WRF hardware

Looking for new hardware to run WRF? Intel or AMD? Check this forum.
Post Reply
calvus17
Posts: 4
Joined: Mon Feb 29, 2016 7:53 am

WRF hardware

Post by calvus17 » Thu Mar 03, 2016 10:49 am

Dear all,

I'd like to run WRF-ARW with two domains on ubuntu 14:

1. Mother domain, Central Europe (30 km resolution, 72*48)
2. Second domain, covering Hungary (6 km resolution, 120*80)

The forecasted time would be 48 hours and one forecast per day. Simulation time should be around 6-7 hours.
I plan to buy the following hardware:

- 1x Intel® Core™ i7-5820K (6 cores)
- Vengeance® LPX 16GB (4x4GB) DDR4 DRAM 3000MHz C15 Memory Kit
- MSI X99A RAIDER


What do you think it would be enough? And what is the advantage using CentOS operating system?

Best regards,

Norbert

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: WRF hardware

Post by meteoadriatic » Thu Mar 03, 2016 11:29 am

That should be much more power than you will need for such small domains, I believe. Ubuntu vs Centos, I don't think you will face any difficulties with any of them.

calvus17
Posts: 4
Joined: Mon Feb 29, 2016 7:53 am

Re: WRF hardware

Post by calvus17 » Thu Mar 03, 2016 12:56 pm

meteoadriatic wrote:That should be much more power than you will need for such small domains, I believe. Ubuntu vs Centos, I don't think you will face any difficulties with any of them.
Hi Meteoadriatic,

Thanks for your reply! So I think that I should rather go with a 3-nested domain and 1:3 domain ratio, e.g. 54 km resolution (120*80), 18 km (120*80) and 6 km (120*80). That would be surely better in forecast quality than the 2-nested. Could you perhaps suggest a possible domain configuration better suitable for that hardware (the budget allows for that power) or should I make some simulation runs when I have it to found out an optimal setting.
Thanks a lot.
regards,
Norbert

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: WRF hardware

Post by meteoadriatic » Thu Mar 03, 2016 5:45 pm

Before we go deeper into answering, one question. What is intendend usage of that setup? Some reanalysis, operational forecast, something else? And what is input data?

calvus17
Posts: 4
Joined: Mon Feb 29, 2016 7:53 am

Re: WRF hardware

Post by calvus17 » Fri Mar 04, 2016 7:27 am

Hi,
It would be operational forecast for agriculturer purposes. So the most important parameter would be the (convective) precipitation in the growing season (which is of course the most uncertain for exact locations, so I also calculate a precipitation probability in the postprocessing). I am using now the above mentioned 2-domain setting with mp_physics 14 (WRF double-moment 5-class scheme) in the second domain and cu_physics 4 (simplified arakawa schubert) on an old laptop based on some published research. The input data is GFS 0.5°, which horizontal resolution is about 30 km in central Europe.
Thanks for your suggestions!

My namelist.input:

&time_control
run_days = 0,
run_hours = 3,
run_minutes = 0,
run_seconds = 0,
start_year = 2016, 2016,
start_month = 1, 1,
start_day = 25, 25,
start_hour = 12, 12,
start_minute = 0, 0, 0,
start_second = 0, 0, 0,
end_year = 2016, 2016,
end_month = 1, 1,
end_day = 25, 25,
end_hour = 15, 15,
end_minute = 0, 0, 0,
end_second = 0, 0, 0,
interval_seconds = 10800,
input_from_file = .true., .true., .false.,
fine_input_stream = 0, 0, 0,
history_interval = 60, 60,
frames_per_outfile = 1000, 1000, 1000,
restart = .false.,
restart_interval = 5000,
io_form_history = 2,
io_form_restart = 2,
io_form_input = 2,
io_form_boundary = 2,
debug_level = 0,
/

&domains
time_step = 120,
max_dom = 2,
s_we = 1, 1,
e_we = 72, 131,
s_sn = 1, 1,
e_sn = 48, 86,
s_vert = 1, 1,
e_vert = 30, 30,
num_metgrid_levels = 27,
dx = 30000, 6000,
dy = 30000, 6000,
grid_id = 1, 2,
parent_id = 1, 1,
i_parent_start = 1, 24,
j_parent_start = 1, 17,
parent_grid_ratio = 1, 5,
parent_time_step_ratio = 1, 5,
feedback = 1,
/

&physics
mp_physics = 3, 14, 3,
ra_lw_physics = 1, 1, 1,
ra_sw_physics = 1, 1, 1,
radt = 50, 10, 30,
sf_sfclay_physics = 1, 1, 1,
sf_surface_physics = 1, 1, 1,
num_soil_layers = 5,
bl_pbl_physics = 1, 1, 1,
bldt = 0, 0, 0,
cu_physics = 1, 4, 0,
cudt = 5, 5, 5,
prec_acc_dt = 60, 60,
/

&dynamics
w_damping = 0,
diff_opt = 1,
km_opt = 4,
diff_6th_opt = 0,
diff_6th_factor = 0.12,
base_temp = 290.0,
damp_opt = 0,
zdamp = 5000.0, 5000.0, 5000.0,
dampcoef = 0.2, 0.2, 0.2,
khdif = 0, 0, 0,
kvdif = 0, 0, 0,
non_hydrostatic = .true., .true., .true.,
moist_adv_opt = 1,
scalar_adv_opt = 1,
/

&bdy_control
spec_bdy_width = 5,
spec_zone = 1,
relax_zone = 4,
specified = .true., .false., .false.,
nested = .false., .true., .true.,
/

&metadata
domain = 'baseline_europe',
model = 'WRF',
model_run = 'operational',
/

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: WRF hardware

Post by meteoadriatic » Fri Mar 04, 2016 5:05 pm

Better idea is to use GFS 0.25°.

Having three domains with outermost of 54km is waste of resources. GFS 0.25° has resolution of about 25km (0.25*111km), why would you want to decrease that by making third domain with 54km resolution? You can pretty safely use under 10km resolution in outermost domain, when you use GFS 0.25° as input data. In that case I would go with standard setup of 12km outer domain and 4km inner domain, which should work OK with your mentioned hardware setup, but even 18/6 should work OK unless you do not worry much about convection parametrization problem at 6km resolution.

calvus17
Posts: 4
Joined: Mon Feb 29, 2016 7:53 am

Re: WRF hardware

Post by calvus17 » Mon Mar 07, 2016 8:13 am

Hi,
Many thanks for the suggestions. I will try the GFS 0.25 with 12/4 km resolution and compare the results (precip) with the 0.5 GFS (30/6 km). I think it will be interesting to see how the differences are developing.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest