Unrealistic rain

Topics related to current and future physics in the WRF as well as any problems you may have.

Unrealistic rain

Postby francesco.pasi » Mon Jul 28, 2008 5:45 am

Hi all sorry to post again the same problem but I lost the thread in the forum change.
Model: nmm v3.0
We are experiencing an unrealistic critical behavior during summer time (convective period)
Model produces unrealistic (= not happened) stationary rain in isolated grid points.
- This is a random behavior meaning that it does not happen always in the same grid point, but there are preferred grid points ( mostly are grid points where there is a sea-land boundary, mountain tops, lakes).
- The behavior is resolution dependent and it seems to be more strong with finer resolutions (see figure 1 and 2)
- The model "goes crazy" in defined atmospheric situations (in the example shown there is some instability and wind is blowing from the north, north-east on land and from the north-west on sea, so there is a sort of low level convergence that might trigger rain).
Does anyone have experienced something similar?
Thanks for help
Francesco

Figure caption:
Figure 1: 6h prec of nmm 12km resolution with BMJ scheme
Figure 2: 6h prec of nmm 8km resolution with BMJ
Figure 3: 6h prec of ecmwf initial data (no signal)
Figure 4: 10mwind pattern

Image
Image
Image
Image
francesco.pasi
 
Posts: 23
Joined: Thu Jul 24, 2008 9:05 am

Re: Unrealistic rain

Postby PBLer » Mon Jul 28, 2008 10:44 am

I don't have as much experience with the NMM, but its probably worth posting the physics and dynamics sections of your namelist. Someone may see an alternative option that might work better for you. For example, if the radiation isn't called frequently enough (radt), the model may continue to heat the surface well after cloud formation and increase the intensity of your convection...
Last edited by PBLer on Mon Jul 28, 2008 11:19 am, edited 2 times in total.
PBLer
 
Posts: 38
Joined: Wed May 07, 2008 5:11 pm

Re: Unrealistic rain

Postby jimmyc » Mon Jul 28, 2008 11:10 am

Is the rain generated from the grid scale microphysics or the convective scheme?

What is happening aloft? Is there a trough in the region?

BMJ will trigger convection if the sounding has some CAPE that can be removed and will likely be triggered as long as rain is produced during the adjustment. If sufficient moisture is present between 850 and 600 hPa, rainfall will be generated.
BMJ can also initiate shallow convection and destabilize the grid point. The persistence of the convergence then may produce large rainfall rates if BMJ is being called frequently.

You probably want to plot the soundings (hourly) at those grid points to diagnose what is happening and why.
The views expressed in this message do not necessarily reflect those of NOAA or the National Weather Service or the University of Oklahoma.
James Correia, Jr
jimmyc
 
Posts: 519
Joined: Tue Apr 15, 2008 1:10 am

Re: Unrealistic rain

Postby francesco.pasi » Tue Jul 29, 2008 7:02 am

Hi thanks for suggestions.
1) I tried with more frequent call to the radiation (nrads, nradl) but there is no significant change. The used value is the recommended NCEP (3600/dt) I pushed it to 1800/dt but no differences.
2) The rain is mainly generated by the microphysics. In the sense that in the grid point convective rain is 30 mm and microphysics 60 mm.
3) No large scale anomaly (no upper trough or else). Critical situation is thermal instability with light winds.
4) I am working on vertical soundings to see what happen

Do you think that if I make less call to the cumulus scheme this might be effective?

I attach you my namelist (quite standard I suppose)
/ &domains
time_step = 30,
time_step_fract_num = 0,
time_step_fract_den = 1,
max_dom = 1,
s_we = 1, 1,
e_we = 128, 140,
s_sn = 1, 1,
e_sn = 232, 240,
s_vert = 1, 1,
e_vert = 35, 35,
num_metgrid_levels = 10,
dx = 0.077373, 0.025791,
dy = 0.076473, 0.025491,
grid_id = 1, 2,
parent_id = 0, 1,
i_parent_start = 1, 44,
j_parent_start = 1, 100,
parent_grid_ratio = 1, 3,
parent_time_step_ratio = 1, 3,
p_top_requested = 10000.,
ptsgm = 42000.,
eta_levels = 1.000, 0.993, 0.986, 0.978, 0.970, 0.960, 0.950, 0.940, 0.930, 0.920,
0.900, 0.880, 0.860, 0.830, 0.800, 0.770, 0.740, 0.710, 0.680, 0.640,
0.600, 0.560, 0.520, 0.480, 0.420, 0.360, 0.300, 0.240, 0.200, 0.160,
0.120, 0.080, 0.040, 0.020, 0.000
tile_sz_x = 0,
tile_sz_y = 0,
numtiles = 1
/

/ &physics
mp_physics = 5, 5,
ra_lw_physics = 99, 99,
ra_sw_physics = 99, 99,
nrads = 60, 360,
nradl = 60, 360,
co2tf = 1,
sf_sfclay_physics = 2, 2,
sf_surface_physics = 99, 99,
bl_pbl_physics = 2, 2,
nphs = 6, 18,
cu_physics = 2, 2,
ncnvc = 6, 18,
tprec = 3,
theat = 6,
tclod = 6,
trdsw = 6,
trdlw = 6,
tsrfc = 6,
pcpflg = .false.,
isfflx = 1,
ifsnow = 0,
icloud = 1,
num_soil_layers = 4,
maxiens = 1,
maxens = 3
maxens2 = 3
maxens3 = 16
ensdim = 144
mp_zero_out = 0
/
francesco.pasi
 
Posts: 23
Joined: Thu Jul 24, 2008 9:05 am

Re: Unrealistic rain

Postby jimmyc » Tue Jul 29, 2008 11:02 am

So you are using nesting from 8km to 2 grid spacing?

