern peatlands that are experiencing permafrost melt ( Christensen et al., 2004 [NPR, JoC] ; Wickland et al., 2006 [JoC] ). Based on the relationship between emissions and temperature at two wetland sites in Scotland, ( Chapman and Thurlow 1996 ) ) predicted that CH₄ emissions would increase by 17, 30 and 60% for warmings of 1.5°C, 2.5°C and 4.5°C (warming above the site’s mean temperature during 1951 to 1980 ), respectively. A model simulation by ( Cao et al. 1998 ) ) yielded a 19% emission increase under a uniform 2°C warming. The combined effects of a 2°C warming and a 10% increase in precipitation yielded an increase of 21% in emissions. In most cases, the net emission depends on how an increase in temperature affects net ecosystem production (NEP), as this is the source of methanogenic substrates ( Christensen et al., 2003 [JoC] ), and on the moisture regime of wetlands, which determines if decomposition is aerobic or anaerobic. Emissions increase under a scenario where an increase in temperature is associated with increases in precipitation and NEP, but emissions decrease if elevated temperature results in either reduced precipitation or reduced NEP.

For a doubling in atmospheric CO₂ concentration, the GCM of Shindell et al. 2004 [JoC] ) simulates a 3.4°C warming. Changes in the hydrological cycle due to this CO₂ doubling cause CH₄ emissions from wetlands to increase by 78%. Gedney et al. 2004 [JoC, SRC] ) also simulate an increase in CH₄ emissions from northern wetlands due to an increase in wetland area and an increase in CH₄ production due to higher temperatures. Zhuang et al. 2004 [NPR, JoC, MoS] ) use a terrestrial ecosystem model based on emission data for the 1990 s to study how rates of CH₄ emission and consumption in high-latitude soils of the NH (north of 45°N) have changed over the past century ( 1900 – 2000 ) in response to observed change in the region’s climate. They estimate that average net emissions of CH₄ increased by 0.08 Tg yr^–1 over the 20th century. Their decadal net CH₄ emission rate correlates with soil temperature and water table depth.

In rice agriculture, climate factors that will likely influence CH₄ emission are those associated with plant growth. Plant growth controls net emissions by determining how much substrate will be available for either methanogenesis or methanotrophy ( Matthews and Wassmann, 2003 [MoS] ( Sass et al. 2002 ) ) show that CH₄ emissions correlate strongly with plant growth (height) in a Texas rice field. Any climate change scenario that results in an increase in plant biomass in rice agriculture is likely to increase CH₄ emissions ( (Xu et al., 2004 ) ). However, the magnitude of increased emission depends largely on water management. For example, field drainage could significantly reduce emission due to aeration of the soil (i.e., influx of air into anaerobic zones that subsequently suppresses methanogenesis, Li et al., 2002 [NPR, JoC] ).

Past observations indicate large interannual variations in CH₄ growth rates ( Dlugokencky et al., 2001 [JoC, ARC] ). The mechanisms causing these variations are poorly understood and the role of climate is not well known. Emissions from wetlands and biomass burning may have contributed to emission peaks in 1993 to 1994 and 1997 to 1998 ( Langenfelds et al., 2002 [JoC] ; Butler et al., 2004 [MoS, SRC] ). Unusually warm and dry conditions in the NH during ENSO periods increase biomass burning. Kasischke and Bruhwiler 2002 [JoC] ) attribute CH₄ releases of 3 to 5 Tg in 1998 to boreal forest fires in Eastern Siberia resulting from unusually warm and dry conditions.

Meteorological conditions can affect global mean removal rates ( Warwick et al., 2002 [JoC] ; Dentener et al., 2003a [JoC, SRC] ). Dentener et al. find that over the period 1979 to 1993, the primary effect resulted from changes in OH distribution caused by variations in tropical tropospheric water vapour. Johnson et al. 2001 [JoC, MoS, SRC] ) studied predictions of the CH₄ evolution over the 21st century and found that there is also a substantial increase in CH₄ destruction due to increases in the CH₄ + OH rate coefficient in a warming climate. There also appear to be significant interannual variations in the active Cl sink, but a climate influence has yet to be identified ( Allan et al., 2005 [JoC] ). On the other hand, several model studies indicate that CH₄ oxidation in soil is relatively insensitive to temperature increase (Ridgwell at al., 1999; Zhuang et al., 2004 [NPR, JoC, MoS] ). A doubling of atmospheric CO₂ would likely change the sink strength only marginally (in the range of –1 to +3 Tg(CH₄)yr^–1 ;Ridgwell et al., 1999 [JoC, MoS] ). However, any change in climate that alters the amount and pattern of precipitation may significantly affect the CH₄ oxidation capacity of soils. A process-based model simulation indicated that CH₄ oxidation strongly depends on soil gas diffusivity, which is a function of soil bulk density and soil moisture content ( Bogner et al., 2000 [JoC, MoS] ; del Grosso et al., 2000 [JoC, MoS] ).

Climate also affects the stability of CH₄ hydrates beneath the ocean, where large amounts of CH₄ are stored (~4 ×10⁶ Tg; Buffett and Archer, 2004 [SRC] ). The δ¹³Cvalues of ancient seafloor carbonates reveal several hydrate dissociation events that appear to have occurred in connection with rapid warming episodes in the Earth’s history ( Dickens et al., 1997 [MoS] ; Dickens, 2001 [NPR, MoS] ). Model results indicate that these hydrate decomposition events occurred too fast to be controlled by the propagation of the temperature change into the sediments ( Katz et al., 1999 [JoC] ;( Paull et al., 2003 ) ). Additional studies infer other indirect and inherently more rapid mechanisms such as enhanced migration of free gas, or reordering of gas hydrates due to slump slides ( Hesselbo et al., 2000 [JoC] ;( Jahren et al., 2001; ) ( Kirschvink et al., 2003; ) ( Ryskin, 2003 ) ). Recent modelling suggests that today’s seafloor CH₄ inventory would be diminished by 85% with a warming of bottom water temperatures by 3°C ( Buffett and Archer, 2004 [SRC] ). Based on this inventory, the time-dependent feedback of hydrate destabilisation to global warming has been addressed using different assumptions for the time constant of destabilisation: an anthropogenic release of 2,000 GtC to the atmosphere could cause an additional release of CH₄ from gas hydrates of a similar magnitude (∼2,000 Gt(CH₄)) over a period of 1 to 100 kyr ( Archer and Buffett, 2005 [MoS, SRC] ). Thus, gas hydrate decomposition represents an important positive CH₄ feedback to be considered in global warming scenarios on longer time scales.

In summary, advances have been made since the TAR in constraining estimates of CH₄ source strengths and in understanding emission variations. These improvements are attributed to increasing availability of worldwide observations and improved modelling techniques. Emissions from anthropogenic sources remain the major contributor to atmospheric CH₄ budgets. Global emissions are likely not to have increased since the time of the TAR, as nearly zero growth rates in atmospheric CH₄ concentrations have been observed with no significant change in the sink strengths.

7.4.4.1 Present-Day Budgets of Ozone and its Precursors

Tropospheric ozone is (after CO₂ and CH₄)the third most important contributor to greenhouse radiative forcing. Trends over the 20th century are discussed in Chapter2 .Ozone is produced in the troposphere by photochemical oxidation of CO, CH₄ and non-methane VOCs (NMVOCs) in the presence of NO_x.Stratosphere-troposphere exchange (STE) is another source of ozone to the troposphere. Loss of tropospheric ozone takes place through chemical reactions and dry deposition. Understanding of tropospheric ozone and its relationship to sources requires three-dimensional tropospheric chemistry models that describe the complex nonlinear chemistry involved and its coupling to transport.

The past decade has seen considerable development in global models of tropospheric ozone, and the current generation of models can reproduce most climatological features of ozone observations. The TAR reported global tropospheric ozone budgets from 11 models in the 1996 to 2000 literature. Table 7.9 presents an update to the post- 2000 literature, including a recent intercomparison of 25 models ( Stevenson et al., 2006 [JoC, MoS, SRC] ). Models concur that chemical production and loss are the principal terms in the global budget. Although STE is only a minor term in the global budget, it delivers ozone to the upper troposphere where its lifetime is particularly long (about one month, limited by transport to the lower troposphere) and where it is of most importance from a radiative forcing perspective.

