Karelia is a 2nd year MS student set to graduate in the Spring of 2020. Karelia is working to understand how a variety of machine learning techniques in combination with seismic attribute analyses can improve the discrimination of deepwater channel facies.
From her 2019 SEG abstract: Seismic facies identification in a deepwater channel complex applying seismic attributes and unsupervised machine learning techniques. A case study in the Tarinaki Basin, New Zealand.
Karelia La Marca-Molina*, Clayton Silver, Heather Bedle, and Roger Slatt
Analysis of the Pipeline dataset, a high-resolution 3D seismic volume, in the Taranaki Basin, Western New Zealand, allows for the identification and characterization of several deepwater architectural elements within a
channel complex. This study focuses on the delineation and characterization of architectural elements using seismic attributes and unsupervised machine learning techniques such as self-organized maps (SOM). These techniques provide a quick and detailed interpretation to better understand the geomorphology and distribution of the seismic facies within the channel complex. Seismic
attributes such as sweetness, co-rendered with a sobel filter (coherence) allow for the delineation of the geometry of the channel complex and identification of possible sand-rich lithofacies. In addition, we explore the application of
clustering analyses to discriminate seismic facies within each channel complex. Three predominant groups are obtained using this classification technique a) Sand-rich deposits b) Siltstone deposits c) Mud-rich deposits. The proper selection of input seismic attributes in SOM along with the use of display options such as horizon probes proves to be a quick workflow to classify and characterize facies within reservoirs.
For the 2019-2020 academic year, Karelia is serving as the vice- president of OU’s SEG Student Chapter.
LinkedIn Profile: Karelia La Marca