There are six levels of computation difficulty in computing wells:
Summary: The confidence in the results is inversely proportional to the difficulty of the computation. The difficulty of the computation is proportional to the abundance of core data, as the "degrees of freedom" for the computation results is constrained. Core quality can be checked against itself by comparing the chemistry of the XRD or FTIR vs. the XRF elements. They must agree or the confidence in the core XRD is very limited. In lieu of XRF, the XRD and FTIR chemistry can be checked against the nuclear spectroscopy (such as ECS™ ) elements. If there are no nuclear spectroscopy measurements but there are XRF core elements, the XRF can be used to guide the prediction of nuclear spectroscopy elements when combined with offset well nuclear spectroscopy. In particular, it is critical to have XRF over carbonate zones in a mixed carbonate and clastic environment, unless a measured nuclear spectroscopy log is available.
WELL 1 is a level four project and is interesting as it has a lot of core XRD & some XRF on the mineralogy. Some of the core is high quality XRD and some is not. The high quality is the data that has muscovite, analyzed recently by Calgary Rock. The "other" quality has muscovite and illite grouped together, analyzed in 2008 by Terra Tek. The problem the analyst faces is what do we believe is correct? How can we use the two different core data sets to validate our log computation?
The objective of computing the well is:
Before we can determine the bottom line we have to do steps 2 and 3. The notes are below.
This computation process will have about 6 or 7 plots, but only a few will be shown here:
We have camouflaged any company names in order to use this example. Notations to add:
The idea of this summary is to intrigue the reader, whom we assume is a Log Analyst or Petrophysicist, into going through the process of interpretation which involves the first three plots.
Since interpretation is complicated by the value of the input data, in terms of incomplete or low quality data, one cannot usually use a single program for all wells and depositional environments. However, that is exactly what we propose to do: use one program for all wells. We want to use one program that has enough flexibility, to fit all the potential applications of siliclastics and carbonate environments, whether conventional or resource plays. Modifications must be made easily to fit the core or pre-conceived ideas and answer all of the "What if…" scenarios you come up with as you interpret the well. In order to use this one program we have to allow for incomplete logging programs, by predicting missing curves.
For example we want to use the Spontaneous Potential (SP) to aid in the estimation of formation water resistivity, Rw. When the SP is missing, as it is on this well, perhaps because the oil content of the mud is too high, we will predict the SP from an offset well that has a water base mud. We have not experienced any problems in making this prediction. However, the SP needs to be baseline-straightened before proceeding with calculations. It should also be bed thickness corrected, hydrocarbon corrected and converted to a Static SP. There is a company in Denver that specializes in that, called PST INC., (Petrophysical Solutions). We have used them for some important wells and they do a credible job. Most of the time,
we will just straighten the SP, and carry on, until it becomes critical that corrections to the SP must be made.
The interpretation process has to have self-checks built in. That is why, not only core is necessary, but curve reconstruction at the most basic level, is necessary. As this well will illustrate, the reconstructed elements will point out a problem before you may notice it on the core mineral cross-checks with the logs.
The basics must be right before the bottom line can be relied upon. The method of starting with elements allows one to cross-check the calculated minerals by reconstructing the elements. Otherwise, how does one know if the minerals calculated or the core minerals are valid? Perhaps core XRD or other data says they are "close enough for country work" but the element reconstruction provides the final check of, are they really as close as they could be? Other methods of interpretation in other software packages will provide the "bottom line" values but the advantage of the spectroscopy method is we can cross check the results at the basic level:
Fig. 1 "ELEMENTS CORE AND RECON FROM PASS 10 05-04-2013 9-46-14 PM.jpg"
Fig. 2 "TOC AND RHOG PASS 10 05-04-2013 10-04-42 PM.jpg"
The plot shown below is from pass 10, so this shows what porosity, perm and Sw should look like when almost finished.
Fig. 3 "CORE SW PERM POR CHECK PASS 10 05-04-2013 10-34-13 PM.jpg".
Sw core and logs are close. Lower Sw [top, lower perm] zone on core reflects the lower porosity measured by the core (?). High core-perm [10,000 too high] is probably from fractures that were not filled with epoxy before measurement. Ground-up samples measured with pulse decay circumvent this problem. Conventional core porosity does not penetrate small capillaries; hence "total" porosity is not measured, but rather just the larger & medium pores are measured. Ground samples circumvent this problem and are expected to be close to the yellow "GRI" porosity measurement.
Now that we have seen what "almost final" results look like, let's start at the beginning to see how we can get to "finished results".
We will walk through the process, including our challenges in order to give you a road map on the logic in doing the interpretation process.
Start with Raw logs, Plot 1. However, we soon see that the original core data may be off depth, so we must put it on depth before we compare the raw logs to core.