The post- 2000 model budgets in Table 7.9 show major differences relative to the older generation TAR models: on average a 34% weaker STE, a 35% stronger chemical production, a 10% larger tropospheric ozone burden, a 16% higher deposition velocity and a 10% shorter chemical lifetime. It is now well established that many of the older studies overestimated STE, as observational constraints in the lower stratosphere impose an STE ozone flux of 540 ± 140 Tg yr^–1 ( Gettelman et al., 1997 [JoC, ARC] ; Olsen et al., 2001 [JoC, ARC] ). Overestimation of the STE flux appears to be most serious in models using assimilated meteorological data, due to the effect of assimilation on vertical motions ( Douglass et al., 2003 [JoC, MoS] ; Schoeberl et al., 2003 [JoC, MoS] ; Tan et al., 2004 [JoC, ARC] ; van Noije et al., 2004 [JoC, MoS, SRC] ). The newer models correct for this effect by using dynamic flux boundary conditions in the tropopause region ( McLinden et al., 2000 [JoC, MoS] ) or by relaxing model results to observed climatology ( Horowitz et al., 2003 [JoC, MoS] ). Such corrections, although matching the global STE flux constraints, may still induce errors in the location of the transport ( Hudman et al., 2004 [JoC] ) with implications for the degree of stratospheric influence on tropospheric concentrations ( Fusco and Logan, 2003 [JoC, MoS] ).

The faster chemical production and loss of ozone in the current generation of models could reflect improved treatment of NMVOC sources and chemistry ( Houweling et al., 1998 [JoC, SRC] ), ultraviolet (UV) actinic fluxes ( Bey et al., 2001 [JoC, MoS] ) and deep convection ( Horowitz et al., 2003 [JoC, MoS] ), as well as higher NO_x emissions ( Stevenson et al., 2006 [JoC, MoS, SRC] ). Subtracting ozone chemical production and loss terms in Table 7.9 indicates that the current generation of models has net production of ozone in the troposphere, while the TAR models had net loss, reflecting the decrease in STE. Net production is not a useful quantity in analysing the ozone budget because (1) it represents only a small residual between production and loss and (2) it reflects a balance between STE and dry deposition, both of which are usually parametrized in models.

Detailed budgets of ozone precursors were presented in the TAR. The most important precursors are CH₄ and NO_x ( Wang et al., 1998 [JoC, MoS, SRC] ;( Grenfell et al., 2003; ) Dentener et al., 2005 [MoS, SRC] ). Methane is in general not simulated explicitly in ozone models and is instead constrained from observations. Nitrogen oxides are explicitly simulated and proper representation of sources and chemistry is critical for the ozone simulation. The lightning source is particularly uncertain ( Nesbitt et al., 2000 [JoC, MoS] ; Tie et al., 2002 [SRC] ), yet is of great importance because of the high production efficiency of ozone in the tropical upper troposphere. The range of the global lightning NO_x source presently used in models (3–7 TgN yr^–1)is adjusted to match atmospheric observations of ozone and NO_x,although large model uncertainties in deep convection and lightning vertical distributions detract from the strength of this constraint. Process-based models tend to predict higher lightning emissions (5–20 TgN yr^–1 ;Price et al., 1997 [JoC, ARC] ).

Other significant precursors for tropospheric ozone are CO and NMVOCs, the most important of which is biogenic isoprene. Satellite measurements of CO from the Measurements of Pollution in the Troposphere (MOPITT) instrument launched in 1999 ( Edwards et al., 2004 [JoC] ) have provided important new constraints for CO emissions, pointing in particular to an underestimate of Asian sources in current inventories ( Kasibhatla et al., 2002 [JoC, MoS] ; Arellano et al., 2004 [JoC, MoS] ; Heald et al., 2004 [JoC, MoS] ; Petron et al., 2004 [JoC] ), as confirmed also by aircraft observations of Asian outflow ( Palmer et al., 2003a [JoC] ; Allen et al., 2004 [JoC, MoS] ). Satellite measurements of formaldehyde columns from the GOME instrument ( Chance et al., 2000 [JoC] ) have been used to place independent constraints on isoprene emissions and indicate values generally consistent with current inventories, although with significant regional discrepancies ( Palmer et al., 2003b [JoC] ; Shim et al., 2005 [JoC] ).

Afew recent studies have examined the effect of aerosols on global tropospheric ozone involving both heterogeneous chemistry and perturbations to actinic fluxes. Jacob 2000 [SRC] ) reviewed the heterogeneous chemistry involved. Hydrolysis of dinitrogen pentoxide (N₂O₅)in aerosols is a well-known sink for NO_x,but other processes involving reactive uptake of the hydroperoxyl radical (HO₂), NO₂ and ozone itself could also be significant. Martin et al. 2003b [JoC] ) find that including these processes along with effects of aerosols on UV radiation in a global Chemical Transport Model (CTM) reduced ozone production rates by 6% globally, with larger effects over aerosol source regions.

Although the current generation of tropospheric ozone models is generally successful in describing the principal features of the present-day global ozone distribution, there is much less confidence in the ability to reproduce the changes in ozone associated with perturbations of emissions or climate. There are major discrepancies with observed long-term trends in ozone concentrations over the 20th century ( Hauglustaine and Brasseur, 2001 [JoC, MoS, SRC] ; Mickley et al., 2001 [JoC, MoS, SRC] ; Shindell and Favulegi, 2002; Shindell et al., 2003 [MoS] ; Lamarque et al., 2005c [JoC] ), including after 1970 when the reliability of observed ozone trends is high ( Fusco and Logan, 2003 [JoC, MoS] ). Resolving these discrepancies is needed to establish confidence in the models.

7.4.4.2 Effects of Climate Change

Climate change can affect tropospheric ozone by modifying emissions of precursors, chemistry, transport and removal (European Commission, 2003 ). These and other effects are discussed below. They could represent positive or negative feedbacks to climate change.

7.4.5.1 Changes in the Hydroxyl Radical Over Time

7.4.5.2 Consequences for Lifetimes

7.4.6.1 Interactions

Stratospheric ozone is affected by climate change through changes in dynamics and in the chemical composition of the troposphere and stratosphere. An increase in the concentrations of LLGHGs, especially CO₂,cools the stratosphere, allowing the possibility of more PSCs, and alters the ozone distribution ( Rosenlof et al., 2001 [JoC, ARC] ; Rosenfield et al., 2002 [JoC] ; Randel et al., 2004 [JoC] , 2006; Fueglistaler and Haynes, 2005 [JoC] ). With the possible exception of the polar lower stratosphere, a decrease in temperature reduces ozone depletion leading to higher ozone column amounts and a positive correction to the LLGHG-induced radiative cooling of the stratosphere. Moreover, ozone itself is a greenhouse gas and absorbs UV radiation in the stratosphere. Absorption of UV radiation provides the heating responsible for the observed temperature increase with height above the tropopause. Changes in stratospheric temperatures, induced by changes in ozone or LLGHG concentration, alter the Brewer-Dobson circulation ( Butchart and Scaife, 2001 [JoC] ; Butchart et al., 2006 [NPR, JoC, MoS] ), controlling the rate at which long-lived molecules, such as LLGHGs, CFCs, HCFCs and halogens are transported from the troposphere to various levels in the stratosphere. Furthermore, increases in the Brewer-Dobson circulation increase temperatures adiabatically in the polar regions and decrease temperatures adiabatically in the tropics.

Climate is affected by changes in stratospheric ozone, which radiates infrared radiation down to the troposphere. For a given percentage change in the vertical structure of ozone, the largest dependence of the radiative forcing is in the upper troposphere and ozone layer regions (e.g., TAR, Figure 6.1). Past ozone depletion has induced surface cooling( Chapter2 ). The observed decrease in stratospheric ozone and the resultant increase in UV irradiance (e.g., Zerefos et al., 1998 [JoC] ; McKenzie et al., 1999 [JoC, ARC] ) have affected the biosphere and biogenic emissions ( Larsen, 2005 [JoC] ). Such UV radiation increases lead to an enhanced OH production, reducing the lifetime of CH₄ and influencing tropospheric ozone, both important greenhouse gases (European Commission, 2003 ). In addition to global mean equilibrium surface temperature changes, local surface temperature changes have been identified by Gillett and Thompson 2003 [JoC, MoS, ARC] ) as a result of ozone loss from the lower stratosphere. Observational (e.g., Baldwin and Dunkerton, 1999 [JoC, ARC] , 2001; Thompson et al., 2005 [PoC, JoC, ARC] ) and modelling ( Polvani and Kushner, 2002 [NPR, JoC, MoS, ARC] ; Norton, 2003 [JoC, MoS] ; Song and Robinson, 2004 [JoC, ARC] ; Thompson et al., 2005 [PoC, JoC, ARC] ) evidence exists for month-to-month changes to the stratospheric flow feedback to the troposphere, affecting its circulation. Model results show that trends in the SH stratosphere can affect high-latitude surface climate ( Gillett and Thompson, 2003 [JoC, MoS, ARC] ).

