| Message/Author |
|
|
 Joe
posted on Thursday, September 15, 2005 - 09:28 am
|
|
|
These are just some old email questions and replies (from Tom) I had that might be helpful for others getting started.
Hi Tom-
I've got a handful of practical questions about CAESAR for you, that I couldn't remember from our earlier conversations and couldn't find in your manual.
1. I don't understand what exactly what is stored in the grain.dat files and how they are organized. The first three columns presumably are some sort of cell reference location and some other integer? The next 100 columns presumably relate to grain size percentages in some way? My understanding was that you stored something like 10 or 12 different layers of sediment at every grid location and then described the grain size distributions in each layer based on percentages of classes defined in the input parameter box. Could you explain this? Also, is there any good reason why there are 16 significant figures? Surely, if these are indeed just percentages or even grain sizes you could afford to collapse the number of significant figures down to 4 or 5 and slash the file size dramatically. Right now, on a 353 X 273 grid your grain.dat files are roughly 30 x larger than the elev. dat files.
Yup,all are on lines, the first three are an index number and the x,y co-ordinates. Then follow 11 numbers for the surface grainsize proportions at that co-ordinate (remember there is a veg fraction - #1 and bedrock #11). Then are 9 numbers for the proportions for the next layer down, then anoterh 9 and another etc.. for each of the 10 subsurface layers. Lots of data I'm afraid...
2. During a model run, in the upper right hand corner a real time sedigraph and hydrograph are reported. Are these values stored anywhere to reconstruct hydrographs and sedigraphs later? If so, at what time step are they stored (as the actual time step would be a hideous amount of data). Can you pull sedigraphs and hydrographs off other points in the basin other than the outlet?
Yes, stored in teh catchment.dat file, in the time step as specified. This file updates - or increments, so it will just append to the end of itself if you stop then re-start, so remember to delete it between runs if you want fresh data. I may as well answer below too here. I
1 20.144 0.0 0.0 17.314239 4.712828 0.970313 0.711106 0.284286 6.358521 2.650343 0.951041 0.49928 0.176522
2 20.075 0.0 0.0 7.92809 1.55892 0.357528 0.224936 0.265442 3.144464 1.425625 0.549803 0.296532 0.10484
3 19.907 0.0 0.0 6.616725 0.969692 0.370578 0.131047 0.200159 2.602081 1.40452 0.543313 0.292072 0.103263
4 19.964 0.0 0.0 8.557235 1.689749 0.539715 0.334547 0 2.747729 1.771237 0.819247 0.483919 0.171091
First col is time, second water discharge (in m3 sec) then two blank cols, then total sediment yeild (m3) in that period (depends what youset the output time step for) then the total for all 9 fractions...
3. I don't understand what is stored in catchment.dat. It looks like it might be structured to store the outputs from the sedigraph and hydrograph? However, for the run I'm looking at now, the values are all zeros. Could this be because the run ended abnormally?
4. Why do the output elevdiff.dat files preserve all the ARC header information except for the lower left and upper right coordinates? The elev.dat file seems to preserve this info?
Erm, they should do..
This is because header information was not read in that version of code... not sure if newer version does?
5. What sort of time steps do you save your output files at if you want to visualize the changes during an indivdual flood event? Do you just use a small output time step for the flood, stop the model and then start it again at a longer time step?
Tricky - possibly the best way is as above.. hard/ a pain to code up more detailed output times..
6. Does the model just loop through the rain.dat file again from the start if it has reached the end of it and is still going? If so, is it only stopped if it reaches an error (e.g. full hard disk) or by the maximum number of itterations?
Nope, it should just run with no water - no flood when it runs out of rain, until it gets to the end of the iterations or time elapsed counter.
7. Are itterations, as you use the term, simply a single time step (which is variable), or do they represent the passes from L-R, R-L, T-B, B-T, etc?
Yup, does the lot during one iteration..
Cheers
Joe
Hope this helps!
Tom |
|
|
| Tom posted on Thursday, November 10, 2005 - 03:28 pm
|
|
|
Thanks Joe - thats really useful.
I realised that the sample data I put on the website was not great, so I have put some new examples, complete with config files up. This shows how CAESAR operates in reach and catchment mode,
Tom |
|
|
| Tom posted on Tuesday, November 29, 2005 - 05:34 pm
|
|
|
> 1) "m" - does this control 'vegetation cover'?
