Application of Geostatistical Seismic AVA Inversion for Shale Reservoir Characterization and Brittleness Prediction with Machine Learning

Abstract

Summary In this study we present an application of geostatistical AVA seismic inversion method for characterization of a unconventional Lower Paleozoic shale reservoir in Northern Poland. The target formations are of a small thickness (up tp 25 meters) and deeply buried (ca. 3 km) what makes their delineation and characterization especially difficult. An application of the iterative geostatistical AVA inversion method allowed for obtaining the high-resolution density, P-wave and S-wave velocity models together with the assessment of the uncertainty on the predictions. The obtained elastic property models were compared with the results of the deterministic simultaneous Amplitude-versus-Offset inversion proving that the application of a such sophisticated (geostatistical) inversion technique is a must while dealing with the thin and highly variable layers. The inverted elastic models where further used to improve the prediction of a spatial distribution of the brittleness index with a machine learning (PSVM) algorithm by integrating well-log data and seismic rock property volumes.