INTRODUCTION

The Miocene (~ 23 to 5 Ma) was part of the general late phase of Cenozoic cooling, but its climate was still a general hothouse situation. Various proxy data suggest it had a warmer and more humid climate than today (e.g., Wolfe et al. 1994a, b; Zachos et al. 2001; Bruch et al. 2004, 2006, 2007; Mosbrugger et al. 2005). In particular, the Miocene equator-to-pole latitudinal temperature gradient was weak, implying that higher latitudes were warmer than they are today (e.g., Wolfe et al. 1994a, b; Bruch et al. 2004, 2006, 2007). Corresponding to warm polar regions during the Miocene, it was commonly assumed that the largescale glaciation of the Northern Hemisphere did not start until the latest Miocene or the Pliocene (e.g., Kleiven et al. 2002; Winkler et al. 2002; Bartoli et al. 2005; St. John and Krissek 2002). However, some studies have raised questions about when significant Arctic sea ice first formed (e.g., Helland and Holmes 1997; Moran et al. 2006). It remains an open debate when sea ice first appeared in the Northern Hemisphere, although most recent evidences seems to support that the Arctic Ocean was already ice-covered in the Miocene (e.g., Moran et al. 2006). If so, polar regions in the Miocene may have been cooler than heretofore assumed.

Increasing carbon dioxide in the atmosphere is the primary agent for future climate change (e.g., Cubasch et al. 2001; Meehl et al. 2007). Thus, the amount of atmospheric CO₂ in the Miocene is relevant to the question of what palaeotemperatures were then. Some studies suggest that the Miocene CO₂ level was close to the pre-industrial concentration (280 ppm) or a little higher (e.g., Pearson and Palmer 2000; Pagani et al. 2005). However, other studies support a pCO₂ being as high as 500 ppm (e.g., MacFadden 2005) to 700 ppm (e.g., Cerling 1991). Retallack (2001) even proposed that atmospheric carbon dioxide was higher than 1000 ppm until the Late Miocene. It is difficult to conceive that high latitudes would have been warm if CO₂ was low in the Miocene, but it also seems improbable that polar regions were ice-covered if CO₂ concentrations were high.

Previous climate model experiments for the late Tertiary concentrated primarily on either the roles of geography and orography (e.g., Ramstein et al. 1997; Ruddiman et al. 1997; Kutzbach and Behling 2004) or on the role of the ocean (e.g., Bice et al. 2000; Steppuhn et al. 2006) as a major influence on climate. In order to adapt their models to Miocene conditions, modellers must specify the concentration of atmospheric CO₂; however the fact that the Miocene CO₂ level is still debated, climate models of Miocene time intervals are quite variable (e.g., Kutzbach and Behling 2004; Steppuhn et al. 2006, 2007). Recently, Steppuhn et al. (2007) presented a sensitivity experiment for the Late Miocene, which analysed the effects of using a concentration of 2×CO₂ (700 ppm) as compared to 1×CO₂ (353 ppm). Even with 700 ppm, the Late Miocene model experiments indicated that the Arctic Ocean would still have been ice-covered (Steppuhn et al. 2007), results which appear to conflict with the fossil record. Furthermore, the heating of high latitudes in these high-CO₂ model experiments occurred at the expense of warming lower latitudes, which conflicts with proxy data (Steppuhn et al. 2007). Steppuhn et al. (2007) concluded that while high CO₂ did not explain warm high latitudes, palaeovegetation might. Miocene vegetation is indeed thought to have contributed to warm high latitudes (Dutton and Barron 1997; Micheels et al. 2007), but climate modellers still run into trouble when trying to understand warm high latitudes in the Miocene (Steppuhn et al. 2006, 2007; Micheels et al. 2007). A low atmospheric pCO₂ is not sufficient to explain the warm Miocene climate, nor does a high concentration explain the distribution of temperatures (Steppuhn et al. 2007; Micheels et al. 2007). But what if CO₂ was intermediate?

Geological processes such as the mountain uplift from the Miocene until today and their influence on climate had repeatedly attracted the interest for model experiments (e.g., Ramstein et al. 1997). In contrast, sensitivity experiments with respect to CO₂ are rare (Steppuhn et al. 2007; Tong et al. 2009). Levels of carbon dioxide in the Miocene are not known for sure (e.g. Retallack 2001; Pagani et al. 2005), but realistic model experiments specifically require CO₂ levels to be properly specified. Climate models are tools to test hypotheses and to understand specific processes – carbon dioxide is a general key factor for climate (e.g., Meehl et al. 2007), but in terms of Tertiary climate modelling the relevance of carbon dioxide is poorly understood. This situation is unsatisfying since the Miocene could serve as a possible analogue for the future climate change if it were better understood (e.g., Kutzbach and Behling 2004). Three basic points still remain open for discussion:

1. How high must atmospheric CO₂ be so that the Arctic Ocean is ice-free in the Miocene?
2. What was the general climate sensitivity o enhanced pCO₂ in the Miocene? Is the Late Miocene comparable to future climate change scenarios?
3. How consistent are different Late Miocene CO₂ scenarios compared to proxy data? Can we estimate a Late Miocene pCO₂ from comparing models and independent proxy data?

To address these questions, we perform sensitivity experiments for the Late Miocene using an earth system model of intermediate complexity Planet Simulator. Based on a reference run, we present sensitivity experiments that consider pCO₂ values ranging from low (200 ppm) to high (700 ppm). In addition, we perform a transitional experiment with a steady increase of CO₂ by +1 ppm starting with 200 ppm and ending up with 2000 ppm. Finally, we use quantitative terrestrial proxy data to validate the model results. The results of our sensitivity experiments contribute to a better understanding of the role of CO₂ for the Cenozoic climate history.