SUMMARY AND CONCLUSIONS

With the earth system model of intermediate complexity Planet Simulator, we perform CO₂-sensitivity experiments for the Late Miocene. Therein, CO₂ ranges from 200 to 2000 ppm, whereas all other boundary conditions remain unchanged. We summarise our modelling results in the following:

An atmospheric carbon dioxide of 1500 ppm is necessary to produce an ice-free Northern Hemisphere in our model. This value is much too high to be reasonable for the Miocene (e.g., Cerling 1991; Retallack 2001). Our sensitivity experiments support evidences for an onset of the Northern Hemisphere's glaciation before the Miocene (e.g., Moran et al. 2006; Kamikuri et al. 2007; Jakobsson et al. 2007).
The climate sensitivity on enhanced concentrations of greenhouse gases is reduced in the Late Miocene as compared to the modern situation. The Late Miocene represents a hothouse climate with a reduced ice cover, which dampens the ice-albedo feedback. However, the climate response on increases in CO₂ is only slightly weaker than in future climate change scenarios (e.g., Meehl et al. 2007). With some limitations, the Late Miocene can serve as an analogue for the future situation.
The Late Miocene simulations with a pCO₂ from 280 to 630 ppm agree reasonably well with quantitative terrestrial proxy data. If CO₂ is higher, the consistency to proxy data is successively getting worse. Based on our results, an 'intermediate' concentration of CO₂ between 360 and 460 ppm is realistic for the Late Miocene.

We present some sensitivity experiments with an earth system model of intermediate complexity. The results should not be over-estimated because of uncertainties in the model (e.g., simplified physical parameterisations) and its setup (e.g., the ocean), but they are quite consistent to other studies. The comparison of model results and proxy data emphasises the need for proxy data from crucial regions such as the high latitudes and Africa. More quantitative climate information from the fossil record from these poorly covered regions would be helpful to better estimate the reliability of climate model experiments.