I've been reading through the documentation and code of the repo a bit more in depth (it's very nicely described!). Something that I think needs some clarification is if this is purely focused on LAM or also global graphs? I guess it could start out with 2D LAM graphs and potentially include global ones later, but it would be good to clarify.
Related to this I found the way graphs from existing works are referenced somewhat inaccurate, as both Keisler and Lam et al. purely propose global, triangular mesh graphs. This relates to the archetypes and the documentation around these (https://mllam.github.io/weather-model-graphs/creating_the_graph.html#graph-archetypes). What is described there are the adaptations of Keisler and Lam et al. to rectangular meshes made in Oskarsson et al.
I don't mind the names being used in the archetypes, but I think the documentation of these should make a more clear distinction between what was actually proposed in these papers and the LAM adaptation by Oskarsson et al. of this. E.g.
19 clarifies this a bit in the docstrings of the archetype, but does not change how these graphs are discussed in the documentation. Because of this I think this issue should stay open.
joeloskarsson
commented
1 month ago
I've been reading through the documentation and code of the repo a bit more in depth (it's very nicely described!). Something that I think needs some clarification is if this is purely focused on LAM or also global graphs? I guess it could start out with 2D LAM graphs and potentially include global ones later, but it would be good to clarify.
Related to this I found the way graphs from existing works are referenced somewhat inaccurate, as both Keisler and Lam et al. purely propose global, triangular mesh graphs. This relates to the archetypes and the documentation around these (https://mllam.github.io/weather-model-graphs/creating_the_graph.html#graph-archetypes). What is described there are the adaptations of Keisler and Lam et al. to rectangular meshes made in Oskarsson et al.
I don't mind the names being used in the archetypes, but I think the documentation of these should make a more clear distinction between what was actually proposed in these papers and the LAM adaptation by Oskarsson et al. of this. E.g.
https://github.com/mllam/weather-model-graphs/blob/e9a0381ddbf422581537bd3256fd1016a93e5b26/src/weather_model_graphs/create/archetype.py#L39-L41
is a bit misleading as this will not create the graph from Lam et al.