No, not directly. It controls the hydrological response to rainfall input. Low 'm' values (0.005) result in a flashy hydrograph - high flood peaks with a short duration, whereas higher 'm' values (e.g.0.02) fan result in lower flood peaks and longer durations. Changing m mimicks the effect that vegetation change has on the hydrology. Moving from a forested environment, where the trees and vegetation soak up the rainfall - to a deforested situation where floods are rapidly transmitted from slopes to channel.
> 2) If it does control vegetation cover, how linear/non-linear
> is it? ie. is it worth doing a run every decimal place
> between 0.05 and 0.2 or would this be a waste of valuable
> computer time (ie. if it is a threshold, i would just run
> certain values)
Its not linear. There is quite a bit of literature on the 'm' value - relating to its use in TOPMODEL. I think there is a review paper by Keith Beven from 1999 (I think its in an ESP&L TOPMODEL special edition) where he lists studies using TOPMODEL and what the corresponding m values are.
> 3)Where did these 0.05 and 0.2 values come from and why are they so?
A mixture of calibration and previous studies (see above paper).
Best method is calibration, where you know the rainfall and the stage/discharge, so you can run several model runs, with different M values and choose the one that matches the actual catchment response the best.
> 4)Why is the slope angle set to 80 for Annecy - is it because
> of the steepness of the slopes and how real is this value for
> the catchment?
Yes - because of the steepness..
I don’t know how real it is! Probably too steep. Though you can have a more realisitic angle (45deg) and set up a bedrock DEM that will not landslide - where there are cliff faces etc..
> 5) I remember you telling me but I can't remember....what
> does the " grass now!"
> button mean/do..apart from growing grass..what is it's purpose?
It allows grass to grow instantly where there is no water - or to be removed instantly where there is water. This is so the model can start with a channwl with no grass, then you can add grass later - to the areas around the channel.
> 6) does the 'grass now' mean that you can change spatial
> distribution of vegetation cover in the catchment or is this
> just wishful thinking on my part?
Wishful thinking I'm afraid.
> 7)finally... For the Yorkshire studies into the Holocene, how
> did you create a rainfall data set that extended this far
> back? Our idea this morning was to take the last 5-10 years
> worth of monthly data, look at the daily distribution and
> take that as the standard distribution and match all past
> monthly records to these records...is there another way?
Rainfall reconstruction is really really tricky. Marco had a go at it for the severn and ran into some problems - if I remember correctly.
For the Swale/Starbotton studies, we took a 10 year hourly data set, and multiplied it by a factor that related to a cliamte proxy - a peat bog wetness index. You can look at the distributions and try to re-construct the rainfall sequence, but its hard - and I don’t know how you could do it. If you know someone who is fairly capable with stats/statisical analysis it should be possible in some shape or form. |
|
|
| Katherine posted on Monday, December 12, 2005 - 02:45 pm
|
|
|
Thanks Tom. I could write a book with the amount of questions I have..
1) when restarting the model from 200days for example,does the rainfall record restart at zero or at 200days
2) How long did it take you to do the 9000 year yorkshire runs? These computers are doing roughly 100model days per real-time week, so we've calculated that to be a lot longer than the 3 year phd! - are there any short cuts or anything you can suggest to make it go any faster (have 3 runs going and none have reached a year yet!)
thank you |
|
|
| Tom posted on Monday, December 12, 2005 - 02:46 pm
|
|
|
Hello Katherine,
> Thanks Tom. I could write a book with the amount of questions I
> have..
!!
> 1) when restarting the model from 200days for example,does the
> rainfall record restart at zero or at 200days
It 'should' restart at 200 days, though I have a suspicion that it does not. It might be worth checking this and seeing what difference it makes.
> 2) How long did it take you to do the 9000 year yorkshire runs? These
> computers are doing roughly 100model days per real-time week, so we've
> calculated that to be a lot longer than the 3 year phd! - are there
> any short cuts or anything you can suggest to make it go any faster
> (have 3 runs going and none have reached a year yet!)
Took about a month. On a Celeron 500 machine a few years back..
