Dynamic reservoir simulation models have become an essential tool in oil field management. They can be used to explore the impact of alternative development and operating strategies, and can even be used to search for strategies that are optimal in some desired sense. Robust strategies can be obtained by use of an ensemble of models that are all consistent with available data but that adequately represent the uncertainty. In this paper we demonstrate an application of a workflow to obtain multiple models that are all consistent with historic time-lapse seismic data. We show how these models can subsequently be used to obtain optimal recovery strategies, and discuss how this approach can be extended to optimization under uncertainty.The history matching step is performed using an ensemble-based methodology and an efficient parameterization of the time-lapse seismic data. The optimization step takes the history matched model (or model ensemble to account for all uncertainty) as input. An approximate gradient computation provides improved strategies. The history matching workflow is demonstrated on a representative sector of a field that is developed with both vertical and horizontal wells and is produced with simultaneous injection of water and CO 2 .Evaluation of the history matching step indicates a good match with the seismic data, indicating the power of the parameterization method for handling very large numbers of seismic data. It is discussed that production data can be matched simultaneously, given proper characterization of measurement errors. The optimization step is performed assuming an oil recovery scenario based on alternating water and CO 2 injection. Appropriate cost models may be employed to arrive at operating scenarios that result in the maximum expected economic value. Results will depend on, amongst other aspects, the chosen economic model, indicating that best operating practices will be region specific.The combination of advanced yet practical workflows for assisted history matching and recovery optimization is the result of many developments over the past 10 years. We argue that the cumulative result of these efforts provides significant value to field developments by enabling consistency with measurements, reduction of uncertainties and improved optimization of operational strategies.