Questions for 25km East Asia domain experiment. – in #12: CCLM Starter Package Support

in #12: CCLM Starter Package Support

<p> Dear all, </p> <p> Hi, This is Lee Donghyun in <span class="caps"> POSTECH </span> , South Korea. <br/> We have succeed to install &amp; run cclm-sp in our local server, <br/> and test with initial data provided( <span class="caps"> NCEP </span> ,Jan~Feb,2000). </p> <p> Then, now I tried to change the domain over East-Asia (as <span class="caps"> CORDEX </span> phase2 domain) with higher resolution <br/> (0.22’ almost 25km). The domain picture is attached. </p> <p> I have three issues while generating 25km East-Asia results. </p> <p> (Q1) Should I simulate long-term rcm project not by cclm-sp but cclm? </p> <p> In a condition with same version of cclm &amp; int2lm, is there any possible difference between cclm-sp &amp; cclm? </p> <p> I guess the only difference is cclm-sp providing more utilities than cclm itself. <br/> If there is not difference between cclm-sp and cclm, can I simulate our project with cclm-sp? <br/> or not, What makes difference between starter package and model itself? </p> <p> (Q2) Issue related determinining dt (time step in seconds) </p> <p> At first time, I used dt=240 as default setting for Starter-package, <br/> but strong cold bias occurs overall region. <br/> At second time, I used dt=120, the cold bias weaken. (See also attached picture) </p> <p> How can we decide appropriate dt? <br/> Should we decide dt based on experiential results? or is there any formula to calculate appropriate dt? </p> <p> (Q3) Issue related appropriate domain </p> <p> When I tried to run the test with our objective domain(396×251, pollon=-63.43 &amp; pollat=55.6) <br/> During the preprocess(int2lm), Error messages occurs related zmaxlon, zminlon. </p> <p> zminlon -179.96501460844 &lt; -176.25 <br/> zmaxlon 179.998320995158 &gt; 174.375 </p> <p> In order to generate results, I change domain region smaller (380×236) and it is working. <br/> Is there a way to genreate 396×251 for my target region? </p> <p> Thank you very much. </p> <p> Best wishes <br/> Lee Donghyun. </p>

  @donghyunlee in #0b49ae0

<p> Dear all, </p> <p> Hi, This is Lee Donghyun in <span class="caps"> POSTECH </span> , South Korea. <br/> We have succeed to install &amp; run cclm-sp in our local server, <br/> and test with initial data provided( <span class="caps"> NCEP </span> ,Jan~Feb,2000). </p> <p> Then, now I tried to change the domain over East-Asia (as <span class="caps"> CORDEX </span> phase2 domain) with higher resolution <br/> (0.22’ almost 25km). The domain picture is attached. </p> <p> I have three issues while generating 25km East-Asia results. </p> <p> (Q1) Should I simulate long-term rcm project not by cclm-sp but cclm? </p> <p> In a condition with same version of cclm &amp; int2lm, is there any possible difference between cclm-sp &amp; cclm? </p> <p> I guess the only difference is cclm-sp providing more utilities than cclm itself. <br/> If there is not difference between cclm-sp and cclm, can I simulate our project with cclm-sp? <br/> or not, What makes difference between starter package and model itself? </p> <p> (Q2) Issue related determinining dt (time step in seconds) </p> <p> At first time, I used dt=240 as default setting for Starter-package, <br/> but strong cold bias occurs overall region. <br/> At second time, I used dt=120, the cold bias weaken. (See also attached picture) </p> <p> How can we decide appropriate dt? <br/> Should we decide dt based on experiential results? or is there any formula to calculate appropriate dt? </p> <p> (Q3) Issue related appropriate domain </p> <p> When I tried to run the test with our objective domain(396×251, pollon=-63.43 &amp; pollat=55.6) <br/> During the preprocess(int2lm), Error messages occurs related zmaxlon, zminlon. </p> <p> zminlon -179.96501460844 &lt; -176.25 <br/> zmaxlon 179.998320995158 &gt; 174.375 </p> <p> In order to generate results, I change domain region smaller (380×236) and it is working. <br/> Is there a way to genreate 396×251 for my target region? </p> <p> Thank you very much. </p> <p> Best wishes <br/> Lee Donghyun. </p>

Questions for 25km East Asia domain experiment.

DL wrote

Dear all,

Hi, This is Lee Donghyun in POSTECH , South Korea.
We have succeed to install & run cclm-sp in our local server,
and test with initial data provided( NCEP ,Jan~Feb,2000).

