By Sumon Bhattacharyya
This article first appeared on LinkedIn, July 2nd, 2016
Let me begin my blog quoting a well known Indian poet, Late Harindranath Chattopadhyay on simplicity… This is possibly one of the most philosophically challenging idea ever simplified by a poet… Understanding and explaining any complex problem using simple explanations is notoriously challenging. The reason for this post is to challenge (not the inclusion of recent complexities and development of the modelling algorithms) the notion that more complicated you make a geological model the better it gets. Having spent quite some time building and reviewing 10s of geological models spreading across 32 countries worldwide for the oil and gas sector, I thought it is high time I should document my understanding of the geological models… I would try to answer the important question as to why not to unnecessarily complicate a model when the solution can be found using simpler methods!!! Why not make a simple model and complicate it if required. There would be several experts who would not agree with my views, but remember this is my view – it is for the reader to judge impartially whether my observations are right or wrong.
So let’s start from the beginning, shall we!! why we need geological models?? The most common answer (as I have heard from several geologists) would be to understand geology better so we can be predictive – absolutely not .. the depositional architecture can be well understood using maps and not necessarily creating a complex 3 dimensional geological models. So why are we making a model – the simple answer to this question is “to understand the flow behavior of the reservoir” which should in theory answer questions related to the productivity of the field (e.g. oil rates/water breakthroughs/Water Rates etc.). If earth was homogenous with no barriers and baffles to flow – trust me there would be no-one investing any money to create tools to understand the complexities of the geological processes.
However, having said that the million dollar question is whether these complexities (which has translated into more complicated algorithms over time) are leading to better and more predictable reservoir models or not. Are we just intellectually satisfied that we have more and more algorithms to play with, without realizing the implications to the predictability of the reservoir models. To illustrate, let us start with my favorite – the seismic data– so for the novices the obvious question is “what are we measuring??” In simplistic terms, we are measuring time. So how does it translate to depth?? As some of my friends would say – Tsch Tsch.. geologist eh!!! that explains!!! Don’t you know Velocity equals distance travelled and time taken .. if we know the velocity it is easy to find the distance!! Is it??? Obviously not!! The velocity variation is too complicated and controlled through depositional elements/ compaction etc. that is poorly understood by the seismic.. So there is a great uncertainty in depth conversion… but this is beyond the topic of discussion. Of course we believe for the time being we know exactly where the reservoir top is .. (Shhhhhh.. by the way, in my experience, in most of the cases the seismic is not good enough to give you strong reflector to interpret the reservoir top, but we should really not be worried.. we might have a shallower reflector and we would believe like intelligent human beings that the top of the reservoir is actually parallel to the shallower layer.. whether this is true or not is a completely different question!!!)… I have far too many geophysicists as friends and I chose to live !!!!
Once we have defined the container (reservoir) it is relatively easy to define the cellular outlay (Is it??) of the grid on which we have to populate the properties. So what are properties we need. Any junior graduate could tell you answer to this question – to understand flow, we need effective porosity, permeability, net-to-gross and water saturation within the reservoir interval. The important question to answer is how??? Hmmm.. that’s where things turn really interesting as none of these are measured directly (at the log resolution) even though they are intrinsic property of the rock. Even if you ignore this fact and believe that these parameters can be measured how do you then populate it. The answer is relatively simple – each rock type has got different textural properties that would lead to different flow behavior and all we need to do is to define these rock types and distribute it in the grid … and hey presto here we are with a wonderful rock type distribution. As geologists, we take pride in naming rock types and we spend years understanding and studying these rocks. However, it becomes crucial we understand how do we distribute these rock types in the model. Interesting one!!! Well for simplification let us call these rock types as lithologies.. Now the interesting thing about lithologies is that on it’s own it has no specific spatial characteristics or context. So what guides the spatial characteristics of the lithologies (if you prefer a fancy term you can call it lithofacies)…in simple words it is the depositional environment that controls the distribution of these lithologies. To give you an example, a sandstone deposited in a fluvial environment has different shape than that of Aeolian sandstones. In short, once you associate a lithofacies with a depositional environment you can then use this information to define the shape, size and extent of the facies (Don’t ask me how accurately – as my learned friends would point out it is indeed within the realms of uncertainty). These are generally referred to as architectural elements. Now this is great!!! Right!!! We can define the architectural elements and we can distribute these in the model and of course within the limits of uncertainty we can be predictive!! Yeah!!!! Unfortunately, this is not as easy as it sounds.. All the depositional architectural elements follow a certain hierarchy and shape which sometimes is really difficult to understand/quantify and model mathematically. In my opinion we haven’t yet uncovered all the mathematics behind geology but the good news is that we are trying.. (If someone told you geology excludes mathematics please point them to me)… However, even knowing this can we actually be predictive??? Think about compaction, diagenesis, complex depositional layering etc. that can effect the characteristics (and connectivity) of these sediments significantly!!! Do we know the exact mathematical parameterization required to define these shape and the changes that happen afterwards??? The important question here is that if we don’t understand all the complexities of the geological processes what are we trying to model??? Even with all the complexities in algorithms.. isn’t the assumption still a simple one!!! By the way, we haven’t yet managed to build a model as complex as the underlying geology!!!
Oops this is getting longer than I anticipated!!! Wait for my subsequent post to unravel the secret of the geological models!! i.e. if you are interested…Watch this space!!!