The program that we use for clustering uses imperial units, rather than SI, for rock type identification so we convert to imperial units.
Fig. 4 "PLOT 1 WELL #1 RAW LOGS AND XRF.png"
The next step after looking at raw logs is to do some "Pre-interpretation" processing, to obtain values for missing data such as Rw, free fluid (in this well we have CMRP_3MS, because it is available and because the late John Kovacs told me it was the best to use; you may have other preferences); we also have TCMR, and we can calculate TOC via TCMR, travel time (Passey), density (Passey), uranium (modified "Passey-Bob", plus Nieto's work [valid for Montney but also works elsewhere], and any other method you may prefer to stack up against core-generated values. We used TOC_RHOB for the initial runs and TOC_CMR for the final run, shown in Plot 11. The TOC_CMR seemed to correspond with core more closely than TOC_RHOB, although they give very similar results.
The "Formulas" give you the necessary equations to solve for these values. However, before you can compare to core, the core must be on depth. [Has been depth corrected].
We use the word "Interpretation" and "Computation" interchangeably. However, real "Interpretation" starts when the "Computation" is finished and is at the end of this summary. There are decisions to make in the computation process so the analyst usually refers to these decisions as interpretation.
The raw logs are OK, so we can proceed with the "Pre-Interpretation plot". In the process we have checked KUT and compare to the values in GAPI units of potassium, thorium and uranium vs. the recorded total GR; also have decided on the TOC calculation that best fits core TOC. We initially selected the delta Rhob-resistivity method that we modified from Roy Benteau's "Passey" method. We also use the SP to derive an Rw using the equations in "Formulas". This well used Invermul™ mud so the SP was not recorded but rather, predicted from an offset well. Note the resistivity plot contains only AF30 and AF60 induction logs so the rest of the measurements did not get transferred, since this is a multi-used, released LAS file. The best way to get Rt is to use PST™ to provide their method of correction to obtain Rt. Else, use the service provider. Normally we get them done at PST but in this case there was insufficient data. They need a full induction suite as well as the service company name to do deconvolution. They use the shallowest reading for bed thickness but our clustering method also provides bed boundaries. We don't have the full induction suite of curves.
PLOT 2 PRE-INTERPRETATION: Note, when SLB measures nuclear spectroscopy, they provide QFM, CAR, CLA, RHOG, and PERM at the wellsite. When using old LAS files, this information may be missing, or for permeability, truncated.
Fig. 5 "CWLS WELL 1 PLOT 2 PRE-INTERPRETATION LOGS.png".
The top potassium [and other elements above the yellow "top"] from core does not look correct. The core does not have XRF so the elements were derived from the chemistry of the XRD. [If one has a choice, XRF should always be run with XRD. Else, there is no way to check the core against core measurements.] Note that aluminum from the XRD chemistry is less than Al from the log. When this happens it can mean the core [bulk] clays are not correct. Kaolinite from the core may be slightly low, for example.
We are now ready to do a computation.
After inputting the curves to the "Program" we obtain the results and make the first of the computation plots.
The big plot below will be discussed in smaller plots so that we can explain the nuances of the interpretation a step at a time. One is interested not only in the bottom line of how much hydrocarbon is there but also how you arrived at that conclusion.
Plot 4, Shown Below, is for reference, not readability. We will re-plot in smaller sections for readability.
Fig. 6 "CWLS WELL 1 BOTTOM LINE PLOT 4.png"
Fig. 7 "PLOT 4 FIRST PASS C8 MINERALS, RESIS, SW, OOIP TRACKS.png"
Fig. 8 "PLOT 4 FIRST PASS C8 MINERALS PERM POR RHOG TRACKS.png"
Now, we will explore the rest of the story, permeability and porosity, with the idea that if there are obvious problems we could fix them before running the next pass.
The Timur perm (red) and the SDR perm (blue) have been truncated in the LAS file.
The next track needs some explanation. It is a series of porosity curves that are "overlain" to create an interpretation. We will show how these curves are overlain in example 3, but for now, just look at the colours:
Porosity looks reasonable. The core PHIE is in the expected range for RCA analysis.
Next we will move to the quality control where reconstructed logs are compared to input logs and see what information we can glean before running the next pass.
Fig. 9 PLOT 4 FIRST PASS C8 MINERALS PHID PHIN M_NPHI VS. NPHI TRACKS.png
Fig. 10 "PLOT 4 FIRST PASS C8 MINERALS M_NPHI ILL T_CLAY CHL SMEC TRACKS.png"
We look at the clays, since these are the biggest contributors to the modeled neutron, M_NPHI, and note that:
Fig. 11 "PLOT 4 FIRST PASS C8 QUARTZ PLAG KSP MUSC.png"
Next, we will review the reconstructed elements. Recall that the elements we looked at in the beginning of this story were the "final" elements. What we are looking at next are the "beginning" elements, which will lead us to the constraints required for the next pass.
