The buildup of atmospheric CO2 since 1958 is surprisingly well explained by the simple premise that 57% of the industrial emissions (fossil fuel burning and cement manufacture) has remained airborne. This premise accounts well for the rise both before and after 1980 despite a decrease in the growth rate of fossil fuel CO2 emissions, which occurred at that time, and by itself should have caused the airborne fraction to decrease. In contrast, the buildup prior to 1958 was not simply proportional to cumulative fossil fuel emissions, and notably included a period during the 1940s when CO2 growth stalled despite continued fossil fuel emissions. Here we show that the constancy of the airborne fraction since 1958 can be in part explained by decadal variations in global land air temperature, which caused a warming-induced release of CO2 from the land biosphere to the atmosphere. We also show that the 1940s plateau may be related to these decadal temperature variations. Furthermore, we show that there is a close connection between the phenomenology producing CO2 variability on multidecadal and El Niño timescales.