mindthegap-erc / carbocat-notes Goto Github PK

A fundamentaly important aspect of how CarboCAT is desgiend is that it is an "event based model" meaning that at each grid cell and within each model time step, it records a series of depositional events; every bit of in-situ production and deposition of transported sediment that happens in the model is recorded as a discrete layer. The particular significance of this is that those events produce layers that are directly analagous to the laminae and beds preserved in the rock record that similarly record a series of depositional (and in nature erosional, diagenetic etc) events.

This analogy allows us, in principle, to compare CarboCAT output reasonably directly with rock strata observed and recorded in outcrop; this is actually quite unusual for stratigraphic forward models because most examples of these models do not record discrete events, but instead operate in a more time-averaged mode, either computing only one layer per time step, or if multiple layers are calculated, only storing one layer is is some averaged function (mean, mode, etc) of the many layers.

I would suggest this aspect of CarboCAT is actually quite crtitical to preserve in new versions we develop, because the ability to compare more directly with real rock is very useful, and worth the cost of slower computation and more memory usage. ANd perhaps there are some clver ways to minimise both of those aspects without losing the layer resolution e.g. merging successive layers with the same lithology

Happy to provide more explanation, but hopefully this post successfully defines the event layering as a feature, not a bug ;-)

I think there is a strong argument to implement erosion, at least very initially, as an extension of the existing cellular automata calculations. This has the advantage of being consistant with what is already done in CarboCAT, so no more or less realistic than the existing CA factory/facies distribution calculation, but also, and perhaps most importantly, very simple to implement...

I suggest something like the following:

1. At each time step, for each empty cell on the model grid (i.e. where the facies code is zero after in-situ deposition)
2. if the number of empty neighbours exceeds a minimum threshold e.g 4 cells), erode at a rate proportional to the number of neighbouring empty cells e.g. 4 empty neighbours, min rate erosion, 8 gives max rate erosion.
3. Add the eroded sediment to the volume available to be transported (but routing might get complicated e.g transport back to just-eroded low point, so transport in a specified direction might be required) or if the erosion is disolution (less realistic than scour erosion IMO, but...) the eroded mass can just disappear into the water column

I think this is so simple to do, at least in the simplest possible form, that we should try it, see how it behaves, if bad behaviour is easily fixed, etc, as a first step. At the very least, we will learn something about how a more sophisticated erosion algorithm will likely need to work...

Happy to discuss, or, as Yoda said, "Do. Or do not. There is no try!"