Naturally fractured reservoirs are currently being considered as potential candidates for geological storage of CO2. Simulations of fractured reservoirs are notoriously challenging. Dual-porosity models are a cost-effective way of representing fractured reservoirs whose fundamental ingredient are transfer functions that represent fracture-matrix interaction in an up-scaled manner. In order to develop accurate transfer functions, it is essential to capture the underlying physics of the fluid transfer. Material properties and dominant processes in CO2 storage differ from the ones in conventional production environments. In this contribution we develop a novel transfer function that accounts for these differences. We first analyse the simplifying hypotheses that are commonly made in the current existing transfer functions. Those simplifications lead to inaccurate results in the context of CO2 storage. We then develop a transfer function for buoyancy displacement based on the timescale of the one-dimensional equation for immiscible two-phase flow in porous media. We analyse how the newly developed transfer functions improve over the current existing ones in simple matrix-block geometries. The results are evaluated against high-resolution numerical simulations of matrix blocks considering realistic physical properties of CO2/Brine systems and fractured rocks. Our results show that the developed transfer functions are able to represent accurately the basic physics of the process, and improve over other existing transfer functions in the literature. The transfer functions are also implemented in a dual-porosity simulator and different CO2 injection scenarios are tested. We show that a careful design of the injection schedule may increase the mass of CO2 that is stored in the matrix block.@en