Then, now I tried to change the domain over East-Asia (as CORDEX phase2 domain) with higher resolution
(0.22’ almost 25km). The domain picture is attached.

I have three issues while generating 25km East-Asia results.

(Q1) Should I simulate long-term rcm project not by cclm-sp but cclm?

In a condition with same version of cclm & int2lm, is there any possible difference between cclm-sp & cclm?

I guess the only difference is cclm-sp providing more utilities than cclm itself.
If there is not difference between cclm-sp and cclm, can I simulate our project with cclm-sp?
or not, What makes difference between starter package and model itself?

(Q2) Issue related determinining dt (time step in seconds)

At first time, I used dt=240 as default setting for Starter-package,
but strong cold bias occurs overall region.
At second time, I used dt=120, the cold bias weaken. (See also attached picture)

How can we decide appropriate dt?
Should we decide dt based on experiential results? or is there any formula to calculate appropriate dt?

(Q3) Issue related appropriate domain

When I tried to run the test with our objective domain(396×251, pollon=-63.43 & pollat=55.6)
During the preprocess(int2lm), Error messages occurs related zmaxlon, zminlon.

zminlon -179.96501460844 < -176.25
zmaxlon 179.998320995158 > 174.375

In order to generate results, I change domain region smaller (380×236) and it is working.
Is there a way to genreate 396×251 for my target region?

Thank you very much.

Best wishes
Lee Donghyun.

View in channel
<p> (Q1) cclm is part of the starter package cclm-sp (You can find the source in <code> cclm-sp/src/cclm </code> ). cclm-sp also contains int2lm and several scripts (see [[cclm-sp:Documentation|Starter Package Documentation]]). </p> <p> (Q2) As a rule of thumb it is linear with grid width. The bias in T_2M is much too large. As it seems to reflect the orography, did you applied the mandatory orographic correction to T_2M before comparing with observations? </p> <p> (Q3) You probably cross the date line in the larger domain. In that case put <code> east_add_in=1 </code> into the namelist <span class="caps"> GRID </span> _IN in int2lm. </p> <p> General remarks: <br/> (!) A careful namelist setup of the model is a major basis for receiving satisfactory results. Since your domain is in the tropics I recommend to visit http://www.clm-community.eu/index.php?menuid=218&amp;reporeid=302 (you need to login on the page first) and check under the section “Model Setup” the <span class="caps"> YUSPECIF </span> and the <span class="caps"> OUTPUT </span> files. These hold the information for a tropical setup for cclm version 4.8_clm17 and int2lm version 1.9_clm3. If you need namelists for newer versions maybe Hans-Jürgen Panitz has these already. </p>

  @burkhardtrockel in #4165bb9

<p> (Q1) cclm is part of the starter package cclm-sp (You can find the source in <code> cclm-sp/src/cclm </code> ). cclm-sp also contains int2lm and several scripts (see [[cclm-sp:Documentation|Starter Package Documentation]]). </p> <p> (Q2) As a rule of thumb it is linear with grid width. The bias in T_2M is much too large. As it seems to reflect the orography, did you applied the mandatory orographic correction to T_2M before comparing with observations? </p> <p> (Q3) You probably cross the date line in the larger domain. In that case put <code> east_add_in=1 </code> into the namelist <span class="caps"> GRID </span> _IN in int2lm. </p> <p> General remarks: <br/> (!) A careful namelist setup of the model is a major basis for receiving satisfactory results. Since your domain is in the tropics I recommend to visit http://www.clm-community.eu/index.php?menuid=218&amp;reporeid=302 (you need to login on the page first) and check under the section “Model Setup” the <span class="caps"> YUSPECIF </span> and the <span class="caps"> OUTPUT </span> files. These hold the information for a tropical setup for cclm version 4.8_clm17 and int2lm version 1.9_clm3. If you need namelists for newer versions maybe Hans-Jürgen Panitz has these already. </p>
BR replied

(Q1) cclm is part of the starter package cclm-sp (You can find the source in cclm-sp/src/cclm ). cclm-sp also contains int2lm and several scripts (see [[cclm-sp:Documentation|Starter Package Documentation]]).

(Q2) As a rule of thumb it is linear with grid width. The bias in T_2M is much too large. As it seems to reflect the orography, did you applied the mandatory orographic correction to T_2M before comparing with observations?

(Q3) You probably cross the date line in the larger domain. In that case put east_add_in=1 into the namelist GRID _IN in int2lm.

