search

MITgcm / xmitgcm

Read MITgcm mds binary files into xarray
http://xmitgcm.readthedocs.io
MIT License
56 stars 65 forks source link

momentum equation reconstruction #58

Open apatlpo opened 6 years ago

apatlpo commented 6 years ago

Hi A postdoc should start sometime early next year on the analysis of LLC4320 surface fields. One objective will be to reconstruct and decompose (spatiotemporal filtering) surface momentum equations. I'd be very keen on using xgcm, would you have Ryan some comments/advice about that?

rabernat commented 6 years ago

It is important to understand that there are two separate packages. This one (xmitgcm) has a single, specific purpose: read MITgcm MDS files into xarray format. This includes llc files, and I have done extensive testing of the performance of reading LLC4320 on pleaides. It is described in detail in the documentation: http://xmitgcm.readthedocs.io/en/latest/performance.html

Once the files are in xarray format, there are many options for how to analyze them, including all of xarray's built in capabilities. My xgcm package (http://xgcm.readthedocs.io/en/latest/) is designed to help with the analysis of finite volume GCMs and is not specific to MITgcm.

In general, both these packages could potentially be very useful to the project you describe. There is also lots of work to do to improve both of them, so I encourage you to try them out, help us identify issues for improvement, and potentially add features that you need for your project. If you describe what you want to do in more detail, I could provide more specific advice.

p.s. I'm curious who you are. You know me by name, but your github handle is just apatlpo. Have we met before?

apatlpo commented 6 years ago

Thanks Ryan, this is very helpful. I don't have access to LLC4320 quite yet, so testing will have to wait for a bit ... unless the file you used for your demos is available somewhere? yes we met, I am aurélien ponte. Sorry for the confusion.

rabernat commented 6 years ago

Hi Aurélien!

unless the file you used for your demos is available somewhere?

Unfortunately it's hard to find places to host hundreds of TB!

apatlpo commented 6 years ago

Ok, I'll wait for the data then. With another postdoc, we've recently analyzed NATL60 with oocgcm and got familiar with xarray. I look forward to give xgcm a spin now.

xy6g13 commented 6 years ago

Hey Ryan, I'm Xiaolong Yu, the postdoc working on LLC4320 with Aurélien! As you might know, we have access to the simulation outputs now (even though there are a few corrupted data). We have been thinking about reconstruction of evolution equations for surface momentum and tracers. Would be great if you could give me some general suggestions. Also, do you know where I can find detailed description of LLC4320? Many thanks.

rabernat commented 6 years ago

We have been thinking about reconstruction of evolution equations for surface momentum and tracers. Would be great if you could give me some general suggestions

A good place to start is with the MITgcm manual describing the momentum equation: http://mitgcm.readthedocs.io/en/latest/algorithm/algorithm.html#vector-invariant-momentum-equations

Also, do you know where I can find detailed description of LLC4320?

As far as I know, none exists. If you want to understand the configuration used when running the model, you can look at the input files here: http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm_contrib/llc_hires/llc_4320/input/

apatlpo commented 6 years ago

thanks Ryan, this is very useful. The most relevant seems to be the data file. For tracer advections scheme, we may need to dig into the code on the other hand.