Hello Guillermo,

Sir thank you for the explanation. But I have a query. I had read from this article(Article The Impact of Spatial Resolution, Land Use, and Spinup Time ...

I have read some papers where both spinup and nudging is applied simultaneously, but I am not able to understand the rationale behind it.

Hello Neeldip,

When you apply nudging you are still needing spinup and restart, but with a larger period than a non-nudged simulation. For instance, it is well known that a good spinup time is 24 hours every 72 hours, but in case of nudged simulation, to let it act properly, it´s recomended a 24 h spinup every 10 days.

Best regards,

Nicolas

Hi Nicolas,

Thanks a lot for the answer.

You have mentioned regarding a 24 h spinup every 10 days if nudging is used.

But spinup will be required, only if I reinitialize the model again after 10 days and wont be required if i do a model restart right?

As to what I know is that spinup is required for the model to develop its own climatology and reduce the influence of the initial conditions. After 10 days, the climatology would be sufficiently develop and hence there would be no requirement for model spinup again after the initial spinup; and that spinup will be only required if I reinitialize the model.

Best regards,

Neeldip

Hi Neeldip,

Exactly, it is reinitialization of the model, not restart. Sorry about that, in my lenguage it´s the same word.

Summary: 24 h spinup with a model reinitialize every 10 days if nudging is used. Otherwise 24 h spinup with a model reinitialize every 72 hours.

Regards,

Nicolas

Thank you very much Nicolas Gonzalez Alonso-de-Linaje sir. My doubts are clear now.

Thank you Sir Guillermo Auad

Neeldip just to add up on the above mentioned information. Spin-up also depend on your domain and grid size since spin-up in simple terms is nothing but the time required by the model to reach a steady or stable ( sorry am not getting the correct term) state. The initial concentrations provided in most models are usually some vague values and hence the model needs some time to reach the actual state.

Thanks Shovan.

If indeed we are performing spinup doesn't that make the effort for getting good ic,bc non useful since eventually we want to develop the domains own climatology and become less dependent on the ic,bc?

Of course but in order to get say get the concentration at Point A at the very first day in very first hour you also need a good estimate of the background or initial concentration at point A. Since the model can as far only estimate the concentration based on the emissions at point A since you still don't have the transport of concentration from neighboring points or the background concentration at very first hour of model run. And since the total concentration at point A is a summation of the in-situ concentration as well as transport and background concentration so until you have good initial concentration your model estimates will be bad. Now expand this one point to multiple points and you will get an idea on why spin-up will depend on size of domain.

You should provide details about your experimental set up to get the right answer. In particular, what is the length of the simulation, its spatial resolution and domain size, and even more importantly which data do you use to initialize the model ? For instance, when performing climate change simulations over several decades, spin-up can be of 1-3 years depending on initialization of soil water content and snow cover. If the initial conditions (especially land) already correspond to an equilibrium state of the model (that is the spin-up was already done at some point), spin-up is only atmospheric and can be much shorter.

Hi Martin Leduc,

My domain is 110*110 grids parent domain with grid resolution of 27 km and grid ratio of 3 for the two nested domains.

I want to perform 1 month simulations with ERA5 reanalysis as IC/BC.

If the length of the simulations are only one month, the only spin-up that needs to be accounted for is the atmospheric one, which is typically only a few days. If I understand well, you are nesting two domains, the first one at 27km resolution (which is about the same as ERA5), and the second one at around 9km. In this situation, I would suggest to use 5 days spin-up in the beginning of each nested integration (so adding up both spin-up would lead to 10 days), and then begin the one-month simulation.

Maybe you could find the following paper of interest :

Article Regional climate model sensitivity to domain size

Hi Martin Leduc,

Thank you for the suggestions and the link.

However, I didn't understand the 10 day concept (so adding up both spin-up would lead to 10 days) . Could you kindly elborate it a bit? It'll be very much helpful for me.

This is really just an example of how it could work, but see section 3.1 in the paper linked above for a more detailed explanation. Another important point, the spin-up does not only depend on your experimental framework but also on what exactly you want to study with your simulations.

Great, I will go through the article.

Thanks again for your suggestions, these things are really useful for a novice.