(RESOLVED) Resolution of meteo input data

[andreas-h]

I'm planning to run WRF(-Chem) using the new ERA5 reanalysis as initial/boundary condition.

My outermost domain (centered at 35N) has 81km resolution, and in 5 nests I'm going down to 1km. The reason for this is that my chemical boundary conditions are very coarse (3x2) and I want to give them time to go down to the target resolution of 1x1km.

ERA5 comes in 0.25 degree resolution. Now this is a lot finer than the 81km of my outer domain.

QUESTION: Should I run ungrib/metgrid on the 0.25 degree ERA5 data, or should I rather get coarser resolution data from ECMWF and use that? (e.g., I could get ERA5 at 0.75 degree or at 1 degree ...) Or should I even combine them, i.e., use coarse resolution for 81km domain and finer resolution for 27km domain?

 

[nepomuzem]

In answer to your question I would say yes, use a coarser resoluted input as by taking a 0.25x0.25° you may be loosing topography features a coarser model is taking into account. Let me explain myself, as the 0.25x0.25° topography is about 3-4 times more resoluted and let's say a mountain is 50km wide in the X axis, then the ERA5 will resolve it in 5 points (with values for example, 0m, 300m, 800m, 500m, 0m) if your model takes the outermost points (1 and 5) both are 0m and then you'll be missing the mountain. A more coarse model may be taking this mountain into account in a single point.

But that depends of course on the area you are dealing with.

 

[Ming Chen]

I agree that you should use ERA5 data. For nesting run, we are usually concerned of the simulation in the innermost domain. And fine resolution input data will provide more detailed information for initial and boundary conditions, which will be helpful for the model to yield better results.

By the way, is there any specific reason why your outermost domain has the grid interval of 81km? With the quarter degree ERA as input, you don't have to run WRF with so many nested domains (I suppose you are only concerned of the high resolution results).

 

[hilboll]

Thanks for the info! (I'm actually the original poster, but my username andreas-h cannot log in any more, I get HTTP500 errors or a blank screen ...)

This is a setup for WRF-Chem, and my chemical boundary conditions are on 3x2 degrees resolution, so I need the extra two outer domains to not jump from 3x2 degree to 9km.

Regarding the spatial resolution of the initial/boundary conditions (ERA5), should I use 1 degree for all domains?

Or should I use 1 degree for the 81km domain and the original 0.25 degree for the inner domains?

Or does metgrid spatial interpolation correctly handle the downsampling from .25 degree to 81km, so I can use the 0.25 degree ERA5 data for all domains without problems?

(In principle, one could run WPS once for each resolution of input data and then combine the met_em files from the different ERA5 resolutions depending on the domain)

 

[Ming Chen]

In this case, I would suggest you stay with the 0.25 degree ERA5 data. Yes METGRID can process the data and put it on 81-km grids.
High-resolution input data is always recommended because we expect it can provide more accurate and detailed information.