A generalized fluidized bed reactor model which covers the three fluidization flow regimes most commonly encountered in industry (bubbling, turbulent and fast fluidization) is proposed. The model is based on probabilistic averaging shown previously (Thompson, Bi, & Grace, Chem. Eng. Sci. 54 (1999).2175; Grace, Abba, Bi, & Thompson, Can. J. Chem. Eng. 77 (1999) 305) to be applicable over a range of superficial gas velocities. In this paper, we extend the model to cases where the volumetric gas flow changes appreciably due to variations in molar flow, pressure and temperature. For the air-based oxy-chlorination process as a case study, it is shown that the volume change affects both the hydrodynamics and the reactor performance. Because the reactions are rapid, almost complete conversion of ethylene is attained immediately above the distributor resulting in an similar to 25% reduction in volumetric flowrate. Using the probabilistic averaging technique, the model tracks the probability of being in the bubbling, turbulent and fast fluidization regimes along the reactor height. The impacts of temperature and pressure variations are also examined. The variable density model gives predictions which compare well with commercial data; ignoring density variations leads to significant underprediction.