The purpose of fine geological models is to retain the vertical heterogeneity measured by well logs and the areal heterogeneity resulting from the different depositional environments. These models can have several million grid cells. The purpose of having upscaled coarser models is to perform fluid flow simulations in an efficient manner. In recent years several papers have presented methods for choosing an optimal upscaled grid suitable for dynamic simulation, given a multi-million cell geological model as a starting point. This thesis describes the effect of the direction in which upgridding is performed on the optimal grid, starting from the upgridding in one-dimension (1D) in all directions (i.e., in either the I direction, J direction, or K direction), moving to upgridding in two-dimensions (2D) (i.e., in the I and J direction while retaining K number of layers, in the I and K direction while retaining J number of rows, and in the J and K direction while retaining I number of columns), ending with three-dimensions (3D) (i.e., in the I, J, and K directions). Streamline based flow simulation (Eclipse FrontSim) was used to perform dynamic simulations on different upscaled grids as well as on the fine model (Society of Petroleum Engineers 10th Comparative Solution Project). The performance of wells (flow rate and water cut), allocation factors between injector/producer pairs and the drainage areas were compared.