How would one alter the boundary condition handling in WRF?

[Pechudin]

Hello beautiful community,

I have tried using WRF to run pseudo-idealized simulations. Basically, what I did was:

1. Fetch some boundary and initial conditions (e.g., ERA5)
2. Replace them with fields uniform in (x, y) space, effectively replacing the input data with a height-dependent sounding
3. Flatten terrain and replace the static fields with simple, uniform values
4. Add a mountain

The simulation results make sense, but after some time artefacts appear on the eastern side (picture of U10 and V10 are included; this is a case of easterly wind with an inversion and a critical level at 2 km height with flow over a 1 km tall, 60 km wide and 600 km long mountain). Additionally, since there is no inflow forcing, the boundary layer speed decreases with time.



The reason for this is that WRF does not allow to have specified values at one edge and e.g., radiation boundary conditions on other sides. All must be specified or none. I was hoping I could get semi-stationary results by making the parent domain really large but the attenuation with time prevents this from happening. Alas, it seems like this will not be possible.

My question is this (well, several questions):

1. Have there been attempts to implement inflow boundary conditions on one side in WRF? I know it would be really helpful for e.g., mountain wave simulations (idealized).
2. How complicated would they be to implement? I have a desire to dabble in WRF code, but it would be nice to have a map which illustrates how various modules and programs interact in WRF (i.e., this routine calls this routine which does X etc.). is there such a map?

Again, I am thinking about investing time into perhaps learning more about NWP by trawling through WRF code. This could be a fun project, but I would like to know where to start (such as a map of how routines interact) to start.

 

[Ming Chen]

Hi, Pechudin,

Thank you for the detailed explanation of your tests. Please see my answers to your questions:

My question is this (well, several questions):

1. Have there been attempts to implement inflow boundary conditions on one side in WRF? I know it would be really helpful for e.g., mountain wave simulations (idealized).

You may know that WRF has entered maintenance status, which means that further WRF development has been frozen. Due to limited human power here in NCAR, I don't think we are able to develop capabilities that implement different boundary conditions for WRF.

1. How complicated would they be to implement? I have a desire to dabble in WRF code, but it would be nice to have a map which illustrates how various modules and programs interact in WRF (i.e., this routine calls this routine which does X etc.). is there such a map?

Please take a look at the document here. Hope it is helpful for you.

Again, I am thinking about investing time into perhaps learning more about NWP by trawling through WRF code. This could be a fun project, but I would like to know where to start (such as a map of how routines interact) to start.

I would suggest that you switch to MPAS, which is a cutting-edge modeling system that has been undergoing robust development. And we provide strong support for MPAS application.

 

[Ojesutofunmi]

Hello beautiful community,

I have tried using WRF to run pseudo-idealized simulations. Basically, what I did was:

1. Fetch some boundary and initial conditions (e.g., ERA5)
2. Replace them with fields uniform in (x, y) space, effectively replacing the input data with a height-dependent sounding
3. Flatten terrain and replace the static fields with simple, uniform values
4. Add a mountain

The simulation results make sense, but after some time artefacts appear on the eastern side (picture of U10 and V10 are included; this is a case of easterly wind with an inversion and a critical level at 2 km height with flow over a 1 km tall, 60 km wide and 600 km long mountain). Additionally, since there is no inflow forcing, the boundary layer speed decreases with time.

View attachment 15226View attachment 15227

The reason for this is that WRF does not allow to have specified values at one edge and e.g., radiation boundary conditions on other sides. All must be specified or none. I was hoping I could get semi-stationary results by making the parent domain really large but the attenuation with time prevents this from happening. Alas, it seems like this will not be possible.

My question is this (well, several questions):

1. Have there been attempts to implement inflow boundary conditions on one side in WRF? I know it would be really helpful for e.g., mountain wave simulations (idealized).
2. How complicated would they be to implement? I have a desire to dabble in WRF code, but it would be nice to have a map which illustrates how various modules and programs interact in WRF (i.e., this routine calls this routine which does X etc.). is there such a map?

Again, I am thinking about investing time into perhaps learning more about NWP by trawling through WRF code. This could be a fun project, but I would like to know where to start (such as a map of how routines interact) to start.

Hi Pechudin
I am interested in using WRF to run pseudo-idealized simulations, what is the best way to reach out to you, i would like some clarifications on some matter.

 

[Pechudin]

You may know that WRF has entered maintenance status, which means that further WRF development has been frozen. Due to limited human power here in NCAR, I don't think we are able to develop capabilities that implement different boundary conditions for WRF.

I did not infact know that. I heard from one of the senior professors that a team (allegedly based in Serbia) is still going to take care of the WRF code?

I would suggest that you switch to MPAS, which is a cutting-edge modeling system that has been undergoing robust development. And we provide strong support for MPAS application.

I'll take a look at that. Might be better to take the plunge and switch to a newer project.

Hi Pechudin
I am interested in using WRF to run pseudo-idealized simulations, what is the best way to reach out to you, i would like some clarifications on some matter.

I think you can hover over the user icon and use "Start conversation". I will tell you what I did, but I have to warn you that I did not get very far, but did learn some interesting stuff.

 

[Pechudin]

Please take a look at the document here. Hope it is helpful for you.

Thank you again for the file, however, one more question - is this call tree different (and how) for the idealized cases (if one wished to modify an idealized case, for example)?

 

[Ming Chen]

There is no specific document for ideal cases. However, the flowchart generally is same for real-data and ideal cases.