The ensemble-mean response of the annual mean ITCZ to CO2-induced warming is to shift to the north, amplifying the pre-existing asymmetry in the basic state, but the spread across models is large. This is consistent with a similarly large scatter in the warming differential between the hemispheres; the magnitude of the shift is also modulated by the input of energy into the equatorial atmosphere, which varies greatly across models. Overall, the inter-model variance in the annual mean changes in the ITCZ latitude is explained equally well by changes in the latitude of the energy flux equator and in the magnitude of the atmospheric energy flux across the equator. The models do not show that the relationship between seasonal excursion of the ITCZ and seasonal changes in energy transport is quantitatively the same as that found between annual mean changes in forced simulation. This is explained in part by a weak relationship between the Hadley circulation and the transport of energy across the equator during much of the year, and in part by the changes in the profile of ascent and in the gross moist stability of the tropical troposphere under global warming.
The CO2-induced northward shift of the ITCZ is not uniform across the year, but is strongest during spring (when the climatological ITCZ is furthest to the south in these setups). The width of the ITCZ (defined as the region of net precipitation) is decreased in the annual mean, because of the sharp reduction in Southern Hemisphere precipitation during spring, but it actually increases during most months of the climatology. The timing of the rainy season is not delayed in the 4xCO2 aquaplanet simulations, but it is in simulations that include the idealized fixed-soil-moisture continent, supporting the hypothesis that a delay of the rainy season in some monsoon regions (as simulated by a majority of comprehensive models) is due to changes in circulation and does not require changes in the surface properties of the land.