This paper describes a multiscale (from global to micron) model for the prediction of atmospheric corrosion. The model has a modular structure, in which the higher scales set the boundary conditions for the lower scales, and the lower scales alter some of the constants in the upper scales. The model has primarily been designed for Australian conditions and so focuses on corrosion by marine aerosols. The upper level modules look at aerosol production by oceans and surf beaches, salt transport and deposition, and cleaning events such as rain and wind, to provide an estimate of salt retention on surfaces. Separate modules that define surface temperature, surface relative humidity, and wetting and drying of deposited hygroscopic salts, enable the prediction of the (three-hourly) 'state' of a surface, where 'state' is defined as dry, wet from rain or wet from the wetting of hygroscopic salts. The state model is combined with a damage model to estimate the progression of damage with time. Currently, damage models are either probabilistic (define the occurrence, growth or death of pits as probability functions) or empirical (define a single relationship between mass loss in a given state on the basis of measured data) in nature, but new experimental and modelling research is being undertaken to develop first-principle models of corrosion under established oxide films. Crown Copyright (C) 2010 Published by Elsevier Ltd. All rights reserved.