Hi there,

Is there any guideline for setting the xwavenum and ywavenum of the spectral nudging?

The default setting is 3, what's the corresponding wavelength value?

Thanks.

Siu-fai

7 posts
• Page **1** of **1**

Hi there,

Is there any guideline for setting the xwavenum and ywavenum of the spectral nudging?

The default setting is 3, what's the corresponding wavelength value?

Thanks.

Siu-fai

Is there any guideline for setting the xwavenum and ywavenum of the spectral nudging?

The default setting is 3, what's the corresponding wavelength value?

Thanks.

Siu-fai

- phfate
**Posts:**2**Joined:**Fri Nov 20, 2009 6:28 am

Hi ,

I have the same problem, I looked Fortran programs that match the spectral nudging but it does not help me to understand to which wavelength value the xwavenum and ywavenum correspond.

Did you find any solution,

thank you

I have the same problem, I looked Fortran programs that match the spectral nudging but it does not help me to understand to which wavelength value the xwavenum and ywavenum correspond.

Did you find any solution,

thank you

- homrani
**Posts:**3**Joined:**Thu Mar 11, 2010 6:10 am

Having read

Regional Climate Simulations over North America: Interaction of Local Processes with Improved Large-Scale Flow

GONZALO MIGUEZ-MACHO, GEORGIY L. STENCHIKOV, AND ALAN ROBOCK

in Journal of Climate, 2005

and

Spectral nudging to eliminate the effects of domain position and geometry in regional climate model simulations

Gonzalo Miguez-Macho, Georgiy L. Stenchikov, and Alan Robock

in JGR, 2004, this is how I understand it:

Compute the size of your domain in both directions to obtain SX (DX times grid spacing) and SY, then divide these by xwavenum and ywavenum to obtain the wavelengths. In their experiment, they had 150 grid points in the X direction and 108 grid points in the Y direction and a grid spacing of 50 km. This means that SX was 7500 km and SY 5400 km. They used xwavenum = 3 and ywavenum = 2 to nudge features with wavelengths of about 2500 km and upwards.

In other words, the choice of xwavenum and ywavenum depends strongly on your domain size. They also used a relaxation time of 5000 s, which would correspond to gph, guv and gt values of 0.0002, i.e. slightly weaker nudging than the WRF namelist default values of 0.0003 (corresponding to a relaxation time of 3333 s, or about 1 hour).

Hope this helps.

Best regards,

Erik Kolstad

StormGeo

Bergen, Norway

Regional Climate Simulations over North America: Interaction of Local Processes with Improved Large-Scale Flow

GONZALO MIGUEZ-MACHO, GEORGIY L. STENCHIKOV, AND ALAN ROBOCK

in Journal of Climate, 2005

and

Spectral nudging to eliminate the effects of domain position and geometry in regional climate model simulations

Gonzalo Miguez-Macho, Georgiy L. Stenchikov, and Alan Robock

in JGR, 2004, this is how I understand it:

Compute the size of your domain in both directions to obtain SX (DX times grid spacing) and SY, then divide these by xwavenum and ywavenum to obtain the wavelengths. In their experiment, they had 150 grid points in the X direction and 108 grid points in the Y direction and a grid spacing of 50 km. This means that SX was 7500 km and SY 5400 km. They used xwavenum = 3 and ywavenum = 2 to nudge features with wavelengths of about 2500 km and upwards.

In other words, the choice of xwavenum and ywavenum depends strongly on your domain size. They also used a relaxation time of 5000 s, which would correspond to gph, guv and gt values of 0.0002, i.e. slightly weaker nudging than the WRF namelist default values of 0.0003 (corresponding to a relaxation time of 3333 s, or about 1 hour).

Hope this helps.

