AMR Refinement Level Question

[Wilfredo Robinson]

Good day Claw Users, 

My name is Wilfredo Robinson, and I am working with EURECOM (France) on a project that uses GeoClaw. I have a question regarding the AMR refinement levels. 

Let's assume that I have a computational domain that is 100km x 100km. I set my coarsest level to 30x30 cells. This means that each side measuring 100km will be divided into 30 cells. This means each cell side is now 100km / 30 = 3.33km. This is my AMR Level 1. 

Now I set my AMR refinement levels as follows: [2, 4]. My AMR Level 2 will now divide my 30x30 computational domain into 60x60 cells, leading to a cell side of 1.66km, twice the resolution of AMR Level 1. Now AMR Level 3 will be have 4 times the resolution that AMR Level 2 has, meaning each cell side will be 0.417km (100km / 30 / 2 / 4 = 0.417km).

I have two questions:

1. Is my interpretation of how AMR works correct? If not, what did I interpret incorrectly?

2. What happens if I make an AMR level lead to a higher resolution than what is actually possible based on the original grid file? For example, let's say my topo file (.asc) has a cell size of 100m (not using degrees here for clarity). If I use this topo file and use an AMR level that makes the cell have a size of 50m, what happens? Does GeoClaw assume its own 50m resolution? Or does it just stick to the minimum possible cell size (100m), which is defined by the original topo file?

Thanks in advance. I await to see if anyone has any feedback for this. 

Cheers, 

Wilfredo

[Kyle Mandli]

Hi Wilfredo,

Sorry for the delay in responding, unfortunately your emails got sent to my SPAM!

1. Your interpretation of resolution is correct.  I also wanted to mention that there is a script in the GeoClaw python tools called resolution.py that helps to calculate this.
2. If you have a resolution that is higher than the given topography GeoClaw automatically handles this for you and integrates the piece-wise bilinear function generated from the best topography data given to find the average value.  Details of this process is in [1].

[1] LeVeque, R. J., George, D. L. & Berger, M. J. Tsunami Propagation and inundation with adaptively refined finite volume methods. Acta Numer 20, 211–289 (2011).

Kyle

--
You received this message because you are subscribed to the Google Groups "claw-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to claw-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/claw-users/bd8d8148-7120-4a70-9cce-a6b5d53b4f45n%40googlegroups.com.

[Wilfredo Robinson]

Hi Kyle,

I see your spam folder does not like me. It and I need to have a serious conversation to sort things out. 

Jokes aside, thanks for the feedback. I'm glad to know my interpretations of resolution is correct, and that there was a resolution.py tool that I did not know about. The detail mentioned in point 2 is also something I will look into. 

Thanks again for your time, and great work with Geoclaw! 

Best, 

Wilfredo