Multiple scenario modelling methodology has been used to study gas injection in a fractured carbonate reservoir to facilitate an increase in productivity and ultimate field recovery as a result of reservoir repressurisation. The complex physical processes involved in the production mechanism, and a relatively sparse data set, highlighted the underlying problem of making predictions based on reservoir models for which a large degree of uncertainty exists. These uncertainties have a significant impact, both in the understanding of the reservoir fracture geology and in the data used for describing the fluid flow processes. A novel approach was developed to represent uncertainty ranges in model forecasts in which multiple history-matched simulation models were constructed to represent the impact of a range in each of the uncertainty parameters on the forecast. The methodology builds on an approach previously reported in Bentley and Woodhead (Ref.1). The results were then combined statistically into probability distributions of recovery profiles. The results show clearly the large impact of data uncertainty on incremental recovery arising from the gas injection scheme. The treatment of uncertainty presented here is a step forward in the modelling methodology for mature assets in which multiple scenario modelling has to be constrained by the necessity to history match all model realisations against production data.