Mathematical models for froth transportation for the froth and attached particles, including froth velocity profile on the cell surface and residence time distribution of the froth, have been developed for the Outokumpu 50 m(3) tank flotation cell used in the primary rougher circuit in the Anglo Platinum Bafokeng-Rasimone Platinum Mine concentrator. The new models take into account both the cell design factors and the cell operating conditions. nu(fh)(r) = J(g)-delta/hf(.)epsilon(f) (.) + (2(.)delta-J(g))/2(.)h(f)(.)epsilon(f) (.) r(0) v(fh)(r) = J(g)-delta/h(f)(.)epsilon(f) (.)r + delta/h(f)(.)epsilon(f) (.)r(e) - J(g)/h(f)(.)epsilon(f) (.) r(e)(2)+r(i)(2)/2(.)r(e) Experimental data were collected from the second cell of the primary rougher circuit at BRPM under a number of different cell operation conditions. The measured froth velocity is compared with the predicted values by the models. It is found that the overall agreement between the measured froth velocity by the SmartFroth program and the simulated froth velocity by the new froth transportation models is reasonable. The froth height above the concentrate launder lip level tended to remain relatively constant under normal cell operating conditions. The froth velocity tended to be linearly correlated with the air rate, as predicted by the model. No evidence suggests that froth velocity vary with the froth height. Froth stability was found to be an important correction factor for froth velocity. (C) 2004 Elsevier Ltd. All rights reserved.