Predicted ice shapes obtained with a fully automated multi-stepping procedure implemented in the ice prediction code ICECREMO2 are discussed in this paper. Several multi-step algorithms are used to generate ice layers on cylinders in rime and glaze ice conditions. Although all multi-step algorithms produce acceptable rime ice results, a multi-step approach based on an ice height criterion appears as the most efficient method for glaze ice predictions. The outcome is compared with results obtained for a NACA0012 airfoil geometry. The influence of the un-iced substrate shape is investigated, general restart criteria for multi-step height-based icing codes are presented and discussed.= ice layer thickness, m b max = maximum ice thickness allowed per step, m c = chord or diameter length, m c i = specific heat capacity of ice, J · kg −1 · K −1 d d = droplet diameter, m g = acceleration due to gravity, m · s −2 h = water layer thickness, m L = typical length of the geometry, m LWC = cloud liquid water content of air, kg · m −3 n step = number of steps P a = ambient pressure, Pa Q x = x component of the water flux, m −2 · s −1 Q y = y component of the water flux, m 2 · s −1 T = temperature in the ice layer, K T ∞ = freestream temperature, K t = time, s t exp = exposure time, s V ∞ = freestream velocity, m · s −1 X = ratio of ice height to chord length β = collection efficiency θ = temperature in the water layer, K κ i = thermal conductivity of ice, W · m −1 · K −1 κ w = thermal conductivity of water, W · m −1 · K −1 μ i = dynamic viscosity of ice, kg · m −1 · s −1 μ ω = dynamic viscosity of water, kg · m −1 · s −1 ρ i = density of ice, kg · m −3 ρ w = density of water, kg · m −3 τ x = x component of shear stress, Pa τ y = y component of shear stress, Pa