VOXI IP - RIP (reconnnaissance IP) data

JoeInman

Has anyone attempted to invert/model RIP data? Specifically RIP data that uses the checkerboard layout; i.e. receiver stations located on a grid say 1 km spacing, as opposed to the wagon wheel layout first utilized and modeled by Kennecott and Gerry Hohmann. I am about to attempt this with a data set...and I am wondering about what is sometimes termed...'pi flips'....when the readings are near pi...3141 mradians. I am planning to invert the individual dipoles at each station rather than combining the two orthogonal dipoles into a single IP value. Obviously EM-coupling must be removed first as well. And what do you do about negative values...negative apparent IP response. Can this actually occur in the real world or is it simply noise? How does Voxi handle this and how did you handle this?

NickWilliams

A few comments Joe:

1. RIP surveys have very little depth resolution, especially on a 1 km grid, so inversion won't add much value - much like for gradient array.
   
2. You definitely need to treat each dipole as its own data point with all 4 TX and RX electrodes entered specifically (don't use pole approximations because of all the angle changes involved)
   
3. Most RIP layouts I have modelled have strong EM-coupling. Theoretically it can be removed by modelling the electrode locations, complete TX wire layout, and the 3D conductivity structure of the earth, but I have yet to see a convincing successful EM-decoupling. Depending on whether you have lots or a few negatives you have to decide whether to leave them in with high error assignments, or remove them. Too many and the inversion will struggle or fail though.
   
4. Negative IP measurements indicate that current is flowing the opposite way from convention. This could possibly happen from certain 3D earth-electrode geometries but more likely it reflects EM coupling, null-coupling, or electrode polarity problems (switched A-B or M-N), none of which are easy to resolve.
   
5. Good luck!
   

RobertEllis2

Regarding negative apparent chargeability your might find this Geosoft Tech Note interesting.

NickWilliams

Nice work Rob. Always glad to see people working on this. For larger dipole sizes I would expect a significant EM response to be present over such a low resistivity earth as your example. Depending on electrode geometry this could cause a large negative apparent chargeability (ie, late off-time) response that would overprint the measured chargeability data and overhwhelm the negative IP sensitivity variations that VOXI-IP and other IP modelling codes can model accurately. You could take your example further by computing the late off-time (IP window) EM response from various grounded dipoles and wire paths, to be sure. I routinely see negative apparent chargeabilities of << -5 mV/V (and as much as -30 mV/V) that are hard to attribute to sensitivity variations alone. As you say we need to assess other sources of negatives, but there are very few tools that can help verify the source.

RobertEllis2

I'm cross posting this email response in case it is relevant to a wider audience. R.

----------------------------------
Hi Joe,

Thanks for your good questions. The branch cut in the amp/phase representation basically means that all inversion internal calculations must be done in the real imag representation. VOXI only uses the amp/phase representation for input, and it would be easy to just input real imag if that is regularly available for SIP data. Is that the case? That is a question for you and Scott. If so I can modify VOXI to accept real and imag form.

Now forgetting about the data form, I recommend you switch Rx1 and Rx2 to flip any negative measurement voltages to positive voltages. Obviously this does absolutely nothing to the physical content of the data. It does however have numerous advantages in terms of visualization, setting error levels, etc.

For amp-phase data, if you have switched Rx1 and Rx2 to ensure only positive data then the phase is limited to +/- pi/2. Phase measurements in IP are usually in the milli-radian domain so it is unlikely that input data are outside the +/- pi/2 radian domain.

VOXI expects phase measurements to be in apparent phase lag in radians, and typically IP data will have a phase lag. Of course, negative apparent IP affects are possible in certain circumstances (see this Geosoft Note) in which case VOXI would expect the phase lag to be negative, which is fine.

VOXI DCIP will correctly invert positive or negative apparent resistivity or chargeability data (or the equivalents in the amp phase domain)

Having said all this, have you performed an inversion using voltage and chargeability as a proxy for phase? If you have not done so please do – you will cause yourself a lot of unnecessary grief and waste an enormous amount of your time if you omit this step.

Cheers,
Rob
---------------------------------------------

JoeInman

Nick...thanks for you comments and of course Rob too. I am working with Rob and with Scott MacInnes at Zonge to sort this out...and I think we are there.....will try an inversion later today to find out. There are some interesting aspect to RIP. In general let's talk real(v) and imag(v) since that is where we are headed and Rob is going to update VOXI IP to include import of real and imag data....because as Rob points out these are the fields VOXI IP actually uses within the routine to do the calculations....regardless of the data input? And Zonge routinely provides this data...but we/I never bothered to use it.

Now in regards to EM coupling...again we use a standard decoupling routine which doesn't actually model the coupling (Rob will get there with this, one of these days, I am confident), but is effective in removing the coupling...3-pt decoupling. So assuming good quality data...negative imag(v) is not due to coupling removal....but can be if the data is noisy? An interesting point here...and I have had many disagreements with peers in this regard......I would rather collect data in the frequency domain...or at least end up there because the decoupling methods are well-known and have been used for some time. And the coupling is in the signal whether it is frequency or time-domain....so I would rather be fully aware of it and remove it in the frequency domain, rather than sample at a long decay time and hope it isn't there...and find that sometimes it may be.

So, Nick....,do you have any RIP survey data/inversions that you can share with me? Or theoretical modeling results?

thanks, Joe

NickWilliams

Hi Joe, glad progress is being made. I agree real & imag is preferable if it can be accommodated by the available codes; however I find noise estimation more challenging for those measures. Unfortunately I can't share my modelling from a former life. The RIP data I have used was all inherited on JV projects where we also had other surveys with better resolution and without fail the other surveys always gave superior modelled results to the RIP data. I also did extensive modelling of real and synthetic CR data and EM coupling in the p223 Marco code which suggested to me that standard 3pt decoupling was unreliable, most likely because it is an approximation which doesn't account for the actual wire path and actual 3D earth structure, but that was about 7 years and 2 companies ago. Unfortunately given how buggy Marco is, I could never be 100% sure of the results anyway. I will watch progress on this closely as I have many more questions than answers and few people seem to talk about it. Sorry I can't be of more help.