OK. This is an important question and there are many ways in which it can be sped up - not all are correct. Basically, the time step is controlled by how much erosion and deposition there is going on within the catchment - for example if there is more than 0.02m of erosion at one point, the model stops that set of calcualtions, halfs the time step, goes back and starts the calcualtion again. Therefore, the timestep can be controlled by just one cell.. You can get a situation where the model is trying to cut away at one little ravine - and this just keeps the time step low. A way to check this is to look at the erosion and deposition file (in Arc View) and see if there are any cells where there has been a lot of fluvial erosion (take care these are not landslide cells) - e.g more than about 2m or so.
If this is the case (as may well be in Annecy) then what you need is a bedrock file. This is the same format as the DEM, but contains the elevations of the bedrock. Below which the model will not erode. As a rough starting guide, what you can do is take the DEM import it into ARCGIS, subtract 2m from it using the raster calcualtor, then save the file, and export it to an ascii format (using rastertoascii). This then prevents the model incising too far. Though of course in some places the bedrock will be at the surface etc..
Grain size and armouring.
Because the model starts with uniform bed material over the whole catchment, there is a period at the beginning of simulations, where it erodes the finer material leaving the coarser material behind. This can take a while, and I refer to this as 'spin up' time. Where the fines are winnowed away. This may take 10-100 years of model operation to get through. One way to check is to plot the sediment coming out of the model and when the angle of the curve declines you should be OK.
Once you have 'spun up' the model, you can then use the elev.txt and (importantly) the grain.txt files from the spin up run, as the starting conditions. It’s a bit like wearing in the model if you like.
Another alternative is to try using larger grainsizes - or checking what graisizes are being used in the model - as if they are too fine (too many sands etc..) then it will just erode away for ever.
Minimum time step
This is a quick and dirty method - that may give some spurious results. However, there is one box in the numerical tab (I think) called min time step. If you set this to say 1 sec, then it will 'force' the model through large floods, when it is running at very very small time steps.
The problem with the above is that it can influence the results. You have to be careful of that.
Hope this helps,
Tom |
|
|
| Frank Engel
posted on Tuesday, January 03, 2006 - 10:38 pm
|
|
|
Hello Tom,
I am a masters student in the US and I am going to try and get CAESAR working for a semi arid watershed in Texas...the idea hopefully is to model watershed change from sediment flushing flows from three dams on the river.
I am sure I will have more questions as I go along, but I have some I can't seem to find quick answers to.
Does CAESAR only use the Einstein-Brown (1950) sediment transport model? My advisor Joanna Curran (formerly Crowe) mentioned that she saw a talk of the model using the Wilcock-Crowe transport model. Is this true? If not, is it feasible to alter the code to use Wilcock-Crowe? (I am new to programming, so I would have to learn quickly to catch up)
Thanks,
Frank |
|
|
| Tom posted on Wednesday, January 04, 2006 - 10:34 am
|
|
|
Hi Frank,
The latest version of the code - as found on this website - uses the Wilcock-Crowe sediment tpt model, so no problems there.
Cheers,
Tom |
|
|
| Tom posted on Wednesday, January 04, 2006 - 05:13 pm
|
|
|
Frank - if you have an email address - or email me directly, and I can send you a recent paper in press - that shuold have more details on the latest model set up.
Tom |
|
|
|
|
Dear Tom,
May I have this paper in press too??
I'm really intresting to learn more about CAESAR...
Cheers,
Pierluigi |
|
|
| Somaieh (Somaieh) posted on Wednesday, September 28, 2011 - 08:49 am
|
|
|
Dear professor Coulthard
Can we apply CAESAR model for urban rivers changes( such as urbanization of the catchment(land-use change)that lead to channel widening or channel narrowing?
sincerely
somaieh |
|
|
| Somaieh (Somaieh) posted on Thursday, September 29, 2011 - 07:54 am
|
|
|
Dear professor
I am currently a Ph.D student in Physical Geography (Geomorphology). I read ur papers about CAESAR model, but i have many questions.I m writting thesis proposal on titled Geomorphic changes of river landscape in urabn areas.
can I apply and use CAESAR model in this case?
cheers
Somaieh |
|
|
|
|
Hi Somaieh,
This may be easier to deal with via email - unless this is easier for you. What aspect of urban rivers are you hoping to model? Is this the effect of urbanisation upstream or studying what happens to rivers in urban areas for example? Is this about flood risk or geomorphology?
Tom |
|
|
| Back to top |
|
