NCAR/esmlab

Esmlab Design Document

Closed this issue · 8 comments

dcherian commented

I saw mention of indices here, so wanted to make sure you were aware of https://github.com/Ouranosinc/xclim

  • 👍1
bradyrx commented

I can't edit that document directly, but wanted to make you aware of https://github.com/bradyrx/esmtools/ as well. I don't think these clash, since esmtools is really a repository for personal functions I use commonly with working with gridded output. It's a kitchen sink for really any sort of domain-specific or domain-agnostic functions for climate analysis.

There might be some things in there worth migrating or drawing from. Particularly functions related to our discussion in bradyrx/esmtools#65. Between climpred and esmtools stats modules I have an rm_poly, rm_trend, fit_poly (returns the actual fit that was made), ACF, autocorr, corr and so on that can be applied to a grid. The native scipy and numpy functions can't of course just ingest gridded products. I have code snippets to update the functions to be optimized for dask per work with @andersy005 and @gelsworth

andersy005 commented

@bradyrx,

I can't edit that document directly,

Let me know whether https://hackmd.io/mwGPvXgXSu2f7iXGCO3m4A?edit doesn't work when you are signed in (I will fix this).

  • 👍1
andersy005 commented

@maboualidev, @matt-long

I just updated the design document with some of information from our discussion this afternoon. When you get a chance in the coming days, please take a look at it, and add the bits that are missing.

huard commented

@andersy005

Hi, xclim dev here. I've gone through the list of issues in esmlab and it looks like we're sharing many objectives. In particular, I'd like to see weighted statistics supported natively by xarray, as well as support for time_bnds aware operations.

Originally, xclim was designed to be a library of climate indicators that are typically used for climate impact studies. We've added some utilities around it, but there's clearly a lot of overlap with other packages. We'd be happy for xclim to focus on climate indicators, include ocean and ice indicators, and outsource other utilities (ensemble stats, clustering, subsetting) to other libraries such as esmlab.

Would it be useful to have an Earth System Modeling and Data Analysis project board somewhere (e.g. pydata) to coordinate efforts?

maboualidev commented