Rotating packed beds (RPBs), as a type of process intensification technology, are promising to be employed as high-efficiency CO2 absorbers. However, the detailed understanding of the liquid flow in the RPB is still very limited. The complex and dense packing of the bed and the multiscale of the RPB make it very difficult to perform numerical simulations in detail, in particular for full 3D simulations. In this paper, a mesoscale 3D CFD modelling approach is proposed which can be used to investigate the liquid flow in both laboratory- and large-scale RPBs in detail and accuracy. A 3D representative elementary unit of the RPB has been built and validated with experimental observations, and then it is employed to investigate the gas-liquid flows at different locations, across a typical RPB, so that the overall characteristics of the liquid flow in the RPB can be assembled. The proposed approach enables the detailed prediction of the liquid holdup, droplets formation, effective interfacial area, wetted packing area and specific surface area of the liquid within real 3D packing structures throughout the bed. New correlations to predict the liquid holdup, effective interfacial area, and specific surface area of the liquid are proposed, and the sensitivities of these quantities to the rotational speed, liquid flow rate, viscosity and contact angle have been investigated. The results have been compared with experimental data, previous correlations and theoretical values and it shows that the new correlations have a good accuracy in predicting these critical quantities. Further, recommendations for scale-up and operation of an RPB for CO2 capture are provided. This proposed model leads to a much better understanding of the liquid flow behaviours and can assist in the RPB optimisation design and scaling up. (C) 2019 Elsevier Ltd. All rights reserved.