General remarks:
(!) A careful namelist setup of the model is a major basis for receiving satisfactory results. Since your domain is in the tropics I recommend to visit http://www.clm-community.eu/index.php?menuid=218&reporeid=302 (you need to login on the page first) and check under the section “Model Setup” the YUSPECIF and the OUTPUT files. These hold the information for a tropical setup for cclm version 4.8_clm17 and int2lm version 1.9_clm3. If you need namelists for newer versions maybe Hans-Jürgen Panitz has these already.

View in channel
<p> Thank you for your kind advices. </p> <p> Although the resolution is eqaul starter-package setting(0.44), their dt was smaller (East Asia domain, 150s) than starter-package (Europe domain, 240s) in the paper <br/> (Regional dynamical downscaling with <span class="caps"> CCLM </span> over East Asia, Dinana Wang, 2013, Meteorol.Atmos.Phys). </p> <p> So for that reason, I am confused about linear with grid width, dt=120s or dt=75s <br/> Should I test with both results? </p> <p> The bias is just calculated by subtraction. <br/> Could you please introduce some materials for me to understand ‘the mandatory orographic correction’? </p>

  @donghyunlee in #4e044ad

<p> Thank you for your kind advices. </p> <p> Although the resolution is eqaul starter-package setting(0.44), their dt was smaller (East Asia domain, 150s) than starter-package (Europe domain, 240s) in the paper <br/> (Regional dynamical downscaling with <span class="caps"> CCLM </span> over East Asia, Dinana Wang, 2013, Meteorol.Atmos.Phys). </p> <p> So for that reason, I am confused about linear with grid width, dt=120s or dt=75s <br/> Should I test with both results? </p> <p> The bias is just calculated by subtraction. <br/> Could you please introduce some materials for me to understand ‘the mandatory orographic correction’? </p>
DL replied

Thank you for your kind advices.

Although the resolution is eqaul starter-package setting(0.44), their dt was smaller (East Asia domain, 150s) than starter-package (Europe domain, 240s) in the paper
(Regional dynamical downscaling with CCLM over East Asia, Dinana Wang, 2013, Meteorol.Atmos.Phys).

So for that reason, I am confused about linear with grid width, dt=120s or dt=75s
Should I test with both results?

The bias is just calculated by subtraction.
Could you please introduce some materials for me to understand ‘the mandatory orographic correction’?

View in channel
<p> The maximum time step allowed is limited by the Courant–Friedrichs–Lewy ( <span class="caps"> CFL </span> ) condition. This is the time step I am referring to. For 0.44 the maximum it is somewhere between 240s and 360s, with 240s on the save side. You can take this for your “rule of thumb” calculation for the maximum time step of higher resolutions. The minimum time limit is given by your computer resources. </p> <p> For the orographic correction you need to apply the moist adiabatic lapse rate (6.5 K/km is mostly used) during interpolation of one data set to the other. The difference between the two orographies is multiplied by the lapse rate and added (or subtracted) to one of the data sets. After that you may compare the data sets. This is a common praxis. However, one should be aware that taking a constant 6.5 K/km is also only a first order approximation. </p>

  @burkhardtrockel in #2d8324f

<p> The maximum time step allowed is limited by the Courant–Friedrichs–Lewy ( <span class="caps"> CFL </span> ) condition. This is the time step I am referring to. For 0.44 the maximum it is somewhere between 240s and 360s, with 240s on the save side. You can take this for your “rule of thumb” calculation for the maximum time step of higher resolutions. The minimum time limit is given by your computer resources. </p> <p> For the orographic correction you need to apply the moist adiabatic lapse rate (6.5 K/km is mostly used) during interpolation of one data set to the other. The difference between the two orographies is multiplied by the lapse rate and added (or subtracted) to one of the data sets. After that you may compare the data sets. This is a common praxis. However, one should be aware that taking a constant 6.5 K/km is also only a first order approximation. </p>
BR replied

The maximum time step allowed is limited by the Courant–Friedrichs–Lewy ( CFL ) condition. This is the time step I am referring to. For 0.44 the maximum it is somewhere between 240s and 360s, with 240s on the save side. You can take this for your “rule of thumb” calculation for the maximum time step of higher resolutions. The minimum time limit is given by your computer resources.

For the orographic correction you need to apply the moist adiabatic lapse rate (6.5 K/km is mostly used) during interpolation of one data set to the other. The difference between the two orographies is multiplied by the lapse rate and added (or subtracted) to one of the data sets. After that you may compare the data sets. This is a common praxis. However, one should be aware that taking a constant 6.5 K/km is also only a first order approximation.

View in channel