Looking through different literature articles and books concerning seismic inversion - mostly post-stack - phrases like "trace-based" and "model-based" keep appearing. I know that model-based in when lower-frequency info from either stacking velocities or low-pass sonic/rho logs are used. However, I do not know what trace-based inversion means or how it's performed. Can someone please help me with this?
I too have found this nomenclature rather confusing. I am pretty sure that trace-based inversion is merely where each seismic trace is inverted independently of surrounding traces. This would take into account most deterministic/probabilistic inversion types (e.g. model, coloured, sparse, etc). In contrast, geostatistical inversion would not strictly be trace-based inversion, as it uses a spatial and temporal variogram to takes into account the expected results of surrounding traces during the inversion, even though it does invert on a trace-by-trace basis. Confusing.
The following paper describes trace and geostatistical inversion methods: Germán Merletti, and Julio Hlebszevitsch. 2003. Geostatistical Inversion for the Lateral Delineation of Thin Layer Hydrocarbon Reservoirs: A Case Study in San Jorge Basin, Argentina. SEG 2003.
I've found the following paper very useful for understanding seismic inversion (geostatistical isn't discussed, though): Dennis Cooke and John Cant. 2010. Model-based Seismic Inversion: Comparing deterministic and probabilistic approaches. CSEG Recorder, Vol. 35, No. 4.
1$\begingroup$ Just to add, trace-based inversions tend to assume that the data can be modeled as a convolutional process. This is a large simplification from the actual wave propagation, seismic acquisition, and data processing steps that need to occur before obtaining each post-stack trace. $\endgroup$– AntonioJun 29, 2017 at 1:04