Wednesday, May 23, 2012

Important Factors in Migration Modeling and Charge Analysis

There is a tendency for researchers to focus on mechanisms and algorithms when it comes to migration modeling. You may have heard debates over Darcy vs IP, and finite element vs finite difference, local grid refinement, etc. at modeling conferences, such as the one later this month in Houston. My opinion is that all these do not matter (compared with other factors we often ignore). The theory of fluid flow in porous media have been figured out more than 60 years ago (Darcy, Hubbert). There is not really any debate that oil and gas will migration up dip as long as there is a carrier bed.

The biggest problem is that many carrier beds are often below seismic resolution. A 5 meter sand is typically not recognizable but will easily divert migration laterally. The presence and absence of carrier beds and their extent and connectivity are basically not observable in most cases and migration modeling based on different assumptions of these will yield very different answers.

The other large uncertainties are paleo-geometries. Typically basin modeling tools back strip layers of sediments to "determine" paleo-strucutre. This more often than not produce the wrong paleo-geometry. This is because (1) basins form before sediments are deposited, the sediments fills the lows in the basin (not the way basin models are constructed), and (2) that depositing 1 km of sediment does not cause basin to subside 1 km, it subsides much less because of isostacy (the mantle material is denser than sediment), and (3) the shape of the basin does not change following exact the shape of the newly deposited sediment layer, because the lithosphere has finite rigidity. For example, building a city like Houston does cause subsidence, but perhaps only a few centimeters. Not taking into account these first order effects will certainly give wrong answers to paleo-migration.

In reality, even present day geometries are often incorrect. Seismic interpretation are often very uncertain. Two interpreters may and often make different structure maps based on the same data. The structure maps can only be treated as "models" themselves, not data. Have you run a migration model using geometries based on 2D seismic and 3D seismic data? Are they not very different? I would even argue, that 3D seismic still is not enough to resolve migration paths exactly, a fault with 5 meter throw may still not be visible on seismic but it will change migration direction.

Seals are as important as carriers - they determine lateral vs vertical migration, yet the parameters that determine seal capacity are uncertain by a factor of 2 to several orders of magnitude. I will perhaps talk about seals in my next post.

Given these large uncertainties, our time is better spend trying to improve the geological model by making and requiring better maps, testing scenarios of carrier presence, extent and taking into account the first order large scale geological processes (isostacy, paleo-bathymetry, flexure, etc) rather than worrying about 3rd or 4th order things like flow mechanisms, and mathematical algorithms. We also should treat our modeling software as tools to help us think about the problems rather than giving us answers.


  1. Excellent and provocative post, as usual, Dr Beta Factor. You make several great points and I certainly agree with your basic thesis that improving the geological model, and being able to quickly run alternative models in order to assess sensitivity to geological variation, should be our first priorities.

    One small quibble, though: several of the concerns that you dismiss as unimportant--such as the choice between capillary and Darcy flow, or local grid refinement capability--may directly affect our ability to improve the geology and test scenarios, by affecting the speed and resolution of the models. For example, if you want to put in a hypothetical 5 m carrier bed, you may need to refine the grid considerably, and you may not be able to afford that resolution everywhere. And once you have such fine features in the model, Darcy flow may be a luxury you cannot afford either, if you want to be able to run several alternative scenarios before breakfast.

    A more subtle point: it is not always possible to predict which aspects of the model will turn out to be important, so we should not restrict our investigation to parameters that we “know” are first order. For example, a recent post of yours emphasized the importance of using a full-lithosphere model as opposed to a base-of-sediment heat-flow boundary in order to properly model transient thermal effects. This is something that most of us did not predict 20+ years ago when we were starting to build basin models. We only realized it gradually, after we went ahead and spent the time and effort of adding a full-lithosphere option to the models. Similarly, flexural isostasy and paleo bathymetry have not always been seen as critical considerations in petroleum systems modeling.

    So my advice is that we go ahead and experiment with all sorts of features in our basin models; and let time and experience inform us which ones turn out to be critical.

    This does not detract from your main argument that when we are trying to tackle a practical petroleum systems problem through modeling, we should focus on the inputs and parameters that will have the greatest impact, and be the most useful in giving us insight into the system.

  2. nice posting.. thanks for sharing.

  3. Welcome to SoftwareCorner, your source for
    pecialized software, development and training Mathematica