The integration gap
Matt HallAgile teams have lots of ways to be integrated. They need to be socially integrated: they need to talk to each other, know what team-mates are working on, and have lots of connections to other agile teams and individuals. They need to be actively integrated: their workflows must complement one another's. If the geologist is working on new bulk density curves, the geophysicist uses those curves for the synthetic seismograms; if the geophysicist tweaks the seismic inversion result, the geomodeller uses that volume for the porosity distribution.
But the agile team also needs to be empirically† integrated: the various datasets need to overlap somehow so they can be mutually calibrated and correlated. But if we think about the resolution of subsurface data, both spatially, in the (x,y) plane, and vertically, on the z axis, we reveal a problem—the integration gap.
This picks up again on scale (see previous post). Geophysical data is relatively low-resolution: we can learn all about large, thick features. But we know nothing about small things, about a metre in size, say. Conversely, well-based data can tell us lots about small things, even very small things indeed. A vertical well can tell us about thick things, but not spatially extensive things. A horizontal well can tell us a bit more about spatially large things, but not about thick things. And in between this small-scale well data and the large-scale seismic data? A gap.
This little gap is responsible for much of the uncertainty we encounter in the subsurface. It is where the all-important well-tie lives. It leads to silos, un-integrated behaviour, and dysfunctional teams. And it's where all the fun is!
† I've never thought about it before, but there doesn't seem to be an adjectival form of the word 'data'.
Reader Comments (4)
If one believes in fractal geometry, is it possible to close this 'integration gap'?
Thanks for the comment, Tooney... I think you're saying 'If geology is scale-free, then there should be no gap', or something like that? I like that thought. I don't think the gap is uncrossable. I can think of a few ways to try to mitigate it:
- Notice how horizontal wells overlap spatially with seismic, if only they had more depth resolution. Doing a small scale seismic reflection experiment along a horizontal well wouldn't be too hard with multi-component receivers. A horizontal seismic profile (HSP).
- Maybe VSPs can be interbred with cross-well methods, with source and receivers all subsurface, to get a super-high resolution VSP. A processing nightmare, I'm sure.
- Build conventional logging tools that try to cross it. We already know we need to get past mud cake or invaded zones, but maybe we need to think about how to get deeper still into the formation. For some low-spatial-frequency goodness. Perhaps that's with sonic tools, but perhaps not.
- I don't think I've ever read about anyone trying to measure acoustic properties along the z-axis of whole cores. Whole cores, not little plugs or sections of core. Before they are sawn up, a core might be a couple of metres long or more. That's creeping up to the seismic band in some places.
It's fun to think about. I was particularly struck, whilst drawing the diagram, how the gap is just about exactly at the scale of a person. It's also the domain of outcrop, and is one reason field work and field trips are so important, even to subsurface geoscientists.
Perhaps you know if anyone is trying any of this sort of thing already?
I was thinking of the work Dvorkin et al have done ( http://pangea.stanford.edu/~jack/FBSept08_03.pdf ) ( The Leading Edge; January 2009; v. 28; no. 1; p. 110-115; DOI: 10.1190/1.3064155 ) with 'cat scanning' small rock samples, building a 'digital' rock from the scan, calculating rock properties from this 'digital' model, then 'upscaling' these properties to 'reservoir scale'...on the basis of a 'fractal' behaviour (statistical stationarity) of rock properties in the reservoir....
The upscaling work is nice as far as it goes, but it seems like it is dependent on models of nature. The tools in my graph are all first order measurements, though some of them are quite abstract (like seismic, for example). I wonder if there's a way to get across the gap with data, rather than just models. Perhaps both approaches are needed: modelling/upscaling because it is practical, and hard data to corroborate it. Good discussion.