- [Voiceover] So fit for purpose models, so your boss has come to the office and he says, "This time I don't want you to build a complicated model. "Last time you spent a lot of time doing that. "This time go ahead and build a very simple model." So you say, "Okay," and then you go back and you're thinking, "What do I need to do?" We're going to talk about that. We're going to talk about how you frame your model, how you handle uncertainties. We're also gonna talk about how you collaborate with other disciplines. Collaboration is key. It's huge in building a model. Model design and how you customize your workflow for different environments. That's also something that we're gonna talk about. So the case of the simple model that you were asked to build, what makes it simple? Is it the process? Is it because it's gonna focus on just one goal? Or could it be because of the tool that you're required to use or that you want or you're gonna use? Are you gonna use a simple layer cake modeler or software capable of 3-D simulation? Any one of these reasons, including time, can be the reason it can be defined simple, but the big picture here is that the model should be simple enough to still convey the geology in a practical manner. If it is too simple, it may be insufficient; and if it too complicated, it may be overdoing what's necessary. So simple models are good. The goal here is always try to preserve the geology. How do you frame your model? So, now you have decided that you're gonna build a model, and you want to frame the model. Part of framing the model would also help you decide if you should still, if you should go ahead and make it a full, 3-D full field geological model, or if you should go on and make a simple layer cake model, because sometimes there may be problems that are complicated. For instance, you may have a field that has no aquifer, and you expect that you're gonna have compaction-drive as the main drive mechanism. So somewhere along the line, compaction is gonna have an impact on some on-field properties, such as permeability. It's gonna have an impact on your ability to maximize the recovery in that reservoir, so what do you do? Do you use a tool that can do that explicitly, or if you're using a simple tool are you gonna figure out a way to replicate that? Those are the kinds of things that come to mind. You know, but when you're framing your model the first step is for you to look at all of the data that you have. Look at your data; do a gap analysis. What does your data tell you, and what does it not tell you? What is it unable to tell you? Those are the things that could help you understand what the key uncertainty in your reservoir. You maybe use a model because you already have established a goal. And an example of a goal would be, for instance, in early development you're about to develop a field, so you go ahead and do a 3-D full field. So, you want to do an analysis of all the uncertainties. That way you can say you touched all the bases. You fully covered the full extent of the reservoir. That's one reason. Most of the time there may be a specific technical or commercial reason why we're doing this. Whichever way that you rank the uncertainties, by impact and probability, and you identify the uncertainties that have an impact on the objective, and together you can build a model that answers that question. That is how in development geology you are able to go right ahead and build a model that at the end of the day provides answers that matter. So let's see some more how you can go ahead and do this. First off, I have a list of some common model objectives that are typically encountered. One of the most common volumetrics, you want to define models of that. It is always encountered in many cases, and so some of the key issues you could also associate from your static uncertainties would be things like compartmentalization. If you have a reservoir that has a lot of form. It's important for your volumetrics that you define it in volumes. If these volumes are on a compartment by compartment basis, or if they're all connected. That's a huge issue. It has an impact on how to go ahead and develop that reservoir. So, we talked about how you identify your uncertainties and errors And then you have to rank your uncertainties, because, most fields, almost everything we do in the sub-surface has an uncertainty associated with it. If you want to account for every single uncertainty, the model may turn out to be too exhaustive, so we'll go ahead here and use a checklist. So I put a sample checklist here with a common example, an issue that has to do with impedance analysis. So you have a fluid contact example here. So you can rank these uncertainties, rank the uncertainty on the basis of the likelihood that it's going to happen or not. If it's going to happen, so, yes, in this case, you dig some wells, you don't know the fluid contact. So, that's a situation you have to deal with. The good thing is you do have an upper and a lower bound. Because you have an upper and a lower bound, it's not as bad as it sounds, so it's in the yellow. So, let's see how the impact, what impact is going to give you. Well, if you are unable to fully evaluate your STOIIP, because the size of the uncertainty is large, if you look at the difference here between the oil down to and the water up to, the difference is quite significant. You're talking about 38 feet. Because it is quite significant, the impact between the lowest volume and the highest volume is quite high, so you have a situation, where our uncertainty is medium, and the impact could be high, and so consequently falls in the red. So what does that tell us? Well, if it had fallen in green, we'd have said we could live with that, because uncertainty is low and impact is low. If it fell in the yellow, we would make a note of it, because it could be something that may have an impact much later in the process, but if you fall in the red, we know we have to do one of the key things we have to model. So that's how, sometimes you can identify what the objective should be, if you don't have a clear objective coming into the modeling part. Once you identify your key objective, you go ahead and build that uncertainty scheme. It could be Monte Carlo based, which is very stochastic, or it could be an experimental design where we use scenarios to capture the range of uncertainties. All through this process, you have to be mindful of collaboration. Collaboration is key. It's important. All the disciplines should sit at the table, because they come back either to use the model, or they use the model or they provide direct input into the model, or can provide suggestions that may be useful. They should sit at the table. The disciplines there, these are disciplines that get to use the results from the model, so they should be able to participate in the frame, and the part where they come in for the results, they should be shown that this is the product from the modeling process. The collaboration is important to them. I have an example here of how you handle uncertainties here. This is After Clayton, a mix between scenario-based uncertainty, and geostatistical uncertainty. So what it actually does here, it is very different from when you have the single deterministic case, and you go ahead and just have one month, and that's it. You move on. Or, the stochastic part, where you define them with a range. Sometimes you have distributions, and you have a normal distribution, and you say at P = 10 and p = 90, you define all of your properties based on distribution, and you use Monte-Carlo, that would give you a stochastic outcome. A scenario-based outcome would define the scenarios that are plausible. It is very geological, because geology happens in scenarios. We sometimes say, these are the characteristics of a braided river, these are the characteristics of rivers that are not braided, or meandering rivers. We don't typically say that these are the characteristics of rivers, and they grade from meandering to braided. We tend to think in scenarios, and I like the information undergo true scenarios, so you may have scenarios and the beauty of this, after you define scenarios, you can then convert your scenarios, into geostatistical, stochastic input, or stochastic models. You end up having a whole lot of scenarios, and you're able to satisfy the jury that requires that you define these scenarios clearly, and all of the other statisticians, who want you to do some statistic work, you'll be able to actually vote by doing a mix of these two methods. This slide just talks about some of the mechanics of designing the model. These are little things that we have talked about before. We have talked about how you have to be able to have the model built with surfaces and faults together, and zones and all that. So these are all things you do. There are a couple of decision, also, you need to take. The choice of modeling method is a decision. The level of complexity of your facies model is a decision. There are several decisions that you get to take, and these all factor into the mechanics of your model. After that, you customize the workflow. There are two main stages that have an impact on customizing workflow, way more than any other stages. Here, let's come back to this slide of mine, and look at stage one and stage four. Those two stages are the big stages where you get most of the customization. Your input data changes. If you're working in unconventional reservoirs versus working in deep water, the type of input data you get changes. The type for a physical model, they vary. If you look here, you see the case for the kinds of things you would do when you're dealing with unconventional setting. You worry about these things, but when you're dealing with, in a conventional setting, you just do your SPROF, your permeability, and weather titration, and you're good to go.