Improving Reservoir Quality Prediction Of Microporous Carbonates Using A Multi-Scale Geophysical Data Analysis Approach

Microporosity is volumetrically the most important porosity type in the giant carbonate reservoirs of Arabia and can significantly impact reservoir quality and ultimate oil recovery. Understanding the impact of microporosity on reservoir quality and accurately predicting the spatial distribution of microporosity at the reservoir scale is required to improve the recovery of remaining hydrocarbons from the reservoirs. Utilizing an integrated data analysis approach applied to multi-scale geological and geophysical datasets from micrometer scale SEM imagery to decameter scale seismic data, we predict the distribution of microporosity at the reservoir grid-block scale. We apply the proposed methodology to Arab-D reservoir equivalent outcrop data from the Upper Jubaila Formation, Saudi Arabia. Thirty-five meters of near-surface well core and a 600 m long 2D near-surface p-wave seismic reflection profile were acquired at the outcrop location. From the core, 106 plug samples are analysed to determine porosity and permeability, ultrasonic velocities, powder XRD compositions, and SEM data. With the seismic reflection data, we perform a near-surface colored inversion to obtain a high-resolution acoustic impedance image of the seismic data. The morphology of micrite crystals that host microporosity was characterized and quantified by analysing SEM data using machine learning image classification tools. We use the resulting data to derive robust statistical relationships between microporosity and texture of micrite microcrystals on centimetre scale geophysical properties with a Self-Organizing Map (SOM) approach for data clustering. A Differential Effective Medium (DEM) model enabled us to correlate acoustic impedance and porosity, and distribute porosity across the 2D seismic cross-section. The key depositional lithofacies identified from core descriptions are bioturbated mudstones intercalated by packstones and grain dominated rudstones and floatstones. Image-based machine learning classification results indicate that microcrystals that host microporosity in this formation were typically homogenous in size but varied in morphological aspects such as granularity and angularity. Based on data clustering results, granularity and angularity of microcrystals appear to be the dominant controls over the lab-measured geophysical properties. This effect is manifested as a simple log-linear relationship between porosity and permeability among the major depositional facies. The DEM fitting parameter effectively represents the velocity-porosity relationship and can be used to predict the distribution of porosity across a seismic cross-section. For the first time, an integrated multi-scale data methodology involving machine leaning tools is applied to the Late Jurassic Upper Jubaila Formation outcrop data. Although we demonstrate the proposed methodology using outcrop data, it can be applied to any subsurface reservoir zone dominated by microporosity.