7.4.6.2 Past Changes in Stratospheric Ozone

Ozone losses have been largest in the polar lower stratosphere during later winter and spring. For example, the ozone hole over Antarctica has occurred every spring since the early 1980 s ( Fioletov et al., 2002 [JoC, MoS] ). Antarctic ozone destruction is driven by climatologically low temperatures combined with high Cl and bromine amounts produced from photochemical breakdown of primarily anthropogenic CFCs and halons. Similar losses, smaller in magnitude, have occurred over the Arctic due to the same processes during cold winters. During warm winters, arctic ozone has been relatively unaffected ( (Tilmes et al., 2004 ) ). The antarctic lower stratosphere is nearly always cold enough to produce substantial ozone loss, but in the year 2002, a sudden stratospheric warming split the early ozone hole into two separate regions (e.g., Simmons et al., 2005 [JoC, MoS, ARC] ). Temperatures were subsequently too high to produce further ozone loss. Following the later merging of the two separate regions back into a single vortex, the dynamical conditions were unsuitable for further ozone loss. This is not an indication of recovery in ozone amounts, but rather the result of a dynamical disturbance (e.g., Newman et al., 2004 [JoC] ). A summary of recent stratospheric ozone changes is given in Chapter2 .

7.4.6.3 Future Changes in Stratospheric Ozone

The evolution of stratospheric ozone over the next few decades will depend on natural, including solar and volcanic activity (e.g., Steinbrecht et al., 2004 [JoC] ; Dameris et al., 2005 [MoS, SRC] ), and human-caused factors such as stratospheric halogen loading, which is expected to decrease over future decades ( WMO, 2003 [NPR] ; IPCC/TEAP, 2005 ). The evolution of ozone will also depend on changes in many stratospheric constituents: it is expected that the reduction of ozone-depleting substances in the 21st century will cause ozone to increase via chemical processes ( Austin et al., 2003 [MoS, SRC] ). However, this increase could be strongly affected by temperature changes (due to LLGHGs), other chemical changes (e.g., due to water vapour) and transport changes. Coupled Chemistry-Climate Models (CCMs) provide tools to simulate future atmospheric composition and climate. For this purpose, a set of consistent model forcings has been prescribed as part of the CCM Validation Activity for Stratospheric Processes and their Role in Climate (SPARC CCMVal; Eyring et al., 2005 [JoC, MoS, SRC] ). Forcings include natural and anthropogenic emissions based on existing scenarios, atmospheric observations and the Kyoto and Montreal Protocols and Amendments. The simulations follow the IPCC SRES scenario A1B ( IPCC, 2000 [NPR] ) and changes in halocarbons as prescribed in Table 4B-2 of WMO 2003 [NPR] ). Figure 7.18 shows the late winter minimum total column ozone poleward of 60° for various transient CCM reference simulations compared with observations. Antarctic ozone follows mainly the behaviour of Cl and bromine in the models. The peak depletion simulated by the CCMs occurs around the year 2000 followed by a slow increase with minimum values remaining constant between 2000 and 2010 in many models. Most models predict that antarctic ozone amounts will increase to 1980 values close to the time when modelled halogen amounts decrease to 1980 values, lagging the recovery in mid-latitudes due to the delay associated with transport of stratospheric air to polar regions. The late return to pre- 1980 values by about 2065 in the Atmospheric Model with Transport and Chemistry (AMTRAC) model ( Austin and Wilson, 2006 [JoC, MoS, SRC] ) is consistent with an empirical model study based on observations ( Newman et al., 2006 [JoC] ). Moreover, increased atmospheric fluxes of CFCs have recently been reported ( Hurst et al., 2006 [JoC] ), which may point to a still later recovery. The CCMs do not predict consistent values for minimum arctic column ozone, with some models showing large discrepancies with observations. In all CCMs that have been run long enough, arctic ozone increases to 1980 values before antarctic ozone does, by as much as 30 years (e.g., Austin and Wilson 2006 [JoC, MoS, SRC] ). This delay in the Antarctic arises from an increased Brewer-Dobson circulation ( Butchart and Scaife, 2001 [JoC] ; Butchart et al., 2006 [NPR, JoC, MoS] ) combined with a reduction in stratospheric temperatures.

Figure 7.18. (a) Minimum arctic total column ozone for March to April and (b) minimum antarctic total column ozone for September to October (both poleward of 60°) in Dobson Units (DU). Simulations of future evolution of ozone were performed by 11 CCMs analysed as part of the CCM Validation Activity for SPARC (http://www.pa.op.dlr.de/CCMVal/). Model results are compared with values calculated from the National Institute of Water and Atmospheric Research (NIWA) assimilated total column ozone database shown as black dots ( Bodeker et al., 2005 [ARC] ). The light grey shading between 2060 and 2070 shows the period when halogen amounts in the polar lower stratosphere are expected to return to 1980 values. Models include AMTRAC: Atmospheric Model with Transport and Chemistry; CCSRNIES: Center for Climate System Research - National Institute for Environmental Studies; CMAM: Canadian Middle Atmosphere Model; E39C: German Aerospace Center (DLR) version of ECHAM4 with chemistry and 39 levels; GEOSCCM: Goddard Earth Observing System Chemistry-Climate Model; MAECHAM4/CHEM: Middle Atmosphere ECHAM4 with Chemistry; MRI: Meteorological Research Institute; SOCOL: Solar Climate Ozone Links; ULAQ: University of L’Aquila; UMSLIMCAT: Unified Model SLIMCAT; WACCM: Whole Atmosphere Community Climate Model.

7.4.6.4 Uncertainties Due to Atmospheric Dynamics

Changes in atmospheric dynamics could affect ozone. For example, sub-grid scale processes such as gravity wave propagation (e.g., Warner and McIntyre, 2001 [JoC] ), prescribed for past and present conditions, may change in the future. Tropospheric climate changes will also alter planetary-scale waves. Together with changes in orographic gravity waves, these waves give rise to the increase in the Brewer-Dobson circulation seen in most models ( Butchart et al., 2006 [NPR, JoC, MoS] ). The magnitude of this effect varies from model to model and leads to increased adiabatic heating of the polar regions, compensating in part the increased radiative cooling from CO₂ increases. Hence, the net heating or cooling is subject to large uncertainty, and available model simulations do not give a consistent picture of future development of ozone, particularly in the Arctic( Figure 7.18 ).

7.5.1.1 Dust

Estimates of the global source strength of bulk dust aerosols with diameters below 10 µm of between 1,000 and 3,000 Tg yr^–1 agree well with a wide range of observations ( Duce, 1995 [NPR, MoS] ; Textor et al., 2005 [ARC] ; Cakmur et al., 2006 [NPR, JoC, MoS] ). Seven to twenty percent of the dust emissions are less than 1 µm in diameter ( Cakmur et al., 2006 [NPR, JoC, MoS] ; Schulz et al., 1998 [JoC, MoS, SRC] Zhang et al. 1997 [JoC] ) estimated that about 800 Tg yr^–1 of Asian dust emissions are injected into the atmosphere annually, about 30% of which is re-deposited onto the deserts and 20% is transported over regional scales, while the remaining approximately 50% is subject to long-range transport to the Pacific Ocean and beyond. Asian dust appears to be a continuous source that dominates background dust aerosol concentrations on the west coast of the USA ( Duce, 1995 [NPR, MoS] ; Perry et al., 2004 [JoC] ). Uncertainties in the estimates of global dust emissions are greater than a factor of two ( Zender et al., 2004 [JoC, MoS] ) due to problems in validating and modelling the global emissions. The representation of the high wind tail of the wind speed distribution alone, responsible for most of the dust flux, leads to differences in emissions of more then 30% ( Timmreck and Schulz, 2004 [JoC, MoS, SRC] ). Observations suggest that annual mean African dust may have varied by a factor of four during 1960 to 2000 ( Prospero and Lamb, 2003 [JoC] ), possibly due to rainfall variability in the Sahel zone. Likewise, simulations of dust emissions in 2100 are highly uncertain, ranging from a 60% decrease to a factor of 3.8 increase as compared to present-day dust emissions ( Mahowald and Luo, 2003 [JoC, MoS] ; Tegen et al., 2004 [JoC, MoS] ; Woodward et al., 2005 [JoC, MoS, SRC] ; Stier et al., 2006a [MoS] ). Reasons for these discrepancies include different treatments of climate-biosphere interactions and the climate model used to drive the vegetation and dust models. The potentially large impact of climate change on dust emissions shows up in particular when comparing present-day with LGM conditions for dust erosion (e.g., Werner et al., 2002 [JoC] ).

