\begin{abstract}
We present in the context of advection of trace species by 3 Dimensional
atmospheric flows,
a comparative test of a hierarchy of finite volume transport schemes
initially derived by Van Leer (1977, J. Comput. Phys., 23:276-299).
Those schemes are conservative by construction and Van Leer proposed
a simple way of insuring monotonicity. One of the scheme, introduced
independently in the atmospheric community by Prather
(1986, J. Geophys. Res., 91:6671-6681), is now considered
as a reference in the  GCM community.
An important aspect of the present work is to perform test simulations
with various spatial resolutions in order to compare the various schemes
at a comparable numerical cost.
 The result is that higher order schemes
are much more accurate than lower order at a given spatial resolution but
much more comparable when the lower order schemes are run on a finer grid
to make the numerical costs equivalent.
 Moreover, the higher moments of the tracer
distribution introduced in the more sophisticated schemes become an issue
when other processes such as chemistry or turbulent mixing are accounted for.
Finally, we suggest that Van Leer scheme I is well suited for transport
of trace species by 3 Dimensional atmospheric winds.
We show the results of applications
to the transport of radon in the GCM of Laboratoire de Météorologie Dynamique.
The GCM implementation of Van Leer scheme I is conservative, positive
and monotonic, and it does not modify a uniform tracer distribution.
\end{abstract}