Presentation on theme: "Going for the RED is an approach used by But for Red to indicate hydrocarbons (like this example) lots of things have to be true. The project I use for."— Presentation transcript:
Going for the RED is an approach used by But for Red to indicate hydrocarbons (like this example) lots of things have to be true. The project I use for my “direct detection” example lies off south Louisiana. It shows several “emergence / subsidence cycles, with fairly obvious overlapped angular disconformities. Current automatic mapping logic can’t begin to handle the complexity, and visual interpretation is sorely needed. Later we will talk about noise, But first let’s consider what is needed to make these “hot spot” amplitudes meaningful. older interpreters who choose to ignore the attribute zealots. My argument starts with inversion. I say the industry misses the excitement of direct reservoir spotting because they stop too soon. Simulating sonic logs to enable accurate stratigraphic tracking should be the goal. More on next slide. Inside, I present “before and after” pairs that are set up for “toggling”. If you have never used this tool, it is worth a visit to use it. Toggling helps one visualize why integration flips the polarities to achieve a simulation of lithology. Multiple well log matches prove it is doing the right thing.
Sonic log from an early shale play, imposed on an ADAPS non-linear inversion (left) and on the integrated final result (right). On the log, low velocities peak to the left (red on the the display). Bed thickness is crucial to stratigraphic correlation, and correlation is the exploration geologist’s most essential tool. In reservoir work, tracking thickness is all-important. Matching sonic logs must be the ultimate goal. All inversions contain “spike guesses” from all interfaces. Unless one is able to resolve this mess in one’s mind, any inversion by itself does not provide attributes that are trustworthy. Any single interface value is dependent on both of the bounding bed velocities. In addition, a “red” from the top interface means one thing while one from the bottom means the opposite. Thus the need for integration. Of course I don’t think others can match the ADAPS non- linear inversion, but that is something else. Examination will show there is little “polarity” correlation between inversions (even mine) and the well log. One must be careful with polarities. Many observers just line up events without regard to color. Note point A where polarities are opposed (some would think this a match). While the integrated match is not perfect, it’s pretty good. Note how well the thickness agrees. As we continue with the older part of the show, be prepared to toggle. All versions of the same original data have been carefully aligned, so two fingers on two arrows is all you need to make an intelligent comparison. Spend some time on this learning experience. A
1 Direct reservoir detection - Where words fail, the well match here says it all. It proves we could have predicted the presence of this particular reservoir before the act of drilling. The project has several wells, and the matches are all quite good. No well data was used as input to our processes, but it is nice for it to illustrate my points. The stars identify what I consider a basal sand, sitting on one of several unconformities (angular disconformities that is). The ability of the system to simulate lithology was crucial in my own interpretation.
Here is well #3. It appears to be tapping below our basal sand (again identified by stars). It is interesting to note that it (the basal sand) has petered out here. This is a problem we should expect in “on-lapping” deposition. Again this is a good match, but not perfect.
AVO probably worked here, but if it did, it was pure accident. The explosion on the gather to the left, would have triggered the AVO logic. It was caused by pre-stack migration that could not handle refractions. How this causes outlandish amplitudes is beyond me, but it is not at all unusual. The refraction(s) are seen developing in the series of gathers shown at the bottom left. Over-corrections make it easy to spot them. Note that the system handles them until a critical point, where the energy explosion occurs. The good results were obtained by using a very deep mute, throwing away almost half of the traces. However that does not mean there is not still a noise problem caused by the “critical angle phenomenon. At that angle, the downward energy is interrupted, and non-reflection events pour in to fill the gap. These mostly consist of “point source” refractions from deeper beds. In my later work I have successfully lifted off this type of noise, and the odds are I could do a better job on this data. Unfortunately that opportunity is gone. Introducing toggling as an educational tool. To really grasp the exploration importance of sonic log simulation one needs to continually compare the before and after display sets. Two fingers placed on the left and right arrow keys are all that is needed, but practice helps. Three sets follow. Please make the effort. I assure you it will be worth your time.
A Unadulterated stack This is the raw stack on comparison #1. Please toggle with my optimized stack.
A Optimized stack input Toggle back w raw Or with the final. Going back and forth here is where the time should be spent, as the integration logic rearranges computed interfaces to simulate lithology. Lobes will disappear as they get combined with their mates.
Back to input (the optimized stack) A Final result Final ADAPS result If this is your first time into the concept, your reaction might be one of disbelief when you see the shifting lobes. All I need to say here is that the well matches are not lying, and there is no way I could make them up. The fact that normal seismic sections don’t represent lithology is worth knowing.