The radiative effect of dust, which, for example, could intensify the African Easterly Waves, may be a feedback mechanism between climate and dust ( Jones et al., 2004 [NPR, JoC, SRC] ). It also alters the atmospheric circulation, which feeds back to dust emission from natural sources (see Section 7.5.4 Perlwitz et al. 2001 [JoC, MoS] ) estimate that this feedback reduces the global dust load by roughly 15%, as dust radiative forcing reduces the downward mixing of momentum within the planetary boundary layer, the surface wind speed, and thus dust emission ( Miller et al., 2004a [JoC, MoS] ). In addition to natural dust production, human activities have created another potential source for dust mobilisation through desertification. The contribution to global dust emission of desertification through human activities is uncertain: estimates vary from 50% ( Tegen et al., 1996 [JoC, MoS, SRC] ; Mahowald et al., 2004 [JoC, MoS] ) to less than 10% ( Tegen et al., 2004 [JoC, MoS] ) to insignificant values ( Ginoux et al., 2001 [JoC, MoS] ; Prospero et al., 2002 [JoC] ). A 43-year estimate of Asian dust emissions reveals that meteorology and climate have a greater influence on Asian dust emissions and associated Asian dust storm occurrences than does desertification( Figure 7.19 ;Zhang et al., 2003 [JoC] ).

Figure 7.19. (a) Chinese desert distributions from 1960 to 1979 and desert plus desertification areas from 1980 to 1999 . (b) Sources (S1 to S10) and typical depositional areas (D1 and D2) for Asian dust indicated by spring average dust emission flux (kg km^–2 per month) averaged over 1960 to 2002 . The percentages with standard deviations in parentheses denote the average amount of dust production in each source region and the total amount of emissions between 1960 and 2002 . The deserts in Mongolia (S2) and in western (S4) and northern (S6) China (mainly the Taklimakan and Badain Juran, respectively) can be considered the major sources of Asian dust emissions. Several areas with more expansions of deserts (S7, S8, S9 and S5) are not key sources. Adapted from Zhang et al. 2003 [JoC] ).

In addition, aerosol deposition affects global ecosystems. Deposition of mineral dust plays an important role in the biogeochemical cycle of the oceans, by providing the nutrient iron, which affects ocean biogeochemistry with feedbacks to climate and dust production ( Jickells et al., 2005 [JoC] ; Section 7.3.4.4 ). Conversely, water-soluble particulate iron over the Pacific Ocean is linked to elemental carbon emissions resulting from anthropogenic activity in Asia ( Chuang et al., 2005 [NPR, JoC] ). The input of trace elements by dust deposition is also of importance to terrestrial ecosystems. For example, it has been proposed that the vegetation of the Amazon basin is highly dependent on Saharan dust deposition, which provides phosphorus, necessary for maintenance of long-term productivity ( Okin et al., 2004 [JoC, ARC] ; Section 7.3 ). The Hawaiian Islands also depend on phosphorus from Asian dust transport ( Chadwick et al., 1999 [JoC] ). Moreover, mineral dust can act as a sink for acidic trace gases, such as sulphur dioxide (SO₂)and HNO₃,and thereby interact with the sulphur and N cycles (e.g., Dentener et al., 1996 [JoC, SRC] ; Umann et al., 2005 [JoC] ). Coatings with soluble substances, such as sulphate or nitrate, will change the ability of mineral dust aerosols to nucleate cloud droplets ( Levin et al., 1996 [JoC] ; Section 7.5.2.1 ).

7.5.1.2 Sea Salt

Sea salt aerosol is a key aerosol constituent of the marine atmosphere. Sea salt aerosol particles affect the formation of clouds and rain. They serve as sinks for reactive gases and small particles and possibly suppress new particle formation. Sea salt is also responsible for a large fraction of the non-sea salt sulphate formation (e.g., Sievering et al., 1992 [JoC] ). The major meteorological and environmental factors that affect sea salt formation are wind speed, atmospheric stability and wind friction velocity, sea surface and air temperatures, present and prior rain or snow and the amount and nature of surface-active materials in the near-surface ocean waters ( Lewis and Schwartz, 2005 [NPR, MoS, SRC] ). The average annual global sea salt flux from 12 models is estimated to be 16,300 Tg ± 200% ( Textor et al., 2005 [ARC] ) of which 15% is emitted into the submicron mode.

New evidence shows that the ocean also acts as a source of organic matter from biogenic origin ( O’Dowd et al., 2004 [JoC] ; Leck and Bigg, 2005b [JoC, SRC] O’Dowd et al. 2004 [JoC] ) show that during phytoplankton blooms (summer conditions), the organic aerosols can constitute up to 63% of the total aerosol. Surface-active organic matter of biogenic origin (such as lipidic and proteinaceous material and humic substances), enriched in the oceanic surface layer and transferred to the atmosphere by bubble-bursting processes, are the most likely candidates to contribute to the observed organic fraction in marine aerosol. Insoluble heat-resistant organic sub-micrometre particles (peaking at 40 to 50 nm in diameter), mostly combined into chains or aggregated balls of ‘marine microcolloids’ linked by an amorphous electron-transparent material with properties entirely consistent with exopolymer secretions ( (Decho, 1990; ) ( Verdugo et al., 2004 ) ), are found in near-surface water of lower-latitude oceans ( Benner et al., 1992 [JoC] ;( Wells and Goldberg, 1994 ) ), in leads between ice floes ( Bigg et al., 2004 [SRC] ), above the arctic pack ice ( Leck and Bigg, 2005a [JoC, SRC] ) and over lower-latitude oceans ( Leck and Bigg, 2005b [JoC, SRC] ). This aerosol formation pathway may constitute an ice (microorganisms)-ocean-aerosol-cloud feedback.

7.5.1.3 Natural Organic Carbon

Biogenic organic material is both directly emitted into the atmosphere and produced by VOCs. Primary emissions from the continents have been thought to be a relatively minor source but some studies suggest that these emissions could be much higher than previously estimated ( Folberth et al., 2005 [JoC, SRC] ; Jaenicke, 2005 [JoC] Kanakidou et al. 2005 [MoS] ) estimate a global biogenic secondary organic aerosol production of about 30 Tg yr^–1 and recognise the potentially large, but uncertain, flux of primary biogenic particles. Annual global biogenic VOC emission estimates range from 500 to 1,200 Tg yr^–1 ( Guenther et al., 1995 [JoC, MoS, SRC] ). There is a large range (less than 5 to greater than 90%) of organic aerosol yield for individual compounds and atmospheric conditions resulting in estimates of global annual secondary organic aerosol production from biogenic VOCs that range from 2.5 to 44.5 Tg of organic matter per year ( Tsigaridis and Kanakidou, 2003 [MoS] ). All biogenic VOC emissions are highly sensitive to changes in temperature, and some emissions respond to changes in solar radiation and precipitation ( Guenther et al., 1995 [JoC, MoS, SRC] ). In addition to the direct response to climatic changes, biogenic VOC emissions are also highly sensitive to climate-induced changes in plant species composition and biomass distributions.

Global biogenic VOC emissions respond to climate change (e.g., ( Turner et al., 1991; ) Adams et al., 2001 [MoS, SRC] ; Penner et al., 2001 [NPR, ARC] ; Sanderson et al., 2003b [JoC] ). These model studies predict that solar radiation and climate-induced vegetation change can affect emissions, but they do not agree on the sign of the change. Emissions are predicted to increase by 10% per °C ( Guenther et al., 1993 [JoC, MoS, SRC] ). There is evidence of physiological adaptation to higher temperatures that would lead to a greater response for long-term temperature changes ( Guenther et al., 1999 [JoC, MoS, SRC] ). The response of biogenic secondary organic carbon aerosol production to a temperature change, however, could be considerably lower than the response of biogenic VOC emissions since aerosol yields can decrease with increasing temperature. A potentially important feedback among forest ecosystems, greenhouse gases, aerosols and climate exists through increased photosynthesis and forest growth due to increasing temperatures and CO₂ fertilization ( (Kulmala et al., 2004 ) ). Increased forest biomass would increase VOC emissions and thereby organic aerosol production. This couples the climate effect of CO₂ with that of aerosols.

