Numerical dissipation of wave height

229 views
Skip to first unread message

Quan T. Nguyen

unread,
Jun 23, 2021, 4:18:47 AM6/23/21
to funwave-tvd
Dear FUNWAVE Community,

I have been working with the FUNWAVE-TVD model to solve wave-breaking problems. And I have been struggling with generating desired wave height for my study area. 

I used the wavemaker function to generate the irregular wave and validated it to the surface wave height data from the WEkEO Copernicus database. However,  the generated wave still did not fit with the WEkEO data I have. The average range of wave height for WEkEO data is about 1.0 - 2.0m, but my generated result was only around 0 - 0.5m. 

Moreover, when I set up the wavemaker to generate irregular waves from left to right, waves appeared near the land size on the right side (attached picture). And the model was blown up around 430s despite my set up for the simulation time as 2100 seconds. 

I really appreciate all suggestion and support. Thank you in advance.

Regards,
Quan Nguyen.
input.txt
2D_domain_Visc_2.png
depth.txt

Matt Malej

unread,
Jun 23, 2021, 6:14:47 AM6/23/21
to Quan T. Nguyen, funwave-tvd
A few quick pointers upon a quick glance, as your setup has several intertwined issues:

- with dx=dy=10m you are not resolving all of your waves; you are asking the wavemaker to generate waves with min period of Tp=1sec (max freq of 1), which would yield not even two points per shortest wave

- with your peak frequency of 0.03 Hz (roughly 404m peak wavelength) do you need to generate waves up to 1 Hz?

- it’s not easy to generate such long peak wavelengths with the internal wavemaker, as you are essentially in the infragravity (IG) band; is your wave field really dominated by IG waves with peak frequency of 0.03 Hz? In similar applications we select min freq as 0.03, max as 0.3 and if you have say 10s waves we impose peak freq of 0.1; lower harmonics are naturally generated

- also, you are trying to generate 5m wave heights in 15m depths, could your waves be immediately breaking inside the wavemaker?

- you set PERIODIC=T, but is your domain truly periodic north-south? It doesn’t appear to be

- finally, if instability persists, you may consider lowering your CFL, then you might even apply BATHY_CORRECTION



On Jun 23, 2021, at 4:18 AM, Quan T. Nguyen <quannguy...@gmail.com> wrote:

Dear FUNWAVE Community,
--
You received this message because you are subscribed to the Google Groups "funwave-tvd" group.
To unsubscribe from this group and stop receiving emails from it, send an email to funwave-tvd...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/funwave-tvd/ed2b6899-27b4-4f1f-b278-0d07c585dc7en%40googlegroups.com.
<input.txt>
<2D_domain_Visc_2.png>
<depth.txt>

iantag

unread,
Jun 23, 2021, 9:28:54 AM6/23/21
to funwave-tvd
Quan - your problem is interesting inasmuch as I have struggled with similar issues as I tried to selectively coarsen a model (to obtain faster runtime). I noticed a drastic ( and puzzling) reduction in travelling  wave amplitudes - that I now attribute to truncation errors. In essence  Matt's  comment about minimum # of points per wavelength is spot on. I suspect you are going to have to significantly increase your grid resolution. Note recommendations for 50-80 points/cells  per wavelength !!

Here's my suggestion. Make an equivalent 1D E-W slice (in your case) and study the effect on grid size on propagated amplitude to shore. You may have to go down to 0.5m spacing or less to resolve the features you seek. Then you can selectively coarsen and see the "truncation" impact. In a "real" beach situation(with  waves  2km to shore)  i have models that work OK at 1m spacing but significantly degrade at 2.5m and 5m spacing ( to the point that amplitudes are easily halved). The degradation is a severe attenuation in the amplitude as wave travel to the beach ( even with Cd=0). The good news - 1D models are fairly quick to run !

Lastly in regards bathy_correction. I've had mixed results - sometimes if works fine - and other times it makes things worse - for reasons I've yet to understand. You might want to consider the breakwater keyword as this allows you to place attenuation (mini sponges)  in flagged cells ( selectively placed at blowup points) - just a small amount of attenuation can tame a troublesome area. ( The root cause is usually abrupt depth changes- so a smooth bathymetry definitely  helps)

Look in this forum for discussion about wave breaking models ceasing to report breaking as grid size is increased. Same/similar issues.

If you do the 1D models - as suggested - please let us know what you find out.

Quan T. Nguyen

unread,
Jun 23, 2021, 1:25:02 PM6/23/21
to iantag, matt....@gmail.com, funwave-tvd
Dear Iantag, 
Dear Matt Malej,

Thank you a lot for your suggestion. Based on your suggestion, I have several changes for this case.
  • For the study domain, I have restricted the area for approximately waves at 10km to shore. And applying an interpolation method to increase the grid size of bathymetry data as Mglob x Nglob = 2666 x 3794 nodes with the resolution of 5m/node.
  • I have done the test with 1D model as iantag's suggestion and modified the wavemaker parameter as Matt Malej's suggestion. Here is what I found out:
    • For dx = 0.5, the model could not even run. 
    • For dx = 1.0, the model could run, but blew up after 75 seconds. 
    • For dx = 2.0, the model could run normally. The generated waves were about +- 2.0m for the first 10km, far from that, the wave height significantly decreased to less than 0.5m (attached picture). 
  • When I used the setting of the wavemaker in the 1D test for 2D calculation, the model could not run even though I changed the DX, DY varied between 1, 2, 5, and 10m.
Then, I will try to test other suggestions. If you have any advice, please let me know. 
Thank you in advance.

Regards,
Quan.
 



You received this message because you are subscribed to a topic in the Google Groups "funwave-tvd" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/funwave-tvd/Of7TOl4xxOU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to funwave-tvd...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/funwave-tvd/a53ecfdf-fe96-44b5-b97a-8fc9509c6ccen%40googlegroups.com.
input_1D.txt
1D_eta_Visc_Cross.png

Ian Taggart

unread,
Jun 23, 2021, 7:58:01 PM6/23/21
to Quan T. Nguyen, matt....@gmail.com, funwave-tvd
Quan - looked at your 1D MODEL. 

PX=PY=2 may not be a good choice for Nglob =3. Set PX=4.  My models can run ok with Nglob=1 (v3.5)
Also try setting method to third order. ( not fourth). (I've read long term solution may be better - with less instability - with third.i tend to use third order for all runs.  )
I see you reduced CFL ... puzzled why you still have blow up in 1d. 
Is it worth trying monochromatic waves ? ( to sort things out). ??


Sent from my iPad
<input_1D.txt>
<1D_eta_Visc_Cross.png>

iantag

unread,
Jun 23, 2021, 10:00:22 PM6/23/21
to funwave-tvd
Hi Quan - just looked at your eta 1D profiles.  There is something 'odd' with the wetting-drying happening ( on dry land ??) here. Suggest - in these 1D models you edit the depths here to take out the surface elevation variations by setting a smooth transition to say (-4m  say). That way waves ( travelling in from the west) can run up to the shore in your 1D model. You could also play with the minimum drying depths - but suggest taking out the wetting drying issues altogether. Sort out the travelling wave issue first.

Getting a nice result at 1m spacing is a necessary 1st step.

Quan T. Nguyen

unread,
Jun 24, 2021, 12:16:39 AM6/24/21
to iantag, funwave-tvd
Thank you a lot for your suggestions. 
I will run some tests with these modifications and update you accordingly.

Quan.

Reply all
Reply to author
Forward
0 new messages