YYYY straight stack Toggle with the optimized stack. But first notice where the arrow is pointing. Straight stack
The optimized stack You should notice glimpses of the basal sand to the right of the arrow. Of course this becomes the input to the inversion/integration logic. Toggle back to the simple. Or toggle with the final Optimized stack
Amazing if true – (and we sure think it is). What we see repeated in this toggle trio is a serious case of tuning and subsequent de- tuning. The basal sand we see here (which checks out on other lines) hardly shows up before NLI. Final inverted and integrated
toggle with the optimized stack. Straight stack Another example.
The optimized stack Again notice that the basal sand is visible now, before inversion. In particular, really look at the right-hand arrow. Toggle back to the simple. Or with the answer Optimized stack
Final inverted and integrated Toggle with the optimized stack.
In-line XX6 and cross-line 2900 I close the toggling session with an intersect to show the merit of visualizing complex results. Of note here is the “bright spot” verification we get when the hydrocarbon indication fades down dip.. the fact that these two independent runs tie so well is further system proof. Associated topics - Strike slip faulting is obvious to me on all these sections, but my goal here was just to show the need to invert and integrate. This topic is covered extensively in the later show introduced below. Major coherent noise discovery. R ecent noise prediction & removal work I have done (both on gulf coast data and on coal seam work in Scotland), has convinced me that strong noise events exist even when the all seems well on the stack. Because of what we saw on the gathers earlier I am certain serious noise still exists here, intermixing with the real data. While the well matches are a lot better than has been seen before they are not perfect, and I believe coherent noise is making its way through. I contend noise removal is the next seismic frontier. I introduce my gulf coast work on the next slide as a case where de-noising uncovered previously hidden continuity that is strong and believable.
A deep well to the south of this Louisiana salt dome project indicates that the prolific shale play strata I show on this well log match extends this far south. (You might notice that final integrated match,) The pinch-out pointed to by the black arrow does not appear on the input. There’s no way my system could have invented that. On the next slide we start back at the shot point format on our explanation.. Making a long story as short as I can, this de- noising goes back to the 3D shot point, iteratively detects refraction events then gently lifts them off the underlying signal. It then produces the treated gathers, applies my own NMO and stacks. It was not told this was a salt dome, but it took it upon itself to sharply delineate the boundaries. This is what you see at the right. The green delimiter is mine. In general, the logic eliminated the noise you will later see crossing this line, and brought in completely new continuity outside the dome. You will see an example on the next slide. Then, on the final I show the same kind of noise removal on Vibroseis data from coal seam work in Scotland. A revolutionary breakthrough? I think so. From the original show
1. 2. 3.3. 4. Here is another before and after – To begin, the green delineation line is in the same place on both. I just cut off the rest of the salt to save space. So let me just cover several points of interest. 1. Unfortunately the strong apparent dips at the edge of the dome were the geologists target. The system threw them out to his intense displeasure. 2. Careful study shows the dips and the events are not the same. This introduces the problem of close overlap of noise and signal, probably the primary challenge to the logic. 3. Point 2 applies here in spades (meaning extremely). Here we see the noise continuing into no-mans land. 4. Both points strongly apply here.To the left of the demarcation, where there is virtually nothing showing on the input, our results are strong. To the right we see nothing, meaning the input was pure noise. Now, reasonable people should see we are dealing with a major discovery here – one that opens new exploration possibilities. (at least that is how it seems to me.) Click on oval to access the original show – (or put adaps.com/PP1.ppsx in your browser) Now on to the Scottish Vibroseis example.
De-noising Vibroseis The contention is that the correlated Vibroseis field record (to the far left) is a mess of overlapping coherent noise and signal, the noise being strong enough to control the stack. The operative target is below the red line. The goal is delineation of a series of coal beds. On first glance, one would think there is a serious statics problem, but once the non-linear process lifts off most of the noise (to the right), we see that statics are not involved. No event shifting was used so the resolution of this jaggedness actually proves the overlapped thesis. As you will see when you continue through the 14 adjacent points, the logic was consistent on its very selective event selection. Note here how it took out the leading lobes just below the red line and then emphasized the ones just below. Once again, the initial jaggedness was the result of the intricate overlay of noise and signal, and eliminating that phenomenon is the essential proof of the logic. The fact that what comes out exhibits the proper event shape is the final verification, since we are not smart enough to force that result. This is a timed show, so relax and pay attention to the detail. Clicking will speed it of course. Click for the Vibroseis show (or put adaps.com/kinc.ppsx in your browser).