We assembled a set of models that allows investigation of local variables that are difficult to measure, validation of mechanistic physical models, and comparison of different numerical solutions. Population balances (PB) for bubbles were combined with local flow modelling in order to investigate G-L mass transfer in an air-water system. Performance of three different impeller geometries was investigated: Rushton (RT), Phasejet (PJ) and Combijet (CJ). Simulations were compared against experimental mixing intensity, gas hold-up, vessel-averaged volumetric mass transfer rates (K(L)a), and local bubble size distributions (BSDs). The simulations qualitatively predict k(L)a's with different impellers at the fully dispersed flow region and gave new insight on how k(L)a is formed and distributed in the stirred vessels. The used bubble breakage and coalescence models are able to describe both air-water and viscous non-Newtonian G-L mass transfer. Difference between experimental mass transfer rates of the three impellers was within experimental error, even trough the flow patterns, gas distribution, and local BSDs differ considerably. The population balance for bubbles was modelled in two different ways, with multiple size groups (MUSIGs) and with the bubble number density (BND) approach. MUSIG calculations took over twice as much computational time than BND, but there was little difference in the results. The Rushton turbine k(L)a was described with best accuracy, which is not surprising since most phenomenological models are fitted based on RT experiments. We suggest that these models should be validated over a wider range of vessel geometries and operating conditions. (c) 2008 Elsevier B.V. All rights reserved.