Speaking at the CSIG’11: ‘Big Data and Big Computing in the Geosciences’ event Michael Gurnis, director of Caltech’s seismological laboratory, presented an overview of 3D and 4D dynamic earth models. These are used to study the ‘large scale space time pattern of the geological record, drivers of sea level change and the forces behind plate tectonics.’ Such processes are best understood by connecting disparate observations in a 4D dynamic earth model—going beyond plate tectonics to model the whole earth from core to surface in a time lapse, 4D supermodel. This, as one can imagine, raises several software and computational ‘issues.’ Gurnis reviewed the state of play, with academic efforts such the GPlates (www.oilit.com/links/1109_26) Python framework for plate visualization, CitcomS (www.oilit.com/links/1109_27 and Rhea/P4est (www.oilit.com/links/1109_29).
The 80 megabyte PowerPoint is a free download (www.oilit.com/links/1109_28) and provides some great imagery from these models. Caltech’s own ‘Rhea’ model is an ambitious attempt to perform full physics modeling across the whole earth. Rhea uses adaptive mesh refinement to adapt model resolution (AMR) to local complexity. A 22 million initial mesh translated into a 65 million cell model that took 15 hours to run on the 6,000 core Texas Advanced Computing Center’s ‘Ranger’ cluster. More from CSIG on www.oilit.com/links/1109_27.
© Oil IT Journal - all rights reserved.