Parker and Levy proposed that geomagnetic reversals result from fluctu
ations in the location of cyclonic convection cells that reverse the t
oroidal flux in mid-latitudes, as exhibited by a mean field dynamo mod
el with differential rotation concentrated at the outer boundary, and
alpha at a moveable interior point. Roberts studied a more realistic m
odel with alpha and omega spread over a volume, but found that it prod
uced oscillatory rather than stationary fields without sufficient meri
dian circulation. This difference in behaviour is explored with a dyna
mo model where the concentrated quantities used by Levy have been repl
aced with broader functions. Stationary solutions are obtained when om
ega is concentrated at a greater radius than alpha, provided alpha is
at sufficiently low latitudes. Toroidal flux is reversed when alpha is
in high latitudes, as required by the Parker-Levy reversal mechanism.
At very high latitudes, the dynamo oscillates, the transition between
stationary and oscillatory modes suggesting a different mechanism for
reversal. When omega is at a smaller radius than alpha, toroidal flux
does not reverse, and the solutions are always steady, so in this cas
e the reversal mechanism will not work. When alpha and omega overlap,
stationary solutions are only possible with alpha at very low latitude
s; otherwise, solutions are oscillatory. Sufficient meridian circulati
on leads to stationary solutions for all locations of alpha with rever
sed toroidal flux when alpha is at high latitudes. I conclude that the
Parker-Levy reversal mechanism can apply provided the cyclonic convec
tion is inside the differential rotation, or sufficient meridian circu
lation exists to stabilise the geodynamo when alpha and omega overlap.
Reversal behaviour also occurs in many instances where the preferred
dynamo mode becomes oscillatory. (C) 1998 Elsevier Science B.V.