Permeability Anisotropy in Brazilian Presalt Carbonates at Core Scale Using Digital Rock Petrophysics

Presalt carbonate reservoirs exhibit high textural heterogeneity and a complex porous system with pores at different length scales. Absolute hydraulic permeability (K_abs) is a property that depends on the measurement direction and observation scale and is, therefore, influenced by horizontal barriers, fractures, and preferential flow paths. The anisotropy of K_abs is an important parameter in reservoir modeling, and its determination is a challenging task that requires the combination of laboratory measurements with well-log data at different scales. This study proposes a method for upscaling measurements from routine core analysis (RCAL) on horizontal and vertical plugs using medical computed tomography in core samples, employing hydraulic flow unit classification and machine-learning techniques to obtain directional K_abs volumes. Flow simulations are performed on these local permeability volumes in two orthogonal horizontal directions and in the vertical direction, with different vertical investigation windows in order to further upscale the permeability information to the well-log scale. The vertical (K_v) and horizontal (K_h) permeability logs are evaluated by facies and compared to RCAL data from plugs and whole core samples. We hypothesize the existence of distinct hydraulic flow units for different directions in the upscaling methodology. The lack of laboratory data at decimeter scales is a limitation when evaluating the results, but the obtained products are satisfactory in the performed comparisons to available RCAL data. A reduction in the K_v/K_hmax ratios is observed with increasing vertical investigation window. The medians of these results are less than 1% in spherulites and vary between 35.64 and 24.33% in stromatolites. There is a higher dispersion of permeability values in spherulites, reflecting the presence of two distinct groups within these facies due to different diagenetic effects, besides a tendency towards isotropy in the horizontal directions. The lower performance in the K_v estimates is related to the inherent uncertainties of the data in this direction, such as the lower CT vertical resolution and the smaller amount of RCAL data. The results can serve as input for predicting K_v at the reservoir scale, either by directly using the permeability logs or by evaluating the behavior of K_v/K_h in facies-based simulations with varying vertical thicknesses. This new approach can be key for accurately upscaling K_v in three-dimensional (3D) reservoir flow models, adjusting the average properties on the sizes and geological classifications of the simulation grid cells.