Under repository conditions, it is likely that radioactive waste canisters will be subjected to both uniform and localised corrosion. Localised corrosion can take many forms depending on the precise physical and chemical environment of the metal at the time of attack, but generally the rates of penetration are much greater than those associated with uniform corrosion. The most likely forms under aerated repository conditions are pitting and crevice corrosion. To ensure adequate radionuclide containment an allowance for these rates must be included in the final canister dimensions. It is considered the best way to predict long term penetration rates is to develop mathematical models, which include all the physical and chemical processes necessary to describe the system and use experimentally determined input parameters relating to these processes. These models must then draw on further experimental data for validation over short timescales. In this paper, we discuss several techniques of modelling long term pit propagation in waste canisters. The complexity of the problem has lead to a number of physical and chemical approximations in the modelling. We investigate the applicability and ranges of validity of several of the more common approximations, both in our own models and in the literature, and compare the predictions with experimental pit growth rates. An investigatation of the sensitivity of the models to the various empirical input parameters indicates which need to be determined most accurately.