Carbon dioxide (CO2) capture and storage (CCS) in unmineable coal seams is regarded as one of the possible approaches to mitigate the ever increasing CO2 concentration in the atmosphere resulting from human activities since the Industrial Revolution. Injection of CO2 into unmineable coal seams not only provides a solution for long term storage of CO2 but it also provides the added advantage of enhancing coalbed methane recovery. Adsorption is the main trapping mechanism for CO2 storage in coal seams where it constitutes to about 95-98% of total storage. Other trapping mechanisms include gas trapped within the matrix structure, free gas and CO2 trapped as a solute in the pore water. Coal is usually highly heterogeneous and contains pores of different sizes: micropores, mesopores and macropores. The physical properties such as permeability, which usually changes with depth and the degree of cleating, complicates the storage capacity estimation process. Injection of highly dense phase CO2 may offer higher storage capacity because of its higher density compared to gaseous CO2. However, there is a lack of verified CO2 storage capacity estimation methodology for coalbeds. Computing storage potential of CO2 is not straightforward due to the highly variable coal properties even in the same coal seam. Therefore, in this paper a statistical framework for estimating the CO2 storage capacity in coal seams is presented with the emphasis on highly dense CO2 conditions. The approach is based on earlier studies, which utilise important in situ parameters to estimate storage capacity in coal seams. These parameters include volatile matter content, moisture, ash, pressure and temperature. Furthermore, several widely used adsorption models for single-and multi-component gas are reviewed. The ability of the various models in predicting the adsorption capacity for different coal types and under various in situ conditions was examined. Dataset consists of adsorption data representing 69 coal types having vitrinite reflectance ranging from 0.25% to 3.86%. Results of analyses of this dataset showed that better estimation can be obtained by expressing adsorption capacity as a power function of pressure rather than assuming a linear relationship between adsorption capacity and pressure while keeping other important parameters unchanged. (C) 2011 Elsevier Ltd. All rights reserved.