ndown approach vs feedback=0 approach : is there any difference ?

[DeebakVijay]

Greetings,
As a part of my work , I am dynamically scaling down the meteorological parameters like wind speed to high resolution. I am using one way nesting for this. I am wondering if there is any difference in " ndown approach" vs "feedback=0 approach ".

One thing that I notice is in ndown approach , the wrf_output files of parent domain becomes the boundary condition for the child domain. Hence the frequency of parent domain output determines the data available for boundary condition in child domain model run.

But , in "feedback=0 approach " both the runs happen parallely and the boundary forcings happen from parent to child domain after each time step of parent domain.

1. Is this the right description about the algorithm ?
2. If so does it make any difference in the solution?
3. Is there any other thing different between each approach?
4. Which would be the right choice for my work?

Thank you. Waiting for your reply.

 

[Willevera]

Greetings,
As a part of my work , I am dynamically scaling down the meteorological parameters like wind speed to high resolution. I am using one way nesting for this. I am wondering if there is any difference in " ndown approach" vs "feedback=0 approach ".

One thing that I notice is in ndown approach , the wrf_output files of parent domain becomes the boundary condition for the child domain. Hence the frequency of parent domain output determines the data available for boundary condition in child domain model run.

But , in "feedback=0 approach " both the runs happen parallely and the boundary forcings happen from parent to child domain after each time step of parent domain. TellPopeyes

1. Is this the right description about the algorithm ?
2. If so does it make any difference in the solution?
3. Is there any other thing different between each approach?
4. Which would be the right choice for my work?

Thank you. Waiting for your reply.

Hello,

Yes, your description is generally correct. In the "ndown approach," the output from the parent domain is used as the boundary condition for the child domain, and the frequency of parent domain output determines the data available for the child domain. On the other hand, in the "feedback=0 approach," both parent and child domains run in parallel, and the boundary conditions are exchanged between them after each parent domain time step.

There can be differences in the solutions between these approaches. In the "ndown approach," the child domain is constrained by the parent domain's output frequency, potentially leading to a loss of high-frequency variability. In contrast, the "feedback=0 approach" allows both domains to run independently, which could capture higher-frequency features. However, it may also introduce inconsistencies if the domains are not well-matched or if there are communication delays between them.

The choice depends on your specific goals and trade-offs. If high-resolution is crucial, the "feedback=0 approach" might be preferred, but it requires careful tuning and can be computationally expensive. The "ndown approach" is simpler but might not fully capture high-frequency information. Assessing model performance and comparing against observations can help in selecting the appropriate approach.