A finite difference model for gas diffusion and bubble trapping in firn is described. The model uses prescribed profiles of density, open and closed porosity, and diffusivity to determine the diffusion and trapping processes. The model is calibrated and tested by using measured air composition in the firn at the DE08-2 site on Law Dome, Antarctica. In particular, we focus on carbon dioxide (CO2), methane (CH4), and sulfur hexafluoride (SF6), which have well-determined atmospheric records (CO2 since 1958, CH4 since 1983, and SF6 since 1978). These trace gases are used to tune the diffusivity-porosity relationship, which is the most uncertain of the model inputs. Modeled trace gas profiles in the DE08-2 firn are improved if allowance is made for reduced diffusion through the most prominent DE08-2 melt layer from the summer of 1989/1990. The relatively rapid growth rate of SF6 in the atmosphere permits good definition of the diffusion reduction due to the melt layer (about 80%). The model quantifies the smoothing effect of the firn diffusion and bubble trapping on atmospheric signals. Gravitational separation in the firn is investigated by comparison of modeled Δ15N2 with observations. The model is used to calculate the isotopic diffusion correction for Δ13CO2 and Δ13CH4. This corrects for the fractionating effects of the firn diffusion process on the different isotopes. The diffusion and gravitational corrections are critical at the measurement precision currently being obtained; for Δ13CO2 the diffusion correction is up to about 10 times the current measurement precision. The diffusion correction is even more significant for Δ13CH4; at over 1? at the bottom of the firn it is more than double the changeover the last decade. The fully corrected Δ13CO2 record from the DE08-2 firn is compared with the history of Cape Grim direct atmospheric measurements with excellent agreement.¿ 1997 American Geophysical Union