xclim: xarray-based climate data analytics

Contributors to this version: Trevor James Smith (@Zeitsperre), Éric Dupuis (@coxipi).

New features and enhancements

• New properties: Bivariate Spell Length (xclim.sdba.properties.bivariate_spell_length), Generalized Spell Lengths with an argument for window, and Specific Spell Lengths with window fixed to '1' (xclim.sdba.properties.threshold_count, xclim.sdba.properties.bivariate_threshold_count). (PR/1758).
• New option normalize in sdba.measures.taylordiagram to obtain normalized Taylor Diagrams (divide standard deviations by standard deviation of the reference). (PR/1764).

Breaking changes

• pint has been pinned below v0.24 until xclim can be updated to support the latest version. (GH/1771, PR/1772).
• numpy has been pinned below v2.0.0 until xclim can be updated to support the latest version. (PR/1783).
• Calendar utilities that have an equivalent in xarray have been deprecated and will be removed in xclim v0.51.0. (GH/1010, PR/1761). This concerns the following members of xclim.core.calendar:
  • convert_calendar : Use Dataset.convert_calendar, DataArray.convert_calendar or xr.coding.calendar_ops.convert_calendar instead.
    • If your code passes target as an array, first convert the source to the target's calendar and then reindex the result to target.
    • If you were using the doy=True option, replace it with xc.core.calendar.convert_doy(source, target_cal).convert_calendar(target_cal).
    • "default" is no longer a valid calendar name for any xclim functions and will not be returned by get_calendar. Xarray has a use_cftime argument, xclim exposes it when the distinction is needed.
  • date_range : Use xarray.date_range instead.
  • date_range_like: Use xarray.date_range_like instead.
  • interp_calendar : Use Dataset.interp_calendar or xarray.coding.calendar_ops.interp_calendar instead.
  • days_in_year : Use xarray.coding.calendar_ops._days_in_year instead.
  • datetime_to_decimal_year : Use xarray.coding.calendar_ops._datetime_to_decimal_year instead.

Internal changes

• Synchronized tooling versions across pyproject.toml and tox.ini and pinned them to the latest stable releases in GitHub Workflows. (PR/1744).
• Fixed a few small spelling and grammar issues that were causing errors with codespell. Now ignoring SVG files. (PR/1769).
• Temporarily skipping the test_hawkins_sutton_smoke test due to strange behaviour with xarray. (PR/1769).
• Fixed some previously uncaught errors raised from recent versions of pylint and codespell. (PR/1772).
• Set the doctest examples to all use h5netcdf with worker-separated caches to load datasets. (PR/1772).

Bug fixes

• xclim.indices.{cold|hot}_spell_total_length now properly uses the argument window to only count spells with at least window time steps. (GH/1765, PR/1777).
• Addressed an error in xclim.ensembles._filters._concat_hist where remnants of a scenario selection were not being dropped properly. (PR/1780).