7.5.1.4 Aerosols from Dimethyl Sulphide

Dimethyl sulphide produced by phytoplankton is the most abundant form in which the ocean releases gaseous sulphur. Sea-air fluxes of DMS vary by orders of magnitude depending mainly on DMS sea surface concentration and on wind speed. Estimates of the global DMS flux vary widely depending mainly on the DMS sea surface climatology utilised, sea-air exchange parametrization and wind speed data, and range from 16 to 54 Tg yr^–1 of sulphur (see Kettle and Andreae, 2000 [JoC, MoS] for a review). According to model studies ( Gondwe et al., 2003 [JoC] ; Kloster et al., 2006 [MoS] ), 18 to 27% of the DMS is converted into sulphate aerosols. Penner et al. 2001 [NPR, ARC] ) show a small increase in DMS emissions between 2000 and 2100 (from 26.0 to 27.7 Tg yr^–1 of sulphur) using constant DMS sea surface concentrations together with a constant monthly climatological ice cover. Gabric et al. 2004 [JoC, MoS] ) predict an increase of the globally integrated DMS flux perturbation of 14% for a tripling of the pre-industrial atmospheric CO₂ concentration.

Bopp et al. 2004 [MoS] ) estimate the feedback of DMS to cloud albedo with a coupled atmosphere-ocean-biogeochemical climate model that includes phytoplankton species in the ocean and a sulphur cycle in the atmospheric climate model. They obtain an increase in the sea-air DMS flux of 3% for doubled atmospheric CO₂ conditions, with large spatial heterogeneities (–15 to +30%). The mechanisms affecting those fluxes are marine biology, relative abundance of phytoplankton types and wind intensity. The simulated increase in fluxes causes an increase in sulphate aerosols and, hence, cloud droplets resulting in a radiative perturbation of cloud albedo of –0.05 Wm^–2,which represents a small negative climate feedback to global warming.

7.5.1.5 Aerosols from Iodine Compounds

Intense new aerosol particle formation has been frequently observed in the coastal environment ( O’Dowd et al., 2002a [JoC] ). Simultaneous coastal observations of reactive iodine species ( Saiz-Lopez et al., 2005 [MoS] ), chamber studies using iodocarbon precursors and laboratory characterisation of iodine oxide particles formed from exposure of Laminaria macroalgae to ozone ( McFiggans et al., 2004 [SRC] ) have demonstrated that coastal particle formation is linked to iodine compound precursor released from abundant infralittoral beds of macroalgae. The particle bursts overwhelmingly occur during daytime low tides ( O’Dowd et al., 2002b [JoC] ; Saiz-Lopez et al., 2005 [MoS] ). Tidal exposure of kelp leads to the well-documented release of significant fluxes of iodocarbons ( (Carpenter et al., 2003 ) ), the most photolabile of which, di-iodomethane (CH₂I₂), may yield a high iodine atom flux. However, the iodine monoxide (IO) and iodine dioxide (OIO) radicals, and new particles are thought more likely to result from emissions of molecular iodine ( McFiggans et al., 2004 [SRC] ), which will yield a much greater iodine atom flux ( Saiz-Lopez and Plane, 2004 [JoC] ). It is unclear whether such particles grow sufficiently to act as CCN ( O’Dowd, 2002 [JoC] ; Saiz-Lopez et al., 2005 [MoS] ). Thus, a hitherto undiscovered remote ocean source of iodine atoms (such as molecular iodine) must be present if iodine-mediated particle formation is to be important in the remote marine boundary layer ( McFiggans, 2005 [JoC, SRC] ).

Aerosol burden and lifetime are also affected by microphysical interactions among the different aerosol compounds as well as by changes in the spatial and seasonal distribution of the emissions. Sea salt aerosols, for example, provide surfaces for conversion of SO₂ into sulphate aerosols ( Sievering et al., 1992 [JoC] ) with consequences for cloud formation ( Gong and Barrie, 2003 [JoC, MoS] ; Section 7.5.2.1 ). A future reduction in SO₂ emissions and the associated reduced conversion of hydrophobic to hydrophilic soot could lead to a prolonged residence time of soot ( Cooke et al., 2002 [JoC, MoS, ARC] ; Stier et al., 2006b [JoC, MoS] ) and increased ammonium nitrate ( Liao and Seinfeld, 2005 [JoC, MoS] ). However, in a transient AOGCM climate simulation with an em-bedded microphysical aerosol module, Stier et al. 2006a [MoS] ) show that the effect on the hydrophobic to hydrophilic conversion can be outweighed by a general shift to low-latitude dry-season soot emissions. Consequently, soot lifetime increases in a future climate despite an enhanced conversion of hydrophobic to hydrophilic soot.

7.5.1.6 Climatic Factors Controlling Aerosol Burdens and Cycling

As discussed above, near-surface wind speed determines the source strength for primary aerosols (sea salt, dust, primary organic particles) and precursors of secondary aerosols (mainly DMS). Progress has been made in the development of source functions (in terms of wind speed) for sea salt and desert dust (e.g., Tegen et al., 2002 [JoC, MoS] ; Gong, 2003 [JoC] ; Balkanski et al., 2004 [NPR] ). Wind speed also affects dry deposition velocities and hence the lifetime of aerosols. In addition, biogenic emissions are strongly dependent on temperature (together with humidity/moisture; e.g., Guenther et al., 1995 [JoC, MoS, SRC] ). Temperature also is a key factor in the gas-aerosol partitioning of semi-volatile secondary organics ( Kanakidou et al., 2005 [MoS] ).

Precipitation directly affects the wet removal and hence the lifetime of atmospheric aerosols. More aerosols decrease the precipitation formation rate, which in turn increases the lifetime of aerosols and results in more long-range aerosol transport to remote regions where wet removal is less efficient. At the same time, precipitating boundary layer clouds maintain themselves by keeping aerosol concentrations low (e.g., Baker and Charlson, 1990 [JoC] ; Stevens et al., 2005 [JoC] ; Sharon et al., 2006 [JoC] ). Precipitation also affects soil moisture, with impacts on dust source strength and on stomatal opening/closure of plant leaves, hence affecting biogenic emissions. Cloud processing is an important pathway in the gas-to-particle conversion. It is the most important oxidation pathway for sulphate aerosols and shifts the aerosol size distribution to larger sizes, such that aerosols are more easily activated in subsequent cloud events (e.g., Hoppel et al., 1990 [JoC] ; Kerkweg et al., 2003 [JoC, MoS] ; Yin et al., 2005 [JoC] ). It is also important in the conversion of hydrophobic to hydrophilic carbon.

7.5.2.1 Aerosol Effects on Water Clouds and Warm Precipitation

Aerosols are hypothesised to increase the lifetime of clouds because increased concentrations of smaller droplets lead to decreased drizzle production and reduced precipitation efficiency ( Albrecht, 1989 [JoC] ). It is difficult to devise observational studies that can separate the cloud lifetime from the cloud albedo effect (see Section 2.4 ). Thus, observational studies usually provide estimates of the combined effects. Similarly, climate models cannot easily separate the cloud lifetime indirect effect once the aerosol scheme is fully coupled to a cloud microphysics scheme, but also predict the combined cloud albedo, lifetime and semi-direct effect.

Evidence for the absence of a drizzle mode due to anthropogenic emissions of aerosols and their precursors comes, for instance, from ship tracks perturbing marine stratus cloud decks off the coast of California ( Ferek et al., 1998 [JoC] ) as well as from analysing polluted compared with clean clouds off the Atlantic coast of Canada ( Peng et al., 2002 [JoC] ). One problem is that most climate models suggest an increase in liquid water when adding anthropogenic aerosols, whereas newer ship track studies show that polluted marine water clouds can have less liquid water than clean clouds ( Platnick et al., 2000 [JoC] ; Coakley and Walsh, 2002 [JoC, MoS] Ackerman et al. 2004 [JoC, MoS] ) attribute this effect to enhanced entrainment of dry air in polluted clouds in these instances with subsequent evaporation of cloud droplets. Similarly, when cloud lifetime is analysed, an increase in aerosol concentration from very clean to very polluted does not increase cloud lifetime, even though precipitation is suppressed ( Jiang et al., 2006 [NPR, JoC] ). This effect is due to competition between precipitation suppression and enhanced evaporation of the more numerous smaller cloud droplets in polluted clouds. Observed lower aerosol concentrations in pockets of open cells ( Stevens et al., 2005 [JoC] ) and in rifts of broken clouds surrounded by solid decks of stratocumulus with higher aerosol concentrations ( Sharon et al., 2006 [JoC] ) are manifestations of two stable aerosol regimes ( Baker and Charlson, 1990 [JoC] ). The low aerosol concentration regimes maintain themselves by higher drizzle rates. However, it is hard to disentangle cause and effect from these studies.

