The rate of increase in the globally averaged atmospheric concentration of CO₂ varies greatly from year to year. �Fast� and �slow� years have differed by 3 to 4 PgC/yr within a decade (Figure 3.3). This variability cannot be accounted for by fossil fuel emissions, which do not show short-term variability of this magnitude. The explanation must lie in variability of the land-atmosphere flux or the ocean-atmosphere flux or both. Variability in both systems could be induced by climate variability.

An association between CO₂ variability and El Niño in particular has been reported for over twenty years and has been confirmed by recent statistical analyses (Bacastow, 1976; Keeling and Revelle, 1985; Thompson et al., 1986; Siegenthaler, 1990; Elliott et al., 1991; Braswell et al., 1997; Feely et al., 1997; Dettinger and Ghil, 1998; Rayner et al., 1999b). During most of the observational record, El Niño events have been marked by high rates of increase in atmospheric CO₂ concentration compared with surrounding years, in the order of > 1 PgC/yr higher during most El Niño events (Figure 3.3). Direct measurements of oceanic CO₂ in the equatorial Pacific over the last 20 years have shown that the natural efflux of CO₂ from this region is reduced by between 0.2 to 1.0 PgC/yr during El Niño (Keeling and Revelle, 1985; Smethie et al., 1985; Takahashi et al., 1986; Inoue and Sugimura, 1992; Wong et al., 1993; Feely et al., 1997; 1999b), mainly due to the reduced upwelling of CO₂-rich waters (Archer et al., 1996). The ocean response to El Niño in the most active region thus tends to increase global CO₂ uptake, counter to the increasing atmospheric concentration. Although it cannot be ruled out that other ocean basins may play a significant role for global interannual variability in ocean-atmosphere flux, the existing oceanic measurements suggest (by default) that the response of the terrestrial biosphere is the cause of the typically high rates of CO₂ increase during El Niño.

Associated variations in the north-south gradient of CO₂ indicate that the El Niño CO₂ anomalies originate in the tropics (Conway et al., 1994; Keeling and Piper, 2000). Typical El Niño events are characterised by changed atmospheric circulation and precipitation patterns (Zeng, 1999) that give rise to high tropical land temperatures (which would be expected to increase Rh and reduce NPP); concurrent droughts which reduce NPP, especially in the most productive regions such as the Amazon rain forest; and increased incidence of fires in tropical regions. Increased cloudiness associated with enhanced south-east Asian monsoons during the late phase of El Niño has also been suggested as a factor reducing global NPP (Yang and Wang, 2000). Typically, although not invariably, the rate of atmospheric CO₂ increase declines around the start of an El Niño, then rapidly rises during the late stages (Elliott et al., 1991; Conway et al., 1994). It has been suggested that this pattern represents early onset of enhanced ocean CO₂ uptake, followed by reduced terrestrial CO₂ uptake or terrestrial CO₂ release (Feely et al., 1987, 1999b; Rayner et al., 1999b; Yang and Wang, 2000).

Atmospheric ¹³C and, more recently, O₂ measurements have been used to partition the interannual variability of the atmospheric CO₂ increase into oceanic and terrestrial components. Analyses based on ¹³C by Keeling et al. (1995) and Francey et al. (1995) reached contradictory conclusions, but the discrepancies are now thought to be due at least in part to ¹³C measurement calibration problems during the 1980s, which have largely been resolved during the 1990s (Francey et al., 1999a). For the 1990s, a range of analyses using different atmospheric observations and/or data analysis techniques estimate that the amplitude of annual peak to peak variation associated with the ocean is about 2 to 3 PgC/yr and the amplitude associated with the terrestrial biosphere is about 4 to 5 PgC/yr (Rayner et al., 1999a; Joos et al., 1999a; Battle et al., 2000 (O₂-based analysis); Keeling and Piper, 2000; Manning, 2001). A similar partitioning was estimated by Bousquet et al. (2000) based on the spatial pattern of CO₂ measurements using the approach described in the next section (3.5.3). However, the various reconstructed time sequences of terrestrial and ocean uptake differ in many details and do not provide conclusive evidence of the mechanisms involved.

The early 1990s were unusual in that the growth rate in atmospheric CO₂ was low (1.9 PgC/yr in 1992), especially in the Northern Hemisphere (Conway et al., 1994), while an extended El Niño event occurred in the equatorial Pacific. Various mechanisms have been suggested, but none fully explain this unusual behavior of the carbon cycle. The slow down in the CO₂ increase has been linked to the predominantly mid- to high latitude cooling caused by the Pinatubo eruption (Conway et al., 1994; Ciais et al., 1995a,b; Schimel et al., 1996), but there is no proof of any connection between these events. Other partial explanations could come from a temporary slow down of tropical deforestation (Houghton et al., 2000), or natural decadal variability in the ocean-atmosphere or land-atmosphere fluxes (Keeling et al., 1995). In any case, the slowdown proved to be temporary, and the El Niño of 1998 was marked by the highest rate of CO₂ increase on record, 6.0 PgC/yr.

Table 3.4: Alternative estimates of ocean-atmosphere and land-atmosphere fluxes.
	Ocean-atmosphere flux	Land-atmosphere flux
Oceanic observations
1970 to 1990
Ocean 13C inventory	-2.1 ± 0.8a
	-2.1 ± 0.9b
1985 to 1995
Ocean 13C inventory	-1.5 ± 0.9c
1995
Surface-water pCO2	-2.8 ± 1.5d
1990
Inventory of anthropogenic CO2	-1.6 to -2.7e
Atmospheric observations
1980 to 1989
O2 in Antarctic firn	-1.8f	-0.4f
1990 to 1999
Atmospheric CO2 and 13C	-1.8g	-1.4g
	-2.4h	-0.8h
Models
1980 to 1989
OCMIP	-1.5 to -2.2i
CCMLP*		-0.3 to -1.5j*
*partitioned as follows:		*partitioned as follows:
Land-use change		0.6 to 1.0
CO2 and N fertilisation		-1.5 to -3.1
Climate variability		-0.2 to +0.9
Sources of data: a Quay et al. (1992). b Heimann and Maier-Reimer (1996). c Gruber and Keeling (2001). d Takahashi et al. (1999) with �0.6 PgC/yr correction for land-ocean river flux. e Gruber (1998), Sabine et al. (1999); Feely et al. (1999a), assuming that the ocean and atmospheric CO2 increase follow a similar curve. f This chapter, from data of Battle et al. (1996). g Updated calculation of Ciais et al. (1995b); Tans et al. (1989); Trolier et al. (1996) (no error bars given). h Keeling and Piper (2000) (no error bars given). i Orr et al. (2000), Orr and Dutay (1999). j McGuire et al. (2001).

IPCC Homepage