[CCLM-Converction Permitting Scale] Korea Domain (3km) simulation configuration – in #9: CCLM

in #9: CCLM

<p> Interesting and challenging task! </p> <p> My first “professional advice” would be to have a look at a recent paper by Brisson et al. <br/> “Modelling strategies for performing convection-permitting climate simulations” <br/> https://www.schweizerbart.de/papers/metz/detail/25/84891/Modelling_strategies_for_performing_convection_permitting_climate_simulations <br/> DOI: 10.1127/metz/2015/0598 </p> <p> One example: They found that the best downscaling strategy for <strong> their case </strong> is to use two nesting steps (25 km and 2.8 km) and an intermediate step (7 km) was found to be redundant. So maybe you can skip the 12.5 km nest. <br/> Also, including graupel parametrization in the <span class="caps"> CPCS </span> had a positive effect (itype_gscp = 4, I think). There is also information on the domain size, so the paper could give you some hints. </p> <p> I recently did some 0.022° runs nested into 0.22° over Svalbard. The scripts defining the set-ups are attached. <br/> The results look good so far, but I didn’t have time to do a deep analysis yet. Surely, the setup is not the final one. <br/> For instance, I didn’t include the graupel parametrization in the <span class="caps"> CPCS </span> . But generally, it looks close to your chooses. <br/> dt=30 seems a little high to me and could yield some <span class="caps"> CFL </span> violations. I started with dt=20 and had to go down to dt=10, but this could be due to the high latitudes in my case. </p> <p> Best regards <br/> Andreas </p>

  @andreasdobler in #a914ca2

<p> Interesting and challenging task! </p> <p> My first “professional advice” would be to have a look at a recent paper by Brisson et al. <br/> “Modelling strategies for performing convection-permitting climate simulations” <br/> https://www.schweizerbart.de/papers/metz/detail/25/84891/Modelling_strategies_for_performing_convection_permitting_climate_simulations <br/> DOI: 10.1127/metz/2015/0598 </p> <p> One example: They found that the best downscaling strategy for <strong> their case </strong> is to use two nesting steps (25 km and 2.8 km) and an intermediate step (7 km) was found to be redundant. So maybe you can skip the 12.5 km nest. <br/> Also, including graupel parametrization in the <span class="caps"> CPCS </span> had a positive effect (itype_gscp = 4, I think). There is also information on the domain size, so the paper could give you some hints. </p> <p> I recently did some 0.022° runs nested into 0.22° over Svalbard. The scripts defining the set-ups are attached. <br/> The results look good so far, but I didn’t have time to do a deep analysis yet. Surely, the setup is not the final one. <br/> For instance, I didn’t include the graupel parametrization in the <span class="caps"> CPCS </span> . But generally, it looks close to your chooses. <br/> dt=30 seems a little high to me and could yield some <span class="caps"> CFL </span> violations. I started with dt=20 and had to go down to dt=10, but this could be due to the high latitudes in my case. </p> <p> Best regards <br/> Andreas </p>
AD replied

Interesting and challenging task!

My first “professional advice” would be to have a look at a recent paper by Brisson et al.
“Modelling strategies for performing convection-permitting climate simulations”
https://www.schweizerbart.de/papers/metz/detail/25/84891/Modelling_strategies_for_performing_convection_permitting_climate_simulations
DOI: 10.1127/metz/2015/0598

One example: They found that the best downscaling strategy for their case is to use two nesting steps (25 km and 2.8 km) and an intermediate step (7 km) was found to be redundant. So maybe you can skip the 12.5 km nest.
Also, including graupel parametrization in the CPCS had a positive effect (itype_gscp = 4, I think). There is also information on the domain size, so the paper could give you some hints.

I recently did some 0.022° runs nested into 0.22° over Svalbard. The scripts defining the set-ups are attached.
The results look good so far, but I didn’t have time to do a deep analysis yet. Surely, the setup is not the final one.
For instance, I didn’t include the graupel parametrization in the CPCS . But generally, it looks close to your chooses.
dt=30 seems a little high to me and could yield some CFL violations. I started with dt=20 and had to go down to dt=10, but this could be due to the high latitudes in my case.

Best regards
Andreas

View in channel