Paleo-climate records in ice cores revealed high variability in temperature, atmospheric dust content and carbon dioxide. The longest CO₂ record from the Antarctic ice core of the Vostok station went back in time as far as about 410 kyr BP showing a switch of glacials and interglacials in all those parameters approximately every 100 kyr during the last four glacial cycles with CO₂ varying between 180-300 ppmv [Petit et al., 1999]. New measurements of dust and the isotopic temperature proxy deuterium of the EPICA Dome C (EDC) ice core covered the last 740 kyr, however, revealed glacial cycles of reduced temperature amplitude [EPICA community members, 2004]. These new archives offer the possibility to propose atmospheric CO₂ for the pre-Vostok time span as called for in the EPICA challenge [Wolff et al., 2004]. Here, we contribute to this challenge using a box model of the isotopic carbon cycle [Köhler et al., 2005] based on process understanding previously derived for Termination I. Our results show that major features of the Vostok period are reproduced while prior to Vostok our model predicts significantly smaller amplitudes in CO₂ variations.

It is investigated how abrupt changes in the North Atlantic (NA) thermohaline circulation (THC) affect the terrestrial carbon cycle. The Lund-Potsdam-Jena Dynamic Global Vegetation Model is forced with climate perturbations from freshwater experiments with the ECBILT-CLIO ocean-atmosphere model. A reorganization of the marine carbon cycle is not addressed. Modeled NA THC collapsed and recovered after about a millennium in response to prescribed freshwater forcing. The initial cooling of several Kelvin over Eurasia causes a reduction of extant boreal and temperate forests and a decrease in carbon storage in high northern latitudes, whereas improved growing conditions and slower soil decomposition rates lead to enhanced storage in mid-latitudes. The magnitude and evolution of global terrestrial carbon storage in response to abrupt THC changes depends sensitively on the initial climate conditions. These were varied using results from time slice simulations with the Hadley climate model for different periods over the past 21,000 years. Terrestrial storage varies between -67 and +50 PgC for the range of experiments with different initial conditions. Simulated peak-to-peak differences in atmospheric CO₂ and d¹³C are 6 and 18 ppmv for glacial and early Holocene conditions. Simulated changes in d¹³C are between 0.18 and 0.30 permil. The small CO₂ changes modelled for glacial conditions are compatible with available evidence from marine studies and the ice core CO₂ record. The latter shows CO₂ variations of up to 20 ppmv broadly in parallel with the Antarctic warm events A1 to A4.

The Siberian permafrost carbon stock has been studied using a newly developed soil model, which takes into account soil freezing/thawing and organic matter decomposition in the form of soil respiration and methanogenesis. The results show that the soil response to a rapid external warming can be a self-sustaining process involving permafrost melting, deep-soil respiration with associated heat generation, and methanogenesis. Most of the soil carbon is thus consumed until there is not enough of it to feed intense respiration and/or methanogenesis. This behavior is manifested only at sufficiently warm climate established after the warming. Carbon consumption in the extremely carbon-rich Yedoma Ice Complex region appears to be moderate due to cold climatic conditions.

In order to investigate the long-term and inter-annual variations in the atmospheric CO₂ concentration record obtained by aircraft measurements over Japan, we have conducted numerical experiments using a transport model with a process-based ecosystem model. The climate-induced anomalies of net biospheric flux account for a significant part of the inter-annual variations in the CO₂ growth rate. The results indicate that year-to-year change in observed vertical CO₂ gradient is mainly caused by the inter-annual variability in atmospheric transport, likely related to El Niño events.

Soil moisture measurements in an Australian tropical savanna show accumulating soil water under three years of Free Air CO₂ Enrichment (FACE).  Most of this accumulation is occurring below the rooting depth of grasses.  Although this increase in stored soil water is only a fraction (< 0.3% yr^-1) of annual rainfall, it is cumulative and may advantage deep-rooted woody plants.

A Time Dependent Inverse (TDI) model is used to estimate CO₂ fluxes for 64 regions of the globe from atmospheric data in the period January 1988–December 2001. These estimated are then used for understanding interannual variability in fluxes and simulating the CO₂ concentrations at various sites. The NIES/FRCGC transport model driven by interannually varying meteorology is used in both part of the analysis. Estimated atmospheric CO₂ concentrations agree closely with those observed at various sites globally.

This paper examines two key feedbacks that operate between the terrestrial carbon cycle, atmospheric carbon dioxide (CO₂) and climate: the positive carbon cycle-climate feedback and the negative CO₂ fertilization feedback.  Both feedbacks affect strongly the growth rate of future atmospheric CO₂, and interact in such a way that the effect of one is notably modified in the absence of the other.

Using discrete air sampling, atmospheric CO₂ and its stable carbon (d¹³C) and oxygen (d¹⁸O) isotopic ratios have been measured since 1994 in a cool-temperate deciduous forest in central Japan influenced strongly by the Asian monsoon. In this paper, the results are shown and the temporal variations on different time scales are discussed.

Is the massive Amazon forest a CO₂ sink, a source or is it in equilibrium?  

There is a large uncertainty in carbon fluxes estimates for the tropics as a whole and in particular for the Amazon region in South America, bringing the attention to the lack of information to call the region a carbon source or sink. The production of scientific consistent and long term data series for the region is a process that has to advance step by step.