An abundance of research works on the lattice Boltzmann method (LBM) have been published, but a clear comparison of simulation results with conventional computational fluid dynamics (CFD) and experimental measurements has not been forthcoming. Moreover, the evaluation, necessary to gauge the use of LBM as an engineering analysis tool, is clouded by the challenge of converting the units of a macroscopic engineering problem to those appropriate for the mesoscopic LBM simulation using lattice units. Here, we simulate natural convection in a 2D porous cavity over a range of Darcy, Prandtl, Rayleigh, and Reynolds numbers with temperature-dependent fluid properties. We use a dual distribution function approach to simulate the evolution of temperature and the velocity field in time in an enclosure filled with an isotropic, homogeneous, and non-deformable porous medium consisting of solid particles. While the LBM algorithm is simple and easy to program, it can violate the physical conservation laws and exhibit various numerical instabilities−attributes which complicate its usefulness as an engineering tool that is on par with commercial macroscopic Navier-Stokes solvers at this juncture.