search

MetOffice / CSET

Toolkit for evaluation and investigation of numerical models for weather and climate applications.
https://metoffice.github.io/CSET/
Apache License 2.0
14 stars 5 forks source link

Use VerPy for visualisation of MET stats #295

Open jfrost-mo opened 11 months ago

jfrost-mo commented 11 months ago

MET outputs statistics into a text based format, however we want everything to be NetCDF/cubes for uniform interfacing. Consider writing a postprocessing script to convert these.

Issue transferred from https://github.com/MetOffice/CSET-workflow/issues/6.

jfrost-mo commented 11 months ago

Note that NG-VER has taken the route of pushing the MET output into a SQLite database and plot from there. The code is already in VERPy, both to put the MET output into the database and to do the plotting. If we are to break the alignment with NG-VER, it needs a deeper discussion.

Original comment by JorgeBornemann

jfrost-mo commented 11 months ago

OK. Issue #7 answers my previous comment.

Original comment by JorgeBornemann

jfrost-mo commented 11 months ago

Before we commit to one path or the other, I really want to answer #11. If VerPy is doing the same kind of plotting as we want, then it makes sense to reuse their efforts. The sticker on our side is being very uncertain about the timelines for VerPy being ready, but if like you say they have working code already, that gives me much greater hope that it is something we can use sooner rather than later.

Original comment by jfrost-mo

jfrost-mo commented 11 months ago

From discussions this week there is an acknowledgement that VerPy will at some point need to move onto GitHub (licence model to be decided), but blockers are that someone has to move it and sort the licencing and in the face of few resources it is not a priority. This means that we might have to develop our own basic plotting solution until VerPy is in a position to be shared.

Original comment by Sylviabohnenstengel

jfrost-mo commented 11 months ago

To make it reasonable for advanced users to customise their plots we will first create a NetCDF and then convert that to SQL for VerPy. NetCDF should be sufficient for our use, even in extreme cases like climate runs and K-Scale.

We will try and stick to standard NetCDF names, but it isn't essential that we be 100% covered by standard.

Original comment by jfrost-mo

jfrost-mo commented 11 months ago

We will want most plotting to be handled centrally within CSET rather than each diagnostic doing their own. This is to promote consistency.

VerPy would be an extra dependency, and one that partners have previously had portability issues with, though I think those have mostly been fixed by now. The main issue currently is with its license, as it is currently not open source. We could pull the code from FCM, but wouldn't be able to use it with the open source bit of the workflow.

Original comment by jfrost-mo

jfrost-mo commented 11 months ago

For now we will just use a custom plot for the point-stat stats, while keeping in mind that we may move to VerPy in future.

Original comment by jfrost-mo