Performance prediction for diatomite reservoirs is very challenging because these reservoirs have a low permeability, contain fractures, and are very compactible. Therefore it has been suggested that reservoir simulation is not suitable for forecasting future hydrocarbon production or identifying production drivers because it is very difficult to obtain a good history match, particularly during waterflooding. In this study we describe how it is possible to effectively model a low permeability, highly compactible waterflooded diatomite reservoir. The impact of pressure dependent permeability on production and cumulative oil recovery in diatomaceous reservoirs was also analyzed. The pressure dependent permeability term accounts for the effect of rock mechanical properties on recovery. This important aspect has been neglected so far and we believe it to be one of the reasons causing the difficulties in matching production history. As a key result of our work we present a new numerical simulation model to effectively predict waterflood performance and guide reservoir management in a diatomaceous reservoir. Incorporating an adequate understanding of the geology in the numerical model proved important to be able to predict performance from wells. We hence used the geopseudo approach, a multi-phase flow upscaling methodology, which accurately models small-scale flow and structure interaction at the lamina level. Hydraulic fractures, dual porosity and dual porosity features were also incorporated in the model. The model was validated by history matching and compared to field performance. Predictions have been confirmed by the ongoing waterflood response.@en