Icing codes are playing a role of increasing significance in the design and certification of ice protected aircraft surfaces. However, in the interest of computational efficiency certain small scale physics of the icing problem are grossly approximated by the codes. One such small scale phenomena is the effect of ice roughness on the development of the surface water film and on the convective heat transfer. This study uses computational methods to study the potential effect of ice roughness on both of these small scale phenomena.;First, a two-dimensional condensed layer code is used to examine the effect of roughness on surface water development. It is found that the Couette approximation within the film breaks down as the wall shear goes to zero, depending on the film thickness. Roughness elements with initial flow separation in the air induce flow separation in the water layer at steady state, causing a trapping of the film. The amount of trapping for different roughness configurations is examined.;Second, a three-dimensional incompressible Navier-Stokes code is developed to examine large scale ice roughness on the leading edge. The effect on the convective heat transfer and potential effect on the surface water dynamics is examined for a number of distributed roughness parameters including Reynolds number, roughness height, streamwise extent, roughness spacing and roughness shape. In most cases the roughness field increases the net average convective heat transfer on the leading edge while narrowing surface shear lines, indicating a choking of the surface water flow. Both effects show significant variation on the scale of the ice roughness. Both the change in heat transfer as well as the potential change in surface water dynamics are presented in terms of the development of singularities in the surface shear pattern. Of particular interest is the effect of the smooth zone upstream of the roughness which shows both a relatively large increase in convective heat transfer as well as excessive choking of the surface shear lines at the upstream end of the roughness field. A summary of the heat transfer results is presented for both the averaged heat transfer as well as the maximum heat transfer over each roughness element, indicating that the roughness Reynolds number is the primary parameter which characterizes the behavior of the roughness for the problem of interest.