Dear all,
I am looking for guidelines (a rule of thumb) in order to estimate the "optimal" number of vertical layers in WRF simulations. It is somehow clear that it depends on the horizontal grid resolution, but may also depend on other settings. I am looking for technical as well as more scientific answers, e.g. how many layers are necessary that WRF is running in a stable way, and that results are more accurate.
I was surprised to find only few papers addressing these questions ...
Help is highly appreciated.
Patrick
Dear Patrick,
The issue of selecting the optimum number of vertical layers in WRF is discussed in section 2.3.3 of the paper
https://file.scirp.org/pdf/ACS_2016042911473822.pdf
In general the problem of selecting the optimum number of levels should be formulated in the framework of the optimum control theory with the appropriate cost functional reflecting the model error, unfortunately I’m not aware about this approach being used in practice.
With my best regards,
Janusz
Thank you very much for the answer. Yes, I fully agree, but also could not find any study where this approach has been applied.
I am asking this question, because one reviewer suggested a specific number of vertical levels but did not indicate how he assessed this value.
It depends on what meterological problem and at which range of forecast you adress. As exemple, for fog prediction it is very important to have a very fine vertical discretization close to the ground surface, the lowest level being at 2m with Dz of some meters in the first 5OOm of the boundary layer. On that field, you can see the work of Tradiff (2007). If you are interested for air quality prediction such vertical resolution can be useful to inject industrial sources at the stack height in the meteorological field.
Ref: Tardif R. (2007): The impact of vertical resolution in the explicit numerical forecasting of radiation fog: a case study. Pure Appl. Geophys., 164, 1221-1240.
Thanks for your answer, Dr. Musson-Genon.
We are mostly interested in getting the total surface precipitation from seasonal forecasts. The goal is to provide drought predictions.
I have no personal experience on that domain. Perhaps you can inspire from the resolution used in the operational weather service for limited aera modelling. At exemple AROME model at the french weather service, wich is abble to predict intense precipitation, has a horizontal mesh of 2.5 km with 90 levels on the vertical for 45 h forecast.
The answer to your question about the optimum number of model levels depends mainly on a scale of the problem at hand. In the absence of a formal optimization algorithm suggested in my first answer we can use a set of the quasi heuristic methods.
For the small scale systems, selection of a grid aspect ratio is based on the considerations of the energy spectrum as discussed in
https://www.geosci-model-dev.net/8/3393/2015/gmd-8-3393-2015.pdf
Some practical observations for the high resolution NWP models are collected in the following working paper
https://www.wmo.int/pages/prog/www/OSY/SOG/SoG-HighRes-NWP.pdf
For the problem of drought simulation which is characterized by a much longer time scale I would consider that the number of levels used in the high resolution NWP models is too large.
Some examples from the hydrological modeling indicate that even 30 levels are considered to be sufficient
https://books.google.ca/books?id=dCm9DAAAQBAJ&pg=PT428&lpg=PT428&dq=the+number+of+vertical+levels+in+the+model+for+drought+simulation&source=bl&ots=ohv2XTnH6r&sig=UxhsTlGfNePvuGY5dmwOt-cmaRw&hl=en&sa=X&ved=2ahUKEwittprLwNrZAhVJyVMKHc2yB2UQ6AEwAnoECAgQAQ#v=onepage&q=the%20number%20of%20vertical%20levels%20in%20the%20model%20for%20drought%20simulation&f=false
I hope that these observations will assist you in answering the comments of a reviewer.
JP
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Atmosphere