Sediment Bed Layers: Age

[ckharris]

HI

Could someone remind me how the sediment-bed-age is supposed to work? Is it supposed to be the date when the deposition BEGAN in the layer? or the date when deposition ENDED in the layer?

I want my sediment layers to be <~ NEWLAYER_THICK; and needed to change the age dating during depositional cycles in order for this to happen.

Not sure if I fixed a bug, or just modified the code to suit my own needs:

IF ((ero_flux(i,j,ised)-settling_flux(i,j,ised)).lt. &
& 0.0_r8) THEN
!
! If first time step of deposit, then store deposit material in
! temporary array, dep_mass.
!
IF ((time(ng).gt.(bed(i,j,1,iaged)+1.1_r8*dt(ng))).and. &
& (bed(i,j,1,ithck).gt.newlayer_thick(ng))) THEN
dep_mass(i,ised)=settling_flux(i,j,ised)- &
& ero_flux(i,j,ised)
bed(i,j,1,iaged)=time(ng)
END IF
! CKH: Jun 2009: think bed age should be set above, not here
! bed(i,j,1,iaged)=time(ng)
END IF
!

Without the changes shown in pink (just moving the age dating from one spot to the other); I did not create a new sediment bed layer in my simple test case that goes through a deposition / erosion / deposition cycle; and my surface layer got thicker than I want it to be.

I have put those changes on one of my CSTMS SVN branch, but am continuing to fiddle around with the sediment bed layering; using some of JPaul Rinehimer's stuff as well as my own changes. The change described above is the only place that I thought I may have found a bug.

My goal is to not have any layers that are too thin, or too thick (except maybe the bottom-most layer).

[jcwarner]

It has been a while since this was coded, but here is what I think the original plan was.
The bed_age was to keep track of the last time any material was deposited into that layer. So the age is constantly updated in the top layer each time if there is deposition. If you move the bed_age to the location you suggest, then it will change the meaning of bed_age to mean the first time the layer was created. It currently means the last time the layer was touched.
The dual dependency on creating a new layer when time > 1.1 dt is to prevent a layer from forming, then eroded, then forming, then eroded. Which sounds like what you want to do. If this is happening, then the dyanmics are rather rapid and perhaps you should consider to reduce your time step (?).

[ckharris]

HI John
thanks for the speedy response.

for this test case, we have sediment deposit; then sit there for several days, then erode to a new steady state, stay there for several days, and then redeposit and remain at that state for several days.

My cycle of erosion and deposition takes place over many days/weeks, but I never got a newlayer created. I have tried to attach figures here: the top(?) figure shows the bed layers with my change to the age dating, and the bottom(?) one shows how bed layers were not created during two cycles of erosion with the standard age dating.

for my application, I'm happy with having the "age" keep track of the time that deposition started for a layer, if it makes the layering work better for us / me.

I am also curious if anyone (Chris or JPaul, for example) is fiddling around with how the sediment bed layers are done.

thanks!
Courtney

[jcwarner]

what did you set as newlayer_thick?
what did you start the top layer at (how thick)?
I think this all might just boil down to IC and user preference.

[ckharris]

HI

The layers were all initially 0.5 cm thick, and newlayer_thick is also 0.5 cm thick in this testcase.

I did have a problem with the initial conditions in that the age of the layers was too old (initialized with an age > model's time(ng)). But, even after I fixed that (now initialize with age of layers = beginning timestep of model), I still needed to redefine the "age" of the deposited layer to mean the age when deposition began; otherwise I didn't get any new layers.

With the standard setup, the
IF ((time(ng).gt.(bed(i,j,1,iaged)+1.1_r8*dt(ng))).and. &
& (bed(i,j,1,ithck).gt.newlayer_thick(ng))) THEN

was never true, because bed(i,j,1,iaged) was almost exactly equal to time(ng) (because it was updated at each time).

With the standard formulation, for our test case, do you think that we should get some extra layers when the second cycle of deposition starts? (around day 200)

thanks, ckh

[jcwarner]

with the standard formulation,
1)if there was a layer that increased in thickness
then decreased to be less than newlayer thick,
then deposition started in that layer,
then a new layer would not form.

2)if there was a layer that increased in thickness
then decreased to still be greater than newlayer thick,
then deposition started in that layer,
then a new layer would form.

3) It could be changed so that a new layer is always formed. But I chose not to do it this way. There were many reasons, and i do not remember them all. But one idea was that I did not want a new layer forming every other time step as a small amount of grains were raining down. That would cause all the underlying cells to merge into 1 layer and then have a bunch of other fine layers on top.
I also wanted to be able to preserve the constant deposition of a river plume. So as the layer began, it would keep increasing in thickness as long as deposition was still occuring.

[ckharris]

HI
Thanks for spelling that out.

I didn't get the behavior listed in (2), using standard ROMS, though it describes our test case pretty well. Until I changed the dating of the layers to mean the date when a deposit was started, I did not get a new layer during the second cycle of deposition in the test case (day 200 in my plots). For that test case, newlayer_thick and the initial layer thicknesses were both 0.5 cm. I expected, at day 200, that 0.5 cm thick layers would start appearing, but they didn't.

[jcwarner]

can you try the standard setup with setting newlayer_thick = 0.001 m and see if you get the behavior in method 2)?

[ckharris]

HI
I only have a minute to post this: one plot is bed thickness using standard age dating; and 1mm (0.001 m) NEWLAYER_THICKNESS. I do not get new layers at day 200 when the second depositional cycle begins, even though there is a big gap in age there.

The other figure is bed_age relative to the beginning of the model run.

Q: What units are bed_age supposed to be (they seem to be seconds), and is dt also in seconds?

Courtney

[jcwarner]

it is difficult to understand those figures.
Can u send the project folder (is this a simple setup test case??)
-j

[ckharris]

In case anyone reads this thread in the future:

John and I talked on the phone today and discussed the sediment bed model. We both agree that some of my changes may eventually be worked into the main trunk, but that they should be looked at more closely before doing so.

My version now has three differences from the version that was on the trunk when I started this project (what I call "standard ROMS"):

1. When the top layer is increased to be at least bottom(i,j,iactv) thick; if the "leftover" in the underlying layer is < some threshold I add the entire underlying layer to the top layer. The "standard" version would leave a very thin underlying layer as its own layer.

2. I set the age of a layer to mean the age when deposition began on it. I needed to do this in order to get new layers created during depositon.

3. I do not make a new layer UNLESS the top layer is at least as thick as (newlayer_thick + bottom(i,j,iactv)). The standard ROMS makes a new layer when the top layer exceeds newlayer_thick in thickness, but in my application this led to the creation of very thin layers.

My code is updated on the cstms svn site on my branch.

The figure is a plot of sediment bed layers. The top line is the sediment - water interface indicating initial deposition, then erosion around day 150, then steady state (energetic conditions) until day 200, then redeposition. The other lines show the interfaces between sediment layers. Twenty layers were used, with newlayer_thick, and initial thicknesses of 0.005 m.