Depth of Investigation (DOI) analysis for IP/Resistivity inversions

PeterDiorio

What approaches are people using to evaluate Depth of Investigation (DOI) for VOXI IP/Resistivity inversion?

Does anyone have guidance to help set parameters when using VOXI?
Specifically I am looking for guidance for setting Starting model /Reference model /Parameter Weighting for each of resistivity and IP and formulation of the DOI function, analogous to the method discussed in Oldenburg and Li 1994, Estimating depth of investigation in dc resistivity and IP surveys.

Have people found this effective?

Are there any alternate approaches that others have found useful?

TaronishPithawala

@PeterDiorio, a loaded question indeed! We've been dealing with just such a case with data from Botswana - trying to determine the validity of a deep target recovered by the inversion.

Adjusting the auxiliary parameters that guide the inversion where data sensitivity is weak (see the WGT voxel associated with your conductivity/chargeability inversion), is a good idea. Though this approach may simply highlight what a particular inversion algorithm will want to recover in the absence of data influence - not necessarily the detection limit of your survey for a particular target.

Assuming you know a little bit about the target body of interest, you could forward model a body buried at various elevations and see when it is detected by your survey. This may be the most straightforward way, as it removes all non-uniqueness inherent in the inversion process itself.

I encourage those with far more DCIP experience to chime in - I'm sure there are some good case studies out there!

PeterDiorio

Thanks Taronish. Completing multiple forward models is a reasonable way forward for a specific case but I was actually hoping for something a bit more generic. For example: often the top of a feature is very clear in the model but the bottom is much less well defined. It can be very hard to determine if I am a I looking at depth-limited body or just the fall-off in sensitivity with depth for this particular array, electrical environment and set of inversion constraints. The Oldenburg-Li paper specifies a couple of approaches but I struggle to map their recipes to VOXI parameters.

TaronishPithawala

@PeterDiorio Can you outline the recipe here?

If you're curious about the sensitivity of your model, you can display the weighting used in the inversion for your specific model.

IanMacLeod

This seems to be a challenging requirement. I don't expect this helps for the current VOXI, but I just saw Andrew Pare (Colorado School of Mines) present a nice paper with an interesting approach at the SEG: Improved imaging of sharp boundaries in DC resistivity. Just fuel for thought as we evolve the technology forward.