Smoke from burning vegetation reduces cloud droplet sizes and delays the onset of precipitation ( Warner and Twomey, 1967 [JoC] ; Rosenfeld, 1999 [JoC] ; Andreae et al., 2004 [JoC] ). In addition, desert dust suppresses precipitation in thin low-altitude clouds ( Rosenfeld et al., 2001 [JoC] ; Mahowald and Kiehl, 2003 [NPR, JoC] ). Contradictory results have been found regarding the suppression of precipitation by aerosols downwind of urban areas ( Givati and Rosenfeld, 2004 [JoC] ; Jin et al., 2005 [JoC, ARC] ) and in Australia ( Rosenfeld, 2000 [JoC] ;( Ayers, 2005 ) ).

Models suggest that anthropogenic aerosols suppress precipitation in the absence of giant CCN and aerosol-induced changes in ice microphysics (e.g., Lohmann, 2002 [NPR, JoC, SRC] ; Menon and DelGenio, 2007 ) as well as in mixed-phase clouds where the ice phase only plays a minor role ( Phillips et al., 2002 [JoC] ). A reduction in precipitation formation leads to increased cloud processing of aerosols. Feingold et al. 1998 [JoC, MoS] and Wurzler et al. 2000 [JoC] ) showed that cloud processing could either lead to an increase or decrease in precipitation formation in subsequent cloud cycles, depending on the size and concentration of activated CCN. Giant sea salt nuclei, on the other hand, may override the precipitation suppression effect of the large number of small pollution nuclei ( Johnson, 1982 [JoC] ; Feingold et al., 1999 [JoC] ; Rosenfeld et al., 2002 [JoC] ). Likewise, Gong and Barrie 2003 [JoC, MoS] ) predict a reduction of 20 to 60% in marine cloud droplet number concentrations and an increase in precipitation when interactions of sulphate with sea salt aerosols are considered. When aerosol effects on warm convective clouds are included in addition to their effect on warm stratiform clouds, the overall indirect aerosol effect and the change in surface precipitation can be larger or smaller than if just the aerosol effect on stratiform clouds is considered ( Nober et al., 2003 [JoC] ; Menon and Rotstayn, 2006 [JoC, MoS, SRC] ). Besides changes in the distribution of precipitation, the frequency of extreme events may also be reduced by the presence of aerosols ( Paeth and Feichter, 2006 [JoC, MoS, SRC] ).

Observations show that aerosols can decrease or increase cloud cover. Kaufman et al. 2005 [JoC] ) conclude from satellite observations that the aerosol indirect effect is likely primarily due to an increase in cloud cover, rather than an increase in cloud albedo. In contrast, model results of Lohmann et al. 2006 [Ambiguous] ) associate the increase in cloud cover with differing dynamic regimes and higher relative humidities that maintain higher aerosol optical depths. On the other hand, the semi-direct effect of absorbing aerosols can cause evaporation of cloud droplets and/or inhibit cloud formation. In a large area with absorbing biomass-burning aerosol, few low-lying clouds were observed when the aerosol optical depth exceeded 1.2 ( Koren et al., 2004 [JoC] ). Increasing emissions of absorbing aerosols from the late 1980 s to the late 1990 s in China also reduced cloud amount leading to a decrease in local planetary albedo, as deduced from satellite data ( Krüger and Grassl, 2004 [NPR, JoC] ). When the combined effect of pollution and smoke aerosols is considered from ground-based observations, the net effect seems to be an increase in cloud cover with increasing aerosol column concentrations ( Kaufman and Koren, 2006 [JoC] ).

7.5.2.2 Aerosol Impacts on Mixed-Phase Clouds

As satellite observations of aerosol effects on mixed-phase clouds are not conclusive ( Mahowald and Kiehl, 2003 [NPR, JoC] ), this section only refers to model results and field studies. Studies with GCMs suggest that if, in addition to mineral dust, hydrophilic black carbon aerosols are assumed to act as IN at temperatures between 0°C and –35°C, then increases in aerosol concentration from pre-industrial to present times may cause a glaciation indirect effect ( Lohmann, 2002 [NPR, JoC, SRC] ). Increases in IN can result in more frequent glaciation of super-cooled stratiform clouds and increase the amount of precipitation via the ice phase, which could decrease the global mean cloud cover leading to more absorption of solar radiation. Whether the glaciation effect or warm cloud lifetime effect is larger depends on the chemical nature of the dust ( Lohmann and Diehl, 2006 [JoC, SRC] ). Likewise, the number and size of ice particles in convective mixed-phase clouds is sensitive to the chemical composition of the insoluble fraction (e.g., dust, soot, biological particles) of the aerosol particles ( Diehl and Wurzler, 2004 [JoC, MoS] ).

Rosenfeld 1999 [JoC] and Rosenfeld and Woodley 2000 [JoC] ) analysed aircraft data together with satellite data suggesting that pollution aerosols suppress deep convective precipitation by decreasing cloud droplet size and delaying the onset of freezing. This hypothesis was supported by a cloud-resolving model study ( Khain et al., 2001 [JoC, MoS] ) showing that super-cooled cloud droplets down to –37.5°C could only be simulated if the cloud droplets were small and numerous. Precipitation from single-cell mixed-phase convective clouds is reduced under continental and maritime conditions when aerosol concentrations are increased ( (Yin et al., 2000; ) Khain et al., 2004 [JoC, MoS] ; Seifert and Beheng, 2006 [NPR] ). In the modelling study by Cui et al. 2006 [JoC] ), this is caused by drops evaporating more rapidly in the high aerosol case (see also Jiang et al., 2006 [NPR, JoC] ), which eventually reduces ice mass and hence precipitation. Khain et al. 2005 [JoC] ) postulate that smaller cloud droplets, such as those originating from human activity, would change the thermodynamics of convective clouds. More, smaller droplets would reduce the production of rain in convective clouds. When these droplets freeze, the associated latent heat release would then result in more vigorous convection and more precipitation. In a clean cloud, on the other hand, rain would have depleted the cloud so that less latent heat is released when the cloud glaciates, resulting in less vigorous convection and less precipitation. Similar results were obtained by Koren et al. 2005 [JoC] Zhang et al. 2005 [NPR, JoC, MoS, SRC] ) and for the multi-cell cloud systems studied by Seifert and Beheng 2006 [NPR] ). For a thunderstorm in Florida in the presence of Saharan dust, the simulated precipitation enhancement only lasted two hours after which precipitation decreased as compared with clean conditions ( van den Heever et al., 2006 [JoC, MoS] ). Cloud processing of dust particles, sulphate particles and trace gases can lead to an acceleration of precipitation formation in continental mixed-phase clouds, whereas in maritime clouds, which already form on rather large CCN, the simulated effect on precipitation is small ( Yin et al., 2002 [NPR, JoC] ). This highlights the complexity of the system and indicates that the sign of the global change in precipitation due to aerosols is not yet known. Note that microphysical processes can only change the temporal and spatial distribution of precipitation while the total amount of precipitation can only change if evaporation from the surface changes.

7.5.2.3 Aerosol Impacts on Cirrus Clouds

Cirrus clouds can form by homogeneous and heterogeneous ice nucleation mechanisms at temperatures below 235 K. While homogeneous freezing of super-cooled aqueous phase aerosol particles is rather well understood, understanding of heterogeneous ice nucleation is still in its infancy. A change in the number of ice crystals in cirrus clouds could exert a cloud albedo effect in the same way that the cloud albedo effect acts for water clouds. In addition, a change in the cloud ice water content could exert a radiative effect in the infrared. The magnitude of these effects in the global mean has not yet been fully established, but the development of physically based parametrization schemes of cirrus formation for use in global models led to significant progress in understanding underlying mechanisms of aerosol-induced cloud modifications ( Kärcher and Lohmann, 2002 [NPR, JoC, SRC] ; Liu and Penner, 2005 [MoS, ARC] ; Kärcher et al., 2006 [NPR, JoC, MoS, SRC] ).

Aglobal climate model study concluded that a cloud albedo effect based solely on ubiquitous homogeneous freezing is small globally ( Lohmann and Kärcher, 2002 [NPR, JoC, MoS, SRC] ). This is expected to also hold in the presence of heterogeneous IN that cause cloud droplets to freeze at relative humidities over ice close to homogeneous values (above 130–140%) ( Kärcher and Lohmann, 2003 [NPR, JoC, SRC] ). Efficient heterogeneous IN, however, would be expected to lower the relative humidity over ice, so that the climate effect may be larger ( Liu and Penner, 2005 [MoS, ARC] ). In situ measurements reveal that organic-containing aerosols are less abundant than sulphate aerosols in ice cloud particles, suggesting that organics do not freeze preferentially ( Cziczo et al., 2004 [NPR, JoC] ). A model study explains this finding by the disparate water uptake of organic aerosols, and suggests that organics are unlikely to significantly modify cirrus formation unless they are present in very high concentrations (compared with sulphate-rich particles) at low temperatures ( Kärcher and Koop, 2004 [SRC] ).

