Using nine chemistry-climate and eight associated no-chemistry models,
we investigate the persistence and timing of cold episodes occurring in
the Arctic and Antarctic stratosphere during the period 1980-2014. We
find systematic differences in behavior between members of these model
pairs. In a first group of chemistry models whose dynamical
configurations mirror their no-chemistry counterparts, we find an
increased persistence of such cold polar vortices, such that these cold
episodes both start earlier and last longer, relative to the times of
occurrence of the lowest temperatures. Also the date of occurrence of
the lowest temperatures, both in the Arctic and the Antarctic, is
delayed by 1-3 weeks in chemistry models, versus their no-chemistry
counterparts. This behavior exacerbates a widespread problem occurring
in most or all models, a delayed occurrence, in the median, of the most
anomalously cold day during such cold winters. In a second group of
model pairs there are differences beyond just ozone chemistry. In
particular, here the chemistry models feature more levels in the
stratosphere, a raised model top, and differences in non-orographic
gravity wave drag versus their no-chemistry counterparts. Such
additional dynamical differences can completely mask the above influence
of ozone chemistry. The results point towards a need to retune
chemistry-climate models versus their no-chemistry counterparts.