Optimal parameter updating and appropriate 4D seismic normalization in seismic history matching of the Nelson field

History matching of reservoirs is very important in the oil industry because the simulation model is an important tool that can help with management decisions and planning of future production strategies. Nowadays, time-lapse (4D) seismic data is very useful for better capturing the fluid displacement in the reservoir, especially between wells. It is now common to integrate 4D seismic with production data in order to constrain the simulation model to both types of data. This thesis is based on a technique for automatic production and seismic history matching of reservoirs by. This technique integrates various tools such as streamline simulation, parameterization via pilot points and Kriging and geo-body updating, a petro-elastic model and the neighborhood algorithm, all in an automatic framework. All studies in this thesis are applied to the Nelson field but the approaches used here can be applied to any similar field. The history matching aim was to identify shale volumes and their distribution by updating three reservoir properties, net:gross, horizontal and vertical permeability. All history matching studies were performed in a six years production period, with baseline and one monitor seismic survey available, and then a forecast of the following three years was made with a second monitor for comparison. Various challenges are addressed in this thesis. We introduce a streamline guide approach in order to efficiently select the regions in the reservoir that have a strong influence on production activity of the wells and 4D seismic signature. Updating was performed more effectively compared to an approach where parameters were changed everywhere in the vicinity of the wells. Then, three parameter updating schemes are introduced to effectively combine various reservoir parameters in order to capture correctly the flow behaviour. The observed 4D seismic data used in this study consisted of relative pseudo-impedance with a different unit compared to synthetic impedance data. This challenge was addressed by introducing normalization. 4D predictions in the vertical well locations and full field simulation cells used in the normalization study and we observed different level of signal/noise ratio in normalized observed 4D maps at the end of study. We include the normalized 4D maps in history matching of the field and we observed that normalization very important. We also compared the seismic and production history matching studies with a case where seismic was not included in history matching (production history matching). The results show that if 4D data is normalized appropriately, the reduction of both seismic and production misfits is better than the production only history matching case