Fig. 12 "PLOT 4 FIRST PASS C8 RECON ELEMENT.png"
Previously we said, "The main effect of quartz being too high is that muscovite is too low." Then we plotted muscovite and found that it was too low. So, how do we explain that potassium is too high, above the Muskwa zone? The sources of potassium are K_spar, illite and muscovite. Illite is OK. That leaves Kspar being too high? But we have seen that it is not. Makes one scratch their head, eh? Look at illite again. There is no good core above the Muskwa except for one point by Calgary Rock and that one point is OK. So it is possible that illite is too high above the Muskwa? ["Interpretation insight": the trump card is that if carbonate decreases then clay increases and, therefore, muscovite increases, as muscovite = total clay - sum (ill + kao + chl + smec). In retrospect, we should have decreased carbonate by setting carbonate = Ca/0.4 instead of Ca (mixture of 0.4 and 0.22, depending on the PEF to determine the amount of calcite (0.4) and dolomite (0.22))]. When we increase muscovite, potassium will increase but decreasing illite will cause potassium to decrease.
We also see that reconstructed aluminum is a bit too high above the Muskwa. Kaolinite has twice the Al compared to Chlorite (Kao Al = 20; Chl Al = 10; Musc Al = 19, Ill Al = 10). So Kao being too high accounts for Al being too high.
We see that reconstructed silicon is too high and we expected that, since quartz is too high.
There is an anomaly in the calcium track for Calgary Rock. The Calcium computed from the carbonate XRD does not match the Ca from the XRF. This probably is a result of [our] including siderite in the carbonates and [our] incorrectly assigning Ca instead of iron to the siderite. ["Interpretation challenge": that anomaly problem led us down the garden path: we thought that since the reconstruction was OK, the carbonate was OK and we could not believe the core that was showing lower XRF and a higher XRD-Ca, than our measured elements showed.]
Note that for all other elements, except Titanium (no rutile in the back-calculation), the elements back-calculated from the XRD are extremely close to the XRF. In fact, they are so close that one data point almost sits on top of the other. That means Calgary Rock is doing a fantastic job in their XRD. Iron is a little different but that results from no pyrite in the back-calculation. So the variances are a result of our abbreviated mineral model, not Calgary Rock's XRD work. Very Impressive!
We have looked at everything so now it is time to recompute, lowering quartz. We tried quartz at a constraint of 60% of the original but we saw little if any change. In the computation flow, carbonate is computed first. Then clay+muscovite is computed. The quartz is what is left over.
[We should have decreased carbonate too but did not. Consequently, we saw little change as shown in plot 5. If we had decreased carbonate then clay+muscovite would increase and quartz would decrease. Since clay+muscovite would increase, and clay was constrained to remain about the same, muscovite would increase and we would have been closer to solving the problem. We are including our "interpretation challenges and insights" in this analysis to try to help the reader with the logical reasoning for their next well. Hindsight helps!]
Plot 5 is shown as a big plot but the scales are not intended to be read; smaller sections of the plot that are readable are shown later.
Fig. 13 "CWLS WELL 1 PLOT 5 SECOND COMPUTATION).png"
Overall, just looking at the reconstructed elements (gold colour), Plot 5 does not look any better than Plot 4. However, we will go through the plot in a standardized way.
OK, let's look at these questions one at a time.
As you can see there is a lot of detail here. The point is that one does not push a button and out pops the correct result.
We think log interpreters will appreciate the detail. Managers will hopefully be left with the impression that this method is a good way to "milk" as much as possible from the logs to provide the best possible petrophysical interpretation. Furthermore, we need good up-to-date log measurements [nuclear spectroscopy, nuclear magnetic resonance, image logs as well as shear, compressional, density with Pe, neutron and natural gamma] as well as core XRD & XRF in order to correctly interpret these complicated resource plays. When we went through the first interpretation pass, we thought we were going to be finished at the fifth plot. However, the analysis method worked so well that it indicated we were not at the correct result yet.
Eventually, we think the result will be a good balance of all the input elements and core. If one did not have core, the process would be more difficult but could be done by paying attention to the reconstruction processes. Furthermore, the analyst would have to be very experienced in the area they were working in to know what was likely from a core point of view, if they did not have nuclear spectroscopy logs, as they would have to guess whether carbonate was present or not, and whether it was likely cementing or not.
We move on to PLOT 11, which is the result of several runs. Through a series of iterations where we keep checking the results on aluminum, silicon and potassium as well as core muscovite, core plag, core k-spar, we will finally come up with a good balance.
