To interpret geophysical anomaly maps, it is necessary to filter out regional and sometimes noise components. Each measured value in a gravity survey consists of different components. Upward continuation (UC) is one of the most widely used filters. The shortcoming of this filter is not to consider the spatial structure of the data, and also the fact that the trial and error approach and expert's judgment are needed to adjust it. This study aims to compare the factorial kriging analysis (FKA) and UC filters for separation of local and regional anomalies in the gravity data of a hydrocarbon field in the southeast sedimentary basins of the East Vietnam Sea. As shown in this paper, FKA method permits to filter out all of the identified structures, while the UC filter does not possess this capability. Therefore, beside general and classic filtering methods, the COMPARISON OF FKA METHOD AND UPWARD CONTINUATION 399 FKA method can be used as a strong method in filtering spatial structures and anomaly component.

In this paper, two methods of kernel bandwidth and wavelet transform are used for simultaneous upscaling of two features of hydrocarbon reservoir. In the bandwidth method, the criterion for upscaling is the cell variability, and by calculating the optimal bandwidth and determining the distance matrix, the upscaling process is performed in a completely non-uniform and unregularly manner. In areas with extreme variability, the bandwidth is considered small enough to maintain the fine scale characteristics of model. Conversely in homogenous areas, with the choice of large bandwidth, the maximum rate of upscaling will occur. The bandwidth upscaling algorithm is an iterative and hierarchical algorithm. The bandwidth method, unlike conventional scale-up methods, focuses on how to upgrid cells and, by determining the optimal averaging window, we will have the least loss information for the fine scale model. Upscaling is a pre-processing to building a simulator model with lower cell number, and thus, reducing volume and computational cost, while maintaining and retaining the basic information of the fine model. Due to the various variability of the reservoir features, the attribute upscaling pattern differs, and in order to show the variability of two features in the upscaling model simultaneously, it is suggested in this paper to upscale two features simultaneously. For simultaneous upscaling, we applied two different approaches; minimum and maximum bandwidth. Moreover, wavelet transformation is applied to upscaling the model. Then, as a result, the variance of the scale-up models based on wavelet is about one-third of the variance of the bandwidth method. Simulation results show that the bandwidth method is a good approach for upscaling the heterogeneous reservoirs.