The scale of heterogeneities in reservoirs is often smaller than the grid s
ize used in large-scale reservoir simulations. Relative permeabilities have
the foremost effect on fluid flow and small-scale fractional flow must be
upscaled to the,grid size in flow simulations. Thus, systems with fractiona
l flow heterogeneities, i.e. relative permeability variations, have to be s
olved and effective relative permeability functions determined. In this pap
er, benchmark analytical solutions are developed for a system consisting of
two media in series where each medium is characterized with a different se
t of relative permeabilities, residual saturations and porosities. The anal
ytical solutions show a significant discontinuity in the saturation and con
centration profiles at the interface of the two media. Numerical results us
ing several weighting schemes are compared against the analytical solutions
for two-phase immiscible and partially miscible systems. It is shown that
single-point upstream weighting requires about 100 grid blocks to capture t
he discontinuity at the interface, whereas third-order TVD weighting requir
es much fewer. Lastly, the validity of the JBN method and harmonic averagin
g for determination of effective relative permeabilities and overall pressu
re drop is tested. Both the JBN method and harmonic averaging cannot reprod
uce the pressure drop across the composite media prior to water breakthroug
h.