The model of Paillard and Parrenin (Earth Planet Sci Lett 227:263–271, 2004) was modified to obtain a closer fit to d18O and CO2 time series for the last 800 kyr. The model performance can be improved if its CO2 sensitivity to I65 insolation is eliminated and if different response times are assumed for ablation/accumulation of ice. Correlations between simulated and experimental time series for CO2 and ice volume V increase from 0.59 and 0.63 to 0.79 and 0.88, respectively. According to these models, terminations are produced by I65 amplification through CO2-T and T-CO2 feedbacks, in synergy with an extra CO2 contribution from the deep ocean. This contribution is strongly dependent on ice-sheet extent and ice volume (or alternatively, CO2 concentration, which is a good proxy of Antarctic temperature) but is insensitive to Southern Ocean (SO) insolation on 21 February (I60). Change of deep SO state may be the ‘‘order parameter’’ for nonlinear deglacial changes. According to these models, 100 kyr periodicity of glacial cycles arises from the characteristic time of Antarctic ice sheet advance to the continental slope

This is a contribution to the TIC-MOC project funded by the 2008–2011 Spanish R + D Plan. [...] The work of one of the authors (C. Herrero) was supported by a CSIC JAE-Predoc scholarship co-financed by the European Social Fund (FSE)

25 pages, 26 figures, 1 table, supplementary material https://doi.org/10.1007/s00382-012-1614-7

Peer Reviewed