We introduce in this paper an algorithm for integration of the transpo
rt-diffusion equations by combining the Galerkin finite element method
with the discretization of the total derivative along the characteris
tics. The present algorithm is an extension of the one proposed by the
author [R. Bermejo, Numer. Math., 60 (1991), pp. 163-194] to integrat
e the transport equation. The algorithm consists of a convective-diffu
sive splitting of the equations along the characteristics. Such a sche
me permits an efficient evaluation of inner products of functions that
take their values in different partitions (grids). The convective sta
ge of the splitting is evaluated in a Q(1)-conforming finite element s
pace, while the diffusive stage can be approximated in higher-order (Q
(2) or Q(3)) finite element spaces. We analyze two time-discretization
schemes for the diffusive stage, namely, the backward Euler and Crank
-Nicolson schemes. Our L(2)-norm estimate shows that, provided h = O(D
elta t), the backward Euler scheme combined with a Q(1)-conforming fin
ite element yields optimal error estimates, whereas the Crank-Nicolson
scheme is optimal when combined with a Q(2)-conforming finite element
.