Best regards,

Erik Kolstad

StormGeo

Bergen, Norway

--------

Erik Kolstad

Bjerknes Centre for Climate Research

Bergen, Norway

http://ewk.no

Erik Kolstad

Bjerknes Centre for Climate Research

Bergen, Norway

http://ewk.no

- erikwkolstad
**Posts:**7**Joined:**Tue Jan 11, 2011 10:25 am

Moin,

I run WRF in hindcast mode, simulating the last 5 years with a coarse domain (30km grid spacing, no nesting). I use FNL as forcing and like to apply spectral nudging.

I like to nudge uv, t and ph above the PBL and above the 18 model level (above ~850 hPa level at standard atmosphere with my eta level settings).

I have one questions on the spectral nudging settings:

1. What does the parameter gfdda_end_h stand for and how do i need to set it if i like to apply spectral nudging during the full five year simulation. The FNL forcing has a time interval of 6h so gfdda_interval_m = 360 ( time interval (min) between analysis times). Will gfdda_end_h = 6 do the job?

My spectral nudging settings are below:

&fdda

grid_fdda = 2,

gfdda_inname = "wrffdda_d<domain>",

gfdda_interval_m = 360,

gfdda_end_h = 6,

io_form_gfdda = 2,

fgdt = 0,

if_no_pbl_nudging_uv = 1,

if_no_pbl_nudging_t = 1,

if_no_pbl_nudging_q = 1,

if_no_pbl_nudging_ph = 1,

if_zfac_uv = 1,

k_zfac_uv = 18,

if_zfac_t = 1,

k_zfac_t = 18,

if_zfac_q = 1,

k_zfac_q = 18,

if_zfac_ph = 1,

k_zfac_ph = 18,

guv = 0.0003,

gt = 0.0003,

gq = 0,

gph = 0.0003,

xwavenum = 4,

ywavenum = 3,

if_ramping = 0,

dtramp_min = 60.0,

Below is my reasoning to derive these spectral nudging settings: Please advise if my understanding of the methods and settings is erroneous or if you have other recommendations.

What parameters to nudge?

I choose to nudge temperature (t), horizontal wind components (uv) and geopotential (ph). I did not nudge moisture (q) in line with a recommendation by e.g. Macho et al (2005) Regional Climate Simulations over North America: Interaction of Local Processes with Improved Large-Scale Flow in Journal of Climate, 200.5

Is there any experience in the community on the benefits of nudging temperature in addition to uv, ph?

Xwavenum, ywavenum settings:First I figured out which scales I want to nudge: scales that are already reliably modeled by the forcing data set and scales on which the regional model may not add value. (e.g. von Storch et al. 2000)

The forcing FNL analysis data has a horizontal grid spacing of 1°, so I assume that it can reasonably well resolve waves with wave lengths of 700 km or above. Thus I wanted to nudge waves with wave lengths > 700 km. Following Eriks earlier posting i calculated the required domain size by multiplying my grid points in x and y with the grid spacing.

So: e_we = 97, e_sn = 75, dx=30km

xwavenum = 3: wavelength=97*30km /3 = 970km

xwavenum = 4: wavelength=97*30km /4 = 727.5 km

xwavenum = 5; wavelength=97*30km /5 = 582 km

ywavenum = 2; 75*30km /2 = 1125km

ywavenum = 3; 75*30km /3 = 750km

ywavenum = 4; 75*30km /4 = 562.5 km

So i went with xwavenum=4 and ywavenum=3 which means that zonal/meridional waves with wave lengths of 727.5 km/750km and above are nudged.

Relaxation Time and nudging factors:

Some papers give non-dimensional nudging factors, e.g. alpha=0.05 in Feser et al. 2001.

This non-dimensional nudging factor relates to WRFs nudging coefficients by:

alpha = dt*guv = 180s*0.0003 s-1 = 0.054 in my case, so my alpha is in the same ball park as Feser et al. 2001.

The relaxation time in my case is 1/guv=1/0.0003=3333=56mins.

Literature:

F. Feser, R. Weisse, and H. von Storch, “Multi-decadal atmospheric modeling for Europe yields multi-purpose data,” Eos, Trans. Amer. Geophys. Union, vol. 82, no. 28, pp. 305–310, 2001.