With regard to aerosol effects on cirrus clouds, a strong link has been established between gravity wave induced, mesoscale variability in vertical velocities and climate forcing by cirrus ( Kärcher and Ström, 2003 [SRC] ; Hoyle et al., 2005 [JoC] ). Hemispheric-scale studies of aerosol-cirrus interactions using ensemble trajectories suggest that changes in upper-tropospheric cooling rates and ice-forming aerosols in a future climate may induce changes in cirrus occurrence and optical properties that are comparable in magnitude with observed decadal trends in global cirrus cover ( Haag and Kärcher, 2004 [NPR, JoC, SRC] ). Optically thin and sub-visible cirrus are particularly susceptible to IN and therefore likely affected by anthropogenic activities.

Radiative forcing estimates and observed trends of aviation-induced cloudiness are discussed in Section 2.6 .In terms of indirect effects of aircraft-induced aerosols on cirrus clouds, Lohmann and Kärcher 2002 [NPR, JoC, MoS, SRC] ) show that the impact of aircraft sulphur emissions on cirrus properties via homogeneous freezing is small. The contribution from air traffic to the global atmospheric black carbon cycle was assessed by Hendricks et al. 2004 [MoS] ). Assuming that black carbon particles from aviation serve as efficient IN, maximum increases or decreases in ice crystal number concentrations of more than 40% are simulated in a climate model study assuming that the ‘background’ (no aviation impact) cirrus cloud formation is dominated by heterogeneous or homogeneous nucleation, respectively ( Hendricks et al., 2005 [JoC, SRC] ). Progress in assessing the impact of aircraft black carbon on cirrus is hampered by the poor knowledge of natural freezing modes in cirrus conditions and the inability to describe the full complexity of cirrus processes in global models.

7.5.2.4 Global Climate Model Estimates of the Total Anthropogenic Aerosol Effect

The total anthropogenic aerosol effect as defined here includes estimates of the direct effect, semi-direct effect, indirect cloud albedo and cloud lifetime effect for warm clouds from several climate models. The total anthropogenic aerosol effect is obtained as the difference between a multi-year simulation with present-day aerosol emissions and a simulation representative for pre-industrial conditions, where anthropogenic emissions are turned off. It should be noted that the representation of the cloud lifetime effect in GCMs is essentially one of changing the auto-conversion of cloud water to rainwater.

The global mean total anthropogenic aerosol effect on net radiation at TOA from pre-industrial times to the present day is shown in Figure 7.21 .Whereas Chapter2 only considers the radiative forcing of the cloud albedo effect, here feedbacks are included in the radiative flux change. In most simulations shown in Figures 7.21 to 7.23 ,the total aerosol effect is restricted to warm clouds except for the simulations by ( Jacobson 2006 ) and Lohmann and Diehl 2006 [JoC, SRC] ), who also include aerosol effects on mixed-phase and ice clouds. The total aerosol effect ranges from –0.2 Wm^–2 in the combined GCM plus satellite simulations ( Quaas et al., 2006 [SRC] ) to –2.3 Wm^–2 in the simulations by Ming et al. 2005 [JoC, MoS] ), with an average forcing of –1.2 Wm^–2.The total aerosol effect is larger when sulphate aerosols are used as surrogates for all anthropogenic aerosols than if multiple aerosol types are considered( Figure 7.21 ). Although most model estimates also include the direct and semi-direct effects, their contribution to the TOA radiation is generally small compared with the indirect effect, ranging from +0.1 to –0.5 Wm^–2 due to variations in the locations of black carbon with respect to the cloud ( Lohmann and Feichter, 2005 [SRC] ). The simulated cloud lifetime effect in a subset of models displayed in Figure 7.21 varies between –0.3 and –1.4 Wm^–2 ( Lohmann and Feichter, 2005 [SRC] ), which highlights some of the differences among models. The importance of the cloud albedo effect compared with the cloud lifetime effect varies even when the models use the same aerosol fields ( Penner et al., 2006 [MoS, ARC] ). Other differences among the simulations include an empirical treatment of the relationship between aerosol mass and cloud droplet number concentration vs. a mechanistic relationship, the dependence of the indirect aerosol effect on the assumed background aerosol or cloud droplet number concentration, and the competition between natural and anthropogenic aerosols as CCN ( Ghan et al., 1998 [JoC] ; O’Dowd et al., 1999 [JoC, MoS, ARC] ). Likewise, differences in the cloud microphysics scheme, especially in the auto-conversion rate, cause different cloud responses (e.g., A. Jones et al., 2001 [Ambiguous] ; Menon et al., 2002a [JoC, MoS, SRC] , 2003; Penner et al., 2006 [MoS, ARC] ).

Figure 7.21. Global mean total anthropogenic aerosol effect (direct, semi-direct and indirect cloud albedo and lifetime effects) defined as the response in net radiation at TOA from pre-industrial times to the present day and its contribution over the NH and SH, over oceans and over land, and the ratio over oceans/land. Red bars refer to anthropogenic sulphate ( Easter et al., 2004 [JoC, MoS] ; Ming et al., 2005 [JoC, MoS] ⁺), green bars refer to anthropogenic sulphate and black carbon ( Kristjánsson, 2002 [NPR, JoC] ^*,+), blue bars to anthropogenic sulphate and organic carbon ( Quaas et al., 2004 [JoC, MoS, SRC] ; Rotstayn and Liu, 2005 [JoC, MoS, SRC] ⁺), turquoise bars to anthropogenic sulphate and black and organic carbon (Menon and del Genio, 2005 [NotFound] ; Takemura et al., 2005 [JoC, MoS, ARC] ; Johns et al., 2006 [JoC, MoS] ; Storelvmo et al., 2006 [JoC, MoS] ), dark purple bars to anthropogenic sulphate and black and organic carbon effects on water and ice clouds ( (Jacobson, 2006; ) Lohmann and Diehl, 2006 [JoC, SRC] ), teal bars refer to a combination of GCM and satellite results (European Centre for Medium Range Weather Forecasts/Max-Planck Institute for Meteorology Atmospheric GCM (ECHAM) plus Polarisation and Directionality of the Earth’s Reflectance (POLDER), Lohmann and Lesins, 2002 [JoC, SRC] ; Laboratoire de Météorologie Dynamique GCM (LMDZ)/ECHAM plus Moderate Resolution Imaging Spectroradiometer (MODIS), Quaas et al., 2006 [SRC] ) and olive bars to the mean and standard deviation from all simulations. Vertical black lines for individual results refer to ±1 standard deviation in cases of multiple simulations and/or results.

*refers to estimates of the aerosol effect deduced from the shortwave radiative flux only

⁺ refers to estimates solely from the indirect effects

All models agree that the total aerosol effect is larger over the NH than over the SH( Figure 7.21 ). The values of the NH total aerosol effect vary between –0.5 and –3.6 Wm^–2 and in the SH between slightly positive and –1.1 Wm^–2,with an average SH to NH ratio of 0.3. Estimates of the ocean/land partitioning of the total indirect effect vary from 0.03 to 1.8 with an average value of 0.7. While the combined European Centre for Medium Range Weather Forecasts/Max-Planck Institute for Meteorology Atmospheric GCM (ECHAM4) plus Polarisation and Directionality of the Earth’s Reflectance (POLDER) satellite estimate suggests that the total aerosol effect should be larger over oceans ( Lohmann and Lesins, 2002 [JoC, SRC] ), combined estimates of the Laboratoire de Météorologie Dynamique (LMD) and ECHAM4 GCMs with Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data reach the opposite conclusion ( Quaas et al., 2006 [SRC] ). The average total aerosol effect over the ocean of –1 Wm^–2 agrees with estimates of between –1 and –1.6 Wm^–2 from the Advanced Very High Resolution Radiometer (AVHRR)/POLDER ( Sekiguchi et al., 2003 [JoC] ). Estimates from GCMs of the total aerosol effect are generally larger than those from inverse models ( Anderson et al., 2003 [JoC, MoS] and Chapter9 ).