I would turn off the cumulus scheme to see what happens. The other option is call the pbl scheme more frequently.

Light winds suggests to me that diffusion may not be strong enough and this is really grid scale noise and storms. Try turning on or ramping up the diffusion.
The views expressed in this message do not necessarily reflect those of NOAA or the National Weather Service or the University of Oklahoma.
James Correia, Jr
jimmyc
 
Posts: 519
Joined: Tue Apr 15, 2008 1:10 am

Re: Unrealistic rain

Postby PBLer » Tue Jul 29, 2008 11:57 am

I don't see anything obviously wrong, but I did notice the following note (at http://www.wrfportal.org/namelist_input_options.html):

Note: If it is desired to run GFDL with a microphysics scheme other than Ferrier, a modification to module_ra_gfdleta.F is needed to comment out (!) #define FERRIER_GFDL.

For the ARW, I have been told that the GFDL radiation should not be used with the Ferrier scheme. This may also be true for the NMM. It could be that the GFDL scheme expects different microphysical species than the Ferrier scheme passes it... Again, I'm not very familiar with these schemes... I doubt the MYJ PBL/scf layer scheme is at fault here, since the sfc layer scheme has the Beljaars (1994) correction for unstable conditions with calm winds...
PBLer
 
Posts: 38
Joined: Wed May 07, 2008 5:11 pm

Re: Unrealistic rain

Postby francesco.pasi » Thu Jul 31, 2008 7:17 am

Jimmyc: sorry for posting wrong: we are not running nested simulations. The namelist I posted is wrong in the sense that the second column is not used (I forgot to take it out). Physical settings are the ones of the first column. I posted two resolutions to show that the problem might be also resolution dependent.
PBLer: I think that for NMM this is not the case as Ferrier and GFDL are the schemes suggested and used operationally by NCEP, so they should work fine.
I show you another case may be of interest with the very standard 12km resolution and GFS init data.
Please note that model domain is not as little as shown so it is not a boundary problem
The first 3 figures show that the model with BMJ is "going crazy" near Croatia while KF with the same data and physics not
The last 2 figures show that the model with KF is "going crazy" over Corse island while BMJ with the same data and physics not
This make me think that the problem is not completly due to the cumulus scheme but is more in depth to the model physics

Image
Image
Image


---------------------------------------------------------------------------------------------------------------------------------------------------------
---------------------------------------- SECOND PROBLEM ------------------------------------------------------------------------
---------------------------------------------------------------------------------------------------------------------------------------------------------


Image
Image
francesco.pasi
 
Posts: 23
Joined: Thu Jul 24, 2008 9:05 am

Re: Unrealistic rain

Postby jimmyc » Thu Jul 31, 2008 10:27 am

Certainly you are getting some kind of grid point storm. But if the problem evolves as both grid spacing and physics are changed then you are looking at a sensitivity to the initial conditions. These models won't get all the details correct, especially for precipitation. No suite of physics schemes ever gets it right all the time ... one scheme is not better than any other. Physics is all about user preference and the intended purpose of the model.

That being said, I still think these are grid point storms (the ones where precip exceeds 20 mm per 6 hours).
So I suggest not using a convective scheme for one of your cases and see if the problem appears again. If it doesn't, it means the cumulus scheme in concert with the initial data is the issue. Not too much to do to fix it. If the problem appears again (precip will be higher in the explicit run but we are looking to make sure that grid point storms are not forming), then it may only be the initial data that is the issue.

How to get rid of grid point storms? The convection scheme is supposed to handle that for you. The only other option is make more frequent calls to the physics schemes. This may not remove the grid point storms but it should reduce the precip.
The views expressed in this message do not necessarily reflect those of NOAA or the National Weather Service or the University of Oklahoma.
James Correia, Jr
jimmyc
 
Posts: 519
Joined: Tue Apr 15, 2008 1:10 am

Re: Unrealistic rain

Postby MeteoAdriatic » Fri Aug 22, 2008 6:48 am

I have same problem with my test setup. On fine grids, In slighly unstable conditions, model forecast precipitation bombs that are very very unlikely to happen in nature.

Example for today:

Image
Image

On parent grids there is no such feature. I run 17km - 7km - 4km nest, and have CU BMJ on 17km and CU off on 7 and 4km grids.

:roll:

BTW, I didn't noticed this behaviour yet on ARW core where I run 2-way nesting 18km-6km. Initial and boundary conditions are fron GFS 0.5° in both cases.
MeteoAdriatic
 
Posts: 14
Joined: Wed Apr 30, 2008 6:07 pm

Re: Unrealistic rain

Postby MeteoAdriatic » Sun Aug 24, 2008 6:36 pm

I tried to swicth on BMJ on 4km after some googleing for "grid point storm". Nothing much different from going without CP.. "Forecast" for tommorow - 143mm of rain in only one hour:

Image

:(

ARW still does good job. What can I try? Any idea?

Thx
MeteoAdriatic
 
Posts: 14
Joined: Wed Apr 30, 2008 6:07 pm

Next

Return to Physics

Who is online

Users browsing this forum: No registered users and 4 guests