H. von Storch, H. Langenberg, and F. Feser, “A spectral nudging technique for dynamical downscaling purposes,” Mon. Weather Rev., vol. 128, no. 10, pp. 3664–3673, Oct. 2000.

Thanks

J

I run WRF in hindcast mode, simulating the last 5 years with a coarse domain (30km grid spacing, no nesting). I use FNL as forcing and like to apply spectral nudging.

I like to nudge uv, t and ph above the PBL and above the 18 model level (above ~850 hPa level at standard atmosphere with my eta level settings).

I have one questions on the spectral nudging settings:

1. What does the parameter gfdda_end_h stand for and how do i need to set it if i like to apply spectral nudging during the full five year simulation. The FNL forcing has a time interval of 6h so gfdda_interval_m = 360 ( time interval (min) between analysis times). Will gfdda_end_h = 6 do the job?

My spectral nudging settings are below:

&fdda

grid_fdda = 2,

gfdda_inname = "wrffdda_d<domain>",

gfdda_interval_m = 360,

gfdda_end_h = 6,

io_form_gfdda = 2,

fgdt = 0,

if_no_pbl_nudging_uv = 1,

if_no_pbl_nudging_t = 1,

if_no_pbl_nudging_q = 1,

if_no_pbl_nudging_ph = 1,

if_zfac_uv = 1,

k_zfac_uv = 18,

if_zfac_t = 1,

k_zfac_t = 18,

if_zfac_q = 1,

k_zfac_q = 18,

if_zfac_ph = 1,

k_zfac_ph = 18,

guv = 0.0003,

gt = 0.0003,

gq = 0,

gph = 0.0003,

xwavenum = 4,

ywavenum = 3,

if_ramping = 0,

dtramp_min = 60.0,

Below is my reasoning to derive these spectral nudging settings: Please advise if my understanding of the methods and settings is erroneous or if you have other recommendations.

What parameters to nudge?

I choose to nudge temperature (t), horizontal wind components (uv) and geopotential (ph). I did not nudge moisture (q) in line with a recommendation by e.g. Macho et al (2005) Regional Climate Simulations over North America: Interaction of Local Processes with Improved Large-Scale Flow in Journal of Climate, 200.5

Is there any experience in the community on the benefits of nudging temperature in addition to uv, ph?

Xwavenum, ywavenum settings:First I figured out which scales I want to nudge: scales that are already reliably modeled by the forcing data set and scales on which the regional model may not add value. (e.g. von Storch et al. 2000)

The forcing FNL analysis data has a horizontal grid spacing of 1°, so I assume that it can reasonably well resolve waves with wave lengths of 700 km or above. Thus I wanted to nudge waves with wave lengths > 700 km. Following Eriks earlier posting i calculated the required domain size by multiplying my grid points in x and y with the grid spacing.

So: e_we = 97, e_sn = 75, dx=30km

xwavenum = 3: wavelength=97*30km /3 = 970km

xwavenum = 4: wavelength=97*30km /4 = 727.5 km

xwavenum = 5; wavelength=97*30km /5 = 582 km

ywavenum = 2; 75*30km /2 = 1125km

ywavenum = 3; 75*30km /3 = 750km

ywavenum = 4; 75*30km /4 = 562.5 km

So i went with xwavenum=4 and ywavenum=3 which means that zonal/meridional waves with wave lengths of 727.5 km/750km and above are nudged.

Relaxation Time and nudging factors:

Some papers give non-dimensional nudging factors, e.g. alpha=0.05 in Feser et al. 2001.

This non-dimensional nudging factor relates to WRFs nudging coefficients by:

alpha = dt*guv = 180s*0.0003 s-1 = 0.054 in my case, so my alpha is in the same ball park as Feser et al. 2001.

The relaxation time in my case is 1/guv=1/0.0003=3333=56mins.

Literature:

F. Feser, R. Weisse, and H. von Storch, “Multi-decadal atmospheric modeling for Europe yields multi-purpose data,” Eos, Trans. Amer. Geophys. Union, vol. 82, no. 28, pp. 305–310, 2001.

