THE CHALLENGE

Seismic inversion and facies classification are two important techniques to understand the subsurface and reservoir characteristics for geothermal energy development.

The main challenges in seismic inversion are the non-uniqueness of the solution, the low-frequency component in the data and the prior information required. There are often multiple possible solutions that can fit the same data and the low-frequency content and prior information are often lacking. Facies classification in seismic data, on the other hand, involves grouping similar rock units based on their seismic reflection pattern and underlying physical properties. The challenge here is that rock properties can vary significantly within a single facies, often below seismic resolution and masked by noise, making it difficult to accurately define and differentiate between them. Both techniques require careful interpretation and validation, and a thorough understanding of the geological setting in order to make an accurate estimate of the formation properties.

Machine Learning (ML) has made significant steps in establishing the relation between seismic data and the causative rock parameters. ML can predict at great speed causative rock parameters from seismic data, but much ambiguity remains.