We have learned from this exercise that there is no direct way to go from elements to minerals, as opposed to mineral groups, which Dr. Michael Herron et al figured out how to do. Consequently, core plus elements and neutron reconstruction as well as a comparison of SWB and Sw are very important, to constrain this normalization process. This comment is being [overly?] candid. It not only takes some experience to do a professional job but it also takes malleable software that can get the analyst to the best result.
Plot 11 is for reference, not readability. Plots that have sections of plot 11 will be shown for readability
Fig. 14 "CWLS WELL 1 PLOT 11 PASS 9.png"
This big plot is impossible to read when compressed to fit the page, so we will look at sections of it.
PLOT 11, Final Pass, Notes:
Fig. 15 "QUARTZ PLAG KSP MUSC AND CORE 05-04-2013 5-27-18 PM.jpg"
Below are the clays
Fig. 16 "CLAY KAO CHL SME TOTAL PASS 10 05-04-2013 5-42-01 PM.jpg"
Next we will look at the expanded views of TOC and porosity.
Fig. 17 "CLUSTER PE RHOB POROSITY HCPV 05-04-2013 5-59-43 PM.jpg"
Brittleness can be calculated if the shear and compressional logs are available. They involve the equations for Poisson's Ratio and Young's Modulus. The results are shown in the last plot, Fig. 18, below.
Fig. 18a Brittleness.
The red curve on the second track from the left shows brittleness increasing to the left. Note the correlation with the orange quartz curve. Brittleness is important when a frac job is planned. It is also important to give a target for a horizontal well.
The final summaries are:
Fig. 18b "WELL 1 FINAL PAY SUMMARY 04-04-2013 4-39-49 PM.jpg"
One can see that both Lobe 1 and Lobe 2 are significantly better than the rest, in terms of [net] PAY1.
Fig. 19 "CWLS WELL 1 FINAL STORAGE CAPACITY 04-04-2013 4-39-49 PM.jpg"
Storage capacity is best in lobes 1 & 2.
Fig. 20 "PLOT 11 WELL 1 FINAL YELLOW SUMMARY BOXES.png"
Note that the yellow summary boxes [quick look method] on Plot 11 ~ confirm the above Final Summary storage capacity, Porosity*metres (PorM) and gives 2.53 for Lobe 1 vs. 2.55 for the "Final Summary" [the colourful summaries above]. Lobe 2 for porosity-m is 1.9 which is very close to the above Final Summary of 1.86 for PHIE. Lobe 3 porosity-m is 0.87 vs. 0.88 above; Lobe 4 is 0.94 vs. 0.99 above. It would seem that one could just use the quick look method and get similar results as the detailed method. However, the quick look method required changes to input grain density of siltstone from 2.71 to 2.75 in order to achieve the close similar values. The detailed method provides a procedure to obtain the relevant computation inputs.
Total porosity and PHIE are very close. Changing the cluster's siltstone/sh input from a grain density of 2.68 to 2.7 changes the PHIE-metres from 1.74 to 1.86. Hence, the HCQL computation is very sensitive to the user-input to each rock type of the cluster's probabilistic grain density.
Fig. 21 "CWLS WELL 1 FINAL OOIP 04-04-2013 4-38-01 PM.jpg"
Fig. 22 "CWLS WELL 1 FINAL HCPV 04-04-2013 4-32-58 PM.jpg"
Of course lobes 1 and 2 dominate this pay picture too, as this is a "reservoir conditions" compared to "surface conditions of STP" for the OOIP.
Fig. 23 "CWLS WELL 1 FINAL FRACTIONAL FLOW 04-04-2013 4-27-58 PM.jpg"
The fractional flow is a simplified method depending on relative permeability's, and says all 4 lobes are important, but Lobes 1 & 2 are clearly the best lobes. Lobe 2 has the best permeability and therefore the best deliverability. Hence, Lobe 2 is the best target for the horizontal.
Lobe 2 has OOIP MMBBL/SECT at 22.4 and Lobe 1 has 26.2.
Lobe 2 has effective-porosity-metres storage capacity of 1.86 Por-m for 14 m; Lobe 1 is higher at 2.55 Por-M but is thicker at a gross of 21.3 m vs. 13.9.
Lobe 2's porosity is 14.6% and Sw is 33% and is not limited by SWB to 9%; Lobe 1 porosity is 13.4% and Sw is 44%; so Lobe 2 is better.
Lobe 2 has net 14 M of PAY 1, out of gross 14 m. Lobe 1 has 21 out of 21. [Pay1 = Phie>6%, no water flags].
Lobe 2 has no fractional water flow compared to Lobe 1 with 1%. This difference is reflected by the water saturation difference. None of the top 4 Lobes will produce water, but the Beaverhill Lake may produce water.
Lobe 4 has a tight streak at the base which will help to isolate water from below if it has lateral extension. If it does not have lateral extension, what look like it might be a "barrier" may not be a barrier.