Lattice gas automata (LGA) were used to estimate errors in transport coefficients, as measured in laboratory experiments with Peclet numbers from 0 to 27.6 (defined relative to channel width), Damk¿hler numbers from 0.18 to ∞, Grashof numbers of 0 or 75, and length/width up to 180. Low Damk¿hler numbers yield long, low-amplitude elution tails, which contain much of the total solute. As a consequence, at Da≈0.18 and KD=8, the solute peak travels at the same speed as the carrier fluid, yielding an apparent KD≈0 after five characteristic diffusion times. Such conditions correspond, for example, to a meter-long path through a 0.5-cm-wide, gas-filled fracture. Buoyancy-enhanced dispersion, found in experiments with horizontal tubes, is confirmed by the LGA analysis; however, a different mechanism is suggested for the enhancement in horizontal fractures. Both kinetic and buoyancy errors can be greatly reduced, or experiments made much smaller, if the first and second moments of a tracer pulse can be measured as functions of time.