Hi,
I am trying to set up WRF model and I would like to specify explicit eta levels. My study will primarily be focused on storms and the boundary layer, so ideally, I would like to have very high resolution upto 3 km altitude. I am using three domains of 9km, 3km and 1km resolutions, with two way nesting. I am unaware of any thumb rules we need to follow or factors we need to keep in mind while deciding the vertical levels. Can you please educate me about them and help me out?
Also, I was told that, in case we don't want to specify eta levels, the number of vertical levels needs to be
Height of model top/ (0.1*Resolution of innermost domain)
Can you please help me understand why this is the case?
My model top will be around 20km (50 hPa), so if I go by this rule, I will end up using 200 vertical levels. Please help me understand why it is necessary to have these many levels.
Thankyou.
Check out this paper https://core.ac.uk/download/pdf/38935571.pdf. It has great insight and results about the impact of vertical resolution on your simulations. They discuss the effect of having a high number of layers below 3.5 Km. Note that 200 layers is impractical and even cloud-resolving simulations can get adequate results with ~ 70-90 vertical layers.
There is an early paper by Prof. Kalnay (MIT, formerly director of NCEP) in the 1980s (I think) that suggested using an exponential function for the altitudes so that the smallest vertical interval is close to ground. This was for the SIGMA coordinate system. She also demonstrated its efficiency.
I think the 100 m resolution target for your grid sounds fairly good. Skamarock et al. (2019, https://journals.ametsoc.org/view/journals/mwre/147/7/mwr-d-19-0043.1.xml) tried a similar nested grid in MPAS with 100 m and 800 m vertical spacing. One issue was resolving mesoscale inertia-gravity waves, which tend to have very shallow angle slopes. So they can be much larger in the horizontal than vertical. But they will cause unphysical flows if insufficiently resolved in the vertical. As the previous answers have suggested, a bit of this decision is trial and error. You need to look at the kinetic energy spectrum of your output etc. to see if it is resolving the structures you want.
A related issue is the consistency between vertical & horizontal spacings. A paper was written by Mahrer in the 90s and another by by Fox-Rabinovich & Lindzen
All the best
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python