As compared with the estimates of the total aerosol effect in Lohmann and Feichter 2005 [SRC] ), some new estimates ( Chen and Penner, 2005 [ARC] ; Rotstayn and Liu, 2005 [JoC, MoS, SRC] ; Lohmann and Diehl, 2006 [JoC, SRC] ) now also include the influence of aerosols on the cloud droplet size distribution (dispersion effect; Liu and Daum, 2002 [JoC, MoS] ). The dispersion effect refers to a widening of the size distribution in the polluted clouds that partly counteracts the reduction in the effective cloud droplet radius in these clouds. Thus, if the dispersion effect is taken into account, the indirect cloud albedo aerosol effect is reduced by 12 to 42% ( Peng and Lohmann, 2003 [JoC, SRC] ; Rotstayn and Liu, 2003 [JoC, SRC] ; Chen and Penner, 2005 [ARC] ). The global mean total indirect aerosol effect in the simulation by Rotstayn and Liu 2005 [JoC, MoS, SRC] ) has also been reduced due to a smaller cloud lifetime effect resulting from a new treatment of auto-conversion.

Global climate model estimates of the change in global mean precipitation due to the total aerosol effects are summarised in Figure 7.22 .Consistent with the conflicting results from detailed cloud system studies, the change in global mean precipitation varies between 0 and –0.13 mm day^–1.These differences are amplified over the SH, ranging from –0.06 mm day^–1 to 0.12 mm day^–1.In general, the decreases in precipitation are larger when the atmospheric GCMs are coupled to mixed-layer ocean models (green bars), where the sea surface temperature and, hence, evaporation are allowed to vary.

Figure 7.22. Global mean change in precipitation due to the total anthropogenic aerosol effect (direct, semi-direct and indirect cloud albedo and lifetime effects) from pre-industrial times to the present day and its contribution over the NH and SH, over oceans and over land. Red bars refer to anthropogenic sulphate ( Easter et al., 2004 [JoC, MoS] ; Ming et al., 2005 [JoC, MoS] ⁺), blue bars to anthropogenic sulphate and organic carbon ( Quaas et al., 2004 [JoC, MoS, SRC] ; Rotstayn and Liu, 2005 [JoC, MoS, SRC] ⁺), turquoise bars to anthropogenic sulphate, and black and organic carbon (Menon and del Genio, 2005 [NotFound] ; Takemura et al., 2005 [JoC, MoS, ARC] ; Johns et al., 2006 [JoC, MoS] ; Storelvmo et al., 2006 [JoC, MoS] ), dark purple bars to anthropogenic sulphate and black and organic carbon effects on water and ice clouds ( (Jacobson, 2006; ) Lohmann and Diehl, 2006 [JoC, SRC] ), teal bars refer to a combination of GCM and satellite results (LMDZ/ECHAM plus MODIS, Quaas et al., 2006 [SRC] ), green bars refer to results from coupled atmosphere/mixed-layer ocean (MLO) experiments ( Feichter et al., 2004 [JoC, MoS, SRC] : sulphate and black and organic carbon; Kristjansson et al., 2005 [JoC, MoS] : sulphate and black carbon; Rotstayn and Lohmann, 2002 [JoC, SRC] ⁺:sulphate only) and olive bars to the mean from all simulations. Vertical black lines refer to ±1 standard deviation.

⁺ refers to estimates solely from the indirect effects

7.5.4.1 Effects on Stability

Changes in the atmospheric lapse rate modify the longwave emission and affect the water vapour feedback ( Hu, 1996 [JoC] ) and the formation of clouds (see, e.g., Section 8.6 ). Observations and model studies show that an increase in the lapse rate produces an amplification of the water vapour feedback ( Sinha, 1995 [JoC] ). As aerosols cool the Earth’s surface and warm the aerosol layer, the lapse rate will decrease globally and suppress the water vapour feedback (e.g., Feichter et al., 2004 [JoC, MoS, SRC] ). The local change in atmospheric stability strongly depends on the altitude of the black carbon heating ( Penner et al., 2003 [JoC, ARC] ).

Absorption of solar radiation by aerosols can change the cloud amount (semi-direct effect; ( Grassl, 1975; ) Hansen et al., 1997 [JoC, MoS, ARC] ; Ackerman et al., 2000 [JoC] ; Ramanathan et al., 2001 [JoC, ARC] ;( Jacobson, 2006; ) Figure 7.20 ). The semi-direct effect has been simulated with GCMs and high-resolution cloud-resolving models, since it is implicitly accounted for whenever absorbing aerosols coupled to the radiation scheme are included ( Hansen et al., 1997 [JoC, MoS, ARC] ; Lohmann and Feichter, 2001 [JoC, SRC] ; Jacobson, 2002 [JoC] ; Menon et al., 2002b [JoC, SRC] ; Penner et al., 2003 [JoC, ARC] ; Cook and Highwood, 2004 [JoC, MoS] ; Hansen et al., 2005 [JoC, MoS, ARC] ). Aerosol heating within cloud layers reduces cloud fractions, whereas aerosol heating above the cloud layer tends to increase cloud fractions. When diagnosed within a GCM framework, the semi-direct effect can also include cloud changes due to circulation effects and/or surface albedo effects. Moreover, the semi-direct effect is not exclusive to absorbing aerosol, as potentially any radiative heating of the mid-troposphere can produce a similar response in a GCM ( Hansen et al., 2005 [JoC, MoS, ARC] ; see also Section 2.8 ). Cloud-resolving models of cumulus and stratocumulus case studies also diagnose semi-direct effects indicating a similar relationship between the height of the aerosol layer relative to the cloud and the sign of the semi-direct effect ( Ackerman et al., 2000 [JoC] ; Ramanathan et al., 2001 [JoC, ARC] ; Johnson et al., 2004 [JoC] ; Johnson, 2005 [JoC, MoS] ). Using a large eddy simulation, Feingold et al. 2005 [JoC] ) show that the reduction in net surface radiation and in surface latent and sensible heat fluxes is the most simple explanation of the reduction in cloudiness associated with absorbing aerosols.

7.5.4.2 Effects on the Large-Scale Circulation

Several studies have considered the response of a GCM with a mixed-layer ocean to indirect aerosol effects ( Rotstayn et al., 2000 [JoC, SRC] ; K. Williams et al., 2001 [JoC] ; Rotstayn and Lohmann, 2002 [JoC, SRC] ) or to a combination of direct and indirect aerosol effects ( Feichter et al., 2004 [JoC, MoS, SRC] ; Kristjansson et al., 2005 [JoC, MoS] ; Takemura et al., 2005 [JoC, MoS, ARC] ). All of these, and recent transient simulations ( Held et al., 2005 [JoC, MoS, ARC] ; Paeth and Feichter, 2006 [JoC, MoS, SRC] ), found a substantial cooling that was strongest in the NH, with a consequent southward shift of the Inter-Tropical Convergence Zone (ITCZ) and the associated tropical rainfall belt. Rotstayn and Lohmann 2002 [JoC, SRC] ) even suggest that aerosol effects might have contributed to the Sahelian droughts of the 1970 s and 1980 s (see Sections 9.5 and 11.2 ). If in turn the NH is warmed, for instance due to the direct forcing by black carbon aerosols, the ITCZ is found to shift northward ( Chung and Seinfeld, 2005 [JoC, MoS] ).

Menon et al. 2002b [JoC, SRC] and Wang 2004 [Ambiguous] ) found that circulation changes could be caused by aerosols in southeast China. In India and China, where absorbing aerosols have been added, increased rising motions are seen as well as increased subsidence to the south and north ( Menon et al., 2002b [JoC, SRC] ). However, Ramanathan et al. 2005 [JoC, ARC] ) found that convection was suppressed due to increased stability resulting from black carbon heating. Drier conditions resulting from suppressed rainfall can induce more dust and smoke due to the burning of drier vegetation ( Ramanathan et al., 2001 [JoC, ARC] ), thus affecting both regional and global hydrological cycles ( Wang, 2004 [Ambiguous] ). Heating of a lofted dust layer could increase the occurrence of deep convection ( Stephens et al., 2004 [JoC] ). It can also strengthen the Asian summer monsoon circulation and cause a local increase in precipitation, despite the global reduction of evaporation that compensates aerosol radiative heating at the surface ( Miller et al., 2004b [JoC, MoS] ). The dust-induced thermal contrast changes between the Eurasian continent and the surrounding oceans are found to trigger or modulate a rapidly varying or unstable Asian winter monsoon circulation, with a feedback to reduce the dust emission from its sources ( (Zhang et al., 2002 ) ).

In summary, an increase in atmospheric aerosol load decreases air quality and reduces the amount of solar radiation reaching the surface. This negative radiative forcing competes with the greenhouse gas warming for determining the change in evaporation and precipitation. At present, no transient climate simulation accounts for all aerosol-cloud interactions, so that the net aerosol effect on clouds deduced from models is not conclusive.