Coal Seam Gas (CSG) is a significant resource; it makes up 90% of the gas production in Queensland, Australia, most of it from the Surat Basin Walloon Coal Measures, (WCM). However, coal connectivity and distribution are poorly understood resulting in production and modelling challenges. Simplistic, homogeneous models cannot adequately explain volumes or flow rates of gas and water. Initially, the focus of the industry was on optimizing drilling, often without the aid of seismic data, relying on correlations based on wireline logs and cores alone. Those correlations were often based on lithology, and connected sand with sand and coal with coal. Those models seem to overestimate flow between wells and volumes of hydrocarbons or water (Cardwell, 2018).
Here, log interpretation is integrated with 3D seismic using a sequence stratigraphic framework. Special attention is placed on the architecture of the WCM revealed by seismic. Seismic mapping of individual coal seams in the Springbok Sandstone or the WCM is difficult on a field scale. Most coals are below seismic resolution, yet have a strong influence on seismic amplitudes. Four methods of spectral decomposition are compared and evaluated on how well they predict known geology at many well locations throughout the 3D seismic survey.
Sequence stratigraphy applied to core and log interpretation integrated with seismic produces an improved view of coal distribution and geometries. Coals are mostly thin and discontinuous and segregated by channels cutting through them that appear to be meandering through flood plains. Coal forms along channels and sands are likely to be present in point bars. Therefore, sands or coals identified in wireline logs are not likely to be connected to a large sand sheet, but are rather individual channels or sand bodies likely to be hydraulically isolated.
In the next phase, it will be investigated how results of spectral decomposition can be used inform geostatistical algorithms for modelling to produce more representative static geological models.