H. von Storch, H. Langenberg, and F. Feser, “A spectral nudging technique for dynamical downscaling purposes,” Mon. Weather Rev., vol. 128, no. 10, pp. 3664–3673, Oct. 2000.

Thanks

J

- Rostock
**Posts:**2**Joined:**Thu Oct 06, 2011 5:59 am

I did a small sensitivity study which showed that a spectral nudging simulation is highly sensitive to the gfdda_end_h parameter.

I would like to use spectral nudging and nudge uv and ph from the FNL analysis.

The forcing of the FNL is updated every 6h, so I want to nudge the updated forcing.

I want to run one year simulation.

How do I need to set the gfdda_end_h parameter?

gfdda_end_h = 6 or

gfdda_end_h = 8760 (24 hours/day*365 days?

What does the gfdda_end_h parameter do?

It is unclear to me and the explanations I have seen so far did not help me to understand it.

My fdda section is below.

Thanks for your help

J

&fdda

grid_fdda = 2,

gfdda_inname = "wrffdda_d<domain>",

gfdda_interval_m = 360,

gfdda_end_h = 6,

io_form_gfdda = 2,

fgdt = 0,

if_no_pbl_nudging_uv = 1,

if_no_pbl_nudging_t = 1,

if_no_pbl_nudging_q = 1,

if_no_pbl_nudging_ph = 1,

if_zfac_uv = 1,

k_zfac_uv = 18,

if_zfac_t = 1,

k_zfac_t = 18,

if_zfac_q = 1,

k_zfac_q = 18,

if_zfac_ph = 1,

k_zfac_ph = 18,

guv = 0.0003,

gt = 0,

gq = 0,

gph = 0.0003,

xwavenum = 5,

ywavenum = 4,

if_ramping = 0,

dtramp_min = 60.0,

I would like to use spectral nudging and nudge uv and ph from the FNL analysis.

The forcing of the FNL is updated every 6h, so I want to nudge the updated forcing.

I want to run one year simulation.

How do I need to set the gfdda_end_h parameter?

gfdda_end_h = 6 or

gfdda_end_h = 8760 (24 hours/day*365 days?

What does the gfdda_end_h parameter do?

It is unclear to me and the explanations I have seen so far did not help me to understand it.

My fdda section is below.

Thanks for your help

J

&fdda

grid_fdda = 2,

gfdda_inname = "wrffdda_d<domain>",

gfdda_interval_m = 360,

gfdda_end_h = 6,

io_form_gfdda = 2,

fgdt = 0,

if_no_pbl_nudging_uv = 1,

if_no_pbl_nudging_t = 1,

if_no_pbl_nudging_q = 1,

if_no_pbl_nudging_ph = 1,

if_zfac_uv = 1,

k_zfac_uv = 18,

if_zfac_t = 1,

k_zfac_t = 18,

if_zfac_q = 1,

k_zfac_q = 18,

if_zfac_ph = 1,

k_zfac_ph = 18,

guv = 0.0003,

gt = 0,

gq = 0,

gph = 0.0003,

xwavenum = 5,

ywavenum = 4,

if_ramping = 0,

dtramp_min = 60.0,

- Rostock
**Posts:**2**Joined:**Thu Oct 06, 2011 5:59 am

It may be a bit late now, but the way I understand this parameter, and the way my more experienced colleagues are using it, it sets the time into the forecast where nudging ends. So if you want it for the whole run, set it to the length of the run (in hours).

Or did you find something different?

Or did you find something different?

- Andre
**Posts:**16**Joined:**Thu Mar 08, 2012 11:15 am

In case someone is stumbling upon this thread looking for advice on setting the wavenumbers, have a look at https://rmets.onlinelibrary.wiley.com/d ... 02/qj.3032

- fryw
**Posts:**5**Joined:**Thu Feb 28, 2019 5:26 am

7 posts
• Page **1** of **1**

Users browsing this forum: No registered users and 8 guests

Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group