| Message/Author |
|
|
|
|
| Notes on version 5.5 beta 6/3/07 |
|
|
|
|
New tab for no lateral erosion (default) to spare having to go through all of one of the two lateral erosion routines with the erosion rate set to 0. There was a need to do this in order to zero some values each iteration, but this has now been removed by the new tab (the default).
Vegetation.
The growth model was allowing veg to grow up through deposited sediment very rapidly (in a day) and not to decay if deleted. Now it lets veg grow up through deposited sediment, and also decay if it remains buried for a while.
TOm |
|
|
|
|
5.5h.cs
Now working on some new ideas partly inspired by the Reduced complexity modelling workshop, recently held.
Firstly there was disccusion that there appeared to be ideal scales at which cellular models operate. In that they can struggle to represent sub bar features etc.. One suggestion as to why is that there are certain length scales - or certain scales in flow structures that shape their size, that scale to vairables such as channel width. Therefore, if we have a model with many cells representing a channel width, with one central bar. Are the processes that are shaping that bar going to be accurately represented by the cumulative of the forces modelled by the individual cells?
For example, with a channel 10 cells wide, might it be reasonable to smooth (for example) shear stress, across a 5-10 cell radius in order to get rid of local fluctuations, and bring in a more 'regional' signal caused by the more generalised flow patterns?
So at present trying smoothing Tau before lateral erosion (seems to make things smoother) by up to 15-20 passes. But not for regular erosion/deposion
Also trying this with smoothing over the bed (1-3 passes)
May also try smoothing only allowing tau to come from contributing cells. |
|
|
|
|
5.5i/j
Working on problem of getting 'knobly' meanders. Is this due to:
1. a concentraion of shear stress on one point?
2. Not restricting erosion in too tight bends or
3. The width calculation not providing an 'average' of width in that section. Local widths leading to excessive local lat erosion rates?
Tom |
|
|
|
|
5.5 j
Found I was getting good smooth meander shapes by smoothing the edge gradient 100 times, also smoothing shear stress (before lateral2 was called) 100 times. THis worked well with the lampetermeanders example - but when on the lowmeander test it resulted in evenly developing but triangular shaped meanders.
I think this is due to 2 edge cells being eroded from diagonal running channel edges, and only one from l-r or u-d channels. Helped by only having lateral erosion from Manhattan neighbours.
May need to re-think how lateral erosion is removed and added from edge cells.
Tom |
|
|
|
|
Notes from expts with 5.5k.cs
using all 8 for edge cells, version running now has -2 and +2 cutoffs for edge values
Still a bit pointy, so now ditched the raidus of curvature calculation part, and just using the edge[x,y] value to calc it. Seems to work better on lowmeander example, but not especially with the lampetermeander example.
Now not using all 8 edge cells did not seem to work any better - if anyting too many edge cells. So back to manhatten neighbours when deterimining edge cells.
Also tried removing Tau smoothing, but seemed to make things notchier again.. now trying it with Tau smoothing over 10 radius, I was concerned that by smoothing Tau over such a large distance it was removing the capapbility of the model to incorporate changes in Tau that occured when there was a cutoff (for example) into how the model then behaved after that.
Now using the lowmeander sample. Changed grid cell size to 10m from 5m, as I wondered if the gradients were a wee bit too steep for getting a nice even meanders to develop. At the moment seems to be running much better. Some sensitivity to the input of sediment from above. Very easy to feed in too much sediment,
Also noted behaviour whereby if too much material is added from the banks then it semi blocks up the channel which causes a glut of water behind - time step drops and eventually it gets eroded away. But can lead to some strange looking effects. Therefore best to run with low lateral erosion factors. |
|
|
|
|
Possible way to calculate edge cells?
Could calculate mean flow depth across a X cell radius then use this to calculate the threshold above which you calculate the edge of the channel.??
Also having success running 100 lat averaging, 10 tau smoothing on lowmeander example, but much better if I moved the grid to 10m resuloution - I wonder if the slope was too great on the 5m example. eg slopes too grerat with |
|
|
|
|
OK, one problem with smoothing Tau before the lateral caclulation is where you have a less defined channel edge. I think this is happening in the Caersws meander simulations. E.g. if the flow is just over bank, then the overbank cells are now the edge cells, and these then have a disproportionately high Tau, as it is smoothed from the main channel cells where there is a high tau. Will try on the lowmeander sample to see if removing the tau smoothing results in notchy bends again!
Also wondering if the above probkem in the Caersws meander section is caused by having too fine sediment in the banks - so its very easily eroded by the bed once the vegetation layer has been breached.
No, still doing it without the Tau smoothing and with constant grain sizes. I wonder if it is because the sediment sizes are too large - so it cannot re-organise the stream bed after a flood.. Trying with lower grain sizes.
Tau smoothing seemed to have some effect on the angularity of bends (lowmeander sample) but not too much... Seems to work well with smoothing iteration of 10.
Having mixed grainsizes also seemed to help - best results so fart with some coarse and some fine. |
|
|
|
|
7/5/07
Interesting on Caersws meanders example, with no recirculation, it creates very wide channels (at the upstream end) where it is recieving no sediment. Downstream it works fine - so I suspect what is happening is that without sediment to infill the inside of meanders - because the lateral erosion scheme has no critical shear stress, it keeps on eroding laterally and widening the channel, despite there not being sufficient Tau to mobilise the bed material to fill in after it - or cut a new smaller channel
Also found that there was an apparent error in the routing shceme - well not an error more that things were not quite correct (probably). Was l-r r-l u-d du. then l-r r-l d-u u-d. Second part should have r-l to start with, as this gives two l-r in a row (in the l-r r-l domain). This may make things better - certainly seems to give smoother meanders and get rid of some errors that I previously thought were to do with the smoothin of the edges |
|
|
|
|
Need to check whether or not differences in lateral erosion are (or are not) caused by having the 'width' term included - as I previously thought. Should check whether (or not) the width has to be included or not?
Last good working version 5.5m - found fix for tracer inputs too.
Problem was the d50 calculating alorithm didnt work correctly when there were tracer inputs too. Now fixed by only calcing d50 from main grainsize. |
|
|
| Back to top |
